首页 >

一种基于机器学习的智能控制数值虚拟飞行方法

梁益铭1,李广宁1,徐敏2   

  1. 1. 西北工业大学
    2. 西北工业大学航天学院
  • 收稿日期:2022-10-10 修回日期:2022-12-13 出版日期:2022-12-14 发布日期:2022-12-14
  • 通讯作者: 李广宁
  • 基金资助:
    国家自然科学基金

A Method Of Numerical Virtual Flight With Intelligent Control Based On Machine Learning

  • Received:2022-10-10 Revised:2022-12-13 Online:2022-12-14 Published:2022-12-14
  • Supported by:
    The authors gratefully acknowledge the support from the National Science Foundation,China

摘要: 本文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,并结合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,本文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,表明本文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,分别进行了传统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、过渡时间以及稳态误差等各个性能指标上均有很大的提高,学习效率越高,则系统的快速性可以得到提升,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, 刚性重叠动网格

Abstract: In this paper, a method of numerical virtual flight with intelligent control based on machine learning is proposed. Com-bined with the case of the Basic Finner projectile model, the proposed algorithm is verified and evaluated. It shows that the pro-posed algorithm is feasible and has a good application prospect. Firstly, a CFD/RBD coupled numerical virtual flight simula-tion model based on overlapping dynamic mesh technology is constructed. According to the case of the Basic Finner projec-tile, the numerical simulation without control is carried out. Compared with the experimental data, the proposed numerical virtual flight simulation algorithm is verified and evaluated. It shows that the numerical simulation algorithm can be used in the design and evaluation of the control parameters in the numerical virtual flight environment. Secondly, The numerical simulations of the Basic Finner projectile's pitch channel are carried out under the traditional PID control strategy and the intelligent PID control strategy respectively. The PID intelligent controller based on 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 over-shoot, rise time, transition time and steady-state error and other performance indicators have been greatly improved, and the higher the learning efficiency, the faster the system, the larger the overshoot, and the smaller the stability error.

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

中图分类号: