15 Replies Latest reply on Nov 4, 2013 7:50 AM by Wolfgang Grzesik

    Mixes meshing + laminate

    Wolfgang Grzesik

      Hi, I came to the point where I started to model with mixed meshes because in my laminated structure are layers with like 0,2 mm height wich takes a lot of RAM to set 2 elements over the heights.So i created midlayers, defined a thickness and bonded them together. I thought I would save lots of calculating time with it. My problem right now is, it doesn't really solve it. Its standing at 0% after 1 hour of solving. The meshing is a lot faster then before.

      I'm on this model for quite some time hoping to finally solve it soon with satisfying results, so i can start my parameter studies.


      The first time I coulda solve it, the bending was too small and i had like no stress in my solarcells. The top layer is still a solid body, because I thought it might be better for load initiating. The load is initiated within the circles on the top layer. It is quarter symmetry for saving ram.

      Using symmetry I took the edge lines of the midplanes and for the solid elements the surface to mirror, I hope that is right.

      I used some contact sets, where I did bond the layers together and also the first midsurface with the solid top layer and the last layer with the frame. Also I did set up contact sets for the midsurface edges with the frame.


      Mixed meshing is relatively new to me, it would be great, if someone could check my model, maybe even run the simulation and tell me if I have geometrical problems in it or made something wrong. I barely did that much complex simulations before.



      Thanks very much for ideas!


      With special Greetings


        • Re: Mixes meshing + laminate
          Jared Conway

          wolfgang, i'm sure you've made some progress with this since posting


          im seeing the first thing and am not super surprised that the first part of your analysis takes a long time. it is building the connectinos between all of your parts and you have a lot of elements. there are over 2 million elements, 5 million DOF in this problem.


          what are you trying to learn with the layering? the are all bonded together, would it be possible to have an equivalent modulus with orthotropic material properties instead?


          if not, what i would do is start a bit simpler with this problem. get it to run with the bottom most shell and a gravity load. add the next one..etc. what you'll do by doing this is make sure that your problem is setup properly and that you're getting reasonable results before you wait for it to build the full model and run it.


          with that being said, overall the problem looks setup ok except that you might want to double check all your contacts. by adding the components one by one or checking them individually, you'll be sure they are applied correctly. also your symmetry conditions, since you're using shells, you'll need to apply the manually. unless something has changed, you can't use the symmetry bc on shells. they have to be done manually using the reference geometry option. however something in the back of my head is telling me it may work in some situations.


          for speeding up the development of this model and setting up contacts, you could consider using sheetmetal plates. the software will automatically extract the mid-plane shells and create the bonds.

            • Re: Mixes meshing + laminate
              Jared Conway

              also, get started with a coarse draft quality mesh, once you have a running problem you can go back and refine it.

                • Re: Mixes meshing + laminate
                  Wolfgang Grzesik

                  Hi Jared, thanks for your reply Yes i made some progress, created also a low quality mesh and could run it but I'm not satisfied at all with the results I get. Its like a maximum displacement of 80 mm wich is by far too much (should be around 30) also it asks me for using large displacements. Setting the element size smaller, it some time takes some hours until it goes past the shell-to-shell connection step in solving but then it takes like 2 minutes and is solved. I have around 130.000 elements wich is still much because I only created the shells so save solving time.


                  I also created the top plate as a shell too, creating an offset-face with offset of 0 and defining it as plate with a thickness of 3,2mm but as top surface. So i can keep my bcs for setting up forces.


                  The symmetry conditions have been updated since SW'2013. You can now also choose edges beside faces in the symmetry option wich makes it possible to make symmetry for faces too.


                  Actually i want the maximum stress in the most stressed solarcells in this quarter and make a parametric study with the influence of loading and cell stress. So I guess defining a full layer with those material parameters wouldn't lead to the results i want.


                  I have no experiences with sheetmetal by now but I'm going for it.


                  When defining the contacts from glass plate the next layers, I am getting the message, that there is a physical gap between the layers which can lead to incorrect results. Actually that can't be true because I created the layers out of the solid bodies that I used before for my solid model. Could be there a problem with that somehow?


                  If that all doesn't work out, i might go back to solid bodies again since the displacements at that model were nearly the same as the testing values.

