Document
NASA Conference Publ&uth 22% LOAN COPY: RETURN TO CHNICAL . LIBRARY :.,.
IRTLAND AFB, N.M. 87117 ‘.
“j?.
,- -----2-i-;-=: - I - -~ - TECH LIBRARY KAFB, NM
llnlllsllllilllill
0099238 NASA Conference Public
NASA Aircraft
Controls Research
Gary P. Beaeley, Compiler Langley Research Center Proceedings of a workshop held at NASA Langley Research Center Hampton, Virginia October 25-27, 1983 National Aeronautics and Space Administration Scientific l d Technical Information Branch III IIIII I lmllllm llllllllllllllllllll II lIllIlIllIIllII !
PREFACE This publication contains the proceedings of the First Annual NASA Aircraft Controls Workshop, which was held October 25-27, 1983, at NASA Langley Research Center. This workshop highlighted ongoing aircraft controls research sponsored by NASA's Office of Aeronautics and Space Technology and provided a forum for critique of ongoing research as well as suggestions for needed research from controls experts or users of control technology. The workshop was initiated in response to a recommendation from the NASA Advisory Council's Informal Subcommittee on Aircraft Controls and Guidance.
About 200 aircraft controls experts from industry, government, and universities participated in the workshop. The workshop consisted of 24 technical presentations on various aspects of aircraft controls, ranging from the theoretical development of control laws to the evaluation of new controls technology in flight test vehicles. It also included a special report on the status of foreign aircraft technology and a panel session with seven representatives from organizations which use aircraft controls technology.
This panel addressed the controls research needs and opportunities for the future as well as the role envisioned for NASA in that research.
Input from the panel and response to the workshop presentations will be used by NASA in developing future programs.
This document contains copies of the visual material presented by each participant, together with descriptive material for each visual. A list of conference attendees is also included.
Use of trade names or names of manufacturers in this report does not constitute an official endorsement of such products or manufacturers, either expressed or implied, by the National Aeronautics and Space Administration.
Gary P. Beasley Langley Research Center iii CONTENTS PREFACE...................,., . , v .?, ?I . . . . . . iii PARTICIPANTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix NASA CONTROLRESEARCH OVERVIEW . . . . . . . . . . . . . . . . . . . . . . . . 1 Duncan E. McIver SESSION I: CONTROL SYSTEM DESIGN REQUIREMENTS FLYING QUALITIES CRITERIA FOR SUPERAUGMENTED AIRCRAFT . . . . . . . . . . . . . 25 Donald T. Berry LARGE AIRCRAFT HANDLING QUALITIES . . . . . . . . . . . . . . . . . . . . . . . 37 William D. Grantham A SUMMARY OF ROTORCRAFTHANDLING QUALITIES RESEARCHAT NASA AMES RESEARCH CENTER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Robert T. N. Chen FLEXIBLE AIRCRAFT FLYING AND RIDE QUALITIES . . . . . . . . . . . . . . . . . . 69 Irving L. Ashkenas, Raymond E. Magdaleno, and Duane T. McRuer IMPLICATIONS OF CONTROL TECHNOLOGYON AIRCRAFT DESIGN . . . . . . . . . . . . . 93 Steven M. Sliwa and P. Douglas Arbuckle RELIABILITY AND MAINTAINABILITY ASSESSMENTFACTORS FOR RELIABLE FAULT- TOLERANTSYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . ...115 Salvatore J. Bavuso SESSION II: PARAMETERAND STATE ESTIMATION AND OPTIMAL GUIDANCE TECHNIQUES PRACTICAL ASPECTS OF MODELING AIRCRAFT DYNAMICS FROM FLIGHT DATA . . . . . . . 135 Kenneth W. Iliff and Richard E. Maine APPLICATIONS OF MODEL STRUCTUREDETERMINATION TO FLIGHT TEST DATA . . . . . . . 155 James G. Batterson and Vladislav Klein MIXING 4D-EQUIPPED AND UNEQUIPPED AIRCRAFT IN THE TERMINAL AREA . . . . . . . . 171 L. Tobias, H. Erzberger, H. Q. Lee, and P. J. O'Brien APPLICATION OF FUEL/TIME MINIMIZATION TECHNIQUES TO ROUTE PLANNING AND TRAJECTORY OPTIMIZATION . . . . . . . . . . . . . . . . . . . , . . . . . . . 191 Charles E. Knox FEEDBACK LAWS FOR FUEL MINIMIZATION FOR TRANSPORTAIRCRAFT . . . . . . . . . . 209 Douglas B. Price and Christopher Gracey IDENTIFICATION OF MULTIVARIABLE HIGH-PERFORMANCETURBOFAN ENGINE DYNAMICS FROM CLOSED-LOOP DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221 Walter C. Merrill V SESSION III: CONTROLSYSTEM DESIGN TECHNIQUES EIGENSPACE DESIGN TECHNIQUES FOR ACTIVE FLUTTER SUPPRESSION . . . . . . . . . . 241 William L. Garrard and Bradley S. Liebst TOOLS FOR ACTIVE CONTROL SYSTEM DESIGN . . . . . . . . . . . . . . . . . . . . 263 William M. Adams, Jr., Sherwood H. Tiffany, and Jerry R. Newsom ALGORITHMS FOR OUTPUT FEEDBACK, MULTIPLE-MODEL, AND DECENTRALIZED CONTROL PROBLEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...281 Nesim Halyo and John R. Broussard PILOT MODELING, MODAL ANALYSIS, AND CONTROL OF LARGE FLEXIBLE AIRCRAFT . . . . 305 David K. Schmidt NONLINEAR SYSTEMS APPROACHTO CONTROL SYSTEM DESIGN . . . . . . . . . . . . . . 329 George Meyer ADAPTIVE CONTROL: MYTHS AND REALITIES . . . . . . . . . . . . . . . . . . . . 343 Michael Athans and Lena Valavani SPECIAL REPORT FOREIGN TECHNOLOGYSUMMARYOF FLIGHT CRUCIAL CONTROL SYSTEMS . . . . . . . . . 365 H. A. Rediess SESSION IV: RECENT EXPERIENCES IN IMPLEMENTATION OF ADVANCED CONTROL SYSTEMS FUNCTIONAL INTEGRATION OF VERTICAL FLIGHT PATH AND SPEED CONTROLUSING ENERGY PRINCIPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...389 A. A. Lambregts FLIGHT TEST RESULTS FOR THE DIGITAL INTEGRATED AUTOMATIC LANDING SYSTEM (DIALS) - A MODERNCONTROL FULL-STATE FEEDBACK DESIGN . . . . . . . . . . . . 411 R. M. Hueschen APPLICATION OF ADVANCED CONTROL TECHNIQUES TO AIRCRAFT PROPULSION SYSTEMS . . . 429 Bruce Lehtinen L-1011 TESTING WITH RELAXED STATIC STABILITY . . . . . . . . . . . . . . . . . 443 J. J. Rising and K. R. Henke AFTI/F-16 DIGITAL FLIGHT CONTROL SYSTEM EXPERIENCE . . . . . . . . . . . . . . 469 Dale A. Mackall EXPERIENCES WITH THE DESIGN AND IMPLEMENTATION OF FLUTTER SUPPRESSION SYSTEMS . 489 Jerry R. Newsom and Irving Abel vi SESSION V: RESEARCH OPPORTUNITIES FOR THE FUTURE RESEARCHOPPORTUNITIES FOR FUTURE COMMERCIALTRANSPORTS . . . . . . . . . . . . 511 J. F. Longshore A BRIEF REVIEW OF AIRCRAFT CONTROLSRESEARCH OPPORTUNITIES IN THE GENERAL AVIATION FIELD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519 Eric R. Kendall RESEARCHOPPORTUNITIES FOR ROTORCRAFT. . . . . . . . . . . . . . . . . . . . . 537 Bruce Blake FIGHTER AIRCRAFT FLIGHT CONTROL TECHNOLOGY DESIGN REQUIREMENTS . . . . . . . . 549 W. E. Nelson, Jr.
MILITARY AIRCRAFT RESEARCH OPPORTUNITIES FOR THE FUTURE . . . . . . . . : . . . 559 Robert C. Schwanz OPPORTUNITIES FOR AIRCRAFT CONTROLSRESEARCH . . . . . . . . . . . . . . . . . 571 Thomas B. Cunningham PROPULSION CONTROLTECHNOLOGY. . . . . . . l . . . . . . . . . . . . . . . . . 585 Edward C. Beattie Vii PARTICIPANTS Imran Abbasy Ernest S. Armstrong c/o Brig Abbasy M/S 161 Cabinet Division Langley Research Center Rawalpindi Hampton, VA 23665 PAKISTAN Irving L. Ashkenas Irvin Abel Systems Technology, Inc.
M/S 243 13766 S. Hawthorne Boulevard Langley Research Center Hawthorne, CA 90250 Hampton, VA 23665 B. R. Ashworth Volkmar K. Adam M/S 125B 564 Logan Place,//7 Langley Research Center Newport News, VA 23601 Hampton, VA 23665 William M. Adams, Jr. Michael Athans M/S 152A M.I.T.
Langley Research Center Room 35-406 Hampton, VA 23665 Cambridge, MA 02139 Robert A. August W. J. Alford Grumman Aerospace Corp.
Dynamic Engineering, Inc.
703 Middle Ground Blvd. Mailstop TO3-05 Bethpage, NY 11714 Newport News, VA 23606 Carl Anderson A. J. Bailey, Jr.
Lockheed-California Company Honeywell 8509 Wyngate Street 7338 Cornelia Drive Sunland, CA 91040 Minneapolis, MN 55435 R. 0. Anderson Siva S. Banda Chief, Control Dynamics Branch AFWAL/FIGC Flight Control Division WPAFB, OH 45433 AFWAL/FIGC Wright-Patterson AFB, OH 45433 Lex Barker M/S 125B Willard W. Anderson Langley Research Center Hampton, VA 23665 M/S 152 Langley Research Center Hampton, VA 23665 Roger Barron General Research Corp.
Robert A. Andes 7655 Old Springhouse Road US Air Force McLean, VA 22102 ASD/BIEFT Wright Patterson AFB, OH 45433 James G. Batterson M/S 161 Ernie L. Anglin Langley Research Center M/S 267 Hampton, VA 23665 Langley Research Center Hampton, VA 23665 Robert I. Baumgartner Lockheed-California Co.
12129 E. El Dorado Avenue 115 P. Douglas Arbuckle M/S 152A Sylmar, CA 91342 Langley Research Center Hampton, VA 23665 Salvatore J. Bavuso Michael P. Bland M/S 130 McDonnell Douglas Corp.
Langley Research Center 66 Chartley Lane Hampton, VA 23665 Bridgeton, MO 63044 Gary P. Beasley Philip D. Bridges M/S 152 Mississippi State University Langley Research Center Department of Aerospace Engineering Hampton, VA 23665 Drawer A Mississippi State, MS 39762 Edward C. Beattie Pratt & Whitney W. Thomas Bundick Mail Stop 121-05 400 Main Street M/S 494 East Hartford, CT 06108 Langley Research Center Hampton, VA 23665 Henry L. Beaufrere Grumman Aerospce Corporation John A. Burns 15 Noel Court Air Force Office of Scientific Huntington, NY 11743 Research AFOSR/NM Chris Belcastro Bolling AFB, DC 20332 M/S 494 Langley Research Center Carey S. Ruttrill Hampton, VA 23665 Technology Applications, Inc.
21 Argall Place Hugh P. Bergeron Newport News, VA 23602 M/S 152E Langley Research Center Anthony J. Calise Hampton, VA 23665 Drexel University Mechanical Engineering & Mechanic8 Donald T. Berry Dept.
NASA Dryden Research Facility Philadelphia, PA 19104 43613:Lively Avenue Lancaster, CA 93534 Michael F. Card M/S 244 Paul W. Berry Langley Research Center The Analytic Sciences Corp. (TASC) Hampton, VA 23665 1 Jacob Way Reading, MA 01867 Preston I. Carraway Old Dominion University Gaudy M. Bezos Department of Electrical Engineering M/S 247 Norfolk, VA 23508 Langley Research Center Hampton, VA 23665 Mr. Edward S. Carter, Jr.
Director of Technology Bruce B. Blake Sikorsky Aircraft Division Boeing Vertol Company United Technologies Corporation P. 0. Box 16858 Stratford, CT 06602 Philadelphia, PA 19142 Charles R. Chalk Neal Blake Calspan Corp.
Deputy Associate Administrator P.O. Box 400 for Development f Logistics for Buffalo, NY 14225 Engineering Code AED- Phillip R. Chandler Federal Aviation Administration AFWAL/FIGL Washington, DC 20591 ASD/AFWAL/FIGL WPAFB, OH 45433 X S. C. Dixon Robert T. N. Chen NASA Ames Research Center M/S 398 Mailstop 211-2 Langley Research Center Moffet Field, CA 94035 Hampton, VA 23665 E. M. Cliff Rolf Duerr VPIdSU Sperry Corp.
3217 North Armistead Avenue Randolph 215 Blacksburg, VA 24061 ' Hampton, VA 23602 Roy B. Cotta John Edwards McDonnell Douglas Corp. M/S 341 M/S 36-81 Langley Research Center 3855 Lakewood Boulevard Hampton, VA 23665 Long Beach, CA 90846< Jarrell R. Elliott Leonard Credeur M/S 152A M/S 152E Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665 Eric T. Falangas Jeremiah F. Creedon Rockwell International M/S 469 3101 S. Bristol #140 Langley Research Center Santa Ana, CA 92704 Hampton, VA 23665 Philip Felleman Luci Crittenden C. S. Draper Laboratory Sperry 555 Technology Square M/S 125B Cambridge, MA 02139 Langley Research Center Hampton, VA 23665 Robert E. Fennel1 Clemson University Thomas B. Cunningham Dept. of Mathematical Sciences Honeywell Inc. Clemson, SC 29631 Systems and Research Center 2600 Ridgway Parkway Dr. Eliezer Gai - MS 60 P. 0. Box 312, MN17-2370 The Charles Stark Draper Laboratory Minneapolis, MN 55440 555 Technology Square Cambridge, MA 02139 Ralph L. Davies Dagfinn Gangsaas Calspan Corporation Boeing Company Buffalo, NY 16047 NE 27 Street Bellevue, WA 98008 Dwain Deets NASA Dryden Flight Research Facility H. D. Garner M/S D-OFD M/S 494 P. 0. Box 273 Langley Research Center Edwards, CA 93523 Hampton, VA 23665 Oded Degani William L. Garrard Lear Siegler, Inc.
Dept. of Aerospace Engineering Astronics Division University of Minnesota 3400 Airport Avenue 107 Akerman Hall Santa Monica, CA 90406 Minneapolis, MN 55455 Xi Kenneth R. Hall John F. Garren, Jr.
Mississippi State University M/S 469 Langley Research Center Dept. of Aerospace Engineering Hampton, VA 23665 Drawer A Mississippi State, MS 39762 Joseph Gera NASA Ames Research Center Nesim Halyo Information and Control Systems Box 273 28 Research Drive Edwards, CA 93523 Hampton, VA 23665 Dan Giesy Kentron M. J. Hamilton Sperry Systems Management Mail 161 3217 N. Armistead Avenue Langley Research Center Hampton, VA 23666 Hampton, VA 23665 Michael G. Gilbert Margery E. Hannah 1048 Willow Green Dr.
M/S 243 Langley Research Center Newport News, VA 23602 Hampton, VA 23665 P. W. Hanson William D. Grantham M/S 398 M/S 152A Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665 Herman J. Hassig Lockheed-California Company Robert Gregory Bendix Aerospace Technology Center P.O. Box 551 9140 Old Annapolis Road/MD108 Burbank, CA 91520 Columbia, MD 21045 Earl C. Hastings, Jr.
Arlene D. Guenther M/S 286 Sperry Systems Management Langley Research Center 3217 N. Armistead Avenue Hampton, VA 23665 Hampton, VA 23666 G. A. Haynes Wiley A. Guinn M/S 256 Lockheed California Company Langley Research Center P. 0. Box 551 Hampton, VA 23665 Burbank, CA 91520 Michael P. Healy Sol W. Gully Grumman Aerospace Alphatech Inc.
140 6th Street 2 Burlington Executive Center St. James, NY 11780 111 Middlesex Turnpike Burlington, MA 01803 Kenneth R. Henke Lockheed-California Company Dave Hahne P. 0. Box 551 M/S 355 Burbank, CA 91520 Langley Research Center Hampton, VA 23665 Larry S. Henry MPC Products Corporation Francis J. Hale 7426 North Linder Avenue N. C. State University Skokie, IL 60077 Box 5246 Raleigh, NC 27650 xii Michael Herb Jer-Nan Juang M/S 230 Analytical Mechanics Associate, Inc.
17 Research Drive Langley Research Center Hampton, VA 23665 Hampton, VA 23185 Ray V. Hood William Karger M/S 158 Rockwell International Langley Research Center NAAO-GB15 Hampton, VA 23665 Los Angeles, CA Jerry R. Karwac, Jr.
Dan T. Horak Sperry Systems Management Bendix Aerospace Technology Center 3217 N. Armistead Avenue 9140 Old Annapolis Road Hampton, VA 23666 Columbia, MD 21045 W. E. Howell Howard Kaufman M/S 494 Rensselaer Polytechnic Institute Langley Research Center ECSE Dept.
Hampton, VA 23665 Troy, NY 12181 G. M. Kelly Bob Huber Analytical Mechanics Association General Electric Company 17 Research Drive Binghamton, NY Hampton, VA 23666 Richard M. Hueschen James R. Kelly M/S 494 Langley Research Center M/S 265 Hampton, VA 23665 Langley Research Center Hampton, VA 23665 George Hunt Eric Kendall Sperry Systems Management Gates Learjet 3217 N. Armistead Avenue 1511 N. West Street, 69 Hampton, VA 23666 Wichita, KS 67203 Kenneth W. Iliff Ralph D. Kimberlin Code D-OF University of Tennessee Space Dryden Flight Research Facility Institute Edwards, CA 93523 103 Autumn Lane, Rt. 2 Tullahoma, TN 37388 Obi Iloputaife Northrop Corporation Martin J. Klepl 1 Northop Avenue Rockwell International M/S 3833185 2178 Paseo Del Mar Hawthorne, CA 90250 San Pedro, CA 90732 Pete Jacobs Charles E. Knox M/S 294 M/S 156A Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665esota S. M. Joshi Fred Lallman M/S 161 Langley Research Center M/S 152A Langley Research Center Hampton, VA 23665 Hampton, VA 23665esota xiii William E. McCain Antonius A. Lambregts Boeing CAC M/S 243 P. 0. Box 3707 Langley Research Center Seattle, WA 98124 Hampton, VA 23665 Beth Lee D. J. McFerron M/S 494 ,Old Dominion University Langley Research Center Dept. of Mechanical Engineering and Hampton, VA 23665 Mechanics Norfolk, VA 23508 Bruce Lehtinen NASA Lewis Research Center Duncan E. McIver NASA Headquarters 21000 Brookpark Road Code RTH-6 MS 100-l Washington, DC 20546 Cleveland, OH 44135 Brad Liebst Dr. Griffith J. McRee University of Minnesota Old Dominion University of Aerospace Eng. Department of Electrical Engineering Dept.
107 Akerman Hall Norfolk, VA 23508 110 Union Street, SE D. T. McRuer Minneapolis, MN 56455 Systems Technology, Inc.
Wai K. Lim 13766 S. Hawthorne Boulevard Hawthorne, CA 90250 Northrop Corporation 1 Northrop Avenue M/S 3833185 Walter C. Merrill Hawthorne, CA 90250 NASA Lewis Research Center 21000 Brookpark Road James F. Longshore MS 100-l Douglas Mail Code 36-49 Cleveland, OH 44135 3855 Lakewood Boulevard Long Beach, CA 90846 George Meyer M/S 210-3 John D. Louthan NASA Ames Research Center Director, Vehicle Projects Moffett Field, CA 94035 Vought Corporation P.O. Box 225907 R. T. Meyer Dallas, TX 75265 Lockheed Georgia Company 865 Cobb Drive William Lynn Marietta, GA 30065 Kentron Technical Center 3221 Armistead Avenue David B. Middleton Hampton, VA 23666 M/S 158 Langley Research Center Dale A. Mackall Hampton, VA 23665 NASA Dryden Flight Research Facility Code E-EDC Dr. Roland R. Mielke Edwards, CA 93523 Old Dominion University Department of Electrical Engineering Michael A. Masi Norfolk, VA 23508 AFWAL/FIGl ASD/AFWAL/FIGL Ernest W. Millen WPAFB, OH 45433 M/S 156A Langley Research Center Hampton, VA 23665 xiv Raymond C. Montgomery Boyd Perry M/S 161 M/S 243 Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665 Kurt Moses Lee Person Flight Systems Division Bendix Corp., M/S 255A Mailstop 218 Langley Research Center Teterboro, NJ 07608 Hampton, VA 23665 Martin T. Moul Richard H. Petersen 104 Yorkview Road M/S 103A Yorktown, VA 23692 Langley Research Center Hampton, VA 23665 Patrick C. Murphy M/S i52A W. H. Phillips Langley Research Center M/S 152A Hampton, VA 23665 Langley Research Center Hampton, VA 23665 Wallace Nelson Northrop Corporation Kenneth S.. Pollock Org. 3833-85 M/S 125B 1 Northrop Avenue Langley Research Center Hawthorne, CA 90250 Hampton, VA 23665 Jerry R. Newsom Joseph Post M/s 243 Sikorsky Aircraft Langley Research Center N. Main Street Hampton, VA 23665 Stratford, CT D. W. Nixon Anthony S. Pototsky General DynamicsIFt. Worth Division Kentron P. 0. Box 748, Mail Zone 2834 3221 N. Armistead Avenue Fort Worth, TX 76101 Hampton, VA 23666 Warren J. North David W. Potts NASA Johnson Space Center AFWAL/FIGL Houston, TX 77058 ASD/AFWAL/FIGL WPAFB, OH 45433 Aaron J. Ostroff M/s 494 Douglas B. Price Langley Research Center M/S 152A Hampton, VA 23665 Langley Research Center Hampton, VA 23665 Joe Pahle NASA Dryden Flight Research Facility Shepherd G. Pryor III P. 0. Box 273 Lockheed-Georgia Company Edwards, CA 93523 86 S. Cobb Drive Marietta, GA 30063 Robert Palmer Naval Air Development Center J. K. Ramage Warminster, PA 18974 WPAFB, OH 45433 Richard S. Pappa Perry N. Rea M/S 230 5 Willowood Drive, 6204 Langley Research Center Hampton, VA 23666 Hampton, VA 23665 xv W. H. Reed III Sue Sawyer Dynamic Engineering Inc. Sperry Systems Management 703 Middle Ground Blvd 3217 N. Armistead Avenue Newport News, VA 23606 Hampton, VA 23666 Dr. Herman A. Rediess Frederick W. Schaefer Vice President Systems Engr. Div. Grumman HR Textron, Inc. 729 Riviera 2485 McCabe Way Mastic Beach, NY 11951 Irvine, CA 92714 Mr. Richard Schoenman Philip A. Roberts The Boeing Company U.S. Air Force Commercial Airplane Company 9301 Harness Horse Court Box 3707 (M.S. 77-19) Springfield, VA 22153 Seattle, WA 98124 John D. Rollins Purdue Univeristy M/S 125B David K. Schmidt Langley Research Center Dept. of Aeronautics and Hampton, VA 23665 Astronautics West Lafayette, IN 47907 Harold M. Rosenbaum Ted Schulman Northrop Corporation Northrop 8900 E. Washington Boulevard 621 8th Street E743/1S Hermosa Beach, CA 90254 Pica Rivera, CA 90660-3737 Robert C. Schwanz Dr. J. Roskam Flight Mechanics Laboratory The University of Kansas Wright Patterson AFB Rt. 4 Box 274 WPAFB, OH 45433 Ottawa, KS 66067 Albert A. Schy Irving Ross M/S 161 Hydraulic Units, Inc.
Langley Research Center 1700 Business Center Drive Hampton, VA 23665 Duarte, CA 91010 John D. Shaughnessy Edmund G. Rynaski M/S 152E Flight Sciences Branch Langley Research Center Research Department Hampton, VA 23665 ARVINICALSPAN P.O. Box 400 Sahjendra N. Singh Buffalo, NY 14225 M/S 161 Langley Research Center Peter Sadler Hampton, VA 23665 Lockheed-Georgia Company Marietta, GA 30063 Charles A. Skira USAF Himankush Saha AFWAL/POTC 462 Lost Rock Drive WPAFB, OH 45433 Webster, TX 77598 Steven M. Sliwa Philip W. Saunders M/S 152A Northrop Corporation Langley Research Center 824 E. Grand Avenue, #7 Hampton, VA 23665 El Segundo, CA 90245 xvi Lawrence W. Taylor, Jr.
Capt. John D. Smith 6340 Americana Drive, Apt. 702 M/S 161 Clarendon Hills, IL 60514 Langley Research Center Hampton, VA 23665 Paul M. Smith Glen J. Tauke Kentron Technical Center Lockheed-California Company 3221 N. Armistead Avenue 8509 Wyngate Street Hampton, VA 23666 Sunland, CA 91040 Ronald H. Smith Sherwood H. Tiffany M/S 247 M/S 152A Langley Research Center Hampton, VA 23665 Langley Research Center Hampton, VA 23665 Richard K. Smyth Milco International .Leonard Tobias 15628 Graham Street NASA Ames Research Center Huntington Beach, CA 92649 Mailstop 210-9 Moffett Field, CA 94035 Cary R. Spitzer M/S 472 Ronald D. Toles Langley Research Center General Dynamics-Fort Worth Division Hampton, VA 23665 103 Valley Lane Weatherford, TX 76086 R. Srivatsan University of Kansas at Langley Robert H. Tolson M/S 494 M/S 103 Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665 George G. Steinmetz Charles N. Valade M/S 156A M/S 125B Langley Research Center Langley Research Center Hampton, VA 23665 Hampton, VA 23665 James F. Stewart Lena Valavani NASA Dryden Flight Research Facility L.I.D.S/M.I.T.
P. 0. Box 273 Rm. 35-437 ‘Edwards, CA 93523 Cambridge, MA 02139 William T. Suit Wallace E. Vander Velde M/S 161 M.I.T.
Langley Research Center 77 Mass Avenue Hampton, VA 23665 Bldg. 33-109 Cambridge, MA 02139 N. Sundararajan M/S 161 V. Variakojis Langley Research Center Douglas Aircraft Company Hampton, VA 23665 Mail Code 36-49 3855 Lakewood Boulevard David A. Tawfik Long Beach, CA 90846 Engineering Director, Flight System Division Marlen Varnovitsky Bendix Corporation Corporate R&D Center, General MC 2/10 Electric ,Teterboro, NJ 07600 Bldg. 5 Room 333C Schenectady, NY 12301 xvii Samuel L. Venneri Yuk K. Woo Langley Research Center U.S. Air Force 6510 TW/TEG Hampton, VA 23665 Mail Stop 236 Dan Vicroy Edwards AFB, CA 93523 M/S 156A Langley Research Center John H. Wykes Hampton, VA 23665 North American Aircraft Operation Rockwell International P.O. Box Robert von Husen Los Angeles, CA 90009 National Transportation Safety Board 1552 Coat Ridge Road Herndon, VA 22070 Thomas A. Waldeck Boeing Wichita 3517 E. Skinner Wichita, KS 67218 Raymond M. Wallace United Technologies Research Center Silver Lane E. Hartford, CT 06108 James F. Watson Engineering Incorporated 41 Research Drive Hampton, VA 23666 Joe Whiting Sperry Systems Management 3217 N. Armistead Avenue Hampton, VA 23666 Carol Wieseman M/S 243 Langley Research Center Hampton, VA 23665 Craig S. Willey Lockheed California Company P. 0. Box 551 Burbank, CA 91520 James L. Williams M/S 161 Langley Research Center Hampton, VA 23665 Charles M. Wilson U.S. Air Force 6510 TW/TEG Mail Stop 256 Edwards AFB, CA 93523 xviii NASA CONTROLRESEARCHOVERVIEW Duncan E. McIver NASA Headquarters Washington, DC First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 INTRODUCTION The intent of this presentation is to provide an overview of NASA research activities related to the control of aeronautical vehicles. A groundwork is laid by showing the organization at NASA Headquarters for supporting programs and providing funding. Then a synopsis of nany of the ongoing activities is some of which will be presented in greater detail elsewhere. A major presented, goal of the workshop is to provide a showcase of ongoing NASA-sponsored research.
Then, through the panel sessions and conversations with workshop participants,it is hoped to glean a focus for future directions in aircraft controls research.
OFFICE OF AERONAUTICS AND SPACE TECHNOLOGY'S GOAL The Office of Aeronautics and Space Technology,which sponsors most of the controls- oriented research for aircraft,publishes a long-range plan. The overall goal is Notice that the purpose of NASA's program in space and aero- stated in figure 1.
nautics is to continue to be long -term contributors toward the continued preeminence of the U. S. in civil and military aerospace activities.
STRENGTHEN THE AGENCY’S AERONAUTICS AND ADVANCED
SPACE R&T PROGRAMS AS EFFECTIVE, PRODUCTIVE, AND
LONG-TERM CONTRIBUTORS TOWARD THE CONTINUED
PREEMINENCE OF U.S. IN CIVIL AND MILITARY AEROSPACE
Figure 1 HIGH-PRIORITY TECHNICAL GOALS/THRUST Figure 2 shows the high-priority technical goals of OAST. It includes all the major thrusts proposed by OAST for the next 5 to 10 years.
Ones of speci- fic interest to the controls discipline are item 4, realize the full potential of advancing technologies for aircraft controls, guidance, and flight systems; item 7, provide technology to enhance flight management and crew effectiveness in aircraft operations and air traffic control systems; and, item 8, provide the technology base for exploitation of the use of modern computers in aeronautics. Other areas relate in terms of systems integration in an interdisciplinary nature. However, those three have generic and specific applications for aeronautical controls.
e
BRING EXTERNAL AND INTERNAL COMPUTATIONAL
FLUID DYNAMICS TO STATE OF PRACTICAL
APPLICATION TO AIRCRAFT AND ENGINE DESIGN
SIGNIFICANTLY REDUCE AIRCRAFT VISCOUS DRAG
OVER THE FULL SPEED RANGE AND IMPROVE THE
UNDERSTANDING OF REYNOLDS NUMBER EFFECTS
AT TRANSONIC SPEEDS
REALIZE THE FULL POTENTIAL OF COMPOSITE
MATERIALS FOR PRIMARY STRUCTURES IN CIVIL AND
MILITARY AIRCRAFT
REALIZE THE FULL POTENTIAL OF ADVANCING
TECHNOLOGIES FOR AIRCRAFT CONTROLS, GUIDANCE,
AND FLIGHT SYSTEMS
l
PROVIDE TECHNOLOGY ADVANCES TO EXPLOIT THE
FULL POTENTIAL OF ROTORCRAK FOR MILITARY AND
CIVIL APPLICATION
Figure 2
HIGH-PRIORITY
TECHNICAL GOALS
. PROVIDE TECHNOLOGY FOR AND FULLY SUPPORT THE
DEVELOPMENT OF ADVANCED MILITARY AIRCRAFT AND
MISSILE SYSTEMS
. PROVIDE TECHNOLOGY TO ENHANCE FLIGHT
MANAGEMENT AND CREW EFFECTIVENESS IN AIRCRAFT
OPERATIONS AND AIR TRAFFIC CONTROL SYSTEMS
. PROVIDE THE TECHNOLOGY BASE FOR EXPLOITATION
OF THE USE OF MODERN COMPUTERS IN
AERONAUTICS
. EXPLOIT THE FULL POTENTIAL OF HIGHLY INTEGRATED
PROPULSION AIRFRAME SYSTEMS
. ADVANCE THE TECHNOLOGY FOR SMALL TURBINE
ENGINES TO A LEVEL COMPARABLE WITH THAT OF
LARGE TURBINE ENGINES
. ESTABLISH THE TECHNICAL FEASIBILITY OF HIGH-
SPEED TURBOPROP PROPULSION
. PROVIDE COMPONENT TECHNOLOGY ADVANCES FOR
FUEL-EFFICIENT SUBSONIC TRANSPORT ENGINES
. PROVIDE SAFETY TECHNOLOGY FOR IMPROVED
DESIGN AND OPERATION OF CURRENT, ADVANCED
CIVIL AND MILITARY AIRCRAFT AND SYSTEMS
Figure 2 (Concluded) NASA ORGANIZATION The overall organization of NASA is shown in figure 3. Most of the aircraft controls research is performed through the Office for Aeronautics and Space The three field centers supported by OAST's program are Ames, Technology (oAsT).
Langley, and Lewis. Although diverse activities occur at all three centers, each center is usually charged with a number of lead roles for specific research thrusts.
ADMINISTRATOR DEPUTY ADMINISTRATOR
r
ASSOCIATE ASSOCIATE ASSOCIATE ADMINISTRATOR ADMINISTRATOR ADMINISTRATOR FOR SPACE FOR AERONAUTICS AND FOR SPACE TRACKING FLIGHT SPACE TECHNOLOGY & DATA SYSTEMS JOHN F. KENNEDY SPACE CENTER I NATIONAL SPACE TECHNOLOGY LABORATORIES I Figure 3 OAST ORGANIZATION The OAST organization at NASA Headquarters is depicted in figure 4.
Aircraft controls are sponsored inboth the Aerospace Research Division (Code RT) and in the Aeronautical Systems Division (Code RJ).
In code RT the research is usually of a general nature and could be applied to several vehicles or categories of vehicles.
The Controls and Human Factors Branch administers research programs for Applied Control Theory and Analysis, Flight Crucial/Fault-Tolerant Controls and Guidance, Spacecraft Controls and Guidance, Flight Management, Flight Simulation Technology, and Space Human Factors. Code RJ sponsors vehicle specific research in each of the indicated areas. Code RJ research often results in a wind tunnel or flight research test of a specific configuration. One way of viewing the organization is that as the generic research of code RT matures it is picked up by code RJ-type programs for validation and fine tuning for specific applications. If fundamental problems are encountered during the vehicle specific research of code RJ, it identifies an area for more effort for code RT.
ASSOCIATE AOMINISTRATOR DEPUTY ASSOCIATE AOMINISTRATOR ASSISTANT ASSOCIATE ADMINISTRATOR SPACE SYSTEMS AEROSPACE RESEARCH AERONAUTICAL SYSTEMS DIVISION
q IVISION I I OIVISION I I I
L L TRANSPORTATION SYSTEMS FLUID 6 THERMAL PHYSICS HIGH PERFORMANCE AIRCRAFT -A I I o Systrm Amlvtis l Comptitiond Fluid Ovnamia l Fighter Yrcmft 6 Missila a Orbiter Exprimut Prowam . Vitcoln Flow Control l Por*nd Lift WISTOLI Airsnft . Adnnad Chamiml ProPllbion l Exprimmtll Auodvnanda . Hvpnonic Vahidn brtb-to-orbit # Fluid Mwhanica of Turbomcbinny l Hi#h Pwformnsl Flight Rsrrsh On-Bard I Combustion 6 Hrt Tnndr . Orbiil Tnmfr . Test Twhniqw SUBSONIC AIRCRAFT I . Awothwmdvnmrda SPACECRAFT SYSTEMS l Aviation SlfRv I
MATERIALS 6 STRUCTURES 1 # Aircraft Flight Gvnanda
. Sysmm Andvsh l Aircraft Opntin~ Svstmm . Spwcmft Tlchnolow Expmimnts . Adnncad llktwial Conclpa . Landnor Flow Control l Lifr Radiction l Pdwnud Composite Synnns SYSTEMSTECHNOLOGY I l Compmita l Svstwm Intqntion Studies . A.r~dmtidty o Spwlab hvimdc . struaura 6 Dynomier ROTORCRAFT I Long Duration Expaua Facilitv l Analysis 6 Svnthab o Ion Auxilirv Ro~ubioa Svdom . A8rodvnrmia 6 Acpunia CONTROLS 6 HUMAN FACTORS I 0 Stmnuml Dynamics SPACE STATION SYSTEMS I Opardnm Problem I l Control Thwrv 6 Alrlvsit l Guidwau mtd Control . Syrtm AndvaN I Flight Crucial firm-aft Controts l Flight Rawch .0pmti0m 6 Guidance . Cradifr suppoll l Spndt Conbob 6 Guidwaca PROPULSION I l Flight kknp~mnt I Flipht Simulation TachnoloSv SPACE ENERGY CONVERSION I Turbomchinrv Componentt 6 lmtvunwntatlon . SPOC~Humn hctwt I Combuston . Inlm. Ducts, and Nozzlr COMPUTER SCIENCE 6 APPLICATIONS o ktwar Tramfr o Photovdmic Enwn Connnlon l Enpinn Llvrumics 6 Control . Elwxronla , Thrrml.to.Elwtrls Connniw I lntrmittwnt Combmtlon Enlist , somnn I row symms ~NpImmt a Automotiw Stirlinfi 6 Gas Turbina 6 Oistribtiion . Communiatiom 0 Computer Scitnsl ENERGY SYSTEMS I Autormion I l Computational Facilitin . Wind Ennn Twbndegv I on0 Svlton l Tmmtrid Fud Cdl Swtems l Elwtric 6 Hybrid Vnhidn Rarrch I Comnution Tachnolopv l Photovolmic Appliutiom Figure 4 AIRCRAFT CONTROLSAND GUIDANCE The three main areas of research sponsored by the Controls and Human Factors Office in the areas of aircraft controls and guidance are: applied control theory, flight-crucial systems, and flight path management and guidance (fig, 5). 'The goal of the applied control theory research is to provide the general tools for design- ing active control systems for many categories of aircraft. The flight-crucial systems research is attempting to develop analytical and mathematical models for ascertaining the validity and probability of failure of electronic active control systems and avionics in general. The overall goal is to provide a methodology for designing and verifying electronic active control systems which have the relia- bility of primary structural surfaces. The flight path guidance research program is aimed at providing fuel-eff,icient trajectories for commercial transports, time- optimal intercept guidance for tactical aircraft,and new enhanced display media for improving the cockpit environment.
1 SENSORS 1 1 LIGHTNING EFFECTS FLIGHT CRUCIAL SYSTEMS , Al?PLIED CONTROL THEORY FLIGHT PATH Figure 5 ACTIVE CONTROLTECHNOLOGY Active control research at NASA embodies many important elements (fig. 6). There is a significant effort involved in the development of synthesis tools for control laws which account for structural flexibility. Several different approaches are being investigated. The control system designs are then evaluated using a variety of analysis tools. Successful candidates are then tested in the wind tunnel or in flight. This process provides a validation of synthesis techniques, analysis tools, and experimental facilities. The goal of active controls research is to improve mission effectiveness by reducing weight, increasing performance, and enhancing passenger acceptance.
ACT -7; CONVENTIONAL f4- I’
T
I--
BENEFITS ACT FUNCTION 0 REDUCED WING LOADS 0 FLUTTER SUPPRESSION t l LOWER,WEI GHT 0 LOAD ALLEVIATION ANALYSIS 0 INCREASED SPAN 0 RIDE QUALITY l HIGHER L/D . STABILITY 0 DYLOFLEX AUGMENTATION 0 STABILITY I I A t I EXPERIMENTS 0 WI ND TUNNEL 0 DAST RESEARCH Figure 6 PARAMETERESTIMATION Parameter estimation is an important part of the control research program (fig.
Control system performance is greatly enhanced by having the most accurate 7 1,.
model of the aircraft possible. Modelling work includes linear and nonlinear analysis of general aviation, commercial transport, and tactical aircraft. Recent cooperative agreements for exchange of data and information with Boeing and with Israel serve to illustrate the importance such work has in the eye of industry and the role that NASA plays in this area.
iASlJF 9. 20 Figure 7 FAULT-TOLERANT COMPUTERS Research for fault-tolerant computers is the prime focus of the flight-crucial Two pioneering computer concepts have been construc- controls program (fig. 8).
ted. The SIFT (Software Implemented Fault Tolerance) computer and the FTMP(Fault- Tolerant Multi-Processor) computer are currently undergoing evaluation in AIRLAR It is hoped that the experience (Avionics Integration Research Laboratory).
gained with studying these concepts will provide an ultimate system reliability that has a probability of failure per flight hour of less than 10eg.
HIGHLY RELIABLE FLfGHT&&& PIONEERING. COMPUTERS ELECTRONIC FLIGHT ‘, x 1;;,:$ DEVELOPED ..,+7:~: CONTROL SYSTEMS.~~ . SIFT- SOFTWARE IMPLEMENTED FAULT TOLERANCE l FTM,P - FAULT-TOLERANT MULTIPROCESSOR lo-” SIFT/FTMP DEStGN GOAL 10-g F ESTIMATE0 OF FAILURE HOUR 10-a c- I. I 1 199c 1+70”1975 1990 1985 .
_ D&VICE SYSTEMS 5 IMPROVEMENT 7 ARCHITECTURE Figure 8 AIRLAB AIRLAB (fig. 9) is a new facility which brings online an impressive set of capabilities for performing fault-tolerant research.
System transient response and recovery rates are investigated during the injection of artificial faults at component, and functional levels.
the gate, Through the study of actual state-of- the-art concepts physically located at AIRLAB, the development of analytical models and emulation methods are a near-term objective.
With tools available for analyzing architectures and hardware component selection, design methodologies can be developed. Additionally, new efforts are underway to model and understand software reliability.
ANALYTICAL ASSESSMENT TECHNIQUES I, < BENEFIT!3 I ^.‘: ,, I, ,*, A”, o LAS FOR USE 8Y **, UNlYERSITiES/INDUSTRY ?
XII TECHNOiiktES I&, I, o CREDIBLE SAFETY, DESIGN PERFORMANCE AND COST PROOF o SIFT ASS&SMENTS o FTMP “^ o NEW T6CHNOLOGY GUIDELINES o OPTICAL DATA DISTRIE~TION “> FOR SYSTEM DESIGNERS ,I ” I ” o GENERIC DATA ON SYSTEM ^” CONFIGURATIONS Figure 9 CONTROLSAND GUIDANCE--FLIGHT PATH GUIDANCE In the area of flight path guidance (fig.
10) there is research in three major segments underway. Tools for computing optimal guidance laws for transport, V/STOL, rotorcraft, and tactical aircraft are being refined.
Current emphasis is on trying to integrate such concepts into the air traffic control (ATC) system and provide 4-D traffic flow management.
Additionally, there is an effort to develop advanced display media concepts for cockpits of the future.
&&
IA
OPTIMAL GUIDANCE .
< ADVANCED CONCEPTS SIMULATOR Frigure 1C - COCKPIT AVIONICS PROGRAM research 11) in the controls and guidance program Cockpit avionics (fig.
emphasizes the development of advanced display concepts. The prime users of this research will be the avionics manufacturers. Active areas of research include advanced display media, display generation techniques, data input/output technol- The goal of most research is toward thin ogy, and cockpit systems integration.
panels which are required for the "All-Glass Cockpit*' concept where electro- mechanical display devices are replaced by computer-generated images. It should be noted that the human factors research program at NASA is cooperating with industry in developing fundamental guidelines for deciding what should be displayed to the pilots.
Figure 11 VEHICLE SPECIFIC CONTROLSAND GUIDANCE TECHNOLOGY Up to now, most of the research that was described was sponsored by the Aerospace Research Division at NASA Headquarters and represents vehicle independent or generic developments. Figure 12 shows some of the vehicle specific research that is sponsored by the Aeronautical Systems Division at NASA Headquarters. Active controls research is performed in transport, general aviation, high-performance, V/STOL, and rotorcraft classes of airplanes. A natural development process would be the development of general design tools which are then used to synthesize con- trol systems for a specific vehicle. The direct application to a particular prob- lem and the experience gained can be used to refocus research in control theory as new challenges are presented.
HIGH’ PERFORMANCE” TRANSPORTS GENERAL AVIATION l FUGHT MANAGEMENT (TCV) l ACTIVE CONTROLS * SINGLE PILOT IFR l HIGH ANGLE-OF-AlTACk ADVANCED DlSPtAYS CONTROL :, * l PROPULiION/FLIGHT <, CONTROL ’ CONTROL MODES FOR : WEAPON SYSTEMS EFFECTIVENESS V/STOL ROTORCRAFT ‘j ,‘,?
l HANDLING QUALITIES/CONTROi l REMOTE SITE NAV/GUID CONCEPTS ~ ,.:k{ ., C” LAWS/DISPLAYS l ADV. COfilTROL/DISPl.AY CONCEi’TS ,..:p ,: ,.y,.-+’ . NAV/GUID SIMULATIONS <*s Figure 12 FLIGHT TEST OF ACTIVE CONTROLSTECHNOLOGY NASA has had an aggressive program in cooperation with industry to investigate and demonstrate the application of active controls technology to commercial trans- ports. One aspect of the program was a flight test demonstration of maneuver load alleviation and relaxed static stability on a Lockheed L-1011 (fig. 13). Analysis and piloted simulations of the technology were validated using the flight test results.
Figure 13 ROTORCRAFTCONTROLSRESEARCH A number of controls activities are being focused upon V/STOL aircraft and control rotorcraft (fig. 14). The nonlinear, inverse theory presented in this conference was sponsored with generic controls research money. Now the tools that have been developed are being used to design control laws which are being flown on the NASA Ames UH-1H. This is a good example of the ideal flow of NASA research in controls: theory enhancement; tool development; simulation and analysis; followed by verification and validation through flight test.
UH-IH IROQUOIS ~ AMES RESEARCH CENTER PRIMARY PURPOSE: .
. FLIGHT CONTROLS & AVIONICS l FLiGHT CONTROL LAWS l INVESTIGATk FLIGHT RELATED PROBLEMS ~ l iNVESTlGATE ROTOR FLlGHi PROBLEMS . UTILITY SUPPORT HELICOPTER ,‘A’ KEY CHARACTERISTICS: > ’ . TWO HELICOPTERS . ONE WITH VSTOLAND EGUlbME& 1’1” l ONE USED FOR ROTOR FLIGHT STtiDIkS . MEDIUM UTILITY HELICOPTER ,, L Figure 14 AFTI/F-16 ADVANCED TEHCNOLOGIES 15) has joint military and NASA funding.
The AFTI/F-16 program (fig. It has been a tremendous success in terms of demonstrating what can be accomplished through the aggressive use of integrated controls technologies. The addition of vertical canards and multi-purpose trailing-edge flans allowed the addition of new control modes including translational flight, nose pointing, and flat turns. Other sys- tems concepts have been evaluated including voice command, heads-up displays, and multi-purpose advanced*panel displays. Work is continuing with the evaluation of advanced combat and maneuvering systems taking advantage of the new operational control modes.
Figure 15 HIGHLY MANEUVERABLEAIRCRAFT TECHNOLOGY(HiMAT) 16) has recently been completed The HiMAT technology demonstration program (fig.
It was used to assess the effective- and was jointly funded by the USAF and NASA.
ness of integrated aircraft design with an emphasis on maximizing transonic maneu- Additionally, composites were vering without compromising supersonic performance.
used to aeroelastically tailor the lifting surfaces to allow deflection into opti- HiMAT is an RPV (Remotely Piloted Vehicle) mal aerodynamic shape under any load.
which permitted the use of risky and advanced technologies in a flight vehicle at greatly reduced cost.
-r.
/ ; ; ^ ..- ; .‘, ,.
.*
: ‘.
?
I I
‘1..
Figure 16 .
X-29A RESEARCH PROGRAM Another joint program between DARPA, USAF, and NASA is the X-29A Flight Research It is being used to investigate aeroelastically tailored, Vehicle (fig. 17).
It features a closely coupled canard, multiple forward-swept-wing technology.
longitudinal control surfaces, and a static margin .of 35 percent unstable.
Because of these features, it has become a major technical challenge for the con- When the joint military/NASA flight tests have been completed, trols specialists.
the aircraft will be retained by NASA and will be used for continuing flight research with respect to integrated control systems.
Figure 17 SUPERAUGMENTED AIRCRAFT CRITERIA The X-29 and Shuttle (fig. 18) are prime examples of highly augmented aircraft.
This level of what is sometimes termed "superaugmentation" results from the requirements for low observables, ultra-high maneuverability, and the extended flight envelope.
However, since the dynamic response of these vehicles is domi- nated by the control system characteristics, as opposed to conventional airplane dynamics, the interfacing with pilots becomes an issue.
Early results indicate that flying qualities criteria need to be reconsidered for this category of air- craft. Such criteria areimportant for being able to design effective controls for piloted vehicles.
Significant research envisioned for the future will undoubtedly be aimed at developing design guidelines through expanding the data base and constructing handling quality criteria for application superaugmented aircraft.
Figure 18 RESEARCH OPPORTUNITIES FOR THE FUTURE What NASA research will be performed in the future? The answer to this question is a major goal of this workshop. The papers presented should yield a good scope of what the current NASA-sponsored aeronautical controls program is.
The panel discussions and interactions with participants will hopefully serve to guide the future directions. Because of this, active participation by all attending the workshop is encouraged and is, in fact, essential for the continued United States preeminence in the area of aircraft controls.
SESSION I CONTROL SYSTEM DESIGN REQUIREMENTS Ray V. Hood Session Chairman FLYING QUALITIES CRITERIA FOR SUPERAUGMENTED AIRCRAFT Donald T. Berry NASA Ames Research Center Dryden Flight Research Facility Edwards, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 SUPERAUGMENTED AIRCRAFT Before proceeding with a review of superaugmented aircraft it would be prudent to define what we mean by super- activities, augmented aircraft. The term is defined below. Early applications of feedback control tended to enhance the basic static and dynamic stability of aircraft in a way that was equivalent to augmenting the basic aerodynamic stability derivatives. The resulting responses were improved but conventional. As basic aircraft stability levels became weaker and the augmentation became more elaborate, aircraft began to depart significantly from classical behavior.
Certain characteristics are highly typical of superaugmented aircraft and are also indicated below.
Definition
l Aircraft with flying qualities that are dominated by the
closed-loop control system rather than
aerodynamic stability
Dominant characteristics
l Large time delays
l Unconventional longitudinal response
l Small, effective roll time constants
AIRCRAFT PITCH RATE POLE/ZERO CONFIGURATIONS An' example of a difference between a classical and super- augmented aircraft is illustrated below in terms of pitch rate transfer functions in the s plane.
These diagrams are highly idealized but the basic features are quite representative. The classical aircraft typically is augmented primarily in damping so the closed-loop configuration is similar to an aircraft with good aerodynamic damping.
Superaugmentation is normally required for aircraft with a basic static instability. Proportional plus in- tegral compensation is typically used to stabilize the aircraft and provide good closed-loop frequency and damping. However, 'the- high gains required result in the conventional attitude numerator being cancelled by a basic aircraft pole and the effective attitude lead being determined by the zero of the proportional plus integral compensation.
Classical Superaugmented DRYDEN SUPERAUGMENTED AIRCRAFT FLYING QUALITIES RESEARCH Current activities of the Dryden Flight Research Facility that pertain to flying qualities of superaugmented aircraft are listed here in chronological order of their initiation. It can be seen that a variety of programs are underway. The highlights of these programs will be covered in the ensuing discussion. However, the descriptions will be brief because of the number of activities and the space available. This paper, then, will be an overview of Dryden superaugmented aircraft flying qualities research.
Program Implementation
F-8 DFBW Experiments
In house
Orbiter flying qualities In house, contractor (STI)
Shuttle FCS improvements
In house,
contractor (CALSPAN)
Nonconventional vehicle Grant - Purdue
flying qualities
AFTI/F-16 In house, Air Force
Fiying qualities and control Contractor study (STI)
system alternatives
Pilot model measurements Grant - U. Cal. Davis
VMS Shuttle evaluation In house, JSC
TIFS pitch rate criteria Contractor (CALSPAN)
F-8 DIGITAL FLY-BY-WIRE (DFBW) FLIGHT EXPERIMENTS The F-8 DFBW was the world's first fully fly-by-wire airplane.
Initial program emphasis was on system reliability and redundancy management. In recent years, the vehicle has been used to investi- gate flying qualities associated with advanced control laws and superaugmentation, as listed below.
The time delay studies inves- tigated the effect of transport delay on flying qualities in land: formation flying, iw3, and in-flight refueling. Highly augmented aircraft typically have large values of equivalent transport delay.
The nonlinear control law investigation was a cooperative program with the British and studied control laws that varied prefilter time constants as a function of feedback error and changed loop structure as a function of task. The PI0 suppression filters study (ref. 4) was an extension of concepts developed for the Space Shuttle.
l Time delays
l Nonlinear control laws (CADRE)
l PI0 suppression filters
OEX ORBITER FLYING QUALITIES EXPERIMENT This effort is under the sponsorship of the Orbiter Experiments Program (OEX). The purpose is to use Shuttle data and flight ex- perience to develop flying qualities criteria for next generation Shuttlecraft and to improve existing Shuttles where feasible. The Shuttle has some unique characteristics and mission tasks; neverthe- and there is much technology less, it is a superaugmented vehicle, transfer between it and high-performance aircraft.
l Generate flight data base for criteria for
current and future Space Shuttlecraft
l Establish flying qualities data “pipeline”
l Use flight data to validate analytic/simulator
studies
DRYDEN GRANT ACTIVITIES Grants under the direction of Dryden that pertain to superaugmented aircraft are outlined below.
The investigator for the Purdue grant is Dr. Dave Schmidt. The integrated pilot-optimal control synthesis simultaneously utilizes an optimal control pilot model and modern control theory to produce control system designs with optimum flying qualities. The optimal control approach to the Neal/Smith flying qualities criteria uses an optimal control pilot model instead of the classical pilot compensa- tion model. The pilot parameter identification techniques study is looking at the use of time series analysis to measure pilot dynam- and workload from flight-test time histories.
ics, strategy, Principal investigator for the University of California-Davis grant is Dr. Ron Hess. This effort is aimed at using existing pilot measurement techniques to obtain data from ongoing flight experi- ments and obtain a flight-validated data base of pilot math model Both classical and optimal control models will be used.
parameters.
Purdue University: (Schmidt)
Develop prediction techniques for flying qualities of
complex, nonconventional vehicles
l Integrated pilot-optimal control synthesis
l Optimal control approach to Neal/Smith
l Pilot parameter identification techniques
University of California - Davis (Hess)
Establish flight-validated pilot model data base
l Analyze F-8 DFBW flight experiments
AFTI/F-16 FLYING QUALITIES FLIGHT EXPERIENCE The AFTI/F-16 is a superaugmented aircraft with direct lift and side force control and a task-tailored multimpde flight control sys- tem. It is a very ambitious program that is striving to evaluate highly advanced control system mechanizations and architectures as well as unconventional control laws. Initial program emphasis has been on checkout of the digital flight control system and functional evaluation of the flight control concepts. Much qualitative flying qualities information has been obtained, and the highlights of this experience are indicated below.
l Utility of task-tailored flying qualities demonstrated
l Technology not available to optimize control modes
l PI0 and roll ratchet tendencies persist
AMES VMS SHUTTLE FLIGHT CONTROL IMPROVEMENT STUDY The Ames Vertical Motion Simulator (VMS) was used in a recent program to study Shuttle flying qualities in approach and landing.
This program was sponsored by the Johnson Space Center.
Dryden parti- cipated because of its background in previous Shuttle approach and landing studies. The program is outlined below.
Objectives
Evaluate proposed changes to Shuttle FCS to improve
flying qualities in approach and landing
Configurations
l Baseline Shuttle
l Shaped pitch rate
l Lead/lag prefilter
l Slapdown system
l Rate command/att hold
l Sink rate command
. c*
Results
l Shaped pitch rate, slapdown, and C* best of mods
l Slapdown and C* eliminated after aggravated maneuvers
due to rate limiting
l Pilots with extensive Shuttle training preferred baseline
system
l Test pilots without Shuttle training preferred shaped pitch
rate
TIFS PITCH RATE CRITERIA FLIGHT RESEARCH PROGRAM The Air Force/Calspan Total In-Flight Simulator (TIFS) very re- cently completed a program under joint Dryden/Langley sponsorship to investigate flying qualities criteria for pitch rate command systems.
The objectives of the program and configurations tested are outlined below. Preliminary results indicate that superaugmented configura- tions were rated level two or three.
Prefilters that tended to re- store more classical aircraft response improved the ratings.
Objectives
l Generate flight data base for improved pitch rate
command systems
l Emphasis on superaugmented aircraft
Configurations
200,000 lb class advanced aircraft
Negative static stability
Neutral static stability
Proportional + integral augmentation
With/without prefilters
Pitch rate augmentation
SUMMARY This review can be summarized as follows.
l Additional data needed to develop superaugmented aircraft
flying qualities criteria
l Dryden activities aimed at increased understanding of
superaugmentation and providing flight-validated data base
l Current effort involves F-8 DFBW, Space Shuttle,
AFTVF-I6 and X=29
BIBLIOGRAPHY 1. Radford;R. C., Smith, R. E., and Bailey, R. E., "Landing Flying Qualities Evaluation for Augmented Aircraft,' NASA CR-163097, Aug. '80.
2. Schmidt, D. K., and Innocenti, M., "Pilot Optimal Multivariable Control Synthesis by Output Feedback," NASA CR-163112, Jul. ‘81.
3. Teper, G. L. et al., "Analysis of Shuttle Orbiter Approach and Landing Conditions," NASA CR-163108, Jul. '81.
4. Powers, B. G., "An Adaptive Stick-Gain to Reduce Pilot-Induced Oscillation Tendencies,' Journ. of Guid., Contr., and Dynam., .-Apr. ‘82, pp. 138-142.
Mar 5. Weingarten, N. C., and Chalk, C. R., "In-Flight Investigation of the Effects of Pilot Location and Control System Design on Airplane Flying Qualities for Approach and Landing," NASA CR-163115, Jan. ‘82.
6. Bacon, B. J., and Schmidt, D. K., 'A Modern Approach to Pilot/ Vehicle Analysis and the Neal Smith Criteria,' AIAA Paper 82-1357, Aug. ‘82.
No.
7. Berry, D. T., "Flying Qualities: A Costly Lapse in Flight Control Design?," Astronautics and Aeronautics, vol. 20, no. 4, Apr. ‘82, 35,54-57.
PP.
8. Berry, D. T. et al., 'In-Flight Evaluation of Control System Pure Time Delays,' Journ. of Aircraft, vol. 19, no. 4, Apr. ‘82, 318-323.
PP.
9. Meyers, T. T., Johnston, D. E., and McRuer, D., 'Space Shuttle Flying Qualities and Flight Control Assessment Study,' Phase 1, NASA CR-170391, Jun. ‘82; Phase 2, NASA CR-170406; Phase 3, NASA CR-17040 10. Shafer, M. F., Smith, R. E., and Stewart, J. F., "Flight Test Experience with Pilot-Induced Oscillation Suppression Filters," AIAA Paper No. 83-2107, Aug. '83.
11. Larson, R. R., Smith, R. E., and Krambeer, K. D., "Flight Test Results Using Non-Linear Control With the F-8 Digital Fly-By-Wire Aircraft," AIAA Paper No. 2174, Aug. '83.
12. Weingarten, N. C., and Chalk, C. R., "Application of Calspan Pitch Rate Control System to the Space Shuttle for Approach and Landing,' NASA CR-170402, May '83.
LARGE AIRCRAFT HANDLING QUALITIES William D. Grantham NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 LONGITUDINAL HANDLING QUALITIES CRITERIA The point to be made here is that some of the present-day longitudinal handling qualities criteria for transport class aircraft do not apply to very large (G.w. = In fact, of the four criteria indicated here, 2,000,OOO lbf) transport aircraft.
of MIL-F-8785C could be said to be in only the short-period frequency requirements Moreover, agreement with the present very large aircraft simulation study results.
if it is conceded that the C-5A has satisfactory longitudinal handling characteris- none of these criteria are applicable to very tics, then it might be concluded that large aircraft.
@@ Large transports simulated C-5A simulated A (a) MIL-F-8785C .* _-I n.(b) AIAA Paper No. 65-780 r rllc. ICC (d) NASA CR-137635 I EFFECTIVE TIMF, DELAY tl IN COMMAND PATH (PILOT QW = 1.5 IUD/SEC) These pitch rate response criteria were developed by Chalk (NASA CR-159236) and "effective time delay "transient peak ratio address such parameters as (q);" "effective rise time (At=t2-tl);" and apply to the dynamic (Aq2/Aql);" and response with the pilot in the loop. This table indicates that the effective time delays of the simulated large transports meet the requirements of this reference for pitch, roll, and yaw. Also, although it is not indicated in this chart, the pitch transient peak ratio requirement was met for all large transports (Aq+q 1 simulated.
None of the large aircraft simulated met the suggested pitch requirements for At (effective rise time parameter). This is quite disconcerting since the referenced limits on At are derived from or related to the constant limits on w,2/n/a used in MIL-F-8785C, and as shown earlier in thefirst figure of this presentation, all of the simulated large aircraft met the level 1 requirement for w,2/n/a.
It should be noted that NASA CR-159236 lists no requirements for "transient peak for the roll and yaw axes.
ratio" or "effective rise time" TANGENT AT MAX SLOPE t, - INTERSECTION OF MAX. SLOPE LINE AND ZERO AMPLITUDE t2 - INTERSECTION OF MAX SLOPE LINE AND STEADY STATE A92 - N TRANSIENT PEAK RATIO *q1 w EFFECTIVE TIME DELAY A92 '1 I A At - EFFECTIVE RISE TIME Large Aircraft Simulation Results Requirements-NASA CR-159236 Augmented Ave Pitch Roil .133 .120 I - COMPARISCNOF SIX LAHOS CONFIGURATIONS TO THE SPACE SHUTTLE SIJBSONIC PITCH RATE REQTJIBEMENTS Because the Shuttle is always operated as a closed-loop system, the conventional ML-F-8785C open-loop aircraft modal format for flying qualities was considered to be inappropriate. Instead, Shuttle pitch axis flying qualities were specified in the time domain by the response boundaries indicated in this chart.
However, the Shuttle specification itself does not correlate well with much of the recent flying qualities experimental data. For example, some selected LAHOS configurations (AFFDL-TR-78-122) were compared to the Shuttle criterion.
Note that it was possible to select some LAHOS configurations that exceeded the boundaries and yet had good (level 1) flying qualities, while others that met the requirements had poor (level 2) flying quali- ties. (Similar results were found for correlations of the Neal and Smith data of AFFDL-TR-70-74.)
Pilot Rating LAHOS
I
4-4 Level 2 Level 2 2.5 - Time, set Time, set (b) LAHOS configurations satisfying the (a) LAHOS configurations which do not meet Shuttle pitch-rate requirements.
the Shuttle pitch-rate requirements.
COMPARISONOF SIX LAHOS CONFIGURATIONS TO THE SPACE SHUTTLE SDBSONIC PITCH RATE ENVELOPE, BTJT USING NORMALIZED o! RESPONSE It has been suggested that the original Shuttle time-history envelope was .developed for angle of attack instead of pitch rate. The figure on the left of this chart shows the a response of LAHOS configurations 2-1, 4-C, and 3-C plotted in the Shuttle time-history response envelope. The responses now fall approximately within the Shuttle envelope with level 1 flying qualities.
The figure on the right of this chart shows a plot of the a responses of UHOS configurations 4-0, 4-3, and 4-4 on the same Shuttle time-history envelope, and all three configurations have level 2 flying qualities. Although the pitch rate responses of these three configurations were shown to. be within the Shuttle envelope in the previous figure, the angle-of-attack responses shown here indicate a very sluggish, unacceptably responsive vehicle.
* CALSPAN suggests that normalized u should be used instead of normalized pitch rate.
Pilot Rating LAHOS Config. Overall Approach 4-o Level 2 0 4-3 Level 2 Level 1 1 4-C 1 Level 11 Level 11 4-4 Level 2 Level 2 y 2.0 : :: 4: ", 1.5 a, ," 4: 1.0 -0 .F -.E .5 k z I $: 7-L I J 0 1 2 3 4 5 0 1 2 3 4 Time, set Time, set (a) Configurations which satisfy Shuttle 6 (b) Configurations which do not satisfy envelope when a is used. Shuttle i envelope when a is used.
COMPARISON OF LARGE TRANSPORTAIRCRAFT SIMULATED TO SHUTTLE PITCH RATE CRITERION ENVELOPE, BUT USING NORMALIZED a RESPONSE This chart presents a comparison of the "angle-of-attack" response for the simu- lated large transport aircraft to the "pitch rate" response criterion developed for the Space Shuttle.
These large aircraft do not correlate wsll with the Shuttle criterion, even when normalized r is substituted for normalized 8. The figure on the right presents the augmented dynamic response for four large aircraft configurations, all of which were assessed by the pilots as having satisfactory (level 1) approach and landing flying qualities. However, when compared to the Shuttle time-history envelope, it would be sluggish responses, concluded that these large aircraft had unacceptably which was not the case.
Large A/C Pilot Rating, Pilot Rating, Large A/C Landing Task Config. Landing Task Config.
1 1 1 Level 1 1 1 Level 2 Level 1
2 Level 2 I 2 I I
Level 1
3 Level 2 I 3 I I
4 Level 1 Level 1
I 4 I I
y 2.0 "u 2.0 am : =: =: u y- 1.5 : 1.5 0 0 a, aJ $1.0 2 1.0 24 -0 : 13 -0 N" = .5 = .5 E E 0' fi z z
I
0 0 0 1 2 3 4 5 0 1 2 3 4 5 Time, set Time, set (a) Unaugmented large transports simulated.
(b) Augmented large transports simulated.
- COMPARISONOF LARGE TRANSPORT SIMULATED AIRCRAFT TO BOEING PITCH RATE REQUIREMENTS The low-speed pitch rate response criterion indicated in this chart and reported in NASA CR-137635 was developed by the Boeing Company for application to the han- dling qualities requirements for supersonic transports. It should be noted that this Boeing criterion differs from the Shuttle pitch rate response criterion presented earlier in that this Boeing criterion allows for much more pitch rate overshoot, and allows for much less initial pitch rate delay.
Upon comparing this pitch rate response and pilot opinion of the very large "subsonic" jet transport of the present ground-based simulation study, it can be seen reasonably well with the Boeing- that the simulated large aircraft results agree developed SST landing approach criterion. Indications are, however, that the minimum satisfactory level of "initial" pitch rate response allowed by this criterion could probably be relaxed for very large (G.W. = 2,000,OOO lbf) transport aircraft.
Large A/C Pilot Rating, Config. Landing Task 1 Level 1 2 Level 1 3 Level 1 -- 4 Level 1 1,r. IN I,-. LIL (a) Unaugmented large transports simulated. (b) Augmented large transports simulated.
COMPARISONOF SHORT-TERM PITCH RESPONSEOF SIMULATED LARGE TRANSPORTAIRCRAFT WITH CATEGORYC REQUIREMENTSOF PROPOSEDMIL HANDBOOK The short-period frequency requirement of 8785C was based upon the premise that is a primary factor affecting the normal acceleration response to attitude changes the pilot's perception of the minimum allowable wsp [that is, limits are placed on Likewise, the physical interpretation of the so-called "control wsp2/(n/a)l.
anticipation parametern [CAP = wsp2/(n/a>] assumes that the dominant concern for a pilot pitch control input is normal acceleration response.
also true that the pitch attitude response to pitch control It is, of course, inputs is of paramount importance, and, whether the appropriate correlating parameter is n/a or 1/Tg2 is a moot point in that data that correlate with l/To2 gener- However, it was observed in AIAA Paper No. 69-898 that ally also correlate with n/a.
the product ~~~~~~ provided a slightly better correlation than CAP. (Physi- cally, represents the separation in phase between aircraft response in %pT02 path and pitch attitude.)
Thus, the combination with criterion of AFWAL-TR-82-3081 %pTe hip* i?' in is presented in this c art along with the characteristics of the simulated very large opinion of these configura- aircraft of the present study. And, since the pilots' approach and landing were, in general, level 2 when unaugmented and tions during it is concluded that the results of the present 6-DOF ground- level 1 when augmented, simulator results are in good agreement with this based WspT02 vs. csp criterion.
NOTE: l/T ; n,(g/V) IN THIS INSTANCE % Large transports simulated C-5A simulated w T
WspTe2
sP 02 H LEVEL ) CLASS .
4SP 4SP (a) Large aircraft, unaugmented. (b) Large aircraft, augmented.
COMPARISONOF SIMULATED LARGE TRANSPORTAIRCRAFT TO LOCKHEED C-5A AND BOUNDARIES FROM FLIGHT TEST The boundary indicated as taken from TN D-7062 was derived using a general purpose airborne simulator (Lockheed Jetstar) with a model-controlled, variable- stability system installed to provide simulation capability. This boundary presents the pilot ratings (PR) for the maximum "roll acceleration" commanded by the pilots for the various roll time constants investigated. The boundary indicates that a roll acceleration capability of approximately 0.12 rad/sec2 or greater was consideredtobe satisfactory (PR < 3.5) by the pilots; and that the pilot ratings rapidly became unacceptable (PR > 6.5) when the roll acceleration capability was decreased below 0.10 rad/sec2.
The boundary indicated for the large transports (G.W. = 2,000,OOO lbf) simulated in the present study (on a 6-DOF ground-based simulator) indicates that a roll accel- eration capability as low as 0.09 rad/sec2 was evaluated as being satisfactory (PR < 3.5). Similar piloting tasks were used in both studies.
Note that the C-5A ground-based simulation results indicate a pilot rating of 4.0 (level 2) for the lateral-directional handling qualities and yet the roll accel- eration capability was greater than 0.2 rad/sec2, indicating that the roll control power was not the reason for the level 2 lateral-directional pilot rating.
w Large transports simulated A C-5A simulated PILOT RATING TND-7062 I I I I I I I I I I ; .04 .06 .08 .l .2 .3 .4 .6 .8 1.0 rad/sec2 L6 'a a max' COMPARISONOF RESULTS FROM LARGE AIRCRAFT SIMULATIONS WITH REFERENCEDRESULTS This chart relates pilot opinion of an aircraft's roll response to the parameter (The term $1 is defined as the maximum bank angle that can be achieved in one $1.
second.) The results of the present large aircraft simulation study (6-DOF ground- based simulator) are compared to the results of TN D-7062 and CR-635 (both reporting results obtained from airborne simulators).
None of these simulation results agrees as to the minimum satisfactory level (PR The results from TN D-7062 indicate that $1 must be greater than < 3.5) of $1.
approximately 6" for level 1 roll response; the results of CR-635 indicate that $q must be greater than approximately 3" for level 1 roll response; and the results from the present ground-based simulation study indicate that the $11 for very large transport aircraft could be as low as approximately 1" and still be considered to have satisfactory (level 1) roll response.
~~~ Large transports simulated, ground-based A C-5A simulated NASA CR-635 (367-80 air- borne simulator) 6- airborne simulator) B- 9- 10 I I I 1 .3 1 3 10 30 q. DEG VARIATION OF PILOT RATING WITH BANK ANGLE ATTAINED IN THE FIRST SECOND Maximum bank angle in the first second after initiation of wheel deflection has been suggested as a figure of merit for roll control systems. The ($1 max) variation of pilot rating with +l,, for the ground-based simulator results reported in CR-635 are indicated in this chart to be a function of effective wheel angle, 6 (The term 6weff is defined as the wheel angle for maximum rolling Weff' moment.)
These indicated lines of constant effective wheel angle suggest that the pilot is rating the bank angle per wheel deflection or roll response sensitivity, more so, or instead of, the parameter $lmax. Another interesting point to be seen from this chart is that a constant pilot rating of 3 (level 1) was obtained at the con- stant value of $1 max/bw = 0.1, while @l,, varied from 3' (6weff = 30') to 9" UWeff = 90"). Note aLso the results of the present large aircraft simula- tion study.
Although values of $lmax of these large aircraft configurations were much smaller than those of the referenced data, the large aircraft 6weff was also smaller and the overall results were the same; a pilot rating of 3 was obtained when (p1/6, = 0.1 (f$lmax pJ 1.5' for gweff = 15").
-------- 50 NASA CR-635 \ --- 75 i i -- -- 90 i Large aircraft results -.-.-.-.-'-.- 15 i i 0 1 2 3 4 5 6 7 8 9 10 BANK ANGLE IN THE FIRST SECOND, 9, , DEG max COMPARISONOF ROLL PERFORMANCE FROM LARGE AIRCRAFT SIMULATED WITH REFERENCEDRESULTS An inadequate "large aircraft data base" has led to handling qualitfes specifi- and, as a result,there is a risk that future aircraft will be over- cation problems, unnecessarily expensive, or possibly inadequate to perform the design mis- designed, sion. For example, considerable effort and expense were initially expended on the C-5A in an attempt to meet a requirement for rolling to an 8" bank angle in one It was later determined from flight tests that the handling qualities of the second.
C-5A were totally acceptable with less than one-half such roll capability.
All four of the figures on this chart relate pilot opinion to the time required Figure (a) shows C-5A roll performance compared to the 8785C to bank 30" (t,+300).
Although this aircraft is considered to have satisfactory roll perfor- requirement.
mance, it would be evaluated as less than satisfactory by the military specification criterion. Boeing suggested a few years ago that the t =3Co criterion should be a function of aircraft landing weight (fig. (b)). Severa ! aircraft in service today meet this criteria but do not meet the MIL-SPEC criteria. Extrapolation of the Boeing criteria indicates that the t$=3Go requirement should be relaxed for heavier Class III aircraft.
Current results of the ongoing large aircraft simulation study are summarized in figures (c) and cd). Results shown in figure (c) indicate that a t,+3Co of less than 6 set should result in "acceptable" roll response characteristics, and that a of less than 4.0 set results in "satisfactory" roll response.
Figure (d) t9=30° indicates that the present large aircraft ground-based simulation results are in good agreement with the airborne simulation results of TN D-7062, wherein smaller Class III aircraft were simulated.
SUMMARY - The short-period frequency requirements of MIL-F-8785C are applicable to the very large transport aircraft simulated.
- The large aircraft simulated in this study meet the requirements of NASA CR-159236 for effective time delay and pitch transient peak.ratio.. However, the requirements of this reference for the effective rise time parameter are believed to be too conservative for very large transport aircraft.
- These large aircraft simulation results are in very good agreement with the criterion of AFWAL-TR-82-3081.
%&32 vs. Csp - A value of the parameter LsA6bax, which is an indication of the roll accleration capability, as low as 0.09 rad/sec2 was considered to be satisfactory for the very large transports simulated. This compares to a value of approximately 0.12 rad/sec2 desired for smaller transports.
- A minimum satisfactory level of the parameter $1 was determined to be much lower for the large aircraft simulated in this study compared to the values determined in previous studies for smaller transport aircraft.
However, the magnitude of required for these large transports was determined to be the same as that $1/6w required for smaller transports; thus, produces satisfactory roll 91&q = 0.1 characteristics.
- Data obtained to date as well as other data indicate that MIL-SPEC requirements for the parameter t,+=300 are too conservative for very large transport aircraft. The results of the present study indicate that a t+=300 of less than 6 set should result in "acceptable" roll response characteristics, and a t,+=300 of less than 4.0 set should result in "satisfactory" roll response.
BIBLIOGRAPHY 1. Military Specification Flying Qualities of Piloted Airplanes, MIL-F-8785C, Nov. 1980.
2. Shomber, H. A.; and Gertsen, W. M.: Longitudinal Handling Qualities Criteria: An Evaluation. AIAA Paper No. 65-780, Nov. 1965.
Aerospace Recommended Practice: Design Objectives for Flying Qualities of Civil 3.
Transport Aircraft. ARP 842B, Sot. Automat. Eng., Aug. 1, 1964. Revised Nov. 30, 1970.
4: Sudderth, Robert W.; Bohn, Jeff G.; Caniff, Martin A.; and Bennett, Gregory R.: Development of Longitudinal Handling Qualities Criteria for Large Advanced Supersonic Aircraft. NASA CR-137635, March 1975.
5. Chalk, C. R.: Recommendations for SCR Flying Qualities Design Criteria.
NASA CR-159236, April 1980.
6. Smith, Rogers E.: Effects of Control System Dynamics on Fighter Approach and Landing Longitudinal Flying Qualities. AFFDL-TR-78-122, vol. 1, March 1978.
7. Neal, T. Peter; and Smith, Rogers E.: An In-Flight Investigation to Develop Control System Design Criteria for Fighter Airplanes. AFFDL-TR-70-74, vol. 1, Dec. 1970.
8. Ashkenas, I. L.: Summary and Interpretation of Recent Longitudinal Flying Qualities Results. AIAA Paper No. 69-898, Aug. 1969.
9. Hoh, Roger H.; Mitchell, David G.; Ashkenas, Irving L.; Klein, Richard H.; Heffley, Robert K.; and Hodgkinson, John: Proposed MIL Standard and Handbook- Flying Qualities of Air Vehicles, vol. 2: Proposed MIL Handbook.
AFWAL-TR-82-3081, vol. 2, Nov. 1982.
