Is there a simple way to determine or express the measurement uncertainty of a device, such as a 1% pressure gauge, 0-100 psi with .5 psi sub-divisions, calibrated against a dead-weight tester with a tolerance of .02% of reading.

I referred to the NIST publication on Measurement Uncertainty and my head hurts. This is more informmation than anyone on this planet needs.

I don't know the answer, but your last sentence is well put...

There has been some chatter about Measurement Uncertainty. I have "Determining and Reporting Measurement Uncertainties", Recommended Practice RP-12, April 1995 from the National Conference of Standards Laboratories.

And I found the below at
http://www.iso.ch

Look for: Guide to the expression of uncertainty in measurement

When reporting the result of a measurement of a physical quantity, some quantitative indication of the result has to be given to assess its reliability and to allow comparisons to be made. The Guide to the expression of uncertainty in measurement establishes general rules for evaluating and expressing uncertainty in measurement that can be followed at many levels of accuracy and in many fields.

1995, 110 p., price group L, CHF 88,50, ISBN 92-67-10188-9

Can some of you help by citing other publications which address the issue?

Also - any 'Practical' comments on MU?

[This message has been edited by Marc Smith (edited 01-15-99).]

Marc,

Is "Determining and Reporting Measurement Uncertainties", Recommended Practice RP-12, April 1995 from the National Conference of Standards Laboratories." you mentioned above available on the web. I would like to see a copy for some research I am doing.

Regards,
Don

Not that I know about. I bought a paper copy a while ago. But then again, I haven't searched for it on the web.

You might want to try this as well.

http://physics.nist.gov/cuu/Uncertainty/basic.html

Regards,
Don

Excellent links, Don! Thanks!

Thanks, Marc, I do my best. No, seriously, I am still researching my MEA and R&R paper and stumble upon these during the search. If any more pop up, I will advise.

Regards,
Don

I'm here to learn and you're providing me with good stuff!

Thanks and you to me as well. The R&R paper mentioned above is currently being proof-read (by a very nice IT lady) and should be ready in a couple of days. When I have it finished, I will forward for your review. If you deem it worthy (?), you can post after your review. Anything else I come across, I will be sure to keep the group informed.

Regards,
Don

Thanks Don, These links are a big help.

David

So far, we have been doing internal calibration and just defining the acceptance criteria. Due to the recent ISO 9000 Quality System audit by the local certification body,they say that we have to state the measurement uncertainty each time we perform calibration based on the clause 4.11. I am trying to seek the correct and simplest method of calculating the uncertainty.I would appreciate if somebody could help me.
In addition to that, how do we know that the measurement uncertainty is acceptable or not.
Thanks

I have been told by statistical experts to do a gage r & r but I don't know what to do from there. What does that tell me?

A search of this forum found this, in no particular order:

http://Elsmar.com/ubb/Forum1/HTML/000089.html
http://Elsmar.com/ubb/Forum1/HTML/000144.html
http://Elsmar.com/ubb/Forum4/HTML/000036.html
http://Elsmar.com/ubb/Forum4/HTML/000047.html
http://Elsmar.com/ubb/Forum4/HTML/000073.html

Regards,
Don

[This message has been edited by Marc Smith (edited 03 April 2000).]

A Gage R&R (or Gage R where automated) tells you how much variation is in your measurement system as a per cent of tolerance used.

There is:

Equipment (Gage) Variation (EV), and
Appraiser Variation (AV)

Let's say your EV is 10% and your AV is 15%. You have 25% of your tolerance taken up in variation.

You also have to consider Measurement Uncertainty - kinda a black art right now.

Thanks for the search forum and all the input gave. I would appreciate if anyone has additional information.

NIST Technical Note 1297, "Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results," can be downloaded off their site. It addresses such subjects as the classification of uncertainty components, combined standard uncertainty and expanded uncertainty. Appendix E is particularly helpful.

Ben Royal

NIST is at: www.nist.gov/ (http://www.nist.gov/)

Also at NIST, check out: www.physics.nist.gov/cuu/Uncertainty/index.html (http://www.physics.nist.gov/cuu/Uncertainty/index.html)

[This message has been edited by Marc Smith (edited 17 July 1999).]

I have been through all these websites and I still am unsure how I am going to express this to a QS Auditor in 3 days for Assessment Registration. Maybe because I'm blonde, but I need down to earth language so I can explain how we are doing it down to earth-I am sure to take a hit for this. Thanks for the help!

NIST covers uncertainty in a technical note. Check http://www.physics.nist.gov/Pubs/guidelines/preface.html

I liked this from:
http://purdue.edu - Link was: /~nocera/project/tutorial

----------snippo-------

Measurement Error and Uncertainty

The difference between the measured value and the true value of an object of measurement is known as the error. The actual value of an error can never be known exactly, only estimated. An uncertainty statement provides an estimate of the possible error in a measurement. The term "error" is almost obsolete in the measurement arena due to it's ambiquity. It has become fashionable in the scientific community to provide uncertainty statements with all measurements due to the useful information it provides. Those still in the habit of providing error statements as opposed to uncertainty statements are looked upon with scorn and ridicule by scientists, engineers, and other technically intellectual types in the scientific community. By utilizing this tutorial, the user shall avoid subjecating himself or herself to terms such as neandrethral-like, doltish, or flat incompetent by the aforementioned persons.

