支持向量机的涡扇发动机试车故障诊断

Application of Support Vector Machine in Fault Diagnosis of Aeroengine Test

针对涡扇发动机在试车过程中缺少有效诊断气路故障方法的问题,为某型涡扇发动机建立了基于支持向量机的气路故障诊断系统。支持向量机算法专门针对小样本集合设计,能够在小样本情况下获得较大的推广能力。该系统建立了发动机非线性稳态模型,生成包含八种典型气路故障的故障样本库,采用支持向量机对故障特征和故障模式进行关联,并用训练好的向量机网络对故障分类。利用该型涡扇发动机试车数据对该系统进行的验证,诊断正确率在80%以上。研究表明该方法能基本满足该涡扇发动机地面试车故障诊断的要求,具有较好的工程应用前景。

To diagnose turbo - fan engine＇ s gas path fault in test , a fault diagnosis system based on Support Vector Machines （SVM） is established . SVM is special designed for small samples set and can obtain good generalization ability despite of insufficient samples. With the nonlinear model of the engine , the emulated fault data base that contains eight representative gas path faults has been built for the engine. SVM is adopted to set up correlation between features and fault patterns and carry out classifier function. The case application in ground test gains commendable results. This method is directive and contributive special for the engine in test.