Development of a Low-NO_x Fuel-Flexible and Scalable Burner for Gas Turbines

Abstract

To reduce dependence on fossil fuels, gas turbine plants using hydrogen/methane blends provide a crucial solution for decarbonizing thermal power generation and promoting a sustainable energy transition. In this context, the development of fuel-flexible burners is fundamental. This work reports the development of a novel burner geometry for gas turbines that can operate with natural gas and hydrogen mixtures (HENG, hydrogen-enriched natural gas) over a wide range of hydrogen content while maintaining low NO_x emissions. The methodology used in this work is multidisciplinary, incorporating (i) CFD numerical simulations to determine the burner’s geometry, (ii) mechanical design for prototype construction (not discussed in the article), and (iii) experimental tests to assess its hydrogen content capacity, stabilization, and pollutant emission characteristics. The geometry was initially optimized through several RANS simulations to enhance reactant mixing and minimize flashback risks. Additionally, some LES simulations were conducted under specific conditions to achieve more accurate predictions and investigate potential combustion dynamics issues. The proposed solution was then transferred into a prototype. Through experimental testing, the burner prototype was characterized in terms of four key performance indicators: (1) the ability to operate with HENG mixtures with more than 20% ${\mathrm{H}}_2$ content, showing a technological trend exceeding 50%; (2) the ability to operate with low ${\mathrm{NO}}_{\mathrm{x}}$ (<25 ppm) and CO emissions within the 30–70% hydrogen volume range; (3) the ability to ignite HENG mixtures with ${\mathrm{H}}_2$ in the 30–70% hydrogen volume range; and (4) the ability to operate with a fluctuating hydrogen content, ±15% over time, while still complying with ${\mathrm{NO}}_{\mathrm{x}}$ and CO emission limits.

1. Introduction

With the aim of reducing reliance on fossil fuels, gas turbine plants that utilize hydrogen/methane blends offer a vital solution for decarbonizing the thermal power generation sector and supporting the sustainability of the energy transition [1]. However, it is crucial to highlight that to significantly lower ${\mathrm{CO}}_2$ emissions, the focus must be on blends with a high hydrogen content. Currently, discussions about fuel flexibility refer to a gas turbine’s ability to operate with hydrogen-enriched blends, which include hydrogen mixed with other gaseous fuels. These blends range from hydrogen-enriched natural gas (HENG) to ammonia, which is a promising carrier for hydrogen.

Despite their tendency to exhibit thermo-acoustic instabilities (pressure oscillations due to the coupling between pressure waves associated with the system’s acoustics and heat release fluctuations), DLE (dry low emissions) combustion technologies are the state of the art to operate gas turbines with pollutant emissions below the limits set by the European Industrial Emission Directives (IEDs). The maximum allowed ${\mathrm{H}}_2$ concentration in commercially available DLE gas turbines depends on the class and the specific combustion technologies adopted by manufacturers [2,3]: on average, 44–63% (by volume) for heavy-duty turbines (100–500 ${\mathrm{MW}}_{\mathrm{e}}$), 43–55% for industrial-use turbines (30–100 ${\mathrm{MW}}_{\mathrm{e}}$), 35% for aero-derivative turbines (1–30 ${\mathrm{MW}}_{\mathrm{e}}$), and 20–32% for microturbines (0.1–1 ${\mathrm{MW}}_{\mathrm{e}}$).

The variation of the maximum allowable concentration of hydrogen in DLE gas turbines arises from differences in combustion temperatures and technologies across different turbine classes. Gas turbine manufacturers are focusing on developing combustion strategies for high-hydrogen content blends that result in low ${\mathrm{NO}}_{\mathrm{x}}$ emissions [4]. Currently, ${\mathrm{NO}}_{\mathrm{x}}$ emissions are often managed by implementing a de-${\mathrm{NO}}_{\mathrm{x}}$ module or derating the machine, which involves reducing its power output. Moreover, incorporating hydrogen into natural gas impacts the start-up and shutdown procedures of the turbines. Additionally, these machines must maintain a suitable level of operational flexibility, enabling them to quickly and safely adjust the power output to respond to fluctuations in energy demand.

As of 2024, the most flexible machines on the market are medium-to-large size turbines, including the NovaLT (5–20 ${\mathrm{MW}}_{\mathrm{e}}$) by Baker Hughes, which is capable of reaching 100% ${\mathrm{H}}_2$ with the implementation of a de-${\mathrm{NO}}_{\mathrm{x}}$ module to limit emissions, and the GT-36 (538 ${\mathrm{MW}}_{\mathrm{e}}$) by Ansaldo Energia, which was certified up to 70% ${\mathrm{H}}_2$ without the need for de-rating and tested up to 100% ${\mathrm{H}}_2$ but at reduced power to control NOx emissions. As for microturbines, there are examples of combustion tests up to 100% ${\mathrm{H}}_2$ (for instance, Ansaldo Energia up to 80% for the AE-T100, DLR up to 100% on a modified TURBEC T100 PH, Genova, Italy), but the current commercial state of the art remains at 20–32%.

There are different technologies on which the current combustors of the various manufacturers are based; among these technologies, the most promising appear to be the ones based on the micro-mixing concept, the sequential combustion, and the ones implementing a trapped vortex configuration.

