Research On Fault Diagnosis Of Gearbox Based On VMD And DBN

Gearbox is an important part of mechanical equipment.It plays an important role in transmission speed and force.Its state will directly affect the operation safety and production efficiency of the entire equipment.Under complex and harsh working conditions,the gearbox is prone to failure.Therefore,the study of fault diagnosis technology that can accurately identify the type of gearbox fault is of great significance to ensure the safe operation of mechanical equipment and avoid accidents.This paper takes the gearbox as the research object.Aiming at the high dynamic characteristics of the gearbox vibration signal and the problem that the early faults of the gearbox are difficult to be accurately identified,the research is carried out from three aspects:signal processing,feature extraction and fault diagnosis.The main content as follows:(1)Variational Mode Decomposition(VMD)method processes the fault signal.The VMD method has the problem of difficulty in selecting the decomposition number K during signal decomposition.Therefore,an adaptive variational modal decomposition(AVMD)method based on the energy ratio is used to select the decomposition number K of the vibration signal.The AVMD is applied to the bearing failure standard data set of Western Reserve University.Experimental results verify the effectiveness of the AVMD method.(2)In order to obtain the characteristic index that fully characterizes the operating state of the gearbox,a feature selection method based on AVMD and ReliefF is used.This method firstly decomposes the gearbox fault vibration signal by AVMD,then extract the time domain and energy entropy of each component signal and the original signal to form a feature vector set,finally uses the ReliefF algorithm to filter out the sensitive feature vector set.Since the classification effect of the Deep Belief Network(DBN)is affected by the number of hidden layer nodes.Use Particle Swarm Optimization(PSO)algorithm to optimize the number of hidden layer nodes of the DBN.(3)In order to verify the accuracy of the PSO-DBN diagnosis model based on AVMD,experiments on the gearbox failure standard data set and verify the accuracy of the method used in this paper.Then,the fault diagnosis of the high and low machine in the gearbox of a certain type of self-propelled artillery collected on the spot was carried out,the diagnosis accuracy rate was as high as 96.67%.Then compare the method used in this paper with the Probabilistic Neural Network(PNN),Kohonen neural network,Support Vector Machines(SVM)diagnostic model based on AVMD.The results show that the method used in this paper improves the recognition accuracy of at least 11.67%compared with the other three diagnostic models.In order to verify the accuracy of AVMD signal processing,the PSO-DBN model using AVMD method is compared with the PSO-DBN model using the Empirical Mode Decomposition(EMD)method.Experimental results show that the AVMD method has higher accuracy in processing the vibration signals of the gearbox of the self-propelled gun,and the recognition accuracy is increased by 21.67%.(4)On the basis of the previous three steps,a Graphical User Interfaces(GUI)system for fault diagnosis of gearbox high and low machine is designed with the help of GUIDE of MATLAB.Experimental results show that the designed GUI system can click the button according to the prompts of the interface to complete the corresponding tasks and functions,which improves the convenience of diagnosis.