基于随机森林算法的旋转机械齿轮组故障诊断  

Fault diagnosis of rotating machinery gearbox based on random forest algorithm

作　　者：王子兰 杨瑞 WANG Zilan;YANG Rui(College of Mathematics and Systems Science,Shandong University of Science and Technology,Qingdao,Shandong 266590,China;College of Electrical Engineering and Automation,Shandong University of Science and Technology,Qingdao Shandong 266590,China)

机构地区：[1]山东科技大学数学与系统科学学院,山东青岛266590 [2]山东科技大学电气与自动化工程学院,山东青岛266590

出　　处：《山东科技大学学报（自然科学版）》2019年第5期104-112,共9页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61603223);山东省自然科学基金项目(ZR2016FB01);青岛市应用基础研究计划项目(16-5-1-7-jch,17-1-1-1-jch);中央高校基本科研业务费专项资金项目(18CX02093A);青岛市博士后研究人员应用研究项目;山东科技大学人才引进科研启动基金项目;西交利物浦大学重点项目建设专项资金(KSF-E-34)

摘　　要：针对单一的分类器用于旋转机械故障诊断时存在准确率不高的问题,提出一种基于随机森林算法的旋转机械齿轮组故障诊断方法。该方法利用随机森林多分类器组合决策树的思想,通过多分类器的组合学习提高故障诊断的准确率,并在风力涡轮动力传动系统故障诊断模拟器系统上进行了多工况多故障的实验验证。首先,收集多工况、多故障的齿轮传感器信号,提取传感器信号的时域特征作为随机森林的输入特征量。然后,利用构建好的随机森林模型进行齿轮组的故障诊断,并将随机森林算法的分类结果与支持向量机方法的分类结果进行对比。通过对故障诊断结果的分析,随机森林算法避免了复杂的寻参过程和传统分类器的过拟合现象,能够处理大规模数据集,通过分类器的组合,提高了故障诊断准确率,并缩短了分类模型的预测时间,具有较好的应用前景。As the accuracy rate of single classifier based fault diagnosis is low for rotating machinery,a fault diagnosis method based on the random forest algorithm is proposed in this paper.Hardware experiments were carried out on wind turbine drivetrain diagnosis simulator(WTDDS)to show the effectiveness of the proposed method.Initially,the sensor signals under multiple operation modes with different gear fault types were collected with time domain characteristics and extracted as input features of random forest.Afterwards,the fault diagnosis of the gearbox was carried out using the constructed random forest model,and the classification result using random forest algorithm was compared with that using support vector machine.Through the analysis of the fault diagnosis results,the random forest algorithm can process large-scale data sets and improve the accuracy of fault diagnosis by avoiding the over-fitting phenomenon of conventional classifier.It can also shorten the prediction time through the combination of weak classifiers with higher prospects as a potential application.

关 键 词：随机森林 旋转机械 故障诊断 决策树 特征提取 

分 类 号：TH132[机械工程—机械制造及自动化]