Scholarlink

<p>Additive manufacturing (AM) offers exceptional design flexibility, enabling the production of complex geometries with minimal material waste. Fused deposition modeling (FDM) is one of the most common and widely spread technologies in AM. Accurate prediction of surface roughness and strength is essential to ensure structural integrity and expand industrial adoption. This study investigates the effects of six process parameters; nozzle temperature, infill density percentage, layer height, printing speed, raster angle, and wall thickness, on the bi-optimization of surface roughness and ultimate tensile strength in polylactic acid (PLA) specimens. A Taguchi <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(L_{32}\)</EquationSource> </InlineEquation> orthogonal array integrated with weighted grey relational analysis is used for bi-objective optimization. Analysis of variance (ANOVA) reveals the infill percentage as the dominant factor influencing performance. To improve predictive accuracy, a hybrid artificial neural network–genetic algorithm (ANN–GA) model is developed. The proposed hybrid ANN–GA achieves superior prediction accuracies of 95% for training and 99% for testing using a 70:30 data split, compared with the traditional ANOVA model, which attains an <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(R^{2}\)</EquationSource> </InlineEquation> of only 29%. These results demonstrate the robustness of the hybrid ANN–GA framework and its potential as a reliable tool for optimizing 3D printing process parameters, thereby improving both mechanical and surface quality outcomes.</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

Hybrid artificial neural network and genetic algorithm-based grey relational analysis for bi-objective optimization of 3D-printed polylactic acid parts

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Essam Kaoud,
  • Mahmoud Heshmat

摘要

Additive manufacturing (AM) offers exceptional design flexibility, enabling the production of complex geometries with minimal material waste. Fused deposition modeling (FDM) is one of the most common and widely spread technologies in AM. Accurate prediction of surface roughness and strength is essential to ensure structural integrity and expand industrial adoption. This study investigates the effects of six process parameters; nozzle temperature, infill density percentage, layer height, printing speed, raster angle, and wall thickness, on the bi-optimization of surface roughness and ultimate tensile strength in polylactic acid (PLA) specimens. A Taguchi \(L_{32}\) orthogonal array integrated with weighted grey relational analysis is used for bi-objective optimization. Analysis of variance (ANOVA) reveals the infill percentage as the dominant factor influencing performance. To improve predictive accuracy, a hybrid artificial neural network–genetic algorithm (ANN–GA) model is developed. The proposed hybrid ANN–GA achieves superior prediction accuracies of 95% for training and 99% for testing using a 70:30 data split, compared with the traditional ANOVA model, which attains an \(R^{2}\) of only 29%. These results demonstrate the robustness of the hybrid ANN–GA framework and its potential as a reliable tool for optimizing 3D printing process parameters, thereby improving both mechanical and surface quality outcomes.