Modeling the response of <scp>UHPC</scp> shear‐critical beams: Integrating nonlinear finite element analysis and artificial neural network

Abstract An analytical framework suited for the analysis of shear‐critical ultra‐high performance concrete (UHPC) members is presented. The numerical methodology utilizes a nonlinear finite element analysis formulation integrated with an artificial neural network (ANN) that characterizes the UHPC tension response based on its mix design. In addition, a novel compression softening model specifically tailored for UHPC is introduced. Both of these behavioral mechanisms are necessary for a realistic assessment of the structural behavior. Special consideration is given to the influence of crack widths and the calculation of crack spacing, specific to UHPC materials. The ANN revealed that the tensile behavior of UHPC is influenced not only by the characteristics of fiber reinforcement but also by the mix design constituents. Validation studies successfully reproduced the response of published experiments on shear‐critical panel specimens and beams. This study also highlights the crucial impact of UHPC direct tension characteristics on the behavior of shear‐critical members. Furthermore, the influence of compression softening on the accuracy of the analytical results was found to be dependent on the magnitude of compressive stresses present.