Research On Vibration Characteristic Analysis And Intelligent Diagnosis Method Of Key Components Of Gearbox

As a key component of a mechanical transmission system,gearboxes are widely used due to their high transmission efficiency and strong carrying capacity.Gears and rolling bearings,as their key components,play the roles of "transmitters" and "joints" in the change of speed,the transmission of torque and the support of the box and the rotation of the shaft system.With the development of machinery equipment in the direction of large-scale,intelligent,and high-speed,condition monitoring system(CMS)is used to collect gearbox status information for its health management.However,the vibration signals of the gearbox and its key components obtained by CMS show the characteristics of energy attenuation and complex time-varying modulation after being transmitted through multiple interfaces and complex paths;In addition,the gearbox and its key components have been in service for a long time in harsh working conditions and extreme environments,causing the latent impact components induced by early failures to be often submerged in strong background noise,making its vibration characteristics extremely complex,and the lack of a priori fault sample data,poses a great challenge for the failure to identify the key components of gears,bearings and so on.In order to solve the problem of accurate and efficient status identification in the absence of the vibration characteristics of the key components of the gearbox and its sample data,based on nonlinear Hertzian contact theory,vibration signal decomposition method,nuclear principal component analysis,DBN deep learning theory,carry out research on the analysis of the vibration characteristics of the key components of the gearbox,the cleaning of vibration signals,the construction of high-quality fault feature sets,the migration of fault sample data,and accurate and efficient fault status identification.The specific content includes:(1)In terms of the analysis of vibration characteristics,starting from the principle of the failure of the key components of the gearbox,the theoretical models of the fault dynamics of the gear and the rolling bearing are established respectively to clarify the failure excitation mechanism and its action mechanism;Starting from the results of the failure,a digital simulation vibration signal model of gears and bearings is constructed,the composition of the failure vibration signal is analyzed,and the vibration characteristics of the key components of the gearbox are explored.(2)In terms of vibration signal cleaning,considering the impact of strong background noise on the vibration signal,the nonlinearity,non-stationarity and modulation characteristics of the fault vibration signal,and the sensitive characteristics of the fault impact energy change,the signal impact energy enhancement and energy evaluation are proposed to extract the key composition,to achieve the purpose of cleaning the vibration signal of the key parts of the gearbox.(3)In terms of research on the construction method of high-quality fault feature set,considering that the fault feature set has the characteristics of high dimensionality,multiple levels,and different quality,a screening method of explicit fault feature index combining neural network and nuclear principal component analysis is proposed,and corresponding fault feature screening criteria are constructed based on this,so as to achieve the dimensionality reduction and quality improvement of the fault feature set.(4)In terms of research on the migration of fault sample data,accurate and efficient state identification,in view of the problems of "difficult selection" and "missing data" of model structure parameters faced in the actual application of intelligent identification algorithms,the structural parameters of the deep belief network model are optimized based on the improved particle swarm algorithm,and the migration of fault data is realized through digital simulation of vibration signals,and the relationship between the proportion of migration data and the accuracy of fault identification is verified through experiments.