Machine learning bridges microslips and slip avalanches of sheared granular gouges

Abstract

Understanding the origin of stress drop of fault gouges may offer deeper insights into many geophysical processes such as earthquakes. Microslips of sheared granular gouges were found to be precursors of large slip events, but the documented relation between microslips and macroscopic stress drops remains largely qualitative. This study aims to quantitatively connect microslips to macroscopic stress fluctuations, including both stress recharges and stress drops. We examine the stick-slip behavior of a slowly sheared granular system using discrete element method simulations. The microslips are found to demonstrate significantly different statistical and spatial characteristics between the stick and slip stages. We further investigate the correlation between the macroscopic stress fluctuations and the features extracted from microslips based on a machine learning (ML) approach. The data-driven model that incorporates the information of the spatial distribution of microslips can robustly predict the magnitude of stress fluctuation. A further feature importance analysis confirms that the spatial patterns of microslips manifest key information governing the macroscopic stress fluctuations. The generalization of ML across granular gouges with different characteristics indicates the proposed model can be applicable to a broad range of granular materials. Our findings in this study may shed lights on the mechanisms governing earthquake nucleation, microslips, friction fluctuations, and their connection during the stick-slip dynamics of earthquake cycles.