Automated Material Accounting statistics System (AMASS)

In this paper the modeling and statistical analysis of measurement and process data is readdressed under a more general framework than that provided in the literature. The derivations are generic in the sense that they do not focus on any particular segment of the nuclear fuel cycle but have application to all segments. From the basic models and associated assumptions, algorithms or rules are derived from fundamental statistical principles by which one can determine the measurement variances of any linear sum of individual items in the inventory difference (ID) calculation. The analysis is then extended by covariance analysis to the examination of sums of IDs to determine the measurement variances to be associated with such sums. The model is further expanded to obtain an estimate of the contribution of the unmeasured process variability to the ID. Finally, both measured inventory errors and unmeasured inventory variances are combined in a single model. Procedures are developed for estimating the parameters of the model and for using the information in determining the significance of single inventory differences and sums of inventory differences. The end result of this effort is a computer program (AMASS) which uses the algorithms and equations of this paper to accomplish the analyses indicated. The actual application of the method to process data is emphasized.