I am trying to model (using Fluent)flow past a prolate spheroid (3d Ellipse) to locate seperation lines at the aft end (masters dissertaion). I have tried both 2D and 3D models. The problem is I cannot seem to get results which show any seperation at all. I have tried all the turbulense models, wall functions etc. I have put a boundary layer in too, but to no avail. I have seen results proving it can be done using CFD but it will not work for me. I have tried all types of mesh on the spheroid surface, but I am restricted to unstructured mesh in the flow domain. Has anyone any idea how to solve this problem? All suggestions greatly appreciated. Thanks.

Boudary layer separation is caused by severe pressure gradients occuring downstream, for instance when the flow is passing through a diffuser. If there is a too high local adverse pressure gradient, this can be a cause of separation. You should be aware of what is the flow regime that you are having. Are you sure in your case separation is expected? You have to take in mind the kind of fluid that you are modelling, if it is a liquid or a gas, if it is compressible or incompressible. By changing the boundary conditions you can enforce separation, but this won't be your original case.

(1). I have said many times here already that it is a bad idea to use commercial cfd codes in graduate school. (2). I must say that flow over a 3-D body is the subject of a post-doctor. (3). Flow over a cylinder is more appropriate in this case. (4). It is hard to say what's wrong in your case. But I am sure that there is something not right. It could be the code, the mesh, the turbulence model, etc. (5). If you have to use a commercial code, my suggestion is: always starts with a working sample case. So, try to get a tutorial sample on the flow over a prolate spheroid from the vendor first. Let them show you how to do it so that you can follow their instructions. (6). If you have been reading my messages then you should already learned that a low Reynolds number turbulence model is required, which in turn will require very fine mesh near the wall. For smooth surface, you might also need fine mesh in the flow direction on the surface, which means a very big mesh size is needed. (7). My suggestion: try the laminar flow case first. But start with a low Reynolds number case. By the way, it is a good idea not to touch a commercial codes in the graduate school, if you are interested in getting a good solution in the first place.

I studied the viscous compressible flow over an OBLATE ellipsoid in my dissertation, titled "A computational study of flow around a rotating disc in flight". The shape was similar to the radome on top of the 767 AWACS. I located the separation lines by putting tufts on the surface. The comparison study for rotating and non-rotating case was truly enlightening!!! My dissertation, which includes a myriad of graphs, can be downloaded as a PDF from my website at,

www.cfd4pc.com/papers.htm

By the way, I wrote the entire program, including mesh generator, flow solver, and post-processor from scratch. I could not agree more with John: If you want to learn something in school, stay away from commercial CFD software. Learning how to use software, any software, is 'training', developing the software yourself is 'education'.

I am planning to make the software to my dissertation available for download some time in the future. If you give me the Re and Mach #, I can run the case for you and we can compare results, that is if you plan to compute in the compressible range (M > 0.3).