10. Hollernan, Euclid C.; and Powers, Bruce G.: Flight Investigation of the Roll Requirements for Transport Airplanes in the Landing Approach.
NASA TN D-7062, Oct. 1972.
11. Condit, Philip M.; Kimbrel, Laddie G.; and Root, Robert G.: Inflight and Ground-Based Simulation of Handling Qualities of Very Large Airplanes in Landing Approach. NASA CR-635, Oct. 1966.
A SUMMARYOF ROTORCRAFT HANDLING QUALITIES RESEARCHAT NASA AMZS RESEARCH CENTER Robert T. N. Chen NASA Ames Research Center Moffett Field, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 OBJECTIVES AND SCOPE The objectives of the rotorcraft handling qualities research program at Ames Research Center, as shown in figure 1, are twofold: (1) to develop basic handling qualities design.criteria to permit cost-effective design decisions to be made for helicopters, and (2) to obtain basic handling qualities data for certification of new rotorcraft configurations. The research on the helicopter handling qualities criteria has focused primarily on military nap-of-the-Earth (NOE) terrain flying missions, which are flown in day visual meteorological conditions (VMC) and instrument meteoro- logical conditions (IMC), or at night. The Army has recently placed a great deal of emphasis on terrain flying tactics in order to survive and effectively complete the missions in modem and future combat environments. Unfortunately, the existing Military Specification MIL-H 8501A (ref. l), which is a 1961 update of a 1951 document, does not address the handling qualities requirements for terrain flying. The research effort is therefore aimed at filling the void and is being conducted jointly with the Army Aeromechanics Laboratory at Ames. The research on rotorcraft airworthiness standards with respect to flying qualities requirements has been conducted in collaboration with the Federal Aviation Administration (FAA). This effort focused, in the recent on helicopter instrument flight rules (IFR) airworthiness criteria and is now past, addressing the airworthiness concerns for such new rotorcraft configurations as tilt-rotor aircraft.
OBJECTIVES SCOPE DEVELOP FLYING QUALITIES AND . NOE OR TERRAIN FLIGHT IN VMC AND CERTIFICATION CRITERIA IMC/NlGHT (ARMY) . TERMINAL AREA IFR OPERATIONS (FAA) .
NOE @ CONTOUR @ LOW LEVEL @ Figure 1 FACTORS INFLUENCING AGILITY FOR TERRAIN FLIGHT In terrain flight, the pilot is often called upon to fly complicated and rapidly changing trajectories to avoid obstacles and to unmask and quickly remask.
The characteristics of the helicopter that permit the pilot to fly these complex trajectories quickly, precisely, and easily are essential to safe and successful operations, and we may define this aggregate of characteristics as agility. Factors influencing agility are many: basic performance potential of the aircraft, engine/ governor dynamics, stability and control characteristics, and cockpit interface (fig. 2). the helicopter must be able to change rapidly the magnitude For quickness, and direction of its velocity vector. Adequate control powers in pitch, roll, and yaw are required for a quick rotation of the thrust vector; adequate installed power and responsiveness of the engine/governor system together with adequate rotor thrust For capability are needed to meet the demand for rapid changes in thrust magnitude.
precision and ease with which the pilot flies those complex trajectories, the heli- copter must have good stability and control characteristics. Interaxis coupling must be minimized so that unnatural or complicated control coordination is not required. Also, proper controller characteristics, flight director displays, and vision aids are needed to assist the pilot in flying the missions in adverse weather conditions or at night.
AGILITY: THE QUALITIES PERMITTING PILOTS TO FLY COMPLEX TRAJECTORIES QUICKLY, PRECISELY, AND EASILY QUICKNESS 0 ABILITY TO RAPIDLY CHANGE MAGNITUDE AND DIRECTION OF AIRCRAFT VELOCITY VECTOR . INSTALLED POWER, RESPONSIVENESS OF ENGINE/GOVERNOR SYSTEM . ADEQUATE CONTROL POWER IN PITCH, ROLL, YAW . ADEQUATE ROTOR THRUST CAPABILITY EASE AND 0 GOOD COMBINATIONS OF STABILITY AND CONTROL, AND PRECISION ADEQUATE PILOT AIDS l ADEQUATE DAMPING, PROPER CONTROL SENSITIVITY . SMALL INTERAXIS COUPLING . ADEQUATE STABILITY . PROPER COCKPIT INTERFACE AND PILOT AIDS - CONTROLLER CHARACTERISTICS - DISPLAYS (IMC) - VISION AIDS (NIGHT) Figure 2 EFFECT OF ENGINE DYNAMICS ON HANDLING QUALITIES The effects of engine dynamics and thrust-response characteristics on helicopter handling qualities have until recently remained largely undefined.
A multiphase pro- gram is being conducted to study, in a generic sense, the effects of engine response, rotor inertia, rpm control, excess power, and vertical sensitivity and damping on helicopter handling qualities in hover and representative low-speed NOE operations.
three moving-based piloted simulations have been conducted on the Vertical To date, Motion Simulator (VMS) at Ames. This series of investigations concentrates specifi- cally on the helicopter configuration with an rpm-governed gas-turbine engine. It was found (ref. 2) that variations in the engine governor response time can have a significant effect on helicopter handling qualities as shown in figure 3.
For the tasks evaluated, satisfactory handling qualities and rpm control were achieved only with a highly responsive governor (which for the model in the study was wn > 7 radlsec). The results indicate that for satisfactory handling qualities, there is a qualified trade-off between engine response time and vehicle vertical damping; however, increases in engine time constant are limited by poor rpm overspeed and underspeed control.
40 knot QUICK STOP DOLPHIN BOB UP AND DOWN EQUIVALENT TENG < 0.2 FOR 6.5 5;3 AVERAGE COOPER- 1 1 .oo SATISFACTORY .
1.a I- HARPER RATING / III / I- 6.5 5.5 .
.
- OPTIMIZED I’ l i- ,,I’ . #- .50 II 3.2 4.0 4.5 $1’ 3.7 g .
Ii. .,w.25 ’ ,I’ SATIS- I’ FACTORY IDEAL ENGINE SATISFACTORY P.R. = 3.0 3.0 2, = -0.65 set-l 0 Z 4 6 8 10 0 -2 -4 -6 w,,, radlsec Z,, set-’ Figure 3 EFFECT OF EXCESS THRUST, ROTOR INERTIA, AND RPM CONTROL ON HELICOPTER HANDLING QUALITIES The excess power requirements (T/W) for the NOE tasks were investigated with various levels of vehicle vertical damping Z,.
Results indicated that the required level of T/W is a strong function of Zw as shown in figure 4, and is minimized ata Z, value around -0.8 rad/sec. In addition to the required engine response time (as previously shown in fig. 3), an excess power level of T/W=l.l is required to achieve satisfactory handling qualities for the bob-up task.evaluated. The thrust response of a helicopter, unlike that of fixed-wing VTOL aircraft, is influenced by several factors, including (1) engine governor dynamics, (2) vertical damping result- ing from rotor inflow, and (3) the energy stored in the rotor, which is a function of rotor inertia. The experimental results (ref. 3) indicate, however, that increases in rotor inertia (thus the stored kinetic energy) have only a minor and desirable effect on handling qualities. The effect on handling qualities of requirements for pilot monitoring and control of rotor rpm can be significant. For a slow engine governor, the degradation in pilot rating in the bob-up tasks was as much as two ratings. It may therefore warrant consideration of techniques to relieve the pilot of the task and concern for monitoring proper rpm.
(BOB-UP TASK) Z, vs T/W ADJUSTED L2 83300 EFFECTS OF INERTIA lo- EFFECTS OF RPM CONTROL (Z,= -0.25sec-'1 Jp = 2000 slug ft2 is g- Z,= -0.25sec-' $ IO- $ 8- SLOW-RESPONSE F DEGRADATION DUE 0 rpm CUEING DEGRADATION DUE TO GOVERNOR DYNAMICS INCREASING 7RT(l/wn) - 6 I I I I I I 1 ’ 1 8 O 200 1000 2000 3000 4000 IDEAL FAST INTER- SLOW MEDIATE Jp, slug ft2 ENGINE - GOVERNOR RESPONSE Figure 4 EFFECTS OF INTERAXIS COUPLING ON HELICOPTER HANDLING QUALITIES One unique characteristic of the handling qualities problems associated with the single main rotor helicopter is the cross'coupling between the longitudinal- and lateral-directional motions such as yawing, rolling, and pitching moments due to col- lective pitch input, and the pitch-roll cross coupling caused by aircraft.angular Recent design trends of augmenting control power with rate in pitch and roll.
increased flapping hinge offset or with a stiffened flapping hinge to increase low-g maneuverability can aggravate those undesirable interaxis cross couplings. To quan- tify their influences on the handling qualities , piloted simulation experiments (refs. 4, 5) and a flight experiment have been conducted (ref. 6). Some of the results are shown in figure 5, which indicates the trends of pilot rating as influ- enced by the level of yawing moment due to collective input Ng,, the ratio of pitching moment caused by collective input to pitch damping MBc/Mq, and the ratio of the rolling moment caused by pitch rate to roll damping Means of reducing Q/Q.
the interaxis coupling through either proper selection of rotor system design param- eters or use of stability and control augmentation systems to improve the handling qualities have also been investigated (ref. 7).
COLLECTIVE INPUT COUPLING PITCH-ROLL CROSS COUPLING MOVING-BASE SIMULATOR m FIXED-BASE SIMULATOR . AVERAGE PILOT RATING A FLIGHT ti AVERAGE PILOT RATING, Ng = 0 --- HUSTON/WARD CRITERIA c c E *t 0 .l .2 .3 -.8 -.6 -2 0 .2 .4 .6 .8 -.4 -M6,/Mq, radhclin LdLP Figure 5 I INTERACTION OF CONTROLLERCHARACTERISTICS, CONTROLLAW, AND DISPLAY In support of the Army's Advanced Digital/Optical Control System (ADOCS) program, a series of piloted simulations (refs. 8-11) were conducted both at the Boeing Vertol facility and in-house on the VMS at Ames to assess the interactive influences of side- stick controller characteristics, level of stability and control augmentation, and a helmet-mounted display which provided a limited field-of-view image with superimposed flight control symbology. A wide range of stability and control augmentation system ranging from the basic lJJ36OAheiicopter SCAS to a SCAS with transla- (SCAS) designs, tional rate command/translational rate stabilization, was investigated. Variations in controller force-deflection characteristics and the number of axes controlled through an integrated side-stick controller as shown in figure 6 were studied. The handling qualities data base developed from this series of experiments for both day visual and night/adverse weather terrain flying tasks will be used not only for the design of the ADOCS demonstrator helicopter but also as design data for future mili- tary rotorcraft such as JVX and LHX.
SCAS DESIGN - COMMAND/STABILIZATION CHARACTERISTICS THREE-DIMENSIONAL FLIGHT SSC CONFIGURATION HMD SYMB : ATTITUDE CMD/ATTITUDE STAB AT + FLIR ‘?- “MC & CONTROLLER CONFIGURATIONS f HMD COLLECTIVE COLLECTIVE + PITCH /‘i- ROLL 4-AXIS I (3+1) PEDALS COLLECTIVE COLLECTIVE - 2+1+1 1 /3+1) COLLECTIVE Figure 6 - COMPARISONOF IMC AND VMC PILOT RATINGS The results of this series of simulation experiments show that an integrated four-axis force controller with small deflection in all axes was preferred to other four-axes devices with no deflection (stiff stick). With small deflection, the pilot's ability to modulate single-axis forces was improved and the tendency to over- control or to produce input coupling was reduced. The results also indicate (ref. 9) that level of stability and control augmentation has a dominant effect on NOE handling qualities. With a high level of augmentation, satisfactory handling quali- ties for NOE tasks were achieved for all the three levels (two to four axes) of However, integrated force controllers having small deflection as shown in figure 7.
the fully integrated (four-axis) controller degraded handling qualities compared to separate controllers for such large-amplitude multiaxis control tasks as a decelerat- ing turning approach to hover and a high-speed slalom maneuver. Mission tasks flown under the simulated reduced visibility conditions received pilot ratings two or more rating points worse than the identical tasks flown under day VMS conditions.
NOE TASK . VMC A IMC (3+l)c 2+1+1 (4 + 0) s 1 F 9 LEVEL 3 (UNACCEPTABLE) LEVEL 2 &-,A (ACCEPTABLE) \\\\\\\\\\\ \\\ ’ \ c LEVEL 1 VMC (SATISFACTORY) s 2t I I I J I I : ,I .V RA/AT AT/AT AT/LV RAIAT AT/AT AT/LV RA/AT AT/AT AT/l PITCH/ROLL AFCS CONFIGURATION Figure 7 INFLUENCE OF LOAD FACTOR AND TURN DIRECTION ON MODE SHAPE With ever-increasing military demands for agility and maneuverability of rotor- craft, there is a need to better understand the flight dynamics of rotorcraft in such large-amplitude, asymmetric maneuvers as steep, high-g turns. To meet this need, an analytical procedure has been developed to permit a systematic investigation of rotor- craft dynamic characteristics in steep, high-g turns (refs. 12-14). Numerical exami- nations of a tiit-rotor aircraft and several single-rotor helicopters with different types of main-rotor systems have been conducted. .It has been found that strong coupling exists , particularly at low speeds, between the longitudinal- and the lateral-directional motions in high-g turns for both the symmetrical and asymmetrical- type rotorcraft; flying qualities and flight-control design analyses based on small disturbances from straight flight are grossly inadequate for predicting flight dynamics in high-g maneuvers. For example, for single-rotor helicopters the direc- tion of turn has a significant influence on the flight dynamic characteristics in high-g turns. Figure 8 illustrates the effects of load factor and turn direction on the eigenvector of the Dutch-roll mode for a study hingeless rotor helicopter in level turns at 60 knots.
l-g STRAIGHT LEVEL AP n AU Aa Ag A0 A/3 Ap A@ Ar Aq A@ Ae 2-g LEFT TURN 2-g RIGHT TURN AU AP AP ACY
I
“; AU Aci Ag Ae Ap Ap Acj
,L Ir
AU Aa Aq Ae Ap Ap A@ Ar t Ar Ar Figure 8 EFFECT OF LOAD FACTOR, TURN DIRECTION, AND UNCOORDINATlUN ON CONTROLRESPONSE The developed analytical procedure also permits a systematic examination of statics and flight dynamics of rotorcraft in various levels of uncoordinated high-g turning maneuvers. Examinations of several rotorcraft indicate (1) that the aircraft trim attitudes in uncoordinated high-g turns can be grossly altered from those for coordinated turns, and (2) that within the moderate range of uncoordinated flight (side force up to +O.l g), the dynamic stability of these rotorcraft is relatively insensitive. However, the coupling between the longitudinal- and the lateral- directional motions is strong, and it becomes somewhat stronger as the sideslip increases. Examinations of the effect of uncoordinated high-g turns on the perfor- mance of a stability and control augmentation system designed using linear quadratic synthesis techniques indicate that the aircraft response with the SCAS on can degrade as sideslip increases.
The influence of sideslip, however, is found to be less drastic than that of either load factor or turn direction as illustrated in figure 9 (ref. 14). In addition, the study also assessed the individual effects of the aerodynamic, kinematic, and inertial coupling on the flight dynamics of rotorcraft in steep turns.
- 1 g FLIGHT (WITHOUT SCAN 2 g RIGHT TURN, - - - 1 g FLIGHT (WITH SCAS) b ny = 0 (WITH SCAS) + 2 g LEFT TURN, ny = 0 (WITH SCAS) -- - ny = 0.10 (WITH SCAS) -..- n = -0.05 (WITH SCAS) Y
sJ;y, , EL -
0 2 4 6 8 10 12 0 2 4 6 8 10 12 TIME, set TIME, set Figure 9 HELICOPTER IFR AIRWORTHINESS CRITERIA We now turn to the rotorcraft certification criteria research. The rapid expansion of civil helicopter operations has led to increasing efforts to assess problem areas in civil helicopter design, certification, and operation. A joint NASA-FAA program was instituted at Ames to investigate the influence of the helicop- ter's inherent flight dynamics, flight control system, and display complement on flying qualities for IFR flight, both in terms of design parameters to ensure a good IFR capability and with regard to the characteristics that should be required for certification. The specific areas of concern that were addressed include (1) the requirements for stable force or position control gradients; (2) the difference in criteria for normal-category rotorcraft, depending on whether the aircraft is to be certified single or dual pilot; and (3) the SCAS and display requirements for decelerating instrument approach to exploit the helicopter's unique capability to fly at very low speeds. Five ground-based piloted simulations and one flight experiment were conducted during a 3-yr period beginning in 1978 (refs. 15 and 16). These experi- ments were summarized in figure 10 in terms of the specific objectives, the task evaluated, and the facility used.
EXPERIMENT SUMMARY - EXPERIMENT TASK OBJECTIVES FACILITY DATE . CONST. SPEED VOR l HELICOPTER MODELS FSAA NOV 78 . DUAL PILOT l SCAS IMPLEMENTATION . CONST. SPEED VOR . STATIC STABILITIES 2 FSAA MAR 79 . DUAL PILOT l SCAS IMPLEMENTATION . CONST. SPEED MLS l FLIGHT DIRECTORS AND 3 . SINGLE AND DUAL CONTROL-DISPLAY FSAA MAR 80 PILOT l CREW LOADING . CONST. SPEED MLS . VALIDATE GRD. SIM.
4 . DUAL PILOT RESULTS FOR STATICS, FLIGHT: SEP 80 SCAS, AND FLIGHT UH-IH DIRECTORS l CONST. SPEED MLS . LONGITUDINAL DOF 5 . DUAL PILOT . STATIC AND DYNAMIC VMS NOV 80 CRITERIA . DECELERATING MLS l INSTRUMENT DECELERATION 6 VMS OCT 81 . DUAL PILOT . ELECTRONIC DISPLAYS Figure 10 INFLUENCE OF LONGITUDINAL CONTROLGRADIENT Regulations (ref. 17) require positive longitudinal control force stability at approach speeds for both transport and normal-category helicopters, regardless of crew loading. This requirement is probably justifiable for rate-damping types of SCAS, although little significant degradation has been shown with neutral or slightly unstable gradients; hence, the neutral gradient, at least, could be considered marginally acceptable. Figure 11 shows the results of this series of experiments, which indicate the trend of handling qualities as influenced by static longitudinal stability for the rate-damping type of SCAS. Note that with this type of SCAS, average ratings in the satisfactory category were not achieved, even at the In commenting about these configurations, the pilots noted most stable level.
increasing difficulties in maintaining trim and controlling speed precisely as the static stability was decreased, but they also noted that the instrument tracking It should be empha- performance was still adequate at least to neutral stability.
sized that a rate-command attitude-hold type of SCAS results in a neutral longitudinal gradient; this type of SCAS was generally rated in the satisfactory category. Hence the requirement of control gradient may have to be linked to the type of SCAS employed, which it currently is not.
EXPERIMENT 0 1 FSAA A 4 UH-1H @ 2 FSAA v 5 VMS
l 3 FSAA + 6 VMS
8 r INADEQUATE
T _
SATISFACTORY 2- STABLE - ) UNSTABLE I I I I I I I o .I -2 -.4 -.2 -1.0 -.8 -.6 1flD. lhec in/l 5 knot 6 Es/v, Figure 11 Most helicopters currently certified for single-pilot IFR operations employ advanced SCAS or displays or both. Of concern is the level of complexity of the SCAS required to achieve a good IFR capability because of the cost, control authority, and reliability factors the SCAS introduces. The influence of SCAS on the IFR handling qualities was therefore investigated. As shown in figure 12, three types of pitch and roll SCAS, among others, were considered: rate damping with input decoupling, rate command-at,titude-hold (RCAH), and attitude command (AC). These cases are primarily for the SCAS incorporated on a machine with neutral basic longi- Note that a rate-damping SCAS does not alter the control position tudinal stability.
gradient, a RCAH SCAS results in a neutral gradient (as described earlier), and the attitude SCAS stabilizes the gradient because of the MO term. As. indicated in figure 12, even at a fairly high level and with input rate-damping augmentation, decoupling, generally has received pilot ratings from marginally adequate to just Attitude augmentation in pitch and roll (implemented either worse than satisfactory.
as RCAR or AC) is required to achieve satisfactory handling qualities for IFR opera- tions in turbulence. With attitude augmentation, the interaxis coupling and turbu- lence excitation are reduced and short-term and long-term dynamics are improved.
EXPERIMENT 0 1 FSAA A 4 UH-1H u 2 FSAA 'I 5 VMS
l 3 FSAA + 6 VMS
/ STABLE BASELINE INADEQUATE GRADIENT I n ADEQUATE E4 :: SATISFACTORY RATE NO RATE SCAS RATE ! ATTITUDE (LONGITUDINAL) DfpPI;G COMMAND COMMAND ATTITUDE DECOUPLING HOLD (NEUTRAL GRADIENT) Figure 12 INFLUENCE OF TASK DIFFICLJLTY Since the pilot rating applies to a control/display combination for a specific task, and since the evaluation tasks varied somewhat across this series of experi- it is instructive to show the influence of task on the ratings.
ments, Ratings from these experiments are compared in figure 13 for similar SCAS characteristics (rate SCAS and attitude-command SCAS) and displays (with and without three-cue flight director displays) as a function of the task evaluated.
It is noted that the addition of three-cue flight directions generally improves ratings.
Also, an increase in the task difficulty (e.g., single pilot or inclusion of an instrument deceleration) results in degraded ratings for equivalent configurations, Specifi- cally, the difference between the dual-pilot and single-pilot tasks is seen to be almost one pilot rating point. A difference in requirements' for single- and dual- pilot operations is warranted. In addition, it may also be seen from figure 13 that a decelerating instrument approach leads to worse ratings than even the single- pilot task with a constant-speed approach. More stringent criteria may therefore be required for decelerating instrument operations.
EXPERIMENT SYMBOLS - RATE SCAS n 2 FSAA OPEN INADEQUATE ATTITUDE SCAS
l 3 FSAA CLOSED -
+ 6 VMS I3 ADEQUATE 0 E4 l SATISFACTORY MLS MLS MLS MLS MLS VOR CONST. CO NST. DECEL.
CONST. CONST. CONST.
SPEED SPEED APPROACH SPEED SPEED SPEED SINGLE DUAL SINGLE DUAL DUAL DUAL PILOT PI LOT PILOT PILOT PI LOT PI LOT , / \ \ * Y FLIGHT DIRECTOR DISPLAYS RAW DATA DISPLAYS Figure 13 PILOT EVALUATIONS OF TILT-ROTOR TRANSITION The first ground-based simulation experiment in a projected series of investi- gations was conducted by NASA and the FAA on the VMS at Ames to perform a preliminary assessment of airworthiness considerations for tilt-rotor aircraft in terminal area operations (fig. 14). Principal variables of the experiment were (1) visual versus instrument approaches, (2) the type of stability and control augmentation, and (3) three conversion profiles ranging from full conversion before the glide slope to full conver- sion on the glide slope. The results obtained in a recent study indicated that, for visual approaches, satisfactory performance within moderate pilot compen- sation was generally achievable irrespective of the conversion profile used; cross- winds and a moderate level of turbulence had a noticeably degrading effect with the baseline XV-15 SCAS but minimal influence with an attitude SCAS. For instrument approaches, the desired performance could be achieved with the attitude SCAS and the conversion profile having all conversion prior to the glide slope. It was also found that the marginally inadequate performance for the profile having all the con- version on the glide slope could be improved to the satisfactory level by adding automatic thrust tilt and three-cue flight directors.
VERTICAL MOTION SIMULATOR CROSSWIND MODERATE TURBULENCE 0 BASIC SCAS INADEQUATE 0 ATTITUDE SCAS OPEN - VFR CLOSED - IFR .._.
z7
- t
ADEQUATE fl SATISFACTORY I I I I! ’ CONVERT CONVERT CONVERT ON BEFORE GLIDE-SLOPE BEFORE GLIDE-SLOPE AND ON GLIDE-SLOPE Figure 14 II I I llllll11llll III Ill I II II Ill1 II Ill lllllll I I SUMMARY In summary, we have briefly reviewed some major rotorcraft handling qualities research projects at Ames Research Center (fig. 15). They were grouped into two categories: (1) military rotorcraft handling qualities research, and (2) civil In the first category, the research rotorcraft certification criteria research.
efforts that focus on determining the effects of engine and thrust response charac- teristics, interaxis coupling, controller characteristics control law/display inter- action, and large-amplitude maneuvers were highlighted. In the second category, efforts to develop IFR airworthiness handling qualities criteria for helicopters and tilt-rotor aircraft were discussed. Before concluding this discussion, it may be worth noting that a joint Army/Navy program is currently under way to update MIL-H-8501A. The objective is to develop mission-oriented handling qualities requirements for military rotorcraft (ref. 18). NASA's role related to the program is (1) to continue working with the Army (Aeromechanics Laboratory) to establish a comprehensive handling qualities data base and design guidelines for land-based military rotorcraft, and(2) to expand the scope to include research on developing rotorcraft handling qualities criteria for Navy shipboard mission tasks.
l MILITARY ROTORCRAFT HANDLING QUALITIES RESEARCH - EFFECTS OF ENGINE AND THRUST RESPONSE CHARACTERISTICS - EFFECTS OF INTER.AXIS COUPLING - INTERACTIVE EFFECTS OF CONTROLLER/CONTROL LAW/DISPLAY - EFFECTS OF LARGE AMPLITUDE (HIGH-g) MANEUVERS l ROTORCRAFT CERTIFICATION CRITERIA RESEARCH - HELICOPTER IFR TERMINAL AREA OPERATIONS - TILT-ROTOR AIRCRAFT Figure 15 REFERENCES 1. General Requirement for Helicopter Flying and Ground Handling Qualities.
Military Specification MIL-H-8501A, Bureau of Naval Weapons, 7 Sept. 1961.
2. Corliss, L. D.: The Effects of Engine and Height-Control Characteristics on Helicopter Handling Qualities. J. AHS, vol. 28, no. 3, July 1983, pp. 56-62.
3. Corliss, L. D.; Blanken, C. L.; and Nelson, K.: Effects of Rotor Inertia and RPM Control on Helicopter Handling Qualities. AIAA Paper 83-2070, 1983.
4. Chen, R. T. N.; and Talbot, P. D.: An Exploratory Investigation of the Effects of Large Variations in Rotor System Dynamics Design Parameters on Helicopter Handling Characteristics in Nap-of-the-Earth Flight. J. AHS, vol. 24, no. 3, July 1978, pp. 23-36.
5. Chen, R. T. N.; Talbot, P. D.; Gerdes, R. M.; and Dugan, D. C.: A Piloted Simulator Study on Augmentation Systems to Improve Helicopter Flying Qualities in Terrain Flight. NASA TM-78571, 1979.
6. Corliss, L. D.; and Carico, G. D.: A Preliminary Flight Investigation of Cross-Coupling and Lateral Damping for Nap-of-the-Earth Helicopter Operations.
AHS Paper 81-28, 1981.
7. Chen, R. T. N.: Unified Results of Several Analytical and Experimental Studies of Helicopter Handling Qualities in Visual Terrain Flight. Helicopter Handling Qualities, NASA CP-2219, 1982, pp. 59-74.
8. Landis, K. H.; and Aiken, E. W.: An Assessment of Various Side-Stick Controller/Stability and Control Augmentation Systems for Night Nap-of-the Earth Flight Using Pilot Simulation. Helicopter Handling Qualities, NASA CP-2219, 1982, pp. 75-96.
9. Landis, K. H.; Dunford, P. J.; Aiken, E. W.; and Hilbert, K. B.: A Piloted Simulator Investigation of Side-Stick Controller/Stability and Control Augmentation System Requirements for Helicopter Visual Flight Tasks.
AHS Paper A-83-39-59-4000, 1983.
10. Aiken, E. W.: Simulator Investigation of Various Side-Stick Controller/ Stability and Control Augmentation Systems for Helicopter Terrain Flight.
AIAA Paper 82-1522, 1982.
11. Aiken, E. W.; Landis, K. H.; Glusman, S. I.; and Hilbert, K. B.: An Investiga- tion of Side-Stick Controller/Stability and Control Augmentation System Requirements for Helicopter Terrain Flight Under Reduced Visibility Condi- tions. AIAA Paper 84-0235, Jan. 1984.
12. Chen, R. T. N.; and Jeske, J. A.; Influence of Sideslip on the Kinematics of the Helicopter in Steady Coordinated Turns. J. AHS, vol. 27, no. 4, Oct. 1982, pp. 84-92.
13. Chen, R. T. N.: Flight Dynamics of Rotorcraft in Steep High-g Turns. AIAA Paper 82-1345, 1982.
14. Chen, R. T. N.; Jeske, J. A.; and Steinberger, R. H.: Influence of Sideslip on the Flight Dynamics of Rotorcraft in Steep Turns at Low Spee‘ds. 39th Annual Forum of the AHS, St. Louis, MO., May 9-11, 1983 (Paper No. A-83-39-56-4000).
15. Lebacqz, J. V.: A Ground Simulation Investigation of Helicopter Decelerating Instrument Approaches. AIAA Paper 82-1346, 1982.
16. Lebacqz, J. V.; Chen, R. T. N.; Gerdes, R. M.; and Weber, .I. M.: A Summary of NASA/FAA Experiments Concerning Helicopter IFR Airworthiness Criteria.
J. AHS, vol. 28, no. 3, July 1983, pp. 63-70.
17. Rotorcraft Regulatory Review Program Notice, No. 1; Proposed Rulemaking, Federal Register, vol. 45, no. 245, Dec. 18, 1980.
18. Key, D. L.: The Status of Military Helicopter Handling Qualities Criteria.
AGARD Criteria for Handling Qualities of Military Aircraft, AGARD CP-333, June 1982, pp. 11-l to 11-9.
FLEXIBLE AIRCRAFT FLYING AND RIDE QUALITIES Irving L. Ashkenas, Raymond E. Magdaleno, and Duane T. McRuer Systems Technology, Incorporated Hawthorne, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 REPORT CONTENTS This presentation covers some of the highlights of IJASA CR-172201, "Flight Con- trol and Analysis Methods for Studying Flying and Ride Qualities of Flexible Trans- port Aircraft." The report itself contains the chapters listed in Fig. 1, and we'll follow this order in our discussion.
Of course, we'll have to limit ourselves to the more significant aspects and forego many of the details that are in the report.
We'll start: with a block diagram representative of a generalized FCS, go into a brief analytic exposition to illustrate a central principle in flexible mode con- trol, list and discuss some of tile pertinent pilot-centered requirements, expose the desired features of the control methodology, and select the methodology to be used.
Then we'll discuss the example Boeing-supplied characteristics and show how we approxlnated these with a reduced-order model and a simplified treatment of unsteady aerodynamics. The closed-loop flight control system design follows, along with first-level assessments of resulting handling and ride quality characteristics.
Some of these do not meet the postulated requirements and remain problems to be solved possibly by further analysis or future simulation.
I, INTRODUCTION
II, GENERAL ASPECTS OF FLEXIBLE VEHICLE CONTROL
III, SYSTEM DESIGN REQUIREMENTS AND DESIRES
METHODOLOGY CONSIDERATIONS FOR FLEXIBLE AIRCRAFT
IV,
CONTROLS AND FLYING QUALITIES ANALYSIS
V, FLEXIBLE AIRPLANE CHARACTERIZATION AND SIMPLIFICATION
VI I FLIGHT CONTROL DESIGNAND ASSESSMENTS
VII, CONCLUSIONS AND RECOMMENDATIONS
REFERENCES
Figure 1 GENEKALIZED FLIGHT CONTKOL SYSTEM FOR TRANSPORTAIRCRAFT INCLUDING FLEXIBLE MODES This block diagram (Fig. 2) illustrates primarily the multiple feedback paths acting on the sensor array and the possible use of secondary control points and limited for- ward loop elements. The primary FCS design task, of course, is to formulate the sensor equalization complex to yield a stable, robust system which meets the direct and implied requirements.
External Disturbances, Flexible Modes Generalized Coordinates i Elevator E Elevator - Forward Loop - Actuation - Equalization I / I System 6e Aircraft Attitude, '8 _ Secondary Dynamics Pitchin Control Secondary -TzG+-- Point Control a, Actuation Acceler- 6, Svstem ation Sensor/Equalization Complex Figure 2 ELEMENTARY FLEX MODE CONSIDERATIONS These equations (Fig. 3) constitute a simplified treatment of the considerations involved in synthesizing a suitable sensor-equalization response.
The first equation represents the rigid-body attitude rate response; the second is the slope of the first is the first oscillatory flexi.ble mode response where $I' bending mode at the sensor station.
Adding these responses yields the third equation with the simplified numerator/ The point is that selection of the sensor location and denominator ratios shown.
corresponding mode slope 4 can be used to directly affect these ratios or the.
equivalent pole-zero ordering.
+;ws Qflex 6,- s2 + 2(&),," + b,; K[s2 + 2q,~qs + ~$1 G(s) = " = s[s2 + 2t3',0@3+ $1 Figure 3 SYSTEM SURVEY FOR QUADRATIC DIPOLE CONTROL the root locus progresses into the right If the zero is greater than the pole, half-plane as in a); if less, it stays in the left half-plane as in b) and the proper choice of feedback gain will then provide enhanced structural mode damping.
This is a simplified explanation of a well-known general principle of flexible mode control, i.e., the desirability of synthesizing a sensor-equalization charac- teristic which exhib-its an alternating numerator/denominator ordering of quadratic pairs (a sawtooth Bode) which creates leading phase "blips" for those modes which are to be controlled. For those modes which are to be largely ignored by the con- trol system, appropriate notch or low-pass filtering might be considered if the modes are not so high in frequency relative to actuator and other dynamics as to make them insignificant anyway. (See Fig. 4.)
OdB Lines for Neutral Stability o 1 Lag- Lead Dipole b) Lead- Log Dipole (2 c I) Figure 4 I I I II I I III I llll1lllllllllllllllllllllllllll1lllllllIIIlII PILOT-CENTERED COMMAND KEQUIKEMENTSAND FLYING QUALITIES In addition to the Foregoing implied requirement, there are direct requirements for minimum satisfactory flying and ride qualities. The flying qualities list shown here pertains to pilot's attitude and acceleration response to elevator input (Fig. 5).
The first two headings refer primarily to attitude control and reflect the pos- sible use of either frequency- or time-domain assessment criteria.
The third heading relates mostly to unwanted acceleration responses which can be self-excited by feedthrough to, and ampli- excited directly by the pilot's remnant, fication resulting from, the pilot's body-arm-controller induced motions, or directly excited by normal closed-loop piloted operation.
The final heading generally relates to either attitude or acceleration responses, although attitude is the more common culprit.
goth synchronous behavior and the PI0 syndrome are assessed later for the derived system, as are pertinent aspects of the preceding items.
FREQUENCY DOMAIN M~b+I/T~E)e-TS f (s) = EQUIVALENT s(s2+2&JJs +w21 BANDWIDTH CLOSED LOOP TINE DOMAIN ENVELOPE BOUNDED TIME PARAHETERS TRP FLEX I BLE MODE EFFECTS REMNANT EXCITATION VIBRATION FEEDTHROUGH PILOT CLOSED-LOOP EXCITATON OF FLEX MODES PI0 CONSIDERATIONS SYNCHRONOUS BEHAVIOR PI0 SYNDROME Figure 5 FQ AND FLEX A/C CONTROLCONSIDERATIONS GOVERNING CONTROL DESIGN TECHNIQLJRSELECTION The summation of certain of the foregoing and of the more complete considera- tions in the report as they pertain to the selection of appropriate design method- ology is listed here (Fig. 6).
In the first place, we have to consider uncertainties and variations in the air- frame poles and zeros due to changes in flight conditions and loading.
Second, we have to utilize and consider many elements which are basically expressed i.n frequency-domain formulations.
Third are the direct and implied control design criteria which can be in time- or frequency-domain formulations, or simply expressed as desirable qualities.
8ased on these and other considerations, the basic control methodology selected comprises conventional, classical, multivariable, frequency-domain analysis tech- niques.
1, KEY AIRCRAFTPARAMETERS WIDE RANGING (LO FREQ) POLES, ZEROS NARROW RANGES (HI FREQ) 2, FREQUENCY DOMAIN FORMULATIONS .
PILOT I/O CONTROL ACTIVITIES, REMNANT, VIBRATION FEEDTHROUGH, PI0 BEHAVIOR .
UNSTEADY AERODYNAMICS .
MODAL FORMULATIONS -FREQUENCY-IDENTIFIED POLES AND ZEROS .
CONTROL ACTIVITY RANGE .
RIDE AND HABITABILITY CONSIDERATIONS AND CRITERIA .
RANDOM GUSTINPUTS 3. CONTROL SYSTEM DESIGNCRITERIA .
FLYING QUALITIES REQUIREMENTS .
FLEX m)DE POLE, ZEROSEQUENCING FOR CONTROLLED MODES .
GAIN STABILIZATION FOR IGNORED MODES .
ENHANCED DAMPING FORm)DESPOSSIBLYCAUSING EXCESSIVE REMNANT PILOT FEEDTHROUGH, PILOT SYNCHRONOUS BEHAVIOR .
PI0 SUSCEPTIBILITY .
CONTROLLER SIfiPLICITY .
CONTROLLER ROBUSTNESS Figure 6 THREE VIEWS OF SUPERSONIC CRUISE AIRCRAFT Before applying these techniques, it was necessary to derive a simplified repre- sentation of the Boeing-supplied data base for the delta wing supersonic cruise air- craft (SCRA), shown in Fig. 7, which included: Modal equations of motion (EOM) - 25 x 25 a.
b. Computer printouts of EOM matrix elements for 6 reduced-frequency sets of unsteady aerodynamics for each of 4 flight conditions C. Mode shape data in a variety of formats: tabulated, interpolated displace- ments and slopes at selected locations on the fuselage centerline; pictorial or perspective views; and contour plots for wing relative displacements out- of-plane d. Numerical frequency response data at 149 discrete frequencies supplied on magnetic tapes for four flight conditions. These "data" are the result of interpolation among the 6 reduced-frequency sets of unsteady aerodynamics SPAN 141.67’ P II !I LENGTH 293.33’ Figure 7 MODE SHAPES FOR TAKEOFF WEIGHT DISTRIHUTION To afford an appreciation for the scope o.f the complete model Eormulation, the total set of centerli.ne elastic mode shapes for the take-off case is shown in.
Fig. 8 in the form of displacement normalized to maximum deflection. In general, the modes are 3-dimensional, and Fig. 8 shows just the cut along the fuselage centerline.
In many cases the maximum deflection is not along the centerline, and there is no corresponding unity value shown for those modes.
Modes one and two (Fig. 8) are rigid-body modes, respectively heave and pitching motion. Mode three is the first structural (bending) mode, and the struc- The i.n-vacua tural modes go up in complexity and frequency as the numbers go up.
frequencies in Hz are as follows.
Mode Frequency (Hz) Mode Frequency (Hz) Mode Frequency (Hz) 3 9 4.44 15 6.44 1.14 4 1.60 10 4.82 16 6.89 5 2.49 11 5.15 17 7.06 6 2.95 12 5.45 18 7.24 7.44 7 3.81 13 5.92 19 8 4.28 14 6.11 20 7.56 For a transport aircraft, this list has a remarkably large number of low-frequency closely spaced modes which can interfere, in one way or another, with piloted con- trol.
Y 11 I. 1, w I I I mm CG -SENSOR- I?::; I I I IL iI- 0 SW moo 1500 2OW 2506 3000 3500 lincherl Figure 8 I II II I I I I lllllllll1lllllllI Ill1 I I II I llllllllll MODE SHAPES FOR TAKEOFF WEIGHT DISTRIBUTION (CONCLUDED) Various body centerline stations and physical poi.nts are identified along the bottom of each plot. The "sensor station" is one chosen by Boeing as being in a fairly stiff region as evident by the fairly flat shape of the various modes in this area. The open circle symbols in Fig.
8 show that there is little change in mode three for the start cruise condition.
Modes nine through fourteen in Fig. 9 are characterized by more lumps and bumps than the first set, and modes fifteen through twenty in Fig. 9 are even lumpier and include some very large spikes. These anomalies appear to be due to ill-conditioned the mass and stiffness elements chosen for the analysis lumped parameters, that is, are not necessarily well conditioned and apparently lead to local resonances which give rise to the discontinuities shown.
However, notice that the area in the "sensor" region, where the structure is relatively stiff, is pretty smooth for all modes.
la- .?5 - = .50- m 8, .25 - it .
o- 1’ -.25 -30 I I .I .I I I II 0 500 1000 1500 2000 2500 3000 3500 I inches I Figure 9 - - COMPARISONOF MODEL C WITH.COMPLETE DISCRETE ROEING DATA The number of elastic modes selected for final retention in .the simplified model underwent a gradual increase from three to seven to ten largely to account for the acceleration response shown in Fig. 10. The reduced-order (10 mode) model "C" shown retains modes 3 to 6, 8, and 11 to 15, and is effected through progressive elimination of successive elastic modes by neglecting dynamic (s2 and s) terms relative to (constant) stiffness terms in each successive modai column. The generalized coordinate to be eliminated, now characterized by only a stiffness term, is expressed in terms of the remaining coordinates. Notice that some' of the retained modal equations are for higher frequency modes than those eliminated. This poses no mathematical problem, the progressive elimination of the equations in question proceeds as described above. However, there is no good physical rationale for neglecting the dynam1.c and s) terms of certain lower frequency modes and retaining those for some (6 higher frequency modes, except that it produces an excellent match as illustrated in Fig.
10, where the high-frequency behavior is reproduced with sufficient fidelity to permit accurate Klde quality analyses to proceed on the basis oE the reduced- order model.
This match is also based on simplified, "distributed" unsteady aerodynamics, meaning that for each degree oE freedom, or matrix column, the corresponding flexi- ble mode frequency was used to assign constant aerodynamics consistent with that value of reduced Erequency.
-100 - Figure 10 MODEL,D TAKEOFF B0DE.S OF PITCH ATTITUDE A final correction was applied ‘to eliminate perceived Inconsistencies in the supplied numerical elevator,fngrtlal properties and produce the "final" (Fig. 11) model "D" attitude responses for two locations.
The rear seat location provides the better sawtooth and is so labeled.
Z,DO wkadkc) .I0 10.00 I -I .I.
8Rear soot “Sawtooth” Bode 0) -8 et Senior b/ 7 cc Figure 11 , _ ,-,: : .. I ‘.. . .
MODEL D TAKEOFF BODE WITH PITCH RATE GYRO ~.r;,:.'. .,.
.,.:, - .'
Figure 12 shows the Bode for &rate gyro ; with' tippica dynamics, at the rear seat station with a:.suggested. clbiure. gain .of OW5. :c .- w(rad/sk) 10.00 0 dB for K,.= .5 - u
p -100 -
a
---.a------------------
-
Rate Gyro Loop T. E 400 o%ec.i soot “Sawtooth” Bode [.7,201 -a,, -( (Ride Quality Modes) Rate Gyro Dynamics : 3’ Figure 12 ROOT LOCUS PLOT OF FIGURE 12 SYSTEM 13 shows the resulting improved damp The corresponding root locus plot in Fig.
13, and 14, which are slightly degraded.
ing of all modes except 11, +o WI5 WI4 WI3
c
WI2 WII Actuator and Rate Gyro Root Locus “Sawtooth” Bode, Ride Quality Modes w4 w3 WSP Figure 13 e'pilot PILOT STATION AUGMENTED ATTITUDE RESPONSE TO ELEVATOR, T-- e feedbacks to 8 but they were not very effective in pro- We also looked at a, f requency modes.
viding additional damping of the lower In fact, the higher fre- quency modes, 6, 8, and 11 to 15, are essentially unobservable by a centerline acceler- ometer. Accordingly, vertical acceleration feedback to the elevator appears to be unnecessary to slightly undesirable; it was therefore eliminated as a pri.mary clo- The remaining basic elevator control loop structure shown below sure possibility.
is si.mple indeed (Fig. 14).
ec a2 8 8 et3 station - Vehicle - 87s L Rate Gyro %s s - Figure 14 ELEVATOR CONTROL LOOP STRUCTURE . ,: The basic FCS-augmented .attitude response at the.piiot's station to control 15. The effective bandwidth, set in this inputs using this system is given in Fig..
case by a 6 dR gain margin.requirement, is abo-ut 0.9 .to 1.0 rad/sec which corre- sponds to satisfactory handlfng ,for this flight ,condfti'on.
.I.
urn I - u(mdk'Y- %" Figure 15 8' RESPONSETO LO-DEG STEP ELEVATOR However, the pilot's station attitude response to a step input in Fig. 16 shows an effective time delay of about O-55 set, greater than allowable even using the most optimistic data. This quite large time delay is also apparent in the Fig. 15 phase characteristics (i.e., using the sLmple approximation 'cw - 90 deg = 1.57 rad where w N 4 = 180 deg, reff k 1.57/3.2 s 0.49). It is directly traceable to bending mode effects as shown at the bottom of Fig.
16, which shows only about a 0.10 set delay in the pitch response at ,the rear seat where the mode slopes are all either basically smaller than, or opposite in sign to, those at the pilot station. Thus the differ- ence is attributable to the natural change in sign .of the mode slopes in going from the rear seat to the pilot's station. The change in sign is a result of the In this sense, the inherent bending mode shapes of a slender flexible body.
associated additional time delay of the pilot's attitude response to a step control aft-surface input is fundamental. Before we decide what, if anything, can be done to eliminate or reduce such additional delay, we need to make further assessments.
l.O- 0; (de@ I I I I I 4 .6 .0 1.0 1.2 Time bed Al Pi/of Station I I I I I !a .6 .0 1.0 1.2 Time bed At f?eoi Seot Figure 16 "SYNCHRONOUS" PI0 POSSIBILITY Relative to PI0 proneness, the Fig. 15 pilot's attitude Rode is such that loop closure can be easily effected with pure gain adaptation on the part of the pilot.
Accordingly, there is no tendency for the "PI0 syndrome" which is characterized by required low-frequency lag adaptation for normal closed-loop operations. Such lag is an easy to accomplish, low-workload behavioral pattern described by pilots as a "smooth, trim-like control action." The trouble arises when, in an attempt to regain control after an upset or other stressful occurrence, the pilot regresses to a Then the pilot-vehicle (with the suddenly pure gain type of proportional control.
changed pilot equalization) may temporarily have too small a gain margin at a "high" frequency oscillatory mode (short period or conceivably a flexible mode).
The Lightly damped peak at about 16.5 rad/sec (Fig. 15) could conceivably be excited momentarily by "synchronous" pilot behavior at thLs frequency. The result- ing PI0 would not be unstable at the more probable lower gain shown in the fragmen- tary root locus of Fig. 17, but could have quite low damping. Of course, the level of pilot gain involved can only be sustained for a short tine before the system diverges at the lower frequency corresponding to the -180 deg phase crossover in Fig. 15. Thus, at best, synchronous PI0 would occur in "bursts" rather than in a sustained oscillation.
18.0 17.5
b
17.0 16.5 16.0 , -2.5 -2.0 -1.5 -1.0 -.5 o- Figure 17 PILOT STATION ACCELERATION (sip) RESPONSETO ELEVATOR INPUTS With respect to vibration feedthrough to the pIlot, the excitation at the pilot's station due to step and ramp elevator inputs is shown in Fig. 18. Clearly, the alleviating influence of a rate limIted surface input is desirable to avoid the high-frequency ringing at about 40 rad/sec and to reduce the amplitude and frequency for a lo-deg elevator ramped in at 30 deg/sec, the of the 16 rad/sec mode. However, effective time delay increment (half of the time to ramp to 10 deg) is 0.17 sec.
Therefore the effects of realistic surface rate limits wiL1 be to accentuate the A better solution effective time delay problem which is already possibly critical.
to the vibration feedthrough problem than Low surface rate saturation is desirable.
aI 10 deg Sfep 0.5 9 rod/xc sip
/-
(a 0.7 I I,, I Time &/ 30 deg/sec Romp Figure 18 ,
as
-
AMPLITUDES WITH SENSOR LOCATION w13 Turning now to ride qualities and Fig. 19, we have to recognize first that the usual Dryden turbulence spectra effectively flatten the asymptotic amplitude response to random gust inputs over almost the entire frequency range.
This is indicated by the long dashed lines in Fig. 19 which reprerient the zero dB line for a still leaving large spikes in the pilot's a, ow unity gust, response at about 16 and 36 rad/sec. As shown in the figure, these same spikes are generally evident along the entire cabin area. This means that ride quality is ageneral problem, not necessarily peculiar to pilot location. Accordingly, a general solution must be found. This could conceivably take one of two forms: use of a secondary control point to damp the offending modes or seat motion attenuation and damping.
The ride mode most evident in Fig. 19 is that at roughly 16 rad/sec, mode 6.
The largest amplitude spike above the zero dB gust lines at 16 rad/sec is about 34 dB (at the rear node) or an amplification factor of 50. This is far too large to be effectively damped by passive seat suspension and cushioning systems.
?lode 6's three-dimensional character is predominantly of a wing torsional nature so it's not surprising that a centerline control has no effect on it.
It is also clear that the wing lift due to such torsional deflection will apply more or less uniformly along the fuselage, which explains the Fig. 19 results. The obvious way to damp this motion is to use the outboard wing movable surfaces responding to motion also sensed at an outboard location. The Boeing data, unfortunately, do not cover symmetric aileron or flaperon inputs to the Longitudinal mode, so this option remains as a possible future exercise.
Rmr Nodt,JIPS Figure 19 RESULTS AND CONCLUSIONS: GENERAL Systematically exposed all design-centered factors to consider (l)(Fig. 20) Control system'functions and roles Flexible mode control.principles ., FCS crkteria an4 desire& for " Pilot-centered command and flying qualities > Bide-qualitfes ,' Controller-centered requirements and deaign.implications Available design methodologies and selection of recommended methods Established fundamental requirements on effective vehicle characteristics (aircraft/controller combination) consistent with simple robust control- lers. -- i.e., pole-zero ordering (2) Translated above into system and subsystem requirements (1) Confirmed a simplified treatment of unsteady aerodynamics which compared very well with the complete treatment (3) ,Developed and demonstrated considerations for selective inclusion/deletion of significant/insignificant modes within reduced-order system8 (3) 0 Demonstrated systematic design/analysis methods to meet requirements; derived a very simple robust system (4)
(1) ASSEMBLED REQUIREMENTS AND DESIRES
(2) REITERATED A CENTRAL PRINCIPAL IN FLEXIBLE MODE CONTROL
(3) DERIVEDAND UTILIZED A SIHPLIFIED,FLEXIBLE AIRPLANE MODEL
(4) DEMONSTRATED SYSTEMTIC DESIGNKTHODS
(5) IDENTIFIED CERTAIN PROBLEMS ENDEflIC TO FLEXIBLE VEHICLES
OF TYPE STUDIED
Figure 20 CONCLUSIONS: SPECIFIC "PROBLEMS" The effective time delay in 0 response to elevator appears to be a generic FJ problem due to low-frequency ending mode(s) as seen at the pilot station (Fig. 21).
l Vertical acceleration feedthrough at the pilot's station can be reduced by lowering the saturation rate of the elevator. However, this adds to the 9 response lag and may not be a viable solution.
Vertical acceleration response to w is high in general and must be reduced for good ride qualities. Analysis ndicates that secondary, outboard sur- f and off-centerline located sensors offer a probable solution which face(s), should also decrease vertical acceleration feedthrough.
0 mode is also involved in "synchronous PIO" possibili- The above "prominent" ties, but these are evident in the pilot's pitch attitude response and may or may not be reduced by the above-suggested secondary control surfaces.
Because of its prominence in ride, synchronous PIO, and vibration feed- response characteristics similar to those of the "prominent" mode through, are likely candidates for simulation research.
Te IN 6p RESPONSE TO ELEVATOR -- A GENERICPROBLEM DUE TO LOW
FREQUENCY BENDINGMODE(S)
VERTICAL ACCELERATION FEEDTHROUGH AT PILOT'S STATION
VERTICAL ACCELERATION RESPONSE TO wg IS HIGH IN GENERAL AND
MUSTBE REDUCED FOR GOOD RIDE QUALITIES
THE ABOVE"PROMINENT"MODEIS ALSO INVOLVEDIN "SYNCHRONOUS
PIO" POSSIBILITIES
BECAUSE OF ABOVEEFFECTS, RESPONSE CHARACTERISTICS OF THE
'PROMINENT" MODE ARE LIKELY CANDIDATES FOR SIMULATION
RESEARCH
Figure 21 RECOMMENDATIONS Expand present study to include off-center line symmetric controls(Fig. 22) Plan and conduct moving base simulation to investigate: (a) LOW-freqUE!nCy tims delay8 in pilot station attitude response asso- ciated with fuselage bending mode8 (b) Motion feedthrough and potential synchronous PI0 due to high-frequency centerline motion Develop more-automated means to achieve simple controllers which exhibit robust characteristics demonstrated herein, e.g.: Automated numerator synthesis for minimum (fixed-form) sensor/ equalization complexes which assure desired zero, pole order, and per- mit maximum spacing between a limited (specified) number of zero, pole pairs Frequency domain optimal performance indices and procedure8 which pre- ordain an optimal controller/aircraft combination satisfying the saw- tooth Rode requirements
STUDYOFF-CENTER LINE SYMMETRIC CONTROLS
PLAN AND CONDUCT MOVINGBASE SIMULATIONTO INVESTIGATE:
(A) TIME DELAYSDUE FUSELAGE BENDINGMODES
(B) FEEDTHROUGH AND POTENTIALSYNCHRONOUS PI0 DUE TO HIGH-
FREQUENCY COCKPITACCELERATION
DEVELOP MORE-AUTOMATED f'iZANSTO ACHIEVE SINPLE, ROBUST
CONTROLLERS
Figure 22 BIBLIOGRAPHY Ashkenas, I. L., Magdaleno, R. E., and McRuer, D. T.: Flight Control and 1.
Analysis Methods for Studying Plying.and Ride. Qualities, of Flexible Transport Aircraft, NASA CR-172201, August 1983. . .
IMPLICATIONS OF CONTROLTEXHNOLOGY ON AIRCRAFT DESIGN Steven M. Sliwa and P. Douglas Arbuckle NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT New controls technologies are now available for implementation with aircraft systems. Many aircraft with state-of-the-art technology in the fields of aerodyna- structures,and propulsion'require extensive augmentation merely for safety of mics, flight considerations in addition to potential performance improvements. The actual performance benefits of integrating the new controls concepts with other new technol- ogies can be optimized by including such considerations early in the design process.
several advanced aircraft designs have run into considerable problems Recently, related to control systems and flying qualities during flight test, requiring costly redesign and fine-tuning efforts. It is no longer possible for the aircraft design to be completed prior to getting the controls specialists involved. The challenge to the control system designer has become so great that his concerns must be considered at the conceptual design level. A computer program developed at NASA for evaluating the economic payoffs of integrating controls into the design of transport aircraft at the beginning will be described.
o NEW CONTROLS TECHNOLOGIES ARE AVAILABLE o MANY NEW AIRCRAFT REQUIRE ADVANCED CONTROLS o EXPENSE OF FINE TUNING CONTROL SYSTEMS FOR CURRENT STATE- OF-THE-ART AIRCRAFT HAS RISEN DRAMATICALLY 8 INTEGRATING CONTROLS INTO DESIGN PROCESS IS BENEFICIAL o A TOOL HAS BEEN DEVELOPED TO EVALUATE THE PAYOFFS OF CONTROLS INTEGRATION INCREASE IN CONTROLSOOMPLEXITY During the past 20 years, the control systems being used on state-of-the-art aircraft have improved significantly. In the 1950's and 1960's, simple control laws were being applied to improve the flying qualities. In contrast, current configura- tions may require extensive augmentation for safety of flight as well as for good flying qualities. Because of this, and because of the increased complexity of all aircraft systems, it has become extremely difficult to fine-tune or adjust control laws during flight test. Redesign efforts currently require significant amounts of engineering, which result in costly delays. Previously, very simple control schemes were used merely for improving flying qualities, and mechanical back-up systems were always utilized in the event of electronic component failure. Now, highly complex laws which rely on the improved reliability of digital and analog circuits use redun- dant systems for back-up modes. These examples illustrate some of the fundamental issues facing a control system design engineer today.
SIMPLE
c\
AUGMENTATION III a USED FOR GOODFLYI NG QUALITIES EXTENQVE AUGMENTATION . EASY TO ADJUST CONTROLLAWS REQUIRED AT FLIGHT TEST FOR SAFETY OF FLIGHT 0 S IMPLE CONTROL 0 VERY EXPENSIVE CONTROLS LAWS WITH TO ADJUST COMPLEX IT\r / MECHANICAL CONTROLLAWS AT BACKUP FLIGHT TEST COMPLEX CONTROLLAWS WITH MULTI PLY L REDUNDANT SYSTEMS -I I ~.. ~~ I I w 1960 1970 1980 COMPARISONOF CONTROLSTECHNOLOGY A comparison of some of the characteristics of early automatic control systems for aircraft and those being applied to current configurations is shown below. Ini- tially, control systems were designed using simple single-loop analyses for aircraft with limited envelopes where rigid airframe assumptions were adequate.
Now flexible aircraft with expanded envelopes have significant aeroservoelastic interactions that cannot be ignored during control system design. The current tendency is to develop digital fly-by-wire control systems utilizing complex multi-input, multi-output design techniques with sophisticated redundancy management.
Clearly, to achieve the full potential of applying these technologies, the controls integration must occur early in the design process.
THEN NOW o MECHANICAL LINKAGES o DIGITAL 6 DOF FLY BY WIRE o SIMPLE YAW DAMPER WAS ONLY m COMPLEX CONTROL LAWS WITH AUGMENTATION HIGH-ORDER COHPENSATORS o SIMPLE SISO DESIGN TECH- . COMPLEX MIMO DESIGN TECH- NIQUES USED DURING CONTROL NIQUES USED DURING CONTROL SYSTEM DESIGN SYSTEM DESIGN a RIGID AIRFRAME ASSUMPTION o AEROSERVOELASTIC INTERAC- TIONS IMPORTANT GOOD o LIMITED ENVELOPE o EXPANDED ENVELOPE o SIMPLE CONTROL MODES o NEW. COMPLEX CONTROL MODES o REDUNDANCY THROUGH MECHANI- o REDUNDANCY THROUGH MULTIPLE CAL STRENGTH SYSTEMS o NEED FOR CONTROLS INTEGRA- TION EARLY IN DESIGN EFFORT EVOLUTION OF CONTROL SYSTEMS The control system components have undergone considerable change and refine- ment. Originally, simple mechanical linkages using cables, pulleys, and push rods were used. As hydraulic boost became popular, it became possible to improve the fly- ing qualities in certain flight regimes by feeding back a sensed variable, such as yaw rate. The control system with simple augmentation still maintained full author- ity through mechanical connections between the pilot and the control surface. In the event of a failure of a control system component, the pilot still maintained control, but with reduced flying qualities. The current trend of fly-by-wire control systems requires redundancy of critical elements since there will no longer be mechanical connections between the pilot and control system as a backup. The concepts of fault tolerance, detection, and isolation are new areas of important research.
CONTROLSTICK CONTROLSTICK n SENSOR t COMPENSATOR SIMPLE AUGMENTATIONUSED FOR MECHANICAL GOOD FLYING QUALITIES CONTROLLER FORCESTICK EXTENSIVE AUGMENTATIONREQUIRED FOR SAFEI-Y OF FLIGHT APPLICATION OF RSSAS To A CURRENT TRANSPORT CONFIGURATION Relaxed Static Stability Augmentation Systems (RSSAS) for transport configura- tions is one application of advanced control systems that may result in significant benefits. Immediate performance gains can usually be realized through a reduction in trim drag. Further gains can be achieved by resizing the horizontal tail due to a reduction in the stability constraint for the inherent aerodynamic stability of the aircraft. Good flying qualities will be achieved by the active control system. The reduction in tail area results in a decrease in aircraft operating weight and drag.
All of these benefits yield fuel savings of 2 to 4 percent for most transport config- urations.
30% DECREASED RESIZED TRANSPORTTO TARE ADVANTAGEOF RSSAS can be achieved by introduc- The greatest benefits of utilizing a RSSAS system ing the concept at the conceptual design stages. A reduction in tail area results in Hence, the wing and engine can be resized, resulting in weight and drag savings.
more weight savings. Additionally, the fuselage and landing gear structure can be redesigned for the lighter weight. In fact, after the airframe modifications, a fur- ther reduction in tail area may be possible, resulting in another round of changes.
These benefits continue to cascade through the design but generally converge rapidly, resulting in a design which takes maximum, synergistic advantage of applying this new technology. If the concept is not introduced soon enough, the full benefits of RSSAS cannot be achieved. In the case of transport aircraft, fuel savings of 6 to 9 per- cent are possible.
ACTIVE CONTROL DESIGN4,
‘\\
PROGRESSIN DESIGN The actual integration of multiple decentralized control systems into a single centralized control system has also been a recent development which will result in augmented operational safety, performance, and capability as well as improved In present practice, each component of the vehicle is designed indepen- economy.
dently. Certain advanced designs require control systems for various aspects, such as flying qualities, engine performance, structural damping, and weapon control. Each subsystem typically has an independent controller which is directed by the crew or flight management computer. It is conceivable that independent controllers could work in harmony; but, it is just as likely that they will conflict with each other.
to integrate all the controls and design each subsystem A preferred approach is controller simultaneously. Such a system will tend to work in harmony in response to crew or computer commands.
INTEGRATED DESIGN OF COMPONENT DESIGN OF PROPULSION, AERODYNAMICS, PROPULSION, AERODYNAMICS, STRUCTURES AND CONTROLS STRUCTURESAND CONTROLS DIRECT CONTROL OF FLIGHT COMPUTER ENGI NE . CREW
I
CREW CONTROL FULL POTENTIAL OF INTEGRATED USE OF CONTROLS Once the use of advanced integrated controls has been hypothesized, there are Modern control theory allows the use of multiple many avenues that can be explored.
effecters allowing such things as wing warping, rolling tails, spoilers, leading-edge devices, Unconven- and thrust vectoring for control and performance enhancements.
tional flight modes, such as target alignment independent of flight path or side can then be contemplated. All of these functions cannot be force excursions, Instead, a total inte- properly used if a separate controller is designed for each.
grated control system design approach should be used to minimize the conflicts and optimize the overall performance.
TRAJECTORYOPTIMIZATION LIFE/ CYCLE EXTENS IONS UNCONVENTIONALFLIGHT MODES NGINE CON TROL ENHANCEDFLYING QUALITIES NOZZLE CONT ROL FLIGHT-PATH CONTROL WEAPON GUIDANCE AND CONTROL CONFIGURATION BENEFITS OPTIMUM PRELIMINARY DESIGN OF TRANSPORTS OPDOT (Optimum Preliminary Design of Transports) is a computer program developed at NASA Langley Research Center for evaluating the impact of new controls technolo- It provides the capability to look gies upon transport aircraft (see reference 1).
at configurations which have been resized to take advantage of active controls and provide an indication of econoniic sensitivity to its use or the requisite assump- tions. Although this tool returns a conceptual design configuration as its output, it does not have the accuracy, in absolute terms, to yield satisfactory point designs for immediate use by aircraft manufacturers. However, the relative accuracy of comparing generated configurations while varying technology assumptions has been demonstrated to be highly reliable making OPDOT a useful tool for ascertaining the synergistic benefits of active controls, composite structures, improved engine efficiencies, and other advanced technology developments.
ACTIVE -=EEL WITH FOR RElAXED STABlUTY ACTIVE CONTROLS OPTIMAL DESIGN METHODOLOGY The approach that is used by OPDOT is direct numerical optimization of an econo- mic performance index. A set of independent design variables is iterated given a set of design constants and data. The design variables include wing geometry, tail geometry, fuselage size, engine size, etc. This iteration continues until the opti- mum performance index is found which satisfies all the constraint functions. The analyst interacts with OPDOT by varying the input parameters to the contraint func- tions or to the design constants. The optimization of aircraft geometry features is equivalent to finding the ideal aircraft size, but with more degrees of freedom than classical design procedures will allow.
WING AREA ECONOMICS WING ASPECT RATIO FIXED GEOMETRIES TECHNOLOGYLEVEL ‘$;:;;EA;N’$;.
ETC.
L/D WT.
GEOMETRYCONSTRAINTS CONTROL POWER \ FUNCTION / ETC.
I- 1
NO + YES CONSTRAINED
I SOLUTION I
PERFORMANCE FUNCTION FLOW DIAGRAM The performance index in OPDOT is computed by having a candidate configuration "fly" an entire mission while satisfying reserve fuel requirements. Industry statis- tics are used for estimating weights and costs. The stability and control analysis is similar to Datcom-type capabilities, and the program computes the interference drag in a general way, making OPDOT sensitive to tail sizing considerations. The flight profile is a multiple-step model of a suboptimal cruise/climb for optimum fuel effi- ciency. The program.is fairly flexible to use and has graphics output to illustrate each configuration.
rPERFORMANCE c ROI * DOC 1 INDEX .
/ COMPUTATION COST I I nwt FEATURES 1 CRUIS; STEP 1 l INDUSTRY STATISTICS FOR COSTS AND WEIGHTS ENGINE
I I
l DATCOM-TYPESTABILITY AND CONTROLDERIVATIVES l MULTIPLE-STEP,SUBOPTIMAL CRUISE/CLIMB @GENERALIZED INTERFERENCE DRAG METHODOLOGY FOR CONDUCTINGSENSITIVITY STUDIES A study is performed by inputting a set of problem parameters and selecting an initial set of independent design variables. OPDOT finds a solution, and that configuration is saved for later comparison.
The analyst then systematically varies a design constant or constraint function, andeach optimum design is stored.
Then a locus of optimum designs can be plotted as a function of the parameter in question.
This plot can be used to determine the sensitivity of a design to applying a new technology, for example, and each point includes the maximum synergistic benefits available for the set of inputs specified.
OPDOT OPTIMAL (Optimize design based on DESIGNS DESIGN CONSTANTS J an economic index) I I- - L- I DESIGN CONSTRAINTS ’ t I CHANGE A DESIGN CONSTANT I OR DESIGN CONSTRAINT - x 103 r INCOME CONFIGURATION 1 REQUIRED FOR A
J
FIXED ROI, $/FLIGHT -- I I I I l I 5000 6000 7ooO 8000 9000 10000 FIELD LENGTH, ft FL,YING QUALITIES STUDY One study that was made with OPDOT (references 2,3) was the evaluation of the impact of minimum acceptable flying qualities upon aircraft design. This is the prime factor which influences aircraft design when RSSAS systems are considered. It is assumed that an RSSAS system will augment the flying qualities up to more than acceptable levels, but provisions must be made in the event the autopilot/augmenta- tion system fails. Transport aircraft will generally have mechanical backups, so they should have sufficient unaugmented stability to assure the flight can be completed after a set of failures. Clearly these requirements, in effect, specify the inherent aerodynamic stability characteristics of the configuration. OPDOT will give the designer and regulators economic sensitivities to these criteria, enabling a proper compromise between safety and economy to be made.
During the course of this study, it was found that many of the criteria being considered for unaugmented flying quali- ties of transports with RSSAS were inadequate or inappropriate for specifying airplane design parameters.
o LEVEL OF UNAUGMENTED FLYING QUALITIES DETERMINES INHERENT STABILITY CHARACTERISTICS o ECONOMIC SENSITIVITIES FOR THESE CRITERIA WERE FOUND o MANY CRITERIA WERE INADEQUATE FOR PROPERLY SPECIFYING THE UNAUGMENTED FLYING QUALITIES IMPACT OF STATIC MARCIN considering the impact of relaxing the static stability A study was made requirement for transport aircraft. A locus of optimum designs is plotted. For the configuration being considered, a savings of 2.5 percent in direct operating cost is possible when compared to a baseline configuration with 5-percent static margin.
This corresponds to a fuel savings of 6 percent. At a certain point, in this case at -7 percent static margin, reducing the static stability constraint yields no further improvements. This is because the control constraints (typically nose-gear unstick during takeoff) override the tendency to make the tail smaller. A certain minimum size tail is required for control, and the center-of-gravity cannot be moved any further aft without sacrificing nose gear steering traction.
5-
\ \ \
4-
\
3-
2-
l-
SAVINGS
O-
IN
-1 -
DOC, %
-2 -
OVERR I DE
-4
0 BASELINEWITH NOMINAL SM (5%)
-5
I I I I
I I
-6
-;0 -15 -L -25 -10 0 5 10 15 20
MINIMUM ALLOWABLE STATIC MARGIN, %MAC
IMPACT OF LOADABILI'N UPON LXX Implied in the static margin sensitivity study was a range of allowed center-of- The control constraints are usually critical on the forward c.g.
gravity travel.
limit, and the stability constraints are usually critical on the aft c.g. limit.
Reducing this range results in savings for all static margins under consideration.
However, most benefits are achieved during the first 50 percent of reduction, indi- cating that if more careful center-of-gravity control is possible, a fuel savings of 2 percent or more is possible.
ALLOWABLE CGTRAVEL
w 1.22 METERS
----a .61 METERS
---a 0 METERS
~=-=-=p&.~-
Q--Q-
8-o
6-
PERCENT SAVINGS
IN DOC
4-
2-
I I I I I I
I
L I I
-15 -10 -5
O -20 0 5 10
15 20
STATIC MARGIN.% mat
IMPACT OF LANDING GEAR LOCATION UPON DOC Also implied in the static margin study was an aft limit for placing the landing gear. Studies have shown that the maximum aft placement for transport aircraft, where the gear and wing are collocated for structural efficiency, is about 65 percent of the mean aerodynamic chord. This is a critical constraint for RSSAS aircraft since it limits how far aft the center of gravity can travel before traction for nose gear steering is lost. Savings of nearly 1 percent in direct operating cost are pos- sible if the gear could be located further aft without structural weight penalty.
This corresponds to a fuel savings of over 2 percent.
M IL-F-8785B CRITERIA
B LEVEL I
10 - -------II LEVEL71 -----t7 LEVELIII 8- PERCENT SAVINGS 6 _ ---,*-” IN DOC ~-- ---=~===4= --
4-
2-
n 0 I Q I I I I o- (y-
.65 .70 .75 .80 .85 .90 .95 1.0
LOCATION OF LANDING GEAR, XLG/C DOC SAVINGS VERSUS TIME-TO-DOUBLE Another unaugmented flying qualities criterion that may be of interest is time-to- This plot illustrates the possible importance of economic sensi- double amplitude.
tivity to a proposed criteria. If a designer or regulator is considering applying a it is easy to see that the economic benefits of constraint of 30 or 40 seconds, relaxing the constraint from 30 to 40 seconds is of little economic consequence.
However, the opposite is true if considering an arbitrary boundary ranging between 2 and 6 seconds. The economic sensitivity information should be used before establish- ing the flight qualities criteria boundary since it significantly impacts the air- craft design.
6-
PERCENTSAVINGS
IN DOC
4-
2-
I I I I I
I
0 10 20 30 40 50 60
TIME-TO-DOUBLE, T2' set
IMPACT OF LOAD ALLEVIATION Gust load alleviation and maneuver load alleviation are active controls concepts Utilization of these technologies impacts the that have potential economic payoff.
design because the structure could be designed to a lower limit load factor resulting Plotted is the savings in empty weight and direct operating in a weight savings.
The dotted line just reflects cost for incremental reduction in limit load factor.
The solid line benefits of the lighter structure for the baseline configuration.
includes resizing the airframe to take advantage of the weight savings from active controls.
OPTIMALLY RESIZED
-- - BASELINE CONFIGlJRATION
E R
”
SAVINGS
7 - OPERATINGEMPTY WEIGHT
WITH
6-
RESPECT
TO
BASELINE,
%
.7 l 8
.5 .6
.l .2 .3 .4
INCREMENTALREDUCTION IN DESIGN LIMIT LOAD, m’s
I I I,,,, I 11.11 I I I I I, mm111.11. _.I. I --I ..,, . ,, ,., . -. . ..--__. --- -.--- --- RESEARCH USING OPDOT Other studies have been performed using OPDOT, including the investigation of the relative benefits of applying general technology improvements to transports and the evaluation of required economic and mission assumptions. Recently, a study was completed which determined the economic viability of canard transports when compared to conventional aft tail configurations. Future studies planned include the comple- tion of vectored thrust integration for transports; multi-body and multi-surface configuration with canard, wing, and aft tail evaluation; and commuter transport technology requirements.
OTHER STUDIES o RANKING OF OTHER GENERIC TECHNOLOGY IMPROVEMENTS
(REFS. ‘-I,51
o EVALUATION OF ECONOMIC AND MISSION ASSUMPTION
(REF. 5)
o DETERMINATION OF ECONOMIC VIABILITY OF CANARD TRANSPORTS (REF. 6) FUTURE STUDIES o COMPLETION OF VECTORED THRUST STUDIES o MULTI-BODY AND MULTI-SURFACE TRANSPORT EVALUATION o STUDY COMMUTER TRANSPORT TECHNOLOGY REQUIREMENTS o EXTEND PROGRAM TO OTHER AIRCRAFT TYPES SUMMARY The integration of controls early in the design process is important because the implication of unaugmented flying qualities during control system failures impacts the aerodynamic design;because it is a requisite for the proposed technology improve- ments to achieve their full, synergistic potential; and, because flight test expense can be saved. Adjustments to the control laws after an advanced technology prototype has been built is no longer -an easy proposition. Hence, it has become increasingly important to include control technologist and design considerations during conceptual design. In this discussion, a computer program developed at NASA Langley was described which utilizes optimization techniques to evaluate economic sensitivities of applying new technologies at the preliminary design level of transport aircraft.
o IT IS BENEFICIAL TO INTEGRATE CONTROLS CONSIDERATIONS INTO BEGINNING OF DESIGN PROCESS CONTROL SYSTEM AND FAILURE MODE ASSUMPTIONS IMPACT INHERENT AERODYNAMIC DESIGN FULL POTENTIAL OF ALL TECHNOLOGIES CAN BE REALIZED fLIGHT TEST EXPENSE WITH RESPECT TO FLYING QUALITIES CAN BE SAVED o A TOOL USING OPTIMIZATION TECHNIQUES HAS BEEN DEVELOPED FOR TRANSPORT AIRCRAFT REFERENCES 1. Sliwa, Steven M.; and Arbuckle, P. Douglas: OPDOT: A Computer Program for the Opti- lnum Preliminary Design of a Transport Airplane. NASA TM-81857, 1980.
2. Sliwa, Steven M.: Impact of Longitudinal Flying Qualities Upon the Designof a Trans- port With Active Controls. AIAA Paper No. 80-1570, Aug. 1980.
3. Siiwa, Steven M.: Economic Evaluation of Flying-Qualities Design Criteria for a Transport Configured With Relaxed Static Stability.
NASA TP-1760, 1980.
4. Sliwa, Steven M.: Sensitivity of the Optimal Design Process to Design Constraints -7nd Performance Index for a Transport Airplane. AIAA Paper No. 80-1895, Aug. 1980.
5. Sliwa, Steven M.: Use of Constrained Optimization in the Conceptual Design of a Medium-Range Subsonic Transport. NASA TP-1762, December, 1980.
6. Arbuckle, P. Douglas; and Sliwa, Steven M.: Parametric Study of Critical Con- stralnts for a Canard Configured Medium Range Transport Using Conceptual Design Opti- mization. AIAA Paper No. 83-2141, Aug. 1983.
RELIABILITY AND MAINTAINABILITY ASSESSMENTFACTORS FOR RELIABLE FAULT-TOLERANT SYSTEMS Salvatore J. Bavuso NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT I _ .- ‘I”.
: A long-term goal of the NASA Langley Research'Centet is the development of a reliability assessment methodology of sufficient powerto enable the credible comparisou of the stochastic attributes,of one ultrareliable system design This methodology, developed over a.lOAyear period, is a' against others.
combined analytic and simulative technique. An analytic component is the Computer-Aided Reliability Estimation capability, third generation, or simply CARE III. A simulative component is the Gate Logic Software Simulator capability, or GLOSS.
This paper focuses on the numerous factors that potentially have a degrading effect on system reliability and the ways in which these factors that are peculiar to highly reliable fault-tolerant systems are accounted for in credible reliability assessments. Also presented are the modeling difficulties that result from their inclusion and the ways in which CARE III and GLOSS mitigate the intractability of the heretofore unworkable mathematics.
RELIABILITY ASSESSMENTGOAL A long-term goal of the NASA Langley Research Center is the development of a reliahi.lFty assessment methodology of sufficient power to enable the credible comparison of the stochastic attributes of one ultrareliable system design against others (fig. ,l). This methodology, developed over a lo-year period, is a combined analytic and simulative technique.