As an example, let's say you designed an experiment to measure the output voltage of a given circuit. The theoretical output of the circuit is 5 volts. You measure 5.02 volts and publish a paper in which you claim the actual value is 5.02 volts with an error of 0.02 volts. You have now opened yourself to flood of spoken and unspoken questions and criticism pertaining to your own personal value and competence in the technologically advanced world.

However, had you published the same paper and claimed that the measured value is 5.02 volts, +/-0.03 volts at a 95% confidence level, and proceeded to list the involved uncertainty factors with estimates of contributions from each as well as thoroughly describing your method for obtaining and combining these factors, you would certainly be praised and hailed as intellectually advanced, equipped with a desirable skill set, and en vogue with all the current trends in the world of measurement science.

Components of an Uncertainty Statement - For purposes of this tutorial, it is assumed that any measurement values used to calculate an uncertainty value have been obtained from a population having a normal distribution. Non-normal distributions require additional attention which is not covered in this material.

Have any of you folks reading this thread actually done any uncertainty calculations???

Or do you read the manufacturers tolerance like the following:

Subject: Re: Q: Certificate of Calibration/Guerra/Naish
Date: Mon, 26 Oct 1998 10:50:29 -0600
From: ISO Standards Discussion

From: PNaish@aol.com
Subject: Re: Q: Certificate of Calibration/Guerra/Naish

I am surprised that you would get a non conformance when it comes from the manufacturer. I have worked with a number of companies and at least 9 registrars who have not given any of the 75 companies we have worked with a non conformance for this.

As far as an accredited lab versus a manufacturer being better, there are good and bad in both. But some manufacturers such as Starrett are accredited by Navlab here in the USA.

When approaching the amount of information that is required, if you get a copy of guide 25 it is easy to understand and helps with knowing what is needed for calibration. However, another approach is to obtain a copy of the ISO certificate from the manufacturer as well as any other certification they have. If you have an internal supplier qualification form have them complete it as well. The using the supplier qualification process and history from previous performance (as demonstrated the first time you sent the equipment out for calibration or the hard data sent by them) you can place them on a dock to stock plan with their certificate of conformance along with their catalogue or data sheet as meeting the requirement until you send it out the first time for calibration.

The uncertainty of the instrument is based upon the catalogue or data sheet which will tell you it is within a tolerance. As long as that tolerance provides you with the accuracy you need, and they have provided a certificate of assurance that they have met it and that it meets international standards you should be able to pass your audit. If you ask for them you can sometimes get a copy of the procedures used by the manufacturer so as to show your registrar the method the manufacturer is using.

I can also send you a sheet with the information that you should be seeing on the certificate after you send it out. You can email me direct if you would like a copy.

Phyllis Naish at Pnaish@aol.com

Mrc,
Is "Determining and Reporting Measurement Uncertainties", Recommended Practice RP-12, April 1995 from the National Conference of Standards Laboratories." you mentioned above available on the web. I would like to see a copy for some research I am doing.

Regards,
Don

Don:

Do you still need a copy of this?

I have found an uncertainty calculator (software type) at the WEB site of a local magazine published here in San Diego. The title is Uncertainty Calculator V2.5 from Compaq Computer Corporation's, Corporate Metrology Department. It is freeware!! I have not had the chance to try it yet but I figure that almost anything would be better than trying to figure out what the NIST 1297 or the RP are attempting to teach me. The magazine is CAL LAB and they are on the WEB at www.callabmag.com (http://www.callabmag.com) The programs are under the downloads tab. Look the site over as there is a lot of good info. I heard a rumor that some company used the program during an audit and the results were accepted by the registar. Only rumor...can not confirm.

[This message has been edited by Bob Hughes (edited 24 September 1999).]

Sorry that site is www.callabmag.com (http://www.callabmag.com)

I coped this off the listserve. Any takers here?

-------snippo--------

From: Charley Scalies
Subject: Q: Calibration Issues for Small Firms /Scalies

Measurement Uncertainty
How do/can very small firms, with no in-house technical expertise in metrology, comply with the ISO90001, 4.11.1 requirement to be sure inspection, test and measuring equipment is "..used in a manner which ensures that the measurement uncertaintly is known..." Alternatively, how would you describe "measurement uncertainty" to a lay person?

Accuracy vs. Precision
How would you describe the difference to a lay person?

Charles J. Scalies

Refer to "Fundamentals of Dimensional Metrology" 2nd edition - Ted Busch, Wilke Brothers Foundation, Delmar Publishers.
This is a very good book for the lay person.

Uses a very good example comparing accuracy,precision and reliability to a target.
Example; you fire 5 shots at a target:

1- Accuracy - did you hit the target?
2- Precision - How close are the five shots grouped together?
3- Reliability - Can you do this consistently.

Note also that there can be varying combinations of accuracy,precision and reliability.

Measurement uncertainty Maximum error that may be added (algebraically) to a reading.
i.e., systematic errors, random errors, human errors . . . .

[This message has been edited by Sam (edited 27 October 1999).]

From: "Thomas David Nichols" <helios@swcp.com>
Subject: RE: Q: Calibration Issues for Small Firms
/Scalies/Nichols

Suppose you have a digital thermometer that displays two decimal places, and the last digit is stable. That is, it doesn't flicker unpredictably 3-7-6-2-9-4, etc. Then the precision of the thermometer is 0.01 degree, or plus/minus 0.005 degree.

