Learning Forum

[Meyzeeh]

Hi, everyone 

I can do mesh independence studies for steady-state simulations but I'm confused about how to do them for transient simulations, can anyone explain the procedure and how it differs from steady-state cases? And what is a time step independence study and can I use steady state results to do a mesh independence study for transient simulation? 

[Federico Alzamora Previtali]

Mesh independence studies involve systematically refining the mesh and observing how the solution changes to determine the appropriate mesh density for accurate results. The approach is the same for Steady or Transient results. Established mesh independence for a steady case may not be appropriate for transient case since you also need to account for temporal resolution in the latter, while steady-state cases may average transient effects (for example, using RANS).

For transient cases, you need to establish that your results are both:

• mesh independent: you are spatially resolving flow features that matter to you and those result do not change significantly with finer mesh;
• time step independent: you are resolving important transient effects and further reducing the time step does not change the results significantly.

[Meyzeeh]

 

Thank you very much for your reply. For the mesh independent should I use the same and fixed time step size for all the meshes? Or I can still use the adaptive where maybe only the minimum time step size is the same in all the meshes? 

I also want to initialize the transient using a steady solution during the mesh-independent study to speed up the simulation. How can I do this with all the different meshes such that they all have the same initial steady state or will I have to run separate steady solutions for each mesh maybe keeping the same number of iterations? 

I’m sorry for asking so many questions.

 

[Federico Alzamora Previtali]

I would suggest having the same time step for mesh independence study so that grid size is the only changing parameter between your simulations.

Good question, I have not tested this myself. Start with 1 steady-state solution and see if you can read it on different meshes (the topology should remain the same, only the number of cells may change).

[Meyzeeh]

Thank you very much again.