17 Replies Latest reply on Nov 18, 2009 7:54 AM by Derek Bishop

    Bonded Connections with Shell Elements

      I have been pushing up the size of shell weldments in terms of quantity of parts, and have been running into more and more issues with the bond connection not working.  Should the "global bonded" connections be automatic for any two shells that are contacting edges?  I have given up on that and am mostly defining the bonds one pair at a time but even then see a lot of failures unless they are perfectly connected (no gap at all and identical length edges).  The worst part of this issue is that a failed bond does not give an error during the mesh which makes it very hard to troubleshoot.

       

      Does anyone have some suggestions or can send me to a guide on how to best set up and bond shell elements for larger studies?

       

      Thanks

      Bruce

        • Re: Bonded Connections with Shell Elements
          Derek Bishop

          Sounds to me like you are on the right path here Bruce. I've also struggled with this a lot. In fact I've wasted hours trying to find where bonds are not working. The problem is the same as with most aspects of Simulation - it is just too unpredictable. There needs to be more explicit guidelines on how to get this kind of thing working properly. I put it down to cheap skate software development.

           

          Having got that off my chest, some suggestions that you might find helpful. Try splitting the  terminating surface where two shells intersect. Sometimes I will extend a surface to intersect the adjacent surface and in that way get rid of the gap. If you do this and get a compatible bond you should not need to define a contact at the intersection.

           

          I guess you worked out that you find where the shell bonds failed by checking the dsplacement plot. I've heard it said failed bonds can also be checked usng the buckling plots but have never worked out how. Sometimes I will add a fixed restraint to the shells and release the restrainsts one by one to find the ne that is not bonded and creates instability in the model.

          • Re: Bonded Connections with Shell Elements
            Bill McEachern

            try 2010 it appears to be substantially more robust and works without much model prep on surfaces/shells - I have not done exhaustive work but I ran a couple of models and no splitting, no manual connections and a pretty decent analysis against the fully split all edge lenght match version. Oh and the split tool is so improved as to not be believable - split multile bodies of the intersection variety the way it would work if the developer where an FEA guy - well almost that would be split all no picking or anything. Anyway, much improved, hassel and tedium much lowered - to maybe the point where it doens't bug me - I would need a big model to confirm that though and I have done really big ones with the old stuff - worlds bigest vacuum chamber roughly 40' x 40' x 26' and prismatic - not round.

             

            In prior versions I would recommend the fully split approach where all connections have the same edge length. I would not use any  "bonded connections" accept to different element types like beams and solids to shells. This aproach mostly works unless higher order geometry is involved and there are workarounds for that as well.

             

            Also in 2010 no manual conteact definitions are required for beam to shell or shell to shell - well at least for the few I tried and I am pretty good at breaking the stuff without a lot of effort. It is the most  encouraged I have been in many a release so far - but the jury is still out on the over all state of things.

             

            You can still break it but it doesn't take much to get an acceptable workaround, at least for what I have tested so far.

             

            If you are using earlier version that 2009, bonded connections on shells won't work if the gap is bigger than about the shell thickness definition.

             

            Derek - all is not lost....it looks alot better.

              • Re: Bonded Connections with Shell Elements
                Derek Bishop

                Hello Bill. I spent a lot of time on Beta testing this year. I'm happy to agree that there are some good things that have been introduced in 2010 including the features that reduce the need to define contacts in studies. But I reported a lot of bugs. Around 90 SR's were sent through. Many of the SPR's are still outstanding. This is where SolidWorks falls down. For me the norm on a project is to spend as much time discovering bugs and convincing my local VAR ad SW that they exist as I do getting the job done. I'll start to smile when that ratio changes.


              • Re: Bonded Connections with Shell Elements
                Roland Schwarz
                What kind of joints are you bonding?  Welds?  Glue?
                  • Re: Bonded Connections with Shell Elements

                    Thanks for the replies.  I am using 2009 64 bit right now, and have been scratching my head this am about trying 2010 version.

                    I am simulating welds for the most part as these are 19' to 45' frame weldments mostly comprised of sheet metal parts welded together ranging in parts numbers from 25 to 150 of various thicknesses.  Also note that I am prepping in Solid Edge and sending a parasolid or step file over to SW and completing missed stuff or changes in SW - so they are not actually SW sheet metal files.  Not sure if that matters.

                     

                    I think I will try 2010 version to see what it does - if I understand you correctly Bill then the global contact of bonded should work, even with edges of different lengths?  When they failed in the past I have been spliting the surface and also placing a manual bonded connection, maybe that was not required...

                    If I can prep the geometry to get a proper mesh and not required to be manually bonded, I am all for that so I'll give it a whirl with 2010.

                     

                    Thanks again for the tips - much appreciated.  These groups are great.

