Friday, May 15, 2009

Thirty Seconds of Confusion

Beep!

"Three thousand for two thousand."

It was a dark night; the landing lights lit up the thick layer of clouds slipping rapidly around us. The plane bobbed rhythmically in light turbulence as we descended at flight idle power. Paul, my First Officer, slid his seat forward and cleared his throat.

"I was talking with what's-his-face the other day, our ALPA communications guy."

I looked over at Paul. "Yeah? What about?"

"Advertising. It occurred to me that we could do a much better job of getting out our message to the general public."

"Definitely."

"So why don't we put out some ads that capture people's attention?" Paul continued. "I was reading a magazine the other day, and there was this really brilliant ad. It featured a lovely voluptuous young lady, and she didn't have any clothes on - this was taken from the back, mind you - but this ad was for -."

PPPPPRRRRRBBBBBBBBTTTTTTT! DING DING!

The cockpit erupted into a cacophony of alarm horns and lights as the autopilot disconnected and the yoke began vibrating furiously. I snapped back to the instruments to find the airspeed dangerously low and the torque still at idle. I grabbed the yoke and shoved the power levers forward. "Set power!" I commanded.

It was obvious what had happened, we had leveled off at 2000 feet during the conversation without me noticing and bringing the power back up. Now, with the engines at full power, the airspeed stopped decaying and started creeping back. The stick shaker stopped momentarily.

"Altitude!" Paul called out.

Crap, we had drifted below 1900 feet. I applied some back pressure to the yoke. PPPRRRBBBBTTTT - the stick shaker started up again and the wings began a light burbling back and forth. Ish, don't want that. I eased the back pressure and gingerly nursed the altitude back to 2000 feet. The airspeed finally crept up to a safe number and I started breathing easier.

Fortunately for my career, this incident took place in a SF340 simulator rather than the JungleBus, and my "First Officer" was actually the director of training for a well-known regional airline. This company does stall training completely differently from most airlines, and in the aftermath of Colgan 3407 and Turkish 1951, Paul invited me to fly their jet and turboprop simulators and experience it for myself. We did departure stalls just after departure, approach to landing stalls while approaching for landing, and high altitude stalls at high altitude. A great many stick shaker events in the real world involve distraction at a critical time while on autopilot, so we did the scenario described above. While all of this might seem quite logical to an outsider, it is actually a revolution in the airline world. Moreover, it is done in apparent contravention of the FAA's Practical Test Standards.

In the last few days, a great many pilots - including some of the commenters on my last post - have been asking what in the world could possess a presumably competent airline pilot to pull up in response to a stick shaker - or for that matter, to use 80 to 120 pounds of force to override the stick pusher that might've saved his life. It's the most puzzling aspect of this crash. Even if the crew was relatively inexperienced, it is drummed into pilots from day one that you don't pull up in response to a stall. What could cause an airline pilot to abandon this most elementary of precepts?

Even if the pilot did survive to answer for his actions, I'm not entirely sure that even he could've given a satisfactory answer to the question. Since he is not around, all we - or the NTSB - can do is speculate about some of the possible culprits and do our best to eliminate them as potential causes of future accidents. There are three primary possibilities that have been discussed: 1) the Captain was a poor pilot to begin with; 2) he simply got confused in the heat of the moment; 3) there was some latent defect in his training. These are not mutually exclusive theories, all three could have come into play simultaneously.

By now it has been widely reported that the Captain failed five checkrides in his career. The first was his instrument checkride, back in 1991, on the partial-panel VOR approach and the NDB approach. He disclosed this failure on his application at Colgan. The next two failures, on his Commercial-Single Engine and Commercial-Multi Engine rides, took place in 2002 and 2004 respectively, and were disapproved for a fairly wide range of tasks. He only disclosed the instrument ride failure on his application. At Colgan, he failed a recurrent Proficiency Check as a First Officer and his upgrade/ATP ride. He also had to repeat a small portion of his initial PC as a new FO in the SF340.