Now suppose you use this thermometer to measure a calibration standard that is known to be within 0.1 degree of 50 degrees. The thermometer's readings vary between 46 and 52 degrees, with a mean of 49 degrees and a standard deviation of 1.5 degrees. This tells you the accuracy of the thermometer.

To summarize, precision is how well you can read the instrument, while accuracy is how close the reading is to the true value.

Measurement uncertainty is a little harder to pin down. Basically, it is the size of the band around a measured value that is known to contain the actual value with some probability, often 90% or 95%. Expressed a different way, it is the band around the true value that will contain the stated percentage of a large number of measurements of the value. It combines all the sources of uncertainty: instrument accuracy, operator errors, the relationship between what is available to measure and what is desirable to measure, the time required to take the measurement compared to the rate of change of the thing measured, the effect of the measurement on the thing measured, and possibly many other factors.

Measurement uncertainty, in turn, has an uncertainty of its own. Your process can tolerate a large error in some measurements, as when room temperature can range from 40 to 120 degrees F, and it never goes below 60 or above 80. In this case, a very simple test can show that your thermometer is good enough, and you really don't need to know exactly how accurate it is. On the other hand, if you are polishing a mirror to a flatness of .05 wavelength of red light, your whole business may depend on how well you can measure flatness.

Either way, though, if the uncertainty in a measurement doesn't matter, the measurement itself doesn't matter.

Thomas David Nichols

Ah! The Hubble space telescope mirror problem?

Reminds me of the metric conversion problem with the Mars probe.

Subject: Re: Calibration Issues for Small Firms /Scalies/Everson/Kozenko
Date: Fri, 22 Oct 1999 15:01:05 -0600
From: Moderator

Subject: RE: Calibration Issues for Small Firms /Scalies/Everson/Kozenko
From:

> Shooting arrows at the target and getting them close the
> bullseye is the accuracy.
>
> Precision is if I am able to have the arrows hit the same small area
> everytime, but they made still be away from the
> bullseye....I was precise, but not accurate..

Mine's similar, but I believe the explanation called for is "degree of precision."

Picture a car tire swinging laterally on a rope, over a 30 foot arc, ten

yards in front of you.

Accuracy would be throwing a football at the swinging tire and at least hitting the darn thing. Anything is accurate if it hits the target. Remember the expression: "Close doesn't count, except in horse shoes and hand grenades." <g>

Diagonally from tire rim to tire rim it's usually about 15 inches (... and my contract with NASA expired so I won't convert that to cm's <G>). The girth of a spirally-thrown football is about, oh, eight inches, so the degree of precision of a throw that goes through the moving tire could be expressed as the required degree of precision of plus or minus one girth. (In other words, in order for the football to clear the center of the tire, it needs to be precisely on center, or no more than one girth-thickness in any direction away from center, along the same plane created by the tire.)

Degree of precision is usually expressed as a set value with a plus-or-minus dimension, but there are cases when it may be expressed as Plus 0 minus X, or Plus X minus 0, or variants in between.

For example, when reaching for a description of the degree of precision in throwing a water balloon, one would rather have a tolerance of Plus 0 so that the splash still would hit the target, which usually happens even at Plus 0 minus 5 feet. Hitting the "20" on a dart board, however, would call for a vertical Plus 9 inches Minus 0 if the bulls eye were considered the "anchor" for the measurement. This one's a complicated example too, because the "20" on a dart board is pie shaped, so the horizontal (side to side) tolerance of the shot increases as the vertical goes from 0 to 9 inches.

Clear as mud?

David Kozenko

-------snippo--------

[This message has been edited by Marc Smith (edited 28 February 2000).]

I wonder how to actually set the acceptance criteria for an IMTE.

Let say a thickness displacement meter used to measure 0.80+/-0.0125mm.
(Measurement uncertainty for the IMTE is +/-0.0004mm)

Question:
Is it correct if I set the acceptance criteria (Acceptable uncertainty) for the IMTE as 30% of 0.0125mm?

Thanks in advance.

From: ISO Standards Discussion <jennejohnn@UWSTOUT.EDU>
Date: Fri, 25 Feb 2000 16:00:25 -0600
Subject: Re: Q: Measurement Uncertainty /Enop/Martins

From: "Frans J.C. Martins" fjcm@icon.co.za

Hi Glover and group,

> From: Pfscott2@aol.com
<snip>
> What is meant by "measurement uncertainty" as in 4.11 General 4.11.1
> last sentence?
> Thanks!
> Glover Enop

Very briefly, measurement uncertainty can be explained as follows: In reality no measurement could be absolutely correct. There is always some doubt present regarding the true value of the measurement. This is generally caused by all sorts of factors, for example, drift, temperature, repeatability etc. Example: A digital gauge indicates a fluctuation of 0.999, 1.000 1.001, 0.999, 1.001, 1.000, The "true" value could be calculated by taking the average of all the readings, and then rounded to the correct resolution, in this case, 1.000, but, depending on the bias, the reading could just as well have been either 0.999 or 1.001! Thus the reported indication is 1.000 ± 1 LSD (least significant digit) In this example we only looked at one factor, which is resolution. Other factors previously mentioned also need to be taken into account to determine the total uncertainty of measurement. The components contributing to the total uncertainty can be classified as either Type A (determined statistically) or Type B (determined by other means other than statistically) All A and B components are then added through a process of Root Sum Squaring to determine the total uncertainty of the measurement.

