Professional Documents
Culture Documents
Abstract
A systematic experimental campaign has been carried out to investigate the scaling of NOx emissions from a
moderate or intense low-oxygen dilution (MILD) combustion furnace operating with a parallel jet burner system
in which the reactants and the exhaust ports are all mounted on the same wall. Its maximum capacity was 20 kW
from the fuel and 3.3 kW from air preheat, with a turndown ratio of 1:3. The burner system was configured to
achieve high dilution of the incoming reactants. A comprehensive data set comprising 191 global measurements of
temperature and exhaust gas emissions is presented, together with temperature contours on the furnace centerline
plane. It was found that, although heat extraction, air preheat, excess air, firing rate, dilution, and fuel type all
affect global NOx emissions, they do not control NOx scaling. The combined effects of these global parameters
can be ultimately characterized by a furnace temperature and a global residence time. A temperature–time scaling
approach, previously reported for open jet diffusion flames, proved to be a useful tool for comparison of NOx
emissions from highly diluted furnace environments regardless of the furnace/burner geometries. Regression-
based predictions found the characteristic temperature to correlate with 85% of the data with an accuracy of only
±50%. The leading-order approach also showed that the jet exit Froude number is of limited value for NOx scaling
in the MILD regime. Because of the weak dependence on temperature observed in the data and the moderate
magnitude of the measured temperatures, it is deduced that the prompt-NO and/or N2 O-intermediate pathways
are of significance comparable to that of the thermal-NO pathway. The analysis also suggests that NOx formation
is controlled neither by kinetics nor by mixing, and hence the conditions inside this furnace approach or span the
range in which Damköhler numbers are of order unity, Da = O(1).
© 2008 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.2 (1-15)
2 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.3 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 3
a warm-up time of about 3 h from a cold state un- verted to temperature units with a USB-TC data log-
til steady-state operation. The inner surfaces of the ger. According to the manufacturer’s specifications,
combustion chamber were coated with a commercial the uncertainties associated with linearization, cold-
product, “Kaowool J-Coat” rigidizer, to improve the junction compensation, and system noise produce a
resistance to erosion caused by high gas velocities. maximum error of ±1.07 ◦ C in a measurement range
The MCF has full optical access through five open- from 250 to 1768 ◦ C. This represents only 0.13% er-
ings (120 × 76 mm2 ) equally spaced along the ver- ror in the worst case.
tical direction on three sides of the furnace. These Eighteen thermocouple ports (see Fig. 1) were
openings can accommodate interchangeable insulat- equally spaced in the vertical direction, through
ing window plugs or UV-grade fused silica windows. which a thermocouple could be traversed. Typically,
The chamber is slightly pressurized to prevent air temperatures were monitored simultaneously by fixed
ingress, and high-temperature gaskets were used to thermocouples in two regions of interest, one at the
seal all surface joints for safety. A U-shaped cool- top section and one at the bottom section of the com-
ing loop with variable heat exchange area was used bustion chamber. The center of the air nozzle exit
to control the heat load. The heat exchanger can be plane is defined to be the origin of a Cartesian co-
inserted through any of the five window openings (po- ordinate system (x, y, z). The top temperature was
sitions A1–A5). measured on the furnace centerline (0, 0, 542.5) and
The configuration of reactants and exhaust ports was denoted as the furnace reference temperature.
was optimized using a computational fluid dynam- The bottom temperature was measured flush with
ics (CFD) modeling study [23]. The furnace was de- the inner surface (0, 140, 42.5) and was denoted as
signed for a maximum capacity of 20 kW from the the reference wall temperature. Temperature profiles
fuel and 3.3 kW from air preheat with a turndown ra- were also measured for a selected baseline case. The
tio of 1:3. The air was preheated with a Leister LE thermocouple probes were traversed in the radial di-
3000 electrical air heater that has a built-in poten- rection (y-direction) across different planes at each
tiometer for air temperature control. The burner con- hole to provide a total of 72 points.
