9 Replies Latest reply on Oct 11, 2016 1:27 PM by Shaun Densberger

    Really, REALLY weird stresses in a study

    Daan Haeyen

      Hi all,

       

      I'm running a study on a shaft and crank, connected by a sunk key. Here is how the model is fixed and the load is applied. All connections are set to "no penetration". There are no parts colliding.

      simu1.jpg

      After running the simulation, this is the displacement of the key, on true scale. Weird, I would think it would rotate more because of the modeled clearance.

      simu3.jpg

      Automatic scale displacement shows the key is indeed rotating as is expected:

      simu2.jpg

      BUT stress plots shows maximum stress in a place on the shaft that isn't even touching the key.

      simu4.jpg

      I'm a teacher in Mechanical Engineering and know quite a bit about Solidworks, and I have colleagues who know Simulation better, but we are all flabbergasted by this result. I would be really thankful if someone can come up with a good explanation for this.

       

      I have attached the complete model to this post.

       

      Thanks, Daan

        • Re: Really, REALLY weird stresses in a study
          Bill McEachern

          I would make it size on size and then run it with surface to surface contact. If you really want the slop then use SWX and the move tool with advanced dynamics and get it as close to the likely initial contact state as you can  - like touching everywhere that counts and try that. Edge contact is not well handled by SWX so tif that is the case it could be quite difficult. This a very tricky analysis even if you think it is easy. ABAQUS would have trouble. then again i have seen this code do so really killer things with tricky contact that a lot of other codes would struggle with in the static study. the other things is that maybe you don't have any deformation - like actual scale - set.

            • Re: Really, REALLY weird stresses in a study
              Daan Haeyen

              Hi, thanks. I made them size on size, but still get stresses in illogical places... How can this be?

               

              simu5.jpg

                • Re: Really, REALLY weird stresses in a study
                  James Riddell

                  From failures I have seen IRL that is about where cracks/failure does begin on this type of feature.  The stresses may/may not be correct but it does follow.

                  • Re: Really, REALLY weird stresses in a study
                    Bill McEachern

                    You think they are on the wrong side? They do not look infeasible to me location wise. Plot the contact pressures and see how that looks. I always do this to make sure it looks reasonable. Using surface to surface contact can make a big difference in the reasonableness in the contact pressure results. The other thing is you want make sure you toggle the stress averaging across parts on and off to see what that does. Do you have friction turned on? Was there a shrink fit contact condition anywhere? Maybe it's just the image but they still don't look size on size- the key in the bottom piece looks smaller than the key way.

                      • Re: Really, REALLY weird stresses in a study
                        Daan Haeyen

                        I changed all the hard internal edges to fillets and used mesh control on them. The resulting stress plot show a totally different image, that looks a LOT more logical as you can see below. The maximum stress is on the complete opposite side now!

                        simu7.jpg

                        Getting rid of these singularities seems to have changed the whole outcome of the study.

                         

                        Bill, how do I plot contact pressures? The friction is off and there is no shrink-fit. The stress-averaging on or off doesn't make a difference.

                        simu6.jpg

                        Thanks, Daan

                          • Re: Really, REALLY weird stresses in a study
                            Bill McEachern

                            Make a new stress plot and pick contact pressure to plot. or edit definition on the plot you show and plot the contact pressure (last item in the selection list.

                              • Re: Really, REALLY weird stresses in a study
                                Daan Haeyen

                                I talked to our reseller about this studies, and he told me that with linear, small displacement studies the contact faces are not rotating. That could explain why in study1, the key hasn't rotated and the stresses are off:

                                study1.jpg

                                He told me that in large displacement mode, the faces do rotate. Now, in the second study, I didn't turn on large displacement. I just added fillets. But somehow, the key is now rotating and the stresses are more plausible. My question: what causes this second study to come out right, as I didn't set large displacement mode. Maybe the program somehow used another solver or something. When I turn on large displacement mode, I even get an error...

                                study2.jpg

                                Can anyone take a look at both studies and see what the cause of this difference is? That would be really great!

                                 

                                As study2 is rather large with the results file included, you can download it here. Study1 is attached to this post

                                 

                                Thanks!!! Daan

                                 

                                PS Bill: I made a contact stress plot, but I'm not really shure what I'm supposed to see here.

                                contact stress.jpg

                                  • Re: Really, REALLY weird stresses in a study
                                    Shaun Densberger

                                    "He told me that in large displacement mode, the faces do rotate. Now, in the second study, I didn't turn on large displacement. I just added fillets. But somehow, the key is now rotating and the stresses are more plausible. My question: what causes this second study to come out right, as I didn't set large displacement mode. Maybe the program somehow used another solver or something."

                                     

                                    It could very well be a contact algorithm issue. In your first study, you have a sharp edge that the solver is going to try to prevent form penetrating (and I'm 100% sure, but I think SW is a pure penalty based method). The radius in the second method means you now have a surface. I highly doubt that the program used a different solver between the two cases. Now, even though that second analysis looks more "right", if your rotations are large (> 5 degrees), then you really need to be doing large displacement to capture the second order terms. Large free-body rotations will result in fictitious stresses under small strain/displacement theory.

                                     

                                    "I made a contact stress plot, but I'm not really shure what I'm supposed to see here."

                                     

                                    A contact stress plot shows the surface-normal stresses (think normal-force / area for an infinitesimal material element). For example, assuming a uniform spatial distribution, a 100 lbf weight sitting on a table with a contact area of 1 in^2 will have a contact stress of 100 psi with a unit vector normal to the surface of the table.

                                  • Re: Really, REALLY weird stresses in a study
                                    Daan Haeyen

                                    Hi, I made some new studies about this situation. As you can see below in an ISO-plot there are some peak stresses forming on (relatively) sharp fillets. I wonder, do the spotty stress distributions indicate that this study is not valid? If so, what more can I do?

                                    study11.jpg

                                    Thanks, Daan