5 Replies Latest reply on Oct 25, 2017 4:12 PM by Bill McEachern

    Numerical Difficulties - Thermal Loads

    Ian Deller
    Ian Deller Oct 24, 2017 4:53 PM

      I am trying to run a steady state thermal analysis on a circuit board. I have...

      - Defined the materials for all components

      - Established contact sets for the thermal resistances between components

      - Established a bonded global contact

      - Defined the temperature of the base plate the model rests on

      - Defined the heat power of the multiple circuit components

      - Meshed the assembly using mesh controls, Jacobian points, & compatible mesh

       

      When I run the study I receive an error saying "The Iterative Solver stopped. Numerical difficulties (Preconditioning failed) Please check thermal loads folder. No results saved."

       

      I noticed that the icon next to one of the heat power loads is black and white while the remaining loads are red. (maybe this is a clue to the issue).

       

      I've read forum posts about numerical difficulties with static loads, but didn't find much about thermal loads. Do you have any recommendations on how to determine which thermal load is causing the issue?

       

      Unfortunately, I am not able to share the files.

       

      -Ian

      • 580 Views
      • Categories: Simulation
      • Tags:
        • Re: Numerical Difficulties - Thermal Loads
          Bill McEachern
          Bill McEachern Oct 24, 2017 5:00 PM (in response to Ian Deller)

          If it was me I would switch to the sparse solver and if the model is large use the draft or first order elements - they are fine for heat transfer. If the sparse craps out then you have a modeling error.

            • Re: Numerical Difficulties - Thermal Loads
              Ian Deller
              Ian Deller Oct 25, 2017 8:05 AM (in response to Bill McEachern)

              Thank you for the speedy response.

               

              I used the sparse solver and the simulation was able to complete, but the results are definitely not what I expected/believe. All of the heat producing elements are in the quadrillion degrees Celsius while other elements are in the negative trillion degrees Celsius. The model does not show any heat transfer from the heat sources to the surrounding mass.

               

              Interesting note: 2 of the heat sources are showing up as a large negative temperature. They are in a group of other identical heat sources. 4 of the 6 heat sources are showing as a large positive temperature, but these 2 are showing as large negative temperatures.

               

              I'm sure I have something set up improperly. Any ideas where I should start to check?

                • Re: Numerical Difficulties - Thermal Loads
                  Bill McEachern
                  Bill McEachern Oct 25, 2017 11:06 AM (in response to Ian Deller)

                  high temperatures usually indicate that things are not touching - no path to remove the heat - so your components are not touching even though you think they are. You can check in the interference detection tool by toggling on the treat coincidence as interference. Fix those and run it again and see what you get.

                    • Re: Numerical Difficulties - Thermal Loads
                      Ian Deller
                      Ian Deller Oct 25, 2017 3:22 PM (in response to Bill McEachern)

                      You hit the nail on the head again.

                       

                      There are a few missing interferences that I will clear up. I'm considering creating a simplified model with cubes instead of detailed chips with all of the connectors modeled. This should make tracking down all the interferences more manageable.

                       

                      Thanks for all your help.

                        • Re: Numerical Difficulties - Thermal Loads
                          Bill McEachern
                          Bill McEachern Oct 25, 2017 4:12 PM (in response to Ian Deller)

                          blocks are the way to go. Export the model as a parasolid, fix all the components and then you can make new configurations of blocks and then they are stay in the same positions. Of course you have to fix all the non coincidences but hey it is sometime easier than dealing with all the feature and mate overhead, especially when you start simplifying. Depends on the model.