10 Replies Latest reply on Jul 15, 2014 3:40 PM by Jared Conway

    Unusual High Solution Times

    Robert Fichera

      Hi there,

       

           I am running a study on a rear upright.  The geometry of the rear is almost identical to that of the front, assembly-wise.  The only difference would be the geometry of the model, more external loads, and only one more model included in the simulation.  I had little to no difficulty running the study on the front assembly, but I am having a lot of difficulty doing the same for the rear.  The solution times are either much higher(when the simulation is successful) or the study does not finish entirely.  I cannot understand why.  There is little to no difference in the contact restraints, loads, and fixtures applied to the model.  In fact, I am using a coarse mesh to try and reduce solution times, and this assembly has less than half the number of elements/degrees of freedom than the front assembly but much higher solution times!  I have also carried my study from the ground up, adding components and making changes little by little to isolate what the issue might be.  The only issue I found was using bolted connections.  I have a no pentration component contact between the two faces that are bolted together, yet I still get a prompt informing me that "contact restraints are not defined between bolted connectors".  I manually implemented conact sets rather than a component contact (running out of ideas), and the simulation failed.  I then used real bolts, no penetration component contact, and received a prompt saying "iteration stopped, model may have inadequate fixtures".  Not making sense to me, and I'm running out of ideas...

       

      Any thoughts about why this might happening?

       

      Thanks

        • Re: Unusual High Solution Times
          Shaun Densberger

          The files you attached are just the assemblies; they do not include the required part files. You can use the 'Pack and Go' function to save a zip file that contains the assembly and all associated part files.

            • Re: Unusual High Solution Times
              Robert Fichera

              This is for the front.  I forgot to include the upright in the last message. Suppress the caliper btw, you do not need it for the simulation

                • Re: Unusual High Solution Times
                  Jared Conway

                  if you do a pack and go and put everything in a single file you'll probably have more takers to download it.

                   

                  when you say "long solve times" can you tell us what you feel is long, and how long you want it to be/expect it to be?

                   

                  also can you give us some info about your hardware?

                   

                  and finally, what is the goal of the analysis. using bolts, using contact > they take time to solve. it is part of the game. but maybe there is a bette rapproach to reach your goal. for example submodeling might be an approach.

                    • Re: Unusual High Solution Times
                      Robert Fichera

                      Thanks for your input guys, I do appreciate it.

                       

                      Solving the front assembly took no more 15 to 20 minutes with everything that was included in the attachments.  The rear would take well over half an hour, and would usually get stuck during the iteration process.  After a certain point I would cancel the analsysis, the iteration number would usually be "-2", although I am not sure why that occurs or what it means.  I understand the process takes time, but something isn't right here.

                       

                      The hardware is just something I added to replace bolted connections.  I would like to simulate the presence of bolts (though it's not critical that I do), and at first I used bolted connection above and below.  My results seemed erroneous and I read that bolted connections experiencing compressive loads can produce innacurate results - hense why I switched to real bolts.  Though what shaun suggest does make sense.

                       

                      When you say "submodeling" do you mean breaking the assembly and analyzing the individual components?  If so, then excluding the bolts, there are only two real components to this model.  I feel as though I should easily be able to deliver a valid set of results.

                       

                       

                      Shaun, the reason you cannot use Standard/Counterbore connections is due to the geometry of the model.  the bolt head hole and the hole at the rear face of the upright are not concentric - although the through holes are (look at the model from the side and you'll see what I mean).  I extruded material outward which was concetric with the bolt head hole to alleviate this issue - which it did.  I'm surprised that the model was not in that condition when I sent it to you - I can re-send the file.

                       

                      Shaun, why should you avoid component to component for No Penetration?

                        • Re: Unusual High Solution Times
                          Jared Conway

                          my suggestion, start smaller and iterate to the point where you have a problem so that you can isolate where the problem exists

                           

                          start with one part, gravity load and a fixture

                          mesh, run

                          then add the next part, leave bonded contact

                          mesh run

                          by the end you should have all your parts, if you're using a coarse mesh, this should solve really fast

                          otherwise you probably have a load or geometry condition

                          then you can start adding contact and bonded contact sets

                           

                          as for shaun's recommendation, his recommendation is to minimize component and global contact when it comes to no penetration because it applies everywhere in your model. instead use no penetration contact sets to limit how much ontact exists in your model.

                            • Re: Unusual High Solution Times
                              Robert Fichera

                              Thank you Jared,

                               

                              I have been doing this

                               

                              This was the process I took:

                               

                              Just the upright, iterated each force one at a time; ended with all four forces - solution time no longer than 5 minutes

                              added the lower ball joint house, bonded, changed the faces to apply the load at the LBJH - worked.

                              applied no penetration contact sets to LBJH, added bolted connection; ran the study first without pre load then with pre-load - much higher solution times, failed without preload.

                               

                              I know, or at least I am quite certain, the issue lies here.  This is why I mention it confuses me why I apply the same constraints to the other assembly without any issues, but here the problem exists. 

                                • Re: Unusual High Solution Times
                                  Jared Conway

                                  "the issue lies here" > what do you mean specifically? bolts + contact? this will have an effect on solve time, you've made it inherently nonlinear. some problems are just more complex than others. have you compared the number of elements? have you compared how many elements are incontact? if either of those are different between the 2 studies, then you can't expect the same solve time.

                                   

                                  note, we have analyses that are "simple" that take hours when taking into account the right effects and that have a good mesh.

                                    • Re: Unusual High Solution Times
                                      Jared Conway

                                      what you should be asking is "if i need fast solve time, what can i do to simplfy things to do the design work and then come back and finalize the design with a more complicated analysis with long solve time at the very end"

                                       

                                      this is why i recommended submodeling which is a feature from solidworks simulation 2013/2014 i believe. this lets you analyze a larger structure and transfer those conditions to a smaller section for analyzing just that section.

                        • Re: Unusual High Solution Times
                          Shaun Densberger

                          I was able to get "VU06 Rear Assembly" to run after re-defining the two bolt connections and one of the remote loads (and user defined CS was missing). I did notice the SW wouldn't allow me to use "Standard or Counterbore with Nut" for the bolt definition (gave an error message stating, "Please select a circular edge for the bolt nut hole"); I had to use a "Standard or Counterbore Screw" definition. After making these changes the model ran without a problem, so I can't say exactly what the issue could be. However, here's a couple pointers to keep in mind:

                           

                          1. A 'No Penetration' contact definition results in a non-linear analysis. Non-linear analyses can take much longer to run than linear analyses due to the iterative nature associated with them
                          2. You need to be careful when you have parts that only have 'No Penetration' contact definitions with other components. I'm 99% sure that the issue you ran into when you created real bolts was due to this (probably the bolt moving along its own axis).
                          3. Generally speaking, you want to avoid 'Component-to-Component' for 'No Penetration' contact definitions and instead set your contact pairs manually.