9 Replies Latest reply on Nov 24, 2017 4:58 AM by Thierry Ouellet

    ***CRITICAL WARNING*** for NON-TOUCHING FACES bonded contact and SHEET METAL ANALYSIS

    Vincent Shebib Loiselle

      Hi,

       

      I realized that making a bonded contact between non-touching faces or edge using the 3 following techniques could produce wrong result :

       

      1. "Non-touching faces" option in component contact
      2. Using the mid-surfaces that are automatically generated when starting a Sheet metal analysis
      3. Using contact set between distant sets

       

      All those techniques could produce wrong result because the relation written between distant sets is false. This relation is displacement equality and rotation equality between the nodes or elements in the sets. This erroneous relation between distant sets locks the rotation of the model at bonded contact points. Sometimes the impact is low, but something it could be crazy high.

       

      In my static simulation (I can't put it online) the deformation was 18 times underestimate when using the automatic mid-surface shells analysis with sheet metal parts.

       

      I'm joining 3 other cases to show the problem.

       

      It's been a mouth I'm trying to explain the problem to technical support... And it is still going. I think it is a failling bridge issue and it needs immediate care.

       

      I see 2 options :

       

      1. Putting a warning every time the solver is started with a non-touching bonded contact condition. This warning should be linked to a well explained help page that will explain in what kind of study we are likely to have errors and that it is more accurate in every case to use touching extended mid-surface shells or touching offset shells.
      2. Removing all those non-touching bonded contact options from the software, those could create failing bridges.

       

       

      This problem is also related to transmission of shear between distant bonded shells and the SPR #:464938. It's the same wrong relations that creates the problem.

       

      Is someone having more examples or infos on this error ?

      Do you also think SW should have a fast fix to inform Users ?

        • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES and SHELL SHEET METAL ANALYSIS
          Vincent Shebib Loiselle

          The error is also well shown in Solution ID S-070909, but it is only shown for bonded contact sets and for a cases were the influence of gaps is low compared to other problems.

          I can confirm that it is also present in component contact "non-touching faces" option and on automatic mid-surface analisys of shell. In the "C beam in rotation" example of the last post, for a gap of 0.79 mm (shell thichness only), we have a 4.6 time underestimated displacement.  

            • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES and SHELL SHEET METAL ANALYSIS
              Vincent Shebib Loiselle

              Here is another embarrassing case :

              The exercise "Basketball Stand" of lesson 8 of the solidwork training course.

               

              This study uses the “non-touching faces” option of the component contact.

               

              The real max displacement is 50% greater than the converge max displacement value of this study. This is because wrong relations that lock rotation at connections are written between the non-touching sets.

               

               

               

              convergence.jpg

               

              Capture.JPG

               

              Knowing all those cases will give wrong results (from 50% to 1800%), I will never do anymore a simulation including one of the 3 options mention above because I will not know even for small gaps if the answer is 2% or 1800 % offset. Also, I know that the result will converge to an error !

               

              I my opinion, we are near to be good for a class action if solidwork does not react rapidly.

                • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES bonded contact and SHEET METAL ANALYSIS
                  Janko Stellaard

                  Hi Vincent,

                   

                  Before talking about class action it is wise to unravel the problem.

                   

                  I took the basketball stand model with gap, with your setting, and checked the contact visualisation plot, and turned on the option include solver generated contacts.

                   

                  The result on the top part looks like this:

                   

                   

                  This means: the entire top surface of the top beam is bonded to the vertical beam. This will indeed generate a conservative displacement because it is much stiffer than the actual bond (edge-to-face).

                   

                  After that, I opened the global bonded contact, and set the gap distance to the distance between the shell of the vertical beam, and the top beam, which is 3.175mm.

                   

                   

                  After that i checked the contact visualisation plot again, and noticed that the 2 pieces of the vertical beam are not bonded anymore. So I added a manual bonded contact between the two parts:

                   

                   

                  If I now run the model again, the result looks exactly like you expected:

                   

                   

                  So it is not the software, but your incorrect interpretation of the bond between the top beam and the vertical beam. The gap distance is a setting that should be considered with care, because more elements will be bonded together than desired if the selected gap distance is too large.

                   

                  With regards,

                  Janko

                    • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES bonded contact and SHEET METAL ANALYSIS
                      Vincent Shebib Loiselle

                      Hi Janko,

                       

                      Thank you for this awser.

                       

                      Here is 2 presicions for that study :

                       

                      1. The setting of the study Basketball_hoop_stand_ORIGINAL WITH GAP wasn't my settings but the setting of the exercise "Basketball Stand" of lesson 8 of the solidwork training course. I haven't changed it.

