Publications

Abstract : Gearbox is one of the most important components in an automobile, enabling power transmission from the engine to the wheels. Gears and bearings are prone to failure. The impending case of failure can be predicted by performing vibration analysis of a gear box, usually done by acquiring data in lab conditions. Objective: This paper proposes an idea to enable fault detection in the gearbox by acquiring data under on road conditions without having to remove the gearbox, thereby simplifying the condition monitoring of a gearbox. Methodology: The experimental studies were conducted on the gearbox in a test vehicle run in real time conditions and the vibration data from the gearbox was acquired using a piezoelectric accelerometer for different conditions of gearbox. The acquired time domain data was normalized and its statistical features were extracted. The classification of the fault class was done by using decision tree (J48) algorithm. Findings: Classification efficiencies as high as 99% were obtained by using decision tree algorithm. Further, normalization of raw data was found to increase the efficiency of the classifier. This observation can be used to make decision trees more efficient. Improvements: This paper has highlighted the concept of on road testing for two fault conditions. Further research work on other fault conditions can be done.