Analysis of Viscous flow and flow separation.
Thread Starter
Join Date: Oct 2007
Location: coventry
Posts: 107
Likes: 0
Received 0 Likes
on
0 Posts
Analysis of Viscous flow and flow separation.
Quoting from Chris Carpenter’s (in my humble opinion) excellent book Flightwise 1:
“Unfortunately viscous flow and flow separation are not amenable to neat elegant mathematical treatment in the way that ideal flow is. We must therefore rely substantially on experience and experimental evidence.”
Given advances in mathematics (?) and computational power is this still the case?
PS- I don’t belong to any of the groups the forum is primarily aimed at…just an interested bystander.
PPS-Please don’t make the mistake that I know much/anything. The above book has made me realise that I should have paid WAY MORE attention in my maths and physics lessons than I actually did.
TIM.
“Unfortunately viscous flow and flow separation are not amenable to neat elegant mathematical treatment in the way that ideal flow is. We must therefore rely substantially on experience and experimental evidence.”
Given advances in mathematics (?) and computational power is this still the case?
PS- I don’t belong to any of the groups the forum is primarily aimed at…just an interested bystander.
PPS-Please don’t make the mistake that I know much/anything. The above book has made me realise that I should have paid WAY MORE attention in my maths and physics lessons than I actually did.
TIM.
Join Date: Dec 2006
Location: Australia
Age: 54
Posts: 44
Likes: 0
Received 0 Likes
on
0 Posts
neat elegant ?
I think not.
Could it be done. (Yes)
Would it be correct. Not the first few hundred experiments with any Long Complicated Formula.
Imagine trying to calculate how a hand full of sand will fall through your hands.
To many variables.
(Imagine also that the Sand granules are not all the same shape weight size Composition etc etc)
Ever hear about the saying that no two snow flakes are the same.
When flow separation happens. it is not just things like wing Surface texture/bumps, shape, speed etc. It is also all the variables of the air itself.
Computational power can calculate how a Start may Go supernova. so it could also be assumed that one could calculate a relatively simple thing like Flow separation etc. it just that it is not a Neat and Elegant formula. (well may be it is to the creator. but to you an I, it would Probably blow our meager 10w fuse).
I hope the statement from the book appears different now
neat elegant ?
I think not
I think not.
Could it be done. (Yes)
Would it be correct. Not the first few hundred experiments with any Long Complicated Formula.
Imagine trying to calculate how a hand full of sand will fall through your hands.
To many variables.
(Imagine also that the Sand granules are not all the same shape weight size Composition etc etc)
Ever hear about the saying that no two snow flakes are the same.
When flow separation happens. it is not just things like wing Surface texture/bumps, shape, speed etc. It is also all the variables of the air itself.
Computational power can calculate how a Start may Go supernova. so it could also be assumed that one could calculate a relatively simple thing like Flow separation etc. it just that it is not a Neat and Elegant formula. (well may be it is to the creator. but to you an I, it would Probably blow our meager 10w fuse).
I hope the statement from the book appears different now
neat elegant ?
I think not
Join Date: Feb 2005
Location: flyover country USA
Age: 82
Posts: 4,579
Likes: 0
Received 0 Likes
on
0 Posts
Well actually -
Since the real flow processes are probabilistic (some molecules will go thisaway, some thataway) it is feasible to design a computer program to model several different molecular scenarios, and then match up a mix of these with empirical data.
I believe that's what CFD (computational flow dynamics) programs do. Certainly not a "neat elegant mathematical treatment" in the sense of a single equation to solve, but it's possible to "thrash the problem to death" via computer.
Since the real flow processes are probabilistic (some molecules will go thisaway, some thataway) it is feasible to design a computer program to model several different molecular scenarios, and then match up a mix of these with empirical data.
I believe that's what CFD (computational flow dynamics) programs do. Certainly not a "neat elegant mathematical treatment" in the sense of a single equation to solve, but it's possible to "thrash the problem to death" via computer.
