基于模态法与LSTM误差补偿的航天器动态变形重构OA
Dynamic Deformation Reconstruction of Spacecraft Based on Modal Method and LSTM Error Compensation Network
针对大型航天器在轨运行期间因热载荷、机动控制等因素引发的结构动态变形问题,提出一种基于模态法与长短期记忆网络(LSTM)误差补偿的航天器动态变形重构方法.该方法首先利用模态法,根据大型航天器上布局优化后关键节点的应变测量信息,对航天器的变形进行初步重构;在此基础上,通过构建 LSTM网络,挖掘应变测量数据中的信息,对模态法变形重构误差进行学习和补偿,从而提高重构精度.仿真结果表明,相较于传统模态法,该方法有效提高了航天器动态变形的重构精度,为航天器实现高精度姿态和形变控制奠定了重要基础.
To address the structural dynamic deformation of large spacecraft caused by thermal loads,maneuvering control,and other factors during on-orbit operations,this study proposes a dynamic deformation reconstruction method based on the modal method and Long Short-Term Memory(LSTM)error compensation network.The method first employs the modal method to achieve preliminary deformation reconstruction using strain measurement data from optimally distributed key nodes on the spacecraft.On this basis,an LSTM network is constructed to mine underlying information from the strain measurement data,learning and compensating for errors in the modal-based reconstruction,thereby enhancing reconstruction accuracy.Simulation results demonstrate that,compared to the traditional modal method,the proposed approach effectively improves the reconstruction accuracy of spacecraft dynamic deformation,laying an important foundation for high-precision control of spacecraft.
宁之成;何英姿;刘潇翔;冯骁;关新
北京控制工程研究所,北京 100094北京控制工程研究所,北京 100094北京控制工程研究所,北京 100094||空间智能控制技术全国重点实验室,北京 100094北京控制工程研究所,北京 100094||空间智能控制技术全国重点实验室,北京 100094北京控制工程研究所,北京 100094||空间智能控制技术全国重点实验室,北京 100094
航空航天
模态法LSTM网络动态变形重构结构应变
modal methodlong short-term memory networkdynamic deformation reconstructionstructural strain
《空间控制技术与应用(中英文)》 2026 (2)
12-22,11
国家自然科学基金资助项目(U23B6002) National Natural Science Foundation of China(U23B6002)
评论