Fuel Processing Technology 152 (2016) 430–437

Contents lists available at ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Influence of external flue gas recirculation on gas combustion in a coke

oven heating system
Stanislaw Gamrat, Jakub Poraj, Jakub Bodys, Jacek Smolka ⁎, Wojciech Adamczyk
Institute of Thermal Technology, Silesian University of Technology, Konarskiego 22, 44–100 Gliwice, Poland

a r t i c l e i n f o a b s t r a c t

Article history: In this study, external flue gas recirculation was considered to reduce nitrogen oxides (NOx) emissions in the
Received 29 February 2016 heating flues of a coke oven battery. The thermal and prompt NO formation was numerically investigated
Received in revised form 4 June 2016 employing an accurate 3-D representation of the heating flue geometry that the most popular Polish coke
Accepted 8 July 2016
oven battery. Originally the developed model was experimentally validated as a transient coupled model for
Available online 19 July 2016
the representative heating flue and the two coke ovens. Then the coupled model was simplified to the heating
Keywords:
flue model only with realistic boundary conditions on the heating flue and coke oven interface. As a result of
Coke oven battery the heating flue model simulations, the range of the reversed flue gas ratio was found for typical thermal loadings
Heating flue of the heating wall to effectively reduce NOx formation. The obtained results showed the significant effect of the
Coke oven gas considered flue gas recirculation on NOx formation reduction. Namely, the recirculation ratio of 0.2 resulted in
NOx reduction 50% of the NOx reduction efficiency.
External flue gas recirculation © 2016 Elsevier B.V. All rights reserved.

1. Introduction
Coke is a fuel that is used in various technical processes, e.g., in iron
production in the blast furnace [1]. Thus, the coke production technolo-
gy in coke ovens is still an important branch of industry. However, those
production processes may be fairly harmful for the environment. There-
fore, it is important to monitor and reduce the emissions from coke oven
gas (COG) combustion.
The coke oven battery, where the coke is produced, consists of coke
chamber charged by blends of coking coals and adjacent heating flues.
In the combustion chamber, the thermal energy, which is necessary
for the process of the dry distillation of coal, is produced by burning gas-
eous fuel. The blast furnace gas (BFG) or purified COG is often used in
this application [2]. In the coking process, the most important problem
is to obtain the appropriate final value and uniform distribution of the
coke temperature within the whole space of the entire coking chamber.
The desired coke temperature in the centre of the coking chamber is
1273 K, which requires a high temperature combustion process. Such
conditions generate a large amount of NOx, which is harmful to the
environment.
Numerical simulations of thermal processes in coke oven batteries
using computational fluid dynamics (CFD) modelling have been used
in various works. A great deal of attention has been paid to modelling
the coking chamber, where the coal-to-coke conversion occurs. Guo
and Tang [3] presented numerical results of the COG flow pattern and
⁎ Corresponding author.
E-mail address: jacek.smolka@polsl.pl (J. Smolka).
profiles of the temperature, density and porosity for a semi-coke bed.
However, this approach does not accurately include the impact of the
combustion process in a combustion chamber. Therefore, numerical
coupled models of the coke and combustion chambers have also been
developed. Smolka et al. [4,5] introduced a numerical model with
time-dependent heat flux profiles transferred from the heating flues
to the coke bed. Jin et al. [6] also developed a three-dimensional tran-
sient mathematical model that included coupled coking and combus-
tion chambers.
In coke oven operations, in addition to effective coke heating, a low-
emission combustion process is essential. To predict the quantity of gen-
erated NOx, an appropriate mathematical model of NOx formation
needs to be employed. Loffler et al. [7] considered the influence of differ-
ent mechanisms of NOx generation on the overall NOx emissions for the
premixed combustion of natural gas in a one-dimensional plug flow re-
actor. As a result, a simplified model of thermal NOx was presented. The
NOx content was computed based on the temperature and species dis-
tribution. Weiss et al. [8] ran calculations of NOx generation in a coke
oven using BFG and COG as a fuel. For this purpose, the prompt and
thermal NOx fields were predicted, and a non-premixed combustion
model was employed. The aim of that work was to examine the impact
of various combustion conditions on NOx generation. Rieger et al. [9]
modelled the formation of NOx in a blast stove for the combustion of
enriched blast furnace gas. That work considered two methods for
modelling the turbulence-chemistry interaction: the eddy dissipation
concept (EDC) and a PDF-flamelet model. The NOx emission simula-
tions using EDC produced more accurate results when compared to
the actual plant emission values.

