A Statistic Sensitive to Deviations from the Zero-Loss Condition in a Sequence of Material Balances

The CUMUFR (cumulative sum of standardized MUFresiduals) statistic is proposed to examine materials balance data for deviations from the zeroloss condition. The time series of MUF-residuals is shown to be a linear transformation of the MUF-time series. The MUF-residuals can directly be obtained by applying the transformation or they can be obtained, approximately, by the application of a Kalman filter to estimate the true state of MUF. A modified sequential test with power one is formulated for testing the CUMUFR statistic. The detection capability of the proposed examination procedure is demonstrated by an example, based on Monte Carlo simulations, where the materials balance of the chemical separation process in a reference reprocessing facility is considered. It is shown that abrupt as well as protracted loss patterns are detected with rather high probability when they occur after a zeroloss period.