Scholarlink

<p>Rapid growth in the demand for such eco-friendly and high-performance engineering materials has prompted researchers to conduct a large number of studies on natural fiber-reinforced polymer composites (NFRPCs). This work discusses an extensive investigation that involved both experimental mechanical characterization and machine learning (ML)-based predictive modelling of alkali-treated hybrid polyester composites with the addition of sugarcane bagasse (SB), pineapple leaf fiber (PALF) and sisal fiber (SF). Five ternary hybrid laminates (A–E) were produced using a constant SF fraction of 50 wt% and various SB/PALF fractions (40/10 to 10/40 wt%). All fibers were treated with a standard 5 wt% NaOH to promote the interfacial compatibility with the unsaturated polyester matrix. Mechanical properties were tested according to the ASTM standards (D638, D790, D621, D256, D2240) which include tensile, flexural, compressive, impact and Shore D hardness properties. Specimen E (50% SF/10% SB / 40% PALF) achieved the highest tensile strength (238.65&#xa0;MPa), tensile modulus (8.8 GPa), flexural strength (104.8&#xa0;MPa), compressive load (3168.4&#xa0;N), impact energy (33.1&#xa0;J/m²), and hardness (96.2 Shore D). All five mechanical properties were predicted using four machine learning algorithms: Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GB) trained on the experimental dataset, with fiber composition input. The overall predictive ability of the ANN was best, with an average R² of 0.9912 and RMSE of 2.47&#xa0;MPa for all of the output properties. The feature importance analysis on SHAP showed that the weight fraction of PALF is the most important predictor variable. The scanning electron microscopy (SEM) results supported the enhanced fiber-matrix adhesion for high PALF compositions while the Fourier transform infrared (FTIR) spectroscopy identified the presence of functional group signatures at 679.40, 1527.78, 1744.58, 2344.22, 3556.04, and 3753.06&#xa0;cm-1. The integrated experimental–ML tool is a powerful and data-driven design tool for optimizing the ternary natural fiber composite formulation for automotive, construction and marine engineering 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

Machine learning-based mechanical prediction and characterization of natural fiber hybrid polyester composites for structural and automotive applications

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月29日

查看全文
  • P. Sundaravadivel,
  • S. Sathees Kumar,
  • Pravat Ranjan Pati,
  • Chinmaya P. Mohanty,
  • Nagaraj Ashok

摘要

Rapid growth in the demand for such eco-friendly and high-performance engineering materials has prompted researchers to conduct a large number of studies on natural fiber-reinforced polymer composites (NFRPCs). This work discusses an extensive investigation that involved both experimental mechanical characterization and machine learning (ML)-based predictive modelling of alkali-treated hybrid polyester composites with the addition of sugarcane bagasse (SB), pineapple leaf fiber (PALF) and sisal fiber (SF). Five ternary hybrid laminates (A–E) were produced using a constant SF fraction of 50 wt% and various SB/PALF fractions (40/10 to 10/40 wt%). All fibers were treated with a standard 5 wt% NaOH to promote the interfacial compatibility with the unsaturated polyester matrix. Mechanical properties were tested according to the ASTM standards (D638, D790, D621, D256, D2240) which include tensile, flexural, compressive, impact and Shore D hardness properties. Specimen E (50% SF/10% SB / 40% PALF) achieved the highest tensile strength (238.65 MPa), tensile modulus (8.8 GPa), flexural strength (104.8 MPa), compressive load (3168.4 N), impact energy (33.1 J/m²), and hardness (96.2 Shore D). All five mechanical properties were predicted using four machine learning algorithms: Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GB) trained on the experimental dataset, with fiber composition input. The overall predictive ability of the ANN was best, with an average R² of 0.9912 and RMSE of 2.47 MPa for all of the output properties. The feature importance analysis on SHAP showed that the weight fraction of PALF is the most important predictor variable. The scanning electron microscopy (SEM) results supported the enhanced fiber-matrix adhesion for high PALF compositions while the Fourier transform infrared (FTIR) spectroscopy identified the presence of functional group signatures at 679.40, 1527.78, 1744.58, 2344.22, 3556.04, and 3753.06 cm-1. The integrated experimental–ML tool is a powerful and data-driven design tool for optimizing the ternary natural fiber composite formulation for automotive, construction and marine engineering applications.