火箭推进

为提高航天发射任务的可靠性和安全性,对液体火箭发动机异常智能检测技术进行了研究。针对传统方法存在的检测准确率低、依赖专家经验和先验知识、所需数据量大等问题,提出运用深度学习方法构建自编码式—生成对抗网络——训练基于发动机健康状态下的多源数据,测试基于输入数据的重构损失和鉴别分数,完成对液体火箭发动机异常状态的智能检测。某型号液体火箭发动机地面热试车实验数据的分析结果表明,该方法取得了96.55的测试准确率,并在利用邻近信息的条件下取得最高100的准确率,可有效用于液体火箭发动机的异常检测。

To improve the reliability and safety of the space launch mission, the intelligent anomaly detection technology of liquid rocket engine was studied.Aiming at the problems of traditional methods such as low detection accuracy, relying on expert experience and prior knowledge, and large amount of data needed, a deep learning was proposed to build an autoencoding generative adversarial network to train the multi-source data under the engine health state.The reconstruction loss and discriminant score of input data were tested, and the intelligent anomaly detection of liquid rocket engine was completed.The analysis results of the ground hot test data for a liquid rocket engine show that this method achieves 96.55 test accuracy, and the 100 accuracy can be obtained by using the neighboring information.It can be effectively used for the anomaly detection of liquid rocket engine.

引言
1 自编码式—生成对抗网络
2 基于自编码式—生成对抗网络的LRE异常检测
3 实验验证
4 结语

图1 AEGAN异常检测流程<br/>Fig.1 Flowchart of AEGAN anomaly detection

图1 AEGAN异常检测流程
Fig.1 Flowchart of AEGAN anomaly detection

图2 数据预处理
Fig.2 Data preprocessing

图3 AEGAN异常检测模型<br/>Fig.3 AEGAN anomaly detection model

图3 AEGAN异常检测模型
Fig.3 AEGAN anomaly detection model

图4 某型液体火箭发动机原理图<br/>Fig.4 Schematic diagram of a LRE

图4 某型液体火箭发动机原理图
Fig.4 Schematic diagram of a LRE

表1 异常检测结果混淆矩阵<br/>Tab.1 Confusion matrix of anomaly detection results

表1 异常检测结果混淆矩阵
Tab.1 Confusion matrix of anomaly detection results

表2 某型液体火箭发动机异常检测结果<br/>Tab.2 Anomaly detection results of a LRE

表2 某型液体火箭发动机异常检测结果
Tab.2 Anomaly detection results of a LRE

图5 某型液体火箭发动机异常检测结果分布图<br/>Fig.5 Distribution of a LRE anomaly detection results

图5 某型液体火箭发动机异常检测结果分布图
Fig.5 Distribution of a LRE anomaly detection results

图6 某型液体火箭发动机异常检测结果ROC曲线图<br/>Fig.6 ROC graph of anomaly detection results for a LRE

图6 某型液体火箭发动机异常检测结果ROC曲线图
Fig.6 ROC graph of anomaly detection results for a LRE

表3 邻域信息聚合的某型液体火箭发动机异常检测结果<br/>Tab.3 Anomaly detection results of a LRE based on neighborhood information aggregation

表3 邻域信息聚合的某型液体火箭发动机异常检测结果
Tab.3 Anomaly detection results of a LRE based on neighborhood information aggregation

图7 不同数量的邻域信息聚合后的异常检测结果<br/>Fig.7 Anomaly detection results after neighborhoodinformation aggregation with different amounts

图7 不同数量的邻域信息聚合后的异常检测结果
Fig.7 Anomaly detection results after neighborhoodinformation aggregation with different amounts

