Scholarlink

<p>The utilization of alkali-activated materials (AAM) derived from industrial waste is increasingly prevalent in engineering as a sustainable construction material capable of significantly reducing carbon emissions. However, the diverse composition of AAM complicates the prediction of its properties, necessitating empirical and somewhat arbitrary material parameter designs. To address this challenge, this study proposes a soft computing framework employing machine learning and heuristic algorithms for predicting the compressive strength and fluidity of AAM, as well as for optimizing material parameters. Microchemical compositions (e.g., CaO, SiO<sub>2</sub>, Al<sub>2</sub>O<sub>3</sub>) serve as input parameters for the first-time prediction of AAM engineering properties. Firstly, a Particle Swarm Optimization-based XGBoost model (PSO-XGBoost) and corresponding evaluation metrics were developed to predict compressive strength and fluidity based on 11 input parameters. Feature importance analysis revealed that water-to-binder ratio (W/B) is the most critical parameter governing AAM properties. Secondly, leveraging the data mapping capabilities of PSO-XGBoost, the Simulated Annealing (SA) algorithm was employed for inverse analysis to optimize material parameters. <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="500_2025_10710_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="136" /> </InlineMediaObject> <EquationSource Format="TEX">\({\text{MAPE}}_{multi - objective}\)</EquationSource> </InlineEquation> reduced the prediction error from 18.84% to 0.27%, demonstrating the efficacy and efficiency of the proposed parameter design methodology.</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

Multi-objective prediction of alkali-activated material properties and material parameter design

  • 期刊论文

  • 发表日期：

    2025年7月1日

查看全文
  • Yueji He,
  • Zhijing Zhu,
  • Jiwen Bai,
  • Dukun Zhao,
  • Rentai Liu,
  • Mengjun Chen,
  • Lianzhen Zhang

摘要

The utilization of alkali-activated materials (AAM) derived from industrial waste is increasingly prevalent in engineering as a sustainable construction material capable of significantly reducing carbon emissions. However, the diverse composition of AAM complicates the prediction of its properties, necessitating empirical and somewhat arbitrary material parameter designs. To address this challenge, this study proposes a soft computing framework employing machine learning and heuristic algorithms for predicting the compressive strength and fluidity of AAM, as well as for optimizing material parameters. Microchemical compositions (e.g., CaO, SiO2, Al2O3) serve as input parameters for the first-time prediction of AAM engineering properties. Firstly, a Particle Swarm Optimization-based XGBoost model (PSO-XGBoost) and corresponding evaluation metrics were developed to predict compressive strength and fluidity based on 11 input parameters. Feature importance analysis revealed that water-to-binder ratio (W/B) is the most critical parameter governing AAM properties. Secondly, leveraging the data mapping capabilities of PSO-XGBoost, the Simulated Annealing (SA) algorithm was employed for inverse analysis to optimize material parameters. \({\text{MAPE}}_{multi - objective}\) reduced the prediction error from 18.84% to 0.27%, demonstrating the efficacy and efficiency of the proposed parameter design methodology.