华南理工大学学报(自然科学版) ›› 2025, Vol. 53 ›› Issue (4): 90-101.doi: 10.12141/j.issn.1000-565X.230781

• 机械工程 • 上一篇    下一篇

基于SSA-BP神经网络的无人机发射参数择优

贾华宇1,2, 郑会龙1,2, 周洪2, 张谦2   

  1. 1.中国科学院大学 航空宇航学院 北京 100049
    2.中国科学院 工程热物理研究所,北京 100190
  • 收稿日期:2023-12-21 出版日期:2025-04-25 发布日期:2024-07-22
  • 通信作者: 郑会龙(1975—),男,研究员,主要从事无人飞行器系统设计、智能协同、智能传感等研究。 E-mail:zhenghuilong@iet.cnedu.cn
  • 作者简介:贾华宇(1991—),男,博士,工程师,主要从事巡飞无人机总体设计、气动优化、发射分系统设计优化等研究。E-mail: jiahuayu@iet.cn
  • 基金资助:
    中国科学院重点部署项目(KGFZD-145-22-02);轻型涡轮动力全国重点实验室基金项目(E31H890206)

Optimal Selection of UAV Launch Parameters Based on SSA-BP Neural Network

JIA Huayu1,2, ZHENG Huilong1,2, ZHOU Hong2, ZHANG Qian2   

  1. 1.School of Aeronautics and Astronautics,University of Chinese Academy of Sciences,Beijing 100049,China
    2.Institute of Engineering Thermophysics,Chinese Academy of Sciences,Beijing 100190,China
  • Received:2023-12-21 Online:2025-04-25 Published:2024-07-22
  • Contact: 郑会龙(1975—),男,研究员,主要从事无人飞行器系统设计、智能协同、智能传感等研究。 E-mail:zhenghuilong@iet.cnedu.cn
  • About author:贾华宇(1991—),男,博士,工程师,主要从事巡飞无人机总体设计、气动优化、发射分系统设计优化等研究。E-mail: jiahuayu@iet.cn
  • Supported by:
    the Key Deployment Project of the Chinese Academy of Sciences(KGFZD-145-22-02);the National Key Laboratory Fund for Light Turbine Power(E31H890206)

摘要:

火箭助推零长发射是无人机发射的重要形式,发射角度、助推器夹角、助推器推力等发射参数的选取直接关系到无人机发射任务的成败。无人机火箭助推零长发射在设计阶段借助工程经验选取发射角度、助推器夹角、助推器推力等关键参数时,存在发射参数迭代择优周期长、设计交互性差、容易造成无人机飞行姿态失稳的问题。该文以某无人机为研究对象,对其发射阶段进行动力学及运动学建模,构建了六自由度非线性模型,基于QT/C++软件编制无人机发射弹道参数化仿真软件,并结合某无人机真实发射试验数据,验证该发射弹道仿真软件的有效性。同时,为解决发射参数自主择优问题,在反向传播(BP)神经网络参数预测模型的基础上引入麻雀搜索算法(SSA)、粒子群优化算法(PSO)、遗传算法(GA)优化模块,提出基于SSA优化BP神经网络的无人机发射参数寻优方法,消除BP神经网络在参数预测过程中存在的过拟合及局部最优效应,对参数预测结果求绝对误差(MAE)、平均百分百误差(MAPE)、均方根误差(RMSE),综合评估SSA-BP对发射参数预测的优越性,并通过发射弹道校核验证发射参数选取的合理性。结果表明,SSA-BP模型对发射参数的预测精度最高、鲁棒性最好,可为无人机发射分系统工程设计阶段的发射参数自主择优选取提供设计依据。

关键词: 无人机发射, 麻雀搜索算法, BP神经网络, 参数寻优, 建模仿真

Abstract:

Rocket assisted zero length launch is an important form of UAV launch. The selection of launch parameters such as launch angle, booster angle, and booster thrust directly affects the success or failure of UAV launch mission. In the design phase of unmanned aerial vehicle rocket assisted zero length launch, key parameters such as launch angle, booster angle, and booster thrust are selected based on engineering experience. However, this approach faces several challenges, including long iteration cycles for optimal parameter selection, poor design interactivity, and the potential risk of causing instability in the UAV’s flight posture. This study focused on a specific UAV and conducts dynamic and kinematic modeling of its launch phase, constructing a six-degree-of-freedom nonlinear model. A UAV launch trajectory parameterization simulation software was developed using QT/C++ software, and the software’s effectiveness was verified through comparison with real UAV launch test data. At the same time, in order to solve the problem of autonomous optimization of launch parameters, sparrow search algorithm (SSA), particle swarm optimization algorithm (PSO), and genetic algorithm (GA) optimization modules were introduced based on the backpropagation neural network (BP neural network) parameter prediction model. A UAV launch parameter optimization method based on SSA optimized BP neural network was proposed to eliminate the overfitting and local optimal effects of BP neural network in the parameter prediction process. The absolute error (MAE), average percentage error (MAPE), and root mean square error (RMSE) of the parameter prediction results were calculated to comprehensively evaluate the superiority of SSA-BP in predicting launch parameters, and the rationality of the launch parameter selection was verified through launch trajectory verification. The results indicate that the SSA-BP model has the highest prediction accuracy and robustness for launch parameters, and can provide a design basis for the autonomous selection of launch parameters in the engineering design stage of unmanned aerial vehicle launch subsystems.

Key words: UAV launch, sparrow search algorithm, BP neural network, parameter optimization, modeling and simulation

中图分类号: