Scholarlink

<p>In this work, a new method of reverse manufacturing technology for aluminum foam is developed based on the assistance of a deep learning algorithm. The generalized regression neural network (GRNN) was used to extract the microstructure features of the aluminum foam samples. The deep convolution generative adversarial networks (DCGANs) are used to generate images by adjusting the noise vector parameter values and the activation function weights. Through the two frameworks mentioned above, the image of aluminum foam with a highly realistic physical structure is generated. The three-dimensional model of the sample is generated from the formed images. Then, the sample of reverse manufacturing was produced by selective laser melting (SLM). The properties of samples obtained from reverse manufacturing and traditional forward manufacturing are tested and compared. After reverse manufacturing, the porosity of open-cell and closed-cell aluminum foams changes tiny, and a high degree of reduction of the sample structure can be achieved. The yield stress values for the three groups of open-cell aluminum foams increased by an average of 16.13&#xa0;MPa, and the plateau stress values increased by an average of 7.55&#xa0;MPa. The yield stress values for the three groups of closed-cell aluminum foams increased by an average of 16.08&#xa0;MPa, and the plateau stress values increased by an average of 8.55&#xa0;MPa. The results show that the reverse manufacturing sample has a higher degree of controllability and superior mechanical properties.</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

Research on Inverse Reconstruction of Porous Aluminum Foam Based on Deep Convolutional Generative Adversarial Network

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Jie Gao,
  • Wenhao Da,
  • Yanli Wang,
  • Lucai Wang,
  • Xiaohong You,
  • Wenzhan Huang,
  • Fang Wang,
  • Yuancong Mi

摘要

In this work, a new method of reverse manufacturing technology for aluminum foam is developed based on the assistance of a deep learning algorithm. The generalized regression neural network (GRNN) was used to extract the microstructure features of the aluminum foam samples. The deep convolution generative adversarial networks (DCGANs) are used to generate images by adjusting the noise vector parameter values and the activation function weights. Through the two frameworks mentioned above, the image of aluminum foam with a highly realistic physical structure is generated. The three-dimensional model of the sample is generated from the formed images. Then, the sample of reverse manufacturing was produced by selective laser melting (SLM). The properties of samples obtained from reverse manufacturing and traditional forward manufacturing are tested and compared. After reverse manufacturing, the porosity of open-cell and closed-cell aluminum foams changes tiny, and a high degree of reduction of the sample structure can be achieved. The yield stress values for the three groups of open-cell aluminum foams increased by an average of 16.13 MPa, and the plateau stress values increased by an average of 7.55 MPa. The yield stress values for the three groups of closed-cell aluminum foams increased by an average of 16.08 MPa, and the plateau stress values increased by an average of 8.55 MPa. The results show that the reverse manufacturing sample has a higher degree of controllability and superior mechanical properties.