22 Replies Latest reply on Jan 21, 2010 12:21 PM by Derek Bishop

    Solidworks vs Roark

    Darrin Dame

      I am trying to verify the accuracy of Solidwork Simulation 2010 by solving an example problem out of Roark's Formulas for Stress & Strain.  The problem is to find the max deflection and stress in a plate that is 0.20 inches thick and 20 inches in diameter.  The plate is simply supported along the edges and as a uniformly distributed load of 3 psi.  Material property given is: E=10(6) and v=0.285.  (This information is needed to create a new material in Simulation.)  I have "fixed" the bottom edge of the plate to simulate the simply supported boundary condition and have tried various meshes - up the the finest mesh - to see if the mesh density was causing my problem.

       

      The deflection and stress I get in Simulation is not close to the correct answer in Roark's (I have tried all of the stress and deflection options).  For the max stress (von Mises) Simulation gives me 8,085.9 psi vs. 9,240 psi in Roark .  For deflection Simulation gives me 3.851e-2 inches vs. -0.0883 inches in Roark.  Any suggestion on how to model this correctly?  I need to verify this simple problem to prove more difficult analysis to be correct.

        • Re: Solidworks vs Roark
          Anthony Botting
          Try using Roark's "simply supported" assumption ("fixed" actually is fixed, making the structure too stiff). It is difficult to do with a full model. Best to cut out 3/4 of the model. Then just put a vertical restraint displacement = 0 on the remaining 1/4 perimeter, and make sure to put symmetry boundary conditions on the cut edges - this will hold the model from experiencing RBM - rigid body mode. I have done alot of these against Roark's and the boundary conditions are key to getting the correct answer. You have to be dead-accurate on the presumed boundary condition or the answer will be way off.
          • Re: Solidworks vs Roark
            Anthony Botting
            please let me know if you need to see a sample 1/4 plate that has the "simply supported" boundary conditions with symmetry bc on the cuts. I have one but it's in 2010 format.
              • Re: Solidworks vs Roark
                Anthony Botting
                Hi Darrin:
                I got von mises stress 9,365 psi (1.3% difference from Roark's). My displacement is 0.251" at center. I don't have my Roark's handy but looked in eFunda for circular plate and it shows 0.261" deflection at center (about about 2.8% difference from eFunda). Note however the plate thickness to diameter ratio starts to violate small displacement theory, so the deflection calculation should be done with nonlinear theory. I can send the model but let me know if you would like it for testing. Note: you should turn-on in-plane stiffening effect anyway (I forgot about that!). I hope this helps.
                  • Re: Solidworks vs Roark

                    I think this analysis should be done with shell elements.  Using:

                     

                    20" diameter disc

                    0.2" thick shells

                    simply supported ("immovable") at the circumference

                    udl = 3 psi

                    E=30x10^6 psi(note: there was a typo here in the original post, but I found the correct number in my copy of Roark)

                    v = 0.285

                    default mesh density

                     

                    I get 9237 psi and 0.088" with "thin" shells checked, and 9242 psi and 0.89" with "thick" shells checked.  I also get 0.088" with Efunda.

                     

                    Regards,

                     

                    Dave.

                  • Re: Solidworks vs Roark
                    Darrin Dame
                    Yes please.  Upload the sample.  I am using 2010 so there will not be a compatibilty issue.  As for your comment on changing the boundry conditions, this was taken from an example problem.  My problem may be on how to model "simply supported" in Simulation.
                  • Re: Solidworks vs Roark
                    Derek Bishop
                    I did an almost identical check in the past and got good correlation of results. Check the way you restrain the plate.
                    • Re: Solidworks vs Roark
                      Anthony Botting
                      Here is the plate. The "Thick" option is available when you "edit definition" on the shell. The results should not matter in this case.The thick option is according to Mindlin Plate Theory to take shearing deformation into account. Since this is a thin-plate, there should be almost no shear in the center of the plate. Also eFunda page is attached. Hope this helps.
                        • Re: Solidworks vs Roark
                          Darrin Dame
                          Ok, I changed my anlysis to a thin shell model and my answers match Roark.  Should you always use shell elements when modeling plates and changed beams to beam elements?  I have a couple of composite plates (1/2 inch thick plates with beams welded to them as stiffeners).
                            • Re: Solidworks vs Roark
                              Anthony Botting
                              Not really. It is a simplification technique that was developed years ago to approximate "thin" structure behavior, due mainly to lack of computational power. A good reason to use them now exists if you believe your product will follow Bernoulli-Euler bending theory for plates, because the shell elements are written for that. Another reason not to use "solid" type elements for thin structures, is the developers recommend a minimum of two (ideally, three) elements through a plate thickness to model the bending behavior correctly. If you tried that with really thin structures, you could easily bring your computer to a halt. There will be huge number of solid elements and equations to solve. Hope that helps identify.
                              • Re: Solidworks vs Roark
                                Brian Zias

