9 Replies Latest reply on Dec 8, 2009 1:18 AM by Mauricio Martinez-Saez

    Simulation, fail to mesh one weld on a shet metal part having two welded corners.

    Mauricio Martinez-Saez

      Have a problem with a sheet metal part which have two welded corners... when generating the mesh for the part, only the second weld is recongnized by the software which only mesh that weld leaving the other "open", therefore the analysis of the part do not produce a valid result.  Any one have this problem before and know how to resolve it?  apparently are some bugs on the sheet metal corner welds..., how I can simulate them in the anlaysis (if the corner welds are not reconized by the mesh),  a bonded contact is not an option since there is a gap betwen the flanges and both are on the same part.


      I can model the part as a solid (no sheet metal) but them the software takes for ever to generate the mesh and run the analysis (it is a very large part of 0.1046" thickness with over 3500 in2 of area), and the assembly have 16 of those parts.


      We are using SW 2009 sp 5.0 with Simulation Professional.

        • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
          Bill McEachern

          As hard as it maybe to believe, it is suppose to work this way. Similar things happen in many places in simulation as designers design the geometry for the right before you weld it state and what is required for simulation is the after you have welded it state. If you keep this in mind you will avoid similar issues in other areas. All other FEA tools have the same issue if that is what you give it.


          Just make a new offset surface and add in the missing surface piece and then just carry on using this as the surface geometry to represent the part and disable the solid, etc. It is explicitly covered in the simulation essentials training course. If in your case the one corner actually was connected then you can break it by adding a mesh control so that adequate resolution will force the mesh to treat it as you have given it -  in other words you got lucky there and the mesh paved over the hole/gap. Not something to be relied upon.

            • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
              Mauricio Martinez-Saez



              I do not understand exactly what you are trying to explain to me.


              I am using SW 2009 sp 5.0, in combination with Simulation Professional (COSMOS Advance Professional),  on this version of SW, there is a Sheet Metal Feature that produce a Weld to joint corners (creating a solid body of the weld), on Simulation, the software automatically recognizes the nature of the part and automatically applies either solid mesh or a mid surface mesh (for sheet metal parts), o a mixed mesh, if the software believe that to be the best way.  This is done automatically (in fact the user can not select a solid mesh for a sheet metal part).


              Well the problem is that I have a part which have two "corner welds"  created using the "corner "weld" feature in SW, and when creating a mesh in Simulation the software only recognizes the second weld, creating a one close corner and leaving the other open, on parts having a single weld it works OK, but as you have more that one weld, the software "do not see" some of the welds.


              If the part is converted to a parasolid and then loaded into Simulation, it mesh perfectly... but I can not use a parasolid, since I am using the "Optimization" module and I need variables controlling the geometry so simulation can run simulations to "optimize" the design.


              On older versions, you use the shell mesh and the surfaces, etc.  as apparently you describe (sheet metal "Corner Weld" feature do not exist before), but in this version, simulation should recognize the corner welds features (it recognize some, so why do not recognize all).


              I believe there is a "bug" either on the SW sheet metal feature or in SW Simulation. If some other people have seen this behavior then is a "bug"

                  • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
                    Mauricio Martinez-Saez

                    Bill, I agree with you that doing the part on "after welding state" or addign a surface on the "gap" as you explain will resolve the issue, that is the way we do it before and woks perfectly.


                    The issue here is that now SW have a new feature to create "corner welds" and when we have a single corner weld on a part, Simulation recognize that weld, and create a mesh of the part including the mesh (which sometimes is recognized as a "solid" during the mesh creation process, and some times as a surface, apparently, depending on the size of the weld), therefore, Simulation recognize the "corner weld",  the problem is why it recognize only one and fail to mesh any other welds (if you have more that one).  I do not believe that is the propper behavior... if the software recognize one weld is supposed to recognize any other welds on the part.  On the other hand, if it is not supposed to recognize the welds and require us to do the model as before, then it is not supposed to recognize any welds.  Is this logic?


