What would make the IRU fail so suddenly the autopilot would automatically disconnect (as per cappt's experience)? What would cause pitch angle to be extremely misleading?

Bad sensors? Icing on sensors? Sensors falling off in flight? Broken gyros? Is this a maintenance issue?

Right now the aircraft is pretty much frozen solid into the ice and snow in northern Sweden. Any investigation regarding instruments left up there would have to wait until spring.. that would be May at the earliest that far up north.

Correct, a single chime caution followed by EFIS COMP MON or AHRS 1/2 failure message. The AHRS failing will remove the FD and kick off the associated yaw damper, disabling the autopilot (calvary charge)
I am not suggesting AHRS/IRU failure here, just one thing of several that will disconnect the autopilot.

Will the FDR record info from both PFD's and the standby info or only the captains/flying pilot?

I would think that normal procedure would be for CPT to fly AHRS1 to his PFD and FO would fly AHRS2 to his PFD, however looking at the text of the discussion between the two crew members, the FO doesn't seem to be aware of the problem with the CPT display. Does anyone else get the feeling that they were both on the same (failed) AHRS? Other than a previous failure of one AHRS, why would both crewmembers be using the same AHRS? Switchology?


As I recall from AF447, there was a problem with only the Cpt data being recorded, and they could only infer what the RH indications probably were. Likely the same on the CRJ.:confused:

During normal conditions the DFDR is provided with attitude information from IRU 1 via the DCU 1. (from the interim report). IRU 1 also feeds PFD 1.

I wonder which IRU feeds the standby attitude indicator? Or is it a separate gyro?

I also wonder about similar accidents where in spite of warning chimes and flags on the PFD, the pilot has just chased the invalid attitude indicator? Something for human factors people and instrument designers to rethink?

bank angle at event:
initially: left aileron, followed by left bank;
then: right aileron until bank 0;
followed by: continuous, hard left aileron and a complete roll over to the left

At a first, superficial view I can't see a discrepancy between the roll inputs and the roll rate. Seem to corelate.

I only know the PFDs in GA airplanes, but at very high or low pitch angles, there are mainly chevrons and air or ground to see, leading attention to pitch instead of bank and making it a bit more difficult to control bank.

The conversation makes me think that there is a good chance that both PFDs were on IRU1, as Machinbird supposed. No call from the FO for need to pitch up. Would normally be the main objective in that situation, approaching VMO.

The stand-by PFD or AI should have a separate IRU.

If you assume that the pilot was reacting to the displayed bank, then his inputs will appear to correlate with the bank angle, naturally. That doesnt, in itself, validate the bank angle. Indeed, if you look at about 23:19:50, thee are 5 to 10 seconds of sustained left wing down aileron (more than 10 deg) at a VERY high airspeed. This should roll the aircraft rapidly left wing down. It doesnt, it stays right wing down, according to the bank angle. So while the aileron correlates witn a pilot trying to correct the bank, the bank doesnt respond to the aileron. Suggesting bank is, as stated by SHK, unreliable.

Oh and at these speeds, the aircraft WILL respond to roll inputs, quickly. The aircraft, at least at this point, is still inside the certified envelope and is not subject to aeroelastic loss of control, as some have suggested might be occurring. By regulation, you have a minimum 15% margin above VD/MD, which in turn has a good margin to VMO/MMO. You have to be going incredibly fast to see the aeroelastic effects become significant.

The elevator also correlates with the displayed pitch - pushing whe pitched up or pitching up, pulling whe pitching down or pitched down - and no one is suggesting that validates pitch.

The standby is indeed truly independent. There were, at least in theory, three independent attitude sources available.

@ Mad FS:

The provided diagrams are labeled:
pitch > 0 means climb, and elevator > 0 commands up
whereas
bank > 0 means RT and aileron > 0 commands LT;
reversed reading for the bank part in the graphs.

Thus, the pitch commands seems to be reaction in opposite direction of the indicated, after a couple of seconds of drifting upwards.

The roll commands are in the same direction of the deflection:
aileron left, with roll rate to the left; then right, with rate to the right, then left with rate to the left.

At 23:19:50 the aircraft is already in a 150 ° left turn. Until then, it seems corelating.

I don't know if bank information in reversed flight is reliable nor how a CRJ behaves at such an attitude.

Andi, as the SHK did state that bank angle is unreliable I wouldn't put too much interpretation into it.

What does strike me (again) is how important it is to have the CVR data as well as the FDR. Without the CVR we might spend hours (well, MORE hours... ;)) trying to decode what the pilots were trying to do. With the CVR at least we have a clue that - as QDM pointed out - their focus was initially on the bank, not the pitch.

/N

As pointed out, there are three sources of attitude information, the left and right displays, plus the standby, when two out of three should agree, with the standby often referred to as the "tie-breaker," since it's assumed that it should agree with the left or the right display, whichever one is correct.

