20 Replies Latest reply on Apr 5, 2013 3:59 PM by Jared Conway

    Solids vs Shells

    Chris Michalski

      Okay, it might be an easy question, but...

       

      I've got x,y,z data from a measured surface.  I create cross-sections and sweep a surface through them (and thicken it to a solid if need be).  (length and width >20x thickness)

       

      I'm running simulations to look at the stresses of laminating this non-flat structure to a similar flat one. 

       

      Is there any reason not to use shells in a case like this?  I know shells should be faster, but SW kept failing to mesh the shell body assembly  but could manage with the solids. Even with purely planar shells it kept finding reasons to complain.  (I didn't screen capture any of the errors so I'd have to recreate them to be specific).

       

      I'm just wondering if it's a case of operator error (highly likely) or are shells somehow limited that makes them problematic for cases such as this with large displacements normal to the surface? (but not large enough to induce large displacement warnings).

        • Re: Solids vs Shells
          Bill McEachern

          Hi Chris,

          It seems really odd. The way the mesher works is that is meshes the surfaces first, sews them together to bound hte region and blows the elements intot he volume. The solid mesher has to mesh the surface fisrt so the fact that you can't mesh 1 surface of the shell make me lean towards finger trouble. What did you do to try and mesh the surface? What I do is make surfaces that all connetec the right way, delete all the solids from the model, assign the thicknesses, shell formulation, materials and them mesh. You can use the faces of the solids as the shell surfaces but if you are not upof the rules it can lead to issues.

            • Re: Solids vs Shells
              Chris Michalski

              Bill -

               

              once my current batch of flow simulations is completed I'll go back and document the exact problem.  It might not be the meshing itself but the solving that was the problem (poor memory).  The simple models I started experimenting with are all dummy parts so I can post those also.

               

              The problem is that the shells don't start out connected.  Think of it as a stack of parts, each a surface defined at the mid plane.  So once I exclude the solid bodies I have no contact between the surfaces.  So it meshes each one individually without regard for the others.  Then as the pressure on the top and bottom of the stack push them together it has to try to resolve the larger mesh of a thick surface (mm) to the finer mesh of a thin surface (um) in contact.

                • Re: Solids vs Shells
                  Jared Conway

                  Hi Chris, based on your description, it sounds like when you ran with solids, it ran because the default contact was set to global and that applied to the faces that were in contact. When you went with shells, they were disjointed and because you didn't add contact sets, the problem didn't solve. There are some ways to get automatic bonding with solids but it depends on the initial modeling.

                   

                  Some things that would help here:

                  1. a picture of the stack

                  2. how you went to mid-plane shells

                  3. the error messages. this will tell us what is failing.

              • Re: Solids vs Shells
                Chris Michalski

                inexperienced operator error (as expected)

                amazing what a difference the automatic transition makes.  I was using it with solids but somehow omitting it for the attempts with surfaces.

                 

                Surfaces save calculation time here but make the post processing more of a hassle.  I have to run a simulation, then re-orient some of the top/bottom faces and re-run.  Otherwise when you plot on the top surfaces you get a mix of tension and compression and your P1 v P3 surface plots are meaningless. 

                I would think SW would be able to put different color schemes on each side of the surface to show both top and bottom at the same time.

                • Re: Solids vs Shells
                  Chris Michalski

                  Okay, follow up question/information.  This simulation is for laminating a flat sheet of glass to a non-flat sheet of glass with a polymer film in between.  The flat sheet has a round hole in it.

                  I got both the solid and shell simulations to complete.  In looking at the meshes they are quite similar.  The biggest difference is that in the solids mesh the polymer sheet takes into account the fact that one side is against the round hole and the other side is not - in the shells mesh SW does not recognize the contact and make the meshes align here.  Both have 15mm elements with 1.04mm tolerance - 39k nodes, 19k elements for shells and 189k nodes, 123k elements for solids

                   

                  When you compare the two results there is a major difference.  This could easily be that I've got things setup wrong with the shells, if so point it out, I'd be glad to admit a mistake here. 

                  The solids model takes into account the deformation of the repeating long-axis ripples of the surface, while the shells doesn't seem to accurately resolve these. 

