Fracture properties prediction of Ultra-High-Performance Concrete (UHPC) after High-Temperature exposure based on Meso-Scale finite element analysis and artificial neural network

High temperature can severely weaken the mechanical properties of UHPC. As the core index for evaluating the material's ability to resist crack propagation, the fracture property should be given more attention. Using the combination of experiments, finite-element models, and artificial neural network models is of theoretical and practical significance for understanding and predicting the fracture properties of UHPC after high-temperature exposure. In this study, two types of UHPC were first prepared. Their fracture properties were investigated after being heated from normal temperature (25 °C) to 200 °C, 400 °C and 600 °C respectively, and then naturally cooled to normal temperature. Subsequently, the matrix's cracking phenomenon was simulated using the Cohesive Zone Model (CZM). For steel fibers, an equivalent failure model was adopted to consider the non-linear deformation relationship between the fibers and the matrix, and then a fracture model of UHPC was established after high-temperature exposure. From a quantitative perspective, the influence of the mechanical properties of the matrix and fibers after high-temperature exposure on the macroscopic fracture properties of UHPC was analyzed. Eventually, based on the experimental and finite-element simulation data, an artificial neural network model capable of predicting the fracture properties of UHPC after high-temperature exposure was constructed. The research conclusions are as follows: High-temperature exposure does not change the fracture morphology of UHPC specimens. For UHPC-containing SF fibers, the loss of strength after high-temperature exposure is slower than that of UHPC without steel fibers. High-temperature exposure remodels the original characteristics of the curve. For UHPC-containing SF fibers, the loss of fracture toughness after high-temperature exposure is slower than that of UHPC without steel fibers. SF fibers can effectively delay the crack propagation in UHPC before the peak load after high-temperature exposure. The calculation results of the UHPC fracture model after high-temperature exposure, which is established based on the CZM and equivalently considers the bond-slip between steel fibers and the matrix, are consistent with the experimental results and can accurately describe the mechanism of the change in UHPC fracture characteristics after high − temperature exposure. The explicit solution proposed based on the artificial neural network model has high accuracy in predicting the fracture properties of UHPC after high-temperature exposure. It can be an important tool for calculating fracture properties in multivariable high-temperature experiments.