ADM, Chaos Theory, and Why There Will Always Be Crashes

  • E-Mail this Article
  • View Printable Article
  • Text size:

    • A
    • A
    • A

Several years ago, I saw a talk by Captain David Cronin. If you don't know the name, you might remember the story. In 1989, Cronin and his crew were climbing through 22,000 feet en route from Honolulu, Hawaii to Sydney, Australia, when the forward cargo door blew off their Boeing 747-100. The entire cabin depressurized fast enough to fill it with mist and the right two engines were taken offline. (Nine passengers were also blown out or killed by the resulting damage).

Cronin said the first thing they did was grab the backup oxygen masks. Unfortunately, Boeing put the O2 tanks for the crew by the cargo door for easy maintenance and they were no longer on board the airplane. Clearly the next action was an emergency decent, which Cronin initiated from memory, but then called for by checklist from his first officer. On the checklist is, GEAR – DOWN. The FO read the item and looked to Cronin before lowering the gear. Cronin chose to leave the gear up, because the plane was already passing through 10,000 feet by then, and he saw no need to make it descend faster. He said (and I'm paraphrasing here), "Someone taught me that you run the checklist only until the task is completed. And then you stop there."

They landed safely, despite extensive aircraft damage and best guesses by the flight engineer on landing speeds with so much stuff broken (the FE's paperwork was blown all over first class during the decompression). What's fascinating about this, however, is that when simulator runs of this scenario were done to see how different crews would react, every crew that put the gear down early went swimming. It seems that, even dumping fuel as fast as possible, the 747 with that much damage and weight and only two engines running is too heavy and draggy to get back over dry land if the wheels are down.

So, the captain broke from protocol … and saved the day.

Now there's an indication the captain of Flight 3407 pulled back on the yoke right after the aircraft warned the crew of an impending stall with a stick shaker. and then tried to fix the situation with a nose-down action. The captain's pull-up may have put the aircraft into a stall/spin and crash. What would make a well-trained aviator do such a thing?

I don't know. But here's how the randomness of reality could make the right thing the wrong thing. Suppose the crew is worried about icing. The captain is thinking about his icing training and the reading he's done on his own. He knows deploying flaps and gear have the potential to cause a problem, so he's spring-loaded. There is ice on the airplane—not more than the Dash 8 can handle, but enough to cause a dangerously-low airspeed at the current power setting. The stick-shaker activates and the plane pitches down to recover from the immanent stall.

What are two indications of a tailplane stall? Tailplane vibration and nose-down pitch. What's the correct recovery action? Full aft elevator.

I'm not saying this is what happened outside of Buffalo that day. It seems an unlikely chain of events, in fact, since tailplane vibration being felt in the yoke applies to aircraft without mechanical controls, not hydraulic ones. The captain should have known that. But it's not impossible that stimulus-response mentally developed in earlier flying led to the wrong action in a different airplane, and that's more the point. An old habit or reaction applied in a new scenario might save the day, or end it.

No matter how long we try to regulate, train or procedure away the winds of chaos, there will still always exist just the right constellation of events where right becomes wrong, wrong becomes right and survival is a matter of a split-second, lucky guess.

Comments (28)

I agree, I think your comments are spot on.

Posted by: daniel schultz | February 21, 2009 9:06 AM    Report this comment

I also concur. I also suspect that as aircraft control and navigation systems become progressively more complex and automated, the permutations of failure modes will also progress leading to accidents the causes of which we can't even imagine right now. An old friend, now retired from the airlines, once pointed out to me that nearly every item in their procedures manual represented the result of someone else's accident. By the way, I've seen the NASA video on tail plane icing and have no doubt that full aft "stick" is part of the proper recovery procedure, but I don't understand why. If the tail plane has stalled, that suggests that it has exceeded its critical [negative]angle of attack. Lowering flaps may increase that negative AOA and precipitate the stall, but it seems to me that pulling back on the elevator control would further increase the negative AOA and aggravate the stall. Could you enlighten me about that, as I am obviously missing something here?

