I try to find and buy a CFD code, of course I hope it can deal with most of internal,external,mulutiphase...etc. problem. I was realy want to hear from any advise from any one.

You may want to visit the home page of the commercial CFD code developers and get more information from them because as a user it is really hard to cover every aspect of CFD applications. And since the information on Internet is free of charge, you don't have to buy anything when you are surfing the Internet. You can get the first hand information directly from the CFD developer. ( By the way, the code is only the begining of the solution, not the end of the answer. CFD field is relatively young. So, before you jump into the water , it's a good idea to do some home work first. If you have un-limited resources, any CFD code is a good code. ) Any black box is a black box. It's hard to say this black box is better than that black box.

If you have certain class of problems that you'd like to solve, I'd suggest that you get trial licenses (1-3 mos) for a nominal fee, run your models/problems on all of them and decide. Normally how do you compare results from different simulations is generally an important question. That is something you should do your homework on. If you have expt data for your kind of problems, you're almost home free. ("Almost", because a number of times, expts can suffer from errors of measurement which a lot of "cfd code validators" seldom pay attention to.)

Of course other factors affecting your decison to go w/ one code or other can be ease-of-use, variety of models offered etc. In my experience, deciding on a commercial CFD code can take anywhere upto three to four months. Good luck.

It is extremely important to assume that the code you are going to use is likely to fail. Unless, the code has been validated for the specific problems you are interested in. Most commercial codes problbly have passed the stage of solving the square cavity flow, back-step flow, pipe-entrance flow. Smart users are critical for the future health of the CFD field. I am still trying to get converging solution from a code which has claimed to be widely used for several years. I had pointed out results from a general purpose code was wrong to my colleague last year. I would say that most of the time the results to your problem would be useless unless the code has been validated for that problem. ( mesh, turbulence models, algorithm are among the problem areas ). Do not assume that a CFD code will automatically solve your problem. No such smart code exists today!!! On the other hand, it is not the end of the world either. A " lesson learned section " would be extremely valuable for new and experienced users if the name of the code can be omitted from the discussion.

Thanks John and Sampige for your nice advise

I dealt with CFD about 10 years ago, then I move my intersting to experimental fluid dynamics. I recently I go back this field and visted some CFD website and some advertisement of CFD code. Some of them seems give you a very nice impression that many problem which we met 10 years ago seems solved thoroughly. It is difficult for me to juge if it is true or it is only because of advertisement,since I lack of knowledge of the development and prograss of CFD. Although I believe that a great step prograss has been made in recent years, I still not so confident if I can rely on some code to solve some engineering problem (thanks your words relate to this).

Do you mean some of provider of CFD code they provide several monthes for trial?

****Do you know Is there any CFD provider accept some homework to do by themself? To show the capacity of their code?

Is there any paper or website talk about an over view of CFD today?

The graphics and the ability to handle complex geometry are two big steps achieved in recent years. With these two accomplishments, it is now possible to explore some real world flow problems. As for the numerical algorithm and turbulnce models used, most codes still use algebraic or two-equation models which are 20 years old ( with some new constants or functions ). There are three ways you can get involved in CFD development and applications 1) write your solver, write your graphics and write your geometry and mesh, (2) write your solver, write your geometry and mesh, use graphic packages, (3) use commercial solver, write your geometry and mesh with the help of commercial codes. In the first step, you do everything. In the last step, you focus on the geometry of your problem, the solver is automatic ( invisible to users )and the graphic is semi-automatic. For specific types of application, it is possible to develope integrated code where the geometry,mesh,solver and graphcis are all in one code. The need for code validation is to check the capability of the code, since the code is still limited by the algorithm and turbulence models used ( that part belongs to research domain ).

Contact the various CFD software vendors: Fluent, CFX, Star-CD...what have you, and ask them for a trial license. If you're from a univ, response may be lukewarm (my experience. Others' experience, not so.) But if you're in a professional setting, most likely they'll give you a trial license. Some may charge a nominal fee. I know some vendors will also run test problems for you on some kind of assurance that you will buy. If there's no such assurance, they'll give you the code to test for yourself on a trial basis of a month or two. At least no harm in asking.

good luck.

