>> It struck me however, please correct me if I am wrong, that since we >> working with global shape function (fe), the integration >> (fe(old),fe(new))_new domain is merely loop over all cells and its >> simple provide to fe(old) the new domains quadrature points! > > Yes, but it's not that simple because you need to find which cell of > the > old mesh and where within that cell a quadrature point lies. > > This is only simple if the new meshes has originated from the old > one by > refinement/coarsening. It is very hard if the two meshes are defined > on > different domains (for example because the domain changes shape in > each > time step).
Actually this is very simple in the deal.II library. There is a function called FEFieldFunction which will behave exactly like an ordinary function, and will interpolate for you the solution of your FE calculation, doing all the dirty work in the background. >> But here comes the obstacle: if the quadrature point for new domain >> is >> outside the domain for the old, does it automatically give zero for >> fe(old)(new quadrature) or i must do a GridTool->is_within check? The function above will throw an exception which you can catch and set it to zero if you like. >> Further, for large deformations one should simply drop the overlap >> functions, the part i am interested is the inertia force preventing >> immediate reaction to the fluid, so then i could actually have a if >> check, doing a quasistatic calculation then ask after, is the >> deformation to large? At that point i choose either step-23, or >> step-28, >> or my version for large deformations, drop the overlap terms. > > I don't know what the overlap terms are you are talking about. > Independently of that, I really believe you should in a first step > try to > take the old solution to the new grid using the SolutionTransfer > class. > Everything else is too complicated to do in the first step. If what you are talking about is a fixed fluid domain in the background, with an immersed body in it, then FEFieldFunction will do the work for you. You will need to cache some information to make it fast, but as a starting step, this is probably the easier thing to do. Luca. -- Luca Heltai <[EMAIL PROTECTED]> http://www-dimat.unipv.it/heltai University of Pavia, Dep. of Mathematics Phone : +39 0382 98 5680, Office: A10 -- There are no answers, only cross references. _______________________________________________
