Scholarlink

<p>This study evaluates the structural and mechanical properties of 3D printed fiber-reinforced composite (FRC) constructs and optimizes them by integrating reinforcement learning with a developed finite element analysis (FEA) model. The structural features of 3D printed FRC con-structs are identified and the geometry is parametrically modeled based on these features. The bending properties of the FRC constructs are assessed by varying the fiber orientation angles using three-point bending tests. The similarity between the bending simulation results from the FEA model and the experimental data is demonstrated, validating the accuracy of the analysis model. The developed FEA model is then integrated with a reinforcement learning (RL) environment to create an optimization framework, and the learning process is automated through the parametric scripting of model. This RL-based optimization technique successfully identifies the optimal fiber angle parameters that maximize the mechanical properties of the FRC constructs. The optimized results demonstrated a 64.51% increase in stiffness compared to the initial design with random fiber angle parameters. The study highlights the effective use of FEA models in RL-based optimization and demonstrates that combining this technique with commercial FEA tools enables efficient design optimization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Mechanical Characterization of 3D Printed Fiber-Reinforced Composite Structures for Reinforcement Learning-Aided Design

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • Goan Woo Hyun,
  • Ju Chan Yuk,
  • Suk Hee Park

摘要

This study evaluates the structural and mechanical properties of 3D printed fiber-reinforced composite (FRC) constructs and optimizes them by integrating reinforcement learning with a developed finite element analysis (FEA) model. The structural features of 3D printed FRC con-structs are identified and the geometry is parametrically modeled based on these features. The bending properties of the FRC constructs are assessed by varying the fiber orientation angles using three-point bending tests. The similarity between the bending simulation results from the FEA model and the experimental data is demonstrated, validating the accuracy of the analysis model. The developed FEA model is then integrated with a reinforcement learning (RL) environment to create an optimization framework, and the learning process is automated through the parametric scripting of model. This RL-based optimization technique successfully identifies the optimal fiber angle parameters that maximize the mechanical properties of the FRC constructs. The optimized results demonstrated a 64.51% increase in stiffness compared to the initial design with random fiber angle parameters. The study highlights the effective use of FEA models in RL-based optimization and demonstrates that combining this technique with commercial FEA tools enables efficient design optimization.