No matter how complex the structure or how sophisticated the computer software it is always possible to obtain most of the important design parameters by relatively simple hand calculations. Such calculations should always be done, both to check the computer results and to improve the engineer's understanding of the critical design issues.

I'm not sure where the term "sanity check" comes from. In my work experience, it seems engineers would have thought themselves insane to depend on results of complex calculations, simulations, finite element or finite difference analyses before looking at something simple first.

We've learned this habit from our earliest days as students, when nearly all problems we faced were of the "given/find" variety: we were given all the inputs and were told to find a single output based directly on the inputs. In the workplace, problems seldom come nicely shrink-wrapped -- we have to find some of the inputs ourselves, and may not even know what outputs are of interest.

Complex analyses of various types are generally developed with the intent of using them again and again. If the analysis you're doing isn't to be repeated, you may find that algebraic solutions have no value. If what you're doing is of general use, develop a general model; if it's specific, use the numbers directly -- even on "the back of the envelope" or "on a cocktail napkin" if that's what's at hand. It's a time-honored tradition!

1. Are the units consistent? (Try solving the problem with numbers and units. You never know -- there may be a stray 12 or 57.296 in there somewhere.)
2. Is it possible that some input affects the result both directly and indirectly? Consider the case of the pattern of fasteners in shear. The shear load borne by an individual fastener is due both to direct shear and to moment about the centroid of the pattern (which you might consider indirect). Is it necessary to break up a problem into smaller, simpler steps in order to isolate these effects?

In either case, try to solve the simplest problem you can with the numbers you have first. If the answers you get make sense, then you take a deep breath and move on. If the answers don't make sense, here are some questions you can ask yourself:

3. Do you have ALL the inputs?
4. Are the inputs reasonable? Where possible, check the numbers against historical data.
5. Is the stated problem reasonable?
6. Are you asking and answering the same question(s)?
7. Did you consider all appropriate natural constraints? Sometimes actuators (for instance) have a limited range of motion -- are you violating such a range?

On manufacturing

1. Meet with manufacturing personnel (especially operators) during the design stage. Here you can do the following:
   • learn what tolerances and finishes are achievable
   • learn process limitations
   • verify fit and function
   • identify changes in geometry or tolerance that can make manufacturing cheaper or faster (i.e. "value engineering")
2. Some organizations have senior designers acting as "checkers" -- these folks will have the most accumulated knowledge of standards, tolerances, and so forth.
3. Parametric modeling software can correct for errors in dimensioning and projection that even an experienced designer might otherwise miss. Even with parametric software, it's still possible to override some automatic corrections.
4. Inspection is analogous to "peer review" in the design stage. The cause of gut-wrenching agony is manufactured tooling with problems related to designs that weren't checked.
5. Prototyping can be more cost-effective even than checking if the part or assembly is critical, high-volume, or expensive.