Artificial immunity-based induction motor bearing fault diagnosis

Abstract

In this study, the artificial immunity of the negative selection algorithm is used for bearing fault detection. It is implemented in MATLAB-based graphical user interface software. The developed software uses amplitudes of the vibration signal in the time and frequency domains. Outer, inner, and ball defects in the bearings of the induction motor are detected by anomaly monitoring. The time instants of the fault occurrence and fault level are determined according to the number of activated detectors. Anomaly detection in the frequency domain is implemented by monitoring the fault indicator bearing frequencies and harmonics, calculated using the bearing dimensions and number of rotor revolutions. Due to the constant fault location and closeness to the accelerometer, the outer race fault in the bearing is the easiest fault type to determine. However, the most difficult fault type to detect is the ball defect. By verification of the detection results, the motor load has very little effect on the fault.