Aircraft Dynamics Models for Use with FlightGear Modeling and Simulation

Maintained by Michael S. Selig, Rob Deters, and Glen Dimock UIUC Applied Aerodynamics Group Dept of Aerospace Engineering University of Illinois at Urbana-Champaign Urbana, Illinois 61801 Last update 3/10/02 Changes

Welcome to the Aircraft Dynamics Models web page. The purpose of this page is to provide a compilation of aircraft dynamics models that are sufficiently complete for use in aircraft flight simulators. The data file format is specific to the free open-source FlightGear flight simulator (FGFS), which includes the UIUC aerodynamics-model code initially developed for the Aircraft that can be currently flown with the UIUC aero model of FlightGear include the following

Beech 99 (last update 3/6/02)

Boeing 747 (last update 3/6/02)

Cessna 172 (four versions, last update 3/6/02)

Cessna 310 (last update 3/6/02)

Cessna 620 (last update 3/6/02)

Cessna T-37 (last update 3/6/02)

Convair 880 (last update 3/6/02)

F-104 Starfighter (last update 3/6/02)

F-4 Phantom (last update 3/6/02)

Learjet 24 (last update 3/6/02)

Marchetti S-211 (last update 3/6/02)

Pioneer UAV (last update 3/6/02)

Piper Cherokee (last update 3/6/02)

Schweizer SGS 1-36 (last update 3/6/02)

Twin Otter (last update 3/6/02)

1903 Wright Flyer (last update 3/6/02)

X-15 (last update 3/6/02)

It should be emphasized that these models are approximations to their namesake, hence they are by no means complete. For instance, many of these aircraft have automatic flight control systems that are not modeled. Moreover, in many cases reasonable parameter estimates must be made because some data is unavailable and/or impractical to obtain. Finally, if all the desired data were available, computers are not yet fast enough to model all of the physics. Users desiring that level of sohpistcation will do better buying the real McCoy. However, having said that, in several cases the aerodynamics are modeled in considerable detail and include table lookups to model a number of nonlinear aerodynamic behaviors (e.g. stall and loss of control) that depend on angle of attack, side slip, and control inputs among other parameters. In the data files, some references are given to the sources used in generating the data. Following the aircraft table, a number of notes are given regarding the aircraft model data and the specific use of these files in FlightGear. This data is released under the GNU GPL.

Aircraft (20 models)

Aircraft	Browse Models / Last update	FGFS ver	Comments	File Authors	Primary Sources
Beech 99	beech99-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
Boeing 747	boeing747-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
Cessna 172	cessna172-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
	cessna172-71-v2 (3/6/02)	0.7.8+	nonlinear model / includes stall curve Cessna 172 as used in fgfs 0.7.1	Scott / Selig	Roskam, Tony Peden (FGFS developer)
	cessna172-73-v3 (3/6/02)	0.7.8+	nonlinear model / includes stall curve Cessna 172 as used in fgfs 0.7.3	Scott / Selig	Roskam, Tony Peden (FGFS developer)
	cessna172-aae319-v4 (3/6/02)	0.7.8+	nonlinear model / uses CL, CD, and Cm curves Cessna 172 used UIUC AAE 319 course	Scott / Selig	Roskam, Ken R. Sivier (UIUC)
Cessna 310	cessna310-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
http://1000aircraftphotos.com/PRPhotos/Cessna620.htm	cessna620-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
Cessna T-37	t37-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
Convair 880	convair880-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Selig	Roskam
F-104	f104starfighter-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
F-4 Phantom	f4phantom-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Selig	Roskam
Learjet 24	learjet24-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Selig	Roskam
Marchetti S-211	marchetti-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Selig	Roskam
Piper Cherokee	piperCherokee-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Roskam
Pioneer UAV	pioneerUAV-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Bray
	pioneerUAV-nl-v2 (3/6/02)	0.7.8+	nonlinear model / includes 10 lookup tables	Scott / Selig	Bray
Schweizer 1-36	sgs-1-36-sprite-nl-v1/ (3/6/02)	0.7.8+	nonlinear model / includes CL, CD and CM curves	Selig	Sim
Twin Otter	twinOtter-nl-v1 (3/6/02)	0.7.8+	nonlinear model / includes CL and CD curves	Scott / Selig	Bragg, et al.
1903 Wright Flyer	wrightFlyer1903-nl-v1 (3/9/02)	0.7.8+	nonlinear model / includes CL, CD and Cm curves	Selig	Culick and Jex, Jex, et al. Selig (airfoil data)
X-15	x15-v1 (3/6/02)	0.7.8+	linear model / constant stability derivatives	Scott / Selig	Jon Berndt (early JSBSim model)

