Fracture networks control the hydraulic and mechanical behavior of rock masses, with direct implications for subsurface engineering, including nuclear waste disposal, geothermal energy, and groundwater management. Conventional Poisson-based discrete f...
Fracture networks control the hydraulic and mechanical behavior of rock masses, with direct implications for subsurface engineering, including nuclear waste disposal, geothermal energy, and groundwater management. Conventional Poisson-based discrete fracture network (DFN) models generate unrealistic topologies dominated by X-shaped crossing fracture intersections, while natural systems exhibit Y-shaped abutting intersections. This topological misrepresentation can lead to significant errors in predicting flow and transport behavior of fracture networks.
This study develops a comprehensive genetic DFN (G-DFN) modeling framework that incorporates simplified kinematic growth rules to generate high-fidelity DFN models while remaining practical for field-scale applications. The development progresses systematically from establishing field evidence, through developing a DFN generation algorithm and calibration methodology, and validating engineering implications through hydraulic analysis and field application.
First, evidence supporting Charles's law growth rates and contact-driven arrest is identified by field-scale analysis of fracture trace length distributions from multiple sites, establishing an observational foundation for the growth rules of the modeling approach. Building on this evidence, a polygon-based G-DFN generation algorithm is developed in which fractures grow according to Charles's law and are arrested upon contact with larger fractures. This approach combines polygonal fracture geometry with purely interaction-driven growth, enabling realistic network topologies to emerge without imposing abutment and predetermined target sizes.
To generate site-specific G-DFN models, a systematic two-stage calibration scheme is developed that links model parameters to features from fracture trace maps through direct calibration of observable parameters and iterative calibration of growth-related parameters. The iterative calibration compares simulated trace map features with field observations and updates model parameters to reduce discrepancies. The methodology is validated through synthetic tests and field applications at the Hornelen Basin (Norway) and the Yongin Demonstration Tunnel (Korea), where calibrated G-DFN models successfully reproduce observed trace map features.
Next, flow simulation and connectivity analysis are conducted to assess the engineering significance of G-DFN models in terms of hydraulic behavior. Compared to conventional Poisson DFN (P-DFN) models, G-DFN models exhibit comparable bulk hydraulic conductivity with stronger flow channeling through fewer intersections concentrated toward large fractures. These characteristics align with flow behavior commonly observed in natural fracture systems, suggesting that P-DFN models may overlook channeling effects.
Finally, the developed G-DFN modeling framework is applied to the Korea Underground Research Tunnel (KURT) site to demonstrate its practical utility. Despite calibration to the same site data, G-DFN and P-DFN modeling approaches produce distinct network geometries and flow characteristics, with G-DFN models more accurately reproducing both the magnitude and variability of field-observed hydraulic conductivity. These results demonstrate that G-DFN modeling produces systematic differences beyond network topology, warranting investigation as an alternative modeling framework alongside conventional DFN approaches.