Colgan 3407 Down in Buffalo

[M-crit]

What would be considered a nominal speed in this config?

With gear down and flaps 5, I would not be slower than maybe 160 or 165, perhaps even faster especially with the ref speeds increased in icing conditions. You would have to be below 181 for flaps 10, or below 172 if you wanted flaps 15.

 

[JetSpeed219]

The increased ref speed switch increases by 15kts. And 134kts for gear down is very slow. Gear is usually coming down around 180. Somewhere around 150 is when flaps 35 is usually selected too.

 

[Papps]

Hate to speculate, but since we are speculating anyway. Is it possible there was a tail stall. as the nose dropped the crew yanked back to correct from tail stall, the tail regained lift, rapid pitch up followed by a wing stall and spin.

 

[glasspilot1]

The reports are that the deice system was turned on. Question for those that fly this airplane: When you turn on the deice system do all the boots activate at once? Or do the wings blow, and then the tail later? I'm just wondering if the wings had been cleared but the tailplane was still iced.

 

[EdAtTheAirport]

Where are you guys getting this spin notion from? I looked at a Q yesterday, and FWIW could not imagine an airplane with that big a tail, on that long of a moment, entering a spin. I could be wrong of course.

 

[WSurf]

The reports are that the deice system was turned on. Question for those that fly this airplane: When you turn on the deice system do all the boots activate at once? Or do the wings blow, and then the tail later? I'm just wondering if the wings had been cleared but the tailplane was still iced.

Its automatic, it sequences from the wings and tail. Works well, there is a knob to select Fast or Slow in the cycling. Also on the Captains overhead panel are the lights that light up once each boot inflates showing you that the system is up and running. Piedmonts Procedure is to Put it on Fast mode anytime you are in icing.

 

[M-crit]

Piedmonts Procedure is to Put it on Fast mode anytime you are in icing.

That is a great idea btw. Why put it on slow? I hope we will be able to find out where the crew had that switch set to- fast or slow.

 

[surplus1]

Colgan 1549 follow up

Surplus 1 said:

As the aircraft levels, the auto-pilot trims nose up to hold the altitude as the speed decreases. The nose up trim continues slowly unnoticed by the crew. The auto pilot is also inducing aileron trim at the same time – to keep the wings level - also unobserved.

Could this accident have been avoided? YES – but not after the upset occurred.
hypothetical.

*****

Well thought out scenario..., doesn't even require any real effects from ice. Be interesting to see what the power setting was while decelerating, while the autopilot held altitude towards the FAF (did they decel into the stall driving in?) Didn't notice the decaying airspeed and approaching stall condition until the high AOA kicked off the autopilot just prior to stick shaker & upset? Full power on the engines hung from the wings (upon stick shaker as first warning to crew) cause initial pitch up?

ILS23 FAF at BUF just under 1500' AGL 4.4 miles from the threshold... upset the aircraft at that altitude and there is no recovery.[/quote]

Thanks for all of your inputs - everyone.

I may have been a bit too vague. The hypothetical scenario was intended to include ice contamination of the airfoil - most probably asymmetrical ice. This would produce stall speeds and critical AOAs significantly higher and lower respectively than normal. The asymmetrical accumulation would induce the autopilot to apply aileron to keep the wings level. I doubt this would have happened in an ice-free environment.

In the hypothetical scenario, auto pilot disconnect is not commanded. [I believe that is consistent with the reports we have so far, i.e., the auto pilot disconnected itself.] It occurs because the autopilot has exceeded its limits (either in roll or pitch or both). My own belief is in roll.

Those who have pointed out that in a tailplane stall the pitch would be nose down are correct. While I didn’t deal directly with that in the hypothetical, I think it was an error on my part that doesn’t fit what we know now – which I did not know at the time.

Given the reported sequence of excursions (pitch and roll) that have since been posted – if they are accurate – I personally do not believe that there was a tailplane stall in this accident sequence. I now believe that the wing stalled. I also believe that one of the two wings stalled before the other. That does NOT mean that I think I am correct. It just means that the available information is pointing my thinking in that direction.