sists of a single air nozzle on the axis of the furnace, The correction to the temperature measurements
and four exhaust ports and four fuel ports arranged for radiation was determined from a steady-state en-
symmetrically in ring patterns (see Fig. 1). The dis- ergy balance on the thermocouple bead surface. The
tance between the center of the air nozzle and the actual gas temperature (Tg ) was determined from the
centers of the exhaust ports and fuel nozzles are 55 measured temperature (Ttcb ) and net radiation flux
and 110 mm, respectively. In this arrangement the in- between diffuse-gray surfaces in an enclosure accord-
coming air stream is separated from the fuel streams ing to
by outgoing exhaust gases, much as in the fuel direct
injection (FDI) concept introduced by Nakamachi et Atcb εtcb N 4 4
w=1 σ G1w (Ttcb − Tw )
Tg = Ttcb + , (1)
al. [24] and in the IFRF burner design [25]. This h
ensures that the reactants are diluted with products where Atcb is the thermocouple bead surface area,
before any combustion reaction can occur. The final εtcb is the thermocouple bead surface emissivity, N
straight sections of the air and fuel supply lines have is the total number of surfaces (i.e., walls) in the en-
length-to-diameter ratios of L/D ≈ 10 and L/d ≈ 18 closure, σ is the Stefan–Boltzmann constant, G1w is
(id = 2 mm), respectively. the fraction of radiation emitted by the thermocouple
bead that is incident on a particular wall surface w
2.2. Temperature and global emission measurements in the furnace enclosure and is absorbed [26], Tw is
the surface temperature of wall w in the enclosure,
Time-averaged temperatures were measured with and h is the convection coefficient calculated for a
bare, fine-wire type R (Pt/Pt–13% Rh) thermocou- sphere [27]. Catalytic and oxidation effects were ne-
ples of 254-µm-diameter wires and 1.2-mm-diameter glected, as well as conductive heat transfer along the
beads under steady-state conditions. This is sufficient ceramic sheath. The G1w correction factor was cal-
for reliable measurements of the mean temperature culated for each probe position. The method led to a
after correction for radiation, given the low temper- maximum correction of 9% and a typical correction
ature fluctuations in the MILD regime, and provides of less than 2% between raw and actual gas tempera-
satisfactory spatial resolution while being sufficiently tures, so that the total uncertainty is within ±1.2%.
robust to avoid probe breakage. For accurate posi- Exhaust gas composition was measured inside
tioning, the thermocouple wires were supported by the exhaust pipe with a TESTO 350XL portable
a 6-mm-diameter rigid ceramic sheath, with the final gas analyzer. The gas analyzer was equipped with
5 mm left exposed. The differential voltage signal was electrochemical sensors with nominal accuracies,
continuously sampled at 1 Hz and automatically con- respectively, for CO (±5%: 0–2000 ppm, ±10%:
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.4 (1-15)
4 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
The parametric study investigated a total of 191 Fig. 2. Schematic diagram of heat exchanger insertion posi-
measurement points under various test conditions. In tions.
all cases, thermal equilibrium was ensured and sta-
ble MILD combustion was achieved, based on the
requirement of no visible flame and NOx emission from 0.67 to 0.98, which is equivalent to an excess air
levels being below 70 ppm throughout the experi- range of 33–2%. The location of the heat exchanger
ment. The oxidant was air, whose temperature was was varied from the top window A1 (centerline at
varied from ambient to 780 ◦ C. Either commercial z = 524.5 mm) to the bottom window A4 (centerline
natural gas (NG) or liquefied petroleum gas (LPG) at z = 176.5 mm) and its exposed surface area varied
was used as the fuel to provide thermal inputs over from 0.015 m2 for 25% insertion to 0.06 m2 for 100%
the range 7.5–20 kW. Table 1 shows typical compo- insertion (Fig. 2). The dilution of the fuel stream was
sition and properties of both fuels. Two alternative- varied, using either N2 or CO2 , over the range 0–76%
sized fuel nozzles were employed, 2.0 or 3.0 mm by mass. Table 2 presents a summary of all test con-
in diameter. The global equivalence ratio (φ) ranged ditions.
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.5 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 5
Table 1
Typical fuel properties and composition
Fuel LHVb (A/F)st c MWd N2 e CO2 e CH4 e C2 H6 e C3 H8 e C4 H10 e C5 H12 e C6 H14 e
NGa 51.154 17.11 17.71 1.278 2.084 91.36 4.364 0.62 0.20 0.055 0.04
LPGa 45.993 15.66 44.07 0.73 0.05 – 1.12 96.32 1.78 – –
a Gas analysis provided by Origin Energy Australia.
b Lower heating value (MJ/kg).
c Stoichiometric air-to-fuel ratio (kg/kg).
d Molecular weight (kg/kmol).
e Mole fraction (Xi ) (%).
