Assessing Measurement Control: Experiences with Nondestructive Assay Methods.

摘要

Demonstration of stability or control of a measurement process over time is often required for critical processes. Measurement control is monitored by calculating measurement errors for a collection of comparison standards over time and producing a Shewhart control chart. However, measurement errors inherently occur one-at-a-time and not in batches. Additionally there is often a non-deterministic drift in the mean measurement error. These facts make it challenging to develop warning and alarm limits for a control chart. Previous studies have suggested using the mean squared successive difference to estimate the variance of one-at-a-time data. This technique can also reduce or eliminate estimation bias due to a fluctuating mean. Application of a control charting methodology based on the mean squared successive difference is demonstrated using data from the nondestructive assay of nuclear materials, and the performance and potential limitations of the method are explored.