Machine learning reveals the spatial and temporal dynamics simulated slip events in a sheared granular fault gouge
Abstract
Granular materials resulting from mechanical processes at the core of mature faults are often observed in nature. The gouge material is known to have significant influences on slip events, as gleaned from laboratory experiments, numerical simulations and Earth observations. However, it's hardly feasible to extract grain-scale information from laboratory experiments. Here we use 3-D discrete element simulations (DEM) to model the stick-slip dynamics in a sheared granular fault gouge and apply a machine learning-based technique developed to process and analyze the large amounts of DEM data. This work aims to reveal through analysis of simulation data the origin of acoustic signals that have been shown in the laboratory to be a fingerprint of frictional state and time remaining before a slip event. Our machine learning analysis shows that dynamic behavior of single flagged particles can predict the frictional behavior of the sheared granular gouge. This analysis reveals the location of specific, individual particles and regions which are highly informative in predicting the frictional state of the system.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.S11E0404R
- Keywords:
-
- 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICSDE: 1910 Data assimilation;
- integration and fusion;
- INFORMATICSDE: 1942 Machine learning;
- INFORMATICSDE: 7223 Earthquake interaction;
- forecasting;
- and prediction;
- SEISMOLOGY