Reconstruction of a granite structure composed of multiple irregular minerals

Accurately reconstructing rock structures using numerical methods is vital in rock mechanics research community, especially when obtaining rock samples is difficult and expensive. The reconstructed models must reflect the comprehensive characteristics of natural rock, including mineral content and spatial distributions. This study employs the bubbling method to reconstruct granite containing multiple minerals in both two- (2D) and three-dimensions (3D), proposing a general procedure for granite structure reconstruction. The bubbling method utilizes numerous bubbles (hemispheres or spheres) of varying sizes and gradually changing properties, which are randomly overlapped to create a heterogeneous plane (2D) or space (3D). The properties of these overlapped areas are adjusted based on the sum of neighboring bubbles' properties, allowing specific regions with extreme properties to be selected and intercepted to form the desired mineral shapes. The results demonstrate that the reproduced granite samples can accurately exhibit the mineral distributions and sizes of real granite, quantified by fractal dimension ( D ) and the hourglass parameter ( V Sum ∗ / V Total ). The proposed method is also suitable for reconstructing anisotropic granite models, with anisotropy described by a fitted elliptic curve derived from ratios between directional mineral sizes and cross-sectional dimensions. Based on these findings, a series of numerical granite models with similar structures were reconstructed and tested. Results indicate that different mineral distributions significantly impact the macroscopic mechanical behaviors, but variability in numerical simulation results decreases with increasing specimen size. The compressive and tensile strength values of the reconstructed numerical models show less variation than those of natural granite specimens. This suggests that, beyond mineral distribution, other factors such as internal defects within natural granite contribute to the observed discrepancies. Additionally, the bubbling method shows great potential for modeling porous structures and offers high computational efficiency.