The Research On Fault Diagnosis Method Of Gear Box Based On VMD And Manifold Learning

Gear box is widely used in many important fields,once the gear and rolling bearing fault damage,may lead to the whole equipment failure serious economic property loss and casualties,so the gear box fault diagnosis has important practical significance.The main steps of fault diagnosis are: vibration signal collection,fault feature extraction and pattern recognition.In order to improve the identification of fault characteristics of nonstationary and nonlinear vibration signals,this paper takes the gearbox fault vibration signals as the research object,and proposes a gear fault diagnosis research method based on VMD and manifold learning.First of all,a new adaptive signal decomposition method,variational mode decomposition(VMD),was selected to extract the characteristics of fault signals for the problems of the gearbox fault vibration signals with non-linear and non-stationary characteristics and difficult to extract the sensitive fault characteristics by traditional methods.At the same time,the basic principle of VMD algorithm is introduced in detail.Aiming at the problem that the selection of parameters of VMD algorithm has a great influence on the extraction of fault features,a swarm intelligence optimization algorithm based on the joint criteria of minimum entropy and maximum kurtosis is proposed to select the two important parameters of VMD.Through the verification and analysis of the measured gear signals,the results show that the method can effectively and accurately extract the fault characteristics.Then the fault features with sensitive information are selected and multi-domain high-dimensional observation samples are constructed from the aspects of time domain,frequency domain and time frequency domain.According to the multi-dimensional complex characteristics of the samples,manifold learning algorithm is selected to reduce the dimension of the fault data.At the same time,the manifold learning algorithm is studied and its advantages and disadvantages are analyzed.Aiming at the problem that the intrinsic dimension of manifold learning and the value of the nearest neighbor domain are not easy to be determined,the false nearest neighbor method is proposed to estimate two parameters at the same time.The correctness and validity of the proposed method are verified by testing and analyzing the manifold learning with the standard test data.Finally,with the measured experimental data of gear box fault as the research object,the feature attributes of gear are extracted and reduced effectively by VMD,and the fault state diagnosis is carried out by combining multiple manifold learning with each classifier,which verifies the validity and feasibility of the fault diagnosis model based on VMD and manifold learning algorithm.