基于无迹卡尔曼滤波的液体火箭发动机故障诊断  

Fault Diagnosis of Liquid Rocket Engine Based on Unscented Kalman Filter

作　　者：许亮[1] 芦弘炜 王闻浩 薛薇 XU Liang;LU Hongwei;WANG Wenhao;XUE Wei(Tianjin Key Laboratory for Control Theory&Applications in Complicated Systems,School of Electrical Engineering and Automation,Tianjin University of Technology,Tianjin 300384,China;China Academy of Aerospace Science and Innovation,Beijing 100083,China)

机构地区：[1]天津理工大学电气工程与自动化学院天津市复杂系统控制理论与应用重点实验室,天津300384 [2]中国航天科技创新研究院,北京100083

出　　处：《载人航天》2024年第4期516-525,共10页Manned Spaceflight

基　　金：国家自然科学基金(61975151)。

摘　　要：针对火箭发动机故障数据难以获取的问题,设计了一种基于无迹卡尔曼滤波(UKF)的液体火箭发动机故障诊断算法。采用MATLAB/Simulink平台搭建了液体火箭发动机故障仿真模型,实现发动机正常运行仿真和预燃室氧阀门故障、氧主泵汽蚀、氢主涡轮叶片脱落3种故障仿真。将正常运行仿真值与设计值、试车值进行了对比。结果表明:模型参数与设计值最大误差不超过5%,仿真精度较高;仿真参数变化趋势与试车值基本一致,且稳态值误差较小。使用UKF算法求取发动机正常运行阈值范围,并对故障序列进行滤波处理,若故障数据连续3次超出阈值区间,且在0.1 s内有至少2个涡轮泵发出报警,则判定故障发生,故障发生时间为第2个涡轮泵报警时间。使用设计算法对3种故障序列进行诊断,判定故障发生时间分别为20.08 s、20.05 s、20.18 s。相比于传统红线阈值算法,文中所设计算法响应更为及时,且误报率较低。A liquid rocket engine fault diagnosis algorithm based on unscented Kalman filter(UKF)was designed.To address the problem that it is difficult to obtain rocket engine fault data,a liquid rocket engine fault simulation model was built using MATLAB/Simulink platform,which realized the engine normal operation simulation and three kinds of fault simulation,including the oxygen valve failure in the pre⁃combustion chamber,the vapor corrosion of the oxygen main pump,and the de⁃tachment of the hydrogen main turbine blades.The normal operation simulation values were com⁃pared with the design values and the test run values.The results showed that the maximum error be⁃tween the model parameters and the design value was not more than 5%,the simulation accuracy was high,the simulation parameter change trend was basically consistent with the test value,and the error of the steady state value was small.Finally,the UKF algorithm was used to find the engine normal operation threshold range and to filter the fault sequence.If the fault data exceeded the threshold interval for three consecutive times and at least two turbopumps alarmed within 0.1 s,the fault was determined to have occurred and the time of fault occurrence was the time of the second turbopump alarm.Using the designed algorithm to diagnose the three fault sequences,it was deter⁃mined that the fault occurrence time was 20.08 s,20.05 s,and 20.18 s,respectively.Compared with the traditional red line threshold algorithm,the proposed algorithm has more timely response and lower false alarm rate.

关 键 词：故障诊断 液体火箭发动机 无迹卡尔曼滤波 故障仿真 红线阈值算法 

分 类 号：V434.3[航空宇航科学与技术—航空宇航推进理论与工程]