CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Node values problems (http://www.cfd-online.com/Forums/cfx/21070-node-values-problems.html)

 Kiddo March 10, 2005 10:08

Node values problems

I do a transient conduction analysis(about 200 run steps and residual target is 1E-4). There're two domains, inner and outer.(the inner one is fully inside the outer one)and the mesh nodes are fully matched at interface(created by ICEM_CFD simultaneously). I export both the domains' temperature data and I found that the temperature at interface nodes are slightly different in different domain!For example: for inner domain x1,y1,z1 359.4359 DegC . . for outer domain x1,y1,z1 359.4285 DegC (note: same coordinate) The temperature should be same for the same node in different domain, right? Thanks for response.

 Glenn Horrocks March 10, 2005 16:56

Re: Node values problems

Hi,

I suspect the small difference is caused by convergence, and as your simulation converges tighter the difference will get smaller.

Glenn Horrocks

 Kiddo March 10, 2005 20:47

Re: Node values problems

Hi Glenn Horrocks, thanks, I do the following trial run cases for the same definition file: 1. parallel run, Converge Target.=1e-4 2. serial run, Converge Target.=1e-4 3. parallel run, Converge Target.=1e-5 4. serial run, Converge Target.=1e-5 (also run transient and steady case for above cases) and found that 1. It doesn't matter with transient or steady. 2. It doesn't matter with the parallel or serial run 3. It always got a slightly difference for the node value at interface node in different domain. And found also the Converge Target affect the result no very apparently. I know CFX is by Finite Volume Method which is a special case of Finite Element Method. Is it possible due to the FVM natively?

 Juan Carlos March 11, 2005 01:02

Re: Node values problems

Hi,

The values are supposed to be different unless the problem is isothermal..

Either node does not represent the value at the boundary, but the mean value of the finite volume (or half volume to be precise) it represents. The smaller the finite volume (finer mesh) the closer these 2 values will be..

Hope this helps, JC.

 Kiddo March 11, 2005 02:08

Re: Node values problems

Hi Juan Carlos, thanks, I refine the interface mesh by 4 times (and it's very dense now). I re-run it and I found that the difference is almost the same as previous mesh.

 Rui March 11, 2005 14:37

Re: Node values problems

Hi, If the temperature gradient is perpendicular to the interface, you have to refine the mesh in that direction for those 2 values become closer.

Depending on how you are creating the volume mesh, refining the surface mesh at the interface may lead or not to a finer volume mesh on the direction perpendicular to the interface.

Regards,

Rui

 Kiddo March 11, 2005 22:44

Re: Node values problems

Thanks, but maybe the mesh size is not the key point. I create a new simple geometry just for test only. Even I refine the volume size globally 5 times and compare the results. The difference is almost the same. I'm wondering there's something inside the solver theroey which affects the node value's difference. Thanks for response again. -Kiddo

 Alex March 12, 2005 06:43

Re: Node values problems

 Kiddo March 13, 2005 20:29

Re: Node values problems

Hi,Alex , I got it ! that's the key point. Thanks!!!!!!!!!!! The reason is described in Solver Theorey-Numerical Discretisation p. 228 Alex, thanks again!

 Kiddo March 13, 2005 20:41

Re: Node values problems

Hi, Juan Carlos: After Alex reminds me seeing the Solver Theorey, and I found you are right . The reason why I cannot get closer difference after refining mesh may be due to the inner domain has a subdomain with a heat source assigned. So there'll be always a gradient over it.

 All times are GMT -4. The time now is 05:52.