Bearing fault diagnosis based on Lasso regularization method

Bearing fault diagnosis in induction motors is an open field of research. The use of the stator current to monitor the bearing condition has some advantages over other signals such as vibration and acoustic emission, but it has proven to be less effective than for another kind of faults. This paper proposes to overcome these difficulties by an automatic classifier that uses a significant amount of information from the current spectra. However, in these conditions, a classifier is prone to overfitting, which is a serious problem that can be avoided using regularization methods such as Lasso. In this work, the Lasso technique is used to improve the performance of a Logistic Regression classifier to diagnose different bearing condition states.