Automation dependency stripped of political correctness.
Jwscud, I don’t know of any particular research re the effect of technology on reducing the accident rate (N.B. not autoflight). Of the few documents which profess trends most restrict their views to one particular subject, which may or may not be correlated with other contributors.
The difficulties in multivariable analysis are often reflected in accident reports which state a wide range of recommendations covering a raft of issues, but without individual prioritisation.
It’s difficult to visualise complex interactions; NASA has attempted this in the previous referenced Aircraft Loss of Control Causal Factors and Mitigation Challenges; however the analysis is based on the outcome and not the contribution of technology or safety activity.
An overriding problem in this is how safety is perceived - predominantly by accidents, failures; whereas the objective is to eradicate accidents – no failures, which if achieved might represent ‘no safety’; as there shouldn’t be anything to measure. Hollnagel discusses this in many documents on Resilience, Safety 1 and Safety 2.
Airbus’s refreshingly different analysis of Commercial Aircraft Accidents 1958-2013 compares categories of accidents vs the advancing technical generations of aircraft.
Thus the reduction in CFIT and LoC accidents might be related to TAWS and Envelope Protection, whilst with little technical activity Runway Excursions remain high (page 15).
However there are inconsistencies; even with the later implementation of TAWS, why does second generation CFIT rate remain high (page 16). This quandary might be more marked by the apparent increasing LoC and Runway Excursion in second generation aircraft which might be as expected by a ‘U’ shaped distribution with age and lack of technical intervention; older aircraft, operating environment, engineering standards, etc. (pages 17-18).
This might indicate that there are greater non-technical issues involved.
I can see the attraction of including CRM in the list, but there may be greater safety value in considering the wider social/operational environment (discussed in previous posts) which may have better capability to influence human behaviour, i.e. in creating the situation which the human has to assess, understand, and manage.
Alternatively, both improved technical and 'non CRM' human input can create safety (the opposite measure of safety as in alf’s designing an accident) by shaping the operational environment. This has some synergy with Engineering Resilience – to build something, but this too involves human activity much earlier in the process – regulation, training, normal behaviour; cf previous hypothetical accident. CRM for regulators and managers; and technology to aid, shape, and guide human behaviour: - decison aiding, not decison making.
The difficulties in multivariable analysis are often reflected in accident reports which state a wide range of recommendations covering a raft of issues, but without individual prioritisation.
It’s difficult to visualise complex interactions; NASA has attempted this in the previous referenced Aircraft Loss of Control Causal Factors and Mitigation Challenges; however the analysis is based on the outcome and not the contribution of technology or safety activity.
An overriding problem in this is how safety is perceived - predominantly by accidents, failures; whereas the objective is to eradicate accidents – no failures, which if achieved might represent ‘no safety’; as there shouldn’t be anything to measure. Hollnagel discusses this in many documents on Resilience, Safety 1 and Safety 2.
Airbus’s refreshingly different analysis of Commercial Aircraft Accidents 1958-2013 compares categories of accidents vs the advancing technical generations of aircraft.
Thus the reduction in CFIT and LoC accidents might be related to TAWS and Envelope Protection, whilst with little technical activity Runway Excursions remain high (page 15).
However there are inconsistencies; even with the later implementation of TAWS, why does second generation CFIT rate remain high (page 16). This quandary might be more marked by the apparent increasing LoC and Runway Excursion in second generation aircraft which might be as expected by a ‘U’ shaped distribution with age and lack of technical intervention; older aircraft, operating environment, engineering standards, etc. (pages 17-18).
This might indicate that there are greater non-technical issues involved.
I can see the attraction of including CRM in the list, but there may be greater safety value in considering the wider social/operational environment (discussed in previous posts) which may have better capability to influence human behaviour, i.e. in creating the situation which the human has to assess, understand, and manage.
Alternatively, both improved technical and 'non CRM' human input can create safety (the opposite measure of safety as in alf’s designing an accident) by shaping the operational environment. This has some synergy with Engineering Resilience – to build something, but this too involves human activity much earlier in the process – regulation, training, normal behaviour; cf previous hypothetical accident. CRM for regulators and managers; and technology to aid, shape, and guide human behaviour: - decison aiding, not decison making.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Wiggy
Well, there are various mature technologies available to both "see" and range-find. Some kind of combination?
Would a computer be any worse at judging these things than a human?
I'm thinking that any autonomous aircraft will of necessity have to use one of the various tech solutions for see and avoid and these should read across onto runway surveillance.
Even a laser rangefinder would do the job.
https://www.youtube.com/watch?v=_qah8oIzCwk
Yes, the system on this video is crude, but it is entirely visual based and incredibly small.
More importantly it is what a bunch of kids and hundreds of pounds can do not Boeing/Airbus with billions of pounds.
Well, there are various mature technologies available to both "see" and range-find. Some kind of combination?
Would a computer be any worse at judging these things than a human?
I'm thinking that any autonomous aircraft will of necessity have to use one of the various tech solutions for see and avoid and these should read across onto runway surveillance.
Even a laser rangefinder would do the job.
https://www.youtube.com/watch?v=_qah8oIzCwk
Yes, the system on this video is crude, but it is entirely visual based and incredibly small.
More importantly it is what a bunch of kids and hundreds of pounds can do not Boeing/Airbus with billions of pounds.
Last edited by Tourist; 21st Jan 2016 at 08:58.
Would a computer be any worse at judging these things than a human?
As AI stands for your idea to work you've pretty much got to forsee and code for all the "unknown unknowns" out there........Are you willing to guarantee you can do that with 10 to the 9 reliability or better...or would it be simpler and cheaper to stick with the idea of a couple of Human Beings up there to try and handle the WTFs, aided by better automation/technology...
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Under that circumstance, the computer will do exactly the same thing as a human pilot.
Crash.
As AI stands for your idea to work you've pretty much got to forsee and code for all the "unknown unknowns" out there........Are you willing to guarantee you can do that with 10 to the 9 reliability or better...or would it be simpler and cheaper to stick with the idea of a couple of Human Beings up there to try and handle the WTFs, aided by better automation/technology...
Safetypee - thank you for the fascinating reading.
Tourist - regarding your idea of a dynamic V1, I agree it is technologically possible. The main question any good analysis would ask is is it worth it? Perf A does not exactly provide massive margins at the moment (look at wet takeoff performance on a limiting runway for example) so I fundamentally don't see there is much more than an extra few hundred kilos of payload to get out of it.
In contrast, the additional risks from cutting margins entirely to the bone is in my view not warranted. Why would anybody waste billions of dollars certifying their dynamic stop-go aid when they could just program HAL with Perf A?
There is a small subset of types and fields where it may be useful - I have operated bizjets capable of sub-1000m landing rolls off 3000m runways, and am well aware that the aircraft could safely be stopped past V1/Vr (as the two are normally the same) given one has 1800m of runway still ahead as one rotates. This is generally less likely to be the case for medium and heavy jets though.
Tourist - regarding your idea of a dynamic V1, I agree it is technologically possible. The main question any good analysis would ask is is it worth it? Perf A does not exactly provide massive margins at the moment (look at wet takeoff performance on a limiting runway for example) so I fundamentally don't see there is much more than an extra few hundred kilos of payload to get out of it.
In contrast, the additional risks from cutting margins entirely to the bone is in my view not warranted. Why would anybody waste billions of dollars certifying their dynamic stop-go aid when they could just program HAL with Perf A?
There is a small subset of types and fields where it may be useful - I have operated bizjets capable of sub-1000m landing rolls off 3000m runways, and am well aware that the aircraft could safely be stopped past V1/Vr (as the two are normally the same) given one has 1800m of runway still ahead as one rotates. This is generally less likely to be the case for medium and heavy jets though.
Join Date: Jan 2014
Location: UK
Posts: 155
Likes: 0
Received 0 Likes
on
0 Posts
Whichever way you cut it computers are rapidly approaching the point where fully autonomous commercial airliners are technologically viable. Indeed there are some very clever people working at this very moment on making fully autonomous air travel a reality.
However what computers lack is imagination, intuition and a sense of self preservation. How many times as a pilot have you had a 'gut feeling' about a situation where something just didn't sit right with you? I'm guessing a fair few times. Lets not forget that computers are only ever as good as the person who programs them.
Though the factor I think will ultimately keep the robot airliner at bay is the passengers. Would you want to put your life in the hands of a glorified calculator? I know I wouldn't.