There are a great many capable pilots who have a checkride bust or two in their past, but a long string of them raises warning flags. The interesting thing is that when you read through the NTSB Human Performance Group's interviews, everyone describes Captain Renslow as a good, consciencious pilot. Many FOs he flew with described him as above-average. Perhaps it's simply a case of not wishing to speak ill of the dead, but if he really struggled on the line you'd think the NTSB could've found someone who would have told them about it. It's hard to reconcile the popular Captain with all the checkride busts. It's possible that he simply wasn't a good test taker. That's not insignificant for our purposes: falling apart on tests can be a symptom of not coping well with pressure, period.

There was plenty of pressure to be had in the last thirty seconds of Colgan 3407. That the stick shaker was a complete surprise is self-evident. We don't know where the Captain's attention was in the moments before stick shaker activation; perhaps looking at the wingtips to see how the deice boots were coping, perhaps around the cockpit to see if anything had been missed during the rushed descent and approach checks. Maybe the long day had got to him and he was simply zoning out. It doesn't really matter; it's very unlikely he had any clue that the stick shaker was coming before it went off. It is difficult to explain to those who have never flown airplanes with stick shakers just how jarring their activation is - even in the sim, much less the real world. The whole idea behind them was to have one signal in the cockpit that is so overpowering and unmistakable that the crew cannot possibly ignore or misinterpret it. Both yokes shake so heavily that you can feel it even if your hands are nowhere near the yoke. Loud clattering noise fills the previously quiet cockpit. The autopilot disconnects with the accompanying lights and aural warnings. In the Q400's case, this is a loud horn that repeats over and over until you acknowledge it by pressing the autopilot disconnect button on the yoke. The Colgan crew never did so - they had their hands full enough already - and that sound must have surely contributed to the chaos and confusion that filled that cockpit in the last 30 seconds.

The sudden cacophony had a clear meaning: do something, now. The Captain indeed reacted very quickly, within half a second. More than a few pilots have suggested that he had tail stalls on the mind. It's possible. He'd just transitioned from an airplane that was known to be susceptible to tail stalls (early models, anyway) and had recently viewed the NASA video on tailplane stalls in recurrent training. The crew had been talking about the icing only a few minutes before. With a tailplane stall, of course, one would not expect to see the stick shaker activate, as that indicates a high aircraft angle of attack and, by extension, a low tailplane AoA. I'm not sure that the distinction would be evident to anyone within the space of a half-second, but by the same token I'm a little skeptical that anyone would think of a tailplane stall within a half second in the first place (much less remember that the corrective action for a tailplane stall is to pull up). It's possible that five or six seconds later the Captain mistook the stick pusher for a tailplane stall (they would feel similar in an aircraft with unpowered flight controls, although not in a Q400) and that's why he fought it. It's very easy to play these parlor games after the fact, having reviewed the NASA video and FAA circulars and discussed among ourselves. At the time, caught by surprise and with little idea of what's going on and events moving far faster than he could really think about them, I rather doubt that the Captain consciously thought about what he was doing, in the same way that the First Officer obviously wasn't thinking about what she was doing when she retracted the flaps in the middle of a stall. Amid the confusion, pure instinct took over.

Why that instinct might involve pulling when new pilots are taught over and over again to push may have its roots in the way that most airlines teach stalls. To begin with, they are not even stall recovery procedures; they are stick shaker recovery procedures. The ATP PTS directs you to recover at the first indication of a stall, which includes the stick shaker. Many pilots will never experience a stick pusher or a real stall in the simulator unless they request it; it was never part of the syllabus at Horizon or NewCo. The maneuver is typically taught and checked well above the ground. The setup is far from realistic: the applicant usually hand-flies and stops trimming well before the stall. The reason to do so is that it makes the recovery easier: the plane won't pitch up when you apply power. The purpose of stall training is really to prepare the student to pass the maneuver on the checkride rather than to prepare them for the possibility of being surprised by a stick shaker on the line.

And this brings us to the most outrageous thing about stall training at many airlines. Applicants are taught to hold their altitude throughout the maneuver. Again, this is due to the Practical Test Standards, which state that an applicant must:
Recover to a reference airspeed, altitude and heading with minimal loss of altitude, airspeed, and heading deviation.
The FAA never defines "minimal loss" of altitude. A great many instructors and check airmen have substituted their own standard, often 100 feet as several Colgan instructors testified. People can be and have been failed for trading altitude for airspeed during a stick shaker recovery. Instead, you are taught to immediately go to full power, and use whatever yoke force is needed to keep the airplane level while it accelerates. This often involves "riding the shaker" for some time. Let me say that again: we are being taught to stay in the shaker for longer than is necessary. Because we stop trimming at such a high airspeed, this can involve significant back force on the yoke until the airspeed increases again. Therefore, you are developing the exact motor memory that, if applied to a real-world situation like Colgan 3407's, will induce exactly the wrong control movements.