Posted by: John Johnson | February 23, 2009 9:51 AM    Report this comment

What Jeff is describing can be termed a "normal accident." I didn't think that up abll by myself: its part of the title of the book "Normal Accidents: Living with High-Risk Technologies" by Charles Perrow. He basically descibes why accidents are a normal state of affairs in what he calls tightly coupled, highly complex systems where operators of said tehcnologies cannot correctly perceive what the exact nature of the problem is (for a variety of reasons like bad information from failed instruments or poorly designed systems, etc.) and more importantly, cannot therefore know what the correct action is.

The book was written the 80's, and one Amazon reviewer who is a commercial pilot said Perrow's conclusion about safety and ATC was off the mark and I tend to agree, but I think the two episodes discussed here by Jeff are much more in line with Perrow's basic tenent. Even being somewhat out of date, its a very interesting read.

Posted by: Joe Marino | February 23, 2009 9:55 AM    Report this comment

I think we are forgetting that this crew was on autopilot, in icing conditions, which is prohibited in every Transport Category Turboprop FAA Approved Flight Manual I am aware. Any ice accumulation can be very detrimental to flying qualities and will increase stall speed, I have seen that up close and personal.

I do remember reading that the aircraft slowed upon gear and flap extension until the stick shaker/pusher went to work.

Certainly, the pilot actions were consistent with recovery from a tailplane stall, except that technique is not appropriate for a stall induced by slow speed.

There is one fact that appears clear, if the aircraft had not slowed into the stall when the gear and flaps were extended, there would have been no stick shaker/pusher.

As we all know, the autopilot will mask the effects of ice buildup. Hand flying in that situation provides an absolute appreciation of exactly how "sloppy" the aircraft can get under such conditions.

As a person with 42+ years as an airline and professional pilot, I question the empahsis the manufacturers/operators put on flying the aircraft on the autopilot. Does that "let the autopilot do it" philosophy discourage the acquisition and maintenance of skills that caused this crew to fly on the autopilot in spite of the AFM prohibition?

Fly your airplane! Keep the speed up and there is no stall, therefor no stick pusher, and no perception of tailplane stall, therefore no pull up into an accelerated stall.

Posted by: THOMAS OLSEN | February 23, 2009 12:19 PM    Report this comment

I agree with Thomas – “Fly your airplane!” As a long-time professional pilot, I find it disturbing how few hours of hand-flying most regional and airline pilots actually perform each year. In bygone days, aspiring airline pilots acquired a good deal of experience in various aircraft and weather conditions before being considered to fly heavier aircraft. The current system often places low-time pilots fresh out of flight school into the right seat of a commuter or regional plane where they will spend most of their time watching the plane fly itself.

While I reserve judgment on the crew of Flight 3407, I think the media emphasis on the effects of icing is misplaced. Icing (as well as all inclement weather) will always be a part of flying. While autopilots are always improving, there is simply no substitute for an experienced pilot at the controls, one who is confident enough to hand-fly the aircraft when necessary, and weather-savvy enough to know how to compensate for whatever conditions are present. Unfortunately, current airline practices make those pilots increasingly harder to find.

Posted by: Joe Brand | February 25, 2009 9:58 AM    Report this comment

For John Johnson. In a normal aft tail fixed wing aircraft, the tail plane produces a trim force opposite of wing lift at most CG postitions. In simple terms, the tail is pushing down to balance the weight of the nose. In a properly designed and loaded aircraft, the goal is to balance the aircraft over the wing (in simple terms), a condition that would create no need for trim force but which would result in a marginally stable airplane. With this in mind, the average aircraft designed to be slightly nose heavy in most conditions. In our average airplane, the horizontal tailplane and elevators produce a small amount of downforce at all times. When the tailplane stalls, that downforce is gone and the intentionally nose heavy aircraft immediatly pitchs nose down. The full aft corrective action reduces the tailplanes angle of attack and unstalls the tail. On the average aircraft, extending flaps increases the nose down pitch which requires even more balancing trim force from the tail, a condition that explains why extending flaps could cause an iced up tail to stall. It is being asked to provide more lift (opposite to the wing) and if it is iced up, it may be doing all it can do before the flaps are extended.