OBJECTIVE: DEVELOP A CAPABILITY TO ASSESS THE RELIABILITY OF ANY FAULT-TOLERANT DIGITAL COMPUTER-BASED SYSTEM, INCLUDING THE SYSTEM EFFECTS OF SOFTWARE Figure 1 COMBINED ANALYTIC SIMULATIVE METHODOLOGY The methodology for performing reliability assessments is based on the utilization of an analytic model that accounts for the long time constants of hardware and/or software failures and a separate analytic model that tracks the short time constants of system fault-handling mechanisms. These models, which are embodied in computer programs, in conjunction with a simulative model, make possible the reliability assessment of large, practical fault-tolerant systems (fig. 2).
The CARE III computer program (codeveloped by the Raytheon Company and the Langley Research Center (ref. 1)) provides an analytic capability. The GLOSS is a simulative capability that provides CARE III with stochastic fault-handling data. The GLOSS concept was demonstrated by application to the CPU of an avionic processor. A generalized GLOSS that provides a user-friendly hardware description language interface is currently being developed. The GLOSS was codeveloped by the Bendix Corporation and the Langley Research Center (refs. 2, 3).
Illrr!!!!
cccc A RRR EEEE . . . . . GGGGG L 00000 sssss sssss C AA RR E ! ! ! G L 0 0 s S C AAAAA RRR EEE ! ! ! G GG L 0 0 sssss sssss ! ! ! G L 0 0 C A ARR E G S S 11111111l CCCC A AR R EEEE . . . . . . . . . GGGGG LLLLL 00000 SSSSS SSSSS COMPUTER-AIDED RELIABILITY ESTIMATION GATE LOGIC SOFTWARESIMULATIDN A SIMULATIVE CAPABILITY AN ANALYTIC CAPABILITY Figure 2 TECHNOLOGICAL DEVELOPMENTLEADING TO CARE III The motivation for developing the combined analytic simulative methodology dates back to 1973. The long-term development of CARE 111 is depicted in figure 3. State-of-the-art reliability evaluators were typical of CARE, a computer program developed by the Jet Propulsion Laboratory, and TASRA (Tabular System Reliability Analysis), developed by Battelle Memorial Laboratories. The Raytheon Company and Langley jointly developed the CARE II, which provided a superset CARE model with an extensive fault-handling model. Langley was also involved in the development of CAST (Combined Analytic Simulative Technique), which provides the current Langley modeling concept. CAST was developed by the Ultra Systems, Inc. CARSRA (Computer-Aided Redundant System Reliability Analysis) was a spin-off from the Boeing ARCS (Advanced Reconfigurable Computer System) study. Langley has also been involved in numerous technology development studies, some of which are depicted in the figure. This long-term involvement has culminated in the development of the CARE III.
y-----yA~~~
S.O.A. 1 1973 1977 PRESENT I *FLIGHT CONTROL SYSTEM ASSESSMENT FLIGHT CONTROL CGMPUTER ASSESSMENT . TRANSIENT-FAULT MODELING @SOFTWARE-ERROR AND REDUNDANCY MODELING Figure 3 PROFOUNDOBSERVATIONS On our way toward developing the specifications for CARE'III, we found that for ultrareliable systems certain factors that previously were of little interest to the reliability analyst now potentially have a significant effect (fig. 4).
This is particularly true of systems with a flight crucial probability of failure of less than 10-g in a l- to lo-hour mission. An example of this observation is the latent (undetected) fault. We also realized that even complex assessment capabilities must be user-friendly; this is always a difficult task for complex capabilities.
PROBLEMS: 1. EVERYTHING IMPORTANT WHEN PF < 10-l 2. PROGRAM VERSATILITY vs CONVENIENCE AND EFFICIENCY Figure 4 HIGHLY RELIABLE FAULT-TOLERANT SYSTEMS TO WHICH CARE III IS APPLICABLE The class of fault-tolerant systems of most interest currently utilizes off-the-shelf processors or:computers (fig. ,5). These systems rely heavily on the ability of-the processors to detect system faults/errors, to identify the fault/error to the smallest reconfigurable unit, and to effect recovery.
,:' CONTROL t PANFI C XKIAL
I II
DATA BUSES
1’
(23 SHARED.
I-
5 DEDICATED I *GENERAL PURPOSE COMPUTER SPACE SHUTTLE SYSTEM SOFTWAREIiIPLfMENTED FAULT TOLERANCE (SIFT) MI L-STD I 10
gag-&a
VLTING ’ m INPUT SENSORS AND ELECTRONICS m OUTPUT CONTROLS AND ELECTRONICS DC-9-80 DIGITAL FLIGHT GUIDANCE SYSTEM FAULT-TOLERANT MULTIPROCESSOR Figure 5 COVERAGE- A MAJOR RELIABILITY DRIVER In ultrareliable fault-tolerant systems, the inability of a system to achieve perfect fault/error handling is often the dominant cause of system failure (fig. 6). The major contributor of diminished fault/error handling is the latent fault /error. The long-term (latent) accumulation of faults/errors poses a severe threat to the system's ability to detect and lnask out anomalies.
The modeling of fault/error handling adds a tremendous amount of additional complexity to the reliability assessment task.
THE PREDOMINANT CAUSE OF FAILURE IN ULTRARELIABLE DIGITAL SYSTEMS HAS BEEN SHOWNTO BE ATTRIBUTED TO FACTORS OTHER THAN HARDWARE SPARES DEPLETION COVERAGE- MEASURE OF SYSTEM'S ABILITY TO HANDLE FAULTS p=>sysT~~ l FAULT DETECTION l FAULT ISOLATION l RECONFIGURATION AND RECOVERY UNDETECTEDFAULT - LATENT FAULT Figure 6 DELINEATION OF HARDWARE AND SOFTWARE FAILURE AND ERROR MODELS The increased complexity is indicated by the number of additional fault/error models that now must be considered. The increase in the number of fault/error models that rmst be accounted for is largely attributed to use of the digital computer (which possess extensive memory capability) and very high system reliability requirements. An extensive memory capability is a two-edged sword in that not only are computational capability and flexibility enhanced, but the likelihood of latent faults and errors occurring is also increased.
Ultrareliability necessitates the consideration of design errors, which previously were considered to be insignificant. Each branch in the trees in Faults are hardware generated, figure 7 represents a fault/error model.
whereas errors are caused by a fault or by software design anomalies. Either one may be permanent or may appear to be transient or intermittent. The common piece-parts reliability analysis is shown as a permanent random hardware failure.
HARDWARE ANOMALY /. \ PERhI: Hf.ITRANS.
// \\ // \\ OESGN - IAB. RANDOM EXTERNALLY OESGN FAB. RANDOM EXTERNALLY INDUCE0 ERROR ERROR FAILURE INDUCED ERROR ERROR FAILURE
//\\
//\\
SGNAL POWER PHYSCAL EMI SGNAL POWER PHYSGAL EMI ERROR FAILURE FAILURE ERROR FAILURE FAILURE SOFTWARE ANOMALY \ / INTJTRANS.
PERM.
/I\ /I\ OESGN CODE EXTERNALLY DESGN CODE EXTERNALLY ERROR ERROR INDUCED ERROR ERROR INDUCED / \ / \ DATA PROCEDURE DATA PROCEDURE PAl7ERN ERROR PAllERN ERROR ERROR ERROR Figure 7 FUNCTIONAL DEPENDENCYTREE FOR A NEAR-FUTURE PROPOSED FLIGHT CONTROL SYSTEM Ultrareliable fault-tolerant systems increase the system reliability by A typical employing redundancy, which further compounds the modeling task.
proposed advanced reconfigurable flight control system would utilize triple voting of units for the sensors, processor memories, and actuator electronics (fig. 8). In this example, the number of units increased from 22 for a nonredundant system to 64 for the fault-tolerant architecture.
MPX A/D + Proc. & - I r Figure 8 POSSIBLE STOCHASTIC MODELING APPROACHES Until recently, the reliability analyst was forced to compromise the analysis of such large systems either by modeling sections of the problem at a time and/or by making simplifying assumptions to keep the size of the reliability model tractable (fig. 9). The difficulty in this approach is that it is time consuming and complex.
Perhaps more important, it is prone to error and is often unreproducible. Reliability models for the advanced reconfigurable system example shown in figure 8, which would include the details previously discussed (fig. 7), would require on the order of millions of states in the Markov modeling sense. For each state, there exists an ordinary differential a Markov model for this system would require the solution of equation. Thus, millions of differential equations, a task that is expensive, if not impossible.
--------• 0 MARKOV (CAST, ARIES, CARSRA, SURF) l COMBINATORIAL (CARE, CARE II) l KOLMOGOROV (CARE III) (REF. 4) Figure 9 ALTERNATE STOCHASTIC MODELING APPROACHES Aside from using the popular Markov technique, two other approaches come to mind. The combinational method is the traditional piece-parts technique (fig. 10). In applying this technique to a fault-tolerant system with a reasonable degree of complexity, one soon learns, as in the development of -CARE II, that the computational aspects become unmanageable and involve nested The Kolmogorov method, in conjunction with a integrals four or more deep.
state aggregation technique, overcomes the computational difficulties of both the Markov and combinatorCa1 techniques.
0 t4fmov (CAST, ARIES, CARSRA, SURF)
____I, e COMBINATORIAL (CARE, CARE II) -b l KOLMOGORW (CARE III) Figure 10 THE CARE III APPROACH The ability of CARE III to provide extensive fault occurrence and fault- handling models is largely attributed to its ability to cope with large state spaces and is made possible by the observation that the time constants associated with fault occurrence are on the order of lo4 hours whereas the time constants of the fault-handling.model are on the order of 10 -3 hours. This wide time separa- ration allows the fault occurrence model to be treated as being independent of the fault-handling model. Thus, the fault-handling model is evaluated without regard to fault occurrences (fig. 11).
The results of the fault-handling model are then combined with the fault occurrence model to produce the desired reliability outputs. The fault occurrence model is solved using Kolmogorov's forward differ- ential equations. The Kolmogorov technique is used ecause the state reduction process discussed above necessarily requires the solution of a nonhomogenous (time-dependent failure rates) Markov process.
APPROACH o DEFINE SYSTEM STATE ONLY IN TERMS OF NUMBEROF EXISTING FAULTS o INDEPENDENTLY EVALUATE TRANSITION PARAMETERS AS A FUNCTION OF DISTRIBUTION OF POSSIBLE FAULT TYPES AND STATES o DETERMINE RELIABILITY USING KOLMOGOROV'S FORWARD DIFFERENTIAL EQUATIONS TASK o NUMBEROF STATES DRASTICALLY REDUCED, TRANSITION RATES NECESSARILY TIME DEPENDENT Figure 11 A MIXED MARROVMODEL AND ITS STATE-REDUCED AGGREGATED RELIABILITY MODEL An ilLustration of the state reduction technique can be seen by observing the reliability model of a two-unit system (fig. 12(a)).
States 0, 1, and F are the fault occurrence states. The states enclosed in the dashed lines are the fault-handling states. The two-unit model is a mixture of a nonhomogenous and a semi-Markov model, which is the type of model CARE III was designed to evaluate. The model that CARE III actually evaluates is the aggregated reli- ability model shown in Figure 12(b). The aggregated model is a nonhomogeneous Markov model. CARE III approximates the mixed process with a nonhomogeneous Markov process and can do so because of the wide separation in time constants in the fault occurrence and fault-handling models. In the aggregate model, the states are strictly fault occurrence states (defines number of failed units). The fault-handling model information contained in the dashed box of the two-unit system is mapped into the time-varying transition rate a'(t). The nonhomogeneous aggregated Markov model is solved using the Kolmogorov solution technique to produce time-varying probabilities of being in states 0, l', and F (the failure state) over the desired mission time.
Although the state reduc- tion wasn't too dramatic for this simple example, in practical assessments, state reductions of 6 orders of magnitude have been estimated.
t = GLOBAL OR MISSION TIME AGGREGATED RELIABILITY MODEL t' = Tlh'lE FROM ENTRY TO STATE A #ITH I = 2.~. LI (t) = A T = TIME FROM ENTRY TO STATE AE (a) (b) Figure 12 CARE III FAULT-HANDLING MODEL .The ability of CARE III to model the fault/error models delineated in figure 7 is made possible by CARE III's single- and double-fault models through the judicious selection of the appropriate transition rates and/or state holding probability density functions. The double fault model accounts for critically coupled coexisti.ng failures, which are user defined. The critically coupled failures, when they exist, are defined by certain combinations of pairs of states in the single-fault model (e.g., failure of two critically coupled units each in state A will cause system failure).
The structure of the single-fault model can be grasped by referring to figure 13, the reliability model of a two-unit system.
Initially, the system is in state 0 and has experienced no failures.
When a failure occurs, the system enters state A, the active latent state.
This arrival is governed by the arrival density Depending upon the x(t >= nature of the failure (i.e., permanent, transient, intermittent, etc.), the fault-handling model will be defined differently. For example, if the failure is intermittent, x(t) would be the probability density function (pdf) for the arrival of an intermittent, and states A and B define the intermittent model where c1 and f3 are constant transition rates into and out of state B.
When the system is in state B, the benign state, the failed unit appears to have healed itself, that is, the manifestation of the failure, a fault, vanishes. However, when the failed manifestation is once again resumed (the fault reappears), the system enters state A, where the failure looks like a It could be detected by a self-test program with pdf permanent failure.
6(t'), and the system would enter state AD, the active detected state.
If a spare exists, the system will purge the faulty unit and switch in the spare (dashed arc to state 1). Alternatively, while in the active state, the fault could generate errors with pdf The system then will enter AE, the dt ' >.
active error state. The intermittent failure could manifest its intermittent state again, and the system would then enter state B the benign error state.
iGi' Although the failure is benign, the error may not be enign and may cause system failure which is denoted by the to F transition (1-c)~(-r).
BE The error detection density is E(T), and 1-C is the proportion of errors from which the system is unable to recover. While in state BE, the error could be detected and corrected. In this event, the system enters state BD (benign detected) by transition CE(T>. At this point, the system may choose to do nothing further with the detected and corrected error and so move to the benign state, or the system may choose to reconfigure out the module containing the error and therefore move to state 1. The dashed arcs are instantaneous transitions. The other transition out of state is to state F, the % single-point failure transition (l-c)E(T). This transition is similar to the In a well-designed fault-tolerant system, (l-c)&(~) kou',"d b: n~~~%:~? If x(t) is the pdf for the arrival of a transient, a would be set to a value greater than zero and B would be equal to zero. The pdf x(t) for the arrival of a permanent failure would be defined so that a = H = 0. The dashed arc going from state AD to A enables the analyst to include the effects of the system decision that the detected fault which took the system from state A to AD was, in fact, a transient. In this regard, the system would not reconfigure out a nonfailed module.
The reader will note that the reliability model has three measures of time associated with it, which necessarily makes the model a semi-Markov process.
This added complexity is required because the behavior of the system is dependent on the onset of the various fault-behavior events. The availability of data for the fault-handling models is unfortunately still poor at best and is often nonexistent altogether. The creation of the data is the subject of a The GLOSS capability alluded to in considerable amount of current research.
figure 2 was used to estimate 6(t') and E(T) for permanent faults in the CPU of an avionic miniprocessor. (See fig. 13.)
Although the literature has often reported that transient faults are by far the most frequently occurring anomaly, virtually no test data exist that can be used for modeling transient occurrences or transient fault handling.
Test data for intermittent faults are also sparse (ref. 5).
In view of the extreme sensitivity that reliability assessments of ultrareliable systems show to best-guess transient and intermittent failure one can only wonder why such data are not abundant.
occurrence data,
t = GLOBAL OR MISSION TIME
t’ = TIME FROM ENTRY TO STATE A T = TIME FROM ENTRY TO STATE AE Figure 13 CONCLUDINGREMARKS The reliability assessment of ultrareliable fault-tolerant systems adds new dimensions of complexity to the assessment methodology (fig. 14). New tools are emerging to assist the reliability analyst to cope with the additional modeling complexities.
The availability of data for these novel tools is, however, slow in coming and will no doubt stunt the progress of developing ultrareliable fault-tolerant systems.
l NOVEL POWERFULASSESSMENTMETHODOLOGIES ARE EMERGING: CARE III AND GLOSS a AVAILABILITY OF DATA IS SPARSE l LACK OF SUFFICIENT DATA WILL STUNT THE GROWTHOF ULTRARELIABLE DIGITAL SYSTEMS Figure 14 REFERENCES Advanced Reliability Modeling of Fault-Tolerant 1. Bavuso, S. J.: Computer-Based Systems. Electronic Systems Effectiveness and Life Cycle Costing, J. K. Skwirzynski ed., Springer-Verlag, 1983, pp. 279-302.
2. McGough, J. G.; and Swern, F. L.: Measurement of Fault Latency in a Digital Avionic Processor Part II. NASA CR-145371, 1983.
3. Bavuso, S. J.; McGough, J. G.; and Swern, F. L.: Latent Fault Modeling and Measurement Methodology for Application to Digital Flight Controls.
Advanced Flight Control Symposium, USAF Academy, Colorado Springs, Colo., August 1981.
Introduction to Probability Theory and Its Applications.
4. Feller, William: Third ed. Wiley and Sons, 1968.
5. O'Neill, E. J.; and Halverson, J. R.:. Study of Intermittent Field Hardware Failure Data in Digital Electronics. NASA CR-159268, 1980.
SESSION II PARAMETERAND STATE ESTIMATION AND OPTIMAL GUIDANCE TECHNIQUES Lawrence W. Taylor, Jr.
Session Chairman PRACTICAL ASPECTS OF MODELING AIRCRAFT DYNAMICS FROM FLIGHT DATA Iliff and Richard E. Maine Kenneth W.
Ames Research Center Dryden Flight Research Facility Edwards, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Virginia Hampton, October 25-27, 1983 The purpose of parameter estimation, a subset of system identification, is to estimate the coefficients (such as stability and control derivatives) of the aircraft Model differential equations of motion from sampled measured dynamic responses.
structure determination, which is another aspect of systems identification, is discussed elsewhere.
Statement of Aircraft Parameter
Estimation Problem
Estimate the coefficients (parameters) of the
aircraft differential equations of motion from
sampled measured dynamic responses
In the past, the primary reason for estimating stability and control derivatives As aircraft from flight tests was to make comparisons with wind tunnel estimates.
became more complex, and as flight envelopes were expanded to include flight regimes that were not well understood, new requirements for the derivative estimates evolved.
For many years, the flight-determined derivatives were used in simulations to aid in flight planning and in pilot training. The simulations were particularly important in research flight-test programs in which an envelope expansion into new flight Parameter estimation techniques for estimating stability and regimes was required.
control derivatives from flight data became more sophisticated to support the flight- test programs. As knowledge of these new flight regimes increased, more complex Much of this increased complexity was in sophisticated flight aircraft were flown.
control systems. The design and refinement of the control system required higher fidelity simulations than were previously required.
Uses of Flight-Determined Estimates
Correlation studies
Handling qualities documentation
Design compliance
Simulation
Flight planning (envelope expansion)
Pilot training
Control system design
Linear analysis
Nonlinear simulation
Pilot in the loop
The maximum likelihood estimator is used to obtain the stability and control This is done by minimizing the cost function J(5) derivatives from flight data.
where the unknown derivatives to be estimated are in the vector 5. The term J is the weighted outer product of the difference between the measured response and the computed response, based on the current value of 5. For the stability and control derivative problem, we can assume the state and measurement equations are linear, although they need not be for maximum likelihood estimators in general.
Maximum Likelihood Estimator
State Equation
ic=Ax+Bu+q
Observation Equation
pi = CXi + Dui + ni
Minimize Cost Function
-g ([)]*R-1 [Zi-Zi (c)l+ Y2N InIRI
J(t)= f, Pi
i=l
Where
q is computed estimate of Zi
5 is vector of unknowns
5 contains only the roll- If we look at the case where the vector of unknowns we can see some of the essential features of the damping and the roll-control power, minimization of the cost function. The cost function is shown here as a function of The these two unknowns for a set of simulated data with added measurement noise.
minimum is shown, as well as the true value used in simulation. The reason for the This is also true for the case of real-flight difference is the measurement noise.
data, where the measurement error may also be caused by modeling error. The maximum likelihood estimate is at the minimum of the cost function.
Cost Function Surface Near Minimum
If we slice through the surface at constant values of the cost function, we can depict the cost function with isoclines. If we are far from the minimum (lowest isocline value), the isoclines are not elliptical. As we approach the minimum, the isoclines become more closely elliptical or nearly quadratic. Most minimization techniques take advantage of the quadratic nature of the cost function near the minimum.
Cost Function lsoclines
1 J \\\\\\
-1
$4 I -2.0 I -1.6 I -1.2 -.8 I -.4 0 .4 -8 1.2 1.6 2.0 2.4 28 3.2 LP The F-14 is a twin-engine, high-performance fighter aircraft that has variable wing sweep capability. The F-14 program addressed improvement of airplane handling qualities at high angles of attack by incorporating a number of control system tech- niques. The first part of the program was dedicated to obtaining flight-determined The flight conditions covered the subsonic enve- stability and control derivatives.
lope of the F-14, which is the complete trimmed angle-of-attack range for Mach numbers of 0.9 and below.
F-14AlRPLANE CONFIGURATION
This figure shows the flight-determined damping in roll(C~p) as a function of angle of attack (a) for low Mach numbers ((0.55) and for a Mach number of 0.9. There CR because was some uncertainty in the accuracy of the wind tunnel predictions of P the tunnel model configuration was different from the flight configuration. These flight data agreed with the trends found in the tunnel; with the proper interpreta- tion, even the magnitudes were in fair agreement.
DAMPING-IN-ROLL ESTIMATES
0 M < 0.55
o M = 0.90
- q -. 6 50 60
0 IO 20 30 40
ANGLE OF ATTACK, DEG This figure shows the flight-determined values of dihedral effect (Gas) as a func- tion of a compared with the results of two different sets of wind tunnel results.
There was some concern about the disagreement of the two sets of wind tunnel results before flight. At low angles of attack, the three sets of estimates are in fair agreement; however, at angles of attack above 15*, the flight data lie between the sets of tunnel data.
DIHEDRAL EFFECT ESTIMATES
FLIGHT DATA, M < 0.55 cl 0 --- LOW RE(l x 106) 1 --- HIGH RE (6.6 x 106) ) W’ND TUNNEL - .002 5 ’ - .004 P DEE’ - .006 - .008 I 1 I I ~- I -.OlO - 0 ,. 20 30 40 50 ANGLE OF ATTACK, DEG The F-14 data in this figure show the sensitivity with which we can determine sta- Rolling-moment coefficient as a result of differen- bility and control derivatives.
(Cg6 ) is shown as a function of angle of attack. It is tial spoiler deflection sP apparent that there is about 10 percent to 20 percent more effectiveness with the direct lift control (DLC) off. The difference between DLC on and DLC off is a small With the DLC on, configurational change. the spoilers are positioned 4O above the with the DLC off, the spoilers are positioned along the wing contour.
wing contour; Therefore, the 4O change in position results in a significant change in spoiler effectiveness, demonstrating the sensitivity with which the parameter estimation method can detect changes in vehicle characteristics that result from changes in con- figuration.
DIFFERENTIAL SPOILER
EFFECTIVENESS ESTIMATES
.ooi
r
0 DLCON o DLCOFF -.OOl 5 ’ tl SP -.002 DE&’ t- -.003 t -.004 1 0 4 12 16 ANGLE OF ATTACK , DEG The highly maneuverable aircraft technology (HiMAT) vehicle is a remotely piloted research vehicle with advanced close-coupled canards, wing-type winglets, and provi- sions for variable leading-edge camber. The flight-test philosophy was to fly the with the control feedbacks set to zero, to obtain sta- vehicle in a stable condition, bility and control derivatives. While these data were being gathered, a control system suitable for unstable flight was being designed, based on wind tunnel tests; Then, with the flight-determined derivatives, the simulator could be updated and the control system adjusted for this update so that the vehicle could be flown safely at a negative static margin. Stability and control maneuvers were performed at Mach numbers from 0.40 to 0.92, at angles of attack up to 100, and at altitudes from 15,000 ft to 45,000 ft. A complete set of stability and control characteristics was obtained for both the longitudinal and lateral-directional degrees of freedom.
HiMAT RPRV BASELINE CONFIGURATION
I- 15.56ft.
The HiMAT vehicle is constructed of advanced composite materials to allow for aeroelastic tailoring and to minimize weight. It is to be flown with a relaxed static margin because the wing deformation then results in a desirable camber shape at high load factor and the time drag is reduced. The vehicle was designed to fly with a sustained 8-g turn capability at a Mach number of 0.9 and an altitude of 25,000 ft, and to demonstrate supersonic flight to a Mach number of 1.4. To attain the Mach 0.9 condition, it is predicted that the vehicle must be flown at a lo-percent mean aerodynamic chord (MAC) negative static margin (unstable). The phi- losophy for testing HiMAT is somewhat different from that for testing production aircraft. Flight-determined stability and control derivatives are to be relied on to keep the wind tunnel program to a minimum. The original simulation data base con- tained the wind tunnel data, supplemented with some computed characteristics.
HiMAT TECHNOLOGY DEMONSTRATION
VEHICLE CONCEPT REMOTELY PILOTED CLOSE-COUPLED CANARD ADVANCED COMPOSITES AEROELASTICALLY TAILORED NEGATIVE STATIC MARGIN DESIGN POINT DEMONSTRATION SUSTAINED 8-G CAPABILITY SUPERSONIC FLIGHT TO MACH OF 1.4 The results of the flight test program showed that damping in yaw (Cnr) was twice the predicted value, yawing moment with respect to roll rate was the opposite (Cn p) sign, and rolling moment with respect to yaw rate (CQ,) was a small fraction of the predicted value. Rudder effectiveness (CnBr ) was 25 percent of the prediction, rolling moment due to rudder deflection (Cg6, ) was twice the prediction, and both yawing moment with respect to aileron deflection (CnGa ) and yawing moment with respect to elevon deflection (C ) were more positive than the prediction. Using the value found n6DE from flight data, the control system was changed markedly from the original control system, which was based on data from the limited wind tunnel program.
FLIGHT TO PREDICTION COMPARISON
(LATERAL-DIRECTIONAL)
(MINIMAL WIND TUNNEL PROGRAM)
DAMPING c” TWICE PREDICTION OPPOSITE SIGN OF PREDICTION sr P SMALL FRACTION OF PREDICTION “r CONTROL C 25% LESS THAN PREDICTION “4 cl TWICE PREDICTION C AND ’ MORE POSITIVE THAN PREDICTION “b ndDE The Space Shuttle is a large double-delta-winged vehicle designed to enter the atmosphere and land horizontally. The entry control system consists of 12 vertical reaction control system (RCS) jets (six up-firing and six down-firing) and eight horizontal RCS jets (four left-firing and four right-firing), four elevon surfaces, a body flap, and a split rudder surface. The locations of these devices are shown in this figure. The vertical jets and the elevons are used for both pitch and roll control. The jets and elevons are used symmetrically for pitch control and asym- metrically for roll control.
Shuttie Configuration
r U~firinglroll Yaw thrusters The flight-determined stability and control derivatives are used to update and refine the control system , modify flight envelope restrictions improve simulations, (placards), and improve flight procedures.
Uses of Estimates From Shuttle
Improve simulation
Control system refinement
Modify placard
Improve flight procedures
One of the interesting examples of where parameter estimation played an important role in the Shuttle program occurred during the first energy management bank maneuver on the first entry of the shuttle (STS-1). The computed response to the automated control inputs with the predicted stability and control derivatives is shown in this The control inputs shown here are the closed-loop commands from the Shuttle figure.
control laws. The maneuver was to be made at a velocity of 24,300 ft/sec and at a dynamic pressure of about 12 lb/ft2.
PREDICTED BANK
MANEUVER FOR STS-1
YAW-JET r SIDESLIP DEG ‘I/\
-A------
10 20 30 40 50 60 TIME. SEC The actual maneuver from STS-1 that occurred at this flight condition is shown in this figure. The flight data show a more hazardous maneuver than was predicted. At this flight condition, the excursions must be kept small. The flight maneuver resulted in twice the sideslip peaks predicted and in a somewhat higher roll rate than predicted. In addition, there was more yaw-jet firing than was predicted, and the motion was more poorly damped than predicted. It is obvious from comparing the predictions with the results of the actual maneuver that the stability and control derivatives are significantly different. Although the flight maneuver resulted in excursions greater than planned, the control system did manage to damp out the oscil- lation in less than 1 min.
With a less conservative design approach, the resulting entry could have been much worse.
Actual Bank Maneuver for STS-1
4- Yaw jet 0 Bank angle, 0 deg Roll rate, deglsec Yaw rate, deglsec Sideslip, o deg Time, set The obvious way to assess the problem with the first bank maneuver is to compare the flight-determined stability and control derivatives with the predictions.
Of all the derivatives obtained from STS-1, the most important one that differed most from predictions at the flight condition being discussed was $J# which is the rolling moment due to the firing of a single yaw jet.
Since the entry tends to monotonically decrease in Mach number, the derivative can best be portrayed as a function of the guidance system "Mach number," which is V/1000. This figure shows LyJ as a function of guidance "Mach number." Only the estimates from STS-1 are shown in these figures.
The prediction is shown by the solid line.
The symbols designate the estimates, and the vertical bars, the uncertainties. The dashed line is the fairing of the flight data.
Roll Due to Yaw Jet Estimates
0 Flight - Prediction - -- Fairing 8,000 c 4.000 T ‘YJ’
' t
ft-lbf I -I w -4,uuu r II 1 Id
L
12,000 - 0 4 8 12 16 20 24 28 Mach number The control system software is very complext it cannot be changed and verified so an interim approach was taken to'eliminate large excursions between STS missions, on future flights. The flight-determined derivatives were put into the simulation data base, and the Shuttle pilots practiced performing the maneuver manually to The maneuver was performed attain a smaller response within more desirable limits.
manually on STS-2 and STS-3. This figure shows the manually flown maneuver from STS-2.
The maneuver appears to be much better behaved, for roll rate (p), yaw The maneuver does rate (r), and angle of sideslip (6) are within the desired limits.
because the derivatives and the input not look like the original predicted response, are different, and the basic control system remains unchanged. Since the response variables are kept low and the inputs are slower and smaller, the flight responses on STS-2 through STS-4 do not show a tendency to oscillate. For STS-5 through STS-8, the control system automatically inputs the commands. The resulting maneuvers look nearly identical to the maneuver shown in this figure.
Bank Maneuver After
Problem Solved
Yaw ma.
w-c I I I I I -4 0 10 20 30 40 50 Tim, w Maximum likelihood parameter estimation techniques were used in the F-14 program to effect control system changes that improved the handling qualities of the aircraft at high angles of attack. The same techniques provided the primary source of infor- mation for the refinement of the control system for the HiMAT vehicle at negative static margin.
The energy management maneuvers have been redefined for the Space shuttle, based on simulations using flight-determined stability and control esti- mates. Moreover, parameter estimation techniques are being relied on for future control system design, placard modification or removal, and flight procedures for the Space Shuttle.
CONCLUSIONS
PARAMETER ESTIMATION IMPORTANT IN FLIGHT TEST
PARAMETER ESTIMATES USED TO
IMPROVE HANDLING QUALITIES
REFINE CONTROL SYSTEMS
UPDATE SIMULATIONS
MODIFY PLACARDS
CAREFUL SCRUTINY OF ESTIMATE NECESSARY
APPLICATIONS OF MODEL STRUCTUREDETERMINATION TO FLIGHT TEST DATA James G. Batterson NASA Langley Research Center Hampton, Virginia and Vladislav Klein George Washington University Hampton, Virginia First'Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 SYSTEM IDENTIFICATION DEFINED The following definition of system identification was suggested by Zadeh'in 1968 The three main objects that this definition connects and is widely accepted today.
are boxed. The harmonic content of the input to the system should be rich enough to excite the important modes of that system. A class of systems from which the model will be chosen is selected through engineering judgment or on the basis of a priori Finally, the decision must be made about a decision criterion specifying knowledge.
which model from the class is equivalent to the physical system under test.
IDENTIFICATION IS THE DETERMINATION.ON THE BASIS OF IInPuT(
AND OUTPUTOF A SYSTEMWITHIN A SPECIFIED CLASS OF SYSTEMS,
TO WHICH THE SYSTEMUNDERTEST IS )EQUIVALENT.J
IDENTIFICATION APPLICATIONS The benefits of airplane identification are shown here. The results of this procedure can be used in several areas indicated.
They are especially important since modern airplanes rely upon digital control and, hence, a good mathemati- today, cal model.
SIMULATOR MODEL CONTROL SYSTEM 4 I-, DESIGN FURTHER I_, BETTER MODEL TESTING BLOCK SCHEMEOF AIRPLANE IDENTIFICATION Airplane identification requires several steps. This presentation will concentrate on model structure determination.
c
FLIGHTDATA
DATAPREPARATION
COMPARABILITYCHECK
NOISEANALYSIS
.
PRELIMINARYANALYSIS
DATAAPPRAISAL
MODELFORMAT
MODEL DETERMINATION
I
PARAMETER EXTRACTION
REGRESSION
MAXIMUM LIKELIHOOD
,
MODELTESTING
I I
STEPWISE REGRESSION The stepwise regression is developed from the "classical" linear regression. It allows for the selection of important terms in the aerodynamic model equation. It linear model or a necessity for adding some nonlinear can show the adequacy of a terms.
ASSUME THE GENERALFORM OF THE AERODYNAMIC MODELEQUATIONS
CAN BE WRITTEN AS
(t)
y(t)= e* + elxlw +8*X2(t) + - - - + 0
Q-l ‘Q-1
THEN FOR EACH OF N OBSERVATIONS
y(i)=80 t 8pl(i) ’ 82x2(i) ’ --- “Q-lx(-j~~ + n(i)
th
OBSERVATION
WHERE n(i) IS THE EQUATIONERRORAT THE i
AS APPLIED TO THE VERTICAL FORCEEQUATION:
2 Q
;- aZ =Cz=CzD+ Cz (a-a(-))
a
qc t c (de -beg)
+c
Z
cy z6
n
t HIGHER ORDERTERMS
TECXNIQUES For the data from large amplitude maneuvers it can be beneficial to approximate the aerodynamic functions by splines rather than Taylor's series expansion.
The polynomial splines are written as functions of the "+" function, (a - ai)m for , knots of ai. This function has value (a - ~i)~ for a > ai and has va ue Cl for a < ai. In case of a function in two variables, a Two-dimensional spline should be used or the data can be partitioned in one of the two variables.
Data partitioning leads to a simplified model.
l ANOTHER REPRESENTATlONOf NONLINEAR MODEL IS A SPLINE
REPRESENTATION:
k
Cz=Cz +Cz a+ C Cz (a'-ai)+
0 a i=l ai
k
+c
qC t C Cz (a-a$ $+
i=l zq 2v
qi
k
+c be +
(a-ail + 6e
c
i=l
czb
“e
ei
l COM6lNE DATA FROM SEVERAL MANEUVERS AND APPLY STEPWISE
REGRESSION WITH POLYNOMINAL SPLINES TO COMBINED DATA SET
l COMBINE DATA FROM SEVERAL-MANEUVERSAND PARTITION AS A
FUNCTION OF ANGLE OF ATTACK, SIDESLIP, etc.
MODEL STRUCTURECOMPARISONAT DIFFERENT ENTRIES The model building using stepwise regression and spline approximation is demonstrated in fitting of the vertical force coefficient. The f.irst three entries into the regression show the effect of each term selected and the improvement in the Ei,t to the data.
Ct - cz loI+ cp 2.0 F a 4-0 + MEASURED - COMPUTED (a) IJ I.!- 12 18 24 30 ANGLE OF ATTACK, a (deg) ANGLE OF ATTACK, a (deg) 2.0 CZ=Ct(O)+CZa +A7fa-a7)++Bsla-016)+O4~/2V a F 1.5 -Q 1.0 0.5 12 18 24 30 ANGLE OF ATTACK, a (deg) I II IIII IllII l11111111llllllllIlIlIllIll11 MODEL SELECTION CRITERIA The stepwise procedure continues in selecting terms as long as they are statis- But for the selection of an adequate tically significant. ,model several criteria should be considered.
THE ACTUAL TERMS SELECTED FOR THE FINAL MODEL DEPENDON
SEVERALCRITERIA:
l THE PARTIAL F VALUE $ OF EACH TERM SHOULD BE
GREATER THAN 5
0 THE F STATISTIC SHOULD BE MAXIMUM FOR THE FINAL
MODEL
l R2, THE SQUAREDMULTIPLE CORRELATIONCOEFFICIENT,
SHOULD BE CLOSE TO 100 PER CENT FOR THE FINAL MODEL
l THE RESIDUAL SEQUENCE SHOULD BE RANDOM AND
UNCORRELATED
LONGITUDINAL PARAMETERSFROM DIFFERENT MANEUVERS GENERAL AVIATION AIRPLANE The comparison of results from small and large amplitude maneuvers is presented for the vertical force coefficient. The spline of first, second, and zero degree was used for the three Eunctions shown.
“Z& -1y Ql 0 O O a0 0 0 I L 0 I I ~~ I I I I I -2 ’ 12 16 20 24 28 0 4 8 Q, deg MODEL VALIDATION GENE% AVIATION AIRPLANE The model determined by stepwise regression from the data of a single large amplitude maneuver is validated by numerically integrating the equations of motion and comparing with the measured data of an independent set.
.
MEASUKED - COMPUTED 9* dtg/stc -20 DIRRCTIONAL STABILITY PARAMETERFROM LOW-AMPLITUDE MANETJVERS AND WIND TUNNEL MEASUREMENTS - ADVANCED FIGHTER of high-performance airplanes can be Nonlinearities in aerodynamics parameters These detected from the measured data by the application of stepwise regression.
The example shows the results can be compared with wind tunnel measurements.
directional stability parameter plotted against the angle of attack as obtained from small amplitude maneuvers and wind tunnel measurements.
< 5’ FOR FLIGHT AND WIND
. PI
.2
TUNNEL
‘h = ‘hTRIM
.l
TUNNEL
cn
P
-.l
-.2
a, dq
III II II IIIIIII lIIlI~l1lllll IIIIllllI I III DIRECTIONAL STABILITY PARAMETERFROM LOW-AMPLITUDE MANEUVERSAND WIND TUNNEL MEASUREMENTS - mvmc~~ FIGHTER (CONTINUED) The directional stability parameter determined from five large amplitude maneuvers is compared with wind tunnel measurements.
< 5’ FOR FLIGHT AND WIND
ISI
TUNNEL Oh = o”
-----FLIGHT
0 10 20 30 40
50 60
a, deg
DIRECTIONAL STABILITY PARAMETERFROM LOW-AMPLITUDE MANEUVERSAND WIND TUNNEL MEASUREMENTS - ADVANCED FIGHTER (CONCLUDED) The directional stability parameter determined from partitioned data is compared with wind tunnel measurements.
.
.2
< 5' FOR FLIGHT.AND WIND
I 4
TUNNEL
.l
----- FLIGHT’
C
“B O
-.l
-.2
0 10 20 30 40
50 60
a, deg
ESTIMATED LIFT CURVE FROM WIND UP TURNS-JET TRANSPORT For the jet transport, nonlinearities in the lift curve can occur at relatively even for small perturbed maneuvers it can be low angles of attack. Therefore, The example shows the necessary to use nonlinear aerodynamic model equations.
results from the wind up turns which were flown for airplane certification.
FRONT C,G, AFT C,G,
1.0
cl
0.8
6 10 I2
ALPHA
(DEG)
SUMMARY The technique presented in this paper has been published in references 2 and 3.
0 INCORRECT STABILITY AND CONTROL DERIVATIVES CAN RESULT
FROMAN INADEQUATE AERODYNAMIC MODEL STRUCTURE.
0 STEPWISE REGRESSION CAN BE USED TO DETERMINE THE STRUCTURE
FOR AN ADEQUATE MODEL.
0 SEVERAL STATISTICAL ANDINFORMATION CRITERIA NEED TO BE
CONSIDERED WHEN SELECTING AN ADEQUATE MODEL.
0 FLIGHT DATA WHICH COVERS A NONLINEAR AERODYNAMIC MODEL
RANGE MAY BE ANALYZED AS A SINGLE DATA SET ORPARTITIONED
INTO SEVERAL DISTINCT SETS.
0 STEPWISE REGRESSION FOR MODEL STRUCTURE DETERMINATION AND PARAMETER ESTIMATION HAS BEEN SUCCESSFULLY APPLIED TO THREE
AIRCRAFT TYPES (SINGLE ENGINE GENERAL AVIATION, UNAUGMENTED
MODERNJET FIGHTER.JET TRANSPORT).
REFERENCES 1. Zadeh, L. A.: From Circuit Theory to System Theory. Proceedings of I.R.E., vol. 50, no. 5, May 1962, pp. 856-865.
2. Klein, Vladislav, Batterson, James G., and Murphy, Patrick C.: Determination of Airplane Model Structure From Flight Data by Using Modified Stepwise Regression.
NASA TP-1916, October 1981.
Determination of Airplane Model 3. Klein, Vladislav and Batterson, James G.: Structure From Flight Data Using Splines and Stepwise Regression. NASA TP-2126, March 1983.
MIXING 4D-EQUIPPED AND UNEQUIPPED AIRCRAFT IN THE TERMINAL AREA L. Tobias, H. Erzberger, and H. Q. Lee NASA Ames Research Center Moffett Field, CA 94035 and P. J. O'Brien FAA'Technical Center Atlantic City, NJ First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 I 111 , .- , . . . . .-.-.. _- -- INTRODUCTION On-board 4D guidance systems, which can predict and control the touchdown time of an aircraft to an accuracy of a few seconds throughout the descent;have been developed and demonstrated in several flight test programs. However, in addition to refinements of the on board system, two important issues still need to be considered. First, in order to make effective use of these on-board systems, it is necessary to understand and develop the interactions of the airborne and air traffic control (ATC) system in the proposed advanced environment. Unless the total system is understood, the advanced on-board system may prove unusable from an ATC standpoint. Second, in planning for a future system in which all aircraft are 4D equipped, it is necessary to confront the transition situation in which some percentage of traffic must still be handled by conventional means.
In terms of 4D, this means that some traffic must still be given radar vectors and speed clearances (that is, be spaced by conventional distance separation techniques), while the 4D- equipped aircraft need to be issued time assignments. How to reconcile these apparent differences and develop an efficient ATC operation is the subject of this paper.
MIXING 4D EQUIPPED AND UNEQUIPPED AIRCRAFT IN THE TERMINAL AREA
--
--
v-\ I--- -
_---
OBJECTIVES The objectives of this study are to develop efficient algorithms and operational procedures for time scheduling a mix of 4D-equipped and unequipped aircraft in the terminal area, and, using the NASA Ames real-time air traffic control (ATC) simulation facility, to evaluate the system operation under various mix conditions.
. DEVELOP CANDIDATE OPERATIONAL PROCEDURES AND TIME- SCHEDULING ALGORITHMS FOR CONTROLLING A MIX OF 4D- EQUIPPED AND UNEQUIPPED AIRCRAFT IN THE TERMINAL AREA l EVALUATE THE SYSTEM OPERATION UNDER VARIOUS MIX CONDITIONS , , , , _. , _ . .
I _ . . . - __._.. -..-~- OPERATIONAL PROCEDURES The basic operational procedure is as follows: the ATC computer generates time assignments for all aircraft as they enter the greater terminal area.
For the 4D-equipped aircraft, the controller assigns the aircraft a route and a The 4D-equipped aircraft generates and flies the 4D route.
touchdown time. The controller was instructed not to alter this assigned time unless necessary for The unequipped aircraft must still be controlled by radar vectors.
safety reasons.
the controllers can use the position of the 4D aircraft to achieve the However, time assignments for the unequipped aircraft.
ATC COMPUTER GENERATES TIME ASSIGNMENTS . CONTROLLER ASSIGNS TOUCHDOWN TIME . 40 EQUIPPED: l AIRCRAFT GENERATES AND FLIES 4D ROUTE l ASSIGNED TIME NOT ALTERED . CONTROLLER ISSUES RADAR VECTORS . UNEQUIPPED: . CONTROLLER USES 40 AIRCRAFT POSITiONS TO ACHIEVE TIMES FOR UNEQUIPPED ON-BOARD SYSTF3 A complete on-board 4D guidance system is a complex entity involving interaction between numerous guidance, control, and navigation subsystems in an aircraft. The integrated collection of these subsystems augmented with special algorithms to provide fuel-efficient time control essentially constitutes the 4D flight manage- ment system of an equipped aircraft. The basic steps in the trajectory synthesis are shown below. For a number of years, NASA has designed and flight tested research systems incorporating various types of time control methods for both STOL and conventional aircraft. These tests have demonstrated the ability to predict and control arrival time accurately under varied operational conditions, achieving arrival time accuracies of f10 sec.
. AIRCRAFT SYNTHESIZES TRAJECTORY 1. HORIZONTAL PROFILE: TURNS AND STRAIGHT LINES 2. VERTICAL PROFILE: LEVEL FLIGHT AND CONSTANT DESCENT ANGLE SEGMENTS 3. AIRSPEED PROFILE: CONSTANT CAS AND DECELERATION SEGMENTS l ARRIVAL TIME ACCURACIES OF +lO set ACHIEVABLE . CONTROLLER CAN VECTOR AIRCRAFT; THEN ASSIGN NEW TIME VIA CAPTURE CAPTURE TRAJECTORIES A 4D-equipped aircraft which has been vectored off its 4D route can be assigned a This figure shows two aircraft revised time and a waypoint to capture the 4D route.
A capture trajectory is shown by a dotted line from position positions P1 and P2.
P1 to the capture waypoint 3. If the touchdown time associated with this trajectory is too early, the aircraft continues to fly according to its last vector where the pilot captures the 4D route via clearance until it reaches position P2, the trajectory shown.
FIXED TRAJECTORY CAPTURED TRAJECTORY SELECTED BY PILOT I RUNWAY PREDICTED CAPTURE TRAJECTORIES WAYPOINT Al RCRAFT POSITION TIME-SCHEDULING ALGORITHMS The 4D-equipped aircraft have the capability of meeting a touchdown-time assignment to an accuracy of a few seconds. It is now desired to use this capability to formulate efficient operational procedures for.the time scheduling of all aircraft in the terminal area. This will be developed in three parts: (1) determine the minimum time separation conditions given the minimum distance separations; (2) determine the interarrival time separations for two consecutive.aircraft to be used in aircraft scheduling; and (3) develop a scheduling algorithm for assigning landing times.
l TRANSLATION OF DISTANCE SEPARATIONS TO TIME l TIME SEPARATIONS AT TOUCHDOWN . INTERACTIVE SCHEDULING ALGORITHMS TRANSLATION OF DISTANCE SEPARATIONS TO TIME The minimum separation distance rules depend on aircraft weight category and are summarized in this figure. These distances can be converted to minimum separation times using speed profile data. The result is the matrix T, where each element is the minimum separation time at touchdown so that at no time when aircraft are along a common path is the separation distance rule violated.
MINIMUM DISTANCE SEPARATION SPEED PROFILES ALONG COMMON PATH COMMON PATH LENGTH: 5 n.mi.
TRAILING A/C 160- LARGE
IY
SMALL LARGE HEAL SMALL I 6 4 3 5 3 3 4 3 3 E Y 1st 6 130 f w TO LARGE LAND HEAVY
I
MINIMUM t11 t12 t13 TIME SEPARATION T = ‘21 t22 f23 MATRIX t31 f32 t33 I I TIME SEPARATIONS AT TOUCHDOWN It is assumed that, if two consecutive aircraft are 4D-equipped, the interarrival times given by T can be used for scheduling purposes. However, unequipped aircraft will need additional time buffers to prevent separation distance violations. If the probability density function of an unequipped aircraft meeting an assigned time via controller vectoring is known (this can be determined in the specific experimental context), then time buffers can be determined to keep the probability These time buffers result of separation distance violation belay a desired level.
in a revised time separation matrix T described below.
BUFFERS ADDED TO PREVENT MINIMUM SEPARATION VIOLATIONS FOR TWO CONSECUTIVE AIRCRAFT AT TOUCHDOWN: IF BOTH EQUIPPED, T’ = (t’ij) = (tij + aa) IF ONE EQUIPPED, T” = (t”ij) = (tij +Sb) IF BOTH UNEQUIPPED, T”’ = (tij’) = (tij +a,) WHEREoG a,< $,< 6, ' INTERACTIVE SCHEDULING ALGORITHMS The previous discussion established the time.separation matrix at touchdown shown as a function of:weight category, and whether or not aircraft are 4D equipped. It is assumed that the feeder fix time for each aircraft is known. Based on this time a desired touchdown time for each and on the desired time to traverse the route, aircraft can be determined. This information can be used to generate an initial time schedule, as described in reference 1. However, in addition to setting up an initial schedule, algorithms are required to revise the schedule. Missed approaches need to be accommodated.. Also, the controllers may need to change the aircraft arrival rate. It may be that they also are required to block out specific time periods from the computer schedule to accommodate a missed approach or a priority landing. These are important aspects of the complete scheduling problem.
WITH TIME SEPARATION CONSTRAINTS CAN NOW GENERATE SCHEDULE . ESTABLISH TOUCHDOWN ORDER . PROVIDE FOR REVISIONS - CHANGE ARRIVAL RATE - MISSED APPROACHES - EMERGENCIES EVALUATION OF SYSTEM OPERATION UNDER VARIOUS MIX CONDITIONS The candidate operational procedures and time schedule algorithms previously described were used in a real-time ATC simulation study of operations under various mix conditions.
. SIMULATION FACILITY . SCENARIO AND TEST CONDITIONS . RESULTS SIMULATION FACILITY The simulation was conducted using the NASA Ames ATC Simulation Facility shown in this figure. It includes two air traffic controller positions, each having its own one was designated arrival control In this study, color computer graphics display.
The controllers each communicate with one or two and the other, final control.
keyboard pilots. Each keyboard pilot can control up to 10 computer-generated aircraft simultaneously. The clearance vocabulary includes standard heading, speed, and altitude clearances as well as special clearances for 4D-equipped Previous studies have This figure also depicts piloted simulators.
aircraft.
utilized one or two piloted simulators which were connected by voice and data link to the ATC Simulation Facility; however, in this study, no piloted simulator was used.
DATA LINK TO FAA TECHNICAL CENTER ATC SIMULATION FACILITY KEYBOARDPILOT, GROUNDCONTROL KEYBOARDPILOT, GROUNDCONTROL STATIONS STATIONS STATIONS STATIONS
f r x
SIMPLE MOVING & FIXED BASE TEST AIRCRAFT PILOTED PILOTED FLOWN AT SIMULATORS AIRCRAFT SIMULATORS CROWS LANDING APPROACHCONTROLLERDISPLAY The route structure and runway configuration investigated are shown in this figure.
Two routes, Ellis, from the north, and Sates, from the south, are high-altitude routes flown by large or heavy jet transport-type aircraft. Aircraft on these routes fly profile descent procedures, but may or may not be 4D equipped. Hence, there is a mix of 4D-equipped and unequipped aircraft of the same speed class along the same route. In addition, low-speed aircraft were considered which flew the Deerpark route from the east, but shared a 5 n. mi. common path length and used the same runway as the jet traffic. The Deerpark traffic was unequipped, and always constituted 25% of the traffic mix. To assist the controller in integrating the 4D-equipped and unequipped traffic, a flight data table (FDT) was provided to the left of the route structure. The information supplied includes aircraft type, route, scheduled touchdown time, and anticipated delay. The main test variable was the mix of traffic. Three mix cases were run: 25, 50, and 75% 4D equipped.
13 : 27 : 12 ID TYPE FIT Rl 4 SA 37oll 00 I1 I44 SA 3824 00 Jl U EL 4305 -09 Tl 4 SA 4429 00 A2 HU EL 4715 05 DEERPARK E2 U SA M2 HU EL ----_ G2 H4 SA 5247 2549 ELLIS AND SATES: JET A/C 4D EQUIPPED OR UNEOUIPPED DEERPARK: LOW SPEED A/C MIX CONDITIONS: 0, 25, 50 75% 4D EQUIPPED EXPERIMENT CONDITIONS In this study, a saturated arrival traffic flow was used. It is assumed that instrument flight rule (IFR) conditions prevail, and that all aircraft use runway 4R; furthermore, no departures, winds, or navigation errors are simulated. For purposes of this study, it was assumed that all aircraft depart the feeder fix at Magnitude departure errors that can be their scheduled departure times.
tolerated as well as the means to provide ground computer assists to nullify departure errors are main issues addressed by current research.
. SATURATED ARRIVAL TRAFFIC FLOW l NO WINDS, NO NAVIGATION ERRORS l ALL AIRCRAFT DEPART AT SCHEDULED TIM& CONTROLLEREVALUATIONS Three research air traffic controllers from the FAA.Technical Center participated in this study. Controllers were asked to compare operations under the traffic mix conditions. The 25% equipped case was rated the condition with the heaviest workload. The main difficulty seemed to be that the controllers were establishing distance spacing of most of the traffic,, and they felt that by not altering the flight path of the 4D-equipped aircraft, they were occasionally.losing some slot time. They were, however, quite pleased with the 50% 4D-equipped case, which allowed for easy handling of the unequipped aircraft. The 75% 4D-equipped case was rated most orderly by all the controllers, but when this many aircraft were 4D- equipped (the only unequipped aircraft were the Deerpark arrivals, which always constituted 25% of the traffic sample), there was "basically nothing to do." The controllers were asked if there was any difficulty in handling the mix of speed classes, the slow traffic on Deerpark and the'jet traffic on Ellis and Sates.
They indicated that spacing behind the low-speed aircraft was sometimes a problem, since they had to allow for a large initiai separation along the common path length. The controllers indicated that the time order information displayed on the flight data table was useful; however, the touchdown time and delay information was not used.
. COMPARISON OF MIX CONDITIONS . CONTROLLING THE MIX . USE OF DISPLAYED TIME DATA AVERAGE NUMBER OF CLEARANCES This figure provides the average number of clearances/aircraft. It can be seen that as more aircraft are 4D ,equipped, the average number of clearances per aircraft This is fairly obvious in the experiment context described, since 4D- decreases.
equipped aircraft were not vectored. They were assigned a touchdown time which was not altered in most cases.
AVERAGE NUMBER OF % EQUIPPED CLEARANCES/AIRCRAFT 0 5.2 4.5 50 2.7 2.4 EFFECT OF 4D ON LOW-SPEED TRAFFIC The previous figure shows the decrease inthe average number of controller clearances as a greater percentage of aircraft are 4D equipped. A major concern is: does the average number of clearances for the.unequipped aircraft increase as the percentage of equipped aircraft increases? The answer to that question is provided in the figure below, which gives the average number of clearance/aircraft for the Deerpark route-only. Recall that the Deerpark traffic was always 25% of the traffic sample, and that all Deerpark is unequipped aircraft. This figure indicates that the average number of clearances given to the Deerpark unequipped aircraft is the same, independent of the mix condition. Also shown is the average time in the system (in minutes) for the Deerpark traffic, which is also seen to be independent of the mix condition.
LOW-SPEED (DEER PARK) TRAFFIC IS 25% OF ALL TRAFFIC IN EACH TEST CONDITION AVG. TIME IN SYSTEM, AVG. # OF CLEARANCES PER Al RCRAFT MIX minsec 0 19:16 6.9 25 18:56 6.5 50 19:05 6.2 75 19:05 6.4 LOSS OF 4D SCHEDULING There was a desire to examine how traffic handling is disrupted if a breakdown of the 4D scheduling computer occurs. To investigate this, during a 75% 4D-equipped run, the FDT was removed from the screen so that the controllers no longer had a display of schedule times and order for aircraft in their sector. Furthermore, all feeder-fix departures from then on would not have any 4D time assignment, and would have to be vectored. The map display which showed aircraft positions was not removed. Initially, there was no change. The 4D-equipped aircraft already in the control sector could still be left alone since they would continue to follow their previously assigned 4D route. This is in contrast to a totally ground-based 4D system in which the ground system generates clearances for every aircraft; when that type of system fails, all aircraft are affected in a short time. The only difficulty experienced with the system tested was that after the failure occurred, controllers continued to allow traffic to depart the feeder fixes at the higher arrival rate for the 75% equipped case, rather than to adjust to the baseline vector arrival rate. If the flow-rate adjustment for new feeder-fix departures is made when the failure occurs, then it seems clear that the use of the on-board 4D system provides a safe transition to the standard vector mode.
DETERMINE EFFECTS OF ATC COMPUTER OUTAGE OBJECTIVE: ACTION: DURING A 75% 4D RUN, FLIGHT DATA TABLE REMOVED OBSERVATION: NO INITIAL CHANGE (BUSY PERIOD). SOME PROBLEMS WITH HIGH ARRIVAL RATE OF NEW ARRIVALS CONCLUSIONS: ONBOARD 4D PROVIDES SAFE TRANSITION TO VECTOR MODE.
NEED TO ESTABLISH AS PART OF PROCEDURE AN IMMEDIATE CHANGE OF FLOW RATE FOR NEW DEPARTURES CONCLUSIONS Algorithms were developed to obtain an initial time schedule and to provide for revisions for a mix of 4D-equipped and unequipped aircraft in the terminal area.
These algorithms were used to develop a candidate set of operational procedures for mixing 4D-equipped and unequipped jet aircraft along the same route, and for mixing different speed classes along merging routes. A basic rule established was not to alter the 4D equipped aircraft once they were assigned a landing time. This procedure resulted in the controllers learning to use the 4D aircraft positions to effectively vector the unequipped aircraft to their assigned landing slot.
However, procedures were also demonstrated to vector the equipped aircraft and to reassign touchdown times. In addition, it was shown that a loss of the ground based 41, system results in a smooth transition to vector operations. Controller evaluations indicated that the 25%-equipped case was the most difficult to handle.
Nevertheless , quantitative data actually showed a decrease in the number of controller clearances with respect to the 0% 4D-equipped case. Controllers felt that the procedure of not altering the 4D-equipped aircraft when so few were equipped was workable, but that it was a more complex task. Nevertheless, fuel was saved even in this case, compared to 0% 4D-equipped aircraft. The controller workload as measured by the average number of clearances per aircraft decreased as the percentage of 4D-equipped aircraft increased. Moreover, this average decrease was not accomplished at the expense of the unequipped aircraft. The number of clearances for the unequipped aircraft as well as the time delays was independent of mix condition.
l DEVELOPED SCHEDULING ALGORITHMS AND OPERATIONAL PROCEDURES l ALL MIX CONDITIONS EFFECTIVELY CONTROLLED . REDUCED CLEARANCES AS PERCENTAGE 4D INCREASED . UNEQUIPPED NOT PENALIZED BY 4D REFERENCE Time Scheduling of a Mix of 4D Equipped and Unequipped Aircraft, 1. Tobias, L.: NASA TM-84327, 1983.
APPLICATION OF FUEL/TIME MINIMIZATION TECHNIQUES TO ROUTE PLANNING AND TRAJECTORY OPTIMIZATION Charles E. Knox NASA Langley Research Center Hampton, VA First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT Rising fuel costs combined with other economic pressures have resulted in industry requirements for more efficient air traffic control and airborne responded operations. NASA has with : an on-going research program to investigate the requirements and benefits of using new airborne guidance and pilot procedures that are compatible with advanced air traffic control systems and that will result in more fuel efficient flight. This paper summarizes the results of flight testing an airborne computer algorithm designed to provide either open-loop or closed-loop guidance for fuel efficient descents while satisfying time constraints imposed by the air traffic control system.
The paper will also describe some of the potential cost and fuel savings that could be obtained with sophisticated vertical path optimization capabilities.
DIRECT OPERATING COST Between 1970 and 1980, the average price paid by airlines for fuel rose approximately 1000%. In 1970, fuel costs represented about 25% of the flights' direct operating costs. In 1980, this percentage rose to between 60 and 70 percent.
In addition, inflation has caused crew costs and other non-fuel airline operating costs to increase. These increased operating costs combined with lower revenue levels arising from recessionary trends in the economy have led to an emphasis on achieving more economical operations through changes in procedures, flight operations, airborne equipment capability, and in air traffic control (ATC) operations.
TIME-BASED METERING PROCEDURES In response to the fuel crisis, t::e Federal Aviation Administration developed several programs to save fuel including an automated time-based metering (TMB) form of air traffic control for arrivals into the terminal This TBM concept provides fleet-wLie (all users) fuel savings through area.
time control by matching the airplane arrival rate into the terminal area to the airport's arrival acceptance rate. This procedure reduces the need for holding and for law-altitude vectoring for sequencing to land. Fuel savings are also achieved on an individual airplane basis by permitting the pilot to descend at his dlmcretion from cruise altitude to a designated metering fix in a fuel-efficient manner. Substantial fuel savings have resulted but air traffic control workload is high mince the radar controller maintains time management for each airplane through either speed commands or path stretching with radar vectors. Pilot workload is Increased since the pilot must plan for a fuel-efficient descent usually by using various rules of thumb.
NASA has flight-tested initsTransportation SystemResearchVehicle (TSRV) Boeing B-737 airplane a flight management descent algorithm designed to increase fuel savings by reducing the time dispersion of airplanes crossing the metering fix at an ATC-designated time by transferring the responsibility of time navigation from the radar controller to the flight crew. Time and were proaided to the pilot for an idle- path (4-D) closed-loop guidance thrust, clean-configured, constant Mach descent with transition to a constant airspeed descent to arrive at the metering fix at a time, altitude, and airspeed predetermined by ATC.
DRAKO METERING \ METERING .'
\ FIX FIX
/’
\
/
-- i--i
\ \ \ kL KIOWA METERING METERING FIX FIX AIRBORNE COMPUTED DESCENT PATH The NASA airborne flight management descent algorithm computes the parameters required to describe a seven-segment cruise and descent profile between an arbitrarily located entry fix to an ATC-defined metering fix.
(Segments 2 and 3 are computed if the flight will be restricted by the ATC 250 knot airspeed limit below 10,000 feet.) The computed parameters are then used by the airplane's navigation and display systems to present guidance to the pilot and/or autopilot.
The descent profile is based on linear approximations of airplane performance for an idle-thrust, clean-configured descent. Airplane gross and weight, wind, nonstandard temperature and pressure effects are also considered in these calculations. To be compatible with standard airline operating practices, the path is calculated based upon the descent being flown at a constant Mach number with transition to a constant calibrated airspeed and speed changes being flown at a constant altitude.
The flight management descent algorithm may be used in either of two modes. In the first mode, the pilot may input the Mach/airspeed descent schedule to be flown, and the descent profile is calculated independent of an assigned metering fix time. If a metering fix time is subsequently assigned, some time error, which must be nuLled by the pilot, may result since an arbitrary specification of the descent speed schedule may not satisfy both the initial and final time boundary conditions.
In this mode, The second mode was designed for time-metered operations.
pilot inputs include the estimated time of arrival to the entry fix and the ATC specified metering fix arrival time. The descent profile is then calculated based on a Mach/airspeed descent schedule, computed through an iterative process, that will closely satisfy the crossing times for both of these way points.
ENTRY TOP OF FIX DESCENT a CD *+* ;HANkit /* SPEED C""'-- CONSTANT FROM CRUISE TO CRUISE MACH DESCENT MAT-H CONSTANT MACH DESCENT AT IDLE THRUST /
#c
SEGMENTS 0 AND @ CALCULATED IF METERING FIX ALTITUDE < 10,000' CONSTANT CAS @ @ DESCENT AT AND DESCENT CAS '250kt IDLE THRUST ,*-- - / / 250kt CA@,' I / METERING FIX * OF DESCENT RESULTS AND FUTURE INVESTIGATION Research flight tests of the NASA flight management descent algorithm in the Denver Air Route Traffic Control Center time-based metered air traffic that time guidance and control in the cockpit were environment demonstrated Descent guidance acceptable to both the pilots and the ATC controllers.
presented on the airborne CRT flight instrumentation allowed the test airplane fix at the proper altitude and speed and to be flown across the metering significantly reduced the time dispersion occurring with other airplanes at The concept of closed-loop guidance time control in the the metering fix.
cockpit could be readily extended, with similar results, to other aircraft However, with integrated electronic navigation and guidance/display systems.
many airplanes flying in the time-based metering ATC environment do not have This research was these integrated electronic guidance and display systems.
then extended to provide the pilots of unequipped airplanes with simplified open-loop 4-D guidance. The issues in this research are a trade-off between performance and pilot workload and acceptance.
DENVER FLIGHT TEST RESULTS @ACCURATE PERFORMANCE (ALT, SPEED, 8 TIME AIRPLANES WITH _+2MIN*+lO SEC)
l POTENTIAL FUEL SAVINGS
l ACCEPTABLE PILOT WORKLOAD
r-V---- ------, I I I . .
SIMULATION AND SIMULATION AND FLEET WIDE FLEET WIDE i SIMPLIFIED 1 SIMPLIFIED I I b FLIGHT TESTS b FLIGHT TESTS .GA/CORP .GA/CORP l ALGORITHM - l ALGORITHM - I I , , I I
l AIR TRANSPORTS l AIR TRANSPORTS
NONINTEGRATED NONINTEGRATED
I I
I I
l MILITARY l MILITARY
GUIDANCE GUIDANCE l l I I INVESTIGATION INVESTIGATION 1 1 I I I ’ I RESEARCH ISSUES RESEARCH ISSUES I
I I
l PERFORMANCE l PERFORMANCE
I I I I COMPARISON COMPARISON I I I I
l PILOT WORKLOAD/ '
I I I ACCEPTABILITY L ---------- -I PROFILE DESCENT HAND-HELD CALCULATOR To determine the feasibility of providing open-loop guidance to the flight crew to make fuel-conservative, time- constrained descents to the metering fix, the NASA descent algorithm was programmed on a small, hand-held programmable calculator. All inputs required by the algorithm are made by the pilot through the keyboard.
All outputs are shown -Ln the calculator display.
Flight tests conducted with NASA test pilots in a T-39A (Sabreline) airplane indicated that it was feasible to fly the descents with open-loop guidance provided to the p-tlot in the form of a DME indication to define the top-of-descent point and the appropriate Mach and airspeed indications to use during the descent. The resulttng mean distance and time errors to actually achieve the predicted speed and altitude conditions at the end of the descent profile were 1.2 n. mi. long and 1.4 seconds early. A question remained, however, if open-loop guidance provided by a hand-held calculator would be pilot acceptable in an operational environment.
, .
. _ ..- PROFILE DESCENT HAND-HELD CALCULATOR UNITED AIRLINES FLIGHT TESTS Joint flight tests were conducted with United Airlines to determine if the concept of using open-loop guidance for fuel-conservative descents with a hand-held calculator during routine flight operations was acceptable to the The results of these tests showed that the majority of the pilots pilots.
participating in the tests felt that the open-loop guidance concept of the calculator provided useful information. Several test subjects felt that they could mentally compute the top-of-descent point and would not save additional fuel through use of the calculator. However, all of the test subjects agreed that the computations necessary to satisfy the metering fix crossing time constraints were too difficult for mental calculation and would require other means (such as the calculator) to provide guidance. All subjects agreed that the workload associated with using the calculator was low and would not interfere with normal crew tasks.
All of the test subjects expressed a concern that the ATC system would not allow them to fly a preplanned descent without being interrupted and thus suffer a fuel penalty. This concern was realized during these flight tests: 68% of the descents were modified with altitude restrictions or speed restrictions by ATC and required recomputation of the descent profile. This statistic emphasizes the requirement that compatibility must exist between the airborne and ground systems to realize significant fuel conservation.
. TEST RESULTS:
CONCEPT WAS ACCEPTABLE TO PILOTS
o HELPFULTO CREWFOR DESCENTPLANNING
o WORKLOAD LOW--DID NOT INTERFEREWITH NORMAL CREW TASKS
o INITIAL TRAINING REQUIREDLESS THAN ONE HOUR
l PERFORMANCE RESULTS--l6 DESCENTS
DISTANCE ERROR
TIME ERROR
SEC, N, MI,
1 EARLY MEAN 1,7
20,8 d 2,l
39 EARLY MAX 6,4
ADVANCED FLIGHT MANAGEMENT CONCEPTS The NASA flight management research activities also include defining the interface and guidance requirements necessary for practical implementation of sophisticated optimal path trajectory calculations. One of the corner stones of this research effort is the "OPTIM" computer program.
This program generates a full vertical path profile including climb, cruise, and descent based upon one of three selectable objectives: minimum cost, mintmum fuel, or fixed time/minimum fuel.
The OPTIM computed profile is generated from solutions of an energy state approach in which range and specific energy are used to describe aircraft state. A cost functional, which expresses the quantity (fuel or operating cost) which is to be minimized, is combined with the aircraft state equations to form a Hamiltonian with energy as the independent variable. As energy is incremented along the trajectory, airspeed and thrust are chosen to minimize the Hamiltonian. In this manner a complete vertical profile is generated along a pre-speciEied horizontal path.
This program presently is being used in a fast-time mode to examine parametric sensitivities and to define potential fuel and cost savings. The program has also been implemented into a real-time piloted simulation to define interface requirements between the pilot, the airborne guidance systems, and the ground-based ATC systems to ensure compatibility and efficiency.
“OPTIM” GENERATES VERTICAL FLIGHT RROFILES THAT MINIMIZE DOC AND SATISFY EXTERNALLY IMPOSEDTIME CONSTRAINTS - MINIMUM COST - MINIMUM FUEL - FIXED TIME, MINIMUM FUEL 0 FAST-TIME ANALYSIS l REAL-TIME SIMULATION - PARAMETRICSENSITIVITY - PI LOT/AIRPLANE SYSTEMSINTERFACE - FLIGHT PLANNING TOOL REQUIREMENTS FOR PRACTICAL - DEFINE POTENTIAL TRIP OPTIMAL FLIGHT PATHS FUEL AND COST SAVINGS - INTERFACE REQUIREMENTS TO ENSURE A/G COMPATIBILITY AND EFFICIENCY DIRECT OPERATING COST MINIMIZATION The direct operating cost (DOC) function used in the OPTIM program is a and the cost per pound of function of the cost of operation per hour Kl fuel K2.
The geometry (speed, altitude, and flight path angle) of the trajectory is a function of the ratio of Kl and KZ rather than the absolute The selection of the value of fuel cost is relatively straight- magnitudes.
forward. However, the selection of the operating costs per hour is much more complex to establish. These costs must be truely time variant costs rather than cyclic costs (i.e., costs associated with take-off and landing are cyclic, not trip time variant). The magnitude of the operating cost may also be biased higher or lower to change trip times (changes airspeed and altitude) to reflect corporate policy. Airline management should select the proper values for Kl and K2 to reflect both optimal flight operations and corporate policy for each flight.
KI (HOURS) + K2 (FUEL USED) DOC q = $/HOUR = $/POUND FUEL
K1 K2
l DOC IS A FUNCTION OF K1 AND K2 l TRAJECTORY IS A FUNCTION OF THE RATIO OF K1 AND K2 l INTERACTIVE SELECTION OF K1 AND K2 IS DESIRABLE AIRLINE TRIP PLANNING An actual flight planning example will illustrate the potential fuel and cost savings that can be achieved by using an optimal path trajectory program like OPTIM.
In this example, an airline has 11 different pre-specified routes between Chicago and Phoenix to allow the flight dispatcher to select the most favorable route considering and winds other atmospheric conditions.
Typically, a flight dispatcher will choose the highest cruise altitude possible, for the given airplane gross weight and ambient temperature, that is consistent with ATC altitude constraints.
For this flight, a route slightly north of the shortest route (designated ATC route) was chosen to take advantage of lower head winds.
Even though this route is nine miles longer than the shortest route, the fuel required to complete the trip and the resulting trip cost were reduced since the time to complete the trip was reduced.
TYPICAL CHICAGO -
PHOENIX ROUTE STRUCTURE
\ SOUTHERN ROUTE ORD - PHX ROUTE SELECTION This figure shows the magnitude of trip time, fuel required, and cost to complete the Chicago to Phoenix trip using a generic commercial tri-jet transport airplane with time variant costs of $600 per hour and fuel costs of 15 cents per pound. The first two cases show a comparison of the trip flown at a cruise altitude of 35,000 feet (chosen by the flight dispatcher) for the shortest distance route and the preferred wind route. By flying the preferred wind route, the trip time was reduced by 3 minutes and 39 seconds resulting in a corresponding decrease of 404 pounds of fuel used and a reduction of $97.10 to the trip cost.
To illustrate how much further trip costs could be reduced, the OPTIM program was run in a minimum cost mode Ear the preferred wind routing. The indicate that a cruise altitude results of this run, listed in the third case, of 24,000 feet should be used. Even though this lower altitude resulted in more fuel used to complete the trip (742 pounds), the total trip cost was reduced by $147.55 due to a 25 minute 44 second reduction of trip time.
It should be stressed at this point that the trip profile (airspeed and cruise altitude) will change as the time cost and the fuel cost ratio is changed. As time costs are reduced or fuel costs increase, the trip profile will change towards a minimum fuel trajectory.
It was interesting to note that with the OPTIM program run in the minimum cost mode on the ATC preferred route, further trip time, fuel, and cost reductions were obtaiiled. This illustrates the fact that to achieve truly both horizontal and vertical path optimization optimal cost and fuel savings, must be obtained together.
TRI-JET AIRPLANE TIME COST$600/HR -- FUEL COST$,15/LB TIME FUEL COST ROUTE DESCRIPTION H:M:S LB $ 1. SHORTEST GROUND DISTANCE 3:46:04 29,100 6625,75 (1263 nmi RANGE): CRUISE ALTITUDE35,000 FT 2. PREFERRED WINDROUTE 3:42:25 28,696 6528,65 (1272 nmi RANGE): CRUISE ALTITUDE35,000 FT 3. PREFERRED WINDROUTE 3:16:41 29,438 6381,10 (1272 nmi RANGE): CRUISE OPTIMPROFILE(+24,000 FT) 4. SHORTEST GROUND DISTANCE 3:15:01 29,134 6320,37 (1263 nmi RANGE): CRUISE OPTIMPROFILE(-24.000 FT) TRIP COST COMPARISON This figure shows trip cost expressed in $/mile as a function of trip length for a generic commercial twin-jet transport airplane operating on three different vertical profiles. The three profiles presented are a standard a minimum cost profile which satisfies current ATC vertical handbook profile, path constraints, and an unconstrained minimum cost profile. The minimum cost profiles each represent a significant savings relative to the handbook profile which calls for a constant airspeed/Mach climb, cruise at a fixed altitude and Mach number, and a constant Mach/airspeed descent. The cost-optimized profile with ATC constraints complies with the ATC-imposed speed limit of 250 knots under 10,000 feet and maintains a fixed cruise altitude. The second cost- optimized profile does not comply with the 250-knot speed limit and gradually increases the cruise altitude as fuel is burned.
The difference in trip costs between the optimized profiles and the handbook profiles are significant - increasing from 3.1% to 3.8% as trip length increases from 500 to 1500 n. mi. for the ATC-constrained profile. For the unconstrained profile, the trtp costs range from 4.3% to 4.5% less than handbook.
( CLIMB IAS/MACH 300/,70 CRUISE MACH/ALT ,76/35k DESCENT MACH/IAS ,76/280 2,60 2.50 0 - 500 1000 1500 2000 TRIP RANGE, n, ml, TRIP FUEL USAGE COMPARISON This figure shows fuel used to complete the trip expressed as pounds/mile as a function of trip distance for the same generic commercial twin-jet used in the previous figure. The three profiles presented are the handbook profile the optimized profiles, with and without the ATC vertical path and constraints. However, the optimized paths were computed based on minimizing fuel usage rather than cost. The resulting fuel savings between the handbook profile and the minimum fuel profile with ATC constraints ranged between 7.4% for a 500~mile trip to 8.8% for a 1500~mile trip. If the ATC vertical profile and speed constraints were eliminated the total savings would range between 8% and 9.5%.
14,c I- TWIN JET AIRCRAFT - 2 13.0 I CLIMB IAWMACH 300/.70 CRUISE MACH/ALT ,76/35k DESCENT MACH/IAS ,76/280 ATC CONSTRAINTS NO ATC CONSTRAINTS I I I TRIP RANGE, n, ml, FUEL SAVED VIA FIXED-TIME OPTION Another option in the OPTIM program is the minimum fuel, fixed-time mode. This mode will be used when a fixed trip time, for either airtine or ATC purposes, is desired. This chart shows the percentage of fuel saved by absorbing known time delays through reduced speeds during the cruise and descent flight segments instead of maintaining normal cruise speeds and absorbing the delay in a holding pattern prior to descent. Curves for 500 n.
.
ml., 1000 n. mi., and 1500 n. mi. trips are plotted to show the percentage of fuel saved for each trip as a function of the amount of time to absorb. The assumption is made that the delay is known at the beginning of the cruise segment, although the OPTIM program can reoptimize the proFile Later in the the later in the flight that the pilot cruise to absorb the delay. However, knows his delay, the smaller the delay that can be absorbed by using speed control.
Significant fuel savings can be obtained with this capability, but may require modification of some ATC procedures and policies to obtain arrival assignments early in the trip.
mi.
