Sensor Location and Selection for Fault Diagnosis and Detection in Nuclear Power Plant Using Signed Direction Graph

Nuclear power plants are expected to achieve comprehensive fault monitoring to ensure secure and stable operation. Sensor placement aims to provide full spatial coverage at target locations within cost constraints. This paper proposes a comprehensive modeling approach to derive the fault-sensor correlation matrix for the deaerator system in nuclear power plants based on SDG theory. The proposed method optimizes sensor placement by effectively integrating the NSGA-II algorithm with the TOPSIS technique, enabling systematic evaluation and ranking of available sensor allocation schemes. Using APROS simulation data, we conducted a comprehensive analysis of fault propagation paths and sensor correlations, establishing a probability model that accurately assesses sensor fault detection performance. In ideal scenarios, sensor placement optimization reveals a theoretical configuration which removes two ineffective locations and identifies about 20 sensors as optimal. Two sensitivity analysis cases assess key placement factors. Sensor cost does not affect placement at specific locations, but failure probability—driven by harsh measurement conditions—significantly influences initial configuration. The SDG-based fault diagnosis and detection method using a compatible pathway accurately classifies different deaerator component failures, showing strong potential for nuclear power system condition monitoring. The sensor placement method proposed in this study demonstrates superior performance over conventional approaches, providing a reference for complex process monitoring and decision-making.