                    Bruce

                      • Re: Bonded Connections with Shell Elements
                        Roland Schwarz

                        http://www.cosmosm.com/pages/news/COSMOSCompanion-TipsTricks.html

                        There's some really good info about modelling welds at the COSMOS Companion site.  I just got a copy last week of issues #116 & 122, which are about techniques for analyzing welds and weld fatigue.  Wading through the registration may take a couple attempts! 

                         

                        One of the recommendations is to model welds as line contact and use analytical methods to size welds from there.

                        • Re: Bonded Connections with Shell Elements
                          Bill McEachern

                          As I said earlier when I do an analysis all the edge lengths are made to be the same and coincident. I just use surfaces - no solids in the model so any reference to sheet metal is of no relevance what so ever.

                           

                          In 2010 there is an edge weld connector that is new. I would not go relying on it unless you have tested it to verify its accuracy and under what conditions it provides reliable results. It would have been better if they had just provided the nodal loads (forces) for axial and shear. Then you could design the weld using conventional methods if you know the loads you need to withstand. Of course, you can always make the joint stronger than the parent material, however that may not be practical. Alternatively, you could do the whole thing in solid with a minimum of elements and get the transmission loads at the weld interfaces and use those. They should be reasonable but don't rely on it for stresses.

                            • Re: Bonded Connections with Shell Elements
                              Anthony Botting

                              Bill, et al:

                              the "Edge Weld" connector in 2010 works really well as a connector - it spans gaps very nicely and is very easy to use. However, I have conducted some simple bend, shear, and normal load testing to try and understand the output and was not able to obtain agreement with hand calculations (i.e, the output appears non-sensical altogether). I turned in the test model and got an SPR: 525838, as posted by Himanshu (technician at SW Corp) as he verified the issue. The way I understand it, the connector algorithm is fashioned identically to the methods outlined in Vince Adams' COSMOSCompanion series. The upshot: for now, I would not recommend using the outputs of the connector until the SPR is addressed. - Tony

                              • Re: Bonded Connections with Shell Elements

                                Bill, one more question;

                                Are you using different thicknesses?  Or is the primary structure mostly the same thickness and done with one surface body, or is it many individual surfaces bonded together using automatic global bonded contacts?  Sorry - just trying fully understand your approach.

                                 

                                New constraints or connections always scare me, and that is one thing I don't like about cosmos is the never ending string of changes on the interface, which means I am not always 100% sure of what it is doing without a lot of tests.  Improvements are good and bad sometimes.

                                  • Re: Bonded Connections with Shell Elements
                                    Bill McEachern

                                    Hi Bruce,

                                    In 2009 making things different thicknesses is an issue - more work not impossible. You just can't split a suface and add a doubler by, say,  doubling the thickness of that patch, like a pipe penetration as an example. In 2009 this short coming is addressed. It worked fine in 2008 but when they moved to surface bodies as the specification item they forgot about accomodating doublers - they were thinking, apparently, of only the sheet metal constant thickness type of components. In 2009 you can specify by patch or by surface body. You can have different components as surface and do the analysis in an assembly - this had issues in previous releases as hiding a surface only component would not hide it from the stress plot. I think this works in 2010 - haven't checked in 2009 but it probably does.

                                     

                                    When I do shell meshing I pay almost zero attention to the global bonded contact condition since it is set by default. I just checked and if it isn't set to bonded the shell edges don't bond if it is say set to free, even withthe same edge lengths. So it has to be set to bonded but that is all I rely on when building a model. So my approach is use surfaces, whether they are in the same part or as parts in an assembly, ensure all edge lengths are the same and use only global bonded contact. Thsi provides a decent analysis. On occaision if two cylindrical surfaces have the same edge lengths they don't bond properly. When that happens I cut away the edges and build an incontext surface to span the gap by whatever means is appropriate (extrude/loft sweep or whatever). That fixes most problems when surfaces edges fail to bond properly.

                                     

                                    On your last comment, it is very sad in my view, that they just chuck this stuff out there with no papers outlining what the approach does and how it performs compared to conventional methods be they FEA or hand calcs. They treat it like CAD where no one cares how, say the surface got there as long as it shows up, where in FEA it matter a great deal to understand this level of detail or you just might kill somebody or wreck some expensive piece of kit. The executive level just doesn't appear get ithat distinction. Some of the rank and file guys get it but they are unable to convince their bosses or so it appears as the approach hasen't changed much. It does get frustrating from time to time. On the up side the innovation rate is pretty astounding but I would trade a low innovation rate for well documented code where I get it done with confidence.

                            • Re: Bonded Connections with Shell Elements
                              Roland Schwarz

                              http://www.cosmosm.com/files/COMPANION/116_Welds.pdf

                               

                              Finally found a working link to the COSMOS Companion weld analysis issue.