Research On Fault Diagnosis And Performance Degradation Evaluation Method Of Rolling Bearing

In recent years,with the implementation of intelligent manufacturing strategy,mechanical equipment has become more intelligent and precise.As an extremely critical part of mechanical equipment,the running state of rolling bearing is closely related to the performance of the whole equipment.According to statistics,about 30% of all mechanical equipment failure events are caused by rolling bearing failure.In addition,based on the easily damaged characteristics of the rolling bearing,if the initial damage and performance changes are not detected as early as possible,it will inevitably cause some losses.Therefore,how to construct a reasonable and effective diagnostic evaluation model to monitor and analyze the service status of rolling bearings has become an important direction of current research.This paper focuses on the study of the health status of rolling bearings,mainly to extract the fault features from vibration data by appropriate processing methods,and to establish fault identification and performance evaluation models.The main research contents are as follows:(1)The vibration mechanism of rolling bearing is studied.Firstly,the basic components of rolling bearing and the functions of each part are summarized,and some failure modes of rolling bearing are described in detail;Secondly,the vibration causes and characteristic frequencies of rolling bearing are analyzed systematically,some types of vibration are listed,and the calculation formulas of characteristic frequencies are given.Finally,the vibration signal processing methods of rolling bearing are analyzed,and the basic principles and algorithm procedures of time-domain,frequency-domain analysis and variational mode decomposition are mainly discussed.(2)According to the parameter-dependent characteristics of the wavelet kernel extreme learning machine(WKELM),which makes the effect of the rolling bearing fault classifier model poor,a fault classification method based on an improved grey wolf optimizer algorithm is proposed to optimize wavelet kernel extreme learning machine.Firstly,the fault signal features are extracted by combining the variational mode decomposition(VMD)method and the singular value decomposition method;Secondly,when initializing the population,the opposition-based learning strategy is introduced into the grey wolf optimizer(GWO),and the levy flight strategy is used to update the individual position of the population,in order to enhance the population diversity and global optimization ability of the algorithm;Finally,the improved GWO algorithm is applied to optimize the parameters of wavelet kernel extreme learning machine to obtain the optimal parameter combination and construct the optimal classifier model.The comparative experimental results show that the proposed method has better fault recognition effect,faster training speed and stronger stability.(3)An improved support vector data description(SVDD)evaluation model is proposed for the performance degradation of rolling bearing.Firstly,considering that the traditional indexes can not completely represent the state change of bearing,the time-domain features and the energy spectrum entropy of VMD selected by the entropy method are used as feature vectors;Secondly,since the subjective selection will affect the accuracy of the evaluation results,the improved GWO algorithm is used to optimize the parameters of SVDD,and the optimal SVDD evaluation model is established to describe the trend of bearing performance degradation;Finally,the threshold is set adaptively to determine the time of early failure.The experimental results show that the evaluation index based on the improved SVDD evaluation method is more effective and can detect minor faults earlier;At the same time,in order to verify the accuracy of the evaluation model,the results are analyzed by VMD and Hilbert envelope demodulation.