                    When SW add a Corner Weld to the part in fact it is adding a "solid" inside the part.  In my oppinion the "corner weld" feature in the sheet metal module of SW is another "incomplete" feature, since it do not allow to define any material for the weld nor any other properties, nor provide any valid information, such as weigth of welding material, number and length of welds, etc, which can be used on BOM's, cost analisis, etc, in fact the only use it have is more "cosmetic" that functional, however if it is recognized by simulation as a "close" gap (even with the same material properties) it will serve the purpose of closing the gap between the surfaces.  In this case we do not want to analyze the structural properties of the weld area, we just want to analyze the part.


                    Try to analyze a part with a single "corner weld" and you will see it work, then add anoter weld and you will see that the mesh do not recognize one of the welds (sometimes after adding the second one, it fails to recognize the first).

              • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
                Roland Schwarz

                In many cases, you don't (or shouldn't) model the actual welds.  You're better off modelling the welds as lines and determining the forces on the welds from there.


                See the following by Vince Adams re: weld analysis.


                  • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
                    Mauricio Martinez-Saez



                    Let me clarify that I am not trying to analyze the "weld area" nor the structural integrity of the "weld" (that is a very complex process that we know how to do and which we do with another application dedicated to that type of analysis or directly by testing samples on the lab).  Also, let me say that we know how to perform analysis of sheet metal parts using mid thickness surfaces, etc. (that is how we do it for years).  However, in SW version 2009, Simulation Professional (former COSMOS Advanced Professional), automatically recognize sheet metal parts (as well as solids with small thickness and large areas) and "automatically" create mid surfaces and perform shell mesh (without the user having any control of that). Also, in the case of assemblies, Simulation recognize solids and create solid mesh on those, ending with a hybrid mesh.  The problems (which we believe are "bugs" or "incomplete" functionality) are:


                    1.- While SW 2009 Sheet metal module have a "new feature" to produce "corner welds", Simulation will only recognize one "corner weld" (and mesh it correctly), but fails to recognize any additional corner welds (if you have more that one),  forget about the validity of the weld area analysis (we know that any FEA will show stress "singularities" on a weld area, unless the weld is model as a separate solid with the proper material properties and including the penetration area, also with his own properties, etc. a complex process to do with a "general" FEA like simulation),  the problem here is that the software recognize and mesh "only one of the welds" (and some times do this in a random way).  We do not believe that is a correct behavior.  The software should recognize and mesh all corner welds or not recognize any.


                    2.- If inside an assembly you model a sheet metal part (no corner welds this time), and them, inside the same assembly, you create a "mirror" of that part (mirror geometry) and them do an study with Simulation, the first part will be "recognize" to be a sheet metal part and then "automatically" the software will create mid surfaces and mesh that part with a shell mesh.  However, the "mirror" part, will not be recognized as sheet metal (even when in fact it is the same part as the other, only with reversed geometry), so Simulation will process that "mirror" part as a Solid (with a solid mesh),  this on large sheet metal assemblies cause a big problem since Simulation takes for ever to mesh those "mirror" parts recognized as solids and then takes a long time to run the analysis.  As mention before, in Simulation 2009, the software on his own make an "intelligent" decision on what type of mesh to use for every part on the assembly.


                    Do you believe that the two aforementioned behaviors are correct?  We belive that this is something to fix by SW.  We know that we can create surfaces on each part and them run the analysis as we do it for years... but if SW add to the software an "improvement" we believe that the "improved functionality" should work in "real life", not just during a "canned" product presentation when they launch the version.

                      • Re: Simulation, fail to mesh one weld on a shet metal part having two welded corners.
                        Roland Schwarz

                        It does sound as if your problem features are no working as-advertised, therefore are in need of a fix.


                        My "non-answer" is that I haven't used the features that are giving you trouble.  What FEA that I do with weldments, I do without modelling actual welds beyond linear bonded contact.  Since presence of 3D-modelled welds does not help me or yield better results, I don't bother with them.


                        I never fully bought into the "one model for all purposes" paradigm.  When I do an FEA, I copy the model and analyze it.  I keep the FEA copy intact as-is for posterity.  Also, less painful to make little tweaks to analysis model (e.g. split faces to apply force or bond) without disrupting designer's progress or adding superfluous features to designer's model.


                        In your case, if the weld area is not being analyzed, I wouldn't waste time with what should-but-doesn't work and proceed w/ simpler representation.