Hacker News Reader

Comment by cockpitherald | Hacker News Reader

This proposed AD would require installing new flight control computer (FCC) software, revising the existing Airplane Flight Manual (AFM) to incorporate new and revised flight crew procedures, installing new MAX display system (MDS) software, changing the horizontal stabilizer trim wire routing installations, completing an angle of attack sensor system test, and performing an operational readiness flight.

https://kokpitherald.com/faa-releases-737-max-review-propose...

View on HN

Why not include a secondary angle of attack sensor. From what I hear some military aircraft have 4 of them in case of failure. I do understand that the bigger issue was not just that the angle of attack sensor failed but that the crew was not informed of how to handle the failure.

They have two AOA sensors. In a particular stroke of genius, the MCAS system would select one randomly at boot and use it exclusively (so you have twice the failure rate of a single sensor and no benefits from an extra one).

And the AOA disagree system the FAA "proposes" installing? That was already an optional extra.

Maybe I need to brush up on my probabilities, but why would you have twice the failure rate?

If each sensor failed 5% of the time and the sensor was chosen at random, wouldn’t the failure rate be the same?

It wasn't chosen at random. It swapped main FC's every boot, and each FC only took one AoA sensor as input. So if one fails, you'd only see it every other flight, and at that, only once you hit a flaps up configuration.

There was no automated cross-check out of the box.

Part of the reason for that setup was that Boeing knew if they implemented a multi-sensor solution, it'd require class D simulator training, which they were trying to avoid at all costs. See the 60 Minutes 737 MAX exposé. They apparently had a whistleblower willing to attest to it.

Imagine flipping a coin. Your chance of tails is 50%. What is your chance of a tails if you flip two coins? It's now 75% clearly (TT, HT, TH but not HH).

Now imagine it's not a coin but a normal distribution. If you sample from it twice then take the minimum of your samples, the chance that the minimum is below the mean is 75%. Just the same as with the coin but in another context.

Obviously the time-before-failure is not normally distributed, nor are the sensors completely independent random variables. But the chance of failure of the system will be higher than one sensor, not double exactly but higher.

What is your chance of a tails if you flip two coins?

Thanks for explaining, but my understanding is the computer only looks at one sensor at a time, it doesn't look at both.

It's wild that this wasn't the result out of the FAA the first pass around with these airplanes. This is basically everything Boeing wanted to and "managed" to avoid with this design, wrapped up and stamped "haha you fools don't do that". So why did it take 346 deaths?

Reply on news.ycombinator.com