[Getfem-users] Uncontinous stress problem (next try)

Hello dear getfem++ users,

I have a question, some of you may know an answer to:

I solve a linear elastic problem using getfem, where first the displacement of a gear tooth under load is computed an from this the von Mises and principle stresses are computed using compute_Von_Mises_or_Tresca(md, …) for von Mises and a slightly adopted own version of it for the computation of the principle stresses. Both functions basically call compute_gradient(mf_U,mf_ST,U,GRAD).

My problem ist, that also if the computed displacement looks very smooth (figure is attached), the computed stresses have local minima and maxima where they should be quite constant. Also at certain single nodes the stress values show unexplainable peaks. As I look for the maximum stress, this peaks are not good. Does anybody know where these local peaks are coming from? How exactly does the interpolation on the quadratic elements work? Is there may be a numerical problem?

Under the following link you can download some pictures of the displacement and stress results:
https://polybox.ethz.ch/index.php/s/cWKcMznQmYQLLda

Following link leads to an example results vtk file:
https://polybox.ethz.ch/index.php/s/ht3dfd5qniYrTZP

And this link to the imported gmsh mesh file:
https://polybox.ethz.ch/index.php/s/hG90v8SeLNlvTb4

As mentioned quadratic elements are used (FEM_PK(3,2)) are used, in order to get continuous stress values at the nodes.

Thank you for any help in advance.

Natanael Lanz

MSc Masch. Ing. ETH

IWF ETH Zürich

Technoparkstrasse 1

PFA E82

CH-8005 Zürich

address@hidden

www.iwf.mavt.ethz.ch

Tel. +41 44 6325714