Interpretation of FE Analysis

[Catjacob]

When I did my FE course at university, I was taught how to do FE formulation (e.g. transforming a PDE into a weak form, defining primary/secondary variables) but not how to apply the results to real engineering problems.

For instance, I know how to solve FE equations to find out say shear forces at nodes of a plate/mesh and find out the shear force distribution between nodes. However, I have no idea of how to “transform” the shear forces at different point inside the mesh/domain into shear force per unit width.

Could anybody give me a clue?

 

[mendinho]

When designing steel you do not need the efforts as far as I know but Von Mises stress.

For the reinforcement of the concrete, if you use shell elements, the software provides you moment (Mx, My,Mxy), shear (Vx, Vy), axial force (Nx, Ny, Nxy) and even torsion per unit of length. If you use 3D elements you will probably need to "integrate" the stresses. At least, that is what I usually do.

Another useful rule to obtain concrete reinforcement efforts is the Wood&Armer rule, which is

Nx*= Nx +/- Nxy
Ny*= Nx +/- Nxy
Mx*= Mx +/- Mxy
My*= Mx +/- Mxy

Finally, use fine meshes and redistribute stress concetration points in linear elastic analysis.

 

[Bobfromoh]

Here's what I think (I'll probably go off in the wrong direction). I think you should forget about the information in the college books and concentrate on whats in the users manual for the FEA software you are using.

The problem I see is that schools emphasize the theoretical side so that studends don't seem to know how to develope an understanding of the practical side of FEA anaylsis. When I first started doing FEA, generally classes weren't available yet. In the job at the time, we were forced to evaluate the equipment with FEA. We went to FEA vendor classes for a few weeks and it seemed like everyone got the hange of using the software.

Since that work 25 years ago (I work for another company now), I have seen "ZERO" engineers come through that are interested in doing FEA. When asked, the answer is "I don't know how to solve those equations" and they give up before they get started. The only people I have seen over the years with an interest in using FEA are the people from those early days.

We leaned to build the model first, converted the solid model to the appropriate FEA element, applied the boundary conditions and loads and ran the analysis. We looked at the results and develpoed ways to make sure the results were accurate. We had access to 2 or 3 FEA packages plus our design was eventually tested. We learned by experience.

Eventually thise equations in the books are relevent to understanding FEA (such as choosing the appropriate element)but I think there is a lot to learn before applying the theory.

 

[prost]

If you don't get the theoretical side of FE in school, where would you get it?

Sounds like you are using beams and/or plates? Is that correct? If you are using commercial FE software, this software should allow you to plot or otherwise extract the results at any point in the body you are analyzing. If you are using your own FE code, then it's a little more difficult, but not hard in general. Say you had a beam, and knew stresses at a point in the beam over the entire section, you can integrate the stresses over the section to get the moment and/or forces at the section.

 

[xerf]

Nowadays it's so easy to use a FEA program, a cave man could do it.

I saw so many persons using FEA codes in a "practical" manner without proper theoretical background...and many of them put very "interesting" questions.

 

[prost]

I would argue a different approach, Bobfromoh. It makes no sense to send someone off to do FE without at least a gradual build up of that person's knowledge of FE theory (university training before doing FE is better, OJT if you have to. Incidentally, I have taken 3 FE courses at universities, all required actual modeling with FE software, so you couldn't say that my theory was completely disconnected from the application, even if the applications were simplified) because so much of the modeling and the results interpretation is tied to the formulation of whatever FE software you are using. For instance, if you know the theory well, you would never blindly model a rubber in your FE model with a linear elastic material with Poisson ratio=PR=0.4999999 if you are using h-element software like Ansys or ADINA (interestingly you could get away with such a high PR with p-version software), because you'd end up with the familiar 'element locking.' If you had little or no knowledge of FE theory, you would have no idea that the stiffness matrix "K" in Ku=f has terms like (1-2*PR) in the denominators, and therefore you'd be at a loss to explain such strange behavior as you see with element locking; in fact, you might just take the results as they are without questioning them. You might say that this an experience problem, while I might argue that it's a problem of ignorance of the FE formulation.

There are so many possible sources of error and many ways you learn to reduce the magnitudes of those errors--those you learn in both university training and OJT; the job gives you far too little time to get the theory, university training gives you far too little time to get the practical experience; with one or the other, you are half a FE analyst--with both, you can make a good FE analyst.

I won't forget to mention the obvious--far more important than knowledge of FE theory is knowledge of elasticity (or nonlinear material if that's what you are modeling) theory, IMO. I find the most difficult part of the FE model is most often specifying boundary constraints; a thorough knowledge elasticity (not just strength of materials) is vital to understanding how to properly specify boundary constraints and, in turn, developing better FE models.