Fuel 380 (2025) 133576

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full length article

On the use of hydrogen in oxy-fuel glass melting furnaces: An extensive

numerical study of the fuel switching effects based on coupled CFD
simulations
Georg Daurer a ,∗, Stefan Schwarz a , Martin Demuth b , Christian Gaber b , Christoph Hochenauer a
a
Graz University of Technology, Institute of Thermal Engineering, Inffeldgasse 25/B, Graz, 8010, Austria
b Messer Austria GmbH, Industriestraße 5, Gumpoldskirchen, 2352, Austria

ARTICLE INFO ABSTRACT

Keywords: The present study examined the potential of utilizing hydrogen as a fuel in an industrial cross-fired oxy-fuel
Glass melting furnace glass melting furnace with a total energy input of 3.9 MW. To this end, comprehensive three-dimensional CFD
Oxy-fuel combustion simulations were performed in Ansys Fluent, employing a coupled numerical model from preceding works.
Hydrogen
A two-step validation process was presented, comprising the validation of both the numerical furnace setup
Computational fluid dynamics
and the modeling of industrial oxy-fuel combustion setups using hydrogen as a fuel gas. Assuming an equal
Fuel switching effects
thermal energy input for both combustion conditions, the fuel/oxidizer ratios of velocity and momentum of the
six PrimeFire burners were found to be significantly affected. Consequently, the low-momentum flames with
natural gas transformed into jet-type flames with a straighter trajectory when transitioning to hydrogen. This
change led to an enhanced turbulent mixing and a reduction of flame lengths by over 25%, while the average
flame temperature increased by up to 82 K due to the accelerated reaction kinetics. In addition, the elevated
flame momentum in the cross-fired burner configuration resulted in a localized rise in maximum side wall
temperatures by more than 20 K. With the exception of these local effects, the global temperature distribution
and heat transfer mechanisms were not significantly impacted. This indicated a comparable melting process
and furnace efficiency, while still allowing for a reduction of total CO2 emissions by 77.5%. Future research
should concentrate on both the impact on glass quality and the evaluation of potential alternative burner
configurations for hydrogen-fired oxy-fuel glass melting furnaces.

1. Introduction One potential solution to reduce the still considerable environ-

mental impact of glass manufacturing is the integration of electric
As one of the most energy-intensive processes in the high- melting [7,8] using green electricity. However, the process of furnace
temperature sector, glass manufacturing is responsible for a consider- startup is problematic since glass is only an electrical conductor in its
able amount of the global greenhouse gas emissions. At the moment,
molten state. Moreover, the current limitations in electrode lifetime and
the required energy to achieve the particularly high temperature level is
pull rate flexibility necessitate further research and development [9].
mainly provided in form of natural gas combustion [1]. Throughout the
In addition to these technological challenges, a permanent and depend-
years, several efficiency measures have been developed, such as the use
of regenerative furnaces, employing oxygen-enhanced or pure oxy-fuel able supply of renewable electrical energy remains critical, particularly
combustion or most lately, including batch/cullet preheating [2–4]. At when considering the fluctuating availability of wind and solar energy
least to some extent, all these concepts aim at optimizing the energy as well as the considerable distances that must be bridged between ma-
efficiency by reducing the flue gas losses. Unfortunately, since the jor electricity generation plants and industrial consumers. Nevertheless,
most recent furnaces are already approaching the estimated theoretical achieving at least a portion of the total energy input via electrodes,
minimum of the specific energy demand, as discussed e.g. in [5,6], usually referred to as electric boosting, has become a common practice
another substantial decrease of the fuel consumption is not expected to enhance energy efficiency and glass quality [1,9].
in the foreseeable future. For this reason, still large amounts of non-
As another promising alternative, green hydrogen can be used as
renewable fossil fuels are being burnt, releasing tons of CO2 emissions
the fuel gas, thereby reducing the greenhouse gas emissions to an
in the process.

∗ Corresponding author.
E-mail address: [email protected] (G. Daurer).

https://doi.org/10.1016/j.fuel.2024.133576
Received 1 August 2024; Received in revised form 25 October 2024; Accepted 28 October 2024
Available online 6 November 2024
0016-2361/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
G. Daurer et al. Fuel 380 (2025) 133576

absolute minimum, given the inherently emerging product gases from CelSian have considerable experience in the CFD modeling of glass
the batch-to-glass conversion (calcination). Due to its excellent and melting furnaces, as they have developed their own software for this
carbon-free combustibility, hydrogen is considered a key energy source specific application. However, their work is primarily industry-oriented
not only in high-temperature processes, but in all fuel consuming and does rarely appear in the open scientific literature. Nevertheless,
and heavy emitting sectors. Therefore, several studies have also been several studies on hydrogen–air combustion setups in other industrial
conducted on other applications, including gas turbines [10,11] and environments are already available in the literature, including the
boilers [12,13]. In the glass industry in particular, a number of practical works of Cellek & Pinarbasi [13] on an 1 MW boiler and Lopez-Ruiz
projects have been launched in recent years to assess the feasibility et al. [31] on a 2.5 MW burner for a steel reheating furnace. Neither of
of integrating hydrogen into production processes. One noteworthy the mentioned works included an extensive model validation to support
example represents the already finalized ‘‘HyGlass’’ project [14] of The their results for hydrogen combustion conditions.
Federal Association of the German Glass Industry (BV Glass) together Reviewing the available literature on hydrogen combustion se-
with the Gas- und Wärme-Institut Essen (GWI). As summarized in the tups for high-temperature industries, a detailed study of a full-scale
final project report, it consisted of four main research tasks: First of hydrogen-fired oxy-fuel furnace has not yet been presented. Significant
all, the glass industry and its production processes were analyzed with simplifications in the model setup were introduced in the few numerical
regard to the potential reduction of CO2 emissions through hydrogen analyses available, and they further have not included a thorough
usage. Secondly, the GWI conducted hydrogen combustion experiments validation process for a hydrogen combustion case. As a result, a
at 1.6 MW within a semi-industrial furnace with regenerative air substantive in-depth evaluation of the fuel switch from natural gas
preheating to evaluate combustion and emission characteristics. Their to hydrogen is urgently needed, especially for glass melting furnaces,
observations indicated a shift in the location of the main reaction which are ideally suited for hydrogen combustion technologies due to
zone and an increase in NOx emissions with hydrogen. Furthermore, their elevated temperature level. In this regard, the use of hydrogen
the impact of different hydrogen enrichment levels and, consequently, oxy-fuel combustion is of special interest, since it allows the combi-
combustion atmospheres on the quality of diverse glass types was eval- nation of particularly high process temperatures, maximum efficiency
uated. Although a definitive impact of hydrogen could be confirmed, and minimum NOx emissions. The present work tackles this aspect
the intricate interaction of multiple chemical processes prevented a and aims at closing the research gap, as an industrial glass melting
clear conclusion. Finally, numerical simulations on a large regenerative, furnace was studied numerically under conventional natural gas-fired
cross-fired glass furnace were performed with a hydrogen enrichment and hydrogen-fired oxy-fuel conditions. To this end, a validated 3D
of up to 50 𝑣𝑜𝑙%, highlighting the importance of an advanced gas con- simulation model of a continuously operated glass melting furnace with
trol skid to maintain the thermal power and equivalence ratio. Other a total energy input of 3.9 MW was developed. It has already been
notable ongoing projects are the ‘‘HyNet’’ project [15] in the UK and the part of previous numerical studies [29,32–34], where, on the one hand,
EU-funded ‘‘H2Glass’’ project [16], which already demonstrated partial the coupled simulation methodology was presented, and, on the other
usage of 100 𝑣𝑜𝑙% hydrogen in real-size glass melting furnaces [17– hand, the effects of various process parameters on the glass quality
19]. An overview over current projects on hydrogen trials in the glass and the furnace operation were assessed. As part of the current inves-
industry was recently provided by The American Ceramic Society in tigation, extensive coupled CFD simulations were conducted in Ansys
Ref. [20]. Fluent for both combustion setups, with a particular focus on the impact
In order to support the development of future low-emission hy- of fuel substitution in the combustion space while maintaining the total
drogen combustion technologies for the hard-to-abate industries, it energy input. Through an in-depth analysis, the study evaluated the
is essential that such practical projects are accompanied by detailed differences in local flame characteristics as well as in global furnace
studies of the underlying combustion and heat transfer phenomena. In operation. Based on the results presented, a comprehensive view is
this regard, several researchers already investigated hydrogen flames in provided that supports the future adaptation of the glass manufacturing
lab-scale setups [21,22], observing differences in the flame stabilization process, thus contributing to improved sustainability.
and flow regime. In order to build the bridge to large-scale industrial
furnaces, the authors of the present paper have previously conducted 2. Industrial furnace setup
extensive experiments [23,24] and numerical simulations [25,26] on
a semi-industrial hydrogen combustion setup, both under air–fuel and The investigated furnace was a continuously operated, cross-fired
oxy-fuel conditions. In contrast to the majority of existing studies that oxy-fuel glass melting furnace with electric boosting, as depicted in
focused on lab-scale flames, which are of limited practical significance, Fig. 1. The furnace was comprised of a glass tank with a melting end
these studies yielded detailed insights into the effect of hydrogen en- and a refining segment, as well as the combustion space, which was
richment on the temperature distribution, heat transfer rates, pollutant the primary focus of this study. The entire interior of the furnace was
formation and flame characteristics in an industrial environment. lined with zirconia containing refractory materials. The combustion
While some large-scale furnaces have already been partially fired chamber was equipped with six transversally arranged low-momentum
with up to 100 vol% hydrogen, to the best of the authors’ knowledge, burners B1–B6 of the type ‘‘Eclipse PrimeFire 300’’, which are com-
the long-term switch from natural gas to hydrogen in an entire furnace monly used in oxy-fuel glass melting furnaces. A schematic image
with several MW thermal input has not yet been realized, mainly of the burner together with its numerical representation inside the
due to economic and technological challenges. Therefore, numerical combustion chamber model was also included in Fig. 1. Their position
simulations play an essential role to evaluate different combustion inside the combustion chamber relative to the furnace length was
scenarios, additionally yielding detailed information of the ongoing specified in Table 1. All burner outlets were arranged parallel to the
physical phenomena inside the furnace. However, recent numerical- global 𝑥𝑧-plane and located within a burner quarl at a distance of
based research on glass melting furnaces has primarily concentrated 235 mm from the internal furnace wall. The corresponding outlet cross
on the modeling and comprehension of the intricate processes occurring sections were 𝐴𝑓 𝑢𝑒𝑙 = 2190 mm2 for the fuel and 𝐴O2 = 4925 mm2
within the glass melt, as e.g. in Refs. [27–30]. As mentioned previously, for the oxygen. The PrimeFire burners were designed to yield a higher
the GWI performed simulations of a large-scale regenerative glass melt- melting efficiency through an improved radiative heat transmission
ing furnace with up to 50 𝑣𝑜𝑙% H2 enrichment as part of the HyGlass from the flame [35,36]. This is achieved by an extremely low flame
project [14]. When an equal thermal power input and constant burner momentum and thus limited mixing rate that results in an intended
equivalence ratios were maintained, no significant differences in the formation of soot particles within the flame. In addition, the fan-shaped
temperature and velocity distribution of the combustion space were flame is characterized by a high flame spread that yields an increased
observed. It is worth noting that companies such as Glass Service and load coverage and more uniform heat distribution [35,36].