                   

                  Are shells just too simplified to accurately deal with deformation perpendicular to their initial state?  Other than the outline of the parts you can't even tell these are the same model.

                  Both images are only of the non-flat part looking at the side touching the laminating film.  The first picture is the solids (18MPa), second is the shells (7.5MPa).

                  B134114034000002 module-700mBar P1 inner -crop.jpgB134114034000002 module-700mBar - surfaces - P1 inner -crop.jpg

                    • Re: Solids vs Shells
                      Jared Conway

                      Hi Chris, got your file downloaded and am working on running the analyses. How long do they take on your computer? My first attempt at using Run all had them all fail on me.

                        • Re: Solids vs Shells
                          Chris Michalski

                          Jared -

                           

                          solids was 6.5 hours, shells 10 minutes, thin shells 16 minutes.  All on an i7 3.2GHz with 12GB of RAM in sw2009

                           

                          that's why I'd really like to use shells, but only if I can figure out how to make them match the resolution of solids.  I'm not sure making the mesh finer would help.

                            • Re: Solids vs Shells
                              Jared Conway

                              woah, for troubleshooting, maybe i'll bring the mesh down. was going to run and then look at your setup, might switch that around. will let you know.

                                • Re: Solids vs Shells
                                  Jared Conway

                                  Chris, you've got soft springs and inertial releif enabled on these studies. Soft springs should really only be used to troubleshoot stability issues. Inertial relief might be ok, but really should only be used if you have zero restraints at all. For example if you wanted to submerge something under water and apply pressure all around.

                                   

                                  I'm guessing the reason you needed to apply them is because only one sheet is being held and the other 2 are unrestrained. While the other 2 are connected with no penetration contact sets to the other (also in the solid there is a small gap between the 2 "top" plates) and there is equal and opposite pressure, the chance of there being a small inbalance in force is high which will cause stability issues in your simulation. This is probably the reason for the long run time as well.

                                   

                                  Could you elaborate on the analysis that you're looking to perform? For a lamination type analysis I think I'd expect a bonding layer between the components and/or a bonded contact but I'm not seeing that. Do they really need to slide with respect to each other? If so, what are the physical restraints and what are you ok with adding as additional restraints to hold things together?

                                    • Re: Solids vs Shells
                                      Jared Conway

                                      Another thought, adding friction for your no penetration contacts MIGHT help, but I'd suggest eliminating the gap first.

                                      • Re: Solids vs Shells
                                        Chris Michalski

                                        Jared -

                                         

                                        thanks for taking a look.  I put in the soft springs because with the pressure on the two plates but one having a hole there is a difference in the area that pressure is applied to (plus with the curvature of the one part there may be a resulting lateral force also).  (And to be quite honest I've come to default to using springs and inertial relief even if not necessary, no real reason, just what I've found to prevent problems). 

                                         

                                        In reality this lamination happens between two rubber bladders.  The easiest way I knew to restrain it overall was simply to pick 3 points on one part and keep them in a plane but otherwise let everything float.  The gap in the solid model is because that B134... plate is warped so mating it to be coincident initially is problematic.

                                         

                                        I could bond the polymer to the flat plate and only have the no penetration between the polymer and the warped plate.  Being inexperienced with mechanical simulations I went the safe route and defined things as no penetration.  Plus because it technically happens under heat and the polymer softens slightly I didn't want to allow the polymer to sustain lateral stresses because it was bonded instead of floating.  I considered adding friction here but I've never used it before and didn't want to open a new can of worms.

                                         

                                        FYI: I reran the shells simulation with the mesh refined to 10mm @ 0.59mm tolerance - 38 minutes, nothing drastically different about the stress plots.

                                          • Re: Solids vs Shells
                                            Jared Conway

                                            i think it was good to check if the element sizes affected the results but i think the big issue is the soft springs and inertial relief. since you know everythign runs with the plate that is fixed start there. the next component add, if you can bond, great, if you can't, go with no penetration with contact. if it won't run, you need to add restraints. one possible set of restraints that might work is symmetry type retraints on 2 edges that meet at a corner. see if that meets your needs. my only concern is the potential for inbalance between the pressures but the only way to over come that would be to create a fixture on one side (potentially elastic to create equal and opposite force) and pressure on the other. once you're good with 2, add the third one with the same process. the one that is warped I would start touching (get as close as you can, one way i've done this in the past is a tangent mate or collision detection). If you have trouble. take a step back and just work with that part and the one next to it fixed in space. you'll learn a lot here about what is needed to get it running.

