The Dutch investigation board headed by Pieter van Vollenhoven, examining the crash of the Boeing 737-800 Turkish Airlines flight TK1951 at Amsterdam Schiphol last week, released its preliminary report.
Simply put it was a combination of a faulty instrument, a breakdown in “artificial intelligence” (AI) of the plane, pilot error, and poor weather.
The instruments and “artificial intelligence”
The plane had a faulty radio altimeter (radalt). At 1,950ft altitude, one of the two radalts suddenly gave a wrong reading which fooled the autopilot into thinking the plane was just few feet above the runway. Since the auto-throttle was also engaged, which is a perfectly normal procedure, the autopilot reduced the engines thrust to ‘retard’ or ‘idle’ mode as we normally experience just before touchdown, except this was 2,000 ft up in the air. Naturally the plane started slowing down.
In the right seat the co-pilot was receiving training in making a landing by automatic pilot. The investigators say that too much time was spent exchanging information between the instructor and the trainee, and not enough was spent checking the actual readings which would have alerted the crew to the problem in time to override the automatic pilot.
Thick fog and low cloud may also have played a part in their failure to realise their actual height.
The plane slowed down in to a stall, which is when the alarms sounded. The pilots tried to spool the engines back up, but it was too late.
Turkish Airlines’ maintenance conflict
Another fact has arisen. Two weeks ago, a serious conflict arose between the union representing maintenance workers and the management of Turkish Airlines. The union warned the management of serious shortcomings in the maintenance system, including inadequate staff to cope with the rapid growth of the airline fleet, which could lead to safety risks in the air.
Sharing the blame
The investigators have concluded that blame for the accident must be shared between Turkish Airlines THY, the pilots, and the manufacturer, Boeing. The involvement of human error means that there will be legal repercussions, civil and possibly criminal, but then the pilots have already paid the ultimate price.
Is time to return to the basics ?
For long airplane manufacturers have tried to reduce the workload on pilots by automating tasks. One could say, they are trying to design the pilot out of the cockpit. The pilots have become increasingly dependent on the computers and their AI. It has become a vicious cycle to the point, many pilots now feel they are flying a computer not a plane.
Since the second radalt was working fine, one would naturally pose the question, why did Boeing build in to their AI some failsafe when the two radalts are giving conflicting information? Should Boeing not have put more thought in to this situation? A simple “whoop whoop” alarm that alerts pilots especially in this situation of conflicting instrument information before putting the engines in to idle?
Across the blogosphere I have seen many Boeing fans criticising the AI on the Airbus “Fly By Wire” system, and citing the June 26, 1988 crash of an Air France Airbus A320 as AI gone haywire.
It appears, that Boeing’s AI is not much smarter either. Is it time for both these plane manufacturers to return towards the basics and give more control to the pilots? Machines are fallible, but it is humans who pay the price.
Share your thoughts via a comment.
In the mean while, Turkish Airlines has made an offer of compensation. Full details are in this release.