Go Back  PPRuNe Forums > Flight Deck Forums > Tech Log
Reload this Page >

AF 447 Thread No. 6

Tech Log The very best in practical technical discussion on the web

AF 447 Thread No. 6

Old 14th Aug 2011, 16:55
  #21 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 77
Posts: 1,457
In other words, a design which requires roll inputs, is going to get some pitch inputs, like it or not, with a normal pilot - add in the surprise factor, turbulance, lack of high alt hand-flying training etc. etc. and this can all add to the amplitude of the pitch inputs, which seem to have been unrecognised by the PF, and the lack of recognition then caused the long duration of that NU input (which, integrated over time, drove the THS movement).
A key element of manual instrument flying is building a mental picture of what the instruments are telling you and then controlling the aircraft with that information. With a PFD, it is much easier than with steam gauges since the majority of data is in a narrow visual span.

My present conclusion regarding the PF's control problems is that his scan was broken and disfunctional. His problems were very likely compounded by an inappropriate grip of the stick from the start, i.e. palm instead of finger tips.

The core of a manual instrument scan is control of the nose attitude and roll attitude-just put the pipper above the horizon line by the appropriate amount (~3 degrees in this case) and level the wings. Then drag in peripheral data from the sides, altitude, heading, and airspeed. Make small corrections as necessary. In the case of AF447, since airspeed was not available, apply suitable power to ensure stable speed.
In just a few minutes of this tedious flying, they would have been well down the road and away from the weather-but this did not happen.
Altitude never entered the scan, nose attitude did not enter (or else an inappropriate response to earlier training did). All the PF's attention appears to have been centered on controlling roll attitude which he was badly overcontrolling.

I can see two approaches to preventing this type of accident.
1. Provide a wing leveler function in Alt 2 law so the PF doesn't have to touch the stick except to maneuver. (The lowest common denominator approach)
2. Emphasize basic manual instrument skills under worst case conditions during recurrent training e.g. flying the S-1 and S-3 basic instrument patterns (by hand of course) at altitude in Alt 2 law. (You can always ask to try flying this just to prove to yourself that you can still do it.)
I am hoping the regulators mandate the second approach.
Machinbird is offline  
Old 14th Aug 2011, 16:59
  #22 (permalink)  
 
Join Date: Aug 2011
Location: Near LHR
Age: 53
Posts: 37
Hi 3holelover:
Originally Posted by 3holelover View Post
I don't believe that to be the case. The stick's neutral position, in both axes, is fairly certain... it's rather easily maintained in that 'detent' - if you will - of one axis while being moved within the other.
Thank you for that, and I'm happy to be corrected. I can only speak from my GA (control column) piloting experience, where no such neutral "detent" or bias exists. Now you mention it, the SS "detent" you mention, sounds rather like some PC joysticks

Having said that, I still believe that even with such a "detent", in a high-stress situation, the effects of adrenaline could easily overpower the PF's fine muscle control which is needed to "respect" such a bias toards neutral pitch. Therefore the only way to ensure that such a neutral bias is respected by a pilot who is "high" on adrenaline, may be to enforce it with some kind of lock. I'm sure that would bring its own set of challenges, however...

-----

Hi Machinbird, Thanks for your thoughts.

Originally Posted by Machinbird View Post
A key element of manual instrument flying is building a mental picture of what the instruments are telling you and then controlling the aircraft with that information. With a PFD, it is much easier than with steam gauges since the majority of data is in a narrow visual span.

My present conclusion regarding the PF's control problems is that his scan was broken and disfunctional. His problems were very likely compounded by an inappropriate grip of the stick from the start, i.e. palm instead of finger tips.
I've done limited IFR training, but your comments certainly make sense to me.

The core of a manual instrument scan is control of the nose attitude and roll attitude-just put the pipper above the horizon line by the appropriate amount (~3 degrees in this case) and level the wings. Then drag in peripheral data from the sides, altitude, heading, and airspeed. Make small corrections as necessary.
Perhaps, in the heat of the situation, with immediate roll inputs being required, this seems to bring us back to the PF being "behind" the aircraft and its initial state (inc. attitude), so that his NU inputs changed that state before he could get a good scan going (if that ever happened).

