火箭推进

增压输送系统是火箭产生推力的核心系统,其工作可靠性直接关系到运载火箭安全性。以某型号运载火箭第三级作为故障检测对象,将遥测数据抽样和预处理,降低数据规模和提高故障检测效率,模拟常见故障和偶发故障扩大故障数据规模,使得故障数据和正常数据均衡,通过添加归一化时间序列区分火箭工况。经处理后的遥测数据用于训练BP神经网络,故障检测结果表明:BP神经网络可以利用火箭遥测数据对增压输送系统进行有效的故障检测,正常数据误报率和故障数据漏报率均较低,且可以满足实时故障检测需求。

The pressurized delivery system is the core system of rocket to generate thrust,and its working reliability is directly related to the safety of launch vehicle. Taking the third stage of a certain type of launch vehicle as the fault detection object,the telemetry data is sampled and preprocessed to reduce the data scale and improve the fault detection efficiency.By simulating common faults and occasional faults to expand the fault data scale,the fault data and normal data are balanced,and the rocket working conditions are distinguished by adding normalized time series. The processed telemetry data is used to train BP neural network.The fault detection results show that BP neural network can effectively detect the fault of pressurized delivery system by using therocket telemetry data,and the false alarm rate of normal data and false alarm rate of fault data are both low,and it can meet the real-time fault detection requirements.

引言
1 数据预处理
2 故障模式的数值仿真
3 神经网络故障分类算法
4 算法测试分析
5 结论

表1 故障检测参数<br/>Tab.1 Fault detection parameters

表1 故障检测参数
Tab.1 Fault detection parameters

图1 某遥测参数运行图<br/>Fig.1 Operation of a certain telemetry parameter

图1 某遥测参数运行图
Fig.1 Operation of a certain telemetry parameter

图2 t时刻正常遥测参数运行图<br/>Fig.2 Operation of normal telemetry parameters at time t

图2 t时刻正常遥测参数运行图
Fig.2 Operation of normal telemetry parameters at time t

图3 t时刻随机故障参数运行图<br/>Fig.3 Operation of random fault parameters at time t

图3 t时刻随机故障参数运行图
Fig.3 Operation of random fault parameters at time t

图4 第1 200抽样点发生A1故障参数运行图<br/>Fig.4 Operation of A1 fault parameter at the 1 200th sampling point

图4 第1 200抽样点发生A1故障参数运行图
Fig.4 Operation of A1 fault parameter at the 1 200th sampling point

图5 B1故障参数运行图<br/>Fig.5 B1 fault parameter operation

图5 B1故障参数运行图
Fig.5 B1 fault parameter operation

图6 C1故障参数运行图<br/>Fig.6 C1 fault parameter operation

图6 C1故障参数运行图
Fig.6 C1 fault parameter operation

图7 D1故障参数运行图<br/>Fig.7 D1 fault parameter operation

图7 D1故障参数运行图
Fig.7 D1 fault parameter operation

表2 ANN分类模型参数<br/>Tab.2 ANN classification model parameters

表2 ANN分类模型参数
Tab.2 ANN classification model parameters

图8 BP神经网络示意图<br/>Fig.8 Schematic diagram of BP neural network

图8 BP神经网络示意图
Fig.8 Schematic diagram of BP neural network

图9 训练集算法输出值及标签值<br/>Fig.9 Output value and tag value of training set algorithm

图9 训练集算法输出值及标签值
Fig.9 Output value and tag value of training set algorithm

图10 测试集算法输出值及标签值<br/>Fig.10 Output value and tag value of test set algorithm

图10 测试集算法输出值及标签值
Fig.10 Output value and tag value of test set algorithm

表3 神经网络故障检测结果分析<br/>Tab.3 Results analysis of neural network fault detection

表3 神经网络故障检测结果分析
Tab.3 Results analysis of neural network fault detection

[1] 廖少英.液体火箭推进增压输送系统[M].北京:国防工业出版社,2007.
[2] SCHWABACHER M,WATERMAN R.Pre-launch diagnostics for launch vehicles[C]//2008 IEEE Aerospace Conference.Big Sky,MT,USA:IEEE,2008:1-8.
[3] RODNEY A.Evaluation of anomaly detection capability for ground-based pre-launch shuttle operations[M]//Aerospace Technologies Advancements.[S.l.]:In Tech,2010.
[4] FERRELL B,BROWN B,HALL D,et al.Usage of fault detection isolation and recovery(FDIR)in CxP launch operations[C]//SpaceOps 2010 Conference.Huntsville,Alabama.Reston,Virigina:AIAA,2010.
[5] SCHWABACHER M,MARTIN R,WATERMAN R,et al.Ares I-X ground diagnostic prototype[C]// Aerospace Conference 2010.Atlanta,Georgia.Reston,Virginia:AIAA,2010.
[6] 李艳军.新一代大推力液体火箭发动机故障检测与诊断关键技术研究[D].长沙:国防科学技术大学,2014.
[7] 陈海峰,赵政社,张少博.基于振动检测的发动机故障诊断算法研究[J].火箭推进,2012,38(6):62-69.CHEN H F,ZHAO Z S,ZHANG S B.Research on liquid rocket engine fault diagnosisAlgorithm according to vibration detection[J].Journal of Rocket Propulsion,2012,38(6):62-69.
[8] 党锋刚,马红宇,李春红,等.液氧煤油发动机稳态故障仿真分析[J].火箭推进,2012,38(2):27-31.DANG F G,MA H Y,LI C H,et al.Steady-state fault simulation and effect analysis of LOX/kerosene rocket engine[J].Journal of Rocket Propulsion,2012,38(2):27-31.
[9] 孙迪钢.基于深度学习的轨道转辙机故障检测系统[D].广州:华南理工大学,2018.
[10] 司佳.面向工业大数据的智能故障诊断方法研究[D].济南:山东大学,2018.
[11] 吴建军,程玉强,崔星.液体火箭发动机健康监控技术研究现状[J].上海航天(中英文),2020,37(1):1-10.
[12] 李京浩.基于数据挖掘技术的液体火箭发动机故障检测和诊断研究[D].长沙:国防科学技术大学,2007.
[13] 张翔,徐洪平,安雪岩,等.液体火箭发动机稳态运行故障数据聚类分析研究[J].火箭推进,2015,41(2):118-122.ZHANG X,XU H P,AN X Y,et al.Clustering analysis for fault data in steady process of liquid propellant rocket engine[J].Journal of Rocket Propulsion,2015,41(2):118-122.
[14] 耿辉,张翔,张素明,等.一种基于聚类分析的液体火箭发动机稳态过程故障诊断方法[J].火箭推进,2014,40(5):86-91.GENG H,ZHANG X,ZHANG S M,et al.Cluster analysis based diagnosis method for fault in steady process of liquid propellant rocket engine[J].Journal of Rocket Propulsion,2014,40(5):86-91.
[15] 李宁宁.粒子群算法研究及其在航天发动机健康管理中的应用[D].天津:天津理工大学,2019.
[16] 谢廷峰,刘洪刚,黄强,等.液体火箭发动机地面试车实时故障检测参数的选取[J].航天控制,2008,26(6):77-81.
[17] 黄敏超,冯心,张育林.应用神经网络识别液体火箭发动机的故障模式[J].推进技术,1993,14(6):32-38.
[18] 黄敏超,张育林,陈启智.动态神经网络在液体火箭发动机故障检测与分离中的应用[J].航空动力学报,1996,11(2):149-152.

备注

引言

1 数据预处理

2 故障模式的数值仿真

3 神经网络故障分类算法

4 算法测试分析

5 结论

期刊信息