|
Precision, Accuracy, Uncertainty: Do I care? |
|
Precision, Accuracy, Uncertainty: Do I care? Yes you do care. You are reading this material because you are involved in particulate Air Sampling and the Calibration/Verification set up measurements, related to Air flow rate (most often in lpm), Temperature (ambient or flow stream) and Barometric pressure. The explanations offered here are informal and simplistic. They are meant to present a quick, easy to remember, overview for interested parties. They may well be found wanting by various Govt. Agency/Learned Institute Scientists or Metrologists1. For those wanting a rigorous discussion, a really superb book is available2. Lest we lead anyone astray, Disclaimer & Limit of Liability Accuracy means how close a measurement is to the true value? “My instrument is ±1% accurate at 10 lpm”, means that the quantity of air flowing is between 9.9 and 10.1 lpm. More than that, you are not to know. When appreciating the accuracy of a device, be aware that some devices have the accuracy quoted as a % of the point (this is good) or % of the entire range (this is not as good). Precision means how close you can get on a repetitive basis. Because of the manner in which air flow is measured the values given for precision and accuracy are usually the same numerical value. However, supposing your calibrator is “broken”. It reads 10 lpm but the true flow is really 9 lpm. Further, you take 10 measurements and they all show up as being between 9.9 and 10.1 lpm. You have an instrument which is “off” by 10 % (±10% Accurate) but has a precision of 1%. In recent years the term Uncertainty has arrived. This is the difference between a True value and a reference standard value. This is just a way of defining the difference between the device that you are using and a better device that it was calibrated against. An example of what uncertainty means to you may be observed by the measurements that might be taken when calibrating an air flow measurement device. Air flow measured by a transfer standard: Uncertainty = 0.5%, Data furnished by calibration laboratory and NIST traceable. Temperature measured with a glass thermometer: Uncertainty = 0.071%, Data furnished by calibration laboratory and NIST traceable. Barometric Pressure measured with a Fortin Mercury Barometer: Accuracy = 0.1%, Data furnished by manufacturer and NIST traceable. Absolute system pressure measured with Mercury Manometer: Accuracy = 0.025%, Data furnished by manufacturer and NIST traceable. Note, by convention, when uncertainty information is unavailable accuracy is used in its stead. Whilst, learned books and standards are available on how to calculate and investigate uncertainty, one simple approach is to take the square root of the sum of the individual uncertainties squared. In the above example, it comes out 0.52%. Is that good? Metrologists say that this is an important number. The smaller it is the better it is. Reference 2 presents an excellent graph relating the various terms. It is reproduced here with the permission of the publishers from Page 10 figure 2.1.
1. Metrology is the science of measurement and includes all theoretical and practical aspects. A practitioner is referred to as a Metrologist. Copyright © 2007 by BGI / Modified:
Friday, November 2, 2007
|
