Single pilot A350s, [i]There[/i]

An open discussion of aviation safety related issues.

Moderators: FrankM, el, Dmmoore

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Wed Aug 18, 2021 8:43 am

The general public has been all over self-driving cars; it's reasonable to expect the same for planes.

I don't think it's reasonable to expect software to have a lower critical failure rate than human pilots. Sure, periodically humans do foolish things that no automated system would ever get into, but computers are subject to different kinds of problems. For example: humans occasionally capture false glideslopes, and don't always notice before getting into the weeds. But computers don't notice things at all, they just chug ahead. Anything that's going to be checked has to be anticipated, and while someone's bound to think of that issue, there's thousands of other possibilities. i remember years ago someone (I think DP) describing a situation where the autoland had decided for some reason to line up 50 feet to the left of the runway (or something of the sort) and needed to be disconnected. That kind of thing is really hard for computers to deal with.

(And while training/testing the system on every reported incident and accident of the past sixty years will help, some, it does little to make the system resilient to something new. And that's assuming everyone resists the temptation to add special-case handling for every such known scenario, which is for many reasons a guaranteed way to fail.)

Meanwhile the rate of critical failures is already low enough to not be statistically significant, so it's very hard to reason about any of this in the abstract.

Regarding Tesla's autopilot... it has killed several people, but it's not so clear that it's been saving people. Last I remember the insurance statistics indicated that related fancy stuff like lane-departure warnings and autobraking and whatnot did not actually save either lives or insurance payouts.

User avatar
3WE
Posts: 8133
Joined: Wed Feb 06, 2008 2:37 pm
Location: Flyover, America

Re: Single pilot A350s, [i]There[/i]

Postby 3WE » Wed Aug 18, 2021 10:58 am

The general public has been all over self-driving cars; it's reasonable to expect the same for planes.
While interest has been high, I think the number of those against autonomous cars is pretty high. The freedom to chose different cars versus being ‘forced’ into drone aeroplanies is a different situation.
Commercial Pilot, Vandelay Industries, Inc., Plant Nutrient Division.

User avatar
elaw
Posts: 2047
Joined: Thu Jul 24, 2014 7:01 pm
Location: Massachusetts, USA
Contact:

Re: Single pilot A350s, [i]There[/i]

Postby elaw » Wed Aug 18, 2021 12:12 pm

I have to say I'm with Gabriel on this aspect.

Many people like the idea of self-driving cars because otherwise they have to drive them. For such people, a self-driving car provides tangible benefits - for example it would allow them to stare at their phone screen 100% of the day instead of just 90%. Or slightly less cynically, save them from a 1/2-hour slog through bumper-to-bumper traffic in and out of the city every work day. And in spite of the statistics, I think most people consider cars safer to be inside than an airplane.

With airplanes of course it's just the opposite. 99.9% of the people who travel on an airplane on any given day are not flying it so they don't perceive a tangible benefit from not having a pilot(s). There could be a tangible benefit in the form of lower ticket prices, but I predict the first airline that publicly says "We've removed pilots from the cockpit to save you money" will be out of business shortly thereafter.

And I agree with him on the software aspect too. With a computer-flown airplane there might be some edge-case accidents where a human could have saved it (or could not, depending on the human - United 232 is a great example) but the computer could not. But there will also be a lot of non-edge-case events that with a human could have been an accident (almost all the "what's it doing now?" for example), but thanks to HAL, end well.

But the above is strongly dependent on having software that is designed and written to a very high standard of quality - much higher than the norm for the software industry. And the question is, would that actually happen? I don't know if any of you read Byte magazine back in the day but they once ran an article titled "Bridging the 90% gap" which posited that most software does 90% of what it's supposed to do (correctly), and bringing that number to 100% can be very difficult. If the software flying a fleet of passenger aircraft only does 90% of what it's supposed to do, things are not going to end well!
HR consultant, Yoyodyne Propulsion Systems, Inc.

