XFEM and machine learning combined approach for failure prediction of microcapsules in cement-based self-healing materials

The cracking behavior of microcapsules in cement-based self-healing materials is essential for releasing healing agents during crack healing. In this study, the extended finite element method and machine learning were combined to predict the competitive behavior of microcapsule debonding and rupture. The fracture criterion based on energy was developed to investigate the interaction of the crack and microcapsule for heterogeneous materials. The artificial neural network was trained on 1031 datasets from extended finite element method (XFEM) simulations, and the mean square error (MSE) was 3.7E-3 and the overall R2 was 0.94 in 16 epochs. With the proposed model, the cracking behavior of microcapsules can be judged accurately and instantly without extensive computation, which can help clarify the release condition of healing agents encapsulated in microcapsules so as to guide the design of intelligent composite materials.