However what computers lack is imagination, intuition and a sense of self preservation. How many times as a pilot have you had a 'gut feeling' about a situation where something just didn't sit right with you? I'm guessing a fair few times. Lets not forget that computers are only ever as good as the person who programs them.
Though the factor I think will ultimately keep the robot airliner at bay is the passengers. Would you want to put your life in the hands of a glorified calculator? I know I wouldn't.
Join Date: Jun 2000
Location: last time I looked I was still here.
Posts: 4,507
Likes: 0
Received 0 Likes
on
0 Posts
A Boeing test pilot told a friend of mine who was a simulator instructor on the 787. "Boeing have designed the 787 assuming it will be flown by incompetent pilots
I recently checked the listings offered in the 'XXXXX for dummies.' library. Couldn't find B787. Perhaps FCOM should be retitled.
What is a fact is that Boeing FCOMs in last 25 years, ever since EFIS FMC a/c were introduced, now contain less & less technical information than previous generation a/c. Added to that is the universal acceptance of CBT self-study of technical training, multiple choice questions and little chalk & talk wash-ups at the end of the day by a knowledgeable engineer. Couple this with the limited knowledge gained during TR courses and the a/c knowledge of today's pilot, compared to yesteryear, is minimal. Don't think, just do. OK, sometimes you can out-think the QRH and dump yourself in a load of poo; but I found it disappointing/frustrating in FFS TR sessions, with non-normals, that the students were not aware about why the QRH was instructing them do make certain selections and take certain actions. It was as if they had dumped all their ground school from a couple of weeks earlier and were following instructions blindly. You might get away with it for a simple single failure, but not always with multiple failures or failures which did not fit into the QRH box.
I recently checked the listings offered in the 'XXXXX for dummies.' library. Couldn't find B787. Perhaps FCOM should be retitled.
What is a fact is that Boeing FCOMs in last 25 years, ever since EFIS FMC a/c were introduced, now contain less & less technical information than previous generation a/c. Added to that is the universal acceptance of CBT self-study of technical training, multiple choice questions and little chalk & talk wash-ups at the end of the day by a knowledgeable engineer. Couple this with the limited knowledge gained during TR courses and the a/c knowledge of today's pilot, compared to yesteryear, is minimal. Don't think, just do. OK, sometimes you can out-think the QRH and dump yourself in a load of poo; but I found it disappointing/frustrating in FFS TR sessions, with non-normals, that the students were not aware about why the QRH was instructing them do make certain selections and take certain actions. It was as if they had dumped all their ground school from a couple of weeks earlier and were following instructions blindly. You might get away with it for a simple single failure, but not always with multiple failures or failures which did not fit into the QRH box.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Tourist - regarding your idea of a dynamic V1, I agree it is technologically possible. The main question any good analysis would ask is is it worth it? Perf A does not exactly provide massive margins at the moment (look at wet takeoff performance on a limiting runway for example) so I fundamentally don't see there is much more than an extra few hundred kilos of payload to get out of it.
I have been spoiled by the military allowing me to play and do things like abort at V1 in various Perf A transport aircraft.
I think the margins are enormous to allow for very poor pilots having a bad day.
The other option of course is to not take the margin at all and just gain the increase in safety.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
The manufacturers have come to accept (rightly in my opinion) that human thinking in the moment is detrimental to the operation of the aircraft.
They would rather you just followed SOPs than figured it out for yourself airborne.
Knowledge encourages thinking.
I think that with the current complexity of a modern airliner, they may well be correct.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
There is no magic there.
Computers can do all this assimilation too.
Do you imagine that the person who programmed the Airbus Autopilot can fly better than it in the cruise?
Do you imagine that the person who programmed the FADEC can run the engines better?
Do you think that the person who designed the various chess computers can beat them?
With neural networks this is extra not true, obviously.
Originally Posted by Tourist
This I just not actually true is it?
Do you imagine that the person who programmed the Airbus Autopilot can fly better than it in the cruise?
Do you imagine that the person who programmed the FADEC can run the engines better?
Do you think that the person who designed the various chess computers can beat them?
Do you imagine that the person who programmed the Airbus Autopilot can fly better than it in the cruise?
Do you imagine that the person who programmed the FADEC can run the engines better?
Do you think that the person who designed the various chess computers can beat them?
