This concerns the peculiar case of Aeroflot Flight 593, which crashed in March 1994 after the pilot let his kids take the controls. The pilot intended to let his kids think they were controlling the plane, surreptitiously adjusting the autopilot heading while the kids each applied a stick input. But the son, who went second, evidently applied enough stick force to temporarily override the autopilot, sending the plane out of control. The pilots then struggled to regain control for a couple of minutes before crashing, killing all aboard.
The key sentence that caught my eye was this:
ISTR that the F-117 was equipped with a “panic button,” such that if the pilot became disoriented during night operations, he could hit the button and the plane would restore itself to straight, level flight. If that’s true, then it showed that it’s possible for certain autopilot systems to recover from arbitrary flight attitudes. However, here’s an animation, with CVR audio, showing what the Aeroflot plane (an Airbus A310) went through in the minutes before the crash; the chaos begins around 0:52.
Would it have been possible for the autopilot on the A310 to recover from the crazy gyrations the pilots had put it through before the crash, if they had simply gone hands-off? Moreover, would it have done so in a usefully rapid manner, or is it programmed for only very gentle corrective actions?
I’d like to see a cite for the purported F117 capability.
USAF is just now testing an emergency recovery system for fighters in general that will take over when it computes you’re about to become committed to impacting the ground. That will be relevant to G-induced loss of consciousness, spatial disorientation in clouds, and also just plain old goofs of perception about turn radius required versus turning room available.
That system is also aimed at aircraft that are vastly more capable of flying any which way versus a transport.
The whole point of Airbus’ automatic envelope protection is that transport aircraft can be flown into attitudes and speeds from which a recovery is physically impossible no matter how much altitude is available. So the only way to win that scenario is to not get there in the first place. IOW, there are points of no return that you can’t fly across & have the airplane survive. Note also that Airbus envelope protection is not a feature of the “autopilot”, but rather the flight control system. It’s lower in the hierarchy of systems.
As a general rule, *current *airliner autopilots are designed for smooth gentle flying through normal maneuvers. They are NOT able to bank rapidly or pull Gs sufficient to recover from weird attitudes quickly enough to prevent severe over/underspeed situations.
Part of the reason for this is that autopilots malfunction precisely by making weird unexpected maneuvers. They are designed to not have the ability to “yank & bank” precisely so they can’t fail by putting the aircraft into jeopardy faster than the humans can recover the situation. In early days they’d often refuse to let go when told to, resulting in pilots having to physically overpower the autopilot’s inputs to the same control mechanism. So autopilots’ “muscles” are deliberately weak so one or two humans can phsyically override them, for hours if need be. Remembering human’s arms eventually get tired, whereas most mechanical servos don’t.
There are (possibly apocryphal) stories of Airbus autopilots refusing to disconnect as well. But in that case it leaves the pilots unable to use their controls to steer the airplane since it all flows through the same confused computer.
As well, autopilots are designed to disconnect whenever they sense an out of tolerance condition within their inputs. In general they lack the redundancy to deal successfully with malfunctioning sensors. e.g. the response to the various airspeed sources’ data values diverging is to disconnect and sound the alarm: “Hey pilots, the airspeeds are wonky & I don’t know which one to trust; here … you fly & figure it out; I’m gonna take a nap.”
In any severe aircraft gyration it’s common for one or another data sensor to deliver flaky data, at least for a few seconds. So it’s almost a sure thing the autopilot will drop off during, say, a stall.
*Could *these systems be made smarter? Yes, eventually. I expect that in 200 years piloting will be a thing of the past & transport aircraft will be flown solely by computers. But I’m not concerned that my job will be automated away in the decade I have left to work. I do expect some computerized disruption to the careers of the young guys now in college.
The typical fighter/attack autopilot was (is?) pretty minimal, good enough to fly a straight & level line and that’s about it. What that book describes is closer to a typical airliner system in that it couples to the nav system for preplanned route, altitude & speed control. PLUS the key feature we’re discussing, automated recovery from confused flight.
I suppose the key difference beween that and the new USAF system I described is the new system is meant to be always on / always armed. And yet it needs to be unobtrusive that it doesn’t overreact to deliberately aiming the aircraft at the ground as is often done in bombing, strafing, low altitude terrain masked flight, etc.
So what I’m getting from this is that despite the claim on the Wikipedia page, it’s highly unlikely that the Aeroflot plane would have self-recovered if the pilots had gone hands-free. Is that an accurate assessment of your post?
Overall my post was about generalities, not that specific aircraft type or accident.
I’ve now actually read that wiki article & am not impressed with that particular cite, coming from a sensationalist TV show as it does.
The A310 was an earlier Airbus design that lacked their signature full-up serious automatic envelope protection. That innovation was introduced with the A320 and all higher numbered models now have it.
All big jets even from the 1960s have features intended to warn of stall or overspeed, and some may have simple-minded stall preventers that would fire in some situations and would be expected to be helpful in a benign straight-ahead stall, but probably not in a crazy upset like they experienced.
In 1960s autopilot designs it was common to be able to engage roll or pitch control separately. Which design has since been deprecated since it enables exactly the problem they got into: if the autopilot is controlling only pitch, it thinks pull back always equals nose goes up & speed goes down. But that’s only true if your bank is near zero. If you manually roll into 60+ degrees of bank, it deciding to pull isn’t going to help & will probably make the situation worse.
Overall I *suspect *the authors of that TV show confused the various models of Airbus and are making unwarranted assumptions which the wiki author parrots without understanding.
So that’s my working conclusion, based on my background knowledge and not much research. It’s worth 100% of what you paid for it.
The likely best answer would come from reading the actual accident report & consulting an A310 aircrew operating manual. Although as noted in the wiki article, there are lots of optional features on any modern jet and there’s no guarantee the manual we might locate online exactly matches the equipment installed on the accident aircraft on that date.
Agree with the above. Unfortunately the accident report is in Russian so it would be difficult to get a clear answer.
Reading a little about the A310 at www.smartcockpit.com it appears that the autopilot roll servo can be temporarily overridden by applying a force greater than 15kg to the control wheel. This function is to allow a pilot to override a malfunctioning autopilot.
I think this may be how the pilot’s son disconnected the roll servo only, in which case if they had let go of the control wheel early in the sequence of events then the roll servo would have re engaged and normal flight would have resumed.
So the statement is correct but with some conditions that aren’t made clear in the wiki article.
I wasn’t able to listen with the sound on, but reading the subtitles really raises questions about the pilots’ abilities. Arguing about whether to roll left of right, arguing about the throttle, arguing about pulling up, etc.
I pictured yakkity sax playing in the cockpit as the pilots fought each other’s control inputs for 2.5 minutes of plummeting.
Hate to speak ill of the dead, but gosh, what idiots.