Interestingly enough, the Beechcraft 1900 turboprop is designed with a few extra tail surfaces and that aircraft is one only a few known to me to fly with the tailplane producing lift instead of downforce at normal CG settings.

Posted by: Matt Nowell | February 25, 2009 10:31 AM    Report this comment

Thank you, Matt, but I don't think I agree with your analysis: I'm well aware that the tailplane in normal flight exerts a downward force ("negative lift" as it were)but to do that the tail plane must have an angle of attack opposite that of the wing, i.e. a "negative" AOA if the wing's AOA is considered "positive". Pulling back on the elevator control raises the nose in normal flight because the elevator's trailing edge is raised, thus increasing the "negative" camber of the tail plane and thus creating more "negative" lift. That means that the AOA of the tail plane (which again is in the "negative" direction, or downward in wings level flight) has to have been further increased. A flying surface stalls when its maximum AOA has been exceeded, regardless of airspeed or direction of flight. Pulling aft on the control wheel cannot but further increase the AOA of the tail plane. Thus I disagree with your assessment that "full aft corrective action reduces the tailplane's angle of attack and unstalls the tail."

In fact, normal corrective action for an aircraft (wing) stall is forward stick to reduce the (negative) lift created by the tailplane. Downforce reduction occurs because of a reduction in the tailplane's (negative) AOA. So forward control movement = lower tailplane AOA and downforce, aft control movement = higher tailplane AOA and downforce.

Posted by: John Johnson | February 25, 2009 11:41 AM    Report this comment

Regarding "Thus I disagree with your assessment that "full aft corrective action reduces the tailplane's angle of attack and unstalls the tail." in John Johnson's post. Some may have seen the following clip on tail icing and stall, but for those that have not, take the twenty minutes or so required to view it and do so. A tail stall caused by ice build-up requires a "pull" not a push (amongst other things).

Scary stuff; but, knowledge is power.


Posted by: John Swallow | February 25, 2009 2:42 PM    Report this comment

John, I think you missed my point: As I said in my 1st post, I have seen that video and I have no doubt that proper corrective action includes full aft stick. What I disagree with is the explanation of WHY that works given by Matt - I do not agree that full up elevator (i.e. full aft stick) REDUCES the tailplane's AOA. Remember that in level flight, the tailplane's AOA has to be a negative number of degrees if the tailplane is going to generate downward force. Any way you cut it, full aft stick has to INCREASE the tailplane's negative number of degrees of AOA. Higher AOA (whether positive or negative is just an arbitrary convention) means closer to stall. My suspicion is that ice accretion bad enough to stall the tail causes such destruction of laminar flow that the tail plane has no effective AOA, and full elevator actually re-establishes some effective negative AOA. Jeff Van West, you started this thing - help us out here!

Posted by: John Johnson | February 25, 2009 8:17 PM    Report this comment


Sorry if I ame across as contradicting your point of view. Didnt' mean that. I concur with your last sentence... That's what I was thinking also.


Posted by: John Swallow | February 25, 2009 9:31 PM    Report this comment

No appology necessary, John, and it was probably I who misunderstood you. The harder I think, the confuseder I git!

