AF447 Vanity Fair article

[Autothrust Blue]

While we're on the subject, why do we do v1 cut after v1 cut in the sim, but never have anyone try to vector us in to a mountain or some creeping systems failure that doesn't have a four step "solution" in The Book?

I get a CFIT escape maneuver just about every time I go to the box, actually - unanticipated and "no time."

 

[PhilosopherPilot]

I'd be more interested in a statistical analysis of TT vs. flaming holes in the ground (or, I suppose, water). To borrow a tired old trope from the "SATs are Racist" crowd..."Testing is great at testing how well you take tests...everything is at the very least open to argument".

Well, he wasn't talking about failed check rides. He was talking about extra lessons due to training difficulty (as well as check rides). A failed (or passed) check ride is one thing, but if someone struggles throughout the training program that's indicative of something other than just how well they test or not.

 

[Boris Badenov]

I get a CFIT escape maneuver just about every time I go to the box, actually - unanticipated and "no time."

You're ahead of us, then. Nevertheless, I rather suspect that we'd all benefit from a lot more "something is wrong, figure it out" and a lot less "YOU WERE .03 MILES OFF CENTERLINE, BUT GOOD JOB LET'S DO IT AGAIN".

 

[BobDDuck]

While we're on the subject, why do we do v1 cut after v1 cut in the sim, but never have anyone try to vector us in to a mountain or some creeping systems failure that doesn't have a four step "solution" in The Book?

This is an issue I have with my current shop's training (and also the last place I was at). Every LOFT/LOE I've ever done has exactly one correct solution to it and everything that happens during the scenario pushes you towards that one solution. There is no thinking required. There is simply a "good solution" and "fail". I'd really like to see a scenario evolve the requires critical thinking and good use of CRM to arrive at a "good solution" but still allow for completion but with a "less good" solution that could be debriefed.

 

[Boris Badenov]

Well, he wasn't talking about failed check rides. He was talking about extra lessons due to training difficulty (as well as check rides). A failed (or passed) check ride is one thing, but if someone struggles throughout the training program that's indicative of something other than just how well they test or not.

Hmm. At first blush, I'd argue the opposite. That is, however they learned, they reached whatever goals were set (whether those goals are rational or useful is another conversation). But maybe I don't know what you mean by "struggling throughout the training program". Struggling how?

 

[PhilosopherPilot]

While we're on the subject, why do we do v1 cut after v1 cut in the sim, but never have anyone try to vector us in to a mountain or some creeping systems failure that doesn't have a four step "solution" in The Book?

In AQP you can develop a training program however you see fit, so long as you can demonstrate through the data that it's appropriate. In theory, if V1 cuts are never screwed up during recurrent you could make the argument that they should be removed from the recurrent syllabus. But since they are A, rarely done in real life and B, very risky, I'll bet no one will ever try to remove them.

The philosophy is "train the rare, check the common." So we train V1 cuts to proficiency, and expect crews to fly ILS approaches without much training.

 

[jynxyjoe]

Well, he wasn't talking about failed check rides. He was talking about extra lessons due to training difficulty (as well as check rides). A failed (or passed) check ride is one thing, but if someone struggles throughout the training program that's indicative of something other than just how well they test or not.

Or with new training regimes you're finding a way to combat complacency. If you're training is good, and combats complacency, I would expect high time guys to get the retest more often than the low time guys.

 

[Boris Badenov]

I don't know enough about the Airbus or this example in particular to say with any certainty at all, but didn't the airplane recover normal flight data prior to the impact? I'm wondering if the FO had just let go of the stick and turned on the autopilot would the plane have recovered to level flight?

I don't know enough about the airplane to comment intelligently, either. My suspicion is that by the time they got airspeed data back, the airplane was far enough out of tolerances that the autopilot would have said "No, thanks!" and kicked right back off. But I don't know. In any case, I'd be happy if "turn stuff off" was the sort of "go-to" option, followed by "turn stuff back on" once you have an idea of WTF is happening.

 

[PhilosopherPilot]

Hmm. At first blush, I'd argue the opposite. That is, however they learned, they reached whatever goals were set (whether those goals are rational or useful is another conversation). But maybe I don't know what you mean by "struggling throughout the training program". Struggling how?

I'm not certain what you're getting at. If I have a crew for an entire phase of training, say the "procedures" phase, they may very well meet the minimum standard in term of "jumped through the appropriate hoops" but still fail a lesson because they just aren't ready to progress due to overall situational awareness or use of TEM application. Typically the two are closely correlated. I've yet to have a few who can meet all of the standards with flying colors, but doesn't show the required skill to proceed.

