Data-driven method to assess the influence of process parameters on the fatigue response of additively manufactured Ti6Al4V

The fatigue behavior of Additive Manufacturing (AM) parts is influenced by manufacturing defects, whose dimensions are primarily determined by the parameters of the AM process, which, in turn, also affect the resulting microstructure, together with heat treatments. This study employs Machine Learning (ML) techniques to forecast the fatigue response of AM parts from the AM process variables and the heat treatment characteristics. Feed-forward neural networks (FFNN) and physics-informed neural network (PINN) models are formulated and verified employing published datasets on AM Ti6Al4V alloy. The results demonstrate that physics-based ML approaches are effective in forecasting the fatigue response of AM components.