Fault detection and isolation for self powered neutron detectors based on Principal Component Analysis

The self powered neutron detectors (SPNDs) play an important role in nuclear reactor monitoring. The 3-D power distribution and parameters used to evaluate the operation condition of reactor and the margin of safety can be determined using the measurement of SPNDs through power mapping procedure. Faulty SPNDs that are either completely or partially failed (hard fault or soft fault) provide incorrect information for monitoring. Correct detection and isolation of the faulty SPNDs are of primary importance to the efficient operation and management of the nuclear reactor. In this study, the methodology of Principal Component Analysis (PCA) is utilized to construct the mathematical models among various detectors at different axial location within the same string. The data used to build the mathematical models are generated by advanced neutronics code SMART rather than measurements. The square prediction error based on the model and the Detector Validity Index (DVI) based on the reconstruction are employed, respectively, to detect the SPND fault and to isolate the faulty SPNDs. The fault detection and isolation scheme is validated with four types of simulated SPND faults, i.e. bias, drifting, precision degradation and complete failure. The simulation results show that the proposed PCA based method can be used in the nuclear reactor to ensure that faulty SPNDs can be detected quickly.