Research On Feature Extraction And Fault Diagnosis Of Rolling Bearing

As one of the most basic general industrial components in mechanical equipment,the rolling bearing plays an important role in ensuring the normal operation of industrial equipment and the safety production of enterprises.Therefore,it is necessary to conduct research on the fault detection and fault identification of rolling bearings.Based on the rolling bearing as the research object and the original vibration signals of the rolling bearing as the source information carrier,the dissertation extracts features from vibration signals of rolling bearings by the time-domain analysis method based on statistical characteristics and the empirical mode decomposition method as the input for the fault detection and fault identification,experiments with the designed fault detection method for the rolling bearing based on LOF-KPCA and the designed fault identification method for the rolling bearing based on SVMKPCA,and then validates the effectiveness and superiority of the proposed methods in the research of the rolling bearing fault diagnosis.The main work and research results of this dissertation are as follows:1.Study the time-domain feature extraction of vibration signals of the rolling bearing.The time-domain analysis method based on statistical characteristics is used to extract the features of the original vibration signals in different states of the rolling bearing,whose five types of time-domain parameter indexes will be used for the subsequent research on the rolling bearing fault diagnosis as the feature input;2.Study the feature extraction of vibration signals of the rolling bearing based on the empirical mode decomposition method.The empirical mode decomposition method is used to obtain several intrinsic mode functions of vibration signals in different states of the rolling bearing.The correlation coefficient method and the variance contribution rate method are used to eliminate illusory intrinsic components and to obtain the true intrinsic components.And the energy of the true intrinsic components is used as the feature vector for the fault diagnosis research in the subsequent chapters;3.Propose a fault detection method for the rolling bearing based on LOF-KPCA algorithms.Considering that the KPCA method is based on the global distribution of the data set for the fault detection and lacks the exploration of the local information of the data set,a local outlier factor-LOF method is introduced for the fault detection based on the KPCA method.The feature input data of different feature extraction methods is used to compare and discuss the performance of the proposed method on the rolling bearing fault detection.Comparative experimental results show that the proposed LOF-KPCA algorithm based on IMF energy performs better in fault detection of rolling bearings;4.Research on the fault identification method of the rolling bearing based on SVM-KPCA.Considering the multi-dimensional input characteristic data of rolling bearings,and the significant advantages of SVM algorithm in processing small sample data,the superiority of SVM algorithm in traditional classification algorithm is comparatively demonstrated.Based on the SVM algorithm,the KPCA algorithm is introduced to improve it.The improved traditional algorithm and the popular classification algorithm-random forest algorithm and XGBoost algorithm are compared to discuss the performance of rolling bearing fault recognition under different feature extraction methods.Comprehensive comparison found that the fault recognition effect of rolling bearings with SVM-KPCA algorithm based on IMF energy is better.