Fuel 217 (2018) 478–489

Contents lists available at ScienceDirect

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

Full Length Article

Effect of producer gas addition and air excess ratio on natural gas flame T
luminescence
⁎
Robertas Navakas , Andrius Saliamonas, Nerijus Striūgas, Algis Džiugys, Rolandas Paulauskas,
Kęstutis Zakarauskas
Lithuanian Energy Institute, Laboratory of Combustion Processes, Breslaujos g. 3, 44403 Kaunas, Lithuania

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents the emission spectroscopy method for registering chemiluminescent radical species of
Flame diagnostics OH∗,CH∗,C∗2 by means of optical flame imaging using optical bandpass filters, corresponding to emission wa-
Chemiluminescence velengths of the radicals of interest, and by scanning the flame using optical fibers coupled to a spectrometer for
Spectroscopy recording the emission spectra. The aim of this research is to analyse the distribution of chemiluminescence
Producer gas
intensity along the flame axis from various angles simultaneously in order to determine the effect of the air
Renewable energy
equivalence ratio (ER) and mixing of producer gas (PG) from biomass gasification into the natural gas flow to
spectral characteristics of flame at specific wavelengths representing formation reactions of OH∗ radical species.
The experiments were carried out with pure natural gas (NG) and air mixture and a premixed air/gas mixtures
(80%NG and 20% PG, 65%NG and 35%PG). Flows of air, NG and PG were premixed before entering the
combustion chamber. For flame emission spectroscopy, two different methods were used: 1) imaging by an ICCD
camera through optical bandpass filters suitable for OH∗,CH∗,C∗2 chemiluminescence intensity registration; 2) by
a combination of a spectrometer and five optical fibers to collect the flame spectra from 5 different angles and in
various heights from the burner outlet. The distribution of chemiluminescent species intensity along the burner
vertical axis was analyzed and zones with most intense OH∗,CH∗,C∗2 generation in flame were identified, and the
effect of addition of PG to NG and ER to the OH∗ chemiluminescence was analysed.

1. Introduction
As the estimated worldwide use of natural gas (NG) is projected to
increase by nearly 12% in 2020, one of the strategic objective is the
reduction of emissions from stationary combustion plants using gases
[1]. Due to these reasons alternative fuels are being sought to partially
or fully replace natural gas. One of such fuel type is producer gas (PG)
created from biomass via the gasification process. During the gasifica-
tion process, low quality solid fuel is converted into more valuable
combustible gases. These gases are mixtures of methane, hydrogen
(H2), carbon monoxide (CO) and inert gas such as carbon dioxide (CO2),
nitrogen (N2) [2]. Due to its composition, producer gas could be used to
substitute natural gas in various combustion devices such as water
boilers, drying units, chemical heating, incineration, ceramic kilns etc.
through combustion of the gas in a burner. Despite the great potential,
the combustion characteristics of PG are different from those of NG.
According to several studies [3,4], PG is more prone to flashback, has
different behavior of flame stability and different dynamic responses.
Also in mixtures of PG and NG the heat release rate profiles are wider
for PG, compared to NG, indicating slower flame propagation in the
combustion chamber. This affects the combustion stability and results
in power loss, cyclic variation, maximum pressure and temperature
decrease in thermal applications when trying to mix or completely
substitute NG with PG [5]. On the other hand, some works have de-
monstrated [6,7], that the hydrogen rich PG advances the gas mixture
ignition and extends the flammability limits, thus ensuring the flame
stability. In this way, burning the fuel with higher air excess ratio the
flame temperature decreases, leading to the decrease in NOx genera-
tion. Due to these reasons, it is necessary to understand the specifics of
the combustion process in order to design novel burners for producer
gas or adapt the existing ones. One of the most promising non-intrusive
combustion process optimization technique is flame chemilumines-
cence monitoring [8–10]. In a chemiluminescent reaction, a part of the
released energy is used to excite an electronic state which is short-lived
and relaxes through a number of mechanisms, including collisional
non-radiative deactivation and spontaneous photon emission. In the
case of photon emission, species emit at a characteristic wavelength,
signaling the presence of the species [11]. Some studies of emission

⁎
Corresponding author.
E-mail address: robertas.navakas@lei.lt (R. Navakas).