Posted by: John Johnson | February 25, 2009 10:13 PM    Report this comment

I think in the video, which I watched, and we're talking about, they speak of a "reattachment of air-flow". They further discuss that if that reattachment occurs enough aft on the rear stab, that a downward force is induced upon the elevator. I don't mean to confuse the original conversation of a negative angle of attack on the elevator that causes a stall. Keep in mind that flaps on the main wing decrease stall airspeed but also increase the angle of attack. These are two seperate arguments. A highly cambered airfoil has more lift. An elevator in an up position duplicates somewhat a higher cambered airfoil (just as flaps do on the main wing). There is confusion between simple increase in AOA and stall speed. Even though flaps on a wing increase drag and AOA, they increase lift moreso. Thus does the elevator increase lift more than AOA would indicate (camber).

Posted by: eric hanson | February 26, 2009 2:40 AM    Report this comment

To help clarify; for any airfoil, lift is proportional to AOA up to stall AOA. This is for a GIVEN airfoil shape. Although deployment of flaps changes the chord line of the airfoil, and hence AOA, it is a better lifting shape and has an entirely different lift to drag curve (called a polar). Although up elevator would seem to increase AOA (which it DOES do), the higher camber decreases stall speed just as flaps do on the main wing. I hope this "second try" makes sense to some. There are such things as "full flying stabilizers" which change AOA but do not camber. Hope this helps.

Posted by: eric hanson | February 26, 2009 2:48 AM    Report this comment

Eric - very interesting points. Perhaps in other words one could consider that the ice contaminated airfoil in a tailplane stall is "stalled" not because of excessively high AOA but because the leading edge ice simply spoils all or most of the airflow over the surface. Full up elevator in this case could recreate some useable camber under the tailplane. I was assuming, probably incorrectly, that a stall = excess AOA, but that may only apply to an uncontaminated, unspoiled airflow. I wonder if a stabilator like on my Comanche would respond less well in that scenario since the camber doesn't change, making icing even more dangerous than it already is?

Posted by: John Johnson | February 26, 2009 11:46 AM    Report this comment

John, My original reply had to be cut short because of the limitation put on post length. I should posted nothing instead of trying to cut out enough to make it fit.

I think I see your point, but I've flown aircraft with a trimable horizontal stab for too long. I'll have to think about this before replying further.


Posted by: Matt Nowell | February 27, 2009 5:10 PM    Report this comment

Yes, Matthew, think about it.

Posted by: eric hanson | February 28, 2009 3:58 AM    Report this comment

I think that NASA has created some of the confusion. They state that you pull back to deal with a tailplane stall. A careful read of the data reveals that the full back instruction is to overcome elevator snatch to the full down by an elevator hinge moment problem. If the elevator has not yet been yanked out of your hands, then I find it unlikely that full aft is the best course of action. The aft elevator recovery is obviously to regain control after the elevator snatch, and has very little to do with angle of attack.

Posted by: Matt Nowell | February 28, 2009 11:24 AM    Report this comment

I didn't here "snatch" as part of the language. I think it does have to do with the reattachment of airflow well back on the stab, and into the actual elevator surface. A true tailplane stall IS what they refer to however. I think this or these aircraft certifiction standards need to be looked at before we confuse pilots with more than one stall recovery method. ....That would be bad. -Eric

Posted by: eric hanson | March 1, 2009 7:15 PM    Report this comment

I'll put up two posts here to accomodate the space restriction.

There are two conditions to consider regarding ice contaminated tailplane stall: first, a genuine aerodynamic stall of the horizontal stabilizer. This is the worst case. More likely is a change in the aerodynamic balance of the elevator. Note that, in fully powered controls, aerodynamic balance is not a factor. Thus powered controls are not subject to the latter condition, but could be subject to the complete stall of the stabilizer.

When the aerodynamic balance of the elevator is changed, there are two possible scenarios. In one, the elevator is "snatched", or pulled from the pilot's hand by aero loads at the elevator. This can be a very powerful event; however the elevator is still functioning and re-cambers the stabilizer into the nose down configuration. In the second case, again more likely, the elevator stick force "lightens", or possibly reverses. When it lightens, the pilot finds that less stick force than normal is required for elevator deflection. In the "lightened" case, the stick force is still increasing with elevator deflection, just not as much as it would in the normal, uncontaminated condition. However, the stick force gradient can actually reverse, meaning that the farther the pilot deflects the elevator, the less force is required.

