A fatigue life posterior analysis approach for laser-directed energy deposition Ti-6Al-4V alloy based on pore-induced failures by kernel ridge

An efficient fatigue life posterior analysis model, a two-step five-feature machine learning one, has been developed on laser-directed energy deposition (LDED) Ti-6Al-4V alloy. The model consists of fine granular area existence prediction model and subsequent fatigue life posterior analysis model; the former adopts ridge classification algorithm and the latter kernel ridge regression algorithm. A comparison of feature importance and correlation coefficient was utilized to improve the generalization ability of the ultimate posterior analysis models via extracting efficient input features from post-mortem analysis, with a focus on microstructure indicators and stress intensity features. The microstructure indicators include pore type and existence label of fine granular area, while the stress intensity is characterized by stress intensity factor range, aspect ratio of pore, the shortest pore distance to the free surface respectively. This is expected to have great potential for fatigue life prediction, specifically laying foundation for fatigue life prediction of LDED Ti-6Al-4V alloy.