It's entirely possible that the Captain was reacting to the stick shaker exactly as he did in the simulator and simply overreacted a bit with the adrenaline rush. It didn't take that much back pressure to start the abrupt pitch-up after the shaker, only about 25-30 pounds according to the Flight Data Recorder. The fact that the autopilot was engaged right up to the stick shaker meant that the plane was trimmed for the speed at which the autopilot disengaged, which certainly didn't help matters when the Captain shoved the power levers forward during the pitch-up. I don't doubt that he was as surprised as anyone that his reaction to the stick shaker induced a 30 degree pitch up and subsequent stall. It was still a recoverable situation at that point; it was fighting the stick pusher the whole way down and retracting the flaps mid-stall that ultimately doomed the crew. These actions may reasonably be attributed to panic at a situation that had quickly spiraled out of control.

So why did stalls come to be taught this way? I think I see the FAA's original reasoning. A lot of training and checking used to be accomplished in real transport category aircraft, many of which reacted very poorly to full stalls. In the interest of safety, the FAA decreed that recovery be initiated at the very first sign of a stall. Simply increasing one's airspeed from a low number to a high number doesn't seem like a very difficult task, and nobody wants crews to be diving transport category jets at the ground in a low-altitude situation, so the FAA added the language about minimum loss of altitude. Transfer this to the simulator, where the element of danger is removed, and many check airmen began treating it not as a survival maneuver but a proficiency maneuver not much different than steep turns.

Many major airlines at least include simulator training to the stick pusher for their pilots, but as far as I know only Paul's regional airline has completely revamped the way they do stall training. They teach their pilots that reducing angle of attack promptly is the most important thing in recovering from a stick shaker, and that this involves both increasing power and lowering pitch to trade some altitude for airspeed. I tried both their method and the traditional method in the sim, and using the new method resulted in far less time spent in the shaker in exchange for altitude loss generally no greater than 200-300 feet (the scenario I described at the beginning of this post was using the traditional recovery method). Just as importantly, this airline trains and checks stick shaker recoveries using the most common scenarios in which real crews have encountered stick shakers: accidental reversion to pitch mode after takeoff, mountain wave at high altitude, leveling off on a non-precision approach, and turning base leg to final approach. Most scenarios involve the autopilot being on and trimming all the way to the low airspeed. They often give students low speed scenarios when they're not expecting them, and make ready use of distraction. In my own case, I knew exactly what Paul was doing when he struck up the conversation about the ad with a naked woman in it, yet I still found myself surprised when the shaker went off. The end result is that if one of their pilots ever finds themselves surprised by a stick shaker at low altitude, it won't be the first time they've had that experience, and they'll have accurate motor memory to call upon for the recovery.

So why haven't more airlines changed the way they train stall recovery? Surprisingly, the FAA isn't standing in the way: they wholeheartedly approved of the changes that Paul introduced to his airline's training program. A lot of it is simply institutional inertia. Until now, few have thought there was a problem that needed addressing. This a symptom of a reactive rather than proactive safety culture at many regional airlines. Another element is cost: many regionals' training programs are all about turning out pilots as quickly and cheaply as possible while maintaining a basic level of competence and safety. When you compare regional airline training syllabi to those at major airlines, you typically see fewer simulator sessions despite having similarly sophisticated aircraft and less experienced pilots. That means that certain things get glossed over, and no "superfluous" training is included. This accident will of course change the way we teach stalls - I fully expect to be using Paul's method next year (and I hope they call it "Paul's method" in recognition of his foresight) - but I do worry that it will take future accidents to expose other weaknesses unless there is a fundamental change in the safety culture at the regional airlines. I'll write more about that in my next post.

22 comments:

VanMan said...

I'm not a pilot, but have traveled enough to amass at least four million frequent flier miles.

I find your analysis well written and very insightful.

Thanks!

Charlie32 said...

