Acoustic Emission Geomechanics of Hydraulic Fracturing In the Laboratory

Successful optimization requires the synergy of numerical models, field observations and laboratory experiments. In the present study, we investigated the effect of triaxial stress ratio on hydraulic fracturing in the laboratory and compared the results to a range of discrete element numerical model outputs.

A triaxial deformation cell (GIC:A ErgoTech Ltd.) equipped to handle a cylindrical rock specimen (length 125 mm, diameter 50 mm) was used to conduct the reported experiments. Acoustic Emission (AE) was monitored by 19 piezoelectric sensors. Amplified waveforms were continuously recorded using the IMaGE Richter acquisition system. After the experiment, continuous AE waveforms were harvested to extract discrete events if the trigger criterion was met. Experiments were conducted for four different stress ratios with the fluid injection rate held constant at 1 mL/min and four different injection rates with the stress ratio held constant at σ1 = 20 MPa and σ3 = 10 MPa. For the four stress states, σ3 was held at 10 MPa and σ1 = 12, 15, 20, and 30 MPa where σ3 was applied parallel to the borehole and σ1 was applied perpendicular to the borehole. For the injection rate experiments, fluid was injected at constant rates of 0.25, 0.5, 1, and 4 mL/min. Samples were loaded up to the desired stress state, first hydrostatically and then differentially. Once the desired stress state was reached, the system was left to stabilize before fluid was injected at a constant rate until the sample was failed and the borehole pressure dropped back to 10 MPa (σ3). Rock specimens were cored from blocks of Westerly granite because of its homogeneous and isotropic structure.

The following figure shows the results of AE locations and fracture surfaces obtained from micro-CT scanning of the rock samples:

(Courtesy of the University of Toronto; Goodfellow et al., 2013)

 

The fracture initiation point varies from experiment to experiment. For experiment SR1, the fracture initiates at roughly the midpoint of the sample, whereas for experiments SR2, SR3, and SR4, the fracture initiates from the bottom edge of the borehole. This results has not been thoroughly investigated at this point, however, we speculate that it is a results of a stress concentration at the bottom of the borehole, which is affected by the changing stress ratio. Another observation is that for low stress ratios, AE source locations are more dispersed, which suggests secondary microfracturing, but as the stress ratio increases, AE locations are less dispersed and appear to be concentrated along one main fracture. The results of sample X-rays are in good agreement with AE source location results and confirm the initial observations. Qualitatively, we can say that the induced fractures show more highly varied surface topography and more branching for low stress ratios.

 

To gain further insight of the geomechanics at play, the laboratory experiments were simulated with XSiteXSite is a newly developed software based on the Lattice method (simplification of the bonded-particle model in which  the finite-sized particles are replaced by point masses and the contacts between particles are replaced by springs that may break). The primary advantages of this method are: 1) full coupling between mechanical and fluid flow calculations, 2) it is based on mechanics, 3) it represents all the dominant mechanisms which occur in fracturing, 4) the inputs can be obtained from standard testing methods, and 5) it is very computationally efficient. XSite is unique in that it can model fully 3D fracture propagation in both homogeneous and inhomogeneous rock masses, with no pre-defined fracture shape or trajectory, and can simulate the interactions between hydraulic fractures and natural fractures and between multiple hydraulic fractures.  It solves rigorously for fracture propagation, including reorientation, stress shadowing, and crossing/non-crossing based on mechanics, not correlations or approximations.

Results of the XSite simulation corresponding to the above laboratory experiments are show below:

The model shows that the micro-cracks are less dispersed for higher stress contrasts.  In other words, as the stress ratio (σ1/ σ3) increases, the micro-cracks tend to concentrate along one main fracture, corresponding well with the lab results.

The acoustic emissions attributes can also be quantitatively compared with the geomechanical model. For example, the largest moment magnitude computed from the lab AE was approximately -7 from which a seismic energy can be estimated. Assuming a seismic efficiency of 0.05% matches the largest events from the fracturing experiments. Using values of the smallest events detected in the different lab tests, the equivalent number of modeled AE can be estimated.  The model indicates more events, which is expected because some events tend to be missed in reality and the lab data would not be complete at the minimum magnitude. Nevertheless, the comparison shows that the relative proportion of events is very similar for all the cases, i.e., the number of recorded events is consistently approximately 15% of the number of modeled events.