Scholarlink

<p>In the present-day manufacturing industries, 3D printing appears as an emerging technology to convert complicated product geometries into real-world models, mainly through the fused deposition modeling method. While fabricating 3D printed products having higher dimensional stability and better surface quality, polylactic acid (PLA), with superior strength and biodegradation properties, has found broad applications in many of the food packaging, healthcare and medical, structural and textile industries. In this paper, based on Taguchi’s orthogonal array, 27 drilling operations are performed to examine the impacts of spindle speed (SS), feed rate (FR) and drill diameter (DD) on material removal rate (MRR), surface roughness (SR), delamination factor (DF), cylindricity (CYL) and circularity (CIR) of 3D-printed PLA polymer. The related main effects plots examine the influences of those drilling parameters on the responses, while analysis of variance results identify their significance on the drilling performance. Higher SS is responsible for higher MRR together with lower SR, DF, CYL and CIR. On the contrary, all the considered responses show increasing trends at higher FR values, while higher DD results in higher MRR, SR and CIR, and lower DF and CYL values. In the subsequent step, the said drilling process is optimized using a multi-criteria decision making tool, i.e. technique for order of preference by similarity to the ideal solution (TOPSIS). Finally, seven machine learning algorithms, i.e. decision tree, random forest, AdaBoost, gradient boosting, extreme gradient boosting, CatBoost and extremely randomized trees are employed to predict the related response values during drilling of 3D-printed PLA polymer. Among them, random forest, AdaBoost, gradient boosting and extremely randomized trees have comparatively better prediction performance with respect to the considered statistical error metrics.</p> Graphical Abstract <p></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

Parametric analysis, optimization and machine learning-based prediction during drilling of 3D-printed polylactic acid polymer

  • 期刊论文

  • 发表日期：

    2025年5月3日

查看全文
  • K. Shunmugesh,
  • Baneswar Sarker,
  • Shankar Chakraborty

摘要

In the present-day manufacturing industries, 3D printing appears as an emerging technology to convert complicated product geometries into real-world models, mainly through the fused deposition modeling method. While fabricating 3D printed products having higher dimensional stability and better surface quality, polylactic acid (PLA), with superior strength and biodegradation properties, has found broad applications in many of the food packaging, healthcare and medical, structural and textile industries. In this paper, based on Taguchi’s orthogonal array, 27 drilling operations are performed to examine the impacts of spindle speed (SS), feed rate (FR) and drill diameter (DD) on material removal rate (MRR), surface roughness (SR), delamination factor (DF), cylindricity (CYL) and circularity (CIR) of 3D-printed PLA polymer. The related main effects plots examine the influences of those drilling parameters on the responses, while analysis of variance results identify their significance on the drilling performance. Higher SS is responsible for higher MRR together with lower SR, DF, CYL and CIR. On the contrary, all the considered responses show increasing trends at higher FR values, while higher DD results in higher MRR, SR and CIR, and lower DF and CYL values. In the subsequent step, the said drilling process is optimized using a multi-criteria decision making tool, i.e. technique for order of preference by similarity to the ideal solution (TOPSIS). Finally, seven machine learning algorithms, i.e. decision tree, random forest, AdaBoost, gradient boosting, extreme gradient boosting, CatBoost and extremely randomized trees are employed to predict the related response values during drilling of 3D-printed PLA polymer. Among them, random forest, AdaBoost, gradient boosting and extremely randomized trees have comparatively better prediction performance with respect to the considered statistical error metrics.

Graphical Abstract