Diagnostic Fault of the Induction Machine with Application of the Neural Network

Abstract

Although the asynchronous machine is known by its qualities of robustness and low cost of construction, however, it happen that it has an electrical or mechanical fault. This article presented rotoric faults which are treated as part of the detection of broken bars, and the inverter faults (switch fault). This detection technique is based on neural network. The simulation was elaborated on Matlab/Simulink that show results used to assess the performance of the technique presented. Keyword: asynchronous machine, diagnostic, rotoric faults, broken bar, five Levels inverter, neural network. 1. I NTRODUCTION The induction machine is largely used in industry, mainly due to its reliability and relatively low cost. The control of the induction machine must take into account machine specificities: the high order of the model, the nonlinear functioning as well as the coupling between the different variables of control. The new industrial applications necessitate speed variations having high dynamic performances, a good precision in permanent regime, and high capacity of overload on all rang of speed and robustness to the different perturbation. The voltage source inverters construct a non controllable function in the power electronic; it is used in the variable application domains. The strategy obtaining by this technique is based on the study of variation speed in induction machine. The strong functional evolution was based, on one hand, the development of components for semi-conductor entirely commendable, ability, and robustness, the other hand, on the quasi generalized used of the technique pulse width modulation (PWM) (1) (2).The five levels inverter structural is more adapted in the ground power application, because the output voltage and current presented a reduction of the harmonic distortion.