Statement of Mike Poole
Transportation Safety Board of Canada

MR. POOLE: Thank you, Dennis. I'm very pleased and honored to have been invited to speak at this symposium and to present jointly with the NTSB.

Certainly the level of cooperation between our two organizations, particularly in the recorder field, which I'm intimate with, has been exemplary and, I think, a real benefit to advancing transportation safety.

I'd also like to point out there's a number of TSB Canada people here at the symposium, including representatives from our Rail and Marine modes as well as our Director General of Investigations, Mr. Bill Tucker, our Executive Director, Mr. Ken Johnson, and our Chairman, Mr. Benoit Bouchard. So, if you have any questions about the TSB, there's lots of us here at your disposal.

As Dennis mentioned, on September 2nd, 1998, SwissAir MD-11 crashed off Peggy's Cove after reporting smoke in the cockpit and declaring an emergency, and although we're still investigating the occurrence, we thought it might be a useful example to demonstrate the multiple sources of recorded data that have been available in the investigation.

The next few pictures are just meant to give you a little bit of an appreciation of what we have to deal with in an accident like SwissAir. This is fairly representative of the wreckage that has been recovered from the ocean floor, and as you can see, most of the pieces aren't very big.

We have several hundred crates of wreckage that have been examined to date, and we have about 88 percent of the airplane recovered by weight. This is the hangar floor in Halifax just a few weeks ago. You can maybe see in the center of the picture there's a little black thing sticking up. That's the center pedestal from the cockpit, and basically we've laid the wreckage out in organized categories to facilitate the analysis.

The tables in the very front of the picture are some of the aircraft's duct work, and it's been fairly painstakingly lock-wired back together piece-by-piece.

This is what's known as the bird cage. It's a 1:1 frame of the nose section of the aircraft that we had built to mock up and reconstruct the nose area, and this is just looking inside the back of the bird cage, and you can see even though we have 88 percent, impressive number by weight, we don't really have that much of the nose area, and, of course, these are the smallest pieces and the hardest to recover, and we're very interested in this area because so far, the investigation has determined heat damage fore and aft of the cockpit door, all in the ceiling area, consistent with a fire in that area, and that's the area in the top of the picture you see.

So, the sources of recorded data that were used in SwissAir include two radar sites, the air traffic control recording, the flight data recorder, the cockpit voice recorder, an optional quick-access recorder that was installed for airline purposes, digital datalink messages that were sent to and from the aircraft, and non-volatile memory from the aircraft's computer that controls the engines.

So, I'll just quickly run through each of these and how they apply to SwissAir. The first is the radar data. It was available from two different centers, and that's useful because radar accuracy is partially dependent upon whether or not you're traveling to and from the station or traversing perpendicular to it.

The radar data was e-mailed to the Safety Board about midnight, two hours after the accident, and we plotted this track, which I was able to take to Peggy's Cove when our go team dispatched at 7 a.m. that same morning, and what we did is just super-impose the radar image over VHF navigational map for easy reference, and this was very useful early in the investigation when generally nobody has very many answers, and a lot of people are asking a lot of questions.

We used the radar information to have specially-equipped remote-sensing aircraft. They had infrared and synthetic aperture radar, and these aircraft back flew the track within days of the occurrence, looking for any pieces that might have fallen off the aircraft and also any evidence of fuel dumping, and the track was also very helpful for us to search out where potential witnesses might have best been able to see the aircraft.

The air traffic control recording, despite aviation's image of advance technology, the ATC recordings are still largely cumbersome, old-fashioned tape recorders, but they do the job, and obviously they provide useful information with respect to aircraft ground communications between the pilots and the controllers.

The aircraft's flight data recorder was recovered in about 150 feet of water, four days after the accident, was located with the aid of its under-water locator beacon or the ULB, which is shown in the inset in the bottom left, and you can see that the ULB was almost detached from the recorder. It was kind of just hanging on by a thread, and this is actually something that's been of concern among the international flight recording community for some time now. Will the beacons stay on in high-impact water accidents?

The box itself didn't see very much damage, jus a few dents in the outer casing, and we did keep the recorder in water and disassembled it in water to prevent any corrosion, and this is just the solid state memory module that was removed from that recorder. Its memory chips for this particular model are encapsulated in foam, and all we had to do was cut that little ribbon connector and plug it into one of our bench surrogate units and recover the information.

We had about 25 people on the recorders group, including the Swiss accident investigators, the NTSB, SwissAir, McDonnell-Douglas, which is now Boeing, ALPA, others, and all these people participated in the review and analysis of the data.

We had 250+ parameters, and they were examined interactively on the computer screens within a few weeks, and this was somewhat unprecedented because we normally have a pretty good idea of which subset of parameters we're most interested in when we start. In this case, we didn't. So, we had to look at every one, and it would, I think, have taken orders of magnitude longer if we'd use the traditional method of printing the data out in tabular or graphical format and then passing these pieces of paper around to everybody.

So, what we did is we had small groups of three to four people. Each of them took a computer station at our lab, and then they produced what was relevant to demonstrate what they found interesting, and I think this is something that will quickly become apparent to the other modes as they start to experience the volumes of data that we have on modern airplanes.

You really have to have some pretty good interactive analysis tools to basically wade through it.

Obviously the data gave the investigation immediate focus and yielded very useful clues as to very specific areas on the aircraft where the fire may have originated.

The cockpit voice recorder was found a few days after the flight recorder was found. It was also kept in water right through the disassembly, and also the beacon on the left side there is precarious. It hung on but barely, and the voice recorder also didn't really receive much impact damage, just a small dent in its outer casing as well.

The internal tape recording, the top right is with the cover on, the bottom left is with the cover off, and you can see the tape's in very good condition. This is a 30-minute endless loop-type recording.

