簡易檢索 / 詳目顯示

回結果列表
研究生： 黃郁傑
Huang, Yu-Jie
論文名稱： 非監督式學習於氨氣微燃燒器化學反應器網路模型之聚類最佳化
Unsupervised learning for cluster optimization of ammonia micro-burner chemical reactor network model
指導教授： 伍芳嫺
Wu, Fang-Hsien
學位類別： 碩士
Master
系所名稱： 工學院 - 機械工程學系
Department of Mechanical Engineering
論文出版年： 2026
畢業學年度： 114
語文別： 中文
論文頁數： 218
中文關鍵詞： 氨氣微燃燒器 、化學反應器網路 、非監督分群演算法 、CFD-CRN轉換框架
外文關鍵詞： Ammonia micro-combustor, Chemical reactor network, Unsupervised clustering algorithm, CFD-CRN conversion framework
相關次數： 點閱：3  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 高解析度計算流體力學（CFD）模擬雖能提供豐富物理化學資訊，卻伴隨龐大計算成本。化學反應器網路（CRN）雖可大幅降低計算負擔，然傳統建構依賴主觀反應區域判斷，難以適應幾何複雜度之變化。本研究建立一套以資料驅動分群為基礎、物理誤差為導向之自動化CFD-CRN轉換框架，比較K 平均分群法（K-means）、高斯混合模型（Gaussian mixture model, GMM）與凝聚式階層演算法（Agglomerative Hierarchical Clustering, AHC）三種非監督分群演算法，並比較資料空間評分策略與誤差菁英評分策略兩種超參數選擇方法對 CRN 預測能力之影響。本研究以單段觸媒、多段觸媒及觸媒結合空腔三種氨氣微燃燒器作為CFD之基礎，將不同幾何設計作為化學場複雜度遞增之驗證序列，探討特徵集規模、分群演算法與超參數選擇對 CRN 建構結果之影響，最後藉由改變入口條件與幾何構型之外推測試，評估超參數組合之泛用性。結果顯示，資料空間評分策略之最佳分群數量，在三種幾何下均固定為5個反應器；誤差菁英評分策略之最優分群數量則隨幾何複雜度增加至15個反應器，且能為空腔滯留區分配獨立反應器，適用演算法之部份主要目標物種預測誤差皆改善10%以上。在演算法比較中，雖然AHC搭配平均連結法（average linkage method）展現出最佳跨幾何穩定性，其層次分辨能力可同時捕捉大型主流群集與小型局部反應熱點。有關特徵集縮編分析卻進一步指出，多段觸媒幾何透過GMM於縮編後除氨氣（NH3）外，其餘物種之預測誤差可降5%至30%，且其CV-RMSE絕對值也具較佳表現，而單段觸媒與觸媒結合空腔幾何則以原規模特徵集為佳。於泛用性測試結果中可知，固定超參數但當入口組成與速度改變時仍能正確重現流場結構，點火後燃燒區溫度誤差可控制在62 K以內，相關物種之莫爾分率誤差亦可控制在10-3至10-5；另外，若以氨氣（NH3）/過氧化氫（H2O2）做為不同燃料與氧化劑系統之極端外推條件，其溫度、水(H2O)、氮氣(N2)、氧氣(O2)等物種之皮爾森相關係數分析，仍有大於或相當0.8的高相依性，同時搭配CV-RMSE證實該超參數亦能捕捉CFD趨勢。

    This study proposes a data-driven, clustering-based automated framework for converting high-fidelity Computational Fluid Dynamics (CFD) results into Chemical Reactor Network (CRN) reduced-order models for platinum-catalyzed ammonia micro-combustors. Ammonia's role as a zero-carbon fuel is growing, yet high CFD computational costs limit large-scale parameter sweeps. The framework automates CRN construction and reduces reliance on subjective engineering judgment.
    Three unsupervised clustering algorithms—K-means, Gaussian Mixture Model (GMM), and Agglomerative Hierarchical Clustering (AHC)—are evaluated across three combustor geometries of increasing complexity: single-stage catalyst, multi-stage catalyst, and catalyst combined with a cavity. A key contribution is a physics-error-oriented elite scoring strategy using CV-RMSE from CRN simulations combined with a Top-K Rank Intersection method, serving as an alternative to conventional feature-space metrics (DB index and Silhouette coefficient) for hyperparameter selection.