http://dx.doi.org/10.1016/j.fuproc.2016.07.010
0378-3820/© 2016 Elsevier B.V. All rights reserved.
 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437
The methods that have been applied for reducing NOx emissions
can be mainly divided into primary and secondary methods [10,11].
Among the primary methods that relay on a proper operation of
combustion process, the most significant are staging air in the com-
bustion process, reburning and flue gas recirculation processes
[12–15]. Currently, recirculation is applied in the considered coke
oven battery as an internal process that occurs through the recircula-
tion window between the upward and downward heating flues. This
method of recirculation works well, however, the amount of recycled
flue gas cannot be controlled. Usually, this process does not allow for
the effective reduction in the generation of NOx. Another problem is
the uncontrolled mixing of flue gas with air and COG. To overcome
these deficiencies, external flue gas recirculation (EFGR) can be con-
sidered. This method is used in many different industrial units [16–
18]. These authors reported a positive impact on the NOx emission
reduction and temperature uniformity in the combustion domain.
This technology can be installed on an existing object, which is an ad-
ditional advantage.
In this paper, a numerical model of NOx reduction using EFGR in
the heating system of a coke oven battery is proposed. Hence, the
main contribution of this paper is to provide the results of an appli-
cation of a well-established method of NOx emission reduction to
the most popular coke oven battery in Poland. To predict the influ-
ence of the coking chamber on the combustion chamber operation,
the heat flux profile was introduced as a boundary condition. Typical
values of the heat flux were obtained from previous work [4] using
coupled models at various instants of the coking process, and the
model was finally formulated as a steady-state process. Due to the
high temperatures that occur in the combustion chamber, the
model described by Loffler et al. [7], in which the thermal mechanism
of the NOx generation has a prevailing role, was also employed in
this study. In EFGR, changes in flow, species and the temperature of
the air and flue gas mixture are taken into account. The presented
Fig. 1. Cross-sections through the heating flue.
431
results of this work demonstrate some tendencies and potential of
the EFGR technology. In particular, the results with a gradually mod-
ified amount of the recycled flue gas are discussed and compared
with cases without EFGR. In general, this primary NOx reduction
method exhibited a very promising effect in the coke oven heating
system for recirculation ratios ranging from 0.2 to 0.25.
2. Numerical model
2.1. Heating flue model
In general, a coke oven battery consists of a series of heating flues ar-
ranged in rows on both sides of the coke ovens. Heating flues are sup-
plied from the bottom with air and the purified COG. The air and COG
are preheated in a ceramic heat exchanger that works as a regenerator
by accumulating heat from the hot flue gases from the combustion
chamber. As shown in Fig. 1, a single combustion chamber is divided
into two heating flues: an upward heating flue, where air and fuel are
supplied, and a downward heating flue, where the flue gas outlet is
located.
In the upper part of chamber, there is a bridge window, where the
gases change their flow direction. In the lower part of the heating flue,
there is a recirculation window designed for internal recirculation be-
tween both flues. The combustion cycle for one direction inside the
chamber lasts approximately 20 min. Then, the flow direction is re-
versed, i.e., the air inlet becomes the flue gas outlet and vice versa. The
whole coking process takes approximately 15 to 20 h.
Therefore, every numerical analysis of the heating system opera-
tion in the coke oven battery should be performed as a time varying
case. This type of approach was presented in the previous authors'
work [4]. The developed model the whole heating system was for-
mulated as a transient coupled model of the heating flue and coke
oven. On the other hand, the combustion process itself can be a
432 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437

steady-state process if a single reverse period is taken into account. heat transfer was modelled using the Discrete Ordinates (DO) model,
Changes in this process originate from the intensity of heat transfer where the Radiative Transfer Equation (RTE) is defined as [19]:
through the heating walls only. Then, a steady-state mathematical
model can be employed that includes different values of the heat ∇  ðIλ ðr; sÞsÞ þ ðaλ þ σ s ÞI λ ðr; sÞ
flux as a boundary condition on the heating walls. In this study, Z4π
σs
such a formulation was used to simulate the NOx emission from ¼ aλ n2 I bλ I λ ðr; s0 ÞΦðs  s0 ÞdΦ0 ð6Þ
þ 4π
the heating flue model. 0

