Document
NASA/TM–2014-218671
Artificial Intelligence Based Control Power
Optimization on Tailless Aircraft
ARMD Seedling Fund Phase I Final Report
Frank H. Gern and Dan D. Vicroy Langley Research Center, Hampton, Virginia Sameer B. Mulani, Rupanshi Chhabra, Rakesh K. Kapania, and Joseph A. Schetz Virginia Polytechnic Institute State and University, Blacksburg, Virginia Derrell Brown and Norman H. Princen Boeing Research and Technology, Huntington Beach, California
December 2014
NASA STI Program . . . in Profile Since its founding, NASA has been dedicated to the • CONFERENCE PUBLICATION.
advancement of aeronautics and space science. The Collected papers from scientific and technical NASA scientific and technical information (STI) conferences, symposia, seminars, or other program plays a key part in helping NASA maintain meetings sponsored or this important role. co-sponsored by NASA.
The NASA STI program operates under the auspices • SPECIAL PUBLICATION. Scientific, of the Agency Chief Information Officer. It collects, technical, or historical information from NASA organizes, provides for archiving, and disseminates programs, projects, and missions, often NASA’s STI. The NASA STI program provides access concerned with subjects having substantial to the NTRS Registered and its public interface, the public interest.
NASA Technical Reports Server, thus providing one of the largest collections of aeronautical and space • TECHNICAL TRANSLATION.
science STI in the world. Results are published in both English-language translations of foreign non-NASA channels and by NASA in the NASA STI scientific and technical material pertinent to Report Series, which includes the following report NASA’s mission.
types: Specialized services also include organizing • TECHNICAL PUBLICATION. Reports of and publishing research results, distributing completed research or a major significant phase of specialized research announcements and feeds, research that present the results of NASA providing information desk and personal search Programs and include extensive data or theoretical support, and enabling data exchange services.
analysis. Includes compilations of significant scientific and technical data and information For more information about the NASA STI program, deemed to be of continuing reference value. see the following: NASA counter-part of peer-reviewed formal professional papers but has less stringent • Access the NASA STI program home page at limitations on manuscript length and extent of http://www.sti.nasa.gov graphic presentations.
• E-mail your question to help@sti.nasa.gov • TECHNICAL MEMORANDUM.
Scientific and technical findings that are • Phone the NASA STI Information Desk at preliminary or of specialized interest, 757-864-9658 e.g., quick release reports, working papers, and bibliographies that contain minimal • Write to: annotation. Does not contain extensive analysis. NASA STI Information Desk Mail Stop 148 • CONTRACTOR REPORT. Scientific and NASA Langley Research Center technical findings by NASA-sponsored Hampton, VA 23681-2199 contractors and grantees.
NASA/TM–2014-218671
Artificial Intelligence Based Control Power
Optimization on Tailless Aircraft
ARMD Seedling Fund Phase I Final Report
Frank H. Gern and Dan D. Vicroy Langley Research Center, Hampton, Virginia Sameer B. Mulani, Rupanshi Chhabra, Rakesh K. Kapania, and Joseph A. Schetz Virginia Polytechnic Institute State and University, Blacksburg, Virginia Derrell Brown and Norman H. Princen Boeing Research and Technology, Huntington Beach, California National Aeronautics and Space Administration Langley Research Center Hampton, Virginia 23681-2199
December 2014
Acknowledgments The authors would like to thank the NASA Aeronautics Research Mission Directorate (ARMD) for supporting this research through a Phase I ARMD Seedling Fund Award.
The use of trademarks or names of manufacturers in this report is for accu rate reporting and does not constitute an official endorsement, either expressed or implied, of such products or manufacturers by the National Aeronautics and Space Administration.
Available from: NASA STI Program / Mail Stop 148 NASA Langley Research Center Hampton, VA 23681-2199 Fax: 757-864-6500
Abstract
Traditional methods of control allocation optimization have shown difficulties in exploiting the full potential of controlling large arrays of control devices on innovative air vehicles. Artificial neutral networks are inspired by biological nervous systems and neurocomputing has successfully been applied to a variety of complex optimization problems. This project investigates the potential of applying neurocomputing to the control allocation optimization problem of Hybrid Wing Body (HWB) aircraft concepts to minimize control power, hinge moments, and actuator forces, while keeping system weights within acceptable limits.
The main objective of this project is to develop a proof-of-concept process suitable to demonstrate the potential of using neurocomputing for optimizing actuation power for aircraft ® featuring multiple independently actuated control surfaces. A Nastran aeroservoelastic finite element model is used to generate a learning database of hinge moment and actuation power characteristics for an array of flight conditions and control surface deflections. An artificial neural network incorporating a genetic algorithm then uses this training data to perform control allocation optimization for the investigated aircraft configuration. The phase I project showed that optimization results for the sum of required hinge moments are improved by more than 12% ® over the best Nastran solution by using the neural network optimization process.
Nomenclature
ANN = Artificial Neural Network DEP = Distributed Electric Propulsion FEM = Finite Element Model HWB = Hybrid Wing Body OREIO = Open Rotor Integration on a BWB Concept by Boeing S&C = Stability and Control
Purpose
This project investigates the potential of applying artificial intelligence methods like neurocomputing to the control allocation optimization problem of Hybrid Wing Body (HWB) aircraft concepts. Researchers from NASA Langley, Virginia Tech, and Boeing Research and Technology are exploring the use of artificial neural networks to develop innovative control algorithms minimizing control power, hinge moments, and actuator forces to keep system weights within acceptable limits.
HWB platforms feature multiple control surfaces, with large control surface geometries leading to large hinge moments and high control power demands. Due to the large number of control surfaces on an HWB, there is no unique relationship between control inputs and resulting aircraft response, i.e. different combinations of control surface deflections may potentially result in the same maneuver, but with large differences in control power. While traditional methods of control allocation optimization may have limitations in exploiting the full potential of controlling large arrays of control devices, artificial neutral networks (ANN) are inspired by biological nervous systems and have successfully been applied to a variety of complex optimization problems (see e.g. Refs. 2 - 4 ).
This project employs a finite element based aeroelastic HWB model, as well as wind tunnel and flight test data from the Boeing X-48 flight demonstrator to build a database that can be used to train an artificial neural network (ANN) to perform control allocation optimization for tailless aircraft at the conceptual design level (Fig. 1).
Figure 1: Boeing’s X - 48 Blended Wing Body demonstrator (NASA Photo).
Background
HWB aircraft concepts offer the potential to achieve significant fuel burn savings of over 25%, with resulting emissions reductions, as well as community noise benefits, therefore countering 6 , 7 the impact of dramatic increases in future air traffic volume. Recent wind tunnel testing indicates that control authority issues still exist (e.g. stall recovery and three-dimensional coupling effects ). Stability augmentation and control power optimization for these aircraft concepts can be enablers for their success.
The use of artificial intelligence to overcome the shortcomings of conventional methods for control allocation has not been explored to a significant extent in the open literature. A literature search yielded very few results, most of them not applicable to this critical need for HWB configurations. Artificial intelligence offers new ways to optimize control power while minimizing hinge moments and structural loads.
The proposed concept applies to the design and development stages of future air vehicles. If the research proves successful, artificial intelligence will be used to develop optimized control laws in a much more efficient manner than by using traditional methods; and these optimized control laws are an enabler for an entire fleet of revolutionary aircraft. Although the control laws will be developed using artificial intelligence methods, there is no need to implement these tools on the aircraft itself; therefore, no certification issues are anticipated for such a solution.
Approach
The project applies artificial intelligence to the HWB control allocation problem to optimize control surface schedules and minimize control power. An aeroelastic finite element model (FEM) provides the baseline for stability and control (S&C) analyses. Boeing Research and Technology provided control surface data, actuator characteristics, and aeroservoelastic modeling support to generate an accurate representation of the HWB baseline. The FEM model has been used to generate a database of hinge moment and actuation power characteristics for an array of control surface deflections.