Fortunately for us experimentalists, but perhaps not for the computer boffins, the track record of CFD in accurately predicting turbulent or transitional fluid flow effects is not actually that good.
G
G
Thread Starter
Join Date: Oct 2007
Location: coventry
Posts: 107
Likes: 0
Received 0 Likes
on
0 Posts
Thanks for the replies. I like the idea of "thrashing to death" with computers. A bit like giving a 100 monkeys pencils and papers and waiting for them to produce The Complete Works of Shakespeare.
TIM
TIM
Join Date: Oct 2002
Location: Heathrow
Posts: 291
Likes: 0
Received 0 Likes
on
0 Posts
Some chaotic systems are deterministic, others are not. For instance, flipping a coin is chaotic, but quite deterministic - you give the same coin the same impulse and it behaves in the same way.
I doubt flow is a deterministic chaotic system. The same situation will produce different results, even though they may be very similar on a large scale. For instance, a wing at a given angle of attack at a given speed, temp etc will produce the same amount of lift, but from a computational point of view there will be differences. For instance the picture of separated flow will be different and unpredictable between two exposures taken a small time apart. This is not true for the coin.
If you could model every particle in the fluid, with a large number of paramaters such as temperature, speed, mass etc then you could possibly come up with a good approximation, however the amount of points and variables is so huge that even a big array of very fast computers is going to take a very long time with even the simplest problem.
Notwithstanding that, the micro and macro worlds are very different and flow is not usually considered to be a bunch of particles. CFD uses elements, but these are many many times the size of the particles that make up the flow, and are subject the different laws - for instance the electromagnetic force is important in molecules, but not usually in flow.
I doubt flow is a deterministic chaotic system. The same situation will produce different results, even though they may be very similar on a large scale. For instance, a wing at a given angle of attack at a given speed, temp etc will produce the same amount of lift, but from a computational point of view there will be differences. For instance the picture of separated flow will be different and unpredictable between two exposures taken a small time apart. This is not true for the coin.
If you could model every particle in the fluid, with a large number of paramaters such as temperature, speed, mass etc then you could possibly come up with a good approximation, however the amount of points and variables is so huge that even a big array of very fast computers is going to take a very long time with even the simplest problem.
Notwithstanding that, the micro and macro worlds are very different and flow is not usually considered to be a bunch of particles. CFD uses elements, but these are many many times the size of the particles that make up the flow, and are subject the different laws - for instance the electromagnetic force is important in molecules, but not usually in flow.
Join Date: Sep 2002
Location: La Belle Province
Posts: 2,179
Likes: 0
Received 0 Likes
on
0 Posts
To the best of my knowledge most, if not all, CFD programs in general use by aircraft manufacturers and others are indeed deterministic - the same set of initial conditions will ALWAYS produce the same solution. There is no chaos theory/random element in the workings of the program.
The 'Holy Grail' of CFD is the fully applied Navier-Stokes equations, solved throughout the flow field; it's a very very difficult problem, and for many practical cases one uses significant approximations, with empirical modelling and an understanding of the limitations of the method used to restrict use to meaningful analyses. Genghis' comment about turbulent flow is a case in point, and most CFD users understand that their tool is limited (but not all CFD advocates, some of whom could easily make a living in the marketing department one suspects).
CFD is very useful as an incremental tool, or for cases where the flow can be more simply modelled. Some conditions (such as the natural stall of a wing) are less amenable.
The 'Holy Grail' of CFD is the fully applied Navier-Stokes equations, solved throughout the flow field; it's a very very difficult problem, and for many practical cases one uses significant approximations, with empirical modelling and an understanding of the limitations of the method used to restrict use to meaningful analyses. Genghis' comment about turbulent flow is a case in point, and most CFD users understand that their tool is limited (but not all CFD advocates, some of whom could easily make a living in the marketing department one suspects).
