Air buss loss at Paris Airshow?

Greg Farris schrieb:

accident was classic pilot error, no fault was found with the FBW system...

The computer was not only not at fault, but it even prevented a catastrophe.

The automatic wing leveller kept the wings levelled, even at full stall.
It did this so perfectly that when the airplane descended into the
trees, both wings were ripped off simultaneously, and the fuselage
"landed" straight ahead and perfectly aligned. From the 120 or so
passengers, *no* one was killed or even severely injured at the impact!
This is a plain miracle.

The three deaths of that accident were not killed at impact, but very
tragically later: There were two passengers in wheelchairs who couldn't
leave the airplane by themselves after the impact. One flight attendant
got aware of this after she had left the plane and returned back into
the plane to help the two to get out. Tragically, at that moment, the
developing fire reached the fuselage and the three were trapped and burnt.

Stefan

"Kev" wrote in message
oups.com...

Danny Dot wrote:
In about 1990 Airbus did low pass at the Paris airshow and lost the
plane.
I recall it had something to do with the throttle software thinking the
pilots were in landing mode and "refused" to go to high power for the
go-around.

Yes, the computers did think the pilot was landing, but the crash was
caused by his being too low and slow. (See other posts for more info
on the latter.)

The Airbus software has modes where its flight control computer laws
are quite different. Some of those computer laws are divided into
Ground, Flight and Landing (Flare) phases.

One claim is that he was trying to demonstrate that the airplane was
unstallable. He had reportedly done this demonstration several times
before at a slightly higher altitude, and it had always worked. Why?
Because the Airbus has what's known as Alpha Protection (pitch related)
and Alpha Floor (thrust related). Too little thrust, at too high an
angle of attack (AOA), and its computers automatically kick in and
override the pilot.

The reason the automatic protection didn't work this time was because
he went below 100', so the computers switched to Landing Mode. That
doesn't mean they do an autoland. It means they think the pilot is
landing the plane and their rules change. The Alpha Floor is disabled
so that a landing is possible at all. By the time the pilot advanced
the throttles himself, it was too late.

In addition, another Landing Mode kicks in when the Bus passes below
50' going down to 30', as he did. The computer starts changing the
stick reference for landing, so that if you have the stick pulled
back', that position soon becomes the neutral spot. This is supposed
to force the pilot to pull back more for flaring.

Regards, Kev

---------links snipped----------

That was not the first or last time that a flight crew got into trouble with
a new control system that they only partially understood; and I am sure that
there will be more to come.

Just as an example, Eastern Airlines lost one of the early Lockheed L-1011
aircraft in the Florida Everglades due to a chain of events which began with
a failed indicator lamp for the nose wheel. The new feature, in the
experience of the crew, was that the autopilot could be dissengaged by a
sharp pull of push on either yoke--and would remain dissengaged until either
pilot engaged it again. That was both a safety feature and a convenience
feature, since it did not require a crew member to continue to forcibly
override a rachetting capstan until the autopilot could be dissengaged. But
the crew did not fully understand the feature, or all of its implications,
at the time. Further, autopilots do operate the trim, but not perfectly,
and they happened to be trimmed very slightly nose down; with the result
that the aircraft gradually drifted down until the wheels contacted the
vegetation and water--and dug in. There were no outside visual referenced
at that place and time--and they still believed that they were on autopilot
at a constant altitude.

That's just one more relatively famous accident.

Peter

Stefan wrote:
From the 120 or so passengers, *no* one was killed or even severely injured at the impact!


There were pax aboard that flight? I thought it was a mfr test hop?

Kingfish schrieb:

There were pax aboard that flight? I thought it was a mfr test hop?

It was a demonstration flight on an airshow. This was the last such
flight in Europe with passengers aboard.

Stefan

"Stefan" wrote in message
. ..
Kingfish schrieb:

There were pax aboard that flight? I thought it was a mfr test hop?

It was a demonstration flight on an airshow. This was the last such
flight in Europe with passengers aboard.

Stefan

It used to be fairly common for an airline with a nearby base to have one of
their aircraft make a low pass over the runway at an airshow--sometimes slow
and dirty (as this was planned to be), other times clean and fast, and
occasionally one of each. It was a way to showcase their presence, and also
provided an interesting contrast of sizes and characteristics...

Peter

Kev writes:

The Airbus software has modes where its flight control computer laws
are quite different. Some of those computer laws are divided into
Ground, Flight and Landing (Flare) phases.

One claim is that he was trying to demonstrate that the airplane was
unstallable. He had reportedly done this demonstration several times
before at a slightly higher altitude, and it had always worked. Why?
Because the Airbus has what's known as Alpha Protection (pitch related)
and Alpha Floor (thrust related). Too little thrust, at too high an
angle of attack (AOA), and its computers automatically kick in and
override the pilot.

The reason the automatic protection didn't work this time was because
he went below 100', so the computers switched to Landing Mode. That
doesn't mean they do an autoland. It means they think the pilot is
landing the plane and their rules change. The Alpha Floor is disabled
so that a landing is possible at all. By the time the pilot advanced
the throttles himself, it was too late.

In addition, another Landing Mode kicks in when the Bus passes below
50' going down to 30', as he did. The computer starts changing the
stick reference for landing, so that if you have the stick pulled
back', that position soon becomes the neutral spot. This is supposed
to force the pilot to pull back more for flaring.

These are all examples of poor design. It's extremely dangerous to
put such features into FBW software unless you can be certain that all
pilots will know as much about the software as the designers did, so
that they'll know all the modes and all the combinations and
permutations of possibilities.

If you aren't going to let the pilot control the plane, why have a
pilot at all? And if the computers are going to second-guess the
pilot's intentions every step of the way without the pilot knowing it,
having a pilot is worse than not having one.

This is a good example of poor human-machine interactions, created by
poor and inadequate design.

--
Transpose mxsmanic and gmail to reach me by e-mail.

Danny Dot wrote:
In about 1990 Airbus did low pass at the Paris airshow and lost the plane.
I recall it had something to do with the throttle software thinking the
pilots were in landing mode and "refused" to go to high power for the
go-around.

Anyone remember the details???

Danny Dot

This really comes down to a simple difference between the way the
autothrottle operates in Boeing airplanes vs. Airbus.

In the Boeing model, if you shove the throttle handles forward, the
engines respond to the throttle setting. If autothrottle is engaged,
it will start to retard the throttles back to the commanded setting and
the pilot will see the throttles moving, realize that autothrottle is
engaged and trying to do something different than what he wants, and he
will grab the throttle handles, shove them back forward and push the
autothrottle disengage button on the side of the throttle handle. This
takes only a second or two to resolve.

In an Airbus model, if you shove the throttle handles foward, the
engines do not respond to the throttle setting (when autothrottle is
engaged), and the throttle handles will just sit there in a position
that differs from what the autothrottle is doing. In the meantime, it
can take a while for the pilot to realize what is going on and the push
the TOGA button to disengage the landing mode and get the computer to
respond to the throttle handle position. This is what happened to the
pilot involved in this accident. By the time he realized what was
going on and pushed the TOGA button, there wasn't time for the engines
to spool up enough to miss the trees.

This is why the difference between a chainsaw and an A320 is 100 trees
a minute.

Dean

Mxsmanic wrote:
These are all examples of poor design. It's extremely dangerous to
put such features into FBW software unless you can be certain that all
pilots will know as much about the software as the designers did, so
that they'll know all the modes and all the combinations and
permutations of possibilities.

For once, I don't see how anyone could disagree. The FAA commissioned a
study of these issues. Check out the site map and the Find All Issues
section.

http://www.flightdeckautomation.com

If you aren't going to let the pilot control the plane, why have a
pilot at all?

Boeing thinks along those lines. They have soft limits vs. Airbus'
hard limits.

And if the computers are going to second-guess the
pilot's intentions every step of the way without the pilot knowing it,
having a pilot is worse than not having one.

