PC-Based Simulation of the F-16/MATV

Matthew J. O’Rourke*, John N. Ralston†, Jeffery W. Bell‡

Bihrle Applied Research, Inc.
Hampton, Virginia 23666

Stanley F. Lash§
Wright Laboratory
Flight Dynamics Directorate, Flight Control Division
WPAFB, OH 45433

Summary developed, D-SIX, a six-degree-of-freedom simulation

environment for the PC. One distinguishing feature of
The Wright Laboratory (WL) F-16 simulation this software is the unique capability to rapidly re-host
model has been re-hosted on a personal computer aerodynamic databases, flight control system (FCS)
running a unique six-degree-of-freedom simulation code and/or other complex computer code which model
environment. This model contains over 1000 two- subsystems such as weapons, sensors and mission
dimensional aerodynamic data tables and a complex equipment. Due to the complex nature of the computer
flight control system (FCS) modeled by thousands of programs which model these systems, it is important
lines of FORTRAN computer code. The FCS model that as much code as possible be reused when a
can emulate several different configurations including a simulation model is re-hosted in another operating
standard Block 40 F-16 and the F-16 Variable In-Flight environment. This is done not only to minimize the
Stability Aircraft (VISTA) with Multi-Axis Thrust level of effort required, but also to ensure that the
Vectoring (MATV) in numerous modes. In addition to accuracy of the model is preserved.
re-hosting the WL F-16 model, the high-angle-of-attack
database was extended to include nonlinear effects at One such simulation model which exists is the
moderate to high angles of attack. These additional USAF Wright Laboratory (WL) F-16/MATV
data greatly improve the ability of the simulation model simulation model1, operated by the Flight Dynamics
to predict the in-flight maneuvering characteristics of Directorate, which includes an accurate representation
the F-16 with active MATV. of both the Lockheed Martin F-16 aerodynamic
database and FCS. The FCS can be configured to
Introduction model both the standard Block 40 F-16C and the
Variable-Stability In-Flight Simulator Test Aircraft
Bihrle Applied Research (BAR) has been (VISTA) with Multi-Axis Thrust Vectoring (MATV),
conducting simulation studies of stalled, departed and shown in Figure 1. Each configuration modeled by the
spinning flight of fighter aircraft for over twenty years. FCS, in most cases, employs the identical code used in
To facilitate the modeling and simulation of the the Lockheed Martin simulation and has been validated
complex dynamic behavior encountered during these to match throughout the entire flight envelope.
and other extreme flight conditions, BAR has

*
Engineer
†
Engineering Manager
‡
Engineer
§
Principal Scientist, Control Dynamics Branch

