High Performance Computing and the Discrete Element Model

Opportunity and Challenge

Gao-Feng Zhao

• Publisher: ISTE Press - Elsevier
• Year: 2015
• Print ISBN: 9781785480317
• eBook ISBN: 9780081008096
• Topic: Technology & Engineering
• Subtopic: Materials Science
• Index: Technology & Engineering

This book addresses the high performance computing of the Discrete Element Model (DEM). It is a comprehensive presentation of parallel implementation of the DEM on three popular parallel computing platforms; the multi-core PC, the GPU computer, and the cluster supercomputer. Featuring accompanying MatLab source this book helps you implement the DEM model for use with high performing technology, for particular implementation of the dynamic failure of solids, granular flow and stress wave propagation through solids.- Features both Pre-processor, Solver, and Post-processor for the DEM- Covers the parallel implementation of DEM on the cluster, multi-core PC, GPU PC- Full of examples of dynamic fracturing, granular flow and stress wave propagation using high performance DEM- Source codes and data files available for hands-on practice

1

Serial Implementation

Abstract

In this chapter, implementation details of a serial DEM code, DICE2D, are described. The target is to provide a programming environment for further parallel implementation of the DEM. Full aspects of DEM coding are covered, such as system design, data structure design, flowchart design, algorithm design and implementation. A number of benchmark examples are designed for verification and debugging purposes.

Keywords

Constitutive model

Contact detection

Damping

Energy calculation

Particle-to-particle constitutive model

Particle-to-particle contact

Serial implementation

Simplified grid cell method

System design

Time step

Wall-to-particle constitutive model

Wall-to-particle contact

In this chapter, implementation details of a serial DEM code, DICE2D, are described. The target is to provide a programming environment for further parallel implementation of the DEM. Full aspects of DEM coding are covered, such as system design, data structure design, flowchart design, algorithm design and implementation. A number of benchmark examples are designed for verification and debugging purposes.

1.1 System design

Unlike the modern commercial DEM codes, which use the object-oriented design concept, DICE2D adopts the process-oriented design concept that is simple to understand and suitable for algorithm research. To release the workload of graphical user interface (GUI) design, MATLAB® was selected as the programming environment. It provides a number of built-in functions to display the particles and computational results. Moreover, the high interactive feature of MATLAB also makes it a good choice for algorithm development and the study of the DEM. The DEM aims to simulate the nature process; therefore, the code is named DICE2D. It is inspired from Einstein’s comment on quantum mechanics: “…God does not throw dice”. To run the code, the first step is to select a folder named DICE2D as the current work directory (Figure 1.1). Then, type “D2D(iEx)” into the command window to run the corresponding example.

Figure 1.1 DICE2D in MATLAB

Figure 1.2 shows the work flow of DICE2D. First, the user needs to input the example ID. Then, the preprocessor will prepare the corresponding model data. The numerical model will be further processed by the DEM solver to obtain the simulation results, which will be finally processed by the postprocessor. Unlike most commercial software, in DICE2D, users must build up their own pre- and post...

Table of contents

• Cover image
• Title page
• Table of Contents
• Dedication
• Copyright
• Foreword
• Preface
• Introduction
• 1: Serial Implementation
• 2: Multi-core Implementation
• 3: GPU Implementation
• 4: DICE2D and Cluster
• Bibliography
• Index