I have always wondered why Simulation allows the user to assign a Plastic material in a Static study. Isn't this in 99% of the cases just asking for trouble? Would it not be better if Simulation would warn the user or outright forbid it to be possible?
Are there any cases where using a Plastic can be justified for a Static run, if so I'd love to hear people's experiences with this!
Your question doesn't make a lot of sense to me. What do you mean by plastic? A polymer based material or a material that exhibits plasticity - ie it bends and stays bent or maybe something else?
A polymer based material.
Hi Kevin: I believe you are expressing concern for how plastic materials can behave - namely, the loading can cause deformation beyond linear-elastic behavior. I can think of a few cases where you might want to use a linear-elastic behavior.
In one case we use the linear-elastic run in pragmatic FEA as a method to transition to a nonlinear run. That is, (as we taught the course), we encouraged a philosophy of "...just getting the model to run...", and "...take baby-steps..", using simple linear-elastic behavior. The turn-around is relatively quick compared to nonlinear, and you can trouble-shoot things like meshing issues, examining behavior due to applied boundary conditions (is the model deforming as expected?), evaluating reaction forces at supports, and so-on. Conversely, I have seen cases where users plow directly into a nonlinear model and uncover scores of issues that are super-difficult to identify - and the turn-around can take a very long time to trouble-shoot since the nonlinear stepping algorithm can take a long time. Only after conducting a linear-elastic run would we encourage setting-up a nonlinear model since the user has presumably fixed any issues and observed some sense of reasonableness in the linear run.
For a second case I have seen some PVC materials exhibit linearity up to about 3 or 4% strain (esp. at low temperature applications), so a linear-elastic set-up might be appropriate if loading is not expected to cause strain greater than these values.
For a third case I have seen customers who purposely do not want the response of their polymer plastic to "get into" the nonlinear response region. For a given loading, they iterate on geometry, wall thickness, adding ribs and so-on and conduct linear-elastic runs to stay under what they consider their "not-to-exceed" strain (which is usually the "proportional limit" of the stress-strain curve, as you may already know). I have observed this design philosophy for some plastic housing assemblies of portable power tools, for example.
However, I appreciate your comment about a warning from the software. It might be a good idea for the software to post a note to the user that linear-elastic behavior is assumed, should they choose a relatively soft material. However, it might be hard to define a threshold for the notice to appear - a good thought.
Hope that helps a little. - Tony
Hi Tony,
I am not sure I agree with more warnings....you need to know something about what you are doing when you use this stuff. I am a big fan of not needing to know much about the numerical aspects (it helps though -particularily in NL analysis) but one does have to understand what the "engineering assumptions" are going into an analysis. - whether its a hand calc or FEA. On Kevin's comment on the modal analysis I totally agree that showing the plot the way they do can introduce uncertainties in peoples minds if they are not that familiar with this type of analysis. And, for an outfit that really wants to simplfy FEA in general this is very schizophrenic on SWX's part. And really, there is no need to show a deformation scale. I remember back in the early days they just showed the mode shape plot as a single shade of gray. Some one really didn't think that one through on on adding the color deformations to the plot. Probably someone on the sales side who thought it looked cooler in a demo or some other misguided reasoning. You can now take the colors out of the mode shape plot. No idea why it is not that way by default.
On plastics/polymers - if stress is the object it makes no difference what the stiffness is nor whether it is linear or non-linear as long as it is about the same order of magnitude (and maybe not even then it is arguable) as there is no "E" in computing either bending or normal stresses if you are "assuming" small displacements - which you are in a staic analysis.. For deflections on the other hand it matters a great deal. A linear assumption on a non-linear material property is a perfectly valid assumption to make as long as you know what its implications are. In fact, I would agrue that no material is prefectly linear - aboslutely none - some are just more linear than others and most are linear enough depending on what your situation is. Being aware and knowledgable about the assumptions you are making in an analysis is quite scarce in a lot of fields of endevor these days and it can have profound consquences (as in bad). In our field testing is a great way to ensure that respect for reality is maintained. One needs to have the right level of skeptisicm with respect to the pretty FEA pictures and the numbers produced. Sorry Kevin, software is never (might be too strong a word) going to be smart enough to be surprise free if no knowledge of the phenomena being simulated is present in the operator.
How can you not want to get into the nonlinear response region, when most plastics if I'm not mistaken are nearly along their entire stress-strain domain nonlinear? can you use a bi-linear stress-strain curve in a Static analysis? I could imagine you could look one up and manually check whether the numbers are below the bi-linear transition. Sounds like an inconvenience but I suppose, it could work.
