[mohsinmkhan]

Can anybody tell me difference between embedding scale and base grid size? on what grounds its has to be choosen ? any predefined value are there? & is there any relation between time step and base grid and embedding scale ?

[ksrivast]

Hello Mohsin,

Fixed embedding refines the base grid as follows : cell_size = base_size/ (2^embedding scale). So, for example, if the base grid is 4mm, after refinement with embedding scale of 2, cell sizes are now 1mm in the fixed embedding volume.

Embedding scales are chosen as per grid resolution requirements of the user. A suggested way is to open an example case similar to your problem of interest and adopt similar grid settings. Increase the embed scale to have enough resolution within small gaps, near spark regions during spark event, etc. It is recommend to perform a grid independence study to ensure you have adequate refinement.

Certain time step limiters are affected by grid sizes : dt_cfl*, dt_move, etc. These dt limiters calculated for each cell reduce as local cell sizes reduce. You would see a corresponding simulation time-step reduction if these refined cells are more influential than the base grid cells.

Please refer to our manual for more information.

Hope this helps,

Sincerely.

[mohsinmkhan]

Originally Posted by ksrivast

Hello Mohsin,

Fixed embedding refines the base grid as follows : cell_size = base_size/ (2^embedding scale). So, for example, if the base grid is 4mm, after refinement with embedding scale of 2, cell sizes are now 1mm in the fixed embedding volume.

Embedding scales are chosen as per grid resolution requirements of the user. A suggested way is to open an example case similar to your problem of interest and adopt similar grid settings. Increase the embed scale to have enough resolution within small gaps, near spark regions during spark event, etc. It is recommend to perform a grid independence study to ensure you have adequate refinement.

Certain time step limiters are affected by grid sizes : dt_cfl*, dt_move, etc. These dt limiters calculated for each cell reduce as local cell sizes reduce. You would see a corresponding simulation time-step reduction if these refined cells are more influential than the base grid cells.

Please refer to our manual for more information.

Hope this helps,

Sincerely.

How can we do grid independence study? & Sir if time step decreases then simulation time will increase, isnt it?

[ksrivast]

Hello Mohsin,

The grid sizes are reduced by half and results obtained from the refined mesh are compared with the coarse mesh to observe the difference. If the differences aren't within an acceptable tolerance, you haven't attained grid independent results yet and must switch to finer mesh resolution. An easy way to refine the grid in the entire domain is to set grid_scale value in inputs.in to 1. This would reduce the base grid by half. Increase the value to refine the mesh further. Please refer to our manual for more information on grid control techniques.

Increase in run times with reduction in time steps is not always the case. If you're having stability issues and the solver is working hard (too many individual iterations) with your current time steps, smaller time steps might make your case more stable and the solver works less to obtain converging results within an iteration. This could result in overall reduction in run times.

Sincerely,

[mohsinmkhan]

Originally Posted by ksrivast

Hello Mohsin,

The grid sizes are reduced by half and results obtained from the refined mesh are compared with the coarse mesh to observe the difference. If the differences aren't within an acceptable tolerance, you haven't attained grid independent results yet and must switch to finer mesh resolution. An easy way to refine the grid in the entire domain is to set grid_scale value in inputs.in to 1. This would reduce the base grid by half. Increase the value to refine the mesh further. Please refer to our manual for more information on grid control techniques.

Increase in run times with reduction in time steps is not always the case. If you're having stability issues and the solver is working hard (too many individual iterations) with your current time steps, smaller time steps might make your case more stable and the solver works less to obtain converging results within an iteration. This could result in overall reduction in run times.

Sincerely,

That was really a nice explanation sir. Thanks a lot.