Simulation Model Development for Icing Effects Flight Training

NASA/TM—2003-212115 SAE–2002–01–1527

Simulation Model Development for Icing

Effects Flight Training

Billy P. Barnhart, Edward G. Dickes, and David R. Gingras Bihrle Applied Research, Inc., Jericho, New York Thomas P. Ratvasky Glenn Research Center, Cleveland, Ohio

April 2003

The NASA STI Program Office . . . in Profile Since its founding, NASA has been dedicated to • CONFERENCE PUBLICATION. Collected the advancement of aeronautics and space papers from scientific and technical science. The NASA Scientific and Technical conferences, symposia, seminars, or other Information (STI) Program Office plays a key part meetings sponsored or cosponsored by NASA.

in helping NASA maintain this important role.

The NASA STI Program Office is operated by • SPECIAL PUBLICATION. Scientific, Langley Research Center, the Lead Center for technical, or historical information from NASA’s scientific and technical information. The NASA programs, projects, and missions, often concerned with subjects having NASA STI Program Office provides access to the NASA STI Database, the largest collection of substantial public interest.

aeronautical and space science STI in the world.

The Program Office is also NASA’s institutional • TECHNICAL TRANSLATION. English- mechanism for disseminating the results of its language translations of foreign scientific and technical material pertinent to NASA’s research and development activities. These results are published by NASA in the NASA STI Report mission.

Series, which includes the following report types: Specialized services that complement the STI • TECHNICAL PUBLICATION. Reports of Program Office’s diverse offerings include creating custom thesauri, building customized completed research or a major significant phase of research that present the results of databases, organizing and publishing research NASA programs and include extensive data results . . . even providing videos.

or theoretical analysis. Includes compilations of significant scientific and technical data and For more information about the NASA STI Program Office, see the following: information deemed to be of continuing reference value. NASA’s counterpart of peer- reviewed formal professional papers but • Access the NASA STI Program Home Page has less stringent limitations on manuscript at http://www.sti.nasa.gov length and extent of graphic presentations.

• E-mail your question via the Internet to • TECHNICAL MEMORANDUM. Scientific help@sti.nasa.gov and technical findings that are preliminary or of specialized interest, e.g., quick release • Fax your question to the NASA Access reports, working papers, and bibliographies Help Desk at 301–621–0134 that contain minimal annotation. Does not contain extensive analysis. • Telephone the NASA Access Help Desk at 301–621–0390 • CONTRACTOR REPORT. Scientific and technical findings by NASA-sponsored • Write to: NASA Access Help Desk contractors and grantees.

NASA Center for AeroSpace Information 7121 Standard Drive Hanover, MD 21076

SAE–2002–01–1527

NASA/TM—2003-212115

Simulation Model Development for Icing

Effects Flight Training

Billy P. Barnhart, Edward G. Dickes, and David R. Gingras Bihrle Applied Research, Inc., Jericho, New York Thomas P. Ratvasky Glenn Research Center, Cleveland, Ohio Prepared for the General Aviation Technology Conference and Exhibition 2002 sponsored by the Society of Automotive Engineers Wichita, Kansas, April 16–18, 2002 National Aeronautics and Space Administration Glenn Research Center

April 2003

Trade names or manufacturers’ names are used in this report for identification only. This usage does not constitute an official endorsement, either expressed or implied, by the National Aeronautics and Space Administration.

Available from NASA Center for Aerospace Information National Technical Information Service 7121 Standard Drive 5285 Port Royal Road Hanover, MD 21076 Springfield, VA 22100 Available electronically at http://gltrs.grc.nasa.gov

Simulation Model Development for Icing Effects Flight Training

Billy P. Barnhart, Edward G. Dickes, and David R. Gingras Bihrle Applied Research, Inc.

Jericho, New York 11753 Thomas P. Ratvasky National Aeronautics and Space Administration Glenn Research Center Cleveland, Ohio 44135 In order to establish this icing effects flight training ABSTRACT capability, NASA Glenn Research Center teamed with Bihrle Applied Research and the Wichita State University A high-fidelity simulation model for icing effects flight in 1998 to develop a flight simulation demonstrator. The training was developed from wind tunnel data for the work also establishes a methodology for developing flight DeHavilland DHC-6 Twin Otter aircraft. First, a flight simulators that incorporate icing effects that can be used model of the un-iced airplane was developed and then by flight training organizations, operators, airframe modifications were generated to model the icing manufacturers, and pilots in safety training programs.

conditions. The models were validated against data records from the NASA Twin Otter Icing Research flight A typical application for such a flight modeling process test program with only minimal refinements being would be to add the capability to existing flight trainers. In required. The goals of this program were to demonstrate such a case, a baseline flight model would already exist.

the effectiveness of such a simulator for training pilots to This effort focuses on the steps needed to augment a recognize and recover from icing situations and to baseline model with appropriate data to model the effects establish a process for modeling icing effects to be used of icing. In the case of the Twin Otter, no appropriate for future training devices.

baseline simulation was available; therefore, as an additional step, one was developed and validated as part INTRODUCTION of the effort. It was important to have both baseline and iced simulations from a pilot training standpoint, so that a In response to a 1997 White House Initiative to reduce pilot could be exposed to the juxtaposition of both aviation accidents, NASA formed the Aviation Safety conditions. Starting from the baseline simulation also Program (AvSP) in 1999. The seven-year program has allowed researchers to assess, in an incremental been tasked to reduce aviation accident rates by 80% by fashion, the level of modification and steps required to 2007 and by 90% by 2017. Accident and incident reports implement icing effects to an existing baseline. To were analyzed to focus efforts on areas of highest return.

