Other Dimensions: Working with Measurement Uncertainty
January 28, 2010
If outside facilities do your calibration work, you’ll notice a value shown as measurement uncertainty on their reports. If it’s not there, get a new source for your work because without it, the numbers they’re reporting are simply readings that have little or no value. The same thing applies if you are doing your own calibration work. Technically speaking, it applies to any measurement made.
Any laboratory that is accredited by a reputable organization will show the uncertainty. If a laboratory claims to work to ISO 17025 or similar standards, it too should show the uncertainty because it is a mandatory requirement of such standards.
How uncertainty values are calculated will be dealt with another time, but for now let’s look at what the final number means and how to use it.
Simplistically speaking, the value for uncertainty represents a range around the reported value within which lies the “true” value for that specific measurement. For example, if the reported dimension is 0.2502 inch and the measurement uncertainty is 0.0002 inch, the true value will be somewhere within 0.2500 to 0.2504 inch. Of course, the true value could actually be the same as the reading, 0.2502 inch, but the odds are it will not be.
If the potential size variation shown in this example is not good enough for your requirements, one way out of the impasse would be to use a laboratory with smaller uncertainty. You could review the uncertainty budget that produced the value to see if there are one or two elements that could be changed to reduce it. The use of more accurate gage blocks in the calibration of the instrument used may reduce the overall uncertainty enough. It’s possible that by replacing the instrument used, you could get it to an acceptable level. Conversely, a completely different procedure may be a way to bring the uncertainty value down to where you can live with it.
One possibility to consider is expectations regarding uncertainty. What you expect may not be achievable in the best of calibration laboratories. Talk with the folks who do your calibration work; they should be able to advise you in this regard. You may wish to find what NIST’s uncertainty is for the same calibration work and use that as a guide. But remember that because NIST is at the top of the measurement pyramid, everyone else’s uncertainty for the same work will be greater-in the everyday world.
In gage calibration work it is not uncommon to encounter situations where the typical uncertainty for a measurement is 50% of the tolerance for a given element when the gage was new. Typical of this situation is pitch diameter measurements where the tolerances were generated long before measurement uncertainty became well known in the gage making industry. When NIST’s uncertainty falls in this range, you are not going to do better elsewhere so you’ll have to review your acceptance criteria to balance it against reality.
I got into a discussion about this subject with a man who operated a calibration laboratory. He had a well-known laser-based measuring device from a respected manufacturer, and he claimed that his laboratory had been designed by a national specialist in the field to provide the optimal environment. I was impressed.
Then I asked the telling question: “On, say, a half-inch thread plug gage, what would your uncertainty be for the pitch diameter measurement?” When he started his answer with “Well, it would probably be about...” I suspected he didn’t know what it was. And when he told me “three to four microinches” I knew he didn’t know. I suggested he get in touch with NIST to show them where they are going wrong because their uncertainty for that measurement was 80 microinches.
He was mixing up uncertainty with repeatability, a common mistake. If you were in a measurement dispute, this man’s number for uncertainty would indicate his readings for size were much better than most-including NIST-even though he couldn’t provide an uncertainty budget to support it. Values for uncertainty like this are essentially mythical guesses.
It’s difficult to resolve measurement disputes as it is, but when you’re up against a mythical level of uncertainty, there’s no hope for anyone.