Bearing Fault Diagnosis of Sorting Machine Induction Based on Improved Neural Network and Evidence Theory
Roller bearing is an important mechanical element of sorting machine induction.It usually has defects in outer race,inner race or balls due to continuous metal-metal contacts in high-speed operating conditions.This paper presents a novel diagnosis algorithm based on improved neural network and D-S evidence theory.Firstly,fault features are extracted through vibration signal analysis.Improved neural network classifier is then constructed to finish primary recognition,which introduces momentum to increase the learning rate.In order to reduce recognition uncertainty,each single classifier is regarded as independent evidence,and they are aggregated by improved Dempsters combination rule.Experiment results show that proposed algorithm can improve diagnosis accuracy and decrease recognition uncertainty.
fault diagnosis roller bearing neural network evidence theory
Wei Chen Qing-xuan Jia Han-xu Sun Si-cheng Nian
School of Automation,Beijing University of Posts and Telecommunications,Beijing 100876,China
国际会议
杭州
英文
185-188
2013-03-22（万方平台首次上网日期，不代表论文的发表时间）