    Results show that the error elite strategy consistently yields lower CRN prediction errors for temperature and major species by better resolving high-gradient zones such as catalytic wall layers and cavity recirculation regions. AHC with average linkage demonstrates the most stable cross-geometry performance. Feature set reduction further reveals that a minimal seven-dimensional feature set improves CRN accuracy for the multi-stage catalyst geometry by removing spatially patchy intermediate species that distort clustering.
    Generalization tests confirm that fixed hyperparameters maintain acceptable accuracy in post-ignition zones under varying inlet conditions, though limitations remain near the wall ignition onset region. The framework offers a generalizable, computationally efficient methodology extensible to other low-carbon fuel systems and complex combustor designs.

    摘要 i 英文摘要 iii 致謝 viii 目錄 ix 表目錄 xi 圖目錄 xii 第一章 緒論 1 1.1 全球減碳壓力與工業燃燒轉型 1 1.2 無碳燃料之燃燒特性及應用發展 5 1.3 微燃燒器技術與氨氣微尺度燃燒特性 7 1.4 人工智慧於燃燒研究與工業應用之發展趨勢 8 第二章 文獻回顧 12 2.1 氨氣微燃燒 12 2.2 CFD降階建模方法 20 2.3 非監督式分群演算法 25 2.4 氣相化學反應機制 30 2.5 研究動機 35 第三章 研究方法 39 3.1 研究框架總覽 39 3.2 Ansys Fluent 40 3.2.1 統御方程式介紹 40 3.2.2 數值模型、網格與邊界條件設定 43 3.2.3 化學反應機構 48 3.2.4 數據提取與預處理 53 3.3 非監督式分群與 CFD–CRN 降階建模擬 60 3.3.1 非監督式分群演算法 62 3.3.2 超參數調查與網格搜尋設計 69 3.3.3 空間連通性與質量平衡修正 73 3.4 性能與誤差評估指標 80 3.5 化學反應器網絡(Chemical Reactor Network, CRN) 88 3.6 化學特徵集規模與遷移性測試 90 3.7 超參數泛用性 94 第四章 結果與討論 97 4.1 CFD–CRN 降階框架之超參數選擇策略比較 97 4.1.1 兩種評估策略之差異 99 4.1.2 分群與燃燒化學場結構之匹配性 104 4.1.3 CV-RMSE量化評估主要目標與次要目標誤差之差異 112 4.1.4 分群演算法適配性與方法論適用邊界 128 4.2 輸入特徵集規模對分群與CRN模擬之影響性 132 4.2.1 特徵集之遷移性測試 133 4.2.2 特徵集規模於多段觸媒幾何對最佳超參數與CRN誤差之影響 149 4.3 超參數泛用性 162 4.3.1 更改入口條件 162 4.3.2 更改幾何與入口條件 169 4.3.3 計算效益評估 181 第五章 結論 184 參考文獻 190 附錄 196

    [1] H. Ritchie, M. Roser, and P. Rosado. "CO₂ and Greenhouse Gas Emissions." Our World in Data. (accessed.
    [2] E. Basson, "2024 World Steel in Figures," ed: World Steel Association, 2024.
    [3] 陳建任, "全球鋼鐵業碳排現況與低碳化發展動向," ed: 經濟部產業技術司, 2023.
    [4] IEA, Net Zero Roadmap A Global Pathway to Keep the 1.5 °C Goal in Reach. IEA, 2023.
    [5] Basic Hydrogen Strategy The Ministerial Council on Renewable Energy, Hydrogen and Related Issues, 2023/06.
    [6] IEA, Global Hydrogen Review 2023. IEA, 2023.
    [7] H. Kobayashi, A. Hayakawa, K. D. Kunkuma A. Somarathne, and Ekenechukwu C. Okafor, "Science and technology of ammonia combustion," Proceedings of the Combustion Institute, vol. 37, no. 1, pp. 109–133, 2019/01/01/ 2019, doi: https://doi.org/10.1016/j.proci.2018.09.029.
    [8] L. Folcarelli et al., "A Chemical Reactor Network methodology for estimating NOx emissions under non-premixed lean hydrogen combustion with water injection," International Journal of Hydrogen Energy, vol. 168, p. 150561, 2025/09/15/ 2025, doi: https://doi.org/10.1016/j.ijhydene.2025.150561.
    [9] I. Group. "Hydrogen." https://www.ineos.com/businesses/ineos-hydrogen/hydrogen/ (accessed.
