| Abstract |
Thermal cracking may play an important role in fracture network (FN) evolution in geothermal reservoirs. During hydraulic stimulation, boreholes and reservoirs can undergo varying degrees of thermal shock and thermal stress. Thermal cracking, in such situations, may couple strongly with hydraulic and background tectonic stresses to influence seismicity rates and fracture network evolution. In conditions where grain-scale ductility is significant (a.k.a. the brittle-ductile transition), crack tip stresses may be partially relaxed, causing crack interactions to occur at shorter length scales than in the purely brittle field. Thermal stress and cracking may play an important role in creating dense networks of grain-scale fractures. Constitutive models that capture strong coupling among microstructural evolution and thermal, mechanical and hydraulic properties at the grain scale on up are needed to describe high temperature geothermal reservoir processes. We perform tri-axial laboratory experiments at 10 MPa confining pressure, on thermal cracking in Westerly granite. We control stress buildup to varying degrees by ramping up temperature in stages under different mechanical boundary conditions. In some experiments, we lock the piston so that differential stress builds as the sample thermally expands, sometimes to failure levels. In other experiments, we release the stress by retracting the piston between each thermal loading stage. We record continuous near-field acoustic data using novel high-temperature piezoelectric sensors stable up to about 500ËšC. Using standard methods of recording acoustic emissions (AEs) by setting triggering levels just above the noise level, we record on the order of several hundred AEs in an experiment. Then, we analyze the continuous data with a signal processing method consisting of short-term/long-term average detections to build a catalog of template waveforms followed by matched filtering. With this approach, we have built a catalog that has about 1000 times more events, mostly below the noise level. We then analyze the temporal clustering statistics of this catalog; we find that the large number of small events are close to random, while small numbers of larger events have strong temporal clustering and cascade-like behavior. Finally, we perform unsupervised feature extraction (using SpecUFEx and hierarchical clustering) to detect/discover temporal patterns in the spectral content of the signals. These show a subtle evolution in the collective sound of fracturing as temperature and stress both increase. While these are initial results, we have found that the combination of near field sensors and high-resolution detection methods show promise for offering new insight and opening new questions into microfracturing processes and network evolution. |