MITSUBISHI, GE, and KAWASAKI exploit the principle of micro-mixing, in which a very rapid mixing between the fuel and oxidizer is achieved by using a large number of very small injectors, and the flame is divided into many small, diffusive, partially premixed or premixed flames. This makes it possible to reduce the formation of nitrogen oxides by reducing the residence time of reactants in the high-temperature regions by shortening the flame. GE has developed the DLN 2.6e multi-pipe mixer system, which features multiple air parallel channels with fuel injections arranged in a crossflow configuration [5]. Because air velocity exceeds the flame speed, the risk of flashback is minimized, even when using mixtures with high hydrogen content, and the short tube length leads to a low-pressure drop. The DLN 2.6e burner was tested under H-class conditions in a facility with up to 50% hydrogen by volume and integrated into a full-can combustion system, showing excellent performance in terms of pollutant emissions. A scaled-up multi-nozzle combustor was operated up to 90% of hydrogen, accumulating over 100 h of operation, with ${\mathrm{NO}}_{\mathrm{x}}$ emissions below 10 ppm (corrected to 15% ${\mathrm{O}}_2$). The MITSUBISHI-HITACHI design [6,7] features multiple clusters, each containing several coaxial or crossflow fuel and air injections. Since the jets are strategically directed to create a converging and diverging swirl flow, each cluster produces its own lifted flame. A central pilot further enhances combustion stability. As fuel and air are partially mixed over a short distance and the flame is lifted, this burner design takes advantage of low-${\mathrm{NO}}_{\mathrm{x}}$ premixed combustion and is at the same time flashback resistant. The clusters are fueled in groups one after the other until the full load is reached. MITSUBISHI-HITACHI currently guarantees operation with 30% hydrogen by volume and is aiming for full 100% hydrogen capability. In the KAWASAKI Heavy Industries burner [8,9,10,11], the airflow passes through small openings while fuel is injected perpendicularly (crossflow configuration). The resulting fuel/air mixture enters the combustion chamber, forming an inner and an outer vortex that stabilizes the flame. Two critical parameters that affect the flame shape and ${\mathrm{NO}}_{\mathrm{x}}$ formation have been identified: the airflow blockage ratio and the penetration depth of the fuel jet into the airstream. The design offers inherent safety against flashback and low ${\mathrm{NO}}_{\mathrm{x}}$ emissions due to the very short residence time of reactants in the micro-flame region. Load adjustments are managed by igniting progressively three rings, each containing many air passages and fuel injections: from idle to 30% load, only the inner two hydrogen rings are activated, and from 30% load to full load, all three rings are used. The burner demonstrated to operate on up to 100% hydrogen in the M1A-17 gas turbine under commercial conditions, while ${\mathrm{NO}}_{\mathrm{x}}$ emissions remain below 20 ppm (at 15% ${\mathrm{O}}_2$ and two bar) regardless of the load.

ANSALDO Energia [12] uses sequential combustion, which allows ${\mathrm{NO}}_{\mathrm{x}}$ emissions to be controlled by appropriately balancing the fuel injection along the path of the working fluid, thus acting on the reactivity and combustion temperature of the individual stages. The first stage is characterized by a lean mixture where the combustion is aerodynamically stabilized, while in the second, the combustion is controlled by self-ignition, and the fuel is injected into the flow of hot gases coming from the first stage. THOMASSEN Energy has developed a premixed combustion technology called “FlameSheetTM” [13], where the flame is aerodynamically stabilized through a trapped vortex. It provides a swirled premixed pilot and a main combustor that can be operated independently. Relative to the main, the mixture undergoes a notable aerodynamic stretching which prevents the pre-ignition of the reactants through a 180° rotation, generating a strong recirculation zone, which represents the aerodynamically trapped vortex that stabilizes the flame.

It is important to note that the accepted ${\mathrm{H}}_2$ content and the level of ${\mathrm{NO}}_{\mathrm{x}}$ emissions are not the only factors that determine the actual flexibility of a gas turbine. Other indicators (key performance indicators) [14] must also be considered: the maximum reduction in electrical efficiency, the maximum ${\mathrm{H}}_2$ content during startup, the minimum ramp-up speed (relative to nominal power), and the fluctuation of ${\mathrm{H}}_2$ accepted (% of ${\mathrm{H}}_2$ by volume or mass per minute compared to the input value). The last three indicators contribute to the dynamic flexibility of the gas turbine: their state of the art in 2020 [14] was, respectively, 5% by volume of ${\mathrm{H}}_2$, 10% of nominal load per minute (at 30% by volume of ${\mathrm{H}}_2$), and ±10% by volume of ${\mathrm{H}}_2$ per minute; the targets to be reached by 2024 were 20%, still 10% (at 70% by volume of ${\mathrm{H}}_2$), and ±15%; the targets to be reached by 2030 are 100%, still 10% (at 100% by volume of ${\mathrm{H}}_2$), and ±30%. The target for ${\mathrm{NO}}_{\mathrm{x}}$ emissions until 2030 remains constant: 25 ppmv (at 15% dry ${\mathrm{O}}_2$). Last, but not least, gas turbine operation must be ensured without knowing the inlet composition of the fuel mixture.

The present results come from an Italian three-year research project (see funding details at the end of the article) aiming to identify a low-emission fuel-flexible burner technology for microturbines (more specifically, the TURBEC T100) with scalable solutions to larger gas turbines, build a prototype, and test it in an off-machine test facility (TRL 3-4), proving the following while respecting emission limits:

• The operability with a hydrogen content of at least 20% by volume (state-of-the-art for microturbines of around 100 kWe);
• The potential to demonstrate a technological trend that allows reaching and exceeding 50% by volume.

These activities are part of a broader work plan aimed at designing and realizing, in a subsequent research program, a 330 kWt prototype to be installed in the TURBEC T100 microturbine (Ansaldo Green Tech AE-T100, Genoa, Italy) at the AGATUR plant (TRL 5), which is equipped with the CCT Laboratory at ENEA Casaccia. The constraints to be respected for defining the new burner are therefore clearly linked to the specific microturbine:

• Thermal power of 330 kWt (100 kWe);
• Operating pressure of 4.5 bar;
• Space constraints for accommodating the new burner and its liner;
• Geometric configuration of the air supply system.

In addition, since considering microturbines as an application makes sense in the context of decentralized energy systems, but it is inherently limiting, the proposed solution must have scalability characteristics toward larger gas turbines.

Hence, the goal of the project is to develop a burner capable of reconciling conflicting needs, transitioning from a fuel composed solely of natural gas to one with a high hydrogen content. In the former case, it must allow complete combustion of natural gas, while in the latter, it must prevent flame flashback in the fuel system, ensuring low emissions of nitrogen oxides (${\mathrm{NO}}_{\mathrm{x}}$) and carbon monoxide (CO) in both scenarios. The present article offers a synthesis of the development of such a fuel-flexible burner for gas turbines with the goal of reaching 0–100% ${\mathrm{H}}_2$ in natural gas blends: the results will be the starting point for another project, aiming to enhance the key performance indicators reached in the first project.

The originality of the article consists of the rare opportunity to look simultaneously at the various aspects and methodologies involved in the design process of a new technology from its conceptualization to its prototyping. The originality of the investigated fuel-flexible burner solution is in the coexistence of two independent lean-premixed swirl burners, respectively, for low and high hydrogen content, and in the integration of the aerodynamic flame stabilization with the micro-mixing concept for the latter. The methodology adopted in this work is multi-disciplinary, involving the following: (1) CFD numerical simulations to define the burner’s geometry, (2) mechanical design to build its prototype, and (3) experimental tests to verify its actual hydrogen content capabilities, its stabilization and pollutant emission characteristics (i.e., the key performance indicators aimed by the project).