I also think the subsequent changing of what instrument data was available and valid (especially airspeed and v/s) could easily have led to confusion over which data to trust & which to ignore. Of course with hindsight, we can see there was "only" incorrect airspeed & intermittent loss of v/s indication - but he didn't know that

I can see two approaches to preventing this type of accident.
1. Provide a wing leveler function in Alt 2 law so the PF doesn't have to touch the stick except to maneuver. (The lowest common denominator approach)
2. Emphasize basic manual instrument skills under worst case conditions during recurrent training e.g. flying the S-1 and S-3 basic instrument patterns (by hand of course) at altitude in Alt 2 law. (You can always ask to try flying this just to prove to yourself that you can still do it.)
I am hoping the regulators mandate the second approach.
Understood, and your approach (1) would have a similar result to my thoughts about a "pitch lock" i.e. the pilot would not have an opportunity to inadvertently change the pitch when controlling the roll, as he/she wouldn't need to control the roll either when the AP disconnected, with that "wing leveler".

My concern about approach (2) is that, while it is much preferable and I am all for pilot training for hand flying, it's different in the shock of an emergency. As someone else here said, a sim or training session where you know you're going to get emergencies thrown at you, or where you know you're not going to get them thrown at you, cause different human responses than a nighttime flight at FL350, with no warning that those emergencies are about to happen. Therefore can we really rely on approach (2), without also having some additional help (like approach (1))?

This type of situation where humans don't take over very well from a "monitoring" role, was discussed in a paper by Dr. Lisanne Bainbridge called "Ironies of Automation". Well worth a read, for those who haven't, IMHO. I'll try to find a link... This isn't specifically a piloting problem; it's a human problem and has also been seen at nuclear power stations etc.

Last edited by Diagnostic; 14th Aug 2011 at 17:23. Reason: Added reply to Machinbird
Diagnostic is offline  
Old 14th Aug 2011, 18:32
  #23 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 77
Posts: 1,457
Hi Diagnostic
One man's emergency is another man's "Oh the ding blab *@% autopilot just quit, guess I'll have to earn my living. Training is the difference. If those pilots had recent experience with the two mentioned instrument training patterns and could master them, then it would have been no big deal.
Machinbird is offline  
Old 14th Aug 2011, 19:19
  #25 (permalink)  
 
Join Date: Aug 2011
Location: Near LHR
Age: 53
Posts: 37
Hi Machinbird,

Thanks again and I do agree that training is the key! I'm just not sure we can rely on some (beancounter-driven) airlines to give pilots all the training (inc opportunities for hand flying) which they really need - hence my wondering about what additional automation help can/should be incorporated, to avoid overcontrolling in Alt law hand flying. Perhaps the answer is "none"; but perhaps further automation help (in addition to more training!) might be useful.

Me, cynical about airline training policies? Surely not!

-----

Hi safetypee,

Thanks for those links, including The Ironies of Automation paper which I mentioned.
Diagnostic is offline  
Old 14th Aug 2011, 20:34
  #26 (permalink)  
 
Join Date: Jun 2009
Location: florida
Age: 76
Posts: 1,059
automation and feedback

Thank you Safety, thank you. The Norman paper is appropriate and a necessary "read" for all here.

Before adding to 'bird's ideas, I must point out that the accident started when air data sensors went south.

So what do we do when that happens? What do we do when the A/P disconnects and autothrottle keeps power setting where it was?

Well, it depends upon the design of the flight control system (FCS), old fashioned or the new FBW systems. It also depends upon crew training and "attitude" of the crew.

As a pilot, I would prefer a more straightforward reversion sequence, similar to 'bird's #1.. No need to go straight to the "direct" law where electrons simply replace tubes, cables and hydraulic lines. Just revert to a very basic "control stick steering" (CSS) and replace insistent efforts by the FCS to "protect you". Provide warning and caution indications of mach, AoA, etc when available, but basically fly attitude that is available from embedded sensors in the FCS. In other words, you have a very capable autopilot type function and have time to assess the loss of air data while maintaining aircraft control. Retired, 'bird, Smilin' and others here have flown with CSS, and it is easy.

Inherent in a FBW system are embedded sensors we used to use for attitude and navigation/weapon delivery. When the confusers fail, these sensors also fail or are ignored, You are SOL at that point. Meanwhile, you should have body rate, acceleration and even attitude sensors embedded. e.g if. you don't depend upon the external navigation/autopilot sensors - inertial, GPS, doppler, pitot-static, AoA vanes/cones, AHRS, etc.

We don't need a "wing leveler", just a "hold the roll and pitch angle when control pressure/movement is relaxed". Throttle/power is manual. You can even have the gee command active and the roll rate command active when you move the stick, as they are inherent in the FBW system.

So reflecting upon Norman's outstanding discussion of feedback in automated systems, that's my story and I'm stickin' to it. He was too early for the AA regional ATR accident when the A/P was feeding in control deflection due to ice build up. When all went to hell in a handbasket, Sad sad, and there was also some distractions in the cockpit for ten or fifteen minutes before A/P disconnect.

Bottom line is there are too many "external" inputs to the FCS after the air data has failed, and the spurious stall warnings and such did not help the crew. With a straightforward reversion to a CSS type system, I think the crew would have handled the situatin better. I do not excuse the pilot's mysterious nose up commands for so long before the aircraft entered Chuck Yeager territory.
gums is offline  
Old 14th Aug 2011, 21:36
  #27 (permalink)  
 
Join Date: Aug 2011
Location: Near LHR
Age: 53
Posts: 37
Hi Gums,
Originally Posted by gums View Post
Bottom line is there are too many "external" inputs to the FCS after the air data has failed, and the spurious stall warnings and such did not help the crew.
Agreed, sir - as others here have said before, a stall warning which turns off even when the aircraft is still stalled, is likely difficult for them to understand. I wonder how many AB pilots here, were taught that the stall warning could be silent with the a/c stalled, before this crash...? Yes, I know the attitude & airspeed should have been clues, but if they're (correctly) ignoring some instruments due to them being wrong (airspeed) or missing (v/s), how do they know they can trust the attitude displays?

Returning to your point about too many (pilot) inputs after the air data failed: Is it reasonable to expect civilian pilots who (whether we like it or not) now spend much of their time monitoring during the cruise rather than hand-flying, to instantly handle a totally unplanned transition to hand-flying, without occasional excessive inputs due to shock / surprise / fear / etc.? Especially with inadequate training for that situation? (Of course we're now seeing some improved training, as mentioned in the BEA report.)

My understanding of your eloquent posts (and please correct me if I'm wrong), is that your military flying was very different, in that you were very rarely using any form of (even basic) AP - yes? If so, then you were always "caught-up" with the exact state of your aircraft (through your ss!), in a way that a civilan pilot monitoring the cruise is much less able to be, unless hand-flying. Or do you believe I'm wrong?

It seems to me that the transition from "normal" to "problem" (it's not even necessarily an emergency, as Machinbird kindly pointed out), occurs to military pilots while flying; but occurs to civilain pilots, if in the cruise, when they are not (actively) flying and hence is a bigger shock and needs more time to catch-up, just because we're human. That's the time when control inputs have a greater risk of being inappropriate, due to the shock factor and lack of preparedness for that transition.

I think I'm probably just repeating PJ2 and his comments about them needing to "don't do something, sit there" - the requirement for immediate roll inputs prevented those initial moments of "catch-up", however.