The trainee has to be able to fly to the standard expected of that lesson, plus demonstrate proficiency in all tasks trained up to that point. If someone is "struggling" they might be able to meet the letter of the standard for today because that's what they've worked on the night before, but they lose the previous days objectives. Or, maybe they can fly the hell out of an approach, but the second a curveball comes (like an automation failure or incorrect input they weren't expecting) they fall to pieces.

It's not black and white, unfortunately.

 

[Boris Badenov]

Or, maybe they can fly the hell out of an approach, but the second a curveball comes (like an automation failure or incorrect input they weren't expecting) they fall to pieces.

Well, and here we get to the question of whether what's being checked is what ought to be checked. It seems to me that multiple, cascading failures and/or "it can't happen" failures are what get people dead...well, that and simple complacency or, yes, I'll go there, in some cases, incompetence. I certainly agree with you that none of it is black and white, anyway. Wait, is that racist?

 

[PhilosopherPilot]

This is an issue I have with my current shop's training (and also the last place I was at). Every LOFT/LOE I've ever done has exactly one correct solution to it and everything that happens during the scenario pushes you towards that one solution. There is no thinking required. There is simply a "good solution" and "fail". I'd really like to see a scenario evolve the requires critical thinking and good use of CRM to arrive at a "good solution" but still allow for completion but with a "less good" solution that could be debriefed.

Our training is like your latter example. A crew doesn't have to solve the problem "as expected" to pass an LOE. As long as they get their without an incident, accident or violation (and it wasn't just dumb luck that they made it), they likely will pass. There may be a solid debrief, but if they arrive without bending metal it's likely okay.

We "drive" the outcomes to get the crew where we want them to go. For instance if a crew tried to divert to a city we don't have planned in the LOE, "ATC" gives them a reason to go somewhere else until they pick the right airport. How they get there is up to them.

 

[jynxyjoe]

I don't know enough about the airplane to comment intelligently, either. My suspicion is that by the time they got airspeed data back, the airplane was far enough out of tolerances that the autopilot would have said "No, thanks!" and kicked right back off. But I don't know. In any case, I'd be happy if "turn stuff off" was the sort of "go-to" option, followed by "turn stuff back on" once you have an idea of WTF is happening.

@SmoothLanderJ might be a good one to talk about this too, because he and I have discussed this when I went from dirty Turboprops (men's airplanes) to shiny jets (princess airplanes). He said there's two ways to fix a problem (probably more) when it comes to young vs: old. Option one: turn everything off and make the airplane do what you want, then re-engage. Option two: finger bang the FMS until you trick the airplane into doing what you want to do.

Now I admit, 4000 hours and four years of college with plenty of knowledge, I was dead set that I was going to "turn the damn thing off" and make it do what I want because I have chest hair dammit. Somewhere about the 4500 hour mark, when a captain I respected showed me that you weren't manly for shutting off the autopilot, you were just choosing not to learn the tool the engineers made for you, I started making myself learn to trick the computer into doing everything I'd normally hand fly. I really want to learn both ways, but I don't know how much more I can learn from hand flying.

If I'm ever feeling froggy I just shut off the auto at 15k and hand fly it in, as long as the captain is game for it. I've really changed my mind into using that automation more but being able to ANTICIPATE everything it does rather than blinking, cocking my head, and saying "What the hell is it doing now?!?" followed by shutting everything off and manhandling it like a bad puppy.

Getting better with the new system, not there yet.

 

[PhilosopherPilot]

Well, and here we get to the question of whether what's being checked is what ought to be checked. It seems to me that multiple, cascading failures and/or "it can't happen" failures are what get people dead...well, that and simple complacency or, yes, I'll go there, in some cases, incompetence. I certainly agree with you that none of it is black and white, anyway. Wait, is that racist?

When was the last accident that was a result of a cascading multiple failure? When was the last accident due to complacency and incompetence?

Answer those questions and you'll see why we focus more on the latter than the former. The former are exceedingly rare, and difficult to prepare for. In fact, I fear that if we trained people to go "full cowboy" and start flipping switches off script it would increase risk, not decrease it.

 

[Boris Badenov]

Somewhere about the 4500 hour mark, when a captain I respected showed me that you weren't manly for shutting off the autopilot, you were just choosing not to learn the tool the engineers made for you,

It might surprise you (or someone, anyway) that I agree with this. I've learned a lot about the FMSes on my Princess Airplanes by keeping my damned thumb off the red "nuke the automation from orbit" button and trying to figure out why it's doing what I don't want it to.

But not when I was plummeting out of the sky at ~10000ft/min. That seems more like the time to, you know, put the old cutlass between the teeth, grab the yoke, and go Neanderthal.

