Ball-Bearing Fault Detection of Squirrel-Cage Induction Motors Based on Single-Phase Stator Current Using Wavelet Packet Decomposition and Statistical Features

This paper presents a multi-stage method for detecting ball-bearing faults in three-phase squirrel-cage induction motors using single-phase armature current signals. The method utilizes wavelet packet decomposition, fast Fourier transform, power spectral density, and statistical analysis to identify faults under various load conditions and supply modes, including direct on line, direct torque control, and v/f scalar control. Fault detection indices are derived from the energy, standard deviation, and variance of both time and frequency domain processed signals. The method was evaluated using the data from both laboratory setup and industrial cases. This approach ensures accurate detection of ball-bearing faults by comparing normalized indices from healthy and faulty states, providing reliable diagnosis even with limited motor accessibility.