The total uncertainty of measurement MUST be reported for any given quantity on a certificate of calibration to in order to satisfy the requirements of ISO Guide 25 (ISO 17025)

Hope this clarifies your query.

Regards in Quality
Frans J.C. Martins

---------snippo--------

From: ISO Standards Discussion <jennejohnn@UWSTOUT.EDU>
Date: Fri, 25 Feb 2000 16:07:38 -0600
Subject: Re: Q: Measurement Uncertainty /Enop/Pfrang

From: pfrang@nicolet.com (Doug Pfrang)

Section 4.11.1, last sentence, states: "Inspection, measurement, and test equipment shall be used in a manner which ensures that the measurement uncertainty is known and is consistent with the required measurement capability."

This means that if your specification requires a measurement to be within a certain tolerance -- say, 0.1 inches -- then your measurement equipment and measurement process must be capable of at least that level of certainty. It does no good to measure something with a 0.1 inch tolerance if your measurement equipment is only good to +/- 0.5 inches.

-- Doug

See also:
and http://Elsmar.com/ubb/Forum4/HTML/000036.html
and http://Elsmar.com/ubb/Forum4/HTML/000047.html
and http://Elsmar.com/ubb/Forum4/HTML/000072.html
and http://Elsmar.com/ubb/Forum4/HTML/000076.html
and http://Elsmar.com/ubb/Forum4/HTML/000120.html

Edited 26 November 2001

I can't re-lookup and change every old link to posts in the old forums. Most of them *should* still be there. If you want to find the post in the New forums, if the link to the thread in the old forums works (most of them should...), look at the thread (topic) title and what forum it is in. Then back here in the New forums - go to that forum and look for the thread topic title - OR - do a Search for the key words from Title (Note - you can search entire threads or just the 'subject' or 'title' - if you look in the Forums search page you'll see the options.

Call me lazy... :rolleyes:

When I receive new calibration certificates, I do something like a validation of the results: total uncertainty + max. error (data from the certificate) must be smaler than a criteria stabilished for each kind of instruments.
Sometime ago, an auditor charged me if I was adding to this total uncertainty, the uncertainties of standards used for calibration. But I'm not, because I believe this uncertainties won't cause influence in my measures at production.
Can anybody tell me if that is the correct procedure??
Thanks,
Yochio

As I understand it uncertainties are additive. But if you have one uncertainty item that is a factor of say 10 or 100 of the others I would say what you did - it's not significant to the total.

This may help you some:


Uncertainty Budget for Thread Pitch Diameter Calibration

Reproducibility: The reproducibility was calculated from thirty readings (10 each from three technicians) utilizing gage blocks, thread wires, super-micrometer and a 1-14 UNS-2A thread set plug. The standard deviation for the reproducibility was calculated at 11.6 micro-inches (.0000116”).

Should add the uncertainty of the S-mic when rec’d - Resolution: The resolution of the super-micrometer providing the value for measurement over wires is 10 micro-inches. We assume a rectangular distribution, that is, the error is equally likely to be any value between 0 & 10u”. The guide to measurement uncertainty instructs to divide by the square root of three to convert a rectangular distribution to one standard deviation of 5.77 micro-inches.

Gage blocks: The blocks used in the calibration of plug gage diameters are grade one. In the calibration of a 1” thread plug gage we need only the 1” gage blocks. We will assume that the block value can be anywhere within the +/- 2u” tolerance of the block. We will also allow this tolerance to expand by the 1.3u” (1/2 of the k=2 value) uncertainty reported on the gage block certification. Therefore the gage block tolerance plus the uncertainty was divided by the square root of three to get to one standard deviation of 1.91 micro-inch.
2” gage block at 2.94 micro-inch (from1.1 & 4)
3” gage block at 3.64 micro-inch (from 1.3 & 5)
4” gage block at 5.25 micro-inch (from 3.1 & 6)
5” gage block at 11.7 micro-inch (from 1.1 & 4, 1.3 & 5)
6” gage block at 14.2 micro-inch (from 1.4 & 4, 3.1 & 6)
7” gage block at 15.4 micro-inch (from 1.3 & 5, 3.1 & 6)
8” gage block at 18.7 micro-inch (from 1.3 & 2, 1.3 & 5, 3.1 & 6)
9” gage block at 20.5 micro-inch (from 1.1 & 4, 1.3 & 5, 3.1 & 6)
10” gage block at 24.1 micro-inch (from 1.3 & 2, 1.1 & 4, 1.3 & 5, 3.1 & 6)

Thread wires: Wires used were supplied at size with an uncertainty value of 12 micro-inches. As this value is reported at two standard deviations will bring to one standard deviation for the completion of this budget.


Force applied: A reproducibility study was conducted on the effect of force in relation to the digital values displayed on the super-micrometer. The study consisted of three operators with ten readings each. This test produced a standard deviation of 5 micro-inches.

Anvil parallelism: The super-micrometers anvils have a parallelism requirement of 20 micro-inches of which is routinely verified through calibration. Assuming the worst condition and a rectangular distribution, we can arrive at one standard deviation of 11.55 micro-inches.

