Optimization of surface roughness and dimensional accuracy in LPBF additive manufacturing

• Laser powder bed fusion Taguchi experiments were conducted with 316L stainless steel. • A machine learning framework, Kriging-WOA, was proposed for the optimization of process parameters. • Better surface roughness and dimensional accuracy were obtained after the optimization. • Confirmation experiments revealed the effectiveness and reliability of the built model. • The main effects and contribution rates of process parameters are analyzed. Laser powder bed fusion (LPBF) is one of the most promising additive manufacturing technologies. It has been utilized in the high level and stringent requirements fields such as aerospace and biomedicine industries. However, compared to subtractive manufacturing, the relatively poor surface finish and dimensional accuracy of the LPBF part hamper its widespread applications. In this work, a data-driven framework is proposed to obtain optimal process parameters of LPBF to get satisfactory surface roughness and dimensional accuracy. The effects of key process parameters on the surface roughness and dimensional accuracy are analyzed. Specifically, a machine learning technique is defined to reflect the dimensional accuracy and the surface roughness of the as-built products under different combinations of process parameters. Considering the limited experimental data, a machine learning model is introduced to predict the surface roughness and dimensional accuracy in the whole process parameters space. Then the predicted value is considered as an objective value when using the whale optimization algorithm (WOA) to search the global optimal process parameters. In the verification experiments, LPBF parts with better surface finish and dimensional accuracy were obtained with optimized process parameters which indicates that the optimized results are consistent with the experimental results.