User avatar
3WE
Posts: 8133
Joined: Wed Feb 06, 2008 2:37 pm
Location: Flyover, America

Re: Single pilot A350s, [i]There[/i]

Postby 3WE » Wed Aug 18, 2021 1:40 pm

***Bridging the 90% gap***
Never read the article, but I chuckle…

Lotus 123 a few 360K floppies…puts numbers (and text) in rows and columns…do some figuring…make some graphs…

Today: Excel version 597 and how many zillion lines of code and Mb…INDEED, it does rows and columns and arithmetic, but it’s pretty damn cumbersome, not_intuitive and AMAZINGLY buggy.

Footnote: yeah, I skipped over VisiCalc.
Commercial Pilot, Vandelay Industries, Inc., Plant Nutrient Division.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Wed Aug 18, 2021 8:44 pm

I'd like to make a few points.

1st, the driver for a pilotless (or fewer pilots) plane is of course money, not safety or self-driving-car-like convenience.

An airline that has a good utilization of its fleet needs AT LEAT 3 crews (that is 6 pilots) for each plane assuming that the plane is being used (flying or loading / unloading pax) 16 hours per day. You have 1 crew on duty, one crew in working days but off duty hours, and one crew not scheduled for that day (rest days, vacations, training, sick).

American Airlines has almost 900 airplanes, so that's at least 5400 pilots, likely more. I have no idea how much the average AA pilot makes a year in total compensation... Say 100K? So you have like 600M just in compensation. Add other benefits that are at least partially paid by the company, employer contributions (state and federal), training (this is huge), hotels, meals and transportation when off base, HR structural costs, lawsuits, etc, etc, etc and the number will grow by a lot into the "billions" range. SO when 3WE says that pilots are an inexpensive safety device, I don't know what he is talking about.

Safety should be positively impacted by such a move, if done correctly, but the safety of transport category planes is so amazingly good these days that even magically eliminating every single death would not be a huge improvement. You will go from like 99.99999% safe to 100% safe (and no, I didn't take the number of 9s out of my hat, that's 1 in 10 million, which is about how many pax die due to an aviation accident for every 10 millions that fly).

But since we are in an aviation safety forum, let's talk about safety.

We have been taking authority and functions out of the pilot and putting it into technology since ever. The removal of the flight engineer is the most clear example, but not the only one. Hydraulics were introduced, firs with mechanical backups and now without mechanical backups. A modern airplane is unflyable unless you have both hydraulic power and electrical power. We always talk about Airbus, but the 777 and 787 are 100% fly-by-sire too, one that, unlike Airbus, emulates traditional control shapes and a handling characteristics, but the controls are still "fake", still an electronic control input device. Remove electricity, and you have nothing left. And the control surfaces are hydraulically operated, Remove hydro power, and you have nothing left. No matter how hard the pilot tries, he is powerless to control the plane without that. That was a consideration also back then, first when hydraulics were introduced, and then again when FBW became a thing. And yet we sorted it out. How many accidents have you heard of where the pilot was left powerless because the computers, the electricity or the hydraulics totally failed them?

Then we already have computers that have full authority, where the pilot has no power to bypass them. In some way, computers that control FBW could be put in this category, although the last computer that you cannot turn off and bypass by going to direct law is a very simple (I think analog) computer that just take the inputs of the sidestick and convert them directly and proportionally in hydraulic commands. But what about the FADEC? The FA stands for Full Authority. This is what seats between the thrust levers and the engines, and they are truly fuel authority, The only thing that the pilot can do if the FADEC turns off or gets stuck, is to shut the engine down. The engine is simply uncontrollable without the FADEC. How many engines shut down, let alone accidents, have you heard were caused by FADEC failures? There have been a few engines shut down because of this, but the FADECs have demonstrated to be extremely reliable, And yes, if a FADEC has a total failure, you lost an engine, but you have another one. Well, the same will have to happen with any other computer.

Finally, in cars we know that the vast majority of accidents by large are related to human factors: DUI, driver distraction, reckless driving, lack of skills, etc. That is why self driving cars will be a giant improvement in car accidents. https://www.youtube.com/watch?v=yjztvddhZmI

Well, it is not different in airplanes. Cars crash much more often that airplanes, but in those rare* occasions when airplanes do crash, it is mostly human-factor related (*rare for transport category, GA planes crash all the time, worst than cars)

These are some of the big crashes since 2018:

Crash: Petropavlovsk-Kamchatsky AN26 near Palana on Jul 6th 2021, missing aircraft impacted edge of coast. Apparently CFIT under loss of situational awareness.

