Fuzzy classifier with automatic rule generation for fault diagnosis of hydraulic brake system using statistical features

This study focuses on the condition monitoring of a hydraulic brake system using vibration signal through a machine learning approach. The machine learning approach has three main steps: feature extraction, feature selection and feature classification. Statistical features were used for the fault diagnosis of hydraulic brake system. Through a feature extraction technique, descriptive statistical features were extracted from the acquired vibration signals. C4.5 decision tree algorithm was used for selecting best features that will distinguish the fault conditions of the brake from given train samples. For feature classification, fuzzy logic was used as a classifier. A necessary rule set was formed automatically by using decision tree algorithm. The generated rule set is fed to fuzzy classifier. The procedure to build fuzzy classifier is also explained and the results were discussed.