A hybrid framework combining damage mechanics and Physics-Informed neural Networks for damage and failure analysis of additively manufactured lattice structures

Additive manufacturing (AM) lattice structures offer high load-bearing capacity with lightweight design, yet their mechanical behavior remains insufficiently understood. This study develops a Continuum Damage Mechanics (CDM)-based framework to describe constitutive response and damage evolution in AM materials, together with its numerical implementation. The model is applied to simulate compression of body-centered cubic with vertical strut (BCCZ) lattices. Results show that the framework effectively captures the behavior of multi-layer lattice structures and accurately estimates compressive modulus and initial ultimate strength. Parametric studies indicate that larger cell size or layer number reduces stiffness and strength, while increasing the number of cells per layer significantly enhances them. To improve predictive efficiency, three machine learning (ML) models are constructed: an artificial neural network (ANN), a weak boundary effects-based physics-informed neural network (PINN), and a physical law-driven PINN. Comparative analysis demonstrates that the two PINN models outperform ANN, as physical knowledge improves prediction accuracy, generalization, and physical consistency while reducing overfitting. This integrated theoretical–numerical–data-driven approach provides new insights for analyzing and predicting the mechanical performance of AM lattice structures.