2.2. Governing equations where I is the intensity, which is defined as a function of the place where
the ray starts r and its direction s, a is the absorption coefficient, σ is the
The general governing equations that describe the combustion pro- Stefan-Boltzmann constant, n is the refractive index, Ibλ is the black
cess in the heating flue are [5] now discussed. body intensity, s' is the scattering direction vector, and Φ is the phase
coefficient. The spectral radiation properties of the optically active gas-
• Energy: es, i.e., CO2 and H2O, were calculated using the Weighted Sum of Gray
Gases (WSGG) model. In this approach, the radiation from the gases is
approximated by a finite number of gray gases of different absorption
!
∂ðρuÞ X coefficients and one transparent gas. The most accurate and theoretical-
þ ∇  ðρvhÞ ¼ ∇  ðk∇T þ τ  vÞ−∇  hi J i þ Sh ly sound procedure for evaluating the values of the WSGG model coeffi-
∂t i
ð1Þ cient is to use the spectral emissivity data of CO2 and H2O molecules
X ZT
p stored in an appropriate database such as HITEMP 2010 [20]. The proce-
J i ¼ −ρDi ∇yi h¼ hi yi hi ¼ cpi dT u ¼ h−
i
ρ dure for computing the radiation constants is cumbersome and time
T ref
consuming. The data available in Ansys FLUENT are those proposed by
[21]. These data are limited to H2O and CO2 partial pressure ratios equal
where ρ is the density, u is the internal energy, t is the time, h is the spe-
to 1 and 2, respectively. Several research groups are focusing on the im-
cific enthalpy, k is the effective thermal conductivity, v is the velocity
pact of the atmosphere on radiative transfer and are developing compu-
vector, T is the temperature, τ is the effective stress tensor, J is the diffu-
tation tools suitable for addressing ratios smaller than one, as can
sion mass flux, S is the source term in the transport equation of the sca-
happen when large amounts of flue gases are recirculated.
lar quantity, D is the mass diffusion coefficient, m is the mass fraction, cp
is the isobaric specific heat and p is the pressure.
2.3. Combustion model
• Momentum:
For the combustion process modelling, the global reaction mecha-
nism was used [5]. The turbulent-chemistry interaction was modelled
∂ðρvÞ using EDC [22]. The two-step COG oxidation reaction mechanism was
þ ∇∙ðρvvÞ ¼ −∇p þ ∇∙τ þ ρg þ S ð2Þ assumed.
∂t

where g is the gravitational acceleration vector, and S is the source term CH 4 þ 1:5O2 ¼ CO þ 2H2 O ð7Þ
in the momentum transport equation.
CO þ 0:5O2 ¼ CO2 ð8Þ
• Continuity:
H 2 þ 0:5O2 ¼ H2 O ð9Þ

∂ρ X C 2 H 4 þ 2O2 ¼ 2CO þ 2H2 O ð10Þ

þ ∇  ðρvÞ ¼ Syi ð3Þ
∂t i C 3 H 6 þ 3O2 ¼ 3CO þ 3H2 O ð11Þ
where Smi is the mass source term of the ith species. C 6 H 6 þ 4:5O2 ¼ 6CO þ 3H 2 O ð12Þ
• Species transport for i − 1 elements:

2.4. Properties of the gases and walls

∂ðρyi Þ
þ ∇∙ðρvyi Þ ¼ ∇∙ρDi ∇yi þ Syi ð4Þ Material properties for the gas domain and walls such as the density,
∂t
thermal conductivity, specific heat and mass diffusivity coefficient were
defined as functions of the temperature. Table 1 shows the general def-
initions of these quantities.
• Two-equation k–ε realizable turbulence:

