TAM A320 crash at Congonhas, Brazil

[TyroPicard]

I guess that puts me firmly in the 'old' category then.....
I teach 'em with a big hammer until they see it my way...

In which box exactly does the landing fit?

Both! You reduce thrust from approach setting to idle, and A/THR disconnects when both T/L are at idle. Simple, pull back, go up , corblimey stuff.
Tyro

[marciovp]

My old question...

I say that I am not a pro or an expert. Just used to fly Cherokees... But my question remains. Sorry for insisting...:

How come the A320 is built in such a way that ate least six experienced, well trained, pilots make the same mistake when landing with a reverser off at least in three occasions (if not eight pilots in four occasions)? There must be something there that is not unforgiving, not pilot friendly. Or it is just a problem of the six (eight) pilots?...

Again, wrong question?...

[Dream Land]

Again, wrong question?...
Today 21:23

Why did a 737 crew confuse a Cabin Altitude warning with a Configuration warning, both good questions that many on this thread are analyzing thoroughly.

[lomapaseo]

I say that I am not a pro or an expert. Just used to fly Cherokees... But my question remains. Sorry for insisting...:

How come the A320 is built in such a way that ate least six experienced, well trained, pilots make the same mistake when landing with a reverser off at least in three occasions (if not eight pilots in four occasions)? There must be something there that is not unforgiving, not pilot friendly. Or it is just a problem of the six (eight) pilots?...

Again, wrong question?...

Yes wrong question

Please discuss the support for your statement that the pilots were well trained or could it be that is the real problem.

[bsieker]

Originally Posted by Flight Safety

When auto thrust is engaged on the A320, switch arm positions CLB, FLX, and TOGA are available as possible selector switch positions.

To be nitpicking, this is not strictly true. Above CL position, thrust lever position always directly selects thrust. (Above MCT with one engine out.) But I take your point.

In selector switch mode, lever position no longer equates to engine power thus the levers have no direct control over engine power, because they no longer function as throttles in any classic sense.

Even with A/THR engaged the thrust levers function as limiters (perhaps throttles, really), limiting the maximum thrust A/THR may set. Since this is rarely used, thrust levers may seem to be working more as a selector than a throttle during flight.

Perhaps not.

At least one A320 pilot (TyroPicard) said he did not think of them as mode selector switches, but rather thinks of selecting idle thrust when pulling them back at the flare. Others may feel different. I cannot judge that.

I also agree with him that most of the times you actually touch the thrust levers, they work as throttles. As long as they stay in CL and A/THR is active, they're not used, and not thought about much.

Typical thrust lever movements are: Using low thrust during taxi, setting medium power to stabilise engines, setting T/O thrust, reducing thrust at acceleration altitude, retarding to idle, selecting reverse idle, selecting max reverse, back to reverse idle, back to forward idle, low thrust for taxiing. (I also count 10, 12 if you count setting and reducing taxi thrust as two movements, more if you do it multiple times.)

In all cases in which you move the levers, thrust actually changes in the direction of lever movement. For all intents and purposes these are throttle levers.

Modes exist where the same gesture (control action) yields different results depending on system state at a time when your attention is not on system state.

Here's a problem with your reasoning:

Pulling the thrust levers to idle yields the same result, regardless of autothrust being engaged or not: it selects idle thrust. So by your own admission, this is modeless.

Bernd

[Dani]

Flight Safety, I thank you very much for your post. Not because I agree with your analysis, but because you finally steer the discussing into the right direction: The human aspect of moving the TL - or not to move.

After going trough your post one thing stroke me most:

...the pilot leave ENG2 TL in the climb detent after pulling ENG1 TL back to idle 2 seconds after the "retard" call in the flare

I didn't know that, because I never laid the CVR beside the DFR. If this is really the case, then it is one reason more for the argument that they didn't really know how "to fly an Airbus".
Let me explain: On the line in every Airline I have been before there are Airbus pilots that wait for the "Retard" call to close the TL. It's not an exception. I guess it's around 80% of all Airbus pilots that went trought an Airbus procedure course, i.e. Toulouse or others. I also have been in Toulouse, so I know that Airbus says that you should go to idle BEFORE the retard call, since the call is only a reminder. Still, I make this observation on the line, that most of the pilots do it after. The result is often a great increase of flare and landing distance.

I always stand under the impression that pilots who do it this way are somewhat not pilots anymore but rather remote controled robots in an "executing mode", not acting like real pilots but listening to orders from the computer and then acting.

If you look at the FDR printout again, you see also very interesting facts in the last landing of the crew in Porto Alegre, executed by the captain in the right seat: Both TL go to Reverse! (although the Reverser was faulty already)

And when you look at the second last landing, in Congonhas (most probably done by the captain in the left seat again): Already there he reverses only one engine. When he does he doesn't retard both TL at the same time! He starts about half a second to retard TL1, then slows down, starts to retard TL2. But then TL 1 needs more than a second (!) more to go to idle than TL2. If you remember how quickly normally people retard engines on an Airbus (ever heard that clacking noise against the metal?), you clearly see that he had problems to fiddle around with these TLs. He most probably spent a lot of time and effort to carefully take one TL and then another because he wasn't really sure about what to do with them.

The fact that the TAM captain did it the same way twice lead me to believe that he was not ahead of his task. They didn't know what to do with that TL, and hence did the wrong thing. He was somehow in the above described "remote controlled" mode, not acting as a conventional pilot anymore.

We have to put more emphasis on conventional pilot training in Airbus aircraft. By using always this "Managed Mode", pilots become submissive to the machine, forgetting basic flying priciples and standard airmanship. We have to include the flare into the approach briefing (like a lot of airlines do): What type of Airbus we fly, what weight it has, where the wind is blowing from and what basic flare concept I will follow. Some mental preparation is needed. And it is needed before top of descent, not at the treshold.

Otherwise if pilots still neglect basic flying principles, I do not see any place for them in a cockpit - period.

Dani

[marciovp]

Insisting... sorry.

Why did a 737 crew confuse a Cabin Altitude warning with a Configuration warning, both good questions that many on this thread are analyzing thoroughly

Three times, with three different crews, the same problem?

Please discuss the support for your statement that the pilots were well trained or could it be that is the real problem.

Good point. WIll not know for sure. Could it be that all three crews had pilots that were not well trained? Tam had the first officer that flew TAM since 1987 and landed safely in Porto Alegre. Of course the second officer had experience with Boeings 737 and 767 but he trained for six months in TAM before being a Comandant. Both had about 13000 hours flying time. In the other accidents I don´t know. But to make my answer short: yes there is a possibility that these six pilots or some of them were not well trained... I resist accepting this coincidence... but yes...it could be.

[GMDS]

PBL:

I introduced Cat. e (rabble-rousing). BOAC tried to introduce another Cat. e. Now you want to try yet another Cat. e. What is this thing with the letter "e"? Could we maybe try "f"?

Sorry PBL, I didn't recall taht sequence. Not wanting to go for the "f", I'll settle for the g-spot then.............

ChristiaanJ:
It might seem that I long for the good old days, like grandpa, but there's more to it. I am a defender of proven designs, not against innovation, but only the ones that bring along improvement. A good example were the new tachymeters in the Auto industry. Citroën tried the digital display first in the early 80's, others followed and today almost all manufacturers went back to the original clock design. Why? They realised that the average Joe and Jane controlled speed better with the classic layout and less accidents were traced to misinterpretation. So, no improvement of a new design, no false pride of manufacturors but good analysis and the old one prevailed. It now seems that the fixed throttle design led to more confusion (just re-read the thread!!) than to any improvement, which might allow the assumption that logic and airmenship of the manufacturor could lead to a reversal to the proven moving throttle design.
Now why do I assume that this is not going to happen?

[bsieker]

Originally Posted by Dani

I didn't know that, because I never laid the CVR beside the DFR.

It has been done for you, on page 12 of the FDR graphs. They have the CVR ttranscript superimposed.


[Re. retarding thrust levers after "Retard" call.]

I guess it's around 80% of all Airbus pilots that went trought an Airbus procedure course, i.e. Toulouse or others.

That was my impression, having seen several cockpit videos of A320 landings. Not one is without the retard call, the usual number varies between 2 and 5 callouts.

So I assumed it was standard procedure to wait for the call. FCOM says:

. At about 20 feet:
- FLARE ..... PERFORM
- ATTITUDE .... MONITOR
- THRUST levers ..... IDLE
[...] "RETARD" callout is generated at 20 feet RA.

So wouldn't getting one or two calls be normal, when 20ft is both the normal flare altitude and the callout altitude?

