Abstract: Concrete materials are inherently multiphase and heterogeneous. Owing to the coupling between intrinsic randomness and nonlinearity, concrete exhibits an extremely complex mechanical behavior under external loadings. Up to now, accurately predicting the failure behavior of concrete structures throughout their entire loading process remains a formidable challenge. Adopting the integral geometry as a theoretical foundation, a novel nonlocal macro-meso-scale consistent damage (NMMD) model has been proposed and further developed through introducing the meso-configuration within the characteristic scale domain of macroscopic material points. Owing to its clear mechanism of two-scale damage evolution and to the physically grounded geometry-energy conversion relationship, the NMMD model exhibits unique advantages in characterizing the strong nonlinearity of solid fracture failure. The current study employs the second-type stochastic harmonic function to characterize the spatial variability of concrete material properties, and, in conjunction with the NMMD model endowed with the net-type meso-configuration, develops a fracture analysis framework tailored for engineering applications. The model proposed is applied to simulate the crack propagation in three-point bending beams and in wedge splitting tests, both with initial notches of varying lengths and without an initial notch. The computational results demonstrate that: with only a single random field (Young’s modulus) introduced, the model proposed can not only reproduce the random wandering of crack paths and the transition of failure modes but also effectively capture the stochastic fluctuation of load-displacement curves, including the peak loads. Without prescribing fracture energy or fracture toughness as input parameters, the stochastic NMMD model directly yields the evolution of fracture energy with increasing the initial notch length; and its nonconstant characteristics are in a qualitative agreement with experimental observations.