CFD is very useful as an incremental tool, or for cases where the flow can be more simply modelled. Some conditions (such as the natural stall of a wing) are less amenable.
Join Date: Oct 2002
Location: Heathrow
Posts: 291
Likes: 0
Received 0 Likes
on
0 Posts
I can't think of a non deterministic CFD program, although I don't have much experience at all. Certainly in my CFD course at Uni, non deterministic programs were never mentioned.
I should have said above that even if you did have the computing power for every particle in the flow, that you would have very different initial conditions for every flow field anyway. How do you define where all the molecules are in the real flow? There would be an infinite spread of possibilities, so making the problem insoluble anyway.
CFD is a hard business, despite spending ages on a fairly simple problem, I got some amazingly strange results from a CFD program (rubbish in, rubbish out!). It takes a lot of knowledge and skill to get it right and even then the results can be misleading.
I should have said above that even if you did have the computing power for every particle in the flow, that you would have very different initial conditions for every flow field anyway. How do you define where all the molecules are in the real flow? There would be an infinite spread of possibilities, so making the problem insoluble anyway.
CFD is a hard business, despite spending ages on a fairly simple problem, I got some amazingly strange results from a CFD program (rubbish in, rubbish out!). It takes a lot of knowledge and skill to get it right and even then the results can be misleading.
Join Date: Feb 2004
Location: purple academy
Posts: 28
Likes: 0
Received 0 Likes
on
0 Posts
I once had a chat with a fellow who designed racing cars. He showed me a report that some of his high priced help had given him. He had hired some very bright CFD specialists, bought some awesomely powerful computers and some very expensive CFD software. Their task was to model the airflow around a racing car - with a view to reducing drag and improving down forces. He had spent a very great deal of money.
He was very pleased with the results. They had managed to predict the drag over the body of the car to within 5%. The flow over the wheels was not quite so easy and they had not got within 25% of the experimental results. I asked what is the proportion of wheel drag to the whole car. It was about 85% of the total.
Some things cannot be solved by maths - however good you are. The design of gas turbine combustors is a classic example. The knowledge is gained by sheer hard graft and laborious collection and processing of data to produce correlations (posh rules of thumb). The correlations can then be used, with healthy scepticism, to 'sort of' predict what might happen if we are lucky.
Some of engineering practice is art. Western education seems to be very much focused on analysis. I blame that Descartes bloke. You can't always make the whole system better by analysing the behaviour of the parts. Synthesis is where it's at - try to understand how the whole thing behaves. Then you will know where you will get the best returns for your efforts.
Micro-analysis of small volumes of flow is about as useful as predicting what the next wave will look like on a storm beaten shoreline. The chances of success are about the same as my winning the lottery tonight and hiring a buxom nurse for my dear old weary Dad.
Turn off your mind, relax and float downstream. Happy Christmas everybody.
He was very pleased with the results. They had managed to predict the drag over the body of the car to within 5%. The flow over the wheels was not quite so easy and they had not got within 25% of the experimental results. I asked what is the proportion of wheel drag to the whole car. It was about 85% of the total.
Some things cannot be solved by maths - however good you are. The design of gas turbine combustors is a classic example. The knowledge is gained by sheer hard graft and laborious collection and processing of data to produce correlations (posh rules of thumb). The correlations can then be used, with healthy scepticism, to 'sort of' predict what might happen if we are lucky.
Some of engineering practice is art. Western education seems to be very much focused on analysis. I blame that Descartes bloke. You can't always make the whole system better by analysing the behaviour of the parts. Synthesis is where it's at - try to understand how the whole thing behaves. Then you will know where you will get the best returns for your efforts.
Micro-analysis of small volumes of flow is about as useful as predicting what the next wave will look like on a storm beaten shoreline. The chances of success are about the same as my winning the lottery tonight and hiring a buxom nurse for my dear old weary Dad.
