LAUSR

This study takes sluice foundation rock mass in Datengxia Hydropower Station, China as an example to examine two-dimensional (2D) discrepancies in fracture geometric factors and connectivity between field-collected and stochastically modeled discrete fracture networks (DFNs). We discover that the trace lengths of field-collected and corresponding modeled DFNs diverge, especially with relatively large lengths. A new variable called minimum spacing sequence (MSS) is proposed, which lists minimum spacing between each fracture midpoint and all the other fracture midpoints. The probability density function curve of MSS shows that the fracture locations do not follow a homogeneous (Poisson) model. The following is performed to examine whether the differences will result in noticeable DFN application errors. The 2D fracture connectivity, which is calculated by depth-first search algorithm, is applied to quantify the discrepancies between field-collected and statistically modeled 2D DFNs. Results show that the statistically modeled DFNs have small clustered fracture path numbers and ratios but with considerably large maximum and average lengths for paths (or for paths longer than certain thresholds) owing to the concentration disadvantage and connection advantage of scattered fractures. We comprehensively compare different 2D DFNs (including field-collected DFN, totally modeled DFNs, DFNs with field fracture size, and DFNs with field fracture locations) and conclude that generating statistically modeled DFNs with identical connectivity features is extremely difficult. Mechanical means that consider connections among fractures are recommended for DFN applications.