(1. 山东理工大学 信息中心，淄博 255000；
2. 山东理工大学 机械工程学院，淄博 255000)
摘 要: 基于高斯过程回归(GPR)模型，对激光粉末床熔融Ti-6Al-4V合金的致密度和表面粗糙度观测数据进行了机器学习，得到了高致密度合金样品的激光功率-扫描速度的工艺优化窗口，并探讨了激光功率-扫描速度对表面粗糙度的影响。结果表明：获得高致密(≥99.5%)合金的激光功率-扫描速度工艺窗口呈梨形，扫描速度比激光功率对致密度影响更大，且高功率条件下适宜的扫描速度范围较宽。降低激光功率和提高扫描速度会单调增加表面粗糙度，且在低激光功率和高扫描速度下该影响更显著。同一激光能量密度下打印的合金致密度取决于具体的扫描速度和激光功率，但表面粗糙度基本相同。优化工艺窗口下样品的表面粗糙度小于10 μm。实验证明GPR预测的优化工艺窗口是可靠的，该方法可拓展应用到其他合金增材工艺优化设计中。
关键字: 激光粉末床熔融；机器学习；激光功率；扫描速度
（1. Information Centre, Shandong University of Technology, Zibo 255000, China;
2. School of Mechanical Engineering, Shandong University of Technology, Zibo 255000, China）
Abstract:A machine-learning approach based on Gaussian Process Regression (GPR) was proposed to optimize the processing window of laser power and scanning speed in the Ti-6Al-4V alloy fabricated by laser powder bed fusion (L-PBF). The effect of laser power-scanning speed on surface roughness of the samples was investigated as well. The predicted results from the model show that the optimized L-PBF processing window for manufacturing fully dense Ti-6Al-4V alloy with relative density ≥99.5% is pear-shaped. It is suggested that the scanning speed is more influential than laser power in relative density of the L-PBFed alloy, and the wide favorable scanning speed range can be obtained in the case of high laser power. The lower power and high scanning speed tend to increase surface roughness monotonously and the effect become more pronounced as power decreasing and scanning speed increasing. The relative density of the L-PBFed alloy depends on the specific scanning speed and laser power rather than a single energy density value. However, the surface roughness significantly depends on the energy density and the same energy density employed leads to the similar surface roughness. The optimized laser power-scanning speed processing window brings about the highly dense alloy with surface roughness less than 10 μm. The further experimental evidence proved that the GPR model established in this study is reliable and can be readily applied to the L-PBF process optimization of other metals and alloys.
Key words: laser powder bed fusion; machine learning; laser power; scanning speed