1
 In a recent study performed by BAR, a new attack. In addition, no dependency of control
high-angle-of-attack database was developed to more effectiveness or basic airframe pitching moment on
accurately model the complex aerodynamic sideslip is modeled. Basic airframe lateral-directional
characteristics displayed by the F-16/VISTA while characteristics of the WL F-16/MATV simulation
flight testing the MATV nozzle.2 A detailed database are modeled with a linear dependence on
description of this database is presented in Reference sideslip. Reference 2 indicates that a highly nonlinear
2. To build upon this effort, BAR has been tasked to lateral-directional database accounts for some of the
re-host the previously mentioned F-16/MATV unexpected lateral-directional flight dynamics
simulation operated by the Flight Dynamics Directorate encountered at high angles of attack during MATV
at WL using the D-SIX simulation environment, while flight testing.
simultaneously updating the high-angle-of-attack
database with that described in Reference 2. This paper Although the effort described in Reference 2
will address the differences between the WL F- produced a high-angle-of-attack database, its structure
16/VISTA high-angle-of-attack database and the high- remained significantly different from that of the WL F-
angle-of-attack database assembled by BAR. In 16/MATV data structure and included no high-speed
addition, experience and lessons learned from re- aerodynamic effects. Therefore, the challenge existed
hosting the FCS will be discussed. Finally, the re- to incorporate the BAR-developed high-angle-of-attack
hosted simulation will be validated against the WL database with the existing low-angle-of-attack and
simulation as well as flight test. high-speed database of the WL F-16/MATV simulation
without compromising the integrity of either. In
Brief D-SIX Simulation Environment Description addition, to avoid discontinuities as the simulation
transitions from the low-angle-of-attack to the high-
D-SIX is a MS Windows-based 6-degree-of-freedom angle-of-attack database, it is was necessary to include
simulation environment designed to address issues the entire low-speed database in one data set, modeling
associated with many of the current complex the Mach and altitude effects as increments taken from
engineering simulators. Coding requirements are the WL database to that unified data set. However, this
minimized with the extensive use of graphical interface required that the low-angle-of-attack data, included in
for all simulation functions such as multidimensional the BAR database and used for the low-speed portion
table-look-up, flexible plotting capabilities, and data of the unified data set, match the low-angle-of attack
editing. Real-time operation coupled with a low-cost, and low Mach data of the WL F-16/MATV database.
accessible control interface and visual output permits This was done to ensure that the Mach and altitude
timely pilot-in-the-loop evaluations of FCS and increments yield values identical to the WL F-
aerodynamic characteristics. Existing modules from 16/MATV database when added to the unified low-
other simulation codes can be incorporated through speed data set.
Dynamic Link Libraries to avoid extensive re-coding
and re-validation. Additionally, many tools associated The changes to the database were made in two
with flight test data are incorporated to allow dynamic, basic classifications: those to the basic airframe data
real-time trajectory replay and flight parameter set and those to the control increments. As stated
extraction for model validation and refinement. The previously, the basic airframe data were taken from the
project-based setup permits simple exchange of low-speed dataset in Reference 2 and the WL data were
simulation models between users, and any capability incremented from that (Mach=0.2) data to create basic
upgrades in the D-SIX simulation environment is airframe increments to all six force and moment
automatically propagated to all existing models. . coefficients as a function of Mach and altitude. The
control increments were treated differently, however,
Aerodynamic Database Structure and Upgrade due to the very different model structures of the
nonlinear control data and the WL control data. To
Typical of most current high-fidelity flight compensate for this, the WL control data were
simulations, the WL F-16/MATV model provides an converted to multipliers on the low speed data as a
accurate representation of controlled, maneuvering function of Mach and altitude. This allows the full
flight below maximum lift. However, like most highly- nonlinear control data to be used with the control
developed fighter simulations, extensive portions of the power augmentations / reductions with Mach and
database have been linearized. In particular, some altitude properly modeled. Great care was taken to
control deflections such as aileron and rudder have ensure that the multipliers are realistically represented
been linearized with deflection, which, for the F-16, are when low-speed data values are small.
known to be nonlinear at moderate-to-high angles of

2
 Flight Control System Re-Host Although this phase of database validation was
completed for the high-angle-of-attack database alone
Both the WL F-16/MATV simulation and the in Reference 2, the database validation time histories
D-SIX simulation environment contain sections that presented in the this paper use the unified data set
perform identical calculations such as equation of described previously. Two examples of the comparison
motion integration, stick inputs, coordinate of the total flight-extracted rolling moment coefficient
transformation, sensor modeling, etc. However, the to that predicted by the updated database are presented
FCS is a function of many of the quantities calculated in Figures 5 and 6 using a flight of the F-16
in the these sections of code external to it. Therefore, VISTA/MATV executing a 35° α roll . Figure 5 shows
the external variables which are input to the FCS code the comparison using the mostly-linear data set in the
in the WL F-16/MATV simulation were identified and current WL model, and Figure 6 shows the comparison
mapped to the appropriate D-SIX variables so that the using the fully non-linear model. Not only does the
FCS could fit seamlessly into the D-SIX environment. comparison show that the nonlinear data set matches
Using code development environments such as much better, but the following section will show how
Microsoft Visual C++, the FCS can be compiled in its these differences are amplified when executing actual
native language (in this case FORTRAN) and linked maneuvers with the six-degree-of-freedom simulation
directly to D-SIX in the form of a dynamic link library. models. For a more detailed description of the process
of database validation with flight test, please refer to
Correlation With Simulation Check Cases Reference 4

