Uncertainty Determination Demonstration Program On MC&A Measurement Systems *

Year
1996
Author(s)
John P. Clark - Westinghouse Savannah River Company
A. Harper Shull - Westinghouse Savannah River Company
Abstract
Statistically propagated limits of error(LOE) for accountability measurements are usually smaller than LOEs derived from historical data. The LOE is no better than the data used in its propagation. Laboratory measurement quality control programs generate estimates of random and systematic errors for the LOE calculations. The uncertainty of measurement system standards and instrument calibrations are often not included in measurement quality control(QC) programs (MCPs) estimates. Therefore, the uncertainty associated with a measurement system is usually underestimated. A program was conducted at the Savannah River Site (SRS) to evaluate a commercial measurement assurance program software package (JTIPMAPƤ)that records, charts, and analyzes control standard measurements to determine and control total measurement uncertainty. The demonstration program involved six types of accountability measurements. The software uniquely uses the uncertainty of the standards, the calibration histories and routine QC data to estimate the total uncertainty of a measurement process. The demonstration program involved: training measurement personnel on the principles of process measurement assurance (PMAP) and the use of the software; technical support in setting up PMAPs on gas mass spectrometry, calorimetry, Fourier transformed infrared(FTIR) spectrometry, alpha PHA spectrometry, diode array spectrophotometry, and mass standards calibrations measurement systems; and determining and evaluating uncertainty estimates for each system. Results of the demonstration program are described and the uncertainties for these measurement systems are summarized in the paper below. The demonstration showed the training and software provided several useful functions such as uncertainty determinations that include the standards used and independent standards that reveal measurement process systematic errors, which produced larger uncertainties estimates than current MCPs. The software will be tested further in pilot programs for D2O measurements, calorimetry and mass standards calibrations.