accomplish this, the PSIM effort was designed to These studies showed that 13% of all weather-related proceed in three phases, a baseline model development accidents were due to airframe icing.

and validation (BASELINE), a horizontal tail icing model development and validation (ICE01), and a full aircraft To address the icing hazard, NASA has developed a icing model development, assessment, and validation number of tools to supplement pilot training. To date, (ICE02).

these tools consist of educational & training videos and computer-based training CD-ROMs. However, a task Once a baseline simulation structure was established, within the System Wide Accident Prevention Project of each phase of the flight model development focused AvSP is currently underway to develop a flight simulator primarily on the development of aerodynamics models that incorporates icing effects for pilot training and control system models. The basis for each of the applications. The purpose of the Pilot Simulator Training developmental aerodynamics and control models was for Aircraft Icing Effects (PSIM) activity is to provide low speed wind tunnel data. These data were used to pilots with ground-based training facilities that provide a populate table driven models of aerodynamics forces, realistic simulation of in-flight icing encounters. This moments, and hinge moments. The models were then capability will provide pilots a pre-exposure to the validated with flight data. To complete the validation, adverse effects of icing on airplane performance, physics-driven modifications are made to the raw wind- stability and control. It will serve as a tool for initial and tunnel data and then reintroduced to the simulation. The recurrent pilot training to provide awareness of the revised model is then reevaluated against an consequences of an icing encounter and the knowledge independent data set until appropriate acceptance of how to best manage potential adverse maneuvers that criteria are met. In this effort, modifications applied to the may result from icing-induced loss of control.

NASA/TM—2003-212115 1 wind-tunnel data were noted and applied to subsequent GROUND REACTION efforts (Figure 1). For example, modifications made to baseline data were applied to ICE01 data during the The ground reaction model provides forces and ICE01 development process. Modifications made during moments to the equations of motion (EOM) that the ICE01 validation would then be applied during ICE02 represent landing gear functionality and its interaction model development etc. This approach allowed with the ground. The Twin Otter’s tricycle landing gear researchers to evaluate procedural modifications and was modeled as three spring and damper assemblies their applicability to the development of a flight model located at each wheel contact point. The model encompassing icing effects with out a priori knowledge of accounted for nose wheel steering and representative their effects on the airplane flying qualities and gear deflection. Since landing gear fidelity was not the performance. focus of this effort, this simple representation was sufficient to allow for take off and landing and appropriate The aircraft chosen for the simulation development representation of the aircraft attitude on take off.

programs was the DeHavilland DHC-6 Twin Otter (Figure 2), a twin engine, high wing configuration. The Twin Otter PROPULSION MODEL is a key asset in the Icing Research program at the NASA Glenn Research Center (Figure 3). The NASA The twin-engine propulsion system of the Twin Otter was Glenn Team has extensive operational experience with modeled using thrust coefficients that were determined this aircraft in icing conditions and with the airplane for the 3-bladed 8.5-ft propeller from the Hartzell modified with artificial ice shapes, and has supplied all propeller chart. These values were truncated within the flight data used during the simulation development effort aircraft power coefficient and advance ratio ranges and described herein (References 1 and 2). modeled in tabular form. The power coefficient was computed from independent (left and right) torque The remainder of this paper provides background pressure commands, propeller rotational speeds, and information pertaining to the structure and development density ratios. Each power coefficient was then used to of the baseline and the ICE01 and ICE02 flight models compute appropriate thrust values to be applied along as well as their validation. The paper also presents a the engine thrust lines. Using thrust line locations and number of the modifications made to the raw wind-tunnel orientations, thrust data were converted to body axis data as a result of the validation process. forces and moment and supplied to the six-DOF EOM.

AERODYNAMIC MODEL FLIGHT MODEL DEVELOPMENT PROCESS The main effort of the flight model development was to The development of a simulator involves modeling establish a successful methodology for modeling the external physical contributions that act on the air vehicle aerodynamic effects of icing conditions. The general to compute a new vehicle state and presenting approach was to use a combination of experimental data observations of those states to the pilot via visual, tactile, from wind tunnels and flight test data to derive the aural, or vestibular means. The flight model, as aerodynamic flight characteristics.

discussed herein, is the component of the simulation that implements the physics of flight and provides state The baseline aerodynamics model was developed using observations to the rest of the simulator. Typical both wind-tunnel data and flight-extracted parameters.

subcomponents of the flight model are equations of The following sections provide details pertaining to data motion, propulsion model, ground reaction, from each source.

aerodynamics model, and flight control model.

Wind Tunnel Data In the development of the baseline Twin Otter simulation these sub-components were assembled and Wind tunnel tests were conducted using a 6.5%-scale implemented in a commercial off the shelf simulation model of the Twin Otter (Figure 4). A number of tests environment, D-Six  , (Reference 3). The following were conducted in the Wichita State University’s (WSU) sections provide a discussion of each component.

7’x10’ low-speed wind tunnel in Wichita, Kansas to identify appropriate static characteristics. Additional wind EQUATIONS OF MOTION AND INTEGRATION tunnel tests were then conducted using the same model at the Bihrle Applied Research’s Large Amplitude A standard thirteen-state, quaternion-based, Newtonian Multiple Purpose (LAMP) Facility in Neuburg a.d. Donau, representation of the six-degree of freedom airplane Germany, where the dynamic damping characteristics of equations of motion formed the basis of the flight model.

the aircraft were obtained. This facility contains a Aircraft position, velocity, and rotational states were multiple purpose rotary-balance rig that is capable of integrated from accelerations computed from force and obtaining dynamic data, both rotational and forced moment contributions provided by aerodynamics, oscillation data, as well as static data, over a large range propulsion, and ground interaction models.

NASA/TM — 2003-212115 2 of angles of attack and sideslip angles. Rotary-balance rotary balance data, nonlinear variations in the forces (wind-axis damping) and forced-oscillation (body-axis and moments due to rotation rate, sideslip angle and flap damping) tests were conducted using this rig. A deflection were mechanized in the model (For example, significant amount of additional static data was also see Figure 6). The body-axis dynamic damping terms, collected. During each test forces and moments acting deduced from the forced oscillation tests, were on the wind tunnel model were measured with a six- mechanized as incremental coefficients that were a component balance. Data reduction and conversion function of the non-dimensionalized body rate, and not yielded aerodynamic coefficients in a wind tunnel axes as derivatives with respect to rate as was classically system. done in the past. This method of handling the body axis damping terms permits the proper modeling of the The WSU wind tunnel data range extended from -12 ° to nonlinearities that are commonly found in the stall and +20 ° angle of attack for 0 ° , 10 ° , and 20 ° sideslip angles post-stall angle of attack regions (Figure 7). As shown in at 0 ° flaps. Data were collected at only 0 ° sideslip angle Figure 7, these damping characteristics were also found to be a function of flap deflection.

for 20 ° and 40 ° flap settings. The Bihrle LAMP test provided an extensive set of static data at each flap An example of the longitudinal component of the deflection for further refinement of the model. These data aerodynamics model structure is shown in the build up of were used to extend the angle of attack range from -20 ° the pitching moment. The pitching moment coefficient to +40 ° and, because of the observed nonlinear modeling reflects its functional dependencies on angle of variations in the aerodynamic characteristics beyond 20 ° attack, sideslip, flap deflection, elevator deflection, of sideslip, were also used to extend the sideslip rotation rate and pitch rate. The total coefficient is functionality to ± 30 ° for all six body-axis force and produced as a sum of several terms that are each moment coefficients.

tabular data with the shown dependencies. The first table, C , represents the static pitching moment The wind tunnel data were analyzed to determine the m BASIC model structural dependencies and to insure the coefficient for the basic clean configuration as a function preservation of all nonlinear effects. The initial definition of angle of attack, sideslip angle and flap deflection. The of the basic airframe was built from the most significant incremental coefficient due to elevator deflection for the functional dependencies, which for this aircraft were clean configuration, ∆ C , is modeled as an additional m DE angle of attack, angle of sideslip and flap deflection.

table with the functional dependencies shown. The Increments were determined for control deflections, aircraft ’ s dynamic damping characteristics are dynamic damping, and power effects.

represented in the next two sets of tables. The incremental pitching moment coefficient due to rotation A merging of the two data sets was accomplished for the about the velocity vector is modeled by the table ∆ C .

m ROT evaluation and determination of control effectiveness The rotational effect due to sideslip for the longitudinal modeling. For the WSU wind tunnel data set, runs were coefficients is symmetrical, such that the increment for a conducted on the basic Twin Otter with primarily full (-26 ° positive sideslip and rotation rate is the same as that for and +14 ° ) and half elevator deflections. Rudder the same magnitude negative sideslip and rotation rate, effectiveness was measured at full deflections. Since a and has been mechanized accordingly by the functional comprehensive set of control surface deflections had dependency on the product of the rotation rate multiplied been tested at the Bihrle LAMP facility, these data (for by the sign of the sideslip angle and the absolute value the elevator and rudder, as well as for the ailerons) were of the sideslip angle. The body-axis pitch rate damping, used to define the nonlinear variation in control authority ∆ C , as mentioned before, is mechanized as an m as a function of surface deflections and combined with Q incremental table that is a function of pitch rate, as well the WSU control effectiveness data result in a as angle of attack and flap deflection.

comprehensive control effectiveness model (Figure 5).

Elevator and aileron effectiveness exhibited flap C = C ( α , β , δ f ) dependencies, and these characteristics were m m TOTAL CLEAN BASIC incorporated as well.

+ ∆ C ( α , δ e, δ f ) m DE The Bihrle LAMP rotary-balance (wind axis) and forced- oscillation (body axis) data were used to define the Twin + ∆ C ( α , Ω b/2V ∗ SGN( β ),  β  , δ f ) m ROT Otter aircraft ’ s dynamic damping characteristics for coordinated and uncoordinated motions, respectively.

+ ∆ C ( α , q c /2V, δ f ) m These data were mechanized separately and combined Q in the simulation using the techniques proposed by Similarly, lateral-directional characteristics were modeled Kalviste (Reference 4) where the implementation of the as a nonlinear multidimensional build up. An example of data is governed by the actual model test motions. For this is the rolling moment coefficient implementation.

the wind-axis damping terms, traditionally referred to as NASA/TM — 2003-212115 3 C = C ( α , β , δ f ) were under predicted with the raw wind tunnel data. To l l TOTAL CLEAN BASIC insure that the full-scale stall angle was properly identified, flight derived and “ flight manual ” data were + [ ∆ C ( α ,  δ a  , δ f )] * SGN( δ a) l DA used to determine where the full-scale airplane stall occurred.

