Back to home
Regression model for crack severity estimation in NPP   AYODEJI Abiodun1,2, and LIU Yong-kuo1   1. Fundamental Science on Nuclear Safety and Simulation Technology Laboratory, Harbin Engineering University, Harbin, Heilongjiang 150001, China.(lyk08@126.com , abiodun.ayodeji@hrbeu.edu.cn)  2. Nuclear Power Plant Development Directorate, Nigeria Atomic Energy Commission, Abuja 900271, Nigeria (abiodun.ayodeji@nigatom.org.ng)   Abstract:In-containment pipe failures in a nuclear power plant are being detected by measuring the humidity in the containment. However, incipient leaks and cracks are difficult to detect because traditional leak monitors are not sensitive to small leak rate changes and cannot be used for low-level leak rates and are limited to post-accident analysis of significant releases. In this work, we present an optimized data-driven Support Vector Regression (SVR) model. The proposed method can be integrated with the existing leak detector to form a robust hybrid diagnostic system, effective for detecting both incipient and large leakage in nuclear plant pipes. The SVR model estimates the size and location of incipient breaks using fault signatures, and the size estimation efficiency is evaluated using the mean squared error values (MSE). To obtain efficient predictive model and minimize false alarm rate, Genetic Algorithm is utilized for feature selection purposes. To demonstrate the method and evaluate the generalization capability of the predictive model, cracks of various severities at the inlet plenum of CNP300 NPP is simulated with RELAP5/SCDAP Mod4.0 code. The SVR’s relative error (MSE) with and without feature selection algorithms were compared using different solver algorithms. The result shows better performance for the model built with features selected by GA. The model also diagnose fault faster than conventional techniques. Keyword:fault diagnosis; support vector regression; feature selection algorithm
	Click here for viewing full text.