Temperature: The gage blocks used are made of chromium carbide which carries a coefficient of expansion that is approximately 4.7 x 10-6. The majority of thread plug gages cycling through our metrology lab are made primarily of steel which carries a coefficient of expansion that is approximately 6.4 x 10-6. For the purpose of this study we will use the 1” block size. We will assume that all components have been stabilized to the room environment of 68o +/- 1o F. As the temperature sensing device has it’s own uncertainty of 05 o, we will use 1.05o temperature error. In this scenario we receive an error of 4.9 micro-inches on the block and 6.7 micro-inches on the thread plug gage. Using the difference of we can provide a rectangular distribution and arrive at one standard deviation of 1.04 micro-inches.
2”, 1.05 o change = 3.5 with one std. dev. of 2.0 micro-inch
3”, 1.05 o change = 5.4 with one std. dev. of 3.1 micro-inch
4”, 1.05 o change = 7.2 with one std. dev. of 4.2 micro-inch
5”, 1.05 o change = 8.9 with one std. dev. of 5.1 micro-inch
6”, 1.05 o change = 10.7 with one std. dev. of 6.2 micro-inch
7”, 1.05 o change = 12.5 with one std. dev. of 7.2 micro-inch
8”, 1.05 o change = 14.3 with one std. dev. of 8.3 micro-inch
9”, 1.05 o change = 16.1 with one std. dev. of 9.3 micro-inch
10”, 1.05 o change = 17.8 with one std. dev. of 10.3 micro-inch


Uncertainty Budget for Thread Pitch Diameter Calibration

All Values expressed in micro-inches

Source ofUncertainty Standard Deviation Limits Uncertainty“u” “u2”
Reproducibility (A) 11.6 134.56
Gage blocks 1.3 2 1.91 3.648
Thread wires 12 6 36
Resolution 10 5.77 33.293
Force (A) 5 25
Anvil parallelism 20 11.55 132.25
Temperature 1.8 1.04 1.082

Sum = 365.833
Combined uncertainty, uc = 19.127
Expanded uncertainty, U = 38.3
2” pitch diameter Expanded uncertainty, U = 38.7
3” pitch diameter Expanded uncertainty, U = 39.2
4” pitch diameter Expanded uncertainty, U = 40.3
5” pitch diameter Expanded uncertainty, U = 45.8
6” pitch diameter Expanded uncertainty, U = 49.0
7” pitch diameter Expanded uncertainty, U = 51.0
8” pitch diameter Expanded uncertainty, U = 55.8
9” pitch diameter Expanded uncertainty, U = 58.9
10” pitch diameter Expanded uncertainty, U = 64.7

Ah! A wizard! The detail is appreciated! I hope you stop by often!

Is there some way to determine the uncertinty right off the Gage R & R calculations?
And how would you suggest we state it in every work instruction? Thanks!!!!!

Measurement uncertainty, as it applies to calibration, is a bit more than a gage r&r. I see that a few posts have already pointed you to the method, but I also saw the word "easy" in there somewhere.

To this end, there is a FREEWARE uncertainty calculator program available. It was developed at Compaq Metrology, and conforms to ANSI/NCSL Z540-2 and the GUM. The program is in two parts, to ease installation from floppy disks.

Go to http://www.proficiency.org and click the "Software Download" page. The program is "Uncertainty Calculator 2.5".

The program will not teach you the ins and outs of uncertainty budgets, such as proper distribution for contributors or effective degrees of freedom, but it does make the calculation process very simple once you do have the knowledge.

I hope this helps.

Ryan Wilde

Originally posted by Marc Smith:
As I understand it uncertainties are additive. But if you have one uncertainty item that is a factor of say 10 or 100 of the others I would say what you did - it's not significant to the total.

Uncertainties are a variation of additive. They are combined in a mind-reeling Root-Sum-of-the-Squares sort of fashion.

The general rule for uncertainties is the factor of 10. If the particular contributor is less than 10% of the total uncertainty, then it is considered unimportant.

Ryan Wilde

I've read a few posts in this forum regarding measurement uncertainty. The sad fact is, there is no easy way to do uncertainty, unless you are a statistician moonlighting as a metrologist. The process is very statistics laden, with interpretive distributions, degrees of freedom, yada yada yada.

I suggest taking a course. Not one of those week-long courses, just an introductory course. I took one, and like most things, it is relatively simple after you learn how. The first step is a big one, but after you actually undertand the concepts, it makes sense and isn't difficult. It is, however, time-consuming.

Best of luck!
Ryan Wilde

Hello,

While calculating Uncertainty of Measurement, we take into account the various sources of uncertainty. For example, Uncertainty due to resolution, due to repeatability & reproducibility , due to Standards used, due to temperature, and so on.

I wonder if there exists a standard (if not exhaustive) list of the uncertainty sources and their frequency distributions (Rectangular/Triagular...) that must be considered for various types of instruments and gages.

Any pointers?

Thanx,

-Atul.

Here you go Atul, here is a tad of what is generally accepted. These are dimensional examples, because that is what I generally deal with.

- Repeatability and Reproduceability is a normal distribution. They are statistically derived, therefore they are calculated to a normal distribution.

-Temperature is sinusoidal in nature, therefore it is a "u-shaped" distribution.

Standard uncertainty - This is also a normal distribution. Actually, if your standards calibration was properly reported, it is two standard deviations.

