Abstract:  Aiming at the problem of difficulty in obtaining and identifying the fault information when the rolling bearing was working, a random forest (RF) based fault identification method for rolling bearings was proposed. First, the rolling bearing vibration signal was collected and the feature vector of the original vibration data based on the time-domain statistical indicators was extracted. Then, the fault identification model based on random forest is established, and the test set was used to verify the classification results. The decision tree with a higher recognition rate was given a greater weight, so that the corresponding decision tree could play a greater role in the future classification process. Finally, the verification set was used to verify the final classification result. The effectiveness of the proposed method was verified through a multi-domain and multi-channel rolling bearing fault feature data set. The research results show that the fault diagnosis of rolling bearings based on random forest can achieve good identification results under different speeds and variable operating conditions, and the classification accuracy is as high as 96%; comparing with the traditional classification of back propagation(BP), k-nearest neighbor classification(KNN), and support vector machines(SVM), the comparison results of the detectors show that the accuracy of random forest classification is significantly higher than that of traditional classifiers at different speeds.
Key words:  rolling bearing; fault identification; feature extraction; random forest (RF)

WANG Lan-lan, ZHU Jie, ZHOU Zheng-ping, et al. Fault identification method of rolling bearing based on random forest［J］.Journal of Mechanical & Electrical Engineering, 2021,38(12):1599-1604.