Crash: Sriwijaya B735 at Jakarta on Jan 9th 2021, lost height and impacted Java Sea, One engine rolled back without the crew noticing it, the AP compensated up to a point where it gave up and disconnected. Plane rolled inverted and crashed.

Crash: PIA A320 at Karachi on May 22nd 2020, impacted residential area during 2nd approach, both engines failed as result of a gear up touchdown

Crash: UIA B738 at Tehran on Jan 8th 2020, lost height after departure, aircraft shot down by Iran's armed forces

Crash: Ethiopian B38M near Bishoftu on Mar 10th 2019, impacted terrain after departure. MCAS horribly designed, pilots could have but didn't save the day.

Crash: Atlas B763 at Houston on Feb 23rd 2019, loss of control on approach, human-induced crazy dive after GA was engaged (likely unintentionally).

Crash: Saha B703 at Fath on Jan 14th 2019, landed at wrong airport (one with a way too short runway)

Crash: Lion B38M near Jakarta on Oct 29th 2018, aircraft lost height and crashed into Java Sea, wrong AoA data, MCAS horribly designed, pilots could have but didn't save the day.

Crash: Global Damojh B732 at Havana on May 18th 2018, stall/spin t shortly after takeoff (we'll never now why)

Crash: Iran Aseman AT72 near Semirom on Feb 18th 2018, impacted terrain, flying blow MSA to avoid clouds, stalled, failed to recover.

Crash: Saratov A148 at Moscow on Feb 11th 2018, lost height after departure, pitot heat left off, crew failed to control the plane under the resulting speed disagree.

User avatar
elaw
Posts: 2047
Joined: Thu Jul 24, 2014 7:01 pm
Location: Massachusetts, USA
Contact:

Re: Single pilot A350s, [i]There[/i]

Postby elaw » Wed Aug 18, 2021 9:26 pm

Well as we've seen, an aircraft isn't always totally uncontrollable after a total hydraulic failure. I actually expect flying an airplane by engine control only is something a computer could do better than 99% of pilots - if it's possible under any given circumstances.

But there's one other interesting question that arises from your post: are electronic engine controllers powered from the aircraft's normal power sources, or independently? What I'm really asking is in a FBW aircraft if there's a total electrical failure in the airframe, is there a chance the (FADEC-equipped) engines can still be controlled? If not I'd say it's a step backwards compared to the scenario I describe above.
HR consultant, Yoyodyne Propulsion Systems, Inc.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Wed Aug 18, 2021 11:01 pm

What I'm really asking is in a FBW aircraft if there's a total electrical failure in the airframe, is there a chance the (FADEC-equipped) engines can still be controlled? If not I'd say it's a step backwards compared to the scenario I describe above.
Forget FBW, which is "stick by wire". FADEC (throttle by wire, if you will) is independent and many non-FBW planes have FADEC.

But the answer I understand is no, in the case of total electric failure the FADEC-equipped engines cannot be controlled. I believe that they will remain at whatever last fuel-metering-valve position they had.

Now, what happens with the engines would be a secondary concern, since you have just lost the ability to adjust pitch and roll too (if FBW), as well as 100% of your flight and navigation instruments.

Asking what is the B plan for a total electric failure is like asking what is the B plan for a main spar failure. It is just not supposed to happen, and the B plan is to enjoy being along for the ride the last few seconds of your life.
That is why you have 2 engines each with its own generator, a generator driven by the APU, a generator driven by the RAT, and an assortment of batteries, together with multiple main electrical system (each driven by a few of these sources) that are protected and isolated.

I have heard of a couple of accidents / incidents related to total loss of hydraulics. I have never heard of one due to total loss of electricity.

