MICROSEISMIC MONITORING: PHYSICAL MODELING AND SOURCE CHARACTERIZATION

This research investigates some fundamental aspects of microseismic monitoring: location and source mechanism. We developed a ray-tracing and diffraction-stack procedure to determine source locations. The location algorithm uses a grid-search technique to find source coordinates. For every possible point source, three attributes (traveltime residual, stacked energy and energy/residual ratio) in a grid area are calculated. Then the location can be found by either choosing the point that yields a minimum traveltime residual or maximum stacked energy or maximum energy/traveltime ratio. Further, focal mechanisms and radiation patterns of simulated microseismic events are examined using Focmec (Focal Mechanism Determination), an open-source program. The location algorithm is developed in a MATLAB environment and tested on physical modeling data from the Allied Geophysical Laboratories (AGL) at the University of Houston.

Three different physical modeling experiments have been conducted using ultrasonic source and 3-component receivers. For the first experiment, a single layer Plexiglas model was used; the second experimental model was built by assembling Plexiglas and aluminum blocks. In the third experiment, a real sandstone rock (57.5 x 43.8 x 17.5 cm) was employed. To determine which method (P versus S waves and travel time versus amplitude) and acquisition design (surface or borehole receivers) is most accurate, we have undertaken variety of tests. Locating events using S-waves is as accurate as with P-waves; however, combining both P and S-waves are the most accurate approach among all experiments. Furthermore, location certainty increases when downhole receivers are included for both P and S-waves.

To increase the speed of the algorithm, CPU and GPU computing was implemented. Locating a single microseismic event with 7 different methods takes 11.4 seconds on single core CPU, whereas, this number is decreased to 4.2 seconds using multi-core CPU computing. Further, implementing GPU computing further decreases the total elapsed time to only 1.9 seconds. There is more than an 80 percent increase in terms of computation time compared to single core CPU.