Calibrating a damage model from lattice discrete results

Lattice modeling of concrete is a discrete mesoscale representation of the material, where constitutive relations are prescribed at a smaller scale compared to traditional continuum-based models.These approaches can capture complex nonlinear behavior at the macroscale while maintaining a simpler and less phenomenological constitutive model at the mesoscale.Although these models come with a high computational cost, they are capable of accurately predicting global mechanical behavior and, in several cases, outperform continuum-based models.For this reason, they are considered valuable for generating high-fidelity databases that can be used in data-driven or coarse-graining approaches.In this study, discrete stress-strain findings from the Lattice Discrete Particle Model (LDPM) are upscaled using a coarse-graining technique based on the averaging of conservation equations.The results are used to calibrate a non-local damage model, where the non-local model's length is prescribed by the width of the area in which energy is dissipated in the LDPM calculations.Multiple coarse-graining lengths, ranging from one to five times the maximum aggregate size, are considered.We conclude that the non-local length should better be directly related to the width of the area where the energy is dissipated in the LDPM calculation.We also observe that the calibrated constitutive model provides consistent responses on other structural geometries, including size effect studies.