臺灣博碩士論文加值系統

本論文提出以深度學習設計貼片天線，以全連接神經網路與反向傳播技術生成貼片天線金屬圖形，以此方法緩解天線設計過程中需進行大量全波電磁模擬的問題。隨著無線通訊技術的進步，對天線的性能標準也日益嚴格。由於製造容易及體積輕巧等優點，貼片天線已成為無線通訊系統中的熱門選擇。然而進行貼片天線設計的過程中，需進行全波電磁模擬獲得天線結構對應的天線響應，而此過程相當耗時的過程。且當前設計天線方案中，大多會以初始金屬圖形的幾何參數縮放進行設計，但這會使金屬圖形的設計空間受到限制。本研究目標是以全連接神經網路生成符合規格需求的貼片天線金屬圖形，以減少在貼片天線設計過程中所需的全波電磁模擬次數，但在設計過程中會遇到全波電磁模擬軟體無法計算與傳遞梯度資訊的問題，因此會設計梯度估計網路解決此問題，並用線上學習訓練梯度估計網路減少蒐集資料集所需的大量電磁模擬。此外，會讓金屬圖形以網格化的形式設計，期望在人為干擾最少的情況下，透過此流程生成創新天線金屬圖形。研究中會探討金屬圖形生成網路與梯度估計網路的設計方案，也會根據不同性能需求設計損失函數。最後，為了驗證此本研究的設計流程，會將設計流程應用於不同基板結構上，還會對不同天線性能需求進行實驗，以驗證設計流程的有效性和靈活性。透過本研究提出的設計流程，我們在考慮n257頻段的情形下，成功設計出24GHz 至 32GHz 頻段範圍內具有濾波效果的四層板貼片天線金屬圖形。

This thesis proposes designing patch antennas using deep learning by generating metal patterns with fully connected neural networks and backpropagation techniques. This approach alleviates the need for extensive full-wave electromagnetic simulations. As wireless communication technologies advance and antenna performance standards become increasingly stringent, patch antennas, known for their ease of manufacture and compactness, have been a popular choice. However, designing these antennas involves time-consuming electromagnetic simulations to obtain antenna responses. This research aims to reduce the number of simulations required by generating antenna metal patterns that meet specifications using neural networks. A challenge arises as electromagnetic simulation software cannot compute and transmit gradient information, necessitating the development of a gradient estimation network, which is trained online to minimize the need for extensive data collection. Furthermore, the metal patterns are designed in a grid format to minimize human intervention, allowing for the creation of innovative designs. The research explores both the metal pattern generation and gradient estimation network, designing loss functions tailored to different performance needs. Finally, to validate the process, it is applied to various substrate structures and tested against different antenna performance requirements to confirm its effectiveness and adaptability. Using the design process proposed in this study, we successfully designed a four-layer patch antenna metal pattern with filtering effects in the 24GHz to 32GHz frequency range, considering the n257 band.

目錄
誌謝 ...........................................................................................................ii
摘要 ..........................................................................................................iii
Abstract.....................................................................................................iv
目錄 ...........................................................................................................v
圖目錄 .....................................................................................................vii
表目錄 ......................................................................................................ix
第一章 緒論 ...........................................................................................................1
第二章 文獻探討與問題定義 ...............................................................................3
2.1 貼片天線設計探討.............................................................................3
2.2 演化演算法進行貼片天線設計.........................................................7
2.3 神經網路技術應用於設計貼片天線金屬之探討.............................9
2.3.1 神經網路模型加速演化演算法設計貼片天線 .......................10
2.3.2 以 RL 設計貼片天線 ................................................................ 11
2.4 問題定義與總結...............................................................................13
第三章 研究方法 .................................................................................................16
3.1 貼片天線金屬圖形設計流程...........................................................16
3.2 基板選擇與可調控金屬區域設定...................................................17
3.3 模型設計...........................................................................................19
3.3.1 MPGN 模型設計........................................................................19
3.3.2 GEN 模型設計 ...........................................................................23
3.4 天線設計流程損失函數...................................................................25
3.5 因應不同基版的 GEN 模型改善.....................................................27
第四章 實驗 .........................................................................................................29
4.1 實驗環境說明...................................................................................29
4.2 實驗項目...........................................................................................29
4.2.1 實驗一 饋入源金屬固定與否的設計比較 .............................30
4.2.2 實驗二 找尋符合此架構之激活函數 .....................................33
4.2.3 實驗三 四層板 n257 頻段天線設計 .....................................36
4.2.4 實驗四 四層板濾波天線設計 .................................................38
4.2.5 實驗五 雙層板濾波天線設計 .................................................40
