← Back to Sessions

Using Deep Learning for High-Resolution Fault Analysis and Stress Characterization at the Forge Site, Utah

Description: 

The precise determination of fault location and geometry assumes major importance in the context of induced seismicity management during reservoir stimulation. By increasing the size of new datasets, we need new techniques to get the most information from these datasets. In this study, we used Machine Learning and Neural Networks to prepare data for focal mechanism determination and stress field evaluation of the FORGE site in Utah. The dataset used for this purpose is derived from dual-borehole seismic arrays located at the FORGE site. Accurately determining focal mechanisms for microseismic events is a challenging task, primarily due to the inherent limitations in geometric coverage. To eliminate this challenge, we used various seismic attributes such as P-wave polarities and P/S amplitude ratio, with the aim of supplying a robust estimation of the focal-plane mechanisms.

Our study focuses on the analysis of 1500 microseismic events recorded during phase 3 stimulation in 2022 at the FORGE site. We used PhaseNet to pick P and S arrival then we developed A Convolutional Neural Network (CNN) by using Southern California Earthquake Data Center (SCEDC) dataset to predict the first motion polarity. By using this approach, we prepared a dataset consisting of 738 events for focal mechanisms inversion. We obtained 681 A and B quality focal mechanisms solutions by using HASH method. The high-resolution focal mechanisms map shows different patterns at different parts of the site area with more strike-slip fault at shallower depth, while mixture of reverse and normal faulting at deeper depth. By utilizing MSATSI package, we estimated the stress field principal axes orientation and relative magnitude at each section. The stress result shows different stress regimes at different sections, also the stress map showed a clear transition in principal axes orientation with depth at the FORGE site area.

Session: Seismic Monitoring, Modelling and Management Needed for Geothermal Energy and Geologic Carbon Storage [Poster Session]

Type: Poster

Date: 5/2/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Ahmad

Student Presenter: Yes

Invited Presentation: 

Authors