    [10] W. contributors. "Haberprocess." https://en.wikipedia.org/w/index.php?title=Haber_process&oldid=1329042096 (accessed.
    [11] W. contributors. "Ostwald process." https://en.wikipedia.org/w/index.php?title=Ostwald_process&oldid=1333723090 (accessed.
    [12] Q. Cheng, A. Muhammad, O. Kaario, Z. Ahmad, and L. Martti, "Ammonia as a sustainable fuel: Review and novel strategies," Renewable and Sustainable Energy Reviews, vol. 207, p. 114995, 2025/01/01/ 2025, doi: https://doi.org/10.1016/j.rser.2024.114995.
    [13] J. K. Moses, T. A. Tivfa, N. A. A. Qasem, and A. B. S. Alquaity, "Combustion characteristics of hydrogen, ammonia, and their blends: A review," Fuel, vol. 388, p. 134404, 2025/05/15/ 2025, doi: https://doi.org/10.1016/j.fuel.2025.134404.
    [14] T. a. I. Ministry of Economy, The 6th Strategic Energy Plan. 2021.
    [15] T. a. I. Ministry of Economy. "Ammonia can be used as fuel! Part 1: Ammonia is commonly used in our daily life, but its potential is not widely known." (accessed.
    [16] C. Fernandez-Pello, "Micro-Power Generation Using Combustion: Issues and Approaches," presented at the Twenty-Ninth International Symposium on Combustion, Sapporo, Japan, July 21–26, 2002.
    [17] A. Epstein et al., "Micro-heat engines, gas turbines, and rocket engines - The MIT microengine project," 06/29 1997, doi: 10.2514/6.1997-1773.
    [18] T. F. Guiberti, G. Pezzella, A. Hayakawa, and S. M. Sarathy, "Mini Review of Ammonia for Power and Propulsion: Advances and Perspectives," Energy & Fuels, vol. 37, no. 19, pp. 14538–14555, 2023, doi: 10.1021/acs.energyfuels.3c01897.
    [19] M. Nauman, J. Pan, Y. Wang, F. Li, A. Oluwaleke Ojo, and A. Raza, "A review of recent advancements in micro combustion techniques to enhance flame stability and fuel residence time," International Journal of Hydrogen Energy, vol. 49, pp. 1165–1193, 2024, doi: 10.1016/j.ijhydene.2023.09.050.
    [20] K. L. Lueth. "Industrial AI market: 10 insights on how AI is transforming manufacturing." https://iot-analytics.com/industrial-ai-market-insights-how-ai-is-transforming-manufacturing/ (accessed.
    [21] A. b. M. QuantumBlack, The state of AI 2025: Agents, innovation, and transformation. McKinsey & Company, 2025.
    [22] Artesis, "AI Predictive Maintenance: Real Data Shows 73% Drop in Equipment Failures," ed, 2026.
    [23] C. Noon, "Blade Runners: GE Vernova Is Deploying AI-Enabled Machines to Boost Wind Turbine Blade Quality," ed: GE Vernova, 2024.
    [24] I. Asaolu, "AI Adoption Straining Global Electricity Grids, IEA Warns," ed, 2026.
    [25] J. S. Jones, "Iberdrola launches AI innovation project to optimise power grids," ed, 2024.
    [26] Y.-H. Li, G.-B. Chen, F.-H. Wu, T.-S. Cheng, and Y.-C. Chao, "Combustion characteristics in a small-scale reactor with catalyst segmentation and cavities," Proceedings of the Combustion Institute, vol. 34, no. 2, pp. 2253–2259, 2013, doi: 10.1016/j.proci.2012.06.058.
    [27] Y.-H. Li, G.-B. Chen, F.-H. Wu, H.-F. Hsieh, and Y.-C. Chao, "Effects of carbon dioxide in oxy-fuel atmosphere on catalytic combustion in a small-scale channel," Energy, vol. 94, pp. 766–774, 2016, doi: 10.1016/j.energy.2015.11.042.
    [28] Y.-H. Li, G.-B. Chen, F.-H. Wu, T.-S. Cheng, and Y.-C. Chao, "Effects of catalyst segmentation with cavities on combustion enhancement of blended fuels in a micro channel," Combustion and Flame, vol. 159, no. 4, pp. 1644–1651, 2012, doi: 10.1016/j.combustflame.2011.11.017.
