Distributed computer-aided engineering : for analysis, design, and visualization

Networking of personal computers and workstations is becoming commonplace in academic and industrial environments. A cluster of workstations provides engineers with a familiar, cost-effective environment for high performance computing. However, workstations often have no dedicated link and communicate slowly on a local area network (LAN), such as the Ethernet. Thus, to effectively harness the parallel processing or distributed computing capabilities of workstations, new algorithms need to be developed with a higher computation-to-communication ratio. Distributed Computer-Aided Engineering presents distributed algorithms for three fundamental areas: finite element analysis, design optimization, and visualization - providing a new direction in high performance structural engineering computing. Cover......Page 1 Series Page......Page 2 Title Page......Page 4 Copyright Page......Page 5 Preface......Page 6 About the Authors......Page 8 Acknowledgment......Page 10 Dedication......Page 12 Table of Contents......Page 14 Chapter 1: Introduction......Page 18 2.1 Introduction......Page 22 2.2 FE Analysis and PCG Solution Method......Page 25 2.2.1 Conjugate Gradient Methods......Page 26 2.2.2 Element-by-Element Approach......Page 28 2.3 Data Distribution for Distributed FE Analysis......Page 31 2.3.1 Basic Definitions......Page 32 2.3.2 Data Distribution......Page 36 2.3.3 Data Movement......Page 45 2.4 Distributed PCG Algorithm......Page 47 2.4.1 Concurrent Vector Operations......Page 52 2.4.2 Concurrent Matrix-Vector Product......Page 54 2.4.3 Redundant Computation to Reduce Communications......Page 58 2.4.4 Coarsening the Task-Granularity......Page 59 2.5 Final Comments......Page 60 3.1 Introduction......Page 62 3.2 Object-Oriented Data Structures......Page 64 3.2.1 Modeling the Hierarchy of Elements......Page 65 3.2.2 Knowledge Base of Connectivities......Page 68 3.2.3 Creating a List of Adjacent Nodes of a Node......Page 69 3.2.4 List of Adjacent Elements of an Element......Page 72 3.2.5 Domain Decomposition......Page 73 3.3 Graphic Visualization Using Hidden Line Removal......Page 74 3.3.1 Hidden Line Removal......Page 76 3.4 Examples and Results......Page 77 3.4.1 Example 1: Plane Stress Problem......Page 80 3.4.3 Example 3: Taj Mahal Structure......Page 89 3.5 Comments on Performance......Page 96 4.1 Introduction......Page 104 4.2 GA-Based Structural Optimization......Page 106 4.2.1 Objective Function Evaluation......Page 111 4.3 Distributed Genetic Algorithm......Page 112 4.3.1 Dynamic Load Balancing......Page 114 4.4 Communication Cost and Performance Estimates......Page 123 4.4.1 Granularity......Page 124 4.4.2 Speedup......Page 126 4.4.3 Augmented Lagrangian GA......Page 129 4.5 Final Comments......Page 130 5.1 Introduction......Page 132 5.2 PVM System......Page 133 5.3 Implementation Model......Page 135 5.4 Example Structures......Page 137 5.4.1 Example 1 :17-Member Truss......Page 140 5.4.3 Example 3: 50-Story Space Structure......Page 143 5.5 Performance Studies......Page 151 5.5.1 Speedup......Page 152 5.5.2 Processor-Farming Model......Page 156 5.5.3 Observations Regarding the PCG Solve......Page 161 5.6 Final Comments......Page 162 6.1 Introduction......Page 164 6.2 GA-Based Structural Optimization......Page 166 6.3.1 CM-5 Architecture......Page 167 6.3.2 Programming Models......Page 169 6.3.3 Execution Models......Page 173 6.4 Coarse-Grained Concurrent Genetic Algorithm......Page 175 6.4.1 Virtual Channels......Page 176 6.4.2 Number of Processors Versus the Size of Population......Page 177 6.4.3 Parallel I/O......Page 178 6.5 Fine-Grained Data Parallel Fitness Function Evaluation......Page 180 6.5.1 Scatter and Gather Operations......Page 184 6.6.1 Example 1: 50-Story Space Structure......Page 185 6.6.2 Example 2: 35-Story Tower Space Structure......Page 186 6.6.3 Example 3: 147-Story Tower Space Structure......Page 189 6.7.1 Overall Processing Time......Page 192 6.7.2 Scalability......Page 193 6.7.3 Iterative PCG Solver......Page 196 7.1 Introduction......Page 200 7.2 Dynamic Analysis of Structures......Page 202 7.3 Subspace Iteration Method for Eigenvalue Extraction......Page 204 7.4 Graphic Animation in a Heterogeneous Environment......Page 205 7.4.2 Parallel Virtual Machine (PVM)......Page 211 7.4.3 Effective Communication Strategy......Page 212 7.5 Example......Page 213 7.6 Final Comments......Page 214 8.1 Introduction......Page 218 8.2 Direct Versus Iterative Solvers......Page 219 8.3 Hybrid Preconditioner for Large-Scale Problems......Page 220 8.4 Operating Systems War: Unix Versus Windows NT......Page 222 8.5 VRML: 3D Visualization on the Internet......Page 225 8.6 Network Infrastructure for Distributed Computing......Page 228 8.7 Final Comments......Page 229 References......Page 230 Author Index......Page 244 Subject Index......Page 250 This book presents an introduction to computer-aided engineering in distributed computing or networked environment. Distributed algorithms are presented for two fundamental areas of C-A engineering, finite element analysis and design optimization, demonstrating their efficiency and effectiveness through numerous examples. The last chapter details concurrent design optimization on parallel systems using genetic algorithms. Distributed Computer-Aided Engineering presents distributed algorithms for three fundamental areas of engineering: finite element analysis, design optimization, and visualization - providing a new direction in high performance structural engineering computing.