Measurement Uncertainty (MU) vs. Level of Certainty - ISO17025 para. 5.1.1

[Bernard Alderson]

In section 5.1.1 of the stardard ISO17025 they ask you to factor in or rate the factors like Human, sampling, conditions. In assessing the level of certainty of measurement and results. My Questions is: Is there a format that can beused for rating the level of certainty (ie - FMEA) and if so does anyone have a system that they use and would not mind sharing it?

[Statistical Steven]

Originally Posted by Bernard Alderson

In section 5.1.1 of the stardard ISO17025 they ask you to factor in or rate the factors like Human, sampling, conditions. In assessing the level of certainty of measurement and results. My Questions is: Is there a format that can beused for rating the level of certainty (ie - FMEA) and if so does anyone have a system that they use and would not mind sharing it?

I would do a measurment study with those factors designed in (DOE) then use Variance Components to assess the different sources.

[Hershal]

Couple of things.....

Measurement Uncertainty (ANS/ISO/IEC 17025 Clauses 5.1.1, 5.4.6.1, 5.4.6.2) is generally described (using Metrology methods and models) in the GUM, which is the Guide to the Expression of Uncertainty in Measurement, available as ISO GUM, or as ANSI/NCSL Z540-2-1997.....

Now.....

MU is basically divided into Type A and Type B. Type A is the statistically driven information, generally, the readings you get, that is, the measurements themselves.....

Type B on the other hand is non-statistically driven, and unfortunately has an INFINITE number of contributors. Put another way, a single system such as FMEA may not be the ultimate answer.

We may be able to provide a bit more help if you give some detail of the application you have, whether for test or calibration.

Also, the first thing you should consider is obtainng training, both in ANS/ISO/IEC 17025 and in MU (assuming you have not yet obtained training). There is a link or two in the 17025 board and the training board (posts by me and others also I believe) on such training.

Not quite the answer you sought I suspect, but more information is needed to provide a better answer.

Hope this helps.

Hershal

[Bernard Alderson]

The laboratory in question is a test laboratory, which carries out phyical testing (validation) on performance fabrics. Tests such as
- Number of ends/10cm (warp)
- number of ends/10cm (weft)
- Tensilte properties
- Tear properties
- Porosities
- Abration resistance

The factors I have isolates are:
- inturpretation of test methods
- Operator competence
- Lab contitions
- Sampling

I now need a way to evaluate the impact of those factors on the outcome of the result to ensure acceptablilty
and the reason for not seeking out training is the cost and bugetry contrains so I will have to go the difficult route.

[Statistical Steven]

Keep in mind the factors you want to isolate and study are hard to measure. How do you measure intepretation of a test method? Seems like training would reduce or eliminate much of the issues. You can study the effects of operator, lab conditions, and sampling if you want to look at differences in the response.

[Ken K]

The factors I have isolates are:
- inturpretation of test methods
- Operator competence
- Lab contitions
- Sampling

I would probably add a couple more factors:
-materials tested
-equipment used to test materials
-gages used to measure samples
-dimensions of sample pieces

But like Hershal mentioned, there can be INFINITE number of contributors. I would concentrate on those that would have the largest impact on the test results.

[Hershal]

For a test lab, you have an advantage.....ISO/IEC 17025 only requires reporting uncertainty....
1. when the customer requests it.....
2. when there is a regulatory requirement.....
3. ifthe uncertainty can push the results potentially into the fail zone.....

But uncertainty development mens you have to look at every part of the process, quantify when/where possible, and establish what the contribution is actually doing to your measurement. THAT is where a test lab actually gets some value from an uncertainty study.....

Put another way, if you are doing the uncertainty only to satisfy some AB auditor, don't bother, take the CAR and save your effort - BUT - if you are using uncertainty analysis as a process mapping and improvement tool, you are using it properly, for a test lab.

Also, for most test methods, the final number is not a source of great concern.....whatever it is, that is what it is.

Hope this helps.

Hershal

[amanbhai]

i work for the physical testing lab & have the same problem.
It is important for all of us since i don't find the person who could teach us these calcualtions.
Even tough we have calculated the MU easonably that we can justify to the assessors. I belevie.