Sad: Atlas 767F down near Houston

[John H Watson]

They said it, then apparently redacted it. I would doubt they would write it if they didn't mean it.

There have been other suggestions like a crew member heart attack (and subsequent slumping on the control column).  Would all crew members be securely belted in at this point?

Although column position is fed into the DFDR, I'm not certain if that data is necessarily recorded. The DFDR may only have had a record of elevator position. (Something may have been lost in translation .. between the guy reading the DFDR data and the guy who issued the statement).

The column sometimes kicks when hydraulics are applied. Might a hydraulic problem have caused a hard nose down input? Have there been incidences of mechanical or A/P failures resulting in large elevator deflections? It's happened on A330s, but they are different beasts.

Another interesting analysis:

https://www.satcom.guru/2019/02/more-questions-raised-for-atlas-5y3591.html

[Hardy Heinlin]

Do they know whether the A/P was engaged during the dive?

I read "column" and "elevator" -- did I miss something?

[emerydc8]

...and a lot of carriers overseas, they are so automation dependent that they don't know, based on their training, when to intervene, and if there is a problem they continue to try to use the automation. We've seen that now in three accidents. Lion Air, Ethiopian, and in fact, Atlas Air, the one that crashed in Houston. The automation was still coupled, the pilots didn't hand fly the airplane when they lost control and even through the recovery they were fighting the automation.

While riding in the observer seat of a flight this evening we were at FL370 in moderate TB. The captain decided to climb to FL390 to get over the tops. During the climb, it went into SPD|VNAV PTH and the A/T came right back to about 75% N1 as we leveled at FL390. It stayed there as the airspeed bled off towards M.70 and headed lower. We have no idea why it happened but intervention was definitely required to prevent a stall. The automation is definitely not foolproof and this was no programming error by the crew.

I so hope this was an automation dependency-related crash so that at least something good could come out of it.

[Hardy Heinlin]

Was there any turbulence while the A/T was staying at 75%?

[emerydc8]

Yes. And IMC conditions, but too cold for ice (-56 SAT). The ECON SPD was .79 and I saw .70 when I said something. It took firewall power to get it back and even then it took about a minute. The throttles were not moving at all by themselves. They eventually disengaged the A/T and then pressed SPD to reconnect. After re-engaging VNAV they seemed to work okay, but we were already out of the TB by then.

[United744]

Was a report filed?

[emerydc8]

This isn't United.

[Hardy Heinlin]

Is there any turbulence mode on the 767 that freezes the A/T to avoid chaotic thrust adjustments?

[emerydc8]

Not that I'm aware of. But there is something to the effect that operation in moderate to severe icing conditions could result in erratic or unreliable autothrottle operation. I just can't seem to find it now. They train this in the sim on a LOFT during climbout. If you don't intervene quickly, you'll blow right through the barberpole when you level off at an intermediate altitude.

[emerydc8]

Found it. We were on top/in this "bomb cyclone" over the midwest https://www.zerohedge.com/news/2019-03-12/bomb-cyclone-back-and-its-ready-explode-hurricane, so it could have been ice crystal icing that caused the A/T issue, even at the colder SAT. It was pretty unusual weather.

[Jeroen Hoppenbrouwers]

Just asking out of curiosity. The autothrottle cannot maintain the desired airspeed, but the engines can, right?

Many control systems are based on two items:

1. A model, using mathematics, to predict what the controlled system will do,
2. A feedback loop, using sensors, to tell the controller what the controlled system actually does.

If the A/T cannot maintain the desired airspeed, this must mean that something goes completely upside down in the feedback loop. Apparently there are elements other than airspeed (which is obvious) that tell the A/T that it is doing fine. Is the A/T model so much based on physics and expectations that, say, a frozen over OAT/TAT sensor will upset the controller enough so that it does not push the throttle forward, even while it physically can? In other words, is the A/T controller always flying blind by the numbers it uses to estimate the required throttle position, and does not cross-check itself with the desired airspeed, but only with the thrust it wants?


Hoppie

[Hardy Heinlin]

On the 744 (and probably on other jets too) the A/T disconnects when a relevant probe heat fails. The control loop is very sensitive. It's not only the pitot probe that provides fractions of a knot to the sensor loop; the airspeed trend too adds nuances to the sensor loop. The system would oscillate if there wasn't any trend data in the sensor loop that feeds the control loop. Without any anticipation function it would chase the target point back and forth. So the data filters need to be fine-tuned very much; any slight sensor error would have a big effect. I guess disabling the control loop in such a case is better than keeping it enabled, driven by wrong sensor data.

[Jeroen Hoppenbrouwers]

Got it. But I keep thinking, if the single sensor that gives you a final good/no-good of the control loop, the KIAS pitot tube, is telling you you're not at all even close -- what do you do? *BING* A/T disconnect, I'd think. Better than the sneaky slowly decreasing airspeed somewhere up near the coffin corner. Apparently this KIAS sensor is not in the loop at all, or so overwhelmed by all the other air data sensors that it has nearly no vote?

The 767 and 747 A/T computers in the FMC may actually be of the same generation, so they may share this "interesting" feature? Has any of our resident 747 drivers ever seen such a hideous thing?

