3 Replies Latest reply on Dec 11, 2008 12:15 PM by Derek Bishop

    Problems with Simulation 2009 SP1.0

    Derek Bishop
      The following notes were sent through to our VAR. The model comprised an assembly with parts of a vehicle mounted crane. The model was contructed for Simulation as a surface for the beams and solids for brackets and bearing plates. An attempt was made to do this two way - one way was to import the whole model into simulation and then exclude the relevant solids. There were problems hiding and showing the excluded bodies. The other way was to delete the solids from the model in the feature tree.

      FEA was done on the attached assembly using SW Simulation 2009-1. The model was meshed as a combination of shells and solids. Problems encountered were as follows:

      It takes a long time to carry out analysis of the model. Over an hour and a half was spent by the program just trying to solve contacts. Running this model as a complete solid was much quicker but there were concerns about the validity of the results with plates represented with one element across its thickness.

      What is the minimum recommended number of elements across the plate thickness to give sensible results? In the Cosmos courses we were advised to ensure there was a minimum of two elements across the plate thickness. This normally creates a solid mesh so small that it becomes unfeasible.

      The program was unable to hide a solid body after the body was excluded from the analysis. This can lead to confusion and error. In 2008 solids were hidden automatically when a component was defined with a shell mesh. This was preferred.

      There were problems trying to redefine surfaces that were previously defined for mesh control after a part was excluded from the analysis or suppressed. Please investigate and advise if there is a problem here.

      When trying to define shell thicknesses the program failed to correctly identify shell faces of a part that was previously excluded or suppressed. Please investigate and advise if there is a problem here.

      Provide recommendations on the best way to mesh this model?

      When a part in an assembly is suppressed, how does the program handle the previously specified mesh control surfaces and contacts for that part. What needs to be done by the software user when this happens to the mesh controls and contacts previously defined?

      Hardware details are as follows: 8 GB of ram, Machine HP xw6400, graphics card Nvidia FX 1700, processor Core 2 Duo Intel E6850.
        • Problems with Simulation 2009 SP1.0
          Derek Bishop
          And some more:

          When meshing a part as a solid, the mesh would not work. The error was one similar to what you get when the mesher has trouble with difficult geometry. Several other users over here are having te same problm. It is related to the problem people are having with meshing a simple solid which can be fixed by changing the results folder.

          and another:

          When I try to obtain contact forces, the forces are not being displayed.
            • Problems with Simulation 2009 SP1.0
              Derek Bishop
              Got a response from Vince Adams on these queries. I learned a lot by dissecting the model he had set up for the FEA. This was helpful so a thanks to Vince. He gave the okay to post the responses on the forum.

              VA

              Hi Derek, I've had a chance to look over your model. Having modeled similar crane structures, I'd model the whole thing as shells. A slightly more traditional modeling approach might be warranted here to get the best shell representation. I've attached the Jib model as a shell model as I would construct it. Note the use of split lines and "weld" surface to connect the plates to the main structure. For expediency, I've modeled all parts as the same (probably wrong) thickness to show you where I'm going. You can use this if you want but you'll have to update the thicknesses. Remember that the doubler plate at the top (bottom?) needs a doubled thickness value. I'd recommend that you get comfortable with the techniques I used here if these are typical of projects you work on. (Approx modeling time from solid model - 20 minutes. Approx solve time of single weldment - 1 minute.)

              To address your other concerns:


              DB
              It takes a long time to carry out analysis of the model. Over an hour and a half was spent by the program just trying to solve contacts. Running this model as a complete solid was much quicker but there were concerns about the validity of the results with plates represented with one element across its thickness.

              VA
              I would not be comfortable with results from a solid mesh of this system without comparison to shell equivalents in test models.

              DB
              What is the minimum recommended number of elements across the plate thickness to give sensible results? In the Cosmos courses we were advised to ensure there was a minimum of two elements across the plate thickness. This normally creates a solid mesh so small that it becomes unfeasible.

              VA
              There are "rules of thumb" but no good answer to this. If your body is in pure membrane loading, one element can capture it. If severe bending, you might need 2-4 elements. In cases such as this (in-between), your general stiffness may be captured with a low aspect ratio single element/thickness mesh. However, I'd be suspicious of stress results. That doesn't mean they'd be wrong. You just wouldn't know it. A test model in solids and shells on a simpler, yet representative model could provide lots of insight.

              DB
              The program was unable to hide a solid body after the body was excluded from the analysis. This can lead to confusion and error. In 2008 solids were hidden automatically when a component was defined with a shell mesh. This was preferred.

              VA
              We went back and forth on this in Product Definition. Since the surface body is now separate from the solid, they don't necessarily know about each other so we wouldn't know to hide the solid. If you convert all of these solids to Sheet Metal parts, they'll midplane and hide automatically. However, I strongly encourage you to review the techniques I used in the attached model and get comfortable with them for these types of problems. Trust me, in SolidWorks, this goes really FAST (I've done it in several other pre-processors and we are MUCH faster.) If you create your models like this, many of your concerns go away.

              DB
              There were problems trying to redefine surfaces that were previously defined for mesh control after a part was excluded from the analysis or suppressed. Please investigate and advise if there is a problem here. I need more detail (sequential steps maybe?) to duplicate and comment on.

              VA
              When trying to define shell thicknesses the program failed to correctly identify shell faces of a part that was previously excluded or suppressed. Please investigate and advise if there is a problem here. I need more detail (sequential steps maybe?) to duplicate and comment on.

              DB
              Provide recommendations on the best way to mesh this model?

              VA
              Done

              DB
              When a part in an assembly is suppressed, how does the program handle the previously specified mesh control surfaces and contacts for that part. What needs to be done by the software user when this happens to the mesh controls and contacts previously defined?

              VA
              They are suppressed. You need to review the icons for actions taken by SolidWorks based on part states. Is this what you were asking? If not, please clarify...
                • Problems with Simulation 2009 SP1.0
                  Derek Bishop
                  One comment I would make regarding the hiding of solids in the FEA page when the solid is replaced by a shell. In Cosmos 2008, there seemed to be no problems with the program recognising when a shell had replaced a solid. So why should it be hard for the software to recognise when a solid is coincident with a surface and allow the use to hide this when they want?