Scholarlink

<p>The adhesion properties of steel fibers to concrete are significant parameters in concrete reinforcement. There have been many explorations on the adhesion mechanism of steel fibers with various geometries and forms. Still, little research has been conducted on predicting the pull-out behavior of hook-shaped steel fibers by artificial neural networks (ANN). This study utilizes ANN in a feed-forward multilayer perceptron with backpropagation training to project the pull-out response of hook-shaped steel fibers. Finite element method (FEM) simulation data were used to train the ANN model. FEM simulations, using ABAQUS software and an interfacial transition zone (ITZ) model, were conducted to simulate fiber-concrete interaction. The ANN model possessed a regression coefficient (R) value of 0.944 and a mean squared error (MSE) of 0.00238. To provide additional justification for the scheme, its predictions were compared to experimental results from literature based on fiber diameter and hook angle as controlling factors. The ANN revealed an average percentage deviation of 12%, justifying the precision in the prediction of the pull-out behavior of hook-shaped fibers. Furthermore, hook-shaped fibers enhanced critical separation and pull-out work values with good energy-absorbing properties. The new method based on FEM and ANN for predicting fiber-concrete interaction introduces a reliable means of optimizing fiber geometries in fiber-reinforced concrete applications.</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

Predicting the adhesion parameters between steel fibers and concrete using neural networks

  • 期刊论文

  • 发表日期：

    2025年7月2日

查看全文
  • Shaoli Li,
  • Lirong Liu,
  • Minjuan Zhou

摘要

The adhesion properties of steel fibers to concrete are significant parameters in concrete reinforcement. There have been many explorations on the adhesion mechanism of steel fibers with various geometries and forms. Still, little research has been conducted on predicting the pull-out behavior of hook-shaped steel fibers by artificial neural networks (ANN). This study utilizes ANN in a feed-forward multilayer perceptron with backpropagation training to project the pull-out response of hook-shaped steel fibers. Finite element method (FEM) simulation data were used to train the ANN model. FEM simulations, using ABAQUS software and an interfacial transition zone (ITZ) model, were conducted to simulate fiber-concrete interaction. The ANN model possessed a regression coefficient (R) value of 0.944 and a mean squared error (MSE) of 0.00238. To provide additional justification for the scheme, its predictions were compared to experimental results from literature based on fiber diameter and hook angle as controlling factors. The ANN revealed an average percentage deviation of 12%, justifying the precision in the prediction of the pull-out behavior of hook-shaped fibers. Furthermore, hook-shaped fibers enhanced critical separation and pull-out work values with good energy-absorbing properties. The new method based on FEM and ANN for predicting fiber-concrete interaction introduces a reliable means of optimizing fiber geometries in fiber-reinforced concrete applications.