Laser powder bed fusion process optimization of CoCrMo alloy assisted by machine-learning

Abstract

Gaussian process regression (GPR) model of machine learning method was employed to identify the optimal process window for high-performance CoCrMo alloy in laser powder bed fusion (LPBF), considering density (≥99%) and surface roughness (≤7 μm) as key parameters. Additionally, the study examined the impact of LPBF parameters on morphology and distribution of defect and surface roughness. Results revealed a tongue-shaped optimal process window, with scanning speed having a greater influence on density than laser power. High laser power reduced surface roughness, and a combination of medium-to-high laser power (160–340 W) and moderate scanning speed (600–1500 mm/s) achieved low surface roughness (Ra ≤ 7 μm). The mean absolute error confirmed the reliability of the optimized process window predicted by GPR.

Keywords

• Laser power bed fusion
• Machine learning
• relative density
• Surface roughness