Hey, Please the response from John. He understand my question.

John,

Do you totally agree with the COBOK's opinion that CFD says "nothing particular" about the numerical method used to solve the "fluid mechanics" problem?

How can you solve fluid dynamics problems on the computer without the application of a particular numerical method? In my view, numerical analysis and fluid dynamics are both important components of CFD. I respectifully disagree with your notion that CFD has nothing to do with the numerical methods used. May be I am missing your point.

(1). Sorry, I could not understand his comment. (2). I tried to read " the lack of understanding of basic things in computational mechanics", but I was not getting anywhere. But, I think, the person is trying to say something. (3). My understanding of CFD is: Computational Fluid Dynamics consists of computational geometry, computational mesh, numerical algorithms, computer graphics in the post-processing. (4). Some technical journals are still using names, such as, numerical heat transfer, numerical methods in fluid dynamics,...etc. So, basically, they are dealing with the same subject. Because, the papers published in these journals also must include discussions of geometry, mesh, solutions, not just the numerical algorithm part. (5). Well, you don't have to understand everything posted here. By the way, when someone says "No, I don't", or "Yes.", or "Yes, I don't" in answering a negative statement, he simply say he agrees with the negative statement. So, it is up to you to interpret the comment. I think, it is not the comment itself, but it is "something the person was trying to say".

I agree with you.

First of all let me define what I mean when I say "terrible" results. When components dissipating heat are cooler downstream then upstream. When you don't need any measurements to understand that results you received are stupid. Why do I call it "terrible"? I think people, dealing with thermal analyses on commercial basis will understand my filing when I have to say to my customer: "Sorry, I can not give you results because they are wrong."

And now my question. Is there somebody who can share with me "good" experience with PHOENICS?

i think COBOK has taken enough abuse the post he was responding to asked an ambiguous (at best) or ignorant (at worst) question. i'll quote the question then the reply:

the question was-- So, do you think that CFD can only perform itself greatly in "fluid" and not other fields? In other words, is CFD not as good as FEM? Because FEM is capable to handle more "things".

and the answer was-- CFD stands for computational fluid dynamics. It says nothing particular about the numerical method used to solve the fluid mechanics problem. So, the door is wide open for everything: finite differences (actually, pioneering CFD works were based upon FDMs), finite/control volumes, finite elements, spectral methods, boundary elements and etc. When someone opposes CFD to FEM, he/she shows nothing but the lack of understanding of basic things in computational mechanics.

the question asked for a comparison of CFD and FEM. whether CFD was inferior to FEM because CFD is only good for fluids while FEM is more widely applicable. COBOK's reply meant to say that a comparison between FEM and CFD is comparing apples to oranges. CFD means 'computational fluid dynamics' and it is the study of solving the equations of fluid mechanics with various numerical techniques (among other things such as mesh gen, numerical analysis, graphics etc) which include FEM. now FEM means 'finite element method' which is means of solving differential equations. FEM is a numerical method while CFD is an application of numerical methods so it is not appropriate to compare them. moreover someone who would compare them is showing a limited knowledge of one or the other or both.

it would be more appropriate to compare FEM to FVM, FDM, BEM or other numerical methods. or to compare the state of the art in CFD to that of computational structural mechanics CSM, computational electromechanics CEM, or computational aeroacoustics CAA etc. then you would be more appropriate.

In my opinion, nowadays CFD can alone be a seperated field. It becomes more and more obvious that CFD and FEM can be treated as different applications although they may share the same basic.

Although CFD is one of the applications of FEM. However, do we ever think about the effort to make CFD to a "method" and not an "application" any more?

Hi. I suggest everybody to read this two papers about the use of CFD for HVAC.

[1]A. J. Baker, et al; "Computational Fluid Dynamics: A Two -Edged Sword", Ashrae Journal, v. 39, n. 08, Aug. 1997, pp.51-58 [2]Qingyan Chen; "Computational Fluid Dynamics for HVAC: Successes and Failures", Ashrae Transactions, v. 103, n.1, Jan. 1997, pp. 178-187.

Baker calls CFD, "Colorful Fluid Dynamics"

Good reading

Carlos Vilela

(1). Could you give us a brief summary of these two papers? It could take a while for us to locate the journals. By that time, we will be talking about different issues. Thanks.

Hi John. These, and many others that i have read, papers presented some cases where the use of CFD for HVAC problems need some especial attention, about modelling equations your are using to study some problem, the chosen grid, the discretizations schemes, the algorithm used to couple momentum-pressure fields, stability parameters, etc. So, the point presented is that CFD is not a "panacea", where you will solve all the problems, no matter how you use it. You have to be carefull, and be attentive to everything CFD involves and what can affect your solution. I have used CFD to simulate heat and mass transfer, natural convection, turbulent flows, and allways had obtained good results.

Thanks for replay.