                    • Re: Mixes meshing + laminate
                      Jared Conway

                      I suspect that this is a true large deflection problem. If you got that message and didn't choose to rerun with ld on, go back and do that, it may help your results. Results with the option disabled I wouldn't trust. Worst case, run this with nonlinear for a true nonlinear solution.


                      Regarding the shells, having everything as shells I think is appropriate for this problem. As for the gap, it is true there will be one. It can cause issues but in the opposite direction as you are seeing because it should stiffen things with the beams used for bonding. Sheet metal may help here. But worst case, a laminate shell or shells closer together/on top of each could also be tried. For a laminate shell, the individual cells would be challenging to get something that makes sense from a setup and solving perspective. Similar to my suggestion above, setting up a test case to prove out your method with quick solve time would be a good one. Also it helps you quickly troubleshoot issues and determine if it is a software issue or setup or model or study issue.


                      I'll have to check out the symmetry for shells, I think there were special situations where is applied. Also, in your model, some of the shell edges were not selected.

                        • Re: Mixes meshing + laminate
                          Wolfgang Grzesik

                          But from mechanical tests, I did with a real pv-module I got a maximum bending of around 30 mm in the mid of the module thats why I dont trust these 8x mm and actually wonder where they come from. ill try to simplify the mesh and work it out until I am close to the testing results, hopefully to reach them. With the solid-model I've been very close to that but also the mesh been a little coarse with already too much of solving time.


                          Quote: "Also, in your model, some of the shell edges were not selected." Yes i saw that after adding the model. Right now I have the face of the solid frame but also 5 edges. Each for every layer with overall two symmetry-conditions. I guess there is no need in connecting the shell edges touching the frame faces because of the global bonded contact.


                          This model is quite tricky somehow. I'd love to go back to simple models as we used to work with while I did study with UGS NX5-7



                          #### Somehow are none of the forces transfered to the cells in this simulation. First principle stress is zero what actually can't be. The cells are around factor 2 young's modulus of the glass so they should change the maximum bending and be at around 100-130 MPa since they do break in the tests with this load. Maybe I should go back to solid bodies, even I do like shells for this problem.

                          ++ Solving the quarter-symmetric part gives simular results no stress in the cells and too high bending, hopefully this isnt a symmetry-problem.

                            • Re: Mixes meshing + laminate
                              Jared Conway

                              the bad results are most likely due to the large displacement option not being selected or the need for nonlinear analysis. 30mm is a large displacement which invalidates the small displacement assumption of linear static analysis.


                              regarding a simple model, that suggestion is mostly for troubleshooting and learning. even in our consulting work, when we have a new category of problem, we start with a small/simple analysis to make sure we have a good idea of how things are going to work before we work on the customer model. and then we start simple. this is the same technique we teach in our training and mentoring.


                              regarding the stress problem, this is one of those cases where i'd suggest checking your method with a simpler analysis to see that it works. if you're still having trouble with that, post it up and i can take a look. i don't think that it is related to symmetry. i think it is more likely due to a contact assumption or missing contact.

                                • Re: Mixes meshing + laminate
                                  Wolfgang Grzesik

                                  Hi, i tried some further but went first back to 3D-Elements. But after 3 days of solving with not satisfying results I went back to shells wondering why the displacements are ok but there is zero stress in the cells.

                                  I also tried one thing and disabled the symmetry... result was the same. I guess its something with the bonding contacts but I somehow dont find the problem.

                                  Deleting the contact sets should only solve the glass layer, but then there are some instability issues.


                                  Right now im trying to solve them with big displacements but i get the message that the option of large displacements is not working with offset-shells.


                                  I also got one fundamental question: How can I set forces on top or botton of a shell? I always wondered how that works, because the shell is in the mid-position of the thought 3d-shell, how does solidworks know if the force is directed from the middle, top or botton side of the shell?

                                  In my example I want to set the forces on the top of the glass plate. Modeled as a midlayer, that would be my problem.

                      • Re: Mixes meshing + laminate
                        Wolfgang Grzesik

                        I'll try that the next days. It feels like some fundamental problem. Now i remodeled the glass plate and set up the glass plate resulting in a always crashing STAR.EXE when solving.

                        Maybe ill start calculating the glass plate, adding layer per layer.


                        I guess when I'm done with this I feel very comfortable modeling with shell-elements, because I have worked with them too much