Job, Macarthur. Air Disaster, Vol. 3. Fyshwick, Australia: Aerospace Publications, 1998. ISBN 1-875671-34-X.
In the early 1970s I worked for a company that sold remote batch computing services on UNIVAC mainframes. Our management visited Boeing headquarters in Seattle to pitch for some of their business (unlikely, as Boeing had their own computer service bureau at the time, but you never know unless you try). Part of the presentation focused on how reliable our service was, averaging better than 99.5% uptime. The Boeing data processing manager didn't seem too impressed with this. He asked, “When you came up here from San Francisco, did you fly on one of our airplanes?” “As a matter of fact, we did.”, answered the president of our company. The Boeing guy then asked, “Well, how would you feel if I told you Boeing airplanes only crash about once every two hundred flights?” The meeting moved on to other topics; we never did get any business from Boeing.

Engineering is an art we learn from failure, and the aviation safety community is the gold standard when it comes to getting to the probable cause of a complicated disaster and defining achievable steps to prevent it from recurring. There is much for practitioners of other branches of engineering to admire and learn from looking over the shoulders of their colleagues in air accident investigation, and Macarthur Job's superb Air Disaster series, of which this is the third volume (Vol. 1, Vol. 2) provides precisely such a viewpoint. Starting from the official accident reports, author Job and illustrator Matthew Tesch recreate the circumstances which led to each accident and the sometimes tortuous process through which investigators established what actually happened. The presentation is not remotely sensationalistic, yet much more readable than the dry prose of most official accident reports. If detail is required, Job and Tesch do not shrink from providing it; four pages of text and a detailed full page diagram on page 45 of this volume explain far more about the latching mechanism of the 747 cargo door than many people might think there is to know, but since you can't otherwise understand how the door of a United 747 outbound from Honolulu could have separated in flight, it's all there.

Reading the three volumes, which cover the jet age from the de Havilland Comet through the mid 1990s, provides an interesting view of the way in which assiduous investigation of anomalies and incremental fixes have made an inherently risky activity so safe that some these days seem more concerned with fingernail clippers than engine failure or mid-air collisions. Many of the accidents in the first two volumes were due to the machine breaking in some way or another, and one by one, they have basically been fixed to the extent that in this volume, the only hardware related accident is the 747 cargo door failure (in which nine passengers died, but 345 passengers and crew survived). The other dozen are problems due to the weather, human factors, and what computer folks call “user interface”—literally so in several cases of mode confusion and mismanagement of the increasingly automated flight decks of the latest generation of airliners. Anybody designing interfaces in which the user is expected to have a correct mental model of the operation of a complex, partially opaque system will find many lessons here, some learnt at tragic cost in an environment where the stakes are high and the margin of error small.

June 2005 Permalink