AoA sensor errors on Lion Air flights JT043 and JT610
 

In 2006 a Boeing 747 taking off from London Heathrow with over 400 people onboard had both stick shakers begin to operate as the aircraft started to rotate. Once in the air, the stick shakers continued to operate, and the pilot’s and F/O’s instruments showed a continuous disagreement in the IAS air speed. The pilots turned back to land at Heathrow, where maintenance engineers diagnosed the problem as an ADIRU problem and replaced the RHS ADIRU. When the pilots tried to take off again, the stick shakers began to operate in exactly the same way, so they aborted the takeoff and taxied to an airport parking area, where they shut down the plane. Data from the flight data recorder did not show anything unusual in any of the sensors or the electronics. The next day, while the engineers were conducting a simulated flight test on the ground by forcing air past the pitot sensors, it was found that the right hand stick shaker was activated even when the AoA vane remained in the horizontal position. The RHS AoA sensor was then replaced and the system was re-tested, showing that the problem was fixed. The aircraft was returned to service, but a ten days later several diagnostic messages started to appear indicating a different type of intermittent “AoA vane” failure on the RHS. When the RHS AoA sensor was replaced again, the problem was fixed again.

The first failed sensor was returned to the manufacturer’s facility in Seattle Washington, where it failed during a test. The problem was found to be a loose main drive gear on the vane shaft that connected the vane shaft to two inductive-type resolvers and an oil-filled damper. The loose drive gear was caused by an improperly torqued set screw that allowed the vane to rotate freely 360° around the main shaft. This caused angle readings that were offset randomly from the zero position, which correlated with the continuous abnormal activation of the stick shakers. A counter-weight was also found to be loose and had a free play of about ±2° rotation.

The second failed AoA sensor was tested at the facility of its different manufacturer in the UK. It also failed its test. The problem was found to be an “open spot” in the older potentiometer-type resolver, which could have been caused by a contaminant particle acting between the brush-type angle contactor and the continuous resistor coil. This “open spot” almost certainly accounted for the intermittent failures reported by the aircraft’s diagnostic messages.

What this incident report shows is that a vane-type AoA sensor is more complicated than just a vane with a resolver on the same shaft. The additional gearing required between the vane and the resolver can be the cause of random offsets in either direction if a gear can slip on its supporting shaft. The offsets can remain constant until sufficient force is placed on the vane to cause the gear to slip on its shaft, which causes a new offset value in the sensor output. This failure mechanism may explain the 22° offset in the LHS AoA reading on the 737 MAX Lion Air JT610 flight of 29 October 2018, which caused the stick shaker to activate and remain activated for the entire flight. While this reported anomalous sensor behavior does not prove that the AoA sensor was at fault in the Lion Air incident, it does provide a working hypothesis that may be tested further.

For further information, see AAIB Bulletin 8/2008, titled G-VHOT, EW/C2006/12/01.

AoA sensor errors on Lion Air flights JT043 and JT610
 

One may ask how the error in the captain's LHS AoA sensor on flights JT043 and JT610 originated. If the error was it present when the AoA sensor was installed wouldn’t the installation test procedure have detected this error?
In response, I believe it may be possible that the testing process used after the installation of the new AoA sensor prior to the JT043 flight may have been the cause of the 22.5° offset. My reasoning is as follows.

It is known that the range of rotation of the Rosemount 0861FL AoA sensor on the B737NG and B737MAX aircraft is ±110°. The captain’s LHS sensor is the same 0861FL sensor flipped over 180° about the horizontal axis.

However, the report quoted above implies that the vane of the AoA sensor has no such end stops, because it states that if the main gear on the vane shaft becomes loose, then the vane can be rotated completely around 360°. Therefore, the end stops limiting the range of rotation of the AoA sensor must be elsewhere inside the sensor housing. Now, a review of the specifications of many types of resolvers on the internet shows that all resolvers have limitations on the ±angles they can be rotated through. This implies that the resolvers have stops inside them to prevent them from being rotated beyond their maximum angles of rotation. This implies that the maximum angle of rotation of the AoA sensor vane is limited by the stops in the resolver, and not by any stops on the vane or the shaft on which the vane is installed. Therefore, if one applies too much force to the AoA sensor vane to pin it against one of the stops, it is possible that this force can cause a slipping of the main gear on the vane shaft, causing the vane to be offset from the resolver while the resolver continues to read the same value because it is up against the stop inside the resolver. Therefore, an offset can be created between the vane angle and the resolver output angle.

