Atlas Air 3591 NTSB Public Docket Opened

[4runner]

Originally Posted by svhar

I flew the 757/767 for 15 years, 2 years in the RHS and the rest in the LHS. I just cannot imagine a scenario where someone could accidentally hit the go-around switches when selecting flaps.

I don’t have that many, but I am current and I can imagine the scenario. The Captain reaches behind the thrust levers to select the flap lever. His watch hits the TOGA paddle without realizing it. It doesn’t become active until flaps indicate 1. We just tried it in the sim last week. I can totally see it happening in the heat of battle with an FO that you’re focused on.

[tdracer]

Did we figure out what instrument panel this freighter had? It was probably mentioned in the closed thread a few months ago but I can't seem to find it. Was it the legacy 767 panel with the round dial airspeed and Trinitron CRT screens? Or a flat screen mod done on the freighter conversion?

Airbubba, I'm only aware of one 767 flat panel upgrade - it's the one that FedEx is getting on it's new build 767F where the six CRTs are replaced with 3 large(r) flat panels - done by an STC. I have some familiarity with the STC since they started implementing that STC in production before I retired (initially they'd take a new 767 with the CRTs, fly it somewhere to remove the CRTs and install the flat panel systems - obviously it was cheaper to just install the STC flat panels during production). If it's the same flat panel STC as FedEx is getting, the logic as to what gets displayed and where is unchanged - there are some interfacing electronics that convert the display inputs as needed for the flat panel instead of the CRTs. But the exact same information is being displayed as with the original CRTs. The standby instrumentation is also unchanged.
Afraid I can't help with the detailed autothrottle logic during a go-around.

[Check Airman]

Originally Posted by Airbubba

The FO was apparently flying on autopilot and had called for flaps 1. There is a little pitch up on the '76 when you select flaps 1 (actually just the LE slats) but the autopilot should have handled that. But if those go-around switches or whatever they are called had been bumped I guess TOGA power would be selected when the flaps were 'out of the up position' (it's been a long time since I took the oral).

The '76 and the '75 are supposed to throttle back to a 2000 foot per minute climb as the nose comes up on the go-around. However, if you try to keep the nose down the plane thinks it needs more power and doesn't throttle back in my experience. There are subtleties and nuances on 757/767 autoflight that in some cases seem to be specific to the engine manufacturer. tdracer probably knows a lot about this.



Did we figure out what instrument panel this freighter had? It was probably mentioned in the closed thread a few months ago but I can't seem to find it. Was it the legacy 767 panel with the round dial airspeed and Trinitron CRT screens? Or a flat screen mod done on the freighter conversion?

From the CVR transcript:



Did the FO go to ALTN EFI? On the legacy panel it wouldn't affect his airspeed display it seems.

From the pictures in the report, it seemed the plane had the speed tape on the side of the ADI. I assume it’s representative.

[derjodel]

Originally Posted by MechEngr

It appears to be if one side control column jammed, say if a mechanic lost a wrench or a bolt worked loose. Not so much if the pilots disagree on how best to fly the plane.

thanks, under mechanical failure it might indeed make sense.

but when pilots disagree (not uncommon in recent crashes!), it reduces the chances of survival from 50% (stronger wins) to 0% (no one wins)

seems like a design flaw to me. Final report might be interesting.

[tcasblue]

Originally Posted by Airbubba

I agree, Atlas training gets good reviews from the folks I've worked with who have passed through there.

The interviews in the Accident Docket give a lot of insight to the hiring process and training. The NTSB does the customary 'we're here to find the cause, not to assign blame' opener. And 'that FAA guy is here to help you, he can't pull your ticket for anything you say during the interview'.

Atlas Director of Human Resources Denise Borrelli's interview starting on page 534 of the 734 page file is interesting as she is asked how the FO's numerous training failures are somehow missed in the PRIA check and the job interview process. Was it just an oversight?



Looking at the interviews, it appears that the plane had the flat screen panel upgrade with the round dial airspeed still there as well. So, even with the loss of the EADI you would presumably still have airspeed indication on the dial.

Here's a link to the interview transcripts, I find the online docket interface with the MS Word table of contents to be a little, well, wonky.

https://dms.ntsb.gov/public/63000-63...168/631157.pdf

Everybody be honest. Would you have let this person get hired if you knew his record?

