结合动态时间弯曲与决策树方法的液体火箭发动机故障诊断  

Fault Diagnosis of Liquid Rocket Engine by Dynamic Time Warping Combined with Decision Tree Method

作　　者：胡小平[1] 韩泉东[1] 李京浩[1] 

机构地区：[1]国防科技大学航天与材料工程学院,湖南长沙410073

出　　处：《国防科技大学学报》2007年第4期1-5,共5页Journal of National University of Defense Technology

基　　金：国家自然科学基金资助项目(50376073)

摘　　要：对某大型液体火箭发动机的热试车数据及通过发动机模型仿真得到的故障数据进行动态时间弯曲分析,得到弯曲路径集,然后结合决策树方法进行了故障检测和诊断。对于故障试车没有出现漏报警和误报警,对于正常试车没有出现误报警。通过与神经网络、支持向量机等方法所得结果的对比,证明该方法可以成功地应用于火箭发动机的故障检测和诊断。Through dynamic time warping analysis to the hot-fire test data and simulated fault data of a certain liquid rocket engine,the warped path sets were obtained.Combined with the decision tree method,fault detection and diagnosis were carried out.Results show that there were no failing report and no misstatement for fault tests,and no misstatement for normal tests.Compared with those results by ANN and SVM methods,the successful application of the dynamic time warping method combined with decision tree was proved.

关 键 词：液体火箭发动机 故障检测和诊断 数据挖掘 动态时间弯曲 决策树 

分 类 号：V448.1[航空宇航科学与技术—飞行器设计]