15 Replies Latest reply on Feb 13, 2014 4:11 AM by Afran Menu

    The rod does not want to fit [Solved]

    Afran Menu

      Hello all,

       

      I'm wondering why this problem doesn't want to converge. I'm trying to find out the force necessary to be applied to lock the barbell. (Please, see the picture).

       

      I'm performing a nonlinear study, but the material remain isotropic, elastic and lineal, and is the same for both workpieces.

       

      For simplicity, I'm running a 2D simplification. The contact pairs are defined as showed in picture.

       

      If instead of applying a force, I solve the problem by applying advanced restrictions and impose the displacement necessary to bring the rod to end position, then the problem converges without mayor problems. The problem in this case is, that I don't trust the results, thus I want to solve this another way.

       

      Looking the problem closely, if I try to solve this as a linear problem, then in the 2D aproximation I obtain a very curious "solution": There is a moment in which the sheet opens too much and loses contact with the bar....

       

      I found a correspondence between both situations: the solution in the non-linear problem stops always when the same deflection point is achieved. So, there is a point where the rod and the sheet lose the contact, which  causes the non convergence of the solution... ?? For me that make no sense, but seems that it is what is happening.

       

      I tried all: restarting the solution varying all posible factors, I redo the contact pair...

       

      Any idea?

       

      Thanks for watching.

        • Re: The rod does not want to fit.
          Jared Conway

          why not cut it down the middle and leverage symmetry to simplify your problem further?

           

          what is the error that you're getting when you try to solve it?

           

          i'd suggest to start a bit simpler

           

          run your analysis without contact to make sure you can get things to move the way you want.

           

          then you can move to adding the contact.

            • Re: The rod does not want to fit.
              Afran Menu

              Hello Jared, Thanks for responding.

               

              The problem get solved quickly enough with the 2D simplification. I know that he problem moves as I want because I already solved it by imposing  the lowest point of the rod to move 22 mm down. Please, click the attached gif to see the full movement.

               

              Now, would like to to solve it again by applying a force.

               

              If I remove the contact restrictions, problem won't to be solved because the rod will move as a free body.

               

              The solution process progresses without mayor problems, but only till it reaches a point, and no matter what modifications I do: the nonlinear solution process progresses stops always at the same point, the point where, in the linear solution, the sheet opens too much and loses contact with the rod.

               

              I hope my description is understandable...

              rod.gif

                • Re: The rod does not want to fit.
                  Jared Conway

                  If I remove the contact restrictions, problem won't to be solved because the rod will move as a free body.

                   

                  >correct, this is a troubleshooting method. to make sure that the full "throw" is acheived.

                   

                  Now, would like to to solve it again by applying a force.

                   

                  >this isn't necessary. just read the reaction force on the displacement. that will tell you how much force it took to create the motion.

                    • Re: The rod does not want to fit.
                      Afran Menu

                      Hi Jared, thanks again for your time,

                       

                      this isn't necessary. just read the reaction force on the displacement. that will tell you how much force it took to create the motion.

                      I am interested in the force and the problem is that I cannot plot force vs time, so I have to review every solution step in order to find the max reaction force in the lowest support (what I find very tedious). Besides, I did it and the results do not convince me (the thickness of the sheet = 1 mm -AISI 304-; rod = 25 mm): max force applied = 3.11 N... about 300 grams seems too little for me.

                       

                      Anyway,  even if the result is correct, I do not consider it unnecessary, just because I want to solve it. Let's say it's now an exercise: something doesn't work and I want to find out what it is. For me that is the best way to learn.

                       

                      Thus: any idea what can I do to achieve the convergence of the problem applying a force??

                       

                      Thanks!!

                • Re: The rod does not want to fit.
                  Shaun Densberger

                  After reading all the posts I have a few comments.

                   

                  1. Solving this with an enforced displacement is the ideal way to do this. Enforced displacement non-linear problems are always more stable when compared to applied force. This is especially true for you model, where you have a "snap-through" effect occurring, which is partly defined by a negative stiffness. This effect can causes serious convergence issues, even if you have an arc-length method being applied in your code. I'm not too familiar with SW non-linear side, but a potential resolution to your model with an applied force could be to increase the number of pseudo time steps, especially around the point that the stiffness goes from positive to zero and negative.
                  2. Can you clarify what you mean by, "try to solve this as a linear problem"? If you're still including contact in your "linear" problem, then did you disable large strain and displacement formulation?
                  3. Even though the problem runs quick enough, it's still probably a good ideal to apply symmetry. If you want to try to get your applied load model to work (by increasing the number of pseudo time steps), then it's worth modifying the model for symmetry. It will also help if your going to try and capture friction in your model.
                    • Re: The rod does not want to fit.
                      Afran Menu

                      Shaun, thank you very much for your great input.

                       

                      I'll try to solve the problem with friction over the night. I'll try it first with enforced displacement: I think it's the best approach because, by applying a force, once exceeded the "midpoint" (let us say, it is the point when the force turns from positive to negative) I think the problem won't to converge no way.

                       

                      Can you clarify what you mean by, "try to solve this as a linear problem"? If you're still including contact in your "linear" problem, then did you disable large strain and displacement formulation?

                      ->I tried to solve with the linear solver, maintaining contact and with the large displacement flag up. Contact and large displacement problems are nonlinear by definition, but I just did the test, and it gave me a clue what was happening.

                       

                      Even though the problem runs quick enough, it's still probably a good ideal to apply symmetry. If you want to try to get your applied load model to work (by increasing the number of pseudo time steps), then it's worth modifying the model for symmetry. It will also help if your going to try and capture friction in your model.

                       

                      Absolutely. With friction the problem will be totally different and every saved resources will be worth.

                       

                      Anyway, It's silly that you can not plot force vs time: In nonlinear problems that goes over 500 steps it is a nonsense to have to examine all solution steps to find what is the max reaction force...

                       

                      Thanks.

                    • Re: The rod does not want to fit.
                      Afran Menu

                      Ok. Solved. I was able to solve the problem by applying a force. Anyway, the solution's convergences seems a little erratic... I must to increase the convergente tolerance from the default value = 0.001 till 0.05, and the solution doesn´t "moves" toward the convergence... I mean with this, that in step X the rod still didn't move and in step X+1 the rod has moved full down.

                       

                      Anyway, it seems like the friction was the key.

                       

                      The solution with an enforced displacement and friction now seems much more coherent also.

                       

                      Thanks to you all.