上海航天（中英文）2025，Vol.42Issue(z1)：62-70,9.DOI:10.19328/j.cnki.2096-8655.2025.S1.008
基于机器学习的运载火箭在线测发故障检测技术
Machine Learning-Based Online Test and Launch Fault Detection Technology for Launch Vehicles
彭逸 1陈晓 1董祥见 1张鲲鹏 1王赞2
作者信息
- 1. 上海航天控制技术研究所,上海 201109||上海市空间智能控制技术重点实验室,上海 201109
- 2. 北京航空航天大学 电子信息工程学院,北京 100191
- 折叠
摘要
Abstract
To address the demands for high-reliability,rapid,and refined interpretation under high-frequency launch operations of launch vehicles,this study enhances comprehensive system state evaluation capabilities by developing an online test and launch fault detection system based on machine learning.Targeting the characteristics of crtical subsystems,specific detection architectures are proposed.For servo system,a generative recurrent network framework is adopted to construct the detection architecture.By online prediction model updating and historical data generation mechanism,this approach resolves challenges of catastrophic forgetting.For inertial and secondary power system,anomaly detection is achieved using One-class SVM and iForest algorithms,respectively.Application demonstrates that in large-scale system tests for launch vehicles,the model achieves 95.5%accuracy with high detection reliability and zero false alarms,significantly improving interpretation coverage and efficiency.The research provides an intelligent solution for launch vehicle health monitoring,effectively supporting enhanced launch mission reliability and intelligent development requirements.关键词
运载火箭/机器学习/神经网络/故障检测/自动判读Key words
launch vehicle/machine learning/neural network/fault detection/automatic interpretation分类
航空航天引用本文复制引用
彭逸,陈晓,董祥见,张鲲鹏,王赞..基于机器学习的运载火箭在线测发故障检测技术[J].上海航天（中英文）,2025,42(z1):62-70,9.
