复合材料科学与工程 ›› 2020, Vol. 0 ›› Issue (7): 68-73.

• 应用研究 • 上一篇    下一篇

基于深度学习的复合材料铺层优化方法

白国栋1, 童小燕2, 姚磊江2*   

  1. 1.西北工业大学航空学院,西安710072;
    2.西北工业大学无人机特种技术重点实验室,西安710072
  • 收稿日期:2019-10-18 出版日期:2020-07-28 发布日期:2020-07-28
  • 通讯作者: 姚磊江(1973-),男,博士,教授,主要从事陶瓷基复合材料方面的研究,yaolj@nwpu.edu.cn。
  • 作者简介:白国栋(1995-),男,硕士研究生,主要从事复合材料损伤方面的研究。
  • 基金资助:
    国家重点研发计划(2016YFB0700503);国家自然科学基金项目(51772244,11072195)

LAYUP OPTIMIZATION METHOD OF COMPOSITE WING BASED ON DEEP LEARNING

BAI Guo-dong1, TONG Xiao-yan2, YAO Lei-jiang2*   

  1. 1. School of Aeronautics, Northwestern Polytechnical University, Xi′an 710072, China;
    2. Science and Technology on UVA Laboratory, Northwestern Polytechnical University, Xi′an 710072, Chin
  • Received:2019-10-18 Online:2020-07-28 Published:2020-07-28

摘要: 复合材料在航空工程中有着广泛的应用,其中铺层优化对提高结构承载能力至关重要。本文中首先运用基于神经网络的深度学习方法快速预测设计结果,然后依据预测结果,通过遗传算法搜索复合材料层合板设计问题的最优解,提高铺层设计效率,最后对比应用深度学习技术与传统优化算法得到的优化结果,证明了深度学习技术在层合板铺层优化中的便捷性和有效性。

关键词: 深度学习, 复合材料, 遗传算法, 多任务学习, 铺层优化

Abstract: Composite wings have been widely used in engineering, the optimization of the ply is essential to improve the carrying capacity of the structure. In this paper, a neural network is trained by samples, and then the neural network mode is used to quickly predict the design result. A genetic algorithm evolution operation is used according to the prediction result to search for the optimal solution of the composite laminate design problem and accelerate the design efficiency of the laminate. Applying deep learning technology to quickly solve the optimal layup order of laminates, and comparing with the optimization results obtained by traditional optimization algorithms, it proves a convenience and effectiveness of deep learning technology in the optimization of laminates. The trained neural network can quickly predict the layering sequence optimization index and provide guidance for the optimization process, which has certain guiding significance for further development of the potential and engineering design of the laminate.

Key words: deep learning, genetic algorithm, multi-task learning, layup optimization, skin design

中图分类号: