Fly By Wire: Possible "early warnings"
Join Date: Jun 2009
Location: Paname
Age: 55
Posts: 21
Likes: 0
Received 0 Likes
on
0 Posts
vapilot2004,
If I'm not mistaken, I think GV is referring to this post by desitter in the AF447 thread (had drawn my attention too):
I haven't checked it (and none replied), but it sure is intriguing
If I'm not mistaken, I think GV is referring to this post by desitter in the AF447 thread (had drawn my attention too):
Application of rudder, instinctual -
what has been said of rudder input in normal flight is very interesting and telling. One should consider the animations the NTSB often publish as part of their public dockets. It would seem that a sudden upset almost always leads to rapid, one would say instinctual, rudder input from the PF. Sometimes this may be justified but from what has been said by pilots here, not often. Perhaps this goes back to training days in small aircraft where more rudder authority is required to escape an upset or establish stable flight in turbulence - the ideas imprinted there could somehow just become part of the airman's natural skills.
what has been said of rudder input in normal flight is very interesting and telling. One should consider the animations the NTSB often publish as part of their public dockets. It would seem that a sudden upset almost always leads to rapid, one would say instinctual, rudder input from the PF. Sometimes this may be justified but from what has been said by pilots here, not often. Perhaps this goes back to training days in small aircraft where more rudder authority is required to escape an upset or establish stable flight in turbulence - the ideas imprinted there could somehow just become part of the airman's natural skills.
Join Date: Jun 2009
Location: Paname
Age: 55
Posts: 21
Likes: 0
Received 0 Likes
on
0 Posts
Gergely,
Just a word of caution there: as you probably know, such a system would only act as a protection if its specificity and detectivity were high enough. On the contrary, if the rates of false alarms and/or false negatives were to high, it might actually hinder the pilot rather than help him.
So the real problem may be feasibility: you speak of "basic airmanship 101", but some detection tasks which are quite obvious for humans are very tricky for computers. So "basic airmanship 101" may actually require "expert computership 999"...
If a system would tell you when you are getting "dangerously close to the limits" but wouldn't restrict you in anyway, that would be an extra layer of protection.
So the real problem may be feasibility: you speak of "basic airmanship 101", but some detection tasks which are quite obvious for humans are very tricky for computers. So "basic airmanship 101" may actually require "expert computership 999"...
Thread Starter
Join Date: Aug 2008
Location: Siófok
Age: 47
Posts: 18
Likes: 0
Received 0 Likes
on
0 Posts
JuggleDan: Yes, this is why such systems aren't common yet 
What is beasic airmanship for any of us isn't that basic things for a computer, but it is getting feasible...
But this is why advancement we seen in last decades opens up new possibilities. I think in middle of 1990s the costs and size (and power and cooling requirements) of an "1 GFLOPS computer" would be make it pretty hard to imagine how that much computing power would be used on an aircraft. Now you can buy a TESLA S1070 system with... Which is 4096 GFLOPS. It is compact, no special power / cooling requirements, safe.
Should I speak about software? I doubt if use of SNNs would be possible / practical about 10 years ago. Just as an example, because we are speaking about "perception" and what our computers can detect about the weather.
We need to analyze a lot of data to make the system reliable, and extensive testing? It would be more useful than SETI@Home, but when concept of smart cockpit, etc. was designed, such options weren't a possibility
How feedback is possible if we keep a sidestick is also very different now.
What is beasic airmanship for any of us isn't that basic things for a computer, but it is getting feasible...But this is why advancement we seen in last decades opens up new possibilities. I think in middle of 1990s the costs and size (and power and cooling requirements) of an "1 GFLOPS computer" would be make it pretty hard to imagine how that much computing power would be used on an aircraft. Now you can buy a TESLA S1070 system with... Which is 4096 GFLOPS. It is compact, no special power / cooling requirements, safe.
Should I speak about software? I doubt if use of SNNs would be possible / practical about 10 years ago. Just as an example, because we are speaking about "perception" and what our computers can detect about the weather.
We need to analyze a lot of data to make the system reliable, and extensive testing? It would be more useful than SETI@Home, but when concept of smart cockpit, etc. was designed, such options weren't a possibility
How feedback is possible if we keep a sidestick is also very different now.
Join Date: Mar 2005
Location: avalon
Age: 50
Posts: 39
Likes: 0
Received 0 Likes
on
0 Posts
with unreliable airspeed data fed into the FBW computer, does the FBW now allows the aircraft to maneuver more than allowed?
