Fault identification in PCA method during sensor condition monitoring in a nuclear power plant

摘要

Principal component analysis (PCA) is applied in this paper for sensor condition monitoring in a nuclear power plant (NPP). Based on the results of fault detection with PCA method, two different fault identification methods are applied simultaneously to locate the faulty sensor. One is the improved weighted contribution analysis (WCA) method which is based on traditional contribution analysis (TCA) of sensors to Q statistics. The other fault identification method is based on sensor validity index (SVI) in which the iterative reconstruction method is applied to locate the faulty sensor more accurately and quickly. Finally, the fault identification abilities of TCA, WCA and SVI are evaluated with sensor measurements from a real NPP. According to the simulation results, the improved WCA method presents better fault identification performance no matter with single or double sensor faults in the testing samples, and with single sensor fault in the testing samples, SVI method not only can verify the fault identification results by WCA method, but also can accurately reconstruct the measurements of faulty sensor as required during fault identification. (C) 2018 Elsevier Ltd. All rights reserved.