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

AF 447 Thread No. 7

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

AF 447 Thread No. 7

Closed Thread

Old 3rd Mar 2012, 01:43
  #781 (permalink)  
 
Join Date: Jun 2009
Location: somewhere
Posts: 451
@RR_NDB:


IMO this configures a SERIOUS and PERSISTENT "DESIGN FLAW" or even,
ERROR
Let's assume there is another - more reliable - system available.
wouldn't it be a great idea to have 3 or more of them iso 2 or 1 less reliable, just because they should not be identical?
They are still operating in 1 and the same environment.

Don't want to suggest the probe was reliable but there have been several modifications and upgrade options available before and since.

I agree with PJ2, just sit and wait (30 sec.) is proven to be the best strategy in case of UAS.
A33Zab is offline  
Old 3rd Mar 2012, 02:20
  #782 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
Originally Posted by A33Zab
I agree with PJ2, just sit and wait (30 sec.) is proven to be the best strategy in case of UAS.
Even better would be no UAS at all. It probably is achievable.
Machinbird is offline  
Old 3rd Mar 2012, 11:54
  #783 (permalink)  
 
Join Date: Feb 2011
Location: Nearby SBBR and SDAM
Posts: 868
Hysteresis

Hi,

A33Zab

PF applied large stimuli to the plane.

I agree with PJ2, just sit and wait (30 sec.) is proven to be the best strategy in case of UAS.


We may say this "technique" is an "hysteresis like" one. No reaction (in the loop) until a given threshold. In absence of other solution I AGREE this is WISE.

Even better would be no UAS at all. It probably is achievable.


As a researcher i have to remove the "probably" in the phrase.

And i am not not focusing on the probes. I look to the System.

The whole "processing" of the probes data IMO should be reviewed.

As i understand the a/c operation degrades EXACTLY when you need most.

Who knows what PF saw in the transitory, just after AP quit?

But this is to be covered in another post i am preparing:

Transient in Feedback Systems (II)
RR_NDB is offline  
Old 3rd Mar 2012, 19:25
  #784 (permalink)  
 
Join Date: Jun 2011
Location: france
Posts: 756
Snoop System, PIO, Transient

Hi to all the pilots and engineers,

I feel a great progress has been made with the help of these forums.

Despite existence of FBW, automation was still missing in the closed loop system "Pilot AND Aircraft". We pilots were not really ALLOWED to use these words "system", "PIO", "transient", "unstability", and we were not allowed to think them, to fly them...

Despite the best place in the armchair (on the ground) is on Lieman airlines (with 2 or 1 pitot tubes and no redundancy ) a great step [correction : flight] has been done to a new aviation with Machinbird and RR_NDB.

Thank you to both, the first as pilot, the second as system engineer
who have overcome the taboos, without falling into dreams.

Last edited by roulishollandais; 4th Mar 2012 at 20:52. Reason: Lieman airlines armchairs
roulishollandais is offline  
Old 3rd Mar 2012, 19:54
  #785 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
While RR_NDB is working on his next concept, let me throw out one of my own.
There is no doubt that the best course of action with UAS in cruise is to do nothing while you size the situation up. With a little more smarts in programming, the flight control system could do the same thing.
Process runs like this:
  • Oh Oh, lost the airspeeds!
  • Wait, we are in cruise, instructions say do nothing and warn the crew that they may have to take over.
  • Power table lookup says cruise power for current flight environment is X, set power to X, maintain Normal law.
  • Monitor airspeeds to see if we can get a valid airspeed.
  • Airspeeds agree on 90 knots? Not in the acceptable range-disregard.
  • Airspeeds now say 271 knots. In acceptable range. Advise crew that we are resuming normal control.
  • End of allowable dead reckoning period (where control would then pass to the crew).
How you handle loss of airspeeds in climb and descent could probably be handled pretty well with a lookup table as well. That only leaves Approach and the transition right after takeoff to climb speed.
Can they build this reasoning into the computers? Can they certify it? If so, most of the problem is solved.
Machinbird is offline  
Old 3rd Mar 2012, 20:36
  #786 (permalink)  
 