Standard / M&TE accuracy - If all you have to go on is a tolerance, then you have one of two options. Usually, you would use a rectangular distribution, because there is an equal probability that the actual accuracy is anywhere in that range. The exception is VERY STABLE GAGES from extremely reliable manufacturers. If you know beyond a shadow of a doubt that your equipment was adjusted to nominal, and it has little to no drift, then you would use a triangular distribution. I have never had the opportunity to use this distribution.

Thermal expansion uncertainty - This is added because although material expansion is characterized, it is not perfect. Example: The accepted formula for steel growth is APPROXIMATELY 11.5µm/m/°C. Due to metallurgical imperfection, there is uncertainty involved in this number. I use 10%, because I tend towards the conservative side. This 10% would be expressed in a rectangular distribution, because you really have no idea where in that 10% the actual growth is.

I hope this gives you just a bit more insight. Deciding on a distribution is somewhat subjective. You have to visualize how the error is likely to occur. Once you can do that, it gets to be quite easy to choose an appropriate distribution.

Ryan Wilde

Here is a website metrology forum (HP equip)that provides some down to earth explanations on measurement uncertainty (even for blondes as user Dawn has requested). In addition they give some info on how these concepts are generally implemented at their facility. http://metrologyforum.tm.agilent.com/basics/uncert.html

Does anyone have an easily understandable step by step method for calculating measurement uncertainty for gage calibration items, i.e. micrometers, calipers, etc.? I have an example of one conducted on a thread plug gage that says WHAT was done, but not HOW it was done. NIST info is like reading an ancient elvin tongue. It gives me a headache. I'm an analytical learner and as such, I need a step by step to grasp the concept and find out the why's. Pleases Help. Dave.

i have read a lot about measurement uncertainty and have quite a few posts and such posted about measurement uncertainty. See the uncertainty directory in http://Elsmar.com/pdf_files/

I learn like you do. All I can say is I have not fould one 'simple' article. I think I can explain it, but I don't have to do the calculations and such so I have no hands-on experience in determining uncertainty per se. However - you do have to look at what uncertainty is - for example, in an R&R, what is uncertainty? In Gage R&R it's the confidence limits. You buy a thermometer and the instructions say it's accurate to +/-2 degrees - your uncertainty is 4 degrees (as I understand it). When you get to an uncertainty budget, it's the combination of all uncertainties of the entire system.

Comments from others??? I'd like to hear a good simple explaination as well.

As another reply says, measurement uncertainty is not an easy subject. Even when calibrating the same type of instrument, YOUR uncertainty will be different from anyone else's because you are in a different facility and using different standards. However, a working understanding of measurement uncertainty is necessary for a calibration lab. And, of course, it is required by ANSI/NCSL Z540-1, ISO 17025, and other standards.

Here are a few publications I have found useful. I have tried to list them in order of increasing difficulty.
"An Introduction to Error Analysis" by John R. Taylor (second edition, University Science Books, Susalito CA: 1997. ISBN 0-935702-75-X) This is an excellent book that starts off easy and gradually works into more difficult areas.
"Calibration: Philosophy in Practice" by Fluke Corporation (second edition, Fluke Corporation, Everett WA: 1994. ISBN 0-9638650-0-5) Although geared to electrical/electronic measurements, the principles are usable in every measurement area. Chapter 22 is about uncertainty statements, but first you will need to read either chapters 20 & 21, or the book by Taylor.
"Determining and Reporting Measurement Uncertainties", National Conference of Standards Laboratories Recommended Practice # RP-12. (NCSL, Boulder, CO: 1995) Somewhat easier to understand than the ISO standard (next), but still very technical. This includes a few practical examples writtem by metrology managers.
ANSI/NCSL Z540-2, "American National Standard for Expressing Uncertainty -- U.S. Guide to the Expression of Uncertainty in Measurement". (NCSL, Boulder, CO: 1997) This is the official US adoption of the ISO guide: very technical, but the authoritative standard.

The NCSL publications are available from the organization (www.ncsl-hq.org). The Fluke book is available from your Fluke sales representative. The Taylor book can be ordered through any major online or traditional book retailer.

Please feel free to contact me directly for additional information.

Graeme C. Payne
ASQ Certified Quality Engineer
Graeme@asqnet.org

in clause no 4.11.1, calibration, it does mention that the measurement uncertainty shall be known and consistent with the measurement capability.

what will be the evidence to proof to the 3rd party auditors in case they ask this question ?

Good question. Any takers?

I would assume the calculations - like a spreasdsheet.

YKT said:
in clause no 4.11.1, calibration, it does mention that the measurement uncertainty shall be known and consistent with the measurement capability.

what will be the evidence to proof to the 3rd party auditors in case they ask this question ?

I note that the original question (by YKT in December 2002) refers to section 4.11.1 of ISO 9001:1994, which is canceled and replaced by the 2000 version. The equivalent section of ISO 9001:2000 is 7.6.
ISO 9001:1994 section 4.11.1 – "Inspection, measuring and test equipment shall be used in a manner which ensures that the measurement uncertainty is known and is consistent with the required measurement capability."
ISO 9001:2000 section 7.6 – "The organization shall determine the monitoring and measurement to be undertaken and the monitoring and measuring devices needed to provide evidence of conformity of product to determined requirements ...
Note: see ISO 10012-1 and ISO 10012-2 for guidance."

(Excuse me while I dig out my auditor hat & put it on ... OK.) :thedeal:

The objective evidence I would look for depends on if the organization is a user of inspection, test and measuring equipment (IM&TE), or if the organization provides calibration services. As most organizations are users, let’s address that first.