Notes

• FAQ Nothing here yet.
• My FGFS command line parameters are here. See more notes below.
• Three flight models in FlightGear.    To make a long story short, the UIUC aircraft models above work only with the LaRCSim flight dynamics model in FlightGear. Our group has written an aero module and a specific file syntax that works with LaRCSim. The initial motivation for the development of the UIUC aero model was to model aircraft in icing conditions together with a Smart Icing System to make flying in those conditions safer (Ref: UIUC Smart Icing System Project). In this mode, FlightGear runs as one client in a network-based simulation together with other displays and supporting codes. Although the UIUC aero was motivated by the aircraft icing problem, it can be used for a wide range of aircraft such as those listed above, in addition to the Twin Otter which is the UIUC benchmark icing aircraft. There are two other flight dynamics models in FlightGear, namely JSBSim, which is default, and YASim. The UIUC code does not include the more general features of JSBSim, which will appeal to a wider audience, including gamers. Details of JSBSim can be found here: http://jsbsim.sourceforge.net and "JSBSim Technical Overview" (http://jsbsim.sourceforge.net/JSBSimCUJ.PDF). For those inclined, we would be happy to see the UIUC models converted into the JSBSim- and YASim-model format where possible. For YASim, however, conversion is probably not feasible because its approach is aero-surface component based and driven by geometry, not stability derivatives and nonlinear aero lookup tables.
• To run these aircraft together with the 3D models shown, the run command line must include additional info that specifies [1] the use of the UIUC aero model (--aircraft option), [2] the path to the aircraft.dat file (--aircraft-dir option), [3] sometimes the x-offset-m, y-offset-m, and pitch-offset-deg options. Also, the aircraft.dat file must be edited to include the full path to the nonlinear aero data on the local system. Run commands for FGFS 0.7.8 are included on the downloads page. Things have changed a little for FGFS 0.7.9, and we don't have the specific information for that yet.
• Publications and Documentation
  • Smart Icing Systems Project using the UIUC aero model inside FlightGear: Bipin Sehgal, Rob Deters, and Michael Selig, "Icing Encounter Flight Simulator," AIAA Paper 2002-0817, AIAA Aerospace Sciences Meeting, Reno, NV, Jan 2002.
  • UIUC Aero Model Users Guide for FlightGear Flightsim. Latest version for FGFS 0.7.8.
• Linear and nonlinear models (and things that can go wrong)
  • "Linear" aerodynamics models are ones with constant stability derivatives (single values given). For these models, aerodynamic forces and moments are calculated by multiplying the angle or rate by the stability derivative to arrive at an aerodynamic coefficient. High rotation rates and/or angles can lead to large aerodynamic coefficients which together with the dynamic pressure and area or area*length scale can generate very large forces and moments. High forces and moments like this, should they occur, can send the model off at unrealistic speeds, generating higher rates and angles and feeding on itself until ... being software, something goes seriously wrong and FlightGear crashes. Such large changes can occur when the aircraft lands (rapidly). The spring forces jolt the aircraft to produce these large rates and angles. One way to avoid these problems should they occur is to not takeoff or land, and if that is not practical then decrease the gear spring constants to reduce the rates. In so doing, it's helpful to keep the damping ratios constant as well, which leads to lowering the damping coefficients proportionately. The landing gear parameters in the files above work quite reasonably. (Note: Takeoff can be subverted by typing ctrl-U in FlightGear. Each ctrl-U lurches the aircraft up 1000 ft at a time.)
  • Having said all that, "nonlinear" models for which the aerodynamic parameters are given as functions of angle of attack, etc are much more desirable. High angles of attack and high rates do not lead to unrealistically high aerodynamic coefficients. Instead with a nonlinear model, stall occurs for high angles of attack in which case the forces are bounded. With a "linear" model that, say, uses CL_alfa=constant, a high angle of attack leads to a high lift coefficient. For instance, these aircraft (those using a constant CL_alfa) can be landed at unrealistically low speeds because the wing never stalls. All of the models above that state "linear model / constant stability derivatives" are of this type. These types of models are the easiest ones to build and are usually good starting points for "nonlinear" higher-fidelity models.
