Determining Calibration Uncertainty with Unknown Reference Standard Uncertainty

howste · Feb 11, 2011

I'm working with a company that wants to calculate the uncertainty of their calibrations for conformity with ISO 17025. Their reference standards are calibrated by the State Department of Agriculture. The calibration certificates for the reference standards don't show any uncertainty values or as-found/as-left values.

Is it possible to calculate the uncertainty of the company's calibrations without the uncertainty data from the reference standards?

:thanks:

Stijloor · Feb 12, 2011

howste said:
I'm working with a company that wants to calculate the uncertainty of their calibrations for conformity with ISO 17025. Their reference standards are calibrated by the State Department of Agriculture. The calibration certificates for the reference standards don't show any uncertainty values or as-found/as-left values.

Is it possible to calculate the uncertainty of the company's calibrations without the uncertainty data from the reference standards?

Any ISO 17025 Lab people who can help?

Thanks!!

Stijloor.

howste · Feb 12, 2011

I've thought that perhaps I could just use the tolerance of the reference standard, and treat it as a rectangular distribution. Would this work? If I do it that way, would I also have to account for drift, or would that already be included in the tolerance distribution?

So many questions...

Hershal · Feb 18, 2011

I agree with you, however, you need to review whether other distributions are more appropriate such as triangular and whether the calculation are right such as square root 12.

Then you can answer with a bit less uncertainty.

BradM · Feb 18, 2011

howste said:
I'm working with a company that wants to calculate the uncertainty of their calibrations for conformity with ISO 17025. Their reference standards are calibrated by the State Department of Agriculture. The calibration certificates for the reference standards don't show any uncertainty values or as-found/as-left values.

Is it possible to calculate the uncertainty of the company's calibrations without the uncertainty data from the reference standards?

I think you would need to develop some estimation of the uncertainty of the reference standards. Just some ideas:

Collaborate with other testing labs.
See if there are any industry standards
Contact N.I.S.T. and see if they have any suggestions

I cannot conceive the dept. of agriculture would not have any available tolerances/uncertainties of their assessments.

howste · Feb 18, 2011

Hershal said:
I agree with you, however, you need to review whether other distributions are more appropriate such as triangular and whether the calculation are right such as square root 12.

Then you can answer with a bit less uncertainty.

Thanks Hershal. I'm reading in the UKAS M3003 document that a triangular distribution is used when combining two rectangular distributions. If this is the case, which two rectangular distributions would be included? Are they the 1) unknown as-left value and the 2) unknown uncertainty of the standard? I would assume that the uncertainty of the standard would be normal.

howste · Feb 18, 2011

BradM said:
I think you would need to develop some estimation of the uncertainty of the reference standards. Just some ideas:

Collaborate with other testing labs.
See if there are any industry standards
Contact N.I.S.T. and see if they have any suggestions

I cannot conceive the dept. of agriculture would not have any available tolerances/uncertainties of their assessments.

After some phone conversations with the Dept. of Agriculture, I found that they do have the uncertainties. Apparently they only report the as-found, as-left, and uncertainties on the certificates if the customer asks for them.

I just received an email with their uncertainties which I'll use instead of trying to estimate them based on the tolerances.

BradM · Feb 18, 2011

howste said:
After some phone conversations with the Dept. of Agriculture, I found that they do have the uncertainties. Apparently they only report the as-found, as-left, and uncertainties on the certificates if the customer asks for them.

I just received an email with their uncertainties which I'll use instead of trying to estimate them based on the tolerances.

Excellent! :agree1:

Thank you for letting us know.

Yea, we love the "Oh, but you didn't ask for it" game.

howste · Feb 18, 2011

OK, one more quick question: If two or more reference standards are combined for the calibration at one value (e.g. weights), do I add the uncertainty of each standard as a separate component in the calculation?

This seems like the way to do it, as it would be similar to stack-up tolerances of an assembly.

Determining Calibration Uncertainty with Unknown Reference Standard Uncertainty

howste

Thaumaturge

Stijloor

howste

Thaumaturge

Hershal

Metrologist-Auditor

BradM

howste

Thaumaturge

howste

Thaumaturge

BradM

howste

Thaumaturge

Similar threads