A sufficient amount of training data is a prerequisite for successfully training an ANN for control allocation and to determine the proper size and architecture of the ANN, both of which are crucial for the success of applying neurocomputing to the problem of control power optimization.
Phase I of the effort focused on the development of a proof-of-concept process demonstrating the feasibility of using neurocomputing for actuation power optimization (Fig. 2).
Figure 2 : Using an artificial neural ne t work for control allocation optim ization.
In the interest of maximizing the results and impact of the Phase I effort, some modeling simplifications were applied. While these limitations did not affect the validity of the Phase I results, they will be removed in the future to improve the value and practical usefulness of the approach. As will be described in the “Results” section, the total hinge moment sum from all active control surfaces was used as a proxy figure-of-merit to minimize actuation power. Future efforts will include the complete actuator dynamics developed by Boeing during the Phase I and therefore use actual actuation power as a figure of merit for the optimization. The Boeing actuator dynamics model was validated through X-48 wind tunnel and flight testing. The aeroelastic finite element analysis used a symmetric half model for an initial 2.5g pitch maneuver analysis.
Accomplishments
The main objective of the Phase I effort was to develop a proof-of-concept process suitable to demonstrate the potential of using neurocomputing to optimize actuation power for aircraft featuring multiple independently actuated control surfaces (see Fig. 3). All the proposed work tasks and milestones were completed on time and within budget. All deliverables from the project partners were received by the Principal Investigator and directly contributed to the success of the project. The project team successfully laid all the necessary groundwork to continue the initial research in a follow-on project.
Figure 3 : P roof - of - concept process to optimize actuation power through aeroservoelastic neurocomputing .
The key accomplishments from the Phase I project are outlined below: Established a complete proof-of-concept aeroservoelastic neurocomputing process to optimize actuation power for a representative 2.5g symmetric pitch maneuver.
Developed a full aeroelastic model suitable for analyzing a representative HWB platform developed by Boeing. The non-proprietary Boeing OREIO (Open Rotor Engine Integration Figure 4 : The Boeing OREIO features up t o 25 control surfaces and high lift devices.
(OREIO = Open Rotor Engine Integration and Optimization on an HWB ) Figure 5 : Finite element model of the Boeing OREIO used for the study.
and Optimization on an HWB ) concept was chosen for this task (see Figs. 4 and 5). Boeing provided control surface geometries and developed actuator dynamics models based on a Simulink analysis model that was validated through X-48 wind tunnel and flight testing.
Details regarding the structural and aeroelastic modeling of the OREIO configuration can be found in Refs. 10 and 11 .
® 12 Generated an aeroelastic trim database containing up to 2,500 Nastran aeroelastic TRIM ® solutions. For this purpose, the baseline Nastran model is used to perform a large number of maneuver-trim analyses with randomly generated linkage coefficients between the ® different control surfaces. Latin Hypercube Sampling is used to generate random Nastran AELINK card coefficients for each individual Nastran SOL144 static aeroelastic TRIM 13 ® analysis. A Matlab script then reads trim variables, control surface deflections, and stores all this information for post-processing and optimization using artificial neural networks.
Investigated neural network topologies (number of layers/neurons) and learning algorithms to train the neural network using the aeroelastic trim database. A neural network with one hidden layer of up to 300 neurons was chosen as the network to perform the actuation power optimization. The hidden layer neurons were trained using a hyperbolic tangent sigmoid transfer function, while a single output neuron featuring a linear transfer function represents the required actuation power.
Optimized the neural network using a genetic algorithm to develop sets of control surface linkage coefficients minimizing the sum of all control surface hinge moments, which was the figure-of-merit used as a representation of total required actuation power.
Quantified the optimization results by demonstrating a more than 12% improvement over the ® best Nastran solution by using the neural network optimization process.