  • Clearly, the best models (above) are the "nonlinear" ones. At a minimum these will include a lift curve that models stall (see the associated files: *CLfa.dat). Even these models, however, must use some constant stability derivatives because finding sufficiently complete nonlinear data as would be obtained from flight tests and/or wind tunnel test is rare. Hence no model is fully "nonlinear", but a little goes a long way.
  • On a related note, nonlinear aerodynamic performance can be obtained from flight tests that provide constant stability derivatives. Typically, with flight tests, the data is presented in terms of stability derivatives because that is what the mathematics prove. These stability derivatives can change depending on the flight condition (e.g. cruise vs loiter and subsonic speed vs supersonic), and hence the data can be nonlinear while still being presented in stability derivative form. In such cases, tables of stability derivatives are provided along with the flight condition for which they were obtained (e.g. see examples in Roskam. For the linear models in the table, the flight condition is listed in the aircraft.dat data file if it was provided in the original reference. An improved model is one that uses the force and moment coefficients as functions of angle of attack, sideslip, control deflections, flight conditions, etc. This type of data can only practically be obtained from wind tunnel tests. The Pioneer UAV model and 1903 Wright Flyer are two such examples.
• Other approximations
  • Most UIUC models currently used the simplest engine thrust model: "engine SimpleSingle". In these cases, the engine thrust is proportional to the joystick throttle control. It is not realistic, but it gets the models flying. The C172 models using Tony Peden's data include more realistic engine performance.
  • For cases when the control deflection limits were not known, estimates were made. These values on the "controlSurface" lines can be adjusted for what might feel like proper handling.
  • Most aircraft are modeled using cruise condition data only. For these aircraft, takeoff and landing configurations and associated changes in aerodynamic characteristics are not currently modeled.
• Some files use "record" lines to produce flight data recorder output. These data are written to the file uiuc_record.dat. Be aware that when flying for an extended period of time this file (if written by using record lines) can become very large.
• These models are not compatible with Microsoft Flight Simulator. Details of their approach are sketchy, but some reverse engineering has been somewhat successful in deciphering their file format. For more info on the MSFS format see the Abacus Flight Dynamics Editor (was http://www.flightsimdownloads.com/premier/fde/).
• An earlier version of this data was prepared by Jeff Scott. He now maintains the website www.aerospaceweb.org
• Flight Characteristics (adapted from Jeff Scott's early info):
  • Beech 99 (v1): Very smooth characteristics.
  • Boeing 747 (v1): Cruise condition is sluggish, but flies well. Aileron response is limited. Increasing the aileron effectiveness from Cl_da from -0.013 to -0.053 (landing condition) improves handling.
  • Cessna 172: All versions fly nicely. Nonlinear ones are preferable.
  • Cessna 310 (v1): Elevator seems too effective.
  • Cessna 620 (v1): Sluggish?
  • Convair 880 (v1): Flies pretty well.
  • F-104 (v1): Difficult to fly. It's called "the Widow maker" for a reason.
  • F-4 (v1): Very difficult to fly. Use nose up trim once in flight.
  • Learjet 24 (v1): Flies very well.
  • Marchetti S-211 (v1): Flies very smoothly, probably one the easiest aircraft to fly.
  • Piper Cherokee (v1): Flies well.
  • Pioneer UAV: All versions fly nicely. Flies coordinated turns without rudder corrections.
  • Schweizer SGS 1-36 (v1): Flies like a metal Schweizer. Lots of elevator and rudder authority, the latter of which is needed to counter the strong adverse yaw.
