Research On Fault Recognition Of Rolling Bearing Based On Machine Learning

As the key component of the rotating mechanical equipment, the rolling bearing plays an important role in the safe operation of the rotating mechanical equipment.Owing to the vibration signal of bearing fault is non-linear and non-stationary, and its sample quantity is small. Selecting appropriate feature extraction and pattern recognition methods is the key for rolling bearing fault diagnosis and improving the fault recognition rate.The fault mechanism and the commonly used fault diagnosis methods of rolling bearing are introduced in this paper. According to the non-linear and non-stationary characteristics of bearing vibration signals, Hilbert Huang transform is selected as the feature extraction method, whose advantages and disadvantages are analyzed in this paper. To solve the problem of existing false weight during the process of empirical mode decomposition, the correlation coefficient is introduced to select the effective intrinsic mode function component for Hilbert transform. The simulation results show the effectiveness of the improved method.As the bearing fault diagnosis belongs to the typical problem of small sample size, support vector machine which adapts to small sample classification is selected as the classifier. Because the bearing signal is non-linear and has various types, the non-linear support vector machine and multi-class support vector machine are introduced in this paper. The parameters of support vector machine are optimized by the improved particle swarm optimization algorithm, and the effectiveness of the improved algorithm is verified by experiments.The marginal spectrum contains a large amount of redundant information which leads to the increasing of the complexity of the classifier, then the sparse coding is introduced, thus fault diagnosis model of rolling bearing HHT-SC-SVM is established.Through abstracting marginal spectrum secondly to establish a unified dictionary library to get sparse representation, and the sparse coefficients are put into the classifier to train and predict.SKF-6205-2RS bearing test rig data is used to analyze and verify the HHT-SC-SVM model which is proposed in this paper. The experiment results show that the fault identification rate of ten kinds of bearing fault in fan end is 98.25%, and the fault identification rate of ten kinds of bearing fault in drive end is 99.5%.Compared with the HHT-SVM model, the fault recognition rate is improved by 9%and 0.5% respectively.