4.3 實驗總結...........................................................................................41
第五章 結論與未來展望 .....................................................................................48
參考文獻 .................................................................................................49
附錄一、PReLU 中參數 a 的更新過程.................................................53
附錄二、天線場型圖 .............................................................................54
附錄三、不同天線金屬設計結果 .........................................................56
附錄四、與其他濾波天線論文比較 .....................................................59


圖目錄
圖 2-1 天線結構示意圖...............................................................................................4
圖 2-2 天線金屬幾何參數示意圖...............................................................................5
圖 2-3 四層板基板結構 [15].......................................................................................6
圖 2-4 網格表示天線金屬分布示意圖.......................................................................8
圖 3-1 本研究的貼片天線設計流程示意圖.............................................................17
圖 3-2 雙層板基板結構與金屬調控區域示意圖.....................................................18
圖 3-3 四層板結構示意圖.........................................................................................18
圖 3-4 四層板基板結構與金屬調控區域示意圖.....................................................19
圖 3-5 四層板饋入點位置.........................................................................................19
圖 3-6 雙層板 MPGN 網路架構 ...............................................................................20
圖 3-7 MPGN 輸出陣列重塑為二維陣列的過程 .....................................................20
圖 3-8 四層板 MPGN 網路架構 ...............................................................................21
圖 3-9 符號函數的輸出輸入輸出關係圖.................................................................21
圖 3-10 三種激活函數輸入與輸出的關係圖...........................................................23
圖 3-11 雙層板 GEN 網路架構圖.............................................................................24
圖 3-12 四層板 GEN 網路架構圖.............................................................................25
圖 3-13 進行濾波天線設計的期望響應範圍 (A)反射係數範圍 (B)增益範圍......26
圖 3-14 加入 BATCH NORMALIZATION 的 GEN 網路架構圖 ....................................28
圖 4-1 雙層板固定饋入源金屬區域示意圖.............................................................31
圖 4-2 雙層板金屬網格與固定金屬區域結合示意圖.............................................31
圖 4-3 沒有固定饋入源金屬的損失曲線圖.............................................................32
圖 4-4 有固定饋入源金屬的損失曲線圖.................................................................32
圖 4-5 MPGN 使用 TANH 做為激活函數進行設計的損失曲線圖..........................34
圖 4-6 MPGN 使用 LEAKYRELU 做為激活函數進行設計的損失曲線圖..............34
圖 4-7 MPGN 使用 PRELU 做為激活函數進行設計的損失曲線圖.......................34
圖 4-8 四層板金屬區域配置示意圖 (A)第一層饋入源固定金屬區域 (B)第二層不配置金屬區域 .............................................................................................................37
圖 4-9 金屬網格與金屬配置區域結合示意圖.........................................................37
圖 4-10 四層板進行 N257 頻段設計的損失曲線圖.................................................38
圖 4-11 四層板進行濾波天線設計的損失曲線圖...................................................39
圖 4-12 雙層板進行濾波天線設計的損失曲線圖...................................................41
附錄圖 1 雙層板進行 N257 頻段設計的天線場型圖..............................................54
附錄圖 2 四層板進行 N257 頻段設計的天線場型圖..............................................54
附錄圖 3 四層板進行 N257 頻段且考慮濾波的天線設計場型圖..........................55
附錄圖 4 雙層板進行 N257 頻段且考慮濾波的天線設計場型圖..........................55
附錄圖 5 四階濾波天線架構 [32]............................................................................59
附錄圖 6 天線的幾何堆疊圖 [32]............................................................................59
附錄圖 7 天線基板架構圖 [32]................................................................................60
附錄圖 8 天線金屬區域配置圖 [32]........................................................................60
附錄圖 9 天線的反射係數和增益圖 [32]................................................................60

表目錄
表 2-1 演化演算法設計天線文獻比較.......................................................................8
表 2-2 演化演算法結合代理模型研究之比較......................................................... 11
表 2-3 RL 進行天線設計研究之比較........................................................................13
表 4-1 MPGN 四層 PRELU 的 A 值變化...................................................................36
表 4-2 不同天線設計方案的模擬次數比較表.........................................................44
表 4-3 雙層板進行 N257 頻段天線設計模擬結果...................................................44
表 4-4 四層板進行 N257 頻段天線設計模擬結果...................................................45
表 4-5 四層板進行 N257 頻段且考慮濾波的天線設計模擬結果...........................46
表 4-6 雙層板進行 N257 頻段且考慮濾波的天線設計模擬結果...........................47

[1]3GPP, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; New frequency range for NR (24.25-29.5 GHz) (Release 15),” Jun. 2018.
[2]L. Lizzi, F. Viani, R. Azaro, and A. Massa, “A PSO-Driven Spline-Based Shaping Approach for Ultrawideband (UWB) Antenna Synthesis,” IEEE Transactions on Antennas and Propagation, vol. 56, no. 8, pp. 2613–2621, 2008, doi: 10.1109/TAP.2008.927544.
[3]X. Li and K. M. Luk, “The Grey Wolf Optimizer and Its Applications in Electromagnetics,” IEEE Transactions on Antennas and Propagation, vol. 68, no. 3, pp. 2186–2197, 2020, doi: 10.1109/TAP.2019.2938703.
[4]Z. Wei et al., “Automated Antenna Design via Domain Knowledge-Informed Reinforcement Learning and Imitation Learning,” IEEE Transactions on Antennas and Propagation, vol. 71, no. 7, pp. 5549–5557, 2023, doi: 10.1109/TAP.2023.3266051.
[5]Z. Wei et al., “Fully Automated Design Method Based on Reinforcement Learning and Surrogate Modeling for Antenna Array Decoupling,” IEEE Transactions on Antennas and Propagation, vol. 71, no. 1, pp. 660–671, 2023, doi: 10.1109/TAP.2022.3221613.
[6]R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-resolution image synthesis with latent diffusion models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 10684–10695.
[7]T. Brown et al., “Language models are few-shot learners,” Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020.
[8]B. Yildirim, “Multiband and Compact WCDMA/WLAN Antenna for Mobile Equipment,” IEEE Antennas and Wireless Propagation Letters, vol. 10, pp. 14–16, 2011, doi: 10.1109/LAWP.2011.2107873.
[9]S. X. Ta and I. Park, “Compact Wideband Circularly Polarized Patch Antenna Array Using Metasurface,” IEEE Antennas and Wireless Propagation Letters, vol. 16, pp. 1932–1936, 2017, doi: 10.1109/LAWP.2017.2689161.
[10]M. Hu and Y. Li, “Wideband Back Cover Microstrip Antenna With Multiple Shorting Vias for Mobile 5G MIMO Applications,” IEEE Transactions on Antennas and Propagation, vol. 71, no. 10, pp. 8290–8295, 2023, doi: 10.1109/TAP.2023.3291802.
[11]C. A. Balanis, Antenna theory: analysis and design. John wiley & sons, 2016.
[12]S. K. Mishra, R. K. Gupta, and J. Mukherjee, “Effect of substrate material on radiation characteristics of an UWB antenna,” in 2010 Loughborough Antennas & Propagation Conference, 2010, pp. 157–160. doi: 10.1109/LAPC.2010.5666237.
[13]L. C. Paul, M. S. Hosain, S. Sarker, M. H. Prio, M. Morshed, and A. K. Sarkar, “The effect of changing substrate material and thickness on the performance of inset feed microstrip patch antenna,” American Journal of Networks and Communications, vol. 4, no. 3, pp. 54–58, 2015.
[14]A. A. Bhoot, S. A. Memon, A. Ahmed, and S. Hussain, “Comparative Study of Microstrip Patch Antenna with Different Shapes and its Application,” in 2019 2nd International Conference on Computing, Mathematics and Engineering Technologies (iCoMET), 2019, pp. 1–8. doi: 10.1109/ICOMET.2019.8673487.
[15]林佑宣, 毫米波封裝濾波器與封裝天線之設計, 國立中正大學, 2021.
[16]C. Mendes and C. Peixeiro, “A Dual-Mode Single-Band Wearable Microstrip Antenna for Body Area Networks,” IEEE Antennas and Wireless Propagation Letters, vol. 16, pp. 3055–3058, 2017, doi: 10.1109/LAWP.2017.2760142.
