CFD - Trends and Perspectives

July 17, 1998, 11:17

In this era of commercial codes, colorful fluid dynamics and quick results it is nice to sometimes sit back and reflect a bit about where CFD stands today and where we will be in 5 or 10 years. I'll give my opinion here. I appreciate any feedback.

Today I see a very strong trend towards commercial codes. It has been happening for a few years, but this last year, with the release of a couple of new unsctructured codes, I think the commercial market has taken a big step forward. I personally know of several companies that now intentionally have put all development on their "in-house" codes on hold and instead they are switching to commercial codes. The same trend has also started in academia. First comes the more applied fields like turbomachinery, combustion engines, etc. and soon we'll probably see more and more pure fluid dynamics guys that use commercial codes. I don't know if this is good or bad, but I think it is inevitable. Hand in hand with this commercialization goes also a more widespread acceptence of CFD in industry, and we now see CFD being used in areas where noone even knew what CFD stood for 5 years ago.

Enough said about commerial codes.

Then we have the never ending problem in CFD - turbulence modeling. 5 or 10 years ago everyone were optimistic and eager to tackle this problem. Or perhaps it was just me beeing younger and more innocent ;-) I don't know. Anyway, people had just started to do serious LES simulations, Germano and others came with interesting new ideas. LES would save the world. At the same time we also began to see some promising Reynolds stress models published by Launder and others. These models were very interesting and relied on a sound foundation. In a few years everyone would have switched to RSM some people argued. But then the years went by and not much happened. LES still has a lot of unsolved problems, and it is still too expensive for most applications. RSM models proved to be very difficult to use and they aren't as general as people hoped. In an attempt to take the best from two worlds there was a short new interest in non-liner eddy-viscosity models a couple of year ago, but that soon faded away. Now recently we've seen a similar sudden interest in explicit algebraic RSM models, but that also seems to fade away. Instead people are now going back to use one-equation models in the boundary layers. We've sort of walked one turn around the problem without coming any closer to a solution. I don't know where we'll head next. Perhaps a new interest in LES? Perhaps a completely new approach based on lattice-gas / cellular-automata methods. A couple of years ago I listened to a very interesting discussion about turbulence models between Prof. Piomelli (LES guru) and Prof. Launder (RSM, EVM guru). Interesting enough Pimelli didn't belive that LES would save the world, he was more positive about the new RSM/EVM models developed by Launder and others. Launder on the other hand, had the opposite view, or at least that was what he said then. He was optimistic about LES and thought that it would become more and more useful. It's not very encouraging when scientists working on one approach belives more in another approach. Perhaps we've come to a dead end here. Where do you think turbulence modeling will head next?

I see a big danger in the commercialization of CFD in relation to the turbulence modeling problem. As everyone who has worked with turbulence modeling knows it is extremly important how you implement your turbulence model in your code. With commercial codes this becomes a well guarded secret, and we might end up in a citation where noone knows which equations they are really solving. There is no danger yet, academia is still very much in the front line of turbulence modeling, but in 5 year it might look different. Or what do you think?

Then we have the other side, numerics, I haven't done any research in this area lately, so I migth be a bit out-of date with this. If I write something wrong someone please correct me. My impression from this side is a bit similar to my view of turbulence modeling. 10 years ago things looked bright. We had multigrid that would speed up things, we had new FEM methods that could perhaps give us a-posteriori error control, we had new discr. schemes and implicit methods that were very promising.... and computer power doubled every 18th month. Looking back at this the area of numerics has had much more success than turbulence modeling. Although a lot of problems remains to be solved, we today we have robust methods that give good results for most types of applications. However, it still takes a lot of experience and knowledge to use a CFD code, or perhaps I should say to interpret results and to improve predictions. Perhaps someone in-tune with the latest developments on the side of numerics could give his view on what will happen now?

You might interepret me as if I'm sceptic about the future of CFD. This is not the case though, I'm just saying that we still have a lot of problems to solve. However, I see one very positive thing happening now, and that is that people from different areas are beginning to talk to eachother. 10 years ago the turbulence guys were using their own small boundary layer solvers and the numerics guys were using very simplified turbulence models, and noone talked to the guy who actually had a problem to solve. Today people are much more aware of that in order to advance from this point we have to collaborate. There is no general solution, and what happens in one field very much influences other fields. I'm speaking in general terms here, but I'm sure you can think of a number of examples yourself. This new era of "global thinking" is very positive!

Opps, this is what happens when you're bored with writing on your thesis. I'd better get back to work now. Please feel free to add your view about CFD today and in the future. I've left out a lot of things that should be said... before I end I will just give my guess of where CFD will be in 10 years from now:

o Turbulence modeling is still a *major* problem

o Commercial codes are completely dominating, however, they have become much more modular and it is easy to add your own parts here and there. Also we will see more "networked" computing with people hiring a "code+computer+consultant" directly over the internet. Visalization has become a real 3D thing (not only flat screens)

o Numerics is a more mature field, less problems left to solve here, but we still haven't got any good error-control.

o New problems in focus - multiphase flows, combustions, multi-physics simulations.

o Perhaps a completely new approach - lattice gas or something like that.

Do you share my views? Did I forget something important?