I'm here to learn, and become a better (GA) pilot, so I'm open to being told that I'm full of &^%$
Diagnostic is offline  
Old 14th Aug 2011, 21:38
  #28 (permalink)  
 
Join Date: Jun 2009
Location: NNW of Antipodes
Age: 76
Posts: 1,331
An "in context" adaption of an excerpt from Ironies of Automation, by Lisanne Bainbridge, Department of Psychology, University College London.
Manual control skills
Several studies have shown the difference between inexperienced and experienced pilots making a flight path change. The experienced pilot makes the minimum number of actions, and the aircraft's flight path moves smoothly and quickly to the new position, while with an inexperienced pilot it oscillates round the target value. Unfortunately, physical skills deteriorate when they are not used, particularly the refinements of gain and timing. This means that a formerly experienced pilot who has been monitoring an automated aircraft's flight parameters may now be an inexperienced one. If he/she takes over he/she may set the aircraft's trajectory into oscillation. He/she may have to wait for feedback, rather than controlling by open-loop, and it will be difficult for him/her to interpret whether the feedback shows that there is something wrong with the system or more simply that he/she has misjudged his/her control action. He/she will need to make actions to counteract this ineffective control, which will add to his/her work load. When manual takeover is needed there is likely to be something wrong with the automated process, so that unusual actions will be needed to control it, and one can argue that the pilot needs to be more rather than less skilled, and less rather than more (sensory) loaded, than average.
The above is just one of many factors mentioned in the above paper that will have had a bearing on the outcome of AF447. The inability of the PF to get immediate stable control of the aircraft was reflected in his actions throughout the final stages of the flight.

Once stalled, the phugoid nature of the aircraft's motion coupled with a similar rolling pattern appears to have resulted in a sensory blockage as to where the aircraft was in the flight envelope, let alone where it may have been - a direct result of factors mentioned above.

Physiological and psychological aspects of this accident will feature heavily in the final report.

As an example the "circadian rthymn" has previously been raised in the MAK/IAC Final Report into the crash of an Armavia A320 EK-32009 near Sochi Airport on 3 May 2006. An adaption of a graphic presented in that report is reproduced below and attempts to quantify the increase in error rates expected at the low point of the circadian cycle which is nominally 0300 LST.

The effects of time zone changes in relation to circadian time is difficult to determine, and in the above graphic the circadian time is shown as Local Solar Time. The social aspects of time zones that are quite removed from normal circadian time, e.g. Paris Local Solar Time and Central European Summer Time may have a modifying effect. Hence the showing of the Paris LST.

P.S. Thanks for the link safetypee.
mm43 is offline  
Old 14th Aug 2011, 21:48
  #29 (permalink)  
 
Join Date: Aug 2009
Location: Germany
Age: 62
Posts: 1,809
Cool

Hi,

Will be interesting to have a circadian rhythm study about this accident:
http://www.airdisaster.com/reports/ntsb/AAR75-13.pdf
See page 17 .....
I doubt the circadian rhythm study will be useful ...
jcjeant is offline  
Old 14th Aug 2011, 22:49
  #30 (permalink)  
 
Join Date: Aug 2007
Location: London, New York, Paris, Moscow.
Posts: 3,633
What on earth are you on about.
glad rag is offline  
Old 14th Aug 2011, 23:15
  #31 (permalink)  
 
Join Date: Jan 2008
Location: Herts, UK
Posts: 729
Originally Posted by Northwest Airlines 1974 727 Accident
Because the use of attitude references is a fundamental of instrument flying which is stressed in Northwest's flightcrew training program, the Safety Board concludes that the flightcrew improperly relied on airspeed indications as a means of determining aircraft performance.
Back to the importance of a scan including attitude rather than airspeed when in any doubt of a/s.
A clear AoA display is another thing - some have suggested here that it is probably yet to be accepted as primary display data

Last edited by HarryMann; 14th Aug 2011 at 23:27.
HarryMann is offline  
Old 14th Aug 2011, 23:52
  #32 (permalink)  
 
Join Date: Aug 2009
Location: Germany
Age: 62
Posts: 1,809
Cool

Hi,

What on earth are you on about.
Check page 17
http://www.airdisaster.com/reports/ntsb/AAR75-13.pdf
the pilots actions and behavior and then .. think about the pilots actions and behavior of the AF447
Do you not see some similarities other than "circadian rhythm" ???
jcjeant is offline  
Old 14th Aug 2011, 23:58
  #33 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 77
Posts: 1,457
A clear AoA display is another thing - some have suggested here that it is probably yet to be accepted as primary display data
That is easily fixed in a few simulator sessions. Take away the airspeeds and then fly a few approaches.