Thus allowing the aircraft to be over stressed or over loaded.
Thus allowing the aircraft to be over stressed or over loaded.
Join Date: Sep 2002
Location: La Belle Province
Posts: 2,179
Likes: 0
Received 0 Likes
on
0 Posts
So at the lowest level, in Direct Law, the flight controls can be manipulated so as to exceed the aircraft's design envelope. But the key thing is that it is then under the flight crew's control - and they will be presumed to be exercising appropriate caution with their control inputs.
The only way to defeat this arrangement is to have all the airspeed etc. data be consistently wrong coming into the FBW system. If it doesn't know there's anything wrong, then it could inadvertently cause a problem. But the same would apply to a non-FBW a/c in the same circumstances - if every piece of data is misleading, but consistent, you'll inevitably be misled.
Join Date: Jun 2006
Location: Canada
Posts: 94
Likes: 0
Received 0 Likes
on
0 Posts
Gergely
It seems like you want to put limits on the limits. The limits were determined by experienced people. Somehow, I think that providing some kind of warning in advance of reaching the limit might prevent the pilot from using all of his ability to control the aircraft.
Join Date: Nov 2000
Location: Montgomery, NY, USA
Posts: 137
Likes: 0
Received 0 Likes
on
0 Posts
What is the point?
I suspect there isn't any.
Some folks in the IT industry seem to think that their particular expertise is applicable to FBW aircraft types.
These folks are truly....out to lunch, in more ways than one.
I suspect there isn't any.
Some folks in the IT industry seem to think that their particular expertise is applicable to FBW aircraft types.
These folks are truly....out to lunch, in more ways than one.
I think everyone agrees that flying is no longer stick and rudder only. I won't argue if this is a good thing or a bad thing ( although I would rather have a stronger focus on pure flying skills and the training that promotes them.) The crew is tasked with interacting with a complex logic system, both in FBW aircraft, as well as non-FBW which have complex auto-pilot systems. When this interaction breaks down, the results are often tragic. The human factors involved in these breakdowns are not specific to aviation. We have seen them in nuclear power plants, transportation systems, communication systems,etc. It can be any system that controls a large and complex functions. And that is my point. We need more focus on how the human mind reacts to a system that is using some level of logic to make a decision. Too often, in all interactions between humans and computers, the interaction hits a "what is it doing now" moment. And in that moment, the human must determine if the logic in the system has broken down, or they don't understand the logic, or maybe a third or fourth or fifth possibility. The trouble is, in aviation, they usually don't have all that much time.
I am sorry if you think that pilots are the only ones who have any serious input into this discussion. Most of us in the IT industry succeed when we listen to everyone, and try to totally understand what the user requirements are and how the logic systems we are building will support those requirements.
Most of my questions on this board are trying to understand the human interaction level, and how the human mind interperts and anticipates logic systems. It's more than just flying. It's understanding the system and the mind that uses it. And I am sorry, but I don' think you need a commercial pilots license to do so.
Truth of the argument is independent of the person making it. E.g. "All other parameters being equal, the lift is proportional to velocity squared" is true even if made by IT consultant. "People who fly Airbus are not pilots anymore but mere system operators and it's all FBW fault" is false even when Airbus TRI/TRE says so. Therefore IT people, planespotters et al can and do make valuable contributions to PPRuNe, just not every time they post.
However, notions that there are enough data to predict the exact behaviour of aeroplane in turbulence and that failures leading to FBW degradation can be predicted and we just need more computing power to calculate when they will occur are false. You can play all day with your ANNs and SNNs or whatever but you can not have the meaningful output without the meaningful input. There's no radar or laser or whatever that will tell you how the air ahead of your aeroplane exactly churns and whether you'll get updraft/downdraft/positive windshear/negative windshear. Most of the time you get all of them in quite a short timeframe. Weatherpersons can estimate general area and general strength of the turbulence but even them don't get it right every time.
For example: one of the reasons of A330 switching to Alternate 2 law is loss of rudder pedals transducer. What inputs would you use to feed your computer that gives you "Caution - rudder pedals transducer failure in 5 seconds, prepare for alternate law" alert as its output?
However, notions that there are enough data to predict the exact behaviour of aeroplane in turbulence and that failures leading to FBW degradation can be predicted and we just need more computing power to calculate when they will occur are false. You can play all day with your ANNs and SNNs or whatever but you can not have the meaningful output without the meaningful input. There's no radar or laser or whatever that will tell you how the air ahead of your aeroplane exactly churns and whether you'll get updraft/downdraft/positive windshear/negative windshear. Most of the time you get all of them in quite a short timeframe. Weatherpersons can estimate general area and general strength of the turbulence but even them don't get it right every time.
