GV – May 1998

Animation is No Accident at NTSB
By Sheldon Liebman

When TWA Flight 800 exploded off Long Island, the question on everyone's mind was "What went wrong?" In order to find out the answer to that question, many different agencies became involved and lots of processes were set in place. In this article, we'll look at one of those processes - simulating the flight at the National Transportation Safety Board.

The creation of simulations doesn't "solve" an accident. Rather, it provides a way for people investigating a tragedy to visualize the events leading up to it. Due to the presence of the Flight Data Recorder, or "Black Box," in an airplane, these events can be portrayed with an extremely high degree of accuracy up to the moment the FDR stops recording. From that point forward, information gathered from the crash scene can be used to help determine what took place.

Dan Bower is an Aerospace Engineer at the NTSB who is directly involved with analyzing data gathered from Flight Data Recorders. Unfortunately, there are many different formats this information can take. The FDR data, says Bower "is different for each airplane and each style of Flight Data Recorder." Because of that, he adds, "as soon as we have an accident, we call the airplane manufacturer to get the conversion algorithms."

The Flight Data Recorder itself is hand carried from an accident site to the NTSB headquarters in Washington, DC, where the information is downloaded to Hewlett Packard workstations. Using these high-speed machines, engineers at the NTSB then convert the raw digital data into useful information. At least 17 parameters are recorded approximately once per second, but newer airplanes and FDRs can provide information on as many as 200 parameters.

The parameters that are recorded include "airspeed, altitude, pitch, roll, vertical acceleration and lateral acceleration," explains Bower. For fast-changing parameters like vertical acceleration, as many as eight samples may be taken in a second. Since this is a serial process, the point within a second that each parameter is measured is different. This means you can't just create "sets" of parameters and build animation frames around them. Each individual parameter must be charted separately and placed into a continuous timeline.

The first goal, says Bower, is to "create time history files." In order to do that, he tries to link the FDR data, which is recorded as elapsed time, with radar data, which is recorded as time of day. "The common piece of information we have is altitude and time," he adds, "and we try to marry those up." Once the time history files are created, they can be exported to a personal computer for further analysis as well as to a Silicon Graphics Onyx workstation where most of the final animations are created.

On Bower's PC, which runs the Windows NT Operating System, he uses multiple software packages to manipulate and animate the FDR data in a variety of ways. "We have some flight performance software that allows simulations with three, four and six degrees of freedom," explains Bower. Within that software and other packages, Bower can animate the data as charts and graphs, using an airplane model or as cockpit instruments changing over time. While these animations don't offer the level of detail or realism that comes from the SGI, they are used to illustrate factual data in the best way possible.

The software that Bower uses can also output data in the format used for the SGI-based animations. Once the data is moved to that machine, one goal is to make the animation look as realistic as possible. For this process, detailed models are used, backgrounds are carefully constructed and terrain models are incorporated into the scene. A lot of time is also devoted to determining which viewpoint or viewpoints to use for viewing the animation.

Since there is no FDR information to use for the final portion of the simulation, other data must be examined. According to Dr. Vern Ellingstad, the NTSB's Director, Office of Research and Engineering, this process was particularly complex and time consuming in the case of the TWA 800 flight.

For the initial analysis, "radar data from five different sources" was used, explains Ellingstad. "Some of the pieces (after the initial explosion) were big enough to follow on radar. However, the flight path that's established by radar is a lot less precise. Essentially, there's a hit every 4.5 seconds rather than every second." Using that radar data, aerospace engineers performed a trajectory analysis using the aerodynamic properties of the pieces. "We tried it a few ways," adds Ellingstad. "One way assumed the engines continued to run, another didn't. Essentially what they were doing was driving those aerodynamic studies to fit the radar data."

In some cases, results from the salvage effort provided final location information for pieces of the plane. In other cases, initial trajectory analyses helped steer the salvage crews to the areas they should search. As long as that process continued, the trajectory data was updated to reflect as accurate a path as possible.

"The data we develop from those analyses are the vital part of the animation," explains Ellingstad. Once it exists, the final animation is used to accurately visualize the event. At the NTSB, they pair their Silicon Graphics Onyx with Vis Lab animation software from Engineering Animation, Inc. in Ames, Iowa. EAI specializes in 3D visualization technology and has devoted a lot of effort to ensuring that Vis Lab is flexible enough to utilize many different types of data.

For the TWA 800 animation, a high quality model of the airplane was brought into Vis Lab and texture mapping was used to add identifying information to the model. A very accurate terrain map of Long Island was obtained from another government agency and that model was also brought into Vis Lab. Time of day and weather conditions helped round out the environment that was used to simulate the flight.

Once the animation data was imported into Vis Lab, it was possible to visualize how the accident occurred. The next step was to determine the best camera positions to use. In this case, the animation was created as a kind of "picture in picture" display, with a long range view showing the whole area and a close-up of the airplane itself positioned in the upper right hand corner. Text information was also incorporated into the animation to illustrate some key parameters.

Vis Lab is also capable of adding special effects to animations that enhance their realism, but this isn't the goal at the NTSB. "We could do a Steven Spielberg kind of thing," comments Ellingstad, "but we try to keep it as close to the empirical data as we can." This doesn't mean all the special effects tools are ignored, however. In the TWA animation, smoke, fire and vapor trails are used to help understand how the events developed.

The last consideration is how much to show of an event. Usually, explains Ellingstad, "we bring an animation up to the event and then it ends. In this particular case, it was important for scientific reasons to go beyond the point of the airplane exploding. We needed to account for the parts getting down to the water."

As we go to press, a QuickTime movie of the accident recreation is available on the NTSB Web Site. If you'd like to view it, point your browser to http://www.ntsb.gov//events/twa800/gallery.htm. At this site, you'll be able to view not only this animation, but also another computer simulation and a number of graphics related to the accident. The amount of work that went into this project is incredible and the results are amazing. At the National Transportation Safety Board, that's no accident.

Back to Recent L&S Marketing Articles