Integrating Machine Learning to Investigate the Effect of Process Parameters on the Toughness of Additively Manufacturing 316L Stainless Steel

Abstract

Abstract. This study investigates the impact toughness of additively manufactured 316L stainless steel using the bound metal deposition (BMD) technique and explores the influence of process parameters including print orientation, outer wall thickness, skin overlap percentage, and printing sequence. Charpy V-notch impact tests were conducted on samples produced with varying configurations, followed by predictive modeling using machine learning (ML). The results demonstrate that a 45 deg printing orientation and increased outer wall thickness significantly enhance impact energy absorption, with a peak value of 49.89 J. The optimal skin overlap was found to be 0%, yielding the most uniform material structure and highest toughness across both infill-first and outer-wall-first strategies. A ridge regression model was developed to predict impact energy based on printing parameters, achieving modest predictive accuracy (mean R2 = 0.151) due to dataset size and variability. Although predictive power was limited, the study highlights the potential of ML in parameter optimization for metal additive manufacturing. These findings provide valuable insights for improving the mechanical performance of 3D-printed metal components, particularly in impact-critical applications.