For example: one of the reasons of A330 switching to Alternate 2 law is loss of rudder pedals transducer. What inputs would you use to feed your computer that gives you "Caution - rudder pedals transducer failure in 5 seconds, prepare for alternate law" alert as its output?
Join Date: Jan 2008
Location: Scandinavia
Posts: 98
Likes: 0
Received 0 Likes
on
0 Posts
411A Wrote:
Most (ie: 99.999%) of people who say they have "IT experience" have only ever played with Windows computers or maybe done a bit of system admin. I suspect the person here has no idea about how safety critical systems work, are designed, are built etc.
The analogy with the car is misleading in a number of cases, firstly anyone can get into a car and drive it without understand how the various systems on board interact and work under various conditions. Studies in Finland have shown that driving with modern "safety" features such as ABS, ESP etc actually cause more problems for persons who are not trained in their use. How many people here have actually been to an ESP driving course? Answer: NONE.
Compared with aircraft where all pilots who fly the Airbus, Boeing etc are explicitly trained to use and understand those systems. As many posters here have pointed out, operating such aircraft under the given situations (ie: failure, degradation of functionality etc) is NOT a problem.
So, unless the original poster has worked with avionics systems at the level where he has a very deep understanding of the theory and practise safety critical systems then any comment is hearsay.
fc101
E145 driver
* some comments and notes above attributable to a friend who works with such systems (railway and avionics)
What is the point? I suspect there isn't any.
Some folks in the IT industry seem to think that their particular expertise is applicable to FBW aircraft types.
These folks are truly....out to lunch, in more ways than one.
Some folks in the IT industry seem to think that their particular expertise is applicable to FBW aircraft types.
These folks are truly....out to lunch, in more ways than one.
The analogy with the car is misleading in a number of cases, firstly anyone can get into a car and drive it without understand how the various systems on board interact and work under various conditions. Studies in Finland have shown that driving with modern "safety" features such as ABS, ESP etc actually cause more problems for persons who are not trained in their use. How many people here have actually been to an ESP driving course? Answer: NONE.
Compared with aircraft where all pilots who fly the Airbus, Boeing etc are explicitly trained to use and understand those systems. As many posters here have pointed out, operating such aircraft under the given situations (ie: failure, degradation of functionality etc) is NOT a problem.
So, unless the original poster has worked with avionics systems at the level where he has a very deep understanding of the theory and practise safety critical systems then any comment is hearsay.
fc101
E145 driver
* some comments and notes above attributable to a friend who works with such systems (railway and avionics)
Join Date: Aug 2007
Location: UK
Age: 32
Posts: 53
Likes: 0
Received 0 Likes
on
0 Posts
I wonder why people aren't as up-in-arms about computer controlling the engines? Surely a FADEC can be just as much of a problem as fly-by-wire? In the same way the alternate law can appear, the FADEC can fail and go into alternate mode. IIRC it would lead to situations potentially much worse than alternate law with flight controls.
Join Date: Jan 2009
Location: Australia
Posts: 135
Likes: 0
Received 0 Likes
on
0 Posts
How many protections do you want on the end protection, the one that really counts? Be interested to see how you you could actually do that without seriously degrading the dark cockpit concept. Be careful of the "what's it doing now?" syndrome becoming the "what protection are we protecting now?" KISS
Join Date: Nov 2000
Location: Montgomery, NY, USA
Posts: 137
Likes: 0
Received 0 Likes
on
0 Posts
I suspect the person here has no idea about how safety critical systems work, are designed, are built etc.
The point I was trying to get to had to do with what I think is an overdependence on technology, and not enough focus on good training and skill levels. I do not think technolgy can fix these issues. We don't need more protections, we need more good pilots, trained well and paid what they are worth. I think that designing a good system that has simplicity at its heart to support them in such a way that they can understand it is the key to safety in any of these systems. I think too many levels of artificial logic make it more difficult for the human logic at the top of the chain (or in the front seats of the plane) to sort through it all. Simplify the systems and train the pilots. And that from an old gray haired IT guy.
I'll shut up now.
Last edited by patrickal; 15th Jun 2009 at 21:28.
Join Date: Jun 2009
Location: Paname
Age: 55
Posts: 21
Likes: 0
Received 0 Likes
on
0 Posts
fc101,
Having done scientific programming and software engineering for about 15 years, I respectfully disagree with you: for a variety of reasons, interacting with computers and computer softwares can sometimes be very taxing, even for highly trained and qualified professionals.