https://doi.org/10.1016/j.fuel.2017.12.094
Received 12 February 2017; Received in revised form 15 November 2017; Accepted 21 December 2017
0016-2361/ © 2017 Elsevier Ltd. All rights reserved.
R. Navakas et al.
spectroscopy with an iCCD camera [12,13] show that strongest che-
miluminescent intensities can be registered at the beginning of the
flame, near the burner outlet. According to [14–17], spectral emission
of hydroxyl radical (OH∗) correlates with the equivalence ratio (ER,
denoted as α ), therefore, it can be used for combustion process mea-
surements and control as it is one of the most important parameters for
effective and clean combustion. A few authors point out that ER in some
cases can be monitored from the OH∗/CH∗ or OH intensity relations
[12,18]. Other authors [19] working on chemiluminescence emission of
premixed syngas flames determined that OH∗/CO∗2 ratio is suitable to
monitor stoichiometry of combustion but only in a certain range of ER.
Hernandez et al. [6] performed research on OH∗ and CH∗ radical
emissions from the combustion of NG and PG and have found that the
presence of OH∗ radical could indicate formation of NOx . The spatial
distribution of radical chemiluminescence intensity values measured in
flame by emission spectroscopy method was also modeled with
CHEMKIN resulting in acceptable similarities with experimental data
[18]. Modeling with GRIMech also confirmed the relation between ER
and OH∗, CH with multiple methane oxidation mechanisms and an
additional sub-mechanism that accounts for the formation and de-
struction of the chemiluminescent species OH∗ and CH∗ [20]. For more
in-depth analysis, the combination of a spectrograph, an ICCD camera
and optical fibers are often used [21]. By measuring a narrow point in
the flame, light from an optical fiber is dispersed into spectra and then
can be analyzed in depth [22]. Parameswaran et al. [23] used a fiber
probe connected to a spectrometer to acquire the radiation from flames
generated with various gas mixtures. The results showed that the OH∗
peak intensity has a strong correlation with ER and with heating value
of the gas mixture as well. However, to obtain more precise results, the
use of optical fibers from multiple angles enables to measure such
properties as the flame radial front (defined as regions with strongest
chemiluminescence emissions) and the precise flame curvature [24,25].
Besides, in most cases flame is turbulent and non symmetrical in in-
dustrial boilers, therefore, using a single sensor to scan a limited flame
region cannot provide adequate diagnostics for, e.g., automated process
control. Therefore, it is necessary to scan the whole combustion area
and to identify the most informative flame regions for positioning the
sensor in order to maintain the optimal combustion regime, including
ER. For these reasons, this work aims to apply a spatial flame chemi-
luminescence scanning technique to compare OH∗,CH∗ and C∗2 intensity
values in the line-of-sight chemiluminescence intensity contour along
the flame axis when using both pure natural gas and its mixture with
producer gas in various proportions. Also, this article presents com-
parison of chemiluminescence along the flame axis from five different
angles registered at same time with optical fibers and flame imaging
through optical bandpass filters using an ICCD camera.
2. Materials and methods
2.1. Experimental setup and procedure
The experiments were performed using the experimental setup
consisting of an air/gas supply system, a gasification reactor system for
producer gas supply, a combustion chamber, and a flame optical ana-
lysis system. Schematic diagram of the experimental setup is presented
in Fig. 1. The images of flame in the burner are shown in Fig. 2. In order
to estimate the flame characteristics of PG addition to NG, the air
equivalence ratio α was set in range from 1.0 to 1.3 to recreate optimal
combustion conditions used in industrial boiler burners. The constant
0.6 kW thermal output of the burner for the mixture was maintained in
all the experiments for result comparability (Table 1). The combustion
process was organized in a combustion chamber made of a 56 cm high
and 6 cm diameter quartz glass pipe. A Bunsen type burner was mod-
ified to work in a closed chamber with premixed gas flow by adding a
premixing chamber and a mesh type bluff body for flame stabilisation.
The burner was fired by mixtures of natural gas (NG) and producer gas
479
Fuel 217 (2018) 478–489
(PG) at different proportions (Table 1). NG, PG and air flows were
connected to the premixing chamber via steel pipes. For each mea-
surement of flame chemiluminescence, an initial volumetric gas flow
was set with flow meters (operating range 0.5 ÷ 20 l/ min) (Table 1). At
the exit of the combustion chamber, the flue gas analyzer probe was
inserted at a 50 cm distance from the burner nozzle. The samplings with
a gas analyzer were made relatively far from the combustion zone to
recreate conditions of an industrial flue gas measurement system. These
result were later compared to the results obtained by using the ICCD
camera in with varying ER. The sampling and composition measure-
ment of flue gas was used to analyse whether the spectral data of flame
emission can indicate the combustion parameters similarly to gas
sampling with a faster response time.
2.2. Preparation of air/gas supply
Producer gas was generated by using a lab scale gasification reactor
[26]. The reactor was filled with pine and spruce wood pellets and
gasified at the temperature of 800 °C. The producer gas generated in the
process was transported from the reactor to the burner by N2 used as
the carrier gas. In all the experiments, the same batch of commercially
available wood pellets was used and producer gas was extracted at
stable operation regime of the reactor. By measuring the gas composi-
tion multiple times, no significant variation in gas composition was
detected, therefore, it was assumed to be constant. Natural gas was
supplied from the district gas distribution grid. The compositions of
producer gas (measured) and natural gas (from supplier data sheet) are
presented in Table 2. Air was supplied from the compressed air system
located in the research facility. Both flows were controlled with flow
meters (operating range 0.5 ÷ 20 l/ min). The inlet parameters for fuel
and air flows are given in Table 1.
2.3. Optical system
The whole combustion process was observed using an optical
system for analysis of spatial distribution of the excited species
OH∗,CH∗ and C∗2 in the flame at atmospheric pressure. The flame images
were captured using an ICCD (Intensified Charge Coupled Device)
camera Andor iStar DH734. The diameter of the photocathode (in-
tensifier) was 18 mm; a pixel size was 13 μ m. The matrix contains
1024 × 1024 active pixels sensitive to 200–800 nm wavelength emis-
sions. The Equivalent Background Illuminance (EBI) is <0.2 e− / pix/ sec.
The readout noise is as low as 2.9 e−. The peak quantum efficiency at
room temperature is 25%. For spectral analysis, the spectrometer Andor
Shamrock SR-303i coupled to the camera was used. The spectrometer
focal length is 303 mm, the aperture f/ 4. The grating resolution was 300
lines/ mm.
In order to compare the spatial distribution of selected radicals, two
types of experiments were carried out:
• Flame imaging: the camera is focused to capture the entire image of
the flame; an optical filter is placed in front of the camera objective,
thereby, the flame images were recorded at different wavelengths
corresponding to emissions of the different radicals (Fig. 1 and
Table 4);
• Flame scanning: 5 optical fibers were placed around the flame in a
ring-shaped holder, with its plane oriented horizontally, to provide
the spatially–integrated spectra obtained from different angles. The
plane was scanned along the vertical axis of the flame and the
spatial distribution of the radicals in the flame was thereby obtained
(Fig. 3).
Both experiments are described in more detail below.
2.3.1. Flame imaging
To obtain a spatially resolved distribution of radical species in the
R. Navakas et al.
flame, the camera was positioned and focused with respect to the flame
in such a way that the flame fits the image frame entirely. To separate
the contribution of the selected radical, the bandpass filter for the re-
spective emission wavelength was placed on the camera objective. In
total, five filters were used in the experiments; the transmitted wave-
lengths and the respective radicals imaged through each filter are listed
in Table 4. The obtained image represents a distribution of the re-
spective radical specie in the flame contour integrated along the line of
sight of the camera.
480
Fuel 217 (2018) 478–489
Fig. 1. Scheme of the experimental setup: a) ar-
rangement of burner setup and camera for flame
imaging through optical filters (side view); b)
arrangement of burner setup and optical fibers for
flame scanning (side view); c) arrangement of
optical fibers in annular holder for flame scanning
(top view).
The achieved resolution of the flame image was 6.6 pixel/ mm.
Single accumulation cycle time was set to 1.266 s. Optimal exposure
was set to 0.1 s and 300 image acquisitions were accumulated into a
single frame for each combustion and imaging regime (fuel mixture
composition, fuel and air flow rates, selected optical filter) in order to
minimize the effect of flame instability.
2.3.2. Flame scanning through optical fibers
In order to record the entire emission spectrum of the region of
Fig. 2. Images of flame in the experimental setup:
fueled by a mixture of natural gas and producer gas
(left) and natural gas (right).
R. Navakas et al. Fuel 217 (2018) 478–489

Table 1 Optical data was acquired analogously to the case of flame imaging
Inlet parameters for fuel and air flows at different content of producer gas (PG) in the through the optical filters. The camera exposure duration was set to 5 s.
mixture (0, 20%, 35%) and air equivalence ratios α .
Each single image consisted of 10 acquisitions. The measurements for
PG in mixture, Natural gas PG flow, Air flow, l/ min each position of the optical fibers were repeated 6 times. This procedure
vol.% flow, l/min l/ min was repeated for each flame condition at 20 measurement planes along
α = 1.0 α = 1.1 α = 1.2 α = 1.3 flame height.
Multiple frames were recorded for each combustion and imaging
0 1.08 0 10.30 11.33 12.37 13.40
20 0.86 0.80 9.94 10.93 11.93 12.92
regime (fuel mixture composition, fuel and air flow rates, selected
35 0.70 1.40 9.67 10.63 11.60 12.57 measurement plane) in order to minimize the effect of possible flame
instability. All 5 channels of optic fibers were balanced to give same
signal intensity by using a uniform light source placed in the center of a
Table 2 measuring plane.
Compositions and lower heating values (LHV) of natural gas (NG) from gas distribution
grid and producer gas (PG) supplied from gasification reactor, vol.%. (NG composition is
taken from the gas supplier’s data sheet). 2.4. Data processing

Component Content in NG, vol.% Content in PG, vol.%

To reduce flame instability for each flame condition and each filter,
H2 — 14.00 ± 0.95 multiple frames were put together resulting in a single image re-
CO — 16.40 ± 1.16 presenting the mean values of spatial distribution of emission in-
CO2 0.08 9.20 ± 0.86 tensities of respective species. To filter out the camera internal noise, a
CH 4 97.5 1.05 ± 0.11 dark frame was taken with a capped camera and the values of this frame
C2 H2 — 0.06 ± 0.019
C2 H 6 1.2 0.05 ± 0.016
where subtracted from the main image. The dark frame was taken with
C3 H8 0.27 0.02 ± 6.4 × 10−3 photocathode set to “off” and the camera was covered with a cloth to
C4 H10 0.08 — prevent accidental light leaks during exposure. Collected raw data was
N2 — 59.22 ± 2.72 saved as Flexible Image Transport System (FITS) files and processed
LHV, MJ/ m3 33.6 5.6 using MATLAB application. To determine how the light intensity
changes along the flame in a vertical direction, each horizontal line of
the pixel intensity values in the frame being analyzed was averaged,
thereby producing the averaged distribution of chemiluminescence in-
tensity along the flame axis, representing relative concentrations of
chemiluminescent species along the burner axis. In parallel, the total
mean intensity values of each image were calculated in order to com-
pare how total intensity of chemiluminescence changes with changing
ER.
The signal to noise ratio was estimated using the following ex-
pression [28]:
Dqe P
SNR = ,
2 2 2
δdark + δsignal + δreadout (1)

where Dqe = 0.25 is the quantum efficiency, P is the number of photons

Fig. 3. Sketch of arrangement of the optical fibers in the annular holder: the range of the
vertical motion along the flame axis (front view; left) and their orientation in the plane of
the holder (top view; right).
interest, an optical fiber coupled to the spectrometer is pointed to the
appropriate direction. The fiber collects the incoming light within the
viewing angle of 15 degrees; the emission is therefore integrated along
the viewing direction of the fiber. In total five fibers were used for
experiments. They were attached to a ring-shaped holder and arranged
at equal angular distances from each other, pointing towards the center
of the ring (Fig. 3). The ring holding the fibers was located to enclose
the flame, with its center located at the vertical axis of the flame and
the ring plane perpendicular to the flame axis. This arrangement pro-
vides the view of the flame from five different directions simultaneously
at the same horizontal plane. The height of the fiber holder was ad-
justable; a number of measurements were made by translating the ring
in the vertical direction along the flame axis, thereby scanning the
entire flame, in the range from 0 to 70 mm height above the burner
nozzle. For recording the entire emission spectrum, the camera sensor is
digitally divided into 5 equal areas to collect light from each of 5 optical
fibers.
incident upon a camera pixel, δdark = 0.2 is the dark noise, δreadout = 2.9
is the readout noise, δsignal = Dqe P is the signal noise [29]. For a count
of 1000 photons per pixel, this gives SNR = 15.55, which is considered a
sufficient signal. For most measurements, the photon count per pixel
was on the order 2 × 105 , resulting in SNR > 200 .
The uncertainty estimation was based on the best available esti-
mation of the overall equipment precision. The mathematical opera-
tions such as mean values calculation were also considered as the other
uncertainty source. The experimental equipment was manufacturer-
calibrated and the traceability of the results was established during that
time. The operation parameters and allowed deviations are well de-
scribed in the standard method and the manufacturer’s manual. The
uncertainty of the instrument operation was estimated from a series of
repeated observations by calculating the standard deviation as the
overall precision of the measurement.
3. Results and discussion
3.1. Flame imaging through optical filters
Fig. 4 presents the spatial distribution of chemiluminescent species
emission intensity in flame acquired with optical filter setup at air
equivalence ratio α = 1.0 with variable mixture composition: 100%
natural gas; NG substituted with 20% producer gas; and NG with sub-
stituted 35% PG. All the emission intensities in figures are presented as
photon counts per exposure.

