Simulation-based investigation on the accuracy of discrete fracture network (DFN) representation

Calibration and validation of the Discrete Fracture Network (DFN) model are always based on the comparison between actual fracture trace mapping and cross-section extracted from the generated model, and some inaccuracies may arise. In this study, the process of fracture data acquisition was simulated; then, the accuracy of the DFN representation technique was investigated. A DFN model (called the original DFN (ODFN) model) was generated based on a natural fracture network. Sampling windows were set inside the ODFN model, and fracture data were collected. A number of additional DFN models, defined as the secondarily constructed DFN (SCDFN) models, were developed. The geometric and mechanical properties of the ODFN and SCDFN models were determined, comparisons between the ODFN and SCDFN models were performed, and the accuracy of the DFN representation technique was analyzed. The results show that the fluctuations in the geometrical properties of these SCDFN models are insignificant and that the mechanical properties of a few SCDFN models considerably differ from those of the ODFN model. These findings indicate that DFN representation is a considerably robust method of characterizing the fracture patterns of in situ rock masses, but care must be exercised when this technique is used in engineering practice.