Superb post, that's all I can say, really well done.
It's nice to see that someone stops pointing fingers at the pilot and sheds some light on the Safety Culture in the background too.

Anonymous said...

Sam, excellent post. I fly Dash 300s in the Caribbean and have been following this crash quite closely. This is by far the most sensible article I've read in the past week or so. If you are not already in your company's training department, you should be!!!

David said...

"It is difficult to explain to those who have never flown airplanes with stick shakers just how jarring their activation is"...it does seem like the cockpit cacophony might have made connected thought difficult. Why did stick shakers get introduced, anyhow?...the high-pitched tone used for stall warning on GA aircraft is pretty hard to miss. And maybe warning horns like the autopilot disconnect should beep a couple of times and then be silent, while leaving a light on to indicate activation.

I wonder if any human-factors work has been done re the effect of warning-systems noise on thought and reaction times.

Bob Collins said...

Was there a degradation of airspeed prior to when the alarms went off? Is a stick shaker generally the first sign of trouble? Or the last?

Will said...

Great post. Let me guess, the regional in whose sim you tried out this scenario rhymes with "Mesaba"?

The truly scary thing about 3407 is, in some context, everything the crew did makes sense.

The captain, thinking he had a tailplane stall, reacted prudently, at least initially.

The FO following a cardinal rule of aviation, if you move something and bad things happen move it back, retracted the flaps she had just deployed.

Or not.

It's obvious, on the ground with a warm cup of coffee and months to think about it, that you wouldn't get a shaker if the tailplane were stalling.

But the captain never experienced a shaker in training and had mere seconds to react.

The lack of training with the shaker and pusher are the most startling revelation to me so far and far more troubling than the checkride failures because while the checkride failures deal with one pilot the lack of shaker experience extends across an entire airline.

Confusion it was then, and the crew were simply unable to untie all the knots they had tied in time.

We'll never know with certainty what they were thinking, sadly. But hopefully more airlines will modify their training procedures as a result.

Sean O. said...

Excellent post! You've reaffirmed some of my thoughts on stall training in real-life situations. I will be going into the sim later this summer for self imposed recurrent training (C-310R) and if this doesn't come up, I will introduce it during the additional issues segment.

Sam said...

Bob - there are warnings before the shaker, all a lot more subtle. There's a trend bar on the airspeed tape which shows the airspeed you'll be at in ten seconds (if I recall correctly), the red low speed cue rises as you get closer to it, and the airspeed turns yellow if the trend bar is at or below the low speed cue. Of course if your attention is turned completely from the PFD, none of these will be seen.

Will- like I said, after reviewing the flight data recorder I'm fairly unconvinced that the pull up really was a reaction to thinking "tailplane stall." It's definitely a possibility, but to my mind no stronger than the chance of initiating a stick shaker recovery the way he was taught in the sim but with a little overcorrection. Again, it wasn't that strong of a pull, 25-30 lbs of force. If he really was immediately thinking "holy crap, a tailplane stall!" like he was in the SF340, you'd expect that he would haul back on the yoke with a *lot* more force.

Since we're never know what he was thinking and since both scenarios are actually plausible (and could happen again), I'd propose that we both change the way we do stall training AND provide more explicit guidance on tailplane stall. IE:

"If you experience a stick shaker and the plane pitches down, it's a wing stall. If you experience a pitch down at an airspeed well above the low speed cue and there is no stick shaker, it's probably a tailplane stall."

That's actually really basic knowledge, but since this happened I've heard quite a number of pilots make really nonsensical statements about tailplane stalls, making me think that this knowledge isn't as widespread as one might think....

Will said...

Yeah Sam, I agree that it's unlikely he instantly thought "tailplane stall", only that viewed in that context it makes some sense. (Like most pilots, I'm troubled by someone botching something so simple as a stall recovery, which we've all been doing since about our third hour of primary training.)

What's so bothersome is from all indications he was reacting the way he'd been trained, perhaps not perfectly but initially in the ballpark, and the whole mess got away from him, which as you've pointed out so nicely may be an indication of a more systemic problem.

In times of stress we revert automatically to our training.

In a way, it's a similar situation to AA191, a DC-10 that lost an engine departing ORD which damaged the slats on the left side. The crew, per their training, decelerated to V2 instead of maintaining a higher airspeed, resulting in a stall. They did everything they were trained to do but the training itself was flawed.

