1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

All things Calibration and Uncertainty

Discussion in 'Gage Calibration and Uncertainty' started by Andy Nichols, Oct 26, 2021.

  1. Andy Nichols

    Andy Nichols Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    5,086
    Likes Received:
    2,553
    Trophy Points:
    112
    Location:
    In the "Rust Belt"
    Hello knowledgeable statisticians!

    I'm trying to understand the relationship between TUR and expressing Measurement Uncertainty and the resulting "k" factor (as reported on a cal cert).

    Can anyone point me to a good printed article or give me the "cliff notes version"?

    TIA
     
  2. John C. Abnet

    John C. Abnet Well-Known Member

    Joined:
    May 23, 2017
    Messages:
    709
    Likes Received:
    510
    Trophy Points:
    92
    Location:
    Upper Midwest- USA
    Good day @Andy Nichols . Regrettably, most of my 'knowledgeable statistician" skills have been forgotten over the past 20 years. Here is a link specific to TUR ..
    https://www.isobudgets.com/calculate-test-uncertainty-ratio/

    Attached are some publicly available documents that I collected in recent years. Hopefully these can prove helpful to you.

    (i'm sure there are others on this forum such as @Jennifer Kirley that can provide REAL help on this topic).

    Best of luck.

    Be well.
     

    Attached File(s): 1. Scan for viruses before using. 2. Report any 'bad' files by reporting this post. 3. Use at your own Risk.:

  3. BradM

    BradM Moderator Staff Member

    Joined:
    Jul 31, 2015
    Messages:
    394
    Likes Received:
    314
    Trophy Points:
    62
    Location:
    Arlington, TX
    Hershal Brewer is your best bet if you can hit him up.

    Let me butcher this a bit...

    TUR is like TAR in that the estimated uncertainty maintains a sufficient ratio to the item being calibrated. So the test accuracy ratio (TAR) is making sure you have a 4 to 1 ratio of accuracies (different than uncertainties).

    The K factor.... is the normal distribution probability (think... 95%, 99%....). So most reported uncertainties use the K=2.

    Meaning... supposed your estimated uncertainty is... .1°C with a K=2. So.... say you measure the unit under test and the values shown is 121.4°C 95% of the time the actual true value is no greater than 121.5 and no lower than 121.3.
     
  4. John C. Abnet

    John C. Abnet Well-Known Member

    Joined:
    May 23, 2017
    Messages:
    709
    Likes Received:
    510
    Trophy Points:
    92
    Location:
    Upper Midwest- USA
    Nice @BradM
     
    BradM likes this.
  5. Andy Nichols

    Andy Nichols Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    5,086
    Likes Received:
    2,553
    Trophy Points:
    112
    Location:
    In the "Rust Belt"
    In sleeping on my question, what I'm really looking for is the "story" behind measurement, calibration and uncertainty etc goes. For example:

    • I have a part diameter with a dimension "X" and a tolerance of +/- 0.001" which I need to measure, because it's considered a key characteristic and we're going to run a capability study, with a view to providing Xbar/R charting with the parts.
    • I propose using a Mitutoyo 293-340-30 Digital Micrometer, Inch/Metric, Ratchet Stop, 0-1" (0-25.4mm) Range, 0.00005" (0.001mm) Resolution, +/-0.00005" Accuracy, Meets IP65 Specifications
    • I get at calibrated at an ISO/IEC 17025 accredited lab. and request a cert with their measurement uncertainty
    • I organize an R & R study per the AIAG MSA handbook... and so on.
    What will the process look like when it comes to considering the T.U.R and the lab's Uncertainty value etc?
     
  6. BradM

    BradM Moderator Staff Member

    Joined:
    Jul 31, 2015
    Messages:
    394
    Likes Received:
    314
    Trophy Points:
    62
    Location:
    Arlington, TX
    Hey Andy!! Let me know if I'm missing this... and I know you're shy and won't tell me. LOL!!!!

    Once you get your device calibrated and they report measured values (what they found when they checked it)... You can estimate what the actual real value is by building an interval around that measured value based on the uncertainty and the K value.

    If all you do is rely on accuracies.... you don't really know quantitatively how good your measurements are. You just have confidence the error doesn't exceed a certain value.

    So in your example of R&R, having uncertainty actual real values can assist in explaining portions of your variance in measurements. Saying (really simple...) you measured ten times, got 5 inches worth of variance. However... the uncertainty of the device could have contributed up to 3 inches, so two inches is attributed to other variance (method; operator; etc.).
     
    Andy Nichols likes this.
  7. BradM

    BradM Moderator Staff Member

    Joined:
    Jul 31, 2015
    Messages:
    394
    Likes Received:
    314
    Trophy Points:
    62
    Location:
    Arlington, TX
    I realized I missed a component in this...

    Suppose we're doing temperature.
    Probe tolerance: ±.025°C
    Nominal temperature: 100 °C
    Measurement uncertainty: .0047 °C

    The tolerance limits would account for the tolerance and uncertainty: so (±.025°C ±.0047 °C) would yield a range around 100 °C of -.0203 °C and +.0203°C.

    With K-2; 95 % confidence.
     
    Andy Nichols likes this.