Computers are amazing tools. In the hands of a skilled, knowledgable engineer, they make wondrous creations possible. CAD/CAE programs have enabled the building and testing of machinery without ever cutting a chip or tightening a screw. Cool stuff! The problem is, computers live in a perfect world. That's not the world of engineering that we live in.

I cannot count the number of times I have reviewed drawings with tolerances impossible to achieve in real-world manufacturing. When asked about them, the designer usually responds with, "That's what our CAD system uses."

News flash: The real world is not perfect.

In the real world, plastics do not hold tolerances of one ten-thousandth of an inch. Six-foot-long lead screws do not have length tolerances of two-thousandths. Even if you could afford the cost to manufacture to such tolerances, normal environmental fluctuations would blow them out of the water.

So just because your CAD software is set to default to four decimal places, don't assume that is the right thing to use. And just because your modeling software can evaluate your creation with line-to-line fits and perfect (nominal) dimensions, don't let it!

Many years ago, when I took a finite element analysis (FEA) class to learn how to use computer modeling for my designs, the instructor made a simple statement that stuck with me. He said to never use the software to do something we couldn't do ourselves.

That, of course, raises this question: If we can do it ourselves, why have the software?

The answer: The computer can do it faster, with fewer mistakes. In other words, the computer can do it better and cheaper.

If that's so, then why do I have to know how to do the calculations too? There is a four-letter answer: GIGO. Garbage In, Garbage Out! Unless I can do the work AND anticipate what the answer should be, I have no way to evaluate the software output. Having the requisite knowledge has saved my butt many times. The computer didn't care when I slipped a decimal place on the input. It still gave me an answer and a pretty picture. But the results did not make sense, and I knew that because I had done a rough calculation beforehand. Doing it myself first saved both embarrassment and catastrophe.

So here is my suggestion. Don't sign off on a design until you know how it will be built and that the design allows for real-world limitations. Perform your modeling and analysis at the extremes of the processess, not at the nominals only. Ask yourself, "Will this be cut with an end mill, a laser, or a hack-saw? Will this be wired with hand-picked components or a Radio Shack(R) variety pack?"

But most important, ask yourself, "Could I make this myself? Could I get out the handbook and calculator and perform the analysis by hand? Do I really understand how this will be made?" If the answer is, "No!", get some help. Don't count on your vendors to read your mind or perform miracles. Remember that computers are tools that amplify the skills of the operators, good or bad!

Tom Solon is a licensed PE with degrees in mechanical engineering and economics from Brown University. E-mail him at [email protected].