Scholarlink

<p>Additive Manufacturing (AM), specifically stereolithography (SLA), is extensively used in the dental industry due to its high accuracy and fine resolution. This study introduces a novel chemo-mechanical post-processing technique combining barrel finishing and chemical treatment to improve the surface roughness of SLA-printed dental parts, essential for clear aligner production. We employed the Central Composite Design (CCD) method to systematically design experiments, examining the effects of rotational speed, time, and chemical concentration on surface roughness. Our findings show that this post-processing technique can enhance surface smoothness by up to 77%, significantly improving the functionality and aesthetic quality of clear aligners. To address the high costs associated with manufacturing and post-processing, we implemented machine learning algorithms. To achieve this, eight weak learners were combined using a Stacking Regressor, and their performance was evaluated with and without hyperparameter tuning. The Stacking Regressor significantly improved accuracy, achieving a Train MSE of 0.02 and a Test MSE of 0.01, with R<sup>2</sup> values of 0.96 and 0.91, respectively. In comparison, the best individual model, K-Nearest Neighbors (KNN), had a Train MSE of 0.01 and a Test MSE of 0.03, with R<sup>2</sup> values of 0.99 and 0.81. Despite the Stacked Ensemble's higher computational cost (0.1032 seconds, 72.6% of total training time), the gains in predictive accuracy justified this expense. Bayesian optimization further improved the ensemble's performance, increasing the Test R<sup>2</sup> to 0.96 with optimized models.</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

Ensemble machine learning for predicting surface roughness improvement in chemo-mechanical post-processing of additive manufacturing dental parts

  • 期刊论文

  • 发表日期：

    2025年3月10日

查看全文
  • Amir Kiadarbandsari,
  • Mohammad Vahid Ehteshamfar,
  • Hamed Adibi

摘要

Additive Manufacturing (AM), specifically stereolithography (SLA), is extensively used in the dental industry due to its high accuracy and fine resolution. This study introduces a novel chemo-mechanical post-processing technique combining barrel finishing and chemical treatment to improve the surface roughness of SLA-printed dental parts, essential for clear aligner production. We employed the Central Composite Design (CCD) method to systematically design experiments, examining the effects of rotational speed, time, and chemical concentration on surface roughness. Our findings show that this post-processing technique can enhance surface smoothness by up to 77%, significantly improving the functionality and aesthetic quality of clear aligners. To address the high costs associated with manufacturing and post-processing, we implemented machine learning algorithms. To achieve this, eight weak learners were combined using a Stacking Regressor, and their performance was evaluated with and without hyperparameter tuning. The Stacking Regressor significantly improved accuracy, achieving a Train MSE of 0.02 and a Test MSE of 0.01, with R2 values of 0.96 and 0.91, respectively. In comparison, the best individual model, K-Nearest Neighbors (KNN), had a Train MSE of 0.01 and a Test MSE of 0.03, with R2 values of 0.99 and 0.81. Despite the Stacked Ensemble's higher computational cost (0.1032 seconds, 72.6% of total training time), the gains in predictive accuracy justified this expense. Bayesian optimization further improved the ensemble's performance, increasing the Test R2 to 0.96 with optimized models.