HYBRID TECHNIQUE FOR FAULT DETECTION AND DIAGNOSIS IN ROTATING MACHINES UTILIZING MATHEMATICAL MODELS

Abstract

A hybrid techniques for fault detection and diagnosis in rotating machines utilizing mathematical model, is proposed in this paper. The methodology is derived using a combination of correlation analysis based on the Ljapunov Matrix and Artificial Neural Network (ANN). This procedure of parameter fault diagnosis uses only measured state variables. The direct measurement of the stochastic excitation is not necessary. Coloured noise stochastic excitation is modeled by a dynamical system excited by white noise. Using the properties of correlation of the output variables, it is possible to derive specific relations involving the physical parameters of the system and the correlation matrices of the measured variables. Faults in the rotor can be detected by monitoring the variation of the physical parameters through a comparison of theoretical and estimated correlation functions. Artificial Neural Network is used as a tool to map such correlations. The numerical approaches for rotor systems modeled with six degrees of freedom are discussed with respect to this methodology. The good results show the viability of further studies in this area. Keywords: Fault Detection, Rotating Machines, Mathematical Models.