Procedure for estimating measurement uncertainty

[eng.luma]

hi all
could any one help, i need to know how i can put a quality procedure for estimating uncertainty of measurments and how to assure the quality of test record??????

[Wayne]

Originally Posted by eng.luma

I need to know how ... estimating uncertainty of measurements ...

The American Measuring Tool Manufacturers Association (AMTMA) has published a book called Searching for Zero. In this book the AMTMA reports the results of round-robin measurement tests on many types of gages by state-of-the-art manufacture's laboratories. Included in the results are the uncertainty values specifically associated with various tools. These uncertainty values from their tests could be incorporated into your local procedures.

Searching for Zero is available for purchase from the AMTMA website at a very small price.

Hope this helps.

[Marc]

Thanks, Wayne!

[Louis Reimer]

Could you tell me what type of equipment are we talking about. Digital displays, Analog. and the application, temeprature, vaccuum, weight, Flow etc. I may have a UC prodedure that I wrote a few years ago.

[Hershal]

Based on your location and organization it appears you are a test lab.....which means that uncertainty works different than it does for calibration.....

That is not a bad thing, it is simply the reality of uncertainty.

One point to consider.....in some fields of testing, uncertainty is valid and will give useful information.....in a field such as agriculture the numbers may have any value, which means they tell you nothing.....but you still need them for accreditation.

First, the procedure should be general in the approach and refer to other sources for specific calculations. Next, you should develop a method of identifying the sources of uncertainty and quantifying them. The quantification will change depending on what you are dealing with.

As an example, if you are testing wood the test results from a moisture content and density test will vary by species and method used. The actual test results - whatever they may be - are your Type A, or random, contributions. Typically for a test lab this is the primary driver of the uncertainty numbers.

List all the Type B, or systemic, contributions.....you will find that calibration of your instrumentation has almost no input whatsoever in a test method, but must be accounted for. Quantify the uncertainties that you can.....document the contributions that cannot be quantified so you have addressed them.....some contributions may be easier with an arbitrary assigned value, just make sure it is realistic.....

Then perform the calculations as a calibration lab would.....for your Type A, if there is only one reading, unfortunately it is the number so a standard deviation may not be possible.

Hope this helps.

Hershal

[eng.luma]

well the lab that am working on it to get the accreditation for; is a plant and fertilizer analysis laboratory; they use spectrophotometer, flame photometer, pH meter, Atomic absorbsion spectrometer, automatic kjeldahl (digestion, distillation and titration ) unit for total nitrogen analysis.

[Louis Reimer]

If you are using recognized test methods like ASTM that specifies limits to the values of the Major sources of UC and specifies the form of presentation of calculated results, the test lab is considered to have satisified the requirements for estimation of UC measurement.

The above statement assumes that your lab is not performing any in-house calibrations on your test equipment. The accredited lab that you are having calibrate your test equipment should be providing you with data and UC related to the calibration.

If you are peforming calibrations in-house, you will be required to apply the UC of the standard to the calibration your are performing. ISO 5725 should be able to provide additional assistance in the UC issue.

Since I'm not a math expert, those that are have given me some useful formula's.

The device under test resolution X 0.29 then divide by 1.7321 then square root it will provide you with the contributing error related to the resoultion of the device. You would do this for accuracy of the standard used for calibration, any other contributing errors, such as evironmental factors.

With the Standard uncertainity that is reported to you, will in most cases be stated as expanded UC of test = K2 (95% confidence factor). You will need to convert this number to a K1 factor and apply that factor then divide by 1.7321 and square it for the contributing error associated to the UC of measurement of the standard.

Again, it is recommended that you review ISO 5725 that comes in four parts so you can better understand UC requirements and how and if it needs to be applied to your operation. This knowledge will allow you to better commuicate with your accredition assessor related to UC measurements.

[Hershal]

One final point on this.....as a test lab, you do NOT have to report uncertainty unless requested by customer or obligated by regulatory authority.....the advantage of a test lab compared to a cal lab!

Hershal