Machine learning bridges microslips and slip avalanches of sheared granular gouges

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