3 ms·
> Are you talking about the flight control computers? One out of three of them, yes. That's what happened in the Quantas flight. Would Direct Law still work un
by pdonis 10y ago
> Are you talking about the flight control computers?
One out of three of them, yes. That's what happened in the Quantas flight. Would Direct Law still work under those conditions?
> If one of the flight control computers crashes, then that is one way of triggering a switch from Normal Law to Alternate Law.
Should that have happened in the Quantas flight described in the article?
> Under Direct Law the pilots inputs are directly translated to control surface deflections.
Hm. If that's true then the obvious thing for the pilots to do would have been to go into Direct Law mode, but I don't see that mentioned in the article.
- foldr 10y ago>One out of three of them, yes. That's what happened in the Quantas flight. Would Direct Law still work under those conditions? Yes, of course. I'm not sure what you're thinking is the alternative. If the aircraft stopped being controllable even under Direct Law when one of the flight control computers failed, then there wouldn't be any redundancy. >Should that have happened in the Quantas flight described in the article? No, because none of the flight control computers crashed. The issue was caused by bad sensor data. >Hm. If that's true then the obvious thing for the pilots to do would have been to go into Direct Law mode, but I don't see that mentioned in the article. It is not what they are trained to do, and probably isn't really a sensible response. Flying the aircraft without any protections isn't particularly safe either.
- pdonis 10y ago> none of the flight control computers crashed. The issue was caused by bad sensor data. I think you're confusing the Quantas flight with Air France 447. From the article on the Quantas flight, after the first dive: "One of the aircraft's three flight control primary computers – which pilots refer to as PRIMs – is faulty. They begin to reset it by flicking the on-off switch." And later: "After QF72's second dive, the number three flight control primary computer faults again."
- pdonis 10y ago> Flying the aircraft without any protections isn't particularly safe either. The plane was flying straight and level when the flight control computer crash caused the automatic control to start doing obviously wrong things. Going to Direct Law at that point would have allowed the pilot to reestablish straight and level flight while they figured things out. That seems like it would have been safer than what happened. If they weren't trained to do that, maybe their training needs to be changed.
- foldr 10y ago>Going to Direct Law at that point would have allowed the pilot to reestablish straight and level flight while they figured things out. That seems like it would have been safer than what happened. It may have been safer in this specific incident. The question is whether it would make things safer to tell pilots to switch to direct law whenever something occurs that they think might be due to a problem with the flight control computer. I suspect you would just see lots of accidents causes by the absence of the protections. It's worth noting that an Airbus has never had a fatal accident as the result of this kind of incident. So the correct response to is probably no response at all. It's easy to get fixated on these kinds of incidents because there's something terrifying about the idea of a crazy computer overruling the pilots, but the actual chance of a fatal accident is way lower than the chance of an accident caused by pilot error owing to an absence of protections. Thanks for the correction about the flight control computer. I'm not sure how reliable the technical info in the article is, though.
- pdonis 10y ago> The question is whether it would make things safer to tell pilots to switch to direct law whenever something occurs that they think might be due to a problem with the flight control computer. In this case the pilots didn't "think" there was a problem with the flight control computer; they knew it, because they were given an indication that it had faulted. It was that positive indication plus the obviously wrong behavior that made it a sound conclusion that the flight control computer was causing the problem. That's a much narrower rule to follow than just "whenever you think it might be a problem with the flight control computer". > It's worth noting that an Airbus has never had a fatal accident as the result of this kind of incident. It depends on how narrowly you define "this kind of incident". If you define it as "flight control computer fault", then you're correct. But if you define it as "airplane not helping pilots to deal with a sudden emergency", then I think Air France 447 falls into the same category. (There has been a fair bit of discussion of that incident elsewhere in this thread.) See further comments below. > the correct response to is probably no response at all "Correct" by what standard? Yes, nobody died as a result of this Quantas incident, but plenty of people were injured, and it could have been worse. Also, in more general terms, if the correct response is always no response at all, what is the human pilot there for in the first place? The reason to have humans in the cockpit is that there will be situations where the computers cannot provide the correct response, and doing nothing at all endangers the passengers. Air France 447 is an example of that. If you're going to have humans in the cockpit to deal with situations of this kind, the airplane needs to be designed with that in mind. I'm not sure that design criterion is being given enough weight. > the actual chance of a fatal accident is way lower than the chance of an accident caused by pilot error owing to an absence of protections Averaged over all flights, yes, of course that's true. But we're not talking about all flights. We're talking about a small subset of flights in which humans, if given proper information, can potentially do better than computers at making the correct response. Unless that subset is empty, which I strongly doubt, there is a potential safety gain in identifying such situations and designing for them.