2
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 1. Schematic illustration of the investigated industrial glass furnace with the employed PrimeFire burner [37].

At the considered operating point of the furnace, the six burn- only a brief overview of the modeling methods is given as the employed
ers were operated with natural gas (NG) at a near stoichiometric boundary conditions and corresponding mathematical equations were
fuel/oxygen equivalence ratio of 𝜙 = 0.99 that corresponded to 2 explained in great detail in the previous works [29,32–34]. For further
𝑣𝑜𝑙%𝑑 .𝑏. oxygen in the dry flue gas. The thermal energy input by the information, the reader is referred to these preceding publications.
burners amounted to 3.34 MW. Furthermore, six electrodes E1–E6 The entire furnace was considered to be in a quasi-steady state and
were integrated into the melting end in order to support the primary thus could be described by means of a RANS simulation approach. This
fining of the glass, which provided an additional electric power input is a reasonable and viable assumption for a continuously operating
of 600 kW directly to the melt. These electrodes were arranged in two glass melting furnace, as adopted in Refs. [27,39,40]. Similar to those
staggered transversal rows in the rear part of the melting end close to previous numerical studies on glass melting furnaces, the respective
the hot spot. The production process was monitored and controlled by steady-state solutions of the two submodels CC and GT were calculated
means of temperature measurements throughout the furnace, including separately and coupled via the melt surface boundary condition in an
seven Type B sheathed thermocouples TC1–TC7 that were installed iterative calculation scheme. It included the transfer of the heat flux
along the center plane. These values also served for validating the distribution (CC → GT) and the temperature distribution (GT → CC).
developed CFD model of the furnace under conventional NG oxy-fuel The coupling scheme was aborted as soon as the relative temperature
conditions, as discussed in Section 4.1. deviation between subsequent coupling iterations fell below 0.5% for
A central part of the present paper was the evaluation of the all cells across the melt surface.
suitability of a standard industrial oxy-fuel burner for pure hydrogen As indicated in Fig. 1, the burners were modeled exclusively in
operation. The numerical study of the fuel substitution effects was terms of the fuel and oxidizer flow paths. To guarantee a fully de-
thus based on switching the natural gas supply of each burner to pure veloped flow at the burner outlet, the flow paths were extended to
hydrogen (H2 ), ensuring an equal thermal energy input by the fuel. a distance of 1.2 m from the actual outlet nozzles, where mass
The fuel mass flow rate of each burner was calculated based on the flow inlets for the fuel and oxidizer were defined. Fuel and oxidizer
respective burner power (see Table 1) and the lower heating values inlet temperatures of 25 ◦ C were assumed together with a turbulent
𝐿𝐻 𝑉𝐶 𝐻4 = 50.01 MJ∕k g and 𝐿𝐻 𝑉𝐻2 = 120.0 MJ∕k g [38], respectively. intensity of 5%. A pressure of 200 Pa below atmosphere was specified
The corresponding oxygen flow rates were adapted accordingly in at the outlet according to the operation of the real furnace. A no-slip
a way that an equal residual oxygen content of 0.676 𝑣𝑜𝑙%𝑤.𝑏. was condition and an emissivity of 0.8 was assigned to the internal walls
obtained in the wet flue gas, resulting in similar burner equivalence of the refractory and the burner. At the outer walls of the refractory,
ratios of approximately 0.99. More details on the numerical modeling a combined convective and radiative boundary condition was applied,
are provided in Section 3. with an ambient temperature of 25 ◦ C, an estimated heat transfer
coefficient of 30 W∕(m2 K ) and a wall emissivity of 0.8. For the tur-
3. Numerical methodology bulent flow field in the combustion chamber, the realizable 𝑘–𝜀 model
by Shih et al. [41] was employed with the enhanced-wall treatment
As part of the numerical modeling strategy, the total furnace was option, a common choice for CFD simulations of complex combustion
divided into two submodels, namely the combustion chamber (CC) processes industrial furnaces [40,42–46]. The combustion modeling in
model and the glass tank (GT) model. This was required due to the both fuel configurations included the partially-premixed Steady Lami-
significant differences in fluid properties and flow field characteristics nar Flamelet (SLF) model with the skeletal mechanism of Peeters [47].
between the turbulent gas-phase and the highly viscous glass melt, The partially-premixed approach was shown to achieve a more re-
which would have resulted in numerical instabilities at the interface. alistic numerical representation of low-swirl air–fuel flames [26] as
The three-dimensional CFD model of the furnace as well as the coupling well as oxy-fuel flames when the nozzle velocities of the fuel and
methodology was already developed and optimized in preceding works the oxidizer are similar [48]. The combination of the Flamelet Model
by Raič et al. [32] and Daurer et al. [29]. Numerical grids with a and the mechanism of Peeters [47] was already tested and validated
total of 3.16 and 0.58 million elements were implemented for the CC extensively for oxy-fuel conditions [25,44,49,50]. Due to its accelerated
and GT models, respectively, which were further characterized by a reaction kinetics, the mechanism better reproduces the combustion
maximum skewness of 0.9 and a maximum aspect ratio of 30.5. A grid characteristics under oxy-fuel conditions [25,51]. In particular, it was
independence study was already presented in Ref. [32]. At this point, validated for hydrogen oxy-fuel combustion at the Institute of Thermal

3
G. Daurer et al. Fuel 380 (2025) 133576

Engineering at TU Graz, which is highlighted in Section 4.1 as part of unfeasible, primarily for economic reasons and also due to technolog-
the overall validation strategy. Thermal radiation was modeled with the ical and safety-related concerns. Consequently, the validation process
Discrete Ordinates (DO) approach and a 4 × 4 discretization. Although for the present investigation has been divided into two stages, referring
the PrimeFire burners are characterized by producing a flame with to the results of previous works [25,29,32]. A schematic illustration
high luminosity, flame radiation by soot particles was neglected in the of the entire validation strategy with the referenced results is given in
modeling approach used. This was because experience in combustion Fig. 2.
engineering has shown that soot radiation is an insignificant contribu- First and foremost, the CFD model of the glass melting process had
tor to the total heat transfer from natural gas flames [52], since these to be evaluated. This was done to ensure that the entire furnace setup,
particles are typically present only in limited concentrations in the including all the applied boundary conditions, material properties, and
flame center, which is a particularly small region compared to the the general numerical modeling approach yielded results that were
volume of the total gas phase. In addition, the amount of soot radiation representative of the ongoing physical and thermochemical phenomena
depends strongly on the local particle concentrations, which are diffi- in the real furnace. For this validation step, the temperature measure-
cult to determine accurately with available numerical models. For this ments TC1–TC7 inside the melting furnace under natural gas oxy-fuel
reason, the modeling approach focused on a precise description of the conditions served as a reference. As described in Refs. [29,32], a highly-
dominant gas radiation, with the corresponding absorption coefficient precise temperature prediction could be achieved by the CFD model,
being calculated according to the recently published weighted-sum-of- with mean deviations in the considered natural gas operation mode of
gray-gases model (WSGGM) of Bordbar et al. [53,54]. This alternative only 23.4 K in the combustion chamber and 25.0 K in the glass tank.
and more accurate WSGGM was integrated into Fluent in form of a In terms of the accuracy, these results were unmatched by previously
user-defined-function (UDF). The UDF retrieved the static pressure, presented simulation models of glass furnaces in the open literature and
temperature and species concentrations of each cell and calculated thus argued for an overall reasonable numerical setup.
the corresponding molar ratio H2 O/CO2 as an input parameter for The second part involved the combustion modeling approach for
the model of Bordbar. It then derived the WSGGM coefficients 𝑏𝑖,𝑗 , 𝜅𝑖 hydrogen oxy-fuel conditions in an industrial environment. For this rea-
and weighting factors 𝑎𝑖 , which were used to determine the final gray son, the hydrogen operation of the industrial oxy-fuel burner ‘‘Messer
gas emissivity and absorption coefficient for a pre-calculated, domain- Oxipyr Hydroflex’’ was investigated both experimentally and numer-
based radiative path length. The calculated absorption coefficient of the ically within a semi-industrial combustion chamber at the Institute
gas mixture was ultimately returned to Fluent for each cell. of Thermal Engineering at Graz University of Technology [23,25]. Its
In the glass tank model, the transversal batch supply at each side relevance for the glass industry is emphasized by the fact that the same
of the furnace was described using a mass flow inlet with an estimated burner was later installed in a batch furnace for glass melting at a
batch temperature of 25 ◦ C. An atmospheric pressure of 1 at m was partner company. There, the impact of different oxy-fuel combustion
specified at the melt outlet after the refiner. The refractory walls were atmospheres on the glass quality was evaluated as part of an internal
modeled analogous to the combustion chamber, including a no-slip investigation. With regard to the CFD study presented in Ref. [25], the
condition and an emissivity of 0.8 at the interior, and a combined applied modeling strategy demonstrated a high degree of accuracy in
convective/radiative boundary condition at the exterior. Due to the reproducing the measured temperature distribution and heat transfer
high viscosity of the melt, the flow in the glass tank was completely rates under hydrogen oxy-fuel conditions. For the operating conditions
laminar and no turbulence modeling was required. The batch-to-glass shown in Fig. 2, including a burner power of 140 k W and six water-
conversion was modeled in terms of temperature dependent material cooled lances, the predicted temperature profile exhibited a mean
properties and an energy sink throughout the melting end that corre- deviation of less than 20 K, while the numerically calculated heat
sponded to the reaction enthalpy of the glass forming reactions and fluxes to the cooling lances differed from the measured values by
amounted to 139.69 k W in total [29]. The DO model was employed an average of only 2.2%. As a result, the combustion and radiation
for calculating thermal radiation effects in three different wavelength modeling approach was directly applied to the simulation of the glass
intervals and the electric potential equation was solved to incorporate melting furnace with hydrogen oxy-fuel combustion. In light of the
the energy input from electric boosting [32]. comprehensive validation strategy that has been employed, it could
The steady-state solutions of both submodels were computed by be reasonably concluded that accurate numerical results were pro-
using double-precision pressure-based solvers. The resulting Poisson’s duced under both natural gas and hydrogen-fired oxy-fuel combustion
equation was handled with the SIMPLE algorithm. All transport equa- conditions.
tions were discretized with a second-order upwind scheme, except from
the radiative transfer equation (RTE), which was discretized with a 4.2. Local effects on flame characteristics
first-order formulation. Equation residuals and several physical mon-
itors were used to assess convergence of each simulation performed. In the following, the numerical results of the fuel substitution study
A common workstation PC with an Intel Core i7-5960X CPU was are presented, focusing first on the influence on the burner operation
employed to calculate the simulation results. Using 8 cores, a converged and the combustion process. The fuel switch resulted in significant
solution of the combustion chamber model was obtained within 24 h, changes in local flame characteristics, primarily due to differences
while a solution of the glass tank model required several days due to in fuel properties such as density, lower heating value (LHV), and
the slow transport of information throughout the melting space and stoichiometric oxygen demand. All observations described were related
refiner. Overall, with approximately 7 to 9 coupling iterations, the final to the conventional oxy-fuel operation mode with natural gas as the
steady-state solution for each fuel configuration was obtained within 6 fuel.
weeks.
4.2.1. Burner operation parameters
4. Results and discussion Given the reduction of volumetric and molar lower heating values
when switching from natural gas to hydrogen, the fuel flow rates of
4.1. Model validation each burner must be increased substantially when maintaining the
thermal energy input into the furnace. With the volumetric lower
A thorough validation of the utilized numerical models is a vital heating value of hydrogen being less than one third of the reference
component in the generation of reasonable results, thereby facilitating value of natural gas, the fuel flow rate per burner increases by a factor
a valid interpretation. The straightforward operation of a full-scale of about 3.3. Correspondingly, the oxidizer flow rates must also be
industrial furnace utilizing pure hydrogen oxy-fuel combustion is still adjusted to maintain the burner equivalence ratio or residual oxygen

4
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 2. Schematic illustration of the performed validation strategy, referring to the findings of preceding works on the glass furnace [29] and the oxy-fuel combustion of
hydrogen [25].

