Scholarlink

<p>This research examines the impact of key fused filament fabrication (FFF) process parameters—infill density, infill pattern, printing speed, and layer thickness—on the dimensional accuracy of PETG and carbon fiber-reinforced PETG (CF-PETG) composites. A Taguchi-based experimental design was employed to efficiently identify optimal parameter combinations for high geometric fidelity. Results indicate that infill density and pattern are the most influential factors. Higher infill densities (50-90%) with gyroid patterns enhanced dimensional stability by increasing internal rigidity, reducing thermal shrinkage, and promoting uniform stress distribution. Optimal parameters—90% infill density, gyroid pattern, 40&#xa0;mm/s printing speed, and 0.2&#xa0;mm layer thickness—achieved dimensional accuracy up to 99.82% for PETG, while CF-PETG showed comparable or superior stability due to carbon fiber reinforcement, which reduced thermal expansion and increased stiffness. Lower printing speeds improved interlayer bonding, and thinner layers (0.10-0.20&#xa0;mm) enhanced surface fidelity and reduced geometric deviations. ANOVA and residual analysis validated the regression models and confirmed strong correlations between process parameters and dimensional accuracy. To further enhance predictive capability, a random forest-based machine learning (ML) framework was integrated with the Taguchi approach. The ML model achieved excellent performance, with R<sup>2</sup> values of 0.962 for PETG and 0.985 for CF-PETG and low prediction errors (RMSE ≈ 0.035% for PETG and ≈ 0.022% for CF-PETG). Robust clustering along the 45° reference line confirmed prediction reliability. The novelty of this study lies in combining Taguchi optimization with ML-based predictive modeling for CF-PETG dimensional accuracy, an approach scarcely explored in literature. The integrated experimental–ML framework offers a reliable and efficient strategy to optimize FFF parameters for fabricating dimensionally stable thermoplastic composites, with potential applications in aerospace, automotive, and high-precision prototyping.</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-Assisted Optimization of FFF Parameters for Enhanced Dimensional Accuracy in PETG and CF-PETG Composites

  • 期刊论文

  • 发表日期：

    2026年2月9日

查看全文
  • Vishwas Mahesh,
  • Prashanthkumar Hadi,
  • Vinyas Mahesh

摘要

This research examines the impact of key fused filament fabrication (FFF) process parameters—infill density, infill pattern, printing speed, and layer thickness—on the dimensional accuracy of PETG and carbon fiber-reinforced PETG (CF-PETG) composites. A Taguchi-based experimental design was employed to efficiently identify optimal parameter combinations for high geometric fidelity. Results indicate that infill density and pattern are the most influential factors. Higher infill densities (50-90%) with gyroid patterns enhanced dimensional stability by increasing internal rigidity, reducing thermal shrinkage, and promoting uniform stress distribution. Optimal parameters—90% infill density, gyroid pattern, 40 mm/s printing speed, and 0.2 mm layer thickness—achieved dimensional accuracy up to 99.82% for PETG, while CF-PETG showed comparable or superior stability due to carbon fiber reinforcement, which reduced thermal expansion and increased stiffness. Lower printing speeds improved interlayer bonding, and thinner layers (0.10-0.20 mm) enhanced surface fidelity and reduced geometric deviations. ANOVA and residual analysis validated the regression models and confirmed strong correlations between process parameters and dimensional accuracy. To further enhance predictive capability, a random forest-based machine learning (ML) framework was integrated with the Taguchi approach. The ML model achieved excellent performance, with R2 values of 0.962 for PETG and 0.985 for CF-PETG and low prediction errors (RMSE ≈ 0.035% for PETG and ≈ 0.022% for CF-PETG). Robust clustering along the 45° reference line confirmed prediction reliability. The novelty of this study lies in combining Taguchi optimization with ML-based predictive modeling for CF-PETG dimensional accuracy, an approach scarcely explored in literature. The integrated experimental–ML framework offers a reliable and efficient strategy to optimize FFF parameters for fabricating dimensionally stable thermoplastic composites, with potential applications in aerospace, automotive, and high-precision prototyping.