                                Darrin, just to add to what everyone else is saying:  to model a simple support it is much easier to make a shell since you can select the edge and do a No Translation.  To keep it a solid, I think you can also create a split line down the mid-thickness, then fix just that EDGE.  In the solid mesh this should allow rotation about those nodes.

                                 

                                Brian

                            • Re: Solidworks vs Roark
                              Joe Galliera

                              It's not always straightforward or easy to translate analytical hand calcs into a model suitable for a 3D (or even 2D) FEA tool.  I understand the underlying reasoning, but in general I suggest not getting bogged down on trying to do this.

                               

                              I submit the attached PDF document that basically will tell you that commercial FEA tools will basically give the same answer within reason.  The difference is in how the program allows one to do things (GUI, meshing, connectors, as examples) and the level of training and understanding of the user on the particular software tool.  Some do things better than others... and the basis of where they came from, how and why they were developed, essentially the history of the company that develops the tool is the root behind why some are better than others.

                                • Re: Solidworks vs Roark
                                  Darrin Dame

                                  I do understand that is not easy to verify theoretical answers with any FEA tool.  What I was trying to prove to myself whether the FEA or the hand calculation is correct.  Since my initial problem was taken directly out of Roark, I was trying to verify how close Simulation would come to the same conclusion.  There is little probability that the two answers will be exact because for the accuracy of a calculator versus the accuracy of a computer.  As an engineer for almost 20 years, I do both hand calculations and FEA calculations to verify a solution or method.  I have been trained in and have used two other FEA programs, so I was determining how Simulation solves the same problem.

                                   

                                  I am now looking to see how Solidwork Simulation handles a model with beam and plate elements (as described above).  I have been reading that this feat is not easily done in Simulation and was asking if there is another discussion thread, or web page, that can help me understand how to analysis this type of model.  I do a lot of analysis on structures that have combination of beam and plate elements.  I will use shell elements when I can do so.   I have modeled each component in a part files and have them joined together in an assembly.  I am analyzing the assembly.

                                   

                                  Any help would be appreciated.

                                    • Re: Solidworks vs Roark
                                      Anthony Botting
                                      I know of the tutorial in the on-line help. There is a framed "building" and is uses shells for the plate glass and beam elements for the beams and columns. Hope that helps.
                                        • Re: Solidworks vs Roark
                                          It has been possible to bond beam elements to shell elements for a release or two, but I usually simulate combined beam/shell models by drawing the beams as shells.  I find this more flexible, as Simulation tends to lose track of anything attached to beam nodes every time you change the beam, causing me to manually have to reattach things (loads, restraints, connections), which is tedious.  I model shells with SolidWorks surfacing commands and use lots of split lines to make sure my shells connect.
                                        • Re: Solidworks vs Roark
                                          Daniel Schulz

                                          "As an engineer for almost 20 years, I do both hand calculations and FEA calculations to verify a solution or method. "

                                           

                                          You are doing what ever good FEA user should do, but almost none do.  When I learned FEA this concept was pounded into us (along with mesh convergence checks).  Almost no one does the convergence checks either.

                                            • Re: Solidworks vs Roark
                                              Joe Galliera

                                              I believe that engineering schools still try to enforce checking analysis with hand calcs, mesh convergence and just common sense, but I think there is too much of a reliance on technology these days that some of these methods are neglected.

                                               

                                              A similar example: How many of you have been at a cashier and handed them some change where the cashier has to punch in the numbers to tell them how much to give back?  Honestly, if there is an error in my favor, I just walk away without saying anything, but if the cashier doesn't give me enough back, then I will definitely speak up.

                                               

                                              I was taught that the engineering doesn't come in when punching in the data, but rather in understanding the results that it spits back.  If it doesn't make sense, then it must then be related somehow to how it was put in, so the experience kicks in to know how to fix the problem.  Yes, an iterative approach for analysis is in most respects the norm rather than the alternative.

                                               

                                              I agree with the post regarding using shells instead of beams, but not always, it depends on what I am trying to get out of the analysis model.  Whenever using either beams or shells, anything other than solids, than typically it requires more manual effort in setting up the problem and interpreting the results.  The up side is that you can actually do the problem on the machine that you have in a reasonable amount of time.

                                                • Re: Solidworks vs Roark
                                                  Derek Bishop

                                                  Joe, in all the industries with which I've been involved as a mechanical engineer, FEA simulations are not accepted as a safe design tool without additional validation. That could take the form of destructive testing, strain gauging or design checks against other traditional calculation methods. The Australian Standard for pressure vessel design AS 1210 states that where ever FEA is used it needs to be checked against other traditional calculation methods and gives Roark as one example of such a method. In addition to pressure vessels, the same appproach is seen in design standards for heavy mobile vehicles like buses, pressure piping, mechanical structual steel, lifting equipment, cranes and tanks.


                                                  There are exceptions. FEA software is used by structural engineers and piping engineers without the need for further validation. In those cases however the software is developed with direct reference to the applicable codes and in close consultation with engineers. SolidWorks Simulation is still a long way off achieving that level of acceptance in the industry.

                                                   

                                                  In case you hadn't noticed, there is a world of difference difference between the consequences of getting the wrong change from grocery shopping and a pressure vessel holding thousands of litres of high temperature toxic liquid rupturing. SolidWorks Simulation sells itself as a tool for engineers. Engineers normally do calculations when a failure will have serious consequences or where there are significant cost benefits by optimising the design. In other words in critical applications. In non critical applications a lot of design work is done by pure intuition. Why waste time doing FEA simualtions if it is not warranted.


                                                  What you are suggesting actually contradicts the following discalimer is SolidWorks Simulation 2010: "Disclaimer: Do not base your design decisions solely on the results from the software. Use this information in conjunction with experimental data and practical experience. Field testing is mandatory to validate your final design. This software helps you reduce your time-to-market by reducing but not eliminating field tests." The reason for the disclaimer is that if something goes wrong the engineer and the company he represents still takes responsiblility for design, not SolidWorks.


                                                  I have found the technical assistance you provide on these forums to be helfpul and hope that continues. I suggest that you avoid making these kinds of uninformed comments in future.



                                                    • Re: Solidworks vs Roark
                                                      Bill McEachern

                                                      Derek,

                                                      What do you think Joe suggested that was uninformed exactly and contradicted the disclaimer?

                                                       

                                                      Solidworks isn't the only FEA vendor with a disclaimer similar to that by the way. It is up to the engineer or more correctly the company process to ensure safety & suitability, along with the constructor and the operator of the equipment.

                                                        • Re: Solidworks vs Roark
                                                          Joe Galliera

                                                          Thank you, Bill.  I stand behind my statements, and I have never contradicted this disclaimer either in casual conversations, email or posts on the internet.  FEA is a design tool, and I feel that SW Simulation captures this spirit very well, and is never a replacement for physical testing and validation.

                                                           

                                                          What I wrote in so many words is that if you don't get a similar answer to Roark's as out of our Simulation tool, then it most likely is due to how the analysis model was set up by the operator in that it doesn't reflect the definition as set by the problem.  Sometimes it is not always straightforward to 'convert' hand calc simplifications to FEA, so I wrote that one should not get bogged down on those details.  I also stand behind the underlying quality of our Simulation tools and solvers.

                                                          • Re: Solidworks vs Roark
                                                            Derek Bishop

                                                            Bill, the need to validate FEA results as specified in various standards would apply equally to all FEA software packages with similar capabilites as SolidWorks Simulation. I suspect you are right that most would carry the same caveat as SolidWorks. No one is suggesting otherwise. The point of my comments was to highlight the need for validation. To state directly or by implication that this is not the case is wrong. That is certainly how I comprehend the comments posted by Joe. If he intended otherwise then maybe he could clarify the comments.

                                                             

                                                            SolidWorks needs to make clear to users of Simulation that it is a tool and is not substitute for a proper understanding of the engineering concepts involved. Users need to be aware of the need for validation of results. The recent discussion in which someone got upset when challenged over his ability to properly use the software is strong evidence to suggest that if SolidWorks is doing this, the message is not getting through.

                                                             

                                                            What do you think about this need for validation? What kind of work do you undertake? How do you go about validating design work done using SolidWorks? How do you think SolidWorks can better convey to users the need for Simulation to be used by suitably qualified personnel?