Xgboost在滚动轴承故障诊断中的应用

摘要

Due to the difficulty for selecting the data feature for different bearings and the low accuracy in fault diagnosis of rolling bearings using single classifier method, a rolling bearing fault diagnosis algorithm with eXtreme Gradient Boosting (Xgboost) based on classification and regression tree is proposed. The Xgboost is an ensemble learning method which contains a variety of classifiers. The accuracy of rolling bearing fault diagnosis is improved by the"boosting"thought of the Xgboost. First of all, the time domain statistical indicators extracted from the vibration signals of the rolling bearings are used as feature vectors. Then, the Xgboost algorithm is utilized for the fault diagnosis of the rolling bearings. Comparing the vibration data obtained in the bearing test on the SQI-MFS testing platform with the diagnostic results of traditional algorithms (SVM, kNN and ANN) and single classification and regression tree, it is concluded that the Xgboost algorithm is superior to the above algorithms, and the computation time is less than that of the traditional boosting decision tree algorithm.%针对不同轴承数据特征选择困难和单个分类器方法在滚动轴承故障诊断中精度较低的问题,提出一种基于分类与回归树的Xgboost(eXtreme Gradient Boosting)轴承故障诊断算法.Xgboost是包含多个分类器的集成学习方法.通过Xgboost的"提升"思想来提高滚动轴承故障诊断的精度.首先,从滚动轴承的振动信号中提取时域特征参数;然后利用Xgboost算法对滚动轴承故障进行诊断.将SQI-MFS实验平台的轴承振动数据,与传统分类器(支持向量机、邻近算法和人工神经网络)以及单个分类回归树的诊断结果相比,结果表明Xgboost在轴承故障诊断率上优于上述几种算法,且计算时间比传统提升决策树算法短.