ESTIMATION OF REMAINING LIFE OF BEARINGS USING ROTATION FOREST AND RANDOM COMMITTEE CLASSIFICATION MODELS – A STATISTICAL LEARNING APPROACH

Abstract

Bearings are considered to be one of the critical elements in all rotating machineries. Bearings are in general used to reduce or minimize the friction in the rotating parts. Strengthening the predictive maintenance of bearings helps to improve the performance of the machines. Hence, bearing prognosis gains its importance in the recent times. This paper emphasis on estimation of remaining life of bearings using classification models through condition monitoring techniques. Vibration signals acquired from the experiments were used to assess the current state of the bearings while in operation. Statistical features were extracted from the signals and the best contributing features were selected for building a classification model with Random forest, Rotation forest and Random committee classifiers. The effectiveness of the classification model built by Random forest, Rotation forest and Random committee classifiers were analysed and compared through a statistical machine learning approach.