481
R. Navakas et al. Fuel 217 (2018) 478–489

PG 0% PG 20% PG 35% Fig. 4. Flame images through bandpass filters at the

emission wavelengths of different radicals (indicated at
the left hand side of the respective image rows) and
different gas mixture compositions (indicated at the top
by volume percentage of producer gas (PG) in the mix-
ture: 0%, 20%, 35%) and air equivalence ratio α = 1.0 .
OH∗
308.9 nm

CH∗
387.1 nm

CH∗
431.4 nm

CH∗2
514 nm

OH∗
282.9 nm

Fig. 5 exhibits the spatial distribution of chemiluminescent species
emission intensity in flame at variable air equivalence ratios
α = 1.0;1.1;1.2;1.3.
The results show that the spectral intensity along the flame in Fig. 4
and Fig. 5 has profile shapes similar to the flame profile as seen in the
visible light. These results coincide with the results obtained by other
authors, who used similar experimental methodology [30,31]. Max-
imum intensity values were located at approximately 2 cm distance
from the burner nozzle. In C∗2 filter case, maximum spectral intensity
was observed at 1 cm and 3–5 cm distance from the burner nozzle
(Fig. 6).
The flue gas analyzer confirmed high CO concentration (Table 3)
that is related to presence of C∗2 in the combustion zone [32]. It is
widely known that high CO concentration influences (reduces) the
flame temperature thus reducing thermal NOx formation. High CO
concentrations may arise due to incomplete mixing and inhomogeneous
concentrations, giving rise to areas of rich fuel mixture. NOx reduction
was also observed when using gas mixture instead of pure NG at mix-
ture proportions of NG 80% and PG 20%. The reason of this effect re-
lates to presence of hydrogen in the mixture. The addition of hydrogen
to lean natural gas mixture is known to lower the NOx emissions
[33,34]. Due to presence of hydrogen in the mixture, a chemilumi-
nescence intensity peak in acquired images reaches the maximum value
closer to the burner. It means that the reactions related to chemilumi-
nescent intensity peak at specific wavelength (Table 4) start faster when
using mixtures. According to Slim et al. [35], hydrogen reacts faster
with oxidant than methane and increases the temperature in the reac-
tion zone, so the combustion reactions end faster and closer to the
burner’s nozzle. The chemiluminescence intensity shift toward the
burner’s nozzle can be seen in the results. The hydrogen presence in

482
R. Navakas et al.
α = 1.0 α = 1.1 α = 1.2
OH∗
308.9 nm
CH∗
387.1 nm
CH∗
431.4 nm
C∗2
514 nm
OH∗
282.9 nm
mixture for combustion was confirmed with gas analyzer data of pro-
ducer gas composition (Table 3). As mentioned in the introduction, the
relation between ER and the chemiluminescence intensity is reported
by other authors [36]. This relation was also seen in the current case.
This linear dependency remains not only for pure natural gas, but for
the mixtures with producer gas as well (Fig. 7).
3.2. Flame scanning by optical fibers
Flame spectra were collected from five different angles to re-
construct the spatial distribution of intensity (Fig. 8). Each optical fiber
signal presented the intensity values in the 200–600 nm wavelength
range. Six bands for OH∗, CH∗ and C∗2 chemiluminescence intensity
determination were selected from all the data and compared at different
flame conditions. Similarly as in the case with optical filters, the che-
miluminescence intensity along the flame height was distributed un-
evenly.
The main chemiluminescent species of interest were the same as
described in Section 3.1 as they are known to emit most intensely
[36–38,12,18] and tend to be related to the air excess ratio. The main
reactions of these radical formation are presented in Table 4. The ex-
periments show that the C∗2 radical emits weaker intensity and thus is
less attractive for practical applications. All the registered flame profile
intensity maximum values were located at the distances between 0.5
Fuel 217 (2018) 478–489
α = 1.3 Fig. 5. Flame images through bandpass filters at the emission
wavelengths of different radicals (indicated at the left hand side
of the respective image rows) and different air excess coefficients
α, as indicated at the top of the image. The content of producer
gas in the mixture for all the flame images is 20%.
and 1.5 cm from the burner outlet meaning that most of the chemical
reactions related to OH∗, CH∗ and C∗2 radical transitions from the excited
state to the ground state [39] occurred near the burner outlet. Also it
can be seen that addition of producer gas makes the reaction zone even
shorter due to presence of hydrogen in the mixture, leaving no re-
cordable signal of chemiluminescence at nearly 4 cm distance from the
outlet (Fig. 8). As stated before, the hydrogen reacts faster with oxidiser
than methane and increases the temperature in the reaction zone so
that the combustion reactions complete faster and closer to the burner
nozzle. The chemiluminescence intensity shift toward the burner nozzle
can be seen in the results (Fig. 8).
The results collected from different angles for various flame condi-
tions show that the chemiluminescence was strongest at the 288 degree
angle from the reference direction. In our case this was the combined
result of a non–uniform distribution of chemical reactions in the flame,
flame instability and a possible burner alignment asymmetry. This re-
sult also confirms meaningfulness of using a multi-spot flame mon-
itoring, because the intensity distribution variations can be significant
for accurate determination of the local excess ratio, temperature and
other parameters. In the OH∗ chemiluminescence emission range at
289.9 nm and 309.8 nm, it was determined that the 309.8 nm wave-
length signal is stronger yet showing similar behavior like that at
289.9 nm. Also, addition of producer gas increased the overall intensity
values, but the relations to the excess air ratio remained unchanged. For
483
R. Navakas et al. Fuel 217 (2018) 478–489