The manuscript is structured as follows. Section 2 describes the final burner configuration. An overview of the burner geometry development phase, based on a CFD approach, is provided in Section 3: the final geometry was identified by means of several non-reactive RANS simulations, which were aimed at verifying the flow rates of reactants through their distribution channels and optimizing the mixing between the fuel mixture and air; a more accurate non-reactive LES simulation was adopted to confirm the reliability of the predicted mixing; then, some reactive RANS and LES simulations were performed to look at flame stabilization and pollutant emissions. Section 4 reports the experimental campaign on the built prototype: attention is posed on flame stabilization and topology, pollutant emissions, flashback and blowout occurrence to identify the equivalence ratio range at which the burner can be operated when changing the hydrogen content in the HENG mixture.

2. The Burner Configuration

An entirely original construction and operating strategy has been developed (Figure 1). This strategy features two independent lean-premixed swirl burners with flame aerodynamic stabilization: a central burner (330 kWt) designed for low hydrogen content mixtures (Figure 2a) and a coaxial crown burner system consisting of ten 33 kWt burners for high hydrogen content mixtures (Figure 2b). This setup enables ${\mathrm{NO}}_x$ emissions to be limited by adopting an appropriate equivalence ratio (DLE strategy), which helps control the adiabatic flame temperature. By managing the flow rates through an active control system, the gas turbine’s operation can be adjusted based on the hydrogen content in the incoming HENG mixture and the machine load.

This article focuses on the development and initial characterization of one of the burners in the coaxial crown system (Figure 2b), which was specifically designed for high hydrogen content (nominally over 50% by volume). To mitigate flashback risks, the burner must be sized to ensure sufficiently high efflux velocities and low residence times in the mixer. To achieve these characteristics, the crown system burners were designed by combining a typical swirl configuration, providing flame aerodynamic stabilization through central recirculation or vortex breakdown, with micro-mixing concepts. The central burner used for natural gas or low hydrogen content experiences fewer issues and can therefore be designed for lower efflux velocities, allowing the complete combustion of higher natural gas content mixtures. The burners’ configuration has been designed to be similar to each other but with the consideration that the smaller burners allow for higher efflux velocities. The similarity between the two types of burners, once the geometric configuration of one of the side burners is defined, significantly reduces the study of the air and fuel mixing system in the larger and less problematic central burner.

The crown system features an even number of burners with alternating swirl flow directions, clockwise and counterclockwise, ensuring that the velocity vectors in the contact zone between two adjacent burners align in the same direction. In its final design (Figure 2b), the burner consists of 18 inclined channels for airflow. Fuel is injected into each channel at two opposing points (with a hole diameter of 0.75 mm), allowing the fuel to penetrate a significant portion of the airflow without diffusing throughout the entire channel height. This dual injection system requires two fuel plenums: the main fuel inlet supplies the lower plenum, which connects to the upper plenum through channels situated between the air conduits. The high number of channels and injection holes enhances mixing through distributed fuel injection.

The fuel and oxidizer mix within a sufficiently long premixing chamber (20 mm) to ensure uniform fuel concentration. The initially cylindrical chamber narrows toward the outlet, which helps accelerate the flow (with an exit diameter of 15 mm). The tangential velocity component initiates rotation in the flow, achieving a swirl flow number (SN) of 1 at the burner outlets. This rotation also promotes the effective and rapid mixing of air and fuel within the premixing chambers. A cylindrical body at the center of the chamber prevents the recirculation of burnt gases due to the depression produced by the mixture’s rotation, thus preventing premature ignition.

The first prototype of the crown system burner was fitted with an additional independent diffusive fuel injection system (with a hole diameter of 0.6 mm) positioned just before the outlet section. Diffusive hydrogen injection has already been adopted in various micro-mixing solutions. The added complexity of this diffusive system allows us to experimentally evaluate its potential for very high concentrations of hydrogen in the fuel mixture. The diffusive injection system consists of two separate plenums separated by a perforated ring to ensure uniform fuel delivery to the injection holes. note that the fuel supply to the first lower plenum is not shown in the figure.

The final burner installed in the laboratory is the product of the assembly of several individual elements fitted together through guide teeth and/or screwed together. To ensure uniform filling of air in the 18 channels, a plenum divided into three parts separated by perforated distribution rings has been provided (Figure 3a). The constructed and machined prototype is shown in Figure 3b: the burner is on the left (about the diameter of a €1 coin), while it is assembled in its feeding system on the right.

3. Overview of the CFD Development Phases

The sizing and optimization of the burners were performed using RANS (Reynolds-Averaged Navier–Stokes) simulations with Ansys-Fluent 2019R1 software, focusing on mixing performance. The computational domain is reported in Figure 4, which depicts a single burner of the crown system uniformly supplied by air and a portion of 36° of the combustion chamber delimited by symmetry planes. The computational grid consisted of approximately five million cells; since the design process planned several numerical simulations, a mesh sensitivity analysis was initially performed to identify the most convenient mesh resolution in terms of accuracy, grid independence results, and computational cost. In Figure 5, three different meshes are reported with an increasing number of computational cells: 3, 5, and 8 million. Panel d of the same picture reports the comparison between the radial profiles of the equivalence ratio at the outlet of the mixing chamber of the burner for the three meshes. Given the high number of simulations to be performed, it was decided to choose the mesh with 5 million elements, which is sufficiently reliable to reduce the computational burden. The “k-$ϵ$ realizable” turbulence model [15] was utilized, due to its robustness, in conjunction with the Eddy Dissipation Concept (EDC) model [16], incorporating a detailed chemical mechanism [17] featuring 46 chemical reactions and 17 species for the reactive simulations. Transport properties were calculated using kinetic gas theory with simulations conducted at an operating pressure of 4.5 bar. A pressure-based solver, the SIMPLE scheme for pressure-velocity coupling, and a second-order upwind scheme for spatial discretization were adopted. The air mass flow was always fixed at 0.02207 kg/s and 834 K, while the fuel mass flow, at 300 K, changed according to the mixture composition and equivalence ratio (0.000569 for 50% ${\mathrm{H}}_2$ and 0.000275 kg/s for 100% ${\mathrm{H}}_2$) for fixed thermal power.

The representative burner of the crown system was analyzed for mixtures containing 50% and 100% hydrogen at a nominal power level of 33 kWt while maintaining a constant air flow rate. It was observed that the average nominal equivalence ratio ($\Phi$) decreased with increasing hydrogen content: 0.5 for 50% ${\mathrm{H}}_2$ and 0.43 for 100% ${\mathrm{H}}_2$ (the values of $\Phi$ provided here are preliminary).

The goal was to design a burner that optimally mixes the reactants, minimizes the risk of flashback, and achieves low ${\mathrm{NO}}_{\mathrm{x}}$ emissions. After identifying the optimal configuration, heat transfer through the burner material was also simulated to account for the actual conditions within the microturbine; indeed, in the TURBEC T100 micro-gas turbine, air enters the combustor at 834 K, while the fuel is supplied to the machine at ambient temperature. Considering heat transfer through the solid material is crucial to evaluate the effect of inevitable fuel preheating before actual mixing: simulations revealed that the fuel heats up and decreases in density as it flows through the conduits and feeding plenum, significantly affecting its distribution among the injection ports.

The evolution of the burner model in the crown system, designed for high concentrations of hydrogen, began with its first geometric configuration. Initially, a simple swirling configuration was selected, including a relatively long premixing chamber without a central body, which was fed by inclined air channels with fuel injection at the bottom. The velocity field at the outlet prevented the return of hot gases, and the length was sufficient to ensure effective mixing. During the initial assessment of the dimensions and footprint of the central burner, it became clear that the lateral burners needed to be very short. This reduction in length created a risky central depression that could lead to the backflow of hot gases into the premixing chamber. Consequently, a central body was added, with its width optimized to achieve a sufficiently high axial velocity profile at the outlet, especially in the adjacent area. However, this configuration revealed that the length of the premixing chamber was inadequate to ensure uniform mixing between the fuel and air. For this reason, an additional fuel injection was added at the top of the air channels. With only the lower injection, the fuel tended to stratify around the central body, whereas with a potential upper injection, it tended to stratify on the outer part of the mixing chamber. This evidently complicates the design, as it requires creating an additional fuel plenum at the top and connecting channels between the upper and lower plenums, but it ensures much better mixing. After several attempts, it was found that the most efficient solution was to incorporate injection holes at the top in only some of the channels, alternating between yes and no. Subsequently, the effect of fuel heating was investigated concerning its distribution between the lower and upper plenums. Along its path, the fuel first passes through the air plenum, which is at 834 K, and then distributes inside the mixer, which is immersed in hot air. This results in a significant temperature increase, affecting its density and the balance of flows feeding into the lower and upper plenums, ultimately leading to fuel enrichment in the central part of the chamber. This effect necessitated the provision of fuel injection holes in the upper part for all air channels.

An example of the resulting equivalence ratio distribution is shown in Figure 6a and Figure 7. At the outlet section, $\Phi$ generally follows a radial pattern, with two inner and/or outer rings having a richer mixture and a central ring with a leaner mixture, due to the inability of the fuel jets to completely penetrate the airflow. The flow at the burner’s exit plane is highly swirled, reaching velocities of around 115 m/s, as illustrated in Figure 6b.

Non-reactive Large Eddy Simulations were also executed for a 50% ${\mathrm{CH}}_4$–50% ${\mathrm{H}}_2$ fuel mixture and 0.5 equivalence ratio. Here, the burner was simulated at the center of an axisymmetric computational domain with the same inlet boundary conditions as in the RANS simulations reported above. The WALE model [18] was used as a subgrid scale model. Again, a pressure-based solver and the SIMPLE scheme for pressure–velocity coupling were adopted. A bounded second-order implicit and a bounded central differencing scheme were used for the temporal and momentum discretization, respectively, while a second-order upwind scheme for spatial discretization was adopted for the other variables. In this case, the geometry of the installed experimental burner was simulated with the real air supply system (Figure 8) with about 10 million tetrahedral computational cells. The upper and lower fuel injection mass flow rates in the 18 channels were taken from the RANS simulations, since the fuel distribution system was not simulated due to the excessive computational cost. The results (Figure 8) show a slightly higher accumulation of fuel around the central body with respect to RANS simulations.

3.1. The RANS Reactive Simulations

Here are the initial results of reactive simulations aimed at highlighting the macroscopic characteristics of the flame, such as its extension and flame front conformation. The simulations are conducted at a nominal power of 33 kW. Two adjacent burners were also simulated (Figure 9), covering a 72° segment of the combustion chamber (36° for each), with periodic boundary conditions imposed and about 2 million tetrahedral computational cells. The boundary conditions were the same as the non-reactive simulations. Two compositions were analyzed, 50 and 100% hydrogen, maintaining constant air flow rate and thermal power. The chamber walls were assumed to be adiabatic and no-slip.

It is observed that in the cases reported here, the average nominal equivalence ratio decreases with increasing hydrogen content (0.5 for 50% ${\mathrm{H}}_2$, 0.43 for 100% ${\mathrm{H}}_2$). The most suitable equivalence ratio to ensure low ${\mathrm{NO}}_{\mathrm{x}}$ emissions while achieving complete combustion with varying hydrogen content in the fuel mixture will be determined experimentally.

Figure 10 and Figure 11 depict the distributions of temperature and H radical mass fraction on a section plane for the conditions examined: the latter quantity was chosen since it can be used to mark the reaction front. The reactive simulations indicate that as the ${\mathrm{H}}_2$ content increases, the flame becomes shorter because of a remarkable increase in the laminar flame speed. This shortening is particularly pronounced when the fuel consists of 100% ${\mathrm{H}}_2$. Additionally, the flow velocities of the reactive mixture, for which the burner was designed, appear adequate to prevent flame flashback in the premixing chamber. Conversely, with a 50% ${\mathrm{H}}_2$ content, the flame is noticeably longer and more stretched. CO and ${\mathrm{NO}}_{\mathrm{x}}$ emissions are reported in Figure 12. Even if the ${\mathrm{NO}}_{\mathrm{x}}$ values are overestimated compared to the experimental results and the LES simulations reported below, the trend of reduction of emissions with increasing hydrogen content can be justified by the fact that the flame front is less extended. Therefore, even if the production rate is higher, due to the generally higher temperatures, the integration of this rate carried out on a significantly smaller volume ultimately produces lower ${\mathrm{NO}}_{\mathrm{x}}$ emissions.

3.2. LES and Combustion Dynamics

The LES simulation of the previous RANS case with 100% ${\mathrm{H}}_2$ and $\Phi =0.43$ has been performed by means of the in-house parallel code HeaRT and ENEA’s CRESCO supercomputing facility [19]. The HeaRT code solves the 3D compressible Navier–Stokes equations with staggered finite-difference schemes, employing the dynamic Smagorinsky and LTSM models for subgrid terms [20]. Convective terms are handled with the ${\mathrm{AUSM}}^{+}$-up method and WENO interpolation, while diffusive fluxes use a second-order central difference scheme.