(It usually goes like this: "fifty - forty - thrirty - twenty RETARD RETARD"), i. e. if you start flare at the "twenty" callout and then pull the levers back, you'd still get the RETARD call. It'd be nice to see at least one A320 landing video without the call

And when you look at the second last landing, [...] When he does he doesn't retard both TL at the same time! He starts about half a second to retard TL1,

The half-second difference is an artifact of the sampling rate (once per second, alternating between left and right lever).

Allowing for that, at the previous landing, he most likely pulled the thrust levers exactly together. And retarded them to idle quickly.

then slows down, starts to retard TL2. But then TL 1 needs more than a second (!) more to go to idle than TL2.

For all we know, TL2 may have been pulled back very quickly, at one sample it's at CL, the next sample it's at IDLE. no way of telling what exactly happened between the two samples.

But I agree, very strange that he didn't pull them to IDLE together.

If you remember how quickly normally people retard engines on an Airbus (ever heard that clacking noise against the metal?), you clearly see that he had problems to fiddle around with these TLs. He most probably spent a lot of time and effort to carefully take one TL and then another because he wasn't really sure about what to do with them.

Sounds like a possible explanation, and certainly deviation from trained procedures. "Most probably" I don't know.

Bernd

[Dani]

Thanks Bernd for that insight information about recorders.

Where do you have your FCOM3 from? Mine says at 30 feet!

LANDING:

At approcimately 30 feet:

Flare ..................................................perform
Attitude...............................................monit or
Thrust levers........................................idle
The retard callout is generated at 20 ft.

Anyway, ANY landing on any aircraft starts at 50 ft. It doesn't mean that you start your flare or your idle there. But most definetly over the treshold ends an approach and starts the landing.

I'm always surprised that even very experienced heavy jet driver wait for very long to retard a trust lever (on any aircraft). It comes from the misconception that you have to keep V appr (Approach Speed) until touchdown. But this is not correct, on a perfect landing you land with slightly lower speed. But somehow they cannot accept that the speed drops below the speed bug.

This is very pronounced in people with turboprop background where they - depending on type - have bigger problems for a smoother landing with lower power settings and low tilted propellers.

btw, the "Retard" call always comes on, even when you have retarded already. Since I normally retard above 30 ft (depending on weights, winds aso), I can verify that on every flight. Sometimes you also get "wild" retard calls, like any other call, like uneven radio call outs. This is not always a sure hint.

Dani

[DozyWannabe]

GMDS:
A good example were the new tachymeters in the Auto industry. Citroën tried the digital display first in the early 80's, others followed and today almost all manufacturers went back to the original clock design. Why? They realised that the average Joe and Jane controlled speed better with the classic layout and less accidents were traced to misinterpretation.

The problem with this, and a lot of other analogies comparing the car industry with aircraft manufacturers is that you end up with an apples-to-oranges comparison. There is a world of difference between a type of display that is technically fancier but turns out in practice to be harder to read than the traditional mechanical dial, and the throttle quadrant of a commercial airliner.

For a start, you have to look at the reasons for the change - the A320 was conceived as a completely modern machine, looking at what was possible with state-of-the-art technology to make the pilot's job easier and the aircraft more economical to operate and maintain. You can bet any amount that you like that AI weren't going to sink the millions required for a project of this magnitude without a significant amount of R&D on the engineering, customer and pilot's view of how this thing would work.

As I said before, modern technology meant that engine control no longer required a physical connection between levers and engines for throttle control - in fact statistically it behaved more reliably than the old arrangement, simply because digital control is not subject to physical entropy through wear and tear in the same manner. To backdrive the levers would require an artificial force-feedback system that would not only incur a weight penalty, but also compromise the simplicity of the system, incurring a maintenance cost penalty as well. At the time, none of the pilots involved in the project objected to the non-moving design that I've learned. Unless proven beyond a shadow of a doubt that non-moving throttles are dangerous - and I'm talking about more than hearsay and conjecture from a small but vocal group of pilots, a majority of whom haven't even flown the aircraft they're criticising.

In this case, I'm not even sure that moving throttles would have made a difference, as the evidence points to a lever that was deliberately left out of position. Even if the levers were of the moving type, the pilot may still not have selected the incorrectly positioned lever to idle, the only difference is in how far he would have had to move it to put it there. Of course, I think that Rananim is talking about the secondary levers that control reverse with the physical interlock. A very clever system, but again, one that goes against the design brief to keep it simpler from a maintenance perspective.

I can't answer ChristiaanJ's question regarding Concorde's system, because I don't know the designers involved on both projects, nor can I find any material about what informed their decision to go one way in the '60s and another way in the '80s.

As an aside, one thing I have noted during my time on here is the number of A320 sceptics who come from the other side of the pond. Would their objections be so vehement if it was Boeing, Douglas or Lockheed that designed and built a similar aircraft to the A320?

[GMDS]

The problem with this, and a lot of other analogies comparing the car industry with aircraft manufacturers is that you end up with an apples-to-oranges comparison

It is in the nature of a analogie not to treat equal things, otherwise it would be a simple comparison. Analogies are used to clarify a idea or thesis. The setback is that people who don't want to understand will always bring up the apples and oranges.

As I said before, modern technology meant that engine control no longer required a physical connection between levers and engines for throttle control - in fact statistically it behaved more reliably than the old arrangement, simply because digital control is not subject to physical entropy through wear and tear in the same manner. To backdrive the levers would require an artificial force-feedback system that would not only incur a weight penalty, but also compromise the simplicity of the system, incurring a maintenance cost penalty as well.

A equally modern design (777) has just as little physical connection to the engine. But you pointed it out nicely: It must be the cost penalty that inhibited AI engineers to consider the backdrive depriving us of a (my interpretation) extremely useful feedback in case of screw up or confusion (as in this incident).

Just to comfort you: When I flew a MD product I was heavily involved in a fervent criticism/discussion about a lack of design/checklist. It has nothing to do with anti-AI feelings. I simply strive for the best pilot-machine interface available and strongly believe that the throttle design with feedback from autothrottle commands is safer than the one without.

Safe flights

GMDS

[DozyWannabe]

A equally modern design (777) has just as little physical connection to the engine. But you pointed it out nicely: It must be the cost penalty that inhibited AI engineers to consider the backdrive depriving us of a (my interpretation) extremely useful feedback in case of screw up or confusion (as in this incident).

Hmm - your phraseology implies that it was monetary cost to Airbus I was referring to, when I was referring to cost in engineering terms, a very different meaning of the word. You should be aware that the anti-AI crowd make hay by ignoring that kind of distinction.

All the differences in implementation prove is that the engineers and pilots on the 777 project came to a different set of conclusions than their counterparts on the A320 project. No doubt both implementations have pluses and drawbacks.

[ChristiaanJ]

GMDS et al,

To backdrive the levers would require an artificial force-feedback system...

There seems to be a basic confusion here, that I've noticed in several other posts as well.
In "conventional" systems, throttle angle is NOT representative of actual engine thrust but of the commanded engine thrust. There is no feedback.

On Concorde (and I presume more recent aircraft as well), the autothrottle computer commands an electro-mechanical unit (inside the pedestal) which directly moves the thrust levers through clutches, rather than electrically signalling the ECU. There is no "artificial force-feedback system".

BTW, the autothrottle unit on Concorde weighed about 7 kg, so a similar unit for the A320 would have weighed about 4 kg. Hardly prohibitive.....

[GMDS]

There seems to be a basic confusion here, that I've noticed in several other posts as well.
In "conventional" systems, throttle angle is NOT representative of actual engine thrust but of the commanded engine thrust. There is no feedback.

Come on ChristiaanJ, we're not as naive as that. Nobody requested a exact feedback of the actual thrust. The request, relating to this incident, would be more of a feedback of the intended command, i.e. "it wants to incresase thrust" as opposed to "it wants to retard thrust". This would have maybe triggered the poor pilot to close the second lever as well, because he would have FELT the lever moving forward as the autothrottle suddenly commanded more thrust, and not idle. In such critical moments, apparently the visual feedback from ECAM proved insufficient.

[DozyWannabe]

ChristiaanJ:
BTW, the autothrottle unit on Concorde weighed about 7 kg, so a similar unit for the A320 would have weighed about 4 kg. Hardly prohibitive.....

True - so there must have been another factor that influenced that decision. The only way to get a definitive answer would be to ask anyone who worked on the A320 cockpit design team. Unfortunately my only contact with anyone involved extends only to software, so I can't help there.

GMDS: The throttle did *not* command more thrust until late in the roll, and then not a significant amount. The thrust remained constant (approach thrust) until the aircraft started decelerating.

