2 Replies Latest reply on Oct 8, 2016 2:48 PM by John Willett

    integer limit with simulation 2016

    Chris Pellegrino

      Small simulation, actually. Mesh details say it's 360k nodes and 250k elements. I do have two bolt connectors and No Penetration between three parts (actually two of the parts are bodies from the same part file.) It is also a 72deg slice of a symmetric assembly, so there is also a cyclic symmetry set up.

       

      This doesn't sound like a hardware issue (32GB 16bit i7 quad core), rather some other sort of limitation in the software. If this is too large of a problem for the software, that is very lame.

       

      I need the no penetration because I need it to solve for the reaction between the parts since that defines the friction between them. I can try without the symmetry, but that seems counterintuitive to why you use symmetry...

       

      Anyone else know what to do to deal with this?

       

      *EDIT (next day) Well, I removed the cyclical symmetry and it ran. WTF? Symmetry is supposed to make it smaller. Bug?

       

      Here we go again.... SO MANY changes in 2016, most of which do nothing to make the tool work better and be more robust, but stuff like this slides by. THIS IS WHAT WE BOUGHT THE SOFTWARE FOR - THE FUNCTIONALITY; NOT NEW ICONS THAT MAKE THEM HARD TO FIND NOW (not to mention the colors...) Come on people..... how much do you get payed? I want your job...

        • Re: integer limit with simulation 2016
          John Willett

          What does "integer limit" actually mean?  I found only one (rather old!) related Knowledge Base entry, S-062259, but it doesn't help.

           

          I have a similar problem with an apparently even simpler simulation -- only bonded contacts, three parts, 713090 nodes, 496334 elements (primarily a result of needed mesh controls on four surfaces).  Currently running SW Premium 2016 SP3.0 (64-bit on a similar Windows 7 SP1 machine, also with 32GB memory and plenty of available disk space).  The error message I get is:

           

          "The solver is currently running out of integer limit. It is recommended that you change the solver from Direct Sparse to Large Problem Direct Sparse (see the study properties).

          Would you like to change the solver to Large Problem Direct Sparse now?"

           

          At the time of the error the solver screen looked like this:

          Occasion of Error Message.png

          I never saw more than about 2 GB of memory usage.  The simulation does run (and quickly) with the Iterative solver, but I suspected the results and wanted to compare with the Direct Sparse solver.

           

          Should I go back (with considerable difficulty!) and try running the problem in SW2015? -- John Willett

          • Re: integer limit with simulation 2016
            John Willett

            OK, I think I found the problem, in my case at least.  I can get the Direct Sparse simulation to run if I decrease the resolution of the aforementioned mesh-control resolution slightly.  The memory usage now maxes out (during the "Decomposition of stiffness matrix" phase, which I never got to before) around 17 GB, which from past experience is pushing my limits.  I guess it really was a memory problem after all...