Table 1 As compared with the overall moderate inlet velocities when using
Burner power in (k W) as well as fuel and oxidizer flow rates in (N m3 ∕h) and mean
natural gas as the fuel, supplying the burner with hydrogen yielded
inlet velocities in (m∕s) for burners B1–B6 in NG and H2 oxy-fuel operation.
a strong acceleration of the flow field close to the burner outlet. In
Position Burner NG oxy-fuel H2 oxy-fuel
addition to this acceleration effect, a change in fuel/oxidizer mixing
𝑥∕𝐿 Power 𝑉̇ 𝑓 𝑢𝑒𝑙 𝑉̇ O2 𝑣𝑓 𝑢𝑒𝑙 𝑣O2 𝑉̇ 𝑓 𝑢𝑒𝑙 𝑉̇ O2 𝑣𝑓 𝑢𝑒𝑙 𝑣O2 characteristics was also indicated when viewing the burner velocity and
B1 0.28 568 57.1 115.4 7.25 6.51 189.5 96.0 24.04 5.42 momentum ratios, respectively. Scrutinizing these ratios is particularly
B2 0.47 634 63.8 129.0 8.10 7.27 211.8 107.3 26.87 6.05
convenient as they are independent from the burner power and thus
B3 0.55 735 73.9 149.3 9.38 8.42 245.2 124.3 31.11 7.01
B4 0.66 601 60.5 122.2 7.67 6.89 200.7 101.7 25.46 5.74 directly characterize the overall burner behavior. In Fig. 3, the ratios
B5 0.78 396 40.3 81.5 5.12 4.59 133.8 67.8 16.97 3.82 𝑣𝑓 𝑢𝑒𝑙 /𝑣O2 and (𝜌𝑣2 )𝑓 𝑢𝑒𝑙 /(𝜌𝑣2 )O2 were plotted over the hydrogen content
B6 0.88 396 40.3 81.5 5.12 4.59 133.8 67.8 16.97 3.82 in the fuel gas when operating the industrial PrimeFire 300 burner with
a mixture of methane and hydrogen. This will be of special interest
for the industry, given that a gradual increase in the hydrogen content
in the fuel is anticipated prior to the operation of industrial furnaces
content in the flue gas. For stoichiometric conditions, the absolute
on pure hydrogen. The underlying specific momentum values of the
oxygen flow rate per burner decreases by approximately 17% because
fuel and oxidizer streams, namely (𝜌𝑣2 )𝑓 𝑢𝑒𝑙 and (𝜌𝑣2 )O2 , were again
the lower stoichiometric oxygen demand of hydrogen even outweighs
calculated directly at the outlet of the burner, using a fuel density
the increase in fuel flow rate. These changes in the flow rates directly
between 𝜌𝐶 𝐻4 = 0.716 k g∕N m3 and 𝜌𝐻2 = 0.090 k g∕N m3 and an
affect the velocities of the fuel 𝑣𝑓 𝑢𝑒𝑙 and oxygen 𝑣O2 at the burner outlet.
Together with a significant change in fuel densities, which influences oxygen density of 𝜌O2 = 1.428 k g∕N m3 .
the inlet momentum ratio (𝜌𝑣2 )𝑓 𝑢𝑒𝑙 /(𝜌𝑣2 )O2 , differences in the turbulent Under conventional natural gas fired conditions, the burner was
mixing phenomena and thus the flame shape are to be expected. characterized by a velocity ratio close to unity. This resulted in a
For the present oxy-fuel glass melting furnace that was equipped particularly low fuel/oxidizer mixing rate, since velocity gradients play
with PrimeFire burners, and adopting a constant wet residual oxygen a major role in the generation of turbulent eddies and, consequently,
content for both fuel setups, the obtained fuel and oxidizer flow rates mixing. In this combustion setup, the diverging oxygen stream domi-
𝑉̇ 𝑓 𝑢𝑒𝑙 and 𝑉̇ O2 and inlet velocities 𝑣𝑓 𝑢𝑒𝑙 and 𝑣O2 were calculated in nated the flow characteristics, as it possessed a 60% higher momentum
Table 1. They were determined based on the given flow rates at than the central fuel jet. When operating the burner with approximately
standard conditions (STP: 0 ◦ C, 1 at m) and the respective burner outlet 60 𝑣𝑜𝑙% H2 , equal momentum of both the fuel and oxidizer streams was
cross sections 𝐴𝑓 𝑢𝑒𝑙 = 2190 mm2 and 𝐴O2 = 4925 mm2 , respectively. obtained. Further increasing the hydrogen content in the fuel mixture
The results showed the aforementioned increase in fuel jet velocities ultimately led to a reversal of the observed behavior. In the case of
by a factor of 3.3 and the reduction of oxygen inlet velocities by 17%. pure hydrogen, the fuel jet momentum was found to exceed the oxygen

5
G. Daurer et al. Fuel 380 (2025) 133576

4.2.2. Flame shapes and temperatures

The elevated fuel/oxygen velocity ratio, in conjunction with the
overall higher velocities, suggested an intensified intermixing of the
fuel and oxidizer streams. When considered alongside the much faster
reaction kinetics observed in hydrogen flames, this should exert a
significant influence on the location and extent of the main reaction
zone. As previously outlined, the flame lengths are of particular impor-
tance in a cross-fired burner arrangement. It is essential to avoid direct
contact between the high-temperature flame region and the opposing
refractory wall to guarantee a safe furnace operation and prevent
structural damage. To enable a detailed and quantitative comparison of
the flame lengths for both fuel setups, the flame had to be defined as
such by means of relatable quantities. For this reason, iso-surfaces were
created at the position of the stoichiometric mixture between the fuel
and the oxidizer inlet streams of each burner. This method for defining
the flame front is commonly employed in the literature [26,55]. The
resulting stoichiometric flame surfaces were illustrated alongside the
temperature distribution in the 𝑥𝑦-burner plane in Fig. 5.
Using natural gas as the fuel, the fan-shaped burner outlet and
Fig. 3. Burner velocity ratio 𝑣𝑓 𝑢𝑒𝑙 /𝑣O2 and momentum ratio (𝜌𝑣2 )𝑓 𝑢𝑒𝑙 /(𝜌𝑣2 )O2 for
methane/hydrogen combustion using the PrimeFire 300 burner.
low mixing rate caused the aforementioned pronounced flame spread.
Consequently, the high-temperature region with peak temperatures
exceeding 1800 ◦ C was concentrated on both sides of each burner.
Conversely, in the case of hydrogen oxy-fuel combustion, the nu-
momentum by approximately 25%. The corresponding fuel/oxygen
merically calculated reaction zones exhibited a more centralized and
velocity ratio attained a value of 4.46. These results indicated a severe
localized distribution, situated closer to the fuel nozzle. Although the
change in the fuel/oxidizer mixing characteristics and thus in the flame
fuel jet velocity and momentum were markedly higher, the length of the
shape. To evaluate this aspect first on a local scale, the flow field close
stoichiometric flame surface was observed to decrease for all burners.
to the burners was visualized in Fig. 4. Streamlines were plotted for
This phenomenon was attributed to the accelerated combustion kinetics
the two opposing burners B1 and B2, which were colored according to
their inlet origin and plotted for a constant time duration to enable and the enhanced mixing characteristics resulting from the higher
a qualitative comparison of velocities. In addition, the global flow inlet velocities and increased inlet velocity ratio. In general, the flame
field surrounding the burners was visualized along transversal sections shape associated with hydrogen operation exhibited similarities to a
below, as it also influenced the burner streamlines to some degree. high-impulse jet flame.
In the natural gas setup, both the fuel and oxygen streams spread out As an alternative approach to quantify flame lengths, iso-surfaces of
steadily after entering the combustion space, following the fan-shaped a specific OH mole fraction can be employed to identify the primary
burner nozzle geometry. This was the desired low momentum flame reaction zone, as these radicals are formed as intermediate species
behavior, characterized by a rapidly expanding flame with a noticeable in all combustion processes. Due to their high reactivity and role in
thermal lift. Upon switching to hydrogen, the flow field demonstrated combustion chain reactions, OH may serve as a marker for the heat
the anticipated dominance of the high-velocity fuel stream. Despite release in flames [56–58]. In the present work, a representative value
lower oxygen inlet velocities, both the fuel and oxidizer streams exhib- of 1 𝑣𝑜𝑙%𝑤.𝑏. was chosen to characterize the main reaction zone. The
ited a jet-like flow structure. This phenomenon was attributed to the final flame length at each burner was then obtained as the maximum
aforementioned acceleration effect of the fuel stream, which possessed extension of the respective iso-surface relative to the burner outlet.
a significantly higher momentum and consequently also accelerated In Fig. 6, the obtained flame lengths for all burners and both flame
and straightened the oxygen stream. As previously stated, the global definitions were illustrated in a schematic view of the combustion
flow field exerted a certain influence on the plotted streamlines, par- space.
ticularly on those from burner B2. While the flame of burner B1 could For the iso-surfaces of the stoichiometric mixture, the average rela-
spread almost freely into the combustion space, burner B2 was notice- tive reduction of the flame length for all burners amounted to 25.9%.
ably affected by the surrounding flow structure that was dominated by At the same time, when characterizing the main reaction zone with
the nearby burner B3. In the hydrogen configuration, the global flow an OH molar fraction of 0.01, an average decrease by only 1.1% was
field resulted in a slight deflection of the flame from burner B2 in both determined, because of a distinct behavior of the burners B2 and B6. At
the negative 𝑥-direction and towards the melt surface. In contrast, the
these burners, the region of high OH mole fractions expanded further
flow field surrounding burner B2 in the natural gas setup was domi-
into the combustion space as a result of alterations in the transversal
nated by a transversal circulation that lifted the flame. Overall, the fuel
flow circulations and an overall increase in OH concentration, which
substitution resulted in a loss of the advantages of the low-momentum
was caused by enhanced O-H chain reactions with hydrogen. However,
burner design, such as enhanced load coverage through a high flame
except from these two burners, the results indicated that the mixing and
spread. Furthermore, the effect of an improved flame radiation due
combustion processes were completed more rapidly than with natural
to the intended presence of soot particles naturally disappears when
using a carbon-free fuel. Conversely, the fueling of low-momentum gas as the fuel. These observations were in good agreement with the
PrimeFire burners in a cross-fired configuration with hydrogen proved findings from the experimental work in Ref. [59], where an obvious
critical in terms of furnace operation. As particularly visible in Fig. 4, decrease in the length of an oxy-fuel flame was observed for a hydrogen
the alteration in flow characteristics yielded a more direct and rapid enrichment of up to 15 𝑣𝑜𝑙% at fixed fuel inlet velocities. In preceding
transport of hot flue gases towards the opposite side wall, which, in works of the authors [23,25], a reduction of flame length was also
turn, would increase the risk of local overheating and may reduce the determined for pure hydrogen oxy-fuel combustion at constant thermal
refractory lifetime. These effects were analyzed in a more detailed way input.
in Section 4.3.1. As the discussion thus far was based solely on the With regard to absolute flame temperatures, sphere volumes with a
velocity field and therefore not directly related to the thermochemical diameter of 0.5 m were placed around each burner outlet, and the
processes occurring within the flame, a deeper evaluation of flame corresponding volume-averaged temperatures were evaluated. These
shapes and temperatures was subsequently conducted. temperature values increased by up to 80 K for the hydrogen-fired

