Strategy for addressing composition uncertainties in a Hanford high-level waste vitrification plant

Abstract

Various requirements will be imposed on the feed material and glass produced by the high-level waste (HLW) vitrification plant at the Hanford Site. A statistical process/product control system will be used to control the melter feed composition and to check and document product quality. Two general types of uncertainty are important in HLW vitrification process/product control: model uncertainty and composition uncertainty. Model uncertainty is discussed by Hrma, Piepel, et al. (1994). Composition uncertainty includes the uncertainties inherent in estimates of feed composition and other process measurements. Because feed composition is a multivariate quantity, multivariate estimates of composition uncertainty (i.e., covariance matrices) are required. Three components of composition uncertainty will play a role in estimating and checking batch and glass attributes: batch-to-batch variability, within-batch uncertainty, and analytical uncertainty. This document reviews the techniques to be used in estimating and updating composition uncertainties and in combining these composition uncertainties with model uncertainty to yield estimates of (univariate) uncertainties associated with estimates of batch and glass properties.