航空学报 > 2023, Vol. 44 Issue (17): 128098-81280986   doi: 10.7527/S1000-6893.2022.28098

基于机器学习的智能控制数值虚拟飞行方法

梁益铭1,2, 李广宁1(), 徐敏1   

  1. 1.西北工业大学 航天学院,西安 710072
    2.西安现代控制技术研究所,西安 710065
  • 收稿日期:2022-10-10 修回日期:2022-11-16 接受日期:2022-12-12 出版日期:2023-09-15 发布日期:2022-12-14
  • 通讯作者: 李广宁 E-mail:lgning@nwpu.edu.cn
  • 基金资助:
    国家自然科学基金(12072278)

Method for numerical virtual flight with intelligent control based on machine learning

Yiming LIANG1,2, Guangning LI1(), Min XU1   

  1. 1.School of Astronautics,Northwestern Polytechnical University,Xi’an 710072,China
    2.Xi’an Modern Control Technologies Research Institute,Xi’an 710065,China
  • Received:2022-10-10 Revised:2022-11-16 Accepted:2022-12-12 Online:2023-09-15 Published:2022-12-14
  • Contact: Guangning LI E-mail:lgning@nwpu.edu.cn
  • Supported by:
    National Natural Science Foundation of China(12072278)

摘要:

提出并实现了一种基于机器学习的PID智能控制策略下的数值虚拟飞行算法,结合Basic Finner导弹标准模型算例,对所提出算法进行了验证和评估分析,表明本文算法可行,并具有良好的应用前景。首先,搭建了一种基于重叠动网格技术的CFD/RBD耦合数值虚拟飞行仿真模型。针对Basic Finner导弹的标准工况进行了无控自由飞行状态的数值飞行模拟,并结合实验结果对所构建的数值虚拟飞行仿真算法进行了验证和评估,表明所采用的数值模拟算法可用于数值虚拟飞行环境下的智能控制参数设计与仿真评估;其次,结合数值虚拟飞行过程对飞行器气动、姿态和位移等参数的实时控制需求,提出了一种基于BP神经网络算法的PID参数在线学习的智能控制器,并针对Basic Finner导弹的俯仰通道,分别对传统PID控制策略和智能PID控制策略下的导弹自由释放后的俯仰角快速稳定控制过程、阶跃式和正弦式俯仰角输入下的导弹跟踪控制过程进行了数值虚拟飞行仿真模拟。研究表明,基于BP神经网络的PID智能控制器能够根据所获得的实时飞行参数,实现控制参数的在线学习和自我优化、调整,相比于传统PID控制器,对于不同输入工况表现出良好的适应性,所关注的控制变量的超调量、上升时间、过渡时间、稳态误差等性能指标均有很大的提高,学习效率越高,则系统响应速度越快,超调量也越大,稳定误差越小。

关键词: 数值虚拟飞行, 机器学习, BP神经网络, PID, 重叠动网格

Abstract:

This paper proposes a method for numerical virtual flight with intelligent control based on machine learning. Combined with the case of the Basic Finner projectile model, the proposed algorithm is verified and evaluated. The results show the feasibility and good application prospect of the proposed algorithm. Firstly, a CFD/RBD coupled numerical virtual flight simulation model based on the overlapping dynamic mesh technology is constructed. According to the case of the Basic Finner projectile, the numerical simulation without control is conducted. Compared with the experimental data, the proposed numerical virtual flight simulation algorithm is verified and evaluated, showing that the numerical simulation algorithm can be used in the design and evaluation of the control parameters in the numerical virtual flight environment. Secondly, numerical simulations of the Basic Finner projectile’s pitch channel are carried out adopting the traditional PID control strategy and the intelligent PID control strategy, respectively. The PID intelligent controller based on the BP neural network can realize online learning and self-optimization of control parameters according to the acquired real-time flight parameters. Compared with the traditional PID controller, the concerned control variable overshoot, rise time, transition time and steady-state error and other performance indicators have been significantly improved, and the higher learning efficiency leads to the faster system, larger overshoot, and smaller stability error.

Key words: numerical virtual flight, machine learning, BP neural network, PID, moving chimera grid

中图分类号: