9 Replies Latest reply on Jun 6, 2009 1:05 PM by Derek Bishop

    Beam simulation is so limited and buggy it's unusable

    Billy Wight
      SolidWorks 2009, Service Pack 2.0
      Dell M6300 Workstation Laptop
      Windows XP 32bit
      2.8 GHz Dual Core (Intel Core2 Extreme)
      8 Gb RAM (I have dual boot for 64bit, that's why)
      512 Mb NVIDIA Quadro FX 3600M

      Does anyone actually use the beam simulation for a mildly complex model with any success? I am modeling a spaceframe (vehicle chassis) and can only get the model to successfully (if you can call it that) run with all joints defined as rigid and all members defined as beams (trusses aren't working).

      The ideal situation is to model the chassis as steel beams of their respective cross-sectional properties and the suspension arms as rigid elements with their joints as hinges (to better simulate how the loads are fed into and reacted by the chassis - the suspension members are connected through spherical bearings and cannot transmit a moment, only forces). Then apply an enforced 1 degree displacement at the front suspension with the rear suspension fixed. Query the results for reaction forces at the rear SPC's and determine torsional stiffness. HOWEVER:

      1) There is no option for applying a remote load, rigid link, or anything similar when using beam elements. My original work-around was to define a new "Rigid" material with a really high modulus, however there is a limit on how high you can make the modulus (and it's not as high as I would like it to be)! My second work-around was to model the suspension as solid members of appropriate diameters and model the torsion beam (the component applying the torsional moment to the chassis) as a 4" diameter beam to minimize its influence (deflection) on the model.

      2) I can model the suspension as rigid if I want to mesh those components as solid (i.e. treat solid as rigid), but that's just ridiculous.

      3) Changing joint definitions to hinges does not work, I end up getting an error saying the stiffness matrix diagonal coefficient is zero at equation 5902. (though my definitions of the hinges should not cause this). I'm looking to have these joints transmit forces, but no moments to the members they are connected to.

      4) Changing above members to truss members rather than beams also does not work, I end up getting some other error that is unexplained (it just says the analysis has failed)

      5) When going through the joints and finding ones to delete (members close enough to one another within tolerance to create a joint there, but not meant to actually have a join there) why can't you select the joint from the graphics window? I have to go through the list one by one, looking at the graphics until the joint I want to delete is highlighted. This is no trivial matter when there's 170 joints and I want to delete 6 of them!

      6) (Somewhat unrelated to the topic, but something worth mentioning) The software doesn't recognize all the structural members in the model! I had to insert a delete bodies feature in the tree and re-create those members again, then it worked. Not a terrible workaround, but it's not something I should have to do...

      7) I've tried deleting the study and starting over, saving as and starting over, etc. all with no success.

      I am lost as to what to do at this point, the only reasonable option now it to give up on Cosmos and use a proper finite element program (I also use Altair HyperWorks). I'm already well over budget on this project time-wise, and the transition to the other program certainly won't help that!

      It seems quite ridiculous that I am being a beta tester for RELEASED software that is already on Service Pack 2! I don't have the time to do this when customers are constantly asking when the project will be done, and it's not an acceptable to answer them with "This software is just too limited and so buggy that it will take longer than expected". It's even more ridiculous that I'm paying in excess of $2000/year for this privilege!

      And one more rant: Why does SolidWorks insist on making Cosmos more user friendly for the non-engineer/analyst? (I know the answer is purely to increase sales) It is so bad at this point (SW 2009), that it is harder for the real engineer to actually understand what is going on - the real engineering is hidden behind so much fluff to make it easier for the non-engineer! This type of analysis is not for non-engineers, if you don't have enough understanding of the theory to be doing the analysis, you shouldn't be doing the analysis!!! Just because you can get pretty, colorful results plots doesn't mean your model (mesh, boundary conditions, assumptions, etc.) is correct and this can have disastrous results...

        • Beam simulation is so limited and buggy it's unusable
          Hi Billy,

          I feel your pain. Not that it helps, but if you think 2009 is bad for beam element based analysis, you should have tried 2008 - it was worse. What I find really strange is that I have an old Windows 3.1 program that fits on a couple of floppies that does beam element static analysis faster and more flexibly than simulation can. (Unfortunately, it is incapable of frequency or buckling analysis. It is also very awkward to use for complex geometry as it has a wireframe graphical representation (neutral axes only), and awful zooming, rotation and panning controls. It gets very difficult to visualize, edit, and even generate complex geometry in the first place). I would be so happy if SW could have just duplicated the analysis functionality of this program. I have submitted a lot of enhancement requests in this regard and am hoping for the best. My comments to your points:

          1) Another work around is to bond beam joints to small shell or solid plates and use available connector options to connect the shells or solids. This is awkward and has an additional limitation - free body forces are unavailable in mixed mesh studies with beam elements!

          2) I think making a super-rigid material for the members you need to "rigidize" is the official work around.

          3) I have had the same problem.

          4) Ditto. Even when the structure I was modelling/analyzing was a truss, setting even one or two elements to be truss elements rather than beam elements caused the analysis to fail.

          5) I can select joints from the graphics menu, highlight members connected to them, and disconnect members I don't want connected. However, it is very laborious. I find that I often get really high stresses in these little areas with a whole bunch of short beams interconnected in weird ways when they are either not supposed to connect or supposed to connect differently. Sometimes you can get everything to connect with one joint by setting the tolerance high enough, and then other stuff gets disconnected!

          6) I haven't noticed this problem.

          7) FWIW, I have found that while you can apparently continue to use an existing study after changing geometry or section sizes, things tend to work more reliably if I create a new study and drag whatever loads and constraints it will allow me to from my old study to the new study. One problem with this is that anything attached to joints (loads or restraints) is generally difficult if not impossible to transfer without manually re-selecting the joints (even though some part of the software knows where the joint was and applies the symbol there).

          I can really see where you're coming from with the rants. What I have been doing lately is submitting lots of questions and problems I encounter to my reseller. So far I am attached to 13 SPRs for 2009 SP 2, a few of which were existing, and all of which are related to beam elements and weldments. While this is time consuming, it is worth while; 3 of my SPRs are apparently fixed in 2009 SP 3.0 (and 1 in SP 4.0).

          In answer to your original question, yes I have used Simulation with success on complex models (models with more beam elements than the one you show in your attachment).

          Hope this helps,

            • Beam simulation is so limited and buggy it's unusable
              Billy Wight

              Stephen, one of the product managers from the SolidWorks Simulation group emailed me to work out some of these problems. I took some time to better document the issues and sent him the models to have a look at (and in doing so I couldn't repeat one of the issues I had found, and found a couple of mistakes of my own). He was quite helpful, and after a few emails back and forth, he came up with quite a few SPR's (7+) and enhancement requests (9+). Hopefully most of the issues will be resolved in the future releases.
                • Beam simulation is so limited and buggy it's unusable
                  Derek Bishop
                  So your initial comment was correct. Simulation 2009 was so full of bugs that in my opinion it should not have been released for use in the market place. As wel as being frustrating it can lead to potentially dangerous designs. 7 SPR's is an excessive number. I came up with 4 or 5 on other issues before Christmas 2008. How much time did you and I spend identifying these things. Pretty hard to convince a client that he should be charged for this time and it really isn't fair to charge them for the time. So the cost is being born by the users of the software or by business. I don't mind one or two bugs in unusual situations but these are bugs affecting some of the most basic aspects of operation of the software. It leads me to conclude there is an inadequate amount of testing of the software before it is being released. I would like to hear the quality / testing program run by Solidworks for the FEA software.
                    • Beam simulation is so limited and buggy it's unusable
                      Valery Rutkovskiy

                      You are 100% correct.
                      For your structure very nice and proven tool Ansys Workbench.
                      You can make even automatic design optimisation!

