27 Replies Latest reply on Dec 6, 2017 6:35 PM by Chris Clouser

    Unconstrained bodies (that look constrained)

    Andrew Cuttle

      I seem to be finding all the weird errors this week.

       

      Running an analysis on a shell model I am getting the error for insufficiently constrained bodies.

       

      However everything looks pretty constrained, I have tried this both with the global bonding and local bonding for the problem areas. Neither works.

       

      Checking the model with the unconstrained body checker and also running an analysis with soft springs active shows the same parts are moving freely.

       

      The mesh looks like these parts have continuity on the webs (though not the flanges) but I would have thought that would only be the case if the contacts were working?

       

      Any thoughts? Thanks

       

      Drew

        • Re: Unconstrained bodies (that look constrained)
          Andrew Cuttle

          Update: To make some progress I end up excluding the problem parts and got the analysis working. But it seems like there are other joints that are lacking constraints too?

          Capture14-7.JPG

          Does Simulation not run true multi-body meshing then? Instead when meshes have continuity it's just bonding nodes together?

            • Re: Unconstrained bodies (that look constrained)
              Bill McEachern

              Generally in shell meshing, for things to be "bonded" with compatible mesh the common edge lengths need to be the same length strictly speaking. The program does an OK job when this is not the case but it is a bit hit and miss in my experience. Also, with shells there is only one bonding technique that is automatic and this is the merging of nodes. Anything else requires a manual contact set definition. You can tell if things are going to bond up as the meshes at a joint edge of two surfaces need to be identical - ie they need to share nodes at the joint edge. Any kind of small gap or different edge length can cause nodes to get laid down that are outside the merge distance and then they are not connected.  The node seeding just takes all the surfaces in the model and separately seeds the edges with the default element size node spacing so if the edges are not the same length it is not possible to guarantee that the nodes will line up sufficiently closely  that the edges will merge. Use the split line feature to ensure the same edge length. Anytime you bond a solid to a shell you need a manual contact set between the entities, a beam to a shell requires a manual contact set. A beam to a solid will sometimes bond up, most times but not every time if conditions are a bit ambiguous. I did a presentation at SWX World   that covered most cases back in say 2012 I think. It is probably around some place. It just about put everyone to sleep so no one may remember. If you really want it and can't find it, let me know and I will go digging.

                • Re: Unconstrained bodies (that look constrained)
                  Andrew Cuttle

                  Cheers, I'll have a look. What you're saying sounds like stuff I've tried but let me take a look and get back to you. I wasn't sure about the edge length thing but thought it might be the case. The test models I've run, sometimes it didn't seem to cause issues, other times it did.

                   

                  Yeah, I'm used to setting up constraint types for different elements, simulation doesn't seem to offer very much in the way of information for what it's doing. No idea what pinball radius it's taking for contacts etc.

                   

                  The paper might be interesting but I'll deal with this thing first.

                • Re: Unconstrained bodies (that look constrained)
                  Andrew Cuttle

                  Update:
                  Half fixed the problem.

                  The beams were all individual knitted surface bodies which I thought might be the issue. However unknitting parts is pain so I just saved them out to iges then re-imported them with knitting turned off. This should also be all joined edges sliced to have equal length to the joint, this has fixed most, though not all, of the mesh continuity issues.

                • Re: Unconstrained bodies (that look constrained)
                  Bill McEachern

                  You can tell by looking at the one edge in your second image that it has problems. The nodes don't line up for all elements.

                    • Re: Unconstrained bodies (that look constrained)
                      Andrew Cuttle

                      Aye, I was aware that the top flange didn't have continuity of the mesh but the webs look like they do and that should be enough to stop the bodies from being unconstrained?

                       

                      In my experience with other software they treat multi-body assemblies as a single body when it comes to produce a mesh as otherwise you cannot guarantee the behaviour of the nodes. To me it looks like the compatible mesh function in simulation keeps the bodies separate and then bonds the neighbouring nodes, which is how it can align some of the edge nodes but not all of them, like on the flange.

                        • Re: Unconstrained bodies (that look constrained)
                          Andrew Cuttle

                          It sounds like it is just bonding nodes rather than truly carrying the mesh on between shells:

                          • Re: Unconstrained bodies (that look constrained)
                            Bill McEachern

                            All surfaces are treated as if they are in the same part. When I use shells in SWX I don't both will the sheet metal automatic functions - if I am lazy and just want to see how it goes sometimes but typically this is unsatisfactory. I typically insert a new part and copy the surfaces into it and use the part for the part for all contiguous surface that make up say a welded component. This approach can sometimes, rarely, lead to post processing issues where it is difficult to get a plot of some areas that end up hidden but its pretty rare. Someitme I would export the surfaces as parasolid and then bring them back in but SWX can't seem to write out the most basic surfaces and read them back in without a raft of errors - doesn't matter what format you use. An SPR was issued for this behavior but I have no faith it would get dealt with anytime soon. They don't really care about shell analysis. I surmise they think it is too advanced for their user base. This surface behavior has been going on for years.They seem to be putting all the effort into the V6 thing.

