[Objective] To investigate the meso-scale fracture mechanisms and crack evolution of hydraulic concrete under tensile loading, and to address the inadequate representation of initial defects and the interfacial transition zone (ITZ) in conventional meso-models, a four-phase 3D meso-scale model consisting of aggregates, mortar, an ITZ, and initial defects is established. The effects of initial defect ratio, ITZ thickness, and aggregate content on tensile strength, failure modes, and energy dissipation are quantified. The innovation of this study lies in explicitly introducing the initial defects with controllable volume fractions within a 3D meso-scale framework and coupling them with ITZ and aggregate factors to enhance the reproducible representation of experimental responses and failure patterns. [Methods] Uniaxial tensile tests on cylindrical specimens of ordinary-strength hydraulic concrete (C40) were conducted to obtain stress-strain curves and tensile strength for model validation. Numerically, a 3D meso-scale geometry was established based on randomly generated aggregates. Aggregates were modeled using a linear-elastic constitutive law, while mortar and ITZ were modeled with a damage-plasticity constitutive law. Spherical initial defects were randomly embedded in the mortar phase to represent pores and micro-cracks. After validating the model against experimental strength, curve morphology, and failure modes, a parametric study was performed to investigate the effects of initial defect ratio (0-7%), ITZ thickness (including a control without ITZ), and aggregate content (10%-40%) on tensile fracture behavior. [Results] (1) The four-phase 3D model well reproduced the mechanical response and crack propagation pattern throughout the uniaxial tensile process. Simulated tensile strength was generally overestimated when initial defects were not considered. (2) Simulated and experimental tensile strengths and stress-strain curves agreed well with a initial defect ratio of 1%-2%. As the defect ratio increased, tensile strength decreased significantly. Cracks tended to concentrate in the mid-region of the specimen and propagate rapidly, exhibiting more pronounced strength weakening, earlier instability failure, and concurrent deterioration in energy dissipation capacity during fracture. (3) Within the small range examined, variations in ITZ thickness had a relatively limited influence on peak tensile strength. However, omitting the ITZ led to systematic deviations in strength and failure mode, making it difficult to correctly capture crack paths and local damage evolution. (4) An increase in aggregate content further reduced the tensile strength and significantly altered crack propagation paths and patterns. This reflected an important modulating role of changes in the proportion of aggregate/ITZ weakened regions on crack evolution. [Conclusion] The proposed four-phase 3D meso-scale modeling method incorporating initial defects provides balanced accuracy in reproducing both mechanical responses and failure patterns. The study demonstrates that for ordinary-strength hydraulic concrete, properly characterizing an initial defect ratio of 1%-2% is crucial for improving the accuracy of tensile fracture simulation. An increase in the initial defect ratio significantly reduces tensile strength and promotes concentrated crack coalescence in the specimen’s mid-region. The ITZ has a non-negligible controlling effect on crack paths and macroscopic responses, and ignoring it will lead to significant errors. Aggregate content further affects strength and crack patterns by altering the proportion of weakened interfaces. The findings can provide references for the determination of meso-scale parameters for hydraulic concrete, risk assessment of dam cracking, and enhancement of structural safety.