11 Replies Latest reply on May 21, 2013 12:49 PM by Jared Conway

    SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?

    Drew Fullerton

      I'm very new to SW Sim Pro - less than a week in fact; however, I was wondering what the pros and cons of converting "stuff" within an assembly, regardless of how complex, into surfaces were - as opposed to leaving them as solids??

      Is speed the only advantage or does accuracy improve as well (or does it suffer)?

       

      Thoughts would be appreciated.

        • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
          Jared Conway

          what "stuff" are you converting? if it is thin in one direction relative to the rest, that would make sense to convert to a shell. you want to go this way for solving speed but also from an accuracy standpoint, you want 2-3 elements across the thickness of a part so to do that with a thin solid, it might be very difficult in a reasonable amount of time.

           

          same thing goes for using connectors and beams and other types of idealization.

          • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
            Drew Fullerton

            Jared, was wondering whether you would weigh in - lol. Been monitoring a number of forums and - there you are!

            "Stuff" - my generic term for just about anything; however, other thatn speed, I am of the opinion that one should use solid bodies within FEA to approximate the real world as best we can - thoughts???

              • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                Jared Conway

                Regardless of what idealization you use, the finite element method is still discretizing the problem so that it is solvable. You could do everything with solids but I'd you can do it in less time with the Sam level of accuracy, why wouldn't you take advantage of it?

                 

                You could argue that building a model for shells or beams might take some pre processing time but my experience is that either you'll start building your parts and assemblies smarter so they are easier to use in simulation or you'll work with models specifically for simulation for conceptual design. In both situations, the work needed is either eliminated or minimized significantly.

                  • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                    Drew Fullerton

                    Good response about building models smarter - currently taking what I get from clients and modifying to suit.

                    Ref same level of accuracy - our models of the real world are an approximation at best - whether I use FEA or do stuff handraulically, as long as they're in the ballpark of what I expect, I'm happy!

                    However, being new to SW sim, I'll have to develop a level of trust I think.

                    For example, did a buckling analysis on a 24" diameter pipe - pretty simplistic result - yet as a structural engineer, I would expect more than a pretty picture of an image waving in the wind!! - lol - just playing devils advocate here! Not sure why the analysis is even a feature of SW? It doesn't appear to take into account slenderness ratios etc - anyway, still have a lot of playing to do.

                      • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                        Jared Conway

                        In our consulting work, basically the first step is to decide whether to work with the model as is or build a model for analysis. Generally we can work with the model as is with some simplification or reassembly but when we work with customers in the conceptual phase, it is really nice to build a model for analysis. And the benefit is usually that can be used as a skeleton for the full assembly. Truly win win!

                         

                        On getting confidence with the software, check out the verification problems in the help. Also julien from solidworks did a bunch more and you should be able to find them with the keyword afnor. One of the advantages of working with customers on consulting projects is that we usually get to see the results. And so far, the simulations have been very close!

                         

                        On buckling, did you run an actual buckling analysis or a static analysis with a compression load? The output of the buckling analysis would be the buckling load factor, whereas in the static analysis it will unlikely predict the buckling. If that is something you'll do often, you may also want to look into nonlinear buckling with sim premium.

                          • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                            Drew Fullerton

                            Confidence is growing as I use it; but since it is only a tool, it has it's limitations - that, I think is the biggest issue, not knowing what the limitations are.

                            Buckling - the steel codes are quite specific in their direction on what and how to check steel members for buckling (we use a process called limit states) - don't think SW does, so a buckling factor is not useful to me in the structural world. Regardless, until I know the math behind it, I will always resort to my excel spreadsheet to check out buckling

                              • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                                Bill McEachern

                                Regardless of what the Handbook of Steel construction says, there are three types of buckling: linear (Euler) buckling, Non-linear buckling (P-delta is a sort of approximation of this) and post buckling. Linear buckling is based on the undeformed shape of the sturcture (what you learn in mehcanics of materials in undergraduate courses). Non-linear is based on the deformed sturcture under some set of applied loads (touched on in some undergraduate courses). Post buckling is you collapse the structure and see at what point the maximum load is reached - this is, in FEA speak,  a some what tricky NL calc depending on what is going on (not covered in undergraduate at all in my experience). SW sim does the first two in Sim Pro with some limitations on the type of loads supported in the NL buckling case - if you apply the loads in a buckling study it just does it for you and it looks like a linear buckling calc unless you see what is happening in the solver window. You need Sim Premium to do the post buckling.

                                 

                                In any event, you need to have models with the right element topologies to get a good approximattion of the stiffness of the structure. Solid elements are contunum elements and theorectically do cover the highest degree of complexity with regards to a complex state of 3D stress in a part: however, they require significant numbers of them to get where you want to go when aspect ratios are large. The general rules are something like this for shells - thickness to spand ratios of less than 10:1 use solids, 10 to 20:1 use thick shells, graeter that 20:1 use thin shells. Fro beams when the length is greater than say 10:1 as compared to the cross section dimensions is worth considering but depends on what you are modeling. For the analysis of buildings, at least in my expereince, they tend to be dominated by structural elements: shells and beams (mostly beams) as opposed to continum elements which tend dominate the analysis of say fittings, connections and many other types of problems.

                                 

                                You can use solids instead of say beams provided you are aware of what is well approximated by there use and what is not.

                                 

                                Hope that helps.

                                • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                                  Patrick O'Hern

                                  One advantage of using shell elements, is that you don't have to worry about whether or not you have enough elements across the thickness.  As Jared stated, you need at least 2-3 elements across your thickness, but it can be easy to forget to check that.  Using shell elements allows you to deal with the other aspects of your study without worrying if you are getting accurate results do to the thickness.

                                    • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                                      Drew Fullerton

                                      ok, thanks for all inputs; here is what I have concluded;

                                      1. Speed, and processing power are not the only advantages of FEA within the SW environment- the biggest advantage is the power of optimization
                                      2. I have arrived at this conclusion based on your responses and use of the product.
                                      3. I would like to offer responses to your inputs - responding in order of participants remarks in this thread:
                                        • Jared, as the primary threads to your responses, you brought up accuracy, building models for analysis + you suggested how I might build confidence in the software. Let me address accuracy first by throwing the following out there, "It is a game of probabilities"
                                          • I'm certain that you appreciate that everything industry does to provide man made product to trades that ultimately form, shape, bend and build are all produced to a range of tolerances. For example, consider steel and aluminum, chemical composition and material properties such as tensile and yield stresses are established by material codes that govern producers ability to sell within a certain market segment. By the time professional engineers get involved with the design of typical structures, "limit states" design for strength and stability is incorporated - we apply probability to the liklihood of primary environmental loads and companion loads occurring at the same time vs the likelihood that we are using an inferior batch of steel, albeit a batch that is within acceptable tolerances. Therefore, how do we measure accuracy when we engineer in a world of "close enough"?
                                          • Building models for analysis  - I like it
                                          • Building confidence in the software - the guide books I have read so far suggest going back to tried and true physics equations (basic priciples) that are embodied in everything we do in the macro world. Check your results against what you know - I find this to be really sound advice and do it as second nature.
                                            • As an observation/aside - many queries I have observed in the forums, seem to revolve around establishing boundry conditions/constraints/restraints not being established correctly - in other words, the users understanding of how to restrain/constrain something in SW to represent a real world constraint is not being understood and translated.
                                        • Bill - good synopsis - but whether I am using AISC, CISC or Euro codes, I know the equations to use, why they are used, their limitations etc - within SW, I don't know squat - is there tech documentation that tells me everything?? What the equations are, what the limitations are etc? Haven't looked, thus the question.
                                          • Can't say I'd want to use SW sim for a building structure - other software is out there that does a really fine job - as a start, I may give it a try with some trusses and compare to other results - all comes back to confidence in the product.
                                        • Patrick - BUT, you really have to think about how to represent the real world constraints, more so with shells than, in my noobie opinion, solids. With solids, there appears to be less idelaization, therefore, less opportunity to screw up the constraints - which seems to be a big issue.
                                      4. Anyway guys, thanks for your thoughts
                                        • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                                          Bill McEachern

                                          On the equations: The main equation being sovled in FEA is F=kx, the rest is just matrix algebra and eigenvalues, at least for the linear stuff. The non-linear is the same but is iterated to equilibrium, and the stiffness matrix is updated to account for the changing stiffness with deformation. Obviously k is the tricky bit. k is formulated by assembling each elements stiffness to compile the over all structures k. The equations to get k for a solid element are the generalized 3D description relating the nodal displacements in terms of linear stiffness and the state of strain (Hooke's law) & hence stress.These types of elements are called continum elements - just a hunk of steel and it doesn't know what kind of structure it is in.  For other element topologies this is not the case. Beam elements are formalated by using the beam equations you know and love, like sigma=Mc/I, P/A, etc, but are resolved to produce the beam element stiffness with respect to nodal displacements.  Shells are the same, they use the realtions of classic structures for plates (Timoshenko, et al)- these are called structural elements because they know what kind of structure they are. If you look at the outputs for these element types you will get the picture.  Any good somewhat deep book ( The Fintie Element Method by Zienkiewicz & Taylor is one that can put you to sleep pretty quick) on FEA will discuss the various element formulations.

                                            • Re: SOLIDS VS SURFACES - IS SPEED THE ONLY ADVANTAGE?
                                              Jared Conway

                                              if you're looking for equations, the help has some as do the training manual and the old cosmosm handbooks that you can get in the solidworks KB.

                                               

                                              but in the end, i think it goes back to your assumptions. overall you're right. the study is only as good as the setup. the setup will be built on assumptions. your knowledge of the system in question will determine whether these assumptions are good or not, thats why the software needs to be run by someone that knows what the expected behavior is.

                                               

                                              looks like you're on the right path for things drew. simulate on!