Interpreting force chain evolution giving rise to granular failure using network science

When a granular material fails under shear, it can be understood as a loss of stability of the force chain network that supports it. The direct observation of grain kinematics and force chain networks allows us to examine how pre-event observations could forecast the location or timing of a future slip event. However, mining consistent patterns from the complex force networks can be challenging. Recently, network science techniques have been used to characterize networks and understand interactions in complex networks. We perform experiments on sheared quasi-2D granular packing undergoing stick-slip failure. The granular packing is made of photoelastic disks, allowing us to measure interparticle forces using photoelastic force measurements and quantify the evolution of the force network during failure. We create a network corresponding to the granular material, with nodes representing the particles and edges representing the contacts between particles, weighted by force magnitude between corresponding pairs of particles. Force networks at consecutive time steps form a multi-layer network. We utilize community detection techniques which identify strongly-correlated clusters of nodes to find patterns of causality. Temporal and spatial community structure are examined to identify patterns as the loading on granular packing accumulates to surpass a failure criterion. Our preliminary results show that there are promising signals in community structures highlighting the area of high risk and time of failure.