Combined Failure Extraction And Classification Of Low-Speed And Heavy-Load Slewing Bearings

As a new type of mechanical component,large-size,low-speed and heavy-load slewing bearings have been widely used in hoisting and conveying machinery,construction machinery,material processing machinery,etc.Slewing bearings rotate in extremely low speed with various directions and angles.Slewing bearings has high bearing capacity and work in complex environment.In order to guarantee normal operation,protect human and mechanical safety,and reduce property loss,it is crucial to conduct real-time condition monitoring and fault diagnosis on slewing bearings.The thesis made a detailed and systematic review on current methods at home and abroad of condition monitoring and fault diagnosis of slewing bearings.Structure characteristics,typical failure,application background as well as research significances were analyzed and summarized in an all-round way.Overall existing research approaches,achievements,advantages and disadvantages,and application value on condition monitoring and fault diagnosis of slewing bearings were listed and compared.The research orientation,theoretically,was established based on the review,which provided support to experimental analysis and research ideas.The thesis,then,summarized and studied the common methods on vibration signal denoising,fault feature extraction,parameter selection and pattern recognition of rolling bearings.As a type of special rolling bearings,the research on rolling bearings could provide referential value to low-speed and heavy-load slewing bearings.The difficulty and the key on feature extraction were to extract the weak fault information from the strong background noise.Time-domain signal decomposition and reconstruction was an effective method to denoise signals,aiming to raise the signal to noise ratio and remain characteristics of original signals.Constituted by characteristic parameters,characteristic vectors could be taken as the representation of original signals and used in machine learning and classification.The practical vibration signals were collected and selected and other physical and chemical parameters were detected.Different from artificial and single fault experiments conducted on simulant test platforms,this thesis focused on combined natural faults based on complicated equipment.Thus,the design of experimental scheme,selection and arrangement of sensors,and establishment of measurement system were more complicated.A great deal of work on data collection was conducted,which provided precious resources to the data analysis and the following study.According to simulation analysis and experimental verification,this thesis,firstly,adopted the advanced piecewise aggregate approximation-neighborhood correlation plot method conducted feature extraction of all operational periods of the slewing bearing.The shape,direction,and parameter information of the three-dimensional ellipsoid-surface-fitting diagrams were taken as characteristics to effectively divide the different failure stages of the slewing bearing.The qualitative judgment on the degree of severity of combined failure was achieved.Based on the data-driven thought,the maximum correlated kurtosis deconvolutionensemble empirical mode decomposition-approximate entropy method was proposed in the perspective of accordance with signal denoising-feature extraction-pattern recognition.The information of three different operation stages of the slewing bearing was recognized and classified based on the parameter characteristics from the phase space,which obtained the high accuracy.In order to highlight the superiority of the proposed method,a series of comparative trials were made,including denoising methods,feature selection methods,etc.