Scholarlink

<p>Hyperelastic heterogeneous materials featuring microstructures display distinctive multiscale mechanical behaviors, rendering them crucial for advanced engineering applications. However, the geometric structural uncertainty inherent in such heterogeneous materials may give rise to unexpected structural damage or failure, which poses a critical challenge to their reliable application in engineering and medical fields. To address this issue, a novel computational framework for quantifying geometric stochastic effects in such materials is proposed. To tackle the computational challenges posed by material heterogeneities, a mesh- and material-independent method is developed to avoid the remeshing process of the geometric model. Subsequently, a backpropagation artificial neural network is trained to approximate finite element results, substantially reducing the costs of stochastic simulations. The proposed method is validated through three benchmark examples, and further verification is carried out with an atherosclerotic artery model. The results highlight the capability of this framework to handle heterogeneous structures, providing a powerful tool for reliability analysis of bio-inspired materials and medical devices.</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

Uncertainty quantification of geometric dimensions for hyperelastic heterogeneous materials based on an artificial neural network

  • 期刊论文

  • 发表日期：

    2026年7月2日

查看全文
  • Chongshuai Wang,
  • Jinglin Liu,
  • Jia Wang,
  • Chenghao Yang,
  • Di Zuo

摘要

Hyperelastic heterogeneous materials featuring microstructures display distinctive multiscale mechanical behaviors, rendering them crucial for advanced engineering applications. However, the geometric structural uncertainty inherent in such heterogeneous materials may give rise to unexpected structural damage or failure, which poses a critical challenge to their reliable application in engineering and medical fields. To address this issue, a novel computational framework for quantifying geometric stochastic effects in such materials is proposed. To tackle the computational challenges posed by material heterogeneities, a mesh- and material-independent method is developed to avoid the remeshing process of the geometric model. Subsequently, a backpropagation artificial neural network is trained to approximate finite element results, substantially reducing the costs of stochastic simulations. The proposed method is validated through three benchmark examples, and further verification is carried out with an atherosclerotic artery model. The results highlight the capability of this framework to handle heterogeneous structures, providing a powerful tool for reliability analysis of bio-inspired materials and medical devices.