Good Grains Gone Bad: How Grain to Grain Interactions Complicate the Onset of Motion

Abstract:

Predictions of the onset of sediment motion are integral components of bed stability and bedload flux estimates. Mechanistic equations for initial motion employ a balance between driving and resisting forces. Driving forces are modeled as functions of the magnitude and duration of turbulence events whereas resisting forces are simply approximated by the grain weight and a static friction angle. Such resistance approximations do not include the effects of grain packing and dynamic interactions with surrounding sediment. To better understand and quantify grain resistance, we used a Discrete Element Method (DEM) model for a single test grain surrounded by a bed of smaller grains. We applied a constant external force on the test grain in each run and progressively increased the force between runs until the test grain moved out of its resting pocket. The DEM model calculated the test grain velocity, position and net force (sum of applied external force and forces from other grains) at time steps of 1×10^-7 s. Despite applying a constant external force, the net force on the test grain fluctuated by three to six orders of magnitude, depending on the run. These fluctuations were driven by the creation and destruction of force chains, and the rearrangement of the positions of surrounding bed sediment. Stick-slip behavior, which has been observed in shear tests of granular material, occurred during test-grain motion. The frequency of stick-slip behavior generally declined with higher applied external forces. Therefore, the onset of grain motion was not continuous, as is often assumed even in the presence of fluctuating applied fluid forces. The duration and magnitude of turbulence fluctuations have received considerable attention but our results suggest that grain resistance oscillations are also important. Whether turbulence and resistance fluctuations are synchronous will likely dictate if grain movement occurs, and we are currently conducting model runs to better understand such feedback mechanisms.