Processing windows for Al-357 by LPBF process: a novel framework integrating FEM simulation and machine learning with empirical testing

Abstract

Purpose This study aims to develop a holistic method that integrates finite element modeling, machine learning, and experimental validation to propose processing windows for optimizing the laser powder bed fusion (LPBF) process specific to the Al-357 alloy. Design/methodology/approach Validation of a 3D heat transfer simulation model was conducted to forecast melt pool dimensions, involving variations in laser power, laser scanning speed, powder bed thickness (PBT) and powder bed pre-heating (PHB). Using the validated model, a data set was compiled to establish a back-propagation-based machine learning capable of predicting melt pool dimensional ratios indicative of printing defects. Findings The study revealed that, apart from process parameters, PBT and PHB significantly influenced defect formation. Elevated PHBs were identified as contributors to increased lack of fusion and keyhole defects. Optimal combinations were pinpointed, such as 30.0 µm PBT with 90.0 and 120.0 °C PHBs and 50.0 µm PBT with 120.0 °C PHB. Originality/value The integrated process mapping approach showcased the potential to expedite the qualification of LPBF parameters for Al-357 alloy by minimizing the need for iterative physical testing.