What the BEA didn’t say about Air France AF 447

05/08/2011 — 11 Comments

The BEA third interim report on the AF 447 accident raises questions

So I’ve read the BEA report from one end to the other and overall it’s a solid and creditable effort. The report will probably disappoint those who are looking for a smoking gun, once again we see a system accident in which the outcome is derived from a complex interaction of system, environment, circumstance and human behavior.

However I do consider that the conclusions, and therefore recommendations, are hasty and incomplete.

This post is part of the Airbus aircraft family and system safety thread.

My immediate concerns revolve around the failure of the BEA to consider fully the effect of interaction between the aircrew and automation as a causal factor. There are five key technicalareas which I believe are worthy of further consideration, these are:

Yes the aircrew’s behaviour showed a lack of experience and failure to follow procedure, but we should remember that they were the product of a certified flight safety management system. So the real question is would we expect a crew of equivalent training and experience to have behaved differently?

To that end I agree with the BEA that there was a less than optimal allocation and management of crew resources, however there is sufficient information that a much more comprehensive and insightful analysis of crew behaviour could be undertaken. In an earlier report the BEA reviewed the response of other air crew to icing events, yet in this report there is no mention, and perhaps the final report will knit these threads together…

Fundamentally this accident poses the question, why are human beings in the cockpit? If the conclusion is that pilots are there to deal with situations outside the predictions of the designers, and all that goes with such an assumption, then it is one that the BEA has resolutely failed to make.

11 responses to What the BEA didn’t say about Air France AF 447

  1. 

    Are you certain AutoTrim is enabled in all Alternate Law and not specific to Alternate “2B”? Is “2B” specific to UAS, a “special case” when Airspeeds are not reliable and hence StallWarn is “spurious”?

    • 
      Matthew Squair 09/01/2015 at 4:08 pm

      Good question, I’ll check and confirm.

    • 
      Matthew Squair 13/01/2015 at 4:31 pm

      For auto-trim:
      1. Alternate law 2B occured, due to the flight control computers rejecting all thee ADRs.
      2. The aircraft was not in ‘Abnormal attitudes’ law, as it could only then be entered due to inertial parameters (due to the prior rejection of ADRs) which didn’t happen.
      3. Therefore auto-trim was still active (in 2B law) which is what we saw.

      Stall warning is enabled for alternate law (including 2B) it uses the AoA probe data and airspeed to calcuate the stall warning. So even in 2B you’d get stall warnings, which are to be considered as valid ‘respected’ is the term Airbus use.

      Not understanding that auto-trim was running in the background while in alternate law was also a contributing factor to the los of an Air New Zealand A320 at Perpignan in 2008 a couple of months prior to AF447. What appears to be happening is that aircrew set and forgot ‘auto-trim’ in these two incidents. As to whether you should have auto trim working in alternate law? Well that’s an interesting question.

      • 

        Well. That is indeed an interesting question. Outside Normal Law, Stall is possible. By definition, when AutoTrim is cycling NU, it is “trimming into the Stall”. In trimming into a “target” Pitch, it is degrading response in the available reversal. Maneuverability is the goal, not “ride” or a “set” target Pitch, (AoA). In AutoTrimming, the Trimmable Horizontal Stabiliser is enabling a bias in other than neutral elevator (AoA). The goal when Stall is possible is equal availability of elevator, not a computer influenced target Pitch. The goal is not “ride” or “cruise” but maneuverability. IMO.

        As to Stall Warn. In the case of Unreliable Airspeed, reject of all three ADRs, how can Stall Warning be reliable, since Vs is computed with airspeed and AoA, not to mention Mach (temp dependent)? What can cause pitot malfunction generally comes with rapid changes in OAT…

      • 

        “Not understanding that AutoTrim was running….”

        Not relevant. Probably not accurate. AutoTrim is not “selected”. It can only be prevented by holding the TrimWheel (on the pedestal) manually. Once released, the AutoTrim function resumes. It is a part of the aircraft handling package, so pilots better know it is running….(they do). Approaching Stall, 447 was trimming Nose Up. Instead of flying elevators only, the aircraft was trimming into the Stall. Without AutoTrim, elevator (Pitch) authority would not have been so authoritative. Do you think AutoTrim in Alternate Law is a mistake? I do.

      • 
        Matthew Squair 13/01/2015 at 8:22 pm

        That’s correct, my poor choice of words. However autotrim is not ‘always’ on, for example it turns on above 100 ft and disengaged on flare, otherwise you’d be trimming into the flare. I think (if memory serves) that when autopilot is engaged the a/p does the trimming. The thing also to remember is that autotrim is common on a lot of modern aircraft. Friend of mine flues 747-400s they also have auto-trim function. There are other reasons why Airbus uses auto-trim that relate to their flight control loops, I’ll ser of i can dig them out.

        All that being said, if you asked the two crews what autotrim was doing on the day, what would be the response? Transparent cuts both ways…

  2. 

    Yes, in landing mode, then ground mode AutoTrim self selects out. But 447 was not in Normal Law, and the Autopilot had quit straightaway when IAS went south. Neither was the a/c landing or taxiing. In fact the a/c had begun a zoom climb of 7000 feet per minute, yet AutoTrim kept increasing Pitch authority Nose Up. That is not a desirable Automatic feature. AutoTrim cannot be selected out, it must be prevented by holding the trim wheel constantly, or it reverts to trimming out elevator. The THS went to All NU, and stayed there at near maximum of fourteen degrees for the duration of the Stalled descent. The Trim makes no noise, as in other aircraft, and other platforms allow switching off.

