FEA is Backward Machining
There are a lot of companies making their first steps into FEA analysis. At first glance, a lot of the tools look a bit like CAD. Three-D interfaces same CAD models plus pretty pictures showing where your part will break. FEA can sometimes look like just an extension to your CAD software.
But really, there are some reasons you often see CAD/CAM/CAE and not just CAD. If you do CAD, chances are you also build real things. So, most CAD users understand that the machining (CAM) takes a different thought process. FEA (CAE) is similar.
One of the common issues when going from your CAD model to a real world part is accuracy. The CAD model can be defined with almost exact precision but you simply cannot be that accurate in the real would. 2.00000 becomes 2.08335. Machine tools can’t make infinitely sharp internal corners. Tools have to be calibrated. Human error can happen. Material properties can be off. Designers can design things that simply can’t be built or ask for tolerances that will cost a fortune. “What you want” is never “what you get”.
So you end up with this cone of accuracy as you move from designed to machined part. The idea is to make this cone as skinny as possible to meet the design requirements and cost.
FEA has the same cone, it just points the other way. I’ve tried to illustrate this in the image below:
Unlike going from an “accurate” CAD model to less accurate machined part, the analysis process goes from” inaccurate FEA model to actual test results.
This analogy is important. Often, the results of FEA are treated almost like real world test results. If you were to give 100 people a CAD part to go make, you will get 100 different real world parts. If you were to give 100 people a CAD part to analyze, you will get 100 different answers. Which are the good ones?
So the real work with FEA is in controlling and understanding this accuracy cone. This is what companies have to focus on when starting with FEA. What level of accuracy do you want at each step of the process? How do you verify you are sufficiently accurate?
I’ll talk more about this in my next blog.
