Hello Everyone,

I have been interested in CFD for a long time now, but I have no idea how to "break" into the market, as I have no practical or industry experience. I have completed an undergraduate level elective course as part of my Bachelor's Degree in Mechanical Engineering, which has given me a grounding in the basic principles. Through this course and other project work in University, I gained experience using the FLUENT CFD package.

Many of the CFD related positions I have seen around the 'net require substantial previous experience. My question is, what are the best routes to follow for finding an introductory level position that doesn't require eons of prior work experience along with complete fluency (no pun intended <g>) with every CFD code under the sun?

I've just completed my Master's Degree in Fire Protection Engineering, and I work for a fire science engineering/building code consulting firm, so I have a good job that pays the bills, but I find myself dreaming a little bit. If an opportunity in the CFD field presented itself, I'd probably go for it without much thought.

Any advice or leads would be greatly appreciated.

Sincerely,

Jeffery D. Mitchell, B Eng

(1). I think you are facing a tough problem. (2). The work required in the CFD operations normally involves geometry construction, mesh generation, solution, and presentation, assuming that you are dealing with routine problems. (3). In old days, when not everyone was playing stock market, there used to be CFD department in a company to deal with the geometry, mesh and the solver. They also support many other group of engineers to operate the code they have developed. So, Those were the golden days. (4). With the commercial codes arrived, like the one you have used, the cost saving idea was used to replace the in-house development cost (that includes the positions you are looking for) to get the company to use the commercial codes. The saving is say 100k now vs 500k of 5 cfd development engineers, that is 400k saving. The jobs you are looking for are not eliminated. (5). Once the commercial codes is in, they try to increase the number of licenses to increase the annual sale. This requires more engineers to use the code. That is fine. But they also make it clear that the code is user-friendly, and any engineer with minimum training can easily use it. As a result, the company increases the number of licenses, and send more engineers to the training courses. (6). The end result is not good. Because, you really needs to know something about the mesh generation, the numerical algorithm and their impact on the solution, not to mention about the geometry and mesh construction. The number of real success is fairly limited. Even if one succeeded in mesh generation, it is likely that he will be trapped in the solution convergence step. (7). With the stock market very hot, the time shrink, the demand to get the cfd solution ahead of time also increases. (I am telling you the real story) With so many possible places to fail in the cfd operation, no one is going to look for a young graduate in this field, unless the company still has the old fashion cfd department where senior engineers can provide necessary guidance to the junior engineers. The code is so easy to use that any engineer can easily learn how to use it. But in reality, you will not be solving simple sample problems. You will be facing problems so difficult that you don't even know whether the code can handle it or not. (8). In the current commercial code operation, it is not possible to do geometry, mesh, solver, and graphics separately by different engineer in defferent department, as was done in the old days. So, you become the superman, and you have to do everything, and that requires a lot experience and it is hard. It is hard to find such experienced engineers and the company suffers from it. (9). You see, CFD is not like a vending maching. It is more like a cookie monstor. (10). With no distributed cfd infrastructure in a comapny, the likelyhood to find the cfd superman is small. When you have problems, you come here and ask questions. There is simply no other senior engineers in a company to answer your commercial code questions. And even if there is one, he is likely very busy working on his own assignment. Cfd question is not like which key to use in the word processing. It could easily take another engineer many days or weeks to find the answer. (11). That is the state of the art of cfd in applications. The problem is: there is no infrastructure to use the cfd graduates, except the code vendors. Even in that case, they are more interested in the expansion of the market share, that is the sales engineers.

I currently work for a CFD company, and I am intrigued by John's response. Let me ask a stupid question. Do you not get any help at all by calling the code vendors user support line. They should be able to help the neophyte user through the learning curve, or, ideally, have the resources to find and fix more exotic problems. In your opinion, would it make sense to purchase user support services from a third party, with some hope of getting faster/better service?

(1). In terms of the user support services, I would say that it is a must for any commercial code. This is because the short course provided by the commercial vendor normally lasts from three days to a couple of weeks. And with so many options available from a commercial code, it is simply not possible to learn all of them. In this area, I would say, normally it takes about six months to master a commercial code if he is allowed to experiment with these options. But sometimes, he is given a problem to solve and a deadline. So, the user support service (related to the operations of the code) is extremely important. It is even better for a third party to provide the training for say three weeks, so that the user can understand the sequence of operations of the code. So, there is a real market there for those who can provide such services to potential users, to get the user up to speed. (2).For the real problem itself, even with my experience, I can't even say whether the problem (the assignment) can be solved by using this code, by this deadline. The geometry is so complex that one has to spend a lot of time to figure out how to model the geometry first. This is important because the current geometry is not the issue, but the future modified geometry would be of interest in order to reduce the flow separation or loss, or something like that. So, you can't just take the CAD model and then feed it into the mesh generation code. You are really dealing with the issue of how to model the geometry of future design to reduce the possible flow separation problem. (3). Anyway, at that point, it is not a good idea to take the problem outside the company, or even to another department for solution. It is really an issue of creating a solution concept, rather than a CFD computer run. (4). I am not working for a CFD company, and our training for a commercial is typically less than a week of time. A phone call or e-mail to support engineer will always resolve the code operation problem, but in general, it will have little impact on the ability to solve the real problem itself. This requires a lot of trial-and-error. I must say that in most cases, the difficulty of the problem is related to the complexity of the geometry. (flow through 3-D blade passage types are just routine cfd runs, even though the 3-D flow field is still rather complicated. But I am not talking about this type of problems. Even for this routine type of problems of 3-D blade passage flows, I can't send the geometry outside the company.) (5). In many cases, the geometry is the product, not the cfd solution itself. But one needs the geometry to run the cfd code, in order to modify the geometry. And it is not acceptable to leak the geometry to a third party.

Thanks for the detailed response.

I can see how, from the users point of view, the twin problems of mesh complexity and code capability are both firmly coupled. After all, they could both prevent a solution from ever being reached. However, they call for different skills. To take the specific example of flow within blade rows, for instance; I have seen several cases where the mesh is not unreasonable, but the solution drifts, or diverges. These problems are often the hardest to resolve, even for the code developers. This is especially annoying to the user if it occurs in features of the code which are advertised to work (and of course, it always happens close to the deadline when the mesh has been built with a lot of effort), and the code vendor seems unresponsive (they are usually backed up too!).

On the other hand, meshing questions usually are a matter of experience i.e. once you figure out by trial and error how to model a certain class of problems, the next one is that much easier. (PLEASE UNDERSTAND that I am NOT underplaying the difficulty or importance of this aspect; in most cases, the meshing takes significantly longer than the CFD solution). I guess this has to be appreciated by the management, that the first one of a kind will take 3-4 times as long as the rest.

There is also a third aspect of using commercial codes, which is to "push the envelope". Most codes have beta features, and "hooks" into the code for user coding to be included. This is another area where prompt and knowledgible user support is essential. There might be common blocks or variables which are not known to the end user, etc., which may make the task easier/possible.