[albcem]

Hi everyone,

As you may know extreme mesh deformations due to body motions can result with very bad quality grids. So either one ends up with a simulation blow up or very small time steps to fulfill the CFL condition.

I have been thinking of and testing an approach that involves the following steps:

1. stop simulation when grid quality is bad enough
2. create a new grid from scratch preserving the latest position of the body in motion
3. map the latest solution from the old grid onto the new one
4. resume simulation on the new grid

I am having a little trouble with MapFields in step 3. Fields are not mapped well or at all. I think part of the problem is that the pointMotionU and cellMotionU files define the way the old grid is deformed, so MapFields may not be interpreting the old grid right. As a side, can I use "-consistent" flag given that the geometric position of the body and its faces are the same, except that the grid topology is different, hence points will not map one-to-one?

How about alternative approaches?:

Star CCM+ utilizes the over-set grid approach in which the interpolation between body-attached and stationary grids is performed at each time step. OpenFOAM does not yet have this capability forcing me to work with what we already have.

I know sliding meshes might be an alternative, but I doubt it will work for non-academic cases, that involve not so ideal convex assemblies - ie a boat with a tall mast. One would accommodate the assembly in huge (relative to boat size) spheres which slide to take care of 3DOF rotation part. I think the cost for interpolation face search would be very high in these cases...



Comments are appreciated.

Cem