FATIGUE LIFETIME ESTIMATION OF ELECTRON BEAM MELTED MICRO-SIZED PARTS BASED ON SURFACES REGENERATED BY MACHINE LEARNING APPROACH

The fatigue response of additive manufactured parts, using the electron beam melting technique, is detrimentally affected by their surface defects. These surface defects, produced by the multiple phenomena during the additive manufacturing (AM) process, are the main origins of premature crack initiation and lead to subsequent fatigue failure. This study focuses on estimating the fatigue lifetime of electron beam melting (EBM) manufactured Ti-6Al-4V micro-sized parts using a combination of machine learning and finite element modelling. A Generative Adversarial Network (GAN) is trained to generate 2D surface profiles of the EBM-manufactured Ti-6Al-4V micro-sized parts based on experimental data taken from the literature. Next, the regenerated 2D surface profiles are randomly used to construct 2D Finite Element (FE) models to find and calculate the stresses at critical defects. Finally, Continuous Damage Mechanics (CDM) and Theory of Critical Distance (TCD) are implemented to estimate the fatigue lifetime. This way, hundreds of simulations are performed using regenerated surface profiles. The obtained results show that using both GAN and finite element simulation makes it possible to numerically reproduce the observed fatigue scatter data which is an inherent characteristic of additively manufactured materials.
EXTENDED ABSTRACT