My friend and employee Asia just started a maternity leave. Since she was also responsible for training our new team member, in her “casual-self” brilliance she left her an “FEA with no stress” checklist. I’ve read it, and it’s so awesome that I simply have to share it with you! It also gives me a chance to say that I’m really proud of Asia and that I will miss her a lot : )
All right – let’s get to it!
A. The geometry
Usually, you will mesh geometry rather than building a mesh from “scratch”. If that is the case, it’s always a good idea to check if the geometry is ok. There are some things that seem trivial (like surface overlap). But not fixing that can cause trouble elements in some places… thing really difficult to find in postprocessing!
Check those before you run an analysis, or even before meshing:
- Are all surfaces have normal in the same direction? We do a lot of shell models and it’s nice if all the “top” surfaces are on the same side. Makes displaying outcomes much quicker!
- Are there any overlapping surfaces? However accurate you do stuff… things just happen. Just check if all is nice and move on!
- Delete all unnecessary lines and points! This is only 2 clicks in Femap (such a cool tool BTW). They don’t get translated into the input file, but better not to have them “flying around” if they are not needed!
- “Rebuild model” – check for errors. Yet another Femap tool. You can “rebuild” your model, and if Femap finds some errors it will throw them at you. This is a nice way of checking if you won’t translate some stupid stuff into your analysis model!
B. Material properties
Setting material parameters can take seconds. In more complex problems far more of course! In either case, it’s important to be aware of the unit system you are using… and the freaking self-weight! Just check those and you should be fine:
- Do orders of magnitude of material parameters align? When you decide on a unit system this is usually pretty simple. You just use meters, inches or whatever while defining the geometry and your “unit of choice” when you define loads. This is not as simple, however! All those choices “combine” in material properties, where you can get some bizarre units for density or Young’s Modulus. This has to be checked and verified!
- Is the self-weight applied? Don’t get me started on that! In most civil-engineering software self-weight is automatically applied. In FEA you most likely have to input density and gravity acceleration somewhere. I had to redo the calculations several times because I missed this! This is such an important thing to check!
- Does material fit the analysis? As you know there are many approaches to material parameters. You can go with the linear or nonlinear material. The nonlinear models can be pretty sophisticated. Some analyses require first using the linear and then nonlinear approach in second analysis run. Geometric nonlinearity is similar, but you can decide if you want to have linear or nonlinear geometry in solver settings. Material nonlinearity is defined… in material properties. It is easy to forget to change the linear for nonlinear material while defining analysis (there are no switches in solver settings that does that!) – hence the checkpoint!
This point might depend on your software a bit. In most FEA packages, you first define the geometry and then assign “properties” to shells, solids, and beams. In some, the properties (like thickness or material) are assigned WHILE creating geometry. If that is your case checking would be a bit different. However, it is still valid to verify if the model is made from materials and thicknesses that should really be there!
- Are properties defined correctly? The fact that the property name says “10mm” does not necessarily mean that it is defined as such. Sometimes you can misclick, sometimes you change something (thinking that you will remember about it!). This way or another it is worth checking if all thicknesses are actually what they should be!
- Are correct materials assigned? This is a rather straightforward check. Usually, when you create a property you will assign material to it. Just make sure you did assign a correct one : )
- Are properties correctly assigned? So we took care that properties have good values, and that good material is assigned to each of them. Now it’s time to check if actual parts of the model have correct properties assigned!
It is already assumed that you have the correct load values. If you don’t check that may take some serious time (depending on the code you are using I guess). However, if the values are calculated/estimated correctly it’s not the end! You still need to check if the load is applied correctly. Let’s take a look:
- Are the loads applied everywhere they should? A simple miss-click and you are in trouble… better to check then!
- Do the loads have good values/directions? Even if you have calculated the load values correctly, you still need to check if they have those values in FEA! A choice between total load/load on area/load on element can greatly influence this – so be careful with what load “type” was applied as well!
- Is the self-weight applied? Yup! It was already mentioned. And it is mentioned again! Go check – this can save you a lot of time later on redoing the analysis!
- Are coordinate systems ok? You may use scripts to create loads. In such cases, you will most likely base the load on a coordinate system. Is it the correct system, and does it have a “zero point” where it should be? Are the axis directions in your model in accordance with what you have assumed in your script?
There are two things here. The more important is, that supports need to be realistic. Far too many models I have seen were simply “stupidly” supported. This is however not what this point checks! It only checks if the well-thought-out support system of yours (that you have figured out before) is actually correctly implemented. There is no way to make a checklist that would allow you to verify, if you have a “proper” boundary conditions – it’s a far too complex problem!
- Are all supports implemented? It’s all great that you have figured out a way to correctly support your model… but did all supports made it do the model in the end? Are they applied on proper directions and in good areas etc.?
- Coordinate system! As with everything else supports are created in a specific coordinate system… just make sure it’s the one that you thought it was!
- Is your model stable? Sometimes in haste we “upgrade” the support system of the model… and by “accident” we can create a mechanism. The solver will show some errors quick enough, but sometimes it makes sense just to check if all is good : )
Again, designing mesh is a tricky thing. Here we just want to make sure that the mesh you wanted to have will actually work in the model : )
- Is the mesh consistent, are there free edges? Sometimes, when you mesh something, elements won’t get connected at nodes. In Femap, this happens for instance when you remesh part of the model with some meshing toolbox command. At first sight, all is good. But there are double nodes in certain places and your model is actually not connected together. This is important, as the solver will (or at least may) throw out errors similar to those with no stable model. Save your time in searching and just check if there are no nodes overlapping, or that mesh doesn’t have “free edges” where there should be none!
- Is the mesh overlapping? That is bad. Maybe you have missed an overlapping surface, or maybe something just went wrong. It’s good to check if there are no overlapping elements – this leads to poor results!
- Are there any local mechanisms? Sometimes where you are using rigid links, or other elements, you can cause a local mechanism (like 3 hinges in a row). This will be a mess to find – just bare this in mind while meshing!
Uff… this is the last one on the pre-analysis list! Let’s take a look at the list, and be done with it!
- Are all necessary contacts implemented? You know, it’s quite easy to miss something in a big model. That may result in an unstable model… just like an error in meshing or boundary conditions. This is why it’s good to be extra careful here… as finding a mistake in a model can take some time!
- Are good regions used? Contact is usually defined between regions, parts or however your soft calls this. It’s great to be extra sure, that we selected the proper areas : )
- Are region directions correctly defined? When you assign contact it can be assigned from a “positive” or “negative” side of a surface. It’s good to check if you are not trying to connect an outside part from the inside of our model!
- Are good offsets assigned? In various occasions, contact won’t happen where the surface is, but in some distance away from the surface (in the “positive” or “negative” direction, as discussed above). This is commonly referred to as offsets. It would be great if those would be correctly defined!
- Initial penetration! Usually, I don’t have models that could generate those, but if I have, I’ve learned to think about it. Sometimes “pushing a model outward” to reduce the initial penetration can (in itself) create insane stresses and strains in your model. By default, I simply elect to ignore the initial penetration, unless there is a good reason not to do so!
Last few lines
I hope you found this useful! If you did please give me a favor and share this posts with your friends and colleagues that might be interested!
If you want to learn more about FEA, don’t miss my free FEA course! You can get it by signing up below this post!