In order to ensure that the data and FCS have Comparison of Simulation Models
been implemented properly, validation runs have been
made in various stages. The first were made in Initially, a simple comparison of the current
comparison to check cases generated using the WL F- WL model and the fully-nonlinear model was done
16/MATV simulation. This was done using the using a 360° roll at the angle-of-attack limiter with the
original database to ensure that the FCS was F-16 in MATV mode, but with the limiter on. Figure 7
transported to D-SIX accurately. The validation runs shows the time histories for the current WL and
were performed simply by driving the re-hosted nonlinear models. Both models are trimmed at
simulation model with the trimmed conditions from the identical states and are driven with the same stick
WL simulation output and the control stick time inputs. The sideslip excursion of the nonlinear model
histories. Figure 2 shows a maximum roll input for the is slightly higher and the maximum roll rate is about 10
Block 40 F-16. Figure 3 shows a roll doublet at 35° α deg/sec lower than that of the current WL model.
with MATV on. Comparison of both time histories However, the two time histories represent very similar
shows good agreement, especially considering that responses (unfortunately, at the time this study was
small differences in the high angle of attack database completed, not flight test data were available for this
exist between the current WL model and the one used maneuver). It now remains to compare the two models
to generate the check cases. in the high angle of attack region where the flight
dynamic characteristics are most nonlinear, i.e., 30 to
Correlation with Flight Test Data 40 degrees angle of attack.

Once it had been established that the WL F- Since it is most important that the simulation
16/MATV simulation was accurately transported into model predicts the actual in-flight dynamic response,
D-SIX, the BAR high-angle-of-attack database was both the current WL model and the fully nonlinear
implemented as discussed previously. To ensure that model were trimmed to the flight test condition and
this database was more representative of the high- driven using the flight test stick time histories. In
angle-of-attack flight dynamic characteristics, the addition, care was taken to ensure that all FCS
Overdrive feature of D-SIX was used to extract the functions were set according to the flight log. All
force and moment coefficients from flight test data weight and inertia information were calculated based
gathered during flight testing of the MATV. These data on fuel state.
were compared to predictions obtained by running the
flight test state parameters and control deflections with Although many test maneuvers were
the new, fully non-linear database. A more detailed simulated, only one is presented here. A 35° α wind-
description of the Overdrive feature is presented in axis roll maneuver (stick-driven) was selected because,
Reference 2. Figure 4 provides a flow chart which like most aircraft, the F-16 exhibits the most nonlinear
summarizes the Overdrive function and process.

