Multi-objective optimization of swirl burner with premixed ammonia/hydrogen flame

Abstract

In this thesis, the objective is to numerically investigate, validate, and optimize a premixed ammonia-hydrogen swirl burner. The fundamental goal is to establish the necessary design criteria that would allow ammonia, which is gaining prominence in the transition toward sustainable energy resources, to be used in industrial applications in an environmentally friendly, safe, and efficient manner. Although the literature includes numerous studies on the combustion characteristics of ammonia–hydrogen blends, most of these investigations remain either experimentally limited or constrained solely to numerical modeling. This study aims to fill this gap by combining a CFD-based approach, detailed validation with experimental data, and multi-objective optimization within a single unified framework. The originality of the work arises from the integration of CFD, validation, and optimization. The role of ammonia in global energy strategies has increased significantly in recent years. Particularly in the processes of reducing carbon emissions and establishing the hydrogen economy chain, ammonia is considered a strategic energy carrier due to its transportability, ease of storage, and carbon-free nature. Its existing large-scale production infrastructure and transportation network help overcome many of the storage and logistical challenges associated with hydrogen. For this reason, ammonia has been positioned as an early-stage transition fuel in the hydrogen economy in many countries, playing a critical role in the shift toward carbon-neutral systems. When transported in liquid form, its high hydrogen content per unit volume and ease of storage make it an attractive option for various energy applications, especially industrial heating, boiler systems, and internal combustion engines. While hydrogen requires either cryogenic storage or high-pressure tanks, ammonia can be liquefied at temperatures close to atmospheric pressure, enabling a less costly and safer distribution infrastructure. However, ammonia's slow reaction rate, long ignition delay, high NOx formation tendency, and unburned ammonia emissions constitute major drawbacks, particularly in pure ammonia flames, where they adversely affect flame stability and environmental performance. Therefore, burner systems that employ ammonia–hydrogen blends must be evaluated in detail with appropriate geometries, suitable mixing ratios, and validated kinetic mechanisms. In particular, NOx formation is one of the most critical environmental parameters in ammonia-based combustion systems. NOx is both a strong air pollutant and toxicologically hazardous. In ammonia combustion, NOx formation intensifies especially in regions with high temperature and excess oxygen. Hence, controlling the flame shape and heat-release profile in ammonia–hydrogen blends is of great importance to meet environmental regulations. The scope of this study is defined as the development of an integrated numerical approach involving cold-flow validation, reacting-flow analyses, kinetic mechanism comparisons, and multi-objective optimization. This integrated structure provides a unique contribution, as cold-flow validation, chemical mechanism evaluation, and design optimization—typically treated separately in different studies—are applied together for the first time in the same ammonia–hydrogen swirl burner. All models and evaluations used throughout the study were conducted based on a real-scale industrial swirl burner, thereby increasing the practical relevance of the results. In the first stage of the thesis, a laboratory-scale swirl burner system characterized experimentally in the literature was modeled numerically, aiming to validate the numerical results using cold-flow experiments. For this purpose, LDA measurements available in the literature and obtained at laboratory scale were used. The formation, location, and size of the central recirculation zone—one of the most important parameters determining flame stability in swirl burners—were examined using both experimental data and CFD solutions. The numerical analyses were performed using Converge CFD®. In the cold-flow CFD analyses, six different turbulence models were compared based on LDA measurements, and the second-order k–ω SST-based RANS approach was found to best represent the experimental data. The selected turbulence model accurately predicted axial velocity profiles near the center and successfully captured turbulence characteristics away from the center. Moreover, by applying an AMR strategy, the mesh was refined only in regions with complex flow features, improving accuracy while reducing computational cost. The results demonstrated that the cold-flow model provides a reliable foundation for the reacting-flow solutions. In the reacting-flow analyses, three different reaction mechanisms—Duynslaegher, Nakamura, and Stagni—were examined based on temperature profiles and emission values to validate the model. Comparisons of these kinetic mechanisms were made alongside experimental literature. Among them, the Nakamura mechanism exhibited the highest agreement with experimental data in terms of both temperature distribution and product species profiles. Additionally, it produced more realistic predictions of NOx compared to the Duynslaegher mechanism and yielded more consistent NH₃ slip predictions, particularly at low H₂ fractions. Therefore, the Nakamura mechanism was selected for the validated combustion analysis models used in the optimization phase. In summary, the validation of the optimization model was assessed under both cold-flow conditions (velocity comparison with LDA data) and reacting-flow conditions (comparison of flame location and temperature distribution), and good agreement with experimental measurements was achieved in both cases. A multi-objective optimization model was developed to systematically explore the design space of the swirl burner using the validated CFD model. Different multi-objective optimization algorithms commonly used in the literature were evaluated. Considering the limited number of analyses, the high cost of validated CFD computations, and physical requirements, the NSGA-II algorithm was selected due to its advantages such as rapid convergence, balanced solution distribution, and low computational cost. Four design parameters were considered in the optimization: confinement ratio, vane number, vane angle, and fuel blend ratio. The optimization was based on three objective functions: increasing specific heat release per unit mass of fuel, reducing NO₂ formation, and minimizing hydrogen consumption. Since these three objectives conflict with one another, the best design is not represented by a single solution but rather by a family of solutions located along the Pareto front. The initial population consisted of 14 randomly selected individuals, each defined within the minimum and maximum allowable parameter limits. Evolutionary processes including crossover, mutation, and ranking