Now, we know from the maintenance records for the aircraft of flights JT043 and JT610 that an installation test was done after replacing the AoA sensor before flight JT043 because the maintenance engineer noted on 27 October 2018 that: “For troubleshooting due to repetitive problem perform replaced the angle of attack sensor in accordance with Aircraft Maintenance Manual (AMM) Task 34-21-05-000-001 and task 34-21-05-400-801 carried out. Installation test and heater system test result good”.

But the Aircraft Maintenance Manual actually specifies TWO types of reference checks that can be performed: 1) A recommended test using a test fixture similar to the one shown below. (Notice that it uses the two tooling holes on the AoA sensor to register the correct angular position). A maintenance technician outside the aircraft sets the AoA sensor vane to the angles 0°, -10°, and °, respectively, and either the same technician, or perhaps a different technician, checks the output of the ADIRU to see if the same angles are provided to the SMYD display. 2) In the absence of a test fixture, a quick check can be done by setting the AoA sensor vane to the angles 0°, -100°, and +100°, the latter of which are the end stops of the vane travel. The output of the ADIRU is again checked to see if the same angles are provided to the SMYD display.

The first reference check cannot cause an offset in the vane-to-resolver output angle. However, the second reference check CAN cause an offset in the vane-to-resolver output angle if the technician setting vane angle applies too much force while setting the vane against the end stop. Specifically, if the last angle to be tested is +100°, then the AoA sensor output will be offset in the positive direction as observed in the JT043 and JT610 flights. This offset will not be observed during the test because the resolver output remains pinned at its +100° output value. Only if the last angle to be tested is different from the +100° end stop setting will an offset be observed in the AoA output during the test.

One further observation. Several posters have commented that the captain’s LHS AoA sensor that had an offset of 22° on flights JT043 and JT610 appeared to have a higher random noise on it than the F/O’s RHS AoA sensor. This may be the result of a defective viscous damper inside the captain’s LHS AoA sensor. This may indicate that the replacement LHS AoA sensor installed on flights JT043 and JT610 was, in fact, a reworked AoA sensor, which may explain why investigators want to review the procedures at the AoA sensor rework facility in Florida as well as the AoA production facility in Minneapolis. And if the sensor was a reworked sensor, perhaps the gearing between the vane and the resolver was not torqued high enough to prevent offsets from being induced by pressure on the vane against the end stops.

Double07, You mention in your post and so one would assume specific/realistic scenarios. On a much earlier post in R&N, there was a link to a basic design which had a <20V AC reference voltage in and (variable) magnetically coupled windings that were centred on zero. What it did say was there was an accurate signal up to ~35 degress, but this got less accurate at ~45. I think this was the limit of the output - wherever the physical rotation stopped.

Oil damping of such a mechanism could I suppose be total, the windings submerged. I'd like to know.

Apropos the F16. I defy any true lover of aviation to stop reading this maiden flight. Sphincter-cringing.

Featured Articles - Electric Jet - How the F-16 Became the World?s First Fly-By-Wire Combat Aircraft

I've had a disturbing idea:

IF:

- Resolver coils inside the vane are at 45 degrees with the horizontal (Forming an X , not a +)
- Software block do not stop AOA calculation if a plausibility check is failed (sin^2 + cos^2 = Vmax^2) and does a funny average (AOA= (arcsin(sin) + arccos(cos))/2) instead of AOA=atan(sin/cos)

THEN:

(real nose AOA= resolver angle - 45 deg)

A short SIN to GND or COS to GND inside any computer (stall management for instance) or connector or wiring loom would cause the reading from the vane to have an offset of 22.5 degrees when the real AOA is mostly zero.