Miracle of miracles, it works!

Properly set up to adjust for gross weight and configuration in the display, you won't even have to remember any numbers.
Machinbird is offline  
Old 15th Aug 2011, 00:13
  #34 (permalink)  
 
Join Date: Jun 2009
Location: florida
Age: 76
Posts: 1,059
Use of autopilot

Thanks for nice words, Diagnostic. All of us can improve our piloting skills, regardless of the thousands of hours and hairy moments.

My understanding of your eloquent posts (and please correct me if I'm wrong), is that your military flying was very different, in that you were very rarely using any form of (even basic) AP - yes? If so, then you were always "caught-up" with the exact state of your aircraft (through your ss!), in a way that a civilan pilot monitoring the cruise is much less able to be, unless hand-flying. Or do you believe I'm wrong?
I flew mostly single-seat planes. Had about 400 hours in a true two-seat interceptor. That plane had attitude hold, heading hold/select, mach hold, coupled A/P to the steering for attacks/missile launches ( and I used it once when firing an actual A2A roket - the Genie). Also could couple the A/P to the ILS for instrument approaches and "monitor", as many airline folks seem to do. My later jets had decent A/P functions and being the only soul onboard, I used it a lot when in IFR and had to switch approach plates or just get a grip on the situation. Of all, the Viper was the worse. Our FCS engineers resisted external inputs to the system. And to make matters worse, we had no warnings that the FCS was not accepting any further AoA commands for "altitude hold". So I lost a buddy over the glass surface of the Salt Lake as his jet gradually descended while he fooled with IFF codes and frequencies and such. The Airbus is the opposite philosophy of the Viper engineers, although I always thot we could have had a much better A/P.

You are correct, Diag, that when not relaxing and letting otto do the flying that we were much more aware of EXACTLY what our planes were doing. Nature of the business, and I would not expect the commercial pilots to fly at the edges of the envelope just to get from point A to point B. Just think about flying 13 or 14 hours over the ocean from the U.S. to Europe or Hawaii in your chair right now. No restroom or help from someone.. Did we use otto? You bet.
gums is offline  
Old 15th Aug 2011, 01:27
  #35 (permalink)  
 
Join Date: Oct 2007
Location: France
Age: 68
Posts: 136
Question to pilots actually flying "jets": by day time, do one see the contrails of planes that are at the same level (so they would be seen "in the thickness direction")? If so, do it happens to fly "inside" them? If so, do it follow a particular smell of the air?

Questions to the "physicists"
- What can be the sectional area of contrails (I mean "primary" ones, not when they develop gradually when the conditions are right)?
- If the conditions are "favorable", could the flue gas of a plane which was "recently" in the same place (in the air mass, for failing to consider the wind) cause ice crystals (or conditions which could generate them) which may "ice" Pitots?

Well, questions may be stupid and irrelevant, but I was wondering if the increase in the occurrence of "Pitot events" could not be due to increased navigation accuracy, so that the planes on defined routes are all flying so exactly in the same place, that they may fly through the contrails of others if the wind is near front or back?
Shadoko is offline  
Old 15th Aug 2011, 02:01
  #36 (permalink)  
 
Join Date: Aug 2009
Location: Germany
Age: 62
Posts: 1,809
Cool

Hi,

Question to pilots actually flying "jets": by day time, do one see the contrails of planes that are at the same level (so they would be seen "in the thickness direction")? If so, do it happens to fly "inside" them? If so, do it follow a particular smell of the air?
A trick used by german fighters attacking US bombers formations by the rear .. was to hide them self in the contrails of the B17 or B24 for approach at fire distance ..
I don't remember in any german pilot records I read the report of particular smell of the contrails or pitot icing (at a average altitude of 7000 meters)
jcjeant is offline  
Old 15th Aug 2011, 10:39
  #37 (permalink)  
 