                                              • Re: Solids vs Shells
                                                Chris Michalski

                                                As I figured, converting the pressures to equal forces on the two sides eliminated the "Large Displacement" warnings.  And I eliminated the soft springs and interial relief.  But it still accentuates the hole in the one plate and misses the local ripples.

                                                I can justify spending time iterating different options though at <10minutes each.  It might be that I'll need a mix of shells and solids to balance accuracy vs time spent.

                                  • Re: Solids vs Shells
                                    Jared Conway

                                    Hi Chris, I got this forum update this morning but I can't find it for some reason.

                                     

                                    As I figured, converting the pressures to equal forces on the two sides eliminated the "Large Displacement" warnings.  And I eliminated the soft springs and interial relief.  But it still accentuates the hole in the one plate and misses the local ripples.

                                    I can justify spending time iterating different options though at <10minutes each.  It might be that I'll need a mix of shells and solids to balance accuracy vs time spent.

                                     

                                    Can you post your latest models so that I can take a look? I'm also not sure i'm following what you're expecting to see vs what you're seeing.

                                      • Re: Solids vs Shells
                                        Chris Michalski

                                        Jared -

                                         

                                        here are the eDrawings results of the solid and surfaces.

                                         

                                        Both are constrained at the same 3 corners of the flat backglass.  Because the mid-plane surfaces don't initially touch I had to add some fixtures to the middle polymer sheet, otherwise fixtures are the same.  In this data both use models apply pressure (not update to equal forces).

                                         

                                        From my understanding, SW calculates deformation and uses that to solve the stresses.  If I go to the beginning information then, and compare the displacement in the Y (perpendicular to the sheets), my two simulations are quite different.  I realize that potentially strains are better than displacements due to fixturing issues (solid fixutred at a corner where as shell fixtured at the midplane of that edge), but those are no better.

                                          • Re: Solids vs Shells
                                            Jared Conway

                                            looks like there is still a setup difference. i think you're going to have to use a combo of shells and solids here because it sounds like you want the top and bottom laminates to dip into the hole. That is only going to happen with that part being a solid. That will solve your initial gap problem because you can have them right on top of the middle laminate.

                                              • Re: Solids vs Shells
                                                Chris Michalski

                                                It's not so much a matter of what I want, I simply want the most accurate representation of reality  

                                                I'd like for the warped plate and inner film to not go into the hole (because in reality there is a slug of soft rubber in there that I'm ignoring for simplicity), but if that hole causes a stress riser I'd like to know that. 

                                                 

                                                Thanks for your help Jared.  I think I'll start experimenting next week with combinations - i.e. can I just keep the complex part a solid and the planar parts as shells.  I've got a dedicated system I can set and forget to run simulations so if it takes 5 hours and a solid model to get accurate data I'm fine with that (management on the other hand...).

                                                 

                                                My original intent was to see if anyone else on the forum had done similar lamination simulations to accelerate my learning curve.  So far it seems to be a niche topic.

                                                  • Re: Solids vs Shells
                                                    Jared Conway

                                                    we've done this before but all the laminates have been bonded at some point so there was less issues with getting the BCs to hold everything togethe while you're doing the bonding. In the end you MAY need nonlinear so that you can hold things together temporarily with forces and then release them once everything is put together. There are some interesting tricks like that in the nonlinear training class. Unfortunately since you can't control the "timing" of loads in linear static, you're kind of stuck with an over restraining situation or similar.

                                                      • Re: Solids vs Shells
                                                        Chris Michalski

                                                        I will go back Monday and try bonding the two planar shells and see if that changes anything.  In reality that flat outer sheet isn't as critical.  In this case the flat plate is tempered glass so it can take the stresses, it is the warped glass which is only heat strengthened that I'm most concerned with.

                                                          • Re: Solids vs Shells
                                                            Jared Conway

                                                            might want to consider holding that part in a way that it can still flatten properly and then move the other parts to that part using a displacement/force. might be more stable that way but to get the full analysis you might need to go with nonlinear. good luck Chris! have a great weekend.