16 Table 3
Content of CO and NOx in flue gas and hydrogen content in gas mixture for different gas
NG
14 NG80%, PG20% mixture compositions (natural gas NG and producer gas PG) and air equivalence ratios α
used in experiments.
NG65%, PG35%
12
Vertical distance, mm

Mixture composition α CO, mg/m3 NOx , mg/ m3 Hydrogen, vol.%

10 NG 1.0 16900 125 0

1.1 150 150
NG
8 1.2 6 173
1.3 4 198
NG 80%,
6 PG 20% NG 80% + PG 20% 1.0 920 155 2.8
1.1 120 160
4 1.2 54 173
NG 65%, 1.3 40 145

2 NG 65% + PG 35% 1.0 2600 118 4.9

1.1 40 141
1.2 5 149
0 1.3 9 146
0 2000 4000 6000
Intensity
a) OH∗ (308.9 nm) Table 4
Wavelengths related to main chemiluminescent radical formation reactions [27].
16
NG Radical Reaction Wavelength, nm Filter transparency
14 NG80%, PG20% ∗
OH∗ CH+O− 282.9 > 65%
NG65%, PG35% 2 →CO + OH
12 308.9 > 15%
Vertical distance, mm

CH∗ C2H+O−
2 →CO2 + CH
∗ 387.1 > 90%
431.4 > 95%
10
C∗2 CH2 + C− → C∗2 + H2 514, 516.5 > 65%
NG
8

6 NG 80%,
PG 20%
4 NG 65%,

0
0 2000 4000 6000
Intensity
b) CH∗ (387.1 nm)
16
NG
14 NG80%, PG20% Fig. 7. Relation between the chemiluminescence intensity ratio of OH∗ at 308.9 nm and
air equivalence ratio; Content of producer gas in the gas mixture was 0%, 20%, 35%,
NG65%, PG35%
respectively. (NG – natural gas, PG – producer gas.)
12
Vertical distance, mm

10 emissions at the CH∗ wavelengths of 387.1 nm and 431.4 nm, it was

NG
determined that the 431.4 nm wavelength signal is stronger yet
8 showing similar behavior like that of 387.1 nm. The addition of pro-
ducer gas increased the overall intensity values slightly less than in the
6 OH∗ case, but the relations to the excess air ratio remained unchanged.
NG 65%, For C2 chemiluminescence case at 514 nm and at 516.5 nm, it was de-
4 termined that both the wavelength signals showed similar behavior,
NG 80%, with intensity at 516.5 nm being slightly higher. The addition of pro-
2
PG 20% ducer gas increased the overall intensity values slightly less than in the
OH∗ and CH∗ case. The relations to the excess air ratio remained un-
0
changed (Fig. 9).
0 2000 4000 6000
To show the effect of the air excess ratio to emission of chemilu-
Intensity minescent species, the experimental points at heights from 0 to 1.5 cm
c) C∗2 (514.0 nm) from the burner nozzle were taken into account. This was the zone,
where the intensity was registered at maximum values. Results were
Fig. 6. The spectral profiles of the flames, acquired at the wavelengths of emissions of
obtained by interpolating the experimental data points. Results show
OH∗ (308,9 nm), CH∗ (387.1 nm) and C∗2 (514.0 nm) radicals as a function of vertical
that emission intensity of OH∗ tends to show the strongest dependence
distance above the burner nozzle and at air equivalence ratio α = 1 with different com-
positions of mixtures of natural gas (NG) and producer gas (PG). on the excess air ratio (Fig. 10).

484
R. Navakas et al. Fuel 217 (2018) 478–489

Fig. 8. Flame vertical profiles obtained by scanning through optical fibers at emission wavelengths of different radicals at different air equivalence ratios α and different gas mixture
composition: natural gas (a, c, e) and a mixture of 80% natural gas and 20% producer gas (b, d, f).

3.3. Comparison of results obtained by flame imaging and flame scanning
Experimental setup for flame imaging through optical filters
(method 1) was compared to the setup for flame scanning with optical
fibers (method 2). In case of method 1, a full instantaneous single point
of view flame image was acquired (Figs. 4, 5). This type of data is useful
for quick determination of chemiluminescent intensity distribution in
the combustion zone of small flames. However, in case of an industrial-
scale flame, this method is difficult to implement, i.e., there is little
practical possibility to acquire a full image of the flame inside the
boiler’s furnace. In this case, it is reasonable to install a single or
multiple optical sensors (method 2) in order to observe emission in-
tensity at a certain specified wavelength in the selected locations cor-
responding to either OH∗ emission or intensity ratios at emission wa-
velengths of OH∗/CH∗ [12,18] that correlates with any of the most
important process parameters, e.g., air equivalence ratio α . To estimate
the ratio of emission intensities by different radicals, we introduce an
intensity Ji (z ,α,c PG) registered at the height z along the flame axis

485
R. Navakas et al. Fuel 217 (2018) 478–489

Fig. 9. Flame radial profiles obtained by scanning through optical fibers at emission wavelengths of different radicals at different air equivalence ratios α and different gas mixture
composition: pure natural gas (a, c, e) and a mixture of 80% natural gas and 20% producer gas (b, d, f).

emitted at the wavelength of the i-th radical, when the process takes J1 (z ,α,c PG )
R (z ,α,c PG) =
place with the given air excess ratio α and the volume fraction of J2 (z ,α,c PG ) (3)
producer gas in the fuel mixture c PG , and the intensity Ii (α,c PG) aver-
aged over the flame height: In our case, the radical with i = 1 is OH∗, and i = 2 corresponds to
CH∗. The ratio of intensities depends on the fuel mixture composition,
Ii (α,c PG) = 〈Ji 〉z = ∫z Ji (z,α,cPG) dz (2)
characterised by the content of producer gas in the mixture. The dif-
ference of intensity ratios, depending on the mixture composition, is
defined as
The intensity ratio is thus defined as

486
R. Navakas et al. Fuel 217 (2018) 478–489

18000 1
NG NG 80%,
16000 PG 20%
NG80%,PG20%
Relative intensity value

14000 0.9
12000
10000 0.8
8000
0.7
6000

Δ(z)
4000
NG 0.6
2000
0
0.95 1.00 1.05 1.10 1.15 1.20 1.25
0.5
Air equivalence ratio (ER)
0.4
Fig. 10. Relation between chemiluminescence intensity ratio of OH∗ at 308.9 nm and air
equivalence ratio; content of producer gas in the gas mixture is 0% and 20%, respectively. 0.3
NG – natural gas, PG – producer gas. 0 5 10 15 20 25 30
z, mm
Fig. 13. Dependence of Δ(z ) on the height z along the flame axis.

