TAM A320 crash at Congonhas, Brazil
I support PPRuNe
Join Date: Jul 2007
Location: Belo Horizonte, Brazil
Posts: 162
Likes: 0
Received 0 Likes
on
0 Posts
Kleyber Lima
The two TAM pilots. On the right Kleyber Lima, with TAM since the 80´s was getting ready to retire. Left Stephanini. Veja magazine said that Kleyber was piloting the plane. Above, how part of the building was after the disaster.
Join Date: Jul 2002
Location: UK
Posts: 3,093
Likes: 0
Received 0 Likes
on
0 Posts
But Rob21 and ChristiaanJ,
The procedure, regardless of locked out TL, is still to bring both TLs to idle before performing any other action before and after the Taiwan and Conghonas incidents. Always has been, and I suspect it always will be.
The procedure, regardless of locked out TL, is still to bring both TLs to idle before performing any other action before and after the Taiwan and Conghonas incidents. Always has been, and I suspect it always will be.
From the perspective of someone dealing with stats daily I’d go along with Rob’s, Marcio’s and ChristiaanJ’s view on them, i.e. not just how many times out of total arrivals was any TL positioned incorrectly, but also how many times did it happen when a thrust reverser was pinned. Both índices are valid for study. If of course such data in thousands of landings were available and reliable and, of course, if AI had a way of monitoring it. Suspect no and no, ideal vs possible.
Join Date: Aug 2007
Location: Brazil
Age: 71
Posts: 131
Likes: 0
Received 0 Likes
on
0 Posts
Maybe not, DozyWannabe
I don't remember the date AI changed the procedure on handling TLs with a locked reverse. Once you left the affected engine TL at idle. Now, you bring both to reverse. What was always, it is not any more. OK, now is just as none of the engines had a reverse inop. But we have one. This is not the same landing. This is not like 30 million landings.
Your LDA will change (extra 55 meters?) doing so.
So it is NOT like any other landing, IMHO.
Rob
I don't remember the date AI changed the procedure on handling TLs with a locked reverse. Once you left the affected engine TL at idle. Now, you bring both to reverse. What was always, it is not any more. OK, now is just as none of the engines had a reverse inop. But we have one. This is not the same landing. This is not like 30 million landings.
Your LDA will change (extra 55 meters?) doing so.
So it is NOT like any other landing, IMHO.
Rob
Last edited by Rob21; 29th Sep 2007 at 01:00. Reason: typo
I support PPRuNe
Join Date: Jul 2007
Location: Belo Horizonte, Brazil
Posts: 162
Likes: 0
Received 0 Likes
on
0 Posts
Something fishy...
Can you make sense of this sequence of news published in VEJA?
This is an atitude that shows the need to learn with the accidents to avoid that they repeat. Captain Ranger Chen, investigative chief in the Aviation Security Council of Taiwan is part of this effort. He is the author of the report of the accident with the Airbus A320 from Transania at the Airport Sung Shan in Taipei, in October 18, 2004. Interviewed by VEJA he says: " At the time of the report, Airbus acepted our recommendation. It was said that the company had developped an alarm system for when the TLs were in an incorrect position during the landing. This did not have to do with the "retard, retard, retard" registered by the black boxes in the Congonhas accident, but one CRC. A continuous alarm that only would stop when the pilots place the TLs in the correct position. The alarm would be triple, besides a loud sound, a red light would go on in the cabin and a mensage would be read at the panel to indicate the situation. It was also said that a Service Bulletin would be served. The Airbus gave up the idea of implementing the system of alarm. We never received the Service Bulletim."
Veja also contacted the General Direction of Civil Aviation in France (DGAC) and the European Agency for Aviation Safety (EASA). If Airbus had sent a Sercie Bulletin, the company had to communicate the two agencies. DGAC and EASA did not receive the Bulletim mentioned by Airbus in the document to the chinese authorities in Taiwan.
The accident of the A320 from TAM in Congonhas is the third case on the same type of airplane when the TLs during landing were in an incorrect position. It ocurred three years after the recommendation from the Security Council of Aviation in Taiwan for the company to improve the alert system.
The vice-president for flight security of TAM, Marco Aurelio de Castro, announced in the CPI that the company is going to buy a equipment that warns the pilots of the incorrect position of TLs during landing. This new alarm is availabe since November 2006, when Airbus informed about its existence in a Service Bulletim sent to airlines that operate this king of plane.
In a practical way. The new alert. That costs 5000 dollars, is an atualization of a program in the computer of the plane, the System of Flight Alert- FWC, the initials in English. The present sound alert of short duration becomes together of a warning light e and text message in the command panel. This sound warning played during three seconds in the accident in Congonhas ("retard,retard,retard") and then stopped. This becase the computer presuposes that if the pilot doesn´t respond to the warning, there is a reazon- an obstacle in the runway or an attepmt to go around, for example.
After the accident with the A320 from Transania in Taipei –one TL was in the wrong position and one reversor was locked out, a situation like the one tragedy in Congonhas – the Security Council of Aviation in Taiwan sugested that the alarm should continue to sound even after the touchdown. Airbus responded that it had created a new system and thau it would send a Bulletim about this "very soon". This was in October 2004, that is, 25 months before the note of atualization sent to the airlines. Besides generating a text message and getting red light on in the panel, the alarm would send a continuous repetitive chime (CRC) – sound alarm veru loud that calls the attention for a grave problem. This would demand a radical reform in the program of FWC. The Airbus prefered only to upgrade the system. To justify, Yannick Maling, director of flight security at Airbus, tells that the alert would not help the pilot to correct the error during the crucial seconds of taking a radical decision.
