Peiyuan Liu, Timothy Brown, William D. Fullmer, Thomas Hauser, Christine Hrenya, Ray Grout and Hariswaran Sitaraman, "A Comprehensive Benchmark Suite for Simulation of Particle Laden Flows Using the Discrete Element Method with Performance Profiles from the Multiphase Flow with Interface eXchanges (MFiX) Code", Technical Report NREL/TP-2C00-65637, January 2016
Abstract: This technical report describes activities performed under a project named “MFIX DEM Enhancement for Industry-Relevant Flows” as part of the National Energy Technology Laboratory’s Crosscutting Technology Research Program’s Transitional Technology Development to Enable Highly Efficient Power Systems with Carbon Management initiative; the report contents also served as a milestone document for Task 2 that project. Five benchmark problems are developed, discussed, and simulated with the computational fluid dynamics and discrete element method code Multiphase Flow with Interface eXchanges (MFiX). The benchmark problems span dilute and dense regimes and consider statistically homogeneous and inhomogeneous (both clusters and bubbles) particle concentrations and a range of particle and fluid dynamic computational loads. Several variations of the benchmark problems are also discussed to extend the computational phase space to cover granular (particles only), bidisperse, and heat transfer cases. A weak scaling analysis is performed for each benchmark problem, and in most cases the scalability of the code appears reasonable up to ~103 cores. Profiling the benchmark problems indicates that the most substantial computational time is being spent on particle-particle force calculations, drag force calculations, and interpolating between discrete particle and continuum fields. Hardware performance analysis was also carried out showing significant Level 2 cache miss ratios and a rather low degree of vectorization. These results provide a baseline to which we can compare future developments of the code as well as a preliminary indicator of where to best focus optimization efforts.