About 1 in the morning, we had the tape out of the voice recorder, but despite many rinses in fresh water, it still smelt fairly heavily of jet fuel. So, rather than adjourn for the night, we decided to go through the night and hand-clean it, and, of course, the voice recording contains all the internal conversations by the crew as well as the acoustic environment on the flight deck and was obviously very useful in understanding what was happening on the flight deck and developing the overall sequence of events as well as it yielded vital clues with respect to the health of the aircraft.

This is a pile of tape that our systems group chairman in Halifax sent us. It was recovered from various pieces of wreckage, both underwater and floating. First, we thought it was a cruel joke that he sent us this, but he was serious. Most of this tape is not very useful, but some of it is from the aircraft's quick-access recorder, which actually can have some useful information since it actually records more sources than the aircraft's flight data recorder.

So, what do we do with the mess of tape like that? No problem. We decided to have a little fun with our Halifax team, and we sent this picture back and told them we'd have it analyzed in just a few minutes. Of course, this is not the real tape. So, don't anybody panic.

What we did do is go through the pile, and we identified all of the quick-access pieces. There are 29 in total, ranging from a few inches to over a hundred inches in length, and we are currently experimenting on ways to recover the data from these segments, and we hope to have some information off them literally in a few weeks.

Datalink, digital messages. Many aircraft transmit both automatic and crew-initiated digital messages to their operations center or to and from their air traffic control facility for operational clearances. This is an example of a typical datalink display unit.

In the SwissAir case, there was a system known as ACARS, which is Communication Addressing and Reporting System, which automatically sent messages to the ground, which were very useful for us in determining the health of the aircraft at various times during the flight. These messages are recorded in what's called service providers before the air passed on to the end destination, the airline or the air traffic control facility, and it's important to get not only the messages as received at the destination facility but also the intermediate messages at the service provider because they, the service providers, strip out systems-related information that the end recipient doesn't care about but the investigation team does.

Non-volatile memory is an increasing source of valuable information. More and more computer-controlled systems have more or less permanent memory, and the data stored can augment the investigation considerably. The SwissAir MD-11 had what's known as full authority digital electronic control computers or FADEC, which stored maintenance information, such as pressure altitudes, speed, engine settings and engine fault codes, and many of you probably know that both recorders stopped while the aircraft was at 10,000 feet as the fire progressed. They did not capture the last six minutes of flight, and the FADEC from the Number 2 engine gave us data in those last six minutes of the flight where the recorders had already stopped. So, in this case, the non-volatile memory was extremely useful.

So, what do we do with all of the data? Well, among other things, one technique that we use is to generate a computer reconstruction of all of the data integrated together, and this is a snapshot of the flight reconstruction that has been developed for the SwissAir investigation.

We have mocked up the primary flight display from the MD-11, which we find helps the investigation team assimilate the information in as familiar a format as the crew would have been presented that information on the flight deck.

We do play these in real time with the voices from both the voice recorder and the air traffic control recording synchronized, and it's kind of like any good movie, any time you watch it again, you might see something you didn't see before, and these reconstructions are also, I think, very useful in developing a common intimacy among the investigation team, and in particular, they tend to bring out the time element of when things happened, which is often difficult to appreciate from a printed page.

This is just another example of the ways that data and in this case the SwissAir can be presented. It's purposeful that you can't read the information on the slide because much of the information hasn't been released yet. It's more to give you the idea of how you can put it together. It's color-coded with respect to the various sources of information, be they the flight recorder, the voice recorder, the air traffic control and so on, and it really serves as a pretty handy reference chart when questions arise during the investigation. It covers -- the top left starts where the crew noticed a smell and covers the last 20 minutes of recording and contains all of the events that we consider significant at the moment.

The future direction. Just a few points in closing. We are headed towards combined recorders, as Dennis mentioned. That's two identical recorders recording both voice and data for redundancy, one installed in the tail and one installed in the nose, and I believe there's some of those on display. These are not new. They have been around for awhile.

Two-hour CVRS up from the historical 30 minutes. In the SwissAir accident, the CVR did not capture the beginning of the flight where we suspect the fire may be initiating. We have a number of questions that will largely remain unanswered because the CVR was insufficient in duration to provide coverage of events like fires where they tend to take a long time to manifest themselves.

The 10-minute independent power supplies for the voice recording, so that we would have continued to capture the acoustic environment on the flight deck for those last six minutes of flight. Recording the datalink messages not only on the ground but also on board the aircraft, so they can be correlated to flight deck activities. The elimination of tape recorders since they no longer meet demonstrated high-impact and fire accidents and have numerous maintenance problems associated with them. Increased parameter lists.

My experience is whenever investigators get in a room for more than five minutes, the parameter list goes up. You just have to live with it. Crash-protected video is currently a topic at the International Flight Recorder Working Group, and we're trying to wrap our hands around the fundamental needs for video. It's not such an easy task, and a very important trend, which is emerging in North America, it's been around in Europe for awhile, which is FOQA, the Flight Operations Quality Assurance, which is simply the airlines' trending flight data routinely to prevent accidents rather than only read the data out after an accident when it's arguably too late, and many people feel that the exploitation of this type of data will allow the industry to address some of the human factors-type accidents that have traditionally been fairly elusive in reducing, and there are also operational benefits for the airlines. So, it's a real win-win situation.

One thing I've learned is never to predict the future because what looks impossible today is often feasible tomorrow. This is from my favorite comedian, the Far Side. It's a print of a theme park where cavemen are visiting future world. The caveman on the right is looking at an advertisement for the potential invention of matches, and his comment is, "Yeah. Right."

On behalf of Dennis and I, thank you very much.