Table 2
Summary of all test conditions
Fuel type Data d Ydil Q̇fuel φ Q̇hx Tair Tfurnace τG
points (mm) (%) (kW) (%) (◦ C) (◦ C) (s)
NG 145 2.0 0 7.6–20.4 0.67–0.98 17–51 20–780 768–1368 2.8–11.4
2 3.0 0 15.6 0.80 18 450 1295 3.9
NG/CO2 8 3.0 38–76 15.6 0.80 16–34 454 1053–1309 1.8–3.1
NG/N2 3 2.0 41–70 7.9–10.5 0.80 25–29 23–445 951–1129 3.7–4.5
5 3.0 33–71 15.6 0.80 18 450 1285 1.6–3.0
LPG 1 3.0 0 15.0 0.80 21 450 1290 3.9
LPG/CO2 10 2.0 31–71 9.9–14.4 0.75–0.81 24–41 458 1031–1261 1.3–3.5
4 3.0 50–74 15.0 0.80 20 450 1268–1306 1.1–2.0
LPG/N2 9 2.0 19–65 14.7 0.80 18–46 447 1046–1298 1.1–3.0
4 3.0 45–70 15.0 0.79 20 450 1285–1326 0.9–1.8
The baseline case of this project was defined as the the sudden decrease in furnace reference temperature
2.0-mm burner with a 15-kW thermal input (Q̇fuel ) of and by the fourfold drop in NOx emissions. The dif-
natural gas, an equivalence ratio of 0.80, and 450 ◦ C ferences seen in the levels of O2 , and consequently
air preheat (Tair ). For this condition the inlet jet ve- of CO2 , during the transition to the MILD regime
locities for air and fuel are U = 29.1 m/s and u = were caused by a slight increase in pressure drop
17.3 m/s, respectively. The reference position for the in the fuel supply system. However, it is clear that
heat exchanger was position A3 (centerline at z = these changes did not significantly influence perfor-
292.5 mm) with 25% insertion. mance. Fine-tuning of the fuel flow rate was typically
The furnace was started from a cold state with a needed to compensate for small pressure variations
nonpremixed flame stabilized by a bluff body. When in the supply line. For excess air levels of 25% or
the temperature at the lower corner of the furnace, more (φ 0.8), CO and UHC emissions were be-
which was identified as safety TC in Fig. 1, exceeded low the detection limit. From the time the furnace
autoignition (≈800 ◦ C) and the exhaust temperature was switched to the MILD regime, about 1.5 h was
exceeded a set threshold (≈600 ◦ C), the furnace was needed to reach steady-state MILD combustion con-
switched to the MILD combustion mode. ditions. Under steady-state conditions, the tempera-
ture difference between the top and the bottom sec-
tions of the furnace was around 110 ◦ C for the base-
3. Results line case. The low temperature gradient throughout
the furnace is an important characteristic of this com-
3.1. Temperature and pollutant emissions during bustion regime. The performance characteristics of
transition the furnace/burner have been discussed previously in
more detail [29]. Together they demonstrate that the
Fig. 3 shows the temperature and emissions mon- present furnace/burner configuration is able to sus-
itored as a function of time for the baseline case. The tain stable MILD combustion, while maintaining high
dashed line represents the time when MILD com- modulation ratios for various thermal input, heat ex-
bustion was activated, the bluff-body was retracted, traction, and air preheat conditions. The investigation
and no flame front was visible [29]. The successful also identified cases in which the MILD regime was
transition between the two regimes is evidenced by established without air preheat.
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.6 (1-15)
6 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
Fig. 3. Temperature and emissions recorded during the tran- Fig. 4. Measured temperature distribution (◦ C) for the base-
sition from conventional to MILD combustion for the base- line case (d = 2.0 mm, Q̇fuel = 15 kW NG, φ = 0.80,
line operating conditions (d = 2.0 mm, Q̇fuel = 15 kW NG, Tair = 450 ◦ C, 25% insertion at position A3) at plane y = 0.
φ = 0.80, Tair = 450 ◦ C, 25% insertion at position A3). The arrows represent the incoming air (solid blue) and fuel
(solid green) streams and the outgoing exhaust (dotted red)
3.2. Thermal field measurements stream. The dashed line shows the heat exchanger (HX) lo-
cation and insertion length. (For interpretation of the refer-
Measurement of the thermal field is a necessary ences to color in this figure legend, the reader is referred to
step in identifying and analyzing potential regions of the web version of this article.)
pollutant formation and/or destruction. Fig. 4 presents
the measured temperature distribution for the base- temperature measurements presented in Fig. 3. The
line case in the x–z plane through the furnace cen- uniformity suggests that a nonadiabatic global fur-
terline. The solid blue arrow represents the incoming nace temperature can reliably represent the reaction
air stream (x = 0 mm), the solid green arrows the zone temperature. Nonetheless, the most representa-
fuel streams (x = −110 mm and x = 110 mm), and tive location for such a measurement will depend on
the dotted red arrows the outgoing exhaust stream the heat exchanger location. The region in the vicinity
(x = −55 mm and x = 55 mm). The dashed line of the heat exchanger is about 100 ◦ C cooler than that
shows the heat exchanger location (window A3) and on the opposite side of the furnace. From the knowl-
insertion length. The temperature contours were gen- edge of heat extracted (Q̇hx ) and material properties,
erated using a Kriging-type interpolation algorithm the average heat exchanger surface temperature (T̄hx )
[30] from the original measurement grid, which con- was calculated assuming one-dimensional conduction
sisted of 3 × 4 × 6 points in each Cartesian direction. to be T̄hx ≈ 70 ◦ C. Hence, local temperatures around
A nonsymmetrical temperature field is noticeable the heat exchanger are expected to be even lower than
inside the furnace and is attributed to the presence 1175 ◦ C. This will result in local extinction and par-
of the heat exchanger inserted 25% through win- tial oxidation near to the cold surface.
