JD-EE

rudderrat and takata, when rudderrat said or quoted "The elevator orders are progressively transferred to the THS through a low-speed integrator to decrease the drag." was this necessarily correct? Is that from documentation or a presumption?

If I think about the apparent intent of the THS it aims to keep the elevator's mean position at neutral. So it's really integrating the elevator's position not the stick input commands. It will continue to drive the THS downwards as long as the PF does not push the stick hard enough for the elevator to go nose down, at which time THS would move. One comment I received suggests it would move nose down faster than it had moved to full nose up.

Lyman

So I see your position from flying. Nevermind my question.

Pilot 'sees Nose Down, and right drop. First thing he thinks is, not does. "I might need to stop that." Stop only, not reverse. That is next, maybe. He is too much the big picture, not enough, gentle, and patient.

He is trying for UP UP UP, if he sees this bird, and sees the bird UP UP UP, he thinks, 'more is wrong than at first I thought'? He is not the one to be flying, a new picture makes him more confused and more stubborn, some one else please to fly, maybe the one who asks, can we go down, now?

When "everything" wrong, the "right" thing can only make this worse, new pilot please. Time later for apologies and courtesy. And perhaps to be fired? get new job.

JD-EE

As a thought experiment I started looking at the data available to the computers that inform the pilot (and autopilot system component) with information about the plane's position, velocity, acceleration, orientation, rate of orientation change, and the rate at which the orientation rate is changing in three dimensional space.

They have GPS.
They have at least one very good ring gyro system and perhaps several more prosaic inertial reference units.
They have pitot tubes.
They have air pressure sensors for altitude.
They have AoA sensors.

And I'm tired enough just now I am probably forgetting something. Modern computational techniques (Kalman filters) can take all these inputs, known data about their goodness for measuring the data about the qualities for various aspects of their data. GPS measures the position (and time) the plane had in three dimensional space at some point in time relatively recent depending on the time scale involved, closer to milliseconds than seconds. It says nothing about velocity without considering the differences of at least two readings and ideally a string of readings for averaging. Accelerometers are good with the rate of change for the plane's three dimensional velocity. And so forth.

Presume some information is lost. That Kalman filter can adjust its parameters to reject or at least partially disregard data that is not present. This degrades the solution; but, you still have a pretty darned good idea of what the plane is doing.

Take the data and work back to extract more data, specifically wind speeds acting on the plane in three dimensions. (An AoA vane on the top of the plane might be handy for deriving a cross-wind data independent of calculated guesswork, by the way.)

Now we step off the cliff. What is the largest wind speed change a plane in normal (non-hurricane) conditions likely to encounter translated to a change in airspeed as measured and corrected? Is it 60kts, 75kts, what? This number gives you a wind velocity change range in which to look for a solution. The plane is without airspeed indication it is willing to accept. But it still knows its orientation and inertial velocity with cross checks from GPS. If flying pitch and power is sufficient to keep the plane flying I suspect AoA and power is even better. The plane knows the AoA. It knows the power. It knows attitude. It knows position. It can work backwards from the pitch/power figures to figure out what airspeed, actually what air velocity set, could have the plane moving the way it is with the pitch and power settings it also knows. It calculates a substitute airspeed of quite remarkable accuracy. (I'd then judiciously feed this back into the Kalman filter for a quality check on pitot readings.)

Presuming no CPU cycle limitations there is no reason for the plane's computers to decouple from most of the automation control of the plane. In fact I am willing to assert that the plane could actually have continued on to Paris and even landed on the mark under computer control with aggressive enough software.

At the very least that suggests even a modest '386 class machine should have been able to guess airspeed well enough to be able to issue a quite adequate stall warning at all times.

I am somewhat bemused by the APPARENT fact that the flight displays only relied on a very small subset of the sensors available. I'm sitting here idly wondering why these computers did not make use of the full suite of adequately accurate instrumentation available on the plane.

(And I suspect I ignited a stink bomb with this observation here among a batch of pilots, mostly good ones at that.)

3holelover

It drives the same speed in both directions, but from it's zero point it's got further to go nose up than it does down.
One of the tests we do on the THS involves testing it's auto function as well as that it'll stop auto-functioning when the wheel up front is held. We position a man in the tail to verify a valve moves to close and the THS actuator stops.

Smilin_Ed

JE EE: Except AoA isn't very useful at cruise airspeeds. At cruise speeds, AoA varies much less per knot of airspeed than at approach speeds. Trying to fly AoA at cruise speeds is like trying to balance on top of a bowling ball. Not easy.

On the other hand, when trying to prevent, or recover from a stall, AoA is exactly what you need.

Smilin_Ed

Tyro: Without any feedback in the stick, it makes it hard doesn't it? The only thing you can do is to try it one way or the other and see what happens, but I sure don't want the system changing what I had set. If I'm trimmed in pitch and have the power set right in cruise, I really don't want to change anything. All I want to do is make very gentle changes to hold altitude but I don't want to change the trim.

3holelover

Why not leave the stick alone and fly with trim?

Smilin_Ed

Because in turbulence you want to preserve that trim setting which will cause the aricraft's positive longitudinal stability to get you back to a correct AoA. Several pilots here have said that simply letting go is the classic way to get out of a difficult situation, and that is true, but once the trim is changed, letting go is no longer an option. In fact, with the neutral longitudinal stability brought about by the autotrim (when hand flying) if you let go, the airplane will hold the existing pitch at one Gee and if it is nose up, you will either mush along or stall when you run out of kinetic energy. Several posters here have commented that the Bus has wonderful positive longitudinal stability. We should take advantage of it.

HazelNuts39

Interesting as personal preferences as to what autotrim should do or not do may be, I fail to understand what relevance to AF447 these have. Between 02:11:40 and 02:12:10 the PF pulled the sidestick fully back. If at any time in the descent he had pushed the SS fully forward 30 seconds and the THS had not moved, then it would have been an issue in AF447.

CONF iture

HN39,

At time 02:11:40 was already mainly full NU.
The relevance is that the trim moved to that position under STALL WRN, it is true to state that the inputs were more UP than DN :



... but is it unreasonable to think that the PF would not have manually trimmed or at least to that extent ... ?
And if he had, then I would not show up here to reply to your post.

As I said earlier, leave to the pilot that responsability to trim if that's really what he was looking for.

Lonewolf_50

mm43, thanks for that trace. It painted a very sad picture for me. At around 2 and 13, you had two pilots making control inputs, and given the way the SS inputs are summed, PNF's inputs (though this was a very late in the game state, maybe not recoverable by that point, somewhere around 10000 feet) of nose down were nulled out by PF.

Horror of horrors, to me, is that even after the control change to LHS the RHS was still making SS inputs.

Lyman

takata has told the THS is inop for trimming up with accelerations above 1.25 g. It will not Auto trim when 'manouerving with this load or more. Elevators only.

Looking at the track of g just after pilot 'Ihave controls.' The Ths has not moving during all g above 1.25, so we should be thankful it has courteously not cause even a faster crash. This THS did as was told. Too bad the pilot isn't counting these chickens.

Lyman

The Ths became the factor mafter excess g load was lost in the climb. It is inhibited above 1.25 g for trimming. So up til the time of Ths starting to trim up (plenty of time to Sa and calm down and do as pj2 says.)

At this much g in the first climb, it wants a new display to tell the pilot "nose down, please?"

Human factors? Fancy words for manhandling the Poodle?

CogSim

Quote:

Originally posted by spagolia:

Consider this bit of conversation on page 90 of the report:

Watch your speed
Watch your speed
Okay, okay okay I’m
going back down
Stabilise
Yeah
Go back down

What does "Go back down" refer to? I think it refers to KIAS not altitude. If so, aft stick fits in with the conversation.

Again on page 95:

The speed?

One second later:

You’re climbing
SV : “Stall, stall”
You’re going down
down down
Am I going down now?
Go down
No you climb there
I'm climbing okay so
we're going down

I think "climb" refers to altitude and "going down" again refers to KIAS. Otherwise, the conversation doesn't make sense.

On page 96:

Amazingly this is the first mention of the word altitude.

What do we have in
alti?

And both the Captain and PNF react with total surprise.

(...) it’s impossible
On alti what do we
have?
What do you mean on
altitude?

On page 98:

Nine thousand feet
Climb climb climb climb
No no no don’t climb

Why would the Captain ask not to climb after PF calls out nine one thousand?

Lyman

Because it is at this point he sees PF climbing (wrong) as PNF is trying to descend. He sees they are both inputting, and hears it on the SV prompt.

It is this PF overcontrol he demands be stopped.

Shadoko

@ CogSim

Climb and Go down are from the French Monte et Descend. There is no ambiguity in French. "Descend" can't be any else. IMHO.
I pasted the French (pages 98 ans 99) beside the English page 95:



If you want to try some "electronic" translation from French, you can have a "à" with Alt+133 and "ç" with Alt+135

CogSim

But its the PNF who has asked for the climb. I think the Captain is overruling the PNF here.

DJ77

Looking at the BEA plots or better at mm43’s re-arranged plots (#1770) pitch response to the few consistent nose-down sidestick inputs appears very weak. Except the last one, pitch curve minima seem more related to high roll angle than stick down input. During these nose-down sick periods the elevator angle never increases above -15 deg. Why would the THS start moving up?
A possible reason could be that in the stall domain decreasing AoA increases a little Nz in an almost unchanged trajectory. Remembering that the pitch control is still C* it is easy to see that an Nz increase will decrement the actuating signal in the control loop. Moreover, the default gains used when CAS is invalid likely play against a fast response.

Lyman

I think I am correct from nine thousand. Both piloits are at cross purposes. I am so sad to say, that Though I may be wrong, or right, it does not matter/ none of the three knew either, and they were There./

CogSim

Thanks for the clarification Shadoko. Why not just use descend in the translation then?