Friends and colleagues

Have we really advanced the subject of CFD as much as we sometimes claim? Its still not a widely accepted design tool not nearly as standard in use as structural analysis codes such as Ansys.

We still struggle to deal with convection and boundedness and we are still dreaming of modelling turbulent flow to any degree of accuracy! (I expect a lot of shots for this one!)

My point is...why is it taking so long to develop robust easy to use CFD tools for industrial applications? Have anyone noticed how CAD software is developing at an incredible pace? Solid modelling is standard in most CAD software, even parametric design...not to speak of the incredible rendering features of this software, capable of producing photo realistic images...even moving animated images!!

Surely this technology must be much more complicated than calculating a face value for a convection problem!!! On the other hand...i may be wrong. Help me understand this.

Are CFD still alive and kicking or are we stagnating on decade old theories and archaic implementations???

You may argue that fluid dynamics is much more complex than solids...I agree. But are we not riding on this as an excuse for advancing CFD to the point that it is aqtually usefull in some way??

just a thought

(1) The reason we still struggle with modelling turbulence is that the problem of why a flow is turbulent is still and unsolved problem (as is defining precisely what you mean by turbulent flow).

(2) The underlying equations of solid mechanics are in a sense "easier" than those of fluid mechanics (even though they are significantly more nonlinear). This is basically because a fluid cannot with stand a shear and so particles are free to move around in complex patterns. A solid on the otherhand can sustain a shear (think of bending a piece of rubber) and there is a one to one relationship between the position of points in the deformed and undeformed states.

(3) Raytracing is much easier than solving the Navier-Stokes equations. Raytracing is not much more than basic trigonometry and geometric optics (this is being a bit unfair) while Navier-Stokes equations are a system of partial differential equations which exhibit extreme instability at high Reynolds number. This instability means that they can only be solved accurately for moderately low Reynolds numbers.

(4) Most research in CFD really arises from basic research into numerical analysis and many of the techniques so derived can take a long time to get taking up by CFD people (if they ever do). A further problem is computing power - nearly all of the present growth in CFD has arisen through cheaper and more powerful computers. The use of tried and trusted methods in such circumstances is unsurprising since the newer techniques often require far greater computer power (in the early stages of their development anyway) than older ones.

I think you have to draw a distinction between research into new methods of solving the Navier-Stokes equations and writing CFD codes to solve specific research problems. If you are doing the latter then there is no point in using the latest state of the art-technique if you don't have to!

Solids are not easy then fluids, simply Ansys and other codes solve problems in very narrow region of "solid's" science - stress analysis. Can these codes solve spreading of cracks with any accuracy? Limits of material fatigue with any accuracy? NO. In these regions they have the same problems like we have in turbulence modeling.

One day Lamb said to his learner - after his death he will ask the Lord about two questions - 1) quantum mechanics and 2) turbulence. And he said, he is doubt about getting answer on second question...

Actually the problems are quite different and many solid mechanics people I know would argue that fluids are easier (for easier read less nonlinear). The main difference to which you are alluding is the fact that the only type of "wave of discontinuity" that can appear in a fluid is a shock - and this is smoothed out by viscosity! A solid on the hand can support a varieity of such waves and non of them can be removed from the solution; i.e. the solid can fracture as you've quite rightly pointed out.

The problem's that arise from fractures in solids are of the form of moving boundary problems which are very difficult to handle numerically in the general case. (There are also issues about the equations changing type which can be even harder).

Turbulence on the otherhand is due to the high Reynolds number instability of the flow and is due, in no small part, to the singular (in the matched asymptotics sense of the word) behaviour of the Navier-Stokes in this limit. Another way of looking at it is, if fractures in solids were related to turbulence, we'd have a theory of turbulence.

Tom.

P.S. In my previous post I was using easier (and hence the quotes) rather loosely and had no intention of implying that solid mechanics was easy. On the contrary I found solid mechanics much harder than fluid dynamics when I was a student. What I was getting at was simply that in many cases the equations are mathematically better behaved than the Navier-Stokes equations (but see my point above).

In this regard, are there any recent review papers on the status of cfd ? The only one I have read is the often quoted 1979 paper of Chapman, "Computational Aerodynamics Development and Outlook", AIAA 79-0129R. That paper was very optimistic about computation of turbulent flows, which has not been realized in practice.

Ok, I suppose we can paraphrase English proverb - there are no hard problems, there are unsolved problems.

Such a discussion was presented at the PVP 2002 conference in vancouver earlier this year on the state and future developments of CFD with delegates from each of the major codes, Prof. Brian Spalding, a guy from ercoftac and so on. Unfortunately what could have been a very good in depth discussion as to the state of CFD and where it was going turned into a sucession of adverts for each of the companies represented. It was most disapointing and from my point of view (a relative novice in the field) a reall let down. It would be good to get such a discussion going on here.

Good idea, perhaps we could try to arrange some kind of "live hearing" here, or in a separate forum, and invite a bunch of CFD gurus to give their opinion on where CFD is heading and answer questions related to this. I think that it could be arranged quite easily... Have you got any suggestions on suitable people to invite?

That would be interesting. You must also invite some people from the industry. It would be good a idea to first request some experts to write an online article. This will give a proper direction to the discussions.

I'm not sure, obviously we need a blend of people involved in the development of the CFD field and also some high level users. The problem with inviting people from big codes is that it inevitably ends up just turning into adverts for theyre respective software. We really need people independent of the large companies (fluent, aet, adapco etc etc..) and more people involved in academia and the smaller more research orientated codes. Oh and brian spalding, hes always entertaining!!

Hi Jonas,

I think we must invite people from the industry because contrarily to what is often believed the big threes (Fluent, CFX, StarCD) are involved in research work either with universities or even on their own. Like any other researchers they do the work to improve and develop CFD further, with the added goal that some of this technology might/will become available in the forthcoming years in the commercial releases. I think that the company heads of technology would be very interestring to hear from, with applications in turbulence, aerocoustics, medical/biomedical, combustion work to quote just a few items I have seen presented in conferences.

Regards

I was at what turned out to be a rather good seminar at the IMechE in London yesterday on this topic "Grand Review on the state-of-the-art in the numerical simulation of fluid flow". There was a lot of potential for really interesting discussion, but unfortunately time was at a premium so only limited questions and discussion times were allowed.

The speakers were all very eminent in the academic world of CFD and several also had good links and even had worked in the industrial CFD world. That might be a good place to start trying to find contributors.

Best Regards Althea

Are there any proceedings of this seminar ? Any website on this seminar where we might get more info ?

Procedings were distributed at the event and may be available via the IMechE. The event coordinator was Gilda Ereira (g_ereira@imeche.org.uk) who may be able to provide procedings or alternatively the IMechE library. The institute website is www.imeche.org.uk and that should have more contact details, although I couldn't find any info on the seminar itself when I looked yesterday. The chairman was Brian Haller from Alstom Gas Turbines, but I don't seem to have any contact details for him. I hope this is sufficient to help you. Best Regards Althea