5 Replies Latest reply on Apr 25, 2018 11:34 AM by Chase Evans

    Nonlinear analysis running slowly with direct sparse solver

    Andrew MacAluso

      I've been working on a nonlinear analysis that's proven to be more frustrating that I anticipated. I had a linear static analysis of the same model completing within about 3 minutes for some deflection information. I wanted to run a nonlinear analysis with a displacement input, but it was taking hours and hours to work thought 1 or 2 time steps. Something didn't seem right, and I suspect my contacts or displacement input was a culprit.

       

      So, I set up my nonlinear analysis to be an exact copy of the linear analysis (except that it was nonlinear, of course). Same material definitions, same loads, restraints, mesh, etc. And, still it churns for an obnoxious amount of time before I cancel it. It seems to be spending all the time on contact interations.

       

      I theorized that the only difference could be the time-stepping and load stepping, so I played with these parameters a few times to see if I could get a different response. After that failed to make a difference, I tried making other changes to the analysis set-up to see what might work. Eventually, I switched to the FFE solver, and it worked. The entire analysis ran to completion within a few minutes.

       

      Now I'm quite confused. I thought the direct sparse was supposed to be more stable, especially with non-penetration contacts. Also, the linear static analysis ran equally quickly with the direct sparse solver, so I wouldn't think that the nonlinear analysis would run several orders of magnitude slower with the same solver.