An Improved Threshold Determination Method Based on PSO Algorithm for Sensor Fault Diagnosis

Abstract

Based on the Kalman filter theory, detecting sensor faults in gas turbine control system and performing output reconstruction for fault signals is a hot focus in recent years. The selection of detection thresholds directly affects the rate and accuracy of fault diagnosis, which is extremely important for the safe operating of control system and its value is usually determined by actual experience and the noise characteristics of the sensor itself. Based on characteristics of the sensor and the observed signals, the cost function of the threshold is constructed by the combination of the leakage alarm probability and the false alarm probability, which is then used as the fitness function in the particle swarm optimization (PSO) algorithm, and the inertia weight in the iterative search process is adaptively changed. So, an improved dynamic weight adaptive threshold determination method based on PSO algorithm comes into being. The simulation comparison of the proposed method with the traditional threshold determination method shows that the timeliness, accuracy and integrity of the method is improved somewhat.