Scholarlink

<p>The present work introduces a machine learning framework to predict and optimise process parameters to control microstructure morphology, defect formation and molten pool features in laser-directed energy deposited maraging steel (18Ni-300), thereby addressing a critical gap that remains largely unexplored. To this end, 100 combinations of process parameters, including laser power ranging 300–750&#xa0;W, scanning speed ranging 400–1000&#xa0;mm/min, and feed rate ranging 1–5&#xa0;g/min, were used to deposit single-track, multi-track, and part-scale structures to predict the optimal conditions that produce the desired microstructure with minimal defects. The predicted results revealed that higher laser power and lower scanning speed produced larger molten pools with columnar-dominated microstructure and promoted columnar-to-columnar transition, whereas lower laser power and higher scanning speed produced smaller molten pools with equiaxed-dominated microstructure and promoted columnar-to-equiaxed transition across the multi-layer build. The defect prediction revealed that higher laser power is associated with fewer defects, whereas lower laser power is associated with more defects at intertrack and at the bulk deposition. Furthermore, the predicted results were validated through experimental investigations at the component scale under different processing conditions, including the molten pool at 200–300&#xa0;W laser power with a 400&#xa0;μm hatch spacing, microstructure morphology at 350 and 650&#xa0;W laser power, and defect formation at 400 and 700&#xa0;W laser power. A process map was then proposed based on decision boundaries derived from the predicted and validated results. This study offers comprehensive insight into controlling microstructure evolution and defect formation across a broader processing window in directed energy deposition, enabling the fabrication of an efficient, application-specific component with enhanced mechanical properties.</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

Machine learning based process optimisation to control microstructural evolution and defect formation in laser-directed energy deposition

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • Aditya Pandey,
  • Mishikari Nizamoddin,
  • Vidit Gaur

摘要

The present work introduces a machine learning framework to predict and optimise process parameters to control microstructure morphology, defect formation and molten pool features in laser-directed energy deposited maraging steel (18Ni-300), thereby addressing a critical gap that remains largely unexplored. To this end, 100 combinations of process parameters, including laser power ranging 300–750 W, scanning speed ranging 400–1000 mm/min, and feed rate ranging 1–5 g/min, were used to deposit single-track, multi-track, and part-scale structures to predict the optimal conditions that produce the desired microstructure with minimal defects. The predicted results revealed that higher laser power and lower scanning speed produced larger molten pools with columnar-dominated microstructure and promoted columnar-to-columnar transition, whereas lower laser power and higher scanning speed produced smaller molten pools with equiaxed-dominated microstructure and promoted columnar-to-equiaxed transition across the multi-layer build. The defect prediction revealed that higher laser power is associated with fewer defects, whereas lower laser power is associated with more defects at intertrack and at the bulk deposition. Furthermore, the predicted results were validated through experimental investigations at the component scale under different processing conditions, including the molten pool at 200–300 W laser power with a 400 μm hatch spacing, microstructure morphology at 350 and 650 W laser power, and defect formation at 400 and 700 W laser power. A process map was then proposed based on decision boundaries derived from the predicted and validated results. This study offers comprehensive insight into controlling microstructure evolution and defect formation across a broader processing window in directed energy deposition, enabling the fabrication of an efficient, application-specific component with enhanced mechanical properties.