宇航学报

宇航学报 ›› 2021, Vol. 42 ›› Issue (1): 61-73.doi: 10.3873/j.issn.1000-1328.2021.01.007

• 飞行器设计与力学 • 上一篇    下一篇

人工智能气动特性预测技术在火箭子级落区控制项目的应用

杜涛，许晨舟，王国辉，宫宇昆，何巍，牟宇，李舟阳，沈丹，程兴，高家一，韩忠华   

  1. 1. 北京宇航系统工程研究所，北京 100076；2. 西北工业大学航空学院翼型、叶栅空气动力学重点实验室，西安 710072
  • 收稿日期:2020-05-13 修回日期:2020-09-25 出版日期:2021-01-15 发布日期:2021-01-25
  • 基金资助:
    国家自然科学基金（11772261）

The Application of Aerodynamic Coefficients Prediction Technique via Artificial Intelligence Method to Rocket First Stage Landing Area Control Project

DU Tao, XU Chen zhou, WANG Guo hui, GONG Yu kun, HE Wei,MOU Yu, LI Zhou yang, SHEN Dan, CHENG Xing, GAO Jia yi, HAN Zhong hua   

  1. 1. Beijing Institute of Astronautical Systems Engineering, Beijing 100076, China; 2. National Key Laboratory of Science and  Technology on Aerodynamic Design and Research, School of Aeronautics, Northwestern Polytechnical University, Xi’an 710072, China
  • Received:2020-05-13 Revised:2020-09-25 Online:2021-01-15 Published:2021-01-25

PDF

561

摘要： 发展了一种基于人工智能算法的气动特性预测技术，在开展部分工况风洞试验基础上，结合少量数值仿真结果，通过机器学习模型预测全部工况气动特性。该方法能够降低研制成本，缩短周期。先后解决了相关函数选择、模型超参数训练、数据检验和“人在回路”应用等关键算法与技术问题，应用于运载火箭子级栅格舵落区控制项目气动研制，获得了设计所需完整的气动特性数据。2019年7月26日火箭飞行搭载试验验证了预测方法的正确性。最后，提出了人工智能技术在气动设计应用的分级概念和标准，划分和识别人工智能的能力，确定阶段性功能，为人工智能与气动设计结合与应用提供参考。

关键词: font-size:10.5pt, ">人工智能；机器学习；气动特性；栅格舵；火箭子级落区控制；技术分级

Abstract: A novel approach of predicting aerodynamic data via artificial intelligence technique is proposed in this article. Based on wind tunnel tests of partial test states, combined with several CFD results, machine learning via Kriging model is used to predict the whole aerodynamic characteristics to shorten the development cycle and reduce the expensive wind tunnel tests as many as possible. After solving several key technical problems such as the selection of correlation functions, hyper parameters training, data verification and application of “man in loop” technique, the complete set of aerodynamic data was obtained successfully and used to the control law design in the rocket first stage landing area control project with grid fins. The correctness of the proposed method was validated by a flight test on 26th July, 2019, which was carried out successfully for the first time in China. At the end, the grading of technology maturity degree for the artificial intelligence technique is presented to evaluate application to aerodynamic engineering design problems.

Key words: Artificial intelligence, Machine learning, Aerodynamic characteristics, Grid fin, Rocket first stage landing area control, Technology classification

中图分类号: