Hi,
Is there a relation between the diameter of cylinder and the mesh size that is to be used in simulations. I am trying to model a 27Dx14D rectangular domain with a cyclinder's centre spaced at 7D from the inlet boundary and the topwall boundry. I am struggling to get a right size of mesh for mesh validation. So far I am using D=0.1m and my base size for mesh is 0.01m with refinement around the cylinder as 10% of base size. For Re=10^6 i get values of cd=0.14 whereas the published data is more in the range of 0.2-0.4 range so i cant validate the mesh.
The arrangement is shown in the attached file.
Hi,
must not be a matter of the mesh. Can you provide more details with regards to your simulation setup?
You may perform some simulations with different base sizes and plot the cd versus the numbers of volume cells. The curve you obtain from this plot should reach an asymptote and give you some indication on how fine the mesh should be. There's no "absolute" answer, the mesh size will depend on the problem you are solving. Usually, a mesh that is a fair approximation of the geometry is a good start. If your results don't match literature maybe you should revise the physics of the problem (i.e. the turbulence model).
Yes your results should depend on mesh size, boundary conditions (inlet profile), turbulence model used, and numerical schemes.
In regards to the mesh you need to continue refining your grid until you reach grid independent results. However interestingly, if your grid quality is bad (skewed cells and not orthogonal to the cylinder surface numerical errors could be affecting your results). So you need to check for grid quality in Star-ccm+. (are you using polyhedral grids?).
I use jg mesh from microval france there are two sizes 11x14 or 12x16 according to the patient regards pierre
Also, since you are dealing with viscous flow, the computed drag coefficient will be strongly dependant on the prism layer. A first guess of its total thickness could be Re-1/2 while the thickness of the first cell should be small enough to satisfy your turbulence model (e.g. with Spalart-Allmaras the wall Y+ should be of order of magnitude as unity) but not too much with respect to the core (volume) cells, otherwise divergence will occur.
If you want to proceed easy, try with Re-1/2 as total thickness, then assign a ratio of 1.1 to the thickness ratio. If your residuals do not converge or diverge, then the aspect ratio of the prism layer cells is too small with respect to your volume mesh (whatever the type) and you should reduce your base size. Otherwise, if they converge but you are unsatisfied with your results, reduce the prism layer thickness or increase the thickness ratio (or change other prism layer properties), then continue to run your simulation.
As such there is no specific answer for size of mesh you have to perform the analysis and check the answer to see the results.
- Badges
- Science topic