It is an axiom that all measurements have uncertainty. As a user of IM&TE, you need to be sure that the uncertainty does not have a significant effect on your product. You need to know certain information for each measurement you are making. Here are some of the items I would consider asking about.
What quality characteristic of the product is being verified?
What is the design nominal value and tolerances?
What is the uncertainty of the measuring system used to verify the characteristic? How is it documented? (Documentation might be as little as the instrument performance specification plus calibration records, or a full measurement system analysis plus gage R&R study, or variations in between.)
Is the uncertainty small enough that it does not significantly affect the measurement results? How is that documented? (What is the risk if an out-of-specification part is passed? What is the risk if an in-specification part is rejected? What is the sensitivity of the next process to variations in this characteristic? What function in the organization made the decision?)
Is the IM&TE used appropriately? Are the users trained? Is it always used the same way? How are these considerations documented?
Do you have any feedback that bears on the appropriateness or correctness of the quality characteristic measurements? How has that affected the process?
Do you have a company guide specifying a minimum accuracy/uncertainty ratio for measurements? Is it consistently followed? How do you handle cases where the best (affordable and appropriate) technology still does not provide a high enough ratio?
Do you have any documentation to show that the frequency of calibration is appropriate?


If the organization is a calibration lab (a supplier of calibration services or calibrated IM&TE) being audited to ISO 9001, then all of the above still applies. However, considering that the "product" is the service of calibration, then all of the product-related sections of the standard come into play as well. The result is that he questions become more detailed, and even more documentation is required. In particular, I would expect to see full calibration results data for all of the organization’s calibration standards, including as-found and as-left data where appropriate. I would also expect to see evidence of use of statistical methods to assure measurement quality.

Since the original question was based on ISO 9001:1994, extending the discussion to ISO/IEC 17025:1999 is probably not appropriate. But to move to that level, the lab not only has to have an effective quality management system in place (ISO 9001), they also have to prove their capability to make the measurements in their scope of accreditation.

I guess that the R&R results ultimately give a picture of how gages and personnel using them interact and what level of "incompetance" can be attributed to the gage or the operator.

This is important in that it leads you to a possible preventive action or continuous improvement.

Calibration/verification is so far from R&R they should never be discussed together.


Dawn Posted:
Is there some way to determine the uncertinty right off the Gage R & R calculations? And how would you suggest we state it in every work instruction? Thanks!!!!!

You can by following the 10/20/30% ranges suggested in the MSA and have a record of such for every gage, or type of gage. Also in my opinion these figures do not need to be included in any work instruction. Supposidly all this is determined during the planning process and handed over to production with their appreciation of being part of the process.

Open question:

Who really needs to know about about measurement system conformance percentages?

Surely not the operator.

Measurement uncertainty is only good during the planning process when gages and fixtures are being decided upon.

I'm sure I'll make a couple of enemies with that statement, but think under the surface.:bigwave:

Can anyone tell me of a recognized source other than MSA thrid edition that goes into a simple explantation and requirements for doing gage R&R's based on the process control limits vs print limits? I been having a battle at my company where there print tolerance is .012, and they are using a .0005 resolution indicator as it meats the 10 to 1 rule. ( never mind the accuracy of the maker) I agree to that , however it is my feeling that I am trying to control a process to .005 and that for the sake of going passed the minimum requirements and taking the resolution restriction out of the measurment process and future requirements customers may have for submitting gage R&R to process not print. :bonk:

Can anyone tell me of a recognized source other than MSA thrid edition that goes into a simple explantation and requirements for doing gage R&R's based on the process control limits vs print limits?


You may refer to the book:
Concepts for R&R Studies - by Larry Barrentine
Available from ASQ Quality Press: http://qualitypress.asq.org/perl/catalog.cgi?item=H1149

Please also search the forums. This topic has been discussed before.
For Example: http://www.elsmar.com/Forums/showthread.php?t=7477

Iday38,

Just perusing this complete thread should give you copious information. Another source for GR&R and uncertainty analysis is:

http://www.measurementuncertainty.org/mu/guide

I hope this helps.

Once upon a time, in a past life, I led 2 laboratories to accreditation and developed a procedure which might explain where the measures for estimating uncertainty are derived from. Then again, it might completely confuse in which case, kindly disregard.

I also have attached TN1297 (from NIST) and two other overseas equivalents. If this ain't enough to put you to sleep, try Seconal. :)

Jon Scott

I also have attached TN1297 (from NIST) and two other overseas equivalents. ...

Jon Scott

Jon,
:thanks:
Thanks a lot, and Welcome to the Cove!

Thank you for your interest in answering my questions. I have to say that the reading I agree is very dry. I have solved my initial problem through the use of gage pak software. I will keep these documents as referance to futher understand measurment uncertaintly

Wow folks.......this thread makes me thingk that I am at NCSLI or MSC at least......

Actually, Graeme has some really good posts in here regarding uncertainty.

My opinion is to take the KISS rule.....keep it short and simple.

Example: A bench tech performs some measurement, takes his $10 calculator (the $10 one because you need square root), and calculate MU. Anything more complex is outside the realm of what cal techs and users typically need.

There are three pieces to MU, concept (the beginning, very simple), formula (the ending, very simple), and identifying/quantifying/plotting (the middle, anything BUT simple).

