Fault Detection for Roller Bearing with Vibration Analysis and Information Fusion in Sorting Machine Induction

Due to continuous metal-metal contacts in heavy and high-speed operating conditions, sorting machine inductions roller bearing easily occurs malfunctions. Therefore, its crucial to make incipient fault diagnosis. This paper presents a novel diagnosis algorithm using vibration analysis and information fusion. In the algorithm, vibration signal for roller bearing is firstly analyzed. Then extracted fault features are used as input eigenvector of constructed neural network classifier. In order to improve diagnosis accuracy, take output information of each single classifier as independent evidence, and aggregate them using improved Dempsters combination rule. Experiment results show that proposed algorithm has high accuracy of 99.5%.