You find interacting with FBW systems easy, and that's a GOOD thing. However, I think you do so not only because of your training and qualifications, but also because these systems have actually been designed by IT people who have done extensive research of cockpit ergonomics and human-computer interactions!
As a matter of fact, I just googled for "Aérospatiale usability", and easily found a 300-page "Human factors for civil flight deck design" book. I guess there must be plenty other books and articles of the same around.
So, expertise in IT and in human-computer interaction is definitely relevant to FBW. However, the point where I disagree with patrickal is that I doubt Airbus waited for any of us to tell them something that is just plain common knowledge in our field... Something tells me they did their homework, and did it long ago!
Compared with aircraft where all pilots who fly the Airbus, Boeing etc are explicitly trained to use and understand those systems. As many posters here have pointed out, operating such aircraft under the given situations (ie: failure, degradation of functionality etc) is NOT a problem.
You find interacting with FBW systems easy, and that's a GOOD thing. However, I think you do so not only because of your training and qualifications, but also because these systems have actually been designed by IT people who have done extensive research of cockpit ergonomics and human-computer interactions!
As a matter of fact, I just googled for "Aérospatiale usability", and easily found a 300-page "Human factors for civil flight deck design" book. I guess there must be plenty other books and articles of the same around.
So, expertise in IT and in human-computer interaction is definitely relevant to FBW. However, the point where I disagree with patrickal is that I doubt Airbus waited for any of us to tell them something that is just plain common knowledge in our field... Something tells me they did their homework, and did it long ago!
Thread Starter
Join Date: Aug 2008
Location: Siófok
Age: 47
Posts: 18
Likes: 0
Received 0 Likes
on
0 Posts
Diaz: Easy to explain the difference, as you see one of the key questions is: If there is a difference between what you know and what you do we can see potential problems.
"Instinct" is the keyword here.
Why it is dangerous? Such instincts can kick in when something bad happens. These instincts are supposed to protect your life, but if you are flying an aircraft they can be dangerous. These protections aren't important in everyday operation when you don't even come close to the limits. But when your instincts try to save you but place you in danger, they can be important. Sadly it is about the same time when you can lose protections...
Jetdoc: I think you posted about a valid problem, the question is: I would say the feedback should be gradual for exactly this reason. If they would see any specific limit when controlling the aircraft it would be a valid risk, if they would see gradual feedback from the aircraft (they would feel it is harder to push the stick in one way) then these added protections could work as extra information.
Clandestino: As you see, the first post speaks about common oppinions about the FBW concept. I doesn't say I agree with them, but it is there to show, that there are many different approaches used to design FBW software and it is one of the few things people do remember about differences between Airbus and Boeing, it is something where they do have oppinions. And this is why I say while consensus isn't reached about what can and what should an FBW system do, there is still some place to discuss possibilities. Which is the point of the thread.
How early a warning can be, and how long you can maintain protections, feedback, etc largely depends on the cause of the problems. Of course the question is: Some would argue that if the level of protection (and warning) could be different in different scenarios wouldn't that confuse pilots who expect a warning? I hope that in most cases you use your knowledge, skills, etc to fly. The protections are needed when some pilots might use instincts and in such cases they can't observe same limits so easily.
Mind if I also point out one thing more: When the pilots are flying the aircraft, they often don't have time for checking charts, doing calculations, and knowing what will be save. When these data can be important for safety any way to see / feel / etc these data can potentially improve safety.
Why warnings and why no constantly displayed data? Because if I recall correctly, too many irrevelant data displayed at once could force the pilot to split his attention and can be dangerous, this is why the amount of displayed extra data should be limited to warnings and alternatives to lost instrument readings. And these options are only present if you can make sure they are reliable enough.
fc101: In first part of your comment, you speak about "how safety critical systems work", then, you say the requirement is experience in how avionics are determined.
And sadly, most people who "played with windows", wouldn't consider the possible risks, and how it can be implemented or other such topics, and probably wouldn't point to anything new.
Let me remind you most Windows systems doesn't know "how long a measured data is valid" based on physical possibilities, windows isn't designed to decide: If I can't measure airspeed, how can it be calculated.
"Instinct" is the keyword here.
Why it is dangerous? Such instincts can kick in when something bad happens. These instincts are supposed to protect your life, but if you are flying an aircraft they can be dangerous. These protections aren't important in everyday operation when you don't even come close to the limits. But when your instincts try to save you but place you in danger, they can be important. Sadly it is about the same time when you can lose protections...