6
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 4. Visualized flow field of the burners B1 and B2 for (a) NG oxy-fuel and (b) H2 oxy-fuel combustion. Streamlines for the top illustration were plotted for a constant time
duration of 0.3 s and colored according to their inlet origin (fuel-red and oxidizer-blue). For visualization reasons, the relative position between the burners is not true to scale
as the distance in the 𝑦-direction was reduced. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Top view of the temperature distribution in the xy-plane of the burners for (a) NG oxy-fuel and (b) H2 oxy-fuel combustion. The surface of local stoichiometry is indicated
in black for each of the burners.

burners, which is indicative of accelerated combustion kinetics and 4.3. Global effects on furnace and process characteristics
elevated flame temperatures. Despite all described effects on the shape
and location of the main reaction zone, the definite lengths of the high- 4.3.1. Velocity and temperature field
temperature regions with at least 1800 ◦ C did not change noticeably. After studying the local effects of fuel switching on the flame region
However, the transversal flue gas velocities exhibited a notable increase close to the burners, the focus was directed towards the influences
in the hydrogen configuration, attributable to the substantial accelera- on the global temperature and velocity field inside the glass melting
tion of the fuel stream. This phenomenon indicated a more rapid and furnace. The flow field was visualized in Fig. 7, where contours plots
direct transport of particularly hot flue gases towards the opposite fur- of absolute velocity and flow directions are shown along several planes.
nace wall, which became already evident in the visualization of burner The illustrated locations included the longitudinal center plane at 𝑦 =
streamlines and velocity fields in Fig. 4. Consequently, as highlighted 0 m, an xy-plane at the melt surface as well as transversal yz-planes
in Fig. 5b, the temperature level between the burners B1 and B5 was through the burners B1 and B4.
elevated significantly, which could result in local overheating of the The comparison of the flow field near the burners showed that
furnace refractories. These effects were reviewed in detail during the the higher fuel velocity and momentum with hydrogen combustion
analysis of the global temperature distribution in Section 4.3.1. accelerated the flow over the entire width of the combustion space. As

7
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 6. Flame lengths in the top view on the xy-burner plane for burners B1–B6 in NG and H2 oxy-fuel operation.

Fig. 7. Velocity field in the glass melting furnace for (a) NG oxy-fuel and (b) H2 oxy-fuel combustion.

a result, enhanced transversal circulation was observed at all cross sec- which might negatively affect the longevity of the refractory materials
tions through burners B1–B6. In particular, at the cross section through and would have to be assessed during practical furnace operation.
B4, no distinct flow pattern and low absolute velocities occurred with In contrast to the flow field observed in the combustion chamber,
natural gas as the fuel, while a strong counterclockwise vortex was the melt flow exhibited minimal influence from the fuel substitution.
present in the case of hydrogen combustion. This vortex dominated the The flow pattern was still characterized by a notable longitudinal
entire flow pattern starting from burner B2 through to the rear end recirculation, which was induced by density differences in the region
of the combustion chamber and was also responsible for the increase of the hot spot in proximity to the boosting electrodes. This aspect
in length of the OH-rich reaction zones of burners B2 and B6. In the was already analyzed in great detail in Ref. [34]. Consequently, it
front part of the combustion space, a clockwise transversal vortex was was estimated that the use of hydrogen did not negatively influence
present with both fuels. The flow acceleration induced by the fuel the melting and fining process. However, in light of the considerable
switch resulted in flue gas velocities in the range of 2–4 m/s near the increase in water vapor content within the gas atmosphere, further
side walls and the furnace crown. However, intensified gas transport experimental investigation is required to ascertain the potential impact
was noticeable throughout the combustion space, since the volume- on the glass quality of different glass compositions.
averaged absolute velocities increased from 1.23 m/s to 1.87 m/s as Analog to the velocity field, the temperature distribution inside the
well. This behavior was anticipated, given the difference in combus- glass melting furnace was depicted in Fig. 8 along predefined locations.
tion stoichiometry between the fuels that also influenced the obtained The fuel substitution did not result in significant differences in
flue gas flow rate. By substituting natural gas with hydrogen in the the global temperature distribution. The volume-averaged flue gas
present configuration, the flue gas volumetric flow rate under standard temperatures amounted to 1482 ◦ C for NG and 1480 ◦ C for H2 ,
conditions increased by approximately 9.7% as the higher fuel flow respectively. Similar results were obtained for the glass tank, where
rate dominated the decrease in the oxidizer amount. All these changes the corresponding average melt temperatures were 1226 ◦ C for NG
in flow rates were already summarized and discussed for hydrogen- and 1225 ◦ C for H2 , respectively. As previously described, the NG–O2
enriched oxy-fuel combustion in terms of flow rate factors in a previous flames were characterized by a lower momentum, which was associated
work of the authors [23]. The overall acceleration of flue gas flow with a notable expansion and thermal lift of the flames. This was
when using hydrogen indicated an enhanced convective mixing and particularly evident in Fig. 8 for the burners B2 to B4. Conversely,
heat transfer, not only to the glass bath but also to the furnace walls, the hydrogen flames spread out more gradually over the entire width

8
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 8. Temperature field in the glass melting furnace for (a) NG oxy-fuel and (b) H2 oxy-fuel combustion.

of the furnace. The increase in velocities led to an interaction of the wall temperatures. In the design and construction of future hydrogen-
opposing flames in the rear part of the furnace, as an upward deflection fired melting furnaces, it is essential to consider such aspects in order
of the flame from burner B5 occurred. To prevent a local overheating to ensure that the final burner design, arrangement, and operation are
at the crown and thus achieve a safe operation of the furnace, such evaluated rigorously.
upward flame deflections should be avoided in general. Taking into A small shift was also observed in the global longitudinal temper-
account the aforementioned enhanced transversal transport of hot flue ature distribution: The average wall temperature at the front of the
gases that yielded an increase in near-wall temperatures (see Fig. 5) combustion space decreased by 5.4 K, while it increased by 4.7 K
as well as the described flame interaction, a detailed analysis of the at the rear. This suggested a slightly steeper longitudinal temperature
interior refractory wall temperatures was performed. To this end, the gradient, with lower values near the batch input in the front and higher
temperature distribution was plotted along the side walls and the crown values in the rear of the furnace. In fact, the experimental study by
in Fig. 9. In addition, differences in wall temperatures were visualized Schwarz et al. [23] also showed a lower temperature level in the
to allow for an easier interpretation of the results. far-field of the flame with hydrogen oxy-fuel combustion, which was
Considering the crown temperatures, the thermal lift of the flames attributed to the increasing flame temperatures and thus an enhanced
from burners B2 to B4 in the natural gas setup yielded maximum heat transfer in the flame-near region. In glass melting furnaces, the
temperatures of 1486 ◦ C right above the center of the right wall. temperature and heat flux distribution plays an important role in the
With the fuel substitution, the location of the maximum shifted towards overall melting and fining process. According to studies by Jebavá
the other side right above of the deflected flame from burner B5. et al. [60,61] and Hrbek et al. [62,63], providing a higher fraction of
Peak crown temperatures amounted to 1484 ◦ C there, which proved the total energy input to the batch input region yields enhanced melting
an insignificant decrease in the maximum temperatures at the crown. space utilization and thus melting performance. In order to obtain
The shift in the location of the peak temperature was attributed to precise information on the longitudinal distribution of temperatures
two factors: the deflection of the flame from B5 and the change in and corresponding heat fluxes for both fuel setups, a more detailed
the transversal flow regime to a distinct counterclockwise vortex. The analysis was carried out subsequently.
average temperature at the crown remained relatively constant when In Fig. 10(a), temperature profiles along the center plane were
hydrogen was used as an alternative fuel. Due to the cross-fired burner plotted for the combustion chamber crown, the melt surface as well as
configuration and the increase in transversal velocities, the discrep- the bottom of the glass tank. Concerning the furnace crown, a relatively
ancies between the two fuel configurations were more pronounced at constant temperature decrease by 20 K was indeed achieved with
the side walls. The maximum temperatures at the left wall increased hydrogen up to a relative length of about 0.7. The corresponding melt
from 1487 ◦ C to 1508 ◦ C, while the temperature level at the right surface temperatures exhibited a lower level only in the front part of
wall only showed minor changes. Furthermore, Fig. 9(c) displayed a the melting tank. In the region of approximately 0.43 to 0.74 in relative
substantial temperature increase close to the outlet of all burners. This furnace length, situated directly below the burners B2 through B5, the
was due to the accelerated reaction kinetics in hydrogen flames, which melt surface temperatures exceeded the reference values with natural
yielded an earlier ignition of the non-premixed fuel/oxidizer streams gas. This was due to the increase in flame temperatures, the reduction
that resulted in a steeper temperature increase combined with higher of thermal lift effects and the occurrence of flame interactions, which
peak flame temperatures. Such a local overheating of the burner quarl resulted in both upward and downward flame deflections. In the rear
close to the outlet is typical for burners that have not been designed part of the furnace, the temperature levels at the crown and the melt
for hydrogen operation. In general, it is critical to monitor and limit surface were unaffected by the fuel switch. Furthermore, the melt
wall temperatures, as exceeding the maximum operating temperature temperatures along the entire bottom of the tank displayed no evidence
of the installed refractory materials can significantly reduce the furnace of being influenced by the alteration in combustion configuration,
lifetime and, hence, dramatically increase costs. There are several possi- which aligned with the initial hypothesis of unaffected melting and
bilities to counter the observed effect of elevated side wall temperatures fining conditions.
in the investigated furnace under hydrogen combustion conditions. One
effective concept could be to re-distribute the burner powers according 4.3.2. Heat transfer and process efficiency
to the CFD results: A partial shift of the fuel input from the burners The heat transfer phenomena in industrial furnaces are rather com-
B2–B4 to the burners B1 as well as B5–B6 would decrease the flame plex, given the multitude of interrelated effects that can occur. The
momentum in the central high-temperature region, slow down the fuel substitution may potentially result in variations in the fundamental
transversal transport of hot flue gases and thus reduce the opposing heat transfer mechanisms, as it affects the flue gas properties through a

9
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 9. Wall temperature distribution along the crown (center) and the side walls (left wall — top, right wall — bottom) of the glass melting furnace for (a) NG oxy-fuel and (b)
H2 oxy-fuel combustion. Absolute temperature differences between the two setups 𝑇𝑊 𝑎𝑙𝑙,H2 − 𝑇𝑊 𝑎𝑙𝑙,𝑁 𝐺 were visualized in (c).

