#1 2019-08-20 16:55:36

m_golbs

Member

From: Germany

Registered: 2009-11-09

Posts: 137

solving method for big modells with STAT_NON_LINE

Hello,

when I use a still higher refinement by mesh, sample 10.000 dkt to 1.000.000 dkt elements, same simple sample modell, I have convergence problems (convergence to a "offset by RESI_GLOB_RELA" [sample not 1e-08] ). I use Newton in STAT_NON_LINE. This is perhaps the normal effect of high dimension and Newton (fast solver versus curse of dimension)? A method like monte carlo for solving are possible in STAT_NON_LINE? Is the NEWTON_KRYLOV a approach to solving this  problems with curse of dimension? What can I use in code-aster as alternative to Newton in big none linear modells?

Thanks Markus

Last edited by m_golbs (2019-08-21 11:55:56)

---

The superfluous chase, is to miss the essential.
Jules Saliège