For the thermal simulation, when is it necessary to have the mesh independence test? When I am using Star-CCM+ to do the thermal simulation, I find it hard to get an accurate temperature even the number of mesh keep increasing. The varying scale is from 0.9 million to 30 million. The simulation geometry model I use here is a cylinder head with a real dimension.So is it different to have the mesh independence test for the thermal simulation compared with common simulation without solving the energy equation.
I am not using Star-CCM+ but Comsol. The meshing problem is universal. You need to refine the meshes structure in the volumes where the thermal phenomena have an important effect and not in the whole domain. Look at your simulation with 900000 nodes and try to find the zones where the gradients are important. It also important to understand where temperature is not converging.
It is always better to perform at least 1 or 2 mesh independence test. This can for example allow you to reduce the mesh size without loosing accuracy and, therefore, reducing the computational effort. Off course you can also discover that your mesh is not fine enough... but at least you will know that your simulation results should not be trusted.
Franco
Also, you could follow an output parameter like a temperature and its dependance on grid.
It is always recommended to perform grid independence test for the cases or numerical models which results depend on grid size. We should always keep one thing in our mind that grid changes are only required only at some places e.g in the steep gradients area. Finally it is up to a uses how he/she is going to design his/her grid.
I have never used star CCM+ but it may happen that grid independence test may not work for some cases. One may choose some experimental data or the data from any published paper to validate the grid.
please refer to following link, hope it will help you.
http://www.computationalfluiddynamics.com.au/convergence-and-mesh-independent-study/
The best time is when you get a stable solution. Don't wait too long for convergence. If you get stable solution not converging then change the grid to 50% of existing grid and check the interested parameter. If it is within 2% or 3% range, then grid-independence test is successful.
- Badges
- Science topic