Macroscale damage model calibrated from discrete mesoscale data with a consistent description of fracture

This study shows that nonlocal models for fracture can be calibrated from field data, e.g., strain fields, stress and other internal variable fields, including their internal length. There is no need to consider specimens of various geometries or size effect data. However, such a calibration should be performed with caution as the fracture process zone ought to be well described. This work starts from a discrete model (LDPM) viewed as a high-fidelity model, and uses it to generate a synthetic data set for the calibration of a macroscale damage model. Compared to the methodology where the internal length and material parameters are obtained from fitting field data altogether, the present study shows that constraining the internal length in the nonlocal model to fit the width of the FPZ prior to calibration yields more accurate results. Consistent structural responses and correct size effects predictions are obtained. • Nonlocal model for fracture is calibrated from strain and stress field data obtained with a high fidelity discrete model. • A proper calibration needs to start from an accurate description of the size of the FPZ. • Internal length is adjusted to the size of the FPZ. • Calibration yields a consistent prediction of size eBect on notched bending beams. • Comparison with results obtained on L-Shaped specimen, including size eBect show a reasonable agreement.