The computational domain is a 60° cylindrical sector, 0.10 m in length, and 0.05 m in radius, as shown in Figure 13. The whole chamber domain is filled with the hot combustion products that come from the corresponding laminar premixed flame (not reported), and the reactant mixture is injected through an annular section using the boundary condition from the RANS simulation (see Section 3.1). The mesh consists of $400\times 300\times 64$ grid points in the axial (z), radial (r), and azimuthal ($\theta$) directions, respectively, resulting in 7.68 million cells. Chemical reactions follow the Glarborg et al. mechanism [21] for ${\mathrm{H}}_2$ with 21 species and 109 reactions. Wilke’s formula [22] and Mathur’s expression [23] are used for viscosity and thermal conductivity, respectively, with preferential diffusion based on Hirschfelder and Curtiss law [24]. The Soret effect (thermal diffusion) is also considered in the model.

Figure 14 displays snapshots of the temperature and heat release rate (HRR) fields in a central z-r slice of the computational domain. The maximum temperature achieved slightly exceeds 2000 K, indicating the ignition zone with intense turbulence fostering efficient combustion. The jet core exhibits a short length ($\approx 1\mathrm{cm}$, which is shorter with respect to the RANS simulation, as expected, due to the flame front wrinkling captured in the LES flowfields), showing a high corrugated, but not fragmented, front where the maximum of the HRR is located. The flame shows also a consistent tip-flapping motion (not visible), and the combustion region is not only confined to the internal shear layer. In Figure 15, the instantaneous turbulent structures are identified through an iso-surface of the Q-criterion colored by temperature. The visualization combines also the time-averaged streamlines in a z-r plane, colored by the NO distribution (averaged in time and in the $\theta$ direction), to provide insights into the turbulent combustion dynamics. The figure reveals flame wrinkling and small-scale turbulence effects, which are characteristic of highly turbulent premixed flames. As the flow develops inside the chamber, the turbulent structures appear to become more elongated in the axial direction. The high-temperature region involves more than half the computational domain, and it slightly decreases near the exit (see also Figure 14). A prominent central toroidal recirculation zone is observed near the axis, which is formed due to the strong swirl imparted to the jet before injection. This recirculation region plays a crucial role in flame stabilization, as it retains hot gases and promotes mixing with the incoming fresh mixture. A secondary toroidal recirculation region near the walls is also visible, indicating natural flow reversal. The maximum NO concentrations never exceed 10 dppm (more than four times lower with respect to RANS prediction) and are predominantly found in these high-temperature regions within strong recirculation zones of the combustion chamber, where the progress variable is expected to be unity. This distribution suggests that in these regions, thermal NO formation is the primary mechanism governing NO production.

To gain deeper insight into the pollutant formation, Figure 16a shows the mean NO, ${\mathrm{N}}_2$O and ${\mathrm{NO}}_2$ concentrations as a function of the progress variable ($\mathrm{c}={\mathrm{Y}}_{{\mathrm{H}}_2\mathrm{O}}/{\mathrm{Y}}_{{\mathrm{H}}_2\mathrm{O},\max }$). Low NO pollutants are observed at low progress variable values, where the temperature is insufficient to drive thermal NO production and reach their highest concentrations ($\approx 7\mathrm{dppm}$) when c approaches 1. This condition, driven by the hydrogen high preferential diffusivity, is achieved in regions where the equivalence ratio reaches its higher value with respect to the nominal one ($\Phi \approx 0.55$). Regarding the other two species, they exhibit a non-monotonic behavior as functions of c. In particular, ${\mathrm{N}}_2$O shows an increasing trend with the progress variable up to $\mathrm{c}\approx 0.98$, where it reaches its maximum and then decreases to 1 dppm at $\mathrm{c}=1$. On the contrary, ${\mathrm{NO}}_2$ presents a peculiar behavior; it rapidly increases, reaching a local peak at $\mathrm{c}\approx 0.05$. Subsequently, the concentration decreases before stabilizing at intermediate values of c. This is in strong agreement with the results of Figure 16b, where ${\mathrm{NO}}_2$ is depleted at $\mathrm{c}\approx 0.2$ to form NO. Finally, as the progress variable approaches unity, ${\mathrm{NO}}_2$ undergoes a steep rise, displaying an exponential-like growth. This trend suggests that ${\mathrm{NO}}_2$ formation is strongly dependent on both the local oxidizing environment and thermal conditions.

The contribution of the six more relevant NO formation pathways to the NO mean rate of production (ROP) is depicted in Figure 16b as a function of the progress variable. These formation pathways and their corresponding reactions are described in

Table 1. Synthesis of the reactions involved in the NO formation pathways.

Reaction #	Chemical Expression	NO Pathway
R52	${\mathrm{NH}}_2+\mathrm{NO}={\mathrm{N}}_2+{\mathrm{H}}_2\mathrm{O}$	${\mathrm{N}}_2$
R51	${\mathrm{NH}}_2+\mathrm{NO}=\mathrm{NNH}+\mathrm{OH}$	${\mathrm{N}}_2$
R55	$\mathrm{NH}+\mathrm{NO}={\mathrm{N}}_2+\mathrm{OH}$	${\mathrm{N}}_2$
R49	$\mathrm{NH}+\mathrm{NO}={\mathrm{N}}_2\mathrm{O}+\mathrm{H}$	${\mathrm{N}}_2\mathrm{O}$
R32	$\mathrm{NH}+\mathrm{O}=\mathrm{NO}+\mathrm{H}$	${\mathrm{NH}}_2$
R28	$\mathrm{N}+\mathrm{OH}=\mathrm{NO}+\mathrm{H}$	Zeldovich
R27	$\mathrm{N}+{\mathrm{O}}_2=\mathrm{NO}+\mathrm{O}$	Zeldovich
R50	${\mathrm{N}}_2+\mathrm{O}=\mathrm{N}+\mathrm{NO}$	Zeldovich
R59	$\mathrm{HNO}+\mathrm{H}=\mathrm{NO}+{\mathrm{H}}_2$	HNO
R48	$\mathrm{H}+\mathrm{NO}(+\mathrm{M})=\mathrm{HNO}(+\mathrm{M})$	HNO
R56	$\mathrm{HNO}+\mathrm{OH}=\mathrm{NO}+{\mathrm{H}}_2\mathrm{O}$	HNO
R62	${\mathrm{NO}}_2+\mathrm{H}=\mathrm{NO}+\mathrm{OH}$	${\mathrm{NO}}_2$
R63	${\mathrm{NO}}_2+\mathrm{O}=\mathrm{NO}+{\mathrm{O}}_2$	${\mathrm{NO}}_2$
R47	${\mathrm{NO}}_2(+\mathrm{M})=\mathrm{NO}+\mathrm{O}(+\mathrm{M})$	${\mathrm{NO}}_2$

