Gearbox Wear Fault Diagnosis Based On Information Fusion

Gearbox is an important transmission component of mechanical equipment,and its operating status directly affects the stable operation of the mechanical equipment.Since most of them work under severe conditions of high speed,heavy load and strong impact,the frequency of gearbox failures is relatively high.Faults and failures caused by gearbox wear are one of the main factors that further cause serious equipment accidents.At present,the gearbox fault diagnosis usually adopts methods such as vibration monitoring,acoustic signal monitoring and oil monitoring.Among them,the monitoring methods of vibration signals and acoustic signals are susceptible to noise,and are not suitable for harsh environments and working conditions with frequent noise.However,oil monitoring has strong anti-interference ability and is suitable for various complicated working conditions.This paper takes the armored vehicle gearbox as the research object,and research on the gearbox wear fault diagnosis algorithm based on oil information.Because of the lower diagnosis accuracy and poorer stability of the single fault diagnosis model,a method of gearbox wear fault diagnosis based on information fusion is proposed.The main work of this paper is as follows:(1)This paper studies the mechanism of gearbox wear failure,systematically summarizes the gearbox wear failure forms and common failure types,and researches on gearbox fault diagnosis technology,comparing the method of vibration monitoring,acoustic signal monitoring and oil monitoring.(2)Based on the oil monitoring information of the gearbox,this paper establishes three feature-level fault diagnosis models of BP neural network,support vector machine and extreme learning machine.The traditional BP neural network has the disadvantages of low recognition accuracy and slow convergence speed,so the paper proposes an improved method of increasing momentum term and adaptive learning rate,and finally compares the diagnosis results of the three feature-level fault diagnosis models.(3)This paper has established three feature-level gearbox fault diagnosis models,regarding the output of a single fault diagnosis model as an independent evidence body and assigning the basic probability distribution function value to each evidence body through the error function or recognition error rate of the model,and the final diagnosis results are obtained by fusing each group of evidences based on the classic D-S evidence theory.Since the classic D-S evidence theory cannot handle the synthesis of highly conflicting evidence,an improved method based on the weighting of evidence distance is proposed.Finally,this paper verifies the anti-interference of the gearbox information fusion system.The simulation experiment results show that the gearbox fault diagnosis method based on the improved D-S evidence theory fusion effectively improves the identification accuracy of the gearbox wear fault types,and enhances the anti-interference ability of the network,which can better meet the actual engineering needs.