Boeing Unveils 737 Max Software Fixes (cnbc.com)

hcs_$reboot shares a report from CNBC: Boeing previewed its software fix, cockpit alerts and additional pilot training for its 737 Max planes on Wednesday, saying the changes improve the safety of the aircraft which has been involved in two deadly crashes since October. By the end of this week, Boeing plans to send the software updates and plan for enhanced pilot training to the FAA for certification approval. After the FAA approves the fix, Boeing said it will send the software update to customers. Among the notable changes to the MAX flight controls:

• The plane's Maneuvering Characteristics Augmentation System, or MCAS, automated flight control system, will now receive data from both "angle of attack" sensors, instead of just one.
• If those disagree by more than 5.5 degrees, the MCAS system will be disabled and will not push the nose of the plane lower.
• Boeing will be adding an indicator to the flight control display so pilots are aware of when the angle of attack sensors disagree.
• There will also be enhanced training required for all 737 pilots so they are more fully aware of how the MCAS system works and how to disable it if they encounter an issue.

Comments

Sensors are physical objects

[Firethorn]

Because the sensors are physical devices, and are this subject to all physical device problems. They can break, corrode, be bent by a physical impact, etc...

They're regularly inspected, which is about the best you can do.

Re:Sensors are physical objects

[PPH]

It might not be the physical sensor. Data from both the LION and Ethiopian flights shows an offset between the two AoA sensors of 22 degrees. Neither appear to be stuck, as they both track airplane movements. But with this offset. Same physical fault causing the exact same offset? Doubtful.

One theory is that the 22 degree figure is pretty close to the value of one bit in the ARINC 429 word for AoA (22.5 degrees). So, software might be flipping a bit. This might be a tough bug to run down.

Re:Sensors are physical objects

[Firethorn]

I was keeping my response simple, but for "flight critical" sensors the general idea is to have at least 3 and use a voting system. For sensors that are 99.X% reliable, the odds that two will be out such that they are throwing the same erroneous value(or at least within error margins) is quite low. Though there are differences between 'simple' sensors that report back a simple voltage or resistance where determining a fault can be difficult, and complex ones like radar, GPS, that are more likely to tell the system they have a problem. The vanes here are simple sensors.

Though with the MCAS it was supposed to assist, not be critical, thus 1 vane being enough. Pilots were supposed to be able to override with just more stick application. That assessment is being challenged, and the 2 vane + alarm thing is Boeing hoping to avoid having to avoid installing another sensor for proper 3 sensor + voting reliability, as the extra sensor will be expensive.

3 good sensors: all good
2 good sensors: all good(less redundancy)
2 good sensors, 1 whack - get fixed after landing
1 good, 1 whack - system unreliable, turn off. Consider landing early.
1 good - 2 whack(different values) - system unreliable, turn off, consider landing early
1 good - 2 whack(same values) - hope you notice before crash/fire. Turn off system. Seriously consider landing early. Last good sensor may or may not be usable(does it have an output you can use?). Consider firing maintainers as it is likely at least one was whack when you took off.
0 good - 2 whack(same values) - same as previous, really. Without minor hope of good sensor being useable.
3 whack - same as previous. Consider firing maintenance department out of a cannon.

Re:Sensors are physical objects

[dgatwood]

It might not be the physical sensor. Data from both the LION and Ethiopian flights shows an offset between the two AoA sensors of 22 degrees. Neither appear to be stuck, as they both track airplane movements. But with this offset. Same physical fault causing the exact same offset? Doubtful.

One theory is that the 22 degree figure is pretty close to the value of one bit in the ARINC 429 word for AoA (22.5 degrees). So, software might be flipping a bit. This might be a tough bug to run down.

It seems unlikely that software would suddenly start flipping a bit repeatedly. That usually implies faulty hardware. The real question is how two pieces of hardware could experience the exact same fault on exactly the same bit.

My money is on thermal expansion of a BGA fastened with lead-free solder.

Re:Sensors are physical objects

[viperidaenz]

No, it's intended to stop a stall from happening by automatically adjusting the stabilizer trim as the elevators don't have enough pitch authority to counteract the pitch-up caused by the more powerful engines.

The system is intended to allow the plane to be certified without redesigning the elevators.

Re:Sensors are physical objects

[PPH]

The LION plane had an AoA system problem on a previous flight. The sensor was replaced. It appears that didn't fix it.

Re:Sensors are physical objects

[Darinbob]

Just your industry standard screwup. A better design is expensive, more testing is expensive, any delay is expensive. To the product managers will push and push and push for you to ship the product. The plan was not designed from scratch, it's an incremental modification of the 737 line and this feature was essentially a patch that was less expensive than a redesign.

Re:Encouraging news. Still nervous.

[viperidaenz]

The amount of pitch up with the newer more powerful engines got to a point where when the plane is already at a high angle of attack, the elevator don't have enough authority to counter act it. The entire rear stabilizer needs to be moved using the stabilizer trim.

Other planes have larger elevators or less pitch-up under full thrust.

A software fix could have used both sensors?

[PhotoGuy]

The depressing (or incriminating?) part here is that the fix didn't require any hardware modifications, as I would have expected. I assumed that there was some cost/weight issue to having the MCAS have access to the left and right sensors. But nope, it could have compared both.
If it can be fixed with a software fix, then it could have been done right from the start without any extra hardware costs of production.
Very damning.
I get so tired of the reports calling clear software/algorithm bugs "computer glitches."
It's akin to blaming every pilot error situation on the plane.
Just as with hardware design flaws, software design flaws should have repercussions for the manufacturer, and not written off as "oh, one of those computer glitches!" If your computers are glitchy, don't put them on my plane, thanks.