Wet runway accelerate-go data provide for a screen height of 15ft at the end of the takeoff distance and only provide that you will make V2 at 35ft at an unspecified further distance down the road. Not much visible margin there. Would you really get a useful load increase out of the thinking time in the stop case?
Operating on private ops using unfactored landing data, I can tell you there is no margin in "advisory" landing distances using "average pilot" techniques either (we used to practice performance landings on empty legs.) I definitely only use (below) average pilot skill and technique and my results are anecdotal but I suspect your V1 aborts would be equally unrepresentative of the certification requirements including worn brakes &c
I personally would bank the increase in safety margin your system would create but I doubt anyone would develop it on that basis.
Operating on private ops using unfactored landing data, I can tell you there is no margin in "advisory" landing distances using "average pilot" techniques either (we used to practice performance landings on empty legs.) I definitely only use (below) average pilot skill and technique and my results are anecdotal but I suspect your V1 aborts would be equally unrepresentative of the certification requirements including worn brakes &c
I personally would bank the increase in safety margin your system would create but I doubt anyone would develop it on that basis.
Join Date: Sep 2006
Location: Required field missing
Posts: 565
Likes: 0
Received 0 Likes
on
0 Posts
Tourist
I am not entirely sure I fully agree with you, but I admit you make really good points and you present things under a sensible and different perspective.
Even thought is understandable that aircraft producers are slimming the FCOMs, I feel, as a pilot, mistreated by that, and the same applies for the ever increasing SOPS that leaves you with almost zero necessity to think.
Right, is progress, is safer. But...
Out of curiosity (I think I've missed it because I can not read every pages), are you a pilot?
Cheers
I am not entirely sure I fully agree with you, but I admit you make really good points and you present things under a sensible and different perspective.
Even thought is understandable that aircraft producers are slimming the FCOMs, I feel, as a pilot, mistreated by that, and the same applies for the ever increasing SOPS that leaves you with almost zero necessity to think.
Right, is progress, is safer. But...
Out of curiosity (I think I've missed it because I can not read every pages), are you a pilot?
Cheers
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Tourist, you completely and utterly missed the point. I suggest you start thinking like a normal person instead of 01011010110 HAL. What sort of collection of cells programmed the airbus stall warning to silence below 60 knots? A human. Was that a silly idea? Yes. Machines are only as good as the human programmer.
I don't think that "machines are only as good as the human programmer" is a clear way of making that point.
"Machines will still be imperfect if programmed by humans" is a better way of making the intended point, and I would agree.
That does not mean that they will not make less errors than humans which is what is important.
You example of the stall warner is a good one.
It is an error that helped confuse pilots already having a problem.
Because it is a machine based fault, however, it can be fixed to never happen again.
Human errors are sadly repeated ad infinitum...
Last edited by Tourist; 22nd Jan 2016 at 03:37.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
1. I don't know enough about the details of this incident to tell you how an automated aircraft would have coped.
Why don't you tell me exactly what an autonomous aircraft could not have done under the QF32 circumstances?
2. QF32 had an utterly non standard crew. Remind me how many were in the cockpit and their background?
The fact that they coped does not mean that the designed number of average crew, which is how 99% of these flights will be carried out, would have coped so why is it a good example?
3. As I have stated many times, autonomous aircraft will not be perfect.
They will still have "pilot" error caused accidents.
These accidents are likely to be different than human pilot caused accidents.
Black swan events are likely to be a higher proportion of those accidents, since that is the real area where humans still hold an advantage.
It may be that QF32 is an example of where a human pilot would be better. I don't know enough about it.
The vast majority of air accidents are not black swan, they are all too familiar repetitive scenarios.
An autonomous aircraft does not have to be perfect, it only has to kill less people than humans, or kill the same number more cheaply to be worth the effort.
For every QF32 there are 10 serviceable aircraft flown into the ground by humans.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
I started out military rotary then military fixed wing and traveled through corporate and airline and now fly an aircraft without an autopilot again because I prefer to play with a toy that requires my input to perform.
I too have seen the reduction in information passed to the aircrew.
I can still remember the pressure at an obscure point in the main gearbox that sets off the Emergency lubrication system on the Seaking helicopter.
By contrast, the tech knowledge required for the Airbus CBT is more like "the ladybird book of planes"
The fact that Airbus manage to be incredibly safe under these circumstances make me think that it is deliberate policy and not degradation of standards that has led to this.
