hi

could you please give your opinion, discuss about the changes,scope,future of CFD in the years to come.

regards

shiva

That is a huge topic. Some big CFD men claimed that the only CFD pacing term is turbulence modeling. Then more better models may appear for that. Also, improvement on mesh flexibility,moving bondary and so on. A CFD headache is the theory and techniques of BC calculation. In addition, how to adopt a high-order scheme in the FVM frame could be anoth headache!

I would say CFD will some get fresh wind (more: a storm) from the computer graphics/gaming industry. Those guys are actually some steps ahead of the still "old school" CFD community, which is based upon pre-modern principles.

The graphics people make heavy use of special hardware (the graphics cards). Real time 3D graphics cannot be thought today without hardware acceleration. However, hardware acceleration for CFD is not known so far. There is a bit of research on that, but the big CFD companies do not yet offer hardware acceleration.

Also, the gaming and graphics programmers have a more flexible, pragmatic approach to simulation which is the actually interesting part. Possibly, next generation CFD will run on the graphics cards with algorithms that tend to be of the lattice-Boltzmann type. Possibly by discarding the local character of the interconnections.

One must also say that our understanding of fluid flow is only basic. And I am sure that the clique of dry old men, controlling turbulence research by controlling the journals in the peer-review, will block any of such fresh winds. Ever heard of "Paradigmenwechsel"? We must wait one generation for this, I think.

regards

Unless you count NEC, Cray, IBM and other makers of supercomputers that is!

Computer graphics and games are about "fooling" the eye by looking right while CFD (should be) about getting the right answer with useful predictions (e.g. what's the weather going to be like in 3 days time).

I mean that when playing around with algorithms that produce visually appealing results, one is completely free with respect to the methods. Only the result counts. And this will one day produce a surprisingly simple algorithm, running on the graphics card, that produces nice images, but one may notice that the results are also good from a scientific point of view.

The supercomputers are today more for prestige and for military/security use. CFD is more like optimizing this particular valve or turbine or wind farm.

I disagree with Rob

CFD is a great science and has great future (for at least 80 .... 100 years ahead)

- fighters, bombers, cargo aircrafts

- helicopters

- stars and interstellar gas dynamics

- processes in the Sun

- airbreathing engines

- cars

- ships

- weather simulating (and I hope and its control)

- fire modelling

The secrets of turbulence is the most complicated problem , it may be compared with Quantum Mechanics

Until there will be wings, CFD will live.

May be someone will invent how to fly without wings, but it will not be during nearest the 50 years

I am very proud that I am CFD-man

I think CFD has a long road ahead.

On one had we have ppl who think "CFD is all about colorful images" and on other hand we have ppl who treat it as a "black box". It will take a lot of time to get rid of such opinions.

Also for me CFD is a great science. There must be misunderstanding. It has a great future, of course. So at what point do you disagree? I did not say that CFD does not have a great future.

Although I hope that it has no future in optimizing fighters. I don't like this military stuff at all.

CFD is one of the few key tools required in military. Currently it is being used to optimise designs on missiles and bombs. Many countries designing fighter aircraft also use it - in fact CFD is becoming an increasingly more important tool in military!

I suspect this posting might be a troll given the absurdity of some of the statements. Nonetheless...

In the early 80s one of the dry old men built a microprocessor with particle exchange rules as hardware microcode instructions. No problems making such devices but one needs a market to make mass production viable. It was not there.

EXA have marketed a commercial particle code for 10 or 15 years. It has been discussed several times before in this group. It works but not as well as traditional approaches for most if not all engineering problems. At one stage they were looking at purpose built hardware (fast general purpose integer microprocessor if I recall correctly but someone please correct me if I am wrong) but, again, beaten by economic considerations.

For many years parallel processing clusters were built of Inmos transputers for communication combined with Intel i860 microprocessors for number crunching. The i860 was also used quite widely as a graphics acceleration chip.

Particle methods have their strengthens and weaknesses. The dry old men have been aware of them for many decades if not centuries. In general, the approach is only going to work when used with large numbers of cheap microprocessors. In the late 70s and early 80s (in the UK at least) there was an expectation this would happen fuelled by the massive drop in price of devices like operational amplifiers which had occured in the 70s. Apart from the transputer, which was never close to being cheap enough, it has not even looked like happening. Microprocessors for computing started expensive and stayed expensive.

Making such devices today is a straightforward engineering task and has been for about 20 years. However, given the way the industry works and the size of the likely market it is hard to be optimistic such devices will ever be produced. The microprocessors+memory on a graphics card have often been mentioned as possible candidates but in terms of bangs for bucks they are not signficantly different from normal cluster hardware.

I admit that my postings were a bit subjective and more addressed to day-to-day CFD work in smaller size engineering companies.

Of course, the big discoveries are already made... and the CFD-methods are clearly separated into distinct classes, each of them with its own language and pros and cons. But this is what I mean with "blocking fresh wind".

That research is some steps ahead of practical work, is clear and there is considerable relaxation time between both. But for me it appears that practical CFD engineers are waiting too long for the scientists to come up with some better methods for handling turbulence. CFD is a comparably small market, at least when looking at the entertainment industry. Why not take advantage of the methods of each other?

CFD entered where standard mathematics failed. You can't solve NS analytically without simplifying it. However, to model the nature, from where we have extracted our principles, we need complicated or should I say complete mathematical models.

And deeper into the future these will get more involved. So, CFD has a great scope, however, it also needs to get developed. We know we may have computers, in the near future, of such high storage capacities and speeds that DNS will be much accessible, but still modelling will always be there since the value of time will increase. Even after decades of CFD, experiments are still inevitable, although reduced. Likewise CFD is going to live and flourish as long we have faith in it, and nothing better is found.

Cheer up, our life is anyway shorter than CFD.

With thanks Vinod Dhiman