Signal pattern matching and automated algorithms for the loose parts monitoring system of the nuclear power plant

Abstract

Monitoring signals in the control room of nuclear power plants come in through multiple channels and are displayed as graphs on the display devices. It is crucial to make correct and reliable decisions on the relationships among the displayed graphs, which are representations of signals from different channels, both by implemented algorithms and human decisions. Simplified schemes to compare signal envelopes using cross correlation and cross entropy are introduced. In this work, algorithms to curve-fit the signal envelope to a polynomial and envelope generation with interval means are also researched. When generating signal envelopes, using interval means instead of interval medians is safe enough and is not seriously influenced by possible outliers. An effective use of eigenvalues for verification of the signal correlation is pursued in this work for a three channel case. Monitoring signals of the nuclear power plants are compared across the channels, or auto-correlation between recurring signals is analyzed to determine the signal closeness or to screen out false signals. Cross correlation of interval means from the bulk data is a better method than comparing the bulk data sets themselves. This is due to the smoothing effect of the signals when interval means are taken in place of the bulk data. A mathematical justification for the smoothing effect is tried using a newly defined smoothing function. The smoothing function shows a behavior that is similar to wavelet functions. The relative errors are also examined with an example from a reference. The Loose Parts Monitoring System(LPMS) monitors the probable damages by traveling loose parts to the internal structures of the nuclear power plant. It involves a lengthy and complicated process to determine the parameters like size, mass, speed, or impact location of the loose part when the signals are collected and processed. In this work, a simple and efficient model for determining the impact location of the loos...