Figure 3 highlights the descent during the last two minutes of the flight with select paraphrased CVR comments overlaid. The descent appears normal until about 1238:31 and approximately 6,300 ft pressure altitude when the go-around mode was activated using one of the Go-around (G/A) buttons on the throttle quadrant: see Figure 4. There was no mention of initiating a goaround by the flight crew on the cockpit voice recorder (CVR), and flight 3591 had been cleared to descend to 3,000 ft. (The normal load factor, an, in Figure 6 shows that the airplane encountered turbulence at approximately 1238:25, six seconds before the recorded flight mode transitioned to G/A. The turbulence was likely associated with the cold front mentioned earlier and included instrument meteorological conditions or IMC.)
The airplane Euler angles recorded on the FDR are shown in Figure 5. The airplane pitch attitude increased from about -1˚ airplane-nose-down (AND) before G/A was annunciated to approximately 4˚ airplane-nose-up (ANU) six seconds later at 1238:37. The airplane then pitched AND to about -49˚ over the next 18 seconds in response to an AND elevator input.
At 1238:48 and 1238:51, the pilot flying (the first officer in the right seat) made comments about the airplane stalling that were recorded on the CVR. However, Figure 6 shows the recorded airplane wing angle-of-attack and airspeed were below -15˚ and above 250 kt, respectively. This is well below the airplane’s wing stall angle-of-attack.
Figure 7 shows an elevator split between the left and right sides of the airplane: the left elevator is associated with the captain in the left seat, and the right elevator is associated with the first officer in the right seat. The elevator deflections are similar until about 1238:46. At that time, the first officer is heard asking a question about airspeed on the CVR, and the elevators begin to split: the captain begins to pull from 2˚ to 8˚ more ANU elevator than the first officer. The split continues until about 1238:56, about one second after the airplane would have broken out of the reported 3,500 ft cloud layer. Both the captain and the first officer subsequently commanded ANU elevator until impact. (The normal load factor recorded by the FDR in Figure 6 is greater than 4g right before impact.)
Subject: Houston Amazon 767 Crash 23 Feb 19
From the net, courtesy of a reliable source.… [i.e. Now, this is no s**t... - Airbubba]
Just FYI… we’ve heard the full cockpit audio and seen the data. Here’s... what really happened (name redacted to protect the innocent!):
During the approach, at about 6,000 FT (being flown by the first officer), the Captain reached around the throttle quadrant to extend the flaps to the next position after being called to do so by the first officer (pilot flying)… very normal.
In many aircraft including the 767, that’s a very odd/difficult repositioning of your hand (from the left seat, all the way around to the right side of the center console), and requires intimate familiarity and slow deliberate motion to do successfully.
Well in any case, it was not done so this time. The captain accidently hit the “go around” switch while bringing his hand around for the flaps, which brought both engines up to full power. In the landing configuration, as this aircraft was transitioning into, that obviously causes a vast increase in lift… and the first officer (pilot flying) used everything he had to force the nose back down.
Still not sure why that occurred, as the crew should have just “gone around” and tried it again when properly configured… but they did not. And that started in motion a chain of events that lead to tragedy.
As the First Officer over-rotated downward, again with the engines at full power, the aircraft quickly accelerated and approached something we’re all trained to handle (at least in good training environments)… an “upset recovery”, countered by NON-AUTOMATION and basic “stick and rudder skills”.
This captain however, in turn, grabbed the controls without using positive command (“I’ve got”, “My aircraft”, or anything normally done), and countered the F/O’s control input by completely hauling his control column full aft… remember, while the F/O is pushing full forward.
In the process of doing that, he broke the “shear pin” on his control column (a device/mechanical safety interlock used to separate a control column from the “innards” of the control architecture in the event one control column is doing something it should not)… and that occurred here.
The captain, a few seconds later, now accelerating downward out of the control envelope of the 767 (remember, all of this started at 6000 FT and probably took less time to get to the fatal point than it did to read this far), recognizes the has no control column and then asks the F/O to pull up, get the nose up, or something to that affect. It isn’t 100% clear what he says.
The F/O then tries to pull aft on his column (going from full forward to full aft), but isn’t getting the response he needs, because the aircraft is out of the envelope of controllability and the controls are “air-loaded” in position.
At about 2000 FT, eventually the trim motors are able to start overcoming the air-load, and the aircraft begins to attempt to arrest its rate of descent… but alas it’s far too little, far too late, and the aircraft impacts about 30-40 degrees nose down, with what is believed to be about 4-5000 FT / minute rate of descent.
All during this time the throttles aren’t touched until somewhere during that last few seconds of flight… which is believed to be what enabled the trim motors to start working. Unclear who does it, and no audio indicates who it was.
Just FYI… we’ve attempted in our 767 simulators to recover from the event with the exact same setup, and thus far we’ve only had success when starting at 8000’ or higher… meaning we are fully established in the “out of control” position at 8000’, recognize it by then, and initiate recovery starting at 8000’.
These guys started the whole thing at 6000’ and were much lower when a true recovery attempt was initiated. No chance, and just shows you how quickly you can get “out of the envelope” when you don’t follow procedure, try some completely erroneous recovery technique, and don’t have a clue what you’re doing.
So many things went wrong with crew coordination, basic flying skills, aircraft envelope awareness, basic procedures, and such… that this will likely go down as one of the absolute worst “pilot error” events ever.
It needs to have serious impact throughout the Amazon flying circus (and associated partnerships), and show people that Jeff Bezos’ attempt to push the envelope at lower cost, all things else be damned, doesn’t apply to aviation.
This accident no doubt was absolutely horrible, and three people lost their lives…one of them (the jumpseater) through absolutely no fault of his own. But making an approach into Houston, TX, it could have been so much worse. In another few miles, they would have been over major population centers and who knows what would have happened then.
Know your aircraft. Know your procedures. And for God sakes, just FLY! It’s not a video game!
Relayed by your PJP Editor
Bert Botta
Aviation Writer
PJPFBO Editor
Email: bert@privatejetpilots
this from the boxes folks, heard the same thing from my insider connection, jumpseater thrown from under his unfastened belt, Captain sheared the control breakaway pulling so hard to override the FO who had his column forward, it is thought as flaps were called for, somehow the Capt inadvertently hit the TOGA levers, which then caused TOGA to activate when flaps "1" were selected, the pitch and startle factor caused the FO to aggressively move his controls nose down, with the captain fighting him to no avail, when the aircraft came clear of the clouds then the FO made some exclamation and began to pull back on the controls, but apparently not aggressively enough , I was not privy to the actual conversation on the CVR, this account was heavily redacted, might be a bit before all the story is heard, the word I heard used to describe it was "shocking"