Results From the Seedling Phase I Effort: Figure 6 shows the probability density function of the training data generated by the FEM analysis for different numbers of test cases. During the network training process, the least squares error between the pre-computed results and ANN predictions is minimized through back-propagation by adjusting the weights and biases of the individual neurons. With each generation, the fitness value is reduced until it is flattening out after about 60 generations, indicating that the neural network is fully trained at this point. A genetic algorithm is used to minimize the absolute sum of the hinge moments.
Figure 6 : T raining data p robability dens i ty function.
® The AELINK control surface linkage coefficients in Nastran define relationships between individual control surface deflections and an “independent” control surface, which in the pr esent case is the center “elevator” surface in Fig. 7. For this study, the range of allowable control surface deflections has been limited to the [-1,1] interval, and the overall analysis data base for the 2.5g maneuver case included 2,500 samples. Future studies will further increase this interval to avoid extrapolation on the critical AELINK coefficients.
Figure 7 : Optimized control surface deflections (in degrees) for a Boeing OREIO 2.5g symmetric pitch maneuver .
The neurocomputing process was applied to a symmetric half model of the Boeing OREIO, featuring eight actuated control surfaces, i.e. seven trailing edge flaps and the rudder (Fig. 7).
To improve engine noise shielding, the OREIO vertical tails are highly canted, resulting in a significant amount of pitch authority for the rudder. To test the potential of the neurocomputing process, optimization of the neural network was performed using both, the AELINK control surface deflection coefficients and the direct values of the individual control surface deflection angles. The obtained control surface deflections minimizing the total hinge moment sum for both cases are highlighted in Fig. 7 and although quite similar, the deflections for both approaches are not identical, thus further highlighting the fact that different control surface deflection schedules may result in similar actuation power requirements.
Table 1 shows that for both cases, the total sum of the hinge moments is less than the best Nastran solution, with more than 12% improvement when using the control surface deflection angles as the input parameter. As a validity check, both control surface deflection data sets Table 1 : 2.5g Maneuver optimization r e sults for the total sum of hi nge moments (in lb - in) as a proxy for actuation power .
® were used as fixed values for a Nastran SOL 144 aeroelastic trim solution to check whether Nastran would reproduce the results when using the predictions from the optimizer. Table 1 shows that the Nastran FEM analysis for the actuation power proxy matched the neural network prediction within numerical accuracy.
Potential Impact on NASA or National Aeronautics Challenges
Reducing actuation power is an enabler for ultra-efficient commercial transport aircraft and therefore directly impacts the National Aeronautics Challenges of simultaneously reducing fuel consumption, emissions, and noise of future air traffic. HWB aircraft and other innovative concepts suitable for this research offer the potential to achieve significant fuel burn savings of over 25%, with resulting emissions reductions, as well as community noise benefits, therefore countering the impact of dramatic increases in future air traffic volume. As a result, the proposed research directly applies to three of the six ARMD Strategic Thrust areas: Innovation in Commercial Supersonic Aircraft Ultra-Efficient Commercial Transports Transition to Low-Carbon Propulsion The process can easily be adapted to other innovative and unconventional configurations.
Potential candidates currently under development by other NASA projects are shown below: Low Boom Supersonic Vehicles: These vehicles have proven to be extremely difficult to trim for cruise conditions, making this an even greater challenge for maneuvers (Fig. 8). In addition, their very thin airfoils require detailed structural and aeroservoelastic models for realistic analyses, which is generally beyond the scope of traditional flight controls models.
Figure 8 : NASA Low Boom Supersonic Transport Concept .
Distributed Electric Propulsion (DEP) Vehicles: The control laws required to ensure a robust transition control in pitch, roll, and yaw, while achieving high cruise aerodynamic efficiency are depending on a new approach for control allocation. Actuation power optimization is crucial on these vehicles to satisfy their stringent power management requirements and weight savings goals. In addition, multiple distributed concentrated masses and control surfaces, combined with high structural flexibility and significant configuration changes make these concepts ideal candidates for the proposed neurocomputing process. An example of a DEP vehicle is shown in Fig. 9.
