CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (

JP June 17, 2008 05:58

Hello; Ive reading about the disvantages of the nature of the mesh. Its suppose that the flux non ortogonal to the cell faces should be afected by a numeric viscosity. How important is that, and when i suppose to orient the mesh to the flow? how is related that viscosity induced with the numerical viscosity asocied to the scheme (upwind, 2 order..) i use? I guess any one could suply me an answer in order to increase the general CFD knowledge.


CycLone June 17, 2008 19:39

If you use a high order scheme (which CFX uses by default) and generate a sufficiently fine mesh, the numerical dissipation is insignificant. If you can generate a mesh which is orthogonal to your flow, you don't need a solution!

The important thing is to use a mesh topology which allows you to resolve the features of interest while minimizing user effort and turnaround time. There are many ways to go about this and a lot to learn as you gain experience. My advice is to focus on techniques that minimize your manual effort, not the solve time. For instance, if you may generate an appropriate hybrid unstructured mesh in an hour and it is 2 million nodes and take 8 hours to run, or spend a day and a half crafting an all hex mesh which runs in 4 hours, you haven't actually save yourself any time. Even if you need to do a lot of runs, solver time is less expensive than engineer time, so you would still be better off with the unstructured mesh and a few extra computers to run in parallel.

Generally, you can throw structured meshes onto parts that lend themselves well to it and unstructured on the more complex bits. The art is in determining which are which, and that comes with experience.

In terms of accuracy, all element types yield accurate solutions, the only question is with what resolution?


All times are GMT -4. The time now is 07:19.