Machine learning based very-high-cycle fatigue life prediction of Ti-6Al-4V alloy fabricated by selective laser melting

Few machine learning (ML) models were applied for very-high-cycle fatigue (VHCF) analysis and these methods encounter limitations in data sparsity and overfitting. The present work aims to overcome data sparsity and propose an easy-to-use and nonredundant ML model for VHCF analysis. Monte Carlo simulation (MCs) is run to enlarge dataset size and a ML method is proposed to investigate the synergic influence of defect size, depth, location and build orientation on Ti-6Al-4V. The coefficient factor that indicates the percentage variation between the predicted and experimental fatigue lives can reach up to 0.98, meaning that the model demonstrates good prediction accuracy.