Feature extraction for the prognosis of electromechanical faults in electrical machines through the DWT

Antonino-Daviu J.A., Riera-Guasp M., Pineda-Sánchez M., Pons-Llinares, J., Puche-Panadero R., Pérez-Cruz J.

Abstract:
Recognition of characteristic patterns is proposed in this paper in order to diagnose the presence of electromechanical faults in induction electrical machines. Two common faults are considered; broken rotor bars and mixed eccentricities. The presence of these faults leads to the appearance of frequency components following a very characteristic evolution during the startup transient. The identification and extraction of these characteristic patterns through the Discrete Wavelet Transform (DWT) have been proven to be a reliable methodology for diagnosing the presence of these faults, showing certain advantages in comparison with the classical FFT analysis of the steady-state current. In the paper, a compilation of healthy and faulty cases are presented; they confirm the validity of the approach for the correct diagnosis of a wide range of electromechanical faults.

回上页