If the NTSB is truly interested in aviation safety, they will not look the other way when it comes to investigating all possible reasons that this copilot was hired.

That includes affirmative action. Maybe that aspect of the investigation will show that it had nothing to do with him getting hired which would provide clarity but maybe it did get him hired. The families of the dead have a right to know along with the flying public.

[B2N2]

Originally Posted by 4runner

I don’t have that many, but I am current and I can imagine the scenario. The Captain reaches behind the thrust levers to select the flap lever. His watch hits the TOGA paddle without realizing it. It doesn’t become active until flaps indicate 1. We just tried it in the sim last week. I can totally see it happening in the heat of battle with an FO that you’re focused on.

Have you considered the FO triggering TOGA while stowing the speedbrake?

[misd-agin]

Originally Posted by slickcity

I have thousands of hours in both seats of the 757 and the 767. When the f/o is hand flying an approach and calls for flaps, the captain many times will reach behind the throttles and around to set the flaps. This avoids contacting the first officer’s hand while he/she is operating the throttles. But it also puts the captain’s wrist in close proximity to the go around switches. I can see this happening.

^^ x2. 10,000+ hrs on the 757/767. I can recall at least 3 incidents of inadvertently activating the G/A switches in 15,000+ hrs on Boeing's with G/A switches. 'Click, click' to deactivate the auto throttles is the rapid response.

It's typically a non event because the FO, with their hand on the throttles, feels an uncommanded forward movement and physically restrains the thrust levers while using the push buttons to disconnect the auto throttles.

[Airbubba]

Originally Posted by misd-agin

^^ x2. 10,000+ hrs on the 757/767. I can recall at least 3 incidents of inadvertently activating the G/A switches in 15,000+ hrs on Boeing's with G/A switches. 'Click, click' to deactivate the auto throttles is the rapid response.

It's typically a non event because the FO, with their hand on the throttles, feels an uncommanded forward movement and physically restrains the thrust levers while using the push buttons to disconnect the auto throttles.

And, in my observation, even if you leave the autothrottles engaged it's easy to overpower the automation and pull the power back manually. At 6000 feet the plane hand flies just fine so if you went up off altitude just pop off the autopilot and ease back down. Would selecting FLCH also bring you back to the altitude in the window? It probably would but there may be some inhibit above 2500 feet RA with flaps out in G/A mode on alternate Thursdays etc. So just pop off the automation and make the plane do what you want then 'Autopilot Center to Command' to recover from the inadvertent go-around switch activation would be my suggestion.

Some folks would have their hands on the throttles at 6000 feet with the autothrottles and autopilot on, others would not. I've seen it done and taught both ways.

In the docket file of interviews starting on page 305 an Atlas FO recalls a possible earlier automation glitch incident with the accident aircraft.

3 Q. Okay. Tell me what it is you want to share with us. You had
4 some information about the accident airplane.

5 A. Well, I don't know if it does. The -- I had an incident
6 coming out of Stockton on -- and I'm not actually sure that it was
7 that date. I think I believe it is. It very well could be --
8 Q. What date are you talking about?
9 A. 11/2 of '18.
10 Q. Got it.

11 A. And I had a incident coming out of Stockton on one of our
12 aircraft. It was -- we departed there in the middle of the night
13 it was 0730Z VFR and normal departure. It was heading off
14 departure with a climb -- a left climbing turn to 7,000 feet and
15 when we went through 1,000 we engaged the center autopilot and
16 somewhere before 2,000 we got a Flight Director command full down,
17 nose down. And the airplane tried to start leveling off and I
18 disengaged the autopilot, hand flew it, we reset the autopilot in
19 the climb out and it was all systems were normal after that.
20 And we did not do an FTR on it. We just took it as a anomaly
21 of the autopilot and it reset and we just continued on.

22 And I can't say if Laz, the guy I was flying with, remembers
23 the incident, you know, if we could verify that that was one
24 hundred percent it but when I was looking through my logbook I was
25 like, “Oh, that's -- that tail number was on that flight.” So it
1 just kind of concerned me.

2 Q. Okay. So when it nosed over did you happen to notice what
3 the flight motor enunciator was telling you at the time?
4 A. I do not.