  • T-37 (v1): Flies very well.
  • DeHavilland Canada DHC-6 Twin Otter (v1): Clean version flies well.
  • Twin Otter "All Ice": To come. NASA Glenn Twin Otter for icing research. This model includes hard points on the wing tips and tail (see additional gear data). Note the degradation in performance after ice accretion begins (icing begins 2 minutes into flight; transition from clean to iced aerodynamics lasts 5 seconds--this is not very realistic, but you can vary these times and the icing severity factor, eta [0=no ice, 1=max ice], to see the impact on performance). Others coming "Tail Ice" and "Wing Ice".
  • 1903 Wright Flyer (v1): Wonder what it was like to fly the 1903 Wright Flyer? Be a test pilot. See the longitudinal static instability. Fly steady, straight and level using some trim to help relieve pressure on the stick, then put in a brief canard pulse (elevator stick) so as not to change the trim speed. With an up canard pulse and hands off the stick, the Wright Flyer will start to pitch upwards until it stalls. Down pulse, and it goes down. This same instability is seen by simply trying to fly straight and level! The rates are low enough that a good pilot with practice can stay ahead of the static instability. Stall can get pretty nasty if both the canard and biplane wing stall at the same time. Typically, however, only the canard will stall first. Full up elevator can be used to fly out of the dive. But in this condition the canard is stalled. Relieving the back pressure will put the canard out of stall and the pull-out will be a little bit faster, but the difference is hard to notice. Keep in mind this aircraft was flown four times before it was damaged on Dec 17, 1903. The first flight was 120 feet in 12 seconds, and the last was 852 feet in 59 seconds. How far can you fly? Be sure to run this with FGFS "enable-auto-coordination" option to model the coupling between the wing warping and rudder to counteract the adverse yaw effect of the wing warping. Later designs, however, the Wright Brothers decoupled the warping and rudder to have more control and the ability to fly in coordinated flight ("cross-controls"). For more info see www.wrightflyer.org. Especially useful in the modeling were two papers by Fred Culick and Henry Jex, and a third involving wind tunnel testing of a full-scale replica. More publications.
  • X-15 (v1): Go easy on the throttle since this is a very over-powered aircraft. Has very slow elevator response, but extremely responsive ailerons. Use elevator trim once in flight. This is a linear version, so extremely unrealistic angles of attack can be used takeoff. Nonlinear models for this and all the others are in the works. Keep in mind this aircraft was designed to be dropped from a B-52 at 45,000 ft and Mach 0.8. Like many of the high speed aircraft, the X-15 had a stability augmentation system consisting of angular rate feedback loops in about all three axes. These are not currently implemented in any of the models.
• The difficult aircraft to fly (F-104, F-4, X-15, 1903 Wright Flyer) are we think most easily flown from the external view + HUD with the 3D models to visualize. In this view the horizon stays level while the aircraft pitches, rolls, etc. Unfortunately, the direction of the view is somehow keyed to the body axis so that when the aircraft yaws the horizon swings back and forth. For these aircraft, a better view would be one that stays behind the aircraft and looks along the flight path but still with a flat horizon. Maybe an additional view option like this is in the works.
• UIUC Aeromodel Credits
  Rob Deters, Glen Dimock, Ann Peedikayil and Michael Selig - Currently active on aero model and UIUC Smart Icing System components
  Jeff Scott, Michael Savchenko, Jay Thomas, Eunice Lee, Elizabeth Rendon, and Sudhi Uppuluri - Past contributors
• Photo Credits
  NASA Dryden Photo Gallery - Schweizer 1-36, Learjet 24
  NASA Ames - 1903 Wright Flyer (replica)
  UAV NAVIAR - Pioneer UAV
  All others are 3D models from Wolfram Kuss: FGFS Airplane Hangar - was http://home.t-online.de/home/Wolfram.Kuss
  

Flight Dynamics Links to more Links

• Smart Icing System Project (was http://www.ae.uiuc.edu/sis and is on the wayback machine)
• FlightGear
• JSBSim
• Aircraft Dynamics (MAE 331), Princeton University, Robert Stengel - includes several good flight dynamics links