Years ago, I wrote about the nature of engineering design in my then CAE Magazine column. A NASA engineer took great umbrage at my calling the process “trial and error.” He sent me angry e-mails and copied much of NASA, saying I was undermining the credibility of engineers by using a term that includes the word “error.” He said this implies that mistakes are part of the engineering process, which, in fact, is thought-out, methodical, and reason-based.

Of course, “trial and error” is not about mistakes — it’s about our inability to predict the behavior of complex systems. We create a design that we think addresses the design challenge. Then we build a prototype and test it.

The difference between desired and actual behavior is the “error.” We use this valuable information to modify the design, and hope that the next iteration yields behavior closer to what we wanted.

I emphasize “hope” because, in many designs, we don’t know what changes will, in fact, bring us closer to our goal. We have experience, intuition, and educated guessing as guides. But seldom do we have hard science or math to back up our iteration decisions. There are few iterative design processes that are guaranteed to produce the desired behavior.

In general, this is not a problem. Experience, skill, and intuition combine in the engineering process to produce great results as documented in Machine Design and other publications and seen throughout the world.

Creating computer-based models that include behavior is where CAD is going and its ultimate purpose: It provides a frame in which the trial-and-error process can take place far more quickly, safely, and inexpensively than in the building of physical prototypes.

This trend helps engineers be more innovative and create new designs more quickly and reliably than ever before. But to say that simulation works, and is a good thing, is not to imply that we have learned how to best optimize. Can we always ensure that the iterative process converges on the best possible solution for the design conditions? That’s math, physics, and other science we haven’t mastered yet.

So when I read a recent white paper published by an engineering-automation software firm that I know and respect, I felt they were giving a false impression by saying: “The goal of simulation-driven design is to converge on optimal solutions as rapidly as possible. By exploring diverse concepts early in the process, engineers can quickly understand the design approach that will best meet performance objectives and use that concept to specify detailed design.”

Yes, that’s the goal. And exploring diverse concepts early in a digital environment, where the exploration is cheap, safe, and fast, is wonderful. What I object to is the claim that by doing so, “engineers can quickly understand the design approach that will best meet performance objectives...” That is too much to promise.

Simulation-based design is the best we can do right now, and the firm that produced the white paper is on the leading edge of all that is good in this field. But it would be a mistake to think that the process now yields optimal results. By all means, simulate and strive to optimize. But do not think for a moment that the current tools do it all.

Joel Orr

Stuck for business ideas?
Drop me a line: joel@joelorrcoaching.com

Edited by Leslie Gordon