5 Replies Latest reply on Nov 27, 2007 6:27 PM by genexxer genexxer

    CosmosXpress - Element size's impact on results

    Jessica Preston
      Hi All, I am currently modeling a relatively simple part incosmosxpress, attached. The part is a mounting plate for fourwheels and it is restrained by two bolts. The problem is that whenthe element sizes are changed, the result vary quite significantlyand it doesn't seem to converge on a answer, although there is anobservable trend. The results are in the attached jpeg. With suchvarying results, its hard to tell which element size to use andhard to gain any confidence in the result. Does any one have anysuggestions or ideas?
        • CosmosXpress - Element size's impact on results
          Jessica Preston
          So does any one have any ideas?
          • CosmosXpress - Element size's impact on results
            Valentin Leung
            The CAE software are only accurate as much as the way you set all the parameters. I don't have any rule to set element size but I just have a quick look on the mesh and think about the ratio of the elements against the size of my model.

            I just open your model, 24550N on some tiny holes? it's huge... may I ask you what kind of usage it is used for? I am just wandering the accuracy of the results to your load... can a 65x20mm steel bar sustain this amount of load?

            I think that if you start to have some goofy results like that, you may reduce your load, for example, start with 5% of your load and then increase and build a graph like you did...

            I don't work a lot with steel so I am a little skeptic about the loading, for me it's way too high and the results are inconsistent because of that.

            just my 2cents thought
              • CosmosXpress - Element size's impact on results
                Lin Shaodun
                I think under the force condition you had defined, the model 's stress condition will be no more in linear area.

                ComosXpress will not produce correct result for non-linear study, that's why the element size impact result so much.

                You may want to have ComosWorks Professional , use non-linear study for your model.
                  • CosmosXpress - Element size's impact on results
                    Vince Adams
                    Let's take a step backwards here...

                    First... a linear analysis doesn't know the concept of 'nonlinear' exists so it will converge on the correct linear solution, regardless of load magnitude.

                    Second, the guidelines for meshing vs. accuracy are that decreasing mesh size will increase stress (and accuracy assuming all your other inputs are valid) up to a point. You should see the stresses flatten out. When the stresses don't converge with increasingly smaller elements, you are probably experiencing a singularity.

                    In this model (see attached) the restraint on the split line is creating such a singularity. In this case, the stiffness in the model goes from infinitely stiff to having a finite stiffness at the boundary of the split line. FEA solvers cannot converge on these conditions. However, away from this unrealistic condition, the stresses range from 0-10ksi, all very reasonable stresses for steel and linear for nearly all of them.

                    It is important to look beyond the reported factor of safety in COSMOSXpress to the location of these FOS values. If they appear at a restraint or a sharp inside corner, a singularity is probably to blame. You should consider trying alternate restraints or adding an appropriate fillet to see if the stress goes away. If you are still unsure, talk to your reseller tech support or post some pix here and ask as Jessica did.

                    The fact is that many problems that seem simple enough require 2 or more parts interacting to get appropriate stress and displacement results. Attempting to work around this with restraints in COSMOSXpress actually requires more FEA knowledge than simply sticking all the parts in the model and letting it go in COSMOSWorks. Please don't mistake the simplicity of the tool for simplicity of the problem. While there are many problems that can be solved reasonably in CX, many can't and I'd encourage all CX users to read up on proper FE model set up, even if you don't have access to these techniques, so you can better understand the ramifications of what you are able to do.

                    Vince
                • CosmosXpress - Element size's impact on results
                  genexxer genexxer
                  so then does cx support multibodies?...which then would imply support of multiple material properties. I use multi bodies in cwxap when possible instead of assemblies.