Fig. 10. Longitudinal profiles of (a) center plane temperatures and (b) specific heat fluxes for NG and H2 oxy-fuel combustion.

substantial increase in water vapor content. More precisely, the laminar No notable distinction was found in either the absolute heat fluxes
thermal conductivity as well the absorption coefficient of the flue gas or the fundamental transfer mechanisms. The dominating share of the
increases with hydrogen as the fuel gas, while the laminar viscosity is heat transfer occurred due to thermal radiation, which was already
reduced. These aspects were scrutinized in great detail in a preceding shown for other industrial furnaces in a number of publications, as
work on hydrogen oxy-fuel combustion [25]. To assess such effects on e.g. in [43,64]. Similar shares were calculated for all the enclosing
an large scale, the convective and radiative heat transfer rates inside the wall surfaces, confirming the plausibility of the simulations performed.
combustion space of the present glass melting furnace were evaluated. Despite the increase in flue gas velocities and the enhanced thermal
The obtained results of absolute heat fluxes 𝑄̇ 𝑡𝑜𝑡 as well as the respective conductivity of the water vapor-rich flue gas resulting from hydrogen
convective and radiative shares were summarized in Table 2. combustion, the convective heat transfer showed a marginal decline

10
G. Daurer et al. Fuel 380 (2025) 133576

Fig. 11. Radiation properties of flue gas atmospheres from NG and H2 oxy-fuel combustion, including (a) the specific spectral radiation intensities for 1500 ◦ C and 𝑠 = 2.5 m
calculated based on the HITEMP2010 database [65] and (b) gray gas emissivities over the gas temperature for different radiative path lengths calculated with the WSGGM of
Bordbar et al. [54]. A residual oxygen amount of 1%𝑤.𝑏. was assumed for all calculations, resulting in molar fractions of water vapor and carbon dioxide of 𝜈H2O = 0.66 and
𝜈CO2 = 0.33 for NG oxy-fuel, and 𝜈H2O = 0.99 for H2 oxy-fuel, respectively.

Table 2 because of the elevated H2 O partial pressure in the H2 –O2 combustion

Comparison of the heat fluxes in the combustion space for NG and H2 oxy-fuel
atmosphere, since this spectral range is dominated by the emission
combustion.
from water vapor molecules. A distinct contribution of the additional
NG oxy-fuel H2 oxy-fuel
carbon dioxide molecules in NG–O2 combustion atmospheres was only
𝑄̇ 𝑡𝑜𝑡 (kW) Rad. Conv. 𝑄̇ 𝑡𝑜𝑡 (kW) Rad. Conv.
observed between 4.2 to 4.8 μm. In this particular wavelength band,
Side walls −275.0 98.84% 1.16% −275.5 98.88% 1.12% a decline in radiative heat transfer is to be expected when switching
Crown −349.9 98.82% 1.18% −349.7 98.93% 1.07%
from natural gas to hydrogen. This could have a definite impact on the
Melt surface −1656.3 98.86% 1.14% −1653.2 98.96% 1.04%
melting process of different glass melts, as the spectral absorption co-
efficient of typical glasses increases significantly above wavelengths of
2.5 μm [66]. However, the absolute radiation intensities with increas-
in its contribution to the overall heat transfer as compared with the ing wavelengths are relatively low in comparison to those observed
natural gas setup. However, from a practical standpoint, the total in the high-intensity region at shorter wavelengths, which limits the
amount of heat transferred to the glass bath is of primary interest. overall effect of this phenomenon. In general, the radiative calculations
In this regard, an insignificant decrease by approximately 0.2% was have shown that water vapor is responsible for the dominating share
determined using hydrogen. This phenomenon was mainly attributed of heat transfer from flue gases at temperatures above 1000 ◦ C. This
to the aforementioned reduction of temperature levels in the far field was also underscored by elevated gray gas emissivities for H2 -O2 com-
of the flames. bustion atmospheres in Fig. 11(b). The difference to NG-O2 combustion
In order to further validate the numerical results and obtain a atmospheres increases both with a rising gas layer thickness and also
deeper understanding of the occurring heat transfer phenomena, ex- with increasing gas temperatures due to a shift in the spectral radiation
tensive calculations on radiative gas properties were conducted. First intensities towards shorter wavelengths.
of all, detailed radiation spectra were determined for both flue gas at- Performing a more detailed investigation of the heat flux distribu-
mospheres by means of line-by-line calculations using the HITEMP2010 tion provided additional information on the furnace’s melting perfor-
spectral database [65]. The calculations were performed using a Voigt
mance under both operating conditions. Several studies of Polák and
profile and a spectral wavenumber resolution of 0.01 cm−1 . The ana-
Němec [67,68], Jebavá et al. [60,61,69,70] and Hrbek et al. [62,63]
lyzed gas conditions included an atmospheric pressure, a representative
already emphasized the importance of the heat input distribution to
temperature of 1500 ◦ C and a gas layer thickness of 𝑠 = 2.5 m, which
the emerging melt flow patterns and thus the achievable glass quality.
was in close alignment to the conditions observed in the glass melting
The consensus was that a helical melt flow pattern is particularly
furnace under investigation. The spectral radiation intensities obtained
advantageous in terms of space utilization and melting performance,
were plotted in Fig. 11(a), with the line-by-line data interpolated
which is induced by a combination of both transversal and small
onto a wavelength grid with a constant spacing of 0.01 μm for en-
hanced visibility. To extend the validity of the conclusions to different positive longitudinal temperature gradients. These conditions can be
temperatures and gas layer thicknesses, corresponding gray gas emis- achieved by redirecting a greater portion of the total energy input,
sivities were calculated using the WSSGM of Bordbar et al. [54]. These amounting to at least 80%, to the batch input region [60,61]. Given
calculations yielded emissivity profiles over the gas temperature for the slight shift in the longitudinal temperature gradient due to the fuel
both atmospheres, which were plotted in Fig. 11(b) for three specified switch in the investigated furnace, it was also anticipated that this
gas layer thicknesses. In all of the evaluations performed, a residual would have an impact on the heat transfer distribution. In order to
oxygen amount of 1%𝑤.𝑏. was assumed and additional sources of CO2 , evaluate this effect, the melt surface was divided into discrete sections
e.g. through calcination processes inside the melt, were neglected. along the furnace 𝑥-coordinate, with each section measuring 0.4 m
Only marginal differences could be observed in the radiative spectra in length. Subsequently, the total heat flux to the melt in each of the
in Fig. 11(a), which supported the previous results on similar heat specified sub-surfaces was divided by the length, thereby yielding the
fluxes transferred. When evaluating the range with particularly high specific heat flux per furnace length coordinate. This quantity was then
radiative intensities, i.e. from 1 to 4 μm, the intensity peaks in the plotted in form of a profile for both fuel setups in Fig. 10(b). The fuel
hydrogen combustion atmosphere were notably higher than the corre- substitution, and with it the difference in flame characteristics, had a
sponding peaks in the natural gas combustion atmosphere. This was notable influence on the heat flux distribution. In the region below the

