Research On Fault Diagnosis Of Gear Transmission Based On Deep Learning Theory

Aiming at the problems of gears and other parts in the gear transmission system that are prone to failure or failure,the paper studies the fault diagnosis of the gear transmission system,mainly including the mechanism of the vibration of the gear transmission system,the transmission route,the failure mode and the fault reason analysis,and the depth analysis.Research on Vibration Signal Feature Extraction and Fault Diagnosis of Learning Theory.And using Lab VIEW test development platform to design a set of fault diagnosis system based on fault diagnosis method described in this paper.For the fault diagnosis of gear transmission system,this paper proposes a fault diagnosis of gear transmission system based on deep learning theory.The deep learning model used is Deep Belief Network(DBNs).Firstly,using the powerful self-extraction feature of DBNs network,the vibration signal of gear transmission system is extracted,then the fault signal is identified by DBNs’ s complex map representation capability.In practical applications,when the DBNs network is directly used to process the original time domain signal of gear vibration,the correct recognition rate can only reach about60%.When a simple Fourier transform is performed on the time domain signal,the DBNs network is used to process the frequency spectrum of the vibration signal.The accuracy rate can reach 99.7%,which confirms the simplicity and effectiveness of the fault diagnosis method described in this paper.Although the fault diagnosis method of gear transmission system based on deep belief network can effectively discriminate complex signals,but the fault recognition rates of different network frames are different.Therefore,the optimization of relevant network parameters becomes a research difficulty.In order to solve this problem,this paper presents their own optimization methods for the parameters of the DBNs network,such as the number of network layers,the number of hidden layer units,the learning rate,and examples verify the feasibility of the optimization method.Finally,the activation function of the DBNs network is optimized and replaced by a nonlinear element.The advantages of the nonlinear element activation function are highlighted by comparison with other activation functions.At last,we use Lab VIEW’s graphic programming language to organize the PC software system,take the laboratory gearbox as the test object,and use the data acquisition hardware of NI company as the technical support,and design a gearbox fault diagnosis system based on deep confidence network.The system can easily implement functions such as data monitoring,data acquisition,data storage and fault diagnosis.It also uses the simulation signal to test the function modules of status monitoring,data acquisition and data storage.Using the fault sample as a test set,the reliability of the fault diagnosis module of the gearbox fault diagnosis system was verified.