NUMERICAL CONSIDERATION FOR MULTISCALE STATISTICAL PROCESS CONTROL METHOD APPLIED TO NUCLEAR MATERIAL ACCOUNTANCY

The multiscale statistical process control (MSSPC) method is applied to explain material unaccounted for (MUF) in large scale reprocessing plants using numerical calculations. Continuous wavelet functions are used to decompose the process data, which simulate batch operation superimposed by various types of disturbance, and the disturbance components included in the data are divided into time and frequency spaces. The diagnosis of MSSPC is applied to distinguish abnormal events from the process data and shows how to detect abrupt and protracted diversions using principle component analysis. Quantitative performance of MSSPC for the time series data is shown with average run lengths given by Monte-Carlo simulation to compare to the non-detection probability ß. Recent discussion about bias corrections in material balances is introduced and another approach is presented to evaluate MUF without assuming the measurement error model.