   Table 1
μ Material properties of the flue gases and heating flue walls (temperatures in K).
ρðv∇κ Þ ¼ ∇  μ þ T ∇κ þ Sðμ T ; κ; vÞ−ρε
Pr k
   ð5Þ Quantity Gaseous subdomain Solid subdomain
μ ε ε2
ρðv∇ε Þ ¼ ∇  μ þ T ∇ε þ C ε1 Sðμ T ; κ; vÞ−C ε2 ρ Density Ideal gas model ρ = 1800
Pr ε κ κ
Specific heat Polynomial function of the c = 0.25116 ∙ T + 768.633
Thermal temperature and the mass-weighted k = 0.000628 ∙ T + 0.9334
conductivity average for the mixture
where κ is the turbulent kinetic energy, μ is the dynamic viscosity, Pr is Dynamic –
the Prandtl number, ε is the dissipation rate of the turbulent kinetic en- viscosity
ergy, and Cε1 and Cε2 are constants. The gaseous phase was modelled as Diffusivity –
an absorbing-emitting medium of the temperature and composition Absorption Weighted sum of gray gas model –
coefficient
that depend on the absorption coefficient. In this work, the radiative
 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437 433

2.5. Heating wall and inlet boundary conditions

The air and COG compositions, mass flow rates and temperatures
that were defined at both inlets of the heating flue for the case without
EFGR are listed in Table 2.
For the boundary conditions at the heating walls, the heat flux
resulting from the coking chamber neighbourhood was assumed. The
value of the heat flux profile was the same for both heating walls and
was determined on the basis of the results obtained from the coupled
model of the heating flue and coking chamber. These results were consis-
tent with the heat demand monitored in the typical coke oven battery. In
the first hours of the coking process, when intensive evaporation occurs,
the average value is 5000 W/m2. In the second stage, the average heat
flux drops to 2000 W/m2. Therefore, the following average values of the
heat flux boundary conditions were employed in this study: 2000, 3000,
4000 and 5000 W/m2. It has to be noted that in the first minutes after
the coke oven charging with wet hard coal, the heat demand is much
higher than the mentioned values. This period was taken into account
in this study.

2.6. External flue gas recirculation Fig. 2. Scheme of the EFGR system.

To simulate the external recirculation process, a fixed value of the inlet

mass flow rate at the oxidizer duct inlet was assumed. This is practically – mass fractions:
the easiest and the cheapest way to implement EFGR in an existing
plant. This means that the amount of the air mass flow rate was gradually
lowered, whereas the flue gas mass flow rate was increased. Moreover, _ rec ∙youti þ m
m _ air ∙yairi
the simulated EFGR system required updates of the mixture composition yini ¼ ð16Þ
m_ in
and the inlet temperature. User Defined Functions (UDFs) in the commer-
cial software program ANSYS Fluent were employed to simulate the ex- where yini, youti and yairi are the mass fractions of the ith component at
ternal mixing according to the scheme presented in Fig. 2. the mixture inlet, the flue gas outlet and the air inlet duct, respectively.
The proposed algorithm to numerically analyse EFGR in the heating
system of the coke oven battery included the following equation groups – total enthalpies:
for the mixture, air and the recirculated flue gas.
– inlet mass flow rates: H in ¼ Hair þH rec ð17Þ

_ air ∙∑yairi ∙cpair ∙T air

H air ¼ m ð18Þ
i
_ in ¼ idem:
m ð13Þ
_ rec ∙∑youti ∙cpout ∙T out
H rec ¼ m ð19Þ
_ air ¼ m
m _ in −m
_ rec ð14Þ i

where Hin, Hair and Hrec are the mixture inlet, the air inlet and the
where m _ in , m
_ air and m
_ rec are the total (mixture) inlet, air inlet and the
recirculated flue gas total enthalpies, respectively, Tair and Tout are the
recirculated flue gas mass flow rates, respectively. The mass flow rate
air inlet and flue gas outlet temperatures, respectively, and cpairi and cpouti
of the recirculated flue gas was computed on the basis of the recircula-
are the air inlet and flue gas outlet specific heats of the ith component
tion ratio that was defined as
gas.
_ rec
m In the last step, the mixture inlet temperature was determined on
R¼ ð15Þ the basis of the inlet enthalpy, the mass flow rate, and the mass-aver-
_ out
m
aged specific heat of the mixed air and flue gas:
_ out is the flue gas outlet mass flow rate.
where m
H in
T in ¼ ð20Þ
_ in ∙cpave
m
Table 2
The mass flow rates, temperatures and mass fractions of the air and the COG inlets of the
heating flue. where Tin is the mixture inlet temperature, and cpave is the mass-aver-
Air COG
aged specific heat of the mixture that was computed from

