6 Replies Latest reply on Feb 1, 2010 2:00 PM by Bill McEachern

    Convergence Criteria and Stepping in a NL Analysis

    Andy Krispie

      Hello,

       

      I'm wondering if it's possible to define convergence criteria in solidworks in a non linear analysis, or if this is something the program automatically decides for itself based on material properties.

       

      Additionally, is it possible to use load-stepping or some sort of deflection controled application of load instead of time-stepping?  For instance, can I have a uniform application of 1kn/step or 1mm-of-deformation/step?

       

      Thanks,

      Andrew

        • Re: Convergence Criteria and Stepping in a NL Analysis
          Bill McEachern
          Yes. Accesst he help file from the NL study's property tab. Pretty decent explanation of what does what.
            • Re: Convergence Criteria and Stepping in a NL Analysis
              Andy Krispie

              Thanks very much for the suggestion Bill - some good stuff in there indeed.

               

              Would I be correct in understanding that  my only option to help encouredge convergence is to change the "singularity elimination factor" from 1 to 0?

                • Re: Convergence Criteria and Stepping in a NL Analysis
                  Bill McEachern

                  Well that is typically a last resort - sort of hail Mary pass.

                   

                  What exactly are you trying to do and what exactly is going wrong with the analys? I take it, the solution is not running to what you think is the end?Often it depends on the physics and the physics can guide you in figuring out what to do. Also there are other numerical procedures that may be more appropriate for what you are tying to do that you maybe unaware of. However, I can't really assist without know more about exactly what you are trying to do.

                    • Re: Convergence Criteria and Stepping in a NL Analysis
                      Andy Krispie

                      I am running a non-linear analysis to examine local buckling.

                       

                      The models run well beyond the defined ultimate material strengths - which is time consuming.  For example, if I use steel with a yield strengh of 350MPa and an ultimate strength of 450MPa, yielding occurs as expected, but the model continues to run to stresses well beyond 450MPa.

                       

                      Perhaps I'm going about it the wrong way, but rather than limiting the applied load, I'd like to be able to apply an arbitrary large load and have the analysis automatically stop and declare failure based on some sort of stress or deformation criteria (say when elements surpass the ultimate material strength).  If not, is there some sort of design check feature I can employ in a non-linear analysis?

                        • Re: Convergence Criteria and Stepping in a NL Analysis
                          Bill McEachern

                          Andy,

                           

                          Are you trying to get the thing to buckle? or are you trying to figure out whether the thing is safe?

                          If you have plasticity turned on you need to have a tangent modulus (defines stiffness post yield) or you need the entire "true" stress strain curve. If you are using steel 200 ksi is a decent number for a tangent modulus - depending of course on what steel. You can figure it out as stress from yield to ultimate over strain from yield to failure. If you have true stress strain curve it would help but it would be a second order effect. The key thing is get those yielded elements to behave like mush (relatively speaking). Next you need to use path/arc length control in the NL solution, otherwise you will never get past the buckling with force control. See the help files for the other settings. It would be easier to help you if stated clearly what your objective is in precise terms. These analysis (post buckling) can be very challenging depending on what you are doing and even impossible in some cases is SWX Sim if it is a shear dominated buckling.

                      • Re: Convergence Criteria and Stepping in a NL Analysis
                        Brian Zias

                        As Bill mentioned the SEF is a last resort.  That's what the book says.  Under the hood I am not sure I understand what that factor actually does.


                        From a convergence standpoint, it's somewhat hard to determine what is not converging.  What error message do you see in the solver status window before it fails?  Those lead to specific suggestions.

                         

                        One thing I have taught is to start simple.  Start with linear material models, bonded contacts, solid mesh.  Then complicate the model one step at a time.  Don't jump into a non-linear contact problem with non-linear material models, shells, and beams.  You get my point, you would have no idea what is causing failure: geometry, material props, contact, solver parameters etc.


                        Regards,

                        Brian