Fault Diagnosis of Linear Control Systems Based on the Discrete Wavelet Transform and an ART2 Neural Network

Abstract

Fault detection and isolation of systems continues to be important problems to be addressed due to the increased complexities of more advanced systems. Early detection and isolation of faults can assist in avoidance of major system breakdowns. Many methods require some model of the plant in order to perform the fault diagnosis. In this paper we present a fault diagnosis method for dynamic systems based on discrete wavelet transform (DWT) and an adaptive resonance theory 2 neural network (ART2 NN). In the proposed method a fault is detected when an error between the system output and the nominal system output cross a predetermined threshold. Once a fault in the system is detected the ART2 NN based fault classifier isolates the fault. The algorithm contains three main steps: fault detection through the threshold test, data preprocessing via DWT, and fault isolation using the fault classifier. The simulation results demonstrate the effectiveness of the proposed DWT and ART2 NN based fault diagnosis method.