And if it's high resistance short, as it is often the case with FOD or chaffing, it may very well pass the installation test because the coil is sending a lot of amps to the short, but burn/short the coil or something after a while (say taxi to runway). That would also explain lack of dampening (something fusing/melting/overheating inside the vane) after the flight. That would also explain why at least two different computers sensed too high AOA, which clearly points to faulty sensors, without having to believe that two sensors failed in a row without any external help, which is an impossible coincidence. That will very well match with the conductive FOD inside computer hypothesis mentioned some posts ago (A/C Cb tripping, invalid AOA, fault SMYD computer...)

question: FDR traces show nose AOA or wing AOA (=nose AOA / 2 from what I read around here)?

I really hope it's not that simple.

Loose Rivets, You asked, "Oil damping of such a mechanism could I suppose be total, the windings submerged. I'd like to know".

The oil damper (or more accurately the viscous damper) in these AoA sensors is a small cylinder filled with oil that slows down the rotation of a shaft extending from the inside of the cylinder to the outside. The outside part of the shaft holds a gear that meshes with a gear on the shaft that holds the AoA sensor vane. Therefore, the entire AoA sensor is not filled with oil, but only the cylinder of the viscous damper..

By the way, sometimes the oil leaks out of this cylindrical damper, allowing the gears and the AoA sensor vane to rotate more freely. This can cause the AoA sensor output signal to have a random noise superimposed on it.

Ecto1, You asked, "Question: FDR traces show nose AOA or wing AOA (=nose AOA / 2 from what I read around here?".

I believe that the FDR traces show the wing AOA, and that the calibration is wing AOA is approximately 0.5 x nose AOA. I do not understand the rest of your post enough to comment any more on it.

I was interested to read something that I was pondering over but could not put into words, and then someone sort of did:
I have experienced weird wiring anomalies in a number of aircraft, and a few avionics systems, that frustrate the maintenance troubleshooters for days and weeks.
I've also learned about how "certain batches of finished work" can be recalled.

If the AoA unit itself is an industry standard piece, it may be that the signal (upon arriving at the computer brain) has gone wrong. The trace that shows one AoA signal going high and one steady on makes me wonder if there isn't signal contamination ... a condition which can be a real bugger to isolate on the ground.
Just a thought.
For an automotive defect that was a real pain to trouble shoot ...
My sister in law's ford excursion (big V 8 engine) would not and could not keep the AC on. but the problem wasn't in the AC system.
One of the coils was bad, so only 7 cylinders were firing. The computer brain in the car thus cut out the AC automatically as due to the engine running roughly/badly, and all of the signals not lining up in nice lines when arriving at that little brain. It's almost as though Ford's version of HAL was saying ...

You want the AC? I can't let you do that, Danielle ...

Posting the link for you, since being a newcomer to this site, you are not yet permitted to post links: Aero 12 - Angle of Attack

Title: AoA sensor errors on Lion Air flights JT043 and JT610
 

It has now been confirmed that the replacement AoA sensor that had a 22° offset on Lion Air flights JT043 and JT610 was a reworked sensor. On 28 November 2018, investigator Nurcahyo Utomo of the Indonesian National Transportation Safety Commission (KNKT) told reporters that the AoA unit that was installed in the ill-fated airplane had previously been fixed by Boeing in Florida. He stated that the team will also visit the Florida maintenance facility to check the procedure that was used to fix the AoA. Reference: Tempco.Co article entitled “Lion Air Crash; KNKT to Further Study Recovered AoA Unit”, dated 28 November 2018.

Unless I've missed it, I haven't seen any report on whether there was a confirmed fault found with the sensor that was removed. Does anyone know?

You are correct. There has been no report to date on whether there was a confirmed fault found with the sensor that was removed.

In two previous posts above I tried to present evidence that the sensor MAY have been at fault after many other posters have been unable to explain how a large angular offset of 22° can be produced by the electronics downstream from the sensor. My post confirming that the replacement sensor was a reworked sensor was merely additional evidence that the sensor MAY be at fault. I was careless, however, in my choice of words. What I should have said was: “It has now been confirmed that the replacement AoA sensor on Lion Air flights JT043 and JT610 was a reworked sensor. This sensor was on the same LHS side as the 22° offset in the AoA angle recorded by the instrument recorder, and may have been the cause of the offset rather than the electronics.“ Even with this additional evidence, it is still not confirmed that the sensor was at fault. This is why the KNKT investigation is visiting the Florida maintenance facility to see if they can confirm that the sensor was at fault.

