近年來為因應全球淨零碳排目標，工業燃燒系統之減碳成為當前迫切議題。氨(NH3)作為零碳含能載體，具高氫密度、可液化儲運、以及成熟之合成與運輸基礎，被視為未來工業燃料之重要候選；然而氨之低層流燃燒速度、窄可燃極限與含氮燃料所生成之高fuel-NOx排放特性，使其難以單獨應用於現有工業燃燒系統，須與甲烷或氫氣混燒以提升火焰穩定性並抑制NOx生成。本研究於100 kW工業先導燃氣爐上，同步進行氨氣分區噴吹減NOx策略之發展，以及窺視孔光譜結合機器學習之燃燒性能預測模型建立。
實驗於成大歸仁校區100 kW工業先導燃氣爐進行，配置預混旋流燃燒機及不同噴吹管徑與軸向距離之噴吹管。爐內量測系統含K型熱電偶多通道排氣分析儀與傅立葉轉換紅外光譜儀，並配合爐頂視角相機以紀錄火焰形態。窺視孔光譜量測系統採用Ocean Insight Flame-S-UV-VIS光譜儀，透過光纖連接於燃燒機後方之既有窺視孔，於不干擾燃燒場之條件下擷取寬頻火焰化學發光光譜。熱電偶量測值於高溫近火焰區另以能量平衡模型進行輻射熱損修正。
氨氣噴吹實驗建立甲烷–氨氣預混燃燒基準案例，並探討混氨比(xNH3)、當量比(φ)、噴吹管徑(d)、噴吹軸向距離與火口直徑比(L/Db)及釋熱率(Q̇)對燃燒特性與排放之影響。預混結果顯示，隨混氨比提升，爐心溫度下降，CO與NO2排放上升，而NOx排放呈先升後緩之趨勢。相較於預混燃燒，氨氣噴吹可使氨氣避開主火焰高溫核心區，降低fuel-NO生成。較大之d可降低氨氣出口流速與噴流動量，其中d= 7.2 mm為較佳管徑，於xNH3 = 0.19條件下NOx較d= 3 mm明顯下降並較Pre-blend條件降低約80%；噴吹位置方面，L/Db = 1.5於xNH3 = 0.19時可使NOx降至約200 ppm，較Pre-blend減少約80%；然而，噴吹位置過度下移至L/Db = 2則因氨氣與主火焰混合不足而CO顯著上升。
光譜結合機器學習部分，本研究建置氫氣平均光譜資料庫、氨氣平均光譜資料庫與氨氣不平均光譜資料庫，並結合高斯過程回歸(GPR)、人工神經網路(ANN)與物理資訊神經網路(PINN)建立燃燒性能與排放預測模型；系統性比較基線扣除、正規化方式、輸入波段與模型架構對預測性能之影響。結果顯示，氨氣混燒資料庫因NH*、NH2*等含氮自由基與較複雜之反應路徑，採用基線扣除可有效提升預測精度；正規化策略則以全域OH*正規化全面優於逐筆正規化，顯示OH*之絕對強度保留了火焰放熱率、當量比與NOx生成等重要跨樣本資訊。輸入波段方面，氫氣混燒以200–550 nm為較佳波段，而氨氣混燒於當量比預測時需使用200–850 nm全波段以涵蓋NH*、NH2*等氨氣特徵訊號。模型比較顯示，小至中等樣本資料庫下GPR全面優於ANN；而於1,104筆之氨氣不平均資料庫下，GPR主導多數輸出，ANN於φ、O2、NO2反超 GPR，PINN之C元素守恆約束則使CO2預測取得最佳。
綜合而言，本研究之氨氣分區噴吹策略可有效降低甲烷–氨氣混燒之NOx排放，並建立噴吹管徑與噴吹位置之設計原則；同時，透過窺視孔光譜結合機器學習模型，可由工業爐既有窺視孔之有限視野光譜訊號預測燃燒操作條件與尾氣排放，10項主要輸出之測試集R²普遍≥0.83、9項≥ 0.90。研究成果可作為未來氨氣混燒系統低NOx 設計、工業燃燒即時監測與智慧化控制之參考依據。

In response to global net-zero carbon targets, ammonia (NH3) has been identified as a promising carbon-free fuel for industrial combustion owing to its high hydrogen density, mature global logistics infrastructure and the potential for green production from renewable electricity. Direct ammonia combustion, however, is hindered by a low laminar flame speed, narrow flammability range, high ignition temperature and elevated NOₓ emissions. Co-firing ammonia with methane partially mitigates these issues but introduces additional challenges, including fuel-bound NOₓ formation via NH₂/NH oxidation pathways, and the limited per-burner real-time monitoring capability of existing industrial systems that rely solely on stack-gas analysis. This study addresses both challenges on a 100 kW industrial pilot furnace through two coupled investigations.