+ [ ∆ C ( α ,  δ r  )] * SGN( δ r) l DR The total lift coefficient of the flight test aircraft was + [ ∆ C ( α , Ω b/2V ∗ SGN( β ),  β  , δ f )]* SGN( β ) l ROT measured for various flap deflections (Reference 2) and is shown in Figure 8. Maximum lift coefficient values + ∆ C ( α , pb/2V, δ f ) l P were also calculated using the stall speed placard information provided in the aircraft as a function of weight + ∆ C ( α , rb/2V, δ f ) l R and flap deflection. Using this information, adjustments were made to the wind tunnel lift coefficient data (Figure The rolling moment coefficient is mechanized as a 9). The associated static and dynamic aerodynamic function of angle of attack, sideslip, flap deflection, characteristics of all six coefficients at stall were also aileron and rudder deflections, rotation rate, as well as shifted to the revised stall angles of attack as a function both roll and yaw rates. The first table, C , represents l BASIC of flap deflection. Experience has shown that the the static rolling moment coefficient for the basic clean general trends in the dynamic data observed for the configuration as a function of angle of attack, sideslip model at its stall angle of attack should be representative angle and flap deflection, and therefore defines the basic for the aircraft in its stall region. The Reynolds number nonlinear static lateral stability of the aircraft. The effect is diminished as angle of attack increases beyond incremental effect due to aileron deflection is the region where the wing stall influences the dynamic represented in the table ∆ C as a function of angle of characteristics, and typically there is no longer any effect l DA by at most 40 degrees angle of attack.

attack, the absolute value of the total aileron deflection, and flap deflection. The output from this table is Flight Extracted Data multiplied by the sign of the total aileron deflection to account for rolling to the left or right. The table ∆ C l DR Since no aerodynamics effects due to propulsion were represents the incremental rudder effectiveness as a measured during wind tunnel testing these effects were function of angle of attack, and the absolute value of the determined from flight data. The flight test data used to rudder deflection. The output from this table is multiplied extract these effects were obtained from the NASA Iced by the sign of the actual rudder deflection to account for Aircraft Stability and Control Program conducted in 1992 yawing to the left or right. Rotational rolling moment (Reference 1), the NASA/FAA Tail plane Icing Program characteristics, representing the wind-axis roll damping, (TIP) conducted in 1995 and 1997 (Reference 2) and the are included in the table ∆ C as a function of four l ROT Pilot Simulator Training for Aircraft Icing Effects (PSIM) independent variables; angle of attack, the product of the program conducted in 2001. The maneuver set used for rotation rate and the sign of the sideslip angle, the the effort consisted of steady state trim points and absolute value of the sideslip angle, and flap deflection.

throttle transitions.

As with all lateral-directional tables mechanized with a ‘ sign of sideslip ’ functionality, the result of the table look- Equation error parameter identification was used to up must be multiplied by the sign of the sideslip angle, extract the effects of propulsion on the longitudinal due to the fact that the functional dependency on sideslip aerodynamics of the Twin Otter (Figure 10). Thrust angle is anti-symmetric with the sign of sideslip, with dependent terms that were functions of angle of attack, positive sideslip angle and a rotation in one direction flap deflection, and thrust coefficient were added to the producing a variation similar, but opposite in sign, to baseline aerodynamics model. For each of these terms, negative sideslip rotating in the opposite direction. The table values, for given independent argument body-axis damping terms for roll and yaw rate are breakpoints, were identified from a concatenated set of represented by the tables ∆ C and ∆ C and are l l data. These estimates were implemented into the new P R mechanized as functions of angle of attack, roll or yaw aerodynamics model, which was exercised with data rate, and flap deflection.

measured in flight. During each iteration, the model predicted total aerodynamic coefficients were used with Modification to Baseline Wind-Tunnel Data coefficients extracted from flight data to compute a residual vector. The root sum of squares of this residual Both the WSU and the LAMP wind tunnel data were vector was then minimized by a nonlinear least squares measured at low dynamic pressures/Reynolds numbers.

algorithm (Reference 5).

As would be expected for thick airfoils like that incorporated on the Twin Otter, it was noted that the stall The approach described above relies on several angle of attack, and, hence, the maximum lift coefficient assumptions: the propulsion model used for extraction NASA/TM — 2003-212115 4 from flight is of good fidelity and the baseline BASELINE VALIDATION aerodynamics model is of good fidelity for thrust independent cases. Figure 11 contains an example of Overdrive Equation Error Analysis the results for the aerodynamic effect of propulsion on the pitching moment of the Twin Otter. A validation tool used in this analysis from the D-Six simulation environment, ‘ Overdrive ’ , allows the validation The resulting propulsion effects on the aerodynamics of the simulation aerodynamic database against flight- were modeled as functions of angle of attack, thrust extracted data using the process illustrated in Figure 12.

coefficient and flap deflection and included as At each time slice, extraction of aerodynamic moment incremental tables for the appropriate coefficients, coefficients from the flight-recorded time history occurs primarily pitching moment, as shown on the right side of Figure 12. Angular rates are numerically differentiated to obtain the angular ∆ C ( α , CT, δ f ), acceleration of the vehicle. After the removal of the m CT inertial effects, the remainder is nondimensionalized to calculate the aerodynamic force and moment coefficients and axial force, experienced during flight. Also, at each time step, flight- recorded states, such as angle of attack, angle of ∆ C ( α , CT, δ f ).

A CT sideslip, control surface positions, etc., are used to exercise the aerodynamic model in accordance with the CONTROL SYSTEM MODEL aerodynamic model specification discussed previously.