                       

                      2. I don't agree that "the entire top surface of the top beam is bonded to the vertical beam"

                       

                      When you  checked the contact visualisation plot, and turned on the option include solver generated contacts you have your image. But to really see the solver generated contact you have to unclic a button to hide the geometry generated contact :

                       

                      solver generated contact.JPG

                       

                      Finaly you get this image, this is the real solver generated contact :

                       

                      solver generated contact 2.JPG

                       

                      When I see this image I don't know that the simulation will be 50 % offset. I even think that it is representing the weld better

                       

                      You have been able to remove the error because you have removed the contact created between 2-3 elements of the top surface of the horizontal beam and vertical beam top edge.

                      solver generated contact 3.JPG

                       

                      The problem I'm talking about will have a more important influence when there is bonded contact in different direction.

                       

                      I don't want to make a class action. I just wanted to make people at solidwork react because I think they are l sending people in the wrong direction to use those non-touching options (without any warning).

                       

                      At least you reconise that  "The gap distance is a setting that should be considered with care" and also that there is an error in the Solidwork exercise 18 of the lesson 8 of the solidwork training course.

                       

                      Please have a look at my other file such as "C beam in rotation".

                       

                      Best regards,

                       

                      Vincent

                    • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES bonded contact and SHEET METAL ANALYSIS
                      Janko Stellaard

                      Well this problem is not one of my priorities. I don't have time to enter a full detailed discussion at the moment.

                       

                      In the case of the basketball stand I did get to the correct solution, so it is not wrong.

                       

                      If you find a problem with the code, the usual procedure is that you contact you VAR. If he can't help you, or the conclusion is that there is a bug, he will send it to SOLIDWORKS so they can fix that problem in a future release. I think you are wasting your time by convincing us of the problem, convince your VAR, so he can send it to SW.

                       

                      It is good to see that you take the discrepancy in the FEA results seriously, and that you try to find the exact (cause of the) problem. On the other hand talking about class action, and the above tone to people who are trying to help finding the cause of the problem does not belong on this forum for professionals in my opinion.

                  • Re: ***CRITICAL WARNING*** for NON-TOUCHING FACES bonded contact and SHEET METAL ANALYSIS
                    Vincent Shebib Loiselle

                    Hi,

                     

                    I would like to explain better what is the source of the error when using non-touching shells for analysis. (Non-touching faces in component contact, sheet metal automatic mid-surface generation and contact set with non-touching sets.)

                     

                    As I said before the relation between node to face or node to node created is wrong for non-touching entities. The created relation is equality of displacement and equality of rotation.

                    1.png

                    Let have a look of the influence of the wrong relations created at distant bonded contact where all contact conditions are in the same direction:

                     

                    2.png

                     

                    When we import this model into SW simulations, the software automatically chooses to mesh in shells and automatically apply non-touching faces bonded contact. Here is what we have:

                     

                    3.png

                     

                    Let’s have a closer look at exaggerated displacement at the end of the beam:

                     

                     

                    4.png<

                     

                    Because of the equality of displacement and rotation of distant nodes, shear stress in the space between entities is not well accounted (the mid-shell has not followed the right path) and this leads to a 65 % less stiff structure for this problem.

                     

                    With the worst 3D mesh we can choose, we have:

                     

                    5.png

                     

                     

                    The exaggerated displacement image gives:

                    6.png

                     

                    Flange mid-plane had followed the right displacement path (shear stress is well accounted), this is the difference. The analysis with non-touching shells will never converge to the right solution.

                     

                     

                     

                    Here is another part of the problem that could lead to even worst results:

                     

                    If there is bonded contact condition between entities that are in different directions and if those entities are close, rotation is restricted by the virtual contact line that can’t rotate. Here is the explanation:

                     

                    7.png

                     

                     

                    8.png

                    9.png

                     

                    Let see an example of this:

                    It’s a bar with 2 equally opposed forces that created a torque. Perpendicular displacement is fixed to have a 2D analysis. There is no other constraint. Soft spring options in ON to be able to solve unstable analysis.

                     

                     

                     

                    10.png

                     

                    Case without bonded contact is working properly: displacement is infinite, but the soft spring option makes the problem solvable (the displacement value is 1730 km).

                    Case with the bonded contact is infinitely stiffer, the displacement is 3.32 mm.

                     

                    Here is why: Virtual contact line can’t rotate because of the wrong relations between distant entities. The material has to deform to be able to rotate.

                     

                    11.png

                     

                     

                    This could cause great offsets in simulation results, even with small gaps. For having a feel of real problems, you can have a look at the first post files (ex. “C beam in rotation”).

                     

                    Vincent