The more information that comes out, the more I see this upset as being remarkably close to the upset sequence that happened to Comair 3272 on its approach to the Detroit area. Up to this point the only major difference I note is the altitude. In the CMR event they were at about 4000 ft (don’t remember exactly) when the upset occurred. If I recall correctly, the time lapse from upset to impact was 17 seconds – based on FDR data.

In this accident sequence - When the autopilot disconnects the nose pitches up immediately (due to the unobserved up elevator trim that I mentioned in the 1st post) and the wing immediately exceeds the critical AOA and stalls. At the same time, one wing stalls more than the other (due to the ice – I believe) and the aircraft rolls in that direction - primarily because the only thing that has been keeping that wing from dropping was the control pressure (aileron) applied by the autopilot.

That would leave us with this unanswered question: What made the autopilot disconnect? I obviously do not know but I do have a theory.

There was a lot of ice on the airplane – much more than the crew realized. There was more ice (for some reason) on the left wing than on the right wing – perhaps behind the boot; Enough to make the left wing drop and the aircraft turn to the left. The autopilot applies enough aileron to prevent the turn (keep the aircraft from banking) or if in a turn, to keep the bank angle from exceeding the norm (about 25 degrees on auto pilot). So far so good.

Note: Shortly before this all happened the aircraft was reportedly vectored to a heading of 260 and cleared for the approach. That appears to comply with the 30 degree localizer intercept angle that controllers are required to give. Continue.

As the aircraft captures the localizer, the autopilot makes another left turn to the inbound heading (I have no plate – but let’s say 230). The autopilot studiously accomplishes this turn. It relives the pressure and the aircraft banks to the left. That’s what it is supposed to do. [All the while more ice is accumulating].

[As you review this keep in mind that both wings don’t necessarily stall at exactly the same time. If one wing is flying and a part of the other is not the aircraft will roll in the direction of the wing that is partially stalled.] Continue.

The next thing the autopilot has to do is stop the turn and hold the localizer. To accomplish this it must apply enough aileron to raise the left wing. It tries but the left wing is heavy with ice and it doesn’t come up. The autopilot tries harder - until it reaches and exceeds its limits. At that instant – the auto pilot disconnects itself.

If this assumption is correct, the aircraft immediately rolls hard left. At the very same time the auto pilot (previously) induced elevator up forces take effect and the nose pitches up hard (+31 deg. according to the reports.)

[Keep in mind that most of us have never seen more than 30 pitch up in actual flight. Perhaps in simulators but not otherwise. This is not ‘normal’ in T-category aircraft. Even if it were, it would not happen suddenly and unexpectedly].

The excursion is so rapid that both shaker and pusher activate almost simultaneously. [The prior post says +31 nose up and 46 deg (left) wing low.]

At this point the upset has occurred and there is really nothing the crew can do to recover within the available altitude. Of course that doesn’t mean that they stop flying. To the contrary – they react immediately in an effort to recover from the unusual attitude.

Taken completely by surprise, the PF (pilot flying) grabs the yoke, moves it full forward and at the same time applies full right aileron to raise the left wing. Full power is applied. [The power application helps to increase the nose up tendency and the PF fights it – as is natural].

The nose rapidly drops through a 76 deg. arc to a -46 deg. nose down pitch and the aircraft accelerates. PF applies up elevator (a whole lot of it) to stop the excessive pitch down. The nose comes up rapidly but most of the aileron input is still there (not consciously). A secondary stall (accelerated) occurs (shaker/pusher activate again) with the aircraft rapidly rolling (almost if not snapping) to 105 deg right bank angle [technically that's partially inverted].

In the effort to recover, control movements are extreme. There is much over-controlling. That's not pilot error, it's to be expected under those conditions. The PF will always be a little behind the aircraft's reaction to control input – this is acrobatic flight but it is not intentional. All of these things occur in 5 seconds or less. Remember - they are not visual - this is all on instruments.