dow A3. The highest temperature region was mea-
sured to be 1300 ◦ C in the region 80 < x < 100 mm 3.3. Effect of fuel dilution
and 200 < z < 300 mm. The temperature differen-
tial is only about 100 ◦ C within about two-thirds of A high degree of dilution of the reactants is one of
the furnace (z 180 mm). This is consistent with the the necessary conditions for establishing MILD com-
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.7 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 7
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.8 (1-15)
8 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
Fig. 6. Relationship between NOx emissions and furnace Fig. 7. Relationship between NOx emissions and exhaust
temperature, Tfurnace , for all tested operating conditions. temperature, Texhaust , for all tested operating conditions.
The solid line represents a linear curve fit for all data. The ex- The experimental data points are grouped according to fuel
perimental data points are sorted according to fuel type, dilu- input power and fuel type. No distinction is made with re-
ent, and burner size. Diamonds: NG, circles: LPG, thin line spect to fuel dilution and burner size. The thick solid black
symbols: no fuel dilution, solid symbols: N2 dilution, thick line represents a linear curve fit for all data. The other thin
line symbols: CO2 dilution, small symbols: d = 2.0 mm, and colored lines are linear curve fits for each fuel input power
large symbols: d = 3.0 mm. bin. Open symbols: all cases burning NG, and solid symbols:
all cases burning LPG.
PSR model show the same general trend observed in fuel type, diluent, and burner size. The solid line rep-
the measurements of decreasing gas temperature and resents a linear fit to all of the data. Although there
NOx emission levels with fuel dilution. However, the is a general trend of increasing NOx emissions with
maximum predicted NOx emissions are only 6 ppm furnace temperature, the wide scatter and low corre-
for both diluents. A sensitivity analysis (not shown) lation coefficient (R 2 = 0.50) show that the depen-
showed that the recirculation stream has little effect dence is weak. It is worth highlighting that very low
on NOx formation at 1280 ◦ C. It also revealed that the NOx levels, <10 ppm, are achieved for furnace tem-
reactor temperature would need to be at least 200 ◦ C peratures below 900 ◦ C. However, at temperatures
higher to reproduce the NOx emission levels observed around 1400 ◦ C, NOx emissions exceeding 60 ppm
in the experiments, which is when the thermal-NO were measured in the MILD regime.
mechanism starts to dominate. Fig. 7 shows NOx emissions data plotted against
Finally, we assess the cases in which the burner di- the exhaust temperature, which, while lower than the
ameter is varied. It is noticeable from Fig. 5 that NOx furnace temperature, avoids the bias in some data due
emissions increase when the burner size is increased. to the location of the cooling loop relative to the ther-
For a constant diluent mass fraction and thermal input mocouple. It can be seen that there is a stronger in-
from the fuel, an increase in nozzle diameter causes fluence of temperature for all data (R 2 = 0.56) and
a decrease in the fuel jet velocity by 44% and a de- for individual fuel input ranges. Even though this
crease in the jet momentum flux by the same amount general correlation is somewhat stronger, it is suffi-
for the same diluent. A lower fuel jet momentum will ciently poor to demonstrate that other parameters are
reduce the entrainment of hot furnace gases and hence also important. This result shows that a representative
change the mixing characteristics close to the burner furnace temperature alone does not adequately scale
exit. In an attempt to confirm this trend, a third fuel NOx emissions.
nozzle of diameter 4.6 mm was also tested. However, Fig. 8 shows NOx emissions plotted against heat
stable MILD combustion could not be sustained for extraction. The heat extraction, Q̇hx , is expressed as a
the baseline operating conditions, and a flame was percentage of the total thermal input. As expected, an
visible. A discussion of the stability of the MILD increase in heat extraction, which conversely means a
regime is beyond the scope of the present paper. decrease in temperature, leads to a decrease in NOx
emissions. Because the heat extraction is only part of
3.4. Global nitrogen oxides emission correlations the net heat output, it is not surprising that the correla-
tion with heat extraction is even weaker (R 2 = 0.39)
Fig. 6 presents the NOx emissions as a function than in Fig. 7.