The concept is that MU mathematically describes the errors associated with a specific measurement of a specific quantity at a specific time/space location....and on and on with the specific bit.

The formula is a root-sum-square. Quantities are squared, added, square root of the total is taken to get to standard uncertainty. For cal labs, the multiplier k=2 is used to get to expanded uncertainty so the confidence level will vary. For test labs, the 95% confidence is the key, so the multiplier (k factor) will vary.

Now, the part in the middle is the challenge. a Gage R&R is a useful tool to determine the Type A uncertainty. Other methods also can be used. Type B is a bit fuzzier. This is usually where temp/RH/vibration (if by a RR track), sand (can be a real concern in some places), whether the tech had coffee yet.....any number of factors can be put here. The key is to identify anything that may actually affect the measurement. Sometimes quantification is easy, like with MU from your accredited third party cal lab. Sometimes an arbitrary number must be used, as is the case sometimes in the testing world. And sometimes, the influence can be either ignored or easily compensated for, so it has no effect.

An earlier post gave the link to the agilent download page where the uncertainty calculator 3.2 is, and I recommend it.

I hope I have made things a bit more clear, but suspect I have made it more muddy. But such is MU, so I have discovered.

Hershal

I feel like I'm beating an horribly injured horse here, but after having read through all these posts, and reading tons of internet information, I'm still confused on how to report the measurement uncertainty on my calibration records.

From what I've gathered, I can calculate a theoretical uncertainty:
Using the manufacturer's accuracy claim of the measurement equipment along with other factors (operator, part, methods, environment, etc.) and come up with (in my opinion) a large uncertainty. In my case it is calculated to be 1.6mm (!)

Then using the equipment, I took 200 measurements and determined the average, standard deviation and a confidence level and say that my uncertainty, with 99% confidence is +/- 0.022mm.

So, can I say that I measured it 200 times, and my measurement uncertainty is known (+/- 0.022mm) and is consistent with the required measurement capability (1.6mm)?

:confused:

Queen,

The one thing I have discovered to be absolutely certain is that given the exact same question, five "experts" will likely arrive at five similar but different answers. That you state you are confused means you are normal. I worry about those who claim to have total understanding of MU.

I repeat...KISS rule - Keep It Short and Simple!

For your example, I didn't see any of the Type B stuff, or what the item being measured should have been, plus/minus MU. That makes comment on the example a bit challenging.

The key is to take small chunks. For example, create a brainstorm list of what can affect you measurement. write down all influences, never mind at first what they are. Go back through and see if you can eliminate some of them. Then go back through and see how many can be quantified. At the end of all the studies and calculations you will likely find that one or two components are the dominant contributors.

Another thing to keep in mind.......if you are a cal lab, your uncertainties will likely be pretty low.........if you are a testing lab, that may not be true. I know some test labs that, if they could get down to 50% MU they would have dinners and have press releases. It is the nature of what those labs do. Cal labs typically have MU below 1% so it is a different world.

Put another way, do the studies, do the calculations..........but don't lose sleep over it. If you can back up what you claim, you should be in pretty good shape.

Hershal

The proper instruments, excellent training on their proper use and maeasuring techniques and calibration are the key.

Mr.Jerry,

Thanks for the help, truly appreciate it.
Advanced Christmas greetings.
I suppose others have personally asked you for similar advise before and i would truly be obliged if you can let me know in brief/point form in what is required of an approved signatory of a RF/Microwave lab under the 17025 scheme.
I have come to know that the technical requirements are more important than the quality aspects.
I know we have to know the standards(both operation and service aspects) by hard which is what i am trying to so so hard now.
And i also know that we have to knw something about the uncertainty measurement(BMC) and all?
Do you have a sample of such a calculation underlying an example say..using a measuring receiver HP 8902S in calibrating a signal generator say a HP 8648C?
I am still blur in getting the infos..accuracy, resolution and so on.
Appreciate the reply.

Rgds
Ruben

Is there a simple way to determine or express the measurement uncertainty of a device.

The American MeasuringTool Manufacturers Association (http://www.amtma.com/) has published a book called: Searching for Zero which is talked about in this thread (http://elsmar.com/Forums/showthread.php?t=16520) in the Book Recommendations Forum. The publication can be ordered at www.amtma.com (http://www.amtma.com/) or by calling (216) 241-7333.

In brief the book gives measurement uncertainties for several common types of measurement tools.

Hi Ruben,

Based on my experience as an Approved Signatory here in Malaysia, the assesors (both, technical & lead) will have their sessions with you. You have to know all the clauses in the MS/ISO 17025 + the SP series documents (which is exclusive to Malaysia only).

Afterwards, they will ask you to perform the test/calibration work for which the scope you are applying for.

And then, they will ask you about the uncertainty in the measurements that you made.

That's pretty much it. But when I had mine last time (and the Lead Assesor was not very kind, well we only have 3 lead assesors in Malaysia), by the end of the day, I felt like I could not care whether I got the signatory or not.

Well, I did-lah at the end :-)

Good luck!

To everyone,:bigwave:
Thank you very much for all the information with regard to the type of measurement uncertainty. I learned a lot.
Thanks again to everyone.:thanks:
Best regards,
raffy:cool:

i would very much appreciete it if an example of calculating MU for calibrating a temperature control device using a Fluke 741B mmultimeter with type K thermocouple wire.

thanks in advance,

gD