The first investigation develops an ammonia-lancing strategy that reduces NOₓ emissions in methane–ammonia co-firing by injecting ammonia downstream of the main flame core, thereby suppressing fuel-NO formation through a SNCR-like reduction route. Three principal parameters—injection tube diameter d, axial distance L from the burner exit and thermal input Q̇—are varied across ammonia molar blending ratio xNH3 = 0-0.47 at equivalence ratios  = 0.90 and 0.95, and their effects on flame morphology, in-furnace temperature distributions and exhaust composition are systematically quantified.
The second investigation constructs a sight-port chemiluminescence spectroscopy framework combined with machine learning that predicts combustion performance and exhaust composition in real time. Three databases are evaluated under different baseline-removal, normalisation and wavelength configurations, and three ML models (Gaussian process regression (GPR), artificial neural network (ANN) and physics-informed neural network (PINN)) are compared. The optimised lancing configuration (d = 7.2 mm, L = 150 mm, Q̇ = 96 kW, xNH3 = 0.28) achieves an approximately 80% NOₓ reduction relative to premixed operation while maintaining CO below 50 ppm, and the proposed multi-model ML framework predicts ten combustion- performance and emission targets with R² ≥ 0.83 on the single-shot ammonia database, demonstrating that low-cost broadband spectroscopy combined with ML can substitute for parts of conventional gas analysis instrumentation and supports a closed-loop architecture for industrial ammonia co-firing systems.

[1] W.S. Chai, Y. Bao, P. Jin, G. Tang and L. Zhou (2021), A review on ammonia, ammonia-hydrogen and ammonia-methane fuels, Renewable and Sustainable Energy Reviews 147, 111254.
[2] H. Xiao, S. Lai, A. Valera-Medina, J. Li, J. Liu and H. Fu (2020), Study on counterflow premixed flames using high concentration ammonia mixed with methane, Fuel 275, 117902.
[3] G.B. Ariemma, G. Sorrentino, R. Ragucci, M. de Joannon and P. Sabia (2022), Ammonia/Methane combustion: Stability and NOx emissions, Combustion and Flame 241, 112071.
[4] T. Egawa, H. Nagahashi, A. Hayashi, S. Fukuba, K. Sato and S. Nakamura (2023), Hydrogen/Ammonia-fired Gas Turbine Initiatives for Carbon Neutrality, Mitsubishi Heavy Industries Technical Review 60(3), 1-10.
[5] S. Mashruk, M.O. Vigueras-Zuniga, M.E. Tejeda-del-Cueto, H. Xiao, C. Yu, U. Maas and A. Valera-Medina (2022), Combustion features of CH4/NH3/H2 ternary blends, International Journal of Hydrogen Energy 47, 30315-30327.
[6] E. Hu, Z. Huang, J. He, C. Jin and J. Zheng (2009), Experimental and numerical study on laminar burning characteristics of premixed methane-hydrogen-air flames, International Journal of Hydrogen Energy 34, 4876-4888.
[7] Y. Chen, B. Zhang, Y. Su, C. Sui and J. Zhang (2022), Effect and mechanism of combustion enhancement and emission reduction for non-premixed pure ammonia combustion based on fuel preheating, Fuel 308, 122017.
[8] I.A. Makaryan, I.V. Sedov, E.A. Salgansky, A.V. Arutyunov and V.S. Arutyunov (2022), A comprehensive review on the prospects of using hydrogen-methane blends: Challenges and opportunities, Energies.
[9] M.-D. Bloj, R.G. Ripeanu, A. Diniță, V.O. Oprea and M. Tănase (2025), Comprehensive review of hydrogen-natural gas blending: Global project insights with a focus on implementation and impact in Romanian gas networks, Heliyon 11, e43090.
[10] J. Arroyo, F. Tovar-Lasheras and A. Gil (2026), Experimental study on the combustion of methane-hydrogen mixtures in a pilot-scale furnace, Energy Conversion and Management 356, 121350.
[11] C.Y. Lien, Non-premixed Swirl Burner Flames of Hydrogen-blended Fuels, Master thesis, National Cheng Kung University, Taiwan, 2024.
[12] D. Wang, C. Ji, S. Wang, H. Meng, Z. Wang and J. Yang (2020), Further understanding the premixed methane/hydrogen/air combustion by global reaction pathway analysis and sensitivity analysis, Fuel 259, 116190.
[13] X. Zhu, A.A. Khateeb, W.L. Roberts and T.F. Guiberti (2021), Chemiluminescence signature of premixed ammonia-methane-air flames, Combustion and Flame 231, 111508.
[14] W. Boyette and A. Tulgestke, Determination of Trace Species Concentrations in Hydrogen and Ammonia Flat Flames Using Fourier Transform Infrared Spectroscopy, National Energy Technology Laboratory (NETL) Technical Report.