Mass flow rate, kg/s 0.0334 0.0022 _ air

m _ rec
m
Temperature, K 1473 1173 cpave ¼ ∑cpair ∙ þ cpout ∙ ð21Þ
Mass fraction O2 0.2068 0.2090
i _ in
m i m_ in
CO2 – 0.1363
H2O 0.0159 0.3420 As mentioned before, the discussed algorithm of EFGR was imple-
N2 0.7773 0.1049
mented into Ansys Fluent using Parallel User-Defined Functions. The
CO – 0.2305
H2 – 0.0883 code after each iteration computed the flue gas mass flow rate and the
CH4 – 0.3090 face-average mass fractions, specific heat and temperature on the
C2H4 – 0.0476 heating flue outlet (see Fig. 1). Moreover, the program read the recircu-
C3H6 – 0.0204 lation ratio from the external file. Then on the basis of these information
C6H6 – 0.0079
and the fixed air properties, the updated values of the mass fractions,
434 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437

specific heat and temperature for the air-heating flue mixture were
computed. These values were then prescribed to the air inlet boundary
condition (see Fig. 1).
The base case for the comparison purposes was computed for the
zero recirculation ratio. Since the mass flow rate of the air-flue gas mix-
ture was assumed constant at the inlet of the heating flue model, more
flue gas recirculated resulted in a lower value of the air excess ratio.
Then the recirculation ratio was increased until the value of 0.3 that
corresponded to the air excess ratio near unity. Moreover, numerical
simulations were performed for the mentioned recirculation ratios in-
volving heat flux values ranging from 2000 W/m2 to 5000 W/m2 as
the boundary conditions for both heating walls.
All the computations were performed employing high-performance
computing cluster installed at ITT, Gliwice, Poland. The approximate
time consumption for a typical case run in parallel on 20-node machine
was 30 h. The convergence of the numerical computations was moni-
tored using the governing equation residuals and the field values.
2.7. NOx formation
During the gaseous fuel combustion process, NOx is mainly pro-
duced due to prompt and thermal mechanisms. The first mechanism
is mostly focused on modelling hydrocarbon radicals such as CH and
CH2. that react very rapidly with nitrogen and, in consequence, NOx
are generated. In the present work, the global kinetic parameters were
used to control the rate of NOx formation without modelling the
many subsequent reactions [23,24]. The second mechanism contributes
more to the overall NOx formation. The source of NOx is nitrogen pro-
vided to the combustion chamber with an oxidiser. This path is
governed by three reactions:
increase with the increase in the recirculation ratio. This was expected
because with the higher recirculation ratio, more flue gases, of which
CO2 and H2O are the main components, are recirculated. The changes
in these fractions in comparison with other recirculation ratios can be
clearly observed near the air inlet because the recirculated flue gas is
added to the air. Moreover, visible changes can be observed in the
downward flue where triatomic gases are well mixed with other flue
gas components. The increases in the CO2 and H2O fractions cause an
abatement of the O2 mole fraction. This causes a decrease in the air ex-
cess ratio. A recirculation ratio of 0.3 results in nearly 0% of the O2 mole
fraction, so further analysis of the higher recirculation ratio was aban-
doned. In this case, the maximum CO2 mole fraction reached approxi-
mately 9%, and H2O mole fraction reached a value of approximately 24%.
The volume-averaged and maximum gas temperatures in the com-
bustion chamber are presented as a function of the recirculation ratio
in Fig. 4. A general observation is that the average gas temperature in-
creased with the growing recirculation ratio. This phenomenon has
two causes. The first cause is the rise of the hot flue gas fraction at the
mixture inlet. The second is the higher content of triatomic gases in
the flue gas mixture, which are not transparent to radiation. Simulta-
neously, this caused a drop in the maximum gas temperature. Obvious-
ly, the highest changes in the maximum gas temperature for the
particular recirculation ratio were obtained for the case with the highest
heat flux value of 5000 W/m2, which was transferred from the heating
flue to the coke oven. The average temperature was practically the
same for that case. In contrast, the maximum temperature of approxi-
mately 2500 K was almost constant for a heat flux of 2000 W/m2. This