[25]. The ${\mathrm{NH}}_2$ path is the most important one, which is followed by the Zeldovich mechanism and ${\mathrm{NO}}_2$. The NO rate of production contributions peaks at approximately the same progress variable ($\mathrm{c}\approx 0.85$) except for the ${\mathrm{N}}_2$O and ${\mathrm{NO}}_2$ paths. In particular, the ${\mathrm{N}}_2$O formation rate exhibits a significant peak at intermediate combustion stages with a steep increase as c progresses and a maximum near $\mathrm{c}\approx 0.9$ before declining. The ${\mathrm{NO}}_2$ pathway exhibits a dual-peak behavior, with a first local maximum at $\mathrm{c}\approx 0.2$, which is followed by a weak decline and a subsequent peak in later combustion phases ($\mathrm{c}\approx 0.8$).

4. Experimental Tests

The experimental campaign described in the following was not intended and designed to provide data for CFD validation but to check the fuel-flexible burner’s key performance indicators aimed in the funded project. Hence, the burner was operated under various conditions to look for the best equivalence ratio ranges of operation at different hydrogen content in the HENG blend. However, experimental observations and measurements also qualitatively validate numerical predictions in terms of aerodynamic stabilization and ${\mathrm{NO}}_{\mathrm{x}}$ emissions.

In the experimental tests described below, the preheating of the combustion air is achieved by utilizing the enthalpy content of the exhaust gases. Specifically, before being mixed with the fuel (methane/hydrogen in various proportions), the combustion air flows through a 2-m-long copper coil, which forms a double spiral inside the combustion chamber and is exposed to the combustion products. The preheated air is then introduced into the burner, where it mixes with the fuel.

The combustion air temperature quickly reaches a steady state within about ten minutes. However, due to its high mass (see diagrams in Figure 3), the burner’s body takes significantly longer (around 60 min) to reach thermal equilibrium, which impacts pollutant emissions. The formation of nitrogen oxides (${\mathrm{NO}}_{\mathrm{x}}$) stabilizes quickly along with the combustion air temperature. In contrast, carbon monoxide (CO) emissions remain high (hundreds of ppm) until the burner body reaches its operating temperature, at which point they drop to a few dozen ppm. Pollutant emissions are measured using an FTIR (Fourier Transform Infrared Spectroscopy) device, which analyzes the absorption of specific wavelengths in the near-infrared spectrum. Each chemical species absorbs radiation at characteristic wavelengths, allowing the absorbed spectral lines to reveal the quantitative composition of the analyzed gas.

4.1. Gaseous Emission Analysis

Exhaust emissions from the combustor were continuously monitored by a GASMET DX4000 (Gasmet Technologies Inc., Vantaa, Finland) portable multicomponent gas analyzer incorporating a low-resolution (4 ${\mathrm{cm}}^{-1}$) Fourier Transform Infrared Spectrometer (FTIR) and a Peltier cooled mercury–cadmium–telluride (MCT) detector with a scan frequency of 10 Hz and a wavenumber range of 900–4200 ${\mathrm{cm}}^{-1}$. A ZrO2 sensor for accurate ${\mathrm{O}}_2$ measurement was coupled. The GASMET portable sampling system (PSS) for gas conditioning and cleaning consisted of a 180 °C heated sample pump (gas sampling rate of 2–5 L ${\min }^{-1}$), particulate filter, and valve. The PSS was connected to the FTIR by a 180 °C heated transfer line.

The sampling probe was mounted 200 mm inside the combustor outlet by a heated stainless steel (SS) transfer line (1 m length and 3.175 mm internal diameter) with temperature control settable up to 300 °C (Pike Technologies Inc., Cottonwood Dr., Madison, WI, USA) and connected to the inlet of the PSS by a 10 m long heated hose to minimize the thermal shock damage of the heated gas cell. The 0.40 L Rh-coated aluminium gas cell with a fixed absorption of 5.0 m path length was kept to 180 °C to avoid condensation. Zero-calibration was achieved automatically by PSS with ${\mathrm{N}}_2$ before starting experiments.

Gas sampling was performed every 20 s interval, and the FTIR spectra collected were stored and processed on a laptop PC using the CALCMET™ Standard software (version 11.10), which analyzes the acquired sample spectrum using a sophisticated algorithm. The software can simultaneously detect, identify, and quantify up to 50 different gas components from a single sample. Quantification was conducted by factory calibration and typically was set up to measure evolved gases such as ${\mathrm{H}}_2$O (0–40% v/v), ${\mathrm{CO}}_2$ (0–30% v/v), CO (0–500 mg ${\mathrm{N}}^{-1}$${\mathrm{m}}^{-3}$), ${\mathrm{CH}}_4$ (methane, 0–100 mg ${\mathrm{N}}^{-1}$${\mathrm{m}}^{-3}$), ${\mathrm{NO}}_2$ (0–50 mg ${\mathrm{N}}^{-1}$${\mathrm{m}}^{-3}$), NO (0–200 mg ${\mathrm{N}}^{-1}$${\mathrm{m}}^{-3}$), and ${\mathrm{N}}_2$O (0–100 mg ${\mathrm{N}}^{-1}$${\mathrm{m}}^{-3}$) with 2% error.

Figure 17 shows the picture of analytical instrumentation. The measured concentration of emitted gaseous species j was normalized to 15 vol.% ${\mathrm{O}}_2$ content by applying the following equation:

$${\mathrm{j}}_{\mathrm{norm}}={\mathrm{j}}_{\mathrm{meas}}·({\mathrm{O}}_{2_{\mathrm{amb}}}-15\%)/({\mathrm{O}}_{2_{\mathrm{amb}}}-{\mathrm{O}}_{2_{\mathrm{meas}}})$$

(1)

where the subscripts norm, meas and amb stay for normalized, measured and ambient concentration, respectively.

