Abstract:
The diagnosis of bearings defects in electric machines has become an important area
of research in recent years. Most studies on the detection of bearing faults using vibration
measurements descended from sensors placed close to the mechanical elements to be
monitored. For this reason, the work of this thesis focused on the automatic classification of
bearing defects of the asynchronous machine with squirrel cage.
We proposed a clustering approach to bearing defects. This approach is based on five
steps: The first step uses the decomposition of the empirical method to separate each vibration signal in the different intrinsic mode functions, where each mode is in a specific frequency band. The second step extracts amplitudes and instantaneous frequencies for each mode in order to identify its frequency band by calculating the Hilbert marginal spectrum. The third step, the feature extraction phase is carried out by using the energy operator Teager-Kaiser (TKEO) .In the fourth step, vectors forms extracted was optimized from optimization by particle swarm which is a stochastic optimization method developed, based on the reproduction of social behavior. Finally, the final step relates to the automatic classification of optimized vectors that can be achieved through the Gaussian mixture model algorithm. This will make the relevance of attributes and automatically selection and classification of bearings defects in asynchronous electric machines.