Figure 9 : Greased Lightning (GL - 10) Distr ibuted Propulsion (D EP) Vehicle flight demonstrator.
Changes to Current Concept of Operation for the Proposed Concept
to Achieve Practical Application
The proposed concept applies to the design and development stages of future air vehicles. If the research proves successful, artificial intelligence will be used to develop optimized control laws in a much more efficient manner than by using traditional methods; and these optimized control laws are an enabler for an entire fleet of revolutionary aircraft. The proposed approach reduces power requirements, hinge moments, structural loads, and therefore overall vehicle weight, therefore allowing us to exploit the full potential of innovative aircraft with multiple distributed control surfaces.
The process is easily applicable to other innovative and unconventional configurations. Without the necessity to implement artificial intelligence on the aircraft itself, no certification issues are associated with such a solution.
Conclusions
During the course of this project the researchers developed a proof-of-concept process to demonstrate the potential of using neurocomputing for optimizing actuation power for aircraft ® featuring multiple independently actuated control surfaces. A Nastran aeroservoelastic finite element model was used to generate a learning database of hinge moments and actuation power characteristics for a symmetric 2.5G pull-up maneuver using various sets of control surface deflections. An artificial neural network incorporating a genetic algorithm then used this training data to perform control allocation optimization for the investigated aircraft configuration.
The phase I project showed that optimization results for the sum of required hinge moments are ® improved by more than 12% over the best Nastran solution by using the neural network optimization process.
References
1. Cameron, D. and Princen, N., “Control Allocati on Challenges and Requirements for the Blended Wing Body,” AIAA Paper 200 0-4539, AIAA Guidance, Navigation, and Control Conference and Exhibit, August 14-17, 2000, Denver, CO.
nd 2. Haykin, S., Neural Networks - A Comprehensive Foundation, 2 Edition, Prentice Hall, July 1998.
3. Ansari, N., Hou, E., Computational Intelligence for Optimization, Kluwer Academic Publishing Group, Norwell, MA, 1997.
4. Tagliarini, G.A., Christ, J.F., and Page, E.W., “Optimization Using Neural Networks,” IEEE Transactions on Computers, Vol. 40, No. 12, Dec. 1991, pp. 1347-1358.
5. Risch, T., Cosentino, G., Regan, C.D., Kisska, M., and Princen, N., “X -48B Flight-Test Progress Overview,” AIAA -2009-0934, 47th AIAA Aerospace Sciences Meeting and Exhibit, January 5-8, 2009, Orlando, Florida.
6. Collier, F. S., Thomas, R, Nickol, C. A., Lee, Chi- Ming, Tong, M. “Environmentally Responsible Aviation - Real Solutions for Environmental Challenges Facing Aviation, 27th International Congress of the Aeronautical Sciences, Paper No. 802, Nice, France, 19-24 September, 2010.
7. Liebeck, R.H., “Design of the BWB Subsonic Transport”, AIAA -2002-0002, 40th AIAA Aerospace Sciences Meeting and Exhibit, January 14-17, 2002, Reno, NV.
8. Gatlin, G.M., Vicroy, D.D., and Carter, M.B., “Experimental Investigation of the Low -Speed Aerodynamic Characteristics of a 5.8- Percent Scale Hybrid Wing Body Configuration,” AIAA th Paper 2012-2669, 30 AIAA Applied Aerodynamics Conference, June 25-28, 2012, New Orleans, LA.
9. Pitera, D.M., DeHaan, M., Brown, D., Kawai, R.T., Hollowell, S., Camacho, P., Bruns, D., and Rawden, B.K., Blended Wing Body Concept Development with Open Rotor Engine Integration, NASA/CR – 2011-217303, The Boeing Company, Huntington Beach, California, November 2011.