[15] X. Wei, M. Gu, S. Li, Y. Wu, M. Chen, J. Liang and Q. Lin (2024), Experimental and numerical studies of NO reduction by ammonia in different methane flame positions, Case Studies in Thermal Engineering 61, 105107.
[16] B. Breer, H. Rajagopalan, C. Godbold, H. Johnson II, B. Emerson, V. Acharya, W. Sun, D. Noble and T. Lieuwen (2023), Numerical investigation of NOx production from premixed hydrogen/methane fuel blends, Combustion and Flame 255, 112920.
[17] S. Schwarz, G. Daurer, C. Gaber, M. Demuth and C. Hochenauer (2024), Experimental investigation of hydrogen enriched natural gas combustion with a focus on nitrogen oxide formation on a semi-industrial scale, International Journal of Hydrogen Energy 63, 173-183.
[18] W. Pan, N. Yao, Y. Chen and L. Kang (2024), Numerical investigation of NOx emission characteristics in air-staged combustion system fueled by premixed ammonia/methane, Journal of the Energy Institute 117, 101857.
[19] S. Xu, C. Dou, S. Tian, L. Xi and H. Liu (2025), Effect of burnout air injection conditions on methane air-staged MILD combustion in a laboratory-scale furnace, Fuel 380, 133179.
[20] P. Stuber, S. Winkler et al. (2017), Staged combustion concept for gas turbines, Journal of the Global Power and Propulsion Society.
[21] Y. Tu, K. Su, H. Liu, Z. Wang, Y. Xie, C. Zheng and W. Li (2017), MILD combustion of natural gas using low preheating temperature air in an industrial furnace, Fuel Processing Technology 156, 72-81.
[22] M. Ferrarotti, W. De Paepe and A. Parente (2021), Reactive structures and NOx emissions of methane/hydrogen mixtures in flameless combustion, International Journal of Hydrogen Energy 46, 34018-34045.
[23] M. Ayoub, C. Rottier, S. Carpentier, C. Villermaux, A.M. Boukhalfa and D. Honoré (2012), An experimental study of mild flameless combustion of methane/hydrogen mixtures, International Journal of Hydrogen Energy 37, 6912-6921.
[24] G. Ali and Y. Zhou (2025), A detailed numerical analysis of NOx formation and destruction during MILD combustion of CH4/H2 blends using a skeletal mechanism, International Journal of Hydrogen Energy 103, 119-131.
[25] J. Cheng, B. Liu and T. Zhu (2024), Experimental investigation of NOx reduction by varying internal flue gas recirculation structures in non-premixed methane combustion, Thermal Science and Engineering Progress 55, 102973.
[26] M.A. Nemitallah, A. Abdelhalim, A. Abdelhafez and M.A. Habib (2024), Experimental and numerical study on flow/flame interactions and pollutant emissions of premixed methane-air flames with enhanced lean blowout hydrogen injection, International Journal of Hydrogen Energy 65, 14-32.
[27] P.W. Chang, Reduction of Nitric Oxides Emission from Hydrogen-Blended Methane and Propane Combustion via Hydrogen Lancing, Master thesis, National Cheng Kung University, Taiwan, 2024.
[28] M. Mayrhofer, M. Koller, P. Seemann, R. Prieler and C. Hochenauer (2021), Assessment of natural gas/hydrogen blends as an alternative fuel for industrial heat treatment furnaces, International Journal of Hydrogen Energy 46, 21672-21686.
[29] G. Coskun, O. Yalçınkaya, Z. Parlak, V. Tür, H. Pehlivan and E. Büyükkaya (2025), Investigation of the hydrogen-enriched methane combustion in a domestic boiler with porous burner on emissions and performance, Fuel 384, 134051.
[30] J. Ballester and T. García-Armingol (2010), Diagnostic techniques for the monitoring and control of practical flames, Progress in Energy and Combustion Science 36, 375-411.
[31] Y. Hardalupas and M. Orain (2004), Local measurements of the time-dependent heat release rate and equivalence ratio using chemiluminescent emission from a flame, Combustion and Flame 139, 188-207.
[32] C.S. Panoutsos, Y. Hardalupas and A.M.K.P. Taylor, Numerical evaluation of equivalence ratio measurement using OH and CH chemiluminescence in premixed and non-premixed methane-air flames, Combustion and Flame.
[33] M.E. Baumgardner and J. Harvey (2020), Analyzing OH, CH, and C2 chemiluminescence of bifurcating FREI propane-air flames in a micro flow reactor, Combustion and Flame.
[34] T. García-Armingol, J. Ballester and A. Smolarz (2013), Chemiluminescence-based sensing of flame stoichiometry: Influence of the measurement method, Measurement 46, 3084-3097.