O þ N 2 →N þ NO ð22Þ

N þ O2 →O þ NO ð23Þ

N þ OH→H þ NO ð24Þ

The first reaction requires very a high activation energy that is used
to break the triple bonds of N2 to atomic nitrogen and nitric oxides. For
this reason, the thermal mechanism becomes important only at temper-
ature higher than 1800 K [25]. In NOx modelling using this mechanism,
the third reaction is also important. This reaction should be taken into
account for stoichiometric and fuel rich regions. The thermal NOx mech-
anism assumes that the rate of formation and consumption of N atoms
are equal, so that the NOx formation can be defined as

kr;1 kr;2 ½NO2

1−
d½NO k f;1 ½N2 k f;2
¼ 2k f;1 ½O½N2  ð25Þ
dt kr;1 ½NO
1þ
k f;2 ½O2  þ k f;3 ½OH

where kf,1, kf,2 and kf,3 are the rate constants for the forward reactions,
and kr,1 and kr,2 are the rate constants for the reverse reactions. To
solve this equation, the concentrations of O and OH must be deter-
mined. In this study, the O and OH radicals were modelled using a par-
tial-equilibrium model [26]. This model accounts for the participation of
the third-body in the creation of the oxygen radicals that is important
when high temperature differences occur within the computational
domain.

