In "Magnum Force", the second offive Dirty Harry movies, Clint Eastwood says, "A man gotta know his limitation." Although used in a different context, these words ring true in our professional lives, both in terms of our individual skills and the tools we use. I'm a CAE engineer who uses FEA on a daily basis. But, I've noticed the change in strategy of CAE software companies and its impact on the decisions that are being made in design. Specifically, they're re-writing their FEA software so it can be integrated with CAD packages any design engineer can use. The idea is to have design engineers do a first-pass analysis, and in theory that's great. It can make for a more efficient design process, and it can lower costs. But the implications of a wrong first-pass analysis can lead to a totally over- or under-designed piece of machinery with underlying issues related to safety and product cost. So, does this strategy serve industry well in the long run?
The answer depends on three things: First, how much time design engineers are willing to spend in understanding the physics/failure mechanisms/weak links of their designs prior to embarking on creating FE models; Next, how honestly they recognize their own limits; and finally, how closely they work with the professional analysts within their companies.
FEA is both a tool and an art form. And, it's an approximation. Knowing which part of a design contributes to its overall strengths or weaknesses is essential. To get it right, you have to carefully plan the proper material model, mesh density in critical areas, and the representation of the boundary conditions. That requires time and experience. Just when you think you've got it right, you have to step back and reassess. And, you have to keep in mind two contradictory, yet equally valid aspects about FEA that every analyst knows: All FEA answers are correct. And, as Vince Adams said in his book, "Building Better Products With Finite Element Analysis," all FEA answers are wrong—we just need to figure out how wrong they are. If the engineer makes a mistake entering the material properties by an order of magnitude, the stress results may be reasonable and correct but the deformation numbers will be off by a factor of 10! In effect, the engineer has inadvertently solved a different problem than the one he started out trying to solve.
Even experienced analysts always fall back to sanity checks like closed-form solutions. I always assume that my results are wrong, and I go through a sanity check on each problem. I ask myself: Is the model linear or not? Are the loads static or dynamic? How are the parts attached or interfaced? Are my results biased by leaving out the variation in material strength or the manufacturing irregularities?
CAE software vendors are not exactly guardian angels in any of this. They don't provide guidance on problem set up, unless you pay for it. They won't walk you through the selection process for the right FEA package that suits your needs. But FEA specialists within your own company will help you. Take them to lunch. And, don't hesitate to use them as a sounding board for your designs and FEA modeling approach. They should be your new best friends.