On Thu, 23 Jul 2015, Paul T. Bauman wrote: > On Wed, Jul 22, 2015 at 7:13 PM, Sahai, Amal <sah...@illinois.edu> wrote: > >> I have a system with 3 variables, representing the x,y,z component of a >> quantity. I solved this system to obtain the component wise distribution in >> a given domain. I would now like to calculate the divergence of this >> distribution at the node points and then output this along with the >> previous 3 variables. How do I compute the divergence at the node points? > > What finite element basis are you using? Depending on the basis, the > divergence isn't going to be continuous across elements, so the nodal value > is well defined. (I'm pretty sure we don't even have an FE that would have > continuous divergence implemented in libMesh.) You could do something like > patch recovery to extrapolate the quadrature point values to the node.
If you want to approximate computed values via a global L2 projection, there's a new little app for that in libMesh now: "calculator". For divergence of (u,v,w) you'd do something like: calculator-opt --inmesh mymesh.xdr --insoln mysoln.xda.gz --calc 'grad_x_u+grad_y_v+grad_z_w' --- Roy ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
