Realistic Representation of Grain Shapes in CFD--DEM Simulations of Sediment Transport: A Bonded-Sphere Approach

Abstract

Development of algorithms and growth of computational resources in the past decades have enabled simulations of sediment transport processes with unprecedented fidelities. The Computational Fluid Dynamics--Discrete Element Method (CFD--DEM) is one of the high-fidelity approaches, where the motions of and collisions among the sediment grains as well as their interactions with surrounding fluids are resolved. In most DEM solvers the particles are modeled as soft spheres due to computational efficiency and implementation complexity considerations, although natural sediments are usually mixture of non-spherical particles. Previous attempts to extend sphere-based DEM to treat irregular particles neglected fluid-induced torques on particles, and the method lacked flexibility to handle sediments with an arbitrary mixture of particle shapes. In this contribution we proposed a simple, efficient approach to represent common sediment grain shapes with bonded spheres, where the fluid forces are computed and applied on each sphere. The proposed approach overcomes the aforementioned limitations of existing methods and has improved efficiency and flexibility over existing approaches. We use numerical simulations to demonstrate the merits and capability of the proposed method in predicting the falling characteristics, terminal velocity, threshold of incipient motion, and transport rate of natural sediments. The simulations show that the proposed method is a promising approach for faithful representation of natural sediment, which leads to accurate simulations of their transport dynamics. The proposed method also opens the possibility for first-principle-based simulations of the flocculation and sedimentation dynamics of cohesive sediments. Elucidation of these physical mechanisms can provide much needed improvement on the prediction capability and physical understanding of muddy coast dynamics.