Join Date: Aug 2007
Location: London, New York, Paris, Moscow.
Posts: 3,633
@ jcjeant

You wrote in your post

I doubt the circadian rhythm study will be useful ...
glad rag is offline  
Old 15th Aug 2011, 11:08
  #38 (permalink)  
 
Join Date: Jun 2000
Location: last time I looked I was still here.
Posts: 4,513
I've not read the whole thread, or report yet. However, there are many references here to lack of airspeed. Surely there was a ground speed readout. Thus there was not a total loss of speed indication. That info, coupled to basic power/attitude V/S could give some idea of what was going on. Easy to say from the arm chair, but I have tried it in the sim on TQ courses. OK, an approach is different to high level flying. Headwind/tail wind etc., but ground speed would change with attitude. Block the pitot tubes and fly an approach on groundspeed. It really emphasised that power/att V/S works. If they were confused by high nose no stall warning, low nose stall warning, then pause, consider and look for other clues. What does the ground speed do with change of attitude? That can at least confirm that the ASI is giving out B.S. At high level there is a lot of thinking time. Remember the BA 747 that had all 4 shutdown in volcanic ash over SE Asia. They had thinking time.
A scenario with loss of a primary parameter such as airspeed should be a recurrent training exercise. It has happened on enough occasions, resulting in a crash, to be included. In all cases the a/c was flying and controllable; either blocked pitot or staic vents. Sadly, many of the 3 year recuurency items are tick boxes. They are not related to real life incidents/accidents. If knowledge of these is not passed on and trained how will we learn and not prevent their re-occurence? After Hudson I know many airlines suddenly included loss of thrust and ditching. It was fun and ticked many boxes. The same after BA's B777 glider, but not Air Trans A330 glider. That would take too long in the sim. 30mins at least, but you would tick many boxes; flight controls, alt flap, alt gear, aircon & press, panic management, and it would be a great CRM exercise. Why was it not done before? Do we have to wait for another high profile crash to consider more meaningful recurrent training? With commercial pressure on crew resources training is diluted to minimums. The downward spiral is tightening.
RAT 5 is offline  
Old 15th Aug 2011, 11:25
  #39 (permalink)  
 
Join Date: Jan 2008
Location: Herts, UK
Posts: 729
Remember the BA 747 that had all 4 shutdown in volcanic ash over SE Asia.
Indeed, but they had a Flight Engineer, remained disciplined, at post, and focused on trying and re-trying everything possible...
Flight Engineer worked tirelessly at re-starting engines, and eventually succeeded... whilst pilots flew, navigated, communicated and Captain thought things through strategically even, as well as tactically.
Even retained flexibility to change those plans at short notice, when engines started failing a second time...
Would in todays cockpit environs, IMHO, to be possible to lose focus and subsequently the aircraft, but then I am not flight crew, so have no right really to pontificate

Last edited by HarryMann; 15th Aug 2011 at 11:42.
HarryMann is offline  
Old 15th Aug 2011, 12:42
  #40 (permalink)  
 
Join Date: Jun 2009
Location: I am where I am and that's all where I am.
Posts: 660
To amplify just a tad on 3holelover's comment, there is a dead band around neutral position. That dead band makes all accidental inputs less likely. That's
the way I've programmed joystick based camera aiming software in the past. (The available joysticks were incredibly "noisy" around neutral position. So I had to put in a dead band.)
JD-EE is offline  

Thread Tools
Search this Thread

Contact Us Archive Advertising Cookie Policy Privacy Statement Terms of Service

Copyright 2018 MH Sub I, LLC dba Internet Brands. All rights reserved. Use of this site indicates your consent to the Terms of Use.