1 N
Δ(z ) =
Nα
∑ jαα=1 ΔR (z,α jα )2 ,
(5)

where Nα is the number of values of α used in the experiment, α jα being

the jα -th value.
Flame scanning eliminates the filter transparency problem because
all the fibers have approximately the same transparence. This method
enables measurements at the same moment from multiple spots if ne-
cessary. Although both methods reveal dependence of OH∗ chemilu-
minescence intensity to the air equivalence ratio, but the curves with
different gas mixture compositions are different (Figs. 7 and 10).
Having calculated the ratio of emission intensities at OH∗ (308.9 nm)
and CH∗ (431.4 nm) wavelengths for both the methods used, it can be
seen that this ratio weekly depends on producer gas concentration and
Fig. 11. Dependence of the ratio of emission intensities at the emission wavelengths of strongly depends on α for method 1 (Fig. 11). For the case of method 2,
OH∗ (308.9 nm) (I1) and CH∗ (387.1 nm) (I2 ) on the air equivalence ratio α for flame of this dependence is less linear due to incomplete averaging over the
natural gas (NG) and flame of fuel mixture consisting of 80 vol.% natural gas and 20 vol.
flame: only 5 fibers were used, sampling emissions from 5 spots (Sec-
% producer gas (PG).
tion 2.3.2), i.e., some part of the light intensity was skipped. Knowing
that the intensity ratio of emissions of OH∗/CH∗ correlate sufficiently
well with the air equivalence ratio α , these dependences were plotted
30
for every flame section along the z axis (Fig. 12). It can be seen from
here that certain zones arise in the flame (at z = 5 and z = 10 mm in our
25 case) where α does not depend or weekly depends on the initial com-
position of fuel mixture but depends on the air excess ratio. These
20 possible locations where the difference R is small, i.e., does not depend
on the concentration of producer gas c PG , appear at these locations
where the most intense emissions of OH∗ and CH∗ radicals were de-
z, mm

15
tected. Consequently, it can be concluded that pointing the emission
PG 0% sensor to this spot will result in lowest variability of R with varying
10 PG 20% producer gas content c PG . This difference is better illustrated by a
parameter Δ(z ) (Fig. 13). Knowing these characteristic locations in
5 flame, the sensors could be placed in the identified spots enabling to
monitor, e.g., changes in the air excess ratio during the process and
possibly implement the real time control for optimal combustion re-
0
4
3
gime. Further analysis of the identified dependencies is the subject of
1.2 1.25
2 1.15 future work.
R 1
0 1.05 1.1 α
1

Fig. 12. Dependencies of the ratio of emission intensities R = I1/ I2 at the emission wa-
velengths of OH∗ (308.9 nm) (I1) and CH∗ (387.1 nm) (I2 ) on the air equivalence ratio α at
different heights z along the flame axis.
ΔR (z ,α ) = R (z ,α,c PG = 0)−R (z ,α,c PG = 0.2).
The differences of emission intensity ratios arising from varying fuel
compositions, are also different in different flame regions; we therefore
introduce the position-dependent difference, averaged over the values
of the air equivalence ratio α :
4. Conclusions
The multi-point imaging of chemiluminescence intensity has been
implemented using an ICCD camera, and used to obtain the spatial,
(4) instantaneous, visible chemiluminescence profiles of premixed flames
fueled by NG and mixtures of NG and PG. Two flame diagnostics
methods were compared: flame imaging through optical filters at the
emission wavelengths of excited radicals and flame scanning by optical
fibers coupled to a spectrometer. The flame scanning appeared to
provide more detailed flame chemiluminescence intensity profiles due
to additional information collected from different angles at same time.

