Scientific Computing and Algorithms in Industrial Simulations (eBook)

Michael Griebel is Director of the Fraunhofer Institute for Algorithms and Scientific Computing SCAI in Sankt Augustin, and holds a chair for Scientific Computing and Numerical Simulation at the Institute for Numerical Simulation at the University of Bonn.Marc Alexander Schweitzer is head of the department of Numerical Software of the Fraunhofer Institute for Algorithms and Scientific Computing SCAI in Sankt Augustin. Furthermore, he is currently Director of Institute for Numerical Simulation at the University of Bonn and holds a chair for Scientific Computing there.Anton Schüller is working at the Fraunhofer Institute for Algorithms and Scientific Computing SCAI in Sankt Augustin.

Preface 5
Contents 7
Part I Methods 9
Calculation of Chemical Equilibria in Multi-Phase: Multicomponent Systems 10
1 Introduction 10
2 Problem Formulation 12
2.1 Non-Ideal Gibbs Function 12
2.2 Stoichiometric Constraints 13
2.3 The Optimization Problem 15
3 Methodology for the Calculation of Chemical Equilibria 16
3.1 Reformulation of the Minimization Problem 17
3.2 Discretization of the H-Problem 18
3.3 Corrector Step 19
4 Automated Detection of Miscibility Gaps 20
5 Results 23
5.1 Gibbs Free Energy Minimization Using BePhaSys 23
5.2 Calculation of Two-Dimensional Phase Diagrams: Interpolation and Parallelization 23
Appendix: The Gibbs Free Energy Function 29
References 30
LC-GAP: Localized Coulomb Descriptors for the Gaussian Approximation Potential 32
1 Introduction 32
2 Potential Energy Prediction Through Machine Learning 33
2.1 The GAP Framework and Gaussian Process Regression 34
2.2 Localized Coulomb Matrix Descriptors 35
3 Results 37
3.1 Comparison of Descriptor Functions on QM7 39
3.2 Larger Datasets and Prediction of Multiple Properties 41
3.3 Distribution of Individual Atomic Contributions 44
4 Conclusions and Future Work 47
References 48
River Bed Morphodynamics: Metamodeling, Reliability Analysis, and Visualization in a Virtual Environment 50
1 Introduction 50
2 RBF Metamodel 53
3 Quantile Estimation 55
3.1 Sensitivity-Based Approach 55
3.1.1 First-Order Approximation 55
3.1.2 Second-Order Approximation 55
3.2 Monte Carlo 56
3.3 Weighted Monte Carlo 57
3.4 Quasi-Monte Carlo (QMC) 57
3.5 Quasi-Random Splines (QRS) 58
4 Numerical Tests 59
5 Visualization in Virtual Environment 62
6 Conclusion 65
References 65
Cooling Circuit Simulation I: Modeling 67
1 Introduction 67
2 Network 68
3 Water Pipes 69
3.1 Continuum Mechanics 69
3.2 Simplifying Assumptions 70
3.3 Discretization and Regularization 72
4 Further Devices 75
4.1 Resistors and Valves 75
4.2 Pumps 77
4.3 Heat Exchangers 79
5 Element Control 82
6 Conclusion 84
References 85
Part II Products 86
Algebraic Multigrid: From Academia to Industry 87
1 Introduction 87
2 From Geometric to Algebraic Multigrid 89
3 The Early Phase of Algebraic Multigrid (1982–1987) 91
3.1 The First Documented AMG Application 92
3.2 The Basics of `Classical' AMG 93
4 The Renaissance of AMG (1995–2000) 95
4.1 Resumption of Major Research on AMG 95
4.2 Towards Industry 96
4.2.1 Computational Fluid Dynamics 97
4.2.2 Streamline Approach in Oil Reservoir Simulation 99
5 The Main AMG Development Phase (2000–Today) 101
5.1 The General Trend 101
5.2 Bridging the Gap 102
5.3 SAMG for Coupled PDE Systems 105
5.3.1 Unknown-Based Approach 105
5.3.2 Point-Based Approach 106
5.3.3 Status of the Solver Framework SAMG 106
6 Industry-Driven Applications 107
6.1 Semiconductor Applications 107
6.2 Multi-Ion Transport and Reaction 110
6.3 Oil Reservoir Simulation 111
6.3.1 The Reservoir Simulation Models 113
6.3.2 Fully Implicit Methods 114
7 Summary, Conclusions and Lessons Learned 119
References 121
Parallel Algebraic Multigrid 124
1 Introduction 124
2 Challenges Imposed by Parallel Computer Architectures 126
2.1 Single Core Performance: CPU Clock Speed and Memory Frequency 126
2.2 Multi-Core CPUs and Shared Memory Parallelism 127
2.2.1 Multi-Core and Memory Access 128
2.2.2 Intrinsically Serial Components 129
2.2.3 Race Conditions 129
2.2.4 Multiple Sockets 130
2.3 Distributed Memory Parallelism 130
3 How SAMG Counters the HPC Challenges 131
3.1 Tuning and Parallelization of Smoothing 132
3.2 Ruge-Stüben Coarsening 134
3.3 Coarse Grid Solution 135
4 SCAI's Parallel SAMG Solver Library 136
References 137
MpCCI: Neutral Interfaces for Multiphysics Simulations 138
1 Introduction 138
2 MpCCI CouplingEnvironment 139
2.1 Aero-Elasticity and Fluid-Structure-Interaction 141
2.1.1 Wing and Spoiler Design 141
2.1.2 Hydraulic Pump Layout 142
2.2 Thermal and Vibration Loads in Turbomachinery 143
2.2.1 Thermal Loads on Ceramic Impeller 143
2.2.2 Life-Time Estimation of Turbine Blades 144
2.3 Vehicle Dynamics and Nonlinear Component Behavior 144
2.3.1 Driving Over Obstacles 144
2.3.2 Wading Simulation for Off-Road Vehicles 144
2.4 Automotive Thermal Management 146
2.4.1 Automotive Thermal Management for Full Vehicles 146
2.4.2 Automotive Thermal Management for Vehicle Manifolds 146
2.5 Component Design in Electrical Engineering 147
2.5.1 Cooling of a 3-Phase Transformer 147
2.5.2 Electric Arc in Switching Devices 147
3 MpCCI FSIMapper 148
4 MpCCI Mapper Solution for Integrated Simulation Workflows 149
4.1 Passive Safety 150
4.2 Forming Tools and Material Properties 151
4.2.1 Lightweight Stamping Tools: Use Forming Loads in Structural Optimization 151
4.2.2 Validation of Material Model Parameters: Compare Forming Results and Experimental Data 151
4.3 Composite Structures and Plastic Components 152
4.3.1 CFRP Workflows: From Draping via Mulling and Curing to Structural Analysis 152
4.3.2 Structural Integrity of Blow Moulded Plastic Components 153
5 Conclusion 153
References 153
Cooling Circuit Simulation II: A Numerical Example 155
1 Introduction 155
2 Application 156
2.1 Cooling System 156
2.2 Circuit Basics and Example 156
3 Concept and Software 163
3.1 Framework and Components 163
3.2 Semi-Automatic Model Creation with Schemparser 165
3.3 Device Modeling and Sensor Mapping 166
3.4 Collection of Measurement Data with PowerDAM 167
3.5 Nonlinear Problem Setup and Solution with MYNTS 168
4 Numerical Tests 170
4.1 Simplified Heat Exchanger 170
4.2 Logarithmic Mean Temperature Difference 177
5 Conclusion 179
References 181
The LAMA Approach for Writing Portable Applications on Heterogenous Architectures 183
1 Introduction 183
2 LAMA 184
2.1 Heterogeneous Memory 186
2.2 Heterogeneous Kernel 188
2.3 Task Parallelism 188
2.4 Distributed Memory Support 190
2.5 Matrices and Vectors 191
2.6 Solver Framework 191
2.7 Extensibility and Maintainability 193
3 Performance Comparison 194
4 Summary 197
Appendix 198
Test Environment 198
Test Matrices 199
References 200
ModelCompare 201
1 Introduction 201
2 Development History 202
3 Capabilities 202
3.1 Detection of Geometry Changes 203
3.2 Detection of MultiParts 204
3.3 Spotwelds and Rigid Body Elements 205
3.4 Detection of Material-ID and Thickness Changes 206
4 Outlook 207
References 207
Rapid Enriched Simulation Application Development with PUMA 208
1 Introduction 208
2 Partition of Unity Methods 209
3 PUMA Framework Design 210
4 Application Examples 212
5 Concluding Remarks 225
References 226
Part III Applications and Show Cases 228
Applying CFD for the Design of an Air-Liquid Interface In-Vitro Testing Method for Inhalable Compounds 229
1 Introduction 229
2 In-Vitro Air-Liquid Interface 230
3 Simulating the Aerosol Conduction System 231
4 Simulating the Liquid Supply System 237
