7 Replies Latest reply on Apr 17, 2013 11:04 AM by Jared Conway

    Frequency Study - Large Model - "STAR has stopped working" error confusion

    Joseph Sizemore

      Hello all,

       

      First time poster.  Here's my situation: I'm running a frequency study on SolidWorks Simulation and - after days of resolving spring and contact errors - I thought I was ready to finally run the study to completion.  Unfortunately, when I returned about 5 hours later to check the progress, I was met with the "STAR has stopped working" error (and accompanying shutdown of the study).  I am baffled on what the problem could be, since NO information whatsoever is given to the cause of the error.

       

      A little background information: I started this model in SW 2012 (not sure the Service Pack number), only to be met with the "contact elements out of range" error, which I learned was a result of a built-in memory limit to the number of contact elements allowed in a study.  This limitation has been removed in the 2013 version of SW.  So, I was able to convince our IT department to let me install SW2013 with the lastest Service Pack (as of today, April 2013).  The assembly I'm studying is quite large.  My guess is there's at least one hundred parts, several hundred virtual springs and virtual pins, and probably one hundred total contacts sets and/or component contacts.  Unfortunately, I do not know of any major "shortcut" I can take (like halving my model and using symmetry).  My plan is to tentatively use the draft-quality mesh to identify areas of interest and refine the mesh after it runs.  With the draft-quality mesh, I'm sitting around 1.5 million nodes.  Out of habit, I always mesh and save before clicking "Run".  I have been able to  successfully mesh thus far.

       

      My machine specs: 

      Fresh Install of Windows 7 - 64bit

      Dual Xeon W5590 CPUs @ 3.33 GHz

      24GB RAM

      1TB HDD  (I'm NOT doing any kind of network storage for this study, everything is local)

      GPU: NVidia Quadro (I'm not sure where to find the model number, but it's a good card nonetheless)

      ... it's a good machine for this

       

      Just to be clear, I have done some searching and found these posts useful:

      https://forum.solidworks.com/message/103726#103726

      https://forum.solidworks.com/message/244022#244022

      https://forum.solidworks.com/message/246039#246039

      https://forum.solidworks.com/message/184640#184640

      https://forum.solidworks.com/message/314772#314772

       

      Here's a list of what I tried to do to fix my problem, based on the posts I linked above:

      1) Change the location of the results folder (to shorten the file path)

      2) Deactivate the Simulation Add-In at Startup

      3) clean all temp files w/ SW Rx

      4) Restart computer

      5) Reopen model

      6) Reload Simulation Add-In

      7) Duplicate study

      8) Delete old study

      9) Rerun

       

      I pressed "Run" about 30 minutes ago, and it typically takes at least an hour before the STAR box pops up.  I'm going to let this run and check back in the morning to see if anything has happened.  In the meantime, is there something I have overlooked here that would help me better understand the error and (hopefully) overcome the error?

        • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
          Joseph Sizemore

          As I feared, my study failed again.  Same error.  Now I have no idea how to proceed.  Any recommendations?

           

          Edit:  After the crash, the "Do you want to send more information about the problem?" box appears and lists out three files that help describe the problem, stored in a temp directory.  I saved these 3 files in case it would be useful.  Are they useful?

            • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
              Bill McEachern

              Hi joe,

              Nice posts. This problem occurs on ocaission and is typically not that easy to solve. It's going to be work. What I usually do is make a drastic reduction in complexity till I get a successful run and then gradually add to the complexity till it goes south. This will allow to fing the offending specification or component. By reduce complexity I do not mean change the analysis intent rather just reduce the number of components and add fixed restraints to ensure model stability as required. The exclude from analysis command is handy for this. I will also increase element size to get it to run as fast as possible and use the tet 4's. The incompatible mesh option can be handy here. You can limit complexity by bonding up contact pairs once they have been confirmed to work, however this does present some risks. Good luck.

            • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
              Jared Conway

              Bills approach is what we teach in our classes and one on one mentoring and use on tech support. Even working with people on the forum and tech support, I would do that in parallel.