                      If you cannot put your hands on - then simulate your structure in "Simulation" using thick shell elements. If you will go by this way check very carefully tubes connections.
                        • Beam simulation is so limited and buggy it's unusable
                          Billy Wight
                          I ended up solving this problem in Altair HyperWorks. It took a bit more setup time than SW Simulation, but I know exactly what's going on as I define everything (there's a lot more user control). Since this project I have met up with the Simulation developers working on the 2011 release and mentioned these issues as well as other to them and they know of the various problems and have either solved them for 2010, or working on them so hopefully it will be more reasonable to preform beam simulation in SW in the near future...
                            • Beam simulation is so limited and buggy it's unusable
                              Hello All, I am one year new to Solidworks and I have been throught the ups and downs of learning the software. If some one could help me on this I would be ecstatic.

                              I think there is a huge problem with Beam and Shell/Solid Connections. I have never been able to produce results in beams in a mixed mesh when the cross-sectional face of the beam is connected to a face of a solid or surface. I created a simple assembly to illustrate my problem.

                              I know I can fix one node of the beam and pull on the other. For my application I need to fix one joint of the beam and apply force to the plate it is connected to.

                              1. I have a square base plate....
                              2. I have a C4 x 5.4 Channel mated to the surface of the base plate
                              3. I created a connection between the beam and the surface of the base plate.
                              4. I fixed the beam at the top beam joint and applied a 1000 lb load to the base plate
                              5. The Cross Section of c4 x 5.4 is 1.59in2. So the max axial stress (520 psi) is wrong for starters (stress = 1000 / 1.59 = 629 psi). But the rest of the beam has 0 psi of axial stress in it. This is clearly an error. I have seen this a hundred times before on my models - and I have been forced to find some other way to go about it - but it is really cumbersome and does not give me a lot of confidence in the software.
                              6. Has there been any updates on connecting beam face cross sections to faces or surfaces?

                              Thanks in advance, Justin
                                • Beam simulation is so limited and buggy it's unusable
                                  Derek Bishop
                                  Hello Justin. I fiddled around with this a lot back here.


                                  I had another try and couldn't even get the beam end to mesh with either a solid or shell plate let alone get results. The problem is in defining contacts and when I run the analysis a warning comes up saying that the model needs more restraints.

                                  I tried to define the beam to plate contacts a number of ways including

                                  Beam body to plate surface
                                  Beam node to plate surface
                                  Beam end surface to plate surface
                                  Using component contacts

                                  I did this modelling the plate as a body and a shell.

                                  I'd be interested to hear how you got the two to mesh.

                                  For many structures you may be able to model it without plates. In my game that is common. Restraints are only applied to the nodes.
                                    • Beam simulation is so limited and buggy it's unusable

                                      I am running with 2009 SP 3.0. Ive learned that the version and service pack GREATLY change the way things run in the simulation world... I would upgrade to 2009 SP 3.0 if you already haven't.

                                      I have completed some welded skids beams and shell parts with what I think is success - but shame on you if you try and create a contact set between the beam and a solid/shell when the beam interfaces the shell by its cross sectional face/end. It gives ridiculous/unusable results. Please let there be someone in SolidWorks land working on this!

                                      Oh yes, the beloved restraints message!! This has absolutely nothing to do with restraints. Do you have a refined mesh control in your mixed-mesh model? THis is usually the issue when I run into that. Either delete the mesh control- or create a coarser mesh and try again.

                                      If that doesn't work - make sure all of your beams that contact a solid/shell part have a clearly defined contact set. When you create a contact set make sure that (1) you select the beam icon (2) select the beam in the top box and (3) select the solid/shell contact face
                                        • Beam simulation is so limited and buggy it's unusable
                                          Derek Bishop
                                          Thanks for that. As I mentioned I don't usually model connections to steelwork structures. It happened to be one of the first examples I tried and failed when I heard you could do mixed meshing.

                                          I do have a degree of success modelling bins and hoppers though with beam stiffeners. I define the contacts in the same way you suggested. I am impressed with SW Simulation capability in this regard.