abarret / advdiffsolvers Goto Github PK

The extrapolated integrator still has some limitations that should be addressed.

• Full integration on both sides of the domain
• Positive level sets corresponding to the "physical" domain.

From #39. It would be good to move some of PointwiseFunction out of the class. It shouldn't be necessary to create an object to do these point-wise function evaluations.

For ExtrapolatedAdvDiffHierarchyIntegrator to be used correctly for level sets that abut the physical wall, we will need to hack the boundary conditions used in the convective operator classes. The boundary conditions need to be consistent with however we are extrapolating the concentration fields.

One possibility is to overwrite the RobinBcCoefStrategy class to fill in boundary coefficient data via the extrapolated concentration field.

As we move away from using cut cell methods, we need to do calculations based on level set values instead of cut cells. For example RBFReconstructCache originally determines stencils based on cell volumes. #35 and #36 try to get around this by adding options based on level set values, but this makes the code more convoluted. We need more clear pathway or perhaps a different class structure to make these switches more clear.