Carlos Vilela

(1). Now I understand the root of the problem, that is CFD VS FEM. (2). Assuming that one is from the structure mechanics side, or even the Engineering Mechanics side, then the top layer of program is "FEM". (2). And there is a very well-known book of " The finite-Element Method". In other words, FEM is on the top of the empire. (3). At the commercial program level, AYSYS is equivalent to FEM. It is a FEM program. (4). In the FEM empire, there are statics, dynamics, heat transfer, electro-magnetics, structures, fluid dynamics,...etc. That is how they write the book, and how they create the empire. (5). It has been established to some degree that one takes a FEM program first, then try to solve the heat transfer problem next. This is especially true in the real world working environment. You rarely say that first look at different methods, then select one from FDM, FVM, FEM, or series expansion method,... and write a program to solve the heat conduction problem. (6). In this FEM empire, the process to solve a heat conduction problem is: learn how to use a FEM program, such as ANSYS, then create a geometry model, after that, generate the FE mesh automatically, set the boundary conditions and select the equation and the solver. Vola!, you have the FEM solution. (7). In the computational fluid dynamics, the empire is completely upside down. (8). CFD empire is created based on the fluid dynamics. The top layer of the empire is "fluid dynamics". Under the fluid dynamics, one can derive many branches of fluid dynamics, such as, inviscid flow, potential flow, boundary layer flow, viscous Navier-Stokes flow, reacting flow, two-phase flow, laminar flow, turbulent flow,....etc. (9). The need to do this is because the fluid dynamics problem is so difficult to solve that one has to look for a suitable method in order to have a converged solution. There is no cook book solution available. (10). So, in the FEM empire side, one can start with a general FEM program and try to solve problems in every engineering or scientific fields. That is the FEM point of view. (11). From the CFD side, one needs to konw how to deal with the turbulent flow, chemical reactions, multi-phase flows,...first, before he even starts looking for the solution methods. And one has to know the governing equations first, before selecting the numerical methods. (12). Well, this is not unique to the fluid dynamics people. If you have to deal with a new and very flexible structure, with some kind of turbulent liquid inside the composite material, and from time to time there will be heat release from the chemical reactions inside the material, then the current FEM empire and program will fail. (13). So, in my opinion, if FEM is acting like a male, then CFD must be a female, a very difficult subject to understand. It is almost impossible to understand a female. The best one can do is to develop some kind of feeling about the female first. (14). CFD VS FEM, a life time issue with no solid answer.

John,

You are right. "CFD VS FEM, a life time issue with no solid answer".

I think you had get to the main point of "How good is CFD".

Thus,

if CFD vs FEM, then the comparision is up to the readers. Because, I found that everyone is giving different opinion. Am I right?

it is very surprising how in a field which is so technical and careful about terms that engineers have allowed the term FEM to become so misused. as i said before FEM is a technique of solving differential equations it has no special ties with modelling structures computationally. however since FEM has been the dominant technique for structural modelling over the last 40 years is has become synonymous with computational structural analysis, CSM. if you read and understand my previous post you'd see that comparing CFD to FEM is like comparing digging holes to shovels. you can use a shovel to dig holes but there are other tools for digging holes. (also there are other purposes for shovels) but the two aren't the same kind of thing so you can't compare them. right know this isn't even an engineering/scientific discussion it is one about the use of language.

if the shovel and holes analogy passed over your head think of it this way: compare CFD and FEM is like comparing CFD and unstructured mesh generation.

I agree with most of what you have to say. I however, disagree with you on the issue of FEM having no special ties with computational structures. The Rayleigh-Ritz method was one of the oldest methods for solving Differential equations. The Finite Difference method (FD) jumped ahead of the variational principle approach on which the Rayleigh-Ritz was based because the FD methods appeared easier to solve with the limited computing that existed a century ago. When one takes calculus of variations, one clearly sees the relationship between variational principles and elasticity problems. So there is a strong historical bond between Rayleigh-Ritz methosd and structures. Since FEM is really the new name for the Rayleigh-Ritz method, I conclude that FEM and structures do have a special historical relationship.

i agree that the finite element method was originally formulated using the rayleigh-ritz variational concept so CSM does have a historical claim to the finite element method but today most FEM practitioners formulate their problems using the method of weighted residuals which is more general ie has no special ties to structures. so today CSM has no special link with FEM other than the historical one. for structural applications the method of weighted residual formulation and the variational formulation are mostly equivalent but in most other applications the method of weighted residual formulation (usually but not always the Galerkin method) is preferred.

No more serious disagreement but just adding a bit more historical epitaph. In 1960 the name finite element was believed to be first used by R.W Clough. In 1963 and 1964 several writers showed that FEM was a form of Ritz's variational method using piecewise linear test functions. In fact, it was not until 1965 that Zienkiewicz and Cheung broadened the applicability of FEM to cover all problems that can be casted in variational form. It is this FEM that you are refering to, i.e post-1965 FEM, even though FEM did exist prior to 1965.

I USED FLUENT, CFX AND PHOENICS AND ALL OF THEM ARE VERY GOOD OR VERY BAD. IT DEPENDS WHAT YOU ARE DOING WITH THEM. I LIKE CFX IF YOU HAVE TO DO A LARGE PROJECT AND YOU NEED TO FIND PROPER RESULTS. PHOENICS IS MORE FOR AN ENGINEER WHO NEEDS A QUICK ANSWER TO A PROBLEM.

Alain http://www.agca.zetnet.co.uk