Page 55 - Machine Learning for Subsurface Characterization

Characterization of fracture-induced geomechanical alterations Chapter 2 41

compressional and shear stiffness. Upon crossing the fracture the wave
experiences attenuation and a time delay, both of which are controlled by the
stiffness of fracture and sonic impedance of the intact material. Therefore, a
decrease in amplitude along with a delay of the arrival time can be used to
identify fractured zones in fractured material.

2 Objective of this study

The study aims to noninvasively map the fracture-induced geomechanical
alteration in a hydraulically fractured geomaterial. To that end, unsupervised
clustering methods will be applied on the laboratory measurements of
ultrasonic shear waveforms transmitted through the fractured geomaterial. The
proposed workflow (Fig. 2.2) can be adapted for improved fracture
characterization using sonic-logging and seismic waveform data. Several
researchers have used machine learning to analyze seismic events, like
earthquakes, volcano activity, and rock stability [7, 8, 9]; however, no known
reference exists that applies clustering methods to noninvasively visualize the
fractured zones and geomechanical alterations in geomaterials.

3 Laboratory setup and measurements
3
The experiments were performed at the IC laboratory (http://ic3db.ou.edu/
home/) at The University of Oklahoma. Present study analyzes the data

FIG. 2.2 Workflow for the noninvasive mapping/visualization of the fracture-induced
geomechanical alterations in a hydraulically fractured geomaterial. Clustering results are
converted into geomechanical alteration index using displacement discontinuity theory.