China Airlines 006—Minor Problem, Major Consequences

En route from Taipei to Los Angeles, a Boeing 747SP experienced a loss of thrust in its number 4 engine. This engine had previously failed on two earlier flights, but had successfully been restarted after descending to a lower altitude. The cause of the shutdowns was not identified. A Boeing 747 can fly perfectly well on three engines, and most airlines permit it to proceed to its destination after losing one engine. But in this case a seemingly minor error in reconfiguring the aircraft after shutting down the engine led to a wild ride, with the aircraft plunging 9100 metres and experiencing accelerations exceeding 5g.

Mentour Pilot looks at what happened and why.


I watch most of his videos. He’s grown into a good presenter.


Great point about how much punishment the 747 could take and still remain flyable. Boeing used to be wonderful!


Here is another near disaster where I am again shocked at the pilots’ behavior. When three independent sets of flight instruments tell you the aircraft’s attitude, it is beyond foolish to decide your kinesthetic senses are correct and the instruments are wrong! Also, to not use the rudder when the plane is obviously yawing due to asymmetric thrust from the engine-out, suggests a complete absence of basic flying skills. As in the last video we saw, where the pilot flying could not assume a heading and achieve an assigned altitude during a go-around (not nearly as complex a maneuver as suggested in the analysis), it appears clear that commercial pilots rely on automation to such an extent that they no longer possess basic skills to do the first thing in an emergency: aviate (then navigate, then communicate).

Two answers to this obvious problem occur to me. First, to require some recurring flight time in simple, small aircraft without automation and with a flight instructor. Second, lots of simulator time with scenarios of complete and especially partial, automation failure.


Amazing. Just amazing. The lowest experience level here is 7 thousand hours! I have a mere 1400 hours and even I know to always trust your instruments unless there is a clearcut disconnect in their performance. NOT because you feel they are wrong but they don’t internally co-incide. That is something you spend a lot of time practicing - finding an instrument that is out and flying with the rest of them.

I believe CW has a point. Many, perhaps most, of today’s commercial pilots have lost real flying skills because they fly automated all the time. Private pilots practice loss of instruments regularly. In a check ride the examiner can, and will, throw in a wrinkle - an instrument or other system failure. I owned a Bonanza that had both alternate vacuum and alternate alternator, so dealing with such problems was far easier than many have. But you still have to know how to fly, what your problem is, fly the doggone plane while you solve the problem, then fly the solution.

This performance is simply unacceptable. Retraining is clearly necessary, and since they got lucky enough to survive the episode, they should ALL be subject to additional training.


We have to remember that the particular incident happened in 1985 – a third of a century ago. And there have not been too many repeat incidents, which suggests that some lessons have been learned and applied.

On the other hand, I once found myself on a long flight sitting next to a pleasant business jet pilot who was returning from delivering a plane to Indonesia. He acknowledged that he would fly the plane manually for maybe 10-20 minutes during take-off & landing – but he was busy the rest of the flight, monitoring all the gauges! Yes, there probably is a need for pilots to spend more time manually controlling the aircraft.

On the other other hand, that video did make clear the incredible complexity of the control panel of a large aircraft, and the astonishing variety of procedures which a pilot may be required to perform in different circumstances. Training! Training! Training!


The number of things that can go wrong as to controlling the attitude of an aircraft is not infinite (the forces which must be acting to produce a given disturbance are well understood). That suggests that an AI system which may be receiving data from the flight control automation or may be completely separate from and independent of it (this connection or not needs to be thought through more deliberatively than off the top of my head), ought to be able to quickly diagnose a problem and tell the pilots what to do to fix it. Again, there are only so many forces acting on an aircraft and so many inputs possible to correct an upset. This should be workable, I think.

In the case in point, once a yaw was detected due to the #4 engine out/asymmetric thrust and the autopilot could not correct it with aileron, an AI system (or, shockingly, a competent pilot) would simply say “apply left rudder”. Had this been done, the rest of the sequence with near loss of the plane & passengers would never have occurred.