Contribution au Diagnostic de la Machine Asynchrone Triphasée par une Approche Modèle

Abstract

Fast Fourier transform (FFT) is a widely used technique for fault diagnosis in induction machines. It works well in high power or constant load torque operating conditions but difficulties arise in variable load torque, speed and supply voltages. The need to find other signal processing techniques for non stationary signals becomes necessary. On this issue, our contribution is the application of discrete wavelet transform (DWT) to electrical signals of induction machine for broken rotor bars and mixed eccentricity faults diagnosis in the case of variable load torque. From the simulation and experimental results, the diagnosis of these faults using FFT becomes very difficult and even erroneous. Indeed, the fault characteristic components have widely spread in wide frequency bands that are proportional to the variation of the load in the case of broken rotor bars and decreased proportionally to the variable load torque in the case of mixed eccentricity. In addition, as the instantaneous values of slip cannot be clearly defined, the exact identification of the amplitude and position of the fault characteristic components in the spectra becomes more difficult. In these cases, DWT can accomplish the task of fault diagnosis using a qualitative analysis by analyzing the increase of the specific detail signals energy and the particular evolution of the fault characteristics components according to the variation in load torque or by quantitative analysis through a fault severity factor (FSF).