I think the modal analysis' displacement plot is admittedly one of a few, but most visible example of Simulation where it just should be more user friendly, as is at the moment it borders on the misleading!
I'm also thinking about mass participation factors which has no form of any warning, when you don't yet have enough mass participation in your system.
Who siad anything about not wanting to get it to non linear analysis? Static analysis is SWX sim though is a linera static analysis. Linear means linear ( yeah yeah there is a NL force controlled NL large diaplacement analysis that is supported in it but it is pretty limited). IF you want to non-linear and you have the material data knock yourself out and do an NL analysis. However, as I said if it is stresses your are after you may not need an NL analysis to get the answer. If it is displacements then you have to do an NL analysis.On mass participation factors: sometime you can not get above 80% in any practical number of modes - just stick some large thin shells/strucutre in your analysis. Mass participation in the in plane directions will be low. You ned to understand what is going on to know that you can live with it. If you are just after the frequnecies the mass participation factors are not of much value unless you plane on doing a linear dynamic study - the warning if it goes anywhere whould go there not in a frequency study.
Agreed, the warning should be once you try to make a Linear Dynamic study!
kevin, it sounds like what you'd like is an extension of the analysis advisor that does some checks for you on certain analyses. for example mass participation factor..etc. that might be good for some users. personally i'm with bill, I've spent a lot of time with the tool and adding more warnings might upset me a bit. but theres not reason we couldn't have an "interactive" mode and "expert" mode or something.
but in the end, i'm with bill. static analysis, you're assuming small displacements anyways, theres not reason you can't use a plastic as a linear model if you're staying in the initially straight part of the stress/strain curve. if you deviate, thats when nonlinear is a requirement.
Fair enough. As you start putting in one warning, before you know it you get 6 dialog boxes when you switch from a Linear to a Frequency analysis!
I'm not a big fan of the analysis advisor though, it creates a chasm between beginner and advanced users. There is, of course, but the advisor puts it in a bit of a negative vibe. Though effective it may be.
Kevin, I come to this a bit late. Plastic ain't linear, that's likely to be the root of the problem.
Take a look at a normalized stress-strain curve for any polymer (by the way, the differences between rubber and plastic are that plastics come in more colors and rubber starts with an r) and compare it to a normalized curve for any metal. Here I take normalized to mean that stress is expressed in terms of its value divided by its "yield". Yield here is in quotes because it is a problematic term for a polymer (beginning with an r or a p). True stress will also work, and its more entertaining. When testing a polymer dogbone specimen in tension you'll need to pull it at a much higher rate than for the metal to get anything reasonably comparible. One more correctly reports the secant modulus for a polymer tension test, not Young's. Some time when you're really quite boooored, take a look at the Four Element Model of Stress Relaxation and Creep. 'Round bed time would be good. Oh, and did I mention property's temperature dependance?
Hi Kevin: I believe you are expressing concern for how plastic materials can behave - namely, the loading can cause deformation beyond linear-elastic behavior. I can think of a few cases where you might want to use a linear-elastic behavior.
In one case we use the linear-elastic run in pragmatic FEA as a method to transition to a nonlinear run. That is, (as we taught the course), we encouraged a philosophy of "...just getting the model to run...", and "...take baby-steps..", using simple linear-elastic behavior. The turn-around is relatively quick compared to nonlinear, and you can trouble-shoot things like meshing issues, examining behavior due to applied boundary conditions (is the model deforming as expected?), evaluating reaction forces at supports, and so-on. Conversely, I have seen cases where users plow directly into a nonlinear model and uncover scores of issues that are super-difficult to identify - and the turn-around can take a very long time to trouble-shoot since the nonlinear stepping algorithm can take a long time. Only after conducting a linear-elastic run would we encourage setting-up a nonlinear model since the user has presumably fixed any issues and observed some sense of reasonableness in the linear run.
For a second case I have seen some PVC materials exhibit linearity up to about 3 or 4% strain (esp. at low temperature applications), so a linear-elastic set-up might be appropriate if loading is not expected to cause strain greater than these values.
For a third case I have seen customers who purposely do not want the response of their polymer plastic to "get into" the nonlinear response region. For a given loading, they iterate on geometry, wall thickness, adding ribs and so-on and conduct linear-elastic runs to stay under what they consider their "not-to-exceed" strain (which is usually the "proportional limit" of the stress-strain curve, as you may already know). I have observed this design philosophy for some plastic housing assemblies of portable power tools, for example.
However, I appreciate your comment about a warning from the software. It might be a good idea for the software to post a note to the user that linear-elastic behavior is assumed, should they choose a relatively soft material. However, it might be hard to define a threshold for the notice to appear - a good thought.
Hope that helps a little. - Tony