Each aerodynamic model element (i.e., pitching moment Control Surface Gearing due to elevator, etc.) is stored and summed as prescribed in the aerodynamic model. By over-plotting Because the Twin Otter control system is reversible and the model predicted coefficients with the flight-extracted comprised of mechanical linkages, the control system total coefficients, differences can be easily identified.

model was implemented using gains to convert yoke, Correlating the discrepancies with the excitation of the wheel, and rudder positions to control surface positions.

individual elements and parameters from the flight time These control surfaces positions were then used to history aids to isolate potential weaknesses in the compute control surface effects in the aerodynamics aerodynamic model.

model.

It should be emphasized that there is no integration Elevator Hinge Moment/Stick Force during an Overdrive run. The states are completely restricted to the values from the time history. The One observation from the icing flight test program was advantage of this approach is that there is no that the difference in elevator hinge moment was one of propagation of error over time. Any differences between the strongest indicators of the reduced tail plane the model-predicted and the flight-extracted values are performance due to horizontal tail icing; with a strictly the result of local error. This ‘ analog-matching ’ progressive hinge moment lightening and reversal as the approach alleviates the problem associated with the tail plane stall margin was reduced. Consequently, an propagation of error over time; any noise or dropout of effort was made to accurately model the elevator hinge signals would not affect the analysis of the subsequent moment characteristics in the PSIM simulator.

events in the time history. Additionally, the flight control system is bypassed in this methodology thus avoiding Hinge moment coefficient data from an Ohio State any confusion between the error caused by aerodynamic University wind tunnel test of a Twin Otter tail plane modeling and the error from the flight control system airfoil section (Reference 6) were used to generate a model. Once the user is satisfied with the aerodynamic model of the forces on the elevator. These wind tunnel model, the entire simulation (including the aerodynamic data were measured for the horizontal tail section alone model and flight control system modeling) can be without the presence of the wing. For modeling validated by running the simulation in open-loop purposes, it was desired to generate tabular data as a controlled by the flight-recorded control surface or pilot function of wing angle of attack, so calculations of the input.

downwash angle corrections (Reference 7) were made for each flap deflection to establish a relation between The validation effort showed that, in general, the model tail angle of attack and wing angle of attack. Hinge correlated well with the flight records, however, some moments were then modeled as a function of angle of lateral-directional aerodynamic characteristics required attack, and elevator and flap deflections. From the scaling to improve the correlation. These included moments, stick forces were computed using a linkage reducing the directional stability and the roll due to yaw model.

NASA/TM — 2003-212115 5 rate term, and increasing the incremental roll due to The effect of ice buildup on the horizontal tail (ICE01) rudder deflection. Examples of the overdrive provided no more than a small influence on the basic comparisons of the final models for the baseline are lateral-directional aerodynamic stability characteristics, shown in figures 13 and 14, for lateral-directional and as expected. This effect is represented by the longitudinal control doublets, respectively. replacement table C and is a function of angle of l BASEI1 attack, sideslip angle, and flap deflection − the same Using the overdrive technique, comparisons of predicted independent variables as the clean configuration stick force were made against stick forces measured in flight. Figures 15 presents a sample set of plots of these CONTROL SYSTEM MODEL comparisons. The model did not have the ability to trim out the initial forces, as was done in flight, so there is a The wind tunnel tests at Ohio State University included a constant bias between the two curves, but otherwise the unique set of data collected for a horizontal tail section correlation is very good as can be seen.

with ice accretions on the leading edge. This data was utilized in exactly the same manner as was done for the ICE01 (TAIL PLANE ICE) SIMULATION baseline to generate a hinge moment model of the ICE01 DEVELOPMENT AND VALIDATION configuration. The tabular data were a function of angle of attack and elevator and flap deflections, the same as AERODYNAMICS MODEL DEVELOPMENT the baseline model, and replaced the tables for the baseline model when it was desired to simulate the iced Wind Tunnel Data condition.

As part of the wind tunnel testing mentioned above, data VALIDATION were collected with ice shapes on the horizontal tail plane. As part of the Wichita State University tests, The validation procedure discussed for the baseline equivalent, simpler, ice shapes were evaluated that were model using the Overdrive process was also done for the demonstrated to produce essentially the same ICE01 configuration. The data modifications that were aerodynamic influences as the actual ice shapes during incorporated into the baseline model, including the the sub-scale wind tunnel tests (Reference 8). These modifications to the stall angle of attack, the flight equivalent ice shapes were subsequently used during extracted propulsion aerodynamic effects, and those the remainder of the WSU tests and for the tests at the arising from the validation effort were all incorporated LAMP facility. To identify the effects of the horizontal tail into the ICE01 model, as well. As a consequence, the ice accretion on the aerodynamic characteristics of the validation showed very good agreement between the Twin Otter, these data were collected at test conditions calculated and flight measured results immediately, identical to the baseline.

which indicated that the effects of the ice accretion were well represented by the wind tunnel test data. The Ice accretions on the horizontal tail (ICE01) alter the Overdrive plots for the ICE01 cases look very similar to basic aircraft ’ s static pitching moment characteristics the ones shown for the baseline model and, (Figure 16), particularly over the negative angle-of-attack consequently, are not reproduced here.

region, as well as the control effectiveness of the elevator. To model these effects, the basic pitching ICE02 (WING AND TAIL ICE) SIMULATION moment table and the incremental pitching moment due DEVELOPMENT AND EVALUATION to elevator deflection table are replaced by two new tables, C and ∆ C , that represent the airplane m m AERODYNAMICS MODEL DEVELOPMENT BASEI1 DEI1 with ice accretions on the horizontal tail. These tables are functions of the same independent variables as the Wind Tunnel Data clean configuration.

