Grid independency test
Can anybody tell me what is the reason for not achieving grid independency?
|
There is a multitude of possible reasons.
If you describe more in detail what you did, what you expected and what you got... |
grid independency
I want to simulate swirling flow in a cylinder and just wanted to do the grid independency test. But I got different plots of swirl and axial velocities near wall and axis region. I just wanted to know the reason for this behaviour. Does this happens or is there any mistakes in my simulation?
|
Is the flow laminar or turbulent? What is your resolution in terms of the smallest scales in the flow? Do you use any type of model? Tell us more about the numerics you are using to solve this flow. Has the code been verified?
There are so many possible explanations for this behavior. You need to provide the parameters of your physics and your numerics for us to help you. |
grid independency
I am simulating turbulent swirling flow using CFX in a cylinder using standard k-e model. meshing is done using icem cfd. I just increase the number of nodes on inlet face by multiples of 5 to find out grid independency.
|
Decreasing the cell size in only one region of the mesh will not reduce the discretization error in the other (constant-size) regions of the mesh.
|
The k-e model can tend to give strange grid convergence results if you are decreasing the first cell height at the walls with increasing mesh refinement. Check to see that your wall functions are valid, or switch to a more appropriate turbulence model.
|
Quote:
This kind of tests are very important to show that the solution you obtained is accurate enough. You should know that your solution is dependent on the size of your grid cells. The smaller is your cells, the more accurate your solution will be. But below a given grid size the solution becomes independent of this size. It means that you stop to gain accuracy below this given grid size. Then it is not anylonger necessary to decrease the grid size. So you have to perform this test to be sure that you have run your computation with the right grid size. Keep in mind that these tests may concern spatial grid size (dx, dy) or time step (dt) To my mind these tests should be always performed before delivering any solution.This is a must. |
Keep in mind, as suggested by Adrian, that some models, such as wall functions, are dependent on the mesh/time-step size. Other models to look out for includes: solidification, melting, evaporation...
Another aspect is that when doing mesh/time-step refinement, the accuracy of the solution increases (and the dissipation goes down). This creates a "conflict" where you are able to achieve convergence for coarse mesh, but fine mesh may exhibit transient behaviour... |
Hello prom,
I just recomend a book verification and validation. There's a very detailed discussion on this topic. Other replies in this forum are also valuable. This book is, Roache, P. J., Fundamentals of Verification and Validation, 2nd eds, Hermosa publishers, 2009. |
Gid
Dear
GID is varies from case to case , you need to conduct a timestep independence study. Reagrds |
All times are GMT -4. The time now is 07:31. |