基于深度不确定性估计网络的低轨卫星互联网故障预测方法

doi:10.11959/j.issn.2096-8930.2022025

摘要/Abstract

摘要：

低轨卫星互联网技术具有创新性强、已有案例少且卫星数量众多、拓扑变化频繁、载荷类型繁杂等特点，其故障类别难以靠专家系统或者工程经验等方法遍历和训练。同时传统的基于单颗卫星、单一功能的故障检测和诊断方法，对于复杂多变条件下的不确定性故障难以进行预测。针对以上问题，提出一种基于深度不确定性估计网络的低轨卫星互联网星座故障预测算法模型，并基于地面星地联试获得的数据对所提方法进行验证。实验结果表明，提出的方法可以提高已知故障的检测准确率，尤其是对未知故障具有更准确的预测能力。

关键词: 低轨卫星互联网, 低轨卫星星座, 故障预测, 深度神经网络, 不确定性估计

Abstract:

Low-orbit satellite internet technology is a highly innovative direction of development with few existing cases.Low-orbit satellite internet systems are characterised by a large number of constituent satellites, frequent topology changes, and complex payload types.The fault categories of such complex systems are diffi cult to be completely mastered with methods such as expert systems or practical engineering experience.Moreover, traditional fault detection and diagnosis methods tailored for a single satellite and a single functionality are diffi cult to predict uncertain faults under sophisticated and transient environment conditions.For the above problems, a fault prediction model based on the deep uncertainty estimation network for low-orbit satellite internet was proposed, and was verifi ed used self-collected data from simulated joint satellite-ground tests.Experimental results showed that the proposed method could improved the detection accuracy of known types of faults and demonstrated particular eff ectiveness in predicting faults of unknown types.

Key words: low-orbit satellite internet, low-orbit satellite constellation, fault detection, deep neural network, uncertainty estimation

中图分类号:

TP393

孙文宇, 张伟嘉, 王立民. 基于深度不确定性估计网络的低轨卫星互联网故障预测方法[J]. 天地一体化信息网络, 2022, 3(2): 89-97.

Wenyu SUN, Weijia ZHANG, Limin WANG. Fault Detection Method of Low-Orbit Satellite Internet Based on Deep Uncertainty Estimation Network[J]. Space-Integrated-Ground Information Networks, 2022, 3(2): 89-97.