[GMDS]

GMDS: The throttle did *not* command more thrust until late in the roll, and then not a significant amount. The thrust remained constant (approach thrust) until the aircraft started decelerating.

I actually don't care WHEN it commanded more thrust: As a matter of fact it did.

I actually don't care HOW MUCH more thrust it commanded and for how long and how constantly: As matter of fact it did and a movement would most probably have been detected ....

Any twist or torsion of grammar or expressions does not avoid the tormenting thesis, that a throttle that would have moved, might just have helped picking up the glitch by the guys in this situation. All the "they should haves" or the "if they would haves" do not help here. Anyone might just screw up one day, maybe even Dani, and at that specific moment one would wish to have the best design to help mending fences.

[DozyWannabe]

GMDS:
Anyone might just screw up one day, maybe even Dani, and at that specific moment one would wish to have the best design to help mending fences.

But "the best design" is itself subjective, unless you want to foot the bill for the HF research to prove it scientifically.

[ChristiaanJ]

GMDS,

Come on ChristiaanJ, we're not as naive as that. Nobody requested a exact feedback of the actual thrust.

Not exactly, but several posts did equate TL position with actual thrust rather than commanded thrust, to the extent that in a reply somebody had to point out that the TL wouldn't budge even if the engine flamed out.

The request, relating to this incident, would be more of a feedback of the intended command, i.e. "it wants to increase thrust" as opposed to "it wants to retard thrust".

My issue was with the word "feedback", since strictly speaking there is (or should be) none. The pilot positions the TL to set a thrust demand, the TL position is converted to an electrical signal and sent to the ECU (or FADEC or whatever), and the latter controls the fuel flow to meet this demand asap. At no time is there a feedback.

Having the autothrottle move the TLs mechanically actually simplifies the system slightly, since the command to the ECU always comes from only one source: the TL position sensors.

[SyEng]

TL Mechanism

Some posters here have jumped to the conclusion that the PF left No.2 TL in the climb detent. That may very well be the case. However, all that can be inferred from the FDR trace is that the FDR continued to record TLA at CLB. I would guess that this is the signal from the EEC in control (1 of 2).

This is how the thrust lever mechanism works: the thrust lever moves a control rod which moves an input/output lever on the artificial feel unit (this provides friction to stop the TLs from wobbling about plus the detents and stops). The other end of the artificial feel unit input/output lever moves another rod which moves the input lever on the throttle control unit. The throttle control unit contains the RVDT resolvers which provide an anologue TLA signal to the ECUs. All of this lot is in the centre pedestal.

Each end of each control rod is located to its lever by a nut and bolt. Each nut is locked by a split pin. There are 4 such nut/bolt/split pin connections between thrust lever and throttle control unit.

Failure of either end connection on the artificial feel unit to throttle control unit rod would result in a normal-feeling TL but a frozen and valid TLA output from the throttle control unit. (Caveat: the TCU incorporates a "safety device" which drives the resolver out of range if the "driving device" fails. I believe this refers to the TCU internal driving device rather than a failure of drive into the TCU but my source runs out of detail here. Will attempt to investigate further.)

Failure of either end connection of the thrust lever to artificial feel unit rod would result in a floppy thrust lever ("look this" ? ) but a frozen and valid TLA output from the TCU.

In addition to the above there are the following possible failure modes:
- failure of the thrust control unit mountings
- failure of the artificial feel unit mountings
- unsignalled mechanical failure of the thrust control unit
- mechanical failure of the artificial feel unit

To my mind, the most likely of the above is a failure of one of the rod-end nut/bolt/split pin connections.

To those who have followed this post so far, firstly congratulations, and, here are some questions.

- Was there any recent maintenance work on the TL linkages?
- What physical evidence is there from the wreckage of the state of the TL mechanism?
- Might a proper spectoscopic analysis of the CVR recording show only one, or both TLs hitting the idle stop?
- Why is the design such that the primary connection is to the "secondary " feel unit function and the secondary connection is to the "primary" control function? Why not have the RVDTs on the TL pivot?

No doubt the official investigation will address all of the above in some detail.

Now, those of you who are not engineers/analytical types, don't get me wrong. The above is not my proposed "explanation" of this accident. It is very, very unlikely that the failure modes that I have identified here would occur. It is also very unlikely that an experienced pilot would neglect to retard the throttles in the flare. My point in posting this is simply that these failure modes need proper investigation. With the evidence available to the investigators it may be possible to rule them out.