As part of the wind tunnel testing mentioned above, data C = C ( α , β , δ f ) were also collected with equivalent ice shapes on the m m TOTAL ICE01 BASEI1 wing, vertical tail, and horizontal tail plane. As was done for the ICE01 testing, these data were collected at test + ∆ C ( α , δ e, δ f ) m DEI1 conditions identical to those for the baseline.

+ ∆ C ( α , Ω b/2V ∗ SGN( β ),  β  , δ f ) m ROT This icing condition (ICE02) produces changes to all of the pitching moment terms (except the propulsion effect + ∆ C ( α , q c /2V, δ f ) term which was initially assumed to be the same as that m Q extracted from the baseline flight records), not just the basic static pitching moment characteristics and elevator + ∆ C ( α , CT, δ f ) m CT effectiveness, as was seen for the horizontal tail ice NASA/TM — 2003-212115 6 alone. This is not surprising, since with ice accretions at CONTROL SYSTEM MODEL the leading edges of both the wing and horizontal tail, it would be expected that the pitch characteristics would all The same elevator hinge moment modeling that was be influenced. All of the replacement pitch tables are developed for the horizontal tail ice case was used for functions of the same independent variables as those for the ICE02 modeling since the ice shapes on the the clean configuration. horizontal tail were the same in both.

C = C ( α , β , δ f ) VALIDATION m m TOTAL ICE02 BASEI2 A validation effort, similar to that done for the baseline + ∆ C ( α , δ e, δ f ) m DEI2 and ICE01 configurations is currently underway.

Because there were no previous flight results for this + ∆ C ( α , Ω b/2V ∗ SGN( β ),  β  , δ f ) m ROTI2 condition, the validation for this case had to wait until new data from flights specifically flown for the PSIM effort + ∆ C ( α , q c /2V, δ f ) m were available (Reference 9). Early preliminary QI2 validation results show generally good correlation for this configuration, which is a promising indication that similar + ∆ C ( α , CT, δ f ) m CT icing simulation models can be developed without a priori flight data. A sample Overdrive plot is shown in Figure Likewise, the ICE02 icing results in replacement tables 17. Except for some steady-state biases in the for all of the terms in the rolling moment equation. Since longitudinal data, which may be due to unmodeled power there are ice accretions on both the wing and vertical tail, effects for this ice configuration, the correlation with the which together largely determine the roll characteristics, initial model is quite good.

this is not a surprising result. All of the terms are functions of the same independent variables as the clean IMPLEMENTATION OF FLIGHT MODEL configuration tables.

DESKTOP REAL-TIME SIMULATION C = C ( α , β , δ f ) l l TOTAL ICE02 BASEI2 The simulation model was constructed in the form of data + [ ∆ C ( α ,  δ a  , δ f )] * SGN( δ a)  l DAI2 tables that were easily incorporated into the D-Six real- time flight simulation software environment for the PC.

+ [ ∆ C ( α ,  δ r  )] * SGN( δ r) l This software includes all of the model independent code DRI2 needed to run the simulation, including the equations of motion, graphics generation, etc. Consequently, real- + [ ∆ C ( α , Ω b/2V ∗ SGN( β ),  β  , δ f )]* SGN( β ) l ROTI2 time flight simulations of the Twin Otter can be performed on a desktop PC with just the addition of some type of + ∆ C ( α , pb/2V, δ f ) l PI2 flight control input devices. These can be as simple as game-type joysticks (but this option would lack the very + ∆ C ( α , rb/2V, δ f ) l important longitudinal force cue) or include a full RI2 complement of force-loaded controls. Out the window Wing stall occurred at a lower angle of attack for the and Twin Otter instrument views are displayed on the ICE02 configuration because of the ice accretion on the computer screen or projection system (Figure 18). A full wing leading edge. Indications are that it would also stall complement of analytical tools are included that provide, earlier for the full-scale aircraft, as well. It was reasoned for example, plotting of any simulation parameter (Figure that the presence of the ice on the wing leading edge for 19) and simulation playback.

this configuration would likely separate the flow, thus promoting wing stall, in a similar manner for both the full- TRAINING COCKPIT LAB STATION scale airplane and for the sub-scale wind tunnel model.

As a consequence, no shift in the wing stall angle of The Twin Otter simulation flight model for icing effects attack was added to the ICE02 database as had been flight training was interfaced with Bihrle ’ s cockpit station done for the other configurations.

and control loading system for further evaluation. The cockpit station is also driven with the D-Six  Simulation With no pre-existing flight data for the ICE02 environment running on a state-of-the-art Pentium configuration from which to extract the equivalent computer. The cockpit station includes out the window propulsion related aerodynamic terms, the terms that graphics, custom multi-channel instrument panel were extracted for the baseline, un-iced model were displays, control yoke column and rudder pedals, multi- assumed initially. As part of the validation effort the lever throttle, and standard gear and flap levers (Figure validity of this assumption will be determined.