5 Q. Okay. And was it a hard over or was it just trying to level
6 off and kind of go to a normal reduced pitch?
7 A. No, it felt like it was pushing over and I'm sure that the
8 command bars being all the way as far down as they could go.

9 Q. Okay. And when you reengaged --
10 A. I wouldn't say it was so abrupt -- it wasn't so abrupt like
11 it was trying to push us out of our seats or anything but it was
12 definitely heading into a pretty good dive. But -- and, you know,
13 we kicked the autopilot off, we kept flying the airplane and, you
14 know, corrected it but it was, you know, very noticed, very
15 pronounced.

16 Q. Okay. When you reengaged the autopilot do you reengage that
17 second or the center autopilot?
18 A. Yes.
19 Q. Okay.
20 A. We reengaged the center autopilot.

21 Q. And anything happen after that?
22 A. No. It was Flight Director was all normal, everything was
23 normal. It flew all the way to CVG with no problems.
24 Q. Okay. Was this written up in the logbook?
25 A. I do not recall.
1 Q. Okay. Any irregularity reports or anything like that filed
2 in conjunction with this event?
3 A. No.

4 Q. Okay. And how certain are you that this was 1217 Alpha?
5 A. I would say really only talking about 80 percent. If Laz
6 cannot verify that he was on that and that was the -- then I would
7 have to look elsewhere but it was -- I'm about 80 percent sure
8 that that's what it was.

9 Q. Okay. You said earlier it was pushing over toward a dive.
10 It never got into a dive or did it actually start to descend?
11 A. It started to descend. We went -- it -- level come off
12 immediately and it started to descend down and we probably got a
13 little bit vertical descent and immediately kicked off and
14 reinitiated the climb.

15 Q. Yeah. I'm trying to visualize this so you turned the
16 autopilot on during the climb then it started to level off and
17 then it continued past the level off into a descent before you
18 disconnected the autopilot. Did I get that right?
19 A. Yes.

In the preamble to this interview conducted over the phone the FAA disclaimer mentioned in my earlier post is given.

19 MR. LAWRENCE: Okay. Oh, by the way, FAA is in the room
20 here. They're a representative on our group but they're not here
21 for certificate action or anything. It's just like a safety --
22 this is just a safety investigation, okay. Just to let you know.

23 MR. ANDREWS: Right.

24 MR. LAWRENCE: Okay. Cool.

[Check Airman]

Originally Posted by Airbubba

And, in my observation, even if you leave the autothrottles engaged it's easy to overpower the automation and pull the power back manually. At 6000 feet the plane hand flies just fine so if you went up off altitude just pop off the autopilot and ease back down. Would selecting FLCH also bring you back to the altitude in the window? It probably would but there may be some inhibit above 2500 feet RA with flaps out in G/A mode on alternate Thursdays etc. So just pop off the automation and make the plane do what you want then 'Autopilot Center to Command' to recover from the inadvertent go-around switch activation would be my suggestion.

Some folks would have their hands on the throttles at 6000 feet with the autothrottles and autopilot on, others would not. I've seen it done and taught both ways.

In the docket file of interviews starting on page 305 an Atlas FO recalls a possible earlier automation glitch incident with the accident aircraft.



In the preamble to this interview conducted over the phone the FAA disclaimer mentioned in my earlier post is given.

Interesting find. That muddies the waters a lot. Also the fact that the AP was engaged the whole time. How does it stay engaged with all those stick forces? Does it not disengage after a certain level of force is applied?

[svhar]

G/A does not arm or engage by someone just looking at the switches or just by brushing them with their fingers or a watch, which is normally worn on the left arm, they need a firm touch. Even so, it should be a non event, autopilot off, autothrottle off and then back to normal.

[deltahotel]

Just press ALT HOLD

[4runner]

Originally Posted by derjodel

thanks, under mechanical failure it might indeed make sense.

but when pilots disagree (not uncommon in recent crashes!), it reduces the chances of survival from 50% (stronger wins) to 0% (no one wins)

seems like a design flaw to me. Final report might be interesting.

every transport category aircraft I’ve ever flown has some sort of elevator break away system installed. 4 different airplanes by 2 different manufacturers had this. A design flaw?????

[4runner]

Originally Posted by exeng

Whilst under line training as a brand new F/O on the 737 I went to disconnect the A/T during an approach. Engines spooled up and flight director looked a bit odd so I shut the thrust levers and switched off the flight director - many chortles from the line trainer in the LH seat. In fairness we were in marginal VMC at the time.

I never repeated that trick and having spent some time in the LH seat on the 757 it would seem quite odd to manage an inadvertent operation of the TOGA switch on the 75 or 76. Even if one did I cannot imagine why one would not recognise what had transpired and act accordingly.


kind regards
Exeng

ive done the EXACT same thing in the 73, during an actual approach. The difference is our ability. The FO in question displayed irrational reactions to non normal scenarios throughout his career. He saw the check board and mistook it for a stall and not max speed.

[derjodel]

Originally Posted by 4runner

every transport category aircraft I’ve ever flown has some sort of elevator break away system installed. 4 different airplanes by 2 different manufacturers had this. A design flaw?????

You did not address my reasoning at all, and simply stating the fact that other types use the same design does not refute the argument.

think about it. It’s guaranteed to cause a crash in case of yoke fight (pilot disagreement)

[Australopithecus]

Originally Posted by derjodel

You did not address my reasoning at all, and simply stating the fact that other types use the same design does not refute the argument.

think about it. It’s guaranteed to cause a crash in case of yoke fight (pilot disagreement)

It can happen in an Airbus too, btw. The first autopilot disconnect switch pressed and held overrides the opposite sidestick*. All of these features have to assume two rational, competent pilots. Its up to vetting, hiring and training to ensure that assumption remains valid

* Although, on edit, the Airbus architecture does allow for a subsequent dueling disconnect button fight..

[AlexGG]

derjodel, probably the reason for shear pins is that the probability of mechanical failure somewhere in the system is higher (or is considered higher) than the probability of pilots having a duel on control sticks.

[TyroleanCondor]

TO/GA switch vs. A/T DISCONNECT

Having flown the 767 and the 747 I wonder about the inconsistent architecture of the TO/GA switch vs. A/T Disconnect.
Pressing the switch which disconnects autothrottle on the 747 would activate the TO/GA mode on a 767!

[infrequentflyer789]

Originally Posted by AlexGG

derjodel, probably the reason for shear pins is that the probability of mechanical failure somewhere in the system is higher (or is considered higher) than the probability of pilots having a duel on control sticks.

More likely, I think, the reason is that the shear pins (or whatever) can (or may, with some probability) effectively mitigate a failure elsewhere.

There is no mechanical/automated system that can cope with the two pilots fighting each other on the controls. Either the pilot who is correct disables (or persuades to desist) the other one, in time to recover, or the ground will intervene and end it.

Two of anything is only for redundancy if one bails out (figuratively or literally), in case of disagreement two is the worst number to have - voting is inevitable tie. Three physical pilots isn't really going to fit in today's aircraft, but in the future we may be able to have a third synthetic pilot, whose job will be: in the event of disagreement, determine which is the errant pilot and remove them from the control path. Pilots or AOA vanes, take your pick - two is not enough to resolve a disagreement.

[derjodel]

Originally Posted by Australopithecus

It can happen in an Airbus too, btw. The first autopilot disconnect switch pressed and held overrides the opposite sidestick*. All of these features have to assume two rational, competent pilots. Its up to vetting, hiring and training to ensure that assumption remains valid

* Although, on edit, the Airbus architecture does allow for a subsequent dueling disconnect button fight..

"All of these features have to assume two rational, competent pilots."

I believe that is a problem. All designs assume "rational, competent pilots". But what is really a "rational, competent pilot"? Whenever something goes wrong, airplane makers are quick to blame the pilots. Presumably, because they didn't act "rationally and with competence". There are many examples, AF447, both MAC crashes etc.

However, I said it before and I'l say it again, there is no real measure of level of "rationality and competence". There is just a binary term, called "licensed pilot'. Since airplanes are meant to be flown by licensed pilots (whatever their level of competency is), either airplanes should be made such that any licensed pilot could safely fly them, or the license itself needs to be redefined and pass criteria should be severely stricter and should include reactions under stress (remember, the "rational" part). The question is, how does one test that without putting the candidate in real danger. Altough eliminating unsuitable candidates by darwin method would work quite well (and this forum is full of pilots so sure in themselves they would not mind to take such test), I'm not sure that's actually going to happen any time soon.

Going back to the design, the whole system is built around the premise of "2 rational, competent pilots". I believe it would be much more appropriate to require that the airplane must be able to fly safely with two parallel, independent systems in command. The fact that these systems are of biological nature should not play any role in the design.

Now just start thinking if the current cockpit design meets safety criteria. If it does, it should not matter if the "systems" in command are biological or not.

But the assumption that biological systems are flawless and the design can make shortcuts it would not with non-biologic systems is, IMO, a major flaw. Recently we've seen many crashes due to this design:
- AF447: biological system identified stall correctly, applied incorrect procedure, the other biological system was not able to monitor flying system's stick commands and was unable to correct. A perfectly flying airplane crashed as a result
- Germanwings: one biological system was temporarily unavailable due to maintenance, the other biological system malfunctioned, prevented access to the controls, and crashed the airplane
- Atlas air: biological system incorrectly identified acceleration due to (uncommanded?) TOGA as stall, and over applied stall recovery procedure; the other biological system correctly asses the situation and applied the right proceudre, which resulted in a yoke duel between the two piloting systems, however, the surface control system was not designed for such duels and broke off (a feature designed to prevent another type of problem), leading to each piloting system controlling only half of the elevator

Now read both of the above scenarios and think of the self-flying planes. Would you fly (as a passenger) on a plane which was designed in the same way, for two AI pilots? If not, why do you insist it's a good design for biological pilots? Try to view the situation without your inherent bias. Ask people who know nothing about airplanes (and don't depend on your salary) what they think of it...

[blind pew]

Sophisticated cottage industry for the unwary?

An interesting thread which is perhaps getting to the root of the problem which is a lack of understanding of how diverse human responses are to sensory overload.
My first introduction was with Tony Farrell doing a test of unusual attitudes under the hood on limited panel where he took control as I was about to tail slide the PA28.
https://www.hampshirechronicle.co.uk...bomber-command
Thats the man in question ex mosquitoes DFC and unlike many of my instructors he allowed students to take a flying machine to its limits. My senses were so overloaded that nearly 50 years on I haven't a clue as to our attitudes.
Shortly afterwards I had a chop test with Duff Mitchell who also had a DFC..only aircraft flying as weather was P poor..again under the hood and a practice GCA which I landed from whilst still wearing the head gear..cross wind and turbulence but it taught me to forget sensory perception.
Fast forward 25 years which included teaching limited aerobatics, a lot of spinning and mountain gliding including being knocked semi concious; I did an AFF and jumped out of an aircraft at 12,000ft..had no recollection of a falling sensation and found it frankly boring.
I was then asked to find out how a colleague had died; he had 25 years of airliners, a similar amount of hang gliders but like myself at that time around a year paragliding. He had a collapse at 200ft agl and was looking up at his wing when he hit the ground instead of throwing his reserve. A few years later I had a similar series of collapses and lost over 1,000ft in seconds whilst looking up at the wing. My senses told me I had lost around 100ft.
The next year I carried out a SIV course at Annecy with Flyeo and its boss Fabio; a small, quietly spoken young frenchman. He taught me more in two hours about how the muscles and senses act with acceleration than what I had learnt in my flying career.
From that I can understand how PF in AF 447 apparently kept back pressure on the stick.
Yesterday arrived the latest addition of Cross Country magazine which includes a study of paraglider pilots deploying their reserve parachute on a zip wire. Scientifically carried out with prior attempted sensory overload before the pilots were released. The technique was to delay their release whilst carrying out an oral reasoning task plus an "unusual" physical upset. The reserve deployments were delayed and sometimes completely erroneous.
I had one emergency smoke training in a DC9 mock up where I had to find a baby; not only a claustrophobic mask/ dense smoke but the sounds of a baby crying which disorientated me.
I failed miserably as I did in the hypoxia exercise carried out in the decompression chamber at Portsmouth.

The industry has bought two crew operation, advanced simulator use at the expense of real flying and packaged the product up with automatics and checklists..this accident won't be the last that is blamed on pilot error!

As to GA button, a mate who was the chief trainer on the 400 hit it when he went to take the auto throttle out on short final. His copilot said he caught three of the levers but with an outer at full chat the 74 did something he hadn't seen before. It was written up in an air safety report.