Scholarlink

<p>Three-dimensional (3D) printing is an emerging manufacturing method for developing lightweight and high-strength composite components useful in many industrial sectors. This study investigates the mechanical properties of continuous carbon fiber reinforced Onyx composites manufactured using a Markforged Mark Two 3D Printer. Design of Experiments (DOE) is applied to estimate the combined influence of fill density, roof layer thickness, wall thickness and added fiber thickness on the tensile strength, elongation and Shore-D hardness of the 3D printed specimens. Tensile strength samples were prepared as per ASTM D638 and tested under controlled loading conditions. The hardness measurements were conducted on 3D printed specimens using a Shore-D durometer. The results evidence show that fiber thickness and fill density were the higher influential parameters on tensile strength. However, increased fiber reinforcement enhanced stiffness and surface hardness up to an optimum level. Polynomial Regression, Gaussian Process Regression (GPR), Support Vector Regression, Random Forest and Gradient Boosting Machine Learning (ML) models were developed using the DOE values to interpret nonlinear interactions between process parameters. Among these, the GPR model exhibited the most consistent prediction performance and was used for response surface analysis and multi-response optimization. The optimized parameter values (40% infill, 1.25&#xa0;mm roof, 6&#xa0;mm wall and 3.5&#xa0;mm fiber addition) provided simultaneous improvement in strength, ductility and hardness. The predicted results exhibited good correlation with experimental observations, confirming the reliability of the developed models. Further, Scanning Electron Microscopy (SEM) analysis revealed improved fiber–matrix bonding, reduced void content and better interlayer fusion in the optimized samples, supporting the observed mechanical behaviour. The integration of experimental evaluation with ML based modelling offers a potential pathway for parameter optimization within the studied material system and process conditions.</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 and optimization of mechanical performance in 3D printed continuous carbon fiber reinforced onyx composites

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月26日

查看全文
  • Manmadhachary A.,
  • Nagavelly Shiva Kumar,
  • Murahari Kolli

摘要

Three-dimensional (3D) printing is an emerging manufacturing method for developing lightweight and high-strength composite components useful in many industrial sectors. This study investigates the mechanical properties of continuous carbon fiber reinforced Onyx composites manufactured using a Markforged Mark Two 3D Printer. Design of Experiments (DOE) is applied to estimate the combined influence of fill density, roof layer thickness, wall thickness and added fiber thickness on the tensile strength, elongation and Shore-D hardness of the 3D printed specimens. Tensile strength samples were prepared as per ASTM D638 and tested under controlled loading conditions. The hardness measurements were conducted on 3D printed specimens using a Shore-D durometer. The results evidence show that fiber thickness and fill density were the higher influential parameters on tensile strength. However, increased fiber reinforcement enhanced stiffness and surface hardness up to an optimum level. Polynomial Regression, Gaussian Process Regression (GPR), Support Vector Regression, Random Forest and Gradient Boosting Machine Learning (ML) models were developed using the DOE values to interpret nonlinear interactions between process parameters. Among these, the GPR model exhibited the most consistent prediction performance and was used for response surface analysis and multi-response optimization. The optimized parameter values (40% infill, 1.25 mm roof, 6 mm wall and 3.5 mm fiber addition) provided simultaneous improvement in strength, ductility and hardness. The predicted results exhibited good correlation with experimental observations, confirming the reliability of the developed models. Further, Scanning Electron Microscopy (SEM) analysis revealed improved fiber–matrix bonding, reduced void content and better interlayer fusion in the optimized samples, supporting the observed mechanical behaviour. The integration of experimental evaluation with ML based modelling offers a potential pathway for parameter optimization within the studied material system and process conditions.