Study on hardening and damage behaviors of Ti6Al4V alloy based on WOA-BP neural network

Ti6Al4V alloy is widely used in aerospace, marine, and chemical industries due to its excellent specific strength and good biocompatibility. Understanding the damage and fracture mechanism of Ti6Al4V is crucial for the practical applications. In this work, the deformation and failure behaviors of Ti6Al4V were studied by digital image correlation method. Post-fracture surface analysis was performed to investigate the influence of stress triaxiality on failure behavior. A machine learning-based identification strategy was proposed to determine the strain hardening and Johnson–Cook damage model parameters. The datasets were obtained by finite element simulations for training the artificial neural network (ANN) models, which were utilized to establish the relation between the mechanical response and model parameters. The effect of ANN structure hyperparameters on prediction performance was discussed and whale optimization algorithm (WOA) could improve the prediction accuracy of neural network model. The results indicated that the WOA algorithm optimized three-layer BP neural network with 16 hidden neurons and activation functions of tansig + tansig can be used to effectively identify the plastic and damage model parameters of Ti6Al4V alloy.