Scholarlink

<p>The study uses machine learning (ML) models to estimate the mechanical properties of Linz–Donawitz slag (LDS)-filled epoxy composite with 0-30 wt.% filler contents. Based on experimental results, the tensile strength decreased from 48.0 to 36.4&#xa0;MPa (30 wt.%), the tensile modulus decreased from 4.50 to 3.50 GPa, and the flexural strength decreased from 22.0 to 10.0&#xa0;MPa, while the impact strength increased from 17.0 to 26.0&#xa0;kJ/m<sup>2</sup>. These changes are attributed to stress concentration and interfacial debonding induced by the fillers, and improvements in energy dissipation mechanisms, respectively. The experimental data were used to train and evaluate Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GB) models using leave-one-out cross-validation (LOOCV) due to the limited dataset size. Among the three models, Gradient Boosting achieved the highest predictive accuracy, with <i>R</i><sup>2</sup> values of 0.9996 (tensile strength), 1.0000 (tensile modulus), 1.0000 (flexural strength), and 1.0000 (impact strength), along with near-zero MAE and RMSE values. Random Forest also demonstrated strong performance (<i>R</i><sup>2</sup> = 0.9384–0.9924), while SVR showed comparatively lower accuracy, particularly for tensile strength (<i>R</i><sup>2</sup> = 0.6298) and flexural strength (<i>R</i><sup>2</sup> = 0.7607). The findings validate the existence of nonlinear relationships between filler content and mechanical properties, which are well captured by ML models, particularly Gradient Boosting, and suggest that these models have the potential to decrease the amount of experimentation for similar material systems, subject to external validation on independent datasets.</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

Machine Learning-Based Prediction of Mechanical Properties of Epoxy Composites Reinforced with Linz–Donawitz Slag

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Pravat Ranjan Pati,
  • S. Sathees Kumar,
  • Chinmaya P. Mohanty,
  • Abhilash Purohit

摘要

The study uses machine learning (ML) models to estimate the mechanical properties of Linz–Donawitz slag (LDS)-filled epoxy composite with 0-30 wt.% filler contents. Based on experimental results, the tensile strength decreased from 48.0 to 36.4 MPa (30 wt.%), the tensile modulus decreased from 4.50 to 3.50 GPa, and the flexural strength decreased from 22.0 to 10.0 MPa, while the impact strength increased from 17.0 to 26.0 kJ/m2. These changes are attributed to stress concentration and interfacial debonding induced by the fillers, and improvements in energy dissipation mechanisms, respectively. The experimental data were used to train and evaluate Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GB) models using leave-one-out cross-validation (LOOCV) due to the limited dataset size. Among the three models, Gradient Boosting achieved the highest predictive accuracy, with R2 values of 0.9996 (tensile strength), 1.0000 (tensile modulus), 1.0000 (flexural strength), and 1.0000 (impact strength), along with near-zero MAE and RMSE values. Random Forest also demonstrated strong performance (R2 = 0.9384–0.9924), while SVR showed comparatively lower accuracy, particularly for tensile strength (R2 = 0.6298) and flexural strength (R2 = 0.7607). The findings validate the existence of nonlinear relationships between filler content and mechanical properties, which are well captured by ML models, particularly Gradient Boosting, and suggest that these models have the potential to decrease the amount of experimentation for similar material systems, subject to external validation on independent datasets.