    In Direct Law, AutoTrim is not available, and in Mechanical, elevators are not operable, Pitch is controlled using manual Trim only. So the pilots had choices. I would assume since the aircraft was in Pitch “load demand” Law, that the AutoTrim remains available to help maintain close to one gee. It appears that is the design consideration here? So the aircraft was trying to keep a comfortable positive one gee as the pilots Stalled the aircraft? Did THS prolong the time to Stall Nose drop? Did the Nose drop at Stall? After Stall, the Pitch remained well up, and at very low forward velocity in the Stall, AoA reached as much as sixty degrees, yet the THS remained at maximum Nose Up Trim? AutoTrim in effect post Stall? Odd.

    Very inscrutable. There is no aural record of any of the three pilots even mentioning Trim setting after Loss of Control. UAS drill calls for Autopilots OFF, and Flight Directors also OFF. In Roll, ailerons were in Direct Law. So AL2B is a “blend”. Do you acknowledge that AutoTrim rolls in an authority bias? Is that desirable when flight controls need equitable agility in all axes?

  3. 

    AutoTrim in Alternate Law(s)?

    Well. That is indeed an interesting question. Outside Normal Law, Stall is possible. By definition, when AutoTrim is cycling NU, it is “trimming into the Stall”. In trimming into a “target” Pitch, it is degrading response in the available reversal. Maneuverability is the goal, not “ride” or a “set” target Pitch, (AoA). In AutoTrimming, the Trimmable Horizontal Stabiliser is enabling a bias in other than neutral elevator (AoA). The goal when Stall is possible is equal availability of elevator, not a computer influenced target Pitch. The goal is not “ride” or “cruise” but maneuverability. IMO.

    As to Stall Warn. In the case of Unreliable Airspeed, reject of all three ADRs, how can Stall Warning be reliable, since Vs is computed with airspeed and AoA, not to mention Mach (temp dependent)? What can cause pitot malfunction generally comes with rapid changes in OAT…

    So. If STALL Warning in AL2B is VALID (to be ‘respected’) how is it Nose Up authority (TRIM NU) is being enhanced automatically? Counter intuitive? Dangerous? If STALL warning is not valid (spurious) why is it activated? Poor programming?

    • 
      Matthew Squair 14/01/2015 at 12:13 pm

      How and why auto-trim came to be part of the automation actually goes back to the fundamental control laws, the entirety of which I can’t do justice to in a comment so if I get the chance I’ll write a post about it. A better (if that’s the word) example of the problem of auto-trim is the Perpignan accident where the nose-down elevator authority was inadequate to overcome the nose-up pitching moment of the auto-trimmed horizontal stabilizer which tragic results. Interestingly, the A380 cockpit has a dedicated THS status indicator right below the PFD to give aircrew visibility of auto-trim.

      • 

        A little more recent example of THS Pitch dominance is 447. Although with Perpignan the corollary is what killed the crew, eg, once the pull up was initiated, the existing THS NU acted in concert with elevators to arrest the dive. There is speculation that NU authority was available, but g PROT prevented the airframe from accepting the load it would have taken to recover. The aircraft struck the water at fourteen degrees Nose Down.

        Once 447’s THS hit maximum Nose Up, (fourteen degrees), it stayed there, all the way to impact. Down elevator was insufficient to budge the slab back off the stop. The Stall was unrecoverable. The 330 mushed all the way down, at descent rates approaching 15000fpm at times.

        This is my point about AutoTrim in alternate Law. With so much Nose up bias provided by the THS, the aircraft Stalled at a remarkably low airspeed, and the Nose did not drop sufficient to cue the crew that Stall had actually occurred. They were never aware the a/c had Stalled!! With the addition of the reversed Warning cues, they were completely in the dark re Stall. If there had been no AutoTrim (and AutoTrim is deadly in UAS recovery, ALT2B), the aircraft would have Stalled at a higher airspeed, there would have been more buffet, and the nose would have dropped briskly, as it should have. There would have been no mystery about the descent (“This is Stall, get the Nose Down”).

        I repeat: in Alternate Law, where Stall is a distinct possibility, and especially in ALT2B, where Stall Warn can be spurious, there is no upside to Trim, None. Only the danger that with augmented Nose Up authority, and any inattention, the aircraft will mush into a gentle but deadly unrecoverable descent. This is borne out by the BEA data, though they do not address it.

        Yes, 447 was without BACK UP Speed system, THS status cue, and no Angle of Attack indicator. Also lacking was any clear drill to address Unreliable Airspeed. Why would any of that be necessary in an Aircraft that “Cannot Stall?” Why indeed…

  4. 

    Hi

    Your paper has to do with what is unaddressed by BEA? To me, that means viewing the data and asking questions BEA did not ask? Fair cop?

    The descent after Stall has many characteristics that BEA reported, but did not comment upon.

    Flying pilot made mostly NU input with his sidestick.

    1. Why? Was he enamored of finding out what was at the bottom of the plunge?

    Assume for just a moment that he did have a reason for pulling back, and it had to do with behaviour of the aircraft that was worse than mushing toward the sea?

    What was the aircraft’s behaviour each time he pushed Nose Down? He was greeted with the STALL Warning, aural and cricket. Was there a change in a/c attitude? The video shows there was. Roll. What do you know of how a wide body behaves when airspeed is less than 60 knots? Right, I don’t know either. What happens when drag reduces in a powered mush? The only time I flew an aircraft in powered mush and dropped the nose, the plane tried to go on her back. That strikes me as more dangerous than mushing level toward the deck. Just food for thought.

    Did you know the 330 is certified to fly without standard Stall entry characteristic behaviour? It is. The reason for the waiver? Protections. Protections allow the nasty behaviour to be certified.
    Wait, in Alternate Law, are all the protections available? NO. One that goes away is Stall PROT. But isn’t Stall PROT what gives the airframe a pass on “no longitudinal stability” at Stall? By regulation? Hmmm……

    2. Let’s discuss number one.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s