4.2. Flame Topology and Anchoring-Premixed Mode Operation

Figure 18 presents a map of the burner’s operating points under atmospheric conditions, which is accompanied by a visible-spectrum image of the flame (captured by a camera) and the combustion air temperature. To counteract the increase in combustion speed caused by higher hydrogen content, the mixture was progressively leaned while maintaining constant power output, as the total fuel mixture quantity remained unchanged for each hydrogen level. This results in a decrease in the combustion air temperature. Notably, each operating point can be reached from adjacent ones by adjusting parameters without flame extinction. Additionally, in a technologically significant aspect, ignition can be achieved directly using a spark generated between two electrodes subjected to a high-voltage current (approximately 20 kV).

From visual analysis, it has emerged that not all operating points can be considered optimal, not only in terms of pollutant emissions, as previously discussed. Certain combinations of parameters (equivalence ratio and hydrogen content) cause the flame to recede significantly, anchoring itself near the central bluff body, which is visible in Figure 3. In Figure 18, it is evident that this central element becomes incandescent, acting as a hot ignition point that stabilizes the flame in its vicinity, leading to progressive deterioration (erosion). The steel material is not suitable for withstanding such thermal stress, as highlighted in the second panel of Figure 18. Furthermore, the flame’s recession and anchoring near the bluff body significantly alter its structure, shifting from an aerodynamically stabilized flame to a jet-type flame anchored to the hot spot.

An example of aerodynamic stabilization is represented by the point ${\mathrm{X}}_{{\mathrm{H}}_2}=0.5$, $\Phi$ = 0.6, where the flame appears to broaden into a mushroom shape at the outlet of the burner (Figure 18). This shape is due to a central recirculation zone generated by the helical motion followed by the fluid particles, which is a motion induced by a series of internal vanes within the burner. The jet, rotating in this manner, tends to spread radially, and therefore, in order to maintain the momentum of the mass unchanged, it slows down and creates a recirculation zone along its axis that traps the hot combustion gases necessary for igniting the reacting mixture. For this reason, it is said that the flame is aerodynamically stabilized, as shown in the upper panel of Figure 19, where it illustrates how the flame moves from the outside (reactants–low OH concentration–black in the photo) toward the axis of the burner (combustion products–high OH values).

When the flame recedes on the bluff body, this stabilizing recirculation is restricted, partially, or even completely destroyed, and the flame changes structure. This phenomenon is evident from the analysis of the fluorescence of the flame front induced by a laser source: this spectroscopic measurement is capable of highlighting the instantaneous structure of the flame front and distinguishing between the reactant and combustion product regions. In a flame front, the ignited mixture, before finally transforming into combustion products, goes through several stages where intermediate combustion products are generated and destroyed, including certain chemical species known as radicals. One of these is the hydroxyl radical (OH), which can be measured through fluorescence induced by a laser source. In the reactants, hydroxyl is not present, but it increases rapidly during the combustion process and then decreases, more slowly than it was formed, in the combustion products. Its asymmetric distribution is functional not only in determining the position of the flame front but also in establishing its orientation, which is not obvious when the flame front is particularly jagged and convoluted, as in turbulent flames. The extremely short lifetime of the fluorescence signal emitted by the OH radical effectively freezes the topology of the front in an instant [26]. This fluorescence signal is obtained by stimulating (illuminating) the combustion zone with a UV laser source that excites the OH molecule, which, upon relaxation to its ground state, emits photons again in the ultraviolet range but at a different (longer) wavelength than the excitation wavelength. Each molecule has its specific absorption and emission wavelengths; therefore, by appropriately selecting the excitation wavelength, the molecule to be traced is exclusively chosen.

In the lower panel of Figure 19, it is possible to observe the distribution of the hydroxyl radical (OH) and the instantaneous configuration of the flame front. For hydrogen concentrations of ${\mathrm{X}}_{{\mathrm{H}}_2}=0.3$ and ${\mathrm{X}}_{{\mathrm{H}}_2}=0.5$, aerodynamic stabilization is maximized, as shown by the inverted cone shape of the flame. For higher concentrations of hydrogen, close to pure hydrogen, the receding of the flame anchor weakens the swirl effect, and consequently, the size of the inverted cone shape of the flame decreases. The same effect is observed for conditions approaching pure methane (the reasons for this are unclear; given the high flow velocities, methane should be closer to blow-off conditions, but at the same time, the equivalence ratio of the conditions is higher; it could be due to reduced heat release within the recirculation, resulting in its smaller expansion).

4.3. Pollutant Emissions at Steady Conditions–Premixed Mode Operation

Figure 20 presents the results obtained by operating the burner at atmospheric pressure under different excess air conditions, which is represented by the equivalence ratio $\Phi$. This ratio indicates the fuel-to-air mass ratio normalized to the same ratio under stoichiometric conditions. The figure also considers varying hydrogen content, ranging from pure methane (${\mathrm{X}}_{{\mathrm{H}}_2}=0$) to pure hydrogen (${\mathrm{X}}_{{\mathrm{H}}_2}=1$), where X represents the molar fraction (or equivalently, the volume fraction) of hydrogen in the methane–hydrogen mixture. The percent errors of the emission measurements reported in Figure 20 are around 1–2% on average.

It is observed that the temperature of the combustion air tends to decrease as the excess air increases (i.e., for decreasing $\Phi$ values). Additionally, optimal equivalence ratios can be identified for each hydrogen content level in the mixture, where carbon monoxide (CO) emissions are minimal, remaining below 50 ppm (noting that regulations impose a 100 ppm limit). Regardless of the hydrogen content, as the mixture is further leaned out, CO production rises rapidly beyond acceptable levels (exceeding 200 ppm). This is due to the higher air content (with a constant fuel amount, as the power remains fixed), which increases the flow velocity of the mixture through the burner. As a result, the residence time of the mixture in the flame decreases, leaving insufficient time for complete fuel oxidation and leading to an increase in partially combusted gases (unburnt species) such as CO. It is important to note that complete fuel oxidation would produce ${\mathrm{CO}}_2$ instead of CO. Of course, if CO emissions rise at some specific operation point, the combustion efficiency drops, and such operation point must be avoided.

Regarding nitrogen oxides (${\mathrm{NO}}_{\mathrm{x}}$), the burner consistently produces minimal amounts, well below the strictest regulatory limits, which must be under 25 ppm. Additionally, it can be observed that leaning out the mixture, and consequently lowering the flame temperature, has a beneficial effect in reducing these pollutants.