After that crash, maintaining V2 or higher if you were already there and happy became procedure.

Perhaps we'll see a shift away from "maintain altitude" to "maintain altitude if possible, get out of the shaker/pusher regime right now, even if you gotta sacrifice altitude to do it" in training.

Reed said...

Sam,

Another great post. A much nicer place than some of the aviation forums where if you aren't a pilot and bring up technical discussion, you will be flamed.

It seems that all the "holes in the swiss cheese" were lined up in this accident. A breakdown in CRM caused a suprise shaker which was then followed by an improper recovery technique. If one of these events were caught earlier or didn't happen in the first place, the crash would most likely not have occured.

The main thing that interests me here is the suprise factor. In the sim, pilots are something to go wrong. They are expecting that V1 cut or the windshear on final. The threat of something going wrong is at the forefront of the mind, because thats what simulator training is mostly about. Dealing with abnormal/ emergency procedures. The crew of Colgan 3407was not expecting the shaker and got hit with it due to a breakdown in focus. It came as a complete suprise. Getting the shaker at the outer marker was probably the last thing on that crew's mind. There is always a higher probability for something to go wrong when it comes as a suprise, even if all the muscle memory is built in and all the training received. This new stall training you speak about in the post sounds great. It seems to mitigate the "suprise" factor as much as possible by showing pilots the shaker in all different situations.

Another point I would like to bring up is the possible difference in control forces between the Saab and the Q. The captain was fairly fresh off the 340 and maybe the control forces were a bit heavier on that aircraft. Being used to the Saab's controls and pulling back with more force than necessary could have been a factor, along with the obvious ones such as the trim situation and the sudden addition of power. The captain pulled back with more force than necessary in the first place, but I am wondering if this is partly due to familiarty with the 340. I'm not sure what the difference in control forces are between the two aircraft but am very curious to know. Take care!

Reed

Jason said...

Sam,

Excellent piece Sam. This tragedy is eerily similiar to an accident involving a Jetstream 41 back in 1994.

http://www.ntsb.gov/ntsb/brief.asp?ev_id=20001206X00616&key=1

The captain was new to his position and to the aircraft. He had failed prior checkrides and the first officer was fresh out of training.

I wonder why this tragedy did not get the same press coverage as 3407. Maybe some lives would have been saved if some of these issues were addressed 15 years ago.

JD

Anonymous said...

Sam,

A well-thought-out piece. Please contact me regarding it.

UAL78@ProPilotWorld.com

Sam said...

Reed - I've never flown the SF340 other than the few minutes in the sim for that stall training, so I can't really make a good comparison between the two airplanes. I will note, however, that the Q400 is known for being a truck. Control forces are quite high, particularly in the ailerons. Now that I think about it, the force required for roll is significantly higher than the force required for pitch, it's kind of weirdly decoupled that way.

Sam said...

Jason-

Five fatalities, like the two fatalities on PCL4701, gets a whole lot less press than 50. Unfortunately the old press maxim applies: if it bleeds, it leads.

I hadn't heard of that accident before. Interesting stuff, there are a lot of similarities.

Anonymous said...

Amazing post as usual, Sam. Your command of the material and its underlying theories is obvious. If your aviation career doesn't currently include work taking advantage of these skills, it should.

As a low-time VFR guy I followed pretty much everything except one detail: what's the difference between the stick "shaker" and "pusher"? The meaning is obviously self-evident but I can't determine a pattern to your seemingly interchangable use of them.

It's a shame this level of analysis will never make it near the mainstream media...

Regards,
Marty

Flying Europe said...

The Shaker is a warning message to the pilots about an impending stall. The aircraft knows it is becoming to slow and alerts the pilots by shaking the control column before it will enter a stall scenario!

The pusher is a mechanical device that physically moves the control column forward to lower the nose of the aircraft when it experiences a stall. It is actually designed for higher altitudes where you can trade altitude for airspeed, thus through the help of the pusher the airplane would get out of the stall even if the pilots would not touch the controls.

On some airplane types the stick pusher is even disabled automatically during appraoch to make sure it does not harm the airplane when it "fires" wrongly.

