Numerical Simulation and Machine Learning for Fracture Mechanisms in Layered Rocks: Spatial Correlation Method and Modified Gaussian Process Regression
ABSTRACT A numerical framework integrating spatial correlation‐based digital core reconstruction, numerical fracture analysis, and modified Gaussian Process Regression (MGPR) with multiple kernel functions is proposed to investigate the anisotropic mechanical and fracture responses of layered rocks. By incorporating spatial correlation length and anisotropy parameters, the proposed method effectively reproduces the discontinuous and irregular distribution of bedding planes. The numerical model precisely represents their anisotropic mechanical properties and fracture mechanisms. The peak compressive strength and its dispersion exhibit a U‐shaped trend in relation to the bedding angle, while the peak tensile strength and both tensile and compressive elastic modulus positively correlate with the bedding angle. The bedding plane significantly affects the dispersion of peak compressive strength more than the elastic modulus, but this trend reverses under uniaxial tension. Dispersion in mechanical parameters is closely linked to the related failure mechanisms. Fracture toughness exhibits a negative correlation with bedding angle, pre‐existing crack length, and supporting span. Sensitivity analysis shows that parameter combinations such as a / R = 0.5, S /2 R = 0.6 (0°); a / R = 0.4, S /2 R = 0.6 (30°); and a / R = 0.4, S /2 R = 0.7 (60° and 90°) result in the smallest dispersion in fracture toughness. The MGPR model can accurately predict fracture toughness across different bedding angles. The proposed framework provides an effective approach for characterizing the anisotropic mechanical and fracture responses of layered rocks and improving the reliability of fracture toughness evaluation.
Authors
- Yuan Ruifu
- Bin Gong (ORCID: https://orcid.org/0000-0002-9464-3423)
- Zhuo Dong
- Yingxian Lang
Institutions
- Chang'an University (CN)
- Henan Polytechnic University (CN)
Publication Details
- Journal
- International Journal for Numerical and Analytical Methods in Geomechanics
- Published
- 2026-11-14
- DOI
- https://doi.org/10.1002/nag.70451
- Primary Topic
- Rock Mechanics and Modeling
- Type
- article
- Field-Weighted Citation Impact
- 0.00