Scholarlink

<p>Micro-scale laser powder bed fusion (µ-LPBF) offers unique opportunities for fabricating high-precision components, yet the submicron particle size of powders makes uniform layer deposition extremely challenging. To address this bottleneck, we introduce a spray-based powder spreading strategy for µ-LPBF, which enables more uniform deposition of submicron powders compared with conventional methods. The new process significantly expands the dimensionality of process variables, coupling spray parameters with laser and drying conditions. To efficiently explore this high-dimensional space, we develop a machine-learning assisted framework that predicts track morphology features with high accuracy (R² ≥ 0.8) even under small datasets, and employ SHAP analysis to link statistical feature importance with the underlying physical mechanisms of spray-based deposition. Furthermore, we integrate the predictive models with a multi-objective optimization scheme tailored to µ-LPBF constraints, identifying parameter sets that minimize surface roughness and track variability. Experimental validation demonstrates that optimized parameters achieve submicron surface quality (Ra &lt; 0.5&#xa0;μm) and stable morphology. This work presents the first demonstration of spray-based powder spreading in µ-LPBF combined with data-driven process design, highlighting a novel pathway toward intelligent control of micro-scale additive manufacturing.</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

Spray-based µ-LPBF with machine learning-guided track morphology prediction and optimization

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • Yuqi Zhang,
  • Min Zheng,
  • Linqiang Liu,
  • Guochen Dong,
  • Rongshi Xiao,
  • Ting Huang

摘要

Micro-scale laser powder bed fusion (µ-LPBF) offers unique opportunities for fabricating high-precision components, yet the submicron particle size of powders makes uniform layer deposition extremely challenging. To address this bottleneck, we introduce a spray-based powder spreading strategy for µ-LPBF, which enables more uniform deposition of submicron powders compared with conventional methods. The new process significantly expands the dimensionality of process variables, coupling spray parameters with laser and drying conditions. To efficiently explore this high-dimensional space, we develop a machine-learning assisted framework that predicts track morphology features with high accuracy (R² ≥ 0.8) even under small datasets, and employ SHAP analysis to link statistical feature importance with the underlying physical mechanisms of spray-based deposition. Furthermore, we integrate the predictive models with a multi-objective optimization scheme tailored to µ-LPBF constraints, identifying parameter sets that minimize surface roughness and track variability. Experimental validation demonstrates that optimized parameters achieve submicron surface quality (Ra < 0.5 μm) and stable morphology. This work presents the first demonstration of spray-based powder spreading in µ-LPBF combined with data-driven process design, highlighting a novel pathway toward intelligent control of micro-scale additive manufacturing.