基于随机RF的集成SVM故障诊断改进算法  

Random RF and Ensemble SVMs Based Fault Diagnosis Method

作　　者：夏丽莎[1] 吕文元[1] XIA Li-sha;LV Wen-yuan(School of Business,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学管理学院

出　　处：《工业工程与管理》2019年第3期85-90,共6页Industrial Engineering and Management

基　　金：国家自然科学基金资助项目(71572113)

摘　　要：针对工业系统监控数据不均衡导致的故障状态难以被识别问题,提出一种基于随机旋转森林的集成支持向量机(RRFESVM)故障诊断算法,通过将监控数据进行属性随机分割、组合、PCA变换和样本有放回重采样,组建多个新训练子集并使用支持向量机算法进行训练,得到多个支持向量机故障诊断基分类器,集成得到强分类器.由此既保证基分类器之间的差异性,又保证故障诊断精度和分类器性能稳定性,从而解决故障诊断易偏置问题,提高作为少数类的故障状态实时诊断准确率。亚轨道飞行器再入过程实验与TE化工过程实验都表明RRFESVM故障诊断算法能够有效提升不均衡数据情况下的实时故障诊断性能。The system failure states are hard to be identified with imbalanced monitoring dataset.A RRFESVM(Random Rotation Forest based Ensemble SVMs) fault diagnosis algorithm was proposed.After the attributes of monitoring data were segmented and combined randomly,the resample step was implemented and Principle Component Analysis(PCA) was used for feature transformation.Then a number of new training subsets were obtained and Support Vector Machine(SVM) fault diagnosis sub-classifiers can be trained with these new training subsets accordingly.Eventually a voting-integrated strong RRFESVM fault diagnosis classifier was developed.In this way,the diversity of sub-classifiers,the fault diagnosis accuracy improvement and the classifier performance stability requirement can be achieved.The experimental results of suborbital reusable launch vehicle return course and Tenessee Eastman(TE) chemical process show that the proposed RRFESVM fault diagnosis algorithm can improve the diagnosis performance on the imbalanced dataset.

关 键 词：数据不均衡 随机旋转森林 集成支持向量机 故障诊断 

分 类 号：F277[经济管理—企业管理]