Posted by: Steven Green | March 2, 2009 8:06 AM    Report this comment

Both of these stick force conditions can induce disaster. In the typical ICTS accident, landing flaps are selected close to the ground (propeller driven standard). Nothing has happened yet...until the pilot detects a need to slightly correct into a "fly down" glide slope indication. He gently nudges the elevator nose down...and gets a substantially larger elevator deflection than he expected for the push he applied. The nose tucks, and he corrects with the obvious, and proper, response: a massive pull to the nose up elevator deflection. The stabilizer is re-cambered, he gets normal force and response, and the nose rises rather dramatically given that the pull force was powered by adrenaline. Now the pitch attitude is too high. So now he responds by applying a significant, but not reckless, nose down push to restore the correct attitude.

Unfortunately, when he does this, he de-cambers the stabilizer; the stick force gradient is again changed, and his not-unreasonable push force results in the elevator traveling all the way to the stops.

The typical ICTS accident (Viscounts, Convairs, Jetstreams, YS-11s) will exhibit a nose dip, a stronger nose rise, and a final pitchover with impact nearly vertical.

You may wish to look at FAR 25.143. Paragraphs (i) and (j) specify the criteria for ICTS protection in new certifications. This is a recent revision; however, a number of authorities have applied these criteria for around ten years now.

Posted by: Steven Green | March 2, 2009 8:14 AM    Report this comment

Please explain that FAR paragraph Steve (simply for all of us). It IS simple to understand your narative, and it clearly involves negative stability. The best pilot is in deep doo doo here, and a B-2 quality autopilot might just start smoking at some point. What is the "big picture" answer here? I hope I was not off target with my posts Steven, can you give some comment on my and John Johnson's input here? I hope I have not misled anyone.

Posted by: eric hanson | March 2, 2009 11:08 AM    Report this comment

FAR 25.143 provides a requirement, during certification, for a zero G pushover maneuver with critical artificial ice shapes attached to the horizontal stabilizer. The requirement is for a push force to exist all the way to zero G, and the that push force to continue increasing at least until 0.5 G is attained. A pull force of no more than fifty pounds should cause prompt recovery. This certification requirement is intended to identify design problems that lead to an ICTS event.

You are correct with respect to negative stability. The major problem with this issue is that very few pilots have experience with unstable aircraft. Even military folks, flying upside down at Mach 2, are still flying a stable aircraft, even if the stability is artificial. The control relationships when one encounters instability are way too different to ascertain quickly enough, which is what leads to the PIO.

Posted by: Steven Green | March 2, 2009 8:55 PM    Report this comment

I think we may have missanalyzed this somewhat. The automated stick shaker activated(and/or stick pusher). This is a built in "safety feature" which I would think had to perform consistantly even with the aritificial ice forms for that certification. The recovery method would be: Decrease angle of attack. Well, the pusher will have done this for him. If one defeats the safety feature by pulling against it quickly, the result is a secondary/or prolonged stall(FDR shows he held up elevator until impact). In a moment of surprise...a pilot's instinct is to put in some input..any input. Perhaps his instinct was just to counter what seemed to him this "elevator snatch". If it were just a shaker, he would take time to analyze (quickly) whether at this airspeed (he just dirtied it), he shouldn't just initiate stall recovery as trained from his very first flight. No judgment intended, we weren't there. My point: a stick pusher may just confuse the very most inate training any pilot should imprint upon themselves. We might rethink all this dumbing down of the flying task with automated stick pushers and other "safety features".