I've never flown the Q-400 or its predecessors. I don't know what if any limits the attitude instrument may have. Are they capable of normal function in acrobatic flight? Can it maintain accuracy with a 105 deg bank angle? Some can, others can't. How much pitch excursion does it take to make the presentation illegible? I have no idea. Can the back up gyro handle that? I don't know. The AS indicator is probably a tape - is it easy to read or does everything become hieroglyphics in the process? I don't have those answers.

You ask: "(did they decel into the stall driving in?) Didn't notice the decaying airspeed and approaching stall condition until the high AOA kicked off the autopilot just prior to stick shaker & upset?"

I think the answer to that question is YES – they decelerated gradually into the stall regime. Why did that happen? Not too difficult.

There was no reason to believe that the airspeed indication was critical. They were intentionally decelerating and IAS was well above stall (under normal conditions) - under normal conditions. Note caveat: I have no idea what a ‘normal’ approach speed for the Q-400 might be but 134 kts does appear to be somewhat slow at the OM (with zero flap) for an aircraft of that size – by about 40/50 knots.

But, conditions were NOT normal. There was likely a great deal of ice on the airframe - in places that they had no way of seeing or knowing. Critical AOA was much lower and stall speed much higher than 'normal'.

Continued below

 

[surplus1]

Continuation

My educated guess is that the autopilot did not disconnect because of the up elevator trim or because it was kicked off by high AOA. I say this because the reports indicate that the autopilot disconnected before the shaker/pusher activated. If that is correct - It disconnected because its ability to control the bank angle had exceeded its limits. That’s my theory – and that’s all it is; a theory.

When the disconnect occurred, the nose up elevator trim caused the initial pitch up. That pitch up caused the wing to stall - very suddenly and immediately.

The reports tell us that an attempt was made to retract the flaps (and maybe the gear). If that is true – it was mostly likely because the PF associated flap extension with the upset. In any event in stall recoveries in this category of aircraft Training often says “max power, flaps up, gear up”. Again, I have no idea what that airline teaches, so that thought is pure speculation. In any case I consider it to be essentially irrelevant in this instance.

Again, I know nothing about this a/c type but my guess is that the elevator trim is electric - and there is no trim wheel in the airplane (something I personally find objectionable – just a quirk of mine). Therefore, it is especially difficult to tell where the elevator trim may be at a given point in time, especially at night in a dimly lit cockpit, when the a/c is being flown by the auto pilot, holding altitude, undergoing power changes and decelerating – all at the same time.

An upset in a transport category airplane is simply not recoverable at an altitude of 1500 ft AGL; certainly not on instruments and totally unexpected. Most if not all upsets are never anticipated.

The very best among us have worked experimental upset scenarios in simulators trying to develop satisfactory recovery techniques (down in MIA if I recall correctly). They developed some ideas but no real solutions and no brilliant recovery technique. Unless there is a great deal of altitude below you, the only thing they managed to prove was that upsets in T-category aircraft are not recoverable. They must therefore be avoided by all possible effort.

Avoidance isn’t cut and dry. Each aircraft is different and one size just doesn’t fit all. Likewise, Icing is a different ball game in each machine. As someone before me pointed out: “In aviation there are three types of ice: Good Ice, Bad Ice and Hazardous Ice.
Good Ice is found in the galley”.

This is NOT the first incident or accident to result from an upset. Perhaps it is the first in the Q-400 but that aircraft can't be so unusual that it 'can't happen'. We all know Murphy’s Law.

Several upsets have occurred in the Brasilia. At least two that I know of resulted in accidents, one of them as tragic as this one. More than one has resulted in structural damage. ALL occurred while the airplane was being flown by the auto pilot and several in VMC. There have been several in versions of the ATR, especially the -42, at least one of which was also a fatal accident in this country. Its obvious or should be, that these new generation turboprop aircraft require their own rules of operating technique and flight crew understanding. The results indicate that some of the invented simulator scenarios just aren’t cutting the mustard. We just can’t keep repeating the same thing over and over an hoping for different results.

