测控技术2025，Vol.44Issue(7)：35-41,7.DOI:10.19708/j.ckjs.2025.04.220
基于Attention-LSTM的卫星姿态控制系统故障诊断方法
An Attention-LSTM-Based Fault Diagnosis Method for Satellite Attitude Control System
摘要
Abstract
To address the issue of low accuracy in diagnosing slow-changing faults in existing methods,a fault diagnosis method for satellite attitude control systems that combines attention mechanisms with long short-term memory(LSTM)networks is proposed.A variable attention mechanism is used to reveal hidden feature informa-tion between high-dimensional variable domains,and the deep temporal feature information is extracted by u-sing powerful temporal information capture capability of LSTM networks.By thoroughly mining key feature in-formation from system data in both the variable and temporal domains,diagnostic accuracy is enhanced.Addi-tionally,to solve the problem of forgetting caused by overly long input sequences and excessive deep temporal features,the attention mechanism and LSTM networks are further combined to increase the weight of important temporal information,which addresses the shortcomings of previous methods that relied solely on attention mechanisms to extract temporal features of slow-changing faults.Finally,the proposed fault diagnosis method is validated and compared using a satellite semi-physical simulation platform and data.The results indicate that the method effectively improves the accuracy of diagnosing slow-changing faults and is suitable for fault diagno-sis in satellite attitude control systems.关键词
故障诊断/长短期记忆网络/卫星姿态控制系统/注意力机制Key words
fault diagnosis/LSTM networks/satellite attitude control system/attention mechanism分类
信息技术与安全科学引用本文复制引用
朱津津,张超,高升,艾晨光..基于Attention-LSTM的卫星姿态控制系统故障诊断方法[J].测控技术,2025,44(7):35-41,7.基金项目
国家重点研发计划(2022YFE0204600) （2022YFE0204600）
辽宁省自然科学基金(2024-MSBA-80) （2024-MSBA-80）
中国科学院沈阳自动化研究所基础研究计划项目(2022JC3K03) （2022JC3K03）
