Scholarlink

<p>This study is an integrated modeling and optimization study of natural resource basalt and flax fiber-reinforced polymer nanocomposites, which would help in improving mechanical behavior of the material by experiment design and computer-based methods for sustainable application. The experiment examines how three input variables fiber orientation, silane treatment, and nanographene can affect two important mechanical output responses, presumably tensile and flexural properties, across a series of different composite formulations. Quadratic models of the mechanical responses were identified using Response Surface Methodology (RSM), and these have been found to be strongly correlated with experimental results R<sup>2</sup> of over 0.95. Four regression algorithms were trained using 5‑fold cross‑validation and evaluated based on R<sup>2</sup>, MAE, and RMSE metrics. The Decision Tree and SVM models yielded the highest test R<sup>2</sup> values of 0.993 and 0.986 for tensile and flexural strength, respectively, whereas Extreme Gradient Boosting (XGBoost), Random Forest (RF) demonstrated stable performance across both properties. To address the black‑box characteristic of these models, Shapley Additive Planations (SHAP) were implemented to interpret feature importance and interdependencies. The analysis revealed that tensile strength and flexural strength was primarily governed by fiber content, and nanoparticles.</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

Application of multiple explainable machine learning models for predicting the mechanical properties of nanoparticle modified basalt and flax fiber composites

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月1日

查看全文
  • S. Jothi Arunachalam,
  • Nagaraj Ashok

摘要

This study is an integrated modeling and optimization study of natural resource basalt and flax fiber-reinforced polymer nanocomposites, which would help in improving mechanical behavior of the material by experiment design and computer-based methods for sustainable application. The experiment examines how three input variables fiber orientation, silane treatment, and nanographene can affect two important mechanical output responses, presumably tensile and flexural properties, across a series of different composite formulations. Quadratic models of the mechanical responses were identified using Response Surface Methodology (RSM), and these have been found to be strongly correlated with experimental results R2 of over 0.95. Four regression algorithms were trained using 5‑fold cross‑validation and evaluated based on R2, MAE, and RMSE metrics. The Decision Tree and SVM models yielded the highest test R2 values of 0.993 and 0.986 for tensile and flexural strength, respectively, whereas Extreme Gradient Boosting (XGBoost), Random Forest (RF) demonstrated stable performance across both properties. To address the black‑box characteristic of these models, Shapley Additive Planations (SHAP) were implemented to interpret feature importance and interdependencies. The analysis revealed that tensile strength and flexural strength was primarily governed by fiber content, and nanoparticles.