Well, not really of course. Interestingly, it seems that a lot of
pilots love the Airbus overrides. They can flick the handles and not
worry too much. Others seem to want to rely more on the usual stick
and rudder skills. You could argue both ways.

Computers aren't all bad. Consider all the traction control etc
computers in cars these days. (I actually don't like them
second-guessing me sometimes, but for the majority of drivers they're a
good thing.)

This is a good example of poor human-machine interactions, created by
poor and inadequate design.

I agree that letting programmers decide how things work is often a Bad
Idea.

Kev

Kev writes:

Boeing thinks along those lines. They have soft limits vs. Airbus'
hard limits.

I agree strongly with Boeing.

The whole justification for a pilot is to have a human being aboard
for unexpected emergencies. Computers are superior to human beings
for handling _foreseen_ emergencies, so pilots aren't needed in those
cases. They are also superior for controlling normal flight. The
real advantage of a human pilot in cockpit, though, is that human
beings are extremely good at dealing with completely off-the-wall
situations for which computers haven't been programmed in advance. If
an aircraft finds itself in an unusual attitude that the programmers
of the on-board computers have not foreseen, the computers will
probably make catastrophically poor decisions, or they will just fault
or reboot. A human being, on the other hand, will adapt and make
decisions that are at least moderately appropriate to the situation,
no matter how bizarre.

Given this, not allowing the pilot to override the computer makes no
sense at all. Why bother with the pilot, if all he can do in a
situation that the computer doesn't understand is watch himself and
his aircraft plummet towards the ground?

To make a pilot useful, you need a button that says "I have the
controls," and turns the computer completely off. The pilot is then
on his own, but in some situations, that may be what saves the day.

Computers aren't all bad. Consider all the traction control etc
computers in cars these days. (I actually don't like them
second-guessing me sometimes, but for the majority of drivers they're a
good thing.)

Computers aren't bad at all. Look at the widespread use of autopilots,
which are simply a type of computer. But you need to be able to
switch off the computer, just as you can disengage the autopilot,
otherwise the time may come when you'll watch helplessly as your
computer kills you and destroys your aircraft with poor decisions.

--
Transpose mxsmanic and gmail to reach me by e-mail.

Greg Farris writes:

Well - we are getting much closer to a defensible response here.
We can accept then that the plane crashed because the pilot flew too low and
advanced the throttles too late. Contributing factors include the pilot's
incomplete comprehension of the aircraft's systems. What the investigation had
to determine then, was whether the pilot's poor comprehension was due to poor
application of his training, or whether the training itself was inadequate.

There's another possibility: The designers of the FBW system had no
clue as to how real pilots react in different situations, and failed
to anticipate what a pilot would do and expect in those situations.
If they had done their job correctly, the FBW would do exactly what a
pilot would expect it to do, and there would be no "modes" for a pilot
to memorize over and above everything else that he already has to
know. Flying isn't a video game, even if some desk-bound geeks at
Airbus might like to pretend that it is.

If a pilot, faced with an unexpected situation does something other than what
his training suggests, and the result is positive, then nothing is said. But if
the pilot does not act in accordance with his training and the result is
negative, then it is fair play to attribute it to pilot error.

What happens if the aircraft is designed to do something
counterintuitive, such as having the movements of the yoke reversed,
and the pilot forgets this (or is never trained about it) and makes a
mistake that leads to an accident? Is it the pilot's fault because
the aircraft behaved like no other that he has ever flown, or the
manufacturer's fault because it designed in features that were in
direct contradiction of a pilot's normal base training?

But then, all of the above is moot when one considers, as resident experts have
explained to us, that Airbus aircraft are not controllable through pilot input
...

They are not controllable outside an envelope that is enforced by the
computers. In this case, you have to wonder just exactly why pilots
are needed at all. If all flying situations are covered by the
computers, the computers can fly the aircraft from start to finish,
and you can dispense with pilots.

This will probably actually happen one day for commercial airliners,
although that day is still quite far away. Pilots of airliners are
increasingly just skilled attendants, not people who actually fly the
plane.

--
Transpose mxsmanic and gmail to reach me by e-mail.