20). The control yoke was connected through an NASA/TM — 2003-212115 7 advanced Input/Output Device (IOD) interface to a 2. Ratvasky, T.P., Foss Van Zante, J. and Sim, A., Fokker loading device for simulation of longitudinal NASA/FAA Tail plane Icing Program: Flight Test control forces, while the lateral control and rudder pedals Report, NASA/TP-2000-209908, March 2000.

were spring loaded. The hinge moment model was used 3. Bell, J.W., Application of a Multipurpose Simulation to generate a model of the forces on the elevator and Design, AIAA Paper 97-3798, August 1997.

hence at the control yoke to supply the pilot with realistic 4. Kalviste, J: Use of Rotary Balance and Forced control forces, including dynamic force characteristics. Oscillation Data in a Six Degrees of Freedom For the iced conditions, the control forces can become Simulation, AIAA Paper 82-1364, August 1982 very large: a full tail stall was experienced during flight 5. Klein, V, "Identification Evaluation Methods," AGARD test with horizontal tail ice at full flap deflection and Lecture Series No. 104.

required 100-172 lbs of pull force for recovery. 6. Gregorek, G.M., Dreese, J.J., and La Noe, K., Additional Testing of the DHC-6 Twin Otter Iced The NASA test pilots that fly the Twin Otter were invited Airfoil Section at the Ohio State University 7 ’ x10 ’ to perform checkout maneuvers with each of the icing Low Speed Wind Tunnel, NASA/CR-2000- conditions and have commented favorably on the training 209921/Vol2, 2000.

cockpit lab station. A more thorough evaluation is 7. Perkins, C.D., and Hage, R.E., Airplane planned that will determine the minimum level of pilot Performance, Stability and Control , John Wiley and cues that are necessary for the icing training task. Sons, 1967.

8. Papadakis, M., Gile-Laflin, B.E., Youssef, G.M., Ratvasky, T.P., Aerodynamic Scaling Experiments CONCLUSION with Simulated Ice Accretions, AIAA 2001-0833, January 2001.

Once a baseline simulation flight model was established 9. Ratvasky, T.P., Blankenship, K., Rieke, W. and and validated, researchers successfully used wind-tunnel Brinker, D.J.: Iced Aircraft Flight Data for Flight data to identify the effects of ice accretion on the Simulator Validation, SAE 2002-01-1528, NASA/TM- aerodynamic characteristics of the vehicle and control 2003-212114.

feel. With these effects identified, specific portions of the flight model were targeted for augmentation to represent the new configuration. In the first case, where tail plane ice occurred, correlation between predicted aerodynamic CONTACT coefficients and those extracted from flight was good. On occasion, biases were evident in comparisons and Bihrle Applied Research, Inc. (www.bihrle.com) indicated deficiencies in the modeling of secondary effects of the propulsion system for example, or any Billy P. Barnhart - Vice President, barbilly@optonline.net thrust effects due to rapid angle of attack or sideslip Edward G. Dickes – Senior Engineer, changes. In addition, a model of the effect of icing on edickes@bihrle.com hinge-moments was developed from wind-tunnel data David R. Gingras – Senior Engineer, and evaluated. When implemented with the simulation dgingras@bihrle.com and control loading hardware, the effects proved to be representative of the airplane. NASA Glenn Research Center The process of using wind-tunnel data to identify the Thomas P Ratvasky – Icing Branch, effects of icing appears to have worked well in the case Thomas.P.Ratvasky@grc.nasa.gov of the ICE02 configuration. In this case, wind tunnel data comparisons were used to identify longitudinal and DEFINITIONS, ACRONYMS, ABBREVIATIONS lateral directional differences, and were used to modify the simulation. At the time of writing this paper, work on Symbols the ICE02 effort is still under way, but preliminary evaluations indicate that collecting wind-tunnel data to The units for physical quantities used herein are assess the effects of icing on the aerodynamics of a presented in U.S. Customary Units. All aerodynamic data vehicle configuration without a priori knowledge is a are referenced to the body axis system.

success.

b: wing span, ft REFERENCES c : mean aerodynamic chord, ft C : axial-force coefficient, Axial force/ q S A 1. Ratvasky, T.P., Ranaudo, R.J., Icing Effects on C : normal-force coefficient, Normal force/ q S N Aircraft Stability and Control Determined from Flight C : side-force coefficient, Side force/ q S Y Data. Preliminary Results, NASA TM 105977, C : rolling-moment coefficient, Rolling moment/ q Sb AIAA-93-0398, January 1993.

l C : pitching-moment coefficient, Pitching moment/ q S c m NASA/TM — 2003-212115 8 C : yawing-moment coefficient, Yawing moment/ q Sb α : angle of attack, deg n β : angle of sideslip, deg C : axial-force coefficient due to pitch rate A Q ∆ : prefix for incremental component C : normal-force coefficient due to pitch rate N Q Ω : rate of rotation about the velocity vector, rad/sec C : side-force coefficient due to roll rate Y Ω b/2V: non-dimensional rotation rate, positive for P clockwise spin C : side-force coefficient due to yaw rate Y R δ a: aileron deflection, positive when right trailing-edge is C : rolling-moment coefficient due to roll rate l P down and left trailing-edge is up, ( δ a right - δ a left)/2, deg C : rolling-moment coefficient due to yaw rate l δ f: flap deflection, positive when trailing-edge is down, R C : pitching-moment coefficient due to pitch rate deg m Q δ h: horizontal tail deflection, positive when trailing-edge C : yawing-moment coefficient due to roll rate n P is down, deg C : yawing-moment coefficient due to yaw rate n δ r: rudder deflection, positive when trailing-edge is left, R p, q, r: body-axis roll, pitch, and yaw rates, rad/sec deg q : free-stream dynamic pressure, lb/ft Function S: wing area, ft SGN( ): sign function, returns 1 if the argument is V: free-stream velocity, ft/sec positive or zero, -1 if it is negative Flight Test Data Baseline Wind-Tunnel Coefficient Baseline Flight Tests Extraction Coefficients Validated Baseline Baseline Mods (Lat-Dir/Power) Ice01 Flight Coefficients ICE01 Updated (Tail Ice) Ice01 Validated Ice01 Ice02 ICE02 Predicted Flight (Full Ice) ICE02 Coefficients Figure 1. – Schematic of Development Process.