July 17, 1998, 11:17	CFD - Trends and Perspectives	#1
Jonas Larsson Guest Posts: n/a	In this era of commercial codes, colorful fluid dynamics and quick results it is nice to sometimes sit back and reflect a bit about where CFD stands today and where we will be in 5 or 10 years. I'll give my opinion here. I appreciate any feedback. Today I see a very strong trend towards commercial codes. It has been happening for a few years, but this last year, with the release of a couple of new unsctructured codes, I think the commercial market has taken a big step forward. I personally know of several companies that now intentionally have put all development on their "in-house" codes on hold and instead they are switching to commercial codes. The same trend has also started in academia. First comes the more applied fields like turbomachinery, combustion engines, etc. and soon we'll probably see more and more pure fluid dynamics guys that use commercial codes. I don't know if this is good or bad, but I think it is inevitable. Hand in hand with this commercialization goes also a more widespread acceptence of CFD in industry, and we now see CFD being used in areas where noone even knew what CFD stood for 5 years ago. Enough said about commerial codes. Then we have the never ending problem in CFD - turbulence modeling. 5 or 10 years ago everyone were optimistic and eager to tackle this problem. Or perhaps it was just me beeing younger and more innocent ;-) I don't know. Anyway, people had just started to do serious LES simulations, Germano and others came with interesting new ideas. LES would save the world. At the same time we also began to see some promising Reynolds stress models published by Launder and others. These models were very interesting and relied on a sound foundation. In a few years everyone would have switched to RSM some people argued. But then the years went by and not much happened. LES still has a lot of unsolved problems, and it is still too expensive for most applications. RSM models proved to be very difficult to use and they aren't as general as people hoped. In an attempt to take the best from two worlds there was a short new interest in non-liner eddy-viscosity models a couple of year ago, but that soon faded away. Now recently we've seen a similar sudden interest in explicit algebraic RSM models, but that also seems to fade away. Instead people are now going back to use one-equation models in the boundary layers. We've sort of walked one turn around the problem without coming any closer to a solution. I don't know where we'll head next. Perhaps a new interest in LES? Perhaps a completely new approach based on lattice-gas / cellular-automata methods. A couple of years ago I listened to a very interesting discussion about turbulence models between Prof. Piomelli (LES guru) and Prof. Launder (RSM, EVM guru). Interesting enough Pimelli didn't belive that LES would save the world, he was more positive about the new RSM/EVM models developed by Launder and others. Launder on the other hand, had the opposite view, or at least that was what he said then. He was optimistic about LES and thought that it would become more and more useful. It's not very encouraging when scientists working on one approach belives more in another approach. Perhaps we've come to a dead end here. Where do you think turbulence modeling will head next? I see a big danger in the commercialization of CFD in relation to the turbulence modeling problem. As everyone who has worked with turbulence modeling knows it is extremly important how you implement your turbulence model in your code. With commercial codes this becomes a well guarded secret, and we might end up in a citation where noone knows which equations they are really solving. There is no danger yet, academia is still very much in the front line of turbulence modeling, but in 5 year it might look different. Or what do you think? Then we have the other side, numerics, I haven't done any research in this area lately, so I migth be a bit out-of date with this. If I write something wrong someone please correct me. My impression from this side is a bit similar to my view of turbulence modeling. 10 years ago things looked bright. We had multigrid that would speed up things, we had new FEM methods that could perhaps give us a-posteriori error control, we had new discr. schemes and implicit methods that were very promising.... and computer power doubled every 18th month. Looking back at this the area of numerics has had much more success than turbulence modeling. Although a lot of problems remains to be solved, we today we have robust methods that give good results for most types of applications. However, it still takes a lot of experience and knowledge to use a CFD code, or perhaps I should say to interpret results and to improve predictions. Perhaps someone in-tune with the latest developments on the side of numerics could give his view on what will happen now? You might interepret me as if I'm sceptic about the future of CFD. This is not the case though, I'm just saying that we still have a lot of problems to solve. However, I see one very positive thing happening now, and that is that people from different areas are beginning to talk to eachother. 10 years ago the turbulence guys were using their own small boundary layer solvers and the numerics guys were using very simplified turbulence models, and noone talked to the guy who actually had a problem to solve. Today people are much more aware of that in order to advance from this point we have to collaborate. There is no general solution, and what happens in one field very much influences other fields. I'm speaking in general terms here, but I'm sure you can think of a number of examples yourself. This new era of "global thinking" is very positive! Opps, this is what happens when you're bored with writing on your thesis. I'd better get back to work now. Please feel free to add your view about CFD today and in the future. I've left out a lot of things that should be said... before I end I will just give my guess of where CFD will be in 10 years from now: o Turbulence modeling is still a major problem o Commercial codes are completely dominating, however, they have become much more modular and it is easy to add your own parts here and there. Also we will see more "networked" computing with people hiring a "code+computer+consultant" directly over the internet. Visalization has become a real 3D thing (not only flat screens) o Numerics is a more mature field, less problems left to solve here, but we still haven't got any good error-control. o New problems in focus - multiphase flows, combustions, multi-physics simulations. o Perhaps a completely new approach - lattice gas or something like that. Do you share my views? Did I forget something important?