TWIN JET AIRCRAFT TAKE OFF WT - 100,OOOlbs FUEL $,iS/lb TIME $350/hour 0 10 20 30 40 50 TIME DELAY - min,
I
SPEED/ALTITUDE FOR CONVENTIONAL AND MINIMUM COST PROFILES A significant problem that must be addressed in the practical implementation of the optimized flight paths is to provide adequate guidance for the pilot or autopilot to fly the vertical profiles computed by the optimization routines. This figure illustrates this problem by comparing the speed and altitude profiles of a conventional handbook climb, cruise, and descent with one computed for a minimum cost flight.
The piloting techniques employed on a conventional "handbook" profile are manageable by the pilot since thrust is generally set to a predetermined value and the vertical flight path controlled by adjusting the pitch attitude of the airplane in reference to maintaining a constant value in either the altimeter, airspeed indicator, or the Machmeter. As shown in this figure, during a the airplane is accelerated to 250 conventional profile (heavy dashed line), knots indicated airspeed (KIAS) shortly after take-off. A constant 250 KIAS climb is maintained until reaching 10,000 feet. Then the airplane is accelerated to 300 KIAS while remaining at approximately 10,000 feet. A constant 300-knot climb is maintained until the desired .70 Mach number is obtained. At this point a constant .70 Mach is flown until reaching cruise altitude. Then the airplane is accelerated at constant altitude to the cruise Mach number (.76 in this example). The descent is flown at a constant Mach number (.76) with a transition to a constant 280 KIAS between 23,000 and 24,000 feet. At 10,000 feet, a constant altitude is maintained until the airspeed is slowed to 250 KIAS. This speed is maintained until entering the terminal area for landing.
The minimum cost speed and altitude profile (heavy solid line) may be contrasted to the conventional profile. When unconstrained by the ATC-imposed 250-KIAS limit (above 10,000 feet) neither airspeed, Mach number, nor altitude is constant during the climb or descent. Conventional guidance and pilot techniques are not adequate to fly these profiles. These profiles may not be acceptable due to increased pilot workload and the uncomfortable feeling of not being in control of the airplane.
200 2io 300 SSO 460 450 5 TRUE AIRSPEED, KT ADVANCED FLIGHT MANAGEMENT CONCEPTSRESEMCH The flight path profiles of the optimized trajectories may differ significantly from conventional profiles as illustrated in the previous figure. Many questions arise about the interface required for the flight crew to fly the airplane along optimal trajectories, particularly, in an airline environment. There are additional concerns about obtaining the full benefits of optimal trajectories within an ATC environment with other air traffic.
NASA is engaged in an advanced flight management concepts research effort. This research will be conducted in two phases. The first phase will be aimed at defining the interface requirements between the flight crew and the airborne systems necessary for executing practical optimal flight paths.
This will essentially be a single airplane problem with no external influences from ATC or adverse weather. The emphasis in this phase will be on guidance and control requirements and pilot and passenger acceptability from an airline operations point of view.
The second phase of this research will be aimed at defining the interface requirements between the airborne system (including the flight crew and airborne electronic systems) and the ATC system. This will be a systems problem in that additional constraints such as ATC requirements, other air traffic, or adverse weather will be considered. The emphasis in this phase of research will be an airborne system flexibility and air/ground communication requirements.
PILOT/ AIRPLANE SYSTEMS I NTERFACE REQU I REMENTSFOR PRACTICALOPTIMAL FLIGHT PATHS l PILOT/FMS INTERFACE l GUIDANCE & CONTROL REQUIREMENTS l COMPUTATIONAL REQUIREMENTS AIRBORNE SYSTEM/ l AIRBORNE SYSTEM FLEXIBILITY ATC SYSTEM INTERFACE l SATISFY ATC CONSTRAINTS REQUIREMENTS . DATA COMMUNICATIONSREQUIREMENTS SUMMARY Potential fuel savings and subsequent cost reductions have been demonstrated with the simplified computations of the programmable calculator with open-loop guidance.
Additional savings may be obtained from closed-loop guidance and with the more complex trajectory computatI.ons that can be provided with an integrated flight system.
Regardless of the sophistication of the airborne system, however, compatibility must exist with the air traffic system. Airborne derived optimal trajectories must retain the profile qualities that produce the desired optimization, but also fit the path constraints necessary for safe air traffic control.
Flight crew response due to external influences, such as other air traffic or adverse weather, will be a key issue in the acceptance and usefulness of future flight optimization airborne systems. Attention must be paid to the air/ground communication interface and the airborne system flexibility to ensure a high degree of systems efficiency.
l POTENTIAL FUEL SAVINGS AND COST REDUCTIONS HAVE BEEN DEMONSTRATED WITH BOTH SIMPLIFIED AND COMPLEX TRAJECTORY COMPUTATIONS.
. INCREASED OPERATING COSTS HAVE NECESSITATED INCREASED EFFICIENCY AND COMPATIBILITY BETWEEN AIR TRAFFIC CONTROL AND FLIGHT OPERATIONS.
. PILOT AND AIRBORNE/GROUND SYSTEMS INTERFACE REQUIREMENTS RESULTING FROM FLIGHT ON NON- STANDARD, OPTIMIZED TRAJECTORIES MUST BE DEFINED.
FEEDBACK LAWS FOR FUEL MINIMIZATION 'FOR TRANSPORTAIRCRAFT Douglas B. Price and Christopher Gracey NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 TRAJECTORY OPTIMIZATION The Theoretical Mechanics Branch has as one of its long-range goals to work toward solving real-time trajectory optimization problems on board an aircraft. This is a generic problem that has application to all aspects of aviation from general aviation through commercial to military. Our overall interest is in the generic problem, but we must focus on specific problems to achieve concrete results. The problem is to develop control laws that will generate appr0ximatel.y optimal trajec- tories with respect to some criteria such as minimum time, minimum fuel, or some combination of the two. These laws must be simple enough to 3e implemented on a computer that can be flown on board an aircraft, which implies a major simplification from the two-point boundary value problem generated by a standard trajectory optimi- In addition, the control laws must aJlow for changes in end condi- zation problem.
tions during the flight, and changes in weather along a planned flight path. There- fore, a feedback control law that generates commands based on the current state rather than a precomputed open-loop control law is desired. This requirement, along with the need for order reduction, argues for the applicationof singular perturbation techniques.
l Sotve Trajaror-y OprCmization Problems on board rhe
aircrafr Cnreal time
l Allow for chan..<n. condftiotls
-weather
- Air traffk control changes
l Feedback control lavv
l S i?t&.ar perturbation techniques
SINGULAR PERTURBATIONS Singular perturbation techniques can sometimes be used to break a big problem down into more manageable parts. For example, a large-order numerical optimization problem can frequently be divided into a series of smaller subproblems that can be solved one by one in a serial fashion. The solutions to these subproblems are then combined to generate an approximation to the solution of the original Large-order problem. This technique is very valuable and has been used successfully in a number of different areas.
The validity of the technique depends on a separation of time scales for (or a decoupling of) the various states involved in the problem. For some problems, this separation of the states occurs naturally, and may even be made obvious by one or more small parameters of the problem. For flight problems, some of the time-scale separations are fairly easy to find and agree upon, but others are controversial at best. Another problem with the technique is that while stable feed- back laws can be generated for the initial boundary layers which correspond to the ascent portion of a trajectory, the feedback laws are unstable in the descent portion.
4 Subdivide l.arse+cak numeri'cal optimkation
problem intoserks ofsmaller subproblems
l Solve subprobkms serially &n put solutions rose&r
to approxfmte solution to ori$$ixil problem
l Diffikukks wtth the technique:
- ~rne-scaleorden'ng/~~a~tion
-Terminal boundary layers -descent
FUEL OPTIMAL TRANSPORTPROBLEM The particular problem to be discussed here is a fuel optimization problem for a The cost function which is to be minimized transport aircraft in the vertical plane.
is the integral over the flight time of fuel flow rate f. The state vector consists of range x; total energy per unit weight E; altitude h; and flight path angle y.
The remaining parameters The controls for the problem are thrust T; and lift L.
D; and weight W (which is considered are velocity V; the force of gravity g; drag a constant for the problem). The fuel flow rate is modelled as a quadratic in thrust The drag with coefficients that are, in general, functions of energy and altitude.
is modelled asa quadratic in lift with coefficients that are also functions of energy The E'S on the left-hand sides of the state equations are "small" and altitude.
numbers that determine the ordering and separation of the states in the singular perturbation formulation of the problem.
=
b3 ‘f(h,E,T) dt
/
to
?Lv cos x
subject to:
cti = U-D) V/W
&,i = Vsii-2 X
L.,k =g(l-wcos X)/WY
SEPARATION OF STATES One way to approach the question of the separation of the various states in the problem is to assume that they can be separated into mutually exclusive time scales, each consisting of one state. This' is, of course, ad hoc and does not bother with the realities of the aircraft dynamics, but has the virtue that it makes the equa- tions easy to solve. The resulting feedback control law is easy to implement, at The modelling of the aircraft dynamics under least for the initial boundary layers.
this assumption is unsatisfactory because altitude and flight path angle are highly coupled. An alternate approach is to recognize this coupling and separate the states into three groups: range as the outer layer, energy as the first boundary layer, and altitude and flight path angle as the second boundary layer. The equations for alti- tude and flight path angle are linearized about the solution from the first boundary layer subproblem so that a feedback solution can still be obtained.
l l%t-ai&J Singular RrtWbatl’ons - separate
layers for alt-rtude and ftight path an.@
- E&e2
- teads to impkmuwabk feedback law
- tinsat-i’sfacmry modelling of A/C dynamics
- Implementable feedback law
- pd model of dynamics
,,,, .., . . . . -.. . -..-.---- LINEARIZED BOUNDARYLAYERS This figure shows a plot of the solution for altitude from the separate boundary layers discussed on the previous slide for a trajectory with initial altitude at point A. The. cu,rve labeled C is the altitude solution from the outer layer and repre- sents the cruise altitude for a transport trajectory. It is a horizontal line because the outer layer solution is a constant altitude cruise. The curve labeled B represents the first boundary layer solution for altitude.
It converges to the cruise equilibrium nicely but does-not meet the initial condition. The actual alti- tude for the trajectory comes from the second boundary layer solution and is shown as the curve starting at A. It meets the initial condition and approaches the first boundary layer solution as they both converge to the cruise altitude.
ASCENT
I-
I .
B-fi ’
2 J
C - OuW@.yer
I’. 20 I’. 50
6.30 6.60 d. 90
T =103 SEC
FEEDBACK WlWS FOR DESCENT The difficulty with singular perturbation techniques for descent trajectories can be shown in this figure which shows the altitude resulting from a singularly per- turbed solution for a descent. The horizontal line again represents the cruise alti- tude and is an equilibrium for the boundary layer equations. The other two curves are the altitude solutions from the first and second layer subproblems computed in the backward direction from the endpoint. Kowever, for descent (and for any terminal boundary layer), the trajectory is moving away from the stable equilibrium as time increases. For this reason, the feedback law that was stable for ascent is unstable for descent. Therefore, any inaccuracies at the beginning of or along the trajectory will result in very large errors at the endpoint. The only reliable way to use the same control law for descent as ascent is to precompute the descent in the negative direction from the desired endpoint. This precludes taking into account any changes in the end conditions after the descent is initiated.
DESCENT
ANOTHER APPROACHFOR DESCENT In order to accommodate the terminal part of a trajectory while maintaining commonality with the analysis used so far, it was decided to change the problem statement by modifying the cost function to include an altitude term multiplied by an adjustable coefficient. The cost function becomes the integral of a convex combina- tion of the fuel flow rate, as before, and the additional altitude term. When the weighting parameter k is zero, this is the original cost function for the problem.
Values of k between 0 and 1 change the equilibrium altitude for the problem to a value lower than that for the original cruise. Thus, the aircraft can be made to descend by changing the parameter k to a nonzero value. The desirable property of always approaching a stable equilibrium is maintained with this technique, and it becomes just an extension of the technique used for ascent. The cost function is no longer that for a fuel optimal problem when k is not zero, but the fuel flow rate is still a part of the cost function. This cost function represents a trade-off between fuel optimization and simplicity of the overall control law.
l Change problem to&et stable fadback law for
descem trajectory
+ Chanp cost functfon -add aktude term
J= i;kl-k) f(E,h,T) +kh] dt
+ Dl’fferent vatties of lz sfve diffeerent~ V&ES for
equilibrium attiuile
EQUILIBRIUM ALTITUDES This figure demonstrates that the equilibrium altitude may be changed in an almost linear fashion by changing the constant k multiplying altitude in the cost function. The equilibrium altitude for this problem is a function of cruise velo- city. The outer layer subproblem (with k = 0) consists of determining the cruise altitude that corresponds to a given cruise velocity. The choice of cruise velocity This figure shows equilibrium altitudes for must be made from other considerations.
four different cruise velocities plotted against the constant k. It can be seen that as k varies butween 0 and .4, the equilibrium altitude changes from the fuel optimal altitude to the ground. The variation for each value of velocity is nearly linear, but the change with velocity is obviously nonlinear. These curves were generated by solving the outer layer subproblem for four different values of cruise velocity for different values of the parameter k.
33000
‘.
-.
0k I
I -.. -:.. -.I. 1
. -L .2 k .3 -".4 .5
DESCENT TRAJECTORY This figure shows a descent trajectory generated by the technique described previously. It is a plot of altitude vs time.with three different curves plotted.
The curve with the straight line segments represents the altitude from the outer layer subproblem. It starts at the fuel optimal cruise level with k = 0. At 100 seconds into the flight, k is varied linearly from 0 to .15,and the equilibrium alti- tude follows it down almost in a straight line. The other two curves are the alti- tude from the first and second boundary layers. The secondary boundary layer alti- tude represents the actual altitude achieved by the simulated aircraft using the feedback control law under discussion. By changing the way the parameter k is varied, descents with different characteristics can be achieved. One common charac- teristic is that the bottom of the descent is always an exponentially stable approach to the new equilibrium altitude.
Mtdfffd Cost Function
I I 1 I I 00 0.30 0.60 0.90 1.20 1.50
T l 103 SEC
COMMENTS The general area of singular perturbation techniques has been shown to offer a good framework for on-board optimal trajectory control. The large numerical problem of computing optimal trajectories requires some simplification and order reduction in :order to hope for an on-board solution. The haphazard use of order reduction tech- niques without considering the implications of separating states which may actually change on the same time scale can lead to control laws based on an improper model of flight dynamics. In particular, altitude and flight path angle must be considered on the same time scale for the fuel optimal transport problem, since they are highly coupled. By linearizing the altitude and flight path angle equations about the solu- tion to the first boundary layer subproblem, they can be considered on the same time reflects and a feedback control law can be developed that accurately the scale, dynamics of the aircraft.
A major problem with singular perturbation techniques has been the inability to derive stable feedback laws for terminal layers without precommuting the terminal boundary layer trajectory. It has been shown that by adding an altitude term to the cost function with a variable multiplier, feedback laws can be generated that always fly toward a stable equilibrium. These laws, though ad hoc in nature, can be used to approximate optimal trajectories. A nice feature of this technique is that the con- trol law used for ascent is continued throughout the trajectory with the only change for descent being a nonzero multiplier on the altitude term.
framewodc for on -board o~fi-nal trajectory
controt
on same tfme state
Atttiude term a cost fitnction treadsto fi.back
law that ICS good fOt- whole trafatory
Need optimal tra+ctorfes for ccmpat+sons
IDENTIFICATION OF MULTIVARIABLE HIGH-PERFORMANCE TURBOFAN ENGINE DYNAMICS FROM CLOSED-LOOP DATA Walter C. Merrill NASA Lewis Research Center Cleveland, Ohio First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 24-26, 1983 A typical engine control design cycle consists of developing a dynamic engine designing a control based simulation from steady-state component performance data, upon this simulation, and then testing and modifying the control in an engine test cellxto meet performance requirements. This design cycle has been successful for state-of-the-art engines. However, for more advanced multivariable engines that exhibit strong variable interactions, this procedure will result in substantial triai and error modification of the control during the testing phase. One method to automate the design process and reduce control modification testing and devel- opment cost would be to identify accurate dynamic models directly from the closed- These identified models would then be used in conjunction with a loop test data.
synthesis procedure to systematically refine the control. Recent advances in closed-loop identifiability (Ref. 1) present a methodology for this direct identi- fication of engine model dynamics from closed-loop test data. This paper describes the application of an identification method (Ref. 2) to simulated and actual closed-loop FlOO engine data (Ref. 3). This study was undertaken to deter- mine if useful dynamic engine models could be identified directly from closed-loop engine test data (Ref. 4). (See fig. 1.)
Determine Multivariable Engine Models
Directly from Closed-Loop Engine Test Data
Figure l.- Identification objective.
The FlOO engine was tested in the Lewis Research Center altitude test ,facility to evaluate the FlOO Multivariable Control (MVC) law (Ref. 3). During the same test period the "Bill of Material" (BOM) control was also evaluated as a baseline/back- up control model. Thus, there were a variety of closed-loop operating records obtained throughout the flight envelope with a number of different power input requests.
(See fig. 2.)
Note that direct control of the engine controls inputs is not possible.
Since this is a closed-loop process, input and output noise will be correlated. Nor- mally, this precludes the use of open-loop indentification techniques which require independence of the inputs and outputs. However, sufficient independence can be guaranteed if the PI control changes during a transient or if the simplified en- gine model generates a full rank, independent desired input.
This latter condi- tion is the case for the FlOO MVC structure and thus allows a direct application of open-loop identification methodology.
C(mTR#S
F-100 Effille
.
OUTPUTS 3
lEAStRElENTS
+
Schodulod
--- i
Proportion81
J-
l nd Integral - +
+ I
Control
Fi
I
I
I
I
I I I I I I I
TmlJST
PLmHr
1I I
COtWAN COIWITIONS Figure 2.- FlOO multivariable control structure.
The Instrumental Variable/Approximate Maximum Likelihood (IV/AML) method is an output error method of time series analysis. It was implemented in a combined iterative/recursive form. The IV/AML method was selected because the method exhibits reasonable convergence for a small number of samples.
The IV/AML method is based upon an approximate decomposition of the maximum likelihood solution to the identification problem (fig. 3).
Approach
W/AM1 Method of Recursive Time
Series Analysis Directly Applied to
Closed-Loop Data
Figure 3.- Identification approach.
Engine dynamics at a steady-state operating point are adequately modeled by a linear state space system. For the FlOO engine -? three-output four-input model written with the "transfer function" form given in Ref. 2 is shown. Engine speeds (Nl and N2) are important dynamic engine variables. Engine exhaust nozzle pres- sure (PT6) is an indicator of engine thrust. The engine inputs are fuel flow (WF), nozzle area (AJ), compressor inlet variable guide vane position (CIVV) and (See fig. 4.)
rear compressor variable stator vane position (RCW).
Czl + A$x, = Bpmx,
(zi + cl>qk = ek
= xk
+ qk
yk
= CNl ,N2,PT61T
Y
u = (WF 9 AJ t CIVV t RCVVI T
Figure 4.- Engine model equations.
The initial values for the A, B, and C matrices of the model were determined from SISO open-loop identification tests performed on an engine simulation. These values were used to start the closed-loop identification procedure.
The model structure was taken from a third-order behavioral model developed in Ref. 5.
Signal-to-noise ratios were determined from actual closed-loop data. Analysis showed the noise levels to be very low.
(See fig. 5.)
l APB X Initial Values from Simulation
l Structure from Behavioral Model
l Noise Estimates from Data
MVC Data 7<SNR<95
BOM Data 22<SNR<800
Figure 5.- Engine model definition.
The IV/AML, method was originally applied to SISO simulated data to determine The method was then applied to open-loop MIMO simulated initial parameter values.
data. From these MIMO tests an additional element of Al was found to be necessary to satisfactorily model PT6. Also, the noise model was found to be very close The engine model found from this MIMO test was then used to to the plant model.
predict engine behavior based upon actual closed-loop engine data.
The FlOO engine was tested in the Lewis Research Center altitude test facility During the to evaluate the FlOO Multivariable Control (MVC) law (Ref. 3).
same test period the "Bill of Material" (BOM) control was also evaluated as a baseline/backup control 'mode. Thus, there are a variety of closed-loop operating records obtained throughout the flight envelope with a number of different The two multivariable data sets used in this report were power input requests.
recorded at an ALT = 10,000 ft, MN = 0.9 condition as the power request was varied (step change) in a small (hopefully linear) range about intermediate engine power. One set corresponds to an MVC control test, the other to a BOM test. Data were sampled at T = 0.05 set for lo-second transients, which yields 200 points for each record in the data sets. (See fig. 6.)
l Simulation (Open Loop)
SE0
MIMO
l Test Data (Closed Loop)
BOM Control
MVC Control
Tz.05; 200 Points
Application of instrumental variable/approximate maximum Figure 6.- liklihood method.
This is Normalized WP from the BOM and MVC control tests is shown in figure 7.
typical of the engine inputs in these tests. Power spectrum analysis of these inputs shows a slightly higher frequency component in the MVC inputs, although However, for both the BOM and more total power is contained in the BOM inputs.
MVC inputs most of the power is concentrated below 6 radians/set.
Note that these inputs are not persistently exciting.
WF-WC WF - BOM 4 5 TIME, WC Figure 7.- Typical test inputs.
The control inputs of figure 7 were used in conjunction with the MIMO model indenti- fied from the simulation (Model 1) to predict engine output.
Comparing the predicted outputs of model 1 with the actual outputs, it was appar- No output was predicted well for either BOM or ent that model 1 was unacceptable.
Slight discrepancies between MVC data. Figure ,8 is typical of the comparison.
simulation and test data cannot account for large discrepancies between predicted and actual outputs.
PREDICTED
e
ec .Ol -
.oo -
0 SIMULATED BOMOUTPUT
I
I I I I I I I I
I
,
-0 1 i? 3 4
5 6 7 0 9 10
TIME, set
Identification results for model 1.
Figure 8.- To investigate this inability to predict engine response, the IV/APE method was applied directly to the closed-loop data producing models 2 (MVC) and 3 (BOM).
Model 1 was used as a starting point. As illustrated in figure 9, model 3 accu- rately reproduces the data from which it was generated (BOM). Model 2 results are the error of all the outputs for models 1, 2, and 3 is less similar. In fact, However, comparing parameters for models 1, 2, and 3 it can be seen that than 1%.
while Al remains essentially unchanged, elements of Bl do change substantially.
This implies a slightly overparameterized model structure which does account for the inability of model 1 to predict BOM and MVC engine data.
4 S 6
nAM# set
Figure 9.- Identification results for model 3.
A procedure was developed to remove the overparameterization. Three parameters The were eliminated and this new structure was applied to the simulation data.
(See fig. 10.)
resultant IV/AML identified model is given as model 4.
.0025 .oooO -.W5 -.0050 -.OW3 -. oa 2 -.olB 0 slMlnAlEo -. elm 0 moDa A MODELS -. ol73 0 AcnML -.o800 a 23456789u) q= Figure lO.- Identification results for models 4 and 5.
When used to predict BOM and MVC output data, model 4 was still unsatisfactory.
Model 4 did predict Nl(MVC), N2(MVC), and N2(BOM). However, Nl(BOM) and espe- cially PT6 for both data sets were not predicted well. The error in PT6 is some- what expected from sensor and input bandwidth consideration. The Nl(BOM) error was not expected, however. Figure 11 compares predicted Nl data using model 4 to Model 4 predicted Nl grossly follows the trend actual closed-loop Nl(BOM) data.
of the simulated data. Thus, it appears that the dynamic portion of model 4 is correct. However, there must then be large discrepancies in some of the model 4 gain terms. These discrepancies are somewhat perplexing since model 4 predicted Nl(MvC) but not Nl(BOM).
JKO-
.ols -
.OlO -
.als -
Alo -
a
c
-.oos -
TO10 -
-.ols -
I I
I I I I I I
L I -iI
-.=O 1 2
3 4 S 6 7 8
9 10
TlMs=
Figure ll.- Identification results for model 5, PT6.
BOM inputs are larger in magnitude than the MVC inputs Recall, however, that fhc Thus, some nonlinear effects may and that model 4 represents linearized dynamics.
This explanation is not entirely satisfactory since be inherent in the BOM data.
Further work to resolve this problem is NZ(BOM) and N2(MVC) were both predicted.
The IV/AML identification method was again utilized to further refine required.
the model parameters for the structure of model 4 using the two sets of experi- mental closed-loop data. The purpose of this final iteration is to identify a single model that can accurately predict both sets of engine test data and, hope- fully, simulation data as well. (See fig. 12.)
D 1 2 345678910
sa
Tuy
Figure 12.- Identification results for model 5, N2.
Again model 4 was used as an initial condition in the IV/AML method applied to the BOM and MVC data. Models 5 and 6 resulted. Both models 5 and 6 fit their respec- tive data sets quite well. Figures 10 to 12, for example, show a good fit of the Similar comparisons to MVC data were BOM data by outputs predicted using model 5.
More importantly, when the BOM model 5 is used to predict obtained using model 6.
the MVC data, the comparison given in Figures 13 to 15 is quite reasonable. Thus, model 5 (or equivalently model 6) represents a model which predicts a class of inputs and can be used with confidence in a control design procedure.
.ozs
r
0123456789Ml
nw=
Identification results for model 5 predicting multivariable Figure 13.- control data, PT6.
The IV/AML method was applied to both open-loop simulation and closed-loop test data of an FlOO turbofan engine.
The method accurately and consistently identified models from both the simulation and test data.
Due to the structure of the BOM and MVC control laws, the engine model is strongly system identifiable and consequently a direct identification approach was used on the closed-loop data.
A third-order model structure was derived and found to be overparameterized.
Three parameters were eliminated by sensitivity considerations. The simplified structure was found acceptable for fitting both simulation and test data. Test model accuracy is limited to 6 radians/set since spectral analysis of the inputs shows limited signal strength above this frequency.
.Ol6
.Ol4
.Ol2
.OlO
.oQB
.W6
A04
A02
-.OW
-v- vvd
- PREDICTEW
I
I I I I I I I I
I
-A02
S6789Y)
0 1 2 3 4
TIME, sa
Figure 14.- Identification results for model 5 predicting multivariable control data, Nl.
Comparisons showed that models identified from simulated data generally predicted Nl(MVC), N2(MVC), and N2(BOM) test response adequately. However, predictions of pT604VC) and PT6(BOM) were poor and Nl(BOM) showed some discrepancies in dynamics. The PT6 differences are attributed to the low- frequency content of the test input signals (ti radians/set), the bandwidth of and the high-frequency nature of the PT6 mode. However, the the sensor, difference in Nl is attributable to a difference in simulated versus actual engine performance. This conclusion is accurately portrayed in a comparison of identified models.
.012
r
I I I I I 1 I I I
I
-.OOd
01234S6789lO
nMS=
Identification results for model 5 multivariable control Figure 15.- data, N2.
Finally, a simplified model determined from BOM data accurately predicted not only BOM but also MVC test response data. This ability to predict engine performance for a class of inputs generates confidence in controls designed from this model. Thus, it is concluded that useful dynamic engine models can be obtained from closed-loop test data using the IV/AML identification method. This identification technique, then, represents the first step in an automated engine control design process. (See fig. 16.)
l Basic lV/AML Worked Well
l Engine Model is SSI
l Third-Order Model Structure
l Simulation Predicts Test Data
l Models from Simulation do not Predict
NICBOM> and PT6CBOM & MVC)
l BOM Model Predicts MVC Data
Figure 16.- Conclusions.
REFERENCES 1. Soderstrom, T.; Ljung, L.; and Gustavsson, I.: Identifiability Condi- tions for Linear Multivariable Systems Operating Under Feedback.
IEEE Trans. of Autom. Control, Vol. AC-21, No. 6, Dec. 1976, pp. 837-840.
2. Jakeman, A.; and Young, P.: Refined Instrumental Variable Methods of Recursive Time-Series Analysis - Part II, Multivariable Systems. Int.
J. Control, Vol. 29, No. 4, 1979, pp. 624-644.
3. Lehtinen, F.K.B; Costakis, W.G.; Soeder, J.F.; and Seldner, K.: Alti- tude Test of a Multivariable Control System for the FlOO Engine. NASA TMS-83367, July 1983.
Merrill, W.: Identification of Multivariable High Performance Turbofan 4.
Engine Dynamics from Closed-Loop Data. NASA TM 82785, June 1982.
5. DeHoff, R.L.; Hall, W.E., Jr.; Adams, R.J.; and Gupta, N.K.: FlOO Multivariable Control Synthesis Program. Vols. I and II, (AD-A052420 and AD-A052346.1 AFAPL-TR-77-35, June 1977.
SESSION III CONTROL SYSTEM DESIGN TECHNIQUES Jarrell R. Elliott Session Chairman EIGENSPACE DESIGN TECHNIQUES FOR ACTIVE FLUTTER SUPPRESSION William L. Garrard and Bradley S. Liebst Dept. of Aerospace Engineering and Mechanics University of Minnesota Minneapolis, MN First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 OBJECTIVE The objective of this discussion is to examine the application of eigenspace design techniques to an active flutter suppression system for Eigenspace design techniques allow the control the DAST ARW-2 research drone.
system designer to determine feedback gains which place controllable eigenvalues in specified configurations and which shape eigenvectors to achieve desired dynamic response. Eigenspace techniques have been applied to the control of lateral and longitudinal dynamic response of aircraft [1,2]. However, little has been published on the application of eigenspace techniques to aeroelastic control problems.
This discussion will focus primarily on methodology for design of full- state and limited-state (output) feedback controllers.
-We do not intend to address the significant difficulties associated with the realization of full- and limited-state controllers. Most of the states in aeroelastic control and some type of dynamic compensator is problems are not directly measurable, necessary to convert sensor outputs to control inputs. Compensator design can be accomplished by use of a Kalman filter modified if necessary by the Doyle- Stein procedure for full-state loop transfer function recovery [3], by some other type of observer, or by transfer function matching.
EIGENSPACE DESIGN TECHNIQUES Eigenspace techniques allow the designer to place closed-loop eigenvalues (X,) and shape closed-loop eigenvectors (v,). We will briefly review the theory.
For a more detailed discussion see Refs. 1, 2, and 4. First we assume the system is controllable and observable and the matrices B and C are of full rank. (In the case of full-state feedback, C = I.) Later the controllability assumption will be relaxed. The above assumptions yield the results shown below. If we have full-state feedback and n controls, we can arbitrarily place all eigen- values and shape all eigenvectors to any desired form. If we have full-state only pole placement is possible. Since any feedback and a single control, attainable eigenvector is in the subspace spanned by (X I-A)'lB, it is impossible to exactly achieve a desired eigenvector in most aircraft control In practice this does not appear to be a serious problem.
problems.
Uncontrollable eigenvalues cannot be moved but an additional element in each edgenvector associated with these eigenvalues can be shaped.
i=Ax+Bu y = cx Dim (x) = n Dim (u) = m Dim (y) = r max (m,r) closed-loop eigenvalues can be assigned (1) max (m,r) closed-loop eigenvectors can be shaped (2) min (m,r) elements of each eigenvector can be arbitrarily chosen (3) Attainable eigenvector in space spanned by (Ihi-A)-'B CALCULATION OF GAIN MATRIX-I We will first describe the eigenspace design technique for full-state feed- back for a system described in standard state-space form. The design procedure consists of determining a gain matrix K such that for all closed-loop eigen- value and eigenvector pairs (A+BK)vi = h V.
il where X is the desired closed-loop eigenvector and v. is the associated closed- loop eigienvector. This is equivalent to finding wi s?ich that (IXi-A) vi = Bwi Once all the Wi 's have been found, the gain matrix can be calculated.
In order to arbitrarily place all the ?.'s and vi's, the control vector will have to be of the same order as the state veitor and B would have to be inverti- ble. In general this is not the case, and the achievable eigenvalues must lie in the subspace spanned by (&I-APB I In general, the desired eigenvector vid will not reside in this subspace.
jr=Ax+Bu y=cx u = Ky n x= C a v chit i=l i i (A + BK)vi=X v.
i 1 V = LiWi i = (Xil-~)-l~ Li K = W[CV]-' CALCULATION OF GAIN MATRIX-II Since the desired eigenvalues are in general not achievable, the w.'s are selected to minimize the weighted sum of the squares of the difference i: etween the elements of the desired and attainable eigenvectors given by the perform- ance index Ji. The term Pi is a positive definite symmetric matrix whose elements can be chosen to weight the difference between certain elements of the desired and attainable eigenvalues more heavily than others. Setting the derivative of Ji with respect to wi equal to zero gives wi. The notation * denotes complex transpose. Once wi is calculated, the achievable eigenvector vi is obtained. If an eigensolution is not to be altered, setting wi = 0 assures that the associated vi and hi remain in their open-loop configurations.
If we desire output feedback, the procedure is easily modified as shown.
In case of a complex eigenvalue, a real-gain matrix results from a simple transformation [1,2, 41.
d Ji = (v. - vi)*Pi(vid - Vi> 3J i -= 0 aw i W. = (Li*PiLi)-l LifPiVid If an eigensolution is not to be altered, w. = 0 UNCONTROLLABLEEIGENVALUES-I In aeroelastic control problems, the states associated with the gust model are uncontrollable. Moore [4] showed that it is possible to use feedback to assign some components of eigenvectors associated with uncontrollable eigen- values. An algorithm for performing such an assignment is given. For an uncontrollable eigenvalue A g' is singular. We can partition the eigenvector vg associated with this eigenvalue as shown below where vgI1 contains only the uncontrollable states.
x uncontrollable g Partition v such that g [A I-A]v =Bw g g g Becomes II contains only uncontrollable states vg UNCONTROLLABLEEIGENVALUES-II The equation Rv I1 = 0 f3 II to be equal to the open-loop is automatically satisfied if we select v states. Since Xg is not an f which contains the uncontro lable portion of v is nonsingular, and by performing the indicated eigenvalue 0 f A=, [A I-AI] cangbe determined in much the same way as for the controllable calculations, wg eigenvalues.
I = [Xg~-~l]-l~l~g - [XgI-AIIB1 pvgLI % or I = ~~~~ + Av l3 minimizing Id = - VgI)*Pg(vgId - vgI) J (v g g yields = cL~*P~L~)-~L~*P~(v~I~ - Avg) ws DAST ARW-2 FLIGHT TEST VEHICLE The DAST ARW-2 flight test vehicle shown below is a Firebee II Drone which has been modified by replacing the conventional wing with a high aspect ratio supercritical wing designed to flutter within the flight envelope. Two control surfaces, an inboard and outboard aileron, are available on the wing. Current plans are to use the outboard aileron for flutter control and gust load allevia- tion and to use the inboard aileron for maneuver load control. Since two control inputs are needed if eigenvector shaping is to be accomplished, it was decided to use both the inboard and outboard control surfaces for flutter control. As will be shown, the inboard aileron is not effective for flutter suppression. However, it is possible to demonstrate eigenspace techniques using this control surface.
Research currently in progress uses the elevator for control of rigid-body modes and the outboard aileron for control of flutter. Planned research will incorporate a leading-edge and a trailing-edge surface.
DASTARW-2SENSOR AND CONTROL SURFACE LOCATIONS SENSORS CONTROL SURFACES VERTICAL ACCELEROMETERS WBL 84 REAR SPAR OUTBOARD AILERON WBL..,82 FROrJT SPAR, e FSS : WBI. 92,REAfl:.SPAR :.,-j l GLA 1MBL;, !x.:.FtjOtj~;,:gy~:, l MLA o FSS STABILIZER VERTICAL ACCELEROMETER e RSS BS 250 e AFCS . GLA e GLA I AFCS e MLA RUDDER ROLL, PITCH,, YAW RATE ANGLE OF ATTACK (NASA) 8 AFCS e RSS 8 AFCS INBOARD AILERON 8 MLA ‘.’ : .’ : .: CHANGED FROM .. ITERATION 2 .:...
PERFORMANCE SPECIFICATIONS The design flight condition for the flutter control system is M=O.86 (275 m/s or 908 ft/s) and an altitude of 15000 ft (475 m). At this flight condition, the uncontrolled wing flutters, and the flutter control system is required to stabilize the wing without exceeding specified limits on rms control surface activity. The control surfaces saturate if these limits are exceeded.
Gain and phase margins must be adequate. The wing flutters at M=O.75 at this It must altitude, and the flutter control system must be activated at M=O.7.
be verified that activation of the control system does not destabilize the wing Also the flutter controller must not result in excessive at this flight condition.
or torsional loads compared with the uncontrolled wing.
increases in bending, shear, Design Condition M = 0.86 h = 15000 ft.
Maximum RMS Control Surface Activity for 12 ft/s Gust Deflection Deflection Rate --__~. - Inboard 100 1300/s Outboard 15O 7400/s Mininum Stability Margins Gain - 6 dB Phase - 45O No large increases in bending, torsion, and shear at M = 0.7 AEROELASTIC MODEL In the aeroelastic model of the wing given below, yf is the vector of displacements of the various flexural modes, y, is the vector of control surface deflections, y is the gust velocity, MS is the structural mass matrix, Cs is the f3 structural damping matrix, KS is the structural stiffness matrix, q is dynamic pressure, and Qc is the matrix of aerodynamic influence coefficients. Qc is calculated as a function of reduced frequency by a doublet-lattice procedure and is approximated by QA, a matrix of rational polynomials in the Laplace operator S.
The matrices Ai are selected to give the best least-squares fit to Q, over a range of reduced frequencies.
+ [Csls + [KS]) [yfl + q[Q,(sjl
( [MS1 s2
LAnl+21 s [Q,(s)I-[A~I + [A11 z + [A21 s+2v -.-- c Km FLEXURAL MODES Initially, seven structural modes were used in the math model of the wing; however, by comparing eigenvalues calculated using lower order models with eigenvalues resulting from a model which included seven structural modes, it was found that flutter could accurately be modeled by including only three modes.
The first mode will be labeled first mode bending, These modes are-shown below.
'the second mode will be labeled second mode bending although it contains some and the third mode will be labeled first mode torsion. Rigid-body modes torsion, were not included.
First Bending Second Bending First Torsional OPEN-LOOP ROOT LOCUS The locus of the roots associated with the flexure modes is shown below.
The lowest frequency mode is associated'with first mode bending, the next highest with second mode bending and the highest with first mode torsion. It can be seen that first mode bending is the unstable mode while second,mode damping increases with velocity. The frequencies- of these modes approach one another with increas- ing velocity. The first torsion mode is not affected by velocity as much as the but it is necessary to include this mode in order for the system other two modes, The wing flutters at a velocity of 787 ft/s (240 m/s). This is about to flutter.
Mach 0.75.
Velocity (m/s) 0 25 A 150 i 230 Imag, set’ A 275 1st Mode Torsion Design Cond. - 2nd Mode Bending Design Cond.
/ Design Cond.
1st Mode Bending I I I -150 -100 -50
Real, set?
WIND GUST AND ACTUATORMODELS AND STATE SPACE FORMULATION A second-order model forced by white noise was used to simulate the vertical Both inboard and outboard ailerons are driven by high bandwidth actuators.
gust.
In the range of frequencies covered by the three-mode structural model, a fourth- order transfer function was shown to give a very close approximation of the actual inboard actuator/aileron transfer function. A third-order transfer function was used for the outboard aileron. The details of these models are given in Ref. 5.
The state space model of the combined system is given below. The vector XF includes the displacements and velocities associated with flexure modes and the aerodynamic lag states. The vector X,includes the states associated with the in- includes the states associated board and outboard actuator models. The vector Xg with the gust model. The vector U is the control input to the actuators, and w is .the scalar white noise input to the gust model. The total system model is 18th order. Note that the open-loop responses of the actuators are decoupled from one another and are not influenced by the motion of the wing (small inertial cross- coupling terms have been neglected) and that the gust states are uncontrollable.
yg/w = Gg(s) y c hi = Gi(d i yco's = Go(S) .
‘01 *11 Al2 Al3 xF *F + B U + 0 X 0 w xC *22 O C C x 0 0 X 0 G A33 g g I I
I
EIGENSPACE DESIGN APPROACHES The initial eigenspace controller was designed by rotating the unstable eigen- values about the imaginary axis and leaving all other eigenvalues and eigenvectors in their open-loop positions.
The results are shown in the table on the following page.
Although this initial design stabilized the wing at both the flutter test condition (M=O.86) and at the condition at which the flutter controller would be initially activated (M=O.i'), the rms inboard deflection rate is near its maximum (saturation) value at the flutter condition. It was felt that the performance might be enhanced by redesign of the control system to reduce the inboard deflection rate. Also the initial design approach did not use the capability of eigenspace techniques to shape eigenvectors as well as assign eigenvalues, and it was desired to exercise this capability. Since the aircraft exhibits satisfactory response at velocities less than the flutter speed, it was decided to force the closed-loop response of the wing at the design condition to approach the open-loop response of the wing at a velocity 656 ft/s (20% less than the flutter speed). The closed-loop eigenvalues associated with the flexure modes and aerodynamic lag states were moved to their open-loop positions at 656 ft/s. The open-loop actuator eigenvalues were the same for both flight conditions and were not moved, and the gust eigenvalues were uncon- trollable and could not be moved. The desired eigenvectors were selected to be the open-loop eigenvectors of the wing at 656 ft/s. The weighting matrices in the per- formance index were initially set at one. This procedure reduced all surfaceactiv- ity only slightly.
It was decided to use eigenvector shaping to shift control surface activity from the inboard to the outboard actuator. The components of the open-loop aeroelastic eigenvectors in the actuator directions are zero for all flight conditions. Since the desired eigenvectors were chosen to be the openeigen- vectors, penalizing the difference between the achievable and desired aeroelastic eigenvectors in the direction of the inboard actuator would reduce inboard activity.
This was accomplished by increasing the weights on these components to 2.5 x 103 while all other weights remained at one. The results are shown in the table on the following page.
COMPARISONOF RMS CONTROLSURFACE ACTIVITY FOR THE EIGENSPACE CONTROLLERS The performance of the initial and final eigenspace controllers is summarized in the table below. The inboard actuator rate for the initial eigenspace design is almost saturated while, with eigenvector shaping, the control effort is shifted from the inboard to the outboard control surface. Failure of both the inboard and outboard aileron was simulated. When the inboard actuator failed, performance failure of the outboard actuator resulted in was not affected very much; however, This indicates that the inboard actuator is not a good an unstable response.
choice for use in flutter suppression.
----.- r_
Controller Design Inbd Defl Inbd Rate Outbd Defl Outbd Rate
r
Unstable 509.0 0.5 108.0 3.3 Roots Rotated About Imag Axis (Initial Design) ~- Eigenvector Shaping 612.0 to Reduce Inbd Rate 0.9 86.0 4.7 (Final Design)
I - _ --
Initial Design 3.5 519.0 / Inbd Failed
L
-A- Final Design 4.4 572.0 Inbd Failed _-_.- Max Allowable 130 15 10 740
I--
CLOSED-LOOP ROOT LOCUS The root locus of the aeroelastic modes for the final eigenspace design is The wing goes unstable at about 1017 ft/s (M=O.96). The open-loop shown below.
flutter speed is 787 ft/s (M=O.75); therefore, the control system results in an increase in flutter speed of about 29%. As in the open-loop case, the first but in the closed-loop case, the eigenvalues associated bending mode goes unstable, The with this mode move to the real axis where one real root goes unstable.
roots associated with the second bending mode are almost unstable at this velocity.
Imag, se?
Velocity (m/s )
b
0 25
1st Mode
A
250 I50
Torsion
0 230
Design ’
A 275
Con&
(Design
\
Cond.)
2nd
0 305
Mode
n 330
Bending _
I I
-200 -150 -100 -50 50 100
Real, se8
FLUTTER BOUNDARY The results of varying altitude while maintaining Mach number constant are used to define the flutter boundary for the open-loop wing and the wing controlled At M=O.86, the uncontrolled wing is unstable with the final eigenspace design.
until an altitude of 6700 m is reached. At the same Mach number, the controlled At M=O.7. the uncontrolled wing is wing is stable for altitudes above 2900 m.
- 1800 m, whereas the controlled wing is stable for stable for altitudes above altitudes above 2100 m.
(25000 (20000
I
I
1’
/
/
1 Des& Condition (15000 /
Altitude-m
/’
(ft) /
/
304E (10000 kr Boundary (5000 ‘0.6 0.7 0.8 0.9 1.0
Mach Number
STABILITY MARGINS Since the inboard aileron was ineffective in stabilizing the wing, and the stability margins with only the outboard loop closed outboard aileron is critical, are shown below. The initial eigenspace design results from rotating the unstable eigenvalues about the imaginary axis and , with the inboard aileron inactive, this is This is identical to a design resulting from linear quadratic regulator theory.
guaranteed to have excellent stability margins (Ref. 6). The gain margin is 6 dB, and phase margins are greater than 60°. The frequency response characteristics of the final eigenspace design are considerably different from those of the initial design. Gain margins are 6 dB or better, but phase margins are less than 20°. The and the dotted lines represent solid lines represent the initial eigenspace design, the final eigenspace design.
180 ----
r-
-1801
Frequency (radlsec) 4Or ---- Frequency ( radisec) OUTPUT FEEDBACK Since the inboard actuator was ineffective for flutter control, it was elimi- nated from the design. This resulted in a system with a single control input and allowed only eigenvalue assignment. An accelerometer was used to measure the motion of the wing, and a feedback compensator'that approximated the frequency response characteristics of the full-state loop transfer function was designed.
Also the effects of restricting the number of states in the feedback controller were investigated. The design approach was to treat this as a problem in output feedback as described earlier. The C matrix was selected to eliminate various states from the output, and a feedback controller was designed using eigenvalue placement techniques. Since the gust states are uncontrollable if P states are fed back, only P-2 eigenvalues could be placed. The following sets of states were eliminated from being fed back: (a) gust (2 states), (b) aerodynamic lags (3 states), (c) actuator (3 states), (d) first bending (2 states), (e) second bending (2 states), (f) first torsion (2 states), and (g) first torsion and aerodynamic lags (5 states). Except in the case where the first bending mode was eliminated, unstable eigenvalues were rotated about the imaginary axis, and eigenvalues associated with the retained states were maintained in their open-loop positions. The positions of the other eigenvalues were not assigned.
Since the eigenvalues of the first bending mode were unstable, these eigenvalues were rotated, and the eigenvalues of the first torsional mode were not assigned.
This mode then went unstable. The results are summarized below. The rms response was very sensitive to the gust states, and attempts to remove these states or alter their eigenvectors resulted in large rms responses. Loads at the wing root were reduced using the full-state design. Note a slightly different gust model was used so the rms results below are not directly comparable with those presented earlier.
I -1 ~ I ' States Otbd Phase Margin Otbd ! Gain iEliminated Margin Db ' Defl Rate 1 (Degrees) .~ Gust 32 456 6.0 +60 - _- _ Actuators 265 ?36 2.0 4.1 -- Aero. Lags 6.6 394 5.7 -50,60 . ___- Mode 1 Unstable Unstable . Unstable Unstable _.. -. -. ._.
-27,18 Mode 2 2.0 260 2.8 -.-- 2.0 Mode 3 271 5.8 -63,+54 .._~~ ~ " - Mode 3 + -54,+72 6.6 430.0 6.1 Aero. Lags +81,+45 Pull State 2.0 259 6 ?60 Compensator 2.72 250.8 6.0 ?60 .__~_~ RMS Responses M = .86 CONCLUSIONS Eigenspace techniques can provide a powerful tool for the design of feed- back control systems for aircraft.
The design technrques described in this paper are easily implemented and are computationally Inexpensive.
The basic problems facing the designer are those of determining where to place eigenvalues and of selecting appropriate eigenvectors.
This usually requires some insight into the system to be controlled; however, in this respect all control system design techniques are the same.
Stability margins must be carefully examined since there are no guaranteed margins as is the case for linear quadratic regulators.
On the other hand, modal decoupling is easily achieved, and certain roots can be maintained in their open-loop configurations if desired.
Output or limited-state controllers can easily be designed.
i I I Bending Torsion ] Shear lbs x 10') (in - lbs x 10') ' (in - (lbs x 10') 2.365 4.131 6.928 Open Loop 2.321 1.029 3.775 Closed Loop Wing Root Loads M = 0.7 REFERENCES 1. Andry, A. N., Shapiro, E. Y., and Chung, J. C., "On Eigenstructure Assignment for Linear Systems," IEEE Trans Aerospace and Electronic Systems, vol. 19, no. 5, Sept. 1983, pp. 711-729.
2. Cunningham, T. B., "Eigenspace Selection Procedures for Closed Loop Response Shaping with Modal Control," Proceedings IEEE Conference on Decision and Control, vol. 1, Dec. 1980, pp. 178-186.
3. Doyle, J. C., and Stein, G., "Multivariable Feedback Design: Concepts for a Classical/Modern Synthesis," IEEE Trans Auto. Control, vol. AC-26, no. 1, --- Feb. 1981, pp. 4-18.
4. Moore, B. C., "On the Flexibility Offered by Full-State Feedback in Multivariable Systems Beyond Closed Loop Eigenvalue Assignment," IEEE Trans Auto. Control, --~ vol. AC-21, Oct. 1976, pp. 682-691.
5. Garrard, W. L., and Liebst, B. S., "Active Flutter Suppression Using Eigenspace and Linear Quadratic Design Techniques," Proceedings AIAA Guidance and Control Conference, Aug. 1983, pp. 423-431.
6. Safanov, M. G., and Athans, M., "Gain and Phase Margins of Multi-Loop LQG Regulators," IEEE Trans Auto. Control, vol. AC-22, no. 2, April 1977, pp. 173-179.
-~~ TOOLS FOR ACTIVE CONTROL SISTEM DESIGN William M. Adams, Jr., Sherwood H. Tiffany, and Jerry R. Newsom NASA Langley Research Center Hampton, Virginia First Annual NASA AirSraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 The objective of the research reported here is to develop efficient control law analysis and design tools which properly account for the interaction of flexible structures, unsteady aerodynamics and active controls.
The next two figures indicate how such tools can be employed to incorporate active controls into the aircraft design process.
DEVELOPMENT. APPLICATION. VALIDATION AND DOCUMENTATION OF
EFFICIENT MULTIDISCIPLINARY COMPUTER PROGRAMS FOR ANALYSIS
AND DESIGN OF ACTIVE CONTROL LAWS
UTOPIAN DESIGN PROCESS The optimum airplane for a given mission can only be achieved when full advan- tage is taken of the economic and/or performance benefits that are achievable from each discipline. One could argue that these benefits can best be realized when the design variables from each discipline are varied simultaneously in the search for an optimum design. Currently, however, the process is to perform the optimization sepa- A given discipline, e.g. aerody- rately, but not independently, in each discipline.
namics, may relax certain design criteria and assume that other disciplines, e.g.
structures and controls, can make up deficits in stability, safety margins, etc., that result.
MULTIDISCIPLINAKY CR I ERIA ANALYSIS T -STRUCTURES -AERODYNAMICS ] PEKFO;hANCE 6 YEq-iiii-& -CONTROLS INITIAL - CUNFIGURATIOI -PROPULSION NO \I OPTIMIZER KKl VARIABLES I PRESENT CONTROLLAW DESIGN PROCESS The following diagram illustrates the current active control law design process. Fixed models are received from the structures and aerodynamic disciplines.
These models may have been purposely designed with deficits in stability, strength and flutter margins. The controls specialists initially determine an estimate of the control law that is required to remove the deficits. An iterative process is then initiated to refine the control law to remove the deficits while satisfying robust- ness, control power and other criteria. If the design criteria cannot be met, or if they can easily be satisfied, the other disciplines repeat their portion of the design process with appropriately modified design criteria and supply the controls discipline with updated models. This process is repeated until it converges upon an optimum design. The remainder of the paper will describe techniques for obtaining initial estimates of the control laws, performing aeroelastic analyses and optimizing the control laws subject to specified design criteria.
CRITERIA FIXED AlRFRAfiE YES \ tLASl IL / ESTIMATE ANALYSIS OF CONTROL
fN0 V
ANALYSIS TOOLS Efficient tools for the analysis of stability and response characteristics of aeroelastic vehicles are necessary before active control law design can be contem- plated. Such tools must properly consider the interactions between flexible struc- tures, unsteady aerodynamics and active control systems. Several computer programs were developed in the 1970's either by NASA or under NASA sponsorship. DYLOFLEX is an integrated system of stand-alone computer programs whichwas developed primarily to perform dynamic loads analyses of flexible airplanes with active controls (ref. 1); it also has stability analysis capability. DYLOFLEX was developed under contract by the Boeing Company and is available from COSMIC (Computer Software Management and Information Center). Several years ago an aeroelastic capability was incorporated into NASTBAN by the MacNeal-Schwendler Corporation (ref. 2). This addition gave NASTRAN the capability to compute unsteady aerodynamic forces and stability and dyna- mic response characteristics of aeroelastic vehicles with active controls. NASTRAN is available from COSMIC. The aerodynamic forces are expressed in transcendental form in both DYLOFLEX and NASTBAN.
Consequently, the equations of motion are not in a form that can be used in linear system analysis. One final tool, ISAC, developed at Langley (ref. 3) is described in more detail on the next chart.
o DYLOFLEX DYNAMIC LOADS OF FLEXIBLE
STRUCTURES WITH ACTIVE
CONTROLS
o NASTRAN NASA STRUCTURAL ANALYSIS
PROGRAM
o ISAC INTERACTION OF STRUCTURES
AERODYNAMICS AND CONTROLS ISAC Flexible structures are represented in ISAC in terms of a modal characterization that is input from an external source such as NASTRAN. Unsteady aerodynamic forces can be either accepted as input or computed internally using a doublet lattice code An option is included to make a rational s-plane approximation to the (ref. 4).
This allows the equations of motion to be unsteady aerodynamic forces (ref. 5).
written in time-invariant state space form amenable to linear systems analysis and design techniques (refs. 5,6). Stability and dynamic response calculations can be made and displayed graphically that include the effects of sensor dynamics, actuator The ISAC program is operable in dynamics and multi-input/multi-output control laws.
Its use is greatly facilitated by the pre- either a batch or an interactive mode.
sence of a data complex and data complex manager for reading, writing, storing, and cataloging of data. The ISAC program is regularly used in NASA Langley-related It is partially documented and has been distributed to several users research.
outside Langley.
I MODE SHAFE I ,Icm\ I MODAL SELECT.
INTERPOLATIONS DEFLECTIONS J DAST ARW-2 PROJECT SUPPORT The ISAC program is being used to support the DAST (Drones for Aerodynamic and The ARW-2, Structural Testing) ARW-2 (Aeroelastic Research Wing Number 2) project.
scheduled for flight tests in calendar year 1985, is dependent upon several active Ini- control functions for safety of flight in some regions of its flight envelope.
tial support involved comparison of NASA and Boeing Wichita analytic predictions of These comparisons were valuable in that they stability and response characteristics.
pointed out the need for modeling improvements in both ISAC and the Boeing Wichita programs. The NASA/Boeing predictions are now in reasonably good agreement although differences remain in predicted gain and phase margins in the flutter suppression control law. Ultimately, the correlation of measured and analytically predicted performance will be documented. These comparisons will, hopefully, provide informa- tion that will help to improve current mathematical modeling techniques. Reference 7 in these workshop proceedings presents the results of several experimental tests involving active controls for which analytical modeling and prediction of control law performance were done in part using ISAC. This chart also depicts some of the graph- ical outputs that can be obtained by use of ISAC.
0 COMPARBON OF NASA AND BDEINQ PREWCTIOM OF DAST ANW-2 STAEJUTY AND RES#)IYSE CHhRACEmSlWS DESIGN TOOLS A number of control law design tools have been developed at Langley: ORACLS (ref. 8) is a system of algorithms for designing linear feedback control laws for linear time-invariant multivariable differential or difference equation state vector models. ORACLS applies some of the most efficient numerical linear algebra procedures to implement Linear Quadratic Gaussian (LQG) methodology. The ORACLS system can be obtained from COSMIC.
MICAD (refs. 9-11) uses goal-oriented strategies to obtain pareto-optimal solutions that satisfy multiple objectives for either deterministic plants or plants with ran- dom parameters. MICAD is, to some degree, a special purpose tool that was developed to design control laws for the lateral degrees of freedom of rigid aircraft. It is, nevertheless, generalizable to a wider class of problems. Documentation of MICAD is planned but a completion date has not been identified. A. A. Schy is directing the Since the 1960's he has advocated the explicit inclusion of development of MICAD.
design criteria in the design process.
PADLOCS (refs. 12-14) and SYNPAC (refs. 15,16) are two collections of algorithms which provide the capability to design implementable low order active control laws for high order aeroelastic aircraft. They allow direct inclusion of design criteria and, consequently, are similar in structure to MICAD. The three programs do, hop ever, differ substantially internally in performance function and constraint formula- tions and in options for obtaining constrained optimization solutions. Documentation of SYNPAC is now in progress.
o ORACLS UPTIRAL REGULATOR ALGORITHMS FOR THE CONTROL OF LINEAR SYSTEMS o HICAD MULTIOBJECTIVE INSENSITIVE COMPUTER AIDED DESIGN o PADLOCS PROGRAMS FUR ANALYSIS AND DESIGN OF LINEAR OPTIMAL CONTROL SYSTEMS o SYNPAC SYNTHESIS PACKAGE FOR ACTIVE CONTROLS ISAC/ORACLS/SYNPAC INTERFACE An approach to control law design commonly employed at Langley in active control applications is illustrated using the ISAC, ORACLS and SYNPAC programs. A model of the plant is defined using ISAC and stored on the data complex. TRANSFR, a module of SYNPAC, is used to examine pole-zero locations associated with candidate feedback paths and to prepare input to an interactive version of ORACLS. These input data include the model of the plant and estimates of the intensities of noise sources in the plant and in the sensor outputs. Full order controllers are designed using ORACLS. The Doyle-Stein (ref. 17) procedure of adding fictitious process noise at the input is employed where necessary to improve the robustness characteristics of the full order controller. Modal residualization/truncation is employed to select an implementable reduced order controller. or other similar design Within SYNPAC, algorithms, the reduced order controller is optimized as shown in the next chart.
SELECT LQG HHW&D-
DESIGN
CONTROLLER SYNTHESIS PACKAGE FOR ACTIVE CONTROLS This chart illustrates the phase of the design cycle initiated after a candidate control law form has been selected. The selection process has been illustrated on the previous page for the modified LQG approach. Other techniques could have been employed to select the control law form such as Nissim's energy method (ref. 18), eigenspace methods (refs. 19 and 20), classical methods, etc. Constrained optimiza- tion techniques are employed to determine values for the free parameters in the fixed form control law which optimize a measure of goodness and allow the design criteria to be satisfied.
DESIGN DESIGN FEEDBACKGAINS VARIABLES CONSTANTS FILTER COEFFICIENTS DESIGN POINTS \ < SENSORLOCATIONS CONTROLLAW FORM y ETC. ETC.
Y PERFORMANCE - FUNCTION -ANALYSl MODULE t --7 I LOAD CONTROL [ ” POWER -- t l-c.
CONTROLPOWER LOADS DAMPING RATIOS ETC.
SENSOR SIGNAL IPUTS TO SYMMETRIC FLUTTER SUPPRESSION CONTROLLAW This and the next three charts show the application of SYNPAC to improve the robustness characteristics of a control law for suppression of symmetric flutter of A more complete description is presented in reference 16.
the DAST ARW-2 aircraft.
This chart depicts the. sensors (vertical accelerometers) and control surfaces that were employed and defines how the sensor signals are separated into symmetric compo- nents. Note that the control law is single-input/single-output.
ACCELEROMETERS 2 PER SEMISPAN AILERON RIGHT WING COMMAND TO > SYMMETRIC CONTROL LAW 'AILERON ACTUATOR ROBUSTNESSMAXIMIZATION In reference 16 a full order controller (25th order) was designed using ORACl,S for a design point at a Mach number of 0.86 and an altitude of 15,000 ft.
Order reduction techniques were used to obtain an implementable 9th order approximation to this controller of the form indicated below. The controller exhibited poor robust- ness characteristics at an off design point at a Mach number of 0.91 and an altitude of 15,000 ft. SYNPAC was employed to improve the robustness characteristics. Design variables Di (i=1,2 ,...,9) were found which maximized the minimum singular value of the return difference transfer function subject to the indicated constraints.
FIND VALUES FOR THE DESIGN VARIABLES. DI. WHERE
( s+D2) h+Dj) ( s +D6S+D7)
---------me- ---_---____
hl /Y) = D T(s)
FS 12 2
( s +D4s+D51 (s +D8S+D9)
AND T(S) IS A FIXED FILTER
SUCH THAT
0 MINIMUM SINGULAR VALUE IS MAXIMIZED
0 CONTROL POWER CONSTRAINTS ARE SATISFIED
0 +6DB < GAIN MARGIN < -6DB
0 40' s PHASE MARGIN s -40'
CLOSED MOP BLOCK DIAGRAM The performance of the controller optimized for robustness will be exhibited by showing Nyquist plots for the initial and optimized controllers.
This chart identi- fies the loop breaking point and symbols employed to represent the pertinent transfer functions.
>o ' OUTPUT
UR ?- " 1:;
INPUT
I I NYQUIST PLOTS OF HG TRANSFEH FUNCTION (M = 0.91, H = 15,000 FEET) The plant is unstable (a complex conjugate pair of unstable poles) at the indi- cated flight condition. Consequently, for stability the Nyquist plot must encircle the (-1) point once' in a counterclockwise direction as frequency varies from 0 to +-co. Increasing frequency is indicated by arrows on the figure. The initial control- ler stabilized the system but exhibited poor gain and phase margins with an accom- panying small minimum singular value (the point at which the minimum singular values After optimization the minimum singu- occurs is indicated by the heavy solid line).
constraints (not shown) were lar value was increased by 26 percent, control power satisfied and gain and phase margin constraints ware met to within a 2.5-percent tolerance.
Imoghmy lmoginary
OPTIMIZED CONTROLLER
INITIAL CONTROLLER
CONCLUDINGREMARK3 Several tools applicable to analysis and design of control laws for aeroelastic vehicles have been identified.
DYLOFLEX and NASTRAN are available from COSMIC.
ISAC, developed primarily for in-house research, is only partially documented and would require substantial modification for use on a computer complex differing from the one at Langley. been distributed to several off-site It has, nevertheless, users. ORACLS is the only design tool that is sufficiently well documented for dis- tribution.
Linear potential flow aerodynamic theory is employed in computing aerody- namic forces. Consequently, the modeling accuracy becomes doubtful at analysis and design points approaching the transonic region. Improvement is needed in the aerody- namic modeling.
l ANALYSlS AND DESIGN TOOLS IDENTIFIED l CKITERIA EXPLICIlLY INCLUUED IN DESIGN PROCEUUHE l CONCERN ABOUT UNMODELED NONLINEAR AERODYNAMIC EFFECTS
l DOCUMENTATION OF ANALYSIS AND DESIGN TOOLS IS UNDERWAY
REFERENCES 1. Perry, B. III; Kroll, R. I.; Miller, R. D.; and Goetz, R. C.: DYLOFLEX: A Computer Program for Flexible Aircraft Flight Dynamic Loads Analysis with Active Controls. Journal of Aircraft, Vol. 17, No. 4, April 1980, pp. 275-282.
A Design Study 2. Harder, Robert L.; MacNeal, Richard H.; and Rodden, William P.: for the Incorporation of Aeroelastic Capability into NASTRAN. NASA CR 111918, May 1971.
A Digital Program for Calculat- 3. Peele, Ellwood L.; and Adams, William M., Jr.: ing the Interaction Between Flexible Structures, Unsteady Aerodynamics, and Active Controls. NASA TM 80040, January 1979.
4. Albano, E.; and Rodden, W. P.: A Doublet Lattice Method for Calculating Lift AIAA Journal, Vol. 7, Distributions on OscilLating Surfaces in Subsonic Flows.
No. 2, February 1969, pp. 279-285.
Airplane Math Modeling Methods for Active Control Design.
5. Roger, Kenneth L.: AGARD CR 228, August 1977.
Active Flutter 6. Mahesh, J. K.; Stone, C. R.; Garrard, W. L.; and Hausman, P. D.: Control for Flexible Vehicles. NASA CR 159160, November 1979.
7. Newsom, Jerry R.; and Abel, Irving: Experiences With the Design and Implementa- tion of Flutter Suppressiun Systems. NASA Aircraft Controls Research - 1983, NASA CP-2296, 1984, pp. 489-509.
8. Armstrong, Ernest S.: ORACLS - A Design System for Linear Multivariable Marcel Dekker, Inc., c.1980.
Control.
9. Schy, Albert A.: Nonlinear Programming in Design of Control Systems with Speci- fied Handling Qualities.
Proceedings of 1972 IEEE Conference on Decision and Control and 11th Symposium on Adaptive Processes, December 1972, pp. 272-279.
10. Schy, A. A.; Adams, William M., Jr.; and Johnson, K. G.: Computer-Aided Design of Control Systems to Meet Many Requirements. NATO AGARD ConEerence Proceedings No. 137 on Advances in Control Systems, May 1974, pp 6-l to 6-7.
11. Schy, A. A.; and Giesy, D. P.: Tradeoff Studies in Multiobjective Insensitive Design of Airplane Control Systems. AIAA Paper No. 83-2273, August 1983.
12. Newsom, J. R.: A Method for Obtainiug Practical Flutter Control Laws Using Results of Optimal Control Theory. NASA TP 1471, August 1979.
13. Mukhopadhyay, V.; Newsom, J. R.; and Abel, I.: A Method for Obtaining Reduced Order Control Laws for High Order Systems Using Optimization Techniques. NASA TP 1876, 1981.
14. Newsom, Jerry R.; and Mukhopadhyay, V.: The Use of Singular Value Gradients and Optimization Techniques to Design Robust ControLlers for Multiloop Systems.
AIAA Paper No. 83-2191, August 1983.
and Tiffany, Sherwood H.: Control Law Design to Meet 15. Adams, William M., Jr.: Constraints Using SYNPAC--Synthesis Package for Active Controls. NASA TM 83264, January 1982.
16. Adams, William M., Jr.; and Tiffany, Sherwood, H.: Design of a Candidate Flutter Suppression Control Law for DAST ARW-2. AIAA Paper No. 83-2221, August 1983.
17. Doyle, J. C.; and Stein, G.: Robustness with Observers. IEEE Transactions on Automatic Control, Vol. AC-24, No. 4, August 1979, pp. 607-611.
18. Nissim, E.; and Abel, I.: Development and Application of an Optimization Proce- dure for Flutter Suppression Using the Aerodynamic Energy Concept.
NASA TP 1137, February 1978.
19. Garrard, William L.; and Liebst, Bradley S.: Active Flutter Suppression Using Eigenspace and Linear Quadratic Design Techniques.
AIAA Paper No. 2222, August 1983.
20. Ostroff, A. J.; and Pines, S.: Application of Modal Control to Wing-Flutter Suppression. NASA TP 1983, May 1982.
ALGORITHMS FOR OUTPUT FEEDBACK, MULTIPLE-MODEL, AND DECENTRALIZED CONTROL PROBLEMS Nesim Halyo and John R. Broussard Information and Control Systems, Incorporated Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT The optimal stochastic output feedback, multiple-model, and decentralized control problems with dynamic compensation are formulated and discussed. Algo- rithms for each problem are presented, and their relationship to a basic output feedback algorithm is discussed. An aircraft control design problem is posed as a combined decentralized, multiple-model, output feedback problem. A con- trol design is obtained using the combined algorithm. An analysis of the design is presented.
ADVANTAGESOF STOCHASTIC OUTPUT FEEDBACK The stochastic optimal output feedback problem 11-81 is a significant extension of the "full-state feedback" LQG problem [9]. Its formulation ad- dresses some important limitations encountered in practical systems and provides a flexibility useful in configuring the control law for ease of implementation.
Some of the advantages of the stochastic output feedback problem are shown below.
Output feedback introduces a rich class of control law structures which can be used in modern control designs.
l DESIGNER CAN SELECT THE STATES FOR FEEDBACK PROVIDES A METHODTO DESIGN OUTER LOOP CONTROL LAWS a ACCOUNTS FOR ACTUATOR DYNAMICS WITHOUT NECESSITY FOR ACTUATOR STATE FEEDBACK ACCOUNTSFOR PHASE SHIFTS INTRODUCED BY PREFILTERS AND OTHER ESTIMATORS WITHOUT NECESSITY OF FEEDBACK a PROVIDES A SYSTEMATIC METHODTO INCREASE OR DECREASE GAINS BY ADJUSTING PLANT AND MEASUREMENT NOISE COVARIANCES a PROVIDES CONSIDERABLE FLEXIBILITY IN THE CONTROLSTRUCTURE IN A MODERNCONTROLSETTING FORMULATION OF THE STOCHASTIC OUTPUT FEEDBACK PROBLEM The discrete stochastic optimal output feedback problem is formulated below. The control Uk feeds back the output Yk through a constant gain matrix K.
The term ZI is the set of gains K for which JN(K) converges to a finite value J(K).
The term 5' is the set of gains which stabilizes the closed-loop system. The opti- mization problem can be posed as: Find a stabilizing gain K* (K* e S) which minimizes the cost J(K), i.e., J(K*) I J(K), K c D.
= $ xk + r uk + wk 'k+l Yk = c Xk + Vk Uk = - K Yk
= WAki VT) = Mki x1s, = so
E(Wk w:, E(Vk E(Xo
E(Wk) = 0 E(Vk) = 0 V;) = E(Wk XL) = E(Vk XL) = 0 E(WI< + U; R Uk) JN(K) = &i-) ,io E(X;+, Q 'k+, J(K) = lim JN(K) < 03 KEV EXAMPLE Some important characteristics of the stochastic optimization problem posed are illustrated in a simple first-order example. In this example, the domain of optimization V is the semi-open interval (0, 2), while the set of stabilizing gains S. consists of the open interval (0, 2). The system is completely con- trollable and output stabilizable. However, a.s illustrated by the example, out- put stabilizability alone does not guarantee the existence of a solution to the optimization problem. The cost function J(K), for this example, has no minimum in V or in S. Furthermore, the example illustrates that the continuity of the cost function J(K). over its domain D is not guaranteed, as K.= 0 is a point of discontinuity. Therefore, it is desirable to determine conditions under which an optimal solution exists.
= Xk + Uk Yk 5 Xk + Vk 'k+l Q=l R=O V=l w=o so = 1
rTQI'+R=l>O c w CT+ v =l>O
O<K <2 K=O J(K) 3- 2- OUTPUT FEEDBACK EXISTENCE CONDITIONS As illustrated by the example, the domain of optimization is not necessar-: ily a closed set, and can be unbounded, although S is always open. Thus, it is necessary to determine conditions under which the minimum cost is attained Such conditions which guarantee the existence of a at an interior point of S.
solution to the optimization problem are shown below. Under these conditions, the domain of optimization V coincides with the stability set S, which in- On the other sures that the optimal gain stabilizes the closed-loop system.
hand, it can be shown that the cost function J(K) is always continuous on S Note that the example considered previously fails to satisfy the con- I103 - dition W 2 E r rT, but satisfies all the remaining conditions. While the conditions 1, 2, and 3 ensure the existence of a stable global minimum, they are not necessary for the existence of a solution to the optimization problem.
However, the class of optimization problems covered is quite broad, and because the existence conditions are expressed in terms of known system parameter matrices, verification is a simple task. Note that the measurement noise and control penalty terms are not necessary for existence, which is a major dif- ference between the discrete and continuous output feedback problems. Also note that Q and W need not be positive definite, but must satisfy 1. Condi- tion 1 is intriguing, as it corresponds to a method of improving robustness in control and filter designs [ll]. The uniqueness of the solution is not ensured, except for special cases such as full-state feedback.
SUFFICIENT CONDITIONS FOR EXISTENCE: 1. FOR SOME E > 0 Q ;? E CT C w 2 E r rT 2. rTQr+R>O CWCT+bO 3.
(C, 4, I-') IS OUTPUT STABILIZABLE LET 1 AND 2 HOLD. J(K) HAS A STABLE MINIMUM IF, AND ONLY IF, (C, 4, r) IS OUTPUT STABILIZABLE For gains which stabilize the closed-loop system, the cost function J(K) can be expressed more explicitly in terms of K, as shown below. An expression which provides more insight can be obtained by considering the incremental cost AJ(K, AK). As the incremental cost is the total change in the cost due to a change AK in the gain, the optimization problem can also be treated as that of finding a AK* which minimizes the incremental cost for a fixed K c S. Due to the almost quadratic form of the incremental cost, a "natural" direction is the one which would minimize the incremental cost if it were actually quadratic in AK. The following theorem exploits this direction, d(K).
Theorem: Let the existence conditions 1, 2, and 3 hold, and K. be in S. Then there exist f3 > 0, a sequence {Ki, i 2 O}, and a limit point, say K*, of the sequence such that and $(Ki) -f $(K*) = 0 J(Ki) ~ J(K*) K = Ki + o! d(Ki) i+l d(K) = p(K) -' rT P(K) @ S(K) CT S(K)-' - K whenever 0 < cx I B.
J(K) = i trip(K) WI + i triKT p(K) K VI KcS P(K) = $(K)T P(K) 4(K) + CT KT R K C + Q S(K) = 4(K) S(K) @(K)T + r K V KT rT + W i;(K) = rT P(K) r + R s(K) = C S(K) CT + V AJ(K,AK) = J(K+AK) - J(K) =--i
tr{2 AKT[?(K+AK) K s(K) - rT P(K+AK) $I S(K) cTI
+ bKT B(K+AK) AK S(K)1 K, K+AK E S NECESSARY CONDITIONS ii K* S^(K*) = rT P(K*) I$ S(K*) CT K* c S - OUTPUT FEEDBACK ALGORITHM Convergence Theorem: Let {Ki, i 2 01 be a sequence of gains obtained from the algorithm, starting with K. E S. Then, any limit point, say K*, satisfies the necessary conditions for optimality, stabilizes the closed-loop system, and J(Ki) ~ J(K*).
a0 = 1 1. CHOOSE K, 6 S z>l i =0 2.
SOLVE THE LYAPUNOV EQUATIONS P(Ki) = IT P(Ki) 9(Ki) + CT K: R Ki C + Q S(Ki) = I S(Ki) $(Ki)T + I Ki V Ki rT + W IF P(Ki) OR S(Ki) IS NOT NON-NEGATIVE DEFINITE GO TO 5 3. COMPUTEd(Ki), Ki+, d(Ki) = ij(Ki) -' rT P(Ki) 4 S(Ki) CT S^(Ki)-' - Ki K = Ki + ai d(Ki) i+l 4. COMPUTETHE COST J(Ki) J(Ki) = i triP(Ki) WI + k tr{Ki P(Ki) Ki Vf SET i = 1 IF i = 0, AND GO TO 2.
tr d(Ki_l)T~(Ki l)d(Ki ,)?(Ki ,) ] GO TO 6 IF J(Ki)-J(Ki-l)<-$ ai-1(2-ai-l) t 5. REDUCEa a.
= ai/z Ki = Ki 1 d(Ki) = d(Ki_1) K = Ki + ai d(Ki) cli+l = ai i = i+l GO TO 2 i+l 6. COMPUTEGRADIENT I = - I d(Ki) I IF 1)$(‘i) // ’ El OR IJ(Ki) - J(Ki-l)l > ~29 ai+l = ai, i = i+l, GO TO 2 7. STOP OPTIMAL DYNAMIC COMPENSATION Most control systems for complex plants use some form of dynamic compen- sation. The dynamic compensator may simply consist of an integral feedback or a rate command structure, or may be a Kalman filter or an observer. Classical control designs make considerable use of dynamic compensation in the form of various filters, washout loops, etc. The basic form of a digital control sys- tem making use of dynamic compensation is shown below. The design of dynamic compensation in an optimal control setting can be imbedded into the optimal output feedback formulation by augmenting the state with the compensator states In this form, the order of the dynamic compensator is a design [12,131 - parameter and can be selected so as to obtain a low order, easily implemented compensator. For systems which are not stabilizable with the available measure- ments, such as some cases of flutter suppression, dynamic compensation is a necessary rather than a simply desirable structure [141. The design of the dynamic compensator can be obtained using the output feedback algorithm pre- sented earlier.
PLANT 'k _ 'k , c (4, r) 'k COMPENSATOR = @ xk + r uk + wk Yk = c Xk + Vk 'k+l ADVANTAGESOF THE MULTIPLE-MODEL OUTPUT FEEDBACK APPROACH While output feedback introduces significant flexibility in the structure of a control law, it does not directly address some of the objectives and re- quirements encountered in designing control laws. The multiple-model output feedback approach provides a design method which can be used to obtain impor- tant design requirements while preserving all the advantages inherent in output feedback. Some of the advantages of the approach follow.
l PRESERVESALL THE ADVANTAGESOF OUTPUT FEEDBACK l PROVIDES A DESIGN METHOD FOR ROBUST CONTROL LAWS a PROVIDES A DESIGN METHOD FOR MULTIPLE CRITERIA l PROVIDES A DESIGN METHOD FOR ACTUATOR FAILURE ACCOMMODATION a PROVIDESA DESIGN METH.ODFOR SENSOR FAILURE ACCOMMODATION WI .- PLANT1 U1 ' (42 r-1) - - , I I W .