Posted by: eric hanson | May 13, 2009 1:07 AM    Report this comment

Rereading comments including my own, I wonder. A canard is a lifting surface (and if you are reading, you needn't have an explanation). When it stalls the plane naturaly decreases AOA (main wing). If a tail stalls...guess what, same natural correct result. Fighting this is deadly. You can give me all the tail stall info from NASA (accurate though it may be), and it tells nothing to a pilot. Back stick was bad in this case gentlemen, and ladies! Airspeed is the cure to most conventional stalls, the pusher offered this and apparently was over-powered. We shouldn't train pilots for years to do one thing and then say "'s a case you can make use of it in a panic at 2,300 feet". So train to recover from a stall all your life, and then do the opposite...Thanks NASA! NASA should recommend decertifying aircraft that show their discovered issues? Don't tell pilots to flip a coin!

Posted by: eric hanson | January 3, 2010 4:56 AM    Report this comment

Regardless of background, experience or education, the Airline Operator is responsible to assure all pilots, when they are assigned to the line, have sufficient training, testing and qualification to assure that they have the knowledge and skill to operate safely in the expected operational environment, and even in the corners of that environment.

This is not the first accident in the last 30+ years that can be laid at the feet of misuse or overreliance on automation. However, this accident finally puts an exclamation point on what is an industry wide deficiency. Why is it that the FAA, the airlines, the unions, the manufacturers, and perhaps the pilots themselves, failed to recognize the general deterioration and/or lack of basic flying skills among air carrier pilots is a problem? How can expect that pilots will have sufficient hand flying instrument skills to fly the airplane when all that "foolproof" automation goes south, when they never get to practice those skills?

Posted by: THOMAS OLSEN | January 3, 2010 3:14 PM    Report this comment

The crew should have been hand flying the airplane in icing conditions in the terminal environment. If they had the skill and appropriate instrument scan, they would have been flying the airplane and closely monitoring the airspeed as part of their instrument scan. There is nothing magical about that.

The ICTS training given to this pilot points out a problem with what can pass for "training" these days. Show a pilot a NASA (or other video or slide presentation) and call it complete. Did the video state the differences between an ICTS and a wing stall? Was the pilot fooled by a possible icing induced down pitch at autopilot disengagement? Was any of that ever discussed?

A properly trained pilot, when the stick shaker activates and a slow airspeed is noted, should have known that what he had encountered was NOT an ICTS, but a wing stall, and that it should be handled exactly the way an "approach to stall" is handled in the simulator during training and checks.

Posted by: THOMAS OLSEN | January 3, 2010 3:33 PM    Report this comment

"A properly trained pilot, when the stick shaker activates and a slow airspeed is noted, should have known that what he had encountered was NOT an ICTS, but a wing stall, and that it should be handled exactly the way an "approach to stall" is handled in the simulator during training and checks." To quote Tom above...and I agree. We'll never know exactly what the pilot was thinking, but one might imply that he was encountering "some ice" anyway, and may have thought about ICTS first when a response was quickly needed. I don't feel comfortable with pilots being asked to second guess instinctual stall training (power on or approach to land stalls). Seems the company training will be scrutinized. You don't have a second chance to make the right moves in this situation.

Posted by: eric hanson | January 3, 2010 7:20 PM    Report this comment

They landed safely cisco certification, admitting all-encompassing aircraft accident and best guesses by the flight architect on landing speeds with so abundant being torn (the FE's paperwork was destroyed all over aboriginal chic during the decompression) cisco training. What's alluring about this, however, is that if actor runs of this book were done to see how altered crews would react, every aggregation that put the accessory down aboriginal went swimming. It seems that, even auctioning ammunition as fast as possible, the 747 with that abundant accident and weight and alone two engines active is too abundant and blah to get aback over dry acreage if the auto are down microsoft certification.

Posted by: Hank Freid | August 12, 2010 9:43 PM    Report this comment

Add your comments

Log In

You must be logged in to comment

Forgot password?


Enter your information below to begin your FREE registration