    [29] A. Hayakawa, T. Goto, R. Mimoto, Y. Arakawa, T. Kudo, and H. Kobayashi, "Laminar burning velocity and Markstein length of ammonia/air premixed flames at various pressures," Fuel, vol. 159, pp. 98–106, 2015, doi: 10.1016/j.fuel.2015.06.070.
    [30] E. C. Okafor et al., "Measurement and modelling of the laminar burning velocity of methane-ammonia-air flames at high pressures using a reduced reaction mechanism," Combustion and Flame, vol. 204, pp. 162–175, 2019, doi: 10.1016/j.combustflame.2019.03.008.
    [31] E. C. Okafor et al., "Experimental and numerical study of the laminar burning velocity of CH4–NH3–air premixed flames," Combustion and Flame, vol. 187, pp. 185–198, 2018, doi: 10.1016/j.combustflame.2017.09.002.
    [32] X. Zhu, J. Du, Z. Yu, Y.-B. Cheng, and Y. Wang, "NOx Emission and Control in Ammonia Combustion: State-of-the-Art Review and Future Perspectives," Energy & Fuels, vol. 38, no. 1, pp. 43–60, 2023, doi: 10.1021/acs.energyfuels.3c03104.
    [33] A. Hayakawa, Y. Hirano, E. C. Okafor, H. Yamashita, T. Kudo, and H. Kobayashi, "Experimental and numerical study of product gas characteristics of ammonia/air premixed laminar flames stabilized in a stagnation flow," Proceedings of the Combustion Institute, vol. 38, no. 2, pp. 2409–2417, 2021, doi: 10.1016/j.proci.2020.07.030.
    [34] H. Rong et al., "Thermal performances and NOx emission studies on an ammonia-hydrogen fueled double-channel outlet micro-combustor for thermophotovoltaic applications," Energy and Built Environment, vol. 6, no. 4, pp. 704–717, 2025, doi: 10.1016/j.enbenv.2024.02.007.
    [35] H. Zhao, D. Zhao, S. Becker, and Y. Zhang, "NO emission and enhanced thermal performances studies on Counter-flow Double-channel Hydrogen/Ammonia-fuelled microcombustors with Oval-shaped internal threads," Fuel, vol. 341, 2023, doi: 10.1016/j.fuel.2023.127665.
    [36] R. Swischuk, B. Kramer, C. Huang, and K. Willcox, "Learning physics-based reduced-order models for a single-injector combustion process," AIAA Journal, vol. 58, no. 6, pp. 2658–2672, 2020.
    [37] M. Cracco et al., Deep learning-based reduced-order methods for fast transient dynamics. 2022.
    [38] P. Pant, R. Doshi, P. Bahl, and A. Barati Farimani, "Deep learning for reduced order modelling and efficient temporal evolution of fluid simulations," Physics of Fluids, vol. 33, no. 10, 2021, doi: 10.1063/5.0062546.
    [39] A. A. V. Perpignan, R. Sampat, and A. Gangoli Rao, "Modeling Pollutant Emissions of Flameless Combustion With a Joint CFD and Chemical Reactor Network Approach," Frontiers in Mechanical Engineering, vol. 5, 2019, doi: 10.3389/fmech.2019.00063.
    [40] M. Falcitelli, S. Pasini, N. Rossi, and L. Tognotti, "CFD+reactor network analysis: an integrated methodology for the modeling and optimisation of industrial systems for energy saving and pollution reduction," Applied Thermal Engineering, vol. 22, no. 8, pp. 971–979, 2002/06/01/ 2002, doi: https://doi.org/10.1016/S1359-4311(02)00014-5.
    [41] S. Matteo, C. Alberto, Ward, and P. Alessandro, "Machine learning clustering algorithms for the automatic generation of chemical reactor networks from CFD simulations," Fuel, vol. 343, p. 127945, 2023, doi: https://doi.org/10.1016/j.fuel.2023.127945.
    [42] V. Paramarta, A. Rahman, L. Priska, R. Roken Gurning, and W. Ayu Purwati, "Comparative Analysis of K-Means and Gaussian Mixture Model in Clustering Global CO2 Emissions," bit-Tech, vol. 8, no. 1, pp. 998–1008, 2025, doi: 10.32877/bt.v8i1.2805.
    [43] A. Özden, R. M. Galassi, F. Contino, and A. Parente, "Clustering-based data-driven multi-fidelity reduced order modeling of ammonia combustion," Proceedings of the Combustion Institute, vol. 41, 2025, doi: 10.1016/j.proci.2025.105881.
