Uncertainty ListServe - NIST TN 1297

[Marc]

CURTIS COATES WRITES:

(IM)Moderator;

This message is in reference to numerous postings on the topic of measurement uncertainty. I'm curious if the term, "measurement uncertainty" referred to in this discussion group is the same as the measurement uncertainty defined in NIST Technical Note 1297 [GUIDELINES FOR EVALUATING AND EXPRESSING THE UNCERTAINTY OF NIST MEASUREMENT RESULTS], or if there are other definitions and guidelines in use in industry.

When dealing with gage certifications from different calibration laboratories, I have found that unless specifically requested, Measurement Uncertainty usually isn't expressed at all. But when it is stated, the form runs the gamut from simply a number, to plus or minus a number, to a formula, to a statement (such as, "does not exceed 25% of the specified tolerance").

In the time I've been managing calibration, I've seen only one certificate that expressed measurement uncertainty as specified in NIST TN1297and that certificate came from a calibration lab in England!

Another question I have pertains to the practical application of the measurement uncertainty when it is "known" (or at least expressed). I seriously doubt the validity of the uncertainty statements I've seen.

For example, one posting from this discussion group stated, "The rule is simple. If the result of a measurement plus or minus the uncertainty falls within the part feature tolerance zone then the part feature is accepted."

I don't mean to quibble, but this statement raised a couple of questions in my mind:
1) Where is this rule stated?
2) Since measurement uncertainty is to be expressed as a total (according to NIST), shouldn't the evaluation be based on the measurement plus or minus *one-half* of the measurement uncertainty?

And as if I wasn't confused enough already, I asked the Quality Manager of a calibration laboratory I was considering as a supplier how he applied measurement uncertainty when calibrating a gage. His response was that he would accept a gage if any part of the measurement and uncertainty bandwidth fell within the tolerance for the gage, even if the base measurement value for the gage was out of tolerance. So who is right (if anybody), and how do I/we prove it?

[WONDER IF WE CAN GET A2LA, NIST, AND NVLAP TO WEIGH IN ON THIS?]

------------------------------------------------------------------------

[HENRIK NIELSEN WAS THE FIRST TO RESPOND TO CURTIS COATES QUESTIONS ON HOW SHOULD WE MEASURE UNCERTAINTY.]

Hi, I am part of the ISO TC 213 group, where we deal with Geometrical Product Specifications. Under the main committee we have Working Group 4, dealing with decision rules, which defines, simply put, how we accept or reject parts in light of measuring uncertainty.

In reply to [CURTIS COATES POSTING], let me offer the following:

It is not surprising that a British certificate would be in accordance with NIST TN 1297 [GUIDELINES FOR EVALUATING AND EXPRESSING THE UNCERTAINTY OF NIST MEASUREMENT RESULTS], since that note is merely the NIST [NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY] implementation of the "Guide to the Expression of Uncertainty [IN MEASUREMENT]", or GUM amongst friends, which is published by ISO and 6 other international, metrology related organizations. All laboratories under the European accreditation organization, EAL, are required to state their uncertainty in accordance with GUM.

As for the decision rules, the document you are looking for is ISO 14253-1, [GEOMETRICAL PRODUCT SPECIFICATIONS (GPS) - INSPECTION BY MEASUREMENT OF WORKPIECES AND MEASURING EQUIPMENT] which set the rule that in order to prove conformance with a specification, you have to reduce the tolerance by the measuring uncertainty and show that the part is inside that interval.

If, on the other hand, you are trying to prove non-conformance (that the part is bad) you have to enlarge the tolerance by the uncertainty and show that it is outside that interval. That leaves us with the uncomfortable situation, that there are parts we cannot prove are either good or bad, but who said it had to be easy?

With regards to the one-half uncertainty question, GUM (and I believe the NIST Note) defines the uncertainty as the half-width of the interval, so you should use the full uncertainty. Think of uncertainty as "How wrong can I be?" and it all makes sense.

Henrik Nielsen

------------------------------------------------------------------------

Dr. Steve Phillips at NIST responds as well:

Hello Dick:

A couple of short comments on MEASUREMENT UNCERTAINTY AND NIST TN 1297.