The pusher can be manually disabled by pushing a button or it can be overpowered by using increased force (this happened in the colgan crash).

To Sam: Very nicely written!

FE

Anonymous said...

My initial reaction, after a piddling 60 hours flight time and a zillion hours in FS9 and x-plane, was the same as everyone else: stall recovery is lesson one! If I was flying, even I could have saved 50 lives that night!

Your post explains one possible reaction. In stall recovery training, loss of altitude was always my secondary concern, and it would really bother me if my trainers insisted that it be a primary concern.

Another possibility, however, ties into your third option (training). Immediate speculation was that this was a tail stall, and that he augered in after performing a normal stall recovery, which apparently does not work in a tail stall situation. IIRC, he had recently received training on tail stall recovery.

Isn't it possible that his recent training had overridden his basic training, and that he misinterpreted the signs, in light of the moderate icing conditions, as indicative of a tail stall, and thus, in his mind, correctly overrode the stick shaker? And shouldn't there be a little bit better training for pilots in cold weather environs to be able to tell the difference between a basic wing stall and a tail stall?

What if the next pilot who experiences a tail stall reacts as if it is a normal stall after all of the press from this accident?

I don't know s*** about tail stalls in airliners, beyond the reports in the news and on blogs. Perhaps you can help us understand the difference, and how pilots can best determine which they are experiencing.

Sam said...

Anonymous 7:39--

Sure, it's possible that he misinterpreted the stick shaker as a sign of a tail stall. If so, it surely wasn't due to recent training on tail stalls: in a tail stall situation, you'd expect to see no shaker and a rapid pitch down. In this case it was the exact opposite, a stick shaker and no change at all in aircraft attitude (until he pulled up). So far as whether better training is required on tail stalls, in general I agree but for the Q400 I don't know. If the airplane isn't susceptible to them, why inject that element of uncertainty in the back of a pilot's mind?

Let's see, I've covered tail stalls briefly in previous posts but I'll give you the more complete version here. A tailplane has a critical angle of attack just like a wing does. Under normal circumstances it gets nowhere near critical angle of attack, because as fuselage angle of attack increases, this is actually a decrease in AoA for the tailplane (think upside-down wing). However! If the tailplane is in the wing's wake, the increased downflow at high wing AoA can also increase the tailplane AoA. Flap extension increases this downflow from the wing. If the tailplane is iced up - and being thinner than the wing, tails are good collectors of ice - then its critical AoA decreases. The most common scenario is an airplane on approach, well above wing-stalling speed, with an iced up tail. The crew extends the flaps, the increased downflow increases the tail's AoA beyond critical, and airflow detaches from the bottom of the surface. In the case of manually powered control surfaces, they can be snatched downward and the airplane pitches down due to the loss of tailplane lift counteracting the airplane's natural pitching-down moment. In order to recover, the pilot must unstall the tailplane by pulling back on the yoke...this increases the effective camber of the tailplane, making it easier for the airflow to reattach. However, doing so against aerodynamic forces takes a lot of effort in airplanes with manual elevator control (C172) or aerodynamic control tabs (DC9). The reason that Bombardier says the Q400 is not susceptible to tailplane stall is because its tail is well out of most of the wing's downflow (tall T-tail) and the elevators are hydraulically powered.

Clear as mud?

Michael Rey said...

Sam- Any chance we can chat?
I'm at CBS News in NY.
Rey@cbsnews.com

Thanks-
Michael

Anonymous said...

FYI, Horizon quit training stall recoveries to a base altitude a while ago. We now just set up via the instructor's commands, achieve shaker, and climb away and clean up through acceleration height, if applicable.

Sam said...

Michael - My airline has a fairly strict policy against talking to the media, and they have an army of lawyers at their disposal. Sorry.


Anonymous 9:35 - I'm glad to hear that, it's a step in the right direction. Overall QX had good training, but I wasn't very impressed with their stall training. It was only later that I realized that they were teaching the industry standard.

Flying Europe said...

Can you not talk to a journalist as a private person? You would be a great spokesperson for all pilots. I know the media wants crashes and flames but there's always the chance that they can inform the public reasonably and this might change things in the industry.

Oh..here's a wordcloud of your blog. http://www.wordle.net/gallery/wrdl/879062/Colgan_Wordcloud