10. Gern, F.H., “Finite Element Based HWB Centerbody Structura l Optimization and Weight Prediction,” AIAA -2012-1606, 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, April 23-26, 2012, Honolulu, Hawaii.
11. Gern, F.H., “Conceptual Design and Structural Analysis of an Open Rotor Hybri d Wing Body Aircraft,” AIAA Paper 2013 -1688, 54th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, April 8-11, 2013, Boston, Massachusetts.
12. MSC Nastran 2008, Quick Reference Guide, MSC Software Corporation, 2 MacArthur Place, Santa Ana, CA 92707, USA, 2008.
13. Rodden, W.P. and Johnson, E.H. (Editors), MSC.Nastran Version 68, Aeroelastic Analysis User’s Guide, MSC Software.
14. Thomas, R.H., Burley, C.L., Lopes, L.V., Bahr, C.J., Gern, F.H., and Van Zante, D.E., “System Noise Assessm ent and the Potential for a Low Noise Hybrid Wing Body Aircraft with Open Rotor Propulsion,” AIAA Paper 2014 -0258, 52nd Aerospace Sciences Meeting, SciTech 2014, January 13-17, 2014, National Harbor, Maryland.
Form Approved REPORT DOCUMENTATION PAGE OMB No. 0704-0188 The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To) 12 - 2014 01- Technical Memorandum 4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER Artificial Intelligence Based Control Power Optimization on Tailless Aircraft ARMD Seedling Fund Phase I 5b. GRANT NUMBER 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) 5d. PROJECT NUMBER .
Gern, Frank H.; Vicroy, Dan D.; Mulani, Sameer B.; Chhabra, Rupanshi; 5e. TASK NUMBER Kapania, Rakesh K.; Schetz, Joseph A.; Brown, Derrell; Princen, Norman H.
5f. WORK UNIT NUMBER 694478.02.93.02.13.56.23 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION REPORT NUMBER NASA Langley Research Center Hampton, VA 23681-2199 L-20510 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S) National Aeronautics and Space Administration NASA Washington, DC 20546-0001 11. SPONSOR/MONITOR'S REPORT NUMBER(S) NASA-TM-2014-218671 12. DISTRIBUTION/AVAILABILITY STATEMENT Unclassified - Unlimited Subject Category 08 Availability: NASA STI Program (757) 864-9658 13. SUPPLEMENTARY NOTES Final Report 14. ABSTRACT Traditional methods of control allocation optimization have shown difficulties in exploiting the full potential of controlling large arrays of control devices on innovative air vehicles. Artificial neutral networks are inspired by biological nervous systems and neurocomputing has successfully been applied to a variety of complex optimization problems. This project investigates the potential of applying neurocomputing to the control allocation optimization problem of Hybrid Wing Body (HWB) aircraft concepts to minimize control power, hinge moments, and actuator forces, while keeping system weights within acceptable limits.
The main objective of this project is to develop a proof-of-concept process suitable to demonstrate the potential of using neurocomputing for optimizing actuation power for aircraft featuring multiple independently actuated control surfaces. A Nastran® aeroservoelastic finite element model is used to generate a learning database of hinge moment and actuation power characteristics for an array of flight conditions and control surface deflections. An artificial neural network incorporating a genetic algorithm then uses this training data to perform control allocation optimization for the investigated aircraft configuration. The phase I project showed that optimization results for the sum of required hinge moments are improved by more than 12% over the best Nastran® solution by using the neural network optimization process.
15. SUBJECT TERMS Actuation power; Aeroelasticity; Artificial intelligence; Fight controls; Finite element; Neurocomputing; Structures 19a. NAME OF RESPONSIBLE PERSON 18. NUMBER 17. LIMITATION OF 16. SECURITY CLASSIFICATION OF: OF ABSTRACT STI Help Desk (email: help@sti.nasa.gov) a. REPORT c. THIS PAGE b. ABSTRACT PAGES 19b. TELEPHONE NUMBER (Include area code) 15 (443) 757-5802 U U U UU Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std. Z39.18