From what we've been told, it reads as if the PF took the aircraft and tied it into a knot chasing an unreliable attitude display. From that point it was probably very difficult to figure out just what the airplane really was doing since it would have been at some extremes of pitch and bank, plus at negative G at times, and all on a dark night with little or no outside references available. Then to look at three displays and to pick the two that were correct would have been pretty difficult.

Is that 90 seconds from start to impact? That would imply a vertical speed of about 20 thousand feet per minute, or 300 knots.

Not at all unrealistic when you consider they impacted at over 500 kt.

I have only flown steam gauges, but it even there it can be an issue understanding the displays as you approach vertical flight. From what I've seen in mockups, modern PFDs are optimized for ~horizontal velocity vectors and may be completely inadequate for useful guidance in vertical flight.

Has anyone here ever done a loop in the CRJ simulator to see what the display does?

During extreme attitudes (-20/+30 or 65 deg bank) the PFD will de-clutter and red chevrons appear pointing to the direction of level flight.

I think this mostly industry standard now on EFIS equipped aircraft.

@AK
Sorry, the post quote function seems messed up so i can't do it "normally"

From 23:19:38 (approx) to 23:19:54, the ailerons are consistently positive. That means they are commanding left-wing-down roll (as noted on the plot and mentioned in your post). During that whole time the bank angle is positive - RIGHT wing down (again, as you note for the sign convention) - and INCREASES, which means rolling further to the right.

So the ailerons should be creating a left roll, but the bank trace rolls RIGHT. The roll commands are NOT in agreement with the bank response - they are in the sense to oppose the bank angle - suggesting crew action in response to the displayed bank, but that the crew action was not reflected in the bank angle recorded.

@Niclas - don't totally dismiss the bank. What SHK has said is that it doesn't seem to be compatible with the aircraft motion. they haven't said its "invalid" and it's probably a good assumption that the FDR was correctly recording the bank angle it was receiving, and that same bank angle (and pitch also) were being displayed to at least one set of cockpit displays. So it may be fundamental in determining why the accident happened. We just have to be careful in relating "indicated" attitudes to the likely "true" attitudes.

Could it be that they encountered roll reversal due to being supersonic or transsonic? This will explain why the aircraft rolls opposite the command roll.

A quick calc of IAS Vs altitude tells(wihout temperature corrections made) Mach topped about 0.95 around FL260. It was quite cold so I guess it came closer to transonic at lower alt, despite a FL-to-CAS-calc dont say so.

7-cylinder man

At least I'm pro enough not to call others for amateurs. So now please enlighten all us amateurs. By the sound of it you got it all figured out.

My question referred to the timeframe between 23:19:38 and 23:19:54 where apparently the aircraft was rolling in the opposite direction of what was commanded. I do agree that loss of one PFD should not lead to the loss of the aircraft and its crew, but I wasn't there and I bet there is a lot of factors that the Swedish Board of Investigation will look into.

I would say that the only one jumping to conclusions right now is you...

7-cylinder man

I am intrigued. What is it in the FDR that is so simple we simpletons haven't figured it out? What was the actual failure according to you? What do you know about the FDR data that tells you about what actually went wrong?

I can agree that loss of one instrument or reading should not lead to loss of an aircraft. Still, that happens. Most of the times during night with no visual references. Just as in this case. That is sad.

On another note, I still wonder what might have broken with such short notice as hinted at on the CVR. The crew offers a few expletives and then the plane crashes within 90 seconds. Things went downhill even faster than any Airbus during malfunctioning FOs.

As Mr 7-cylinder man says, there are no clear indication on the CVR what went wrong. Some hints by the FO that they disagree on what direction to bank... maybe, depending on how you read it. Whatever happened it did so very sudden (as I read the CVR) and the crew was more occupied with trying to fly the plane than communicating with anyone. Is it possible that this lack of communication between the pilots indicates that they had the same (wrong) mental picture of the situation? May they both have been fed the wrong information somehow?

I truly hope that there is a conclusion drawn from this accident when the final report is published. On first sight it looks very complex. It appears that we are into a what / who failed first endgame.

I have two cars, one is state of the .. and if something quits .. you pay BMW for a new box. The other is 1968 and of UK origin. If something quits you proabably knew about the possibility of failure, and the appropriate action anyway.

Times have changed.

The nasty part about an attitude system failure is the crew can follow erroneous indications until it becomes apparent that the aircraft is in a very bad place.

By this point the other attitude sources will also be showing unusual attitudes and the crew will likely be in a situation they have never seen or trained before.

Lack of air data makes it difficult to demonstrate such a failure in a sim with good fidelity. Should we do certain upset training in military trainers?

Were there alerts for attitude systems disagree? If so, they seem not to have gotten the crew's attention in time:(

Yes there are alerts. The alert would be a chime on the CVR.

How about equipping crews with night vision goggles, or pilots carrying their own? They aren't that expensive, and being able to see a visible horizon at night once the SHTF might be a life saver.

I don't mean they should be worn routinely at night, just available for quick donning.