               

              Can you share what it is? I've never worked with that many virtual connectors or contact sets in a frequency study. I've come close in a static study though. The long solver start was confirmed as a function of the contact mapping so that seems reasonable.

               

              I thought it might be a bug I saw about problems over 1 mill dof, but it only occurs if you have less than 16gb of ram. So that can't be it.

               

              I'd get this over to your reseller while you start nailing down what part is causing the issue. And look at updating the model so you can leverage automatic bonding.

              • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
                Joseph Sizemore

                Bill & Jared,

                 

                Thanks for your comments.  I am currently working on a simplified version of my model: I removed 63 parts and all the connectors and contacts that go with them, reducing my # of elements to 1.1M (I thought it would be more of a reduction, but it was not).  So far, no problems, but it's still establishing spring connectors, so I should know something about its success in another hour or so.

                 

                I did identify two parts there were not restrained at all, by fact that they were "hidden", but not "excluded from analysis".  I've been changing things so often, I'm not surprised this popped up.  Still, I should know soon if these parts were the likely problem.  My method has since been to go down my parts list in simulation and check that each part has some contraints on it.  Tedious, but it does help some.

                 

                Edit: well, it looks like I made it beyond the "establishing spring connectors" step, STAR is now doing the "establishing stiffness matrix" part.  So, it seems that could've been part of the problem.  I will probably stop the solver once it reaches the actual solution steps and try going back to the big model version, since this simplified run won't tell me anything realistic.  If those two parts are/were the problem, I might have found the fix!  Will report back later today.

                 

                Edit: Update.  I excluded the two "floating" parts I found earlier from the big study and STAR finished establishing the springs and presumably the stiffness matrix, but ultimately failed with the same "STAR has stopped working" error.  Typically, STAR displays an error box as does Windows, but this time only Windows displayed its generic error message.  Therefore, I have reverted to my simplified model again and am going to let this one run to completion to isolate a group of responsible parts, hopefully.

                • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
                  Joseph Sizemore

                  Update: I am still unable to complete my study simulation.  While fixing the aforementioned errors, today when I run my model with the Direct Sparse solver w/ tet-4 mesh @ 4.2M DOF I get the following error:

                  "Stop: 10 - The solver encountered numerical difficulties"

                  when the progress bar was displaying: "Element mass calculation" @ 91% (36.4% overall), which seems an odd time to display this message.  At this point, my only option is to click OK, which exits the solver and the study fails.

                   

                  Also, I get the "Out-of-Core solution! Expect Delays" warning in the solver box as well.  I looked up this issue, and it seems to be a memory problem according to the SW help files, yet my machine was only using 19/24 GB RAM maximum @ about 2% CPU usage.  This seems odd to me. 

                   

                  Can anyone explain the above two error messages and what I should do to fix them?  The first error is especially vague.  I'm going to try the FFE iterative solver now.

                    • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
                      Bill McEachern

                      The sparse solver has an "unofficial size limit" and 4.5M DOF is likely too big. I think they are working on it. I would stick to the FFE solver for large problems. Some problems require the sparse solver - hope this is not the case for your model - mixed mesh beams and shells in particular. If you have a large solid element model the FFE solver is the place to be. You may still get an error with it but it will be because there is something a miss in the model. We have done very large solid element models with the FFE solver with contact and all kinds of connectors. You have the option of contracting us to fix up your model but I would suspect you want to sort it out yourself.

                      • Re: Frequency Study - Large Model - "STAR has stopped working" error confusion
                        Jared Conway

                        Not every error/warning message is documented unfortunately. I am not seeing anything for the first 2 but the 3rd seems reasonable based on the DOF. The only solution is going to FFE or bringing the number of DOF down. Windows controls CPU and RAM so you don't have any control over that other than disabling hyperthreading to force cores to be used to their max.

                         

                        At this point, like Bill mentioned, I think we'd have to see what you're working with to make more suggestions.