487
R. Navakas et al.
Chemiluminescent intensity in flame distributed non uniformly in both
measurement methods due to different chemical reactions occurring at
different flame locations. The locations of known excited state radical
species were determined by locations of maximum emission intensities
at specific wavelengths. In all cases, the maximum intensity was ob-
served at vertical distances and from the burner nozzle. It was assumed
that the reactions from described chemical kinetics mechanisms are
most intense at these locations. Presence of hydrogen in the gas mixture
arising from producer gas resulted in decreased NOx emissions, higher
temperatures, and chemical reactions proceeded faster when burning
NG with addition of PG. This caused chemiluminescence intensity va-
lues higher and more concentrated at distances near the burner nozzle.
It was determined that the OH∗ intensity at 309.8 nm has strongest
intensity in all the experimental results. After determination of the
highest intensity zones in flame, a relation between the local ER and
OH∗ emission intensity was estimated by interpolating the experimental
points. Both cases of experiments showed that there is a strong relation
between the ER and the OH∗ chemiluminescent intensity. In case of
flame scanning, multiple angles provide sufficient information for fur-
ther analysis. Also the obtained results confirm that scanning the flame
from multiple angles using optical fibers coupled to a spectrometer is a
reliable method for determining the characteristics of the combustion
process such as ER and can be applied in industrial boilers equipped
with gas burners. Consequently, it can be concluded that pointing the
emission sensor to this spot will result in lowest variability of R with
varying producer gas content c PG . Flame monitoring from multiple
angles can be applied in monitoring and control systems in order to
adjust burners for firing mixtures of PG and NG maintaining clean and
efficient combustion.
Acknowledgements
This research was funded by a grant (No. P-MIP-17-26, Title:
“Investigation on syngas assisted combustion for flexible and low
emission operation in industrial boilers”) from the Research Council of
Lithuania. The authors also want to express appreciation to the COST
Action CM1404 “Chemistry of Smart Energy Carriers and Technologies
(SMARTCATS)”.
References
[1] E. U.S. Energy Information Administration, International Energy Outlook 2016, Vol.
0484, May 2016. arXiv:EIA-0484(2013), doi:DOE/EIA-0484(2014). URLhttp://
www.eia.gov/forecasts/ieo/pdf/0484(2016).pdf.
[2] Punnarapong P, Sucharitakul T, Tippayawong N. Performance evaluation of pre-
mixed burner fueled with biomass derived producer gas. Case Stud Therm Eng
2017;9:40–6. http://dx.doi.org/10.1016/j.csite.2016.12.001. URL http://
linkinghub.elsevier.com/retrieve/pii/S2214157X16301460.
[3] Richards G, McMillian M, Gemmen R, Rogers W, Cully S. Issues for low-emission,
fuel-flexible power systems. Prog Energy Combust Sci 2001;27(2):141–69. http://
dx.doi.org/10.1016/S0360-1285(00)00019-8. URL http://www.sciencedirect.com/
science/article/pii/S0360128500000198.
[4] Candel S. Combustion dynamics and control: progress and challenges. Proc Combust
Inst 2002;29(1):1–28. http://dx.doi.org/10.1016/S1540-7489(02)80007-4. URL
http://www.sciencedirect.com/science/article/pii/S1540748902800074.
[5] Tsiakmakis S, Mertzis D, Dimaratos A, Toumasatos Z, Samaras Z. Experimental
study of combustion in a spark ignition engine operating with producer gas from
various biomass feedstocks. Fuel 2014;122:126–39. http://dx.doi.org/10.1016/j.
fuel.2014.01.013.
[6] Hernández JJ, Lapuerta M, Barba J. Flame stability and OH and CH radical emis-
sions from mixtures of natural gas with biomass gasification gas. Appl Therm Eng
2013;55(1):133–9. http://dx.doi.org/10.1016/j.applthermaleng.2013.03.015. URL
http://www.sciencedirect.com/science/article/pii/S1359431113001725.
[7] Noble DR, Zhang Q, Lieuwen T. Hydrogen effects upon flashback and blowout.
International colloquium on environmentally preferred advanced power generation
September 5–8, 2006, Newport Beach, California, Proceedings of ICEPAG2006,
2006 2006:24012.
[8] Zizak G, Flame emission spectroscopy: fundamentals and applications, CNR-TeMPE,
Istituto per la Tecnologia dei Materiali e dei Processi Energetici Via Cozzi 53, 20125
Milano, Italy; 2000 1–29[Online; accessed 2016.12.26]. URL http://eprints.bice.rm.
cnr.it/id/eprint/48.
[9] Parameswaran T, Hughes R, Gogolek P, Hughes P. Gasification temperature mea-
surement with flame emission spectroscopy. Fuel 2014;134:579–87. http://dx.doi.
488
Fuel 217 (2018) 478–489
org/10.1016/j.fuel.2014.06.016.
[10] L.C. Haber. An investigation into the origin, measurement and application of che-
miluminescent light emissions from premixed flames [Ph.D. thesis]. Virginia
Polytechnic Institute and State University; 2000. URL https://theses.lib.vt.edu/
theses/available/etd-03142001-144036/.
[11] Karnani S, Dunn-Rankin D. Visualizing CH∗ chemiluminescence in sooting flames.
Combust Flame 2013;160(10):2275–8. http://dx.doi.org/10.1016/j.combustflame.
2013.05.002. URL http://www.sciencedirect.com/science/article/pii/
S0010218013001752.
[12] Józsa V, Kun-Balog A. Spectroscopic analysis of crude rapeseed oil flame. Fuel
Process Technol 2015;139:61–6. http://dx.doi.org/10.1016/j.fuproc.2015.08.011.
URL http://www.sciencedirect.com/science/article/pii/S0378382015301235.
[13] Chong CT, Hochgreb S. Flame structure, spectroscopy and emissions quantification
of rapeseed biodiesel under model gas turbine conditions. Appl Energy
2017;185:1383–92. http://dx.doi.org/10.1016/j.apenergy.2016.01.003. clean,
Efficient and Affordable Energy for a Sustainable Future. URL http://www.
sciencedirect.com/science/article/pii/S0306261916000143.
[14] García-Armingol T, Ballester J, Smolarz A. Chemiluminescence-based sensing of
