Scholarlink

<p>The fiber-reinforced polymer matrix composites (FRPMCs) are extensively adopted in engineering, building and construction applications for their high strength, durability, lightweight and excellent thermal properties. Choosing appropriate material combinations and the composite properties effects of hybrid reinforcements and fillers, however, are still difficult. In this work, eco-friendly fillers such as biochar, biosilica and granite dust at varying weight fractions were combined with polyester resin as the matrix reinforced with glass, bamboo and rock wool fibers to create hybrid FRPMCs. Experimental characterization was used to investigate the physical, thermal, structural and microstructural properties of the composites produced. Based on these experimental results, the properties of density, water absorption, flammability and thermal conductivity were chosen as target properties to be predicted. With these as input features, a hybrid Message Passing Neural Network with TabTransformer, optimized by Multi-strategy Boosted Aquila Optimizer (MPNN-MuBoAOpt-TT), was developed and trained for their efficient prediction. Data pre-processing, such as data augmentation, normalization and SI unit standardization, was used prior to the creation of the model. Using the Chaotic Initialized Multi-objective Reptile Search Algorithm (Chao-MORSA), the best composite formulation was identified. The BS3 composite sample had the best overall thermal and physical properties among all the samples tested. The results indicated that the proposed model had a strong predictive capability for each individual target property, with values of R<sup>2</sup> of 0.9902 for density, 0.9917 for water absorption, 0.9901 for flammability and 0.9881 for thermal conductivity. Overall, this study provides a suitable framework for engineering sustainable, high-performance hybrid composites with tailored physical and thermal characteristics.</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

Experimental Characterization and AI-Based Prediction of Physical and Thermal Properties of Hybrid Fiber-Reinforced Polymer Matrix Composites Using Deep Learning Framework

  • 期刊论文

  • 发表日期：

    2026年7月18日

查看全文
  • Nithyapriya Selvaraj,
  • Sampathkumar Velusamy,
  • Rajkumar Tharmalingam,
  • Vivek Sivakumar

摘要

The fiber-reinforced polymer matrix composites (FRPMCs) are extensively adopted in engineering, building and construction applications for their high strength, durability, lightweight and excellent thermal properties. Choosing appropriate material combinations and the composite properties effects of hybrid reinforcements and fillers, however, are still difficult. In this work, eco-friendly fillers such as biochar, biosilica and granite dust at varying weight fractions were combined with polyester resin as the matrix reinforced with glass, bamboo and rock wool fibers to create hybrid FRPMCs. Experimental characterization was used to investigate the physical, thermal, structural and microstructural properties of the composites produced. Based on these experimental results, the properties of density, water absorption, flammability and thermal conductivity were chosen as target properties to be predicted. With these as input features, a hybrid Message Passing Neural Network with TabTransformer, optimized by Multi-strategy Boosted Aquila Optimizer (MPNN-MuBoAOpt-TT), was developed and trained for their efficient prediction. Data pre-processing, such as data augmentation, normalization and SI unit standardization, was used prior to the creation of the model. Using the Chaotic Initialized Multi-objective Reptile Search Algorithm (Chao-MORSA), the best composite formulation was identified. The BS3 composite sample had the best overall thermal and physical properties among all the samples tested. The results indicated that the proposed model had a strong predictive capability for each individual target property, with values of R2 of 0.9902 for density, 0.9917 for water absorption, 0.9901 for flammability and 0.9881 for thermal conductivity. Overall, this study provides a suitable framework for engineering sustainable, high-performance hybrid composites with tailored physical and thermal characteristics.