Health monitoring of automotive clutch system by using Bayes algorithms

Abstract

Clutch is an essential component in an automotive transmission system that helps in power transmission from the engine to the gearbox. Continuous operation of the clutch leads to degradation, damage, and reduced lifetime of internal components. Such factors can provoke the occurrence of various faults, which left unmonitored will result in clutch damage and seizure. Hence, continuous monitoring of the clutch is necessary to minimize unwanted breakdowns. This paper presents a condition monitoring technique based on vibration analysis to monitor the fault occurrence of different components. A machine learning approach is carried out to classify various faults. Such as fingers worn, pressure plate broken, pressure plate worn, friction material loss, and tangential strip bent. Feature extraction is performed on the acquired vibration signals using statistical learning process. The most important and significant features are selected from the extracted features by J48 decision tree algorithms. Further, feature classification is done with the help of Bayes based classifiers and the obtained results are compared to predict the best in class classifier for real-time.