Abstract:
The reliability and availability of electromechanical systems are increasingly required
in the industrial sector. Unexpected shutdown in the electrical machine can lead to unplanned costly shutdowns and damage to equipment or even danger to people. Electrical
machines are basic elements in many electrical systems and among all types of motors,
induction motors are an important part of many industrial processes because of their advantages : simple construction, robustness and high performance.
The variety of measured physical quantities makes it possible to analyze the defects of
electrical machines differently, as it appeared in the works of this thesis where the techniques of the analysis of the stator current, the analysis of the Park vector and the mechanical vibration analysis are used. The signal approach is well chosen with the default that
suits it, as well as the choice of the diagnostic algorithm and on which it is developed,
adapted and implemented.
The data processing taken from the various measured signals requires prior processing
in order to make a decision on the healthy or faulty cases of electrical machines. The diversification of the signal processing methods usually used for the diagnosis of defects in the
asynchronous machine was one of the subjects of this thesis. From the Fourier transform
and its derivatives, the DFT, the sliding DFT, the Zoom-FFT and Goertzel, to the algorithms
of MUSIC, Zoom-MUSIC and the Wavelets, a survey that was necessary to conclude on the
possibilities of these methods to be considered for an online diagnosis.
To achieve the diagnostic online goal, the exploitation of the lateral lobe leakage phenomenon when applying the sliding DFT is presented as a new method, the bar breakage
and eccentricity defects are considered as typical cases to show the virtue of the advanced
study.