were then applied. Due to limitations on the number of analyses, the population size and number of iterations were carefully balanced. In the fuel-mixing optimization, the equivalence ratio and total thermal power were kept constant, while the volumetric ratio of ammonia–hydrogen was varied to evaluate combustion characteristics. With the integration of Converge CFD® and Dakota, CFD analyses for all candidate individuals were executed automatically. All individuals were normalized and compared, and the resulting Pareto fronts clearly illustrated the relationships between design variables and objective functions. The results showed that in some regions of the design space, increasing heat-release rate inevitably increased NO₂ levels, whereas in other regions, obtaining low NO₂ required higher hydrogen consumption. This clearly demonstrates that achieving a high-efficiency, low-emission, and low-hydrogen-consuming design in ammonia–hydrogen burners inherently requires multi-objective trade-offs. Because the design parameters addressed in the optimization directly influence the physical behavior of the burner, the manner in which each parameter shaped the design space was also evaluated. The confinement ratio plays a critical role in flame stability and NO2 formation because it determines swirl intensity and the size of the recirculation zone. Increasing the confinement ratio enhances the mass of hot products recirculated toward the burner center, raising temperatures in the flame root and influencing both specific heat release and NO₂ formation. The vane number and vane angle are fundamental geometric parameters that determine swirl strength and mixing quality. Particularly at high vane angles, an increase in swirl frequency promotes mixture homogenization but also broadens the reaction zone, which in turn affects heat release and NO₂ emissions. The role of the fuel blend ratio in the optimization is also one of the significant findings of the study. Increasing the hydrogen fraction in ammonia–hydrogen blends typically elevates reaction rates and stabilizes combustion. However, this also increases hydrogen consumption, posing disadvantages from economic and sustainability standpoints. Therefore, optimizing the hydrogen fraction is essential not only for performance but also for operational cost. The optimization results revealed that a performance-oriented approach and an approach targeting reduced hydrogen consumption correspond to substantially different regions of the design space. Thus, the presence of a Pareto front provides flexibility to select a solution based on user-defined priorities. According to the optimization results, the best physically feasible solution was identified as D052. It offers high performance in terms of specific heat release while keeping NO2 emissions at acceptable levels and ensuring reasonable hydrogen consumption. In the final part, the effects of boundary conditions such as inlet temperature and inlet pressure on D052 were evaluated. These analyses are critical for understanding how flame behavior changes under varying operating conditions in industrial systems. Increasing inlet temperature strengthened flame stability, expanded the reaction zone, and elevated heat release. In contrast, increasing inlet pressure significantly increased NO2 formation, indicating that ammonia–hydrogen blends in high-pressure systems require more careful control. The parametric studies confirmed that the low-emission and stable characteristics of D052 make it a suitable reference geometry for such analyses. The results demonstrate that system design must carefully consider not only geometry and fuel composition but also boundary conditions. The observation that NOx formation increases beyond a certain inlet temperature suggests the need for determining an optimal inlet temperature in future studies. The influence of inlet pressure exhibited an even more pronounced chemical sensitivity than temperature. Pressure increase accelerated NO₂ formation mechanisms, as higher pressure increases molecular collision rates and elevates reaction rates of intermediate species that convert to NOx. Therefore, even mixtures with nominally low hydrogen content tend to produce substantial NO₂ under high-pressure conditions. This finding indicates that ammonia–hydrogen blends must be optimized particularly carefully for systems operating at elevated pressures. Moreover, increasing pressure narrowed the reaction zone, significantly altering turbulence–chemistry interactions. The accurate numerical capture of this physical behavior demonstrates the reliability of the combustion model under high-pressure conditions. Another important outcome of these analyses is the observation that operating conditions are as critical as geometric optimization. Focusing solely on geometry or fuel composition during design is insufficient; inlet pressure, inlet temperature, and the overall operating window must also be optimized simultaneously to achieve minimum NO2, high stability, and high efficiency. Therefore, future industrial applications of ammonia–hydrogen blends may require adaptive control strategies and feedback-based combustion control systems tailored to operating conditions. In conclusion, this thesis presents a unique numerical framework that holistically examines the combustion behavior of ammonia–hydrogen blends in swirl burners. Cold-flow validation showed strong agreement with experimental data; reacting-flow analyses demonstrated accurate prediction of temperature profiles; and multi-objective optimization generated new design recommendations for various applications. The insights provided by the model may contribute to the design of future ammonia-fueled systems that are more efficient, lower in emissions, and safer. Controlling NO2 formation is a critical requirement for sustainable combustion technologies, and the findings of this study provide an important foundation in this regard. Overall, the proposed approach constitutes a concrete step toward integrating ammonia into low-carbon energy systems, paving the way for high-performance and environmentally friendly combustion solutions through appropriate blending strategies, balanced geometry, and carefully managed operational conditions. This study also highlights the need for further development of kinetic mechanisms used for ammonia–hydrogen burners, as well as the limitations of existing RANS-based turbulence models, which may be insufficient in certain cases. Advanced turbulence models may yield more accurate results. In the future, the use of AI-assisted optimization methods may allow faster evaluation of complex mechanisms and more efficient exploration of intricate design spaces. Moreover, the application of LES, DNS, or hybrid models may contribute to a deeper understanding of turbulence–chemistry interactions. In the long term, the development of high-fidelity numerical models may accelerate the industrial adoption of ammonia-based combustion systems and significantly contribute to carbon reduction in the energy sector.