Kare11 reports on the NTSB plan for bridge analysis, but doesn’t get the engineering aspects quite right (I know, how terribly shocking and surprising, right?) They say:
He said that investigators were going to take a look at any unique design features that could have created a shift of 50 feet. To assist investigators they’ll be using an advanced computer program called Finite Element Analysis. This program allows the investigators to take elements of the bridge off and watch it collapse in the computer program.
“Finite element analysis” (frequently abbreviated FEA), also known as the Finite Element Method (FEM), is not a computer program but a computational technique.
The idea is simple: a complicated structure, in which it would be difficult to compute stresses by hand with analytical expressions or formulae, is broken up into smaller pieces (the “finite elements”). The stresses are calculated in each piece with special attention given to the restriction that boundaries or corners of pieces have to be consistently calculated for the several different elements they touch. The whole thing is then put back together so you get a picture of the stresses in the whole, complicated structure.
A computer program, the most popular are Abaqus and Ansys but there are many others, is used to do the calculations in steps. First you have to program in the shape of the object (in this case, the bridge) and its intrinsic material properties for each part of it (specific to steel, concrete, etc.), then tell the computer what sorts of external forces are on the structure for doing the calculations. The computational part then takes place and you have many options for analyzing the results. In this case, you would probably get back a colorful picture of the bridge where the colors would represent different stress levels in different parts of the structure, and you would look for places where the stresses were most dangerously high. You could then do what is sort of suggested in the quote above, and take away parts of the structure (like the parts you think might have failed first) and look at how the stresses increase in the other parts to try to model progressive failure of the bridge. You could also put in defects, cracks and flaws in the structures to see how they affect the stresses. I suppose you could eventually put together a movie of the bridge collapsing, but such things are normally done for the benefit of the uninformed–the information the engineers need is in the numbers signifying the stress levels in the different parts of the bridge itself.