 

[typhoonpilot]

This is an issue I have with my current shop's training (and also the last place I was at). Every LOFT/LOE I've ever done has exactly one correct solution to it and everything that happens during the scenario pushes you towards that one solution. There is no thinking required. There is simply a "good solution" and "fail". I'd really like to see a scenario evolve the requires critical thinking and good use of CRM to arrive at a "good solution" but still allow for completion but with a "less good" solution that could be debriefed.

A well crafted LOFT/LOE scenario has 3 to 4 possible solutions. One is the best; one is okay but not necessarily the best; one is pretty bad; and one is just stupid. The key is having 3 to 4 events throughout the LOFT/LOE. One event in the pre-flight ( MEL, passenger issue, etc ); One minor problem once airborne ( insignificant to medium level caution message ); and then one major issue ( preferably not something that is "land at the nearest suitable airport" because that doesn't really require much thinking ). Then have a choice of airports with varying approaches and varying weather conditions. Notams in the weather package given at the briefing stage need to be remembered, specifically ILS outages and runway closures, although those are still available on paper in the flight deck for review. It's best if the MEL or the minor master caution have some performance impact for landing.

Each one of the 3 events ( Pre-flight, minor airborne, and major airborne ) should have 6 possible choices for the instructor so that pilots going through training do not know what to expect. That should all be scripted so rogue instructors can't go off the reservation, so to speak. The events also need to fit into the systems that are required to be hit during that phase of AQP.

Takes a good team a few months to develop and test each phase, but if the airline is willing to devote the resources then the pilots get some real benefit from the lessons.


Typhoonpilot

 

[Boris Badenov]

When was the last accident that was a result of a cascading multiple failure?

Well, of course, they're not mutually exclusive. In fact, I'd imagine they most often come together. The accident we're talking about seems like one to me. Yeah, only one set of sensors failed, but then that failed other systems, which failed other systems, which is where the complacency/incompetence saw its chance...

 

[typhoonpilot]

We actually has one phase that ended with holding going into DXB, a fairly common occurrence. Weather was bad and the fuel was set so there were limited options. It helped the pilots think about possible choices if such a scenario was ever encountered. Well guess, what? Right after the phase ended there was one night when it was exactly that scenario for real. There ended up being something like 6 pans and 3 maydays for fuel because it got so bad, but overall the pilots did an excellent job at managing the situation. Mostly because they'd just seen a very similar scenario in their recurrent training.


TP

 

[PhilosopherPilot]

Well, of course, they're not mutually exclusive. In fact, I'd imagine they most often come together. The accident we're talking about seems like one to me. Yeah, only one set of sensors failed, but then that failed other systems, which failed other systems, which is where the complacency/incompetence saw its chance...

I hear ya, but in this case if the FO had just reverted to "pitch and power" all would've been fine. The number one rule is to fly the damn airplane. If your airspeed is incorrect, pitch to a reasonable attitude and add/reduce thrust as required. He had good motors. He had good wings and flight controls.

I guess I just don't see this as a cascading failure that should've caused an accident.

 

[jynxyjoe]

It might surprise you (or someone, anyway) that I agree with this. I've learned a lot about the FMSes on my Princess Airplanes by keeping my damned thumb off the red "nuke the automation from orbit" button and trying to figure out why it's doing what I don't want it to.

But not when I was plummeting out of the sky at ~10000ft/min. That seems more like the time to, you know, put the old cutlass between the teeth, grab the yoke, and go Neanderthal.

Oh God. Now I know @seagull said the narrative was off and I'm guilty of not looking into the NTSB report yet, so keep that in mind. The narrative of that stupid accident where (if I understood the sequence correctly) he loses the flipping A/S indicator and just goes ape on the yoke made me cringe for 3 pages. SCAN YOU BASTARD SCAN!

To me this seems like a very strange situation for an american pilot to judge, because the stick and rudder and BAI foundation just isn't there. There's no scan, the FO's are completely saturated and no one can fly this A330 with a missing A/S indicator. The Captain couldn't be up there in time to correctly judge the situation. So I'm approaching this a little superficially and from the foundation level, because I can't get my head around the two Air France Baby's decision making.

I hear ya, but in this case if the FO had just reverted to "pitch and power" all would've been fine.

Amen.

 

[jynxyjoe]

Well, of course, they're not mutually exclusive. In fact, I'd imagine they most often come together. The accident we're talking about seems like one to me. Yeah, only one set of sensors failed, but then that failed other systems, which failed other systems, which is where the complacency/incompetence saw its chance...

I'd also mention the pilots responded to a single failure and caused the cascading failure. Sometimes in a sim during a brief an instructor will tell you, "Hey I'm doing single failures... if you find yourself in a multiple failure situation you've probably done it to yourself. Take a breath, retrace, and fix it."