Figure 2. – DeHavilland DHC-6 Twin Otter Airplane.

Figure 3. – NASA Glenn Research Center Icing Research Aircraft.

Figure 4. – 6.5%-Scale Twin Otter wind tunnel model.

NASA/TM — 2003-212115 9 Figure 5. – Incremental pitching moment coefficient as a function of angle of attack, elevator and flap deflections.

Figure 6. – Rotational rolling moment coefficient.

Figure 7. – Influence of Flap deflection on roll damping.

NASA/TM — 2003-212115 10 Baseline C vs L a/c α α α α a/c 3.5 3.0 2.5 dF=0 dF=10 2.0 dF=20 dF=30 1.5 dF=40 1.0 0.5 0.0 -10 -5 0 5 10 15 Figure 8. – Flight Test Total Lift Coefficient vs. Angle of Figure 9. – Comparison of wind tunnel data and final Attack for Various Flap Deflections.(Reference: simulation model Normal Force curve.

NASA/TP-2000-209908) Flight Data Aerodynamics Trim, Power Model Transitions Compute Aerodynamic Predicted Result Coefficients Estimated Parameters Compute True Result Residuals Nonlinear Least Squares Minimization Figure 10. – Parameter Identification logic diagram.

Figure 11. – Effect of propulsion (C ) on pitching moment coefficient.

T NASA/TM — 2003-212115 11 , , Speed, Altitude, α β Linear Angular Rates (P, Q, R) & Accelerations Angular Rates Control Deflections (Nx, Ny, Nz) (P, Q, R) Flight Test Data 2nd-Order Numerical Aerodynamic Differentiation Model Angular Acceleration Summation & Store Results Mass Properties Remove Inertial Contribution & Other External Other External Forces/ Forces/Moments Moments (Thrust/Thrust Vectoring) Flight-extracted Sim-predicted C , C , C & C , C , C & Nf A Y Nf A Y C , C , C @ c.g. C , C , C @ c.g.

l m n l m n Overplot Comparison Identify Differences & Refine Database Figure 12. – OverDrive block diagram.

NASA/TM — 2003-212115 12 Figure 13. – Comparison of flight and simulation Overdrive results for final lateral-directional model.

Figure 14. – Comparison of flight and simulation Overdrive results for final longitudinal model.

Figure 15. – Comparison of flight and simulation Figure 16. – Effect of ice accretion on static Pitching Overdrive for elevator hinge moment and stick forces - Moment Coefficient, flaps = 0.

elevator doublet; baseline model (no trim in sim).

NASA/TM — 2003-212115 13 Figure 17. – Overdrive run with ICE02 showing degree of correlation of the initial model, prior to validation, with flight results.

NASA/TM — 2003-212115 14 Figure 19. – An example of plotted variables available from the simulation.

Figure 18. – Sample Desktop Sim Screens showing one aircraft external view and instrument panel.

Figure 20. – BAR ’ s Cockpit Station.

NASA/TM — 2003-212115 15 Form Approved REPORT DOCUMENTATION PAGE OMB No. 0704-0188 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 Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.

2. REPORT DATE 3. REPORT TYPE AND DATES COVERED 1. AGENCY USE ONLY ( Leave blank) Technical Memorandum April 2003 4. TITLE AND SUBTITLE 5. FUNDING NUMBERS Simulation Model Development for Icing Effects Flight Training WBS–22–728–20–01 6. AUTHOR(S) Billy P. Barnhart, Edward G. Dickes, David R. Gingras, and Thomas P. Ratvasky 8. PERFORMING ORGANIZATION 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) REPORT NUMBER National Aeronautics and Space Administration John H. Glenn Research Center at Lewis Field E–13767 Cleveland, Ohio 44135 – 3191 10. SPONSORING/MONITORING 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) AGENCY REPORT NUMBER National Aeronautics and Space Administration NASA TM—2003-212115 Washington, DC 20546– 0001 SAE–2002–01–1527 11. SUPPLEMENTARY NOTES Prepared for the General Aviation Technology Conference and Exhibition 2002 sponsored by the Society of Automotive Engineers, Wichita, Kansas, April 16–18, 2002. Billy P. Barnhart, Edward G. Dickes, and David R. Gingras, Bihrle Applied Research Inc., Jericho, New York 11753; Thomas P. Ratvasky, NASA Glenn Research Center. Responsible person, Thomas P. Ratvasky, organization code 5840, 216–433–3905.

12b. DISTRIBUTION CODE 12a. DISTRIBUTION/AVAILABILITY STATEMENT Unclassified - Unlimited Subject Category: 05 Distribution: Nonstandard Available electronically at http://gltrs.grc.nasa.gov This publication is available from the NASA Center for AeroSpace Information, 301–621–0390.

13. ABSTRACT (Maximum 200 words) A high-fidelity simulation model for icing effects flight training was developed from wind tunnel data for the DeHavilland DHC-6 Twin Otter aircraft. First, a flight model of the un-iced airplane was developed and then modifications were generated to model the icing conditions. The models were validated against data records from the NASA Twin Otter Icing Research flight test program with only minimal refinements being required. The goals of this program were to demon- strate the effectiveness of such a simulator for training pilots to recognize and recover from icing situations and to establish a process for modeling icing effects to be used for future training devices.

14. SUBJECT TERMS 15. NUMBER OF PAGES Aircraft icing; Flight simulation; Flight simulators; Flight safety 16. PRICE CODE 19. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 17. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT OF ABSTRACT OF THIS PAGE OF REPORT Unclassified Unclassified Unclassified NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89) Prescribed by ANSI Std. Z39-18 298-102