User avatar
3WE
Posts: 8133
Joined: Wed Feb 06, 2008 2:37 pm
Location: Flyover, America

Re: Single pilot A350s, [i]There[/i]

Postby 3WE » Thu Aug 19, 2021 2:32 am

I have never heard of one due to total loss of electricity.
We have a rich aviation history in greater Flyover.

http://libraryonline.erau.edu/online-fu ... R85-03.pdf
Commercial Pilot, Vandelay Industries, Inc., Plant Nutrient Division.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Thu Aug 19, 2021 4:41 am

I have never heard of one due to total loss of electricity.
We have a rich aviation history in greater Flyover.

http://libraryonline.erau.edu/online-fu ... R85-03.pdf
Nice find. A convoluted one, too. The total loss of electricity, that led to losing control of the plane in IMC after the attitude indicator stopped working, was the result of:
- Mechanical failure of the left-engine generator.
- The FO incorrectly turning off the right-engine generator.
- Their election of continuing to their destination, which they knew was IMC, and they estimated was a bit more that 30 minutes away, relying only on the battery that was certified to provide at least 30 minutes of essential power, instead of returning to the airport that they had just taken off which was 1 circuit pattern away in VMC.
- They failed to bring the right generator back on line (because they didn't follow procedures and checklists)
- They didn't shed as much non-essential load as they could have to extend the battery endurance (because they didn't follow procedures and checklists)

The first problem was a technical fault. The other 4 were human faults that a computer would not have made.

Unlike what we discussed above, this airplane doesn't rely on electric power for flight controls, so had they been in VMC they could have continued flying VFR (the FO's air data instrument didn't need electricity to fly, so they had FO's airspeed, altitude, and vertical speed; they also had wet compass). However, losing the attitude indicator in IMC is more or less than losing the flight controls: You cannot control the plane in either case.
Not that for a computer lacking a source of electrical power being VFR would have been much of an advantage, though.

User avatar
3WE
Posts: 8133
Joined: Wed Feb 06, 2008 2:37 pm
Location: Flyover, America

Re: Single pilot A350s, [i]There[/i]

Postby 3WE » Thu Aug 19, 2021 12:17 pm

Nice find. A convoluted one, too.
Most are.
Commercial Pilot, Vandelay Industries, Inc., Plant Nutrient Division.

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Thu Sep 16, 2021 2:25 am

I could go through the last couple months of AVH and make a list of minor occurrences that a computer could turn into a catastrophe, but there's no real point. Ultimately there are two fundamental problems: (a) computers run software and software has bugs, and (b) computers have no judgment, common sense, or ability to improvise.

There is some progress on "AI" techniques that can provide some semblance of judgment, sometimes, but there are at least two problems with using these in safety-critical systems anytime in the reasonably foreseeable future: one is that they aren't reliable, and the other is that it's in general not possible to account for why they choose what they choose; they are black boxes (not in the flight recorder sense) and there's basically zero assurance they won't at some point run off the rails. Sure, you can have a human supervising them, but then you're back to having a pilot, except with an even worse case of not having currency in stick & rudder skills than already happens today.

There are people working on more transparent AI/learning mechanisms, sometimes with some success (but mostly not so much) but even so, that's current research, not anything you'd want to count on in production anytime soon.

All the automation we currently have reverts to manual control if it finds itself broken or encounters a situation it doesn't understand or is outside its designed operating environment. Removing that last recourse is a very big deal.

(And regarding total electrical failure, https://avherald.com/h?article=4c1cc3f6&opt=0 -- that could easily have been fatal, especially if it happened to a crew of AF447 caliber, and would almost certainly be fatal for an automated aircraft.)

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Thu Sep 16, 2021 9:15 am

I could go through the last couple months of AVH and make a list of minor occurrences that a computer could turn into a catastrophe, but there's no real point. Ultimately there are two fundamental problems: (a) computers run software and software has bugs,
There ways around that, from formal verification to bug-resilience (like writing several different programs by different teams using different coding languages and running them on different hardware, but all under the same functional specs)
and (b) computers have no judgment, common sense, or ability to improvise.
Is that a good or a bad thing, accident wise? How many accidents were CAUSED by bad judgment (or judgement at all instead of just following established procedures with a cold blood), improvisation, and what seemed to be common sense that in the aftermath and in hindsight ended up not being so sensible?
All the automation we currently have reverts to manual control if it finds itself broken or encounters a situation it doesn't understand or is outside its designed operating environment. Removing that last recourse is a very big deal.
Yes. Currently. You would not just remove this last resource in a current design.
(And regarding total electrical failure, https://avherald.com/h?article=4c1cc3f6&opt=0 -- that could easily have been fatal, especially if it happened to a crew of AF447 caliber, and would almost certainly be fatal for an automated aircraft.)
Well, 2 things: 1) the Boing 777 was NOT designed to be autonomous and 2) even in this incident it was STILL being flown through computers.

