6 Replies Latest reply on Apr 11, 2013 3:02 PM by Jared Conway

    Equilibrium is not achieved - how to improve setup?

    Chris Michalski

      Okay, I've searched and found other discussions where equilibrium wasn't achieved but they were all assemblies.

      I have a single part - flat sheet of glass as a solid - input thermal data from a flow simulation and temperature dependent material properties. 

      Add 2 X/Y restraints to prevent it flying off into space.

      Add gravity.

      Use a virtual wall to represent it sitting on a flat surface.

       

      The temperature differences should cause it to distort and bow up in certain places (bowing down being resisted by the virtual wall).

       

      I keep getting an "Equilibrium is not achieved" dialog to save up to the current results.

       

      Is there a way to force it to keep calculating to see if it does converge?  I'm not in a time crunch so letting it run longer wouldn't be an issue - if it gave me that option.

       

      I can run it without the virtual wall and it solves, but in that case the glass droops all over the place as if only supported at the corners.  This suggests the virtual wall is ultimately linked to the problem.

       

      Suggestions?

      equilibrium not achieved.jpg

        • Re: Equilibrium is not achieved - how to improve setup?
          Jared Conway

          "I keep getting an "Equilibrium is not achieved" dialog to save up to the current results."

           

          before you get to that point, click on pCG reset and conv_plot you might have some options. same goes if you switch to direct sparse but those may have been added in later versions of the software. in earlier versions, only nonlinear had any control.

           

          regarding virtual wall vs no wall, thats a good test. so that means your problem solves but it is the contact between the virtual wall and the component causing the issue. i'm assuming you mean with no soft springs or inertial relief and that in general the results look correct?

           

          you may want to try with a real wall instead of virtual just to see if there is a difference. do you get any results at the end when it fails? how do they look?

            • Re: Equilibrium is not achieved - how to improve setup?
              Chris Michalski

              I tried the PCG_Reset and set the maximum iterations way up (it normally hits the limit around 250 but I set it up to 1000 anyway). (didn't help)

              I had it on Auto so I'm running it now with Direct Sparse selected.  The Direct Sparse hasn't brought up any options yet (stop and pause, no PCG_Reset or Plot)

              Depending on that I'll try swapping the virtual wall for a real wall.

               

              After my other recent adventures  I started simple this time, no springs or interial relief.

              I originally had 3 corners restrained for no vertical motion - but the glass is soft enough that it droops because of gravity.  Hence my virtual wall. Why virtual instead of real?  Because it was there   Again, not very experienced in non-thermal simulations so I took the first viable option.

               

              With a virtual wall the results look reasonable, I'm just concerned that it's not converged so "looking right" and "being accurate" aren't necessarily the same thing.

              • Re: Equilibrium is not achieved - how to improve setup?
                Chris Michalski

                SPARSE didn't help - resulted in large displacements that are unrealistic.

                 

                When you say real wall, I'm assuming you mean a virtual wall with specified stiffness?  Or is this a newer option than 2009?

                  • Re: Equilibrium is not achieved - how to improve setup?
                    Jared Conway

                    real wall, meaning modeling a wall and applying friction. fix the wall in space.

                     

                    when you ran sparse, did you get the large displacement warning? if so, the results would basically be junk unless you ran it with the large displacement option and/or nonlinear. if you do get this message, that would explain the FFE problem and the results you're seeing. most likely a setup issue.

                      • Re: Equilibrium is not achieved - how to improve setup?
                        Chris Michalski

                        it gave a large displacement warning and I told it solve based on small displacements because the FFE came in under 1mm.  I'm trying again with large displacement selected.  For some reason now it changes temperature and complains about being outside the range of my material definition.

                         

                        If I put in a real wall I'd have to rerun the thermal portion though right?  IIRC you have to use the same geometry for the thermal and static study in order to transfer in the thermal conditions.

                          • Re: Equilibrium is not achieved - how to improve setup?
                            Jared Conway

                            the temp changes, i'm not sure about. sounds wonky but i haven't played a lot with imported temperatures and pressures from flow. usually i use it as a gauge and then use the BC temperature using an average temperature.

                             

                            on the other part, you're correct. in 2012/2013 things changed, but back in 2009, you were stuck with the geometry that you had. also you couldn't swap to shells which I think was added in 2012/2013 as well.