Sensor Fault Detection and Diagnostics in Small Modular Reactor and Advanced Analytical Perspectives: A Review

Abstract

Reliable measurement from sensors is essential to guarantee the safety and efficiency of small modular reactors (SMRs). Nonetheless, reliable and safe SMR operation hinges on precise sensor measurements, which can be compromised by nascent faults such as biases, drifts, and noise. These faults develop progressively, are difficult to detect early, and may propagate through highly coupled reactor dynamics, impairing the performance of SMR control systems. To address the above problems, this article presents a comprehensive survey of sensor fault detection and diagnosis (FDD) methods suitable for SMRs. It considers three classical FDD methodologies: 1) modeling techniques (observers, Kalman filter (KF), parameter estimation); 2) signal processing approaches (wavelet, time–frequency analysis); and 3) statistical techniques [principal component analysis (PCA), dynamic PCA (DPCA), independent component analysis (ICA)]. FDD Detection quality is evaluated based on detectability, identifiability, and detection delay. The analysis shows that a single FDD method would be insufficient for diagnosing faults in SMRs, and a hybrid framework with three layers of physics-based residual generation, wavelet-based denoising, and explainable-artificial intelligence (XAI)-based classification with an adaptive feedback mechanism is proposed. Physics-informed neural networks (PINNs), digital twins (DTs), and cybersecurity-aware FDD are introduced as emerging directions, and the implementation difficulties are also presented.