Early Detection of Leakage in Reactor Coolant System of Nuclear Plant Using Kalman Filter.

This paper concerns leak detection in a reactor coolant system in a PWR nuclear plant. The purpose of this simulation is (1) to develop a detection system using a Kalman filter to shorten the required term for detecting compared to the conventional method and (2) to examine the possibility of determining the covariance in system and observation equations by simulation using normal operating condition data. Two procedures are compared : one is to estimate the leak value using a Kalman filter (method I), and the other is to select one leakage model from several. presupported models using the Kalman-filter-based method combined with Bayes' theorem (method II). The results show that (1) the leak value can be estimated quantitatively by method I, (2) method II requires 1/3 the time taken with the conventional method, and (3) it is possible to optimize the covariance by simulation.