Parallel Discrete Element Simulation and Its Application to the Study of Solid-Liquid Flow Behavior

In recent years the widespread use of the Discrete Element Method (DEM) in engineering has generated increasing research interest across a variety of fields. As a result of rapid and continuing developments in computer science, DEM is now being applied to the modeling of physical phenomena and engineering problems of ever-increasing complexity. Solid-liquid flow behavior simulation is one ubiquitous application. However, dynamic behavior in such systems is difficult to predict due to complex interactions at the solid-liquid interface, which invoke considerable computational overhead. Since the method is constrained by contemporary processing power, an efficient Discrete Element Simulation (DES) system is needed for solving large-scale solid-liquid interaction problems. This paper undertakes to develop and apply such a system in the simulation of both Self-Compacting Concrete (SCC) and wet granular flow behavior. Three strategies are implemented to optimize existing DES procedures for computational speed; the result is an in-housed parallel DES system, KNIGHT&ANNE/IRIS 2.0 developed specifically for accelerated performance in solid-liquid flow simulation. Several numerical benchmarks are applied to both shared and distributed-memory platforms, indicating substantial performance improvements. A two-phase model is then developed for simulating SCC flow behavior. Various rheological experiments - the V-funnel flow test and the L-shaped box test - are modeled from packing to flowing, and DES handling of the simulation is shown to provide an adequate representation of empirical data. This comparison is also used to propose corresponding DES parameter values and ranges for simulation of SCC and mortar flow. A liquid-modified interaction model is proposed for the simulation of wet granular systems, and tested on both wet and dry particulate flows down an inclined channel. The level of congruence found between simulated and empirical data sets confirms the physical model to be reasonably accurate.