Impact of pseudo-genetic approach on discret fracture network geometry and connectivity.

@inproceedings{Bonneau2GM2013,
 abstract = { This paper presents a stochastic sequential approach to simulate Discrete Fracture Networks (DFNs). Our approach drives DFNs simulations in order to reproduce both the network statistics (such as fracture dip, strike and density) and the hierarchical organization of fracture systems. Our method reproduces a continuous transition between a “dilute” and a “dense” fracturing regime. Natural fractures appear and grow gradually when the stress intensity reaches the fracture toughness. Older fractures, because of their influences on both stress field and rock coherence, impact later fracture growth and initiation. In early stage of our process a few 3D discrete fractures are simulated. Later fractures are implanted preferentially at older fracture tips. This sequential seeding of 3D objects bring a hierarchical organisation in the simulated DFN. In the end, a radial growth process is run to calibrate the fracture geometry (size, shape) and enhance linking structure. Each fracture is associated to two ellipsoids that bounds the impact of fracture shadow zone and constraint accumulation zone. It will be use for both the sequential fracture nucleation and the radial growth. (1) The probability for later fractures to be initiated is enhanced (respectively decreased) by constraint accumulation zones (respectively by shadow zones). (2) Large fractures inhibit the growth of surrounding small ones. (3) Fracture growth is re-oriented according to constraint accumulation zones overcoming and (4) stops when the growing fracture intersects a larger fracture. We use simple rules in order to reproduce both the fracture radial growth and the growth by linkage. We argue that the classical power law length distribution law naturally emerge from the fracturing process. We show on 3D examples that the high number of linking structures created allows DFNs to reach the percolation threshold with less dense network than those obtained by classical purely stochastic approaches. The average density of fractures at percolation is found to vary by a factor of 2 between classical and pseudo genetic simulations. },
 author = { Bonneau, Francois AND Caumon, Guillaume AND Renard, Philippe AND Sausse, Judith },
 booktitle = { Proc. 33rd Gocad Meeting, Nancy },
 title = { Impact of pseudo-genetic approach on discret fracture network geometry and connectivity. },
 year = { 2013 }
}