图/表 10

图1

图2

图3

图4

图5

表1

表2

表3

图6

表4

参考文献 22

[1]	吴巍 . 天地一体化信息网络发展综述[J]. 天地一体化信息网络, 2020,1(1): 1-16.
	WU W . Survey on the development of space-integrated-ground information network[J]. Space-Integrated-Ground Information Networks, 2020,1(1): 1-16.
[2]	肖永伟, 孙晨华, 赵伟松 . 低轨通信星座发展的思考[J]. 国际太空, 2018(11): 24-32.
	XIAO Y W , SUN C H , ZHAO W S . Discussion on the problem of LEO communication constellation system design[J]. Space International, 2018(11): 24-32.
[3]	何辞, 张亚生, 孙晨华 ,等. 低轨星座组网及地面IP路由技术适应性分析[J]. 天地一体化信息网络, 2020,1(1): 36-41.
	HE C , ZHANG Y S , SUN C H ,et al. Analysis on low-earth-orbit constellation networking and adaptability of ground IP routing technology[J]. Space-Integrated-Ground Information Networks, 2020,1(1): 36-41.
[4]	孙晨华, 章劲松, 赵伟松 ,等. 高低轨宽带卫星通信系统特点对比分析[J]. 无线电通信技术, 2020,46(5): 505-510.
	SUN C H , ZHANG J S , ZHAO W S ,et al. Comparative analysis of GEO and LEO broadband satellite communication system[J]. Radio Communications Technology, 2020,46(5): 505-510.
[5]	TAKAKI R , HASHIMOTO M , HONDA H ,et al. ISACSDOC:automatic monitoring and diagnostic system for scientific satellite(space development and AI)[J]. Journal of Japanese Society for Artificial Intelligence, 2006,21.
[6]	ZHANG K , JIANG B , SHI P . Adaptive observer-based fault diagnosis with application to satellite attitude control systems[J]. Second International Conference on Innovative Computing,Information and Control (ICICIC 2007), 2008,4(8): 1921-1929.
[7]	BALDI P , BLANKE M , CASTALDI P ,et al. Combined geometric and neural network approach to generic fault diagnosis in satellite actuators and sensors[J]. IFAC-PapersOnLine, 2016,49(17): 432-437.
[8]	陈辛, 魏炳翌, 闻新 . 基于支持向量机的卫星执行机构故障诊断研究[J]. 中国空间科学技术, 2018,38(2): 47-55.
	CHEN X , WEI B Y , WEN X . Study of support vector machine based faults diagnosis for satellite's actuators[J]. Chinese Space Science and Technology, 2018,38(2): 47-55.
[9]	TALEBI H A , KHORASANI K , TAFAZOLI S . A recurrent neuralnetwork-based sensor and actuator fault detection and isolation for nonlinear systems with application to the satellite's attitude control subsystem[J]. IEEE Transactions on Neural Networks, 2009,20(1): 45-60.
[10]	黄瑾, 刘洋, 钟麦英 ,等. 利用随机森林算法的卫星控制系统故障诊断[J]. 宇航学报, 2021,42(4): 513-521.
	HUANG J , LIU Y , ZHONG M Y ,et al. Fault diagnosis of satellite attitude control systems using random forest algorithm[J]. Journal of Astronautics, 2021,42(4): 513-521.
[11]	钱昭勇, 曹裕华, 张雷 . 基于CNN-LSTM的在轨卫星故障预测分析[J]. 军事运筹与系统工程, 2021,35(1): 57-63,72.
	QIAN Z Y , CAO Y H , ZHANG L . Fault prediction and analysis of the on-orbit satellite based on CNN-LSTM[J]. Military Operations Research and Systems Engineering, 2021,35(1): 57-63,72.
[12]	董静怡, 庞景月, 彭宇 ,等. 集成LSTM的航天器遥测数据异常检测方法[J]. 仪器仪表学报, 2019,40(7): 22-29.
	DONG J Y , PANG J Y , PENG Y ,et al. Spacecraft telemetry data anomaly detection method based on ensemble LSTM[J]. Chinese Journal of Scientific Instrument, 2019,40(7): 22-29.
[13]	ZHAO M H , ZHONG S S , FU X Y ,et al. Deep residual shrinkage networks for fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2020,16(7): 4681-4690.
[14]	DEVRIES T , TAYLOR G W . Learning confidence for out-ofdistribution detection in neural networks[J]. arXiv preprint arXiv:1802.04865, 2018.
[15]	KIUREGHIAN A D , DITLEVSEN O . Aleatory or epistemic? Does it matter?[J]. Structural Safety, 2009,31(2): 105-112.
[16]	CHEN K Y , CHEN L S , CHEN M C ,et al. Using SVM based method for equipment fault detection in a thermal power plant[J]. Computers in Industry, 2011,62(1): 42-50.
[17]	HE K M , ZHANG X Y , REN S Q ,et al. Deep residual learning for image recognition[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[18]	SIMONYAN K , ZISSERMAN A . Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.
[19]	SZEGEDY C , IOFFE S , VANHOUCKE V ,et al. Inception-v4,inception-resnet and the impact of residual connections on learning[C]// Thirty-first AAAI conference on artificial intelligence.[S.l.:s.n.], 2017.
[20]	ZHANG X Y , ZHOU X Y , LIN M X ,et al. ShuffleNet:an extremely efficient convolutional neural network for mobile devices[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 6848-6856.
[21]	KINGMA D P , WELLING M . Auto-encoding variational bayes[J]. arXiv preprint arXiv:1312.6114, 2013.
[22]	HE K M , ZHANG X Y , REN S Q ,et al. Delving deep into rectifiers:surpassing human-level performance on ImageNet classification[C]// Proceedings of 2015 IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2015: 1026-1034.

遥测数据状态	数据量/条	是否参与训练
正常状态	12 124	是
故障状态1	5 312	是
故障状态2	5 769	是
故障状态3	4 793	是
故障状态4	4 902	否

算法模型	正常状态	故障状态1	故障状态2	故障状态3
SVM	99.13%	88.64%	90.25%	96.42%
随机森林	99.17%	96.78%	98.36%	98.29%
ResNet18+Softmax	99.47%	98.89%	99.04%	99.15%
ResNet18+U_KL （本文方法）	99.63%	99.26%	99.40%	99.51%
ResNet18+U_JS （本文方法）	99.70%	99.58%	99.67%	99.27%

算法模型	正常状态	故障状态1～3	故障状态4
SVM	99.16%	91.76%	56.57%
随机森林	99.21%	97.85%	60.72%
ResNet18+Softmax	99.32%	99.26%	59.22%
ResNet18+U_KL（本文方法）	99.28%	99.04%	98.14%
ResNet18+U_JS（本文方法）	99.10%	99.21%	98.76%

约束权重	F1-score
α₂/α₁=0	99.08%
α₂/α₁=0.001	99.19%
α₂/α₁=0.01	99.58%
α₂/α₁=0.1	97.34%
α₂/α₁=1	82.55%