    [44] Y. Bao, H. Du, W. S. Chai, D. Nie, and L. Zhou, "Numerical investigation and optimization on laminar burning velocity of ammonia-based fuels based on GRI3.0 mechanism," Fuel, vol. 318, 2022, doi: 10.1016/j.fuel.2022.123681.
    [45] "GRI-Mech 2.11." http://combustion.berkeley.edu/grimech/new21/version21/text21.html (accessed.
    [46] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, and M. Goldenberg. "GRI-Mech 3.0." http://www.me.berkeley.edu/gri_mech (accessed.
    [47] R. Li, A. A. Konnov, G. He, F. Qin, and D. Zhang, "Chemical mechanism development and reduction for combustion of NH3/H2/CH4 mixtures," Fuel, vol. 257, 2019, doi: 10.1016/j.fuel.2019.116059.
    [48] Pedro Azevedo, Isabel Cabrita, and C. Pinho, "REDUCED MECHANISM FOR COMBUSTION OF HYDROGEN AND METHANE WITH NITROGEN CHEMISTRY," Computational Thermal Sciences, vol. 6, pp. 541–551, 2014.
    [49] Ansys, Chapter 1: Introduction to Ansys Fluent.
    [50] 魏詠翔, "觸媒輔助氨燃料於微通道燃燒之數值模擬研究," 碩士, 航空太空工程學系, 國立成功大學, 台南市, 2025. [Online]. Available: https://hdl.handle.net/11296/b3udg8
    [51] T. E. ToolBox. "Emissivity Coefficients of Common Materials: Data & Reference Guide." https://www.engineeringtoolbox.com/emissivity-coefficients-d_447.html (accessed.
    [52] M. Warner, "The Kinetics of Industrial Ammonia Combustion," DOCTOR, School of Chemical and Biomolecular Engineering, The University of Sydney, 2013.
    [53] O. Deutschmann, L. I. Maier, U. Riedel, A. H. Stroemman, and R. W. Dibble, "Hydrogen assisted catalytic combustion of methane on platinum," Catalysis Today, vol. 59, no. 1, pp. 141–150, 2000/06/10/ 2000, doi: https://doi.org/10.1016/S0920-5861(00)00279-0.
    [54] Google. "Normalization." https://developers.google.com/machine-learning/crash-course/numerical-data/normalization?hl=zh-tw (accessed.
    [55] W. contributors. "k-means clustering." https://en.wikipedia.org/wiki/K-means_clustering (accessed.
    [56] K. P. Murphy, Machine Learning: A Probabilistic Perspective. The MIT Press, 2012.
    [57] F. Murtagh and P. Legendre, "Ward’s Hierarchical Agglomerative Clustering Method: Which Algorithms Implement Ward’s Criterion?," Journal of Classification, vol. 31, no. 3, pp. 274–295, 2014, doi: 10.1007/s00357-014-9161-z.
    [58] W. contributors. "Davies–Bouldin index." https://en.wikipedia.org/w/index.php?title=Davies%E2%80%93Bouldin_index&oldid=1351382857 (accessed.
    [59] W. contributors. "Silhouette (clustering)." https://en.wikipedia.org/w/index.php?title=Silhouette_(clustering)&oldid=1335669403 (accessed.
    [60] W. contributors. "Root mean square deviation." https://en.wikipedia.org/w/index.php?title=Root_mean_square_deviation&oldid=1345654531 (accessed.
    [61] R. Fagin, A. Lotem, and M. Naor, "Optimal aggregation algorithms for middleware," presented at the Proceedings of the twentieth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems, Santa Barbara, California, USA, 2001. [Online]. Available: https://doi.org/10.1145/375551.375567.
    [62] H. K. M. David G. Goodwin, Ingmar Schoegl, Raymond L. Speth, and B. B. Weber. "Cantera: An Object-oriented Software Toolkit for Chemical Kinetics, Thermodynamics, and Transport Processes." https://www.cantera.org (accessed.
    [63] 簡愷安, "利用雙氧水輔助氨氣於含空腔微管道內燃燒之模擬研究," 碩士, 機械工程學系, 國立成功大學, 台南市, 2026.
    [64] W. contributors. "Pearson correlation coefficient." https://en.wikipedia.org/w/index.php?title=Pearson_correlation_coefficient&oldid=1359658086 (accessed.

    QR CODE