Uncertainty analysis for multivariate state estimation in safety-critical and mission-critical maintenance applications

摘要

The Multivariate State Estimation Technique (MSET) has been developed at Argonne National Laboratory (ANL) and applied for real time surveillance applications for the purposes of signal validation, sensor operability validation, equipment health monitoring, incipient component fault annunciation, and process anomaly identification. Although MSET was originally developed for applications in the commercial nuclear industry, it has recently been spun off for applications in fields such as aerospace, manufacturing, transportation, robotics, and ship propulsion. Notwithstanding these types of successful applications of MSET in industry, it is necessary for safety-critical and mission-critical applications of MSET to have reliability analysis methods, including a propagation-of-uncertainty tool, which is needed to support safety evaluations in a variety of industries, and technical-specification-change requests in the case of the nuclear industry. For these and related applications, a general purpose uncertainty analysis tool for MSET has been developed that uses Monte Carlo simulation with Latin Hypercube Sampling. For any new application of MSET, the uncertainty analysis tool developed here may be used to investigate quantitative propagation-of-uncertainty behavior for all sensors under surveillance. In addition to supporting safety analysis of surveillance systems that are based on MSET, the tool developed here can be employed in parametric studies to support system designers in evaluating the relative value of adding new sensors to an engineering system during early design stages or for equipment or facility upgrades.