This is an atitude that shows the need to learn with the accidents to avoid that they repeat. Captain Ranger Chen, investigative chief in the Aviation Security Council of Taiwan is part of this effort. He is the author of the report of the accident with the Airbus A320 from Transania at the Airport Sung Shan in Taipei, in October 18, 2004. Interviewed by VEJA he says: " At the time of the report, Airbus acepted our recommendation. It was said that the company had developped an alarm system for when the TLs were in an incorrect position during the landing. This did not have to do with the "retard, retard, retard" registered by the black boxes in the Congonhas accident, but one CRC. A continuous alarm that only would stop when the pilots place the TLs in the correct position. The alarm would be triple, besides a loud sound, a red light would go on in the cabin and a mensage would be read at the panel to indicate the situation. It was also said that a Service Bulletin would be served. The Airbus gave up the idea of implementing the system of alarm. We never received the Service Bulletim."
Veja also contacted the General Direction of Civil Aviation in France (DGAC) and the European Agency for Aviation Safety (EASA). If Airbus had sent a Sercie Bulletin, the company had to communicate the two agencies. DGAC and EASA did not receive the Bulletim mentioned by Airbus in the document to the chinese authorities in Taiwan.
The accident of the A320 from TAM in Congonhas is the third case on the same type of airplane when the TLs during landing were in an incorrect position. It ocurred three years after the recommendation from the Security Council of Aviation in Taiwan for the company to improve the alert system.
The vice-president for flight security of TAM, Marco Aurelio de Castro, announced in the CPI that the company is going to buy a equipment that warns the pilots of the incorrect position of TLs during landing. This new alarm is availabe since November 2006, when Airbus informed about its existence in a Service Bulletim sent to airlines that operate this king of plane.
In a practical way. The new alert. That costs 5000 dollars, is an atualization of a program in the computer of the plane, the System of Flight Alert- FWC, the initials in English. The present sound alert of short duration becomes together of a warning light e and text message in the command panel. This sound warning played during three seconds in the accident in Congonhas ("retard,retard,retard") and then stopped. This becase the computer presuposes that if the pilot doesn´t respond to the warning, there is a reazon- an obstacle in the runway or an attepmt to go around, for example.
After the accident with the A320 from Transania in Taipei –one TL was in the wrong position and one reversor was locked out, a situation like the one tragedy in Congonhas – the Security Council of Aviation in Taiwan sugested that the alarm should continue to sound even after the touchdown. Airbus responded that it had created a new system and thau it would send a Bulletim about this "very soon". This was in October 2004, that is, 25 months before the note of atualization sent to the airlines. Besides generating a text message and getting red light on in the panel, the alarm would send a continuous repetitive chime (CRC) – sound alarm veru loud that calls the attention for a grave problem. This would demand a radical reform in the program of FWC. The Airbus prefered only to upgrade the system. To justify, Yannick Maling, director of flight security at Airbus, tells that the alert would not help the pilot to correct the error during the crucial seconds of taking a radical decision.
Last edited by marciovp; 29th Sep 2007 at 01:03. Reason: typo
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by PBL
I was reconstructing what I take to be AI's position from the comments that Malinge made. It is a coherent reconstruction (as it has to be; I couldn't imagine AI or any other airframe manufacturer presenting dumb arguments on such an important issue).
- in that spirit:-Recommended Strategy for Airbus (if the case goes to court):-
1. Putting too much emphasis on the logical process of decision is inadvisable - it invites accusations of mistake (actually taking a wrong decision, i.e. DECIDING not to install any extra warning) as opposed to mere negligence (i.e failing to decide in good time on the best FORM of any such warning, even though development work etc. had been completed on the one eventually used).
2. Therefore the preferred posture should be to admit to treating the issue as less than 'top urgent,' for the reason that no serious accidents had occurred as a result of the problem, nor did any appear likely to happen.
3. This should be coupled with pressure on the airline to show what extra training on TL handling it put in place after the earlier incidents; and on the airport authority to indicate what runway modifications etc. it commissioned to allow for any such overruns. The answers in both cases are likely to be 'Sorry, none.'
That's all on the basis of what I know so far, though. If I actually WAS back in my working days, and involved in helping to prepare Airbus' case, there's some questions I'd need answers to before finalising any recommendation (or, indeed, before getting involved at all). These are:-
Additional Questions for Airbus:-
1. At some time between the Taipei incident and the Congonhas accident you apparently recommended changes to A320 throttle-handling procedure in the event of one reverser inop. - specifically, that both handles should be pulled back not just to 'idle,' but beyond 'idle' into full reverse.
2. Did the decision to recommend this new procedure have anything to do with your findings following previous overrun incidents like Taipei? If so, WHAT were your findings, WHEN did you obtain them, WHO did you tell, and WHY did they prompt the change to 'both levers to full reverse'?
If, there does turn out to be some connection (in the form of a software or hardware glitch which hasn't yet been disclosed) that would very much change the picture.
Last edited by RWA; 29th Sep 2007 at 01:49.
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by marciovp
Can you make sense of this sequence of news published in VEJA?
I think one can make some sense of it - but the implications are huge.
"This sound warning played during three seconds in the accident in Congonhas ("retard,retard,retard") and then stopped. This becase the computer presuposes that if the pilot doesn´t respond to the warning, there is a reazon- an obstacle in the runway or an attepmt to go around, for example."
So far so good. Trouble is, though, that the computer doesn't appear to stop 'pre-supposing' there.
1. On the one hand, it appears to 'assume' that the pilot intends to land to the extent of allowing the deployment of reverse thrust.
2. On the OTHER hand it appears to 'assume' that the pilot does NOT intend to land in a much more serious way, by preventing the deployment of ground spoilers and the operation of autobraking.
And there is no manual override available to the pilots, especially in the vital area of spoilers.
"This would demand a radical reform in the program of FWC. The Airbus prefered only to upgrade the system."
If that's true as stated by VEJA - that Airbus just put a clearcut control system anomaly like that in the 'Too Hard' basket - this case could get very 'big.'
PS - Brazil's girls entirely deserve their success - they completely out-played the USA. I expect them to win the whole thing.
Last edited by RWA; 29th Sep 2007 at 02:04. Reason: Add PS
Sun worshipper
Join Date: Dec 2001
Location: Paris
Posts: 494
Likes: 0
Received 0 Likes
on
0 Posts
(rant ON)
marciovp,
you forgot to show the mothers' pictures...and the wives....or the children.
And you omitted the tears of the sisters...
You see, you forget a lot.
Where have descended to ?
A "people" magazine subsidiary ?
As for statistics, to-day, for close to a total of 250 airplanes, 7 have a de-activated T/R.
So the *rarity* of this configuration is quite a lot less than you'd imagine.
Why ?
Because T/Rs are built for failure, because
1/of the dangers inherent to a faulty one in flight (Remember Lauda 767 ?)
2/It is not required - by regulation on normal performance.
3/of possible mishandling.
About the change in procedure,
Funny how arguments keep on re appearing, the explanations conveniently forgotten. No, the change in the T/L handling procedure has nothing to do with Tai Pei but more precisely with standardisation. It came from the airlines wo asked AI precisions on the selection of reverse on a blocked T/R. With the approval of the certifying authorities, the procedure was both simplified - in terms of handling - and put back in line with the normal operation. I have to add that the same question was asked BCA, with similar result and application.
Coming to the subject of BCA, throttles mishandlings similar to those cited here have happened...Surprise ! they are not mentioned here any more and our friends so eager to defend dead pilots and improve flight safety have not asked that Boeing airplanes should have the same warning. Why not all airplanes, why stop here ?
Next accident, go demonstrate for placards in the flight deck
TO CLIMB, PULL THE STICK AFT" and *FOR LANDING PULL LEVERS BACK*!
You'd probably feel better knowing that the pilot won't forget these instructions !
If something is *fishy*, it is the turn that this thread has taken :
The "good people" who are on their high horses do not care a fig about improving flight safety at all but about what arguments are going to be used in a court of law in order to obtain more dough from AI and their insurance. Very interesting for PPrune, Hey ?
I for one cannot stand the wolf pack mentality that has been shown in the last two pages of this thread, highly qualified specialists -PBL and bsieker- being rubbished by the same people whose chances of survival in the future they want to improve. People who'd do a lot better if they'd just had a look on their publications (not even that ,The list of their publications).
Talk about ungratefulness and shortsightedness.
And pure lunacy.
(rant OFF)
marciovp,
you forgot to show the mothers' pictures...and the wives....or the children.
And you omitted the tears of the sisters...
You see, you forget a lot.
Where have descended to ?
A "people" magazine subsidiary ?
As for statistics, to-day, for close to a total of 250 airplanes, 7 have a de-activated T/R.
So the *rarity* of this configuration is quite a lot less than you'd imagine.
Why ?
Because T/Rs are built for failure, because
1/of the dangers inherent to a faulty one in flight (Remember Lauda 767 ?)
2/It is not required - by regulation on normal performance.
3/of possible mishandling.
About the change in procedure,
Funny how arguments keep on re appearing, the explanations conveniently forgotten. No, the change in the T/L handling procedure has nothing to do with Tai Pei but more precisely with standardisation. It came from the airlines wo asked AI precisions on the selection of reverse on a blocked T/R. With the approval of the certifying authorities, the procedure was both simplified - in terms of handling - and put back in line with the normal operation. I have to add that the same question was asked BCA, with similar result and application.
Coming to the subject of BCA, throttles mishandlings similar to those cited here have happened...Surprise ! they are not mentioned here any more and our friends so eager to defend dead pilots and improve flight safety have not asked that Boeing airplanes should have the same warning. Why not all airplanes, why stop here ?
Next accident, go demonstrate for placards in the flight deck
TO CLIMB, PULL THE STICK AFT" and *FOR LANDING PULL LEVERS BACK*!You'd probably feel better knowing that the pilot won't forget these instructions !
If something is *fishy*, it is the turn that this thread has taken :
The "good people" who are on their high horses do not care a fig about improving flight safety at all but about what arguments are going to be used in a court of law in order to obtain more dough from AI and their insurance. Very interesting for PPrune, Hey ?
I for one cannot stand the wolf pack mentality that has been shown in the last two pages of this thread, highly qualified specialists -PBL and bsieker- being rubbished by the same people whose chances of survival in the future they want to improve. People who'd do a lot better if they'd just had a look on their publications (not even that ,The list of their publications).
Talk about ungratefulness and shortsightedness.
And pure lunacy.
(rant OFF)
Join Date: Jul 2007
Location: the City by the Bay
Posts: 547
Likes: 0
Received 0 Likes
on
0 Posts
I do think that most contributors of this thread are not in it for personal profit or even recognition. I do think that it is because they are driven by the gravity of this particular accident. In the back of our minds surely is the thought that nearly 200 people died in this tragedy. I do believe that is a major reason this thread has attracted so much attention and contribution. I do believe that most avid readers and contributors to this thread do this out of respect for those who died and feel a need to know how it can be avoided.
We have researchers that have taught us how to logically think out a problem. We have professional pilots that have taken a lot of effort to educate those (such as myself ) who are not pilots and provide us with their expertise. Especially those who are current on Airbus A320.
Some ideas are thrown about, a few personal volleys traded. Inevitable as people get involved. Sound minds are questioned. Through all this , I do believe its our quest to discover the truth of what happened and how to best prevent a recurrence. Some can be thanked for thoughtful contribution. Others contributed something important by chance.
We have Brazilians driven by the love of their country to seek out who is really responsible for this Brazilian tragedy. I said responsible, not necessarily in order to assign blame. I am Taiwanese American. I read how a veteran pilot said he nearly lost it on landing at Sungshan and remembered Transasia had an overrun there and found a link to it. Some Airbus pilots found and read the final report to that accident and wrote about the alarming similarity to what happened at Congonhas. Thats just what I meant when I said some contributed by chance. Whereas others have put a lot of thought and experience into their contributions and they are to be thanked for their efforts.
I do think that this thread has helped the relevant authorities in their quest for the truth and actions taken or to be taken to avoid a recurrence in general and at Congonhas in particular.
And I say let the ideas continue to flow and the questions continue to be asked. Personally I like to think that there are no stupid questions , only stupid answers.
Edit: Why am I saying this now? Just to keep us all focused and less personal volleying.
We have researchers that have taught us how to logically think out a problem. We have professional pilots that have taken a lot of effort to educate those (such as myself ) who are not pilots and provide us with their expertise. Especially those who are current on Airbus A320.
Some ideas are thrown about, a few personal volleys traded. Inevitable as people get involved. Sound minds are questioned. Through all this , I do believe its our quest to discover the truth of what happened and how to best prevent a recurrence. Some can be thanked for thoughtful contribution. Others contributed something important by chance.
We have Brazilians driven by the love of their country to seek out who is really responsible for this Brazilian tragedy. I said responsible, not necessarily in order to assign blame. I am Taiwanese American. I read how a veteran pilot said he nearly lost it on landing at Sungshan and remembered Transasia had an overrun there and found a link to it. Some Airbus pilots found and read the final report to that accident and wrote about the alarming similarity to what happened at Congonhas. Thats just what I meant when I said some contributed by chance. Whereas others have put a lot of thought and experience into their contributions and they are to be thanked for their efforts.
I do think that this thread has helped the relevant authorities in their quest for the truth and actions taken or to be taken to avoid a recurrence in general and at Congonhas in particular.
And I say let the ideas continue to flow and the questions continue to be asked. Personally I like to think that there are no stupid questions , only stupid answers.
Edit: Why am I saying this now? Just to keep us all focused and less personal volleying.
Last edited by armchairpilot94116; 29th Sep 2007 at 03:38.
Join Date: Sep 2000
Location: Bielefeld, Germany
Posts: 955
Likes: 0
Received 0 Likes
on
0 Posts
Some people are showing themselves to have some difficult with basic statistics (including someone who claims to deal with them every day, which is a little worrying!)
Let me take up Rob's invitation. I'll phrase it in words you might hear in a first class in statistics. One can look up the words in a basic intro, of which there are many on the WWW. And let me try to perform the calculations marciovp and Rob21 would prefer to see.
We wish to investigate the relative non-occurrence of an event X. First step is to choose an appropriate sample space.
Event X is: reduction of thrust on all engines to idle on an aircraft at flare. The most obvious sample space we can choose is: all landing aircraft. We may do that. But we are particularly interested in this thread in certain aircraft, namely the A320, and we are much less interested in, say, B737 or B747.
So we can choose: (1) all landing A320 aircraft.
Or we can choose: (2) all landing A320 aircraft with a yellow-painted tail fin and a pilot whose name is Bill.
Or we can choose: (3) all landing A320 aircraft with a reverser locked out.
No. (1) is the widest sample space. At the time of decision about including the "Taipei" warning, there were two known non-occurrences of X amongst, say, 30 million landings. Which is what I said.
marciovp and Rob21 prefer to look at (3). They are welcome to do so. Let's see if I can help.
Rob21's intervention showed he is concerned with what is called the *correlation* between two events. These events are
* non-X: failure to retard all thrust levers to idle on landing
* Y: at least one thrust reverser locked out; or maybe
* Z: exactly one thrust reverser locked out.
Rob21, and I presume marciovp, are really interested in what relation not reducing thrust on both engines might have with reversers placarded inop. A legitimate concern.
Let's see what they have to do to reason about this *statistically*.
The correlation between non-X and, say, Z is usually measured by calculating from four figures: the total number of events in the sample space; the total number of non-X events in the sample space; the total number of Z events in the sample space; the number of (non-X & Z) events in the sample space. (These four figures form a sufficient set. From these, you can also calculate the number of (X & Z) events as well as the number of (X and non-Z) events and the number of (non-X and non-Z events) if you want.)
The first point to notice is that for Rob21 to get where he wants, he has to perform the same calculation he has
chided me for performing: he has to determine the number of non-X events in the sample space.
But he is also going to stumble across a difficulty: because he has no figure for Z, he doesn't have the third and fourth numbers he needs to calculate a correlation.
So he is stuck.
Since it is not possible for marciovp and Rob21 to calculate what they would prefer to calculate, I have no idea why they would suggest that *I* should be doing it. I prefer to stick to the possible; besides, they need the number I produced in any case.
But maybe we can get further with Lemurian's intervention. He said that in a fleet of 250 aircraft, there are currently 7 with a reverser locked out. If we take that to be typical, that might suggest some 850,000 total A320 landings with at least one reverser locked out. So now marciovp and Rob21 have figures they can use!
Total size of sample space: 30 million
Non-X events: 2
Y events: 850,000
(Non-X and Y) events: 2
Thank you Lemurian!
Now we come across is a deeper reason that one cannot do what marciovp and Rob21 are suggesting. We cannot derive any meaningful *statistical* conclusion. That is: the number of total events non-X that one has registered is too small. Any Tests of Significance (I capitalise it because this is another technical term which I suggest that people look up) would show that any correlation derived from a sample of two are not significant.
Fact is, any correlation exhibited here has to come from considerations very different from considering the numbers. So Rob21 and marciovp are chiding me for not doing something that would be inappropriate in any case.
Now let's look a little beyond the sparse data. The purpose of calculating the proportions of these things is, remember, that we are AI trying to determine what to do about these events. We really, really, really, don't want any non-X events.
Remember, we already have a warning about non-X events on the aircraft. Do we want another one? Do we want the one we have to be different? And when we change things, we have to weigh the chances that we introduce more (and different) difficulties, in other words the phenomenon well-known to programmers of reliable systems dealing with rare events of fix one bug; introduce two bugs more.
I am not going to discuss this further here, because my original purpose was to present someone else's reasoning, not my own view. These are the considerations, somewhat amplified, that M. Malinge introduced in his deposition. And, whatever your views of AI's decisions, they are valid considerations, and generally agreed amongst the experts in this field to be so.
PBL
Let me take up Rob's invitation. I'll phrase it in words you might hear in a first class in statistics. One can look up the words in a basic intro, of which there are many on the WWW. And let me try to perform the calculations marciovp and Rob21 would prefer to see.
We wish to investigate the relative non-occurrence of an event X. First step is to choose an appropriate sample space.
Event X is: reduction of thrust on all engines to idle on an aircraft at flare. The most obvious sample space we can choose is: all landing aircraft. We may do that. But we are particularly interested in this thread in certain aircraft, namely the A320, and we are much less interested in, say, B737 or B747.
So we can choose: (1) all landing A320 aircraft.
Or we can choose: (2) all landing A320 aircraft with a yellow-painted tail fin and a pilot whose name is Bill.
Or we can choose: (3) all landing A320 aircraft with a reverser locked out.
No. (1) is the widest sample space. At the time of decision about including the "Taipei" warning, there were two known non-occurrences of X amongst, say, 30 million landings. Which is what I said.
marciovp and Rob21 prefer to look at (3). They are welcome to do so. Let's see if I can help.
Rob21's intervention showed he is concerned with what is called the *correlation* between two events. These events are
* non-X: failure to retard all thrust levers to idle on landing
* Y: at least one thrust reverser locked out; or maybe
* Z: exactly one thrust reverser locked out.
Rob21, and I presume marciovp, are really interested in what relation not reducing thrust on both engines might have with reversers placarded inop. A legitimate concern.
Let's see what they have to do to reason about this *statistically*.
The correlation between non-X and, say, Z is usually measured by calculating from four figures: the total number of events in the sample space; the total number of non-X events in the sample space; the total number of Z events in the sample space; the number of (non-X & Z) events in the sample space. (These four figures form a sufficient set. From these, you can also calculate the number of (X & Z) events as well as the number of (X and non-Z) events and the number of (non-X and non-Z events) if you want.)
The first point to notice is that for Rob21 to get where he wants, he has to perform the same calculation he has
chided me for performing: he has to determine the number of non-X events in the sample space.
But he is also going to stumble across a difficulty: because he has no figure for Z, he doesn't have the third and fourth numbers he needs to calculate a correlation.
So he is stuck.
Since it is not possible for marciovp and Rob21 to calculate what they would prefer to calculate, I have no idea why they would suggest that *I* should be doing it. I prefer to stick to the possible; besides, they need the number I produced in any case.
But maybe we can get further with Lemurian's intervention. He said that in a fleet of 250 aircraft, there are currently 7 with a reverser locked out. If we take that to be typical, that might suggest some 850,000 total A320 landings with at least one reverser locked out. So now marciovp and Rob21 have figures they can use!
Total size of sample space: 30 million
Non-X events: 2
Y events: 850,000
(Non-X and Y) events: 2
Thank you Lemurian!
Now we come across is a deeper reason that one cannot do what marciovp and Rob21 are suggesting. We cannot derive any meaningful *statistical* conclusion. That is: the number of total events non-X that one has registered is too small. Any Tests of Significance (I capitalise it because this is another technical term which I suggest that people look up) would show that any correlation derived from a sample of two are not significant.
Fact is, any correlation exhibited here has to come from considerations very different from considering the numbers. So Rob21 and marciovp are chiding me for not doing something that would be inappropriate in any case.
Now let's look a little beyond the sparse data. The purpose of calculating the proportions of these things is, remember, that we are AI trying to determine what to do about these events. We really, really, really, don't want any non-X events.
Remember, we already have a warning about non-X events on the aircraft. Do we want another one? Do we want the one we have to be different? And when we change things, we have to weigh the chances that we introduce more (and different) difficulties, in other words the phenomenon well-known to programmers of reliable systems dealing with rare events of fix one bug; introduce two bugs more.
I am not going to discuss this further here, because my original purpose was to present someone else's reasoning, not my own view. These are the considerations, somewhat amplified, that M. Malinge introduced in his deposition. And, whatever your views of AI's decisions, they are valid considerations, and generally agreed amongst the experts in this field to be so.
PBL
Last edited by PBL; 29th Sep 2007 at 08:33. Reason: All at threes and fours
Join Date: Jul 2007
Location: Germany
Posts: 556
Likes: 0
Received 0 Likes
on
0 Posts
Statistics vs. Causality
PBL,
thanks for your clarifying iportant aspects about statistics and statistical (in)significance.
Another point may always be if you can show a causal link between the locked reverser and the crew mishandling the thrust levers in the flare. (I stress again: the mistake was made while pulling levers to idle in the flare, reminded to by the "Retard!"-calls, and not while selecting reverse).
To show causality you need a number of necessary causal factors, and you need to show that these are sufficient.
A well-known way to check for necessary causality is application of the counterfactual test, one of its informal representations may be:
"B is a necessary causal factor of A, if, had B not happened, A would not have happened, either."
Thus, to show that the locked reverser was a causal factor of the flight crew not pulling thrust lever #2 to idle, the following must be true:
Had the thrust reverser not been deactivated, the crew would have pulled the thrust lever #2 to idle.
How can we tell if that is true?
We can try to speculate phycologically that the knowledge about the locked-out reverser, and the reminder shortly before landing about "remember, only one reverser" may have influenced the crew's actions, but this still will not satisfy the counterfactual test, although perhaps some probabilities may be arrived at by human psychology experts.
Secondly, even if we take it as true, we are certain that it is not sufficient. There were hundreds of thousands of landings with locked reversers and proper handling of the thrust levers. So we need at least one other necessary causal factor to explain why the crew made this particular mistake in this particular instance. One can imagine high work load, inexperience on type, CRM issues, training, ... as necessary causal factors, but it seems doubtful that we can ever show causal sufficiency.
I am sure the investigators will look at all these in detail and work with experts in the field and may find it necessary to further stress the importance of adhering to SOPs and of discipline on the flight deck, as matters of training.
If and when the official investigation points to such causes, which will have to pass the "IN/JS" criterion (Individually Necessary / Jointly Sufficient), I may augment my Why-Because Graph with nodes showing the factors leading up to the crew mistake.
Bernd
thanks for your clarifying iportant aspects about statistics and statistical (in)significance.
Another point may always be if you can show a causal link between the locked reverser and the crew mishandling the thrust levers in the flare. (I stress again: the mistake was made while pulling levers to idle in the flare, reminded to by the "Retard!"-calls, and not while selecting reverse).
To show causality you need a number of necessary causal factors, and you need to show that these are sufficient.
A well-known way to check for necessary causality is application of the counterfactual test, one of its informal representations may be:
"B is a necessary causal factor of A, if, had B not happened, A would not have happened, either."
Thus, to show that the locked reverser was a causal factor of the flight crew not pulling thrust lever #2 to idle, the following must be true:
Had the thrust reverser not been deactivated, the crew would have pulled the thrust lever #2 to idle.
How can we tell if that is true?
We can try to speculate phycologically that the knowledge about the locked-out reverser, and the reminder shortly before landing about "remember, only one reverser" may have influenced the crew's actions, but this still will not satisfy the counterfactual test, although perhaps some probabilities may be arrived at by human psychology experts.
Secondly, even if we take it as true, we are certain that it is not sufficient. There were hundreds of thousands of landings with locked reversers and proper handling of the thrust levers. So we need at least one other necessary causal factor to explain why the crew made this particular mistake in this particular instance. One can imagine high work load, inexperience on type, CRM issues, training, ... as necessary causal factors, but it seems doubtful that we can ever show causal sufficiency.
I am sure the investigators will look at all these in detail and work with experts in the field and may find it necessary to further stress the importance of adhering to SOPs and of discipline on the flight deck, as matters of training.
If and when the official investigation points to such causes, which will have to pass the "IN/JS" criterion (Individually Necessary / Jointly Sufficient), I may augment my Why-Because Graph with nodes showing the factors leading up to the crew mistake.
Bernd
Join Date: Aug 2007
Location: Brazil
Age: 71
Posts: 131
Likes: 0
Received 0 Likes
on
0 Posts
Thanks, PBL
Great class, I was not trying to teach the priest how to say the Mass. I was trying to teach the Pope...
I'm sorry, I was not chiding you, or better, this was not my intention. After showing so much patience, I can only applaud your effort.
Now, if I may, let me give you an example. With this example maybe I can make myself a bit clearer.
Supose I am a potential buyer of an A-320 type airplane. I know now that around 1% of the flights today will have trouble with reverse. So to help in my decision to buy A or B, I want to know WHY all three (or four) accidents/incidents with probable causes being mishandling TLs on landing, had problems with reverse (locked). So, to convince me this should not be a concern, the "salesman" demonstrate to me that operations with a "locked" reverse are normal (MEL, MMEL, etc,etc...). So I ask him how many accidents/incidents happened caused by misheld TLs on landings with both engine's reverse operational. He could not answer me that.
On a simple way, I want to know what are the chances of TLs being misheld when landing my new airplane with one reverse inop. Very low? It would be, scientifically speaking, IF I knew WHY it happens (mishandling TLs), but since I don't know why, I don't know HOW to avoid it. So chances are very high this can happen with my new airplane on the first flight with one reverse inop.
Although the number of flights with reverse problems is impressive, I know that reverse are not essential. If I buy this airplane, I will tell my pilot to consider BOTH reverse inop, when one is "locked". The salesman could not demonstrate to my satisfaction that mishandling TLs has no relation to having one engine reverse "locked".
I will play safe. One reverse inop, no reverse at all. And fix it.
Anyway, the crew I hired has mixed feelings. The Captain prefers I get a B-737, the Copilot prefers A-320. I know which one is more economical to operate, but my wife wants the safest one. Tough call...
This story is pure fictional. Some info are based on facts, the fiction part is me playing the part of a potential buyer, of course...
Again, PBL, thanks for you patience. You demonstrated to be a real gentleman and a Master.
Rob
Great class, I was not trying to teach the priest how to say the Mass. I was trying to teach the Pope...
I'm sorry, I was not chiding you, or better, this was not my intention. After showing so much patience, I can only applaud your effort.
Now, if I may, let me give you an example. With this example maybe I can make myself a bit clearer.
Supose I am a potential buyer of an A-320 type airplane. I know now that around 1% of the flights today will have trouble with reverse. So to help in my decision to buy A or B, I want to know WHY all three (or four) accidents/incidents with probable causes being mishandling TLs on landing, had problems with reverse (locked). So, to convince me this should not be a concern, the "salesman" demonstrate to me that operations with a "locked" reverse are normal (MEL, MMEL, etc,etc...). So I ask him how many accidents/incidents happened caused by misheld TLs on landings with both engine's reverse operational. He could not answer me that.
On a simple way, I want to know what are the chances of TLs being misheld when landing my new airplane with one reverse inop. Very low? It would be, scientifically speaking, IF I knew WHY it happens (mishandling TLs), but since I don't know why, I don't know HOW to avoid it. So chances are very high this can happen with my new airplane on the first flight with one reverse inop.
Although the number of flights with reverse problems is impressive, I know that reverse are not essential. If I buy this airplane, I will tell my pilot to consider BOTH reverse inop, when one is "locked". The salesman could not demonstrate to my satisfaction that mishandling TLs has no relation to having one engine reverse "locked".
I will play safe. One reverse inop, no reverse at all. And fix it.
Anyway, the crew I hired has mixed feelings. The Captain prefers I get a B-737, the Copilot prefers A-320. I know which one is more economical to operate, but my wife wants the safest one. Tough call...
This story is pure fictional. Some info are based on facts, the fiction part is me playing the part of a potential buyer, of course...
Again, PBL, thanks for you patience. You demonstrated to be a real gentleman and a Master.
Rob
Last edited by Rob21; 29th Sep 2007 at 13:13. Reason: typo
The Reverend
Join Date: Oct 1999
Location: Sydney,NSW,Australia
Posts: 2,020
Likes: 0
Received 0 Likes
on
0 Posts
I can not understand why anybody has a problem with retarding both thrust levers to idle on landing, or reverse on the bus, in spite of an inop reverser on an engine that has been locked out.
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by HotDog
I can not understand why anybody has a problem with retarding both thrust levers to idle on landing
In each case the CVR recordings alone prove that the pilots were utterly confused as to what had happened. The Taipei CVR transcript is typical of the others:-
"1959:37 CAM1 No brake
1959:37 TWR (communication between
TWR and UNI831)
1959:39 CAM1 No brake
1959:40 UNI831(communication between
TWR and UNI831)
1959:43 TWR (communication between
TWR and UNI831)
1959:44 CAM1 No brake
1959:46 CAM1 No brake
1959:47 UNI831(communication between
TWR and UNI831)
1959:50 CAM1 No brake at all
1959:50 TWR (communication between
TWR and UNI831)
1959:53 CAM1 Brake
1959:54 CAM2 What’s going on sir
1959:55 CAM1 I have no idea
1959:57 CAM1 Wow
1959:57 CAM (sound similar to impact)
1959:58 CAM (sound similar to impact)
1959:59 CAM chime (single chime)
2000:01 CAM2 Uh
2000:03 CAM1 Wow
2000:03 CAM (sound similar to impact)"
So I for one am very pleased that Airbus have finally started offering an additional warning, and that TAM (and hopefully many other airlines) intend to install it ASAP.
PBL, I tend to agree with you that statistical analysis cannot really help in this area - but for a different reason. Airline travel is statistically so safe, whatever sort of aeroplane you're flying in, that it would be half a lifetime before the figures alone showed that one particular type was significantly any less safe than another.
On the other hand, though, I once had the job of chairing the monthly meeting of the Accident Committee in a town I helped to build. No shortage of accidents there (well, fewer than most towns, as the roads were newer). But if we detected an accident pattern (at a particular intersection, maybe, or on a particular stretch of road) we discussed it, thought about it, often a couple of us physically went out there and walked or drove repeatedly along and around the stretches of roads concerned.
We usually found something simple - a misplaced or obscured sign, a tree that needed trimming or felling, an awkward bend or turn that would be improved by putting in an island, things of that sort. By those means we usually found a way of reducing the number of red crosses on the monthly map in due course; in those locations anyway.
An unscientific approach in the extreme. All I can say is, it tended to work surprisingly well, surprisingly quickly, and surprisingly often.
I very much hope that similar methods (crude and empirical though they may be) are being tried by devoted people at Airbus and at the HQs of the various airlines operating A320s - flying and re-flying 'one reverser inop.' landings on simulators - or even in real aeroplanes - 'chewing the rag' with pilots who've had to carry out the manoeuvre; doing whatever they can think of to find out not just WHAT happened, but HOW on earth it could POSSIBLY have happened.
Last edited by RWA; 29th Sep 2007 at 18:21.
Per Ardua ad Astraeus
Join Date: Mar 2000
Location: UK
Posts: 18,579
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by RWA
it has happened on A320s not once, but three times
Join Date: Jul 2007
Location: Germany
Posts: 556
Likes: 0
Received 0 Likes
on
0 Posts
RWA,
this is all very nice, but how would you look for obvious signs for why a crew would do this mistake in the simulator? Why would you expect to find anything, if 840,000 landings with inop reverser went fine? You would require on the order of that many simulated landings to witness only one case of the problem.
BOAC,
that post by 4holerpoler seems a bit strange, not least because he said that
There is no way on the A320, barring a major malfunction or Alpha Floor Protection (which cannot activate on the ground), that an engine produces thrust above the thrust lever setting.
IF there had been this malfunction, there would have been a report and an investigation, otherwise the pilot would have been in his duties, and the thrust lever mishandling would not have been a causal factor of the incident, except when this malfunction would have been triggered by the thrust lever being "forgotten" in CL.
Bernd
this is all very nice, but how would you look for obvious signs for why a crew would do this mistake in the simulator? Why would you expect to find anything, if 840,000 landings with inop reverser went fine? You would require on the order of that many simulated landings to witness only one case of the problem.
BOAC,
that post by 4holerpoler seems a bit strange, not least because he said that
Originally Posted by 4HolerPoler
the engine that still had it's TL in the Climb detent went to TOGA.
IF there had been this malfunction, there would have been a report and an investigation, otherwise the pilot would have been in his duties, and the thrust lever mishandling would not have been a causal factor of the incident, except when this malfunction would have been triggered by the thrust lever being "forgotten" in CL.
Bernd
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by bieseker
RWA,
this is all very nice, but how would you look for obvious signs for why a crew would do this mistake in the simulator? Why would you expect to find anything
this is all very nice, but how would you look for obvious signs for why a crew would do this mistake in the simulator? Why would you expect to find anything
All I know is, I reckon that I'd have a damn sight better chance of getting useful (meaning preventive) information my way, by legwork and experimentation and even just plain following hunches and having ideas, than I'd have just reading FDR data and carefully working out percentages that prove beyond doubt that three or four almost identical runway overruns with the exact same type of aeroplane, all with identical mech. problems (one reverser inop.), all with highly-experienced pilots, just couldn't have happened - and that 199 people just couldn't have been killed at Congonhas.........
Because all that is easily proved from the figures to be statistically impossible........
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by BOAC
you may want to up that to at least 4? See '4holerpotholer' 23July 19:30 (around post #412 at the moment).
"This, to me, is so relevant; happened to me when I was a brand new skipper, first flight after a month's vacation, new FO, dispatch with one reverser inop. First two sectors went without incident but on the third (FO's leg) I reminded him of the inop reverser. Retrospectively he removed his hand compeltely from that thrust lever, it was midnight, dark cockpit. At the flare & "Retard" call he only closed the "operative" TL. The aircraft squawked "Retard" at least five times and then after the two second latch the engine that still had it's TL in the Climb detent went to TOGA. Aircraft yawed significantly and came dangerously close to the edge of the runway (we were at about 5 feet AGL). I took control, whammed the other TL closed and got the aircraft back on the black stuff. Was a long runway, CAVOK & wind calm. Glad it wasn't on a short, wet strip.
"I know the "wait until the investigation" but that's it for me."
Direct 'echo' from the Congonhas CVR:-
"18:48:21.0 FWC twenty.
18:48:21.6 FWC retard.
18:48:23.0 FWC retard.
18:48:24.5 CAM [sound of thrust lever movement]
18:48:24.9 CAM [sound of increasing engine noise]
18:48:25.5 GPWS retard
18:48:26.3 CAM [sound similar to touchdown]
18:48:26.7 HOT-2 reverse number one only.
18:48:29.5 HOT-2 spoilers nothing."
In both cases, a last-minute warning about not reversing the wrong 'un. The pilot (at Congonhas, anyway) will have had his left hand, his feet, and above all his EYES and 110% of his attention fully occupied on getting the touchdown achieved as straight and as soon as he possibly could, given the wet, not-so-long runway. So, under such pressure, 'Reverse one side only' translates in your mind as, 'Concentrate on the live one, forget the inop. bugger......."
And then the 'Retard' call stops, you get the nosewheel down OK but then there are no spoilers, no autobrakes, the runway is unreeling fast.........and there's no further warning about the throttles, they're past and gone, they don't even occur to you as the possible problem........and your off-sider is new to A320s.......twenty seconds to live.........
Can it really have been that simple?
Last edited by RWA; 29th Sep 2007 at 17:41.
Join Date: Jan 2005
Location: France
Posts: 2,315
Likes: 0
Received 0 Likes
on
0 Posts
PBL and friends,
I'll quote PBL:
Sounds nice, but it has a glaring hole in it.
We're not interested in the "two known non-occurrences of X", we're interested in the real figure for "non-X".
We have no figures for the number of occasions where a (non-X and Y) event occurred as you have defined it, resulting in:
"No spoilers, no brakes"
"WTF?"
"Close the damn throttle, man!!"
...combined with decent weather and a nice long runway.
That would result in some red faces up front, but unlikely to result in a "known non-X event". It would still have been a (non-X and Y) event.
Add a few more of those, and your sample of non-X events is out of the window.
By the time I would start looking at, say, 10 of those non-X events in the Y sample space, I'm down to something like 10 to minus 5.
That's why I'm with RWA here.
Even his idea of sim flights might work...
Brief them carefully and extensively on having to use a different procedure when one TR is locked out. Then throw a well-thought-out sample of sh!t at them (if they know what to expect, it won't work). With a bit of "luck", somebody may make the mistake, and we can ask him what happened.
I'll quote PBL:
At the time of decision about including the "Taipei" warning, there were two known non-occurrences of X amongst, say, 30 million landings.
Rob21's intervention showed he is concerned with what is called the *correlation* between two events. These events are
These events are
* non-X: failure to retard all thrust levers to idle on landing
* Y: at least one thrust reverser locked out; or maybe
* Z: exactly one thrust reverser locked out.
Total size of sample space: 30 million
Non-X events: 2
Y events: 850,000
(Non-X and Y) events: 2
Rob21's intervention showed he is concerned with what is called the *correlation* between two events. These events are
These events are
* non-X: failure to retard all thrust levers to idle on landing
* Y: at least one thrust reverser locked out; or maybe
* Z: exactly one thrust reverser locked out.
Total size of sample space: 30 million
Non-X events: 2
Y events: 850,000
(Non-X and Y) events: 2
We're not interested in the "two known non-occurrences of X", we're interested in the real figure for "non-X".
We have no figures for the number of occasions where a (non-X and Y) event occurred as you have defined it, resulting in:
"No spoilers, no brakes"
"WTF?"
"Close the damn throttle, man!!"
...combined with decent weather and a nice long runway.
That would result in some red faces up front, but unlikely to result in a "known non-X event". It would still have been a (non-X and Y) event.
Add a few more of those, and your sample of non-X events is out of the window.
By the time I would start looking at, say, 10 of those non-X events in the Y sample space, I'm down to something like 10 to minus 5.
That's why I'm with RWA here.
Even his idea of sim flights might work...
Brief them carefully and extensively on having to use a different procedure when one TR is locked out. Then throw a well-thought-out sample of sh!t at them (if they know what to expect, it won't work). With a bit of "luck", somebody may make the mistake, and we can ask him what happened.
Join Date: Sep 2007
Location: Melbourne
Posts: 180
Likes: 0
Received 0 Likes
on
0 Posts
Originally Posted by ChristianJ
That would result in some red faces up front, but unlikely to result in a "known non-X event". It would still have been a (non-X and Y) event.
The most important 'accidents' are the hundreds every day that DON'T happen - because skilful pilots realise what is going wrong, and put things right in good time.
Unfortunately these do not figure in the news stories - leave alone the 'statistics'........