[petsc-users] Solvers for phase field fracture

[Blaise A Bourdin]

Hi Avtar,

I am assuming that you are referring to spatial variations of the phase-field variable and that the fracture problem that you are solving is well posed (i.e. displacement controlled, thermal loads, work controlled). As we explained in our 2008 book, force driven problems are often ill posed. You may use an arc-length formulation, of minimize the fracture energy subject to prescribed work, and go back to a force-drive evolution (subject to a safe-load type of condition).


In a staggered scheme (alternate minimizations, block  Gauss Seidel, adjoint-based quasi-Newton, call it whatever you like), the elasticity problem can become stiff, but remains really manageable.

Feel free to try my implementation https://bitbucket.org/bourdin/mef90-sieve for comparison.

Blaise

On Jan 4, 2018, at 12:12 AM, Avtar Singh <er.bhullersingh at gmail.com<mailto:er.bhullersingh at gmail.com>> wrote:

Hello Petsc Users,

I am solving a fortran framework to simulate phase-field fracture problem. Upto the failure point the code is working fine. But as the crack-phase starts to propagate, there is abrupt changes in the Stiffness Matrix, so determinant becomes nearly equal to zero.  Hence, the code crashes.

I tried with mumps package with pc lu. Also tried, superlu, gmres and fgmres with jacobi, bjacobi, lu. But the problem still persist.

Can anyone suggest, Which solver and preconditioner should i use?

Thank you

--
Avtar Singh
Research Scholar
Multiscale Mechanics and Multiphysics Lab
IIT Roorkee, Roorkee

--
Department of Mathematics and Center for Computation & Technology
Louisiana State University, Baton Rouge, LA 70803, USA
Tel. +1 (225) 578 1612, Fax  +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin







-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20180104/d033cd68/attachment-0001.html>