Scholarlink

<p>Direct additive fabrication of biomedical dental devices is an effective and cost-efficient alternative process to the conventional manufacturing process. Still, due to the disorganized mesh topology and poor scan quality conditions, the geometric fidelity and slicing consistency of CAD models obtained through 3D scanning is frequently impaired. This study is strategic in its approach to enhance the accuracy of CAD analysis through the assessment of the effects of environmental as well as operator-specific scanning parameters. Design of Experiments (DoE) framework Response Surface Methodology (RSM) is used to give the best combinations of the parameters and hence do away with tedious manual experiments. The reconstructed CAD geometries are compared to reference models in order to measure the deviations. A hybrid predictive model of RSM and Artificial Neural Networks (ANN) is adopted that are then optimized to finesse with machine learning based on prescribed constraints. Comparative appraisals identify the best modelling method amongst the alternatives that are tested. Even though the scanning processes are manually performed, close control will reduce the variability introduced by the operator, which is considered a limitation of the study. Also, mesh refinement of optimized CAD results is performed to eliminate duplicate vertices and analyse structural regularity and fabrication precision. The analysis demonstrates the high interdependence between scan conditions and model accuracy that provides useful insights to optimize scanning processes under limited working conditions like human–robot collaborative environment.</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

A Multifaceted Machine Learning Application in Biomedical Device Fabrication: Synergy of 3D Scanning and 3D Printing

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Anmol Sharma,
  • Pushpendra S. Bharti

摘要

Direct additive fabrication of biomedical dental devices is an effective and cost-efficient alternative process to the conventional manufacturing process. Still, due to the disorganized mesh topology and poor scan quality conditions, the geometric fidelity and slicing consistency of CAD models obtained through 3D scanning is frequently impaired. This study is strategic in its approach to enhance the accuracy of CAD analysis through the assessment of the effects of environmental as well as operator-specific scanning parameters. Design of Experiments (DoE) framework Response Surface Methodology (RSM) is used to give the best combinations of the parameters and hence do away with tedious manual experiments. The reconstructed CAD geometries are compared to reference models in order to measure the deviations. A hybrid predictive model of RSM and Artificial Neural Networks (ANN) is adopted that are then optimized to finesse with machine learning based on prescribed constraints. Comparative appraisals identify the best modelling method amongst the alternatives that are tested. Even though the scanning processes are manually performed, close control will reduce the variability introduced by the operator, which is considered a limitation of the study. Also, mesh refinement of optimized CAD results is performed to eliminate duplicate vertices and analyse structural regularity and fabrication precision. The analysis demonstrates the high interdependence between scan conditions and model accuracy that provides useful insights to optimize scanning processes under limited working conditions like human–robot collaborative environment.