NYT article as of today is the first I have seen:

No more details and that is the only report, Indonesian leak, maybe a pinch of salt needed.

Also doesn't explain how a defective sensor was replaced and still the a/c is getting dud data on that side. Either there are a whole batch of defective sensors out there (so we should be seeing failure on NGs as well) or there were two problems co-occurring.

The sensor that was removed could have been defective in a way that doesn't produce a wrong AoA.
Actually the maintenance log indicates that before the AoA sensor replacement there was no "Airspeed disagree" or anything else that would indicate an "AoA disagree" but rather air data missing intermittently for the captain.

So to me it seems very likely that during that replacement the offset error/ AoA disagree was introduced.
But it's not at all clear if the new sensor was the problem or the act of replacing it introduced an cabling or other error down th eline.

Leak? More a reported statement by "Mr. Nurcahyo, the KNKT head of air accident investigations"

"Mr. Nurcahyo, the KNKT head of air accident investigations, ...

After the crash, the replaced angle of attack sensor was shipped to Minnesota, home of Rosemount Aerospace, the Boeing subcontractor that made it, Mr. Nurcahyo said. He and other Indonesian investigators went to Minneapolis in December. The sensor, he said, was deemed defective."

https://www.nytimes.com/2019/04/02/w...-lion-air.html

That's a somewhat ambiguous statement. Clearly the sensor was deemed (or at least suspected) to be defective at the time it was removed (otherwise why bother?).

So it's not clear whether the comment refers to that assessment, or to the result of subsequent examination by the manufacturer.

My two posts above explained how a large offset angle could be induced in the AoA sensor during the testing process after sensor replacement. This would have been made more likely by a failure during a rework process to tighten a set screw to a required torque in an internal gear of the AoA sensor. This led me to postulate that the replacement sensor may have been a reworked sensor, which was subsequently confirmed by an actual report. This is all consistent with the latest report that the sensor failed a test at the manufacturer in Minneapolis. This does not mean that the original sensor as manufactured by Rosemount was defective. It merely means that a shoddy rework process in a repair facility may have left the sensor susceptible to change during the installation testing process. One of my original posts above describes how this may have happened.

All of the above are theoretically possible. None have actually been demonstrated to be the case.

Double07, thanks. While it at first it may seem unlikely there could be an oil filled unit, I was mindful of oil immersed transformer windings which have stood the test of time. It also would have allowed lubrication of far fewer parts. Dash-pots (if that term is allowed for oil as well as air) are historically prone to leakage, so I'm surprised at the added complexity.

I used to reach with my hat to test the vanes and was impressed with the exquisite smoothness of the rotation.

I'd imagine opening one of these units would be akin to working on a watch. I wonder if in the history of their manufacture, it was fully realised by the technicians just how critical the output data was. I recall one of our fleet of six BAC 1-11's used to have the igniters, come at a noticeably lower airspeed than the other five. It was the first indication of AoA being too high. Shake and then Push coming next. Always hand flying to ToC it was sometimes difficult in that aircraft when hot and heavy.

I don't particulary fancy the theory of a mechanical offset because of loose bolt (slipping shaft). People do know how to manufacture square shafts and splines and keyed shafts and safety wire and plenty of other tricks. I don't know how much is one of those vane things, but no less than 3000 dollars for sure. Probably way way more. two cilindrical shafts with a screw tightening one aginst the other is not a proper way to do the job of critical torque transmission without slipping (well, maybe in toys).

In other words, I would not expect any connecting part of a rotary sensor with an all round shaft. Proper way: Splined (if it needs adjusting), keyed, or at least with a D shape.

Also, a mechanical offset does not explain the previous electric gremlins. (FEEL DIFF PRESS, SMYD computer failures and others).