Research On Fault Classification Algorithm Of Rolling Bearing For Rotating Equipment Based On NN And Xgboost

Researching the theory and methods of mechanical vibration,explaining various complex motion phenomena in mechanical structural systems,achieving effective control of complex equipment vibration and noise,and effective use of vibration are important means to improve the performance of mechanical equipment.Rolling bearings are one of the wearing parts of rotating equipment.Accurate and timely fault diagnosis is essential to ensure the reliable operation of rotating machinery.This paper studies the classification algorithm of rolling bearing faults in rotating equipment.Aiming at the problem of bearing fault detection,this paper proposes a bearing noise fault classification method based on neural network filtering and integrated learning classification.Aiming at the principle that equipment operation status abnormal information or fault information is always superimposed on equipment health status information,this paper studies the algorithm of bearing health status signal stripping based on neural network adaptive filtering,and the algorithm promotion strategy based on integrated learning,based on learning The combined strategy improves the effect of strong classifiers,and a bearing fault classification algorithm based on neural network filtering and integrated learning classification is proposed for rolling bearing fault detection and classification.Filter out the normal components of the bearing signal through the neural network filter;then extract the signal features,use the estimation method to select the distinguishing features,and construct new features through the multi-scale feature extraction method;finally,input the feature signals into the integration based on the combination strategy The quasi three-level classification classifier is learned for training and testing,and the effectiveness of this method is verified through experiments.Aiming at the feature selection problem of rolling bearing fault classification,this paper studies the feature selection algorithm based on gradient lifting decision tree,and proposes a feature selection algorithm XGB-C-FS based on XGBoost.Based on the importance evaluation of XGBoost model,through the combination of the two methods of sequence forward addition,floating backward elimination and feature correlation screening,the algorithm has high importance.It is enough to represent the feature set of the sample with more comprehensive features as the optimal feature subset.By comparing with CDET,XGBoost and other algorithms that select features based solely on importance ranking,the classification accuracy of the classifier is used as the evaluation index to verify the quality of the selected feature subset under different feature selection methods.The feature extraction and classification algorithms proposed in the paper were trained and tested under different types of bearing running noise data sets,which verified the superiority of the algorithm.