Where do I belong?

Sometimes I'm not sure if I know where I belong. In the corporate arena, or in front of a computer? I guess I can't be blamed. After all, how many people go to work like me feeling as if they are a PhD student? Yet I am not a PhD student, nor would you classify me as a researcher, and I'm getting no guidance at that. But then again, I guess you just do what you gotta do and just see where the wind blows you know.....

So I've been learning something new everyday lately. And with everything new I learn, I learn that there's more things that I don't know. Now, that's kinda freaky you know, cos it really should be a nutty professor who's doing such things and worrying about how to save the world, not a government engineer!

Just yesterday I was trying to get the gist of nonlinear static analysis. I mean, that's the stuff that most finite element programs can do nowadays, so its no biggie. We learnt that in school too, all that Newton Raphson strategy with automatic load incrementation, stiffness update, iteration, energy norm termination blah blah. But Professor Filippou did warn that its not so simple, you gotta know when to use which parameters and which strategy. I thought it was all baloney, just turn on all of the controls and put the maximum parameters, that always works in the classroom right? Well, in the real world, your computer software (and hardware possibly) is gonna blow up in your face if you don't know what you're doing. And what matters worst is, what parameters do you even start with?

Today I moved on to eigenvalue analysis. There was this chapter on Guyan reduction. Basically that's good 'ol static condensation that you learn in school, just condense out all the undesired degrees of freedom to make your solution go faster. Simple stuff? Again wrong! In a real model you may be looking at thousands of degrees of freedom. Let the computer decide? Turns out that's the least accurate way....

Let's not even talk about meshing. Use the automatic mesher in the computer? Yeah it churns out very nice symmetrical meshes. But computational mechanics is not about aesthetics. Nice mesh often equals poor mesh, especially if your dumb automeshing software always takes the easy way out by mixing tria and quad elements, resulting in stiffness locking. Guess you can't rely on automatic meshers despite their claimed sophistication....

And there's more to come: nonlinear contact analysis, damage modeling, geometric nonlinearity etc. And I always wondered how difficult could it be to use a general purpose finite element program? Pretty damn difficult, if you want to push it to the limits and there's nobody else around to help you....