Micro seismic data and coring studies suggest that hydraulic fractures interact heavily with natural fractures creating complex fracture networks in naturally fractured reservoirs such as the Barnett shale, the Eagle Ford shale, and the Marcellus shale. However, since direct observations of subsurface hydraulic fracture geometries are incomplete or nonexistent, we look to properly scaled experimental research and computer modeling based on realistic assumptions to help us understand fracture intersection geometries. Most experimental analysis of this problem has focused on natural fractures with frictional interfaces. However, core observations from the Barnett and other shale plays suggest that natural fractures are largely cemented. To examine hydraulic fracture interactions with cemented natural fractures, we performed 9 hydraulic fracturing experiments in gypsum cement blocks that contained embedded planar glass, sandstone, and plaster discontinuities which acted as proxies for cemented natural fractures.
There were three main fracture intersection geometries observed in our experimental program. 1) A hydraulic fracture is diverted into a different propagation path(s) by a natural fracture. 2) A taller hydraulic fracture bypasses a shorter natural fracture by propagating around it via height growth while also separating the weakly bonded interface between the natural fracture and the host rock. 3) A hydraulic fracture bypasses a natural fracture and also diverts down it to form separate fractures. The three main factors that seemed to have the strongest influence on fracture intersection geometry were the angle of intersection, the ratio of hydraulic fracture height to natural fracture height, and the differential stress.
Simply put, the most significant finding of this research is that fracture intersection geometries are complex. Our results show that bypass, separation of weakly bonded interfaces, diversion, and mixed mode propagation are likely in hydraulic fracture intersections with cemented natural fractures. The impact of this finding is that we need fully 3D computer models capable of accounting for bypass and mixed mode I-III fracture propagation in order to realistically simulate subsurface hydraulic fracture geometries.
Field results in Steam Assisted Gravity Drainage (SAGD) heavy oil operations suggest formation permeability changes during production operations. We investigate this process using samples constructed from loose Athabasca sand. Our results indicate that permeability changes (absolute and relative) and endpoint oil and water saturation variation are a function of loading boundary conditions. Triaxial loading paths (increasing mean stress) show increased permeability with sample dilatancy upon failure at low confining stress (100 to 400 psi) that diminishes when run at higher confining stress (800 psi). Residual oil saturation increases during compaction and decreases during dilation while initial water saturation decreases during compaction and increases during dilation. Radial extension tests (decreasing mean stress) result in larger increases in permeability upon failure and dilatancy, and lower residual oil and higher initial water saturations. Finally, failure induced by unloading through pore pressure increase, run under fixed displacement, fixed stress, or mixed axial versus confining boundary conditions, show minimal permeability changes until the effective stresses approach zero. However, the increase in initial water saturation and decrease in residual oil saturation are as significant as or greater than what was seen in the triaxial compression and radial extension tests. Overall, results show that lab experiments support increased permeability due to steam injection operations in heavy oil, and more importantly, the observed reduction in residual oil saturation implies SAGD induced deformation should improve recovery factors.
Shear failure plays an important role in fracture propagation in poorly consolidated formations. In past work, the effect of shear failure on the permeability and porosity of sands has been inconclusive. In this paper, a three-dimensional discrete element model is used to estimate changes in porosity and permeability due to mechanical deformation of sand-packs. Using pore network fluid flow simulations, the effect of shear failure and stress anisotropy on the permeability anisotropy and dilation of the granular specimens is analyzed.
Mechanical deformation data from experiments conducted on cylindrical sand packs is used to calibrate the simulations. Using these calibrated models, anisotropic stress regimes are modeled to mechanically deform the simulated samples. Post-processing tools have been developed to observe the preferential orientation of failure planes. Pore network fluid simulations describe the orientation of the permeability tensor. The principal permeability values of an anisotropically stressed sample are described in 3-D.
Deformation of the samples induces shear failure planes, which are preferentially oriented along the maximum horizontal stress direction. Deformed samples with the same minimum horizontal stress (50 psi) but increasing maximum horizontal stresses (50 psi to 300 psi) show an increase in the permeability in the maximum horizontal stress direction by 28-38%. An increase in horizontal stress anisotropy is shown to result in an increase in permeability in all directions due to dilation and failure. The effect of confining pressure, grain size distribution and sorting on failure, permeability anisotropy and dilation is also studied.
The resultant permeability anisotropy induced by shear failure and dilation is of vital significance for simulating fracturing and flow processes in soft rock formations.