These considerations apply both when the burner operates at full capacity, processing a fuel flow capable of generating a thermal power of 7.3 kW, and at reduced flow rates equivalent to 75% of the nominal power. Naturally, reduced flow rates decrease the velocity of the reacting mixture, leading to flashback issues when the flame speed increases. This explains why, at reduced power, the burner does not provide data above 50% hydrogen content: beyond this threshold, the flame speed increases to a point where flashback occurs (despite attempts to counteract it by reducing the equivalence ratio), necessitating a forced shutdown of the device.

Figure 21 reports the measured pollutant emissions at a higher pressure of 3 bar absolute. Tests revealed a narrow operating range in terms of equivalence ratio and hydrogen content due to the onset of flashback phenomena with a slightly wider range at 75% of nominal power. Such flashback issues are mainly due to the central body of the burner that becomes too hot and even incandescent. However, the trends in pollutant emissions and combustion air temperatures remain similar to those observed during atmospheric pressure operation. The percent errors of the emission measurements reported in Figure 21 are around 1–2% on average.

4.4. Pollutant Emissions at Fluctuating Hydrogen Content-Premixed Mode Operation

An additional test to evaluate the operational effectiveness of a fuel-flexible burner consists of feeding it with a mixture whose hydrogen concentration varies sinusoidally. Practically, starting from a fixed point with 50% hydrogen content, a 15% variation in hydrogen content was imposed sinusoidally, and the response of the burner was observed in terms of flame stability and pollutant emissions. The same test was also conducted from a base hydrogen content value of 30%. In both cases, the flame remained stable and the emissions were extremely low, returning to stable conditions once the fluctuation was eliminated. The results are shown in the images in Figure 22.

4.5. Pollutant Emissions of the Diffusive Micro-Mixing Burner at Steady Conditions

One final test of the burner was related to its operation in a diffusion mode, where the mixing of fuel and oxidizer occurs outside the burner through the jet’s fluid dynamics.

The main advantage of this type of flame is the absence of the flashback phenomenon. Figure 23 shows the results in terms of pollutant emissions for atmospheric conditions and for powers of 100% and 50% of the nominal power of 7.3 kW. The percent errors of the emission measurements reported in Figure 23 are around 4% on average. Under these conditions, the pollutant levels are below the limits set by the regulations while still achieving significant hydrogen concentrations. In this case, as well, an operating point can be reached either by varying the parameters from a nearby point or by direct ignition via spark, which is another requirement for burners processing large flows of hydrogen in the reacting mixture.

In diffusion combustion conditions, it can be observed (see Figure 24) that the absence of flashbacks preserves the central body of the burner, which never becomes incandescent. However, the operating range in terms of the equivalence ratio is significantly narrowed compared to that of the premixed flame because of blow-off phenomena that tend to blow the flame off, especially at high air flows, which are necessary to reduce the nominal equivalence ratio.

It should be noted that in diffusion mode, the actual equivalence ratio of the reacting mixture is determined by the intrinsic mixing between fuel and oxidizer, which is regulated by the fluid dynamics of the flow. Therefore, only a nominal equivalence ratio can be indicated, which is calculated on the basis of the ratios of the masses of fuel and oxidizer injected separately into the burner.

5. Discussion, Conclusions, and Future Work

A combustion strategy and geometric configuration with the potential to operate across a wide range of natural gas/hydrogen mixtures have been developed through RANS and LES simulations until reaching its final geometry. Specifically, the chosen solution features two independent lean premixed (DLE) burners, one for low and one (crown system) for high hydrogen content, with a burner selection and air modulation system. The implemented combustion stabilization mechanism is aerodynamic (flame propagation) and relies on vortex breakdown induced by a swirling flow. The crown system system burners, designed for high hydrogen content, implement this technology in the context of micro-mixing.

In this article, attention is focused on the development of the crown system burner, which is designed for HENG mixtures with high hydrogen content. A prototype (P3.7 H2-High) representative of one of the crown system burners was developed for experimental testing purposes before moving on to the complete prototype.

The simulated cases showed low ${\mathrm{NO}}_{\mathrm{x}}$ levels with a tendency for increased CO emissions under leaner conditions and at reduced power levels. The experimental validation phase was crucial in assessing the actual achievement of some key performance indicators (KPIs) [14]. Performance tests were conducted at 1 and 3 bar with preheated air temperatures matching those expected in the microturbine. These tests confirmed the potential of the crown system burner (in both premixed and diffusion modes), as detailed below in terms of the KPIs:

1.

Ability to operate with HENG mixtures containing more than 20% hydrogen by volume (state-of-art for 100 kWe microturbines) with a technological trend exceeding 50%. Specifically, the tests demonstrated good performance within a 30–70% hydrogen volume range while meeting emission limits.

2.

Ability to operate with ${\mathrm{NO}}_{\mathrm{x}}$ emissions below 25 ppm at 50% hydrogen by volume. Tests showed ${\mathrm{NO}}_{\mathrm{x}}$ < 25 ppm within the 30–70% hydrogen volume range.

3.

Ability to ignite mixtures with a hydrogen content greater than 20% by volume. Tests confirmed ignition capability in the 30–70% hydrogen volume range.

4.

Ability to operate with a hydrogen content varying by ±15% over time while still complying with ${\mathrm{NO}}_{\mathrm{x}}$ and CO emission limits. Tests demonstrated this capability at least within the nominal composition range of 30–50% hydrogen by volume with an actual effective range of 25–58%.

Despite the previous KPIs, some issues have been identified. Certain parameter combinations (equivalence ratio and hydrogen content) cause significant flame retreat, anchoring near the central bluff body of the burner, which becomes incandescent. This results in a localized ignition point that anchors the flame nearby, leading to progressive material degradation (erosion). Additionally, the retreat and anchoring of the flame near the bluff body significantly alter the flame structure, shifting from an aerodynamically stabilized flame to a jet flame anchored at the hot spot.

In a next three-year project, efforts will focus on improving the achieved KPIs and addressing both the excessive heating of the central burner body and burner control issues. Two possible strategies have already been identified for the latter, both based on eliminating the central burner, initially designed for low-hydrogen-content mixtures, to create a uniform burner plate consisting of burners of the same type as those in the ring system and resized to reduce efflux velocities.

The first solution involves burners similar to the P3.7 H2-High prototype, capable of operating in both diffusion and premixed combustion modes, integrated with an exhaust gas recirculation system (either actual EGR or aerodynamically induced). The second solution retains a multi-burner configuration but operates exclusively in diffusion micro-mixing mode.