of the furnace reference temperature (see also Fig. 1). There is no visible correlation with fuel type, dilu-
The experimental data points are sorted according to ent, or burner size, either in Fig. 6 or in Fig. 8. No
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.9 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 9
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.10 (1-15)
10 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
previous work by Weber [16] are also displayed. The [NOx ] [mol/m3 ]
fuel jet exit Froude number (Fre ) is the ratio of jet fρrct EINOx
momentum flux to buoyant forces and is calculated = XNO XNO . (4)
from u2 /gd, where g is the gravitational acceleration. MWNO X + MWNO2 X 2
NOx NOx
From Fig. 9a, it is evident that some data depart
from the correlation for the present study (NG+LPG) Adapting the definition presented by Turns and
by almost an order of magnitude. Indeed, several data Myhr [35] for open flames to enclosed multiple-jet
sets are essentially independent of the Froude number, systems, the global residence time (τG ) is charac-
indicating that other parameters affect NOx emissions terized by the combustion chamber volume and the
and the role of buoyancy is negligible. This is consis- volumetric flow rate of hot reactants as
tent with the Froude number exceeding unity by three 4VMCF fρrct
orders of magnitude (Fre > 103 ), which suggests a τG [s] = , (5)
π d 2 unρfuel
momentum-controlled system. Even those data sub-
sets that do scale with Fre , such as the correlation where VMCF is the MCF volume, f is the fuel mixture
(NG reference) represented in blue in Fig. 9b, exhibit fraction, ρrct is the density of the mixture of reactants
a slope slightly different (∝Fr0.68
e ) from those found at furnace temperature, d is the fuel nozzle diameter,
earlier [32]. Here, ∝ is used to indicate proportional- u is the fuel inlet velocity for a single nozzle, n is
ity. Costa et al. [34] found similar departures for their the number of round fuel nozzles, and ρfuel is the fuel
momentum-controlled open jet methane flames. density at ambient temperature. The use of the furnace
Figs. 9b and 9c show that the slope for NG as the volume to represent the reaction zone is a reasonable
fuel (∝Fr0.75
e ) is significantly different from that of approximation for MILD combustion conditions and
LPG as the fuel (∝Fr0.38 e ). The dependence on fuel is consistent with the global treatment used in other
type is expected from the model of Røkke et al. [32]. studies involving furnaces [14]. The fuel mixture frac-
The effects of fuel dilution can also be observed in tion is calculated from f = φ/φ + (A/F )stoic and
Fig. 9c, when the N2 dilution cases (∝Fr0.41 e ) and implicitly incorporates the air mass flow rate. Here,
the CO2 dilution cases (∝Fr0.30e ) are separately cor- (A/F )stoic is the value of stoichiometric air-to-fuel
related. ratio. The denominator in Eq. (5) is directly related to
Fig. 10 shows the variation of the NOx emission the fuel mass flow rate, which in turn determines the
index, EINOx , with the momentum flux ratio of air, fuel input power, also referred to as the firing rate.
Gair , to fuel, Gfuel . Once again, the experimental A convective time scale, d/u, is commonly used as
data points are sorted according to fuel type, diluent, the scaling parameter in momentum-controlled open
and burner size. Here Gair /Gfuel = nṁair U/ṁfuel u, jet flames. For those flames, d/u is directly propor-
where ṁair is the air mass flow rate and ṁfuel is the tional to residence time because it controls flame vol-
fuel mass flow rate. It is evident that no obvious cor- ume. The convective time scale strongly influences
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.11 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 11
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.12 (1-15)
12 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
Table 3
Regression coefficients for Eq. (6)
Database Parameters A B C Temperature range
(ln(mol/cm3 s)) (1/s) (K) (K)
Turns et al. [36] Tflame , τG 1.1146 −0.7410 −16,347 1900–2350
Tfurnace , τG −6.141 −0.387 −3,634 1050–1650
Present study τG −8.106 −0.733 0 N/A
Tehxaust −2.249 0 −6,681 760–1070
4. Discussion
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.13 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 13
the thermal-NO route. Rather, the fuel type depen- differences between CO2 and N2 dilution. However,
dence is further evidence that the prompt-NO mech- the results from Fig. 10 demonstrate that, although
anism is significant. The dependence on fuel type mixing has some influence, it does not control global
has also been observed in Fig. 9. The relevance of NOx yields. Therefore other effects, such as those due
the prompt-NO pathway in the MILD regime has to chemical kinetics, also influence the NOx forma-
been speculated on by a few researchers [5,17,25, tion in this furnace.
