基于深度强化学习的三维变形机翼反设计方法

苏敬; 孙刚; 陶俊

doi:10.7638/kqdlxxb-2024.0123

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

基于深度强化学习的三维变形机翼反设计方法

研究专栏——飞行器气动/多学科不确定性量化与稳健优化设计 Ⅱ | 更新时间：2026-05-28

- 基于深度强化学习的三维变形机翼反设计方法
- An inverse design method for three-dimensional morphing wings based on deep reinforcement learning
- 空气动力学学报 2024年42卷第10期页码：84-97
- 作者机构：
  
  复旦大学航空航天系,上海,200433
- 作者简介：
  
  苏敬（1998—），男，安徽宿州人，硕士研究生，研究方向：飞行器设计. E-mail：21210290013@m.fudan.edu.cn
  陶俊*，副教授，研究方向：空气动力学，飞行器设计. E-mail：juntao@fudan.edu.cn
- 基金信息：
  
  飞行器基础布局全国重点实验室开放基金项目（ZYTS-202501）
- DOI：10.7638/kqdlxxb-2024.0123
  中图分类号： V244
- 纸质出版：2024
- 稿件说明：
移动端阅览
苏敬, 孙刚, 陶俊. 基于深度强化学习的三维变形机翼反设计方法[J]. 空气动力学学报, 2024, 42(10): 84−97.

SU Jing, SUN Gang, TAO Jun. An inverse design method for three-dimensional morphing wings based on deep reinforcement learning[J]. 2024, 42(10): 84-97.
苏敬, 孙刚, 陶俊. 基于深度强化学习的三维变形机翼反设计方法[J]. 空气动力学学报, 2024, 42(10): 84−97. DOI： 10.7638/kqdlxxb-2024.0123

SU Jing, SUN Gang, TAO Jun. An inverse design method for three-dimensional morphing wings based on deep reinforcement learning[J]. 2024, 42(10): 84-97. DOI： 10.7638/kqdlxxb-2024.0123.

摘要

三维变形机翼在可变工况下如何自主变形以达到气动性能要求，且满足任务自适应变形的基本功能是一个具有重要意义的问题。本文提出了一种基于强化学习的三维变形机翼反设计（reinforcement learning inverse design

RLID）框架，并将其应用于可变工况的自适应变形飞行任务中。选取类别/形状变换函数设计三维变形机翼，并采用拉丁超立方抽样方法对变形设计空间进行抽样，从而获取样本点；通过计算流体力学求解得到对应的气动参数，并通过深度置信网络代理模型构建从变形设计参数到气动参数的输入-输出模型。针对可变工况环境，基于无监督学习的深度Q网络（deep Q-network

DQN）强化学习智能体可为机翼实时提供变形策略，结果满足预期气动性能要求约70%，平均气动性能达到要求的98%以上。此外，本文将DQN智能体与基于贪心的条件生成对抗网络（greedy-basedconditional generative adversarial network

G-CGAN）智能体进行了对比，结果表明，本文所提出的RLID框架在多变工况条件下能够提供可靠的变形策略，且相较于G-CGAN，DQN智能体更注重整体任务的收益。

Abstract

It is of great significance to find out how to deform a three-dimensional deformed wing independently to meet the requirements of aerodynamic performance and the basic function of mission adaptive deforming under variable operating conditions. In this study

an RLID (reinforcement learning inverse design) framework is proposed and applied to the reverse design of three-dimensional morphing wings for adaptive morphing flight missions under variable operating conditions. The CST(class-shape function transformation) parameterization method is chosen to define three-dimensional morphing wings

and the Latin hypercube sampling method is used to sample in the design space and generate sample points. Computational fluid dynamics simulations are performed to obtain corresponding aerodynamic parameters

and the deep belief network surrogate model is constructed to map the input-output relationship between morphing design parameters and aerodynamic parameters. To address the variable operating conditions

a DQN (deep Q-network) reinforcement learning agent

leveraging unsupervised learning

is used to provide real-time morphing strategies

and the results meet about 70% of the expected aerodynamic performance requirements

and the average aerodynamic performance reaches more than 98% of the requirements. Furthermore

the design results via the DQN agent are compared with those via the G-CGAN (greedy-based conditional generative adversarial network) agents. The results indicate that the proposed RLID framework efficiently obtains a satisfactory strategy of morphing wings under variable operating conditions and that the DQN agent focuses more on overall task rewards than the G-CGAN agent.

关键词

Keywords

references

CHU L L, LI Q, GU F, et al . Design, modeling, and control of morphing aircraft: A review[J ] . Chinese Journal of Aeronautics, 2022 , 35 ( 5 ): 220 − 246 .

BARBARINO S, BILGEN O, AJAJ R M, et al . A review of morphing aircraft[J ] . Journal of Intelligent Material Systems and Structures, 2011 , 22 ( 9 ): 823 − 877 .

REICH G, SANDERS B . Introduction to morphing aircraft research[J ] . Journal of Aircraft, 2007 , 44 ( 4 ): 1059 .

WEISSHAAR T A. Morphing aircraft technology - New shapes for aircraft design[R]. NATO/OTAN, 2006.

BONNEMA K, SMITH S. AFTI/F-111 mission adaptive wing flight research program[C]//4th Flight Test Conference, San Diego, CA. Reston, Virginia: AIAA, 1988: 2118.

PENDLETON E W, BESSETTE D, FIELD P B, et al . Active aeroelastic wing flight research program: technical program and model analytical development[J ] . Journal of Aircraft, 2000 , 37 ( 4 ): 554 − 561 .

