Deriving DataDriven Models That Relate Deterministic Surface Topography Parameters of AsBuilt Inconel 718 Surfaces to Laser Powder Bed Fusion Process Parameters

Abstract

Laser powder bed fusion (LPBF) parts often require postprocessing prior to use in engineering applications to improve mechanical properties and modify the asbuilt surface topography. The ability to tune the LPBF process parameters to obtain designer asbuilt surface topography could reduce the need for postprocessing. However, the relationship between the asbuilt surface topography and the LPBF process parameters is currently not wellunderstood. In this paper, we derive datadriven models from surface topography data and LPBF process parameters using machine learning (ML) algorithms. The prediction accuracy of the datadriven models derived from ML algorithms exceeds that of the multivariate regression benchmark because the latter does not always capture the complex relationship between the asbuilt surface topography parameters and the corresponding LPBF process parameters in a single bestfit equation. Datadriven models based on decision tree (interpretable) and artificial neural network (noninterpretable) algorithms display the highest prediction accuracy. We also show experimental evidence that thermocapillary convection and melt track overlap are important drivers of the formation of asbuilt surface topography.