Research On Fault Diagnosis Of Rolling Bearing Based On Variational Mode Decomposition

With the rapid development of China's high-speed railways,the operation speed has been continuously improved,and its safety issues have received more and more attention.The health of the rolling bearings has a direct impact on the safety of the train.Therefore,it is of great significance to study the fault diagnosis method and fault pattern identification with the goal of rolling bearings.The main contents are as follows:Firstly,Review the development history of rolling bearing technology,and introduce the feature extraction method and pattern recognition algorithm.Analysis of the causes and characteristics of bearing failures provides a theoretical basis for the analysis below.Secondly,this paper introduces the theoretical basis and theory of variational mode decomposition(VMD)in detail.The performance of the VMD algorithm against two types of modal aliasing(intermittent event interference or modal aliasing caused by the close proximity of modal component frequencies)and noise robustness is analyzed by simulation signals,and empirical mode decomposition is performed.(EMD)relative,the results show that the VMD algorithm is more advantageous.The influence of the penalty factor and the number of modes on the decomposition performance of the VMD algorithm is quantitatively analyzed.The results show that the selection of the number of modes has the greatest impact on the decomposition performance.Through the rolling bearing simulation signal,the influence of the initial center frequency on the decomposition efficiency and performance of the VMD algorithm is analyzed.The results show that the iterative efficiency of the two different initial center frequencies is different,and the center frequency deviation of the obtained modal components is larger.For different penalty factors,the center frequency of the modal components obtained by the algorithm decomposition is regional convergence;As the penalty factor increases,the algorithm iterative efficiency appears as a range stable phenomenon.Then,based on the above analysis,based on the above analysis,a VMD optimization method based on trend estimation guidance is proposed.That is,the decomposition signal is processed by the trend estimation method,the initial guidance center frequency required for the initialization of the VMD algorithm is obtained,and the VMD algorithm initialization is completed by using the obtained parameters,thereby solving the problem of VMD algorithm parameter selection.Finally,the effectiveness and efficiency of the proposed method are verified by simulation signals.The accuracy of the proposed method is verified by the full life cycle acceleration test bed signal and the high-speed train wheel running test platform signal,and compared with the traditional method.Further illustrates the advantages of the proposed method.Finally,in order to realize the automatic identification of the rolling bearing failure mode and ensure the recognition accuracy,a rolling bearing fault combining the optimized VMD algorithm with the composite multi-scale permutation entropy(CMDE),the maximum correlation minimum redundancy(mRMR)method and the extreme learning machine(ELM)is proposed.Pattern recognition method.The multi-scale dispersion entropy(MDE)is introduced.According to the defects of the MDE method,the feature extraction of rolling bearings based on composite multi-scale dispersion entropy(CMDE)is proposed.The CMDE and MDE are compared and analyzed by simulation signals,and the CMDE and the optimized VMD algorithm are combined to acquire the characteristics of the signal.Aiming at the problem of high-dimensional and redundant multi-scale feature set,the mRMR algorithm is introduced to calculate the importance and separability of each scale feature in the vector set,select better features,construct a new fault feature vector set,and combine the extreme learning machine(ELM)realizes the identification of failure modes.Finally,the measured signals verify that the proposed method can accurately and reliably realize fault identification and classification.