首页|A few-shot deep learning framework for predicting high-velocity impact response of ultra-high molecular weight polyethylene fiber-reinforced composites

A few-shot deep learning framework for predicting high-velocity impact response of ultra-high molecular weight polyethylene fiber-reinforced composites

扫码查看
原文链接
  • NETL
  • NSTL
  • Elsevier
Ultra-high molecular weight polyethylene (UHMWPE) fiber-reinforced composites have been extensively employed in aerospace situations requiring outstanding impact resistance. In real scenarios, high-velocity im-pactors may strike the UHMWPE laminates from diverse angles, possibly resulting in unpredictable protection performance changes. However, when considering this issue, existing experimental and simulation methods are resource-intensive, while theoretical approaches lack broad applicability. To this end, this study establishes a deep learning framework, trained on a small dataset, to predict not only the ballistic limit but also the penetration process of the UHMWPE composite plates subjected to oblique impact by flat-nosed projectiles. The effectiveness of the deep learning model is fully validated by both experimental observations and numerical simulations. It is demonstrated that the model consisting of CNN, BiLSTM, and Attention mechanism captures the ballistic limit velocities, contact force histories, and residual velocity evolutions with a reasonable correlation (i. e., the average R~2 ≥ 0.97) using only 30 samples, addressing the challenge of limited experimental data in ballistic research. Additionally, the prediction time of the current model is significantly reduced to 20.7 s, compared to the hours required for conventional experimental or computational efforts. Based on the model, it is indicated that for the UHMWPE composites of any thickness, the ballistic limit initially decreases and then increases with the impact angle; the physical mechanism underlying this non-monotonic trend is attributed to the nonlinear variation in the contact force histories. Overall, by revealing the mechanisms governing angle-dependent ballistic performance, this paper provides a novel data-driven approach and new physical insights for optimizing impact-resistance materials and structures based on the UHMWPE composites.

Deep learningSmall dataBallistic protectionImpact angleUHMWPE composites

Haibo Ji、Yongqian Zhang、Xin Wang、Liutong Qin、Zengshen Yue、Bingyang Li、Zhen Li、Han Meng、Pengfei Wang、Rui Zhang、Tian Jian Lu

展开 >

State Key Laboratory of Mechanics and Control for Aerospace Structures, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, PR China||MIT Key Laboratory of Multifunctional Lightweight Materials and Structures, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, PR China||Advanced Materials and Energy Center, China Academy of Aerospace Science and Innovation, Beijing 100088, PR China

Shanghai Aerospace Control Technology Institute, Shanghai 201109, PR China

Advanced Materials and Energy Center, China Academy of Aerospace Science and Innovation, Beijing 100088, PR China

Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Horn, Kowloon, Hong Kong, 999077, PR China

Advanced Materials and Energy Center, China Academy of Aerospace Science and Innovation, Beijing 100088, PR China||College of Engineering, Peking University, Beijing 100871, PR China

State Key Laboratory of Mechanics and Control for Aerospace Structures, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, PR China||MIT Key Laboratory of Multifunctional Lightweight Materials and Structures, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, PR China

State Key Laboratory of New Textile Materials and Advanced Processing Technologies, Wuhan Textile University, Wuhan 430200, PR China

展开 >

2025

10.1016/j.ast.2025.110152

Aerospace science and technology

Aerospace science and technology

SCI
ISSN：1270-9638
年,卷(期)：2025.163(Aug.)
  • 44