48]. In the predictions of Mancini et al. [48], only The PSR calculations from Fig. 5 indicate that the
5% of global NOx emissions were calculated to be assumption of a well-stirred reactor (Da 1) used
of prompt-NO origin. On the other hand, it has been by some researchers [52,53] does not appear to be
observed that the N2 O-intermediate mechanism can reasonable in the present MCF system. This is con-
represent up to 65% of the total NOx formation in sistent with the recent modeling study of Mancini et
lean-premixed methane combustion in a jet-stirred re- al. [54]. They also found that the combustion chamber
actor at atmospheric pressure in a temperature range cannot be considered simply as a well-stirred reac-
from 1140 to 1570 ◦ C [49]. Further assessments are tor. They proposed a reactor network model, which
required to quantify the contributions of each path- consisted of 22 PSRs, to predict NOx emissions from
way to the overall NOx emissions. a semi-industrial-scale furnace. Their results empha-
Several parameters seem to alter NOx formation sized the importance of the mixing characteristics
in MILD combustion conditions and hence add to the close to the burner exit. This implies that the NOx for-
complexity of the scaling problem. The previous sec- mation process is not controlled by chemical kinetics.
tions have outlined the influence of heat extraction, air In summary, the analysis described above high-
preheat, oxygen concentration, dilution, and fuel type lights the fact that neither momentum nor kinetics
on global NOx emissions. While all these parameters has a controlling influence on NOx emissions, but
are not independent from each other, the combined that rather both are important. This suggests that NOx
effect can be ultimately associated with a character- formation in the present furnace may be within the
istic nonadiabatic temperature and residence time, as regime where mixing time scales are comparable to
shown in Fig. 11. However, the fairly poor fit found chemical time scales, and hence Damköhler numbers
in the regression analysis presented in Fig. 12 demon- are of order unity, Da = O(1). Therefore, in order to
strates that temperature and residence time influence, accurately predict NOx emissions from MILD com-
but do not control, the scaling of NOx . It is not pos- bustion systems, any model will need to include the
sible to isolate one or two controlling parameters, be- effects of both mixing and chemistry.
cause many parameters seem to have an influence of
comparable magnitude.
Although the Froude number has been used as a 5. Summary and conclusions
scaling parameter in furnace conditions with conven-
tional flames [16], Fig. 9 suggests that it is not the Global measurements of temperature and exhaust
most relevant parameter to scale NOx emissions from gas emissions obtained for 191 conditions in a 20-kW
furnaces operating in the MILD regime. This is be- MILD combustion furnace operating with a parallel
cause buoyancy forces are particularly weak in an jet burner system have been presented. The main find-
environment with low temperature gradients. Never- ings of the analysis are as follows:
theless, some dependence on Fre is evident, which
implies that fuel jet momentum is still an important 1. The presence of a water-cooled heat exchanger
parameter. The data for the two different burner sizes dominated the temperature gradients and caused
shown in Fig. 5 give further evidence of the influ- a nonsymmetrical temperature differential of
ence of momentum. A lower fuel jet momentum will about 100 ◦ C across the furnace.
lead to a change in the jet trajectory, much like that in 2. Overall NOx emission values as low as 5 ppmv
the strong-jet/weak-jet model of Grandmaison et al. dry at 3% O2 and as high as 66 ppmv dry at 3%
[50], and hence reduce the amount of dilution with O2 were recorded in the MILD regime for fur-
hot combustion products prior to mixing with the ox- nace temperatures of up to 1370 ◦ C.
idant stream. These changes will affect local species 3. Fuel dilution of up to 76% by mass, with either
concentrations and temperatures in the reaction zone, CO2 or N2 , reduced NOx emissions by up to
altering the ratio between mixing and reaction times, 48% and 10%, respectively. These levels of di-
i.e., the local Damköhler number (Da = τmix /τchem ). lution did not affect furnace temperatures signifi-
A low local Damköhler number can favor nonequilib- cantly, with a maximum reduction of only 56 ◦ C.
rium effects, such as superequilibrium O atoms [51]. The fuel type, diluent type, and fuel jet mo-
Both entrainment levels of hot combustion products mentum were all found to influence global NOx
and nonequilibrium chemistry may account for the yields. However, neither mixing nor chemistry
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.14 (1-15)
14 G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–•••
effects were found to be dominant, suggesting [4] A. Cavaliere, M. De Joannon, Prog. Energy Combust.
that the characteristic Damköhler number spans, Sci. 30 (2004) 329–366.
or is on the order of, unity. The failure of PSR [5] M. Katsuki, T. Hasegawa, Proc. Combust. Inst. 27
calculations to predict the NOx emissions further (1998) 3135–3146.
