Scholarlink

<p>Selective laser melting (SLM) of Ti-6Al-4&#xa0;V provides design flexibility and high material utilization but is highly sensitive to process parameters, which makes it challenging to consistently fabricate fully dense components. This study presents of extensive compilation of data from the literature linking five key process parameters for measured part density: laser power, scan speed, layer thickness, hatch distance, and spot size. Six supervised machine learning models were trained and evaluated for predictive accuracy: linear regression, support vector regression, random forest, gradient boosting, <i>k-</i>nearest neighbors, and an artificial neural network. The model performance was assessed using standard regression metrics for direct comparison between approaches. The results show that gradient boosting achieved the highest accuracy (<i>R</i>² = 0.86, lowest error values), followed by random forest. The remaining models demonstrated moderate to poor predictive capability. Feature importance analysis revealed that hatch distance and layer thickness were the dominant factors influencing density, which reflects their critical roles in melt-pool overlap and interlayer bonding. The findings highlight that data-driven methods can effectively capture complex nonlinear relationships in SLM processing and provide a powerful tool for optimizing parameters, reducing experimental effort, and supporting the development of robust manufacturing for high-performance Ti-6Al-4&#xa0;V components.</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

Comparative analysis of machine learning algorithms for density prediction in selective laser melting of Ti-6Al-4 V

  • 期刊论文

  • 发表日期：

    2025年11月25日

查看全文
  • Yi-Jen Huang,
  • Li-Shang Lin,
  • Sanaz Hadidchi,
  • Amir Reza Ansari Dezfoli

摘要

Selective laser melting (SLM) of Ti-6Al-4 V provides design flexibility and high material utilization but is highly sensitive to process parameters, which makes it challenging to consistently fabricate fully dense components. This study presents of extensive compilation of data from the literature linking five key process parameters for measured part density: laser power, scan speed, layer thickness, hatch distance, and spot size. Six supervised machine learning models were trained and evaluated for predictive accuracy: linear regression, support vector regression, random forest, gradient boosting, k-nearest neighbors, and an artificial neural network. The model performance was assessed using standard regression metrics for direct comparison between approaches. The results show that gradient boosting achieved the highest accuracy (R² = 0.86, lowest error values), followed by random forest. The remaining models demonstrated moderate to poor predictive capability. Feature importance analysis revealed that hatch distance and layer thickness were the dominant factors influencing density, which reflects their critical roles in melt-pool overlap and interlayer bonding. The findings highlight that data-driven methods can effectively capture complex nonlinear relationships in SLM processing and provide a powerful tool for optimizing parameters, reducing experimental effort, and supporting the development of robust manufacturing for high-performance Ti-6Al-4 V components.