.
.
d za PLANT P P Xp Yp % ) cP - (q& r,)
1 /lr
/
1’1
+qY, ,
I P-4
.
L - DYNAMIC , J ( z COMPENSATOR - C- -K2 w#Jz, rz' FORMULATION OF THE MULTIPLE-MODEL OUTPUT FEEDBACK PROBLEM The multiple-model output feedback formulation considers the problem of igning a fixed control law to meet design object ives expressed in terms of des - various plant models, measurement models, and performance criteria as shown below. The control law structure can contain dynamic compensation and output feedback. For example, the design objective of insensitivity to variations in some plant parameters can be addressed by selecting plant models ($s, r.)
which include these variations. Some types of actuator, sensor, or ot er +l iJlant subsystem failures can be addressed in the design by appropriate selection of
the parameters rj, Cj, $j, V., and Wj. Various other design objectives can be
addressed in a similar manne 4.
Let Sj be the set of gains whick stabilizes the jth plant model, while Vj is the set of gains for which the jt cost remains finite. The intersection S of the S*'s determines the control gains which stabilize all the plant models, whi e the intersection V of the Vj’s is the set on which the total i cost J(K) is finite. The optimization problem can be posed as: Find a gain K* which stabilizes all the models (i.e., K* E S) and minimizes the cost J(K), i.e., J(K*) i J(K), K F V.
= +j xjk + rj ujk + wjk lsjsp
'jk+l Y. =c.x +v lsjip Jk J jk jk U -KY =-KC.X - K V.
jk = 0 J jk Jk Jj(K) = lim K E Vj N- & p E(xik+l Qj 'jk+l ' 'ik Rj 'jk) < m k-0 J(K) = f Y. J.(K) yj >o KcV j=l J J v= v. s= s.
R R
j=l J j=l J MULTIPLE-MODEL OUTPUT FEEDBACK EXISTENCE CONDITIONS As for the case of the (single-model) output feedback problem, the exis- tence of a solution to the optimal control problem posed is not always ensured.
However, as seen by the sufficient conditions given below, a solution does exist for a large class of optimization problems. It should be noted that the constraint that the optimal gain stabilizes the closed-loop models excludes problems where no stabilizing gain exists, so that this class of problems must be treated separately. The sufficient conditions for the problem posed can be It can be shown that obtained by extending the results for output feedback.
the cost function J(K) is always continuous on S, but not necessarily on V.
However, for the class of problems satisfying the sufficient conditions, V and S are equal. The conditions given here are not necessary for the existence of a stable global minimum, and the uniqueness of a solution is not ensured.
Nevertheless, the class of problems included is broad enough to cover most parameter sensitivity objectives, control, or sensor failure accommodation objectives, as well as other significant objectives.
SUFFICIENT CONDITIONS FOR EXISTENCE
1. FOR SOME c > 0, AND ALL Qj 2 E CT C. wj 2 E r. 2
jw J J J J 2. I'; Qj rj + Rj > 0 cj wj c; + vj > 0 lsjgp 3.
S IS NON-NULL LET 1 AND 2 HOLD.
J(K) HAS A STABLE MINIMUM IF, AND ONLY IF, P s = n sj CONTAINS AN ELEMENT j=l MULTIPLE-MODEL INCREMENTAL COST AND NECESSARYCONDITIONS For gains which stabilize all the plant models, the cost function J(K) can be expressed in terms of the gain K, as shown.
Similarly, the incremental cost AJ(K,AK) or the change in the cost due to a change in the gain, is seen to resemble a quadratic form in AK.
The necessary conditions can be easily obtained from the incremental cost by letting dK approach zero. The direction d(K), which would minimize the incremental cost if it were a true quadratic form, is selected to obtain an algorithm. It is seen that the direction d(K) is the solution of a'linear equation which requires a larger number of compu- tations than the output feedback case. Also note that setting p equal to one results in the output feedback equations.
KcS J(K) = i jf, Yj[tr{Pj(K) Wjl + trjKT pj(K) K vjl] Pj(K) = oj(K)T Pj(K) 9j(K) + C3 KT Rj K Cj + Qj Sj(K) = ~j(K) Sj(K) Q,(K)T + rj K Vj KT ri + W.
J Bj(K) = r; Pj(K) rj + Rj = Cj Sj(K) C; + V.
'j (K) J AJ(K,AK) = k trj2 AKT jf, vj[i;j(K+AK) K sj(K) - ri Pj(K+AK) @j Sj(K) c:] + E Yj AKT Gj(K+AK) AK Sj(K)1 K K+AK E S j=l
g(K) = jfl Yj‘j(K) K‘j(K) - I': Pj(K) ~j Sj(K) CJ
yj pj(K) d(K) jj(K) = - g(K) E j=l NECESSARY CONDITIONS P y. P.(K*) K* jj(K*) =
1 y. r-7 P.(K*) @j Sj(K*) CJ K* e S
jil J ^J j=l J J J MULTIPLE-MODEL OUTPUT FEEDBACK ALGORITHM z>l, i=O 1. CHOOSE K,eS, ao=l, . . . , 2.
SOLVE THE LYAPUNOV EQUATIONS FOR j = 1, P Pj(Ki) = ~j('i)T Pj(Ki) ~j(Ki) + C3 K: Rj Ki Cj + Qj
Sj(Ki) = $j(Ki) Sj(Ki) ~j(‘i)T + rj Ki Vj K; ri + Wj
IF Pj(Ki) OR Sj(Ki) IS NOT NON-NEGATIVE DEFINITE GO TO 5 3. SOLVE FOR d(Ki), Ki+, Y- h.) d(Ki) Sj(Ki) = - $Ki) E j=l J J 1 K = Ki + ai d(Ki) i+l 4.
COMPUTETHE COST J(Ki) J(Ki) = i E Yi [tr{Pj(Ki) Wj + tr(KJ Pj(Ki) Ki Vj/] j=l IF i = 0, SET i = 1 AND GO TO 2
IF J(Ki)-J(Ki-1 )<-i ai-1(2-ai-1) ,f Yj trfd(Ki-l)T?j(Ki-l)d(Ki-l)?j(Ki-l)~
j=l GO TO 6 5. REDUCE a Ki = Ki-1 a. = CXi/Z d(Ki) = d(Ki-1) i = i+l K = Ki + ai d(Ki) ai+, = ai GO TO 2 i+l 6. CHECK CONVERGENCE
IF II$(Ki) // ’ ~1 OR IJ(Ki) - J(Ki_l)i > E?, ai+l = ai, i = i+l, GO TO 2
7. STOP DECENTRALIZED CONTROLFORMULATION It is well known that the optimal decentralized control problem for linear plants with Gaussian statistics and quadratic cost criteria does not necessar- ily result in a linear system, and.when constrained to linear systems may result in infinite order systems [15-171. Given these negative results, it is natural to constrain the class of decentralized controllers to linear systems of fixed finite order, i.e., the class of decentralized controllers with .fixed- order dynamic compensators as local controllers [.lS$. However, since the dy- namic compensator problem can be imbedded into the output feedback problem, it suffices to consider the class of decentralized output feedback controllers.
Furthermore, the decentralized output feedback problem can be posed as a con- strained output feedback problem, where the gain K is restricted to block diag- onal form [18]. Let S be the collection of block diagram gains (of appropriate dimensions) which stabilize the decentralized system, while V is the set of block diagonal gains for which the cost J(K) is finite. Then the problem can Find a stabilizing gain K* c S which minimizes the cost function be posed as: over V, i.e., J(K*) 5 J(K), K E V. It can be shown that if a block diagonal stabilizing gain exists with rT Qr + R and CWCT+ V positive definite, and if there exists some E > 0 such that Q > E CTC and W > E r rT, then the optimal decentralized control problem posed has a solution.
The necessary conditions are easily obtained from the incremental cost [18].
‘k+,
= ’ ‘k + b rR Ullk + wk ’
a=1 Y = '2 'k + vRk ak U 1sRsL ak = - Ki?, %k J(K) = lim m h ,i b E(X;+, Qg,Xk+' + ';k Rg ',k) -0 R=l = 4 Xk + r uk + Wk Yk = c Xk + Vk 'k+l Uk = - K Yk cl c= :
r = (r,, . . . . rL)
.
( >
cL K = BLOCK DIAG iKa. lsIIsL{ J(K) = lim K E v
Fkoo & ,I, E(X;+, Q 'k+, + u; R 'k) < O3
MULTIPLE-MODEL DECENTRALIZED CONTROLALGORITHM One approach to obtain an algorithm for the decentralized control problem posed is to close the control loop over all the local controllers, except one, solve the resulting unoonstrained output feedback problem; then iterate .on the next local controller.
While this method does not make use of some of the analytical tools developed, which would result in a more efficient algorithm, a solution to the combined decentralized multiple-model output feedback can be obtained using the previously developed algorithm for multiple-model problems.
1.
CHOOSEA BLOCK DIAGONAL K" in S i=l a=1 2. SOLVE THE MULTIPLE-MODEL OUTPUT FEEDBACK PROBLEM FOR THE SYSTEMS C ic r) j = 7, 2, . . . .
ajs 4j - v R, R r!L1j KI1' t'j' tj PI f; WITH WEIGHTING MATRICES )Qj + 1 CT,. Ki: Ra,j K;, Ce, j = 1, . . . .
PI &'#a. R J R R j = 1, . . . , pI AND STATISTICS )Vej j = 1, 2, . . . . pI Rj ' iT .
i+l w. - a,$R ri,j Kg, Va,j KiI r,jj j = 1, . . . . p1 TO OBTAIN K J 3.
IF R = L, GO TO 4 SET K; = K;+' R=R+l GO TO 2 = K; 4. SET K;+' l<R<L IF IJ(Ki+') - Jo < El AND I~(Ki+')lj I ~~ STOP 5. i=i+l g=l GO TO 2 AN APPLICATION TO RESTRUCTURABLE CONTROLS The optimal stochastic multiple-model, decentralized control, output feedback and dynamic compensation problems formulated provide powerful techniques to inves- tigate a 1arge class of control system design problems. The stochastic output feed- back, dynamic compensation, and decentralized control problems. provide a wealth of control structures; however, the "best" structure(s) for a given design problem On the other hand, are not determined and depend on the practical constraints.
the multiple-model formulation provides a powerful technique to describe design objectives in an optimal control setting, with computable algorithms and imple- mentable structures.
As an illustration, the combined multiple-model decentralized control algor- ithm is used to design a simple aircraft control law which accommodates some While the performance of a control law under types of control actuator failures.
normal conditions is a primary design goal, a practical control design must also The consider the implications of various scenarios, such as actuator failures.
restructurable control problem addresses methods of modifying the structure of the law to accommodate failures. However, the detection of the exact nature of the failure requires a period of time. Depending on the actual failure, during this period of time the aircraft may be forced into a condition from which re- covery is difficult, and sometimes not possible. Therefore, it is reasonable to restructure the control system in stages. As soon as the existence of a failure is known, or even highly likely, the system may be restructured into a control law which can accommodate a large number of failed components. This first stage restructuring can provide the valuable time necessary to identify the exact failure, decide the best second-stage structure, and implement it before the aircraft is forced into a possibly irrecoverable condition. The multiple-model formulation provides a control design method where the law isatleast stable in the failed condition as well as under normal circumstances, when possible. Since the control law is stable for the case of no failure, a false alarm does not produce harmful effects. In the following example, a wing leveler with dynamic compensation and a decentralized structure is modeled with normal and failed aileron for failure accommodation, and at two airspeeds to provide insensitivity.
DISTURBANCE SENSOR MEASUREMENTS 3 1AIRCRAFT t------ t. --__ __ -------d
r
FAILURE L FACTOR 6, r H r------- 1 I VARYING FLIGHT CONDITION V=135 kt V=165 kt HGT HEIGHT FAILURE ACCOMMODATION RUNWAY FLIGHT CONTROL DESIGN EXAMPLE The advantages offered by the'multiple-model decentralized approach are demonstrated using an aircraft digital flight control system design problem.
The control problem is the simplified design of the lateral dynamics, inne,r- loop control system for the NASA ATOPS research aircraft (a Boeing 737). The aircraft model includes the body-axis states, v, p, r, and 4. Also included in the model are aileron and rudder actuators dynamics, a one-state dynamic compensator, and aileron and rudder control states caused by weighting the control difference in the quadratic cost function. Noisy sensor measurements The dynamic comperisator state and control states are noise- are p, r, and 4.
free measurements. The dynamic compensator state is quadratically weighted to "follow" the aircraft v state. The closed-loop eigenvalues using optimal out- put feedback at one trinnned flight condition (V, = 135 kn, Wt = 85,000 lbs, ho = 1,000 ft.) are shown below.
The other table shows the closed-loop eigen- values for the multiple-model, decentralized design at the same flight condi- tion with the same quadratic weights and noise covariances. The Dutch roll mode has the lowest damping in both designs.
OUTPUT FEEDBACK DESIGN MULTIPLE-MODEL DECENTRALIZED DESIGN MAPPED EIGENVALUES (135 KN) MAPPED EIGENVALUES (135 KN) REAL IMAG. REAL IMAG.
- .521 0.00 - .436 0.00 - .903 1.47 - -925 1.93 - .903 -1.47 - .925 -1.93 - 1.53 1.72 - 1.08 1.22 - 1.53 -1.72 - 1.08 -1.22 - 1.25 0.92 - 1.23 0.00 - 1.25 -0.92 - 2.62 0.00 -14.4 0.00 -13.4 0.00 -36.5 0.00 -42.1 0.00 FEEDBACK GAINS FOR SINGLE- AND MULTIPLE-MODEL DESIGNS The lateral dynamics flight control design for the multiple-model case uses four models: two models at 135 kn and 165 kn with the aileron opera- tional, and two models at 135 kn and 165 !cn with the aileron failed. The design is decentralized as four gains in the output feedback gain matrix are forced to be zero. The controls are the dynamic compensator control u, In the multiple-model decentralized aileron rate 6'A, and rudder rate 6'R.
design, no control states are fed back to the dynamic compensator, and aileron and rudder control state crossfeed gains are forced to be zero.. The primary differences in the two gain matrices are the 4 and p gains to 6R which change sign. The fixed-gain multiple-model design stabilizes all four models. The single-model design causes the closed-loop system with the aileron failed to be unstable.
OUTPUT FEEDBACK DESIGN CONTROLGAIN MATRIX COMP. r 6A 6R P @ 0.0032 -3.47 0.829 0.661 0.721 -3.94 -7.09 -3.96 -2.94 0.698 -0.499 3.77 -0.447 0.0055 -2.77 MULTIPLE-MODEL DECENTRALIZED DESIGN I CONTROLGAIN MATRIX L COMP. r 6A BR P 4J QJ -0.80 -0.527 -4.58 0.731 0. 0.
&'A 0.104 -2.49 -4.51 -1.96 -1.94 0.
6-R -0.516 0.721 7.74 0.860 0. -4.05 SINGLE-MODEL DESIGN SIMULATION - NO. FAILURES A linear simulation of the single-model optimal output feedback design is shown below.
Roll attitude is initially 5 dtig at the beginning of the simulation and is smooth1.y returned to zero by the control system.
Lateral velocity v is kept small, and the dynamic compensator state is similar to V.
- DYN COMP ---Z_I__ .----_- I - 1.
i----..- 1 1 1 L-- ____ ________.. - _-- .-..- TIME SEC TIME SEC
r
DA CMD DEG DEG '0. 10 .TIME SEC TIME SEC SINGLE-MODEL DESIGN SIMULATION - FAILED AILERON The linear simulation shows the aileron failing 1.5 set into the simu- lation and remaining fixed at approximately 2 deg. The closed-loop aircraft system is unstable and roll attitude is seen to diverge.
DYN -- \ ROLL COMP DEG [ --.'\ -4t ..---Lee- ._.J.--.____l--.-_--L---~.~'l -61--__.-.eJ __.---.-J-...-.-.m-.mi. ...__. .-_..!........-b!
0 10 0 10 TIME SEC TIME SEC I DR ;_- -__.--.__--_ . _----- _-----.
CMD DEG --- '-‘.\ .__.. / ._-- -1L .___. LL_ ~.~L.-~~~I~~
-I--.. _-..-I ___ -I-.--.__- _I
0 0 TIME SEC TIME SEC FIRST-STAGE RESTRUCTURING: MULTIPLE-MODEL DECENTRALIZED DESIGN - FAILED AILERON The linear simulation shows the aileron failing at 1.5 set into the simu- lation and remaining fixed at that level in the following period. The single- model output feedback design controls the aircraft until 10 sec. As the closed- loop system is unstable in this condition, the aircraft continues to roll past the level wings condition. It is assumed that by 10 sec.,or 8.5 sec. after the aileron failure, the decision that a failure has occurredismade, and the first stage restructuring is engaged. The control law simulated after 10 sec. is the multiple-model decentralized design. As shown by the simulation, the restruct- ured control arrests the roll of the aircraft, and brings it to a non-zero but easily manageable and stable bank angle, providing the time necessary for the second-stage restructuring.
r
DYN ROLL COMP DEG .
-7 -6 I TIME SEC r 3i- DR DA CMD CMD + DEG
I
I
TIME SEC TIME SEC REFERENCES Axaster, S., 1. "Sub-Optimal Time Variable Feedback Control of Linear Dynamic Systems with Random Inputs", Inter. J. of Control, Vol. 4, No. 6, Dec. 1966, --- pp. 549-566.
2. "Optimal Linear Regulators with Incomplete State Feedback", Rekasius, Z. V., IEEE Trans. Automatic Control, Vol. AC-12, June 1967, pp. 296-299.
3. Kosut, R. L., "Suboptimal Control of Linear Time-Invariant Systems Subject to Control Structure Constraints", IEEE Trans. Automatic Control, Vol.
AC-15, October 1970, pp. 557-563.
4. Levine, W. S., and Athans, M., "On the Determination of the Optimal Constant Output Feedback Gains for Linear Multivariable Systems", IEEE Trans. Auto- matic Control, Vol. AC-15, February 1970, pp. 44-48.
5. Kurtaran, B., and Sidar, M., "Optimal Instantaneous Output-Feedback Con- trollers for Linear Stochastic Systems", Inter. J. of Control, Vol. 19, No. 4, April 1974, pp. 154-157.
6. O'Reilly, J., "Optimal Instantaneous Output-Feedback Controllers for Discrete-Time Linear Systems with Inaccessible State", Inter. J. of System Science, Vol. 9, No. 1, Jan. 1978, pp. 9-16.
7. Sobel, K., and Kaufman, H., "Application of Stochastic Optimal Reduced State Feedback Gain Computation Procedures to the Design of Aircraft Gust Allevi- Proc. of IFAC Conference, vol. 2,Helsinki, Finland, 1978, ation Controllers", pp. 1127-1233.
8. Horisberger, H. P., and Belanger, P. R., "Solution of the Optimal Constant Output Feedback Problem by Conjugate Gradients", IEEE Trans. Automatic Control, Vol. AC-19, No. 4, August 1974, pp. 434-435.
9. Athans, M., "The Role and Use of the Stochastic Linear-Quadratic-Gaussian Problem in Control System Design", IEEE Trans. Automatic Control, Vol.
AC-16, December 1971, pp. 529-552.
10. Halyo, N., and Broussard, J. R., "A Convergent Algorithm for the Stochastic Infinite-Time Discrete Optimal Output Feedback Problem", Proc. of the 1981 JACC, Charlottesville, VA, June 1981, p. WA-1E.
11. Doyle, 3. C., and Stein, G., "Robustness with Observers", IEEE Trans. Auto- --- - matic Control, Vol. AC-24, August 1979, pp. 607-611.
12.
Johnson, T. L., and Athans, M., "On the Design of Optimal Constrained Dynamic Compensators for Linear Constant Systems", IEEE Trans. Automatic Control, Vol. AC-15, December 1970, pp. 658-660. - 13. Mendel, J. M., and Feather, J., "On the Design of Optimal Time-Invariant Compensators for Linear Stochastic Time-Invariant Systems", IEEE Trans.
Automatic Control, Vol. 20, No. 5, October 1975, pp. 653-656.-- 14. Broussard, J. R., and Halyo, N., "Active Flutter Suppression Using Optimal Output Feedback Digital Controllers", NASA CR-165939, May 1982.
15. Witsenhausen, H. S., "A Counterexample in Stochastic Optimum Control", SIAM J. of Control, Vol. 6, No. 1, Feb. 1968, pp. 131-147.
16. Whittle, P., and Rudge, J. F., "The Optimal Linear Solution of a Symmetric Team Control Problem", J. of Applied Probability, Vol. 11, No. 2, June 1974, pp. 337-381.
17. Sandell, N. R., Varaiya, P., Athans, M., and Safonov, M. G., "Survey of Decentralized Control Methods for Large Scale Systems", IEEE Trans. on Automatic Control, Vol. AC-23, No. 2, April 1978, pp. 108-128.
18. Caglayan, A. K., Halyo, N., and Broussard, J. R., "The Use of the Optimal Output Feedback Algorithm in Integrated Control System Design", Proceedings of the IEEE 1983 National Aerospace and Electronics Conference NAECON 1983, Vol. 2, 1983, pp. 1242-1251.
PILOT MODELING, MODAL ANALYSIS, AND CONTROL OF LARGE FLEXIBLE AIRCRAFT David K. Schmidt School of Aeronautics and Astronautics Purdue University West Lafayette, IN First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 INTRODUCTION The issues to be addressed in this presentation are threefold. The first deals with the question of whether dynamic aeroelastic effects can significantly impact piloted flight dynamics. If so, when and how does this come about, and is if one were to explore this there a potential design problem? For example, what mathematical model would be appropriate to use in the problem experimentally, simulation? What modes, for example, should be included in the simulation, or what linear model should be used in the control synthesis? The second question deals with the appropriate design criteria or design objectives. In the case of active what should be the design objectives for the control control, for example, Finally, synthesis if aeroelastic effects are a problem. if unacceptable charac- teristics are to be eliminated through active control, what is the achievable performance improvement for practical systems?
(See fig. 1.)
The outline of the topics to be presented includes a description of a model analysis methodology aimed at answering the question of the significance of higher order dynamics. Secondly, a pilot vehicle analysis of some experimental data will be presented that addresses the question of "What's important in the task?"
The experimental data will be presented briefly, followed by the results of an open-loop modal analysis of the generic vehicle configurations in question.
Finally, one of the vehicles will be augmented via active control and the results presented.
ISSUES
l CAN DYNAMIC AEROELASTIC EFFECTS SIGNIFICANTLY IMPACT (PILOTED) AIRCRAFT FLIGHT DYNAMICS?
WHEN - How?
Is THERE A POTENTIAL PROBLEM?
l WHAT ARE APPROPRIATE DESIGN CRITERIA OR OBJECTIVES?
l WHAT IS THE ACHIEVABLE PERFORMANCEIMPROVEMENT VIA ACTIVE CONTROL?
Figure 1 WHAT AFFECTS VEHICLE TIME RESPONSE Linear system theory tells us that a system's response to a particular input may be represented mathematically as the summation of contributions from each of the system's modes. Each mode's contribution, furthermore, may be analytically thought of as a term in the partial-fraction expansion of the transform of the system's response. The significance of the eigenvalues of the system to its response is well known. However, equally significant is the residue associated with each system eigenvalue. For real vehicles, the response theoretically includes an infinite summation over all of the system's modes.
However, practically speaking, only a finite number of these modes contributed significantly to the vehicle response. Furthermore, for a conventional aircraft and considering a short period approximation, for example, only one mode is used to approximate the vehicle's response. Furthermore, stating handling qualities specifications in terms of the modal damping and frequency was sufficient in this case to specify acceptable and unacceptable time responses. It is clear then that when the higher order modes of the system (or the eigenvalues and residues of those modes) are such that they significantly contribute to the time response of the system, those modal contributions must be considered to accurately reflect the system's dynamics.
(See fig. 2.)
CLASSICAL EXAMPLE: RIGID BODY &)/a,(s) (E,G, GUST PULSE) R1 + as) = s + x s Y2A2 + [Ji s hi] X’s AND ZEROS (EIGENVECTORS) Rj’S FUNCTIONS OF AND FOR w !kE THE EFFECTS OF
A’S ON Ri’S WAS ENOUGH TO ALLOW STATING HANDLING
QUAL, SPECS, ON A’S ONLY Figure 2 VEHICLE TIME RESPONSE (CONTINUED) Now consider the same example response, that is, pitch rate to a gust pulse, represented mathematically in the time domain with the usual state equations.
Transforming these equations into modal coordinates and expressing the partial- fraction expansion of the transform of the response, we see that the expansion may be determined directly from the parameters in the system's time-domain modal state representation. Specifically, in fact, the residues associated with each mode are simply a product of the elements obtained from the modal observability matrix and the elements obtained from the modal disturbability or controllability matrix, depending upon which input is being considered.
All these results are developed for the input represented mathematically as an impulse, or the residues are the system's impulse-response residues for the input selected. (See fig. 3.)
= K N(sm)/D(sn)
LET $$-$
.
or X =Ax+Ba
i(t) = Cl, . . . O] i cl IN MODAL COORDINATES .
q = h;+Da = Y(t) = c;i
ht)
D = T-'B C
= cl, 0 . . . O]T
i(s) = Y(s) = C[SI - A-j-'D = f
Now
i=i n -Ait ii(t) = 1 c d AND iie i=i n -hit Ri = Cidi = Ri e where x i=l Figure 3 VEHICLE TIME RESPONSE (CONCLUDED) Furthermore, the residues associated with the system's step response are easily obtained in terms of the previously determined impulse residues and the Finally, this system's step response and eigenvalues for the particular mode.
each mode's contribution to that response may alternatively be considered as the: and the contribution of each mode to that term area under the impulse response, is shown in figure 4.
l RESIDUES FOR THE SYSTEM'S STFP RESPONSE ARE = eigenvalue 'i R; = Ri/hi i = 1, . . . . n .
R; = impulse residue l AREA UNDER IMPULSE RESPONSE DUE TO MODE I IS 'it (e Ai ' = R~/Xi -1) Figure 4 APPLICATION OF THE METHODOLOGY one must determine the appropriate system To apply the technique presented, inputs that are important in the application, as well as determine the significant physical response variables of the system that are important in the piloted task.
Inputs in question include the pilot stick input and atmospheric turbulence.
Important vehicle responses might include rigid-body attitude and rate, sensed attitude and rate including elastic deformation effects, flight path angle, Furthermore, the acceleration at various locations on the vehicle, and so forth.
inputs just cited may not be well modeled by white noise, for example. Therefore, evaluating the pure impulse response of the aircraft is not as meaningful as the response evaluated with appropriate input characteristics included. The input characteristics that are significant are the limited bandwidth properties of the These input characteristics pilot's input as well as the atmospheric gust spectrum.
may be incorporated with the vehicle math model to form what might be referred to as an integrated dynamic model. Modal analysis is then performed on this model such that controllability, disturbability, observability, and, in particular, modal residues may be assessed. (See fig. 5.)
VEHICLE MATHMODEL
+
INTEGRATED
I
MODAL DYNAMIC PILOT BANDWIDTH I - ANALYSIS
I MODEL
+
I
I GUSTSPECTRUM -' / SIGNIFICANT CONTROLLABILITY DISTURBABILITY PHYSICAL OBSERVABILITY RESPONSE 1 - VARIABLES (TASK) Figure 5 SYSTEM RESPONSESOF IMPORTANCE The question now turns to what vehicle responses are significant in this problem. As shown in figure 6, both rigid-body attitude as well as indicated attitude, which includes the local elastic deformation of the vehicle, may be considered significant response variables. Determining the important vehicle responses in a longitudinal task will now be considered.
VEHICLE RIGID-BODY AXIS
k<
/ LOCATIONOF MEASURED SLOPE
-
(E,G, COCKPIT>
'IND
-LOCAL ELASTIC SLOPE
////////
INSTANTANEOUS
VEHICLE C,G,
CENTERLINEOF
DEFORMED VEHICLE
Figure 6 PILOT VEHICLE ANALYSIS To better understand the important vehicle responses in the longitudinal axis, a pilot vehicle analysis was performed on a set of generic vehicle dynamics. An optimal-control pilot model was used in this evaluation with essentially "standard" model parameters. Details of this analysis may be found in reference 1. The pilots in the experimental setting were to perform a pitch-attitude tracking task, and this same task was evaluated analytically as well. The observations available to the pilots were both the indicated attitude of the vehicle, as measured at the cockpit, as well as the commanded attitude that the subjects were to follow. The issue is the selection of the appropriate pilot objective, or the appropriate vehicle response that the pilot was attempting to control. Was the pilot attempting to minimize indicated attitude, which included the elastic deformation of the vehicle, or is rigid-body attitude the response that he is attempting to control?
The geometry of the basic vehicle may be considered to be as shown in figure 7.
Seven different sets of generic vehicle dynamics were evaluated, where the first configuration represents a vehicle similar to the B-l. The remaining config- urations may be thought of as having the same geometric characteristics, but the material properties of the structure are changed such that the in-vacua mode frequencies of the structure are modified from Configuration 1, or the baseline.
The resulting eigenvalues of the dynamic configurations are represented in Table 1.
Figure 7 CONFIGURATION SUMMARY Shown in Table 1 are the in-vacua mode frequencies of the first two elastic modes included in the vehicle models. Mode 1 represents the first fuselage bending mode, while mode 2 has a mode shape that would correspond to the second fuselage bending mode. These mode frequencies were varied parametrically, and the resulting aeroelastic vehicle model was obtained in each case. The eigenvalues of the vehicle are listed in the table as well. Each of the modes was identified from the first four configurations.
its eigenvector or mode shape. Note, in particular, Configurations l-3 arise from a monotonic reduction in vibration frequency of the first fuselage mode.
Configuration 4 (although perhaps unrealistic physically) has a reduced mode frequency associated with the other elastic mode. Further- more, comparing Configurations 3 and 4 in terms of their eigenvalues indicates that both have unstable phugoid modes and approximately equivalent short period eigenvalues, and also they exhibit roughly similar aeroelastic mode eigenvalues.
Table 1 - CONFIGURATION JN-VACUO FRF s PHUGOID SHORT PERIOD EI ASTIC Mw RODE 1. MODE 2 HODE+ MODE* FIRST+ SECOND+ I oUFNC'E I ~(BASELINE) 13,7 21,2 (,02, ,031 (,53, 2,8> (,05, 13,3) (,02,21,4) 2 9,2 21,2 (O,, ,06) (,52, 2,6) (,09, 8,8) (,02,21,4) 3 6.2 21,2 (+,09)(-,08) (,52, la81 (,20, 509) (,02,21,4) 4 13,7 (+.15)(-,13) (,69, 106) (005, 13,3) (,ll, 6.0) 4,8 5 ~ lo,7 (+,05, -003) (055, 2,4) (all, 1083) (O,O, 9.8) 9,3 E to,, ,051 (,54, 2,6) 11,7 11.7 (808, 1187) (0,0,11,6> 7 (+,18)(-,15, l.70, 1.4) (019, 7.3) (0.0, 609) 689 6,9 *MODAL PARAMETER NOTATION, COMPLEX (j,wN), REAL l-P), ALL FREQUENCIES IN RAD/SEC SUMMARY OF EXPERIMENTAL RESULTS Shown in Table 2 is the summary of tracking scores and subjective rating Of significant importance is the associated with the seven dynamic configurations.
pilot comment associated with Configuration 5. He specifically stated that he was attempting to ignore the oscillation that he observed in the display, and he attempts to control the rigid-body attitude. Note in the results a monotonic degradation in tracking performance and subjective rating as the elastic mode frequencies in Configurations 1-3 are reduced.
Table 2 Cooper-Harper RMS Error (deg) Configuration Conents (mean + la) Rating (mean _+ lo) 1 1.2 ? 0.6 1.6 f 0.4 Very'nice; No problem.
Little oscillation; More diffi- 2 1.0 f 0.5 2.0 + 0.3 cult than Cl; Slight control response lag.
Difficult; Required high concen- 3 5.7 f 1.1 5.9 f 1.9 tration; PI0 problem; Extreme response lag.
Little more difficult than Cl; 4 1.9 f 0.3 3.1 f 1.1 Slightly sluggish attitude response.
5 1.2 f 0.5 1.9 i 0.4 Not difficult, little more oscil- lation, but could ignore it and fly rigid-body; Like config. 2.
6 1.5 + 0.7 2.0 f 0.5 Pretty good; Same as 2.
7 7.6 f 2.8 6.7 i 1.6 With severe oscillations, virtu- ally uncontrollable. Abrupt con- trol inputs led to disaster.
COMPARISONOF ANALYTICAL AND EXPERIMENTAL RESULTS Shown in figure 8 is the comparison of the tracking performance obtained analytically with the pilot/vehicle analysis and the performance obtained experimentally, shown in the previous table. Note in the case of Configuration 3, the low tracking score predicted from the model under the assumption for example, This error, as that the subjects were attempting to control the displayed error.
you recall, included the elastic contribution to the displayed attitude. On the other hand, modeling the task as a rigid-body attitude control task results in the analytical tracking errors as shown in the figure.
SIMULATION A ANALYTICAL clANALYTICAL W. J,(B1,,) Cl C2 C3 C4 C5 C6 C7 CONFIGURATION Figure 8 COMPARISONOF ANALYTICAL AND EXPERIMENTAL SUBJECTIVE RATINGS Shown in figure 9 is the excellent correlation between experimental subjective ratings and the ratings obtained analytically from a model-based metric.
The metric used was simply the magnitude of the quadratic cost function obtained naturally in the modeling process. These results, and the results of the previous figure, indicate that rigid-body attitude is the primary control variable in the closed-loop pitch tracking task. Also, experimental and analytical results indicate clearly that as elastic mode frequencies coalesce with the rigid-body modes, the tracking performance is significantly degraded.
.
SIMULATION
A ANALYTICAL, Jp(e~>
A
Q
.
.
6 0
Cl C2 C3 C4 C5 C6 C7
CONFIGURATION
Figure 9
I’
OPEN-LOOP VEHICLE ANALYSIS Performing the modal analysis outlined previously on the seven configurations results in the modal residues shown in figures lo-13 for Configurations l-4. (See Ref. 2 for complete results.) Along with other system's responses, these results indicate cle?rly the contribution of the first aeroelastic mode to the rigid-body pitch rate (8,) pilot impulse response. We refer to these residues as pilot impulse residues because they include the important characteristic of limited pilot bandwidth. This is modeled simply as an impulse passed through a first-order lag with time constant representative of that of the pilot. Note the monotonic increase in the residue in the rigid-body pitch rate associated with the first aeroelastic mode, as the frequency of this mode is reduced (Conf. l-3). This residue in the case of Configuration 3 is actually larger than the residue associated with the "rigid-body" short period mode. Reiterating, in the case of Configuration 3, the rigid-body pitch rate response is dominated by the first aeroelastic mode, where dominance is defined in terms of residue magnitude.
PILOT IMPULSE RESIDUES
CONFIGURATION 1
= 8.25 (g)
ZRj
PH SP El E2 PH SP El E2 = 2.39 (g,
.CRi
PH SP El E2
PH SP El E2 = 1.43 (rad) = 0.49 (r’ad) CR:
ZRi
w 1.0
I
$ .5 z E
irind’
= .o PH SP El E2 PH SP El E2 .
Figure 10
PILOT IMPULSE RESIDUES
CONFIGURATION 2
PH SP El E2 PHSPEl E2 PH SP El E2 PH SP El E2 TR, = 1.59 b-ad) IND PH SP El E2 PH SP El E2 Figure 11
PILOT IMPULSE RESIDUES
CONFIGURATION 3
= 0.53 (rad) CRi w 1.0 .5 z E x .O PH SP El E2 PH SP El E2
.-
il;n-EJ
‘I, P PH SP El E2 PH SP El E2 = 0.24 (t-ad) CR-j I& 1 .o i= Y .5 = .o m---l PH SP El E2 PH SP El E2 Figure 12 OPEN-LOOP VEHICLE fwuxxs (C~NCL~IBED) Note, on the other hand, the residues for rigid-body pitch rate for Configuration Although the contribution of this elastic mode to the response is measurable, 4.
the response is still dominated by the short-period mode. This result, along with the results for Configuration 3, explains why the tracking performance and sub- jective rating for Configuration 3 were so drastically inferior to those of Configur- This was true in spite of the fact that these two configurations had ation 4.
roughly comparable eigenvalues. Clearly, the eigenvalues alone do not completely explain the results obtained experimentally.
PILOT IMPULSE RESIDUES
CONFIGURATION 4 &I, = 8.85 (g)
CRj
w 1 .o
z 6
IND .5 z E .O PH SP E-l E2 PH SP El E2 = 53.76 (g’s)
CRi
w 1.0 %z z P z PH SP El E2 PH SP El E2 = 1.29 (rad)
CR-i
w 1.0 !3 c .5 IND z $ .o PH SP El E2 PH SP El E2 Figure 13 MODIFICATION THROUGHMODAL CONTROL Given a linear system's dynamics, a control law can be determined such that the closed-loop eigenvalues and eigenvectors are modified to exhibit more desirable characteristics (fig. 14). The number of closed-loop or augmented system modes that may be "placed" is equal to the number of measurements available for feedback, and the freedom to specify the mode shapes, or eigenvectors associated with these modes, depends on the rank of the control vector. (See Refs. 3, 4). Control laws based on this theoretical concept may be implemented with constant gain measurement- feedback architecture, or they may be synthesized with linear quadratic Gaussian (LQG) optimal control (Ref. 5). In the case of measurement feedback, if insuf- ficient measurements are available, the unspecified system modes may be unstable.
Conversely, control synthesis using LQG invokes the asymptotic properties of such To explore controllers and theoretically guarantees augmented system stability.
the achievable performance that might be obtained through such modal control concepts, we have augmented one of the seven vehicle configurations considered previously (Configuration 2) with a constant gain feedback controller with gains determined directly from the eigenspace assignment goal.
EIGENSPACE ASSIGNMENT (MODAL CONTROL) GIVEN THE LINEAR SYSTEM A GAIN G MAY BE FOUND SUCH THAT IF (A + BGCW, =A;Gi DYNAMICS i = Ax + Bu XiAND~ ; i = 1. ( ,M HAVE DESIRABLE CHARACTERISTICS MEASUREMENTS Z = Cx; DIM Z = M CONTROLS U = Gz; DIM U = R Figure 14 UNAUGMENTED VEHICLE MODES Shown in figure 15 are the mode shapes associated with three of the four modes of interest for Configuration 2 discussed previously. Recall that this configura- tion differs from the baseline in that the first fuselage bending mode frequency is reduced to approximately 9 radians per second. The baseline on the other hand had a first elastic mode of about 13 radians perjsecond. It is evident in the figure mode actually includes a significant amount of elastic that the "short period" deformation. In contrast, the first aeroelastic mode also reflects the presence of rigid-body attitude in its mode shape. It is due to these modal characteristics that the rigid body response was degraded and the elastic mode's contribution was significant in the rigid-body pitch rate.
SHORT FIRST SECOND PERIOD ELASTIC ELASTIC ( E= .52, wn = 2.57) ( 5= .08, W" = 8.8) ( 5 = .02, wn = 21.4) Figure 15 / UNAUGMENTED VEHICLE STEP RESPONSE Shown in figure 16 is the rigid-body pitch rate step response for The contribution of the first aeroelastic mode in this time ' Configuration 2.
response is clearly evident. The eigenspace assignment goal used for augmenting these dynamics included increasing the short period frequency slightly and increasing the damping of the first elastic mode from 0.08 to 0.20.
In addition, the eigenvectors associated with these two modes were modified.
The.short-period eigenvector was to represent pure rigid-body response, while the first elastic eigenvector was selected for purely elastic deformation.
.2000 .oooo 2 -.2000 n .z x - -.40GO -43 -;6000 -.I3000 0 .ooo 1.000 2.000 3.000 Ll .ooo
TIME CS)
Figure 16 AUGMENTEDVEHICLE MODE SHAPES Shown in figure 17 are the eigenvectors of the augmented vehicle modified the short period mode approaches that of a through modal control. Clearly, In this , while the first elastic mode is purified as well.
"rigid" vehicle example, only 4 measurements were selected for feedback (i.e., two accelerometers, along with pitch rate and pitch appropriately positioned in the fuselage, attitude gyros). Consequently, only 4 eigenvalues (or two modes) were specified The phugoid mode and the second aeroelastic mode were not placed in this case In addition, a control but could be if more measurements are made available.
law with limited bandwidth should be selected such that the second elastic mode would be attenuated.
SHORT FIRST SECOND PERIOD ELASTIC ELASTIC (F,= .53, IA = 2.8) ( 6 = .20, w" = 8.8) ( 6 = .02, wn = 20.8) n Figure 17 AUGMENTEDVEHICLE STEP RESPONSE Shown in figure 18 is the step response of the augmented vehicle. When compared to that in Configuration 2 the reduction of the contribution of The long the elastic mode to this rigid-body pitch rate response is evident.
period divergence of this response is due to the phugoid mode instability.
This demonstrates one of the shortcomings of implementing modal control through Additional measurements or measurement feedback alone as cited previously.
equalization are required to stabilize the phugoid mode.
.0500 .oooo -.0500 ii w n l W -.lOOO -.lSOC -.2ooc I I I I 3.000 q .ooo 1.000 2.000
TIME (Sl
Figure 18 EFFECT OF AUGMENTATIONON RESIDUES Shown in figure 19, finally, is the effect of the augmentation on These residues are comparable residues for the pitch rate impulse response.
with those shown previously for the seven configurations. Compared to the unaugmented vehicle (Configuration 2), the results for the augmented vehicle clearly indicate the dominance of the short-period mode in this response.
In conclusion then, we see that handling characteristics, as measured by tracking performance and subjective rating in the tracking task, were significantly degraded due to the presence of dynamic aeroelastic effects.
The rigid-body attitude angle was shown to be fundamental in the vehicle's response in this task. Furthermore, this response may be dominated by "aeroelastic" modes in severe cases. Clearly, from these results, a rigid-body mathematical model of the vehicle is inappropriate. Finally, the ability to modify the modal characteristics of the vehicle through modal control or eigenspace assignment appears to have merit in this application. Multiple control surfaces and an appropriate sensor complement will be required to implement practical modal controllers. Appropriate design criteria for flexible vehicles might be expressed in terms of allowable residue magnitudes of the higher order mode.
AUGMENTATION OF PITCH RATE RESIDUES IMPULSE RESPONSE = 7.9 RAD/S c
Ri
-0" 1.0 WY . 5,- E 0-l . OL PH SP PH SP El E2 MODES Figure 19 REFERENCES "Pilot Modeling and Closed-Loop Analysis of Flexible 1. Schmidt, K.K., Aircraft In the Pitch Tracking Task", AIAA Paper No. 83-2231, Guid.
and Control Conference, August 1983.
2. Schmidt, D.K., "A Modal Analysis of Flexible Aircraft Dynamics With Handling Qualities Implications," AIAA Paper No. 83-2074, Atm.
Flight Mech. Conference, August 1983.
and Crossby, T.R., Modal Control Theory and Applications, 3. Porter, B., -- Barnes and Noble, New York, 1972.
4. Moore, B.C., "On the Flexibility Offered by State Feedback In Multivariable Systems Beyond Closed Loop Eigenvalue Assignment," IEEE Trans. Auto. Control, Vol. AC-21, No. 5, October 1976, pp.
689-691.
5. Harvey, C.A., Stein, G., and Doyle, J.C., "Optimal Linear Control (Characterization of Multi-Input Systems)," Office of Naval Research CR 215-238-2, March 1977.
NONLINEAR SYSTEMSAPPROACH TO CONTROL SYSTEM DESIGN George Meyer NASA Ames Research Center Moffett Field, CA First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 NONLINEAR CONTROL SYSTEM DESIGN METHODS Consider some of the control system design methods for plants with nonlinear If the nonlinearity is weak relative to the size of the operating region, dynamics.
Fixed-gain design may be feasible even for then the linear methods apply directly.
significant nonlinearities. It may be possible to find a single gain which provides If the non- adequate control of the linear models at several perturbation points.
that fact may be used to obtain a fixed-gain linearity is restricted to a sector, controller, Otherwise, a gain may have to be associated with each perturbation point A gain schedule K(p(v)) is obtained by connecting the perturbation points by a Pi' p(v), of the scheduling parameter v (i.e., speed). When the sche- function, say duling parameter must be multidimensional, this approach is difficult; the objective of our research is to develop an easier procedure.
i = f (x, u, 2) 1. SMALL OPERATING REG’ION - LINEAR METHODS 2. FIXED-GAIN DESIGN a. ONE GAIN FOR ALL {pi} b. ONE GAIN FOR SECTOR NONLINEARITIES 3. GAIN SCHEDULING pi = p(Vi), K(P(v)), V = SCHEDULING PARAMETER 4. NONLINEAR TRANSFORMATION OF STATE AND CONTROL SIMPLIFICATION THROUGHCOORDINATE CHANGE We attempt to simplify the design problem by simplifying the representation of the plant. Here is an example. The state equation is nonsingular, but a fixed-gain design is impossible. The nonsingular matrix E represents rotation in the plane of x through the angle I). The change of control coordinates from u to v results in a globally constant, linear system.
Even an approximate cancellation of E could be quite helpful, since the resulting system may be nearly constant and the fixed- gain design may be applicable.
xeR *, ucR *, EE7 = I i = E (J/(t)) u $=O--+i=u,i=Kox,STABLE =m h;= ‘U, Ii= -K,x, UNSTABLE lb NEW CONTROL COORDINATES: vi= v, GLOBALLY v = E(W) u APPROXIMATE TRANSFORMATION: v=E($)u-;= E (6$(t)) v FIXED GAIN MAY WORK BRUNOVSKYCANONICAL FORM The first step in the design approach is to try to transform the given system into something more simple, i.e., ideally, a set of decoupled strings of integrators. This set is called the Brunovsky canonical form for controllable, constant, linear systems.
The number of strings is given by the number of control axes m, and the number of In general, integrators (dots in the figure) is given by a Kronecker index Ki.
= n. Two examples are shown for the case of n = 12 and m = 4. According to CKi the set of Kronecker indexes is invariant under nonsingular transforma- linear theory, tions and feedback.
GIVEN i = Ax + Bu, xeR”, ueRm, (A, B) CONTROLLABLE CAN BE TRANSFORMED WITH y=Tx, v=Rx+Wu, (T, WI NONSINGULAR INTO m DECOUPLED STRINGS OF INTEGRATORS, LENGTH = KRONECKER INDEX EXAMPLE : n=l*, m=4 * w 0 v1 - v1 - - l w l v2 v2 - v3 - v3 - v4 - v4 -- KI = (4,4,2,2) KI = (3,3,3,3) OUTLINE OF DESIGN PROCEDURE In principle, the design procedure is to transform the natural representation of the plant into the corresponding Brunovsky form, then design a control law for the canonical system, and finally pull the law back into the natural v = g(y) coordinates in terms of which law must be implemented.
CANONIC COORDINATES NATURAL COORDINATES u = W(x, v) i=f(x,u) 4 ) i=A,y+B,v Y=T(x) KI = (4,4,4,2) u = W kg U (x))) v = cl(Y) DEVELOPMENTOF THE DESIGN APPROACH Before this design approach becomes practical, several issues must be resolved.
(1) When can a given system be transformed into a linear model, how does one construct the required transformation, and how feasible is it to implement the resulting algorithm on flight computers?
(2) What is the structure of the complete control system that includes the linearization step?
(3) How robust is the resulting design?
(4) How can the constraints on control and state be enforced?
These issues have been explored both theoretically and experimentally. The results are summarized in the following figures.
1. TRANSFORMATION EXISTENCE a.
b. COMPUTATION c. IMPLEMENTATION 2. STRUCTURE OF COMPLETE CONTROL SYSTEM 3. COMPLEXITY AND ACCURACY OF MODEL - ROBUSTNESS a. EXACT STATE SPACE b. TRUNCATED STATE SPACE 4. ENFORCEMENT OF DESIGN CONSTRAINTS 5. EXPERIMENTS a. FORTRAN REAL-TIME (FLIGHT COMPUTER, HYDRAULICS) b.
SIMULATION C. FLIGHT 3-34 LIE BRACKETS The key theoretical result from the existence of linearizing transformations is the result of work by Krener (ref. l), Brockett (ref. 2), Jakubcyzk and Respondek (ref. 3), and Hunt and et al. (ref. 4). The necessary and sufficient conditions are best expressed in terms of lie brackets.
A lie bracket (f, g) constructs a new vector field from the old ones f and gi A set is involutive if the brackets do not create new directions.
VECTOR FIELDS f, g: R”+ R” LIE BRACKETS $f-gg [f, 91 = INVOLUTIVE SET hl,..., hr) IF [hi, hjI E SPAN (hl, mms , hr ) CONDITIONS FOR LINEARIZABILITY It is necessary to find control coordinates which appear linearly in the state 1.
equation. For aircraft this means angular acceleration instead of ailerons, elevator, and rudder.
The resulting system must have linear-like controllability.
2.
3. The fields gm) must satisfy an involutivity condition. For (f, g,, g, - - 0, n = 6, m = 2, and the Kronecker index set KI = (3, 3). Then, the example, let six vector fields must span the state space and the first four must be involutive 1. ic=f(X,“)~ic=f(X)+~ gi(X)U* 2. LINEAR - CONTROLLABLE 3. INVOLUTIVE EXAMPLE: i=f(x) +q (xh,++g* (xhlq’ KI = (3,3) I INVOLUTIVE STRUCTUREOF THE COMPLETECONTROL SYSTEM The structure has the form of an exact model follower. The linearizing transfor- mation is done in the WT-map. Through this map the plant is seen as a set of decoupled strings of integrators. The same set is employed as the dynamics (A,, Bo> of the model servo, where the desired motion is defined by means of the input r* and the, in general, nonlinear control law. The control of modeling inaccuracies and other disturbances is accomplished by the regulator which operates on the error ey and outputs corrective control 6v. It may be noted that the regulator works into the simple canonical dynamics and, in effect, the gain scheduling is done automatically by the WT-map.
--a------- r r*
h(e)
iI
I 1 LAW WT-MAP Al RCRAFT MODEL SERVO REGULATOR HELICOPTER EXPERIMENT The objective of this experiment is to investigate the effectiveness and realism of the design approach and to uncover potential problem areas. The current work is with the UHlH helicopter equipped with the VSTOIAND avionics system including The model used in the design is a rigid-body the Sperry 1819B flight computer.
Inertial coordinates nonlinear force (fF), and moment (f") generation Process.
v) of position and velocity vectors, body attitude matrix C, and angular veloc- (r, The moment controls uM ity w form the state. are the roll and pitch cyclic and The collective is the power control up.
the pedals.
STATE: r V X= E X C R3 X R3 X SO(3) X R3 C w CONTROL: U= dCR3XR STATE EQUATION: ; =v .
v = fF(x, u) c = S(0) c ; = f”(x, u) A CANONICAL MODEL FOR THE HELICOPTER There is a pair of strings, The canonical model chosen is shown in the figure; each four integrators long. This pair represents the two horizontal channels The In addition, there is a two-integrator string for altitude h.
(TX, r 1.
fourthYs,tring is for the heading $. The canonical controls are the second deriva- The tives of horizontal acceleration, vertical acceleration, and yaw acceleration.
First, the transformation is computed by means of two Newton-Raphson.trim..routines.
controls are computed and yield the given accelerations (wbc, h) at the given state.
The Then the attitude C is computed for the given horizontal acceleration.
Jacobian matrices needed by Newton-Raphson are computed numerically.
CANONICAL MODEL .
..
ax ax Vx TX ax - ..
aY,, - & ;v .
ii--d-oh KI = (4,4,2,2) TRANSFORMATION a. ON-LINE NEWTON-RAPHSON MOMENT TRIM fM (r,v, C, wb, u) = &bc UC =
hM (‘4 c,Wb, li,, \;r,,&
f!j (r,v,C,wb,u)= tic I-- b.
ON-LINE N-R FORCE TRIM f E (r,v,C,o,u,.,) = ah - t&e) = h F (LV,$,ah ,O.O) SUMMARY - UHl EXPERIMENT The experin.ent has progressed to the point that the following observations can be made.
(1) The implementation of the control scheme is practical. The complete code takes less than 22 msec on the Sperry 1819B flight computer (2) Both FORTRAN and real-time simulations (real flight computer and hydraulics) are consistent with theory. The tracking accuracy along the trajectory including hover, climb, descent and high-speed flight indicates that the UHl model routinely used for manned simulations is linearizable to a degree where a fixed-gain controller is possible. There is robust- ness with respect to weight, center of mass, moment of iner- tia, and force and moment models, but there is sensitivity The actuator dynamics (15 to errors in wind estimates.
rad/sec) may be commuted with the TW-map for regulator bandwidth below 2 radlsec (3) Further research is needed to develop rigorous methods for including high-frequency dynamics and explicit enforcement of control and state constraints 1. IMPLEMENTATION IS PRACTICAL SAMPLING TIME = 50 msec NEW JACOBIANS FOR N - R TRIM EVERY 5TH SAMPLE COMPLETE CODE = 22 msec ON SPERRY 1819B 2. FORTRAN AND MANNED SIMULATIONS ARE CONSISTENT WITH THEORY TRACKING ACCURACY IMPLIES LINEARIZABILITY ROBUSTNESS ACTUATOR DYNAMICS 3. PROBLEM AREAS METHOD FOR INCLUDING SERVO DYNAMICS METHOD FOR EXPLICIT ENFORCEMENT OF CONSTRAINTS REFERENCES 1. Krener, A. J.: On the Equivalence of Control Systems and the Linearization of Nonlinear Systems. Siam J. Control, vol. 11, no. 4, Nov. 1973, pp. 670-676.
2.
Brockett, R. W.: Feedback Invariants for Nonlinear Systems. A Link Between Science and Applications of Automatic Control. Proceedings of the IFAC Congress (Helsinki, Finn.), vol. 2, 1978, pp. 1115-1120.
3. Jakubczyk, B.; and Respondek, W.: On Linearization of Control Systems. Bull.
Acad. Polon. Sci., Ser. Sci. Math. Astronom. Phys., vol. 28, 1980, pp. 517-522.
4. Hunt, L. R.; Su, R.; and Meyer, G.: Global Transformations of Nonlinear Systems.
IEEE Trans. on Automatic Control, vol. 28, no. 1, Jan. 1983, pp. 24-31.
ADAPTIVE CONTROL: MYTHS AND REALITIES Michael Athans and Lena Valavani Massachusetts Institute of Technology Boston, Massachusetts First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ADAPTIVE CONTROL: MYTHS AND REALITIES In recent years, the area of adaptive control has received a great deal of attention by both theoreticians and practitioners. Considerable theoret- ical progress was made in the design of globally asymptotically stable adap- tive algorithms and in unifying different design philosophies under the same mathematical framework.
In this vein, it was found that all currently existing globally stable adaptive algorithms have three basic properties in common Cl]: (1) positive realness of the error equation, (2) square-integrability of the parameter adjustment law and, (3) need for "sufficient excitation" for asymptotic parameter convergence. Of the three, the first property is of primary importance since it satisfies a sufficient condition for stability of the overall system, which is a baseline design objective. The second property has been instrumental in the proof of asymptotic error convergence to zero, while the third addresses the issue of parameter convergence.
Positive-real error dynamics can be generated only if the relative degree (excess of poles over zeroes) of the process to be controlled is known exactly; this, in turn, implies "perfect modeling." This and other assumptions, such as absence of nonminimum phase plant zeros on which the mathematical arguments are based, do not necessarily reflect properties of real systems. As a result, it is natural to inquire what happens to the designs under less than ideal assumptions. In particular, this paper will be concerned only with the issues arising from violation of the exact modeling assumption which is extremely restrictive in practice and impacts the most important system property, stability.
THEME A variety of adaptive control algorithms which reflect different philo- sophical approaches to control system design have been suggested in the literature.
Such algorithms as a rule tend to improve control system performance due to enhanced information obtained while the system is in operation. In addition, in the late 1970's, certain classes of adaptive algorithms repre- senting the majority of those available (among which the self-tuning regula- tors, model reference adaptive controllers, and the so-called dead-beat algorithms) were proven theoretically to be globally asymptotically stable.
In practice, however, such algorithms would almost surely result in unstable physical control systems, as recent research has indicated.
l MYTH WE HAVE A VARIETY OF ADAPTIVE CONTROLALGORITHMS THAT (1) IMPROVE CONTROLSYSTEMS PERFORMANCE (2) ARE PROVEN TO BE GLOBALLY STABLE l REALITY ABOVE ADAPTIVE ALGORITHMS ALMOST SURELY WOULDRESULT IN UNSTABLE PHYSICAL CONTROLSYSTEMS BASIC PROBLEM The basic problem behind the apparent inconsistency between theory and practice can be attributed to the fact that existing adaptive control algo- rithms have been focused almost exclusively upon performance improvement, without due consideration for system robustness as required in the presence of unmodeled dynamics and/or other unstructured modeling errors. In fact, fundamental system concepts, such as operating system bandwidths, have been ignored in the design of adaptive algorithms that tend to invariably adjust the parameters in response to output errors, regardless of their origin. As a result, although at first sight performance seems to be greatly improved, the system bandwidth grows without bound, and eventually the hard con- straints imposed by the presence of high-frequency unmodeled dynamics are The final result is violent instability of the controlled system violated.
that makes apparent at the same time the nonlinear nature of the overall feedback adaptive loop.
' MODELS HAVE LIMITATIONS; STUPIDITY DOESNOT ' EXISTING ADAPTIVE CONTROL ALGORITHMS HAVE FOCUSED UPONPERFORMANCE IMPROVEMENT . STABILITY/ROBUSTNESS ISSUE WASNEGLECTED - HIGH-FREQUENCY UNMODELED DYNAMICSIMPOSEHARD LIMIT UPONCONTROL SYSTEM BANDWIDTH l ADAPTIVE CONTROL SYSTEM BANDWIDTH CAN GROW WITHOUTBOUND - PERFORMANCE LOOKSGREAT - SYSTEM EVENTUALLY BREAKSINTO VIOLENT INSTABILITY.
ADAPTIVE ALGORITHMS CONSIDERED The algorithms considered in the present study can be classified into one of the following three categories: (1) model reference adaptive control algorithms, otherwise known as direct adaptive control [2-71; (2) self- tuning regulators, otherwise known as minimum variance controllers 18-101; (3) dead-beat algorithms designed for discrete-time systems, otherwise referred to as projection or least-squares algorithms 1111.
They all differ in the parameterization of the controller and, hence, the form of the resulting error equations, in the way they synthesize the control input and in the specific realization of the parameter adjustment laws. The common features of the above seemingly fundamentally different algorithms include the assumption of minimum phase plant zeros, the basic signal correlation of the learning mechanism, and the exact knowledge of the plant relative degree. The latter has proven crucial in obtaining global asymptotic stability proofs for all the above-mentioned algorithms.
l COMMON THEME: GLOBAL STABILITY PROOFS AVAILABLE . MODELREFERENCE ADAPTIVE CONTROL MONOPOLIET AL NARENDRA ET AL MORSEET AL LANDAUET AL - -' - --' - -' -- . SELF-TUNING CONTROLLERS ASTROMET AL EGARDT, LANDAUAND SILVIERA --' . DEAD-BEAT ADAPTIVE CONTROL GOODWIN,RAMADGE,AND CAINES STRUCTLJRE OF ADAPTIVE CONTROL The general structure of an adaptive control system is shown in the first figure below. The term g(s) represents the nominal (low-frequency) plant transfer function, whose parameters are considered unknown.
The term K(s) represents the compensator whose parameters are adjusted on-line on the basis of information generated by the adaptive logic block; its basic compo- nent is the "learning mechanism" that adjusts the compensator gains either directly or by identifying them first on the basis of the error e(t) and the signals r(t), u(t), and y(t), along with their associated auxiliary state variables. The term a(s) represents a multiplicative high-frequency model- ing error whose frequency profile is shown in the second figure below. The term J,(s) has been implicitly assumed to be identically zero at all frequen- cies in all the algorithms that have been proven to be globally asymptoti- cally stable.
g(s): LOW-FREQUENCY MODEL OF PLANT l(s) : HIGH-FREQUENCY MODELING ERROR K(s) : COMPENSATOR WITH ADJUSTABLE PARh4ETERS INSTABILITY MECHANISMS When the ideal assumption of exact modeling is violated, at high frequencies, i.e., E(s) # 0, two mechanisms of instability were identified in the algorithms studied [12]. The first is the so-called "phase instabil- ity" that arises as a result of high-frequency inputs to the plant. For sufficiently high frequencies, the unmodeled dynamics contribute enough lag so that the total phase shift of the overall loop reaches 180", at which point the feedback becomes positive and instability occurs. The second mechanism is referred to as "gain instability" and is due to persistent unmeasurable output disturbances and/or nonzero steady-state errors. In this case, the adaptive control system feedback gains keep drifting to increasingly larger values with a resulting increase in bandwidth; as a result, the high-frequency dynamics get excited, and the closed-loop system becomes unstable.
' INSTABILITY DUE TO HIGH-FREQUENCY INPUTS - HIGH-FREQUENCY DYNAMICS YIELD +180° PHASESHIFT - . INSTABILITY DUE TO PERSISTENT UNMEASURABLE OUTPUT DISTURBANCES - ADAPTIVE CONTROL SYSTEM FEEDBACK GAINS DRIFT AND GET LARGE - CONTROL SYSTEMBANDWIDTH INCREASES - HIGH-FREQUENCY DYNAMICS GET EXCITED - CLOSED-LOOP SYSTEM BECOMES UNSTABLE ESSENCE OF PROOFS The previously discussed error mechanisms can be better visualized in the figure below which is a generic representation of the error system complete with adjustment mechanism in the feedback path.
Here C[y(t), u(t) 1 u(t)] are linear operators that generate auxiliary state and D[y(t>, variables through stable filters , F(s) can represent a shaping filter, and 4 represents a matrix of constants.
The essence of global stability proofs is captured in the figure shown, with a positive-real transfer function in the forward path and a (passive) adaptation mechanism in the feedback [l]. All globally stable adaptive algorithms construct a positive-real function based upon the nominal plant model, which governs the error dynamics [l]. However, the positive-real condition is always violated in real applications due to unmodeled dynamics.
l REF: VALAVANI, PROC. JACC, 1980 ( REF. 1) l ALL GLOBAL STABILITY PROOFS AND ASSOCIATED ALGORITHMS CONSTRUCT A POSITIVE-REAL FUNCTION BASED UPON NOMINAL PLANT MODEL d(t) NOMINALLY CONTROLLED PLANT . PITFALL: POSITIVE-REAL CONDITION ALWAYS VIOLATED IN REAL APPLICATIONS IJNMODELED HIGH-FREQUENCY DYNAMICS - CAUSE: SIMPLEST MODEL REFERENCEADAPTIVE CONTROL STRUCTURE For the sake of example, the simplest model reference adaptive control structure is depicted below [3].
The model and plant transfer functions are given respectively by Stable filters - generate from auxiliary state u(t) and g(t) P(s) variable vectors s(t) and xy(t), which multiply feedback gain vectors b(t) and ky(t), as shown.
These gains, along with k,(t), are adjus- ted according to the equation in the square.
The overall scheme looks indeed very simple for real-time implementation!
' STUDIED BY NARENDRA AND VALAVANI Model ' ADAPTIVECONTROL GAIN ADJUSTMENT /-zy- NUMERICAL EXAMPLE The simple structure outlined on the previous page was employed for the adaptive control of a nominally first-order plant with a pair of complex poles. The reference model, the nominal plant complete with unmodeled dynamics, and the adjustment laws were as described below. The digital simulation results for a number of different reference input and disturbance combinations corroborate the foregoing discussion.
i,(t) =Ylr(t)e(t) ky(t) =y,y(t)e(t)
l UNMODELED DYNAMICS - 9,(s) = a,(s) = s2+30s+229 POLESAT: s=-15+j - - k(s) = I1*(s) = 2 s +85+100 POLESAT: ~-438.33 INSTABILITY DUE TO HIGH-FREQUENCY INPUT The figures below show the plant output and adaptive gain evolution, respectively, when the reference input was chosen to have a d.c. and a sinusoidal component as shown below, in the absence of any output disturb- The input frequency was precisely the frequency at which the ance.
"nominally controlled plant," with unmodeled dynamics has 180" Q(s), phase shift. The output displays an exponential-type oscillatory growth, while the parameters keep drifting and finally diverge.
. DATA: k(s) = ys) r(t)=0.3 + 1.85 sin 16.lt; d(t)=0
‘7
ADAPTIVE GAINS ; k Y INSTABILITY DUE TO SINUSOIDAL DISTURBANCE With the same reference model, plant, and unmodeled dynamics, the reference input was now chosen to be a simple d.c. input of amplitude 2 The frequency an output disturbance was added to the experiment.
(units); of the disturbance was lower than that which would cause a 180' phase lag.
The figure below shows the plant output and parameter evolution, respec- it looks as if the plant output For a very extended period of time, tively.
has converged to within a satisfactory deviation from the desired output; Finally, both plant and parameters however, the parameters keep drifting.
This is the gain instability mechanism that break into abrupt instability.
is indicative of the nonlinear nature of the overall adaptive loop.
l DATA: a(s) = a,(s) r(t)=2.0 d(t)=O.S sin8t
I!
INSTABILITY DUE TO SINUSOIDAL DISTURBANCE (CONCLUDED) In the present experiment, the only difference from the preceding example is that the magnitude of the disturbance is' smaller (by a factor of 5). the same trends are observed, although now the plant output Again, deviates much less from the reference model output and the parameter evolu- tion is different in shape and magnitude. the final instability However, comes about in an almost identical manner.
l DATA: k(s) = a,(s) l r(t)=2.0 d(t)=O.l sin8t L.c OUTPUT c TIHF: is 100 .oo rfl0.00 :i 160.00 ‘2.lD .OO 320.00 Q.ClO flO.lJO w3 I INSTABILITY DUE TO CONSTANT DISTURBANCE This is a regulator-type experiment for the same setup as before.
There is a zero reference input and a d.c. disturbance of amplitude 3 Notice that the output behaves almost ideally for an infinitely (units).
long period of time compared to the system time constants. However, after about 1000 seconds some arrhythmia-type behavior is observed, after which point violent instability occurs in both plant output and parameter values.
We remark that the parameters have continued to drift while the output was displaying a satisfactory behavior. The fact that the gain instability has taken such a long time to develop may have been a reason why the phenomenon was not discussed in the literature earlier. However, it was indeed observed in some applications to chemical processes 1131.
' DATA: k(s) = p d(t)=3 r(t)=0 OUTPUT y, ADAPTIVE GAINS RULES OF THUMB In general, although the time at which instability occurs may be large, Factors that can prolong its onset are the fact remains that it does occur.
increased frequency separation between nominal plant model and unmodeled high-frequency dynamics and decreasing'disturbance amplitudes.
At present, there is no systematic way to prevent such a phenomenon from occurring. The reason is because the adaptive loop is highly nonlinear and, therefore, rigorous mathematical analysis is not easy. However, certain cures have such as the use of dither signals in the reference inputs to been proposed, stabilize the drifting process, exponential forgetting factors, and dead-zones to prevent the adaptation mechanisms from causing the parameters to drift [14-191. None of the suggested methods, however, seems to hold general validity at the present time.
l TIME AT WHICHINSTABILITY SHOWS UP CAN BE LARGE - BUT SYSTEMIS UNSTABLE ' INSTABILITY TIME INCREASES - AS FREQUENCY SEPARATION OF LOW-FREQUENCY MODELED DYNAMICS AND HIGH-FREQUENCY UNMODELED DYNAMICS INCREASES - AS DISTURBANCE MAGNITUDE DECREASES l CERTAIN CURES MAY WORK - DITHER SIGNALS - EXPONENTIAL FORGETTING - DEAD ZONE NOT CLEAROF GENERAL VALIDITI . ADAPTIVE LOOPHIGHLY NONLINEAR - ANALYSIS NOT EASY FUTURE RESEARCHDIRECTIONS Clearly, a lot more research is needed to understand the complex interplay between time-domain and frequency-domain quantities in an on-line adaptation mechanism.
Key concepts such as finite-time identification need to be understood and developed further; input constraints and disturbance constraints have to be formulated and taken into consideration as part of the overall design along with a more precise mathematical representation of modeling uncertainty, perhaps in terms of a modeling error a:(w) as suggested below. To summarize, maximum allowable system bandwidth should be reflected in any design problem either explicitly or implicitly, and the mathematics of any adaptive algorithm should try not to violate the hard constraints imposed by it.
l EXAMINE FUNDAMENTAL ISSUES - INTEGRATED TIME-DOMAIN AND FREQUENCY-DOMAIN APPROACH l FINITE-TIME IDENTIFICATION PHYSICAL O<t<T -- l INPUT CONSTRAINTS: r(t)e R ' DISTURBANCE CONSTRAINTS: d(t)e D NOMINAL MODEL: p",(s) MODEL ERROR BOUND: ~(wl> Ig(j4-iTWl I NEED II E(u) TO LIMIT BANDWIDTH MYTHS AND REALITIES In conclusion, adaptive control algorithms as they now stand are deceptively simple and promising; they are proven theoretically stable and they improve performance by overcoming the conservativeness that nonadaptive designs typically have to contend with. Unfortunately, however, the advantages that these algorithms enjoy are based on mathematical assumptions that are always violated in practice. Moreover, the "ideal" properties that they seem to possess are very nonrobust to even subtle violations of the underlying assumptions. Consequently, they may result in unstable systems when applied in real engineering problems.
l MYTH: LET US RUSH TO IMPLEMENT ADAPTIVE CONTROLLERS: GOOD PERFORMANCE, PROVEN STABILITY l REALITY: THE MATHEMATICAL ASSUMPTIONS THAT LEAD TO GLOBAL STABILITY PROOFS ALWAYS VIOLATED IN REAL LIFE.
WATCH OUT: INSTABILITY MAY EVENTUALLY SET IN.
1. Valavani, L., "Stability and Convergence of Adaptive Algorithms: A Survey and Some New Results," Proceedings of the Joint Automatic Control Conference, Vol. 1, pp. WA2-B, 1980.
mnopoli, R. V., "Model Reference Adaptive Control with Augmented Error 2.
Signal," IEEE Transactions on Automatic Control, Vol. AC-19, pp. 474- 484, October 1974.
3. Narendra, K. S. and Valavani, L. S., "Stable Adaptive Controller Design- Direct Control," IEEE Transactions on Automatic Control, Vol.
AC-23, pp. 570-583, August 1978.
4. Feuer, A. and Morse, A. S., "Adaptive Control of Single-Input Single-Output Linear Systems," IEEE Transactions on Automatic Control, Vol. AC-23, pp. 557-570, August 1978.
5. Landau, I. D., "An Extension of a Stability Theorem Applicable to Adaptive Control," IEEE Transactions on Automatic Control, Vol. AC-25, ppa 814-817, August 1980.
Narendra K. S., Valavani, L. S. and Lin, Y. H., "Stable Adaptive 6.
Controller Design, Part II: Proof of Stability," IEEE Transactions on Automatic Control, Vol. AC-25, pp. 440-448, June 1980.
7. Landau, I. D. and Silveira, II. M., "A Stability Theorem with Applications to Adaptive Control," IEEE Transactions on Automatic Control, Vol. AC-24, pp. 305-312. April 1979.
8. Astr8m, K. J., Borisson, U., Ljung, L. and B. Wittenmark, "Theory and Applications of Self-Tuning Regulators," Automatica, Vol. 13, No. 5, pp. 457-476, September 1977.
9. Egardt, B., "Unification of Some Continuous-Time Adaptive Control Schemes," IEEE Transactions of Automatic Control, Vol. AC-24, pp.
588-592, August 1979.
Kreisselmeier, G., "On Adaptive State Regulation," IEEE Transactions on 10.
Automatic Control, Vol. AC-27, pp. 3-16, February 1982.
11. Goodwin, G. G., Ramadge, P. J. and Caines, P. E., "Discrete-Time Multivariable Adaptive Control," IEEE Transactions on Automatic Control, Vol AC-25, pp. 449-456, June 1980.
Rohrs, C. E., Adaptive Control in the Presence of Unmodeled Dynamics, 12.
Doctoral Dissertation, Dept. of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, August 1982.
-2- 13. Dumont, G. and Belanger, P. R., "Successful Industrial Application of Advanced Control Theory to a Chemical Process," IEEE Control Systems Magazine, Vol. 1, No. 1, pp. 12-16, March 1980.
14. Krause, J., Adaptive Control in the Presence of High Frequency Modeling Errors, LIDS-TH-1314, M. S. Thesis, Dept. of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, May 1983.
15. Ioannou, P. and Kokotovic, P., "Singular Perturbations and Robust Redesign of Adaptive Control," Paper presented at the IFAC Workshop on Singular Perturbations and Robustness of Control Systems, Lake Ohrid, Yugoslavia, July 1982.
16. Orlicki, D., Valavani, L. and Athans, M., "Comments on Bounded Error Adaptive Control," LIDS-P-1320, Massachusetts Institute of Technology, August 1983.
17. Peterson, B. and Narendra, K., "Bounded Error Adaptive Control," IEEE Transactions on Automatic Control, Vol. AC-27, No. 6, pp. 1161-1168, December 1982.
Anderson, B. D. O., 18. "Exponential Stability of Linear Equations Arising in Adaptive Identification," IEEE Transactions on Automatic Control, Vol. AC-22, No. 1, pp. 83-88;' February 1977.
19. Anderson, B. D. 0. and Johnstone, R. M., "Adaptive Systems and Time- Varying Plants," International Journal of Control, Vol. 37, No. 2, pp. 367-377, February 1983.
SPECIAL REPORT FOREIGN TECHNOLOGYSUMMARY OF FLIGHT CRUCIAL FLIGHT CONTROL SYSTEMS Rediess H. A.
H . R. Textron, Inc.
Irvine, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 INTRODUCTION
A survey of foreign technology in flight crucial flight controls
is being conducted for NASA Langley Research Center as a data base for
planning future research and technology programs. Only Free World
countries were surveyed, and the primary emphasis was on Western Europe
because that is where the most advanced technology resides. The survey
includes major contemporary systems on operational aircraft, R&D flight
programs, advanced aircraft developments, and major research and
technology programs. The information was collected from open
literature, personal communications, ,and a tour of several companies,
government organizations, and research laboratories in the United
Kingdom, France, and the Federal Republic of Germany. This paper
provides a summary of the survey results to date.
Some of the figures are taken from a briefing to the NASA
Administrator by Mr. Kenneth Szalai from Ames Research Center, Dryden
Flight Research Facility, on the technology tour of Europe that
Mr. Szalai and the author conducted in 1983. These figures are used
with the permission of Mr. Szalai.
This survey was conducted under contract NASl-17403, and the
Technical Representative of the Contracting Officer was Mr. Cary
Spitzer, NASA Langley Research Center. The material presented herein
solely represents the findings and opinions of the author and is not to
be construed as being endorsed by the U.S. Government or representa- '
tives of the National Aeronautics and Space Administration.
FLIGHT CONTROL SYSTEMS EVOLUTION
Flight controls technology has undergone tremendous evolution over
the past three decades. Figure 1 illustrates many of the key
milestones in the evolution as represented by major R&D systems and
"nonflight critical fly-by-wire" refers
operational systems. The term
to systems that are commanded by electric or optical signals, but loss of
thode signals is not likely to cause the aircraft to crash. Typically,
there is also a mechanical/hydraulic path for primary control or as a
backup system. Flight critical fly-by-wire means that loss of that
system is likely to cause the aircraft to crash.
aircraft,
Although th; chart emphasizes U.S. several _key develop-
ments in Europe are included, and those of current interest are
discussed subsequently. The Concord had a very profound effect on
It represented the
European flight controls technology in two ways.
first, and as yet only, high authority SAS/CAS in commercial transports
and provided a very important experience base for the UK and France
technologists and managers.
That has helped influence an early
commitment to fly-by-wire in the Airbus. Secondly, it accelerated the
development of the technology in France because the French engineers
gained valuable experience working directly with the Marconi engineers
on the Concord system.
There is now a solid base of DFBWtechnology in
and in Europe and wide-spread commitment to DFBW for military aircraft
some- cases, commercial transports.
FLT. CRITICAL FBW NAR ,D”LE.
.-- .-- 0 AiiiST~dT~ RiC l S-7671A.310 DFCS . DC.940 DFCS I 0 F.18 OFCS 0 CH.53E DFCS NON-FLT. CRITICAL AFCSIFSW DC-10 AUTOPILOT ANALOG l MIRAOE 2000 FLT. CRITICAL FSW ----- -- --- -- l BELL 214 ST ELEVATOR 0 AH.10 TAIL ROTOR 0 CU.47 VALT NON.FLT. CRITICAL 0 C.141 FSW I I FBW Cl 3NCORD 0:: ,4’1 ‘WEAD 0 0.47 F:BV ----- -- --- -- .-- Y- HIOH AUTHORITY OF.15 SASICAS cc >NCORO l 53A .
----- -- --- -- .-- l H.34 LIMITED AUTHORITY SAS l IO4 0 1 , 1900 1990 10 0 FIRST FLIGHT YEAR Figure 1
JA-37 AND AIRBUS A-310
The Swedish JA-37 Viggen fighter aircraft (Figure 2) developed by
The aircraft design
Saab-Scania underwent its first flight in 1974.
features a single-channel high-authority digital automatic flight contra
system (DAFCS) provided by Honeywell and mechanical primary FCS.
Functions provided by the DAFCS include a control augmentation system, and control stick steering), attitude hold (pitch, roll, heading, The aircraft contains altitude hold, and automatic airspeed control.
three primary control surfaces (right/left elevon and rudder) which are
controlled by the pilot via the mechanical PFCS, by the DAFCSvia
secondary series servo, and via automatic or manual parallel and series
trim actuators.
The Airbus A-310 transport, currently in production, was first
flight tested in 1982, and features a mechanical primary flight control
and a digital automatic flight control system.
system, DFBW spoilers,
The spoiler system is dual-channel fail safe with identical active and
monitor channels and uses dissimilar hardware (processors) and
software.
Operational Aircraft
JA-37
Swedish (SAAB-Scania)
First flight 1974
Single-channel full-authority
digital automatic FCS,
mechanical reversion
Airbus A31 0 Multinational (Fr, FRG,
Spain, UK)
First flight 1982
Mech primary controls, DFBW
spoilers, digital autopilot
Figure 2
AIRBUS A-320
Airbus Industries (AI) is in the detailed design stage in the
short/medium range A-320 transport (Figure
development of a 150-seat,
Mechanical
3) featuring a quadruplex DFBWflight control system (FCS).
control rudder and mechanical backup pitch trim are retained to permit
Tests in the Airbus A-300
safe landing in the event of power loss.
flight test aircraft have verified that it is possible to land in this
configuration. The system design includes dissimilar redundancy
in both hardware and software of the same general type used in the A-
310 spoilers. The A-320 will also incorporate relaxed static stability
to at least the neutral point and possibly negative static stability.
A flight test program is underway using the A-300 test bed aircraft
to evaluate the use of RSS and a side-stick controller on the A-320.
The evaluations will determine the engineering operational and
certification issues of such systems on civil aircraft. The engines
will incorporate full-authority digital engine control integrated with
the flight management system.
Operational Aircraft
Under Development
Multinational (Fr, FRG, Airbus A320
Spain, UK)
Detailed design in progress
First flight 1986
Quad DFBW - dissimilar
redundancy hardware
and software
Mech backup on rudder
and pitch trim
ACT: relaxed static stability
Side-stick controller
A-320 DFBWSYSTEM
A more detailed description of the A-320 DFBW system design
All primary flight control surfaces
features is shown in Figure 4.
horizontal stabilizer, ailerons, and roll spoilers) are
(elevators,
quadraplex digital fly-by-wire using dissimilar redundancy in both
The rudder control is mechanical, and the
hardware and software.
tail plane trim has a mechanical backup to provide emergency landing
The secondary controls (slats, trim, speed brakes, and lif
capability.
dumpers) are commanded electrically.
Mechanical control: - rudder Electrical signalling - tail plane trim - slats Fly-by-wire: (alternative control) - trim (yaw, pitch and roll) - elevators - speed brakes/lift dumpers . horizontal stabilizer .
Hydraulic actuation of all surfaces .
poilersllift dumpers Figure 4
AIRBUS INDUSTRIES FLIGHT CONTROL SYSTEMEVOLUTION
Airbus Industries (AI) planned development of a transport family is
illustrated in Figure 5.
Based on a continued, vigorous research and
development program, including full-scale experimental testing,
advanced technologies are progressively introduced in aircraft designs
providing practical, evolutionary changes rather than revolutionary.
The next transport is the 150-seat A-320 described previously. A
a two-
series of wide-body aircraft is in the preliminary design stage:
engine short/medium range TA-9; a four-engine long-range TA-11; and, a
two-engine medium/long range TA-12. The first two of these, TA-9 and
TA-11, are expected to incorporate full-authority digital fly-by-wire systems on all surfaces, extensive use of active controls, and reduced energy systems.
1st Flight 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 Continuing R&D Proaram l Mechanical . DFBW spoilers redundancy hardware 8 software l Full DFBW l Mechanical l Extensive back-up in active controls rudder & l Reduced pitch trim energy l Side stick systems control Figure 5
TORNADO AND MIRAGE
Figure 6 illustrates basic characteristics of the multinational
Tornado and French Mirage 2000/4000 aircraft currently in production.
The Tornado, a joint UK, FRG, and Italian project, underwent its
first flight in 1976. The flight control system includes both analog
and digital computing. The primary flight control function is
performed by a command/stability augmentation system (CSAS) which is a
triplex analog FBW maneuver demand system (Ref. 1). While no
mechanical revision is provided for the rudder and spoilers, it is
retained for the ailerons for safe return upon loss of CSAS computing.
A dual digital autopilot/flight director (AFDS) integrated with the
CSAS provides outer loop control. The AFDS uses cross-comparison
techniques for failure detection and a signal consolidation scheme to
It also provides a fail
provide triplex commands to the CSAS.
operational flight director capability to enable the pilot to monitor the
autopilot performance and fly the aircraft manually if the autopilot
malfunctions.
The first flights of the Dassault-Brequet Mirage 2000 and 4000
were conducted in 1978 and 1979, respectively. With no mechanical
revision capability, both include a flight-critical analog FBW flight
control system with digital autopilot. The 2000N version is nuclear
hardened fitted with terrain-following radar.
The Mirage 4000 features
relaxed static stability and automatic variable camber to optimize
performance.
Operational Aircraft
ornado l Multinational (UK, FRG, Italy) l First flight 1976 ya l Analog CSAS, dual digital autopilot, mechanical reversion
I
Mirage 2000/4000 l French (Dassault-Breguet) I l First flight 1978,1979 l Analog FBW, digital autopilot, no mechanical reversion l Relaxed static stability/auto variable camber (4000)
Figure 6
JAS-39 AND LAVI
Flight critical DFBWflight control systems designs are under
development for opertional fighter aircraft in both Sweden and Israel
(Fig. 7). Saab-Scania of Sweden is developing the JAS-39 Gripen
advanced strike fighter. Lear Siegler, Inc., will design, develop,
and manufacture the flight control system. Under subcontracts, Moog
Aerospace in cooperation with Saab Combitech will design the primary
flight actuators, and Lucas Aerospace will supply the maneuvering-flap
control actuation system. The JAS-39 will be a flight-critical triplex
DFBWsystem and, thus, contains no mechanical backup capability. The
fighter, scheduled for first flight in 1987, is being developed for
specific mission needs of Sweden and may not favorably compete for an
international market.
The Israeli Aircraft Industries is developing the LAVI tactical
fighter to replace the A-4 and Kfir C2 aircraft with first flight
scheduled for 1986. The flight controls, to be designed by Lear
Siegler (Moog), will be a digital fly-by-wire system with relaxed
static stability and include an analog but no mechanical backup system.
Advanced digital avionics systems will be incorporated to operate with
interactive multifunction displays/controls, fire control integrated
with internal and external sensors, and enhanced active/passive self-
defensive systems. As planned, much of the design and systems would be
supplied by U.S. companies.
Operational Aircraft
Under Development JAS-39 Swedish (SAAB-Scania) First flight 1987 Triplex DFBW (Lear Siegler), no mechanical backup LAVI Israel (IAI) First flight 1986 Triplex DFBW (Lear Siegler), analog backup, no mechani- cal reversion ACT: relaxed static stability Figure 7
SKYSHIP 600
In the UK, Marconi Avionics, under contract to Airship Industries, is
developing a digital fly-by-light (DFBL) flight control system for
High inherent immunity
application to the Skyship 600 (see Figure 8).
to EM interference is achieved by a 1553 optical data bus between the
FCC and the actuator drive system (ADS) and by providing dedicated
electrical power at the ADS from a hydraulically driven electronic
generator. The ADS includes a microprocessor to locally handle the
failure detection and isolation. The actuators are duplex electric
incorporating two samarium cobalt DC servomotors mounted on a common
Torque is supplied by only one shaft, each fed by separate power.
the second is activated after failure of the first.
motor;
Operational Airship
Under Development
l UK (Airship Industries, Marconi)
Skyship 600
First flight 1983 (with DFBL)
Digital fly-by-light (DFBL)
All four tail surfaces
Active/standby with pilot select
Microprocessor-based
FCS computer
1553 optical data bus
Figure 8
TIAWAN F-104
The Aeronautical Research Laboratories of the Aeronautical Industry
Development Center (AIDC) of the Republic of China in Tiawan has
initiated a program to develop a modern digital flight control system
The system will be a half-
to upgrade 100 F-104 aircraft (Fig. 9).
authority dual digital command augmentation system (CAS) and stability
The existing
augmentation system (SAS) for pitch, roll, and yaw.
mechanical system and a new direct electrical command system will
provide emergency backup capability. The prototype development
contract for five aircraft systems is now under competition, and the
first flight is expected to be early in 1987.
Operational Aircraft
FCS Upgrade
F-l 04
Republic of China-Taiwan (AIDC)
FCS under competition
First flight 1987
Dual digital CAWSAS, mechanical
and direct electrical backups
Figure 9
F-104 CCV AND JAGUARDFBW
Among the European R&D flight programs are the German F-104CCV and
the United Kindom's Jaguar DFBWaircraft shown in Figure 10. The
purpose of the German demonstration program was to investigate
stability and control characteristics of a supersonic aircraft (Ref.
A single-seat F104G was modified as a control-configured vehicle
2) l
(CCV) with a newly developed full-authority quadruplex system while
retaining the original system as a mechanical backup. After initial
flights starting in December 1977 to evaluate the DFBWsystem, various
degrees of destabilization were achieved by adding aft ballast and a
canard. The highest instability reached in normal flight was up to 22%
mean aerodynamic chord at an angle of attack of 11 degrees. The flight
tests were highly successful in demonstrating aircraft controllability
in a highly unstable configuration.
The Jaguar program was initiated to demonstrate a safe, practical,
full-authority DFBWflight control system. This activity is of
interest since it represents the first pure digital fly-by-wire system
The program was initiated in 1977 under with no dissimilar backup.
the technical sponsorship of the RAE and under contract to British
Aerospace. Marconi Avionics furnished the flight control system. While
more descriptions will follow, basically the FSC is a full-authority
quadruplex DFBWsystem with optically coupled data transmission. The
initial flight of aircraft was conducted in October 1981.
R 81 D Flight Programs
German (MBB)
F-104 CCV
First flight 1977
Quad DFBW, full-authority, mechanical
reversion
Relaxed static stability
Jaguar
UK (RAEIBAe, Marconi)
First flight 1981
Quad DFBW, no mechanical reversion
Optical interchannel data links
Figure 10
JAGUARDFBWSYSTEMARCHITECTURE
The overall system architecture is shown schematically in Figure
11 (Ref. 3). Quadruplex computers and primary sensors were used to
satisfy specifications requiring survival of any two electrical
failures in the system and reliance on majority voting rather than self
Sensors of lower redundancy
monitoring within each redundant element.
were used for those functions not necessary for safety of flight. A
sextuplex or duo-triplex first-stage actuation scheme was selected to
The two additional conform with stringent redundancy specifications.
actuator channels are driven by the actuator drive and monitor
computers which were independently voted versions of the FCC's outputs.
Comprehensive built-in-test features were included to measure the
system functional characteristics. While designed to run
synchronously, the system has been operated asynchronously for
continued periods without observable degradation.
J- L------J Figure 11
JAGUARDFBW- COMPUTING AND MONITORINGARCHITECTURE
The basic system computing and monitoring architecture is presented
in Figure 12 which illustrates a simplified primary control path (Ref.
Quadruplex primary sensors, those necessary for flight safety, 4) l are interfaced with four identical flight control-computers (FCC) which
process these as well as less critical sensor signals into commands for
control of the actuators. Cross-channel data transmission is achieved
by optically coupled serial data links. This scheme enables each
computer to carry out bit for bit identical control law implementation.
Voting and failure rejection logic contained in each computer satisfies
the requirement for surviving two sequential failures of all critical
sensors. The actuation architecture required six independent servo
drive signals. To avoid the cost and complexity of a full six-channel
system, the four FCC's were augmented by dual analog actuator drive and
monitor computers (ADMC) which utilize independently voted versions of
the FCC output signals to drive the additional two channels. Failed
FCC channels are detected and latched out, and then the ADMC averages
the remaining good FCC channels. These additional channels are
mechanized to eliminate any interchannel failure propagation between
the six parallel redundant output interfaces.
Flight Control Computers I 1’ I .
Servo Input Control input -- sensor - law a sensors - voting computing drives . , .
-I
Servo ’ xl Actuator Drive & Monitors c 1 Figure 12
JAGUARDFBW-DUO-TRIPLEX ACTUATOR SCHEME
The basic specifications requiring that first-stage actuation has
only two independent hydraulic supplies with no interconnect and that
the system survives a hydraulic failure followed by an electrical system
failure or the converse, led to the selection of duo-triplex first-
stage actuation system design. While a quadruplex configuration would
have offered an attractive one-to-one interface with the FCC's,
designers were concerned with mechanizing some form of fast reaction
actuator monitoring and channel isolation scheme to prevent
uncontrolled surface movement in the event of an electrical followed by
a hydraulic failure.
Each of the five control surface actuation
systems is similar, and Figure 13 illustrates the operation (Ref. 4).
Each system contains six servovalves. An interactuator mechanical
link assures that the spools move uniformly, which effectively sums the
six servovalve outputs.
Thus, failures in two channels are overridden
by the other four. A separate hydraulic supply feeds each trio of
servovalves and is also routed to the corresponding jack of the
conventional tandem power control unit. A hydraulic supply failure is
absorbed because the three associated servovalves are unable to oppose
the correctly operating channels.
Figure 13
JAGUARDFBW- SOFTWARE DEVELOPMENT PROCESS
The use of common software in the flight control system presented
the potential of a generic error leading to a safety critical loss of
control. Therefore it was necessary to provide maximum software
visibility to facilitate thorough testing and functional auditing
supplemented by clear requirements definition, during the design phase,
detailed documentation, and stringent production and configuration
control procedures (Ref. 5). The key documents controlling the software
design are the System Requirements Document (SRD), which controls the
and the Software Structure Development (SSD).
design implementation,
The SSD defines the running order of the modules within each program
segment and is designed to assure strict sequential data flow.
The overall software development process is depicted in Figure 14.
The SRD's are interpreted to produce software module design specifi-
cations which in turn are used for module coding. A module test
specification is written by an independent programmer to minimize error
The module code is tested, and the results are documented.
carry-over.
Senior programmers audit all module documentation to assure that the
design requirements are satisfied, the design rules observed, and the
test process followed. When the module coding is completed, the modules
are assembled and loaded into the hardware for integration tests. All
of the software documentation is subjected to strict configuration
control with changes authorized only through a formal change request
process.
Software requirements I System analysis I I I I- I Module testing assembler Program Module t
1 chqcks
estimates code I t I Configuration Assemble . F control modules 4 checks I Loader verification t into integration hardware tests Figure 14
JAGUARDFBW- SYSTEMINTEGRITY APPRAISAL
The basic integrity of the system was achieved by the selection of
the. system architecture in conjunction with standard design practices,
performance testing, and assessments of operationai/safety
considerations. For the Jaquar DFBWprogram, these procedures were
extended to include an integrity appraisal or system audit as outlined
in Figure 15. The main elements (Ref. 5) were:
100% coverage single-fault FMEA
Multiple-fault FMEA for specific combinations
Flight resident software integrity appraisal
Appraisal of specific functions
Configuration inspection
Qualification program
Burn-in program
These were supplemented by secondary analyses- shown below the main
elements in the figure. As part of the integrity appraisal, various
functions and features of the system were subjected to technical
evaluations as required from results of mainstream failure mode and
effects analysis (FMEA) and/or engineering findings. While the
appraisal was conducted by a team knowledgeable in the specific design, they reported to senior engineers.
System integrity Appraisal of Software specific integrity - appraisal functions Configuration inspection - Module, Voter, Microprogram System connector architecture monitor integrity chassis and appraisal appraisal appraisal LRU FMEA Qualification testing and - inspection BITE coverage analysis I Burn-in
I
Reliability analysis I J Figure 15
JAGUARDFBW- SYSTEMQUALIFICATION PROCESS
Once the design objectives had been specified as subsystem or
elements, such as the FCS, it was necessary to integrate these elements
into a functional system exhibiting the characteristics of the basic
design requirement while assuring that no adverse intersystem reactions
were present and verifying that the common software used contained no
generic or other design defects. These tasks were conducted using a
ground test rig, the aircraft, and an independent software audit, interrelated as shown in Figure 16 (Ref. 5).
The ground test rig was used to (1) verify the control laws by
pilot assessment, (2) integrate the hardware, software, and ancillary
equipment, (3) validate the final software before flight, and (4) gain
overall system confidence. In addition, it served as a pilot training
aid and as a preflight test bed.
The aircraft ground tests included complete checkout and test of
the installed flight control system, electro-magnetic compatibility
testing, aircraft systems testing, and simulated lightning tests.
It was considered essential that an independent software test by a
disinterested group be used to supplement the rig and aircraft tests.
The group was responsible for emulation of the flight control computer
using a general purpose machine and for manual code analysis.
System requirement -------- r- Hardware tests 1 requirement ] i
i
J 4
Model for use Design I on system rig &test I Independent -I Machine b software code I check I I I + i + I Firmware i Test on systems rig in computer 7 I I Figure 16
AUGUSTA A-129 AND T-2 CCV
Flight test programs are being conducted by both Italy and Japan
(Figure 17).
The Italian Augusta A-129 helicopter has been undergoing tests, and
five prototype vehicles are to be manufactured in anticipation of
production. The A-129 features a DFBW tail rotor (nonflight critical)
The design includes a digital
but retains other mechanical controls.
autopilot and an integrated multiplexing system using microprocessors
for aircraft/firing systems control.
Under contract to the Japanese Defense Agency, Mitsubishi has built
a control-configured vehicle version of the T-2 advanced trainer for
use as a research aircraft. The T-2 CCV has composite all-flying
canards located on the inlets ahead of the wing leading edge and a
composite ventral fin located on the fuselage center line. The flight
control system is triplex digital with mechanical backup. The first
flight was conducted in August 1983, and the aircraft is scheduled for a
two-year experimental flight test program by the Japanese Air Self
Defense Force.
R & D Flight Programs
Agusta A-l 29 l Italy l First flight 1983 l DFBW tail rotor, digital autopilot, mechanical rotor controls l Multiplex data bus/integrated avionics/ flight control T2 CCV l Japan (Mitsubishi) l First flight 1983 l Triplex DFBW, mechanical reversion l All moving canard/RSS Figure 17
ADVANCED AIRCRAFT DEVELOPMENT
Britain and France are currently competing for leadership of a new
generation European combat aircraft advance development program which
may lead to a joint development with Germany for 1990's fighter
aircraft (Fig. 18).
Led by British Aerospace, a seven-member industrial consortium
has an agreement with the British Ministry of Defense for government
funding up to and including first flight for the development of an
Agile Combat Aircraft (ACA) technology demonstrator called an
experimental aircraft program (EAP). Both West German and Italian
aerospace companies have contributed some funding, and while not
committed, the West German and possibly the Italian governments may
fund a second demonstrator aircraft. The ACA flight control system
design by Marconi Avionics would be quadruplex digital fly-by-wire
(DFBW) with no mechanical backup and no dissimilar redundancy.
(Marconi considers that, while not required for military application,
dissimilar redundancy is necessary in commercial aircraft for
certification purposes.)
France has a comparable program since Dassault-Berguet (D-B) has
begun manufacture of one technology demonstration aircraft, Avion de
Combat Experimental (ACX). It will be a DFBWdesign with no mechanical
backup and include electrical and fiber optics data busing, voice
control system, holographic displays, and provision for antiturbulence
ride control in the automtic computer-controlled flight control system.
The Federal Republic of Germany has need for a fast-reaction
fighter, and their special requirements are prompting them to consider
an entirely new fighter airframe called the TKF-90 which would employ
existing avionics technologies to minimize costs. Two German
companies, Messerschmett-Boelkow-Blohm (MBB) and Dornier, are pursuing
test programs to satisfy the German Air Force needs but neither is
committing to a flying demonstrator.
MBB is using a modified Saab
Viggen as a test bed to investigate various performance envelopes and
is testing vectoring nozzle canards and other advanced control
features. MBB has considerable experimental background in fire and
flight control systems resulting from their F-104 CCV test bed.
Dornier in conjunction with Northrop has an ND-102 design and is
using a modified Alpha jet to test a new transonic wing and to experi-
ment with direct side force controls and maneuvering flaps/slats.
Thus,
at the time of this writing, there are three fighter design
plans among the UK, France, and FRG. Considering the economics
involved, it is likely that more than one or possibly two will be fully
Both the UK and France seek partnership with the FRG, and it
developed.
is likely that at least the TFK-90 will be a compromise between the ACA
and ACX programs. MBB of West Germany is, in fact, a partner on the
ACA and is being actively pursued as a partner in France's ACX program.
MBB could become the catalyst for a European-wide project with British
While Aerospace and Dassault-Berquet and themselves as principals.
specific mission requirements would be compromised, such a triumvirate
would create an attractive production market and serve as a formidable
obstacle for U.S. competitors.
Italy (Aeritalia-Macchi) and Brazil (Embraer) are jointly
developing the AMX fighter with first flight scheduled for 1984 and
delivery in 1987. Initial production is expected to provide 1.85
The electronic flight control
aircraft for Italy and 80 for Brazil.
system designed by Marconi provides duplex analog fly-by-wire control
of the tail plane, spoilers,
and rudder together with mechanical
elevators and ailerons. The design also incorporates automatic pitch,
roll, and yaw stabilization. The equipment comprises two dual-
redundant flight control computers based on 16-bit microprocessors
To optimize
organized for specially developed fail-safe software.
hardware requirements, analog computing is used for the actuator
control loops, pilot command path, and rate damping computations.
Digital computing is used to handle gain schedules, electronics trim,
and airbrake integrators. System performance is monitored by redundant
processors in the flight control computers.
Advanced Aircraft Development
UK/FRG (BAe, MBB, Marconi) First flight 1986 Quad DFBW, no mechanical reversion, no dissimilar redundancy Fr, FRG (Dassault-Breguet, MBB?)
First flight 1986 DFBW, no mechanical reversion, fiber optic data bus Ride control, voice command, holographic HUD Italy, Brazil (Aeritalia/Marcchi, Embraer, Marconi) First flight 1984 Duplex analog FBW, digital gain sched/monitoring (tailplane, spoilers, rudder) Figure 18 REFERENCES The Development of Multiple Redundant Flight 1. Corney, J. M.: Aeronautical Control Systems for High Integrity Applications.
Journal, vol. 84, no. 837, Oct. 1980, pp. 327-338.
Korte, U.: Flight Test Experience with a Digital Integrated 2.
AGARD Guidance and Control System in a CCV Fighter Aircraft.
CP-321, Lisbon, Portugal, October 1982, pp. 22-l to 22-11.
Ground Flight Testing on the Fly-by-Wire 3. Smith, T. D., et al.: Jaguar Equipped with a Full-Time Quadraplex Digital Integrated AGARD-CP-321, Lisbon, Portugal, Flight Control System.
October 1982, pp. 23-l to 23-20.
4. Marshall, R. E. W., et al.: Jaguar Fly-by-Wire Demonstrator Paper presented at the Integrated Flight Control System.
1981 Advanced Flight Control Symposium, Colorado Springs, CO, August 5-7, 1981.
and Smith R. B: Flight Clearance of the Jaguar 5. Daley, E.
Fly-by-Wire Aircraft. Proceedings of the Royal Aeronautical Society Symposium, Certification of Avionic Systems, April 27, 1982.
SESSION IV RECENT EXPERIENCES IN IMPLEMENTATION OF ADVANCED CONTROL SYSTEMS William E. Howell Session Chairman FUNCTIONAL INTEGRATION OF VERTICAL FLIGHT PATH AND SPEED CONTROLUSING ENERGY PRINCIPLES A. A. Lambregts Boeing Commercial Airplane Company Seattle, Washington First Annual.NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT A generalized automatic flight control system has been developed which integrates all longitudinal flight path and speed control functions previously provided by a pitch autopilot and autothrottle. In this design, a net thrust command is computed based on total energy demand arising from both flight path and speed targets. The elevator command Is computed based on the energy distribution error between flight path and speed. The engine control is configured to produce the commanded net thrust. The design incorporates control strategies and hierarchy to deal systematically and effectively with all aircraft operational requirements, control nonlinearities, and performance limits. Consistent decoupled maneuver control is achieved for all modes and flight conditions without outer loop gain schedules, control law submodes, or control function duplication.
STATE-OF-ART AUTOPILOT/AUTOTHROTTLE Virtually every automatic flight control system (AFCS) in use today has been designed using a single-input, single-output control strategy. Although the elevator control loop may be closed on either speed or flight path, the autopilot path control modes have become the most widely used. Automatic speed control using the throttles was then developed as a natural complement to the autopilot path control modes.
Unfortunately, this historic system evolution has not resulted in optimum AFCS capabilitSes and performance. For certain flight conditions, the autopilot path control at constant thrust produces speed instabilities, while autothrottle speed control at constant elevator results in path instability.
Together, they can provide stable flight path and speed control. However, the operation is far from coordinated teamwork. Generally, the autopilot is designed first and satisfactory path control using the powerful elevator is often obtained at the expense of speed control. This leaves the autothrottle
in a no-win situation: poor speed control with acceptable throttle activity
or acceptable speed control with objectionable throttle activity.
For large autopilot path change commands, the conventional control strategy breaks down because thrust will limit, causing loss of autothrottle speed control, with the ensuing risk of stalling or overspeeding the aircraft.
l Elevator and throttle each control ----------1 one variable.
Autothrottle i I I I I r 1 J+o+&q~, L ---------- Airplane r-----------
!
Elevator I
I Se
I b Control I * I I L I I I !
.
I I Autopilot 1 :------------~ LIMITATIONS OF THE TRADITIONAL DESIGN APPROACH The traditional single-input, single-output design approach has a number of other fundamental design limitations. Modes that have not specifically been designed to work together are generally incompatible. This has led to a proliferation of specialized control modes of limited use.
Still, pilots complain that the operation of the present generation automatic control systems is often contrary to the way the pilot uses the controls. The lack of short-term coordination of elevator and throttle commands results in undesirable and inefficient controller activity.
This is especially true for the autothrottle system in energy-management-type situations.
Clearly, using the throttles to control speed is not an ideal control strategy. H. A. Soul6 summed it all up in the title of his article (ref. 1), "The Throttles Control Speed, Right? Wrong!"
In the present designs, the thrust control loop is subject to extreme variations in loop qain and dynamics due to variations in aircraft weight and engine characteristics for different conditions. Poor system robustness causes wide performance variations.
The fragmented bottom-up design approach makes it difficult to efficiently implement general design requirements, such as limiting of speeds and maneuver rates, because provisions have to be made in many places.
Research Background
Traditional Single-Input - Single-Output
Design Approach Has Inherent Deficiencies:
l Each mode combination is new problem
l Performance of one mode depends on other engaged mode
l Deficiency of one mode often basis for design of another
l Lack of control coordination leads to undesirable coupling
l Control coupling limits achievable damping
l Difficult to deal effectively with performance/safety limits
l Many difficult and, iterative development programs have not
resulted in fundamental design improvements
FUNCTIONAL OVERLAP The numerous efforts to overcome the limitations of the conventional autopilot and autothrottle designs have contributed much to the overall AFCS complexity, but little to the improvement in the fundamental design methodology. This is illustrated in this figure showing the functional Basic flight path and speed control overlap in the latest generation AFCS.
This results in capability now exists in all three computer systems!
unnecessary pilot workload, the need for numerous operational ambiguities, sensors, computers, interfaces, elaborate configuration control, and high cost Obviously, the underlying system architecture is not ideal.
of ownership.
A top-down system synthesis capability, using an efficient control strategy providing maneuver decoupling and the correct control priorities, has been missing in the classical design methodology. Also, design tools using modern control theory remain inadequate for synthesis of multi-mode/ multi-flight condition systems that can be implemented readily with today's hardware and software technology.
Typical Control Function Overlap, Conventional AFCS
Flight Control Computer l Autoland / l Vertical Flight Autothrottle Performance Computer l Thrust Limiting l Flare Retard l Navigation l Performance l Path Definition
Y /
DESIGN OBJECTIVES FOR THE INTEGRATED SYSTEM The limitations of the traditional autopilot and autothrottle design have During the NASA Terminal Configured Vehicle been encountered over and over.
Program guidance and control experiments using the NASA B-737 aircraft, which was equipped with an advanced autopilot and autothrottle, it was concluded that fundamental system improvements could only be obtatned by a multi-input, As a result, NASA funded research work at multi-output design approach.
Boeing during the 1979-1981 period to develop a fully integrated vertical flight path and speed control concept.
The primary objective was to devise a methodology for the design of a largely generic elevator and thrust connnand computation algorithm, providing decoupled flight path and speed maneuver control for any required control A pilot-like control strategy, including energy management modes.
considerations, needed to be developed to achieve effective elevator and throttle coordination, along with an appropriate hierarchy of control modes to deal with thrust limiting and provide complete protection against stall and overspeed. A more robust system design was desired to reduce sensitivity to engine characteristics. Further design objectives are noted below.
Integrated Vertical Flight Path and Speed
Control Design Objectives
l Improve operational capabilities and performance
- Decoupled maneuver control
- Complete stall and overspeed protection
- Normal acceleration limiting
- Simpler and more effective mode control
- Consistent operation in all modes
l Reduce system complexity, weight, volume
- Eliminate functional overlap
- Integrate control laws, minimize software
- Reduce hardware
l Reduce cost
- Development and certification
- Modifications and recertification
- Maintenance
WGN APPROACH The NASA-sponsored conceptual research work resulted in one very promising generalized design approach based largely on point mass energy control considerations. The basic longitudinal point mass airplane equation of ) (shown below), indicates that the motion, solved for thrust required (TRE by thrust, while at constant aircraftls energy rate is mainly contra 9 led thrust (Es-O), the elevator control results in equal and opposite responses of flight path angle and longitudinal acceleration (short-term D-constant).
The elevator control is, in essence, energy conservative.
It follows then, that from-an energy management point of view, ener gy and the the thrust should be used to control the total elevator to control the desired energy distribution between ihe flight-path angl.eeand acceleration, or altitude and speed.
Based on Energy Considerations
Potential flight path angle yr, Equations: Trequired = weight *‘(SIN7 + V/g,’ + drag L J \ I Total energy rate Tlevel flight ay = _ a\i/g For T = constant: -
ah a6 (short term)
e Therefore: - Thrust used to control total energy - Elevator used to control energy distribution Result: Total Energy Control System (TECS) - General control strategy for all modes - Normalized, airplane independent control law - Commonality with proven 7yp display - Very simple, effective design GENERIC TOTAL ENERGY AND ENERGY DISTRIBUTION CONTROL LAW The previous considerations were used to synthesize a generic flight path The total energy error of angle and longitudinal acceleration control law.
the aircraft is controlled by a thrust command proportional to the integral of the sum of flight path angle error y,.and normalized acceleration error V,/g, relative to the targets yc and Vc/g: = YTI upYE) ATC S Likewise, the energy distribution error is controlled by commanding a change in flight path angle and in turn an elevator change proportional to the integral of the difference of V,/g and yc: = KIL.KEI (;/g-y,) "e C S In addition, proportional terms of (y+$/g) and (i/g-y) are used to achieve the desired control damping. The elevator command further includes pitch control damping terms that stabilize the short-period dynamics, while the specific net thrust command is scaled in proportion to aircraft weight to form the total net thrus.1 command.
Natural decoupling of flight path and longitudinal acceleration maneuvers is achieved by feeding forward either command to both controllers and by selection of the proper relative gains.
Basic Total Energy Control System
Airplane Airplane I Dependent Independent ml- Design ! Weight Fesign IJ ’ ’
Yc i t, ~~l~Engine~---T
, x 9 j \ Net Thrust Command ;;Gdjc
, , l Normalized Net Thrust Command Pitch Damping (T,,,) . (; - Cont) i IMPLEMENTATION OF FLIGHT PATH AND SPEED MODES Using this generalized design approach, all needed flight path and speed control modes of the AFCS can be provided without replication of functions and with the proper hierarchy. The resulting system is called "total energy control system" (TEcS). For the outer loop flight path and speed control modes, the altitude and. speed errors are simply normalized to form the yc and Vc signals.
The control law calls for an exponential reduction of the altitude and speed errors with a time constant T inversely proportional to Kh, KV. Thus, to preserve decoupling for simultaneous altitude and speed maneuvers and to maintain the correct relative energy relationship between altitude and speed errors, Kh = KV must be chosen.
All flight path modes ultimately produce a flight path angle comand All speed control modes produce a V,/g. The minimum and maximum yc * speed limiting modes VHIN and VMAX enter downstream of all other speed modes and by autonomous engage logic provide stall and overspeed protection at all times. A more detailed system description is presented in reference 2.
TECS Architecture and Mode Hierarchy
+Tc Generalized Thrust and Elevator Command Computation h/g USE OF ENERGY COMMAND RATE LIMITS The total energy control system meets all functional, operational, and performance requirements of the overall automatic flight control system.
General system requirements are met by generic design solutions, implemented at strategic points in the signal processing, so they apply to all intended modes and do not have to be replicated.
Vertical Acceleration Limiting - Normal acceleration limiting has been implemented at one central point in the control law to cover. all. modes except GO-AROUNDand FLARE, by simply rate limiting the yc, since h = Vy.
This time history illustrates a case where the vertical acceleration is limited to 3.2 ft/sec2 (0.1 CJ) during the executing of simultaneous commands Analogous to the rate limit in the yc in the ALTITUDE and CAS modes.
path, an equivalent rate limit is provided in the V,/g signal path.
This rate limit provides vertical acceleration limiting during s.peed control when thrust is limited, reduces controller activity induced by turbulence, and results in smooth thrust buildup for large speed maneuvers.
Enerqy Exchange Maneuvers - The equal (energy) rate limits on yc and V,/g assure that for dual opposing commands, the system first exchanges energy to the maximum extent possible before commanding thrust to satisfy the required final total energy state. Thus, the system can handle any energy management problem efficiently and avoids unnecessary thrust applications, as the time response shows.
ALTICAS Modes, Combined
Descend/Acceleration
B-737 TECS ALTCMD 10 000 10 000 ALT 9600 9600 fi 9200 9200 ft 270 CAS CASCMD 270 VERT 4 50 HDOT ACCEL 0 0 ttlsec2 -4 -50 ws THROTTLE 55.0 6 ELEV 30.0 2 deg 5.0 -2 deg Negligible Throttle Response! 1 Time, set SPEED CONTROL PRIORITY In the past, there have been numerous incidents where thrust limiting due to large autopilot path commands caused loss of speed control, resulting in The TECS design solves this problem elegantly airplane stall or overspeed.
and without adding control laws or complex system reconfiguration.
The flight path angle error crossfeed to the elevator command computation is switched out, and the thrust command integrator is put in HOLD when the flight path command causes the thrust limit to be reached. In this configuration, the speed control is continued through the elevator, and the speed response dynamics remain the-same as in the linear case because a change in flight path angle Ay = -V,/g is commanded that provides the same acceleration due to the gravity component along the flight path as the engine provides in the linear case. This figure shows the response to an altitude change command, causing the thrust to go to idle, while speed is being maintained through the elevator. Subsequently, the command to reduce speed is executed by temporarily changing the flight path angle, until the speed is captured. Finally, linear control is resumed when the altitude target is approached, and a thrust rate command is developed to drive the throttles out of the aft limit. During the speed change maneuver at idle thrust, the deceleration command was limited to the available total energy rate, represented by ytV/g, thereby preventing the flight path angle from going positive temporarily. Alternatively, the available total energy rate can be shared to provide simultaneous execution of flight path and speed commands at reduced rates.
ALTXAS Modes, Descent With Subsequent
Deceleration
(100% Deceleration B-737 TECS Priority) ALT 16 000 16 000 ALT CMD 13 000 13 000 fl 10000 10 000 ft LDescent Rate Reduced to Decelerate CASCMD 310 ~~ 310 CAS . .. ..1........ . . . . . . . . . .“.
l . . .
280 ------ --~~~ ‘C. .- .- 280 -..
kn 250 -2250 kn VERT 4 50 HDOT ACCEL ftld -4 -50 ftls THROTTLE 55.0 8 ELEV b2 deg PATH CONTROL PRIORITY Similar to the provisions to deal with thrust limiting due to large flight path commands, the TECS includes provisions to allow continued path control when the thrust limits due to large speed commands. This figure shows the system's capability to decelerate at idle thrust and capture the glide slope without overshoot while still at high speed. The flight-path-angle-based control provides inherent system adaptation to speed, resulting in consistent path captures and allowing error-free path tracking while changing speed.
In this case, the speed command was reduced to 100 kn before capture of the glide slope. However, the VMIN mode limits the speed reduction to -1.3 VSTALL at each flap setting. As the flaps are lowered beyond 15", the throttles advance to arrest the deceleration rate, and they stabilize the airspeed when the final flap setting is reached.
High-Speed Glide Slope Capture With Flap
Extension and VM~N Control
737 TECS GSE 0.70 0 GAMMA 0.00 -2 deg .0.70 -4 deg ALT 2000 200 CAS ft 0 100 kn FLAP P 40 Down GEAR deg 0 UP THROTTLE 55.0 6 ELEV 30.0 deg 5.0 ’ -2 deg 20 40 60 60 100 120 140 160 180 :*200 *..
Time, set GO-AROUNDPERFORMANCE The GO-AROUNDmode is implemented by simply switching to a fixed 10" flight path command (one line of software). Upon engagement of the mode, the commanded flight path angle causes a coordinated rapid throttle advancement and pitch-up. After the throttles reach the limit, the speed control reduces the pitch rate and normal acceleration to zero. The altitude loss after mode engagement is 31 ft for this condition. As shown, approximately a 0.5-g peak incre- mental load factor is achieved.
This can easily be adjusted, as desired.
Speed is maintained within -1 kn, and the entire maneuver is very smooth. The attained steady-state climb rate and pitch attitude depend on the aircraft thrust and drag condition.
For a heavyweight aircraft, or low available thrust, the 10" flight path angle cannot be achieved. In that case, a limit thrust climb with speed controlled through the elevator will result. For light weight, high available thrust,orlow drag conditions, theyc = loo is achieved with only partial power. The partial power avoids excessive climb gradients and pitch attitudes.
The system is designed to safely handle engine failures and aircraft configuration changes. In case of flap retraction without a corresponding increase in the commanded speed, the VMTN control takes over to maintain . .-.
-lo3 "STALL' Go-Around Mode - Engage at 100 ft;
Weight, 560 000 lb; Flaps 30”; Altitude loss 31 ft
B-747 TECS ALT 2000 151 CAS 1000 147 ft 0 143 kn VERT 5 15 THETA ACCEL -5 5 ftlsec2 -15 -5 deg HDOT 50 15 GAMMA 10 5 ftlsec -30 -5 deg THROTTLE 80 5 ELEV 40 -5 deg 0 -15 deg 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00 Time, set COMPUTER-AUGMENTED MANUAL CONTROL The total energy COrltrOl system is well suited for providing flight path angle based computer-augmented control because of the system's decoupled control characteristics. In this mode, also called Velocity Vector Control Wheel Steering (VCWS). a rate of change of flight path angle command (yc) proportional to the column force is developed. The column force signal is gain scheduled with l/V to provide constant stickforce per "g", and integrated to develop the flight path angle command yc. An inertial flight path angle feedback yI = h/VGROUNU iS used t0 develop the basic error Signal yE, and provide long-term control relative to an inertial reference. This yc signal is also fed forward to the thrust and elevator comand to obtain the desired augmented response lag of y relative to yc without causing speed perturbations.
The time history shows responses tailored to yield xy = 2.5 set, the fastest response achievable with parallel servos while avoiding elevator over- control and stickforce reversal. It should be noted that this ~~ is too long for closure of the primary pilot loop using y display. To overcome this problem, the yc is displayed along with y. The yc had been shown on the NASA I3737 to be a satisfactory primary pilot display for the flight path angle CWS mode (ref. 3). This mode can be used on various transport aircraft to obtain virtually identical and optimized handling characteristics.
Velocity CWSResponses to 12-lb Column Pull/Push
6747 TECS Flaps 30° Gear Down + 12 lb Pull/Push 1 GAMMA GAMCMD 0.10 0.00 deg -0.10 -1 deg 8 VERT COLUMN 20 FORCE ACCEL -8 ftls2 lb -20 CAS 174 6000 ALT 170 5000 4000 ft kn 166 4 ELEV THROTTLE 80 -4 deg deg 0 0 6 12 18 24 30 36 42 48 54 60 Time. set CONTROL IN WIND SHEAR AND TURBULENCE To optimize the system's performance in wind shear and turbulence, numerous trade-offs and issues were analyzed, i.e., relative flight path and speed tracking, flight path and speed tracking performance versus control activity, and proper control reconfiguration when reaching performance limits.
A major advantage of the total energy and energy distribution control concept is that it takes maximum advantage of the control effectiveness of thrust and elevator by providing conflict-free decoupled control.
Where possible, the system uses airmass-referenced feedback signals, filtered to remove atmospheric noise in the frequency range where the controls cannot respond effectively.
The simulated time history shows how a severe wind shear of -5 kn/sec causes instant bleedoff of the airspeed and a slower departure from the commanded altitude of 5000 ft. This signifies a sharp loss of aircraft energy. The system responds by bringing the throttles forward quickly. Even when limit thrust is reached, the total energy rate remains negative, causing the VWTWspeed control mode to engage and the flight path control to be abandoned. The aircraft captures the minimum speed command (1.3 Vstall or QREF) undershoot free and settles into a steady-state descent. In this case, the filtering of the airspeed related feedbacks was adjusted to yield quick response to wind shear.
Response to Severe Wind Shear
Wind shear - 5 kn/sec
B747
ar Down ALT 20 HDOT 11 60 ft/s CAS 175 kn 135 ALPHA 16 15 THETA deg 0 -5 deg ELEV THROTTLE 80 deg 0 0 10 20 30 40 50 60 70 Time, (set) CONTROL IN WIND SHEAR AND TURBULENCE (CONT’D) For the responses below, the same wind shear condition was run with the filtering of the airspeed control feedbacks adjusted to reduce controller activity in turbulence by -50%. As a result, the throttles respond more slowly to the wind shear and more airspeed is lost.
When the thrust limit is reached, the path control is abandoned and speed control is continued through the elevator. Since the total energy rate remains negative, causing the Vc of the VMIW mode to be more positive than Vc of the selected speed mode, the VnIW mode engages and the minimum speed is recaptured. In the process, the airplane comes very close to stall, but stays closer to the commanded altitude for a longer period of time, before settling into the steady descent.
In a similar situation on the glideslope, the system will stay at the VHIN speed until the wind shear subsides to the point where the total energy rate once again becomes positive. At that time, the glideslope will be recaptured first before the pilot-.selected speed will be reestablished.
As these time histories indicate, the best insurance in case of severe wind shear is to have excess airspeed energy, together with provisions that prevent the airplane from stalling.
Response to Severe Wind Shear
B747 Wind shear - 5 kn/sec Altitude/CA!&Control ear Down ALT 5000 20 HDOT 3000 -20 ft 1000 -60 ft/s CAS 195 kn 115 ALPHA 16 15 THETA 8 5 deg 0 -5 deg ELEV THROTTLE 80 deg 0 deg 50 60 70 80 90 0 40 Time, (set) MODE CONTROLPANEL DESIGN The integration of functions and enhanced operational capabilities of the TECS is reflected in the example mode control panel layout below.
The modes have been chosen to provide all necessary tactical and strategic capabilities. All modes are operable over the entire flight envelope. Any flight path mode is compatible with every speed mode. The panel provides the For example, when either complete mode and control configuration status.
flight path or speed control is abandoned due to thrust limiting, the At the appropriate VARIABLE indication on the bottom of the panel is lit.
same time, the THRUST LIMIT light on the left of the panel lights. The thrust rating mode selection has been integrated in this panel because the performance capability of the vertical flight path and speed control modes depends directly on the selected thrust rating mode.
The CAS and Mach modes can be preselected for automatic mode transition during climb-out and descent.
The T-NAV, V-NAV, and L-NAV modes allow the navigation/performance computer and lateral paths to be input to the TECS for target speed, target vertical, execution. Reference 2 discusses the operational aspects in more detail.
- AiTITUDE THRUST MODE' TEMP SEL CAS MACH FPA -TRACK 1 5 0 p-jqT] TT;T;n LIIJ EPR REF N H L qfgsEL:i: 3s RTG pzq pq p-1 TURN R VARIABLE . INTEGRATED THRUS-I RATING l PERFORMANCE LIMIT INDICATION l INTEGRATED ALT SEL/ALT CLEARANCE TECS ENGINE CONTROL The first major objective of the integrated flight and propulsion control system was the development of a more accurate net thrust command computation.
The second major objective of the design was the elimination of arbitrary engine thrust variation due to changes in environmental or flight conditions.
Therefore, in the TECS design, all major engine environmental dependencies have been compensated to make the engine produce the desired net thrust on comand.
The net thrust command is converted into an EPR command and used for closed-loop engine control. This involves dependencies on the pressure ratio 15 = p/p0 and Mach number. Further, the EPR command is used to predict the steady-state throttle position comand. To maintain constant loop gain, the total air temperature dependency has been compensated. The closed-loop EPR control law also serves to reduce the variation in the response dynamics of the engine and the effects of throttle actuation nonlinearities.
Feedback of throttle position and EPR in excess of selected limits to the thrust command computation serves to prevent engine overboost and ineffective throttle control in the low thrust region.
In the future, customization of the automatic flight control system to work with specific engines can largely be eliminated by designing the electronic engine controls to control to a net thrust command and to compensate for all major thrust environmental dependencies (ref. 4).
TECS
-m
Thrust ---+ Servo - Engine No. 2 - Command Comp Averaging
I
. 1 * A EPR No. 2 Highest’ EPR No. 1 Select . e AFCS FUNCTION OISTRIBUTION The integration of all vertical flight path and speed control functions, together with future integration of lateral directional functions, allows a more efficient system architecture using fewer sensors, computers, inter- faces, and a function distribution that is closer oriented toward airline needs.
As shown, all automatic control modes and safety-oriented functions are which will be designed with the consolidated in the flight control computers, necessary redundancy to provide the required integrity of the critical control functions.
The consolidation of all automatic control functions in the flight control computer (FCC) leaves the navigation/performance (flight management) computer navigation, performance, with strictly airline-operations-oriented functions: and-path definition. The resulting flight control targets can be transmitted No control loops would be closed in the navigation/ to the FCC for execution.
performance computer. This design eliminates functional overlap and provides consistency of operation for all modes.
AFCS Function Distribution Using TECS
Architecture
(NO FUNCTIONAL OVERLAP)
Airline Operations- l Performance Oriented Functions l Altitude/Vertical Speed Pilot Workload Relief l Heading/Track and Safety-Oriented l Vertical Path Control Functions TOTAL ENERGY CONTROLSYSTEM PAYOFF The payoff of the total energy control system development is evident: the design provides proper integration of the flight path and speed control functions, resulting in optimum thrust and elevator control efficiency. The design eliminates the numerous limitations of the previous state-of-the-art autopilots and autothrottles. The system implementation is simple, without functional overlap or operational ambiguities, requiring less software and less hardware.
Safety has been enhanced by complete stall and overspeed protection in case of operational errors and severe wind shear.
The potential for cost reductions is substantial.
The design is largely generic. For example, transfer of the, complete TECS design from the 8737 to the B747 simulator, including adaptation and checkout required only 6 engineering man months.
of innerloops and VMIN/V~X schedules, No energy control concept related changes were needed. The total energy con- trol system is scheduled to be evaluated in flight on the NASA B-737 aircraft under the Air Transport Operating System Program, in the summer of 1984. The Boeing Company will provide the system definition and under contract assist NASA with checkout of a complete TECS simulation, flight software specifica- tion, software test, and flight test.
l Fully integrated, generally applicable design
l Improved performance for all modes
- Decoupled path and speed maneuvering
- Energy efficient thrust control
- Uniform stability bandwidth, transient responses
l Enhanced operational capabilities
- Complete safety and maneuver envelope limiting
- Configuration control when thrust limited
- Reduced pilot workload -
(simpler mode control, VCWS)
l Control law software reduced = 75%
l Fewer sensors, computers
l Large cost reductions
- Development, flight test, procurement, maintenance
REFERENCES 1. H. A. Soul& "The Throttles Control Speed, Right? Wrong!" AIAA Journal of Astronaut. & Aeronaut., vol. 7, no. 72, Dec. 7969, pp. 74-75 and 81.
2. A. A. Lambregts, "Operational Aspects of the Integrated Vertical Flight Path and Speed Control System," SAE Paper No. 831420, October 1983.
3. A. A. Lambregts and 0. G. Cannon, "Development of a Control Wheel Steering Mode and Suitable Displays That Reduce Pilot Workload and Improve Efficiency and Safety of Operation in the Terminal Area and in Windshear," AIAA Paper No. 79-1887, August 1979.
4. A. A. Lambregts, "Integrated System Design for Flight and Propulsion Control Using Total Energy Principles," AIAA Paper No. 83-2561, October 1983.
FLIGHT TEST RESULTS FOR THE DIGITAL INTEGRATED AUTOMATIC LANDING SYSTEM (DIALS) - A MODERNCONTROLFULL-STATE FEEDBACK DESIGN R. M. Hueschen NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 DIALS INTRODUCTION A goal of the Advanced Transport Operating Systems (ATOPS) Program at Langley Research Center is to increase airport capacity. One on-going effort contributing to this goal is the development of advanced autoland systems which can capture the localizer and glideslope with low overshoot and safely do so for short finals (1.5 to 2.0 nautical miles). Another capability being developed is the ability to fly selectable and steeper glideslopes with the potential for noise reduction and wake vortex avoidance. Systems with improved performance in turbulence and shear wind encounters are also desired. One autoland system designed to have these capabilities and flight tested within the ATOPS program is called the Digital Integrated Automatic Landing System (DIALS). The DIALS is a modern control theory design performing all the maneuver modes associated with current autoland systems: localizer capture and track, glideslope capture and track, decrab, and flare. The DIALS is an integrated full-state feedback system which was designed using direct-digital methods. The DIALS uses standard aircraft sensors and the digital Microwave Landing System (MLS) signals as measurements. It consists of separately designed longitudinal and lateral channels although some cross-coupling variables are fed between channels for improved state estimates and trajectory commands. The DIALS was implemented within the 16-bit fixed-point flight computers of the ATOPS research aircraft, a small twin jet commercial transport outfitted with a second research cockpit and a fly-by-wire The DIALS became the first modern control theory design to be successfully system.
flight tested on a commercial-type aircraft. Flight tests were conducted in late 1981 using a wide coverage MLS on Runway 22 at Wallops Flight Center. All the modes were exercised including the capture and track of steep glidescopes up to 5 degrees.
ATOPS TRANSPORTATION SYSTEM RESEARCHVEHICLE (TSRV) The figure below shows a picture of the ATOPS research aircraft referred to as the TSRV which was used to flight test the DIALS.
Flight test data of sensors and selected computer outputs were recorded on magnetic tape for postflight processing.
In addition, on-line strip chart recorders provide real-time readouts of software selectable variables for in-flight system analysis and performance evaluation.
.I^ ‘3 ADVANCED TRANSPORT OPERATINGSYSTEMS(ATOPS) TRANSPORTATION SYSTEMSRESEARCH VEHICLE (TSRV) ,, ,, ' * DIALS AUTOMATIC OPERATIONS The pictorial illustrates the automatic operations performed by DIALS.
Prior to beginning the DIALS operations, the pilot selects the desired glideslope and desired reference airspeed. Then the aircraft is flown towards the runway centerline at a desired heading or ground track angle up to +60° from the runway heading. The DIALS will then automatically initiate the capture maneuver when the capture criteria, which consider aircraft parameters, desired trajectory, and wind conditions, are satisfied. Localizer and glideslope capture can occur independently, in any order, or simultaneously. Upon completing the capture mode, the localizer and glideslope track modes are engaged to maintain the aircraft along runway centerline and the selected glideslope. At an altitude of 250 feet, the decrab maneuver is engaged.
The control law commands the aircraft into a sideslip maneuver, a maneuver that pilots often use, while maintaining the track along runway centerline. In an alti- the flare maneuver is engaged.
tude range of SO to 150 feet, The exact engagement altitude is a function of the glideslope angle, the vertical velocity, and other air- craft parameters. During the flare maneuver, the aircraft is commanded to follow a fixed trajectory in space to a prespecified touchdown point on the runway. The fixed trajectory was chosen as a means to enable precision or low dispersion touch- down. With a precision touchdown capability, an aircraft can reliably decelerate and exit the runway at its earliest convenience and reduce occupancy time.
CLOSE-IN CAPTURE OF LOCALIZERAND STEEP MLS GLIDESLOPE DECRAB AND ADVANCED FLARE 3" ILS GLIDESLOPE PRECISION TOUCHDOWN THE DIALS DESIGN AND SYSTEM DESCRIPTION The DIALS was designed using stochastic modern control theory and direct-digital design methods. The system equations were linearized about.a nominal glideslope, and a set of constant gains was determined according to a quadratic cost function. The gains were then used in one basic control law which accommodated all the autoland modes. The control law generates control commands at 10 Hz, which is one half that use&in- the digital baseline system on-'the research. aircraft.
The sensors used for the DIALS flight tests were the digital MIS signals (azimuth, elevation, and range), pitch, roll, and yaw rate from standard rate gyros, body-mounted accelerometers, vertical velocity determined from MIS processing, calibrated airspeed, engine-pressure-ratio, throttle position, stab.ilizer position, barometric altitude, Pitch, roll, and yaw from an and radar altitude during flare.
inertial platform were used although the design was intended to use standard vertical and directional gyros. However, these gyros were not available on the test vehicle.
DIALS has a full-state constant gain Kalman filter which estimates the states of The wind estimates are fed directly the aircraft and steady, gust, and shear winds.
into the control law to provide improved performance to changing wind conditions.
The longitudinal and lateral control laws were each designed independently of one another. Each design was a multi-input multi-output problem resulting in inte- grated or coordinated controls. The controls coordinated in the longitudinal channel were throttle, elevator, and stabilizer while those coordinated in the lateral were the aileron and rudder.
q FULL-STATEDIRECT-DIGITAL DESIGN USING MODERN CONTROL THEORY l One Basic Control Law For All Modes l lo-Hz Sampling Rate q SENSORS l Digital MLS Signals (AZ, EL, R) l Attitudes & Attitude Rates l Body-Mounted Accelerometers l tY MLS 8 Calibrated Airspeed l EPR's, Throttle Position, & Stabilizer Position l Barometric Altitude & Radar Altimeter During Flare q FULL-STATE KALMAN FILTER (CONSTANT GAINS) WITH WIND ESTIMATESTHAT ARE USED DIRECTLY IN CONTROL LAW, q LONGITUDINAL CONTROL LAW HAS INTEGRATED THROTTLE,ELEVATOR,8 STABILIZER AND LATERALLAW HAS INTEGRATED AILERON & RUDDER, MODIFICATIONS TO DIALS NOMINAL DESIGN One basic control law structure was used for the DIALS but some modifications were made to the nominal design as a result of evaluation iti a nonlinear simulation to achieve desired performance for the various control modes.
For the glideslope capture mode, at the instant of engagement, the desired vertical velocity along the glideslope was commanded to the desired value through an easy-on rather than letting the command to the control law be a step command.
In the glideslope mode, the vertical position error is integrated and used in the control law when this mode is engaged. Also the gains on vertical position and vertical velocity are increased by means of an easy-on for tighter tracking.
Some similar changes were made for the localizer track mode. At track mode, engagement gains on the lateral position are increased through an easy-on, and the lateral position and roll errors each are integrated to eliminate position standoff and to drive roll attitude to wing level. The roll integrator insures that the con- trol law will achieve the desired crab angle and zero sideslip in steady-state wind conditions.
During the decrab mode, the aircraft heading is commanded to align with the run- way through trajectory commands. To insure a sideslip condition, the gain on the roll integrator used during localizer tracking was ramped to zero, and the heading error was integrated and fed back to the control commands.
To insure close tracking of the flare trajectory, the gains on altitude, alti- tude rate, and pitch rate were increased by means of an easy-on at the initiation of the flare mode.
q G/S CAPTURE MODE l COMMANDED VERTICAL VELOCITY AT ENGAGEMENT WITH EASY-ON o G/S TRACK l ADD INTEGRATOR ON VERTICAL POSITION ERROR l INCREASEGAINS (WITH EASY-ON0~ VERTICAL POSITION 8 VELOCITY ERRORS) q LOC TRACK l ADD INTEGRATORS FOR LATERAL POSITION 8 ROLL ERRORS l INCREASEGAIN (WITH EASY-ON) ON LATERAL POSITION ERROR q DECRAB 9 ADD INTEGRATOR TO HEADINGERROR . REMOVE FEEDBACK OF ROLL INTEGRATOR WITH EASY-OFF q FLARE l INCREASEGAINS (WITH EASY-ON) ON VERTICAL POS 8 VEL 8 PITCH RATE BLOCK DIAGRAM OF FEEDBACK LOOP The block diagram shows the feedback loops associated with the DIALS longitu- dinal and lateral control laws. The Kalman filter processes the MLS and aircraft sensor data to determine estimates of the aircraft states and winds. There are 16 aircraft and wind states for the longitudinal channel and 13 for the lateral channel.
The Kalman filter makes path predictions of the aircraft at the next iteration. These predictions are dIEEerenced with the fltght path generator path, and the.result is fed back to the controls to provide path lead information. The aircraft state estimates are differenced with desired trajectory from the flight path generator to form trajectory error signals. The desired trajectory signals are also fed back to the control law to provide the control law information about trajectory deviations from the nominal glideslope.
Using the wind estimates and the above described signals, the control commands are generated ten times a second. The commands enerated for the longitudinal controls are elevator position 6,, throttle rate 8 and stabilizer rate 6 . By choosing throttle rate as a command, it was p%Aible to achieve satisfaztory throttle rate activity by appropriate weighting of it in the quadratic cost function. Stabilizer rate was weighted in the cost so that this command acted primarily as a trimming function. By using rate commands, no penalty is incurred in the cost function for position deviations from the nominal values for the stabilizer and throttle. However, for the elevator, the desire is to keep it nominally near the neutral position so that maximum authority is always available. T$e commands from the lateral control law are aileron position fa and rudder rate 6 . The three rate commands were integrated at a 20-Hz rate to provide smoother positfon commands to the aircraft servos. The stabilizer position is added to the elevator command. In the design, the stabilizer rate was intended to control the trim logic in such a manner as to turn the stabilizer trim motor on and off on the test vehicle; however, due to built-in restrictions on the direction of stab- ilizer movement as a function of elevator position, the stabilizer trimming was achieved by logic driven by the elevator deflection (logic that existed on the base- line test vehicle).
MLS KALMAN FILTER SfNSORS _ ; Aircraft State Estimateh i2 TRIM UP TRIM DOWN 1 PATH PREDICTION CONTROL -I CONfUNDS DIALS LOCALIZER CAPTURE - FLT 361, RUN 7 The capture of the localizer using a 30-degree intercept angle is shown below.
The This flight was conducted in 12 knot crosswinds with 2 to 3 knot gust variations.
top graph is a plot of the deviation of the aircraft from the runway centerline, the and the bottom graph is the aircraft yaw middle plot is the aircraft roll attitude, The capture maneuver was or true heading with respect to runway centerline.
The aircraft captures the localizer with no initiated at 5 seconds on the plots.
The aircraft heading overshoot and was settled on runway centerline by 35 seconds.
indicating that the estimate of smoothly achieves the crab angle with no oscillation At 97 seconds into the flight, a crosswind used by the control law is accurate.
Incidently, this magnitude lateral wind shear of 8 knots/100 feet was encountered.
of wind shear is that specified in the FAA Advisory Circular No. 20-57a for certifica- tion of autoland systems. The aircraft heading again smoothly changed to the runway The aircraft deviated approximately 25 feet from the heading as the wind diminished.
runway centerline during the shear but corrected back to centerline within 10 seconds after the shear stopped.
200 r L LOCALIZER DEVIATION _ TIME [SEC) l$?JL- 160 90 &go_ 100 yo 20 ,-sQ FEET 0) , , , IpTl 1 IY I 11 I I 1 -i-l I Iy-‘A’ I ’ ‘-’ I ’ ’ ’ 1 ‘/
/ 30' INTERCEPT
12 KNOT CROSSWIND 2-3 KNOT GUSTS TIME (SEC) 60,, BO,, * 100,//h 120 /o,140 ROLL, 20 90 p o’c , , ls”Vyl IY~/I I +-r- 1 1 1’ Iv/- I I ‘7”’ I I I- I I I DEG I/ r -L v -20 _ \ DECRAB YAW I DEVIATION _ \S 90 60 T1ME8:SEC1 100 ,-&2!’ --Go 160 FROM o[ .
I ’ ’ 1 ‘1-k’ --,I ,J. A-- RUNWAY DEG t SHEAR -20 _ DIALS 3-DEGREE GLIDESLOPE CAPTURE AND TRACK - FLT 361, RUN 7 The graphs below show the capture of the 3-degree glideslope which begin at 24 The top plot seconds during the time the localizer capture was being completed.
shows the aircraft deviation from the desired glideslope during the glideslope cap- and during the flare mOde, ture and track mOde, it shows the deviation from the flare trajectory. The middle graph is a plot of the aircraft pitch attitude, and the bottom The glideslope is captured graph shows the vertical velocity from MIS processing.
with no position overshoot although some overshoot does occur in the desired glide- slope vertical velocity. This velocity overshoot is not present in the steeper glideslope captures and was not shown in the developmental simulation runs suggesting that an error was present in the glideslope capture criteria flight software. The shortness of time for the capture when compared to simulation times and the capture.
times for the steeper glideslope captures, to be discussed in subsequent figures, suggest that the capture maneuver was initiated too late or too close to the glide- slope. The flare maneuver is initiated at 130 seconds into the flight. The pitch attitude reaches about 2 degrees for good nose wheel clearance, and the vertical velocity is smoothly reduced to an acceptable touchdown sink rate.
VERTICAL TIME [SEC1 PATH b DE;;;;IONO-~ , , I c 12 KNOT CROSSWIND 2-3 KNOT GUSTS PITCH,
F
DEG TIME [SEC1 o I TINE [SEC) I I I DIALS FLIGHT DATA - FLT 361, RUN 7 The flight data shown below are plots of calibrated airspeed (top graph), throttle position (middle graph), and total elevator position (bottom graph) which includes the mechanical downrigging of the elevator due to stabilizer trim position.
This is companion data for the test run discussed in the previous two figures.
The reference airspeed for this run was set to 130 knots.
The data shows that the measured airspeed was somewhat higher than the selected reference by 5 to 8 knots.
DIALS had a tendency to control the airspeed about 3 to 5 knots higher than the selected reference.
The tendency may be related to the way the desired airspeed is achieved by DIALS. The current airspeed is estimated in DIALS by adding its estimate of ground speed and wind along the longitudinal axis.
If this estimate is lOW, the DIALS will advance the throttle to increase inertial speed and thus airspeed.
The slightly higher airspeed may be due to a bias in the measurement or a low estimate of airspeed. Coordination of the throttle and elevator is shown in the plots. At glideslope capture (24 seconds), the elevator goes positive (trailing edge down) to pitch the aircraft down while the throttle is immediately reduced towards glideslope thrust. Although the glideslope pitch is achieved at 30 seconds, the elevator continues to adjust to counteract the pitching moment due to thrust changes.
During flare, the elevator moves to pitch the aircraft up while the throttle decreases to satisfy a commanded reduction in airspeed.
Note that the throttle is not driven to idle but to a position to achieve the commanded airspeed.
AIRSPEED, KNOTS 12 KNOT CROSSWIND t 2-3 KNOT GUSTS r THROTTLE
-L/--
POSITION, TIME (SEC1
1 *~
20( 20 90 60 6gr-x 100 120 140 DEG , , ( 1 ’ ’ ’ 1 L-r 0 t GLIDESLOPE CAPTURE 10- /- TOTAL ELEVATOR 3vr-nfl 0 ?jr' , ~~~~~dvv-~~~~~~o~~~2~~~~qo POSITION, I 20 90 6 160 , , , I VI I lll~lll~lll~lll~lll~lll DEG DIALS 4.5-DEGREE GLIDESLOPE CAPTURE AND TRACK - FLT 361, RUN 8 The variables plotted below are the same as those plotted two figures earlier.
These plots show the glideslope capture and track and flare for a 4.5-degree glide- The slope. This flight test encountered 15 knot crosswinds and 2 to 3 knot gusts.
40- to 80-second time period, which included portions of the localizer capture and The glideslope capture begins at 113 seconds and is completed track, has been omitted.
The flare maneuver at 126 seconds with no overshoot in position or vertical velocity.
begins at 177 seconds and touchdown occurs at 192.seconds for a 15-second flare duration. The flare maneuver was initiated at 134 feet above the touchdown point.
The flare is automatically initiated at higher altitudes for the steeper glideslopes to reduce the sink rate to reasonable levels before the lower altitudes are reached for safety purposes. The pitch attitude at touchdown was 2.5 degrees, and the touch- down sink rate was 2 feet/second.
VERTICA: PATH TIHE (SEC) DEVIATION 160,q 200 FEET 0 120 _ y.1110 ,,.ieo ,, 15 KNOT CROSSWIND 2-3 KNOT GUSTS -90 FILS VERTICAL - TIME [SEC1 120 1110 160 160 200 47 I III I III 1 III I III \ f’ ’ DIALS FLIGHT DATA - FLT 361, RUN 8 The graphs below are plots of additional variables for the same run discussed in the previous figure. The reference airspeed selected for this flight was 126 knots.
The control of the baseline autothrottle on the test vehicle can be seen prior to glideslope capture which occurred at 113,seconds. The DIALS again controls the air- speed above the selected reference similar in magnitude to that discussed for the previous Run 7 of Flight 361. However, the variations in airspeed are much smaller than variations of the baseline control. A small positive elevator pulse is seen at 113 seconds along with a throttle reduction to initiate the glideslope capture.
However, DIALS "sees" an airspeed error and immediately advances the throttle.
The estimated thrust is fed back to the elevator control, and it moves positive to counter- act the thrust increase and to push the nose of the aircraft down. At flare, the elevator goes negative to pitch the nose of the aircraft up while very little change is seen in throttle. The thrust must be maintained to add energy to the system to achieve the reduction in sink rate. The airspeed is being reduced to the commanded value without throttle change by the pitch-up maneuver.
TIHE (SEC) 15 KNOT CROSSWIND 106 2-3 KNOT GUSTS t
uor
THROTTLE POSITION, DEG 20 0 t- GLIDESLOPE CAPTURE \ FLARE a- TOTAL ELEVATOR POS ITIONb DEG -.-...-_..- ._..
.
DIALS 5-DEGREE GLIDESLOPE CAPTURE AND TRACK - FLT 364, RUN 7 The graphs below show flight data for the steepest glideslope capture and track flight tested with DIALS, a S-degree glideslope. This run was flown in strong wind conditions. The crosswind component was 20 knots and the headwind component was 10 knots with gust variations of 8 to 10 knots. The variables plotted are the same as those plotted in earlier figures. The glideslope capture occurs at 57 seconds and is completed by 70 seconds. The capture was achieved with essentially no overshoot in the glideslope or desired vertical velocity. The flare maneuver was initiated at 124 and a positive pitch attitude of 3 degrees was obtained. However, the touch- seconds, down was not completed due to lateral position drift during decrab. The drift was caused by rudder limiting built into the servos for safety purposes during these flight tests. The rudder limit was encountered due to the 20-knot crosswinds.
YO
VERTICAL F
PATH TIME REC) DEVIATION, 1'1 FEET '
/
CROSSWIND- 20 KNOTS
-40 -I
HEADWIND - 10 KNOTS 10 8 10 KNOT - GUSTS PITCH, DEG TIME (SEC1 TIME (SEC) lfi0 DIALS FLIGHT TEST DATA - FLT 364, RUN 7 The variables plotted below are additional data for the flight just discussed in the prior figure, and the variable-s are the same as those plotted in earlier figures. The reference airspeed for this flight was selected at 129 knots. Again, airspeed errors are biased to the positive side although the errors are closer to the selected airspeed than in runs discussed earlier. The throttle position plot shows that throttle limiting occurred at 10 degrees during the periods of 82 to 95 The lower throttle limit is set seconds, 104 to 106 seconds, and 121 to 125 seconds.
to 10 degrees when the system is not in the flare mode to keep the engine spooled up.
In flare, the throttle is allowed to go to idle, zero degrees. In spite of this limiting, the integrated control maintains stability and track of the glideslope. At the throttle increases rather than decreases as was the case for flare (124 seconds), the 3-degree glideslope because more energy must be added to the system to reduce the high sink rate that occurs along the 5-degree glideslope. Even at the point of near touchdown, the throttle is increasing while airspeed was decreasing.
i- AIRSPEED, KNOTS CROSSWIND- 20 KNOTS t HEADWIND - 10 KNOTS 109 L 8 - 10 KNOT GUSTS THROTTLE POSITION, DEG TOTAL ELEVATOR POSITION, 160 DEG v \‘I c DIALS FLIGHT TEST PERFORMANCE - LOCALIZER AND GLIDESLOPE Below is a pictorial to illustrate the performance improvement of DIALS compared to conventional ILS systems. The improved performance was attributable both to the use of the ML.5 (the MLS signals have more accuracy and expanded coverage over 1LS)and It shows that overshoot performance is better by an the advanced control law design.
order of magnitude. DIALS achieved the improved glideslope overshoot with the The time required from additional capability to fly selected and steep glideslopes; localizer capture initiation to engagement of the track mode was a factor of 3 less than the conventional system time. The time to capture the glideslope (for the 3-degree glideslopes since conventional autolands for steeper glidescopes do not exist) was a factor of 4 less for DIALS. The reduced time to capture the localizer and glideslope provides the capability to fly shorter final approach paths ,and, more path flexibility.
in general, The low overshoot during localizer capture The capability to fly steeper glide- will allow closer spacing of parallel runways.
slopes reduces the noise level on the ground and also provides a means for trailing aircraft to avoid wake vortices by selecting a glideslope steeper than that of the aircraft in front of it. The use of parallel runways spaced closer and the capa- bility to fly short approach paths should result in greater terminal area capacity.
HORIZONTAL OVERSHOOT VERTICAL OVERSHOOT
/’
c ’
MOREFLEXIBLE APPROACH PATHS 4' ,H 4' GREATER :APAC :ITY VERTICAL OVERSHO CLOSER RUNWAYS IMPROVEDPERFORMANCE IN WINDS MULTIPLE ANGLES (WAKE VORTEX) DIALS FLIGHT TEST RESULTS The figure below summarizes the number and types of captures and automatic landings that were achieved during the DIALS flight tests. Three-, four-and-one-half-, and five-degree glideslopes were successfully captured. For 21 captures, the mean overshoot of the glideslopes was 4.6 feet with a standard deviation of 2.3 feet.
The numbers in parentheses at the right represent a conventional system. The localizer captures were initiated from ground track angles of 20, 30, 40, and 50 degrees. The statistics of 41 captures resulted in a mean overshoot of 24.2 feet with a standard deviation of 25.7 feet. Typical overshoot distances for a conven- tional system are shown in parentheses at the right. The decrab maneuver was success- fully performed in crosswinds up to 12 knots. Successful decrabs in higher cross- winds were prevented due to rudder limiting in the experimental system of the test vehicle. Ten completely automatic (hands-off) landings were achieved from both the 3- and 4.5-degree glideslopes. In addition, seven additional landings were made in which the pilots had only a slight column input and did not disengage the automatics.
The statistics for the ten hands-off landings were a mean touchdown vertical velocity of -2.4 feet/second with a standard deviation of 0.7 feet/second. The numbers to the right are, respectively, the mean and standard deviation for an advanced flare designed by classical methods and previously flight tested on the same aircraft using MLS signals.
o HAVE SUCCESSFULLY CAPTUREDAND TRACKED 3, 485, AND 5" GLIDESLOPES MEAN OVERSHOOT = 4,6 FT (40-60) 21 CAPTURES Q = 2,3 FT o HAVE SUCCESSFULLY CAPTUREDRUNWAY CENTERLINE (LOCALIZED) AT 20", 30", 40" AND 50" INTERCEPT ANGLES MEAN OVERSHOOT = 24,2 FT (300-1000) 41 CAPTURES 0 = 25.7 FT a HAVE PERFORMED SUCCESSFULLY DECRABMANEUVERSIN CROSSWINDS UP TO 12 KNOTS o PERFORMED TEN COMPLETELY AUTOMATIC (HANDS-OFF) LANDINGS FROMBOTH 3" AND 4.5' GLIDESLOPES, PLUS SEVEN ADDITIONAL WITH SLIGHT COLUMNINPUT t-2,34) MEAN& = -2.4 FT/SEC 10 LANDINGS u= 67 FT/SEC (1,411 DIALS PLANS A goal of the advanced control law research has been to develop methods and tech- niques whereby modern control theory can be applied readily to practical applica- tions. While excellent experimental results were obtained from the DIALS flight the full-state design with its accompanying Kalman filter can be rather test, cumbersome resulting in large software memory requirements. Also certain practical considerations, such as the dynamics of existing sensor filters and stability aug- mentation systems, are difficult to incorporate in full-state'designs. These prac- tical considerations tend to make full-state designs less practical- for direct commercial applications. Thus, the plans are to use the DIALS results as a bench mark and to continue development of modern control theory methods for more direct solutions to practical problems. With the recent development of a digital computer algorithm that efficiently and reliably solves the output feedback problem, the design of control systems with modern control theory becomes more practical. Successful 3-D guidance and control designs were recently achieved using this algorithm. Thus, the design and the development of an autoland system using output feedback techniques have begun. For this design, a third-order complementary filter along with MLS process- ing is being used to provide estimates of velocity and position rather than a Kalman filter.
The existing SAS (yaw damper) on the test vehicle will be used rather than providing this damping in the design as was done with DIALS. The output feed- back design allows inclusion of SAS dynamics and filter dynamics in the model with- out having to feed back these states as is necessary in full-state design. This system will be designed as a Type 1 system whereas DIALS was designed as a Type 0 system with integrators added to the various modes during development to achieve Type 1 properties. The new design will use the integrated control surfaces like DIALS as listed in the figure below. In addition, more attention will be given to methods to account for computational delays than was done in DIALS. Full nonlinear simulation development is planned along with flight tests in late 1984 or early 1985.
q NEWMOREPRACTICAL DESIGN USING LSF TECHNIQUES l Direct Digital Design l Use Third-Order Complementary Filter (Developed for ICAOMLS Processing) Rather Than Steady-State Kalman Filter l Use Existing SAS (Yaw Damper) . Design as Type 1 System (Path Integrators) l Integrated/Coordinated Controls For Elevator, Throttle And Stabilizer For Long. Axis And Aileron And Rudder For Lateral Axis . Design To Account For Transport Delays (Computational) q DEVELOPBY MEANSOF FULL NONLINEARSIMULATION INCLUDING SENSORS NOISES AND WINDS, q PREPARES/W REQUIREMENTS FOR FLIGHT TESTS IN CY 84.
BIBLIOGRAPHY (1) Halyo, N., "Development of an Optimal Automatic Control Law and Filter Algorithm for Steep Glideslope Capture and Glideslope Tracking," NASA CR-2720, Aug. 1976.
(2) Halyo, N., "Development of a Digital Automatic Control Law for Steep Glideslope Capture and Flare," NASA CR-2834, June 1977.
(3) Halyo, N., "Development of a Digital Guidance and Control Law for Steep Approach Automatic Landings Using Modem Control Theory Techniques," NASA CR-3074, Feb. 1979.
(4) Halyo, N., "Terminal Area Automatic Navigation, Guidance and Control Research Using the Microwave Landing System (MLS): Part 5 - Design and Developmemt of a Digital Integrated Automatic Landing System (DIALS) for Steep Final Approach Using Modern Control Techniques," NASA CR-3681, April 1983.
(5) FAA Dept.
of Transportation, "Automatic Landing Systems," FAA Advisory Circular 20-57a, 12 Jan. 1971, p. 2.
APPLICATION OF ADVANCED CONTROLTECHNIQUES TO AIRCRAFT PROPULSION SYSTEMS Bruce Lehtinen NASA Lewis Research Center Cleveland, Ohio First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 24-26, 1983 CONTROLS FOR ADVANCEDTURBINE ENGINES The trend in advanced aircraft turbine engine design is toward more sophisticated cycles and mechanical complexity. This is done primarily to achieve improved thrust-to-weight ratios and improved specific fuel consumption. Control system complexity has also increased due to the increase in the number of engine varia- bles which must be scheduled or controlled. Hydromechanical controls are rapidly being replaced by full-authority digital electronic controls in order to handle the added computational burden. The digital computer, however, does allow more sophisticated control algorithms to be used. In this paper, two Lewis Research Center sponsored programs will be described which involve the application of ad- vanced control techniques to the design of engine control algorithms. Multi- variable control theory has been used in the FlOO MVCS (multivariable control to design controls which coordinate the control inputs for synthesis) program improved engine performance, and it presents a systematic method for handling a com- plex control design task (fig. 1). Methods of analytical redundancy are aimed at increasing the control system's reliability.
The FlOO DIA (detection, isolation, and accommodation) program will be described, which investigates the uses of software to replace or augment hardware redundancy for certain critical engine sensors.
GOALS OF - IMPROVED PERFORMANCE AND EFFICIENCY LEADTO - INCREASEDCOMPLEXlTYAND DECREASED RELIABILITY ADVANCEDCONTROLTECHNIQUES . MULTIVARIABLECONTROL THEORY - COORDINATE CONTROLVARIABLES FOR IMPROVED PERFORMANCE - SYSTEMATICAPPROACH TO COMPLEXITY . ANALYTICAL REDUNDANCY - SOFTWARE REPLACESHARDWAREREDUNDANCY - IMPROVE RELIABILITYWlTH RESPECT TO SENSOR FAILURES Figure 1 . .
APPLIED ADVANCED ENGINE CONTROLTECHNIQUES The interrelationship of the two programs is depicted in figure 2. Activity in applying multivariable control methods to engine controls began in the mid 1970's (refs. 1 and 2). Motivated by these early results, a comprehensive program (FlOO MVCS) was begun to demonstrate the benefits of using control theory to design a full envelope control for an FlOO engine and to verify the design with both simu- lated and actual engine testing. Figure 2 shows the NASA LeRC facilities which were used in performing the evaluations of the MVC logic. The facilities consist of 1) a research-type control computer on which the control algorithms are pro- grammed, 2) a real-time (hybrid) simulation of the FlOO engine, and 3) an altitude test cell in which a full-size engine can be run. The FlOO DIA program, which is now beginning the evaluation phase, built upon early theoretical iork (ref. 3) in analytical redundancy as applied to flight control systems. The FlOO MVCS control forms the basis for the control logic used in the DIA program, with sensor fail- ure DIA logic being incorporated with it to produce the overall control. As in the MVCS program, the DIA logic will utilize the LeRC facilities for overall per- formance evaluation.
THEORY PROGRAMS FACILITIES Figure 2 FL00 MVCS PROGRAM PARTICIPANTS Figure 3 shows the organizations involved in the FL00 MVCS program and outlines the activities involved in the various phases of the program. The contracted portions, funded and monitored jointly by NASA LeRC and the Air Force Wright Aero- nautical Laboratories, were carried out by Pratt and Whitney Aircraft (P&WA) and Systems Control, Inc. (SCI). The overall program objective was to demonstrate the benefits of using linear quadratic regulator (LQR) synthesis procedures in design- ing a practical multivariable control system that could operate a turbofan engine through its operating envelope. P&WA provided a digital simulation of the FlOO, a set of linear design models, and performance criteria on which the design was based. SC1 conducted the overall multivariable control design, which was then evaluated by P&WA on the digital engine simulation. NASA LeRC programmed the control logic on a research control computer and evaluated it using its,real-time hybrid FlOO engine simulation. Upon successful evaluation with the simulation, NASA conducted full-scale altitude tests to verify proper operation throughout the engine's flight envelope. References 4 to 7 document in detail the complete program AFAPL CONTRACTING TECHNICALMONITORING ENGINEMODELS MULTIVARIABLECONTROLDESIGN CONTROLCRITERIA ENGINETEST SUPPORT ENGINETEST SUPPORT Figure 3 STRUCTUREOF FlOO MULTIVARIABLE CONTROL The LQR-based control logic was designed to meet the following criteria. Prima- rily, the logic must protect the engine against surge and maintain speeds, pres- sures, and temperatures below maximum limits. Airframe-engine inlet compatibility requires adhering to minimum burner pressure limits and maximum and minimum air- flow limits at certain flight conditions. The control must keep thrust and spe- cific fuel consumption within tolerance for specified engine degradations. The engine must accelerate and decelerate rapidly and repeatably and must remain stable in the presence of external disturbances. The basic structure of the FlOO multivariable control logic is shown in figure 4. The five manipulated engine inputs are fuel flow, exhaust nozzle area, inlet guide vanes, compressor variable Primary sensed engine outputs are fan geometry, and compressor exit bleed airflow, speed, compressor speed, main burner pressure, afterburner pressure, and fan tur- bine inlet temperature. Basic components of the control are: 1) reference point schedules and transition control logic, which produce desired state and output and approximate control vectors, 2) gain schedules, which produce feedback matrix elements as functions of flight conditions, 3) proportional and integral control loops which produce acceptable steady-state and transient engine behavior without operating limit exceedance, and 4) engine protect logic that places absolute limits on engine inputs to assure safe operation in the test cell despite sensor or logic failures.
I I PT2 x, ENGINESTATES l-r2 I NS ” r 1 y, ENGINEOUTPUTS INTEGRAL 31 L
7 GAINS
LIMITS I 3 --- - INTEGRALTRIM - Figure 4 CCNTROL SYSTEM SCHEMATIC FOR ALTITUDE TESTS The MVC logic shown in figure 4 was programmed in fixed-point ASSEMBLY language on a l&bit minicomputer and debugged while controlling the LeRC real-time The simulation was a nonlinear, component-level repre- hybrid engine simulation.
sentation of the engine and included lumped-volume and rotor dynamics. It accu- rately represents the engine's operation across the entire flight envelope. After the same logic was used to conduct the alti- completion of the hybrid evaluation, tude test evaluation, as shown in figure 5. Bill of material (BOM) hydromechani- cal engine actuators were modified to allow input of electrical commands from the and suitable research sensors were provided for feedback signals.
minicomputer, Portions of the BOM control system were retained to serve as backup control and for engine start-up. A complete steady-state and transient evaluation was per- The LQR-based control logic performed formed over the entire flight envelope.
In addition, the real-time simulation was used periodi- well at all conditions.
cally during the altitude tests for rapidly solving any logic problems encoun- tered. The modular interface system between the computer and simulation or engine greatly facilitated this mode of operation.
8x6 HYBRID FACILITY ALTITUDETESTFACILITY (PSL) FlDD ENGINE
(
XD-II I r A RESEARCH 1 BACKUP SENSED ACTUATORS a SYSTEM > - I --- I/ ACTUATOR COMMANDS Figure 5 TYPICAL FlOO MULTIVARIABLE CONTROLPERFORMANCE IN LARGE PLA TRANSIENT Figure 6 shows an FlOO engine transient response test performed during the alti- tude tests at a simulated altitude of 10,000 feet and Mach number of 0.6. The input is a power lever angle step (snap) from 50° to 83O (maximum, non-after- burning). This transient caused a number of MVC logic functions to be exercised: transfer from fan speed integral control to fan turbine inlet temperature (FTIT) limit control, regulator and integral control gains being varied as functions of compressor speed, use of the FTIT estimator output, and control of the exhaust nozzle area to control fan discharge AP/P, a fan air flow parameter. Initially, the control maintains the desired engine operating point by keeping fan speed and fan discharge.AP/P on desired schedules. During the transient, engine outputs generally follow their desired trajectories. Fuel flow is modulated to keep FTIT at or near its allowed limit during the initial portion of the transient. At steady state, the logic has closed down the nozzle area and trimmed fuel flow so that both fan speed and fan discharge AP/P are on schedule. This transient was one of over ninety performed using the multivariable control, with inputs being a wide variety of PLA trajectories, afterburner ignitions, and flight condition excursions. The MVCS program demonstrated that digital engine controls can be successfully designed using techniques based on LQR theory. As demands for engine performance lead to engine designs which have larger numbers of control variables, LQR methods will be increasingly useful for algorithm design.
ALTITUDE - 10 Ooo ft, MACH a6, 500 TO 830 PlA SNAP - SENSED - - -- SCHEDULED -.--- LIMIT ------ ESTIMATE FAN DISCHARGE, . 163 ftZ 28 tr I I I I I I I FUEL FLOW. t!&---- lblhr ’ D t .-l-l.. .l I I i I TIME, set PM SNAP Figure 6 AESOP - INTERACTIVE COMPUTER-AIDED CONTROL SYSTEM DESIGN A tool that was developed at Lewis and used during the FlOO MVCS program to verify and update the contractor's multivariable control designs was the AESOP computer program (Algorithms for Estimator and Optimal regulator design). An interactive program which solves the LQR and Kalman filter design problems for time invariant systems, it is an outgrowth of an earlier batch program LSOCE (ref. 8). As shown the user typically accesses AESOP by using a light pen to select in figure 7, desired AESOP functions from a menu displayed on a terminal screen. Available functions fall into the categories shown: open-loop system analysis (controll- ability, observability, eigenvalues, etc.), LQR and Kalman filter design, system response to noise inputs (both open- and closed-loop system covariance matrices), for step and initial condition system transient responses (open and closed loop, and transfer functions and frequency responses (system poles, zeroes, and inputs), generation of open- and closed-loop Bode plots).
Graphic output can be produced either at the terminal screen or plotted off-line. AESOP also aids the user by A user's manual has been checking the validity of requested function sequences.
prepared (ref. 9).
AESOP
I
TRANSFER FUNCTIONS OPEN-LOOP AND FREQUENCY SYSTEMANALYSIS SYSTEM RESPONSE TO NOISE INPUTS Figure 7 SENSOR FAILURE DETECTION AND ACCOMMODATION The relative immaturity of digital electronics compared to hydromechanical con- trols has raised concerns with respect to control system reliability. Past stud- ies have shown that the least reliable parts of a digital electronic engine con- trol system are the sensors. For this reason, as a follow-on to the FlOO MVCS program, the FlOO DIA program (sensor failure detection, isolation, and ficcommo- dation) was initiated to develop algorithms for enhancing-digital control system reliability by using analytical redundancy. The general concept of analytical redundancy in the context of an engine control system is illustrated in figure 8. Assume that the reliability of turbine inlet temperature sensor T4 is insufficient to meet mission-reliability goals. The normal procedure (hardware redundancy) would be to add two additional temperature sensors and voting logic to determine if and when a sensor has failed. The analytical redundancy approach is to incorporate a model of the engine which relates T4 with other sensed engine variables (for example, P4 and N). The model is then used to generate an estimate of T4 which can be used with a statistical testing pro- cedure to detect and isolate a T4 sensor failure. Once a failure has been the estimate can be used to replace the failed sensor and allow accept- detected, able but possibly degraded control performance. In the FlOO DIA program, con- tractors P&WA and Systems Control Technology (SCT) have: 1) quantified sensor failure types and frequency of occurrence, 2) determined the relative criticality of various failures, and 3) developed a sensor failure DIA algorithm for the FlOO engine (ref. 10).
T4 III i I i N P4 ANALYTlCAL REDUNDANCY HARDWAREREDUNDANCY 0 1 SENSOR---T4 0 3 SENSORS+T4 0 REFERENCE MODELDETECTS FAILURE 0 MAJORITYVOTE DETECTS FAILURE 0 REDUNDANT INFORMATKIN (P4, N) GENERATES T4, EST SENSOR FAILURE DETECTION, ISOLATION, AND ACCOMMODATION CONCEPT The structure of the FlOO DIA logic and the manner in which it interfaces with the MVC logic are shown in figure 9. The two portions of the DIA logic are an on-line state estimator and detection and isolation logic. The estimator incorporates an full envelope model of the engine which is updated in real time. Esti- accurate, mator residuals are continuously monitored by the detection logic, and if a failure isolation and accommodation algorithms are initiated. During normal is detected, an estimate of state x is fed to the LQR gain portion of the unfailed conditions, multivariable control and the integrally controlled engine output variablesy are sent directly to the integral control portion. This insures that any possible bias in the estimate of y will not cause a shift in the desired engine operating point.
Once a failure has been detected and isolated, the on-line estimator is reconfigu- red to exclude the bad sensor, and an estimate of the failed sensor signal is sent to Due to the modular design of the MVC logic, the only change the integral control.
required after addition of the DIA logic was to eliminate the FTIT estimator, a task now taken over by the DIA logic.
U X ’ Y FloO D ) SENSORS * ACTUATORS ENGINE t ----_-.-_--m--w- ----_--_--_._ r I Dl3ECTION AND I MULTIVARIABLE I Icnl fi-r~nhl I nrlr 13ULnI IUIY LUUIb I CONTROL LOGIC I I , ACCOMMO- i .s RESIDUAL / ENGINEPROTECT BATION -....-.- -----------i- / 4 ; LQR --------s--m- e i ESTIMATOR INTEGRALCONTROL --------------- TRANSITION CONTROL
‘- L------------ -__--__-_- ------A
DIA. LOGIC PLAJ FLIGHTCOND.-I Figure 9 DIA LOGIC SEQUENCE The specific algorithm used by the FlOO DIA logic begins with the generation of residuals for each of five engine measurements using an estimator which is Detection and isolation algorithms designed to work with all sensors present (fig. 10).
are'different, depending on whether a hard (out-of-range), sudden within-range shift) or soft (slow drift, slowly increasing noise intensity) failure is being detec- A hard failure is detected (and isolated) by simply comparing the sensor ted.
residual against a threshold. A soft failure is detected by computing the weight- ed-sum-squared of all residuals for N past observations and comparing that value .
against a threshold. To isolate a soft failure, a generalized likelihood ratio hypothesis test is performed, using residuals computed by a bank of five "off- line" to compute which sensor is most likely to have failed. Each of estimators, the five off-line estimates has one sensor input left out. The acconnnodation procedure consists of reconfiguring the on-line estimator by changing the gain matrix and omitting the bad sensor signal from its input plus resetting of the estimator's initial conditions. The FlOO DIA logic has been successfully evalu- ated on a detailed non-real-time engine simulation while coupled to the MVC logic. The logic has been coded for a Lewis-developed microprocessor-based com- puter control facility and will subsequently be evaluated while controlling a real-time hybrid simulation.
(1) GENERATE RESIDUALS WITH ON-LINE ESTIMATOR (2) DETECTFAILURE - HARD : RESIDUAL > THRESHOLD - son : WEIGHTEDSUM-SQUARED RESIDUALS (WSSR) > THRESHOLD (3) ISOLATE FAILURE Wfl-H GENERALIZEDLIKELIHOOD RATIO TESTS - HARD : ON-LINE RESIDUAL TEST -SOFT : OFF-LINE USING BANK (5) OF ESTIMATORS (4) ACCOMMODATE FAILURE - RECONFIGUREON-LINE ESTIMATORTO EXCLUDE BAD SENSOR - REINITIALIZE ESTIMATOR Figure 10 SENSOR FAILURE DIA LOGIC IMPLEMENTATION The Lewis computer control facility being used in conjunction with the FlOO DIA program is based on an Intel 8086 16-bit microprocessor and replaces the minicomputer facility used during the FlOO MVCS program. It includes both an interface unit which allows information interchange between a simulation or an engine and a moni- toring unit which displays and records information during a test. Figure 11 shows the basic configuration of the facility for implementing the DIA logic.
Rapid control update requirements led to the use of two microprocessors, one for the control logic and one for the DIA logic, which communicate through interrupts and transmit data over a multibus. Specialized D/A and A/D allow communication with the engine, also through the bus. The timing diagram in figure 11 shows the sub- tasks performed by each processor and the inter-processor interrupt timing. The MVC processor processes the sensed inputs and sends an interrupt to begin the DIA processors's detection algorithm. Both processors then operate in parallel until the Kalman filter (estimator) has updated the state estimates, at which time an interrupt from the DIA processor allows the estimates to be used by the MVC proc- essor in the multivariable control calculations. The DIA processor then searches for a possible soft failure, and only if detected, begins the isolation calcula- tion, requiring the updating of five additional Kalman filters and the use of the GLR algorithm. Languages used in programming the logic were both ASSEMBLY lan- guage and FORTRAN with special machine language procedures developed for certain time-critical tasks.
----- ~TCOMPUTER SYSTEM: PLA ..-- -_-.. -- AID X I ENGINE PROTECT7
4 4
\ LQRAND PROCESSTRANSITION INTEGRAL INPUTS CONTROL CONTROL
J
i - I l L-ONE CYCLE
-I
TIMING DIAGRAM Figure 11 FUTURE EFFORTS IN ADVANCEDPROPULSION CONTROL Advanced control related activities which are or will soon be underway at LeRC are outlined in figure 12. The FlOO DIA logic will be evaluated on a real-time engine simulation throughout the flight envelope and will then be tested in the LeRC An operating-point control design will be performed for the 'altitude facility.
FlOO engine using an alternate frequency-domain method (the multivariable Nyquist Also, the use of modern robust array) and compared to the existing MVC design.
control methodology to design engine controls will be investigated with an eye toward decreased sensitivity to modeling errors and simplified control algo- rithms. Contracts have just been awarded to investigate how best to incorporate Finally, robustness into the initial design of sensor failure DIA algorithms.
in-house computer-aided control design capability (such as the AESOP program) will be enhanced so as to be better able to interactively design and analyze multi- variable control and failure detection algorithms for future propulsion systems.
FUTURE EFFORTS IN ADVANCED PROPULSION CONTROL
l EXPERIMENTAL EVALUATIONOF DIA LOGIC
. MNA - LQR DESIGN METHODCOMPARISON
l ROBUST CONTROLDESIGN APPLICATION
l ROBUST SENSOR DIA LOGIC
l ENHANCEDINTERACTIVECAD FOR PROPULSION CONTROL
SYSTEMS
Figure 12 Michael, G.J.; and Farrar, F.A.: Development of Optimal Control Modes for 1.
Advanced Technology Propulsion Systems. United Aircraft Corp., East Hartford, CT, UARL-N911620-2, May 1974.
Weinberg, M.S.: Multivariable Control for the FlOO Engine Operating at Sea 2.
Level Static. Aeronautical Systems Division, Wright-Patterson Air Force Base, OH, ASD-TR-75-28, Nov. 1975.
Failure Accommodation in Digital Flight 3. Montgomery, R.C.; and Price, D.B.: Control Systems Accounting for Nonlinear Aircraft Dynamics. Journal of Aircraft, vol. 13, no. 2, Feb. 1876, pp. 76-82.
DeHoff, R.L.; Hall, W.E., Jr.; Adams, R.J.; and Gupta, N.K.: FlOO Multi- 4.
variable Control Synthesis Program - Vol. 1. Development of FlOO Control Systems. Systems Control, Inc. (VT), Palo Alto, CA, AFAPL-TR-77-35-Vol.-l, June 1977.
Szuch, J.R.; Soeder, J.F.; Seldner, K.; and Cwynar, D.S.: FlOO Multi- 5.
variable Control Synthesis Program - Evaluation of a Multivariable Control Using a Real-Time Engine Simulation. NASA TP-1056, 1977.
6. Lehtinen, B.; Costakis, W.M.; Soeder, J.F.; and Seldner, K.: Fl.00 Multi- variable Control Synthesis Program - Results of Engine Altitude Tests.
NASA TMS-83367, July 1983.
7. Soeder, J.F.; FlOO Multivariable Control Synthesis Program - Computer Implementation of the FlOO Multivariable Control Algorithm. NASA TP-2231, 1983.
Geyser, L.C.; and Lehtinen, B.: Digital Program for Solving the Linear a.
Stochastic Optimal Control and Estimation Problem. NASA TN D-7820, March 1975.
Lehtinen, B.; and Geyser, L.C.: AESOP: An Interactive Computer Program for 9.
the Design of Linear Quadratic Regulators and Kalman Filters, NASA TP-2221, 1983.
10. Beattie, E.C.; LaPrad, R.F.; Akhter, M.M.; and Rock, S.M.: Sensor Failure Detection for Jet Engines Final Report. NASA CR-168190, May 1983.
L-1011 TESTING WITH RELAXED STATIC STABILITY J. J. Rising and K. R. Henke Lockheed California Company Burbank, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 Wind tunnel and flight tests indicate that fuel savings of 2 percent can be achieved by c.g. management for an L-1011 with the current wing configuration. The normal c.g. location is at 25 percent MAC as shown in figure 1. The maximum fuel saving occurs for a c.g. location of 35 percent MAC. However, flight at 35 percent requires that the c.g. range be extended aft of the 35-percent point. Flight at c.g; locations aft of 35 percent requires a pitch active control system (PACS) so that handling qualities are not significantly degraded. Figure 1 shows that the near- term PACS was flight tested with the c.g. at 39 percent MAC.
DRAG BENEFIT = .lM (L/D) PITCH AUGMENTATION REQUIRED I NEAR-TERM
‘I
- TEST RANGE CURRENT WING -S/N 1001 (% C.G.)
(39) (41) i + 2 NEUTRAL POINT 23% AC.G.
/ CUYRENT RANGE - I STATIC STABILITY (%I Figure 1 The near-term pitch stability augmentation system consists of a lagged pitch The damper serves to provide rate damper with washed-out column feed-forward loop.
the necessary short-perfod frequency and damping characteristics while also suppress- ing turbulence effects, and the feed forward .was designed to "quicken" the pitch rate response and reduce stick force gradients without affecting system stability. A block diagram of the system is shown in figure 2.
-------------0-----------me- r , DIGITAL (80-HZ ITERATION RATE) I PREFILTER GAIN IA0 I I I 1 I I Ka 7hg!s+1 .03s+1 _ I E%m?)
----------------------w--B I r I I I SERVO POWER I LIMIT J CURVE SERVO +1.2 I 1.25 AIRFRAME -+ ’ -A==-( I 4.8 (%a I ca> (&gL) I I L,------em 7 FEED-FORWARD GAIN I LOOP i ‘, ; ’ r”;;l( ;“N::E:) I (A%$ I I ,WASHOUT,+ LAG ) PREFILTER I . 6 COL I .4 .nh NOTES: ‘ANALYZED WITH AND WITHOUT WASHOUT.
I / \ “ANALYTICAL APPROXIMATION FOR SAMPLE AND HOLD I L ------------A AND TRANSCRIPT DELAYS IN DIGITAL COMPUTER.
Figure 2 The schedules of pitch rate feedback gain and lag time constant were chosen so that the augmented stability L-1011 has good short-period frequency and damping characteristics for the complete center-of-gravity range (25 to 39 percent iZ) investi- gated in flight test.
The gain and lag schedules are defined as a function of cali- brated airspeed. Characteristics with and without augmentation are shown in figure 3.
= 0.83, h = 33,000 ft, W = 360,000 lb M AILERONS ACTIVE ----AILERONS INACTIVE AUGMENTED &j-SEC-’ C.G. =,,%E+ 1’ UNAUGMENTED ho,, - SEC -’ Figure 3 In cruise, the near-term augmentation system without feed-forward compensation provides slightly higher maneuver stability column force gradients at the relaxed static stability aft limit than the basic airplane has at mid c.g. without augmenta- tion. The effect of feed-forward compensation is to reduce the column force gradi- ents to levels comparable to the basic unaugmented airplane at typical c.g. locations.
The initial force.gradients comply with MIL-F-8785C (ref. 1) requirements; however, there is a deviation from linearity which results in a short-term negative gradient in the 1.6- to 2.0-g range which is unacceptable in terms of MIL-F-8785C requirements.
While these negative gradient character- These characteristics are shown in figure 4.
istics may not be desirable, they are not uncommon to Class III transport configura- It is suggested that some narrow tions which cruise at Mach numbers above 0.8.
region of negative gradient may be acceptable as long as a substantial force level is maintained and there is adequate buffet onset or otherwarning prior to limit load factor.
M= 0.83, h =
33,00 ft, W = 360,000 lb
MAXIMUM BUFFET ACTIVE AILERONS \/ PITCH FEED . DAMPER FWD OFF OFF OFF OFF ON OFF ON ON
LOAD FACTOR-g’s
Figure 4 The near-term pitch active control system was ultimately evaluated in an actual flight test program utilizing the Lockheed in-house research airplane, L-1011 S/N 1001.
The near-term PACS flight test aircraft is fully instrumented for performance This aircraft has been equipped with extended wing tips of flight test programs.
and an aileron active control system. The increase in wing aspect ratio increases the aerodynamic efficfency of the aircraft, and the aileron active control system provides wing load alleviation which results in lower structure weight. This approximately a increased aerodynamic efficiency and lower structural weight provide A unique feature of the basic L-1011 longitudinal control 3-percent fuel savings.
system is the flying stabilizer with a geared elevator. Modifications made to the L-1011 for the near-term PACS program are shown in figure 5.
Three pilots participated in the evaluation, which concentrated on two cruise The altitude conditions and an overspeed condition (VMC) at high dynamic pressure.
test program covered 47 hours of flying, 14 for flutter clearance, and 33 for flying Flying qualities were evaluated at c.g.'s from 25 to 39 per- qualities evaluation.
cent E, where the airplane with active ailerons is close to being neutrally stable (1 to 2 percent static margin) at high-altitude trim conditions.
EXTENDED WING EXTENDED WING ACTIVE CONTROL ACTIVE CONTROL AILERONS AILERONS TIPS TIPS STABILIZER STABILIZER TRANSFERABLE WATER TRANSFERABLE WATER BALLAST SYSTEM BALLAST SYSTEM Figure 5 The pitch active control system utilizes a series servo which connects to the pitch control system in such a manner that the PACS commands are fed into the stabilizer power actuators without reflecting these commands into the pilot input control column. This is shown in figure 6.
PILOT INPUT )PILOT SERVO SERIES SERIES --_-- SERVO Figure 6 Flight test evaluation of the L-1011 with the near-term PACS operative shows a significant improvement in pilot ratings resulting in acceptable ratings to the flight test limit of 39 percent MAC which was 1 percent from the neutral stability The variations of pilot ratings with center-of-gravity locations with and point.
without augmentation are shown in figure 7.
PILOT W UNACCEPTABLE A,lLERONS ACTIVE SMOOTH AIR 43
--------_-------___----_---- LL-
UNSATISFACTORY SATISFACTORY
l-
32 34 36 38 40 42 44 24 26 28 30 CENTER OF GRAVITY - ‘K-c Figure 7 The fuel saving benefits to be gained by c.g. management are highly dependent on the aircraft wing configuration. Figure 8 shows that for advanced wing concepts aft movement of the c.g. location may result in fuel savings up to 4 percent. This 4 percent is based on increased aerodynamic efficiency determined by analyses and wind tunnel tests. The maximum benefits occur when the c.g. location is at approxi- mately a -lo-percent static stability margin. An advanced PACS that has a reliabil- ity of 10-q is required for flight at this negative stability margin.
PITCH -AUGMENTATION REQUIRED
ADVANCED
CIJRRFNT I
--....I.-.
C.G. RANGE
DRAG*
CURRENT
t
E 3ENEFIT
WING
(%) ‘5”
I I NEUTRAL I
POINT
*AM (L/D)
I I I I V I I 10 0 20 -10
STATIC STABILITY(%)
Figure 8 Figure 9 shows the control model used in an eigenstructure placement control law synthesis technique.
With the feedback matrix [F] closed, the state-space equation becomes {ii> =[ A + BFC] lx} + {Z) [F] is computed such that the eigenvalues and eigenvectors In the synthesis process, of [A + BFC] at any c.g. condition are nearly the same as those for [A] with the c.g. at 25 percent is.
M = rJ$ KUsz Ki, 3
incremental pitch attitude, This set of gains operates on the four feedback signals: incremental speed, normal acceleration, and pitch rate.
I I {;l=[A+BFC](x/+{z} Figure 9 Plots of the feedback gains indicated that gain scheduling could be expressed as polynomial functions of dynamic pressure q and of horizontal stabilizer trim position 6hT. Comparisons of least-mean-square (LMS) values determined the order The second-order polynomial of polynomials to be used for each set of curves.
format + esHT2 K = a + bq + cq2 + dsHT Pseudo-inverse matrix was found to provide satisfactory curve fits for all gains.
operations efficiently determined the coefficients to yield the least-squares fit for each set of gain values. Scheduled feedback curves of the pitch rate input are shown in figure 10.
-0.08 CASES DYNAMIC PRESS 0 3 254 A 4 461 -0.32 l 461 .
0 6 203
K8
0 7 218 Q 8 234 - SEC.
9 258 v 0 255 0 11 285 -0.80 -1.20 -0.40 -2.00 1.60 -2.80 ‘HT- DEG Figure 10 Figure 11 shows conglomerate short-period poles on the S-plane for all flaps-up Nearly all closed-loop poles fall within the objective flight conditions.
boundaries: l Short Period Mode an 5 2w, 0.W 5 0.5 5 5 5 0.8 where w. is the short-period frequency of the corresponding open-loop configuration at 25 percent E. A few of the short-period damping ratios are higher than 0.8 but are considered to be acceptable.
YY -!- V 1.60 * 1.2;
jWdh
0.80 SEC-’ /\ I- 0.40 I I 0.00 tr -1.60 -1.20 -0.80 -0.40 0.00 0.40 bN” SEC-’ Figure 11 Control of the phugoid mode requires a velocity component, Ku u, in the feedback signal. Because of frequent velocity changes associated with changing trim conditions, however, use of a velocity sensor is undesirable. Knowledge of the transfer functions implicit in the equation {xl = [IS-Al-l [B] cu) provides the transfer function relating u and 8, as shown in the figure.
The com- posite signal B can be closely approximated by passing 0 throeh a lag-lead network as shown in figure 12 resulting in the following: K3 (y+l) B -= b2s+l) I 1
KIGl
e
U b b K”
K2G2
P
b
b
Ke
Figure 12 Figure 13 shows conglomerate phugoid poles on the S-plane for all flaps-up flight conditions.
Nearly all closed-loop poles fall within the objective boundaries: l Phugoid Mode wn 5 0.16 (40-second period) 0.2 I < 5 0.5 0.16(
h-q”
0.080
SEC-1
- SEC-’
c(4.l
Figure 13 With the feedback and feed-forward loops closed, the system equation is (;r) = [A + BFC] {x) + D(w) where {w) is the input vector comprising column displacement.
Figure 14 shows the PACS control model with feed-forward loop closed.
x0 I I {iii = [A + BFC](x/+ [D] (WI Figure 14 The feedback transfer function KFbGFP(S) is a composite of all four feedback signals, as shown in the figure. It should be noted that the total transfer function from the input 6, is the same regardless of which signal is designated to be fed back. This transfer function is J&NNz(S) + KuJ!Ju(S) + F&,(S) + K&(S) D(S) The frequency variant part of this transfer function can be simplified for the purpose of computing GFF(S), because the PACS is not sensitive to this function.
Simplifying assumptions permitted the approximation: = l/(S/wsp + 1) GFF A block diagram of this logic is shown in figure 15.
FEEL FEED-FORWARD AIRCRAFT SPRING TRANSFER TRANSFER FUNCTION J CURVE FUNCTION 6H J’ b KAGA - KFB GFB 4
FEEDBACK
TRANSFER FUNCTION Figure 15 The secondary gain scheduling is used to compensate for: l Pitch-up at high-math/high-g flight conditions l Outboard aileron symmetric effects when the aileron active control system (AACS) is activated.
The secondary gain scheduling causes the system "to think" it has an additional aft increment by adding an increment to the stabilizer gain-scheduling signal, This occurs when the pitch-up phenomena or the AACS mode is sensed in accord- ance with the block diagram of the secondary-gain controller. The increment produced by the bank angle (Ce6PT2(l - cos 0)) is designed to "straighten" the force gradi- ents during high-g turns. The secondary gain controller block diagram logic is shown in figure 16.
0.25 6a + MACH Krll ’ tL ’ 2 LL 4 6 0.7 0.8 a MACH AACS; FLA + 10 + 0.33O ON &AACS + * ) 6HT * 20s + 1 TO GAIN SCHEDULERS Figure 16 Figure 17 shows that the advanced pitch active control system with pitch excursion compensator completely removes the unstable dip in column dip force gradient characteristics. The data show that column maneuver force gradient characteristics The initial force gradients are essentially are satisfactory for all c.g. locations.
and they also fall in the middle of MIL-F-8785C the same for all c.g. locations, specified design limits (ref. 1).
I I I I 3 1.2 1.4 1.6 1.6 2.0 “‘s - LOAD FACTOR - g’s Figure 17 With the advanced pitch active control system engaged, the response of the airplane to a severe vertical gust (heavy thunderstorm magnitude of 54 fps) is shown in figure 18 to be essentially the same for all c.g.'s from 25 percent E (15 percent stable) to 50 percent E (10 percent unstable).
M = 0.83 = 0.54 HIGH (W/a) cL TRIM CG = 25%E ANGLE OF AlTACK - 4 - DEG “FRL I I I I I 12 16 20 0 4 8 TIME - SEC Figure 18 - The advanced PACS signals can be grouped into four categories.
l Feedback signals to provide desired stability l The feed-forward signal to provide the desired column-force gradient l Primary gain scheduling signals to compensate for flight condition changes l Secondary gain scheduling to provide additional stability compensation and force-gradient compensation during special flight conditions Figure 19 shows these signals as used in the advanced PACS.
SYMBOL I SIGNAL TYPE USE I I COLUMN FORCE FEED-FORWARD COLUMN FORCE FC GRADIENT I I I NORMAL ACCELERATION SHORT PERIOD Nz 8 PITCH RATE FEEDBACK MODE PITCH ATTITUDE PHUGOID MODE PRIMARY GAIN COMPENSATION HORIZONTAL STABILIZER SCHEDULING FOR FLIGHT CONDITI-ON FLAP SWITCH CHANGES ANGLE OF ATTACK SECONDARY GAIN COMPENSATION SCHEDULING FOR PITCH-UP * BANK ANGLE AND AACS MACH NUMBER M OUTBOARD AILERON OPERATIONS Figure 19 The advanced PACS is shown by the solid lines in figure 20. Inputs to the controller are: Feedback = Normal acceleration
NZ
8 = Pitch attitude i, = Pitch rate Feed forward = Column force FC Primary gain scheduling q = Dynamic pressure 6 = Stabilizer trim angle HT Secondary gain scheduling ~1 = Angle of attack = Bank angle G m = Mach number The controller output is provided in this planned flight test mechanization to two series servos which have position-summed outputs that limit hardover stabilizer deflections to +3/4 degree if one series servo fails. The total PACS authority is 21.5 degrees of stabilizer authority.
PACS AUTHORITY AT - 1 DEG TRIM SETTING k1.5 DEG __---- -- Figure 20 The purpose of the pitch-attitude synchronizer is to remove bias offsets from the pitch-attitude signal, and thus to avoid saturating the series servos. The integrator provides a reference level which tends to command the airplane as if it were in an attitude-hold mode. It has two modes of operation which are pilot optional: (1) fixed reference, or (2) controlled ref=rence, selected by the as shown in the circuit below.
switch S2, With S2 = C, the system is in the fixed reference mode and provides a reference attitude 8~ equal to the aircraft trim attitude 8T. A block diagram of the pitch synchronizer function is shown in figure 21.
- S, = T = (F + T) C S5 Figure 21 The advanced pitch active control system was evaluated on the Langley visual motion system (VMS) simulator pictured in figure 22. This is a general purpose simu- lator consisting of a two-man cockpit mounted on a six-degree-of-freedom synergistic motion base. Motion cues are provided by the.relative extension or retraction of the six hydraulic cylinders of ,the motion base.
Washout techniques are used to return the motion base to the neutral- point once the onset motion.cues have been commanded.
.
Figure 22 The handling qualities evaluation at each flight condition covered the c.g. range from 25 to 50 percent E, for which the control law was designed; this represents a stability range in cruise of 15 percent positive static margin to 10 percent unstable.
Results of the simulation in figure 23 show that the advanced flight control system completely fulfills the function for which it was designed.
Pilot ratings indicate that handling qualities of the augmented airplane with c.g. at 25 percent ;E (10 percent statically unstable) are as good as the basic unaugmented airplane with c.g. at 25 percent E (15 percent statically stable). The results are most impressive at high-speed conditions where handling qualities of the unaugmented airplane quickly degrade to unacceptable levels for c.g.'s aft of 40 percent E.
Flight condition 10 is an intermediate altitude cruise condition. At this con- dition, maneuvering stability about trim is essentially linear.
M = .83
CALM AIR
= 1.4 X lo6 LB) C‘ = .392 (w/g 0 PILOT 0 PILOT 1 4 PACS OFF l v PILOT 5 A PILOT 2 PACS ON 0 •I PILOT 3*
--- ------------------ g
----------
P
v
ii
J ---+-;-~~g- ---- 8--A----q z
V V
d
E
I$
A
B A
A s F s I I I I I I I I !4 28 32 36 40 44 48 52 56 60 CG -%C Figure 23
-
I
REFERENCE 1. Military Specification, Flying Qualities df Piloted Airplanes, MIL-F-8785C, November 1980.
AFTI/F-16 DIGITAL FLIGHT CONTROL SYSTEM EXPERIENCE Dale A. Mackall NASA Ames Research Center Dryden Flight Research Facility Edwards, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ,,. ,. ,.. _ . - .-..--- II, 8.8-8 I I- ABSTRACT The Advanced Fighter Technology Integration (AFTI) F-16 program is investigating the integration of emerging technologies into an advanced fighter aircraft. The three major technologies involved are the (1) triplex digital flight control system; (2) decoupled aircraft flight control; and (3) integration of avionics, pilot dis- In addition to investigating improvements in fighter per- and flight control.
plays, the AFTI/F-16 program provides a look at generic problems facing highly formance, integrated, flight-crucial digital controls.
An overview of the AFTI/F-16 systems is followed by a summary of flight test experience and recommendations.
AFTI/F-16 The Advanced Fighter Technology Integration, AFTI/F-16, is a joint Air Force, NASA, and Navy program.
The Air Force's objective is the integration of emerging technologies into a single test bed fighter aircraft.
The technologies include a triplex, dual fail/operate digital flight control system; decoupled flight control.; and integration of the cockpit functions, avionics, and flight control system.
NASA’s primary goals for the program were to assure safety of flight of the vehicle and provide an independent assessment of these advanced technologies. The primary contractor is General Dynamics, Fort Worth, Texas.
The AFTI/F-16 is a modified full-scale-development (FSD) F-16.
Most of the changes were made to the on-board electronic systems, flight control, and avionics.
However, two external modifications were made. Vertical canards were installed below the engine inlet to support decoupled aircraft control.
A dorsal falring was added to house additional avionics and instrumentation.
AFTI'S INTEGRATED TECHNOLOGIES The digital flight control system is the heart of the technologies integrated In the area of fault tolerance, the dual fail/operate system provides into the APTI.
dual fail/operate capability 95% of the time for computer CPU faults. Dual fail/ operate was a goal in addressing dual failures of sensors and discrete inputs for The control law design includes mission-specific and decoupled-, the APT1 program.
The primary goal is to improve aircraft survivability in an attack.
control modes.
Another goal to reduce pilot workload is approached by integrating avionics and flight control functions with cockpit displays. One selection of a mission phase switch configures avionics, weapons, and flight control to a specific mode.
The multipurpose displays allow for a centralized information display for all systems.
These main technologies resulted in two spinoff technology goals for the flight and a software-intensive design.
control system, asynchronous computer operation, Asynchronous operation is a major design characteristic which affects the dual fail/ operate capability, decoupled control law design, and the pilot's display of flight control status. The primary purpose of software-intensive design is to avoid addi- tional unique hardware in accomplishing the dual fail/operate requirement.
Dual fail/operate with avionics Standard 81decoupled Air-to-surface gun Software intensive design digital flight control system RELIABILITY AND FAULT TOLERANCE Experience with the AFT1 F-16 digital flight control system has highlighted the relationship of fault tolerance and reliability. The software-intensive design of AFT1 for aircraft control and fault detection emphasizes the role software plays in overall system reliability. Hardware reliability, based on the replication of hard- ware components, has been the only contributor to system reliability numbers and loss of control probabilities. The reliability of the software and the functions, such as fault detection algorithms, designed in software must be considered in overall system reliability.
Software reliabil- Software reliability continues to be difficult to determine.
ity for AFT1 was accomplished through test and configuration control. Adequate soft- ware reliability was determined indirectly after confidence in system operation was achieved by successful completion of verification and validation testing.
The relationship of fault tolerance to reliability comes through the software.
The proper detection of hardware component failures by the software is essential to support the reliability figures for the hardware. If a single failure which caused loss of control went undetected, the reliability for the system would,be greatly reduced. Reliability is a function of the fault detection software. Here again, reliability is assured through the testing process.
Total system reliability must consider the basic hardware architecture, the soft- ware testing process, and the fault detection algorithms which reside in the soft- ware. It is these last two considerations which will determine the validity of loss of control probabilities such as 1 X 10e7/hr.
SYS relia Hardware reliability self-test) l Architectural desig 9 Environmental test l EMI compatibility The AFT1 F-16 fault-tolerant design consists of two separate sets of checks to increase fault tolerance and provide the dual fail/operate capability. The first set The second set of checks consists of selection and fault detection on input signals.
of checks selects values and provides fault detection on the computers and actuators.
The purpose of the input signal selection and fault detection is to increase the system's ability to tolerate failures through cross-channel monitoring of the redun- Unique characteristics of the AFT1 design include: signal selection dant hardware.
based on the failure status of higher level devices, such as the inverter; an aver- a 15% failure threshold allowance; and reconfiguration of aging selection routine; control laws for unresolved dual-sensor failure.
Computer and actuator fault detection work together to provide the highest proba- bility for valid aircraft command in the face of single and multiple failures. The primary aspect is to accomplish the dual fail/operate capability centered on self- The self-test feature is responsible for identifying the test computer coverage.
last good computing channel when only two channels remain and their commands are not tracking. A unique characteristic of the output selector and fault detection design is the choosing of one computer's command , with a 15% failure threshold allowance, to This design choice resulted from asynchronous computer opera- control the actuators.
tion and interfacing constraints with the actuators.
In-flight Fault Detection Serial digital data from lefl Serial digital data from right Input selection and fauft detection Output selection and fault detection Selection: Use one channel’s value.
Selection: Goodsignal average Channel used is function of computer Fault detection: Goodsignal comparison; and hydraulic system failures tracking to 15% of full-scale failure Fauft detection: Goodsignal comparison; threshold tracking to 15% failure threshold Reconfiguration: Loss of a sensor’s Reconfiguration: Computer self-test data causes reconfiguration of run for dual failures.
control laws or default to safe value ASYNCHRONOUS COMPUTER OPERATION Although not intended to be a primary objective of the AFT1 program, the investi- gation of the asynchronous computer operation became a major activity. The asynchro- nous architectural concept started with the intent to increase EM1 immunity and over- all system fault tolerance. It was believed that concerns about asynchronous opera- tion (testability, data congruency, and nondeterministic operation) would be allevi- ated as the design matured. Considerable engineering effort went into designing and qualifying the DFCS with much being learned about asynchronous computer operation.
Despite considerable effort and improvements in the qualification process, concerns for testability remained because anomalies related to asynchronous operation occurred in flight testing. Asynchronous operation, coupled with the complexities of decoupled control and dual fail/operate capability, resulted in an increased design task, extended qualification period, and marginal testability. After envelope expan- sion, flight test evaluation of the DFCS for mission performance did not identify any new anomalies related to asynchronous operation.
Channel A Channel B Channel C 1Time Concerns Concept - Increase Immunity to EMIllightning - Random computer relationship, non.
deterministic - Increase computer channel independence - Incongruent data sets due to sampling - Increase fault tolerance over synchronized skew system - Testability; assuring rellable operation for all conditions Results - Deslgn task complicated by asynchronism - Qualification time extended, repeatability poor - Complex interactions due to sampling skews - Flight test operatlons affected due to marginal testablllty FLIGHT CONTROL 'MODE STRDCTURR 'The ART1 digital flight control.system consists of eight flight control modes, The four mission-specific categories four standard and four decoupled options.
i include normal, air-to-air gunnery, air-to-surface gunnery, and air-to-surface bomb modes. Mission-specific mode selection is accomplished through a mode panel or through hands-on selectors on the throttle. Mode selection configures both flight Decoupled mode options. are selected through a CCV lever on the control and avionics.
right-hand side stick controller. Decoupled o&ions include pointing, translation, and direct force in both pitch and yaw axes. The decoupled options also include enhanced maneuvering modes utilizing the pitch stick. The decoupled air-toyair gun mode provides an adaptive mode for the pitch stick, changing control structure based This allows for control optimization of gross acquisition and on pitch rate errors.
fine tracking in the air-to-air mission.
In addition to the advanced decoupled modes, reconfiguration modes are included to provide dual fail/operate capability for sensor failure; however, there is a loss A reconfiguration mode is derived from the of decoupled and mission-specific modes.
standard normal mode for dual failures of all primary feedback sensors, both longitu- dinal and lateral-directional. Reconfiguration modes use either synthesized sensor information or zero values as required. The digital system is backed up with an independent analog reversion mode. This provides protection for common mode errors which could cause loss of the digital system.
Decoupled Decoupled air.
Controllers normal to-surface bomb SP.Pitch stick SP.FPMOPRME SP-FPME - SP.FPME SRRoll stick SRRoll rate SR-Roll rate P.Pedals P.Translation P.Flat turn T.Throttle twist T-Translation I T-Direct lift I I FPME. Flight path maneuver enhancemen!
t PRME- Pitch rate maneuver enhancement Reconfiguration Standard normal to-surface gun to-surface bomb Normal act.
SP.Normal act.
Pitch rate - SP.Normal act. SP-Pitch iate SP.Pitch rate SR-Roll rate AOA SR-Roll rate SR-Roll rate P-Rud deflect P-Flat turn Roll rate T-None Yaw rate I --- .--- - pMZ-2 Landing t “$ gear Takeoff and Takeoff and land land + t Analog reversion mode (from any digital mode) DECOUPLEDCONTROLOPTIONS Decoupled control options consist of pointing, translation, and direct force for both pitch and yaw axes. Pitch axis modes use elevator and flap commands, and yaw axis modes use canards, rudder, differential elevator, and aileron commands. Point- ing allows for changing aircraft attitude without affecting flight path. Pointing angles are equivalent to changes in angle of attack and sideslip for each axis.
Translation commands a constant velocity without affecting aircraft attitude. Angle of attack and sideslip vary with the command. The direct-force modes command accel- erations, affecting flight path, while keeping angle of attack and sideslip con- stant. The evaluation of decoupled options centered around weapon effectiveness while increasing aircraft survivability.
t
P
“+--
a
“=&--d
“+=+ “4
P
v-4
Vertical kanslation Lateral translation Direct lift Direct sideforce
-L -L 1
VT‘- VT-. VT-
0 (I
a
Pitch pointing Yaw pointing An example of the interactions which occurred between the redundancy management functions and control law is illustrated below. Triple analog input sensors are sampled asynchronously and compared within each computer. If the inputs are within an established trip level, the control laws in each computer use an average value for Because the system is asynchronous, the average value used by the triplex inputs.
the control laws is slightly different in each computer. The complex control law amplifies the difference in generating an output com- structure, with its high gains, mand. The output command is monitored by each computer to assure that the differen- ces between output commands are within a given difference (that is, trip level). The amplified differences generated by the high-gain control law function cause nuisance If three output surface commands fail within failures in the output command monitor.
one computer, a channel is failed. This was particularly evident in the advanced and The control gains were reduced to pre- decoupled modes during ground qualification.
vent nuisance failures as a consequence of the redundancy management/control law interactions.
Triplex outpd from stick Time Tlme Sampled stick Selected stick Stick Pitch Input to computers Inputs for each output stick -computer A-B> E B-C > L output A-C> E monltor Time Control law gain Differences between Output monltor ampllflcatlon computer channels sees false magnlfled failures The digital flight control software design and test activities are summarized as The design, test, and redesign cycle is a top-down design with bottom-up testing.
accomplished through a configuration control/ discrepancy reporting process. Top- down design began with a function breakdown based on the design specification. The functional breakdown was carried directly into a structured software design. The lowest breakdown of the software is termed a unit and is required to have one exit and entry point and be less than 100 lines of code. Strict documentation, com- menting, and software design reviews were essential to the software design process.
Software testing took a bottom-up approach, beginning with unit testing. The module and the component testing were accomplished as the necessary units were integrated. Testing to this level is done by the software design team, with a final integrated software package going to an independent test group. The indepen- dent test group performed verification (proper software implementation of the functions) and validation (proper system level operation testing of the flight control system). Detailed specification and design documents were used by the test team to assure proper testing of all flight control functions. Details of the.
qualification can be found in references 1, 2, and 3.
The configuration control process provided the means to document discrepancies found in test and to correct the discrepancies in the software or hardware as needed.
The key to the process is a systems-wide approach covering control laws, fault tol- erances, avionics, hardware, and software. Interdisciplinary knowledge and resolu- tion of problems are essential in such heavily integrated systems.
Top-down design Bottom-up test Total s/w package/ Specification I I independent validation J Functional breakdown J I Computer components 1 Computer components/ 1 (highest software function) programmers
I
I s Modules 1Module test/programmers1 J Units Unit test/programmers I I A FLIGHT TEST RESULTS: GENERAL Flight testing of the AFT1 F-16 was successfully accomplished over a 13-month period in 1982 and 1983. A total of 118 flights were flown by pilots from four organizations: Air Force, NASA, Navy, and General Dynamics.
All major objectives including envelope expansion for high angles of attack and Mach num- were completed, combat mission evaluations of decoupled control, and structural load bers up to 1.2, clearance for the decoupled motions.
Low flight rates early in the program were due to anomalies of the basic aircraft as well as to the AFT1 unique systems.
Thirteen software releases were made during flight test to the digital flight control system. Software changes were made to correct discrepancies and provide improvements in flying qualities, fault-tolerant operation, and structural-load-limit An efficient software change process is required to provide safe, timely items.
changes needed to accomplish flight test objectives.
T12 3o r T13 TXX I) Software release number
-
= T9 T7 7 4 9 8 20 14 27 JASONDJFMAMJ -T l 118 total fliahts by Air Force, NASA, Navy, and General Dynamics alots l Thirteen releases of flight control software l Full envelope expansion of three separate flight control modes l Air-to-air and air-to-ground mission evaluations of decoupled control options Built-in test (BIT) is a highly automated test sequence that assures the digital flight control system (DFCS) is free of hardware failures prior to takeoff. BIT is run prior to each flight and takes approximately 2.5 min. Two failures of the hard- ware were detected by BIT during flight testing. The first was a failure of the flap actuator, the second involved memory chips which didn't meet timing specifications at cooler temperatures. Nuisance failures of BIT occurred a number of times. The cause is believed to be EMI.
In-flight fault detection is accomplished by comparing the three values for tracking among the different channels. The only real failure was an input signal which was traced to a pushed-back pin in the aircraft wiring. The 15 false failures were due to design deficiencies rather than actual hardware failures. The design deficiencies, which resulted in both temporary (resettable) and permanent loss of flight control redundancy , were corrected in subsequent software releases.
The asynchronous computer architecture affected a wide range of developmental activities including design, software/system qualification, and flight test opera- tions. Initially DFCS qualification was not full proof because of the dependence of failure modes on computer skew. Testing at predetermined "worst case" computer skew improved testing results; however, some deficiencies still escaped detection. Ground operations during aircraft preflight were impacted by the asynchronous computer architecture. The most common problem resulted in DFCS failures, requiring reset by pilot or cycling of aircraft electrical power.
The false failures, not hardware induced, were the result of the design deficien- cies associated with asynchronous computer operation. The design deficiencies resulted from the coupling of unique computer skews with characteristics of the flight environment, such as sensor noise. Undetected during qualification, these in- flight failures resulted in envelope and flight control mode limitations until they were corrected by software changes.
The software configuration control process details the procedures equivalent to the maintenance procedures for hardware, but in the software environment. Maintain- ing safe, operational software requires specification, design, test, and documenta- tion for every change. Software change, specification, and time line for incorpora- tion directly involved flight test planning.
Testing and documentation provide details for operating characteristics and/or restrictions.
The 13 flight test software releases, in which design, coding, and test of the changes were performed at General Dynamics, Fort Worth, supported the needed changes for flight test.
The first four releases provided full envelope capability for the AFT1 vehicle in all flight control modes.
The remaining nine releases modified the control system's control laws to improve flying qualities and the fault-detection function to improve reliability.
Software errors are software design or coding errors which escaped detection by the software verification and system validation testing. Two of the errors resulted from a loosening of the detailed established procedures. The third error was caused by a coding error involved in the correction of a previous false in-flight failure condition. The error was found when the in-flight failure condition recurred. It was found that the configuration control process provided excellent software opera- tion in the face of constant change. The testing and documentation process, when strictly followed, will detect software design and coding errors.
-_-- I
Flight Test Results: Fault-Tolerant Design
And Software Maintainability
Fault tolerance Software maintenance l Built-in test l Software configuration control process - Details configuration/operating - Comprehensive, automated P-min preflight check restrictions - Some nuisance failures due to EMI - Provides specification, design, test and documentation process l In-flight fault detection - Interfaces to flight test planning - Fifteen false failures detected l Software changes, always - Three failures, cause not known - One real failure detected - 129 total changes in 13 separate releases l Asynchronous computer operation, problems - Changes to both fault-tolerantlcontrol- - Adversely affected DFCS qualification law designs - Accomplice to in-flight failures l Software errors occurred - Three errors found in flight releases - Process keeps serious errors from affecting flighj operations FLIGHT TEST RESULTS: CONTROL LAWS A primary objective of the AFTI F-16 program was the evaluation of a multimode digital flight control system with decoupled aircraft control. The six different decoupled options and right-hand control options were evaluated with the decoupled feature best suited for a given task identified.
The adaptive control law, which uses pitch rate error to optimize performance in gross acquisition and fine tracking , was shown to be the best option for the air-to- The adaptive gain control law was implemented using the right-hand air combat task.
controller; decoupled pointing with the pedals and twist grip showed no significant improvement for the air-to-air task.
The best feature for the air-to-ground task was again through the pitch stick with improved flight path stability and ride smoothness in turbulence. Direct side force or flat turn which is commanded through the rudder pedals improved the task by reducing pilot workload for obtaining lateral axis solutions.
Problems with roll ratcheting affected all the advanced combat modes. Prefilter tuning was not suf- ficient to completely resolve the problem.
The standard normal mode, configured for gear-down, provided a greatly improved mode for power approach. Using more of a pitch-rate command system versus the nor- mal acceleration command system on the F-16's, improvements in flight path and angle- of-attack stability were made.
The need for and design of the analog reversion mode to protect against common mode failures proved most interesting. Although the analog reversion mode (ARM) was never engaged because of digital system failure, flight test experience indicates ARMS are needed. Complexity of the ARM becomes a primary issue; a simple ARM cannot provide protection at envelope extremes which are possible with the digital systems.
Furthermore, the relaxed static characteristic requires a certain level of augmen- tation. The simplified reversion mode used on AFT1 provided get-home capability and level 2 flying qualities for landing as specified. However, simulation and flight test indicated a more capable ARM is needed to cover transitions from the envelope extremes possible with the digital control system (ref. 4).
Flight Test Results: Control Laws
l Air-to-air combat task - Longitudinal axis pitch stick control excellent based on adaptive, pitch rate error, gain system - Decoupled pointing showed no significant improvements for task completion l Airtoground task - Longitudinal axis pitch stick control improved, better flight path stability/ride quality - Flat turn, direct side force useful, simplified task, reduced pilot workioad - Roll ratcheting problem in all combat modes l Power-approach task - Longitudinal axis control improvement over F-16, better flight path/angle-of-attack stability, precise attitude control l Analog reversion mode - No automatic engagement of mode - Design of mode too simplified, reduced failure envelope, compromised flying qualities FLIGHT TEST RESULTS: SUMMARY Successful flight test accomplishments included envelope expansion for the pri- mary flight control modes for stability and control and for structural loads.
Envelope expansion included testing for low-speed, high-angle-of-attack, and high- speed-to-Mach-l.2 conditions. Evaluation of the advanced control modes, the final goal, was accomplished in the last 15 flights.
Advanced control options were evaluated in a variety of air-to-air and air-to- surface combat tasks. Advanced flight control modes for the right-hand controller gave the best performance of the decoupled control options; flat turn showed signifi- cant improvements over conventional control methods.
The asynchronous computer architecture proved to be one of the most interesting aspects of AFTI/F-16 because of its wide-ranging effects. The fact that a given architectural design feature can affect design, qualification, and flight test is noteworthy. Flight test was culminated with no fault-tolerant-type anomalies affecting flight test operations.
highly integrated control Representing a state-of-the-art, flight-crucial, system, AFT1 provided the opportunity to find weaknesses in the developmental process resulting from the new technologies. Design and testing tools to support the devel- opment of increasing complex systems need to be developed. The goal is to develop Tools that tools which can support a generic set of digital control applications.
assist in the design and testing of the fault-tolerant and software aspects of new highly integrated systems would prove beneficial.
l Successful accomplishment of flight test goals l Advanced control options improve aircraft performance l Asynchronous computer architecture gave difficulties, costly l Flight-crucial controls/integrated systems stress developmental process FUTURE CONSIDERATIONS The AFT1 F-16 program provided the engineering community another look at the development of a highly integrated flight-crucial system. Considerable knowledge was gained in the development and flight test of decoupled aircraft control, asynchronous computer operation, and flight-crucial software. As the dust settles and the results are reviewed, several areas for further consideration surface.
The time for a fault-tolerant system design tool has come. As any new technology begins to succeed and grow, design tools are needed to increase engineering produc- tivity and provide better, safer product designs. Information exists to develop a design tool which documents fault-tolerant designs and allows for systematic test approaches that can increase operational reliability. The tool would allow for early integration of the fault-tolerant design with the control law functions to avoid costly downstream changes. The design tool would essentially be an expert system which would help guide the engineer in the specification, design, and qualification of a fault-tolerant design.
The development and the use of software and system-level testing tools also need to be applied to the development of flight-crucial controls. By increasing test coverage, automating the testing process, and providing integral configuration operational reliability and development time could both be improved.
control, Further consideration is also being given to the primary AFT1 technologies. The use of AFT1 flight control laws in an automated-flight fire control system is one example. Increasing weapon delivery accuracy while increasing aircraft survivability will be a primary emphasis of AFTI's second phase, Automated Maneuvering Attack System (AMAS).
l Faulttolerant design tool - Design documentation - Systematic testing capability - Integration of control functions - Architectural studies - Expert system l Software/system testing tools - Increase/measure test coverage - Automate testing process - Provide configuration control l Decoupled control in automated attack l Increase survivability l Weapon accuracy REFERENCES 1. Mackall, D.; Ishmael, S.; and Regenie, V.: Qualification of Flight Critical AFTI/F-16 Digital Flight Control System, AIAA Paper No. 83-0060, Jan. 1983.
2. Mackall, D.; Regenie, V.; and Gordoa, M.: Qualification of the AFTI/F-16 Digital Flight Control System, Paper presented at the IEEE National Aerospace and Electronics Conference NAECON, 1983, Dayton, Ohio, May 17-19, 1983.
3.
Griswold, M. R.; Gordoa, M.; and Toles, R.: AFTI/F-16 DFCS Development.Summary - A Report To Industry, Verification and Validation Testing. Proceedings of the IEEE 1983 National Aerospace and Electronics Conference NAECON 1983, vol. 2, 1983, pp. 1236-1240.
4. Ishmael, S.; Regenie, V.; and Mackall, D.: Design Implications From AFT1 F-16 Flight Test, IEEE/AIAA 5th Digital Avionics Systems Conference, Seattle, WA, Oct. 31 - Nov. 3, 1983.
EXPERIENCES WITH THE DESIGN AND IMPLEMENTATION OF FLUTTER SUPPRESSION SYSTEMS Jerry R. Newsom and Irving Abel NASA Langley Research Center Hampton, Virginia First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 ABSTRACT A considerable amount of research has been conducted on the application of active Because of its impact on safety of controls to increase aircraft performance.
flight, flutter suppression is probably the active controls concept furthest from practical implementation and, therefore, requires significant attention. This attention spans both analytical and experimental studies. Research efforts at NASA have been directed towards the development of analysis and design methodology and the correlation of experimental results with analytical predictions. The purpose of this paper is to discuss some experiences with the design and testing of several flutter suppression systems. Emphasis will be on the experimental activities.
ACTIVE CONTROLSTECHNOLOGY The application of active controls technology (ACT) to reduce aeroelastic response of aircraft structures offers a potential for significant payoff in terms of aerodynamic efficiency and weight savings. To reduce the technical risk associated with this new technology, research was begun in the early 1970's to advance this concept. The technical.program encompasses three areas: control law synthesis, aeroservoelastic analysis, and experiments aimed at verifying both the analysis and synthesis methodology. In the area of control law synthesis, classical methods are being applied where applicable. The latest "state-of-the-art" optimal methods are being refined and applied to the aeroservoelastic case. Innovative approaches are being developed to take highly theoretical synthesis methods which result in complex (high- order) control systems and modify these methods in the design of simpler (low-order) control systems. Strategies are being developed to investigate the sensitivity of the resulting control systems to uncertainty and to incorporate this knowledge into the design cycle. Analysis methods include a comprehensive program (DYLOFLEX) (ref.
1) for calculating the loads on an aeroelastic vehicle equipped with active controls. The evaluation of vehicle static and dynamic stability is being accom- plished using programs and methods developed in-house at LaRC. The experimental pro- gram is aimed at validating the analysis and synthesis methods by comparison with wind tunnel tests and flight results using a remotely piloted drone. The flight test program, called DAST (Drones for Aerodynamic and Structural Testing) (ref. 2), has become the focal point of the experimental validation.
ACT -7;' CONVENTIONAL fi !I SYNTHESIS
II
u FUNCTION BENEFITS .FLUllER SUPPRESSION OREDUCEDWING LOADS l LOAD ALLEVIATION l LOWERWEIGHT @RIDE QUALITY l INCREASEDSPAN . STABILITY *HIGHER L/D AUGMENTATION 0 STABILITY I I ACTIVE FLUTTER SUPPRESSION Active flutter suppression is a concept to increase the flutter speed of a vehicle through the use of active feedback control. The active control system consists of (1) control surfaces, (2) sensors, and (3) control laws. Through proper selection of the control surfaces and sensors and design of the control laws, the damping of the The aeroelastic system can be augmented and thereby increase the flutter speed.
benefit to be derived from flutter suppression is usually reduced structural weight.
4---z
DAMPING
SPEED
CONTROL LAW DESIGN PROCESS The development of methodology to design flutter suppression systems has been an This development ranges from analytical syn- integral part of this research program.
thesis techniques to the overall control law design process. A flowchart of the The first element approach to the overall control law design process is shown below.
of the process is the selection of design objectives (i.e., gain margin, phase margin, etc.). The second element is the selection of a design point (i.e., Mach number and altitude). Control law synthesis is then performed at the design point.
The next element, analysis, provides information on the performance of the control law at off-design flight conditions. If the design objectives at the off-design then a gain scheduler which may he a function of Mach flight conditions are not met, number and/or dynamic pressure is evaluated. If a gain scheduler will not meet the design objectives, then a path back to control law synthesis is selected.
SELECT
DESIGN OBJECTIVES
I
OPTIMAL CONTROL LAW
.-
SYNTHESIS
[YES
CONTROL LAW SYNTHESIS The major emphasis in the development of design methodology has been in the area of control law synthesis. To accomplish the objectives of control law synthesis, as stated below, the practical problems in control law implementation must be The historical development of control law synthesis methodology has pro- recognized.
ceeded from classical techniques to optimal control theory to optimization techniques Both unconstrained and constrained optimization techniques have been W;,,,“,“’ l Beginning recently, emphasis is being given to the use of constrained optimization techniques since several design objectives can then be satisfied simultaneously.
l OBJECTIVE:
. DESIGN A LOW-ORDERCONTROL LAW FOR A HIGH-ORDER SYSTEM
TO MEET SEVERAL DESlGN OBJECTIVES
. LOW ORDER- SIMPLE IMPLEMENTATION
EH ORDER-CHARACTERISTIC OF AEROELASTICSYSTEMS
l METHODOLOGY DEVELOPMENT
l CLASSICAL TECHNIQUES
. OPTIMAL CONTROL THEORY/ORDER REDUCTION
. OPTIMIZATION TECHNIQUES
l UNCONSTRAINED OPTIMIZATION
. CONSTRAINEDOPTIMIZATION
WIND TUNNEL STUDIES Wind tunnel studies of aeroelastic models have been a cornerstone of the NASA research program. Presented in this chart are a number of models that have been used to demonstrate active control concepts on a variety of configurations. ,The Delta- wing model was an early experimental demonstration of flutter suppression (ref. 9).
The B-52 model was tested in support of a USAF/Boeing flight study on active controls Wing load alleviation was studied in support of a USAF/Lockheed program (ref. 10).
using a C-5A model (ref. 11). The DAST ARW-1 model was used for a variety of flutter suppression studies including an evaluation of a control system that would ultimately be tested on a remotely piloted research flight vehicle. Control laws were synthe- sized and tested on the model using classical, aerodynamic energy, and optimal methods (ref. 12). The F-16 and YF-17 model tests have shown active flutter suppres- sion to be a promising method for preventing wing/external store flutter (refs. 13 and 14). Use of active controls is especially attractive for fighters because of the multitude of possible store configurations. These studies are part of an Air Force Flight Dynamics Laboratory/General Dynamics/Northrop/NASA cooperative effort. A cooperative effort was also conducted with the McDonnell Douglas Corporation on a DC-10 derivative wing. Increases in flutter speeds in excess of 26 percent were demonstrated. This study is reported in reference 15.
DAST ARW-1 DELTA-WING MODEL Experimental studies have made a major contribution to the active control technology The Delta-wing model (whose photograph is in this program developed at NASA.
chart) was the first experimental demonstration of flutter suppression in this At a Mach number of 0.9, increases in the flutter dynamic pressure country (ref. 9).
One ranging from 12.5 percent to 30 percent were demonstrated with active controls.
of the major contributions of this wind tunnel program was the development of minia- These actuators paved the way for future wind tunnel tests ture hydraulic actuators.
To evaluate the performance of an active flutter of aeroelastically scaled models.
suppression (AFS) system, subcritical response techniques must be employed. Three different methods were used to determine subcritical response of the Delta-wing model, Analytical methods were used to predict and the results are described in reference 9.
and the results agreed reasonably well with both open-loop and closed-loop stability, However, for the closed-loop case, it was necessary to use a control the experiment.
surface aerodynamic correction factor that was derived using measured hinge moment data.
,DAST MODEL The aeroelastic model used for this study was originally built to support the DAST flight program (ref. 2). The objective of the wind tunnel study was to demonstrate a 44-percent increase in flutter dynamic pressure. Two control laws were designed (ref. 12). One control law was based on the aerodynamic energy method, and the other was based on the results of optimal control theory. At Mach 0.95, a 44-percent increase in flutter dynamic pressure was achieved with both control laws, thereby validating the two synthesis methodologies. Experimental results indicated, however, that the performance of the systems was not as good as that predicted by analysis.
The results also indicated that wind tunnel turbulence is an important factor in both control law synthesis and experimental demonstration.
SIGNIFICANCE I 0 44 % INCREASE IN FLUTTER DYNAMIC PRESSURE l VALIDATED SYNTHESIS METHODOLOGY a WiND-TUNNEL TURBULENCE EFFECiS ; ’ l HIGH-FREQUENCY CONTROL/ STRUCTURE INSTABILITIES CONTROL LAW PERFORMANCE An illustration of flutter suppression performance for the DAST model is shown below.
On the left, the spectrum of outboard peak accelerations for the wing with system off is compared to that for the system on. The model was being excited by tunnel turbulence. The data were measured at a dynamic pressure just below the system-off flutter boundary at M = 0.90. The decrease in amplitude and shift in the maximum response frequency resulting from the control law is evident. Also presented is a plot of flutter dynamic pressure as a function of Mach number. Data for both system off and system on are shown. The flutter suppression system is most effective (i.e., provides the largest increase in flutter dynamic pressure) at the higher Mach The effectiveness is significantly reduced at lower Mach numbers. A dis- numbers.
cussion of these results is given in reference 12.
CONTROL SYSTEMON
CONTROL SYSTEMOFF
Y g
0 CONTROL SYSTEMON
/zh, RiS () I I I I
CONTROL SYSTEMOFF
‘i. g
kPa
RMS I I I I 1 I
i ’ .6 .7 .8 .9 1.0
0 40 0
FREQUENCY, Hz MACH NUMBER
DC-10 MODEL A cooperative study was conducted with the Douglas Aircraft Company to apply control law design methods developed by NASA to a realistic transport configuration and to provide a rapid transfer of research technology to industry. These studies were an extension of previous wind tunnel tests performed by Douglas (ref. 16). The aeroelastic model (shown in the photograph on this chart) is representative of a wing which has a 4.27-m-span increase over the standard DC-10 wing.
Two control laws were designed at NASA Langley using different design methods (ref..
15). Both control laws resulted in a 59-percent increase in flutter dynamic pressure. The performance of the control laws as a function of gain and phase was also evaluated. Calculations performed prior to wind tunnel testing predicted all experimental trends. both structural damping and phase During the wind tunnel tests, characteristics of the actuator were identified as very important factors related to the effectiveness of the control laws. In addition, a correction factor was used to account for control surface effectiveness and did improve the correlation between measured and predicted characteristics.
!
P .’ ’ :o 58 # Il&tEASE IN FLUTTER 0 <DYNAMIC PRESSURE ‘: ^ , ‘-0 PER;ORMANCE AS A FLI~CTION OF I,“““: .,GAIN AND PHASE ‘, ^ .+ ANALYSis PREDICTED ALL _!,~XPERWIENTAL TRENDS ^I: < II i’, : “9 STRUCTUFjAL DAMPING EFFECTS * L ‘: x, :I & ACTljATOR DYNAMICS ^ ,‘. CONTkL SURFACE AERODVNAMIC .CORhECTION ’ ,,, ‘:A’, 2: MEASUREDAND PREDICTED STABILITY BOUNDARIES AS A FUNCTION OF SYSTEM GAIN AND PHASE FOR DC-10 WIND TUNNEL MODEL Measured and predicted stability boundaries in terms of flutter velocity versus system gain and phase are presented below. Three or four distinct flutter modes are exhibited, depending on phase angle.
For all phase angles analyzed, a decrease in flutter velocity is shown for mode 3 at low values of gain. At negative phase angles, the reduction in flutter velocity is more pronounced. The velocity at which mode 8 goes unstable is nearly independent of system gain and phase.
The mode 4 instability is aggravated by negative phase angles and stays relatively fixed for positive phase angles.
At phase angles of +20" and above, a new flutter mode result- ing from a coupling between the feedback filter mode and the first wing bending mode becomes critical. A detailed discussion of these results can be found in reference 15.
-CONTROLLAW NO. 1 MODE 0 ANALYSIS 0 3 0 8 0 EXPERIMENT 04 ti EXPERIMENT (NO FLUTTER) A FILTER >
$~~+~oo\~@~t) Ifly ;/yy
0 .5 1.0 1.5 0 .5 1.0 1.5 Kg 0 -5 do ls5 Kg .5 Kgl.o O 115 DAST: WHAT IS IT?
The concept of the DAST program (ref. 2) is to provide a focus for evaluation and improvement of synthesis and analysis procedures for aerodynamic loads prediction and design of active control systems on wings with significant aeroelastic effects.
Major challenges include applications to wings with supercritical airfoil and tests emphasizing the transonic speed range. The program requires complete solutions to real-world problems since research wings are fabricated and flight tested. Because of the risky nature of the flight testing, especially with regard to flutter, target drone aircraft are modified for use as test bed aircraft.
PRINCIPAL RESEARCH AREAS EMPHASIS
l ACTIVE CONTROLSYSTEMS EVALUATIONS . TRANSONIC REGION
. AEROELASTI C EFFECTS
. AERODYNAMIC LOADS MEASUREMENT
. STRUCTURALINVESTIGATIONS
. STABILITY AND PERFORMANCE STUDIES
-._- ------ DAST: HOW DO WE DO IT?
DAST uses an Air Force version of the Firebee II target drone as the basic test bed.
The standard Firebee wing is removed and replaced with the research wing of as depicted in this chart, involves an air The operational sequence, interest.
launch from beneath the wing of a carrier aircraft; a free-flight test phase of between 20 and 40 minutes (depending on Mach number and altitude); followed by a mid- During the free-flight air retrieval by helicopter via a parachute recovery system.
phase, a test pilot controls the vehicle from a ground cockpit. An F-104 aircraft is used as chase, and the copilot of this aircraft serves as a backup flight controller for the drone in case of a malfunction with the uplink system. Data from the experi- ments are provided in real time to the ground by means of a pulse-code-modulated telemetry system. Experimenters provide real-time assessments of the status of the This assessment is based on research wing and its associated active control systems.
the response of the wing to control surface sweeps and pulses. Flight tests are per- formed at the NASA Dryden Flight Research Facility located at Edwards Air Force Base, California.
, ,: ; .DRONETESTVEkitCLE : DAST: WHAT ARE WE DOING IT WITH?
Two transport-type research wings have been built for flight testing. The first wing, Aeroelastic Research Wing No. 1 (ARW-l), was designed for M = 0.98 cruise and was purposely designed to flutter within the flight envelope. Tests of the first research wing configuration have been terminated due to loss of the aircraft result- ing from vehicle systems problems. However, valuable flutter data and test technique experience were acquired. References 17-20 provide a description of these results.
The wing fabrication and test planning for the second research wing (ARW-2) have been sponsored by the NASA Aircraft Energy Efficiency program. This design involved what is believed to be the first exercise of an iterative procedure integrating aerodynamics, structures, and controls technologies in a design loop resulting in flight hardware. Evaluation of multiple active controls systems operating simultaneously, the operation of which is necessary to preserve structural integrity for various flight conditions, is the primary objective of the flight tests on this fuel-conservative-type research wing.
DAST RESEARCHWINGS ,,-ARW-2 ARW-1 l FLUTTERWITHIN FLIGHT ENVELOPE l ACTIVE FLUTTERSUPPRESSION SYSTEM l SUPERCRITICAL AIRFOIL ARW-2 . FUEL CONSERVATIVEWING DESIGN l HIGH ASPECT RATIO (AR = 10.3) l LOW SWEEP(A = 25O) l ADVANCED SUPERCRITICAL AIRFOIL l MULTIPLE ACTIVE CONTROLSCRITICAL TO FLIGHT OPERATION l FSS l MLA l GLA l RSS CORRELATION OF MEASUREDAND PREDICTED DAMPING AND FREQUENCYVARIATIONS (ARW-1) The frequency and damping of the dominant mode for the symmetric case are shown below. The analysis and flight test data are for a test altitude of 4.56 kilometers. The change in frequency with Mach number is predicted well for both the FSS-off and FSS-on cases. However, analysis overpredicts the damping for both the FSS-off and FSS-on cases.
The experimental flutter speed is extrapolated to be approximately M = 0.80 for the FSS-off case.
An actual flutter point was encountered for the FSS-on case at M = 0.82. Other data comparisons can be found in reference 17.
ALTITUDE =4.57 km: SYMMETRIC
2ot CL--~-~--O--
FREQUENCY,
0 0
Hz
t O "
A I I V
ACTUALFLUTTER POINT
0 A /
.04-
/
DAMPING, /
/
T .*08- 0 /
/
40 ANALYSIS TEST
.12 - 0' FSS-OFF 0
,/.
0,' 0 FSS-ON --- 0
I
A I
I
.16
y
.70 .80 .90
MACH
ARW-2 RIGHT SEMISPAN IN LAB PRIOR TO WIND TUNNEL TESTING The ARW-2 wing panels have been fabricated and are being used to support two ground tests. The left semispan has been used to conduct a hardware-in-the-loop test of the active control system electronics. The right semispan shown in the photograph below has been tested in the Langley Transonic Dynamics Tunnel to obtain unsteady pressure distributions. This is believed to be the first measurement of unsteady pressures on The pressure measurements from the wind tunnel will a flexible supercritical wing.
be compared against those measured during the flight tests. A secondary objective of the wind tunnel test is to investigate possible angle-of-attack effects on the flutter boundary at high transonic speeds.
CONCLUSIONS A large amount of expertise has been acquired through the analytical and experimental studies conducted to date. Many lessons have been learned that help guide the future research directions. A few of these lessons are shown below. The first three les- sons are technical in nature and have been or are presently receiving attention.
