This week is exciting, as I finally venture into the world of finite element analysis through COSMOS Works.  I have waited for such a long time! Unlike many young engineers nowadays, I didn’t receive a foundation on finite element analysis while in college.  Sure, we did have a class on numerical methods, where finite element methods were briefly discussed, but nothing else. Years later I had a brief encounter with a demo version of ALGOR, while my boss was evaluating their software, considering whether to buy it or not. The examples they provided really blew my mind. I had never seen anything like that! Unfortunately, my experience was cut short when my boss decided not to buy the software, since he couldn’t see any practical use for it. It didn’t matter much; I had already been bitten by the FEA bug, and wished really hard that someday I could learn more about it and become acquainted with software such as that one. Fast forward to these days: I’m still crazy about learning FEA, and this time I’m doing a bit more about it than merely wishing.  Here I am, trying to teach myself some FEA and COSMOS Works at the same time.

Of course, my very first experience had to be with COSMOSXpress, which is similar to COSMOS Works, but very limited in capabilities. I have the impression that these limitations, plus the fact that it’s so easy to use and everything takes place “automagically” behind the scenes may make it appear a bit fishy to some, but I have to disagree. I think the usefulness and reliability of FEA software such as COSMOS Works has a lot to do with the way we design our study, as well as how we interpret the results obtained from it. While it’s true that it wouldn’t be wise to take these results as Gospel for a final design, and we’ll always need to do some other calculations and testing, I really believe that finite element analysis can save a lot of time and effort and bring us closer to a final, much better design. 

I don’t think I really understand all the Math behind the solution, but what I do understand is that, basically, a finite element analysis requires creating a simplified model where the original part has been broken down in smaller pieces, called elements. The points connecting adjacent elements between them are then called nodes, and for each of these nodes there will be a set of equations that describe small displacements at the element level and that will need to be solved simultaneously with those of all the other nodes in the model in order to determine stress distribution, for instance, in the whole component. This process is called meshing or creating a mesh, and although it occurs automatically inside of COSMOS Works,  the users still have the ability to control how this mesh is created, thus affecting the accuracy of the results.

So far, I’ve learned about two different kinds of elements that are available in COSMOS Works: Solid Element and Shell Element.  The solid elements are used to model, well, solid components, parts or assemblies that have significant thickness relative to the rest of their dimensions. The shell components are used to model surfaces, thin walled components, sheet metal parts, etc. Each one of these elements is available as a first order element (also known as draft quality) or second order element (known as high quality), depending on the number of nodes on each of their edges. Solid elements are also known as tetrahedral elements.  The following image shows a first order tetrahedral element on the left and a second order tetrahedral element on the right. Notice that the second order element has an extra node for each edge, and this allows its surfaces and edges to deform in a curvier manner, conforming more accurately to the model’s surfaces, fillets, rounds, etc. For this reason and because there are more nodes and equations involved, second order elements provide more accurate results than first order elements.

Besides selecting what kind of elements to use, it is possible to select how fine the mesh will be. In general, the finer the mesh, the better the results, but it will also require more time and computational resources.  The following image shows a coarse mesh of solid elements.

And here is the same component using a finer mesh.

I’m not completely sure, but I think this is something new for 2008, because I can’t obtain the same results when I try it in 2007: the ability of automatically combining and transitioning the mesh size for certain areas of the model with high curvature. Notice in the following image how the mesh is finer around the holes and transitions to a much coarser mesh for the rest of the model.

This option can be found by selecting Alternate for the Mesher, under Options, when creating the mesh.

There certainly is a lot more to learn about FEA and COSMOS Works, and the more I learn about it, the more fascinating it appears. I wonder about examples of real life use of finite element analysis to design products, and what other methods are applied in order to verify the accuracy of the results. According to some notes I was reading, finite element methods are not precisely new; they’ve been around for over 60 years or so, yet, somehow it seems like a novelty. Or perhaps it’s just a novelty to me? Is FEA something that everybody uses nowadays?