3 Replies Latest reply on May 13, 2016 11:43 AM by Keith Frankie

    Getting nonsense results from simulation

    Noah Thomlison

      Hello everyone,

       

      I am having an issue running what should be a simple simulation to my part. The results that I am getting are completely nonsense and when I probe for max/min they are both something like -1.#io+000 for all values (displacement, von missus)

       

      This is the first part that I have run a simulation on which is made up of surfaces instead of bodies but I do not think that overly matters though in general I have limited exposure to surfaces. I have assigned all the surfaces thicknesses and material properties within the surface manager and believe my fixtures and external loads are correct.

       

      If anyone has some helpful tips or insight into what is causing this I would really appreciate the assistance.

       

      Noah

        • Re: Getting nonsense results from simulation
          Keith Frankie

          A bunch of the shell edges aren't bonded correctly.  The excessive displacements are stopping the nonlinear solver.

           

          Turn off the 'large displacement' option and run the study.  Enable the 'Deformed Result' option and you can see massive deformations on non-attached parts.

           

          You can use the contact visualization plot to help verify bonded edges.

           

          vis.png

           

          deform.png

            • Re: Getting nonsense results from simulation
              Noah Thomlison

              Kieth,

               

              Thank you for the assistance. Did you use the contact visualization to see the areas which are not bonding correctly?

               

              Noah

                • Re: Getting nonsense results from simulation
                  Keith Frankie

                  Bonding is often a problem.  I started by trying the linear solver and the results made it clear bonding was the issue.  I then used the contact visualization plot to get a better sense of what was bonded and what was not.

                   

                  Make a list (perhaps mentally) of all the bonding surfaces.  Then either 1) manually define each one or 2) check each one on the contact visualization plot to make sure it is bonded.

                   

                  This looks like the surfaces were imported.  You might be able to extend surfaces in SW "up to surfaces" to get everything to bond automatically.

                   

                  This follows the "crawl - walk - run" method for solving a study.  Start with a well defined (fixed) linear study and work your way up to the accurately setup nonlinear study.  See The Crawl Walk Run system with SolidWorks Simulation Part 1 - YouTube

                   

                  Actually I think this part is fairly "lucky" - usually with this level of bonding issues you'd find one part that wasn't bonded at all, and this would prevent the linear solver from completing.  If this were to happen you can "exclude from study" all but one of the components, verify that works, then add components back in one at a time.  You could do this in a frequency study, which solves quickly.  If a component is only bonded in one area it will flex around a lot, and this will easily be seen by animating the mode plot.