KUDVA J N . Overview of the DARPA smart wing project[J ] . Journal of Intelligent Material Systems and Structures, 2004 , 15 ( 4 ): 261 − 267 .

AJAJ R M . Flight dynamics of transport aircraft equipped with flared-hinge folding wingtips[J ] . Journal of Aircraft, 2020 , 58 ( 1 ): 98 − 110 .

KOTA S, FLICK P, COLLIER F S. Flight testing of FlexFloil TM adaptive compliant trailing edge[C ] //54th AIAA Aerospace Sciences Meeting, San Diego, California, USA. Reston, Virginia: AIAA, 2016: 0036.

VALASEK J, TANDALE M D, RONG J . A reinforcement learning - adaptive control architecture for morphing[J ] . Journal of Aerospace Computing, Information, and Communication, 2005 , 2 ( 4 ): 174 − 195 .

GARNIER P, VIQUERAT J, RABAULT J, et al . A review on deep reinforcement learning for fluid mechanics[J ] . Computers Fluids, 2021 , 225 : 104973 .

GONG L G, WANG Q, HU C H, et al . Switching control of morphing aircraft based on Q-learning[J ] . Chinese Journal of Aeronautics, 2020 , 33 ( 2 ): 672 − 687 .

YAN B B, LI Y, DAI P, et al . Adaptive wing morphing strategy and flight control method of a morphing aircraft based on reinforcement learning[J ] . Journal of Northwestern Polytechnical University, 2019 , 37 ( 4 ): 656 − 663 .

LI R, ZHANG Y, CHEN H . Reinforcement learning method for supercritical airfoil aerodynamic design[J ] . Acta Aeronauticaet Astronautica Sinica, 2021 , 42 ( 4 ): 523810 .

LAMPTON A, NIKSCH A, VALASEK J. Reinforcement learning of morphing airfoils with aerodynamic and structural effects[C]//AIAA Infotech@Aerospace 2007 Conference and Exhibit, Rohnert Park, California. Reston, Virginia: AIAA, 2007: 2805.

MNIH V, KAVUKCUOGLU K, SILVER D, et al . Human-level control through deep reinforcement learning[J ] . Nature, 2015 , 518 : 529 − 533 .

XIAO Z X. Numerical simulation of navier-stokes equation for complex flow and application of turbulence model[D]. Xi'an: Northwestern Polytechnical University, 2003(in Chinese).

CHEN Z G, SHEN Q X, CAI R Q, et al . The flux differential splitting upwind differential format solves the three-dimensional Navier-Stokes equation for incompressible flow, and the Wigley mathematical ship shape is calculated around the ship[J ] . Shipping, 1995 ( 04 ): 8 − 16 (in Chinese).

ROE P L . Approximate Riemann solvers, parameter vectors, and difference schemes[J ] . Journal of Computational Physics, 1981 , 43 ( 2 ): 357 − 372 .

BAO W C, WANG W Q, CHEN X, et al . Numerical analyses of aeroacoustic characteristics of tiltrotor considering the aerodynamic interaction by the fuselage in hover[J ] . Aerospace Science and Technology, 2023 , 141 : 108558 .

SPALART P, ALLMARAS S. A one-equation turbulence model for aerodynamic flows[C]//30th Aerospace Sciences Meeting and Exhibit, Reno, NV. Reston, Virginia: AIAA, 1992: 439.

LAFLIN K R, KLAUSMEYER S M, ZICKUHR T, et al . Data summary from second AIAA computational fluid dynamics drag prediction workshop[J ] . Journal of Aircraft, 2005 , 42 ( 5 ): 1165 − 1178 .

LI J, GAO Z H, HUANG J T, et al . Aerodynamic optimization system based on CST technique[J ] . Acta Aerodynamica Sinica, 2012 , 30 ( 4 ): 443 − 449 (in Chinese).

KULFAN B. A universal parametric geometry representation method - “CST” [C]//45th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada. Reston, Virginia: AIAA, 2007: 62.

SHEIKHOLESLAMI R, RAZAVI S . Progressive Latin hypercube sampling: An efficient approach for robust sampling-based analysis of environmental models[J ] . Environmental Modelling Software, 2017 , 93 : 109 − 126 .

PINAYA W H L, GADELHA A, DOYLE O M, et al . Using deep belief network modelling to characterize differences in brain morphometry in schizophrenia[J ] . Scientific Reports, 2016 , 6 : 38897 .

CHEN R X, YANG X, HU X L, et al . Planetary gearbox fault diagnosis method based on deep belief network transfer learning[J ] . Vibration and Shock, 2021 , 40 ( 01 ): 127 − 133 (in Chinese).

HAO J Y, YANG T P, TANG H Y, et al . Exploration in deep reinforcement learning: from single-agent to multiagent domain[J ] . IEEE Transactions on Neural Networks and Learning Systems, 2024 , 35 ( 7 ): 8762 − 8782 .

SU J, SUN G, TAO J . A novel inverse design method for morphing airfoil based on deep reinforcement learning[J ] . Aerospace Science and Technology, 2024 , 145 : 108895 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于条件扩散的变形结构气动热预测方法

强化学习方法在翼型拍动实验中的应用

类HyTRV升力体阻尼/交叉动导数特性分析

风扇叶片颗粒侵蚀特性的多因素分析与模拟

大展弦比飞行器突风响应数值计算