Scholarlink

<p>The accurate forecasting of mechanical characteristics of natural fiber-reinforced composites were essential for development like structural materials. Advancement of natural fiber–reinforced hybrid composites were frequently impeded expensive, time-consuming, and resource-demanding processes. Although machine learning (ML) presents robust option for expediting prediction, conventional models frequently operate ‘black boxes,’ so constraining reliability and lacking scientific insights essential for genuine material design. This research addresses a significant gap by formulating and validating a comprehensive framework that integrates experimentation, explainable machine learning (ML), and finite element analysis (FEA). Date palm fruit stalk fiber (DPFSF) reinforced epoxy composites incorporating areca nut husk biofiller were experimentally fabricated and evaluated to elucidate and predict their mechanical characteristics. Three supervised ML models Gradient Boosting, k-Nearest Neighbor and Random Forest was trained and constructed using comprehensive experimental dataset. Random Forest model demonstrated greatest prediction, attaining R<sup>2</sup> values of 0.90 for Tensile Strength (TS) and 0.95 for Flexural strength (FS) during validation. To tackle “black box” issue of machine learning, SHapley Additive exPlanations (SHAP) analysis was utilized, identifying filler percentage and length as primary determinants influencing strength estimates. This study demonstrates that explainable machine learning is a viable and sustainable method for optimizing hybrid composites based on natural fibers.</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

Prediction and interpretation of mechanical properties of date palm fiber reinforced areca nut husk biofiller composites using explainable machine learning

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • J Srinivas,
  • Sunitha Tappari,
  • L Ganesh Babu,
  • K Ram Prasad,
  • Rallabandi Revathi,
  • S. Sathiyavathi

摘要

The accurate forecasting of mechanical characteristics of natural fiber-reinforced composites were essential for development like structural materials. Advancement of natural fiber–reinforced hybrid composites were frequently impeded expensive, time-consuming, and resource-demanding processes. Although machine learning (ML) presents robust option for expediting prediction, conventional models frequently operate ‘black boxes,’ so constraining reliability and lacking scientific insights essential for genuine material design. This research addresses a significant gap by formulating and validating a comprehensive framework that integrates experimentation, explainable machine learning (ML), and finite element analysis (FEA). Date palm fruit stalk fiber (DPFSF) reinforced epoxy composites incorporating areca nut husk biofiller were experimentally fabricated and evaluated to elucidate and predict their mechanical characteristics. Three supervised ML models Gradient Boosting, k-Nearest Neighbor and Random Forest was trained and constructed using comprehensive experimental dataset. Random Forest model demonstrated greatest prediction, attaining R2 values of 0.90 for Tensile Strength (TS) and 0.95 for Flexural strength (FS) during validation. To tackle “black box” issue of machine learning, SHapley Additive exPlanations (SHAP) analysis was utilized, identifying filler percentage and length as primary determinants influencing strength estimates. This study demonstrates that explainable machine learning is a viable and sustainable method for optimizing hybrid composites based on natural fibers.