Measure For Measure

Anyone who’s ever held a micrometer has experienced that little nagging doubt: “What am I measuring?”

Most of the time, from a shop floor perspective, it’s an attribute that tells you something about the process you’re trying to control, rather than the part’s quality itself. And that can lead to a surprisingly common “hall of mirrors” effect where you measure, measure and measure again, and end up no farther ahead in understanding what’s going wrong (or right) with your process.

There are lots of ways to screw this up, but I’ll concentrate on three of my favourites:

1 Using the wrong measuring tool. The first time I saw this was as a teenager watching a QC inspector use a microm-

eter to measure a polypropylene gasket. Gaskets are designed to deform, so how do you get a repeatable measurement with the vice-like squeeze of a handheld “mike”? Crazy as it seems, the reason for using this tool was that the print value was specified with tolwerances in thousandths of an inch, so it was assumed that the measuring tool had to have a similar resolution. The results plotted to a beautiful bell-shaped distribution, which seemed to validate the measuring procedure. They were actually measuring the standard deviation of the operator’s thumb and forefinger pressure on the micrometer ratchet, but hey, the curves look great, right? This one is as much the fault of the customer’s inability to understand basic QA procedures. The solution, by the way, was an optical comparator with careful temperature control in the measuring booth.

2 Getting stupid with specifications. This one is a favourite of young design engineers. It really impresses the boss when they draw up a spec sheet as long as their arm for a two-cent part. Why is it necessary to know the dielectric strength of a spacer designed to keep stainless steel brackets from chafing during shipping? Scratch the engineer or tech (inventors do this too) and you’ll find that it comes from a materials handbook that spells out all the properties of the material they’ve chosen, and they’re going to hold their processor to every last one of them. This issue requires the people bidding the job to not just read the spec and cost in the superfluous testing, but to sit down and ask if that tolerance really needs to be held.

3 Cascading errors. By this I mean the influence one parameter has on others. A rectangular part that’s out of spec by going trapezoidal will show unacceptable tolerances in both part height and width, but neither spec may be really at issue — it’s the squareness of the part that kicks the other two parameters out of whack. While it may be strictly complete to report every out of tolerance measurement, if the part is out of square, you may as well stop, because nothing else will make sense, either. The solution is to think about the order in which you make your measurements.

There are other ways for things to get screwed up, but I think you get the idea. Question everything…gently, politely, but don’t assume that because it’s written down, it makes sense. QA, Engineering, Production, and the customer can all be wrong…especially the customer!