1 Reply Latest reply on Sep 7, 2017 8:43 AM by Siavash Khajehhasani

    Mesh Independence

    Eddie Chua

      Hi all,

       

      I'm a new user to Solidworks so please forgive me if I'm asking stupid questions here.

       

      Recently I have a chamber model that I would like to simulate pressure, and achieve a deflection study as part of my final year project.

      However along the way I've learnt that a mesh independence study should be conducted before I carry out an accurate simulation.

       

      From what I know (or picked up on youtube, so sorry if I'm missing out on something) is that I should carry out many reiterations of a simulation while slowly moving from a coarse mesh size to a fine one.

      From there I will plot a max stress vs mesh size graph, and eventually the graph should reach a point where decreasing mesh size will not affect max stress too much.

       

      I have tried doing this on my chamber (550x500x20-ish size), going from 64mm element size to 4mm. The results seem to not have achieved the aforementioned point, and so if I'm understanding this, I should go even further and decrease the sizes.

       

      However the problem is that my computer seems to be unable to handle that many elements (it's reaching 7m DOF) and I don't know what I can do to complete my mesh independence study.

       

      So my question is, how am I able to complete this study without a supercomputer, assuming I'm understanding this correctly.

       

      Thank you!

        • Re: Mesh Independence
          Siavash Khajehhasani

          Hi Eddie,

          As you mentioned, grid independence solution check 'must' be achieved prior to any FEA/CFD post-processing. The general trend is to reduce mesh density, more close to the region of interest and less globally, otherwise you have to pay the expensive computational cost, which I guess you reached to that point. In SW Sim FEA you can take advantage of mesh controls to increase mesh density. While doing this, you should pick at least one monitoring parameter in your analysis in order to monitor its deviation compared to previous mesh resolution case. For general mechanical engineering purposes, this deviation should become less than 'about' 8%; for other sensitive applications such as aerospace/military a lower margin is desirable. For FEA analysis, you should also be careful about 'stress singularity', wherein a high stress value appears at a sharp re-entrant and by increasing mesh density in that area it escalates considerably, marches towards infinity and will never converge. If that's the case, you need to get rid of singularity at the beginning using various techniques such as adding a small fillet etc or in worst case scenario, ignore the stress results close to that region.

          Hope it helps,

          - Siavash