A Bayes learning approach for monitoring the condition of suspension system using vibration signals

Abstract

The suspension system is an important component in an automobile, aimed to provide better comfort and good road handling characteristics during vehicle motion on uneven road conditions. Varying load conditions, prolonged operations, continuous road shocks absorption, and time-based degradation of internal components can create faults in a suspension system. Fault occurrences in the suspension system can damage the internal components that lead to malfunctioning of the system causes failure and endanger vehicle safety. Thus, condition monitoring has become an essential part of identifying fault occurrences in a suspension system. This paper is focussed on condition monitoring of suspension system using vibration signal analysis and faults are classified with the help of machine learning-based Bayesian classifiers. A test setup is fabricated to simulate the working of the suspension system under different load conditions with one good and six faulty conditions. Vibration signals are recorded and used to extract statistical features. Further J48 decision tree algorithm was utilized to identify the most significant features during feature selection process. Bayes classifiers such as NaiveBayes and BayesNet classifiers were used to find the type of fault occurrence from the selected features. The overall results of the aforementioned classifiers are compared and the best in the Bayesian classifier is suggested for real-time application.