11
G. Daurer et al. Fuel 380 (2025) 133576

Table 3 increase. The reduction of heat transfer to the melt by 0.2% was
Comparison of energy fluxes (kW) in the glass melting furnace for NG and H2 oxy-fuel
reflected in a slight decrease in glass outlet temperatures after the
combustion.
refiner, which resulted in a corresponding reduction of the enthalpy
NG oxy-fuel H2 oxy-fuel 𝛥𝑟𝑒𝑙 (%)
flux. However, due to the negligible extent of alteration, the impact
+𝑄̇ 𝐿𝐻 𝑉 3333.0 3333.0 –
of the fuel configuration on the eventual melt output was regarded as
CC −𝐻̇ 𝑓 𝑔 −761.0 –771.0 +1.31%
−𝑄̇ 𝑤𝑎𝑙𝑙,𝐶 𝐶 −909.1 –910.0 +0.10%
inconsequential.
−/+𝑄̇ 𝑚𝑒𝑙𝑡 −/+1656.3 −/+1653.2 −0.19%
4.4. Environmental and economic aspects
+𝑃𝑒𝑙 600.0 600.0 –
−𝐻̇ 𝑔𝑙𝑎𝑠𝑠 −1441.9 –1439.8 −0.15%
GT In order to extend the framework of the presented study and pro-
−𝑄̇ 𝑐 ℎ𝑒𝑚 −140.0 –140.0 –
−𝑄̇ 𝑤𝑎𝑙𝑙,𝐺𝑇 −677.4 –676.8 −0.09% vide context at an operational level, it was also necessary to consider
environmental and economic aspects. At least since the introduction
of emission trading systems, these aspects are closely correlated and
were therefore discussed subsequently with regard to the present glass
hydrogen flame of burner B1, a significantly lower energy input was melting furnace.
achieved into the melt. In turn, an increase in energy input occurred The release of greenhouse gases associated with industrial high-
in the fining region below the burners B2 through B6. A local peak temperature processes represents a significant contributor to global
in specific heat flux was observed in the vicinity of burner B2, with emissions. The principal motivation for switching from natural gas to
values reaching approximately 240 k W∕m. This could be attributed to hydrogen is, of course, the reduction of CO2 emissions. The majority
a slight downward deflection of the flame from burner B2, which was of the greenhouse gas emissions from today’s glass industry are the
a consequence of the formation of a distinct transversal vortex and also result of energy supplied in the form of burned fossil fuels. How-
the interaction of opposing flames, as previously described. The results ever, during glass production, a notable share of the greenhouse gas
were generally consistent with the anticipated enhancement of heat emissions directly results from the melting process. These calcination
transfer in the flame-near region. In contrast, the rapidly expanding gases mainly consist of CO2 and water vapor [71,72]. In typical soda-
natural gas flames of the utilized Primefire burners indeed resulted in a lime glass melts, CO2 is released from soda ash (Na2 CO3 ), limestone
more uniform heat flux distribution, which should be reflected in an en- (CaCO3 ) and dolomite (MgCO3 ⋅CaCO3 ) during reactions that occur
hanced glass quality. However, when focusing on the batch input region at temperatures above 650 ◦ C, as already described in a previous
exclusively, defined by 𝑥 ≤ 0.2 m or 𝑥∕𝐿 < 0.23, almost no difference work [32]. In the investigated glass melting furnace with the examined
was found between the fuel setups. Relative to the total energy input operating conditions, including the production capacity, glass type
into the melt, only 16.42% and 16.24% was provided to this region by and cullet content, the daily CO2 emissions amounted to 20.4 𝑡CO2 ∕𝑑.
natural gas and hydrogen combustion, respectively. This low amount About 15.8 𝑡CO2 ∕𝑑 thereof were caused by the combustion of natural
of heat introduced to the melting zone was identified as the primary gas. The rest, amounting to 4.6 𝑡CO2 ∕𝑑 or 22.5%, resulted from the
factor influencing the melt flow pattern, which was largely governed ongoing calcination reactions. By enforcing a gradual increase in the
by an unfavorable longitudinal melt recirculation. An enhancement of hydrogen content in the fuel, the combustion related greenhouse gas
the melting performance may be attained by reallocating the burner emissions could be effectively reduced. Eventually, when implementing
powers towards a focus on the front of the furnace, i.e. providing more pure hydrogen oxy-fuel combustion, these emissions are completely
fuel to the burners B1 and B2. At least, switching to hydrogen appeared avoided. For the present glass melting furnace, this would amount to
to have no significant influence on the melting capability. a reduction of CO2 emissions by 77.5% or 15.8 𝑡CO2 ∕𝑑. Nevertheless,
With regard to the furnace’s overall energy efficiency, it was nec- due to the release of calcination gases, the overall CO2 emissions of the
essary to determine the principal heat and energy fluxes involved process cannot be reduced to zero regardless of the fuel being used. In
in the process. The energy balances were derived separately for the order to achieve the greatest possible reduction of total emissions, it
combustion chamber (CC) and the glass tank (GT), which were linked is necessary to implement hydrogen combustion in combination with
via the transferred heat flux at the melt surface. In the CC model, an increase in cullet content in the batch. This approach allows for the
energy was supplied in form of the heat release from NG or H2 oxy- recycling of materials that have already undergone the glass-forming
fuel combustion 𝑄̇ 𝐿𝐻 𝑉 . Only a part of this energy was transferred to reactions, thereby eliminating additional sources of CO2 and further
the glass bath 𝑄̇ 𝑚𝑒𝑙𝑡 and thus sustained the melting process. The rest reducing the energy input required.
included the energy losses in form of the ejected hot flue gas 𝐻̇ 𝑓 𝑔 and A comprehensive assessment of the economic aspects was chal-
the wall losses 𝑄̇ 𝑤𝑎𝑙𝑙,𝐶 𝐶 . In the GT model, the transferred melt surface lenging due to the inconsistency and regional variability in the costs
heat flux 𝑄̇ 𝑚𝑒𝑙𝑡 as well as the electric power from the electrodes 𝑃𝑒𝑙 of both the fuel gases and emission certificates. Nevertheless, a pre-
formed the energy input. This positive contribution was balanced by the liminary estimation was conducted based on the current situation in
final extracted product glass enthalpy flux 𝐻̇ 𝑔𝑙𝑎𝑠𝑠 , the chemical reaction the European Union, including latest exchange prices of around 40
heat for the batch to glass conversion 𝑄̇ 𝑐 ℎ𝑒𝑚 and the thermal losses e∕MWh for natural gas (TTF) [73] and 62 e∕𝑡CO2 for carbon emission
through the walls 𝑄̇ 𝑤𝑎𝑙𝑙,𝐺𝑇 . Due to their negligible share, the respective permits (EUA) [74]. Unfortunately, the current market is still lacking
enthalpy fluxes of the fuel, oxidizer and batch materials were not an international hydrogen index or reference exchange price. Efforts
considered. The numerically calculated results for the described energy have yet been made by individual institutions in Germany to introduce
balances with their respective relative difference 𝛥𝑟𝑒𝑙 were summarized a reference index for hydrogen, such as the Hydex proposed by E-
in Table 3. Bridge Consulting [75] and the Hydrix proposed by EEX [76]. Both
The constant energy input via oxy-fuel combustion and the heating pricing indexes are updated on a regular basis, but differ in their
electrodes, in conjunction with an equal glass pull rate, resulted in respective calculation strategies. The Hydex is production cost-oriented
no significant differences in energy fluxes due to the fuel substitution. and is calculated for green, blue, and gray hydrogen. In contrast, the
Most notably, an increase in flue gas losses by 1.31% was observed. Hydrix is based on actual market prices in Germany and is available
This was caused by the aforementioned increase in the flue gas flow for green hydrogen only. Recent prices for green hydrogen were in the
rate, which amounted to 9.7% when related to standard conditions. range of 245 e∕MWh (Hydrix) and 150 e∕MWh (Hydex). Due to the
The elevated flue gas volume was partially offset by a reduction of higher relevance for industrial consumers, who are dependent on the
molar heat capacities by 6.5%, coupled with a slight decline in outlet respective gas market, the Hydrix price was adopted for all subsequent
gas temperature. Consequently, flue gas losses only exhibited a minor considerations. Based on the current price situation, a transition from

12
G. Daurer et al. Fuel 380 (2025) 133576

natural gas to hydrogen appeared to be unfeasible from an economic • The increase in fuel jet momentum with hydrogen accelerated
standpoint, given the projected increase in fuel costs by a factor of the entire flow field inside the combustion space, as the average
6.13. With the considered soda-lime-silica glass melt, which resulted velocity increased by 52%. The arrangement of the jet-type flames
in approximately 203 kgCO2 released per ton of product glass, and a in a cross-fired configuration resulted in slight interactions of
cullet content of 73%, the total costs for CO2 emission permits could opposing flames. In addition, these flames caused an increase
be reduced by 78% when switching to hydrogen. Assuming a similar in maximum side wall temperatures by 21 K, which might be
furnace operation with comparable lifetimes and maintenance costs critical in terms of exceeding the operating temperature of the
with both fuels, this would lead to an increase in total operation costs refractory materials.
by a factor of 4.44. A break-even analysis with constant fuel prices • The global temperature distribution did not indicate substantial
demonstrated that the CO2 emission certificates would have to increase differences, although a minor increase in the longitudinal gradi-
to a value of approximately 1000 e∕𝑡CO2 in order to maintain total ent was observed due to lower temperatures in the batch input
operation costs throughout the fuel switch. This figure may appear region and higher temperatures in the fining region. However, the
to be an unreasonably high expense, but it is, in fact, within the melt flow was not significantly affected, as it remained dominated
range of recently estimated social costs of CO2 in studies on climate by an unfavorable longitudinal recirculation flow.
change impact. In the particularly comprehensive study of Kikstra • Despite the higher water vapor content in the flue gases, similar
et al. [77], social costs of CO2 as high as 3000 $∕𝑡CO2 were proposed heat transfer characteristics were obtained with hydrogen. Ac-
in certain scenarios. It is unfortunate that a corresponding intervention cordingly, the transferred energy fluxes and the resulting process
by global policymakers into CO2 emission pricing is not anticipated in efficiency were not affected considerably, while the combustion-
the near future. In contrast, when current natural gas and emission related CO2 emissions vanished and the total emissions could be
allowance prices were used as a reference, the break-even costs for reduced from 20.4 𝑡CO2 ∕𝑑 to 4.6 𝑡CO2 ∕𝑑.
hydrogen would have to decrease to values well below 60 e∕MWh.
From a fluid mechanical and thermal standpoint, the presented re-
The most realistic break-even scenario in the future will be somewhere sults indicated an adequate furnace operation under hydrogen oxy-fuel
in between, resulting from both a substantial reduction of hydrogen conditions. Nevertheless, the potential chemical interactions between
costs and a simultaneous increase in the costs of natural gas and the elevated amount of water vapor, the refractory materials, and the
CO2 emission permits. It has to be emphasized that for glass melting heated product have yet to be investigated. Future research should thus
and other high-temperature processes being concerned with process- focus on the potential impact on glass quality, given the current lack of
related greenhouse gas emissions, the actual break-even point is further comprehensive understanding regarding its dependence on combustion
influenced by specific process parameters, such as the glass composition atmospheres. In addition, alternative burner configurations should be
and cullet content. evaluated for hydrogen-fired oxy-fuel glass melting furnaces, which
Overall, several interesting concepts have been proposed to estab- take into account the differences in fuel jet velocities and flame shapes.
lish an economically viable utilization of hydrogen as a fuel gas in Ultimately, a clear and transparent pathway for the installation of in-
the industrial sector. Zhang et al. [78], for example, suggested the frastructure for green hydrogen has to be identified at a regional level.
combination of solid-oxide electrolysis cells with H2 –O2 combustion, This is essential in order to facilitate the cost-effective and large-scale
which results in a closed cycle and thus theoretically avoids any waste supply of this alternative fuel gas.
heat and emissions. However, it is challenging to identify a clear global
strategy due to considerable regional differences in the availability of
CRediT authorship contribution statement
green electricity, the demand for high-temperature process heat, and
the costs of natural gas and greenhouse gas emissions. Consequently, it
Georg Daurer: Writing – original draft, Visualization, Validation,
is essential that policymakers, primary energy suppliers, and industrial
Software, Methodology, Investigation, Formal analysis, Conceptualiza-
consumers collaborate at a regional level to develop perspectives for
tion. Stefan Schwarz: Writing – review & editing, Validation, Re-
ramping up a hydrogen economy that considers the entire value chain.
sources, Investigation, Formal analysis. Martin Demuth: Writing –
In the near future, progress in this direction will remain dependent on
review & editing, Validation, Resources, Funding acquisition, Data
targeted funding programs that focus on the installation of hydrogen
curation, Conceptualization. Christian Gaber: Validation, Resources,
infrastructure and the further development of hydrogen combustion
Data curation. Christoph Hochenauer: Writing – review & editing,
technologies.
Supervision, Resources, Project administration, Conceptualization.