[17]A. Alieldin et al., “A Triple-Band Dual-Polarized Indoor Base Station Antenna for 2G, 3G, 4G and Sub-6 GHz 5G Applications,” IEEE Access, vol. 6, pp. 49209–49216, 2018, doi: 10.1109/ACCESS.2018.2868414.
[18]X. Jia and G. Lu, “A Hybrid Taguchi Binary Particle Swarm Optimization for Antenna Designs,” IEEE Antennas and Wireless Propagation Letters, vol. 18, no. 8, pp. 1581–1585, 2019, doi: 10.1109/LAWP.2019.2924247.
[19]B. Liu, H. Aliakbarian, Z. Ma, G. A. E. Vandenbosch, G. Gielen, and P. Excell, “An Efficient Method for Antenna Design Optimization Based on Evolutionary Computation and Machine Learning Techniques,” IEEE Transactions on Antennas and Propagation, vol. 62, no. 1, pp. 7–18, 2014, doi: 10.1109/TAP.2013.2283605.
[20]V. Grout, M. O. Akinsolu, B. Liu, P. I. Lazaridis, K. K. Mistry, and Z. D. Zaharis, “Software Solutions for Antenna Design Exploration: A Comparison of Packages, Tools, Techniques, and Algorithms for Various Design Challenges,” IEEE Antennas and Propagation Magazine, vol. 61, no. 3, pp. 48–59, 2019, doi: 10.1109/MAP.2019.2907887.
[21]“CST Studio Suite,” Dassault Systèmes. Accessed: May 07, 2024. [Online]. Available: https://www.3ds.com/products/simulia/cst-studio-suite
[22]“Ansys HFSS | 3D High Frequency Simulation Software.” Accessed: May 07, 2024. [Online]. Available: https://www.ansys.com/products/electronics/ansys-hfss
[23]D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114, 2013.
[24]I. Goodfellow et al., “Generative Adversarial Nets,” in Advances in Neural Information Processing Systems, Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, and K. Q. Weinberger, Eds., Curran Associates, Inc., 2014.
[25]C. Cui, W. T. Li, X. T. Ye, P. Rocca, Y. Q. Hei, and X. W. Shi, “An Effective Artificial Neural Network-Based Method for Linear Array Beampattern Synthesis,” IEEE Transactions on Antennas and Propagation, vol. 69, no. 10, pp. 6431–6443, 2021, doi: 10.1109/TAP.2021.3069467.
[26]J. H. Kim and S. W. Choi, “A Deep Learning-Based Approach for Radiation Pattern Synthesis of an Array Antenna,” IEEE Access, vol. 8, pp. 226059–226063, 2020, doi: 10.1109/ACCESS.2020.3045464.
[27]E. A. Karahan, A. Gupta, U. K. Khankhoje, and K. Sengupta, “Deep Learning based Modeling and Inverse Design for Arbitrary Planar Antenna Structures at RF and Millimeter-Wave,” in 2022 IEEE International Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting (AP-S/URSI), 2022, pp. 499–500. doi: 10.1109/AP-S/USNC-URSI47032.2022.9887077.
[28]Q. Wu, W. Chen, C. Yu, H. Wang, and W. Hong, “Machine-Learning-Assisted Optimization for Antenna Geometry Design,” IEEE Transactions on Antennas and Propagation, vol. 72, no. 3, pp. 2083–2095, 2024, doi: 10.1109/TAP.2023.3346493.
[29]A. M. Montaser and K. R. Mahmoud, “Deep Learning Based Antenna Design and Beam-Steering Capabilities for Millimeter-Wave Applications,” IEEE Access, vol. 9, pp. 145583–145591, 2021, doi: 10.1109/ACCESS.2021.3123219.
[30]Y. Bengio, N. Léonard, and A. Courville, “Estimating or propagating gradients through stochastic neurons for conditional computation,” arXiv preprint arXiv:1308.3432, 2013.
[31]S. Santurkar, D. Tsipras, A. Ilyas, and A. Madry, “How does batch normalization help optimization?,” Advances in neural information processing systems, vol. 31, 2018.
[32]R. Lu et al., “SIW cavity-fed filtennas for 5G millimeter-wave applications,” IEEE transactions on antennas and propagation, vol. 69, no. 9, pp. 5269–5277, 2021.