Mr. or Ms.?? i'm using flotran of ansys to solve a compressible problem with viscosity. can ansys capture the boundary layer with the mesh being refined? the reason i ask so is that i didn't find the equations of the boundary layer in the theory preference. by the way, how to rifine the mesh when solving a compressible problem. as to my problem, the solution goes well with coarse mesh, but doesn't work with refined mesh. is there some principles for refining mesh in finite element method? such as, when refining mesh, how to pre-estimate the gradient of the grids according to the variables and equations you are solving? wait for your help warmly. thanks!

langma. langma@netease.com

(1). We don't know what is in the code you are using. (2). But I think you can try to find out the original author of the code and look at his background first. (3). Once you have that information, you can decide whether to contact the author or not and find out why the fine mesh does not work. (4). It is extremely important to know the original author of the code, because the code is just portion of the instructions to perform the author's idea. You should try to get some explanation from the support engineer of the code vendor. You also should try to find out whether it is necessary to refine the mesh or not. My impression is finite-difference and finite-volume methods require mesh refinement to obtain the true solution, while the finite-element method is based on the minimization process to obtain the solution.

Mr. Chien:

thank you very much for your replying in time.

i think the boundary layer is important in my problem, because i mainly want to get the pressure , film coefficient and temperature on the wall due to the fluid field. i've visited xansys for help, but no answer for the time being.

so i hope anyone here, who uses flotran of ansys, can lend me a hand to handle with this problem.

Are you saying that grid refinement is not necessary for FEM approach? Also, I have not heard anyone talk about grid refinement studies for 3D problems. Is grid refinement necessary for 3D?

Thanks, Chuck

My previous message was actually meant for John. Sorry, Chuck

(1). That's my impression about FEM. (2). My feeling is that the mesh size of 100x100x100 is required to have numerically accurate solution for simple flow through a turbine blade passage in order to capture the secondary flow and loss. (3). This is consistent with my study in early 70's with 2-D flows, which required 100 cells across a channel when the grid points were properly arranged. (with a low RE model) (4). Here we are not talking about the accuracy of the turbulence model, only the numerical accuracy is involved. (that is you need roughly one Mega cells for mesh independent solution in 3-D. With wall function, it takes less amount of mesh points, but the usefulness of the solution is rather limited)