<p>Due to strict sampling limitations in the Dunhuang Grottoes conservation areas, specimen collection is severely restricted. Additionally, the collected samples exhibit considerable variability in their physical and mechanical properties. Traditional numerical modeling methods cannot accurately reconstruct the actual internal morphology of gravel particles. To address these challenges, this study proposes a novel approach for constructing high-precision digital conglomerate core models. Firstly, three-dimensional (3D) laser scanning technology was applied to accurately capture detailed morphological data of gravel particles. Based on this data, a comprehensive digital library of gravel particles was established through computational modeling. This digital library includes particles of various sizes, shapes, and surface features. Subsequently, digital conglomerate core and numerical models were constructed. These models were built according to actual particle distributions and gradation characteristics obtained from real geological strata. Thus, the resulting models realistically represent structural features and possess high geometric accuracy. Furthermore, the mechanical parameters of the digital conglomerate models were calibrated. Nano-indentation experiments were conducted to measure mechanical properties at the microscale. Numerical simulations employing element-size upgrading methods were also used. This approach enabled accurate parameter transfer from the microscale to the macroscale. Finally, digital conglomerate core models with specified complex particle gradations and high precision were successfully developed. The proposed methodology effectively overcomes the limitations associated with traditional sampling and numerical modeling. It significantly improves the accuracy, reliability, and practical applicability of conglomerate characterization research.</p>