Computers Speed Air-Crash Probes

BEFORE sunrise on March 28, Flight 571 pierced the thick mist that shrouded its final approach to Edmonton (Alberta) International Airport. Seconds later, the Air Canada DC-8 cargo plane rolled onto a runway visible for less than an eighth of its length, scraped the right outboard engine, skidded off the tarmac, veered back on the runway, and ground to a stop on a nearby taxiway. ``At first, the captain couldn't comprehend that he left the runway,'' says flight investigator Harry Boyko of his first conversation with the crew at Edmonton International. ``Within weeks of the incident, he became comfortable with how we described the last 200 feet of his descent.''

As the investigator in charge for the Canadian Aviation Safety Board, Mr. Boyko and his three colleagues learned much about the incident by measuring the runway marks and interviewing the crew in the long hours that followed.

Yet the fundamental question - how the incident happened, and why - remained a mystery for a team of computer scientists to solve.

Before Boyko filed his preliminary report, Mike Poole, superintendent of computer systems engineering for the Canadian Aviation Safety Board, watched the descending DC-8 from a computer laboratory in Ottawa. Mr. Poole and his team of scientists relived the sights and sounds of the incident from a high-speed computer work station.

The ability to reconstruct flights visually, using digitally recorded data, enhances and hastens the work of investigators. Boyko says this is the most significant technological breakthrough of his 20 years as a pilot and accident sleuth.

In cases where total destruction of the aircraft occurs, he says, ``the investigation would just be a guessing game'' without this equipment.

Flight 571 is one of many such incidents members of the safety board will investigate this year. The International Civil Aviation Organization requires them to analyze occurrences taking place in Canadian airspace and over international waters when a Canadian registered plane is involved or the majority of passengers are Canadian citizens. During 1988, the safety board reviewed more than 500 aviation mishaps, including the Air India crash off the coast of Ireland, the Arrow Air disaster at Gander, Newfoundland, and the Alitalia crash near Milan, Italy.

The safety board has gained a worldwide reputation for developing space-age tools to determine the causes of airline crashes. The work of Poole and his team of scientists answered questions that puzzled investigators of the DC-8 incident. In a sense, their work provided the most detailed eyewitness account of it.

An engineering work station, which runs computer-aided-animation programs, and a scientific computer are the safety board's new investigative tools. These computers, together with several electronic recording and measuring devices, received their flight test when the DC-8 skidded off runway 01 in late March.

At Edmonton, safety board investigators removed the disabled aircraft's two flight recorders - the cockpit voice recorder (CVR) and the flight data recorder (FDR). CVRs pick up exchanges between the ground and flight crews, as well as most sounds heard by the captain and co-pilot. FDRs track the altitude, pitch, roll, and engine performance of the aircraft, for example. Both voice and flight data are recorded on quarter-inch, magnetic tape.

For Poole, the laboratory analysis of the DC-8 incident began when investigators brought the brightly painted flight recorders back to the computer-systems-engineering laboratory from Edmonton.

The Sundstrand flight data recorder takes periodic readings from 62 flight and cockpit instruments and records them on magnetic tape. ``An instrument tape recorder retrieves those magnetic signals from the FDR for processing with the electronic instruments and computers,'' Poole explains.

Other labs use separate playback hardware to decode each brand of flight data recorder. The Canadian Aviation Safety Board uses a single ``data acquisition box,'' which samples signals at up to 100,000 times per second, and some high-powered software to convert flight information to data suitable for enhancement by computer.

Software developed by the safety board and Prior Data Sciences of Ottawa allowed Poole and his team of analysts to pull digital wave forms - expressed as 1's and 0's - directly into a Hewlett-Packard scientific computer.

This means that Canadian Aviation Safety Board analysts can edit data that comes from flight recorders on-line. ``A sudden maneuver or hard landing can distort a signal, causing loss of data,'' says Poole. ``Typically, those few seconds of data can be crucial to the investigation.''

Careful analysis allowed the safety board to restore the distorted data for Flight 571 accurately and quickly. ``The safety board can recover poor-quality data using software,'' says Poole. ``The industry [the airlines and other safety labs] is forced, out of tradition, to painstakingly analyze distorted data using oscilloscopes and oscillographs.''

Once Poole and the scientists culled the information from the DC-8 flight recorder, they began to interpret decoded flight data. Software, which includes mathematical routines, plotted the altitude, pitch, roll, engine performance, and airspeed of the DC-8 over time. This software hastened the process of verifying flight data by replacing calculators, drafting curves, and graph paper used elsewhere in the industry with a high-speed computer.

``With so many variables changing over time, it was difficult to get a sense of what happened to Flight 571 from the graphs alone,'' says Poole. ``To get a firsthand look at the crash, we transferred the decoded and refined flight data to a second Hewlett-Packard computer, which drives the graphics work station. Scientists at the lab, safety board investigators, and the crew of the DC-8 could then watch an animated reconstruction of the landing.''

At first, the crew's recollection of the landing didn't confirm what the flight data recorder showed. ``Once they watched the reconstruction,'' says investigator-in-charge Boyko, ``they were able to understand what went wrong with the approach.''

A full-color, high-resolution monitor recreated runway 01 at Edmonton during the waning moments of Flight 571. It simulated the approach, landing, and instrument-panel readings of the DC-8. An instrument tape recorder reproduced the sound.

``We can sit in the cockpit or follow the aircraft in a chase plane,'' says Poole, explaining the reconstruction.

``If we have witnesses who said they were standing on the ground at a specific location the moment the aircraft flew over, we can direct the work station to put them at that location and have the flight pass overhead just as it did for the eyewitnesses.''

Had a chase plane tailed Flight 571 on its approach to Edmonton International, a pilot with an infrared scope would have described the scene that Poole, safety board investigators, and the crew of the DC-8 have watched over and over again: On final approach, the DC-8 heads for the center line on a glide slope. Fifty feet above the ground, the left wing drops. The pilot rights the aircraft. It drifts left. Moments before touchdown the left wing drops again causing the No. 4 engine to strike the tarmac. The aircraft skids off - then bounces back on the runway and comes to rest on a nearby taxiway.

``We could not have identified the cause of the accident without the reconstruction,'' says Boyko.

``The last few seconds of an approach are second nature for most pilots. The pilot didn't have a clear recollection of what he had done,'' the former aviator explains. ``The reconstruction refreshed his memory.

``The speed and accuracy with which the board now recovers flight data allowed Air Canada to alert its crews quickly about hazardous conditions that may exist for other aircraft,'' Boyko says.

Two weeks after the incident, Air Canada issued a bulletin for its flight crews to beware of illusions that can occur in low visibility weather.

``We were very satisfied with the reconstruction,'' says Joe Galliker, senior flight safety officer for Air Canada. ``It confirmed what our graphics people showed by hand.''

The safety board determined what happened to Flight 571 days after the incident. Labs with less-sophisticated electronic equipment could have taken weeks to determine what happened, if at all.

``Speed in determining the cause of an accident is essential for Air Canada as well as the safety board,'' says Mr. Galliker. ``We would prefer to see a flight reconstruction by the next day.''

``You can spend hours analyzing graphs of flight data and still lack an appreciation of what happened,'' says Don Langdon, chief of systems engineering at the Canadian Aviation Safety Board. ``By modeling the aircraft and showing its gyrations on a computer screen, you can get a clear and common understanding of the accident in minutes. Knowing and agreeing on what happened the team can then work on finding out why.''

The final report on Flight 571 is due this fall.