On Mar 23, 2013, at 1:43 PM, David Knezevic <dknezevic@...> wrote:
> Regarding (2): Are you thinking of adding a libMesh example that implements transport of a signed distance function using DG? Or are you thinking about adding core functionality related to this, e.g. a new System subclass for handling this? (IIRC it's not just standard advection problem in this context, since you need to renormalize after every time-step to ensure it remains a distance function
more along the lines of the core functionality. a system, provided along with the library, that discretizes a scalar advection equation for a level-set function & implements the renormalization. You are right, advection will move the interface properly but rescaling is then required I think to properly update the distance.
As it is a hyperbolic advection equation DG seems like a natural choice, but SUPG could also be used…
But I was thinking since the level-set scheme is central to a lot of interface tracking stuff (not just X-FEM) direct library support might be worthwhile. Especially since the renormalization is not trivial for a parallel mesh!