4.1 Clogging in Liquid Channels 239
5 Simulating an Aerosol Sampling Box 240
6 Conclusions 243
References 243
A Mapping Procedure for the Computation of Flow-Induced Vibrations in Turbomachinery 244
1 Introduction 244
2 Nonlinear Harmonic Method 245
3 Mapping of Pressure Excitations 246
3.1 Periodic Models and Nodal Diameters 248
3.2 Deriving Excitation and Responding Shape 251
3.3 Summary 252
4 Application Example 252
4.1 Harmonic CFD Simulation 254
4.2 Mapping 254
4.3 Harmonic Structural Analysis 257
5 Conclusion 259
References 261
Molecular Dynamics Simulation of Membrane Free Energy Profiles Using Accurate Force Field for Ionic Liquids 263
1 Introduction 263
2 Computational Methods 264
2.1 Simulation Details 264
2.1.1 Technical Details 264
2.1.2 Force Field Development for [C2MIM][EtSO4] 265
2.2 Umbrella Sampling 267
3 Results and Discussion 269
3.1 Force Field Development for [C2MIM][EtSO4] 269
3.1.1 Density 269
3.1.2 Self-Diffusion Coefficients 270
3.1.3 Heat of Vaporization 270
3.1.4 Shear Viscosity 271
3.2 Free Energy Profiles 271
4 Outlook and Conclusion 274
4.1 Outlook: Towards Fully Automated Force Field Development 274
4.1.1 Case Study: Automated Parameterization of Ethylene Oxide 277
4.2 Conclusion 279
References 279
The cloud4health Project: Secondary Use of Clinical Data with Secure Cloud-Based Text Mining Services 283
1 Introduction 283
2 Developing a Secure Cloud-Solution for Medicine 285
2.1 Existing Cloud Solutions for Medicine 285
2.2 Requirements for Cloud Infrastructures Arising from Patient Data Processing 287
2.3 Security Mechanisms 288
2.4 Secure Cloud Infrastructure 290
2.4.1 Secure Clinical Gateway to the Cloud 291
2.4.2 Data Processing Flow 292
2.4.3 End-to-End Encryption 293
2.4.4 Multi-Tenancy and No Data Persistence 295
3 Clinical Text Mining Solutions 297
3.1 Short Literature Overview 297
3.2 General Architecture of the Text Mining Services 299
3.3 Overview Use Cases 301
3.3.1 General Use Case Process Model 302
3.4 Mining Endoprosthetic Surgery Reports 303
3.5 Mining Pathology Reports 306
4 Discussion 309
References 311
Dimensionality Reduction for the Analysis of Time Series Data from Wind Turbines 314
1 Introduction 314
2 Time Series Characteristics in Wind Energy 316
2.1 Numerical Simulations of Wind Turbines 316
2.2 Condition Monitoring of Wind Turbines 318
3 Exploration of Time Series Data from Numerical Simulations 319
3.1 Virtual Sensor Data from Wind Turbine Simulations 319
3.2 Nonlinear Dimensionality Reduction for Time Series Analysis 321
3.3 Diffusion Maps 322
3.4 Numerical Results 323
4 Anomaly Detection Based on Linear Dimensionality Reduction for Condition Monitoring Sensor Data from Wind Turbines 326
4.1 Sensor Data from Rotor Blades 326
4.1.1 Pre-processing 327
4.2 Anomaly Detection in Sensor Data 329
4.2.1 Model of Undamaged State 329
4.2.2 Deviation from the Undamaged State 330
4.3 Methodology for Damage Detection 331
4.4 Numerical Results 332
5 Conclusions 335
References 335
Energy-Efficiency and Performance Comparison of Aerosol Optical Depth Retrieval on Distributed Embedded SoC Architectures 337
1 Introduction 337
2 Method 338
3 Implementation 340
4 Embedded Low-Energy System 342
5 Benchmarks 344
5.1 Benchmark Environment 345
5.2 Performance Benchmarks 345
5.3 Energy Benchmarks 347
6 Discussion 353
7 Outlook 353
References 354
Part IV A Short History 355
The Fraunhofer Institute for Algorithms and ScientificComputing SCAI 356
1 Foundation of the GMD, the First Decade (1968–1977) 358
2 Numerical Simulation, Multigrid and Parallel Computing (1978–1991) 359
3 SCAI: Algorithms and Scientific Computing (1992–2001) 362
4 SCAI as a Fraunhofer Institute, the First Years (2001–2009) 366
5 New Fields of Research and New Business Areas (2010–2016) 368
References 369