On 11/21/2014 11:29 AM, Ronghai Wu wrote:
I compared the displacement calculated by sfepy and by Comsol (fine mesh, both for free boundary condition and fixed bottom boundary condition). No matter what boundary condition it is, sfepy result and Comsol result agree with each other perfectly. I mean sfepy free boundary result == Comsol free boundary result, sfepy fixed bottom result == Comsol fixed bottom result. But free boundary result =! fixed bottom result. So, I still have three questions, please forgive my ignorance on FEM:
Maybe you have not the umfpack solver installed. In that case superlu (which is budled in scipy) is used.
With no boundary conditions, the solutions can contain _arbitrary_ translations and rotations of the whole domain as a rigid body. So the agreement of comsol and sfepy is a coincidence, maybe the linear solver uses the same strategy when dealing with singular matrices.
Assuming I use strain = pb.evaluate('ev_cauchy_strain.2.Omega(u)', mode='el_avg'), if the strain result is the strain at cell centers?
It is an average in the elements, so yes, it can be said that it is a strain at cell centers.