A free lunch for two or a free gift could be more effective to bring users for a test drive ?

I think it is important to remember what exactly you are trying to do. Commercial codes are pretty good at predicting things like overall pressure drop, mass flow splits, and other integrated properties like that. But, if you are looking for things like pressure at a given point, don't expect to get anything resonable. I'm not saying that you won't, but sometimes you will be really close(~5%) and other you will be really far (~100%).

Think from the veiwpoint of the software developer. For the most part (99%), they are developing software to make money. They are not doing it to solve all problems. Of course they want to keep imporving the codes, but they're not going to do it at the expense of profits. To that tend, what is developed is what is requested by their customers. And in this age of faster times for product development, customers generally want 1 thing:

Most REASONABLE answers, in the shortest amount of time, with the least amount of effort. (why do you think there is sucah a massive push to Tetra meshing instead of using the same resources to improve code accuracy?)

Note that I say resonable. This is becasue compaines may have a bunch of designs to choose from. They want to know which ones would perform best. All things being equal (ie Boundary conditions, mesh distribution, etc.), CFD in its current state will tell them that. Then, they can eliminate designs down to 1 or 2. These last designs can then be tested to get real performance. Thus saving time and money on excessive testing.

Now, I realize that some people are getting beyond this point. I'm extremely happy about that. But, CFD is at a stage that FEA was probably 5-10 years ago. As the technology improves and more people understand what CFD can do for them, the emphasis will begin to change from speed to accuracy.

This is just my 2 cents.

I think you are right. Many users of FEA have a rather naive idea about the commercial CFD codes. They normally think that they can handle the CFD code and the problem just like they are solving the FEA problems.

Some of code claim that 'no approximation'. How we can certify reliability of a code, by experimental test? It will be high costs and impossible for a CFD company.

Some of code claim that 'no approximation'. How the reliability of a code can be certified , by experimental test? It will be high costs and impossible for a CFD company.

When you are selling hamburgers, people will ask " where's the beef". When you are selling cars, people will ask " do I get the free gift for test driving the new model." When you are writing codes, people will ask " can I have a free copy for beta testing your code." When you are selling a code, people will ask " can your code solve this problem." There are several answers to this question; (1). yes, the code can handle this problem quite well. As a matter of fact it is the main reason why we developed this code. We solve this problem under various conditions and we are very pleased with the result. (2). we have not solved this problem. but we think the code is definitely capable of solving this problem. (3). No, we don't think the code can handle it right now, but we are doing our research on it. If it is a hamburger, it is a hamburger not a cheeseburger. The user is not interested in your hamburger shop, he is interested in the hamburger itself. If the developer of the code includes 5 to 10 test cases in the code itself, then the user should be able to run the demo by himself and convince himself whether he likes it or not. To say that the code can solve steady, unsteady, laminar, turbulent, incompressible, compressible, single, multi-phase, non-reacting, reacting flows is just like trying to sell your hamburger shop to the user, not the hamburger itself. It is up to the user to tell you what he likes in the hamburger. You certify and validate the solution of the problem ( through the use of the code) as if you are the user. The point I am trying to make is : Are you selling the code ( engine ) to the CFD application developers? or Are you providing answers ( games ) to the users ? My feeling is that the current general purpose CFD codes are developed for the experts in the CFD application groups in large companies where the ability to produce the results are interpreted as the capabilities of the engineer, not the capability of the code. ( only a few can get the result successfully from the same code they use.) This brings up another issue of " where and who should carry out the validation of the solution through the use of the code ", The experts in the CFD application group in the large company ? or the CFD code developer ? Regardless of how it is done, the issue will have great impact on the quality of the hamburger in the future.

hello Lans!!

I have a 2 year experience in CFD, and all of my work was done and is doing with NEKTON. First, i use version 2.8, and now version 3.2 This code works very well with multiple physical scenarios such as laminar, newtonian, non-newtonian, transitional and turbulent flows, multiphase, etc. You can find more details in Fluent.com web site.

Hope it helps...

Rodrigo.