[6] H. Tsuji, M. Morita, A.K. Gupta, M. Katsuki, K. Kishi-
supports this hypothesis, as does the finding that
moto, T. Hasegawa, High Temperature Air Combus-
the Froude number scaling revealed a limited de- tion: From Energy Conservation to Pollution Reduc-
pendence of NOx emissions on jet momentum. tion, CRC Press, 2003.
4. A global NOx production rate, calculated based [7] B.B. Dally, E. Riesmeier, N. Peters, Combust. Flame
on a global residence time and a characteris- 137 (2004) 418–431.
tic furnace temperature, was found to give the [8] M. Flamme, Appl. Therm. Eng. 24 (2004) 1551–1559.
best scaling of the data across all conditions as- [9] P.R. Medwell, P.A.M. Kalt, B.B. Dally, Combust.
sessed. However, the dependence on temperature Flame 148 (2007) 48–61.
was found to be much weaker than that expected [10] H. Zhang, G. Yue, J. Lu, Z. Jia, J. Mao, T. Fujimori, T.
Suko, T. Kiga, Proc. Combust. Inst. 31 (2007) 2779–
based on the thermal-NO mechanism alone. This
2785.
trend also applied to other MILD combustion [11] R. Weber, J.P. Smart, W. Van der Kamp, Proc. Com-
systems, irrespective of the furnace/burner geom- bust. Inst. 30 (2004) 2623–2629.
etry. Furthermore, the regression analysis found [12] S.R. Turns, Prog. Energy Combust. Sci. 21 (1995) 361–
that neither temperature nor global residence 385.
time provides a sufficiently strong correlation to [13] A.D. Al-Fawaz, L.M. Dearden, J.T. Hedley, M. Mis-
be considered dominant. The leading-order ap- saghi, M. Pourkashanian, A. Williams, L.T. Yap, Proc.
proach showed that, although the jet exit Froude Combust. Inst. 25 (1994) 1027–1034.
[14] T.-C.A. Hsieh, W.J.A. Dahm, J.F. Driscoll, Combust.
number has been used in the past for enclosed
Flame 114 (1998) 54–80.
conventional flames, it is not the appropriate pa-
[15] A. Sayre, N. Lallemant, J. Dugue, R. Weber, Proc.
rameter for NOx scaling in the MILD regime. Combust. Inst. 25 (1994) 235–242.
This is consistent with the weak role of buoyancy [16] R. Weber, Proc. Combust. Inst. 26 (1996) 3343–3354.
in a furnace with such low temperature gradients. [17] A. Sobiesiak, S. Rahbar, H.A. Becker, Combust.
5. It is clear that no single NOx production mech- Flame 115 (1998) 93–125.
anism is dominant in the present MILD com- [18] J. Newby, B. Cain, T. Robertson, in: Proceedings of
bustion conditions. The nondominant role of the the 2nd International Seminar on High Temperature
Zel’dovich thermal-NO mechanism is consistent Combustion in Industrial Furnaces, Stockholm, Swe-
den, 2000.
with the moderate furnace temperatures (800–
[19] M. Nishimura, T. Suzuki, R. Nakanishi, R. Kitamura,
1400 ◦ C). Hence, it is deduced that the prompt- Energy Convers. Manage. 38 (1997) 1353–1363.
NO and/or N2 O-intermediate pathways are of [20] F.K. Besik, S. Rahbar, H.A. Becker, A. Sobiesiak, U.S.
significance comparable to that of the thermal- Patent 5,772,421 (June 30, 1998).
NO pathway. [21] B.E. Cain, T.F. Robertson, J.N. Newby, U.S. Patent
6,638,061 (October 28, 2003).
[22] T. Suzuki, K. Morimoto, U.S. Patent 4,439,137 (March
Acknowledgments 27, 1984).
[23] G.G. Szegö, B.B. Dally, G.J. Nathan, F.C. Christo,
in: Australian Symposium on Combustion and the 8th
The authors thank Graham Kelly and Bob Dyer for Australian Flame Days, Melbourne, Australia, 8–9 De-
their technical support in all stages of this project. The cember 2003.
financial support of the Australian Research Council [24] I. Nakamachi, K. Yasuzawa, T. Miyahara, T. Nagata,
(ARC) and the School of Mechanical Engineering of U.S. Patent 4,945,841 (August 7, 1990).
the University of Adelaide is also gratefully acknowl- [25] R. Weber, A.L. Verlaan, S. Orsino, N. Lallemant, J.
edged. Inst. Energy 72 (1999) 77–83.
[26] R. Siegel, J.R. Howell, Thermal Radiation Heat Trans-
fer, fourth ed., Taylor and Francis, New York, 2002.
