Discussion Forum

The COMSOL applications Library contains plenty of applications, and the Cylinder roller contact is one application in Structure Mechanics Module->Verification Examples, which has computed the deformation of contact of two rigid body.
The computed pressure is precisely close to the analytical one, it's cool. However, when I try to reduce the force load (corresponding parameter is 'Fn') from 35kN to 20kN, the application encountered a convergence problem.
Then I changed parameters in the application:
1 Steady state solver->Separation step->Non-linear method to 'Auto highly non-linear (Newton)'.
2 Thickness of Structure Mechanics Module (corresponding parameter is 'Fn') from 1mm to 100um.
3 distance between the contact pair (corresponding parameter is 'dist') from 1mm to 100um
4 Point load to boundary load.
5 Geometry repair tolerance to 'relatively 1e-9', Extrapolation tlerance to 1e-3, Structure Mechanics Module->contact 1->Equation view->multiphysics contact tolerance (solid.cnt1.tolcontact) from 4.14e-6 to 4.14e-8.
6 remesh the grid
Now the force has been reduced from 35kN to 360N, but I can't advance anymore, since the mesh troubles: I've tried to make the original mesh finer and all free triangles mesh, but the original mesh has the best performance. The phenemena (Separation Step 1 of displacement field, comp1.u, is not convergence) confused me: Making mesh finer or rebuilding a very fine mesh based on finer gemotry make negative contributes to the convergence of this application.
I wonder if the problem is that the force equation solved gives a tiny displacement to other equation when the force loaded is very small.