You are here
Reg. Convergence and Mesh Density
Hi
I am using a FORTRAN FE code for 4 quad elements with a viscoelastic material and with contact analysis. The routine follows a Newton Raphson iteration scheme and which makes the residual zero for convergence. I am using direct sparse matrix solver from the FORTRAN math library.
When I use relatively lesser number of elements, about 1500 the solution converges and results are as expected. However, for the same analysis if I use more elements say about 4000, then the analysis starts to diverge midway. The residual starts to increase exponentially and the solution doesn't converge.
What could be the reason for this and how can I overcome it? Any suggestions would be helpful.
Thanks,
Shriram
Research Assistant
University of Florida
P.S. I even use an automatic time stepping scheme, but it doesn't help.
Try to reduce the load steps.....
Dear Shriram
Try to see the results via post processing.
Its look like, with increasing number of elements, your model is
getting some local region (convergence problem) problem, may be your
model is not able to capture such local behavior with less no. of
elements. Hence, your model is getting converged with less no. of
element.
If this is the case, you can try with increasing no. of
loading step. and also the tolerance limit also (just to check whether
still your model is diverging or started converging). Later on you can
reduce such tolerance limit as per your requirement.
Attached link may be further helpful to you
http://solidmechanics.org/text/Chapter8_3/Chapter8_3.htm
regards
Vijay