The calculation of measurement uncertainty as described in NIST TN 1297 is equivalent to that of the ISO Guide to the expression of Uncertainty in Measurement (commonly called GUM). Additionally, the ISO guide has recently become an American national standard: ANSI/NCSL Z540-2-1997. Hence all three of these documents describe the same unified approach to the calculation of measurement uncertainty.

To my knowledge, NIST does not tell customers how to use the uncertainty value in the acceptance or rejection of products. However, in ISO TC213/WG4 there is a Final Draft International Standard document 14253-1 [PART 1:] "Decision rules for proving conformance or non-conformance with specification" which gives default rules for this issue. In short, it states that for a sell to "prove conformance" the measurement result must be at least one expanded uncertainty (with k = 2) inside the specification limit. Similarly, for a buyer to "prove non-conformance" the measurement result must be at least one expanded uncertainty (with k =2) outside the specification limit.

Best Regards,
Steve Phillips
Precision Engineering Division, NIST

[MY COPY OF THE DRAFT STANDARD DEFINES K AS THE "COVERAGE FACTOR". "k: NUMERICAL FACTOR USED AS A MULTIPLIER OF THE COMBINED STANDARD UNCERTAINTY IN ORDER TO OBTAIN AN EXPANDED UNCERTAINTY (SEE GUM CLAUSE2.3.6)"]

--------------- The Source ------------------

INVITATION TO PARTICIPATE

The "gdandt" mail list and forum is for design, manufacturing, and inspection professionals and managers interested in dimensional measurement and control of discrete parts and processes.
If you work in industry, government laboratories and facilities, or academia and are interested in these topics, welcome.

Our goal is to improve knowledge and communication in dimensional design, manufacturing, and inspection. We want to invite you to join us in this moderated on-going discussion on issues of importance to our profession. Comments and submissions from members are collated and posted weekly; you will NOT be overrun with messages. Requests for urgent assistance will be handled on a direct basis and not held for the weekly distribution. We particularly welcome questions and requests for help from industrial personnel, responses to those questions, suggestions for discussion, and news items (subject to verification).

We rely on our subscribers to set the agenda and tone with the following caveats:
1)NO COMMERCIALS,
2) courtesy and comity at all times no matter how great the temptation,
3)no politics, no religion, no sex (at least not until cloning reaches production levels).

This list is maintained and monitored by MIMETEK and supports D2M2, the Dimensional Design, Manufacturing, and Measurement Association. For more information on MIMETEK and D2M2 visit ##DEAD LINK REMOVED##

If you have suggestions on how to improve this mail list (even perfection can be improved, else how could Continuous Quality Improvement support so many careers?), e-mail them ##DEAD LINK REMOVED##.

To join, just send a message to ##DEAD LINK REMOVED## with the word "SUBSCRIBE" as the subject and your name, company, and address in the body.

To escape our clutches (as if anyone would ever want to), send an e-mail to ##DEAD LINK REMOVED## with "unsubscribe gdandt" as the subject.

Join us and share your knowledge (and maybe learn a thing or two).

[Hershal]

The other replys have been really good, so I won't add there.....but will clarify the "does not exceed 25%" part of the original post.

That comes from ANSI/NCSLI Z540-1-1994 section 10.2.b, which is a restatement of MIL-STD-45662A requirements, otherwise known as the "4:1" rule.

Hershal

[Graeme]

A quick addition -- while the TN 1297 is NIST's "internal use" version of the GUM, the United States' "legal equivalent" is ANSI/NCSL Z540-2-1997, U. S. Guide to the Expression of Uncertainty in Measurement. (NCSL International, Boulder, Colorado, www.ncsli.org)

If a US calibration or testing activity is (or is preparing for being) accredited to ISO 17025:1999 then they should have a copy of either the ANSI standard or the ISO document in order to conform to requirements. I'm not sure the NIST version would be acceptable, because it is neither the official international version (ISO) or the US legal equivalent.

[Marc]

Thanks for the clarification, Graeme!