AI-enhanced elasto-plastic damage model for recycled aggregate concrete under multiaxial stress

In this study, an elasto-plastic damage model (EPDRAC) for recycled aggregate concrete (RAC) is proposed, capable of predicting the behaviour of RAC subjected to a multiaxial state of stress. The proposed model used four parameters: α, β, γ, and critical strain energy release rate (R_c). Parameters α and β are used to predict the behaviour of concrete in tension and compression, respectively, while γ is used for predicting volumetric dilatation, and R_c controls the damage growth rate. These parameters are functions of concrete compressive strength (⁠$fc′$⁠), its initial elastic modulus (E_o), and normalised invariants of strain $I1ε3$ and $J2′e32.$ Initially, artificial neural networks (ANN) were used to estimate the compressive strength and modulus of elasticity of RAC. Furthermore, parameters α and β were then estimated using ANN. The proposed model was calibrated and validated using the experimental data generated during the course of the study as well as available in existing literature. The proposed model was able to capture the pre- and post-peak behaviour of RAC accurately. The integration of ANN for parameter estimation significantly enhanced the proposed model’s performance compared with prior models, both by the authors and in the existing literature.