There are 2 things that I am confident: 1) A properly designed autonomous plane will still surprise us with new "creative" ways of crashing. 2) A properly designed autonomous plane will avoid more crashes (that today happen due to human factors) than the new crashes it will cause.

Yes. Even with today's technology (meaning hardware and software that could be developed today, not hardware and software existing in airplanes today),

I said it there and I keep it here: The main limitation to an autonomous plane are not technical but the public resistance, the certification complexity, and liability concerns (a human error is paid by the airlines, a computer error will be paid by the plane manufacturer, so for Airbus and Boeing it is cheaper to let humans keep crashing planes rather than computers, even if humans crash 10 times more often).

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Tue Sep 21, 2021 9:37 am

There ways around that, from formal verification to bug-resilience (like writing several different programs by different teams using different coding languages and running them on different hardware, but all under the same functional specs)
I'm aware of that... multi-version programming does not work as well as you might think in practice, because people tend to reinvent the same problems.

Formal verification is great, but it's still extraordinarily expensive and requires computer science PhDs who can also write code, of which there's not a large supply. And even then, it's only as good as the specification, and unfortunately things like "does not crash" are kinda difficult to specify formally.
and (b) computers have no judgment, common sense, or ability to improvise.
Is that a good or a bad thing, accident wise? How many accidents were CAUSED by bad judgment (or judgement at all instead of just following established procedures with a cold blood), improvisation, and what seemed to be common sense that in the aftermath and in hindsight ended up not being so sensible?
It is unknown, maybe unknowable. Like I said, I could go make a list of incidents from AVH that a computer could turn into a catastrophe. However, ISTM that if you go through AVH and count up the incidents where the consensus of serious opinion is people shaking their heads (like e.g. the false glideslope capture one a few days ago) and the ones where the consensus of serious opinion is people clapping, I think you'll find more of the latter.
All the automation we currently have reverts to manual control if it finds itself broken or encounters a situation it doesn't understand or is outside its designed operating environment. Removing that last recourse is a very big deal.
Yes. Currently. You would not just remove this last resource in a current design.
Well, obviously. And yet, there's no silver bullet for doing it, either.
(And regarding total electrical failure, https://avherald.com/h?article=4c1cc3f6&opt=0 -- that could easily have been fatal, especially if it happened to a crew of AF447 caliber, and would almost certainly be fatal for an automated aircraft.)
Well, 2 things: 1) the Boeing 777 was NOT designed to be autonomous and 2) even in this incident it was STILL being flown through computers.
Sure. But the important point is: Boeing thought it was impossible. Therefore, there would have been no procedure for an automated aircraft's computer to follow, so no matter how carefully you prepared it, it would be needing to make things up as it goes along, and computers really can't do that.

(except to some extent with deep learning stuff, but as I was saying before there are good reasons not to allow that anywhere near operational control of anything)
There are 2 things that I am confident: 1) A properly designed autonomous plane will still surprise us with new "creative" ways of crashing. 2) A properly designed autonomous plane will avoid more crashes (that today happen due to human factors) than the new crashes it will cause.
I can agree on (1) but not (2), although there's an awful lot of wiggle room in that "properly".

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Tue Sep 21, 2021 5:20 pm

A few remarks (mostly reiterations).

  • We have to take inti account not only the new crashes the technology will create but the current crashes that it will avoid. In most crashes today human factors is still a causal factor. Distractions, lack of discipline, fatigue, lack of skills, simple errors, not following procedures, loss of situational awareness, startle, tunnel vision, confirmation bias, gethereitis, panic, suicide, etc, etc, etc... A computer is not susceptible to any of those.
  • While we can have an idea of the accidents that would be avoided removing the pilot, by looking at current statistics, we cannot do the same with the accidents that would be caused by computers by looking at the current statistics, because airplanes today are not designed to be autonomous and all the automation is supposed to be there to help the pilot, who remains in charge at all times (can switch automation off) and who, as per the airplane design, is the last resource in case of automation failure. Obviously these "features" would not be part of an autonomous planes.
  • Historically, every increase in automation comes with an increased safety record (and with new accidents that would not have happened without the automation, but which are much fewer than the former)
  • I agree that I don't want a deep-learning AI system making judgement calls and taking decisions in a cockpit. An adaptive control system is good, but that's it.