It should also be noted that upsets have occurred in heavy jets as well and in small jets too – also involving autopilot operations and situational awareness.

Unless we choose to make ourselves aware of history and act differently as a result of our knowledge, history will surely repeat itself.

It is not my purpose to affix blame or to be critical of anyone. Its not about blame, its about cause and effect.

It is my purpose to do whatever I can to avoid unnecessary repetition of tragedy or even white-knuckle operation. I am willing to err on the side of safety in that effort. I also believe in the axiom – knowledge is power.

If you see error in this phase 2 hypothetical please point it out. I won’t be embarrassed if I’ve made a mistake. At this point I am willing to rule out the tailplane stall theory.

“Aviation in itself is not inherently dangerous. But to an even greater degree than the sea, it is terribly unforgiving of any carelessness, incapacity or neglect.” — Captain A. G. Lamplugh,
PS. Sorry for being so long winded.

 

[bailout]

The part i dont get, is if there were no pitch control input changes prior to the AP disconnect, why would it pitch up? They were already trimmed for level flight (pitch) and possibly coming out of the initial intercept turn (roll), but still 'in level trim', so if the roll excedence were to cause the disconnect, it seems that the pitch shouldnt change? I think i understand when you say that when they were slowing the trim was compensating for the required trim until it reached a limit but usually when you get a AP disconnect because of limits, the plane stays where it was, whether it be level pitch, or in a 20 degree bank turn, or what have you because thats where it was when it disconnected.. Granted you have to take immediate control to prevent something from happening further. You might be right about the wing stall but the nose would dive and the pilot would yank bank which would address the pitch up reports.

 

[VF45]

If you cob the power with full nose up trim, will the Q400 pitch up?

 

[glasspilot1]

Its automatic, it sequences from the wings and tail. Works well, there is a knob to select Fast or Slow in the cycling.

Thanks for the reply Surf.

 

[surplus1]

The part i dont get, is if there were no pitch control input changes prior to the AP disconnect, why would it pitch up? They were already trimmed for level flight (pitch) and possibly coming out of the initial intercept turn (roll), but still 'in level trim', so if the roll excedence were to cause the disconnect, it seems that the pitch shouldnt change? I think i understand when you say that when they were slowing the trim was compensating for the required trim until it reached a limit but usually when you get a AP disconnect because of limits, the plane stays where it was, whether it be level pitch, or in a 20 degree bank turn, or what have you because thats where it was when it disconnected.. Granted you have to take immediate control to prevent something from happening further. You might be right about the wing stall but the nose would dive and the pilot would yank bank which would address the pitch up reports.

As I see it, the key to your statement is in the first sentence: "The part i dont get, is if there were no pitch control input changes prior to the AP disconnect, why would it pitch up?"

Actually there were lots of pitch input changes - all made by the auto pilot and unobserved by the crew. [If the darn airplane had a trim wheel they might have noticed it turning] Additionally, the autopilot was struggling to keep the wing from increasing the bank angle (due to the ice on that wing), which they also did not know.

The aircraft was in a descent - through 6000, 4000 to 2300. Power most likely retarded for the descent. Altitude is pre-selected to 2300.

As we reach 2300, altitude captures and is held by the autopilot. What happens now to airspeed? Unless power is added, airspeed will decrease.

If the autopilot was NOT engaged, and you didn't touch the yoke - what would happen to the nose as the airspeed decreases? It would drop off wouldn't it? If you did not want that to happen but you still wanted to slow down. what would you do? Trim nose up.

Now put the autopilot back in the equation. You told it to hold 2300 ft, and it does. As the airspeed decreases how does it do that? It trims the elevator nose up, to hold the altitude.

