The Fault Diagnosis Of Gear Based On Manifold Learning Method And Radial Basis Function Neural Network

The development direction of modern machinery and equipment is toward equipment production automation,operation efficiency,complex structure and large scale,which makes the connection between equipment and equipment more inseparable.Once a mechanical device fails,it will affect the operation of the whole production system.The losses directly or indirectly will increase exponentially.This shows the importance of the high reliability,low failure rate and high safety of the equipment to modern industrial production.In mechanical equipment,gear transmission is the most extensive way of transmission,and it has the characteristics of constant transmission ratio,compact transmission structure and large torque transmission.To prevent the failure of mechanical equipment running,prolong the working time of equipment.On gear transmission system testing and fault diagnosis,and early prediction and detection of faults in the system for safety production and improving economic efficiency,it is of vital significance.In this paper,a gear fault diagnosis method is proposed,which combines the manifold learning algorithm with the RBF neural network.According to the gear vibration signal,the ensemble empirical mode decomposition can adaptively decompose the local information,which can get the signal time domain and frequency domain so that this method using EEMD to extract the fault feature signal of the original feature vector of fault diagnosis.Because the original feature extracted by this method has the disadvantages of high dimension and information redundancy,it will affect the classification and recognition of the neural network.The local tangent space alignment algorithm in manifold learning algorithm is a good nonlinear signal processing algorithm,which can do dimension reduction for redundant signals very well.The combination of LTSA and EEMD can extract features in vibration signals very well.Therefore,a feature extraction method based on ensemble empirical mode decomposition and manifold learning is proposed in this paper:First,EEMD is used to decompose the collected gear vibration data automatically,and the eigenvalues of the covariance matrix corresponding to these IMF are calculated by the intrinsic modal components decomposed,and the eigenvalues are used to form the original feature set of gears;Then,this paper further extracts the original feature set by LTSA,and gets thenew feature set.Finally,this paper inputs it into the trained RBF neural network for gear fault recognition.This paper also analyzes the mechanism and characteristics of gear vibration,and introduces the important application of artificial neural network in the field of fault diagnosis.The gear fault diagnosis method is verified on the gear fault simulation test platform by collecting the gear vibration signal.The results show that the fault diagnosis method has good effect and has wide application prospects.