The manufacturers believe that knowledge will lead to thinking instead of performing like automatons in the event of abnormal events. (They may well be correct that modern aircraft are just too complex for us to understand enough for us to make good decisions under pressure.)
They would rather you just did what the ECAM says, in effect performing as the avatar for the computer.
Under these circumstances humans are really just error vectors.
Unable to help situations, but able to hinder.
I say remove the vector.
Last edited by Tourist; 22nd Jan 2016 at 08:12.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Tourist - regarding your idea of a dynamic V1, I agree it is technologically possible. The main question any good analysis would ask is is it worth it? Perf A does not exactly provide massive margins at the moment (look at wet takeoff performance on a limiting runway for example) so I fundamentally don't see there is much more than an extra few hundred kilos of payload to get out of it.
All the current performance figures including V1 are based upon worst acceptable pilot having longest acceptable reaction time, worst acceptable engines etc etc.
A possibly very large advantage (depending on age of engines) of dynamic V1 is that the actual acceleration rate and actual engine thrust on the day would be used rather than worst possible case.
All the little safety factors added to temperatures, thrusts, could be quite reasonably removed as well as the human failing related fudge factors.
It would also allow go no-go decisions in real time utilising information about whether full thrust reverse would be available which could make an enormous difference to the critical point.
i.e. If the reason for the alert is cabin fire rather than engine failure, it is reasonable to assume that thrust reverse will work, therefore it could safely make the decision to stop thus avoiding taking a fire airborne at a much higher speed.
If however the problem is an engine failure, it is no longer reasonable to expect full reverse to be effective, so the aircraft may have already passed the critical point.
I suppose another side bonus would be that the accel data would flag up any weight finger trouble issues as well.
If the briefed mass and known thrust don't meet the expected acceleration figures, the aircraft will know it very early.
No more incorrectly set reduced thrust take-offs.
If the aircraft was not performing as expected, the aircraft would know it pretty much instantly and either add power or abort nice and early without causing an incident.
Edited to add This last could obviously be added to current aircraft without difficulty! Why isn't it?
Last edited by Tourist; 22nd Jan 2016 at 08:10.
This last could obviously be added to current aircraft without difficulty! Why isn't it?
or maybe it is technically very difficult and/or the regulators are unable to work out how it will be reliable/safe enough and/or the costs involved outweigh the benefits?
Reality check required...
Originally Posted by Tourist
By contrast, the tech knowledge required for the Airbus CBT is more like "the ladybird book of planes"
).
Originally Posted by Tourist
The fact that Airbus manage to be incredibly safe under these circumstances make me think that it is deliberate policy and not degradation of standards that has led to this.
Join Date: Sep 2007
Location: USA
Posts: 601
Likes: 0
Received 0 Likes
on
0 Posts
If you think about it, even a fully autonomous aircraft still relies on a control system that functions based on a set of instructions created by a group of humans that designed the controls using their own best judgement on how to deal with any particular situation they thought the aircraft might encounter.
This approach works very well in most cases, but there are the rare situations where a skilled pilot with the ability to make split-second decisions can do a better job. When it comes to ensuring the safety of commercial airline passengers, the added cost of having two qualified pilots in the cockpit is money well spent.
This approach works very well in most cases, but there are the rare situations where a skilled pilot with the ability to make split-second decisions can do a better job. When it comes to ensuring the safety of commercial airline passengers, the added cost of having two qualified pilots in the cockpit is money well spent.
Join Date: Jan 2014
Location: N5109.2W10.5
Posts: 720
Likes: 0
Received 0 Likes
on
0 Posts
Hi Tourist,
I think even Airbus recognises that ECAM doesn't provide all the answers. It's a pity crews like yourself feel you can't use your technical knowledge and experience to anticipate what ECAM will tell you to do later. e.g. Cabin ALT shows rate of climb & outflow valve fully closed = Suspected door seal leak. You can't prevent cabin ALT from climbing - but no ECAM yet.
Do you sit there and wait for ECAM "EXCESS CAB ALT" to tell you what to do or do you anticipate the problem and initiate a precautionary descent?
They would rather you just did what the ECAM says, in effect performing as the avatar for the computer.
Do you sit there and wait for ECAM "EXCESS CAB ALT" to tell you what to do or do you anticipate the problem and initiate a precautionary descent?