3. Results and discussion

In Fig. 3, the mole fractions of the gas mixture species such as CO2
and H2O are presented as a function of the recirculation ratio. The field
is depicted in the cross-sectional view through burners (see Fig. 1). In Fig. 3. Mole fraction fields in the cross-sectional view shown in Fig. 1 for (a) CO2 and (b)
Fig. 3, it can be observed that the amounts of CO2 and H2O generally H2O.
 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437
Fig. 4. (a) Volume-averaged and (b) maximum temperatures in K of gases in the heating
flues.
value dropped slightly with the increase in the recirculation ratio. Gen-
erally, the recirculation process caused higher drops in the maximum
temperature for higher heat flux values. In contrast, the volume-aver-
aged temperature recirculation caused more significant changes in the
lower heat flux values. For a heat flux value of 5000 W/m2, the vol-
ume-averaged temperature increased by approximately 30 K. For a
heat flux of 2000 W/m2, the increase was approximately 125 K.
A similar tendency can be observed in Fig. 5, in which the inlet and
outlet gas temperatures are presented for the various recirculation ra-
tios. The temperatures increased like the volume-averaged temperature
in Fig. 4(a) as the recirculation ratio increased. However, this tempera-
ture increase was strongly dependent on the heat flux value. The
highest inlet temperature differences occurred for a heat flux of
2000 W/m2. For the case without recirculation, the air inlet temperature
was 1473 K, as for the other cases. However, for the recirculation ratio of
0.3, the inlet temperature of air-flue gas mixture increased to above
1700 K. Hence, the inlet temperature rose by approximately 250 K. In
the case with the highest heat flux of 5000 W/m2, the change is less
than 50 K. Between the cases of the lowest and the highest heat demand
for the 0.3 recirculation ratio, the inlet gas temperature difference
reached over 200 K. The outlet temperature was not as affected by the
recirculated flue gases. The largest value of the gas temperature differ-
ence recorded was approximately 120 K that obtained for a heat flux
of 2000 W/m2. Generally, the recirculation caused greater increases in
those temperatures for the higher heat flux values.
435
Fig. 5. Temperatures in K of gases at the (a) inlet and (b) outlet of the heating flue.
The temperature distributions obtained for the two extreme bound-
ary values of the heat flux boundary conditions are illustrated in Fig. 6.
These results are consistent with those presented in Fig. 4. For the
2000 W/m2 heat flux illustrated in Fig. 6(a), the maximum flue gas tem-
perature in the upward heating flue was practically the same for all of
the recirculation ratios. However, the high temperature region signifi-
cantly increased as the amount of recirculated flue gas increased. More-
over, the maximum temperature in the downward heating flue was
significantly higher for the highest recirculation ratio. This caused the
temperature to be much more uniform as the recirculation ratio
increased.
For the case of the highest boundary heat flux presented in Fig. 6(b),
the maximum temperature significantly dropped with more
recirculated flue gases. It is clearly visible in flame region in the lower
part of the upward flue. In addition, the entire field became more uni-
form. However, it must be noted that that the characters of these fields
were very similar.
Fig. 7(a) presents the influence of EFGR on the rate of the NOx gen-
eration for a typical heat flux transferred to the coke oven. The highest
nitrogen oxides emission, which was approximately 1000 PPM, oc-
curred for the case without EFGR. At the coke plant, the emissions
were 30% lower than the value obtained through the numerical calcula-
tions. Then, the value gradually dropped to approximately 200 PPM
when 30% of the flue gas was recirculated. However, it is worth noting
that the emission reduction was obtained for the highest recirculation
window that corresponded to a low value of the air excess ratio. In
436 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437
Fig. 6. Temperature field in K in the cross-sectional view shown in Fig. 1 for wall heat
fluxes of (a) 2000 W/m2 and (b) 5000 W/m2.
practice, a case with 20% recirculated flue gas should be considered. A
NOx generation reduction of nearly 50% in the emissions can be obtain-
ed in that case. In this case, the flue gas recirculation led to the oxygen
content decrease in a primary combustion zone and it ensured a more
uniform distribution within the whole combustion zone. As a result of
the oxygen dilution that substantially slowed down the NOx forming re-
actions, the NOx reduction was observed. The other aspect that influ-
enced the NOx reduction was lowering the peak flame temperature
which led to a reduction of the thermal NOx formation as well. Fig.
7(b) shows the molar fraction fields of NOx. It can be observed that
the recirculation implementation eliminated the generation of large
amounts of NOx in the region located in front of the flame. For recircu-
lation ratios of 0.2 and 0.3, this region is practically imperceptible.
4. Conclusions
To develop a profound understanding of the results obtained from
the numerical analysis of a coke oven heating system, the temperature
and species distribution in the flue gas domain were described as a func-