flame stoichiometry: influence of the measurement method. Measurement
2013;46(9):3084–97. http://dx.doi.org/10.1016/j.measurement.2013.06.008. URL
http://www.sciencedirect.com/science/article/pii/S0263224113002418.
[15] Muruganandam T, Kim B-H, Morrell M, Nori V, Patel M, Romig B, Seitzman J.
Optical equivalence ratio sensors for gas turbine combustors. Proc Combust Inst
2005;30(1):1601–9. http://dx.doi.org/10.1016/j.proci.2004.08.247. URL http://
www.sciencedirect.com/science/article/pii/S0082078404002863.
[16] Tripathi MM, Krishnan SR, Srinivasan KK, Yueh F-Y, Singh JP. Chemiluminescence-
based multivariate sensing of local equivalence ratios in premixed atmospheric
methane–air flames. Fuel 2012;93:684–91. http://dx.doi.org/10.1016/j.fuel.2011.
08.038. URL http://www.sciencedirect.com/science/article/pii/
S0016236111005102.
[17] Huang HW, Zhang Y. Digital colour image processing based measurement of pre-
mixed CH4+air and C2H4+air flame chemiluminescence. Fuel 2011;90(1):48–53.
http://dx.doi.org/10.1016/j.fuel.2010.07.050. URL http://www.sciencedirect.
com/science/article/pii/S0016236110004096.
[18] Garcia-Armingol T, Hardalupas Y, Taylor A, Ballester J. Effect of local flame
properties on chemiluminescence-based stoichiometry measurement. Exp Therm
Fluid Sci 2014;53:93–103. http://dx.doi.org/10.1016/j.expthermflusci.2013.11.
009.
[19] García-Armingol T, Ballester J. Influence of fuel composition on chemiluminescence
emission in premixed flames of CH 4 /CO2 /H2/CO blends. Int J Hydrogen Energy
2014;39(35):20255–65. http://dx.doi.org/10.1016/j.ijhydene.2014.10.039. URL
http://www.sciencedirect.com/science/article/pii/S0360319914028420.
[20] Panoutsos C, Hardalupas Y, Taylor A. Numerical evaluation of equivalence ratio
measurement using OH* and CH* chemiluminescence in premixed and non-pre-
mixed methane–air flames. Combust Flame 2009;156(2):273–91. http://dx.doi.
org/10.1016/j.combustflame.2008.11.008. URL http://linkinghub.elsevier.com/
retrieve/pii/S0010218008003623.
[21] Docquier N, Candel S. Combustion control and sensors: a review. Prog Energy
Combust Sci 2002;28(2):107–50. http://dx.doi.org/10.1016/S0360-1285(01)
00009-0. URL http://linkinghub.elsevier.com/retrieve/pii/S0360128501000090.
[22] Romero C, Li X, Keyvan S, Rossow R. Spectrometer-based combustion monitoring
for flame stoichiometry and temperature control. Appl Therm Eng
2005;25(5–6):659–76. http://dx.doi.org/10.1016/j.applthermaleng.2004.07.020.
[23] Parameswaran T, Gogolek P, Hughes P. Estimation of combustion air requirement
and heating value of fuel gas mixtures from flame spectra. Appl Therm Eng
2016;105:353–61. http://dx.doi.org/10.1016/j.applthermaleng.2014.11.034. URL
http://www.sciencedirect.com/science/article/pii/S1359431114010321.
[24] Ma L, Wu Y, Lei Q, Xu W, Carter CD. 3D flame topography and curvature mea-
surements at 5 kHz on a premixed turbulent Bunsen flame. Combust Flame
2016;166:66–75. http://dx.doi.org/10.1016/j.combustflame.2015.12.031.
[25] Floyd J, Geipel P, Kempf AM. Computed tomography of chemiluminescence (CTC):
instantaneous 3D measurements and phantom studies of a turbulent opposed jet
flame. Combust Flame 2011;158(2):376–91. http://dx.doi.org/10.1016/j.
combustflame.2010.09.006.
[26] Striūgas N, Zakarauskas K, Grigaitienė V. Comparison of steam reforming and
partial oxidation of biomass pyrolysis tars over activated carbon derived from waste
tire. Catal Today 2013;156(1):67–74.
[27] Ballester J, Garcia-Armingol T. Diagnostic techniques for the monitoring and con-
trol of practical flames. Prog Energy Combust Sci 2010;36:375–411.
[28] CCD signal to noise ratio, Andor Learning Center. URL http://www.andor.com/
learning-academy/ccd-signal-to-noise-ratio-calculating-the-snr-of-a-ccd.
[29] Andor Technology Ltd, iStar 734 Series. Specification sheet.
[30] Dribinsk V, Ossadtchi A, Mandelshtam VA, Reisler H. Reconstruction of
Abel–transformable images: the Gaussian basis–set expansion Abel transform
method. Rev Sci Instrum 2002;73(7):2634–42.
[31] Hernández J, Lapuerta M, Barba J. Flame stability and OH and CH radical emissions
from mixtures of natural gas with biomass gasification gas. Appl Therm Eng
2013;55(1):133–9.
[32] Kathrotia T. Reaction kinetics modeling of OH∗, CH∗ , and C∗2 chemiluminescence
[Ph.D. thesis]. Ruprecht–Karls–Universität, Heidelberg; 2011.
[33] Anderson T. Hydrogen Addition for Improved Lean Burn Capability on Natural Gas
Engine, Rapport SGC 134.
[34] Tunestal P, Christensen M, Einewall P, Andersson T. Hydrogen addition for im-
proved lean burn capability of slow and fast burning natural gas combustion
chambers. SAE Tech Pap 2002:7–8.
[35] Slim BK, Darmeveil H, Dijk GV, Last D, Pieters GT, Rotink MH, et al. Should we add
R. Navakas et al.
hydrogen to the natural gas grid to reduce CO2 emissions? proceedings of 23rd
world gas conference. 2006. p. 1–15.
[36] Parameswaran T, Gogolek P, Hughes P. Estimation of combustion air requirement
and heating value of fuel gas mixtures from flame spectra. Appl Therm Eng
2016;105:353–61. http://dx.doi.org/10.1016/j.applthermaleng.2014.11.034.
[37] Higgins B, McQuay MQ, Lacas F, Rolon JC, Darabiha N, Candel S. Systematic
measurements of OH chemiluminescence for fuel-lean, high-pressure, premixed,
laminar flames. Fuel 2001;80(1):67–74. http://dx.doi.org/10.1016/S0016-
489
Fuel 217 (2018) 478–489
2361(00)00069-7.
[38] Nori Venkata, Seitzman Jerry. Chemiluminescence measurements and modeling in
syngas, methane and jet-a fueled combustors, aerospace sciences meetings.
American Institute of Aeronautics and Astronautics; 2007. http://dx.doi.org/10.
2514/6.2007-466.
[39] Hardalupas Y, Orain M, Panoutsos CS, Taylor MK, Olofsson J, Seyfried H, et al.
Chemiluminescence sensor for local equivalence ratio of reacting mixtures of fuel
and air (FLAMESEEK). Appl Therm Eng 2004;24(11):1619–32.