基于决策树方法的液体火箭发动机稳态段故障诊断  

Steady state fault detection and diagnosis of liquid propellant rocket engines based on decision tree method

作　　者：韩泉东[1] 胡小平[1] 王艳梅[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：《火箭推进》2007年第3期26-30,42,共6页Journal of Rocket Propulsion

基　　金：国家自然科学基金资助项目50376073。

摘　　要：以泵压式液体火箭发动机为研究对象,针对故障特征难以提取的问题,提出了在大量试车数据的基础上,应用决策树方法提取故障特征并进行稳态段故障检测与诊断的思路,对某型发动机大量热试车数据进行了实例分析,将得到的故障检测和诊断结果与神经网络等方法所得结果进行了比较。并利用仿真数据对该方法的性能进行了进一步验证。研究结果表明,决策树方法是一种有效、可靠的液体火箭发动机故障检测与诊断的新途径。Because of noise and the coupling of many factors in the hot fire tests of rocket engines, the detection of fault features was always difficult. The decision tree method was applied to analyze steady test data for liquid rocket engines. Detection and diagnosis results were compared with those of ANN method. Then the simulated data were used to verify the performance of decision tree method. It showed that decision tree method was an efficient, reliable and novel approach for the FDD of LRE.

关 键 词：液体火箭发动机 故障检测与诊断 数据挖掘 决策树 

分 类 号：V434[航空宇航科学与技术—航空宇航推进理论与工程]