Fracture propagation in unconsolidated and poorly consolidated sands has been shown to be a result of shear failure accompanied by greater leak-off and plastic deformation (Khodaverdian & McElfresh 2000). Mechanical deformation of formations leads to development of shear failure zones (Agarwal and Sharma 2011a). Localization of shear failure zones in the sand samples can lead to dilation of samples (Larsen & Fjaer 1999). This shear band formation has been observed in both experiments and simulations (Bruno et al. 1991; Bardet & Proubet 1992; Horabik et al. 2000; Zhou & Chi 2003; Hu 2004; Zhai and Sharma 2005; Alonso-Marroquin et al. 2007; Riedel et al. 2008; Castelli et al. 2009; Huang et al. 2010; Schall & Hecke 2010; Agarwal and Sharma 2011b).
Mechanical deformation can also lead to variation in the fluid flow response through the samples. Various experiments have been conducted by numerous researchers to describe the influence of stress anisotropy, stress path, compaction and mechanical deformation of rock on the permeability evolution of the rock samples (Hyman et al. 1991; Bruno et al. 1991; Zhu et al. 1997; Bai et al. 2002; Wong 2003; Dautriat et al. 2007; Clavaud et al. 2008; Dautriat et al. 2009; Khan 2009; Olson et al. 2009). Numerous computer simulations have also been performed to establish the effect of mechanical deformation on permeability evolution in rocks (Bruno et al. 1996; Karacan et al. 2001; Li & Holt 2002; Wong 2003; Holt et al. 2005; Yaich 2008).
Bruno (1994) reported a comprehensive account of stress-induced permeability anisotropy and damage in sedimentary rock and the underlying micromechanics.
As compressive stress is first applied to a sample, micro-fractures occur at grain boundaries and are predominantly oriented parallel to the primary loading axis. The micro-cracks are initially distributed uniformly throughout the sample and later coalesce and concentrate in narrow deformation bands.
Successful shale gas and tight oil applications require multiple hydraulic fracture treatments in a given horizontal well. Key design considerations are the spacing between injection zones, the size of the treatments, and the injection parameters (primarily rate and viscosity). Other important parameters are whether to inject into multiple zones simultaneously or sequentially, and whether to coordinate fracturing between multiple adjacent horizontal wells. We investigate various
injection scenarios using two different "pseudo-3d?? boundary element, displacement discontinuity method based hydraulic fracture simulators. A single fracture model couples the injection of non-Newtonian fluid, Carter leak-off behavior, arbitrary non-planar propagation, and height growth in a 3-layered media. The second model is designed to be more computational efficient and stable by simplifying the fracture fluid flow calculation. Fracture pressure variation through time is
approximated by summing the inverse of width cubed over the created fracture network, and wellbore pressure is equated to the predicted summative pressure drop in the system. Results indicate that sequential fracturing geometries are highly dependent on injection zone spacing relative to fracture height - when fractures are closely spaced relative to stress shadow size, fracture path diversion and intersection as well as net pressure effects are likely. This strong mechanical interaction scenario promotes fracture complexity (which can be an advantage in some formations), but reduces the penetration distance of fracture wings away from the wellbore. Simultaneous fracturing of multiple stages is also strongly influenced by the spacing to height ratio - the closer the spacing the more difficult it is to get substantial growth from all completed zones.
Field experience in shale gas and some tight gas sandstones suggests significant interaction between hydraulic fractures and natural fractures. Experimental analysis of this problem has mostly focused on frictional interfaces with regard to the natural fractures. We have performed tests to examine the effect of cemented natural fractures on hydraulic fracture propagation. The motivation for this type of work is that core observations from the Barnett and some other shale gas plays suggest that natural fractures are largely cemented (or healed) and trend obliquely or orthogonally to the present day hydraulic fracture direction. We embedded planar glass discontinuities into a cast hydrostone block as proxies for cemented natural fractures. Consistent with theoretical predictions, our results show that oblique embedded fractures are more likely to divert a fluid-driven hydraulic fracture than those occurring orthogonal to the induced fracture path. Hydraulic fracture - natural fracture interaction took three forms - 1) the hydraulic fracture bypassing the natural fracture by propagating around it (via height growth, not curving), 2) the hydraulic fracture arresting into the natural fracture and then diverting along it and sometimes kinking off the end of it, and 3) a combination of bypass and diversion. We also saw some leakage of fracture fluid along the interfaces of multi-layer blocks, even though such interfaces were perpendicular to the maximum compressive stress, suggesting an analogue for laminated and shaly sedimentary sequences.