However, even though the last lesson is nontechnical in nature, it certainly needs to receive more attention. Several of the future thrusts listed below are being researched at the present. These include the use of transonic time plane unsteady aerodynamics, applying flutter suppression methodology to other active control functions, and synthesis of multiple active control systems. The other thrusts are not presently being emphasized but are still on the list of future work.
FUTURETHRUSTS
LESSONSLEARNED
l UNSTEADYAERO THEORYNEEDS l TRANSONIC TIME PLANE UNSTEADY
AERO
l CONTROLSURFACE
.ARBITRARY MOTION 0 SYSTEMATIC METHODSFOR
LOCATING CONTROLSURFACES
l ACCURATEDEFINITION OF AND SENSORS
ACTUATORDYNAMICS
l APPLY FLUllER SUPPRESSION
l ACCURATETURBULENCE MODEL METHODOLOGY TO OTHERACTIVE
CONTROLFUNCTIONS
l CLOSERCOOPERATIONBETWEEN
0 SYNTHESIS OF MULTIPLE ACTIVE
AEROELASTICIANAND CONTROLS
CONTROLSYSTEMS
ANALYST
0 CONTROLCONFIGURED VEHICLES
REFERENCES Perry, B. III; Kroll, R. I., Miller, R. D.; and Goetz, R. C.: DYLOFLEX: A 1.
Computer Program for Flexible Aircraft Flight Dynamic Loads Analyses With Active Controls. J. Aircraft, Vol. 17, No. 4, April 1980, pp. 275-282.
Drones for Aerodynamic and Structural 2. Murrow, H. N.; and Eckstrom, C. V.: Testing (DAST)--A Status Report. J. Aircraft, Vol. 16, No. 8, August 1979, pp.
521-526.
3. Newsom, J. R.: A Method for Obtaining Practical Flutter Suppression Control Laws Using Results of Optimal Control Theory. NASA TP 1471, Aug. 1979.
4. Gangsaas, Dagfinn; and Ly, Vy-Loi: Application of a Modified Linear Quadratic Gaussian Design to Active Control of a Transport Airplane. AIAA Paper No.
79-1746, August 1979.
5. Mahesh, J. K.; Stone, C. R.; Garrard, W. L.; and Hausman, P. D.: Active Flutter Control for Flexible Vehicles. NASA CR-159160, Nov. 1979.
6. Nissim, E.; and Abel, I.: Development and Application of an Optimization Procedure for Flutter Suppression Using the Aerodynamic Energy Concept. NASA TP 1137, Feb. 1978.
7. Mukhopadhyay, V.; Newsom, J. R.; and Abel, I.: A Method for Obtaining Reduced Order Control Laws for High Order Systems Using Optimization Techniques. NASA TP 1876, 1981.
8. Newsom, J. R.; and Mukhopadhyay, V.: Application of Constrained Optimization to Active Control of Aeroelastic Response. NASA TM 83150, June 1981.
9. Sandford, M. C.; Abel, I.; and Gray, D. L.: Development and Demonstration of a Flutter Suppression System Using Active Controls. NASA TR 450, December 1975.
10. Redd, L. T.; Gilman, J., Jr.; Cooley, D. E.; and Sevart, F. D.: A Wind Tunnel Study of B-52 Model Flutter Suppression. AIAA Paper No. 74-401, April 1974.
11. Doggett, R. V., Jr.; Abel, I.; and Ruhlin, C. L.: Some Experiences Using Wind Tunnel Models in Active Control Studies.
NASA TM X-3409, August 1976, pp.
831-892.
12. Newsom, J. R.; Abel, I.; and Dunn, H. J.: Application of Two Design Methods for Active Flutter Suppression and Wind Tunnel Test Results. NASA TP 1653, May 1980.
13. Peloubet, R. P., Jr.; Haller, R. L.; and Bolding, R. M.: F-16 Flutter Suppression System Investigation. AIAA Paper No. 80-0768, May 1980.
14. Hwang, C.; Johnson, E. H.; and Pi, W. S.: Recent Developments of the YF-17 Active Flutter Suppression System. AIAA Paper No. 80-0769, May 1980.
15. Abel, I.; and Newsom, 3. R.: Wind Tunnel Evaluation of NASA-Developed Control Laws for Flutter Suppression on a DC-10 Derivative Wing. AIAA Paper No. 81-0639, April 1981.
16. Winther, B. A.; Shirley, W. A.,; and Heimbough, R. M.: Wind Tunnel Investigation of Active Controls Technology Applied to a DC-10 Derivative. AIAA Paper No. 80-0771, May 1980.
and Pototzky, Anthony S.: Comparison of Analysis and Flight 17. Newsom, Jerry R.; Test Data for a Drone Aircraft with Active Flutter Suppression. AIAA Paper No. 81-0640, April 1981.
18. Edwards, J. W.: Flight Test Results of an Active Flutter Suppression System Installed on a Remotely-Piloted Research Vehicle. AIAA Paper No. 81-0655, April 1981.
Application of a Flight Test and Data Anaysis 19. Bennett, R. M.; and Abel, I.: Technique to Flutter of a Drone Aircraft. AIAA Paper No. 81-0652, April 1981.
20. Newsom, Jerry R.; Pototzky, Anthony S.; and Abel, Irving: Design of the Flutter Suppression System for DAST ARW-1R - A Status Report. NASA TM 84642, March 1983.
SESSION V RESEARCH OPPORTUNITIES FOR THE FUTURE Duncan E. McIver Moderator
RESEARCH OPPORTUNITIES
FOR
FUTURE COMMERCIAL TRANSPORTS
J. F. Longshore
Douglas Aircraft Company
Long Beach, California
First Annual NASA Aircraft Controls Workshop
NASA Langley Research Center
Hampton, Virginia
October 25-27, 1983
MD-100 COCKPIT This photo of one of the MD-100 flight deck configurations being discussed with the airlines illustrates the flexibility in display formats afforded by today’s CRT technology. All of the formats shown can be extensively tailored to phase of flight and aircraft configuration and can be configured to provide unique data to the flight crews for special situations.
MD-100 PRIMARY FLIGHT DISPLAY This photo illustrates the typical primary flight display format. Note that this display includes all air data, flight control system mode annunciation, radio altitude and decision height, as well as the traditional attitude director functions. The navigation display on an adjacent CRT is typical of those found in recently certified flight management systems.
MODE ANNUNICATOR -VERTICAL SPEED SELECTED - ALTITUDE AIRSPEED PREDICTOR - ALTITUDE - PREDICTOR SELECTED AIRSPEED - - BAR0 SETTING RADIO HEIGHT SELECTED DECISION HEIGHT MD-100 MULTIFUNCTION DISPLAYS These display formats are typical of the MD-100 subsystem displays. A synoptic system display is used for the fuel, pneumatic, electrical, and hydraulic systems. These can be called up by the crew. Color and symbology changes are used in the synoptics to identify system configuration and annunciate system failures.
ENGINE PARAMETERS IN MFD 1 PHASE-OF-FLIGHT DISPLAY IN MFD 2 FLIGHT DECK/FLIGHT CONTROL SYSTEM CONCEPTS 0 The objective of the ongoing NASA fault-tolerant architectural studies is to identify architectural concepts that have sufficient reliability for full-time fly-by-wire applications. Additional effort will be required to develop the methodology and criteria to be used in certifying such systems. NASA would function as the facilitating agency working with the air transport community and the FAA to define these methods and criteria.
0 The impact of failures and the appropriate corrective action in future highly integrated fault-tolerant aircraft systems will be difficult to determine. Artificial intelligence (expert systems) concepts may be applicable to the aircraft system management computers required to manage such systems. Expert systems require extensive data bases to be effective. NASA research would focus on building failure mode and effects and degraded mode data bases and on defining and validating expert system concepts for aircraft system management.
FAULT-TOLERANT CONTROL SYSTEM CERTIFICATION CRITERIA AND METHODOLOGY ARTIFICIAL INTELLIGENCE/EXPERT SYSTEM CONCEPTS FOR AIRCRAFT SYSTEM CONTROL AND DISPLAY SYSTEMS AIRCRAFT/NATIONAL AIR SPACE SYSTEM COMPATIBILITY 0 The air transport industry has invested considerable effort in the design and development of fuel- saving flight management systems. To be effective, these systems must be allowed to control the climb, cruise, and descent flight profiles without undue air traffic control interference. The FAA is in the beginning phases of defining a national air space system aimed at increasing air space and airport capacity through the use of increased automation. This system could potentially increase air traffic control constraints to the detriment of efficient aircraft operation. NASA research would be directed at achieving maximum synergy between the desires of air transport operators to minimize costs and the desires of the FAA to increase air space and airport capacity.
0 The FAA national air space system plan makes increased use of automation and assumes increased aircraft densities in airport approach and departure areas. NASA research would be directed at investigating approach and departure route saturation sensitivities to abnormal conditions such as weather, in-flight emergencies, etc.
AIRCRAFT TRAFFIC CONTROL FLIGHT PROFILE AND PROCEDURE CONSTRAINTS AIRPORT/AIR TRAFFIC CONTROL APPROACH AND DEPARTURE SATURATION SENSITIVITIES DISPLAY RESEARCH 0 The objective of the generic non-airframe peculiar display research activities would be to develop standard display formats that would be accepted and used by the air transport community.
0 The objective of the display format human factors studies would be to develop standards pertaining to the use of colors, synoptics, etc., for communicating aircraft status and situation to the flight crew.
GENERIC NON-AIRFRAME PECULIAR DISPLAY FORMATS 0 PRIMARY FLIGHT INSTRUMENTS 0 APPROACH,ROLLOUT,TAXI,TAKEOFF,AND DEPARTURE AIDS . AIR TRAFFIC CONTROL RELATED DISPLAYS HUMAN FACTOR DISPLAY FORMAT STUDIES l FORMAT INFORMATION TRANSFER EFFICIENCY l WORK LOAD ANALYSIS l DATA REQUIREMENTS l COLOR STANDARDS
IL
A BRIEF REVIEW OF AIRCRAFT CONTROLSRESEARCH OPPORTUNITIES IN THE GENERAL AVIATION FIELD Eric R. Kendall Gates Learjet Corporation Wichita, Kansas First Annual NASA Aircraft Controls Workshop NASA LaFgley Research Center Hampton, Virginia October 25-27, 1983 CONTROLSTECHNOLOGY REVIEW The review process itself is part of a feedback control system (Figure 1).
The work already accomplished by NASA on flight test programs (Block A) and on trade studies (Block B) must be reviewed to determine the potential controls technol- ogy benefits available to the general aviation industry (Block C). General aviation industry constraints (Block D) must be defined and applied to determine the currently useable controls technology (Block E). Any shortfall between the required technology (Block F) and the useable technology shows up as a technology deficiency (Block G) and identifies the future research opportunities (Block H). Additional future research by NASA (Block J) can increase the controls technology benefits and ulti- mately nullify the technology deficiency.
GENERAL AVIATION NASA INDUSTRY (A) MILITARY 8 LARGE COMMERCIAL TRANSPORT AIRPLANE TEST PROGRAMS
h
NASA-SPONSORED
I
(J) (H) $ (G) v I FUTURE FUTURE TECHNOLOGY RESEARCH* RESEARCH DEFICIENCY OPPORTUNITIES Figure 1 MILITARY AND COMMERCIALTEST PROGRAMS A significant number of flight test programs related to ACT have been conducted during the last 25 years. Some of these are shown in Figure 2. Most are The technology is related to very large airplanes with flexible structures.
'acronym saturated'. Richard Holloway defines the more commonly used acronyms and explains system functions in Ref. (1). Those used here are:
ccv . . . . . ..Control Configured Vehicle
FMS. . . . . ..Flutter Mode Suppression GASDSAS . ..Gust Alleviation & Structural Dynamic Stability Augmentation System ALDCS. . . ..Active Load Distribution Control System AS . . . . . . ..Augmented Stability WLA. . . . . ..Wing Load Alleviation Figure 2 NASA-SPONSORED TRADE STUDIES A few of the many trade studies sponsored by NASA are shown in Figure 3. These all relate to commercial transports or commuter airplanes. The STAT program, which started in 1978, was reported by Louis Williams of NASA Langley at the 1982 SAE Commuter Aircraft and Airline Operations Meeting in Savannah, GA (Ref. 4). The report contains much material relevant to the application of advanced technologies in the general aviation industry. ACT benefits were explored on two candidate airplane designs. Controls technology benefits need to be separately identified.
1970 - LOWWING LOADING STOL STUDYt2) ,,,,.,,,,,,,.,,,,.,,,NASA/BOEING WICHITA 1972 - APPLICATION OF ADVANCED TECHNOLOGIES TO LONG-RANGE TRANSPORT AIRCRAFTt3),,,,,,.,,..,,,.,NASA/BOEING SEATTLE 1978 - SMALL TRANSPORT AIRCRAFT TECHNOLOGY NASA/INDUSTRY (STAT) PROGRAM(') IIIIIIIIIIIIIIIIIIIIIIIIII~IIIIIII 1982 - INTEGRATEDAPPLICATION OF ACTIVE CONTROLS TECHNOLOGY TO AN ADVANCED SUBSONIC TRANSPORTt5) Illll1llll111I111111111111111, NASA/BOEING SEATTLE Figure 3 CONTROLSTECHNOLOGY BENEFITS (TEST RESULTS) Some of the benefits which have been obtained as a result of the flight test programs are listed in Figure 4. Active control systems on the B-52 and on the C-5A are incorporated as retrofits to production airplanes. The L-1011 systems permit increased wing span which leads to improved cruise performance.
l ECP 1195 ON B-52 REDUCES FATIGUE 8, ALLEVIATES GUST LOADS
. ALDCS ON C-5A IMPROVES FATIGUE LIFE BY GUST AND MANEUVER
LOAD REDUCTIONS
. L-1011 SYSTEMSREDUCE LATERAL GUSTDESIGN LOADSAND PERMIT
WING TIP EXTENSIONWITH NO BEEF-UP
Figure 4 Cotd~RoLs TECHNOLOGY BENEFITS (TRADE STUDIES) NASA-sponsored trade studies have shown significant synergistic design benefits for a wide range of commercial airplane types. Some results from these studies are shown in Figure 5 for STOL, commuter, and subsonic commercial transport airplanes.
GLA. . . ..Gust Load Alleviation PAS. . . ..Pitch Augmentation System AAL . . . ..Angle-of-Attack Limiting
l REFJ2) STOL STUDY SHOWS 10% TO 30% GROSS WEIGHT
REDUCTION WITH MECHANICALFLAP 'GLA'
(F.L, x2500 FTJ
l REF,(3> MACH0,98 AIRPLANE GROSS WEIGHT REDUCED
BY 11% THROUGH USE OF 'AS'
. REF, (4) 'ACT' BENEFITS CONVAIR& LOCKHEED COMMUTER
AIRPLANE DESIGNS
. REFJ5) 10% IMPROVED FUEL EFFICIENCY ON SUBSONIC
TRANSPORT THRU USE OF PAS, AAL & WLA
Figure 5 GENERAL AVIATION ‘ACT’ CONSTRAINTS The constraints listed in Figure 6 are typical of some which might be specified by the general aviation industry. Coordination within the industry is required before these can be considered as an official industry input. However, judging from the complexity levels of systems in use today, constraints such as these will produce significant future research opportunities. The need to be compatible with manual (unpowered) primary flight control systems was addressed by Dr. Jan Roskam and others in Ref. (6).
This considered the use of a separate surface stability augmentation system for general aviation aircraft. Design philosophies and hardware implementation schemes were defined and evaluated in Ref. (7).
l MUST BE SIMPLE
l MUST BE EASY TO MAINTAIN
. MUST NOT BE SAFETY CRITICAL
. MUST BE COMPATIBLE WiTH MANUAL
(UNPOWERED) PRIMARYFLIGHT CONTROL
SYSTEMS
. MUST IMPROVEAIRPLANE SALES
POTENTIAL
Figure 6 GENERAL AVIATION FLIGHT TESTS G.eneral aviation constraints such as those just mentioned have been recognized for some time. As a result, several NASA 'ACT' programs have been directed specifi- A recent review has been pre- cally towards the general aviation type of airplane.
sented in Ref. (8) by Dr. David Downing and others of KU. A few of the programs are listed in Figure 7.
VRS........ ..Vertica 1 Ride Smoothing GPAS....... ..Genera 1 Purpose Airborne Simulator RSS........ ..Relaxe d Static Stability SSSAS...... ..Separat e Surface Stability Augmentation System c-45 111.....1,1...0.1,1. VRS c-140 . . . . . . . . . . . . . . . . . ..GPAS/RSS BEECH99 . . . . . . . . . . . . . . ..SSSAS Figure 7 1o-3 BUMP SIZE The potential need for ride quality control on some commuter airplanes can be seen by comparing the estimated ride qual'ty of various types of unaugmented air- !I planes. Figure 8 shows the estimated lo- bump size for three types normalized .to that of a commercial transport flying at an altitude of 35,000 feet. It is apparent that the commuter with its relatively high response to vertical gusts flying in the more gust-prone lower altitude bands will present a rougher ride in turbulent condi- tions. Technology trends in many of the emerging new commuters are towards simplicity, and it may be some time before ride quality systems are generally accepted. Research should continue to take advantage of the rapid developments in ' electronics and controls to make these systems more attractive for future high- technology commuter airplanes.
1 COMMERCIALTRANSPORT, 35,000"
2 BUSINESS JET, 45,000' 3 COMMUTER, 10,000' I TECHNOLOGYTREND b .SIMPLE FLAPS .PARTIAL GEAR RETRACTION l CONVENTIONAL CONTROLS -----_ --- -a
LCL -
Figure 8 -. - . -.._ ._.- -. - -.~L_-. _ TECHNOLOGY CHOICES General aviation airplane designers tend to use simpler control technologies than those employed on military and large commercial transport airplanes. As a result, a larger reserve of well-proven controls technologylis available as an alternative to the adoption of advanced state-of-the-art controls technology. This is depicted in Figure 9.
Figure 9 - SOME GENERAL AVIATION PREFERENCES A brief list of some general aviation airplane desjgn preferences is.presented in Figure 10. This is included to emphasize a point that in many instances a simple technology is chosen over a more complex and more effective one. The tendency to use simple.flap systems and low wing loadings is an example qf the 'trade towards Quite often there is a tendency to reject any beneficial simplicity' approach.
external features if they are considered detrimental to styling, and the preference is to eliminate avionic systems rather than to add them. Clearly much research will be required to produce ACT benefits which are marketable in the general aviation sector.
. SIMPLE FLAPS 8 LOWWING LOADINGS
. CONFIGURATION FEATURES WHICH ARE BOTH
BENEFICIAL & STYLISH (E,G,,WINGLETS)
. MINIMAL DEPENDENCE ON AVIONIC SYSTEMS
(E,G, YAWDAMPERS,STALL PREVENTION)
Figure 10 SIMPLE FLAPS AND LOW WING LOADINGS The 'trade-towards-simplicity' tendency just noted is seen from the data pre- This compares the stall speeds and wing loadings of some general sented in Figure 11.
aviation airplanes with those for commercial transports and advanced high-lift air- In general, the landing CLMAx for planes. the general aviation types is around 1.8, which is achievable by simple single-slotted partial span flaps. The commercial trans ports have a landing CLMAX close to 2.8, which requires more complex flap arrange- ments. The general aviation landing stall speeds are kept to an acceptable level by using lower wing loadings than are commonly used by the commercial transports.
Clearly there is some tradeability towards more complex flaps to obtain the cruise benefits of a higher wing loading. This is an example of an available technology not being fully exploited due to the reference for simplicity. Similar trends were noted by Dr. Jan Roskam, Ref. (9 P, who proposed new airfoils, higher wing loadings, and a new look at general aviation airplane design.
, K-IS , I I I I I I 1 1 I I I 1 I, 40 50 60 70 so 90 Icm IJO MAX LANDING WING LOADING (LB/FT~) Figure 11 THE MAGNITUDE OF AIRPLANE PERFORMANCE BENEFITS .*.
Quite often the magnitude of airplane cruise performance benefits available While such improvements cannot be ignored, from ACT is small (e.g., 2 or 3 percent).
the cost effectiveness of systems needed to obtain them must be carefully considered.
A basic requirement must be that the improvement will be sustained throughout the life of the airplane and that the benefits definitely will be felt by the airplane owner.
To give some indication of performance improvement 'detectability', data on airplane fleet performance variability are presented in Figure 12. This is a histo- gram of incremental percentage fuel flow gathered from forty-one new production airplanes all of the same model designation and all flown on the same route by Data were corrected for observed ambient conditions.
production flight test crews.
About half the measurements are contained within +2% of the nominal value.
'improver' ACT system might make a small Even though a cruise performance but statistically significant improvement to the fleet picture, it may not be of any practical significance to a particular one-airplane operator.
A . - - . . . . - ..---_I--
I
._-_-_-.- __-.-_-.---- ..,__ --.-
t
I
I
-XI
-;5
+6 +4
l b
a %.FUEL FLOW
'BETTER
WORSE
Figure 12 ACT USED IN TOUGH COMPETITIVE SITUATIONS competitive pressures will accelerate the Williams (Ref. 4) wrote that "...
use of technological advances...". The data in Figure 13 confirm this statement and show how ACT was introduced in a tough competitive situation between two general aviation business airplanes. The cruise performance of each airplane is comparable with airplane A's passenger miles pe-r pound of fuel used being better than B's on the short range. The higher wing loading and winglet used on airplane 'A' more than offsets the low wing loading and supercritical section of airplane 'B'. Then, airplane A's balanced field length was made comparable with B's by introducing automatic performance reserve (APR) and automatic spoilers. These change engine thrust and spoiler setting without a direct command from the pilot and therefore can be classified as active control systems.
3,420 4,200 'WINGLET EFFECT IIJCLUDED, (1) BFL IMPROVED WITH APR 8 AUTOSPOILERS.
Figure 13 GENERAL AVIATION ACT SUGGESTIONS Based on the very brief discussion of some general aviation design trends and preferences (still to be coordinated within the industry sector), Figure 14 It’seems that avionic presents a summary of suggestions for ACT activities.
cruise performance improvers will not sell easily unless the advantages are large.
Since many general aviation airplanes have low wing loadings and fly at relatively low altitudes,the emphasis should probably be on ride quality improvement and gust alleviation systems. Retrofittable systems could be attractive since few airplanes are likely to be designed with optimal structures and no.growth capability.
. 'CRUISE IMPROVERS'SHOULD:
A) COMPLY WITH CONSTRAINTS
B) COMPETE WITH FLAP/WING-LOADING
TRADE-OFF.
c> PRODUCE SIGNIFICANT REDUCTIONIN
CRUISE FUEL-FLOW,
. RIDE CONTROL 8, GUST ALLEVIATION SYSTEMS MAY BE
MOST LIKELY 'ACT' FUNCTIONSTO FIND APPLICATON
. RETROFITTABLELOAD ALLEVIATORSMIGHT BE ATTRACTIVE
FOR PROVIDING AIRPLANE GROWTH CAPABILITY WITH
MINIMUM STRUCTURAL BEEF-UP, Figu.re 14 GENERAL AVIATION CCV Information and discussion presented so far might give the impression that the general aviation industry is ultra-conservative and not likely to adopt any significantly new controls concept in the forseeable future. However, even though there may be a natural reluctance to adopt a complicated avionics ACT system, the field of 'non-electronic' CCV technology might be regarded differently. The new designs introduced by Beech and by the Gates-Piaggio team at this year's National Business Aircraft Association (NBAA) show in Dallas show once again that . ..competitive pressures will accelerate the use of technological advances..."
{'Ref. 4). Figure 15 shows the competitive 'canard' and 'three-surface' designs relative to a conventional configuration. NASA research has been strongly sup- portive of these unconventional designs. Much is left to be done.
THREE-SURFACE CONVENTIONAL ___---_ Figure 15 GENERAL AVIATION RESEARCH OPPORTUNITIES IN THE FLIGHT CONTROLSTECHNOLOGY The research opportunities in the fields of ACT and CCV for general aviation This review has attempted to forsee some of these opportunities are enormous.
by assessing general aviation needs and trends relative to the currently available Coordination within'the general technology. A few ideas are listed in Figure 16.
aviation industry and between industry and NASA should be intensified in the near term to try to provide NASA with a more complete and representative feedback.
A) OVERALL - CONTINUE STAT FOR SMALLER G.A, AIRPLANES (40 PAX,) - DETERMINE POTENTIAL 'ACT' CONTRIBUTION - CATALOGALL ACT BENEFITS, B) AVIONIC SYSTEMS - EXPLORE FEASIBILITY OF RETROFITTABLE LOAD ALLEVIATORS & RIDE QUALITY IMPROVERS, - RUN SIMULATOR STUDIES & FLIGHT TESTS.
- DEVELOP ANALYTICAL METHODS AFFORDABLETO GENERAL AVIATION USERS C> AIRPLANE CONFIGURATIONS - INCLUDE CANARDS& THREE-SURFACEAIRPLANES IN A) & B) ABOVE, - CONTINUE WIND TUNNEL TESTS TO DETERMINE THE STABILITY AND CONTROLCHARACTERISTICS OF UNCONVENTIONAL AIRFRAME/ PROPULSIONARRANGEMENTS Fiaure 16 REFERENCES 1. Hol.loway, Richard B., "Introduction of CCV Technology Into Airplane Design," AGARD-CP-147, vol. 1, June 1974, pp. 23-'1 to 23-16.
2. Holloway, R. B., Thompson, G. O., and Rohling, W. J., "Prospects for Low-Wing- Loading STOL Transports With Ride Smoothing," Journal of Aircraft, volume 9, no. 8, August 1972, pp. 525-530.
3. Boeing Co., "Study of the Application of Advanced Technologies to Long Range Transport Aircraft," NASA CR-112093, May 1972.
4. Williams, Louis J., "Advanced Technology for Future Regional Transport Aircraft," SAE 1982 Transactions, vol. 91, section 3, 1982, pp. 2481-2497.
5. Boeing Co., "Integrated Application of Active Controls (IAAC) Technology to an Advanced Subsonic Transport Project," NASA CR-165963, December 1982.
"An Investigation of Separate Surface Stability Augmentation 6. Roskam, J., et al., Systems for General Aviation Aircraft," University of Kansas, Center for Research, Inc., August 1972.
7. Roskam, J., "Design Philosophy & Hardware Implementation of Separate Surface Automatic Flight Control Systems," University of Kansas, Center for Research, Inc., May 1974.
8. Downing, D., Hammond, T., and Amin, S., "Ride Quality Systems for Commuter Aircraft," NASA CR-166118, May 1983.
9. Roskam, J., "New Airfoils and Higher Wing Loadings: A New Look at General Aviation Airplane Design," University of Kansas, Center for Research, Inc., May 20, 1974.
RESEARCHOPPORTUNITIES FOR ROTORCRAFT Bruce Blake Boeing-Vertol Company ' Philadelphia, Pennsylvania First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 HELICOPTER. ROTOR VIBRATORY LOADS Helicopter vibration reduces crew performance, comfort, component life, It originates primarily in unsteady aerodynamic loads on the and reliability.
rotor blades. Both lift and drag vary periodically because blade angle of attack and local velocity change as the blades rotate relative to the direction of flight and travel above and through the vortex wake, as depicted in this figure. Some harmonic components of the periodic blade loads are transmitted to the fuselage, where they produce airframe vibration. The absorbers and isolation systems currently used to reduce this vibration have undesirable weight penalties and often do not achieve vibration levels that are fully satisfactory.
A concept known as higher harmonic control is capable of suppressing vibration through active control of blade pitch at frequencies above the normal once-per-revolution pitch changes.
Successful results have been obtained in wind tunnel tests conducted at Boeing Vertol and by other researchers as well as in flight tests at Hughes Helicopter.
“ADVANCING” BLADE: HIGH VELOCITY CONCENTRATED TIP VORTEX “RETREATING” BLADE: LOW VELOCITY PRIMARY SOURCES SECONDARY SOURCES ROTOR BLADE SPEED ARTICULATED BLADE DIFFERENTIAL MOTIONS ELASTIC BLADE VELOCITIES INDUCED DEFLECTIONS BY VORTEX WAKE I ACTIVE CONTROL To achieve near-term results, most current R&D efforts have chosen to obtain the required pitch harmonics using the conventional swashplate control configuration rather than individual blade actuators in the rotating system.
The swashplate is already used by both the pilot and the automatic flight control system to provide thrust control by setting the average blade pitch, called collective, and to provide trim and flight path control by creating a one-per-rev variation in blade pitch, called cyclic. The new control will superimpose higher frequency oscillations on the standard swashplate motions to control blade pitch and aerodynamic loads at the vibration harmonics and the adjacent harmonic frequencies. The harmonic hub loads are not to be eliminated completely, but reduced and rephased until their effects in the airframe cancel the effects of other vibration sources, particularly the unsteady aerodynamic loads on the fuselage itself. Higher harmonic control requires an automatic controller because the harmonic blade pitch which achieves optimal results varies greatly with flight condition. The automatic system must analyze vibration measurements in real time to determine the optimum harmonic control, in terms of amplitude and phase of several frequencies.
OPPORTUNITIES CURRENT . FUSELAGE VIBRATION REDUCTION l R8D EMPHASIS
I I
. POWER REDUCTION 9 FLIGHT ENVELOPE EXPANSION (RETREATING BLADE STALL ALLEVIATION] *FATIGUE LOAD REDUCTION BLADES CONTROL SYSTEM - BLADES - CONTROL SYSTEM . IMPROVED HANDLING OUALITIES BY REDUCING APPARENT CONTROL LAG THE EXTENT OF THE BENEFITS ATTAINABLE DEPENDS ON ROTOR AND FUSELAGE DYNAMICS AND WILL VARY AMONG HELICOPTER DESIGNS I ImllmIIIllIllllllIIl III1
VIBRATION REDUCTION 'i wI'ND TUNNEL DEMONSTRATION
The benefits of active control in vibration treatment are shown in this figure. An experimental investigation involving a lo-foot diameter hingeless rotor was conducted at the Boeing Vertol V/STOL wind tunnel. This chart shows that the simul,taneous application of three harmonic control inputs resulted in a dramatic reduction in the three vibratory hub,loads presented here.
These results were achieved through closed-loop control using hub strain gage balance loads as feedback,parameters. The demonstrated response time was 0.075 sec. Typically, higher harmonic control requires a frequency response that is 10 to 20 times greater than primary flight control to provide suppression of higher harmonic loads, particularly during maneuvers and in turbulence. Continued operation at these high frequencies requires improved bearing and seal designs to avoid rapid wear. Furthermore, precise control of the pitch actuation system at higher harmonic frequencies is required for effective vibration reduction.
WIND TUNNEL DEMONBTRATION . IO-FOOTD!AMETER HINGELESSROTOR . THREE HARMONICS APPLIED SIMULTANEOUSLY . INPUTS SELECTED BY ACTIVE (CLOSED-LOOP) CONTROLLER WITH 0.075 SEC. RESPONSETIME WITHOUT HIGHER HARMONIC CONTROL HIGHER HARMONIC CONTROL n/REV VIBRATORY HUB LOADS VERTICAL PITCHING ROLLING FORCE (LB) MOMENT (FT.LEl) MOMENT (FT-LB) CONTROL ACTUATORRESPONSE Flight control power actuators used in all current helicopters have a response bandwidth sufficient for good handling qualities and automatic flight control system performance. Since rotor dynamics dictate a sharp cutoff at rotor rotational speed, there is no need for high system response. Vibration control through higher harmonic pitch demands that control system output be predictable and well defined in both gain and phase at frequencies up to 30 to 40 Hz for small aircraft.
Accurate control at 90 Hz has been experimentally demonstrated in the wind tunnel although at actuator amplitudes and service lives considerably below flight vehicle requirements.
FREQUENCY RANGE OF Vl8RAtlON CONTROL MUCH HIGHER THAN CURRENT FLIGHT CONTROL PRACTICE CONTINUOUS OPERATING FREQUENCIES OF FLIGHT CONTROL ACTUATORS b/REV VIBRATIONCONTROL VREV VIBRATION CONTROL OF DEMNSTRATED IN \ FULL-SCALE AIRCRAFT REQUIRED IMPROVEMENTS Required improvements include: l Feedback measurements, algorithm design, and closed-loop response to keep pace with the rapidly changing blade pitch requirements needed during maneuvers. Large transient vibration levels are very objectionable to the pilot. It is necessary to understand the effect on aircraft performance and component loads.
l Substantially upgraded control system performance.
l Service life characteristics suitable for production aircraft.
l REAL-TIME COMPUTATION OF HIGHER HARMONIC INPUTS TO KEEP UP WITH RAPID CHANGE OF FLIGHT CONDITIONS l LARGE INCREASE IN PITCH ACTUATION POWER AND RATE CAPABILITIES . DESIGN FOR GOOD RELIABILITY AND MAINTAINABILITY l PRECISE CONTROL OF PITCH ACTUATION SYSTEM AT HIGHER HARMONIC FREQUENCIES SYSTEM BENEFITS Since vibration has always been a generic problem with helicopters, all research efforts to date have concentrated on reducing aircraft vibration.
Unlike other vibration control devices such as isolators and absorbers which add weight and therefore reduce payload, higher harmonic control offers the Redistribution of blade unique opportunity of more than paying its own way.
section lift and drag over the rotor disc can significantly reduce power required and increase the usable flight envelope.
It may also be possible to reduce fatigue loads and improve handling qualities. It may not be possible to achieve all these simultaneously.
However, the potential benefits to helicopter users are so large that this additional tool now available to helicopter designers must be explored to its fullest.
ACTIVE CONTROL USES ROTOR BLADE PITCH TO CONTROL THE UNSTEADY AERODYNAMIC LOADS COLLECTIVE +; CONVENTIONAL BLADE PITCH CONTROL USED BY THE PILOT AND THE AUTOMATIC FLIGHT CONTROL SYSTEH
I
ACTIVE CONTROL ACTIVE CONTROL: HIGHER FREQUENCY COMBINED TOTAL BLADE AZIMUTII % DEC For advanced rotorcraft applications, especially those with more than one mode of operation, improved integration between systems results from including thrust/power management in the AFCS. Generally, functions related to blade pitch control are performed within the aircraft system. By uniting thrust and simpler, more flexible mode transition and selection are power management, possible. The engine control may then act solely as a power control with required limitations, resulting in a more adaptable engine.
, CONTROLCONFIGlJRATlONFOR ADVANCED APPLICATIOIIS . ADVANCED HELICOPTER . TILT RUTOR . X-WING . CONVERTIBLE ENGINE APPLICATIONS . POWER PANAGWENT ACCCWLlSHED IN AFCS . RESPONSIBILITY FOR ROTORCOllTROL (SPEED : PITCH) IS ENTIRELY WITHIN AFCS - NOT SPLIT As YITH CONVENTIONAL SYSTEM . RULTI-MIDE INTERACTIONS AND IllCREASED INTEGRATION EOUI~NTS ARE BEST HANDLEDWITHIN AFCS . ENGINE CONTROLPROVIDES GAS GENERATOR WVERnlffi I LIRITING L SIRPLER ENGINE CONTROLFUllCTIONlNG As PWER COWTROLLER , EACH APPLICATIO)I WES NOT NEED A TAILORED ENGINE COnTROL a FEWERENGINE AIRFRAM INTERCONNECTIONS . PROVIDES HIGH LEVEL OF REDUNDANCY OVERALL INTEGRATION GOALS Primary goals for integrating engines and flight controls are to improve handling qualities and system performance. Aircraft control response.
characteristics may be enhanced through selective inputs.to the engine control. Reducing pilot workload by providing simpler engine cockpit controls and eliminating manual backup are important consjderations. Automatically optimizing engine performance and,establishing diagnostic. requirements aimed at improving mission reliability and safety are also engine-related goals of an integration program.
. IMPROVED HANDLING QUALITIES IMPROVE AIRCRAFT CONTROL RESPONSE CAPABILITY SIMPLIFY PILOT CONTROL . ELIMINATE MANUAL BACKUP . OPTIMIZE SYSTEM PERFORMANCE OPTIMIZE ENGINE PERFORMANCE IMPROVE MISSION RELIABILITY/SAFETY DEVELOP DIAGNOSTIC REQUIREMENTS INTEGRATION OBJECTIVES Program objectives should be to optimize response of the integrated system and to upgrade overall engine performance. Specifically, the program should be aimed at improving torsional stability of the engine-rotor/drive when considering multi-mode configurations and engine-out conditions; emphasis should be placed on increasing rotor thrust response to decrease rotor speed droop during rapid maneuvers, gust rejection, and precision hover conditions.
Engine and control diagnostics and trend monitoring represent important considerations toward improving engine/aircraft availability. Methods for displaying power margin in conjunction with power assurance and techniques for continuously optimizing system performance should be addressed. An additional objective of a program should be engine failure recognition and indication with subsequent corrective action by the flight guidance system.
OPPORTUNITIES l OPTIMIZE RESPONSE INCREASED TORSIONAL STABILITY OF ROTOR/DRIVE SYSTEM IMPROVED ROTOR THRUST RESPONSE TO CONTROL - RAPID MANEUVERS - GUST REJECTION - PRECISION HOVER . UPGRADE ENGINE PERFORMANCE ENGINE AND CONTROL DIAGNOSTICS TREND MONITORING POWER MARGIN/POWER ASSURANCE PERFORMANCE OiTlMlZATlON FAILURE RECOGNITION
ADDITIONAL RESEiWCH TOPICS
. DEVELOP LOW-.ALTITUDE ATMOSPHERIC TURBULENCE MODEL AERODYNAMIC ENVIRONMENT BELOW 100 ft. NEAR OBSTRUCTIONS IS NOT WELL QUANTIFIED . HUMAN FACTORS FLIGHT SIMULATION VISUAL AND MOTION CUE REQUIREMENTS METHOD(S) F9R MEASURING PILOT WORKLOAD l SENSOR TECHNOLOGY APPROACHES TO USE OF MULTISPECTRAL IMAGING ANALYTICAL REDUNDANCY WIRE FINDERS . PARAMETER IDENTIFICATION COMPLEXITY OF REQUIRED MODELS HAS RETARDED PROGRESS RELATIVE TO FIXED-WING AIRCRAFT REFERENCES 1. Shaw, John and Albion, Nicholas: Active Control of Rotor Blade Pitch for Vibration Reduction: A Wind Tunnel Demonstration.
Vertica, vol. 4, no. 1, 1980, pp. 3-1-l.
Shaw, John and Albion, Nicholas: 2. Active Control of the Helicopter Rotor for Vibration Reduction. J. Amer. Helicopter Society, vol. 26, no. 3, July 1981, PP. 32-39.
FIGHTER AIRCRAFT FLIGHT CONTROLTECHNOLOGY DESIGN REQUIREMENTS W. E. Nelson, Jr.
Northrop Corporation, Aircraft Division Hawthorne, California First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 -- - This figure represents the evolution of control technology. As we compare the airplane, we notice that they current day fighters with the Wright brothers' had achieved control technology and mastered their applications without electronics.
But the future demands further emphasis of pursuing the aspects of control,with the extension of research going into exotic concepts such as vortex management.
Truly, the day of the pilot controlling the aircraft will diminish. As weapon system integration and flight path and navigational integration become more heavily His primary task will be selecting automated, the pilot will become a manager.
his secondary task will be as backup. He will the right functional requirements; act in the fashion he presently performs, that of a pilot.
This figure humorously depicts the challenge the future pilot will have.
However, it is not unattainable, as work on the AFT1 and the present F-18 has The research needed is to establish standards that the designer demonstrated.
can utilize to evaluate his concepts.
Active control elements in flight control technology encompass many technical disciplines. Electronic chip development will result in achieving potential architectures of control laws operating in real time during flight. The high densities and improved computation times will allow greater design flexibility for fault-tolerant applications. Data communications, actuation technology, and electrical power concept, when combined with the elements of the computer, will lead to the indicated technology objectives.
Flight control technology has evolved to a total systems engineering discipline as depicted in this figure. The mission requirements set the needs. The available electronic technology provides the capability. The interface with all avionic and other aircraft subsystems increases the flexibility of the control capability. But hidden inside the effort is an item that can deter the design performance, the validation testing. Thus, a need exists for desiqn standards to relate the desired methods and procedures to be used in the design effort of future vehicles.
Here is a specific example of a control application that is in its infancy. The payoff on potential aircraft can be related into reduced structural weight.
However, further research is needed to explore various concepts and to demonstrate them in both wind tunnel and flight tests.
BATTLE DAMAGE CONSIDERATIONS
FOR ADAPTIVE FLUTTER SUPPRESSION
a SUSTAIN FLUTTER FREE MRATlDU FOLLOWINQ AN ARRAY OF BATTLE DAMAOE STATES l ENHANCE THE EXIttIm) ADAPTIVE ALGORITHM BY SELF-REPAIRIMB CowCEm ‘AYOFF p&q l IMPROVED COMBAT EFFECTIVENESS l DEVELOP ALDORITHMS FOR BATTLE DAMAQE l RAPID BATTLE DAMAGE REPAIR - DETECTION - VOTINQ 8CHEME - ISOLATION - VARIANCE COMTARIW l REDUCED PEACE TIME COST - CONTROL RECONFlDURATlON - REDUNDANTCONTROLSUBFACE OPERATION -------------- l EMPHASIS DN Y F-17 WIND TUNNEL TEST MODEL l INCREASED SOFTWARE COMPLEXITY l CREW INTERFACE ISSUES l COMPUTATIONAL (FRAME TIME) PENALTY The role of "in-flight simulation" needs to,be revived. This area is valuable to relate new design features of control law, systems operation, and interaction with the pilot in a near real-life environment. and to reflect the needed design changes into the new vehicle before a large change impact of cost and schedule is imposed on the project. A new airframe platform is needed to replace the antiquated T-33.
FUTURE FLIGHT CONTROL SYSTEM NEEDS - IMPROVED IN-FLIGHT SIMUIATION
0 VARIETYOF HANDLING QUALITIES ISSUESNEEDFURTHER STUDY-
LATERAL SENSITIVITY, CONTROL HARMONY, PILOT MODES, CRITERIA
0 AVIONICS/PROPULSION INTEGRATION REQUIRE FLIGHT INVESTIGATION
0 HIGH "G" CAPABILITY REQUIRED
0 SOFTWARE INTENSESYSTEMS REQUIRED
I
"Artificial intelligence" (AI) is now the term used to mean what we once referred However, much effort needs to be applied in this to as computer capability.
area to determine the best approaches and resulting payoffs in using the AI The figure indicates examples of concept during real-time computer operation.
applications.
APPLICATIONOF AI METHODS TO FCS ANDAIRCRAFTTECHNOLOGIES
(NEARAND LONGTERM)
PVI, PILOTDECISION AIDIWG,PILOT AS SYSTEM MANAGER
FAULT-TOLERANT COMPUTING (REAL TIME) ANALYSIS/SYNTHESIS PROGRAM (NON-REAL TIME)
INTEGRATED PROPULSION/CONTROL SYSTEMS
ADAPTIVECONTROLLERS
OPTIMIZATION WITH REGARD TO HANDLING QUALITIES, TERRAIN
FOLLOWING AND AVOIDANCE, TURBULENCE, FLUTTERSUPPRESS ION
AND LOADALLEVIATION, ETC,
"SUPER-MANEUVERABLE" AIRCRAFT
AVIONICS
ROBOT KS
MANUFACTURING QUALITYASSURANCE AND CONTROL
INDUSTRIAL CONTROL
FACTORY OF-THEFUTURE c MILITARY AIRCKAFT KESEARCHOPPORTUNITIES FOR THE FUTURE Robert C. Schwanz Air Force Wright Aeronautical Laboratories Wright-Patterson Air Force Base, Ohio First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 DESIGN OF DECENTRALIZED CONTROL SYSTEMS: "INTEGRATED CONTROL" The objective is to develop a methodology for the design of control systems for interacting dynamical systems which employ only local measurements and control devices.
ENVIRONMENTAL TRAJECTORY OPTIMIZATION ENGINE CONTROL CONTROL OPTIMAL WEAPON DELIVERY MINIMUMTIME CLIMB ELECTRONICS STEADY STATE OPERATION TRANSIENT CONTROL MINIMUMFUEL LIFESUPPORT MINIMUMTIME INTERCEPT MODERECONFIGURATION .
DIFFERENTIAL GAMES .
.
FLYING QUALITIES .
REDUCED STATIC STABILITY NOZZLE CONTROL SIX DEGREES OF FREEDOM ALTITUDE HOLD ALPHA-G LIMITING HEADING HOLD MODERECONFIGURATION .
POWER CONTROL STEADY STATE OPERATION TRANSIENT CONTROL CONTROL CONFIGURE0 WEAPON GUIOANCE .THROAT EXITPRESSURE SHOCK POSITION AN0 CONTROL RIDECONTROL THREAT IDENTIFICATION STRUCTURAL FATIGUE REDUCTION THREAT AVOIDANCE MINIMIZEMISS (;- MANEUVER LOADCONTROL DISTANCE .
FLUTTER MODECONTROL STORE MANAGEMENT .
SEEKER POINTING CAMBER CONTROL AND INITIATION - HISTORICAL PERSPECTIVE Relative to the design of decentralized control systems, recent publications by the IEEE (e.g., ref.
1) and results of military development programs indicate that serious engineering problems prevent reliable control of interacting dynamical systems.
“LARGE-SCALE SYSTEMS AND DECENTRALiZED CONTROL"
l .(SPECIAL ISSUE, iEEE TRANS. AUTO. CONTROL, APRIL 1978)
l iNEFFICIENT OPERATION OFLARGE SCALE INTERCONNECTED SYSTEMS l , LACK OF FUNOAMENTAL UNOERSTANOING OF PHYSICSMOOELS l LACK OF COORDINATED CONTRZLSTRATEGIES . USE OF DETERMINISTIC STATICSTRATEGIES ON STOCHASTIC OYNAMICSYSTENlS l EXISTING TOOLS FORCENTRALlZED CONTROL AREINAPPROPRIATE l SERVOMECHANISM THECRY l RECENT THEORY - MAXIMUM T?RINClPLE . LYAPUNOV STABILITY - ESTIMATION - DYNAMIC PROGRAMMiNG l IMPLEMENTATICN REIIUIREMENTS UNKNOWN . NEEOFOR AN0 COSTOF COMMUNICATION CHANNELS l FIOELITYAND RELIABILITYOF INFORMATiON l ALLOWABLE TIME OELAYSIN INFORMATION . ..
_ .
AVAILABLE DECENTRALIZED CONTROLDESIGN METHODS The current practice is to employ both hierarchical and heterarchical design procedures. They are evolutions of the centralized control theory design methods of the last 30 years. Current development applications depend heavily upon insight gained from centralized design of large-scale systems. Research activity seeks to understand relationships among the decentralized control theories.
CENTRALIZED CONTROL (DEVELOPMENT) DECENTRALIZED CONTROL (RESEARCH) \ a CLASSICAL SERVO-MECHANISM STABiLITY METHODS RESEARCH l MODEL APPRDXlMATION I \ - BOOE - RrJOT LOCUS e LINEAR GUADRATIC ‘2 - l\!ASHSTRATEGY DEVELOPMENT RESEARCH NON-CLASSICAL .
- STACKELBERG STRATEGY - EIGENVALUE I EIGNEVECTOR I?
OPTIMAL DECENTRALIZED FEEDBACK ASS!GNMENT - CONSTRAINED OUTPUT FEEOEACK - LYAPUNOV STABILITY - FiXEOCONTROL STRUCTURE - SEPARATION THEOREM - LINEARGUADRAT:C REGULATOR OPTiMAL DECENTRALIZED FILTERING - KA’LMAN FILTER - t RESEARCH AND DEVELOPMENTNEEDS IN DESIGN OF DECENTRALIZED CONTROL SYSTEMS Research needs include the development of (1) clearer mathematical relation- ships among the various existing methods and (2) the practical significance of the Witsenhauser nonlinear counter-example for linear Gaussian design. Development needs include the definition of (1) suitable dynamics problems of varying complexity for testing of new methods and (2) an efficient computerized methodology that minimizes mathematical complexity and presents the physics essentials of both problem and solution. This method should store user experience in a data base.
FLYING QUALITIES OF ADVANCED VEHICLES The objective is to manually employ the decentralized control of many on-board systems to achieve full dynamics control in highly maneuverable vehicles.
HEAD-DOWN DISPLAYS HEAD-UP l-77+ NAVIGATION
SYSTEM fNTEGRATlON EMlPHASlZES CONTROL
USING LARGER MEASUREMENT SETS
HISTORICAL PERSPECTIVE; FLYING QUALITIES/CONTROL SYSTEM DYNAMICS Recent development programs have experienced manual control problems as the flight control system increases in complexity. If increasing complexity occurs in many control systems, can the manual control problem become easier?
PROBLEM AREA PROBLEM SOURCE PROBLEM SOLUTION BLENDEO AOA I INSTALL PITCH RATE TAKE OFF AND LOAD FACTOR COMMAND SYSTEM, LANDING COMMAND SYSTEM, AOA SIGNAL FADE INCREASE SIGNAL FADE OUT TIMETO1.1 SEC OUT TIME = .Ol SEC TOO SENSITIVE TO CAS MODIFICATION ROLL RACHETING SMALL INPUTS TIMEDELAY FCS PITCH TIMEDELAY Pi0 PRONE AT TOUCHDOWN IN PITCH ADJUSTED PITCH PROBLEMS EXCESSIVE INITIAL ?
RESPONSE DELAY AT TOUCHDOWN Pi0 PITCH PROBLEMS ? ?
AT TOUCHDOWN ,I MANUAL TERRAIN URGEAMPLITUDE RESIDUAL OSCILLATION FOLLOWING PITCH DAMPING FROM PILOT INPUTS ELIMINATED LANDING APPROACH PI0 ?
?
HISTORICAL PEPSPECTIVE; FLYING QUALITIES/DISPLAY DYNAMICS Recent develoiment programs have also been characterized by displays designed with inattention to d-isplay/computer dynamics. Displaying additional information from many decentralized systems may decrease manual flight safety.
FLYING QUALITIES - LEVEL 1 .. PilOT INPUT
TOAIRPLANE RESPONSE400 MS
0 TYPICAL PILOT TRACKING TIME DELAY, c= 300 MS
. EFFECT OFC.25 CM QUANTIZATION OFERROR DISPLAY =sa c = 34% = 100 MS
0 TYPICAL DISPLAY TIME DELAYS
DISPLAY AIRCRAFT FLIGHT PATH MARKER, 30 MS; AOAINDICATOR, 50 MS a TARGET PREDICTOR, 70 rv~s PITCH LADDER, 50 MS CALCULATION, 20 MS REFRESH NAVMAPCHANGE, 1 SEC DATE UPDATE, 40 MS; REFRESH, 1 SEC
CD
HISTORICAL PERSPECTIVE: FLYING QUALITIES/VEHICLE DYNAMICS Much of the nonlinear dynamics information in the MIL SPEC/MIL STD 8785C (ref.
2) must be made more quantitative to manually control many dynamical.processes in highly maneuverable tasks. What does a pilot require to utilize nonlinear dynamic phenomena? What is too much of a good thing? What does he want if some system failure occurs?
LINEAR + NCNLINEAR HISTORICAL PERSPECTIVE: FLYING QUALITIES/VEHICLE DYNAMICS (CONCLUDED) The MIL SPEC/MIL STD 8785C (ref. 2) describes many design objectives as eigenvalue/eigenvector relationships.
Typical of these are the n/cc versus wn charts describing manual control requirements at "moderate frequencies."
13.8 n/0 g’s/RAD RESEARCHAND DEVELOPMENTNEEDS IN FLYING QUALITIES OF ADVANCEDVEHICLES Research needs include the parametric characterization of nonlinear systems: Volterra series Least-squares errors projection Least-squares error/orthonormal projection Energy concepts Development needs include the evaluation of experimental procedures and data used to characterize flying qualities boundaries for centralized/decentraliied control; the definition of metrics to identify manual/automatic system interface boundaries; and the creation of a generic, nonlinear dynamics, manned flight simulator.
REFERENCES 1. Athans, M.: Guest Editorial on Large-Scale Systems and Decentralized Control, IEEE Transactions on Automatic Control, vol. AC-23, no. 2, April 1978, pp. 105-106.
2. Military Specifications, Flying Qualities of Piloted Airplanes, MIL-F-8785C, 5 November 1980.
OPPORTUNITIES FOR AIRCRAFT CONTROLSRESEARCH Thomas B. Cunningham Honeywell, Inc.
Systems and Research Center Minneapolis, MN First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 AIRCRAFT PROBLEMSTHAT DRIVE CONTROLTECHNOLOGY In the course of this discussion, I'd like to deal with what I consider to be four categories of aircraft problems which drive our technology.
First are highly unstable vehicles that are emerging. Second, control tech- nology is being driven by the current thrust and ultimate benefits that can be achieved by expanding the flutter boundary of combat aircraft. Third, we are witnessing, both in the fixed-wing and the rotary-wing world, a new emphasis on low-level penetration. as I'll discuss, will have some This, direct impacts on control technology. Fourth, in general, we along with other members of the avionics community have suffered some setbacks in our attempts to demonstrate "ilities" in the past. NASA and DOD have made great strides in this area; however, improved "ilities", if I can use the word as such, will continue to drive our control research technology.
. HIGHLY UNSTABLE VEHICLES l FLUTTER SPEED BOUNDARY EXPANSION . LOW LEVEL AUTOMATED FLIGHT . IMPROVED “ILITIES” L CONTROLRESEARCHAREAS Breaking down our discussions into technology research areas, I'd like to deal with control from a system theoretic standpoint, particularly multi- variable control theory and adaptive control. Many of the aircraft techno- logy drivers I discussed earlier are going to affect what we do in the future of physical device research, particularly actuators, but also sensors, computers, and data transmissions. Some overall system concepts are going to be impacted by these directions, such as fault-tolerant archi- tectures and some continued work required in the analytical fault-tolerant areas. This latter category is technology that we have pretty well in hand in our controls area, but other avionics system people really need the bene- fit of our experiences. Two other areas I'll just touch on basically are what control technology will have as an impact for future flight research and some comments on higher order language software.
SYSTEM THEORETIC l MULTIVARIABLE CONTROL THEORY . ADAPTIVE CONTROL PHYSICAL/DEVICE RESEARCH . ACTUATION l SENSORS l COMPUTERS l DATA TRANSMISSION SYSTEM CONCEPTS l FAULT-TOLERANT ARCHITECTURES . ANALYTICAL FAULT TOLERANCE OTHERS . FLIGHT RESEARCH . HOL SOFTWARE BANDWIDTH INCREASES DUE TO RELAYED STATIC STABILITY Beginning with highly unstable aircraft, we may be witnessing a signi- ficant increase in bandwidth requirements for fighter aircraft, in particu- lar due to the desire to achieve the benefits of low drag and maneuvera- bility at supersonic speeds.
This chart shows a simple single-input/single-output Bode relation- For instance, we can think of instability being measured by the ship.
magnitude of the root that is farthest to the right in the right half plane. If we denote that by the letter "a" in a frequency response plot, we'd like to be 6 dB in gain above this position as a rule of thumb.
Another rule of thumb for control design engineers is that we'd like to be in the area of 20 dB per decade roll-off in the region of crossover. This simple relationship would then put the crossover frequency at 2 a. Like- wise, for stability and robustness, we'd like to be at least 6 dB below zero gain at a frequency of 4 a.
Two examples of the loop requirements to stabilize the vehicle are the X-29 and the F-16. The only one that is in production is the F-16 fighter.
Its worst flight condition in terms of stability contains a root at +2, which means our crossover frequency is at 4 radians per second, and we must However, if we be rolled off to the tune of 6 dB at 8 radians per second.
look at the X-29 Forward Swept Wing fighter, we see that at its most unsta- ble flight condition, it has a root at +7, which is approximately 3 l/2 times what the F-16 represents. This impacts our control bandwidth because we need a crossover at 14 radians per second, and we must be rolled off 6 dB at 28 radians per second.
Historically, flight control designers have had 6 dB to model systems that have the highest accuracy near crossover, i.e., 2 a and 2 a half of decade either side. In the case of the X-29 and potentially the emerging this could easily mean that we need to know in ATF class of fighters, greater detail the dynamics of the vehicle about 3 l/2 times higher in frequency than we've ever dealt with before. This would logically include more detailed knowledge of the bending/aeroelastic characteristics of the vehicle.
GAIN LOOP b . 21 44 FRE(IUENCV = UNSTABLE ROOT FREQUENCY EXAMPLES: F.16FIGHTEA X.29FORWARD SWEPT WING FLIGHT RESEARCH One of the initial impacts that high bandwidth control poses is in the area of flight research, and the goal as stated earlier is to achieve model uncertainty reduction in the vicinity of control bandpass.
In the case of highly unstable vehicles, this means that we have to deal with higher and higher frequencies, and one of the ways we achieve model uncertainty reduc- tion is through a series of flight examinations and parameter identifica- tion. The payoff, obviously, is to achieve an ultimate lowering of the eventual actuator rate excitations for high bandwidth control.
GOAL -‘MODEL UNCERTAINTY REOUCTION IN CONTROL PASSBAND CHALLENGE -‘PARAMETER ID AT HIGHER FREQUENCIES PAYOFF + LOWER ACTUATOR RATE EXCITATIONS FOR HIGH BANDWIDTH CONTROLS SENSORS Let 'us look at how this nature of highly unstable vehicles impacts various hardware components. Historically, flight control sensors have had ample bandwidth. We at Honeywell feel that if more bandwidth is needed, then expansion is already feasible within the state of the art.
Another issue that is cropping up here more and more is the notion of shared flLght control/navigation/weapon delivery sensors. The placement issue is the one that we have to deal with in the flight control community, and the real goal here is to achieve a robust multivariable control for sensor fault tolerance and high bandwidth performance in the face of ill-placed and dispersed sensors. This technology development should continue.
l FLIGHT CONTROL SENSORS - BANDWIDTH EXPANSION FEASIBLE l SHARED FLIGHT CONTROL/NAVIGATION/WEAPON DELIVERY SENSORS -PLACEMENT ISSUES: ROBUST MVC FOR SENSOR FAULT TOLERANCE HIGH BANDWIDTH PERFORMANCE ACTUATION TECHNOLOGY The next technology area relating to controls is in the area of actua- We have three issues that we must deal with here.
tors.
The first issue is fault tolerance. There is an emerging class of technical developments dealing with digital servo electronics which will indeed make the actuators more fault tolerant.
This is required, because if one looks at the reliability analysis of a current flight control system, we see that from a research standpoint we have been dealing effectively with computers and somewhat effectively with sensors, but the real reliability bottleneck in aircraft flight control is the actuator system.
The next issue is this bandwidth expansion that we've been discussing.
Although I'm not an expert in actuator technology, I'll pose a couple of issues and concerns that I have. One concern is whether we have the hydrau- lic technology available to achieve the high authority and rate/bandwidth we need for unstable vehicles. I'll also note that electromechanical actuators are emerging, which might give us high authority and expanded rate and band- width. This technology development should be pursued.
Third, there is the rate expansion itself for highly unstable vehi- cles. One possible solution, within the system theoretic framework, is the capability to formalize an optimal blending, if you will, of multivariable control solutions which allows us to share the rate requirement among a num- ber of different surfaces.
. FAULT TOLERANCE - DIGITAL SERVO ELECTRONICS . BANDWIDTH EXPANSION -HYDRAULIC +COMPRESSlBILITY LIMITS - ELECTRO-MECHANICAL + EMERGING . RATE EXPANSION - MVC SOLUTIONS FOR OPTIMUM BLENDING COMPUTERS Another important element in our control loop is the onboard digital flight computer. Computer technology is being driven by a number of differ- ent applications, flight control being one. For instance, throughput is really being driven by signal processing requirements on various aircraft and spacecraft applications. The word length related to the accuracy of the computations is being driven more by other avionics functions, such as navi- gation. Memory, likewise, is being driven by other functions also. Some of the storage requirements for digital landmass data, etc., come to mind.
One area that is currently being driven by flight control is fault tolerance. In the emerging set of research issues from a hardware stand- point, one needs to look at integrated-functions architectures such as integrated flight/navigation and integrated flightlpropulsions.
Software on the other hand is being driven to higher order languages such as Ada. We have to be careful in the controls community not to accept massive, unreliable compilers that aren't necessary to do control work. One possible solution that is being explored by the software community is to allow defined subsets of higher order languages to be used by highly flight crucial functions like flight control.
In summary, control problems in general do not drive computer technol- Certainly the first three issues I've listed here are being driven by ogy* other requirements. Fault tolerance is currently being driven by flight control, but as we'll see when we get into the impact of low-level penetra- tion, we will find that other avionics systems indeed need the same kind of fault tolerance that flight control currently does.
. THROUGHPUT + DRIVEN BY SIGNAL PROCESSING . WORD LENGTH + DRIVEN BY NAVIGATION . MEMORY + DRIVEN BY OTHER FUNCTIONS . FAULT TOLERANCE HARDWARE: INTEGRATED FUNCTIONS ARCHITECTURES SOFTWARE: HOL COMPILER SUBSETS BOTTOM LINE + CONTROL PROBLEMS 00 NOT DRIVE COMPUTER TECHNOLOGY MULTIVARIABLE CONTROL THEORY I would like to comment on aircraft research problems that drive the control theoretical developments.
Two areas I'll talk about are the actual basic mathematical theory itself and additionally how we might look at a research area for applying these techniques. First, in my discussion of unstable vehicles, I've alluded to uncertainty reductions and new regions in frequency that have yet to be dealt with. Likewise, in our development of mathematical theory, we have to formalize ways of dealing with uncertain- For instance, if we can generate a mathematical description of ties.
uncertainty in a given frequency range, we should seek to formalize our control design problem to deal directly with that level of uncertainty.
There is a body of research being conducted at Honeywell and other places actually dealing with what we call structured uncertainties.
The other side of the structured uncertainty issue is the design itself. One design technique we're looking at is direct frequency domain optimization that attempts to deal with a formalized approach to achieve performance goals in the face of uncertainty constraints.
Uncertainty representations for practical aircraft control design prob- lems have not been developed beyond simple examples. Therefore, formalized uncertainty boundaries for practical examples should be examined.
The next item concerns the state of the art in analysis and design itself for the whole hierarchy of preliminary design to flight control development, to actual production of flight control systems. We in the control theoretic community have been remiss for decades in not providing the end-user, i.e., the guy who's actually going to design a flight control system, with the proper CAD tools to do his job. And even though we've taken a useful step back towards the frequency domain, the ultimate flight control designer will never use the tools if all he has is a published paper out of the Transactions on Automatic Control.
Third, more directly to the flight control problem itself, I alluded earlier to some sharing of surfaces to achieve some overall higher bandwidth Coupling this idea with the proliferation of control effecters on goals.
board modern aircraft, one should look at ways to formalize the control law design.
Once again, a driving goal is to achieve higher bandwidth control for highly unstable vehicles. In addition, emerging classes of work include the areas of integrated flight and propulsion control and also some rather higher risk areas dealing with robust reconfiguration of control surfaces in combat situations, which could have high payoffs.
. DEVELOPMENT: -STRUCTURED UNCERTAINTIES - DIRECT FREQUENCY DOMAIN OPTIMIZATION . APPLIED: - FORMALIZE0 UNCERTAINTY BOUNDARIES -ANALYSIS AND DESIGN TOOLS - CA0 -MULTIPLE EFFECTOR BLENDING FOR COMMON GOALS HIGHLY UNSTABLE VEHICLES INTEGRATED FLIGHT/PROPULSION ROBUST RECONFIGURATION I ADAPTIVE CONTROL RESULTS There have been some discussions about the usefulness of the adaptive control theory that has been developed over the last couple of decades. I'd like to go on record as saying that despite these emerging reservations, we - at Honeywell are pretty proud of the adaptive control work that we've per- formed in the past, particularly under NASA sponsorship. This chart shows some of the more recent efforts at the Honeywell Systems and Research Cen- ter. This figure shows some successful adaptive work that we've done with NASA and also a very effective technology spinoff. Without going into detail, we essentially used the techniques we developed for the F-8 aircraft demonstrated at Dryden in 1977) to and the CH47 helicopter (which were design an off-shore drilling ship positioning concept that is now in produc- tion.
ADAPTIVE CONTRACTSWITH HONEYWELL EXXON/HONEYWELL DEVELOPMENT (1979) F-8C DESIGN (1974 - LRC) FLIGHT TEST (1977 - DFRC) DEEPWATER OFFSHORE DRILLING SHIP ADAPTIVE CONTROL VALT DESIGN (1977 - LRC) ACTIVE FLUTTER CONTROL Active flutter control has some obvious payoffs that have been recog- nized for a number of years. The technology necessary to perform this is not mature and also the notion of active flutter mode control, particularly the frequencies we're dealing with, is a risky business.
It is one problem to stabilize an airfoil flutter condition on a wing with known characteris- tics, however, from a practical standpoint, future fighter aircraft flutter mode control needs to deal with wings that,have numerous dynamic configura- tions due to the different kinds and placements of stores.
ADAPTIVE CONTROL This leads us to a body of work that exists in adaptive control which has some demonstrated potential.in dealing with this. The issue is to iden- tify very rapidly the onset of a dynamic change in a configuration and to modify the control structure accordingly. A concept, based upon maximum likelihood estimation, has been demonstrated in wind tunnel tests of active flutter control with wing store changes.
Another payoff area for adaptive control deals with the prospect of doing onboard real-time reconfigurations due to battle damages for future military aircraft.
RAPID PARAMETER IDENTIFICATION: . ACTIVE FLUTTER CONTROL WITH CHANGING CONDITIONS . BATTLE DAMAGE RECONFIGURATION - ADVANCED AUTOMATIC TERRAIN FOLLOWING/TERRAIN AVOIDANCE The final two areas are combined treatment of low-level penetration and This figure points out one area of great interest improved "ilities."
within DOD; this is the notion of combined terrain following/terrain avoid- As shown schematically, this involves flying ance and threat avoidance.
very low altitudes automatically while incorporating high g maneuvers, and a high level of flight safety. All of the appropriate pilot interface, avionics subsystems which interface to attack this problem become flight This would certainly include the guidance portion of the safety crucial.
flight control area, navigation, and portions of the mission management.
. ALGORllHi”S FORAUTOMATIC, SIMULTANEOUS TERRAIN FOLLOWING~ERRAIN AVOIDANCE -LOW ALTITUOES -HIGH, MANEUVERING 0 PILOT INTERFACE . FLIGHT SAFETY l SENSITIVITY ANALYSIS 0 COYPUTATI0NA.L REOUIREMENTS &-j-$# ;;PAbz GROUNO PATH PROJECTION OFPRESELECTEDPATH LOW-LEVEL PENETRATION In addition to the fixed-wing terrain following/terrain avoidance and threat avoidance, there is an emerging set of concepts that the Army would like to develop for a single-seat light helicopter for nap-of-the-Earth The challenge that I believe exists is to somehow integrate flight, flight.
navigation, propulsion, weapons, etc., in a fault-tolerant system to allow us to do nap-of-the-Earth flight with high performance and a high degree of safety.
One research area is sensor blending. I think our estimation techno- We also need more creative failure management solu- logy applies here.
Again, I think estimation technology applies, such as expanding upon tions.
things like analytic redundancy, plus development of emerging AI concepts based on expert systems and advanced planning. Also there is this critical need to automate verification and validation of systems, both the hardware and the software, for future flight management systems. And finally in the area of flight path management, we need to somehow go down the road and actually install trajectory optimization capabilities that operate in real time for onboard applications. In addition to that, there are some front-end decision making functions that perhaps could be handled by AI planning concepts.
EXAMPLES . AUTOMATIC TF/TA/OA . SINGLE-SEAT LIGHT HELICOPTER - NOE CHALLENGE 0 FAULT-TOLERANT INTEGRATED FLIGHT/NAV/ PRDPULSIONMTEAPDNS RESEARCH AREAS 0 SENSOR BLENDING - ESTIMATION . CREATIVE FAILURE MANAGEMENT - ESTIMATION, Al 0 AUTOMATED VERIFICATION AND VALIDATION . FLIGHT PATH MANAGEMENT -TRAJECTORY OPTIMIZATION, Al PROPULSION CONTROL TECHNOLOGY Edward C. Beattie United Technologies Corp., Pratt & Whitney Group East Hartford, Connecticut First Annual NASA Aircraft Controls Workshop NASA Langley Research Center Hampton, Virginia October 25-27, 1983 Turbine Engine Control Evolution Control systems for both commercial and military gas turbine engines are being transitioned in an orderly fashion from pure hydromechanical to full-authority digital electronic control in order to obtain the associated operational and Additional benefits will be available with further performance benefits.
control system technology and through increased advancements in electronic integration with aircraft systems. At Pratt & Whitney, supervisory digital electronic control systems are in current operational service, and full-authority will be in service on upcoming models of digital electronic control systems commercial and military engines.
FULL. AUTHORITY SENSORS I ACTUATORS HYDROMECHANICAL I CONTROL Digital Electronic Control Systems Are Incorporated in Pratt and Whitney Military and Commercial Engines Supervisory digital electronic control systems are in current operational service on the FlOO engine in F15 and F16 ajrcraft and on the JT9D-7R4 engine in Boeing 767 and Airbus A310 aircraft. Full-authority digital electronic control systems have been developed for the F1.OO engine, available for future FL5 and F16 aircraft and for the PW2037 engine in the Boeing 757. The PW2037 engine and control system will be certified by the end of 1983. Pratt & Whitney is also developing a full-authority digital electronic control system for its new PW4000 engine which will be available for future versions of all wide-bodied aircraft.
Supervisory
l FIOO - F15, F16
B-767, A31 0, A300-600
l JT9D-7R4 -
Full Authority
l PW2037 - B-757
l FIOO DEEC - F15, F16
. PW4000 - B-767, B-747, A310, A300, MD100
Electronic Control System Has Substantial Experience Pratt & Whitney's digital electronic control systems have substantial operational The current supervisory control system for the FlOO engine and test experience.
The JT9D-7R4 engine's has over 1.5 million hours of in-service operating time.
supervisory control system, recently introduced into service on Boeing 767 and Airbus A310 aircraft, has over 130,000 hours of operating time in revenue Pratt & Whitney has also conducted or participated in, with NASA and service.
the Air Force, a number of electronic control system technology development and demonstration programs dating from the early 1970's. These programs have played a key role in developing digital electronic control system technology to a state With the advent of the of readiness for incorporation in production engines.
full-authority electronic control system, continuing technology programs are required to obtain the benefits of control system technology advancements.
- 1.500.000 + flight hours, F15 and F16 engines F 100 supervisory JT9D-7R4 supervisory - 767 airline service; 130,000 + engine hours JT8D EEC reliability evaluation - 300,000 + hours on 727 aircraft JT8D EPCS, 1975 - Boeing demonstration at Boardman, Oregon TF30 IPCS, 1975 - Fl 11 flight test at Edwards AFB Fl 00 DEEC, 1978 - Fl 00 altitude test at Tullahoma - F401 altitude test at NASA Lewis F401 FADEC, 1979 JTSD EPCS, 1980 - 747 flight test at Boeing FIOO DEEC, 1980 - F15 flight test - F15 flight test Fl 00 DEEC, 1983 FIOO DEEC, 1983 - F16 flight test - Ground, altitude and flight test PW2037 full authority, 1983 JT9D Supervisory Electronic Control System The JT9D-7R4 Supervisory Electronic Control System incorporates a .conventional hydromechanical control system coupled with a digital eletxronic control which trims the hydromechanical control to provide accurate control of the desired Digital communication links with the cockpit provide engine power setting.
simplified cockpit power setting procedures. The electronic control for the JT9D Design of the engine is mounted on the engine's fan case and is air cooled.
electronic control was completed in the late L97O's, and the engine was introduced into service in 1982 on Boeing's 767 aircraft and in 1983 on Airbus' system is similar to that of A31@ aircraft. Operation of the JT9D-7R4 control the FlOO supervisory control system.
Fuel control :,‘, .I ,“‘:%::.
,> Supervisky control L.-r.*.. Ij r _ - -’ / PW2037 Control System The PW2037 Control System incorporates a dual-channel, full-authority digital electronic control and utilizes full redundancy of all inputs and outputs. As with the JT9D-7R4 supervisory control system, digital communication links with the cockpit provide for optimum aircraft/engine control communication and simplified power setting procedures. electronic The controi design was completed in 1981 and incorporates advancements in circuit integration which provide for a substantial increase in functional capacity, compared to the supervisory control, with a significantly less than proportional increase in count. Engine and control system certification will be completed in parts - December, and introduction into service on Boeing's 757 aircraft is 1983, scheduled for November 1984.
NTROL SYSTEM
Electronic Engine Control
>
,, I, j, ‘, ,’ ..$’ “*,
‘X
<
”
#$buaCchannel Full-Authority Digital electmnic
>,c,,
,/$,c ~Co&rbl (FADEC)
X2+’ ^
*u,&‘,’
‘$~~~~‘Du~l redundancy 6f all inputs and outputs
II
. .- FlOO Digital Electronic Engine Control (DEEC) System A full-authority digital electronic control system has been developed for the FlOO engine. This control system incorporates a single-channel electronic control and provides hydromechanical backup capability to meet mission reliability requirements for single-engine applications.
The electronic control design incorporates the same electronic component technology as the PW2037 control. DEEC control systems have successfuliy completed sea level and altitude engine testing and flight testing on NASA's F15 and Air Force F16 aircraft.
_-._- -.. .-
__-_-.. -, .__... - . . .-
I
Propulsion Control: Research Needs and Opportunities Significant advances in propulsion control system technology have led to the full-authority digital electronic near-term incorporation of controls in Substantial propulsion system benefits can be obtained production engines.
in control system technology. lntegration of through continuing advancements aircraft and engine control systems and functions can provide optimized hardware and control modes. Development and application of optical configurations high-temperature electronics, and continuing advancements in interfaces, large-scale circuit integration are likely candidates for electronic control improvements to weight, and design provide size, reliability benefits.
Continuing development of advanced sensor and actuator concepts is required to evaluate potential benefits of optical interfacing and design concepts for increased reliability. NASA has sponsored a number of programs concerned with the development of advanced concepts for detection, isolation, and accommodation Continuing work is required in this area to develop optimum of sensor faults.
redundancy management concepts. Adaptive ccntrol modes can contribute to more predictable system operation and may show significant benefits for integrated control modes. Improvements in propulsion system modelling aircraft/engine techniques become increasingly important as advanced propulsion system design concepts are developed for performance and operability improvements. Finally, need to be developed for a software tools and documentation/testing concepts software development methodology which can meet high-level-language-based regulatory agency certification requirements.
l Aircraft/propulsion control system integration
l optics
l High-temperature electronics
l Large-scale integration
l Sensor and actuator concepts
l Redundancy management
l Adaptive control modes
l Propulsion system modelling
l High-level-language-based software methodology
--- -- __ __._,__^~. - -.. -.- - --- ---.-_._--.--:. _ .- .,
.=
--~. I 2. Government Accession No. 1 3. Recipient’s Catalog No.
1,' NASA CP-2296 4. Title and Subtitle NASA AIRCRAFT CONTROLSPESEARCH - 1983 7. Author(s) Gary P. Beasley, Compiler 9. Performing Organization Name and Address NASA Langley Research Center ll. Contract or Grant No.
Hampton, VA 23665 12. Sponsoring Agency Name and Address Conference Publication National Aeronautics and Space Administration 14. Sponsoring Agency Code Washington, DC 20546 t 15. Supplementary Notes 16. Abstract This publication contains the proceedings of the First Annual NASA Aircraft Controls Workshop, held October 25-27, 1983, at NASA Langley Research Center.
This workshop highlighted ongoing aircraft controls research sponsored by NASA's Office of Aeronautics and Space Technology and provided a forum for critique of ongoing research as well as suggestions for needed research from controls experts or users of control technology. The workshop was initiated in response to a recommendation from the NASA Advisory Council's Informal Subcommittee on Aircraft Controls and Guidance.
About 200 aircraft controls experts from industry, government, and universities participated in the workshop.
The workshop consisted of 24 technical presentations on various aspects of aircraft controls, ranging from the theoretical development of control laws to the evaluation of new controls technology in flight test vehicles. It also included a special report on the status of foreign aircraft technology and a panel session with seven representatives from organizations which use aircraft controls technology, This panel addressed the controls research needs and opportunities for the future as well as the role envisioned for NASA in that research.
Input from the panel and response to the workshop presentations will be used by NASA in developing future programs.
-- __--~---. .,_- -. -- - -__ ___--- 7. Key Words (Suggested by Author(s)) 18. Distribution Statement Aircraft controls Handling qualities Unclassified - Unlimited Parameter identification Digital control systems Subject Category 08 ---- ~ 22. Price’ 9. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages Unclassified Unclassified 608 A99 --__ ’ For sale by the National Technical InformationService, Sprlngfleld. Virginia 22161 *U.S. GOVERNMENTPRINTINGOffICE: 1984 - 739- 010 6 0 REGIONNO. 4