[27] F.P. Incropera, D.P. DeWitt, Fundamentals of Heat and
References Mass Transfer, fourth ed., Wiley, New York, 1996.
[28] S.R. Turns, An Introduction to Combustion: Con-
[1] IEA/OECD, CO2 Emissions from Fuel Combustion: cepts and Applications, second ed., WCB/McGraw–
1971–2000, Organisation for Economic Cooperation Hill, Boston, 2000.
and Development and International Energy Agency, [29] G.G. Szegö, B.B. Dally, G.J. Nathan, F.C. Christo,
Paris, 2002. in: The Sixth Asia–Pacific Conference on Combustion,
[2] J.M. Beer, Prog. Energy Combust. Sci. 33 (2007) 107– Nagoya, Japan, 20–23 May 2007, pp. 231–234.
134. [30] J.C. Davis, Statistics and Data Analysis in Geology,
[3] S. Pacala, R. Socolow, Science 305 (2004) 968–972. Wiley, New York, 1986.
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001
ARTICLE IN PRESS CNF:6965
JID:CNF AID:6965 /FLA [mNY4a; v 1.89; Prn:5/03/2008; 14:39] P.15 (1-15)
G.G. Szegö et al. / Combustion and Flame ••• (••••) •••–••• 15
[31] Gas Research Institute, GRI-mech 3.0, available at [43] A.M. Dean, J.W. Bozzelli, in: W.C. Gardiner Jr.
http://www.me.berkeley.edu/gri-mech/. (Ed.), Gas-Phase Combustion Chemistry, second ed.,
[32] N.A. Røkke, J.E. Hustad, O.K. Sonju, F.A. Williams, Springer-Verlag, 2000, p. 143.
Proc. Combust. Inst. 24 (1992) 385–393. [44] S. Gordon, B.J. McBride, Report No. NASA RP-1311,
[33] N. Peters, Prog. Energy Combust. Sci. 10 (1984) 319– NASA Lewis Research Center, Cleveland, OH, 1994.
339. [45] M. Oberlack, R. Arlitt, N. Peters, Combust. Theory
Model. 4 (2000) 495–509.
[34] M. Costa, C. Parente, A. Santos, Exp. Therm. Fluid
[46] C.P. Fenimore, Proc. Combust. Inst. 13 (1970) 373–
Sci. 28 (2004) 729–734.
380.
[35] S.R. Turns, F.H. Myhr, Combust. Flame 87 (1991) 319– [47] C.T. Bowman, Proc. Combust. Inst. 24 (1992) 859–
335. 878.
[36] S.R. Turns, F.H. Myhr, R.V. Bandaru, E.R. Maund, [48] M. Mancini, R. Weber, U. Bollettini, Proc. Combust.
Combust. Flame 93 (1993) 255–269. Inst. 29 (2002) 1155–1162.
[37] G.J.R. Newbold, G.J. Nathan, D.S. Nobes, S.R. Turns, [49] R.C. Steele, P.C. Malte, D.G. Nicol, J.C. Kramlich,
Proc. Combust. Inst. 28 (2000) 481–487. Combust. Flame 100 (1995) 440–449.
[38] P.J. Coelho, N. Peters, Combust. Flame 124 (2001) [50] E.W. Grandmaison, I. Yimer, H.A. Becker, A. So-
503–518. biesiak, Combust. Flame 114 (1998) 381–396.
[51] J.F. Driscoll, R.-H. Chen, Y. Yoon, Combust. Flame 88
[39] S. Kumar, P.J. Paul, H.S. Mukunda, Proc. Combust.
(1992) 37–49.
Inst. 29 (2002) 1131–1137.
[52] M. De Joannon, A. Cavaliere, T. Faravelli, E. Ranzi,
[40] I.B. Özdemir, N. Peters, Exp. Fluids 30 (2001) 683– P. Sabia, A. Tregrossi, Proc. Combust. Inst. 30 (2005)
695. 2605–2612.
[41] T. Plessing, N. Peters, J.G. Wuenning, Proc. Combust. [53] M. De Joannon, A. Saponaro, A. Cavaliere, Proc. Com-
Inst. 27 (1998) 3197–3204. bust. Inst. 28 (2000) 1639–1645.
[42] J.A. Wünning, J.P. Wünning, Prog. Energy Combust. [54] M. Mancini, P. Schwoppe, R. Weber, S. Orsino, Com-
Sci. 23 (1997) 81–94. bust. Flame 150 (2007) 54–59.
Please cite this article in press as: G.G. Szegö et al., Scaling of NOx emissions from a laboratory-scale mild combustion
furnace, Combust. Flame (2008), doi:10.1016/j.combustflame.2008.02.001