User avatar
elaw
Posts: 2047
Joined: Thu Jul 24, 2014 7:01 pm
Location: Massachusetts, USA
Contact:

Re: Single pilot A350s, [i]There[/i]

Postby elaw » Tue Sep 21, 2021 6:16 pm

But just think, we already have this: https://www.npr.org/2021/06/01/10021962 ... tonomous-d

:(
HR consultant, Yoyodyne Propulsion Systems, Inc.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Tue Sep 21, 2021 8:11 pm

Well, that type of autonomous flight does pose a safety risk that can result in many casualties....

.... by design.

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Sun Dec 12, 2021 1:06 am

  • While we can have an idea of the accidents that would be avoided removing the pilot, by looking at current statistics, we cannot do the same with the accidents that would be caused by computers by looking at the current statistics, because airplanes today are not designed to be autonomous and all the automation is supposed to be there to help the pilot, who remains in charge at all times (can switch automation off) and who, as per the airplane design, is the last resource in case of automation failure. Obviously these "features" would not be part of an autonomous planes.
That's true to some extent, but it's also not hard to find occurrences that seem very difficult for a computer to handle given what, realistically, computers can and cannot be expected to do. This one seems like a good candidate, for example: https://avherald.com/h?article=4f10cac3&opt=0 ("S7 A21N at Magadan on Dec 2nd 2021, unreliable airspeed")

Then again, pretty sure that one would have killed a lot of human crews as well.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Sun Dec 12, 2021 6:17 am

  • While we can have an idea of the accidents that would be avoided removing the pilot, by looking at current statistics, we cannot do the same with the accidents that would be caused by computers by looking at the current statistics, because airplanes today are not designed to be autonomous and all the automation is supposed to be there to help the pilot, who remains in charge at all times (can switch automation off) and who, as per the airplane design, is the last resource in case of automation failure. Obviously these "features" would not be part of an autonomous planes.
That's true to some extent, but it's also not hard to find occurrences that seem very difficult for a computer to handle given what, realistically, computers can and cannot be expected to do. This one seems like a good candidate, for example: https://avherald.com/h?article=4f10cac3&opt=0 ("S7 A21N at Magadan on Dec 2nd 2021, unreliable airspeed")

Then again, pretty sure that one would have killed a lot of human crews as well.
Climb thrust, pitch 10 degrees, wings level or in turns are needed, limit bank to 15 degrees.
That is easy for a computer.

Instead, we got this:
according to FDR the roll angles varied between 49.8 degrees and -91.1 degrees, pitch angles from 43.8 to -23.9 degrees
That is human performance.

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Fri Dec 24, 2021 7:53 am

Eh. Airframe covered with ice, controls don't respond like they're supposed to, computers can't adapt. With a human crew, you might die. With a computer, you will: increasingly wild oscillations until something breaks.

User avatar
Gabriel
Posts: 3659
Joined: Sat Feb 09, 2008 2:55 am
Location: Buenos Aires, Argentina

Re: Single pilot A350s, [i]There[/i]

Postby Gabriel » Fri Dec 24, 2021 8:02 am

Eh. Airframe covered with ice, controls don't respond like they're supposed to, computers can't adapt. With a human crew, you might die. With a computer, you will: increasingly wild oscillations until something breaks.
Except there is a thing called adaptive control.

User avatar
ocelot
Posts: 689
Joined: Thu Jan 15, 2009 9:26 pm
Location: /bin/cat

Re: Single pilot A350s, [i]There[/i]

Postby ocelot » Sat Feb 19, 2022 12:01 pm

Eh. Airframe covered with ice, controls don't respond like they're supposed to, computers can't adapt. With a human crew, you might die. With a computer, you will: increasingly wild oscillations until something breaks.
Except there is a thing called adaptive control.
Not holding my breath.

(It seems that incident was somewhere in between what we each thought, btw. But there are plenty of incidents.)


Return to “Aviation Safety Discussion Forum”

Who is online

Users browsing this forum: No registered users and 7 guests