Subspace Aided Parity-Based Robust Data-Driven Fault Detection in Pakistan Research Reactor-2

Abstract

This article is concerned with FD (Fault Detection) in PARR-2 (Pakistan Research Reactor-2) using a subspace aided parity-based FD scheme. The safety is of vital importance for nuclear reactors and in time fault diagnosis is necessary for safe operation. Conventional model-based FD approaches required the mathematical model of the process. For complex systems like nuclear reactors, the modeling of the system is too much complicated. Due to the availability of huge process data of the reactor and largely inaccessibility moreover as the complexity of the process model, data-driven approaches are effective fault diagnosis techniques for reactors. Subspace aided parity-based data-driven FD approach is a simple, efficient FD approach and has required less online computations. By using a subspace-aided approach, an optimized parity vector is identified directly from the process data instead of the identification of the system model. The identified parity vector is utilized to compute residual generator that ensures robustness against system noises and disturbances and sensitivity to faults. The parity-based FD scheme is successfully implemented for PARR-2. Two possible faults in PARR-2 that are external reactivity insertion fault and control rod withdrawal fault are considered and detected successfully. GLR (Generalized Likelihood Ratio) based threshold setting is used for efficient FD and reduce false fault detection rate.