Análise de falhas em máquinas elétricas usando transformada wavelet e sinais acústicos

Abstract

Signal processing techniques are widely used to solve several scientific problems, when applied to three-phase induction motor signals, they enable pattern recognition and fault identification. Three-phase induction motors play an important role for industries and their malfunction can lead to unwanted interruptions in the industrial process. In order to avoid such problems, it is convenient for industries to carry out fault diagnosis. Therefore, this work proposes a method of detecting faults in three-phase induction motors for different operating situations, namely, bearing failure, voltage unbalance and with different torques on shaft. The technique uses the wavelet transform applied to the engine's acoustic emission signals for each operating situation and then performs the comparison of energy and entropy of the wavelet coefficients extracted from the fault engine signals with the same parameters of the healthy engine. Finally, the technique of support vector machines is used to perform the validation and comparison between three wavelet families, the Daubechies, symlets and coiflets, which all obtained 100% accuracy in the classification, validating the proposed technique.