Turn off your mind, relax and float downstream. Happy Christmas everybody.
Join Date: Sep 2007
Location: Some sunny place with good wine and good sailing
Posts: 237
Likes: 0
Received 0 Likes
on
0 Posts
The answer to your question is yes.
As you say, laminar, non-compressible flow can be neatly represented by very elegant mathematical models. Flows around even complicated shapes can be modelled using the different models like building blocks - very satisfying mathematics, and probably fair to give credit to Soviet aerodynamicists as they were absolute masters at it.
These neat models do not work for compressible or turbulent flow. To model these sorts of flows, "finite element analysis" is used. Basically the entire flow field is divided into an imaginary mesh of cells ("elements"). If these cells are small enough, the flow within that cell can be represented by simple mathematical models (temperature, density, velocity are assumed constant in the cell). Thermodynamic and mechanical theory are then used to predict the characteristics of the flow in the neighbouring cell.
This model only works if the key assumption of uniform density and velocity is accurate. And that is only the case if the cells are small enough. The smaller cells you have, the bigger computer you need. The need for powerful computers is probably why US aerodynamicists started to overtake the Soviets in the 1970s and 1980s. The computing power needed today is enormous - most of the world's biggest computers are chugging away on finite element analysis of one sort or another - whether predicting weather patterns, or the flow around a gas-turbine blade, or the wing of an F1 car.
As you say, laminar, non-compressible flow can be neatly represented by very elegant mathematical models. Flows around even complicated shapes can be modelled using the different models like building blocks - very satisfying mathematics, and probably fair to give credit to Soviet aerodynamicists as they were absolute masters at it.
These neat models do not work for compressible or turbulent flow. To model these sorts of flows, "finite element analysis" is used. Basically the entire flow field is divided into an imaginary mesh of cells ("elements"). If these cells are small enough, the flow within that cell can be represented by simple mathematical models (temperature, density, velocity are assumed constant in the cell). Thermodynamic and mechanical theory are then used to predict the characteristics of the flow in the neighbouring cell.
This model only works if the key assumption of uniform density and velocity is accurate. And that is only the case if the cells are small enough. The smaller cells you have, the bigger computer you need. The need for powerful computers is probably why US aerodynamicists started to overtake the Soviets in the 1970s and 1980s. The computing power needed today is enormous - most of the world's biggest computers are chugging away on finite element analysis of one sort or another - whether predicting weather patterns, or the flow around a gas-turbine blade, or the wing of an F1 car.
A friend of mine does this sort of modelling for a living, and he's very good at it.
But, from a chat with him recently, to do a combined flow / thermodynamic / aeroelastic analysis of a single stage of a (gas turbine engine) compressor, at a single flow condition, using a 256 node computer cluster dedicated to this sort of task - would take about 3 days of continuous running, after a few weeks of data input.
Multiply that by the number of stages, and the number of flow conditions - and you get into a massive amount of computing time and power. And this is to analyse something that's already been designed! Now multiply it all by the number of design iterations required to get it right, and you start to understand why we experimentalists, not to mention hair-a***d old fashioned designers with 30+ years of relevant experience, are likely to remain in gainful employment for the foreseeable future!
G
But, from a chat with him recently, to do a combined flow / thermodynamic / aeroelastic analysis of a single stage of a (gas turbine engine) compressor, at a single flow condition, using a 256 node computer cluster dedicated to this sort of task - would take about 3 days of continuous running, after a few weeks of data input.
Multiply that by the number of stages, and the number of flow conditions - and you get into a massive amount of computing time and power. And this is to analyse something that's already been designed! Now multiply it all by the number of design iterations required to get it right, and you start to understand why we experimentalists, not to mention hair-a***d old fashioned designers with 30+ years of relevant experience, are likely to remain in gainful employment for the foreseeable future!