[35] S. Yan, Y. Gong, Q. Guo, G. Yu and F. Wang (2024), Numerical study of CH chemiluminescence and heat release rate in methane inverse diffusion flame, Fuel 357, 129963.
[36] Y. Huang and Y. Yan (2000), Transient two-dimensional temperature measurement of open flames by dual-spectral image analysis, Transactions of the Institute of Measurement and Control 22(5), 371-384.
[37] P.R.N. Childs, J.R. Greenwood and C.A. Long (2000), Review of temperature measurement, Review of Scientific Instruments 71(8), 2959-2978.
[38] X. Li, Q. Huang, X. Luo and P. Wang (2022), Thermocouple correction method evaluation for measuring steady high-temperature gas, Applied Thermal Engineering 213, 118673.
[39] M.F. Modest (2013), Radiative Heat Transfer, 3rd edition, Academic Press.
[40] B. Leckner (1972), Spectral and total emissivity of water vapor and carbon dioxide, Combustion and Flame 19, 33-48.
[41] C.E.A. Finney, C.S. Daw, T.A. Fuller, T.J. Flynn and C.W. Kulp (2015), Opportunities for the next generation of optical boiler diagnostics, Proceedings of the American Flame Research Committee Industrial Combustion Symposium.
[42] G. Ronquillo-Lomeli and A.-I. García-Moreno (2024), A machine learning-based approach for flames classification in industrial Heavy Oil-Fire Boilers, Expert Systems With Applications 238, 122188.
[43] H. Cai, Z. Zhu and D. Zhou (2024), Study of tuyere combustion flame temperature in vanadium and titanium blast furnaces by machine vision and colorimetric thermometry, Metals 14, 499.
[44] H. Yang, Y. Fu and J. Yang (2022), Review of measurement techniques of hydrocarbon flame equivalence ratio and applications of machine learning, Measurement Science Review 22(3), 122-135.
[45] A. Müller, V. Ersoy, J. Menser, T. Endres and C. Schulz (2025), Real-time analysis of flame chemiluminescence spectra for equivalence ratio and gas composition using neural network approaches, Applications in Energy and Combustion Science 23, 100345.
[46] Y. Zhang, P. Zhao, X. Liu, W. Xiong, Y. Lai, M. Davies, J.R. Willmott and J. Yang (2025), Multi-feature engineering and machine learning for equivalence ratio prediction in methane premixed flames, Engineering Letters 33(6).
[47] L. Zheng, T. Yang, W. Liu, Y. Lai and J. Yang (2024), Enhancing accuracy of flame equivalence ratio measurements: An attention-based convolutional neural network approach for overcoming limitations in traditional color modeling, Sensors.
[48] T.F. Guiberti, N.N. Shohdy, S. Cardona, X. Zhu, L. Selle and C.J. Lapeyre (2023), Chemiluminescence- and machine learning-based monitoring of premixed ammonia-methane-air flames, Applications in Energy and Combustion Science 16, 100212.
[49] L. Mazzotta, X. Zhu, J. Davies, D. Sato, D. Borello, S. Mashruk, T.F. Guiberti and A. Valera-Medina (2025), Assessing the potential of a chemiluminescence and machine learning-based method for the sensing of premixed ammonia-hydrogen-air turbulent flames, International Journal of Hydrogen Energy 100, 945-954.
[50] N. Agwu, J. Davies, D. Sato, S. Mashruk and A. Valera-Medina (2025), Machine learning driven chemiluminescence-based modelling of combustion parameters in premixed swirling NH3/H2 flames, International Journal of Hydrogen Energy 145, 717-731.
[51] M. Mousavi, C. Caldwell, J. Baltes, F.A. Parizad, M. Aljasem, B.J. Lee and N. Karimi (2026), Physics-informed neural networks in clean combustion: A pathway to sustainable aerospace propulsion, Chemical Engineering Research and Design 226, 258-281.
[52] A. Taassob, R. Ranade and T. Echekki (2023), Physics-informed neural networks for turbulent combustion: Toward extracting more statistics and closure from point multiscalar measurements, Energy & Fuels 37, 17484-17498.
[53] F. Frohberg, P. Kandel and A. Ghani (2025), Physics-informed neural networks for reacting flows: Species reconstruction with finite rate chemistry from sparse and noisy velocity measurements, Proceedings of the Combustion Institute 41, 105957.
[54] C.E. Rasmussen, Gaussian processes in machine learning, in Advanced Lectures on Machine Learning, Springer.
[55] T. Akiba, S. Sano, T. Yanase, T. Ohta and M. Koyama (2019), Optuna: A next-generation hyperparameter optimization framework, Proceedings of the 25th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2623-2631.
[56] Y. Liu, G. Wang, Y. Li, S. Xu, X. Chen and D. Luo (2026), Machine learning prediction of flame temperatures in ammonia combustion within a porous burner, International Communications in Heat and Mass Transfer 172, 110270.