3
flight-dynamic characteristics near the stalled region.
The flight is near trim at 35° α and descending slightly Conclusions
at 25,400 ft. The true airspeed is approximately 260
fps and the power is in military setting (this is the same The F-16 model currently used at Wright
flight used to generate the data in Figures 5 and 6). It Laboratory has proven to be a highly effective tool for
should be noted that a significantly large amount of predicting flight response of the F-16 at low-to-
right stick was input by the pilot at the beginning of the moderate angles of attack. However, due to portions of
maneuver to counter an apparent roll asymmetry which the aerodynamic model that are represented by
is included in the nonlinear database. linearized data, enhancement of the database near stall
and at high angles of attack and sideslip was required to
Figure 8 shows the response of the current WL increase simulation fidelity. This has been
F-16/MATV model to stick inputs identical to those of successfully demonstrated, but this study indicates that
the flight data shown in the same Figure. The further investigation of dynamic effects, especially near
correlation is poor, showing lack of angle-of-attack stall, should be undertaken to further increase model
hold (aircraft simulation pitches to over 60° α after 8 fidelity.
seconds) and a much lower wind-axis roll rate is
exhibited by the simulation model. The inability to In addition to showing how high angle of
hold angle of attack is particularly detrimental in this attack modeling of the current Wright Laboratory F-
case, due to an FCS mode that was set which inhibits 16/MATV model can be enhanced, this study has
rolling maneuvers at higher angles of attack. It should shown that today’s personal computers provide the
computational power necessary to accurately re-host
be noted that a 35° α wind-axis roll maneuver can be
the most complex simulation models of fighter aircraft
executed using the current WL model without pitching being assembled. Re-hosting, evaluating and updating
to higher α, however, the required stick inputs would complex simulation models using PC-based tools such
be significantly different. as those presented in this paper will significantly
increase the efficiency of many aspects of aircraft
Figure 9 shows the same comparison using the development, from preliminary design to flight test, by
fully non-linear model. The angle-of-attack hold transferring real-time, high-fidelity simulation from the
compares well to that of flight test, as does the wind- expensive and shared-resource mainframes and
axis roll rate. However, the sideslip, although showing workstations to the inexpensive and readily-available
similar characteristics at the beginning of the maneuver, personal computer.
becomes far too oscillatory. Since check cases
evaluated in Reference 2 and the data presented herein References
had indicated that the static data are representative,
modeling deficiencies in the dynamic data were 1. McKeehen, P., “ F-16/VISTA High-Angle-of-
suspected, specifically in the body-axis damping terms. Attack Aerodynamic Modeling and Analysis,” WL-
This anticipated result was based on the revised low TM-91-304, February 1991.
speed database’s rigorous application of test data. The
original linearized F-16 forced oscillation test data 2. Kay, J., Ralston, J., Lash S., “Development of a
exhibited a wide range of non-repetitive values in the Non-Linear, Low-Speed Aerodynamic Database for
post stall region, and some judgment was required to the F-16/VISTA,” AIAA-97-3576, August 1997.
define the baseline configuration from these data.
Based on these and other considerations, the confidence 3. Baxter, J. et. al., “Flight Test Demonstration Report
level in this portion of the model was minimal. To for the F-16 Multi-Axis Thrust Vectoring Program,”
illustrate the sensitivity of the post stall motion to these WL-TR-95-3086, January 1995.
damping terms, the maneuver was repeated changing
only the dynamic derivative which models body-axis 4. Ralston, J., Kay J., “The Utilization of High Fidelity
rolling moment due to yaw rate, or Clr . The stability of Simulation in the Support of High-Angle of Attack
this derivative were increased between 20° to 45° α, Flight Testing,” AIAA-96-3354, July 1996.
but left unchanged elsewhere. Figure 10 shows the 35°
α wind axis roll with this change. Notice that this
simple change positively affected the correlation of all
three parameters presented, distinguishing the strong
dependence of high-angle attack maneuver modeling
on dynamic data.

4
 Figure 1.- Variable-Stability In-Flight Simulator Test Aircraft with Multi-Axis Thrust Vectoring.

Figure 2.- Comparison of re-hosted model to WL model - F-16 Block 40 maximum roll command check case.

5
Figure 3.- Comparison of re-hosted model to WL model - F-16 with MATV on roll doublet check case.

α, β, Speed, Altitude, Linear

Angular Rates (P, Q, R) & Angular Rates Accelerations
Control Deflections (P, Q, R) (Nx, Ny, Nz)
Flight Test Data

2nd-Order
Aerodynamic Numerical
Model Differentiation

Angular
Acceleration
Summation &
Store Results Mass Properties Remove
Inertial Contribution &
Other External Other External Forces/
Forces/Moments Moments
(Thrust/Thrust Vectoring)

Sim-predicted Flight-extracted
CNf, CA, CY & CNf, CA, CY &
Cl, Cm, Cn @ c.g. Overplot Cl, Cm, Cn @ c.g.
Comparison

Identify Differences
&
Refine Database

Figure 4.- Flow chart representation of the Overdrive feature of D-SIX.

6
Figure 5.- Comparison of flight-extracted to simulation- Figure 6.- Comparison of flight-extracted to simulation-
predicted rolling moment coefficient using current predicted rolling moment coefficient using fully-
WL database. nonlinear database.

Figure 7.- Comparison of current WL model and fully-nonlinear models executing a limiter roll.

7
Figure 8.- Comparison flight test and simulation Figure 9.- Comparison flight test and simulation
predicted 35° α rolling maneuver using predicted 35° α rolling maneuver using nonlinear
current WL database. database.

Figure 10.- Comparison flight test and simulation predicted 35° α rolling maneuver using nonlinear database with
increased roll due to yaw near stall.