G
Join Date: Mar 2007
Location: Middle of nowhere
Posts: 228
Likes: 0
Received 0 Likes
on
0 Posts
I don't use CFD for aerodynamics, but I do use it for other applications with turbulence.
Modelling turbulence and flow separation is all really a question of scale. If you want a precise model of what each particle is doing within the flow then you will need a super super computer, a microscale model and a very very long time-even with today's computers.
However, if you are interested in so called 'averaged quantities' e.g. lift, drag, pressure field etc. then current CFD software is eminantly capable of doing the job. Flow separation will be adequately predicted and lift/drag should be predictable to within a percent or so.
The basic starting point for all CFD models are the Navier-Stokes equations which predict the flow of a fluid with viscosity. For turbulence, the most common approaches at the moment use the so-called two-equation models (k-omega model or k-epsilon model). Which of these models to choose is far from straightforward and is not something that the amateur can or should do.
There are other turbulence models coming along at the moment-some are better proven than others.
If you put some of the keywords above into Google or Wikipedia it should be easy to find some more information. Another good source of information (if you can get it online) are the ERCOFTAC Best Practice Guidelines which have several examples of turbulence and flow separation modelling. ERCOFTAC is the European Research Community on Fluid Flow, Turbulence and Combustion.
Hope that helps. If you're still awake feel free to pm me if you want to know more.
Bri
Modelling turbulence and flow separation is all really a question of scale. If you want a precise model of what each particle is doing within the flow then you will need a super super computer, a microscale model and a very very long time-even with today's computers.
However, if you are interested in so called 'averaged quantities' e.g. lift, drag, pressure field etc. then current CFD software is eminantly capable of doing the job. Flow separation will be adequately predicted and lift/drag should be predictable to within a percent or so.
The basic starting point for all CFD models are the Navier-Stokes equations which predict the flow of a fluid with viscosity. For turbulence, the most common approaches at the moment use the so-called two-equation models (k-omega model or k-epsilon model). Which of these models to choose is far from straightforward and is not something that the amateur can or should do.
There are other turbulence models coming along at the moment-some are better proven than others.
If you put some of the keywords above into Google or Wikipedia it should be easy to find some more information. Another good source of information (if you can get it online) are the ERCOFTAC Best Practice Guidelines which have several examples of turbulence and flow separation modelling. ERCOFTAC is the European Research Community on Fluid Flow, Turbulence and Combustion.
Hope that helps. If you're still awake feel free to pm me if you want to know more.
Bri
Join Date: Dec 2004
Location: YORKSHIRE
Posts: 13
Likes: 0
Received 0 Likes
on
0 Posts
Flow separation
I am just an ordinary pilot but I am sure you guys (and gals) accept honest questions.
Also, I am not sure if this question follows on from this thread but here goes. Given that vortex generators are meant to reduce stalling speed and VMCA (presumably by re energising the flow) what do you think of a Seneca that appears to have an earlier flow breakaway i.e. onset of buffet 20 kts before the clean stall and similiar with flap. And this aircraft does have vortex generators fitted. The stall speed appears to be 5 kts higher clean but about the same with 25 degrees flap compared to a non vortex fitted aircraft.
I am beginning to think they must have fitted them the wrong way round.
Incidentally, Seneca's with badly fitting doors are believed to cause buffet on the tailplane but who knows for sure without seeing the vortices.
Also, I am not sure if this question follows on from this thread but here goes. Given that vortex generators are meant to reduce stalling speed and VMCA (presumably by re energising the flow) what do you think of a Seneca that appears to have an earlier flow breakaway i.e. onset of buffet 20 kts before the clean stall and similiar with flap. And this aircraft does have vortex generators fitted. The stall speed appears to be 5 kts higher clean but about the same with 25 degrees flap compared to a non vortex fitted aircraft.
I am beginning to think they must have fitted them the wrong way round.
Incidentally, Seneca's with badly fitting doors are believed to cause buffet on the tailplane but who knows for sure without seeing the vortices.
