4 Replies Latest reply on Jan 2, 2015 12:38 PM by Jared Conway

    simulation, p-adaptive, a-adaptive

    1 1

      stop when RMS von mises stress change is 2% or less.

      update elements with relative strain eneregy error of 2% or more


      What mean above comments?



      loop?   what means?


      target accuracy?


      accuracy bias??


      I can't understand above those in help text.

        • Re: simulation, p-adaptive, a-adaptive
          Seckin Uslu

          These are mathematical methods for convergence.




          H- adaptive about mesh.


          P-adaptive about energy.


          With these method you will see how your works convergence.


          However if you sure your works , you don't need to use these methods.

          • Re: simulation, p-adaptive, a-adaptive
            Mike Pogue
            • H adaptive means to solve the problem at one element size, then solve again at a smaller element size and compare the solutions.
            • P adaptive means to solve at one element order (linear, quadratic, cubic, quartic ...) then the next and compare the solutions
            • A loop is solution.
            • The program will refine the mesh in areas of high stress gradient in order to converge. Converge means to find a solution that does not change with respect to mesh size/order.
            • RMS is root mean squared. It is a measure of the average change in element stress, but corrects the problem that if some go down and some go up, the average could be zero, even though there were big changes. You calculate it by subtracting the stress in each region on this loop from the stress in the same region on the last loop and squaring it. Now do the same for all the regions. Now add them up. Now take the square root. It is identical to the standard deviation assuming the mean is zero.
            • Strain energy is the volume integral of Hooke's law. It's how much energy is stored in your system from the internal spring forces resisting your loads. As a convergence parameter, it has the advantage that it will converge even though there may be singularities in your model, as infinite stresses become concentrated in infinitesimal elements.
            • Target accuracy is how much change in these parameters you are willing to tolerate between loops before calling it good enough. If you run enough loops at high enough order or small enough elements, the change should approach zero in theory. This is the single most important fundamental fact underlying Finite Element Analysis.
            • Accuracy bias, if I remember correctly, is whether you'd rather cheat towards accuracy in local, high stress gradient regions, or total (global) accuracy. As  the program refines the mesh.