支撑向量机在航空发动机故障诊断中的应用研究  

Applied Research of Support Vector Machines in the Fault Diagnosis of Engine

作　　者：许将军[1] 侯宽新[2] 高丽霞[1] 

机构地区：[1]中国民航飞行学院航空工程学院,四川广汉618307 [2]中国民航飞行学院广汉分院,四川广汉618307

出　　处：《自动化技术与应用》2011年第4期69-71,共3页Techniques of Automation and Applications

摘　　要：航空发动机故障样本有限,利用传统的统计识别方法故障诊断,正确率不高。支撑向量机能解决小样本的故障分类识别问题。研究Support Vector Machine(简称SVM)核函数对识别精度的影响,并把SVM与最大似然法、马氏距离法、最小距离法进行比较,结果表明SVM核函数对故障识别正确率影响不大,基于SVM的航空发动机故障诊断精度高于传统的统计识别方法。As lack of fault samples of engine,the accuracy of traditional classification methods is always unsatisfactory.Support vector machine（SVM）is a good method for solving limited sample problem.the influences of different kernel functions on classification accuracy are researched,and the classification results by SVM are compared with that of other methods（such as Maximum likelihood method 、Minimum Distance method、Mahalanobis Distance）.Experiment results show that the SVM method has better recognition accuracy than traditional algorithms,and the recognition accuracy is almost identical for different kernel functions.

关 键 词：支撑向量机 核函数 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]