If you do nothing else it will continue to trim nose up and the airspeed will continue to decrease - until one of the two reaches the level required to hold that altitude with that amount of power. In other words, if the power available matches the power required to hold altitude at that speed, the auto pilot will stop triming and you will be in level flight with x amount of nose up trim and y amount of power - stable. If the power available/used remains less than required - airspeed will continue to deteriorate slowly until the auto pilot runs out of available trim capability or you exceed the critical AOA and stall.

I am only saying how it works - not what they did. I assume that they added enough power to hold the desired speed (or what they thought would hold it).

Everything is now stable and the auto pilot has stopped triming nose up -but - it has NOT removed much if any of the trim previously input. You're still at 2300 ft and headed for the localizer. Glide slope is not alive and has not captured.

As you approach the localizer the autopilot begins a turn to capture it. The auto pilot is still flying. What happens now?

As the bank increases more back pressure is required to hold the altitude. The auto pilot provides in by additional nose-up elevator trim. Something else also happens - the airspeed begins to deteriorate (slow) again. You decide to accept the slightly lower airspeed and you don't add more power - it's already where you think you want it.

Meanwhile bank angle is increasing and elevator is trimming nose-up (slowly). Bank angle tries to increase beyond 25 deg. (due to the ice) until the auto pilot exceeds its limit and kicks off. The nose pitches up and the wing stalls.

Instantly the bank increases from 25 to 45 degrees - further increasing the stall speed. What happens to stall speed as bank angle increases? Yep, its higher than in was before. It was already on the edge - you just didn't know it.

Just as you act to correct the bank the wing stalls. You call for full power and all that up elevator trim is still there. Where does the nose go before you can stop it? All the way to +31 degrees nose up and you have the yoke full forward. There's a lot of power there and some very big props.

Down goes the nose - very rapidly. Your control input has taken effect There is still too much elevator trim for that amount of power and you're pushing hard.

The instant you see 45 degrees nose down you reverse the pressure and pull like hell. You still have almost full aileron cranked in to pick up the left wing. You havent thought at all about trim. Who would?

Just as you apply that back pressure you now enter an accelerated secondary stall. The aircraft snaps to the right - exceeding the vertical bank angle. It is completely out of control - plus you just lost a thousand feet.

The rest ...............

It's all hypothetical ... just a theory.

 

[surplus1]

If you cob the power with full nose up trim, will the Q400 pitch up?

I've never flown it so I don't really know.

But, if it reacts like other propeller airplanes that I've flown the answer is yes. It would nose up even if the elevator trim was neutral. It's a prop/thrust line thing. Plus it's a high wing, which ususally aggravates that tendency.

 

[trainer8]

Hi Surp,

Very good assesment--your last couple posts. I concur with your assesment, except on a couple minor points.

1) Chealander confirmed AP disconnect occurred at stick shaker/pusher and not from AP sensory loads.

"The NTSB's Chealander confirmed that the autopilot was engaged until the stick-shaker, and the stick-pusher kicked in, signaling the start of an event that could have, and did have dire consequences. He went on to say that when this happens, the autopilot automatically disengages, putting the aircraft back in the hands of the pilot."

2) I wouldn't rule out tailplane icing. Having flown this type for 8100 hours in various icing conditions including moderate icing (tennis ball size ice sculptures on the ice detection probes on an ILS into SWF), it's ability to fly in moderate ice never produced any aerodynamic instability while accreting ice. I would be very surprised if the NTSB ruled that this would be the cause.

The application of wing flap (Dash-8s have full span wing flaps) from 5 to 15 appears to have induced the upset, which fits the NASA tailplane icing flight testing scenarios. (http://www.part135.com/TailplaneIcing.html) Chealander (NTSB) also indicated that the flight crew followed all AFM procedures pertaining to flight into known icing conditions.

What seems to be conspicuous was the situational awareness of airspeed degradation to onset of stick shaker.

Overall, I think your assessment is very good.

T8