5. Conclusion Declaration of competing interest

The influences of a fuel switch from natural gas to hydrogen was The authors declare that they have no known competing finan-
studied numerically in an industrial oxy-fuel glass melting furnace. cial interests or personal relationships that could have appeared to
For this reason, extensive coupled simulations were performed with influence the work reported in this paper.
a validated CFD model in Ansys Fluent. The analysis included an
evaluation of the local effects on the flame characteristics as well as Acknowledgments
the global effects on the furnace operation. The following conclusions
were drawn:
The present work was funded by the Austrian Research Promotion
• With a constant thermal fuel input, the fuel substitution yielded a Agency (FFG) as part of the project ‘Hydrogen HTP’ (Y2 grant project
significant change in fuel/oxidizer momentum and velocity ratios no. 898364, eCall no. 45643452/Y3 grant project no. 907487, eCall no.
of the studied low-momentum Primefire burner. This affected the 50865028). The support is gratefully acknowledged by the authors.
flame in a way that characteristics similar to that of a jet flame
were obtained, including a reduction of flame broadening and Data availability
thermal lift as well as an increase in turbulent mixing rate. Con-
sequently, the mean flame length decreased by 25.9%, whereas The data that has been used is confidential.
the average flame temperatures increased by up to 80 K.

13
G. Daurer et al. Fuel 380 (2025) 133576

References [24] Schwarz S, Daurer G, Gaber C, Demuth M, Hochenauer C. Experimental inves-

tigation of hydrogen enriched natural gas combustion with a focus on nitrogen
[1] Sissa AQ, Delgado Sancho L, Roudier S, Scalet BM, Garcia Muñoz M. Best oxide formation on a semi-industrial scale. Int J Hydrog Energy 2024;63:173–83.
available techniques (BAT) reference document for the manufacture of glass: http://dx.doi.org/10.1016/j.ijhydene.2024.03.190.
industrial emissions Directive 2010/75/EU: integrated pollution prevention and [25] Daurer G, Schwarz S, Demuth M, Gaber C, Hochenauer C. Evaluation of numer-
control. LU: Publications Office of the European Union; 2013. ical modeling and combustion characteristics of hydrogen oxy-fuel combustion
in a semi-industrial furnace. Fuel 2024;369:131690. http://dx.doi.org/10.1016/
[2] Lubitz G, Beutin E, Leimkühler J. Oxy-fuel fired furnace in combination with
j.fuel.2024.131690.
batch and cullet preheating. Amsterdam; 2000, p. 69–78.
[26] Daurer G, Schwarz S, Demuth M, Gaber C, Hochenauer C. Experimental and
[3] Beerkens R. Energy saving options for glass furnaces & recovery of heat from
numerical analysis of industrial-type low-swirl combustion of hydrogen enriched
their flue gases and experiences with batch & cullet pre-heaters applied in
natural gas including OH* chemiluminescence imaging. Int J Hydrog Energy
the glass industry. In: Ceramic engineering and science proceedings. Vol. 30,
2024;80:890–906. http://dx.doi.org/10.1016/j.ijhydene.2024.07.119.
Columbus, Ohio: John Wiley & Sons, Inc.; 2009, p. 143–62. http://dx.doi.org/
[27] Li L, Lin H-J, Han J, Ruan J, Xie J, Zhao X. Three-dimensional glass furnace
10.1002/9780470529010.ch14.
model of combustion space and glass tank with electric boosting. Mater Trans
[4] Beerkens RG, van Limpt H, Lankhorst A, van Santen P. Future energy-efficient
2019;60(6):1034–43. http://dx.doi.org/10.2320/matertrans.M2019044.
and low-emissions glass melting processes. In: Ceramic engineering and science
[28] Cravero C, Marsano D. Numerical simulation of melted glass flow structures
proceedings. Vol. 33, Columbus, Ohio: John Wiley & Sons, Inc.; 2012, p. 15–32.
inside a glass furnace with different heat release profiles from combustion.
http://dx.doi.org/10.1002/9781118217450.ch2.
Energies 2023;16(10):4187. http://dx.doi.org/10.3390/en16104187.
[5] Conradt R. The glass melting process-treated as a cyclic process of an imperfect [29] Daurer G, Raič J, Demuth M, Gaber C, Hochenauer C. Detailed comparison of
heat exchanger. In: Advances in fusion and processing of glass III. Ceramic physical fining methods in an industrial glass melting furnace using coupled
transactions, The American Ceramic Society; 2003, p. 35–44. CFD simulations. Appl Therm Eng 2023;232:121022. http://dx.doi.org/10.1016/
[6] Beerkens RG, van Limpt J. Energy efficiency benchmarking of glass furnaces. j.applthermaleng.2023.121022.
In: Ceramic engineering and science proceedings. Vol. 23, Urbana, Illinois: John [30] Jebavá M, Cincibusová P, Tonarová V, Němec L. Influence of energy distribution,
Wiley & Sons, Inc.; 2002, p. 93–106. http://dx.doi.org/10.1002/9780470294727. melting temperature, kinetics and space geometry on glass melting efficiency. J
ch7, ISSN 0196-6219. Non-Cryst Solids 2024;646:123227. http://dx.doi.org/10.1016/j.jnoncrysol.2024.
[7] Trier W. Glasschmelzöfen - konstruktion und betriebsverhalten. 1st ed.. Berlin, 123227.
Heidelberg: Springer-Verlag; 1984, http://dx.doi.org/10.1007/978-3-642-82067- [31] Lopez-Ruiz G, Blanco JM, Peña A, Romero-Anton N, Azkorra-Larrinaga Z. CFD
0. study of flameless combustion in a real industrial reheating furnace considering
[8] Staněk J. Electric melting of glass. Elsevier Scientific Publishing Company; 1977. different H2/NG blends as fuel. Int J Hydrog Energy 2024;49:1362–74. http:
[9] Zier M, Stenzel P, Kotzur L, Stolten D. A review of decarbonization options for //dx.doi.org/10.1016/j.ijhydene.2023.11.149.
the glass industry. Energy Convers Manag: X 2021;10:100083. http://dx.doi.org/ [32] Raič J, Gaber C, Wachter P, Demuth M, Gerhardter H, Knoll M, Prieler R,
10.1016/j.ecmx.2021.100083. Hochenauer C. Validation of a coupled 3D CFD simulation model for an oxy-
[10] Cappelletti A, Martelli F. Investigation of a pure hydrogen fueled gas turbine fuel cross-fired glass melting furnace with electric boosting. Appl Therm Eng
burner. Int J Hydrog Energy 2017;42(15):10513–23. http://dx.doi.org/10.1016/ 2021;195:117166. http://dx.doi.org/10.1016/j.applthermaleng.2021.117166.
j.ijhydene.2017.02.104. [33] Raič J, Wachter P, Hödl P, Demuth M, Gaber C, Gerhardter H, Prieler R,
[11] Pashchenko D. Hydrogen-rich gas as a fuel for the gas turbines: A pathway Hochenauer C. CFD simulation aided glass quality and energy efficiency analysis
to lower CO2 emission. Renew Sustain Energy Rev 2023;173:113117. http: of an oxy-fuel glass melting furnace with electric boosting. Energy Convers
//dx.doi.org/10.1016/j.rser.2022.113117. Manag: X 2022;15:100252. http://dx.doi.org/10.1016/j.ecmx.2022.100252.
[12] Yang H, Lin X, Pan H, Geng S, Chen Z, Liu Y. Energy saving analysis and thermal [34] Daurer G, Raič J, Demuth M, Gaber C, Hochenauer C. Comprehensive and
performance evaluation of a hydrogen-enriched natural gas-fired condensing numerically efficient CFD model for bubbling in an industrial glass tank. Chem
boiler. Int J Hydrog Energy 2023;48(50):19279–96. http://dx.doi.org/10.1016/ Eng Res Des 2022;186:82–96. http://dx.doi.org/10.1016/j.cherd.2022.07.044.
j.ijhydene.2023.02.027. [35] Honeywell International Inc. PrimeFire 300 | Honeywell. 2024, URL:
[13] Cellek MS, Pınarbaşı A. Investigations on performance and emission character- https://process.honeywell.com/us/en/products/thermal-solutions/burners-
istics of an industrial low swirl burner while burning natural gas, methane, and-heat-exchangers/glass-and-oxygen-burners/primefire-300.
hydrogen-enriched natural gas and hydrogen as fuels. Int J Hydrog Energy [36] Eclipse Inc. PrimeFire 300 product brochure. 2014, URL: https://docuthek.
2018;43(2):1194–207. http://dx.doi.org/10.1016/j.ijhydene.2017.05.107. kromschroeder.com/download.php?lang=en&doc=46699, [Accessed 02 October
[14] BV Glas. HyGlass - Wasserstoffeinsatz in der glasindustrie. 2024, URL: 2024].
https://www.bvglas.de/dekarbonisierung/hyglass-wasserstoffeinsatz-in-der- [37] Eclipse Inc. PrimeFire 300 operating instructions, edition 7.13. 2013, URL: https:
glasindustrie, [Accessed 10 October 2024]. //docuthek.kromschroeder.com/download.php?lang=en&doc=46702, [Accessed
02 October 2024].
[15] HyNet. HyNet North West. 2024, URL: https://hynet.co.uk/, [Accessed 10
[38] NIST Office of Data and Informatics. NIST chemistry webbook. National Institute
October 2024].
of Standards and Technology; 2023, http://dx.doi.org/10.18434/T4D303, URL:
[16] H2GLASS. H2GLASS - Decarbonising our future. 2024, URL: https://h2-glass.eu/,
https://webbook.nist.gov/chemistry/.
[Accessed 10 October 2024].
[39] Carvalho M, Oliveira P, Semião V. Three-dimensional modelling of an industrial
[17] IChemE. Energy: Using hydrogen for glass. 2022, URL: https://www.
glass furnace. J Energy Inst 1988;61:143–56.
thechemicalengineer.com/features/energy-using-hydrogen-for-glass/, [Accessed
[40] Abbassi A, Khoshmanesh K. Numerical simulation and experimental analysis of
10 October 2024].
an industrial glass melting furnace. Appl Therm Eng 2008;28(5):450–9. http:
[18] Bönsch C. H2Glass Partner Steklarna Hrastnik achieves breakthrough with //dx.doi.org/10.1016/j.applthermaleng.2007.05.011.
hydrogen use in glass production. 2023, URL: https://h2-glass.eu/2023/12/20/ [41] Shih T-H, Liou WW, Shabbir A, Yang Z, Zhu J. A new k-epsilon eddy
h2glass-partner-steklarna-hrastnik-achieves-breakthrough-with-hydrogen-use- viscosity model for high reynolds number turbulent flows. Comput & Fluids
in-glass-production/, [Accessed 10 October 2024]. 1995;24(3):227–38. http://dx.doi.org/10.1016/0045-7930(94)00032-T.
[19] Spitra J. Premiere: SCHOTT produces optical glass with 100 percent [42] Casal JM, Porteiro J, Míguez JL, Vázquez A. New methodology for CFD three-
hydrogen. 2024, URL: https://www.schott.com/en-gb/news-and-media/media- dimensional simulation of a walking beam type reheating furnace in steady state.
releases/2024/schott-produces-optical-glass-with-100-percent-hydrogen, Appl Therm Eng 2015;86:69–80. http://dx.doi.org/10.1016/j.applthermaleng.
[Accessed 10 October 2024]. 2015.04.020.
[20] McDonald L. Hydrogen trials show promise at glassmaking plants - The [43] Prieler R, Mayr B, Demuth M, Holleis B, Hochenauer C. Prediction of the
American Ceramic Society. 2023, URL: https://ceramics.org/ceramic-tech- heating characteristic of billets in a walking hearth type reheating furnace using
today/video-hydrogen-trials-show-promise-at-glassmaking-plants/, [Accessed 10 CFD. Int J Heat Mass Transfer 2016;92:675–88. http://dx.doi.org/10.1016/j.
October 2024]. ijheatmasstransfer.2015.08.056.
[21] Cozzi F, Coghe A. Behavior of hydrogen-enriched non-premixed swirled natural [44] Mayr B, Prieler R, Demuth M, Hochenauer C. Modelling of high temperature
gas flames. Int J Hydrog Energy 2006;31(6):669–77. http://dx.doi.org/10.1016/ furnaces under air-fuel and oxygen enriched conditions. Appl Therm Eng
j.ijhydene.2005.05.013. 2018;136:492–503. http://dx.doi.org/10.1016/j.applthermaleng.2018.03.013.
[22] Marragou S, Magnes H, Poinsot T, Selle L, Schuller T. Stabilization regimes and [45] Landfahrer M, Schluckner C, Gerhardter H, Zmek T, Klarner J, Hochenauer C.
pollutant emissions from a dual fuel CH4/H2 and dual swirl low NOx burner. Int Numerical model incorporating different oxidizer in a reheating furnace fired
J Hydrog Energy 2022;47(44):19275–88. http://dx.doi.org/10.1016/j.ijhydene. with natural gas. Fuel 2020;268:117185. http://dx.doi.org/10.1016/j.fuel.2020.
2022.04.033. 117185.
[23] Schwarz S, Daurer G, Gaber C, Demuth M, Prieler R, Hochenauer C. Experi- [46] Mayrhofer M, Koller M, Seemann P, Prieler R, Hochenauer C. CFD investigation
mental investigation of the combustion characteristics in oxy-fuel combustion of of a vertical annealing furnace for stainless steel and non-ferrous alloys strips
hydrogen-enriched natural gas on a semi-industrial scale. Int J Hydrog Energy – A comparative study on air-staged & MILD combustion. Therm Sci Eng Prog
2024;49:323–37. http://dx.doi.org/10.1016/j.ijhydene.2023.07.268. 2022;28:101056. http://dx.doi.org/10.1016/j.tsep.2021.101056.

