7 Replies Latest reply on Mar 31, 2014 11:50 AM by Mark Wylie

    Non-linear static studies with bolted connectors and their reaction forces

    Mark Wylie

      Hi guys,

       

      I am working on quasi-static loading for a bolted electronics chassis. There are mounting bolts (QTY 6) and there are the bolts which hold the chassis together.

       

      I am performing force balance checks under 1 G but not getting the reaction forces I would expect and wondering if anyone has had similiar issues. I am using the iterative solver with 1 time step for 1 second. Mesh is very fine. I then apply gravity. My process/results are as follows:

       

      Model 1: Reaction forces on 6 mounting bolts only

      Chassis is globally bonded. 6 bolts are modelled uing the bolted connectors. Non-Penetration contact set up on bolted joint surfaces. 1 G applied in Z.

      Result: Reaction forces are correct on 6 bolts. Forces  = Mass x gravity.

       

      Model 2: Reaction forces on 6 mounting bolts. Bolts holding chassis together (QTY 60) now included.

      Chassis still gloablly bonded but all joint surfaces where bolts are present have non penetration contact set up on bolted joint surfaces. 1 G applied in Z..

      Results: Reaction forces on the 6 bolts has increased by a factor of two! The mass of the additional bolts is not included so dont understand why I am getting this result.

       

      Is this a convergence error?

       

      Anyone have any idea what might be causing this?

        • Re: Non-linear static studies with bolted connectors and their reaction forces
          Shaun Densberger

          Can you post a model of your system (with 6 bolts and with 60 bolts)?

          • Re: Non-linear static studies with bolted connectors and their reaction forces
            Jared Conway

            do you need this to be done in nonlinear for some specific reason? seems like a pretty straight forward static analysis.

             

            in your second simulation, what happens if you suppress the other 60 bolts? do you go back to the "right" answer?

             

            also when you say "reaction force" what specifically do you mean? axial force in the bolt? do the displacements for your problem look right? it sounds like they really shouldn't change.

              • Re: Non-linear static studies with bolted connectors and their reaction forces
                Mark Wylie

                Hi Jared.

                The linear static analysis worked fine when the gravity was applied in each axis separately but when I combined the XYZ axis loading the reaction forces on the 6 mounting bolts were way off. I believe that the bolts in static don’t handle shear and bending forces well?

                 

                The non-linear model seemed to handle the combined loading. If I supress the 60 bolts my model returns to correct reaction forces on the 6 mounting bolts.

                 

                I have built the model by increasing the number of bolts (10, 20, 30 ,60 bolts) The more bolts that are added the larger the discrepancy. I am now working on increasing the amount of time steps. This is a quasti-static model (static study) so I believe that some sort of pseudo time is employed. I’m not sure how the time steps affects the static study? Can you comment on this?

                 

                When I say reaction force I mean either axial in Z or shear in X or Y when applied in those axis separately. I can also calculate ball park reaction forces on the 6 bolts for combined XYZ by hand under some assumptions.

                 

                Thanks,

                 

                Mark

                  • Re: Non-linear static studies with bolted connectors and their reaction forces
                    Jared Conway

                    i think it would be worth putting together a test model that you can post here and also post some hard numbers. something isn't lining up. i have a feeling it has something to do with what you're reading and where you're reading it vs the simulation outputting something incorrectly. i'm also not sure if your bolts are all defined properly, are you using preload and you haven't answered if the displacements look reasonable. and are the reaction forces at your fixtures equal to your loads?

                     

                    as for some of your other questions:

                    bolts in shear/bending in static vs nonlinear, i wouldn't expect them to be handled any differently or come out with different results.

                     

                    time stepping is used in nonlinear, in static analyiss everything is applied in 1 shot. as noted above, for this problem, i don't see any need for a nonlinear analysis because they are both static and you don't have any nonlinearities as far as i can tell that need to be handled with a nonlinear analysis.

                      • Re: Non-linear static studies with bolted connectors and their reaction forces
                        Mark Wylie

                        Hi Jared,

                         

                        I will work on getting something together but due to company regulation I cannot post the design to a public forum. It may take a few days.I am reading the connector reaction forces for the final time step. I am not using preload in the study because I want the applied force on the bolt for my margins of safety calculations.

                         

                         

                        The displacements on the metalwork is consistent with what you would expect and match the static study of the model, its just the bolts.

                         

                        This is a statement from my VAR with regards to linear versus non-linear "The non-linear study would be more accurate for bolt connectors especially if there was significant bending and shearing".

                         

                        I have made some progress since your last post. for a Quasi-static load case with bolts I found that:

                         

                        Linear model works fine for load applications in plane axis. X, Y, Z but when combined the results (force) on the bolts blow up. Also, this is not quasi-static loading because it is time independent. I need time dependency but slow enough to dump inertial effects.

                        Non-linear model using a direct sparse solver with all bolts in place gave results approx 50 % accurate

                        Non-linear model using FFEPlus gave better results (same model)

                        The more bolts and contact sets I had the larger the error in the bolts (convergence error? tried to mesh this out but no change)

                        The quasi static model was sensitive to time steps. I can hone in on the correct (or close to) reaction forces by adjusting the time steps.

                         

                         

                        My VAR also suggested now using "allow penetration" contacts as opposed to "non-penetration" contacts for bolted surfaces (i.e if a plate is bolted together then the contacting surfaces should be "allow penetration". Can you confirm this please?

                         

                        Thanks in advance Jared and sorry for not being able to provide hard figures!

                          • Re: Non-linear static studies with bolted connectors and their reaction forces
                            Jared Conway

                            bolts more accurate in nonlinear, not sure about that one. its possible. in general the world is nonlinear so nonlinear is closer than linear.

                             

                            allow penetration, don't know why that would be used for a bolted connection. that means there is no interaction at the bearing surface.

                             

                            adjusting time steps, what happens if you use the automatic time step selection (developer recommendation)

                              • Re: Non-linear static studies with bolted connectors and their reaction forces
                                Mark Wylie

                                Hi Jared.

                                 

                                Yes, I agree, allowing penetration should not be used. I have come to a work around to achieve the solution.

                                 

                                My model works correctly when the g check in X,Y and Z seperately. My X,Y and Z components all add up (i.e. the Z- components in the X direction are consistent with the applied G through the COG and the moments created about it).

                                 

                                If I copy the model and apply in XYZ simultaneously the model does not add up. Its 30 % lower reaction forces in axial but shear are accurate.

                                 

                                Let me work on creating a similiar study which I can post. Thanks again for your fast responses.

                                 

                                Best regards,

                                 

                                Mark