7 Replies Latest reply on Jun 17, 2010 4:40 AM by 1-LB43TI

    Mesh Refinement

    Scott Pratt

      Hello All,

      I wanted to see what people usually do when trying to get their models to mesh. Here is what I do in order:

       

      1. Change the Mesh Density to be the finest setting.

       

      2. Select the "Curvature based mesh" option and change it to the finest setting.

       

      3. Increase the minimum number of elements in a circle.

       

      4. Decrease the element size growth ratio.

       

      Is this what everyone else does? If these don't work, what else should I try? Thoughts? Thanks.

       

      -Scott

        • Re: Mesh Refinement
          Simon Richardson

          Hi Scot-

           

          is this meshing for FEA within SW- such as Cosmos,  or is it for a 3rd party program?

           

          We use a 3rd party program- Moldflow - to analyse our injection molded part designs.

           

          Unless we are targeting a specific feature or area, we find that using the "Coarse" option on the "Save As" STL options from SW is best. If we try to use a "finer" tolerance (= tighter mesh) we get way too much data, and the analysis will take way too long to process, with not much gain in analysis accuarcy.

           

          If you are getting gaps in your mesh, I believe that there are other 3rd party  mesh repair tools available- eg STL Doctor which you can use to tidy up the stl file before you import it into your analysis program-

           

          hope this helps-

          • Re: Mesh Refinement

            Don't change to curvature unless you are having problems or your geometry dictates it.

             

            Add mesh controls in areas of interest, usually where you highest stress occurs but not at singularities causing infinite peak stress.

             

            You don't need to change your global mesh to the finest settings, in fact it is best not to. You want to coarsen the mesh in the areas not of interest so save computational time.

             

            Do a rough error check to see approximately if your mesh refinement is ok. Then when you are happy confirm using a convergence study. I tend to do this manually not using adaptive p or h elements.

             

            As for the growth ratio, I normally use about 1.2 for all my mesh controls.

              • Re: Mesh Refinement
                Scott Pratt

                Hey Stewart,

                I change my settings to the curvature based mesh when the standard mesh results in one or more elements failing. I have attached a picture of what I am analyzing and I am not sure why the standard mesh doesn't work. Even at the finest setting less than half of the members meshed.

                 

                How do I do an error check on the mesh? Also, what do you mean about convergence study? I am new to Simulation and trying to learn my way through it. Thanks.

                 

                -Scott

                  • Re: Mesh Refinement

                    It looks like you are using solids, is that correct?

                     

                    I would use shells for that type of work. You could try beam members but based on your attachments it is not the path I would take.

                     

                    Create a copy, midplane offset all your solid structural members. Then trim / extend to suit. Try and mesh again.

                     

                    If you must use solids, have to tried the failure diagnostics to see what is wrong? Try and isolate the parts that don't mesh and see if you can get them to mesh on their own.

                      • Re: Mesh Refinement
                        Scott Pratt

                        Hey Stewart,

                        Yes, my assembly is made up of solids. I created a weldment for the main truss structure, I have a sheet metal bracket and a machined post for the fixed attachment points, and a solid to represent a simplified mass load. What I am trying to do is apply a wind load to the structure to see how well it holds up and then try to optimize it to reduce costs. Since I could not get this to work, I removed the mass load and the attachment members and analyzed the truss structure treating it as being made out of beams. I got it to work and it seems reasonable (see attached pic).

                         

                        I would like to get it to work with the load masses. I am having issues with analyzing the truss members as beams with the sheet metal and similar solids connected to it. Is there a trick to doing this?

                         

                        What do you mean by using "shells"? Solid members that are constructed using the shelling feature? Or create the members out of surfaces?

                         

                        I have tried to do the failure diagnostics and it is really vague. It pretty much says to reduce the element size or increase the number of elements per circle. Most of the time I do this, it still doesn't fix it.

                         

                        Thanks.

                         

                        -Scott

                          • Re: Mesh Refinement

                            Scott Pratt wrote:

                             

                            What do you mean by using "shells"? Solid members that are constructed using the shelling feature? Or create the members out of surfaces?

                             

                             

                            Yeah, like I said, based on your attachments I would probably use shells not beams. Create members out of surfaces, surfaces are treated as shell elements in the simulation environment.

                             

                            Shell elements would be my recommendation.

                             

                            Scott Pratt wrote:

                             

                            I have tried to do the failure diagnostics and it is really vague. It pretty much says to reduce the element size or increase the number of elements per circle. Most of the time I do this, it still doesn't fix it.

                             

                             

                            That is probably what you need to do though. You will find it may be one small face or slither where the problems are caused (you can check under tools). You may also have an intersection that the mesh generation fails on and you need to reduce the mesh size. Like I said, open the parts that don't mesh by themselves and see if they mesh ok with the same mesh properties. If it does it is something to do with the merging of the mesh between parts. If is doesn't mesh try and reduce the size in areas of fine detail. Once it works apply these settings to that part in you assembly simulation.

                             

                            Based on your model my money would be on an intersection.