应用深度学习的卫星典型单机故障预警平台设计  

Design of Fault Warning Platform for Typical Satellite Equipment Using Deep Learning

作　　者：刘鹏[1] 何鹏 张昊鹏 王志会[1] 张芸香[1] 季业[3] LIU Peng;HE Peng;ZHANG Haopeng;WANG Zhihui;ZHANG Yunxiang;JI Ye(Beijing Institute of Spacecraft System Engineering,Beijing 100094,China;Beijing Shenzhou Aerospace Software Technology Co.,Ltd.,Beijing 100094,China;Beijing Institute of Control Engineering,Beijing 100094,China)

机构地区：[1]北京空间飞行器总体设计部,北京100094 [2]北京神舟航天软件技术股份有限公司,北京100094 [3]北京控制工程研究所,北京100094

出　　处：《航天器工程》2024年第5期131-138,共8页Spacecraft Engineering

摘　　要：为了及早发现卫星典型单机的趋势性异常,基于Docker容器和微服务架构设计了卫星典型单机故障预警平台,包括数据预处理、样本数据增广、模型训练及验证、模型软件自动构建等系统,具备良好的可扩展性。在此基础上,重点研究并给出了应用深度学习的典型单机预警模型构建与软件自动化封装运行方法。利用针对北斗卫星2个典型单机构建的实例模型和15份故障样本,验证了故障预警平台设计的有效性。文章的研究成果可为最终实现卫星系统的智能运维提供参考。In order to early detect the trend anomalies of typical satellite equipment,a fault warning platform for typical satellite equipment is designed based on the Docker container and microservice architecture,including systems such as data preprocessing,sample data augmentation,model training and validation,and automatic construction of model software,etc.,with good scalability.On this basis,the typical fault warning model construction method using deep learning and automatic encapsulation method of software are mainly studied and presented.Verification is conducted using the example models constructed for two pieces of typical equipment of Beidou satellite and 15 fault samples,which proves the effectiveness of the fault warning platform design.The research results of this paper can provide a reference for the final realization of intelligent operation and maintenance of the satellite system.

关 键 词：卫星典型单机 故障预警 深度学习 微服务 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]