Fault Diagnosis Method Of Gearbox Based On Fault Diagnosability And VMD

The gearbox is used as a rotating mechanical equipment to adjust the speed and transfer torque.Its normal operation will directly determine whether the machine can work properly.The gearbox runs for a long time under the condition of high speed and heavy load.It is more likely to appear various types of damage,which will affect the normal operation of the vehicle,and then bring serious economic losses and even casualties,so it is very meaningful to monitor the running state of the gearbox in real time and to diagnose the fault.In this thesis,the gearbox fault diagnosis is realized according to three steps: signal acquisition,signal processing and fault identification.The contents of the study are as follows:(1)Gearbox fault type and vibration mechanism for determining the research scheme is analyzed.This thesis analyzes the structure characteristics,fault type,fault mechanism and fault characteristics of the gearbox.The common methods of optimal arrangement of gearbox sensors and fault diagnosis are summarized.Combined with the actual needs,the research scheme of this thesis is defined.(2)An sensor arrangement method of gearbox sensor based on fault diagnosability is constructed.At present,when optimizing the position of the sensor,there is little research on whether the possible fault signal of the gearbox can be identified and separated.The fault diagnosability is applied to the optimal sensors arrangement of gearbox.The different arrangement schemes are evaluated by using the comprehensive index composed of three evaluation criteria: singular value ratio,fault diagnosability and average acceleration amplitude.The optimal sensor arrangement scheme is determined.Firstly,the modal analysis of the gearbox is carried out to extract the modal modes,and then the K-means clustering algorithm is used to classify the dynamic similarity of shape modes of degrees of freedom at important modes.Secondly,the effective independent average acceleration amplitude method is used to select the initial measuring points.Then,the possible fault signals are measured at the positions of these nodes,and the corresponding fault spectrum is obtained by Fast Fourier Transform.The density function of the corresponding fault is obtained by using the Kernel density estimation,and then the fault diagnosability at each position is judged by Kullback-Leibler divergence.Finally,the comprehensive index is used to evaluate the optimal scheme.The feasibility of the proposed method is verified through the ZDH10 gearbox fault diagnosis setup.(3)A variational mode decomposition(VMD)which is optimized with information entropy and ensemble kurtosis,and particle swarm optimized support vector machine(PSOSVM)method is constructed.In order to realize the accurate diagnosis of bearing fault in gearbox and solve the problem of artificial determination of variational modal decomposition parameters,firstly optimize the VMD parameters by using the principle of minimum of product of the reciprocal of ensemble kurtosis and the information entropy.Decompose the original fault signal by the above optimized VMD and obtain some established intrinsic mode functions(IMFs).Select the IMF with the minimum product of the information entropy and the reciprocal of ensemble kurtosis as the best IMF,and extract its fault features to form the feature vector and input PSO-SVM to classify the fault type.Both simulation signal and the actual bearing data are applied to verify the proposed method.