tion of the heat flux boundary condition and recirculation ratio. This nu-
merical investigation presented an increase in the volume-averaged gas
temperature and a decrease in the maximum gas temperature. Both
phenomena became more distinct at higher recirculation ratios. More-
over, the use of EFGR produced a more uniform temperature field
throughout the combustion chamber. This side effect of EFGR may pro-
vide a beneficial impact on the quality and repeatability of coke produc-
tion. Furthermore, the positive influence of the application of EFGR on a
reduction in NOx was confirmed. For the recirculation ratios that were
Fig. 7. (a) The NOx emissions in PPM and emission reductions and (b) the mole fraction
fields of NOx in the cross-sectional view shown in Fig. 1 as a function of the
recirculation ratio.
used, the NOx pollution abated linearly with EFGR intensification. For
a recirculation ratio of 0.3, the recirculation efficiency exceeded 80%.
Therefore, EFGR also appears to be a crucial primary method to reduce
NOx emissions in the coke industry.
Acknowledgements
The authors gratefully acknowledge the financial support of the pro-
ject Competent mechanical engineers for energetic sector no
POKL.04.01.02-00-131/12. The work of JS and WA was partially sup-
ported by the statutory research fund of the Faculty of Power and Envi-
ronmental Engineering.
References
[1] M.A. Diez, R. Alvarez, C. Barriocanal, Coal for metallurgical coke production: predic-
tions of coke quality and future requirements for cokemaking, Int. J. Coal Geol. 50
(2002) 389–412.
[2] R. Razzaq, C. Li, S. Zhang, Coke oven gas: availability, properties, purification and uti-
lization in China, Fuel 113 (2013) 287–299.
[3] Z. Guo, H. Tang, Numerical simulation for a process analysis of a coke oven, China
Particuol. 3 (2005) 373–378.
[4] J. Smolka, L. Slupik, A. Fic, A.J. Nowak, L. Kosyrczyk, 3-D coupled CFD model of a pe-
riodic operation of a heating flue and coke ovens in a coke oven battery, Fuel 165
(2016) 94–104.
[5] J. Smolka, A. Fic, A.J. Nowak, L. Kosyrczyk, 3-D periodic CFD model of the heating sys-
tem in a coke oven, Int. J. Numer. Methods Heat Fluid Flow 24 (2014) 891–906.
[6] K. Jin, Y. Feng, X. Zhang, M. Wang, J. Yang, X. Ma, Simulation of transport phenomena
in coke oven with staging combustion, Appl. Therm. Eng. 58 (2013) 354–362.
[7] G. Loffler, R. Sieber, M. Haraseka, H. Hofbauer, R. Hauss, J. Landauf, NOx formation in
natural gas combustion — a new simplified reaction scheme for CFD calculations,
Fuel 85 (2006) 513–523.
[8] C. Weiss, J. Rieger, B. Rummer, Formation and control of nitrogen oxide in the
heating system of a coke oven, Environ. Eng. Sci. 29 (2012) 555–562.
[9] J. Rieger, C. Weiss, B. Rummer, Modelling and control of pollutant formation in blast
stoves, J. Clean. Prod. 88 (2015) 254–261.
 S. Gamrat et al. / Fuel Processing Technology 152 (2016) 430–437
[10] J. Warnatz, U. Maas, R.W. Dibble, Combustion, fourth ed. Springer, Berlin, 2006.
[11] K. Annamalai, I.K. Puri, Combustion Science and Engineering, CRC Press, Boca Raton,
2006.
[12] B. Staiger, S. Unterberger, R. Berger, K.R.G. Hein, Development of an air staging tech-
nology to reduce NOx emissions in grate fired boilers, Energy 30 (2005) 1429–1438.
[13] K.-T. Wu, H.T. Lee, C.I. Juch, H.P. Wan, H.S. Shim, B.R. Adams, S.L. Chenc, Study of syn-
gas co-firing and reburning in a coal fired boiler, Fuel 83 (2004) 1991–2000.
[14] J. Baltasar, M.G. Carvalho, P. Coelho, M. Costa, Flue gas recirculation in a gas-fired
laboratory furnace: measurements and modelling, Fuel 76 (1997) 919–929.
[15] F. Qian, C. Chyang, C. Yeh, J. Tso, Effect of operating conditions on NOx and CO emis-
sions in a pilot-scale vortexing fluidized- bed combustor with flue gas recirculation,
Chem. Eng. Technol. 36 (2013) 268–276.
[16] R.K. Wilk, Low-Emission Combustion, SUT Publishing House, Gliwice, 2002.
[17] E. Croiset, K. Thambimuthu, A. Palmer, Coal combustion in O2/CO2 mixtures com-
pared with air, Can. J. Chem. Eng. 78 (2000) 402–407.
[18] J. Li, W. Yang, W. Blasiak, A. Ponzio, Volumetric combustion of biomass for CO2 and
NOx reduction in coal-fired boilers, Fuel 102 (2012) 624–633.
[19] ANSYS Fluent Theory Guide, Relaese 15.0, ANSYS, Inc, 2014.
437
[20] L.S. Rothman, I.E. Gordon, R.J. Barber, H. Dothe, R.R. Gamache, A. Goldman, V.
Perevalov, S.A. Tashkun, J. Tennyson, HITEMP, the high-temperature molecular
spectroscopic database, J. Quant. Spectrosc. Radiat. Transf. 111 (2010) 2139–2150.
[21] T.F. Smith, Z.F. Shen, J.N. Friedman, Evaluation of coefficients for the weighted sum
of gray gases model, J. Heat Transf. 104 (1982) 602–608.
[22] B.F. Magnussen, On the structure of turbulence and a generalized eddy dissipation
concept for chemical reaction in turbulent flow, Nineteeth AIAA Meeting, St. Louis,
1981.
[23] G.D. Soete, Overall reaction rates of NO and N2 formation from fuel nitrogen, 15th
Symposium on Combustion 1975, pp. 1093–1102.
[24] K. Kuo, Principles of Combustion, Wiley, New Jersey, 2005.
[25] M. Drake, S. Correa, R. Pitz, W. Shyy, C. Fenimore, Superequilibrium and thermal ni-
tric oxide formation in turbulent diffusion flames, Combust. Flame 69 (1987)
347–365.
[26] W.P. Adamczyk, S. Werle, A. Ryfa, P. Kozołub, Application of the computational
method for predicting NOx reduction within large scale coal-fired boiler, Appl.
Therm. Eng. 73 (2014) 343–350.