Initiation of Internal Erosion in Earth Dams: A Particle-Scale Computational Approach

Australia is known as the driest populated continent in the world, but with periods of high rainfall and flooding followed by long droughts. Earth dams are the number one supplier of water for irrigation, hydropower, and clean water, as well as the major infrastructures for flood control, amongst other purposes. Internal erosion accounts for about 50 percent of dam failures in Australia and across the world. Such failures could be catastrophic, as they often occur without noticeable precursors, posing significant risks to public safety and downstream infrastructures. In this study, we incorporate the Discrete Element Method (DEM) coupled with Computational Fluid Dynamics (CFD) to simulate soil samples under internal erosion as representative elements for dams. The outputs of simulations are evaluated using a statistical machine learning (ML) method to better assess the triggers of internal erosion based on spatiotemporal patterns in particle-scale and sample-scale parameters, such as particle velocity, particle-particle contact force, and fluid-particle coupling force, as well as kinetic and total energy during the initiation of erosion process. Understanding these patterns and correlations at the particle scale may assist in (macro-scale) engineering monitoring and mitigation strategies.