High Performance Simulation for Industrial Paint Shop Applications

Preface
Contents
Part I Introduction and Background
	1 Introduction
	2 Background
		2.1 Computational Fluid Dynamics
			2.1.1 Fundamentals
			2.1.2 Governing Equations
			2.1.3 Discretization Techniques
				2.1.3.1 Grid-Based Methods
				2.1.3.2 Particle-Based Methods
		2.2 High Performance Computing
			2.2.1 Fundamentals
			2.2.2 Shared Memory Parallelism
			2.2.3 Distributed Memory Parallelism
			2.2.4 General-Purpose Computing on Graphics Processing Units
		2.3 Automotive Paint Shop
			2.3.1 Overview
			2.3.2 Challenges
Part II Grid-Based Methods
	3 Overview
		3.1 Finite Difference Method
			3.1.1 Formulation
			3.1.2 Grid Discretization
		3.2 Electrophoretic Deposition Coatings
	4 Simulation of Electrophoretic Deposition Coatings
		4.1 Background
			4.1.1 State of the Art
			4.1.2 Formulation
		4.2 General Idea
			4.2.1 Numerical Modeling of EPD
			4.2.2 Grid Discretization
		4.3 Simulation of EPD Coatings
			4.3.1 Implementation of Numerical Model
			4.3.2 Overset Grid Implementation
				4.3.2.1 Grid Ω16h
				4.3.2.2 Grid Ω8h
				4.3.2.3 Grid Ω2h
				4.3.2.4 Grid Ωh
				4.3.2.5 Discussion and Resulting Overall Algorithm
		4.4 Experimental Evaluations
			4.4.1 Validation with Analytical Data
			4.4.2 Validation with Industrial Data
			4.4.3 Performance Discussion
		4.5 Summary
Part III Volumetric Decomposition Methods
	5 Overview
		5.1 Fundamentals
		5.2 Drawback
	6 Volumetric Decomposition on Shared Memory Architectures
		6.1 Background
			6.1.1 State of the Art
			6.1.2 Basic Architecture
		6.2 Parallel Simulation of Electrophoretic Deposition
			6.2.1 Outer Parallel Layer
			6.2.2 Inner Parallel Layer
				6.2.2.1 Identifying Critical Vertices
				6.2.2.2 Constructing the Volume Decomposition
				6.2.2.3 Integrating Bottlenecks
		6.3 Experimental Evaluations
			6.3.1 Speedup for the Reeb Graph Construction
			6.3.2 Speedup for the Entire Simulation
		6.4 Summary
	7 Volumetric Decomposition on Distributed Memory Architectures
		7.1 Basic Architecture
		7.2 Implementation of the Distributed Algorithm
			7.2.1 Workload Distribution
			7.2.2 Memory Optimization
			7.2.3 Load Balancing
		7.3 Experimental Evaluations
			7.3.1 Test Environment and Considered Data Set
			7.3.2 Speedup in the Reeb Graph Construction
			7.3.3 Speedup in the Entire Simulation
		7.4 Summary
Part IV Particle-Based Methods
	8 Overview
		8.1 SPH Fundamentals
			8.1.1 Formulation
			8.1.2 Internal Forces
			8.1.3 External Forces
		8.2 SPH Variants
			8.2.1 Basic Variants
			8.2.2 Predictive-Corrective Incompressible SPH
		8.3 SPH and High Performance Computing
			8.3.1 CPU Parallelization
			8.3.2 GPU Parallelization
	9 SPH on Multi-GPU Architectures
		9.1 Background
			9.1.1 Basic Architecture
			9.1.2 Motivation
		9.2 Advanced Load Balancing
			9.2.1 General Idea
			9.2.2 Using Internal Cache
			9.2.3 Using Pointers
		9.3 Experimental Evaluations
			9.3.1 Experimental Setup
			9.3.2 Dam Break Simulation
			9.3.3 Spray Wash Simulation
		9.4 Summary
	10 SPH Variants on Multi-GPU Architectures
		10.1 Background
		10.2 Distributed Multi-GPU Architecture
		10.3 Optimization Techniques
			10.3.1 Load Balancing
			10.3.2 Overlapping Memory Transfers
			10.3.3 Optimizing Particle Data Representation
			10.3.4 Optimizing Exchange of Halos
		10.4 Experimental Evaluations
			10.4.1 Experimental Setup
			10.4.2 Dam Break Simulation
			10.4.3 Water Splashing Simulation
		10.5 Summary
Part V Conclusion
	11 Conclusion
References
Index