다중 매개변수 공학 시스템을 위한 시뮬레이션 기반 최적화 프레임워크: 결정질 실리콘 태양전지와 밀리미터파 안테나 사례 연구 = Simulation-Driven Optimization Frameworks for Multi-Parameter Engineering Systems: Case Studies in Crystalline Silicon Solar Cells and Millimetre-Wave Antennas

다국어 초록 (Multilingual Abstract)

The optimization of complex engineering problems has been hindered by the high computational cost of reliable simulation processes, and thus, the separation of design variables. This kind of complexity was apparently, clear evident in the case of the simulation of photovoltaic (PV) and millimetre wave (mmWave) antenna, which need reliable simulation tools, but do not facilitate the optimization of multiple variables. This thesis aims to utilize simulation tools for optimization.
The first case study illustrates the physics-informed optimization process for crystalline silicon solar cells using the physics-informed surrogate model, which is the integration of the Deep Feed-forward Neural Network (DFNN) model and the Personal Computer One-Dimensional (PC1D) device simulation tool. The computational-intensive PC1D simulations result in a considerable amount of data required for the search process, thereby making the process filtered by incorporating the physics-informed feasibility. The DFNN model efficiently searches the design space, and the results are verified by the re-simulations using the PC1D device simulation tool. The proposed technique possesses excellent predictability (R² = 0.999) and obtains the optimal result, where the calculated efficiency is 29.44%, and the result is physically simulated and verified. The computational expense is reduced by a factor of 4.3 when compared to the complete parametric search. The second case study relates to the simulation-based optimization of a wideband dual-polarized U-slot coupled patch antenna operating at 60 GHz. The simulation study aims at port-isolation improvement with simultaneous maintenance of the radiation properties using the mesh-convergence study and simulation-based guided optimization.
These two presented case studies can be taken to represent the unified simulation-based approach to the respective designs, encompassing simulation-assisted designs and the exploration of the high-dimensional spaces. The current work offers some real-world optimization strategies independent of the context since the focus of this study is both robustness and reliability compared to the computational optimality.

목차 (Table of Contents)

• CHAPTER 1: INTRODUCTION 1
• 1.1 Background and Motivation 2
• 1.2 Simulation-Driven Optimization in Engineering Systems 3
• 1.3 Research Gap 3
• 1.4 Research Objectives 4
• 1.5 Thesis Organization 4
• CHAPTER 2: Case Study 1: Physics-Guided AI Framework for Multi-Parameter Optimization of Solar Cells: Application to Crystalline Silicon Devices 5
• 2.1 INTRODUCTION 6
• 2.1.1 Conventional Solar-Cell Optimization Approaches 6
• 2.1.2 Artificial-Intelligence-Based Solar-Cell Modelling 7
• 2.1.3 Limitations of Purely Data-Driven AI Approaches 9
• 2.1.4 Physics-Guided Artificial Intelligence for Solar-Cell Optimization 10
• 2.1.5 Related Work and Positioning of This Study 12
• 2.2 METHODOLOGY 14
• 2.2.1 Overview of the Physics-Guided Optimization Framework 14
• 2.2.2 Sensitivity-Based Parameter Selection and Automated Dataset Generation 20
• 2.2.3 Physics-Guided Feasibility Filtering 21
• 2.2.4 Data Pre-processing and Normalization 21
• 2.2.5 Surrogate Model Development and Hyperparameter Optimization 22
• 2.2.6 ML-Based Optimal Design Extraction and PC1D Validation 25
• 2.3 APPLICATION & RESULTS 27
• 2.3.1 Description of the Application Model 27
• 2.3.2 Dataset Characteristics and Parameter Space 29
• 2.3.3 Surrogate Model Performance Evaluation 34
• 2.3.4 ML-Based Optimal Design Extraction 41
• 2.3.5 PC1D-Based Physical Validation 43
• 2.3.6 Computational Efficiency and Comparison with Existing Works 45
• CHAPTER 3: Case Study 2: Enhancing Wideband Dual-Polarized U-Slot Coupled Patch Antenna Through Maximized Port Isolation at 60GHz Band 50
• 3.1 INTRODUCTION 51
• 3.1.1 60 GHz Band Characteristics 52
• 3.1.2 Microstrip Patch Antennas in mm Wave Applications 52
• 3.1.3 Circular Patch Antennas 52
• 3.1.4 Substrate Considerations for mm Wave Antennas 53
• 3.1.5 Dual Polarization and Port Isolation 53
• 3.1.6 Related Work on 60-GHz Dual-Polarized Antennas 53
• 3.2 METHODOLOGY 58
• 3.2.1 Antenna Geometry and Design Variables 60
• 3.2.2 Simulation Environment (HFSS Setup) 67
• 3.2.3 Airbox Study 69
• 3.2.4 Mesh Convergence Analysis 72
• 3.2.5 One-Variable Sensitivity Sweeps 76
• 3.2.6 DOE Optimization Sweep 78
• 3.3 APPLICATION & RESULTS 82
• 3.3.1 DOE-Optimized Electrical Performance 82
• 3.3.2 Radiation Characteristics of the Optimized Design 83
• 3.3.3 System-Level Validation (Isolation vs Modulation Requirement) 84
• 3.3.4 Comparison with Reference Design 86
• CHAPTER 4: CROSS-CASE DISCUSSSION AND CONCLUSION 89
• 4.1 Cross-Case Synthesis of the Simulation-Driven Optimization Framework 90
• 4.2 Comparative Analysis of Optimization Outcomes 91
• 4.2.1 Computational Efficiency and Scalability 91
• 4.2.2 Physical Consistency and Validation Reliability 91
• 4.3 Methodological Insights and Design Implications 92
• 4.4 Limitations of the Present Study 93
• 4.5 Future Research Directions 93
• 4.6 Final Conclusions 94
• REFERENCES 95
• 요약 (국문초록) 107
• ACKNOWLEDGEMENT 110