5 Replies Latest reply on May 16, 2018 1:11 PM by David McCullough

    Is it possible to run goal optimization without re-meshing between scenarios?

    David McCullough

      I am performing a flow simulation study and would like to run a goal optimization study to determine the location for a component to experience minimum torque. I have the goal optimization set up to what i believe is correct, but the issue is each scenario re-meshes the problem from scratch.

       

      Is there any way to use the mesh generated during the first scenario to solve the remaining scenarios?

       

      The part has some intricate details that require several mesh refinements and a very fine mesh. The cell count is on the order of 5 million. I believe using the same mesh from the first scenario could do two things: 1) it would make the results more comparable and less mesh dependent. 2) Significantly reduce the solve time due to unnecessary re-meshing.

       

      I have tried using the "take previous results" option, but checking this box causes the software to crash for some reason.

       

      Thanks for any suggestions!

        • Re: Is it possible to run goal optimization without re-meshing between scenarios?
          Bill McEachern

          5 million cells sounds crazy for Flow Sim but hey I am not one to judge. If you are coming from another codes I would say you need about 1/10 the number of cells for a similar level of accuracy with flow sim versus an unstructured solver of the body fitted type. I can't see how you can have a geometry change with out remeshing and local remeshing is not an option. Did you make an effort to lower the cell counts with local mesh controls? Are you sure your parameter of interest requires 5 million cells to get to mesh independence or more accurately avoid a bad optimization? Maybe it does but not many problems I have come across require that many.......

          • Re: Is it possible to run goal optimization without re-meshing between scenarios?
            Chris Michalski

            In order to move the component you have to re-mesh.  To change from solid to fluid to eliminate that existing part, and fluid to solid to create the part in a new location requires creating a new mesh.  Flow can't simply convert a cell from solid to fluid or vice versa.  The only way remotely possible would be if you had a mesh control region based on an identical dummy part previously in each possible location such that the mesh was defined as a local fluid that looked like the solid.  And unless you know in advance where you want to experiment with placing that part in the future it's not truly an optimization but just a comparison of multiple choices...  Not possible.

            If you're looking at optimizing the location of this one part, do ALL OF those fine features affect the outcome?  Or can you simplify the mesh around some and not see the difference as far as this part is concerned?

            • Re: Is it possible to run goal optimization without re-meshing between scenarios?
              David McCullough

              So what i am trying to do is determine the center of pressure of a model rocket. I have a sketch with a center line along the axis of the solid body and the only parameter changing is the distance from the base of the model to a point along the center axis. So there is no physical geometry change to the solid body. Only a change of location of an "imaginary" point.

               

              Perhaps there is a better way to accomplish this end goal and perhaps that should be my question.

               

              As for the number of cells, even at 5 million it does not appear to be completely mesh independent as the torque goal value changes a little with every refinement. 5 million is just the limit of RAM my machine has. I could likely work on using local meshes and limiting how much each ares gets refined, but the vast majority of refinement is already at the fins of the model, which i believe would need the most refinement.