14
G. Daurer et al. Fuel 380 (2025) 133576

[47] Peeters T. Numerical modeling of turbulent natural-gas diffusion flames (Ph.D. [63] Hrbek L, Jebavá M, Němec L. Energy distribution and melting efficiency in
thesis), Delft, The Netherlands: TU Delft; 1995. glass melting channel: Diagram of melt flow types and effect of melt input
[48] Schluckner C, Gaber C, Landfahrer M, Demuth M, Hochenauer C. Fast and temperature. J Non-Cryst Solids 2018;482:30–9. http://dx.doi.org/10.1016/j.
accurate CFD-model for NOx emission prediction during oxy-fuel combustion jnoncrysol.2017.12.009.
of natural gas using detailed chemical kinetics. Fuel 2020;264:116841. http: [64] Mayr B, Prieler R, Demuth M, Hochenauer C. Comparison between solid
//dx.doi.org/10.1016/j.fuel.2019.116841. body and gas radiation in high temperature furnaces under different oxygen
[49] Prieler R, Demuth M, Spoljaric D, Hochenauer C. Evaluation of a steady flamelet enrichments. Appl Therm Eng 2017;127:679–88. http://dx.doi.org/10.1016/j.
approach for use in oxy-fuel combustion. Fuel 2014;118:55–68. http://dx.doi. applthermaleng.2017.08.054.
org/10.1016/j.fuel.2013.10.052. [65] Rothman LS, Gordon IE, Barber RJ, Dothe H, Gamache RR, Gold-
[50] Mayr B, Prieler R, Demuth M, Hochenauer C. The usability and limits of the man A, Perevalov VI, Tashkun SA, Tennyson J. HITEMP, the high-
steady flamelet approach in oxy-fuel combustions. Energy 2015;90:1478–89. temperature molecular spectroscopic database. J Quant Spectrosc Radiat Transfer
http://dx.doi.org/10.1016/j.energy.2015.06.103. 2010;111(15):2139–50. http://dx.doi.org/10.1016/j.jqsrt.2010.05.001.
[51] Prieler R, Mayr B, Viehböck D, Demuth M, Hochenauer C. Sensitivity analysis of [66] Choudhary MK, Purnode B, Lankhorst AM, Habraken AF. Radiative heat transfer
skeletal reaction mechanisms for use in CFD simulation of oxygen enhanced in processing of glass-forming melts. Int J Appl Glass Sci 2018;9(2):218–34.
combustion systems. J Energy Inst 2018;91(3):369–88. http://dx.doi.org/10. http://dx.doi.org/10.1111/ijag.12286.
1016/j.joei.2017.02.004. [67] Polák M, Němec L. Glass melting and its innovation potentials: The combination
[52] Stephan P, Kabelac S, Kind M, Mewes D, Schaber K, Wetzel T, edi- of transversal and longitudinal circulations and its influence on space utili-
tors. VDI-wärmeatlas: Fachlicher träger VDI-gesellschaft verfahrenstechnik und sation. J Non-Cryst Solids 2011;357(16):3108–16. http://dx.doi.org/10.1016/j.
chemieingenieurwesen. Springer reference technik, Berlin, Heidelberg: Springer; jnoncrysol.2011.04.020.
2019, http://dx.doi.org/10.1007/978-3-662-52989-8. [68] Polák M, Němec L. Mathematical modelling of sand dissolution in a glass melting
[53] Bordbar MH, Wecel G, Hyppänen T. A line by line based weighted sum channel with controlled glass flow. J Non-Cryst Solids 2012;358(9):1210–6.
of gray gases model for inhomogeneous CO2-H2O mixture in oxy-fired http://dx.doi.org/10.1016/j.jnoncrysol.2012.02.021.
combustion. Combust Flame 2014;161(9):2435–45. http://dx.doi.org/10.1016/ [69] Jebavá M, Němec L. Role of the glass melt flow in container furnace examined
j.combustflame.2014.03.013. by mathematical modelling. Ceram - Silik 2018;62(1):86–96. http://dx.doi.org/
[54] Bordbar H, Fraga GC, Hostikka S. An extended weighted-sum-of-gray-gases 10.13168/cs.2017.0049.
model to account for all CO2 - H2O molar fraction ratios in thermal radiation. [70] Jebavá M, Hrbek L, Cincibusová P, Němec L. Energy distribution and melting
Int Commun Heat Mass Transfer 2020;110:104400. http://dx.doi.org/10.1016/j. efficiency in glass melting channel: Effect of configuration of heating barriers
icheatmasstransfer.2019.104400. and vertical energy distribution. J Non-Cryst Solids 2021;562:120776. http:
[55] Poinsot T, Veynante D. Theoretical and numerical combustion. 2nd ed.. R.T. //dx.doi.org/10.1016/j.jnoncrysol.2021.120776.
Edwards, Inc.; 2005. [71] Conradt R. II.24 - The industrial glass-melting process. In: Hack K, editor. The
[56] Kathrotia T, Riedel U, Warnatz J. A numerical study on the relation of OH*, CH*, SGTE casebook (second edition). Woodhead publishing series in metals and
and C2* chemiluminescence and heat release in premixed methane flames. In: surface engineering, Woodhead Publishing; 2008, p. 282–303.
Proceedings of the European combustion meeting 2009. Vienna, Austria; 2009. [72] Shelby J. Introduction to glass science and technology. 2nd ed.. The Royal
[57] Lauer M, Sattelmayer T. On the adequacy of chemiluminescence as a measure Society of Chemistry; 2005.
for heat release in turbulent flames with mixture gradients. J Eng Gas Turbines [73] TRADING ECONOMICS. EU natural gas TTF - Price. 2024, URL: https://
Power 2010;132(061502). http://dx.doi.org/10.1115/1.4000126. tradingeconomics.com/commodity/eu-natural-gas, [Accessed 21 October 2024].
[58] He L, Guo Q, Gong Y, Wang F, Yu G. Investigation of OH* chemiluminescence [74] TRADING ECONOMICS. EU carbon permits - Price. 2024, URL: https://
and heat release in laminar methane–oxygen co-flow diffusion flames. Combust tradingeconomics.com/commodity/carbon, [Accessed 21 October 2024].
Flame 2019;201:12–22. http://dx.doi.org/10.1016/j.combustflame.2018.12.009. [75] E-Bridge Consulting GmbH. Hydex hydrogen index. 2024, URL: https://e-bridge.
[59] Oh J, Noh D, Ko C. The effect of hydrogen addition on the flame behavior of com/competencies/energy-markets/hydex/, [Accessed 21 October 2024].
a non-premixed oxy-methane jet in a lab-scale furnace. Energy 2013;62:362–9. [76] EEX European Energy Exchange AG. EEX-transparency: HYDRIX Germany.
http://dx.doi.org/10.1016/j.energy.2013.09.049. 2024, URL: https://www.eex-transparency.com/hydrogen/germany, [Accessed
[60] Jebavá M, Dyrčíková P, Němec L. Modelling of the controlled melt flow in a 21 October 2024].
glass melting space - Its melting performance and heat losses. J Non-Cryst Solids [77] Kikstra JS, Waidelich P, Rising J, Yumashev D, Hope C, Brierley CM. The social
2015;430:52–63. http://dx.doi.org/10.1016/j.jnoncrysol.2015.08.039. cost of carbon dioxide under climate-economy feedbacks and temperature vari-
[61] Jebavá M, Hrbek L, Němec L. Energy distribution and melting efficiency in ability. Environ Res Lett 2021;16(9):094037. http://dx.doi.org/10.1088/1748-
glass melting channel: Effect of heat losses, average melting temperature and 9326/ac1d0b, Publisher: IOP Publishing.
melting kinetics. J Non-Cryst Solids 2019;521:119478. http://dx.doi.org/10. [78] Zhang S, Zhang N, Smith R, Wang W. A zero carbon route to the supply of
1016/j.jnoncrysol.2019.119478. high-temperature heat through the integration of solid oxide electrolysis cells
[62] Hrbek L, Kocourková P, Jebavá M, Cincibusova P, Němec L. Bubble removal and and H2-O2 combustion. Renew Sustain Energy Rev 2022;167:112816. http:
sand dissolution in an electrically heated glass melting channel with defined melt //dx.doi.org/10.1016/j.rser.2022.112816.
flow examined by mathematical modelling. J Non-Cryst Solids 2017;456:101–13.
http://dx.doi.org/10.1016/j.jnoncrysol.2016.11.013.

15