Join Date: Oct 2009
Location: UK
Posts: 1,271
Hi Machinbird,
Power table lookup says cruise power for current flight environment is X, set power to X,....
Can they build this reasoning into the computers? Can they certify it?
Yes & yes they did.
It's called QRH Unreliable Airspeed tables and the central processors are organic.
rudderrudderrat is offline  
Old 3rd Mar 2012, 21:05
  #787 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
It's called QRH Unreliable Airspeed tables and the central processors are organic.
Yes, but you have to wake up the organic central processors. Sometimes they use the wrong programming, sometimes they are slow. sometimes they jump to the wrong spot in the instructions. Lets keep the organic computers in the situational loop, but give them a chance to come up to speed before they are put on the line.
Machinbird is offline  
Old 4th Mar 2012, 00:14
  #788 (permalink)  
 
Join Date: Jun 2009
Location: somewhere
Posts: 451
@Machinebird:


Yes, but you have to wake up the organic central processors. Sometimes they
use the wrong programming, sometimes they are slow. sometimes they jump to the
wrong spot in the instructions. Lets keep the organic computers in the
situational loop, but give them a chance to come up to speed before they are put
on the line.
Do you advocate to remove all organic material from cockpit?
A33Zab is offline  
Old 4th Mar 2012, 01:11
  #789 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
Do you advocate to remove all organic material from cockpit?
Good sense of humor A33Zab.

At the rate they crash drones, that is not going to happen any time soon.
Machinbird is offline  
Old 4th Mar 2012, 09:07
  #790 (permalink)  
 
Join Date: Jul 2009
Location: France - mostly
Age: 79
Posts: 1,689
Originally Posted by Machinbird
Process runs like this:

Oh Oh, lost the airspeeds!
Wait, we are in cruise, instructions say do nothing and warn the crew that they may have to take over.
Power table lookup says cruise power for current flight environment is X, set power to X, maintain Normal law.
Monitor airspeeds to see if we can get a valid airspeed.
Airspeeds agree on 90 knots? Not in the acceptable range-disregard.
Airspeeds now say 271 knots. In acceptable range. Advise crew that we are resuming normal control.
End of allowable dead reckoning period (where control would then pass to the crew).
Very attractive at first sight. We just have to settle on a few details of implementation:
(1) Oh Oh, lost the airspeeds! >>> Based on a single speed (the median)?
(2) Wait, we are in cruise, instructions say do nothing and warn the crew that they may have to take over. >>> Do you mean what the system currently does - maintain current pitch and power and sound 'cavalry charge' - or keep A/P and A/THR engaged, so that A/THR will increase power to Max CLB?
(3) Power table lookup says cruise power for current flight environment is X, set power to X, maintain Normal law. >>> Maintain Normal law means maintaining protection for high AoA and overspeed. Both will not function when IAS drops to low value. High AoA thresholds at low IAS are too high for protection at cruise Mach. There will be no stall warning.
(4) Monitor airspeeds to see if we can get a valid airspeed. Airspeeds agree on 90 knots? Not in the acceptable range-disregard. Airspeeds now say 271 knots. In acceptable range. Advise crew that we are resuming normal control. >>> What is the 'acceptable range' ?
(5) End of allowable dead reckoning period (where control would then pass to the crew). >>> What happens here? Isn't the crew always in control within the flight envelope?

P.S.
What I agree with: When the system drops A/P and A/THR, it could initially set pitch and power corresponding to level flight in still air, and maintain wings level.

Last edited by HazelNuts39; 4th Mar 2012 at 10:37. Reason: P.S.
HazelNuts39 is offline  
Old 4th Mar 2012, 16:07
  #791 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
(1) Oh Oh, lost the airspeeds! >>> Based on a single speed (the median)?
As soon as the system cannot figure out which one of the speeds, if any, is "telling the truth."
(2) Wait, we are in cruise, instructions say do nothing and warn the crew that they may have to take over. >>> Do you mean what the system currently does - maintain current pitch and power and sound 'cavalry charge' - or keep A/P and A/THR engaged, so that A/THR will increase power to Max CLB?
Keep A/P and A/THR engaged so that A/THR will be able to set the lookup table power.
(3) Power table lookup says cruise power for current flight environment is X, set power to X, maintain Normal law. >>> Maintain Normal law means maintaining protection for high AoA and overspeed. Both will not function when IAS drops to low value. High AoA thresholds at low IAS are too high for protection at cruise Mach. There will be no stall warning.
Maintain Normal law means just that. Plug in last valid airspeed and Mach number. Live and die by that. This is a dead reckoning exercise. How long you feel you can safely fly while dead reckoning the airspeed determines the period before you dump control to the pilots. This is likely the key certification issue. 3 seconds sure, no big deal. 20 seconds yeah, why not. 60 seconds, I'm beginning to get uncomfortable. How did we fly in the old days before autothrottles anyway? I can tell you I was not jockeying the throttles in cruise.
(4) Monitor airspeeds to see if we can get a valid airspeed. Airspeeds agree on 90 knots? Not in the acceptable range-disregard. Airspeeds now say 271 knots. In acceptable range. Advise crew that we are resuming normal control. >>> What is the 'acceptable range' ?
To be determined by test flights or using existing criterion if deemed acceptable.
(5) End of allowable dead reckoning period (where control would then pass to the crew). >>> What happens here? Isn't the crew always in control within the flight envelope?
This is where you play the cavalry charge if no valid airspeed is found. A/P and A/THR drop.

Last edited by Machinbird; 4th Mar 2012 at 17:59. Reason: Add existing criteria for airspeed acceptance.
Machinbird is offline  
Old 10th Mar 2012, 02:36
  #792 (permalink)  
 
Join Date: Feb 2011
Location: Nearby SBBR and SDAM
Posts: 868
Ref. Post Mar 4 @ 13:07 by Machinbird

Hi,

Sounds as an evolution:

1) The "wait and see" ("hysteresis") has a time out.
2) The "Organic processors" would have some time to prepare carefully to enter (and act) when "inside the loop"
3) The "graceful degradation" of System + crew is improved during the transient (when the System is being affected by the disturbance)

Fine biz!
RR_NDB is offline  
Old 10th Mar 2012, 08:48
  #793 (permalink)  
PJ2
 
Join Date: Mar 2003
Location: East of the sun, West of the moon
Age: 71
Posts: 2,345
Machinbird;

Re, "As soon as the system cannot figure out which one of the speeds, if any, is "telling the truth." "

It doesn't need to. A power setting and pitch attitude is all you need and as someone suggested here, a recent history if the automation is to do it, (but no thanks!). Speed indication is now irrelevant. An AoA would help but it isn't necessary.

As discussed by Owain Glyndwr and to HN39's point regarding autoflight system programming when the AP and AT drops offline, the A330 fails/failed 'gracefully'. When disconnected by the airplane and by the instinctive disconnect buttons on the thrust levers, the autothrust reverts to the thrust-lock mode, maintaining the power setting at disconnect.

Almost all the time the thrust is appropriate. Here, the thrust had dropped slightly but to no short-term effect.

The checklist requires that the autothrust be disconnected, (press the AT button on the FCU or the instinctive buttons and THR LK is removed), then move the thrust levers out of the CLB detent and set thrust according to the tables. About 88% N1 or 1.2 EPR would be a good start, if I recall.

In Alternate Law pitch must be manually controlled so there is perhaps room for some form of "maintaining pitch" but I think this is taking automation precisely where we don't want to take it..."more", to cater to reduced thinking and skill levels.

The airplane is eminently flyable at cruise altitudes even in heavy turbulence, (done it...it's like any other airplane including the DC8)...tiny movements on the stick, (or control column), wait..., set the thrust, wait..., read the checklist, wait for the airspeed to return, as it did within 30 seconds in all previous events. And if it doesn't then complete the checklists, ECAMs, STATUS pages and secure the aircraft for manual flight, call the captain and decide whether to return to RIO (or..?) or continue.

RR_NDB, this response isn't "the academice ideal", this is what crews are trained to do and is absolutely middle-of-the-road SOPs. That it isn't done sometimes is not a counter-example argument for a sense that this is "expecting a lot", nor is it an argument for further-reduced standards and more automation.

In climb/descent and except for close in approach (when dirty), one levels off, and does the same thing. The powerpoint presentation to which Turbine D provided us the link is a superb explanation of the "new" (2006) process. The powerpoint referenced by Turbine D is at http://www.iag-inc.com/premium/Airbu...ableSpeeds.pdf

BTW, I quite appreciate that typing this stuff out is SO much easier. Like others here I'm sure, I've had a few "events" and they are never straightforward and the adrenaline always comes.

Last edited by PJ2; 10th Mar 2012 at 09:01.
PJ2 is offline  
Old 10th Mar 2012, 14:36
  #794 (permalink)  
 
Join Date: Jun 2011
Location: france
Posts: 756
Devil Artificial Intelligence is not KISS

Originally Posted by HazelNuts39
to level flight in still air.
oh I have severe turbulence...

Originally Posted by HazelNuts39
... and maintain wings level.
Oups ! I was just turning to avoid the mountains in the clouds...
roulishollandais is offline  
Old 10th Mar 2012, 14:58
  #795 (permalink)  
 
Join Date: Jun 2011
Location: france
Posts: 756
Wink vocabulary

Thank you to allow us to speek non-linear systems, transitory stages during the flight, closed-loop systems, normality of oscillations, pluridisciplinarity knowledge, etc.

Could we generalize the use of "effective aircraft" or "effective system" to refer to the system (pilot & aircraft) ? I find it very clear to the mind, without excessive simplification.

Congratulations for your next speed sensor. Did PPRuNe already found a manufacturer for your many inventions : I am impatient to see the advert !

Monsieur Henri Pitot will soon be allowed to sleep...
roulishollandais is offline  
Old 10th Mar 2012, 15:55
  #796 (permalink)  
 
Join Date: Aug 2011
Location: Grassy Valley
Posts: 2,123
Yes, it is a leap from armchair to RHS in Bonin's shoes. Since BEA have not issued their Final, it is disconcerting to see so many verdicts of guilty. For that matter, BEA will issue data, and will not condemn. The data may, but BEA is not charged to place blame.

My over arching beef with the investigation is the length of time they take. Also the slim releases of data. To me, it is not for science, but for politics. Time not only heals, it shelters. When the stage has been arranged so completely with damning evidence, one wonders if it is a true picture, or merely a cynical attempt to protect some survivng interests.
Lyman is offline  
Old 11th Mar 2012, 07:01
  #797 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
Originally Posted by PJ2
Re, "As soon as the system cannot figure out which one of the speeds, if any, is "telling the truth." "

It doesn't need to. A power setting and pitch attitude is all you need and as someone suggested here, a recent history if the automation is to do it, (but no thanks!). Speed indication is now irrelevant. An AoA would help but it isn't necessary.
PJ2, I am not sure you understood the intent of the procedure I posted. Its purpose was to basically outline the key functional elements of a computer program that would ease the transition between Normal law autopilot controlled flight and Alternate Law, pilot controlled flight. Rather than drop the aircraft into the pilots lap all at once, the transition could be spread over some time.

I know that pilots worry about their job security when computers are performing things that pilots consider their turf (such as emergency procedures.) I also know that sleepy pilots in cruise in the middle of the night can be slow to come up to speed. Best move is to provide all possible assistance to make the transition as painless as possible. If you get a little help doing your job, it isn't a bad thing, as long as the automatic help does not make things worse. 9 times out of 10, you will be alert and the help won't matter. It is that other time at 2:05 in the morning that you might appreciate the help.

The key thing to recognize is that the aircraft could probably handle the complete UAS procedure by itself for a period of time, assuming proper computer programming. The A330 is a fairly early version of Fly By Wire and despite its great success, does not mean it could not be made better. We need to pick out rough spots in the man-machine interface and improve them. This may be one of those.
Machinbird is offline  
Old 11th Mar 2012, 08:25
  #798 (permalink)  
 
Join Date: Jun 2009
Location: NNW of Antipodes
Age: 75
Posts: 1,328
Originally posted by Machinbird...
Rather than drop the aircraft into the pilots lap all at once, the transition could be spread over some time.
That suggestion raises a deep sense of distrust in pilots that believe that when the automatics "fail", their own ability to take-over should not be questioned.

I'm not disputing that point, but when you are the air-frame manufacturer, your profit margin goes down the "big white telephone" every time someone stuffs up in that department. In that respect, I'm pretty sure that Airbus Industries will be moving (limited only by the time taken for regulatory approval) to do just as you suggest.

Perhaps I could refresh your memory to something posted eons ago in AF447- Thread No.4. The wording might be different, but the intention is just the same.
mm43 is offline  
Old 11th Mar 2012, 18:20
  #799 (permalink)  
 
Join Date: Jul 2009
Location: Not far from a big Lake
Age: 76
Posts: 1,396
Originally Posted by MM43
The wording might be different, but the intention is just the same.
Yep, it was a good idea then and is still a good idea. This type adaptation of the FBW system is appropriate to improving the human interface.

The idea of the system suddenly shrieking, "I don't know how to compute this, YOU have it." rubs the flight instructor in me the wrong way.
Machinbird is offline  
Old 11th Mar 2012, 19:30
  #800 (permalink)  
 
Join Date: Jul 2009
Location: France - mostly
Age: 79
Posts: 1,689
Looking back at an earlier post makes me wonder whether 20 seconds or so of dead reckoning to delay the cavalry charge would have made much difference in this case.
HazelNuts39 is offline  

Thread Tools
Search this Thread

Contact Us Archive Advertising Cookie Policy Privacy Statement Terms of Service