Jetdoc: I think you posted about a valid problem, the question is: I would say the feedback should be gradual for exactly this reason. If they would see any specific limit when controlling the aircraft it would be a valid risk, if they would see gradual feedback from the aircraft (they would feel it is harder to push the stick in one way) then these added protections could work as extra information.
Clandestino: As you see, the first post speaks about common oppinions about the FBW concept. I doesn't say I agree with them, but it is there to show, that there are many different approaches used to design FBW software and it is one of the few things people do remember about differences between Airbus and Boeing, it is something where they do have oppinions. And this is why I say while consensus isn't reached about what can and what should an FBW system do, there is still some place to discuss possibilities. Which is the point of the thread.
How early a warning can be, and how long you can maintain protections, feedback, etc largely depends on the cause of the problems. Of course the question is: Some would argue that if the level of protection (and warning) could be different in different scenarios wouldn't that confuse pilots who expect a warning? I hope that in most cases you use your knowledge, skills, etc to fly. The protections are needed when some pilots might use instincts and in such cases they can't observe same limits so easily.
Mind if I also point out one thing more: When the pilots are flying the aircraft, they often don't have time for checking charts, doing calculations, and knowing what will be save. When these data can be important for safety any way to see / feel / etc these data can potentially improve safety.
Why warnings and why no constantly displayed data? Because if I recall correctly, too many irrevelant data displayed at once could force the pilot to split his attention and can be dangerous, this is why the amount of displayed extra data should be limited to warnings and alternatives to lost instrument readings. And these options are only present if you can make sure they are reliable enough.
fc101: In first part of your comment, you speak about "how safety critical systems work", then, you say the requirement is experience in how avionics are determined.
And sadly, most people who "played with windows", wouldn't consider the possible risks, and how it can be implemented or other such topics, and probably wouldn't point to anything new.
Let me remind you most Windows systems doesn't know "how long a measured data is valid" based on physical possibilities, windows isn't designed to decide: If I can't measure airspeed, how can it be calculated.
Facts often don't support the common opinion.
My point is that proposal you made in the first post is completely unrealistic. Turbulence is unpredictable. Failures that degrade FBW are unpredictable (and extremely rare, too). No amount of computing power can change that.
My point is that proposal you made in the first post is completely unrealistic. Turbulence is unpredictable. Failures that degrade FBW are unpredictable (and extremely rare, too). No amount of computing power can change that.
Join Date: May 2000
Location: Glorious West Sussex
Age: 76
Posts: 1,020
Likes: 0
Received 0 Likes
on
0 Posts
Gergely
Let me give you an example of one of these extra layers that already exist. You have experience of the "amateur side" - how do you know when you are getting close to the IAS limit for your aircraft? You look at the ASI and compare the needle position with the limit mark on the scale. You then use pilot judgement to assess how soon you will reach the limit, and whether you need to take action.
In a modern cockpit such as the Airbus FBW, the ASI has a "speed trend" indication in the form of a yellow arrow extending up or down from the present speed. If the speed is steady the arrow does not show. If the speed is changing the length of the arrow shows you how quickly the change is happening, and predicts what the a/c speed will be in 10 seconds. So in this FBW cockpit the pilot has a visual 10 seconds warning before reaching the limit, and before the next protection becomes necessary. This is plenty of time to take action.
Regarding weather radar, in most cases the pilot will have 20 minutes warning of storms - why do you need more?
Regarding changes of flight control laws - the a/c cannot predict that this will be necessary, because the change is not necessary until a failure occurs. So no advance warning is possible.
If a system would tell you when you are getting "dangerously close to the limits" but wouldn't restrict you in anyway, that would be an extra layer of protection.
In a modern cockpit such as the Airbus FBW, the ASI has a "speed trend" indication in the form of a yellow arrow extending up or down from the present speed. If the speed is steady the arrow does not show. If the speed is changing the length of the arrow shows you how quickly the change is happening, and predicts what the a/c speed will be in 10 seconds. So in this FBW cockpit the pilot has a visual 10 seconds warning before reaching the limit, and before the next protection becomes necessary. This is plenty of time to take action.
Regarding weather radar, in most cases the pilot will have 20 minutes warning of storms - why do you need more?
Regarding changes of flight control laws - the a/c cannot predict that this will be necessary, because the change is not necessary until a failure occurs. So no advance warning is possible.
