Modeling fluid migration through physically-based fracture networks

Understanding of the geometry of fracture networks directly affects our ability to model many sub-surface flow systems. Realistic fracture network geometries are generated by including the mechanics of fracture formation within the network simulator. Specifically, this thesis: (1) Derives a simple criterion for predicting the propagation behavior and linkage of fractures that propagate into one another. Experimental data show that the criterion accurately predicts the occurrence of crossing in eleven different brittle materials. (2) Derives fluid-limited fracture velocities. Growth rates are primarily controlled by the hydraulic conductivity, the storage, and the initial flaw length. Growth rates are summarized in dimensionless plots of fracture length versus time. (3) Discusses the limitations of current numerical fracture propagation models and suggests that the degree of curvature that develops between two mechanically interacting fractures is limited by numerous factors including stress state, material anisotropy, fracture surface roughness, and inelasticity. (4) Develops a computationally efficient numerical model of fracture network formation incorporating current understanding of the physics of rock fracture. The model accurately predicts the evolution of experimentally generated fracture sets. Given an initial flaw geometry, only the velocity exponent controls the growth of the fracture set. The velocity exponent relates fracture propagation velocity to stress concentration at the fracture tip. This parameter controls the extent to which fracture growth is concentrated within zones or clusters. Fracture clustering is less sensitive to the initial flaw density. And (5) explores the physical parameters controlling fluid flow through the simulated networks. Flow characteristics of pseudo-symmetric physically-based networks are shown to be qualitatively similar to the flow characteristics of bond percolation networks, but more sensitive to the scale of measurement. Average network flow characteristics are independent of initial flaw density over the range considered and independent of velocity exponents less than or equal to 1. Velocity exponents greater than 1 result in extensive connected pathways at significantly lower fracture densities. Appropriate values of the velocity exponent for natural fracture networks are uncertain.