Hoppie

[Hardy Heinlin]

Apparently this KIAS sensor is not in the loop at all, or so overwhelmed by all the other air data sensors that it has nearly no vote?

I would say the latter is true. The KIAS sensor is even the primary one. No airspeed control without airspeed data. The (744) FMC has a cruise-airspeed-thrust database though. You can see the database thrust data on the VNAV CRZ page, and on the EICAS (magenta line) when the A/T is disengaged and VNAV is engaged. However, this feature is not designed for permanent, exact airspeed control; and not at all for climb/descent or transition phases. It's rather useful for manual flight in turbulence, or when the A/T fails. In level flight.


|-|ardy

[John H Watson]

NTSB Intial Report review...

https://www.washingtonpost.com/national/cargo-plane-appears-to-have-hit-turbulence-before-crash/2019/03/12/c2737674-44e1-11e9-94ab-d2dda3c0df52_story.html?noredirect=on&utm_term=.bc09e1f4bfa3

[emerydc8]

This coincides with what the friend of a person on the investigative committee told to me last night. He said they are probably going to be looking hard at Atlas's training department and the experience level of pilots they are hiring. With the labor strife over there, and the pilot shortage, it sounds like they are scraping the bottom of the barrel to keep up with the massive attrition.

Pilot Error Suspected in Fatal Atlas Air Cargo Crash

Investigators exploring likelihood that crew accidentally increased thrust on approach to Houston airport, sources say

Andy Pasztor
March 15, 2019 4:06 p.m. ET

National Transportation Safety Board experts, these people said, are focusing on a likely sequence of events that started with the crew of the Boeing Co. 767 approaching Houston's George Bush Intercontinental Airport on Feb. 23 inadvertently commanding dramatically increased engine thrust. Turbulent air could have jostled the arm of one of the pilots, causing the engines to rev up to takeoff power, one of these people said.

The sudden surge in thrust, which the safety board disclosed in an earlier factual update, forced the nose of the plane to pitch upward and startled the cockpit crew, according to these people. Almost immediately, according to the preliminary data released by the safety board, the crew responded by sharply pushing down the nose of the aircraft.

The board previously said the nose was pointed downward at a 49-degree angle with the plane still about 30 miles from the airport, creating a much steeper descent than a normal landing approach.

The seemingly disoriented crew failed to regain control—despite commands to pull up from the jet's high-speed dive—and the wide-body plane plowed into a marshy area.

The safety board said the crew had the required training and medical certificates.

The safety board hasn't issued any final conclusions, and the leading theory currently pursued by investigators could change as more information is developed. A spokesman for the board said it had no comment beyond the factual update released earlier. Boeing had no immediate comment.

Atlas Air Worldwide Holdings Inc. and the union representing its pilots both declined to comment, citing the ongoing investigation.

But at this point, such cockpit slipups are considered the most likely cause of the crash. The plane was skirting around some storm cells before the plunge, but people familiar with the details said the turbulence didn't cause any structural damage or lead any system to malfunction.

In its update, the safety board said the "engines increased to maximum thrust" as the plane was flying at roughly 6,000 feet. After a brief nose-up movement, according to the update, the Boeing 767 entered a steep descent in a "generally wings-level attitude until impact with the swamp."

Many airline and aerospace industry officials have watched the probe closely because the 767 model is widely used as a passenger jet around the globe.

The Atlas Air flight, en route to Houston from Miami, was flying cargo for Amazon.com Inc.

The last fatal U.S. airliner crash also was a cargo flight. In 2013, a United Parcel Service Inc. Airbus A300 slammed into hill while approaching to land in Birmingham, Ala., killing both pilots. The safety board determined that a series of pilot errors and violations of safety procedures caused that accident. The aircraft descended too quickly toward a runway shrouded by clouds, and the pilots waited until the last seconds to try to initiate a go-around.

—Alison Sider contributed to this article.

Write to Andy Pasztor at andy.pasztor@wsj.com

[John H Watson]

Turbulent air could have jostled the arm of one of the pilots, causing the engines to rev up to takeoff power, one of these people said.

Do they mean someone hit the GA switch?

[emerydc8]

That thought came to mind. I can't find any data in my manuals regarding whether or not the A/P would even disconnect if you manully moved the controls while it's engaged. I have tried several times in the real plane to override the ailerons (45 degrees on the yoke) with the A/P engaged to see if it would disengage but I couldn't get it to do it.

From a JAL accident report.

(3) Manual Override of Autopilot
"Manual override" means that when the autopilot is engaged, a force applied by the pilot to the control column makes the control surface angle different from that of the autopilot command. The term "manual override" is used with this meaning in the remainder of this report.
With the Boeing 747-400D airplane, when only one autopilot is engaged, the pilot can manually override the autopilot by applying a force on the control column. The autopilot does not disengage due to the manual override.

This is pure speculation, but what would have happened if both pilots were pushing forward with the A/P engaged and TOGA thrust being applied and at some point the A/P became disengaged? Nose dive?

[Hardy Heinlin]

Max thrust and 49-degree pitch down -- to me that sounds like suicide rather than lack of pilot training.

[emerydc8]

That's what a bunch of us first thought too, but apparently (according to my source) that has been ruled out. I think we'll know soon enough.