[1] 赵松波.基于改进优化算法的液体火箭发动机故障检测与诊断研究[D].天津:天津理工大学,2019.
[2] 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.
[3] 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,2020,37(1):1-10.
[4] MARZAT J,PIET-LAHANIER H,DAMONGEOT F,et al.Model-based fault diagnosis for aerospace systems:a survey[J].Proceedings of the Institution of Mechanical Engineers,Part G:Journal of Aerospace Engineering,2012,226(10):1329-1360.
[5] 晏政.航天器推进系统基于定性模型的故障诊断方法研究[D].长沙:国防科学技术大学,2013.
[6] DJEBKO K,PUPPE F,KAYAL H.Model-based fault detection and diagnosis for spacecraft with an application for the SONATE triple cube nano-satellite[EB/OL].https://www.researchgate.net/publication/336045357_Model-Based_Fault_Detection_and_Diagnosis_for_Spacecraft_with_an_Application_for_the_SONATE_Triple_Cube_Nano-Satellite,2019.
[7] CHA J,KO S,PARK S Y,et al.Fault detection and diagnosis algorithms for transient state of an open-cycle liquid rocket engine using nonlinear Kalman filter methods[J].Acta Astronautica,2019,163:147-156.
[8] GUEDDI I,NASRI O,BENOTHMAN K,et al.VPCA-based fault diagnosis of spacecraft reaction wheels[C]//2015 XXV International Conference on Information,Communication and Automation Technologies(ICAT).New York:IEEE,2015.
[9] 刘英元,陈海峰,耿直,等.液体火箭发动机振动故障特征信号提取方法[J].火箭推进,2019,45(1):77-82.
[10] 张翔,徐洪平,安雪岩,等.基于马氏距离的液体火箭发动机稳态过程故障程度评估方法[J].计算机测量与控制,2015,23(8):2745-2748.
[11] 张翔,徐洪平,安雪岩,等.基于聚类分析的液体火箭发动机稳态过程故障程度评估方法[J].导弹与航天运载技术,2015(4):24-26.
[12] 聂侥.基于过程神经网络的液体火箭发动机故障预测方法研究[D].长沙:国防科学技术大学,2017.
[13] 孙成志,闫晓东.基于神经网络和证据理论的火箭发动机故障诊断[J].宇航总体技术,2020,4(4):20-30.
[14] 彭军,郭晨阳,张勇,等.基于深度学习的航空发动机部件故障诊断[J].系统仿真技术,2018,14(1):20-24.
[15] AN J,CHO S.Variational autoencoder based anomaly detection using reconstruction probability[EB/OL].https://www.semanticscholar.org/paper/Variational-Autoencoder-based-Anomaly-Detection-An-Cho/061146b1
[16] GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets[C]//28th Neural Information Processing Systems(NIPS).Montreal,Canada:[s.n.],2014.
[17] SABOKROU M,KHALOOEI M,FATHY M,et al.Adversarially learned one-class classifier for novelty detection[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York:IEEE,2018.
[18] SCHLEGL T,SEEBÖCK P,WALDSTEIN S M,et al.F-AnoGAN:fast unsupervised anomaly detection with generative adversarial networks[J].Medical Image Analysis,2019,54:30-44.
[19] 张克明,蔡远文,任元.基于生成对抗网络的航天异常事件检测方法[J].北京航空航天大学学报,2019,45(7):1329-1336.
[20] JOLICOEUR-MARTINEAU A.The relativistic discriminator:a key element missing from standard GAN[EB/OL].https://www.semanticscholar.org/paper/The-relativistic-discriminator3A-a-key-element-from-Jolicoeur-Martineau/dd2ebc42a1a4491b4179dec0ca8686d5c66f6bfe,2018.
[21] HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[22] SHEN X,AGRAWAL S.Kernel density estimation for an anomaly based intrusion detection system[EB/OL].https://www.researchgate.net/publication/221188648_Kernel_Density_Estimation_for_An_Anomaly_Based_Intrusion_Detection_System,2006.

备注

引言

1 自编码式—生成对抗网络

2 基于自编码式—生成对抗网络的LRE异常检测

3 实验验证

4 结语

期刊信息