中国空间科学技术

中国空间科学技术 ›› 2025, Vol. 45 ›› Issue (3): 131-142.doi: 10.16708/j.cnki.1000.758X.2025.0045

• 论文 • 上一篇    

基于GA-BP的大气帆气动函数拟合与控制矩阵计算

林瑞1，刘晓文1，丁纪昕1,2，徐明1,2,*   

  1. 1.北京航空航天大学 宇航学院，北京102206
    2.北京航空航天大学 沈元学院，北京100191
  • 收稿日期:2023-12-28 修回日期:2024-05-09 录用日期:2024-05-13 发布日期:2025-05-15 出版日期:2025-06-01

Aerodynamic function fitting and control matrix computation for atmospheric sail based on GA-BP

LIN Rui1，LIU Xiaowen1，DING Jixin1,2，XU Ming1,2,*   

  1. 1.School of Astronautics，Beihang University，Beijing 102206，China
    2.Shen Yuan Honors College，Beihang University，Beijing 100191，China
  • Received:2023-12-28 Revision received:2024-05-09 Accepted:2024-05-13 Online:2025-05-15 Published:2025-06-01

PDF (PC)

2

摘要： 面向航天器低成本、特殊编队的任务需要，针对利用低轨大气阻力的帆式航天器展开研究。为验证大气帆技术的原理可行性，建立了基于仿风筝飞行器的模型。该模型利用地面牵引帆面维持滞空，模拟低轨运行环境，同时借助飞行器两侧可旋转的分布式副帆产生气动控制力矩，完成姿态位置的控制。通过对不同副帆转角和姿态下的飞行器进行气动仿真得到数据集；再使用遗传算法优化的反向传播神经网络模型算法（GA-BP算法）对仿真数据进行训练，得到气动函数的网络模型，其中各项参数的预测相关系数R2均大于0.98（除滚转力矩为0.91）。再将所得的网络模型应用于飞行器的力学平衡方程，逆向求解得到副帆转角与飞行器相对位置对应的控制矩阵。由控制矩阵及副帆转角极限规范了飞行器的安全运行范围，可为大气帆飞行器的实际控制提供参考。

关键词: 分布式大气帆, 气动力, 控制矩阵, 遗传算法, 神经网络

Abstract: Aiming at the need of low-cost and special formation requirements for space missions, the research focuses on the sailcraft utilizing low orbit aerodynamic drag. The practicability of atmospheric sail technology is ascertained via the model based on towing the imitation kite sailcraft to simulate the low orbit environment. And the aerodynamic torque, generated by the rotation of the distributed sub-sails, is applied to complete the position and attitude control of the sailcraft. The dataset is obtained by aerodynamic simulation of sailcraft with different rotation angles of sub-sails and attitudes of sailcraft, which is subsequently used to train an intelligent model based on Genetic Algorithm-optimized Back Propagation Neural Network (GA-BP). The aerodynamic function network model is derived from the dataset, where the predicted correlation coefficient R2 of each parameter is greater than 0.98 (except for the rolling moment of 0.91). The resulting network model is employed in the mechanical equilibrium equations of the sailcraft, leading to the inverse calculation of the control matrix corresponding to the sub-sail rotation angles and the sailcraft relative positions. The control matrix and the boundaries of the sub-sail angles regulate the safe operation range of the sailcraft, which can provide a reference for the actual control of the atmospheric sailcraft.

Key words: distributed atmospheric sail, aerodynamic drag, control matrix, genetic algorithms, neural networks