                              • Re: Unconstrained bodies (that look constrained)
                                Andrew Cuttle

                                Hi Bill,
                                Slept on it and no great solutions came to mind but I've got some stuff to follow up.

                                 

                                I don't use any of the automatic function either. There is so much pre-processing needed to tidy the model up for analysis I have been converting everything to surfaces manually and then creating them as shells in simulation. I prefer to do things myself rather than automatically as I like to have the control over what it's doing.

                                 

                                This is also an assembly saved out as a part.

                                 

                                Not sure if any of these are likely to be contributing to my issues but I'd think it shouldn't. I've not had any of these issues before this model.

                                 

                                Might just be me being grumpy and old but that's pretty depressing to hear, shells should be a key part of anyone's FEA process I'd say for the majority of applications (some things will only ever be solid bodies obviously). If it's too advanced for the user base, mmm, think if they cannot run a shell analysis they probably shouldn't be doing any FEA. Seems to be a trend with a lot of software these days to plug FEA to people who don't have the experience - different topic.

                                  • Re: Unconstrained bodies (that look constrained)
                                    Bill McEachern

                                    I too fall into the sometimes grumpy old man category as well. I have made the product definition people at SWX aware of the shell issues years ago - like 4 maybe 5. It just doesn't get on their radar because the only people that bring up the issues are the more serious FEA types and there just are not that many using this code. I got fed up and got ABAQUS. It has its own set of usability issues but once you figure out the secret decoder ring thing it does work. My biggest problem with ABAQUS with shell modeling is that Solidworks can't write out a decent surface file without vast numbers of errors even on simple things like all planer surfaces with nothing but 90° corners. You can test this by building some collection of surfaces in Solidworks - I suggest using a some solid and extracting surfaces and then adding some trim and extend features to make a surface representation that would be suitable for a shell model, save it out as say parasolid (any format is equally bad) and then bring in right back in and run import diagnostics. It will have a ton of surface and gap errors. A bug gets filed and nothing will happen for decades. Just did this again recently.

                                    The real tragedy is the SWX Sim interface has the potential to be most productive interface for FEA model set up  - it is really elegant and easy to figure out. But the dolts at SWX consider themselves a CAD company so they listen to CAD users about FEA and they don't necessarily understand FEA in the way an analyst does. The program is brilliant at solid elements with tet 10's and small displacement contact - at least for model set up and even the solving is pretty decent. The post processing is pretty good but as soon as model size gets big the post processing is agonizingly slow. SWX knows this too. I know people and companies that are figuring out how to get SWX SIm results in to another post processor just to speed this up. In ABAQUS CAE for instance you can pretty easily get the data in but ABAQUS CAE wants integration point data but SWX writes out nodal data so then you are into a whole interpolation thing to try and get the data over which if you are serious has issues that need to be overcome. Can probably be done but it requires figuring out how many integration points are in a swx tet 10 and where they are located and recreating them from the element nodal data and then do the import. I also talked to people familiar with Paraview. They were not optimistic about that being any faster with the SWX data. I suspect that SWX uses the VTK  tool kit( Paraview is built on the VTK tool kit)  for post processing as it comes out of SANDIA and SANDIA and SRAC (where the SWX sim FEA code was born) had a joint program back in the late 1990's to evolve the FFE solver for NL work. So it would seem a reasonable guess that maybe that is where their post processing tools come from but maybe not.  As soon as you get off what the program is really tuned for the trouble starts and by that I mean things are far less robust. Anyway such is life. However, with enough trickery I can usually get a shell model to behave in SWX but its unpredictable how much mucking about one will have to do, so it is very hard to predict how long a job will take. It makes the tool sub optimal for consulting work in my view.

                                      • Re: Unconstrained bodies (that look constrained)
                                        Andrew Cuttle

                                        That's pretty irritating/depressing, I won't hold my breath for great shell modelling any time soon then.

                                         

                                        W.r.t the issues you've had with SW surface exports have you seen Rhino 3D. It's my tool of choice for complex curvature surfaces but often my projects were tying into SW geometry. Rhino had the advantage that it would open native SW files without having to export them.

                              • Re: Unconstrained bodies (that look constrained)
                                Chris Clouser

                                I've done a lot of FEA on complicated weldments similar to what you're showing.

                                 

                                I always just used solid elements.  Mainly to avoid some of the problems you're having.

                                 

                                You have to go through a mesh refinement process to ensure the solution is converging, but I didn't have much problem with this approach.  The solve time is going to go up, but there's always plenty of things to do while you're waiting.  Occasionally there are issues getting a body or two to mesh properly, but that doesn't happen too often.