See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/23310483

Photodiode-based sensor for flame sensing and combustion-process

monitoring

Article  in  Applied Optics · November 2008

DOI: 10.1364/AO.47.005541 · Source: PubMed

CITATIONS READS

10 1,104

4 authors:

Luis Arias Sergio N. Torres

University of Concepción University of Concepción
35 PUBLICATIONS   116 CITATIONS    89 PUBLICATIONS   1,495 CITATIONS   

SEE PROFILE SEE PROFILE

Daniel Sbarbaro Oscar Farias

University of Concepción University of Concepción
132 PUBLICATIONS   3,096 CITATIONS    21 PUBLICATIONS   116 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Implementation and validation of an early warning system for hydrometeorological hazards based upon unmanned aerial vehicles View project

An Enhancement to the Constant Range Method for Nonuniformity Correction of Infrared Image Sequences View project

All content following this page was uploaded by Luis Arias on 12 January 2016.

The user has requested enhancement of the downloaded file.

 Photodiode-based sensor for flame sensing
and combustion-process monitoring

Luis Arias,1,* Sergio Torres,1 Daniel Sbarbaro,1 and Oscar Farias2

1
Electrical Engineering Department, University of Concepcion, Concepcion, Chile
2
Mechanical Engineering Department, University of Concepcion, Concepcion, Chile
*Corresponding author: luiarias@udec.cl

Received 13 May 2008; revised 7 August 2008; accepted 5 September 2008;

posted 9 September 2008 (Doc. ID 95628); published 0 MONTH 0000

A nonintrusive low-cost sensor based on silicon photodiode detectors has been designed to analyze the
formation and behavior of excited CH
and C2
radicals in the combustion process by sensing the spectral
emission of hydrocarbon flames. The sensor was validated by performing two sets of experiments for both
nonconfined and confined flames. For a nonconfined oil flame, the sensor responses for the axial intensity
were highly correlated with the measurements obtained with a radiometer. For confined gas flames the
ratio between the signal corresponding to C2
and CH
was successfully correlated with the CO pollutant
emissions and the combustion efficiency. These results give additional insight on how to prevent an in-
complete combustion using spectral information. The fast response, the nonintrusive character, and the
instantaneous measurement of the needed spectral information makes the proposed optical sensor a key
element in the development of advanced control strategies for combustion processes. © 2008 Optical
Society of America
OCIS codes: 040.1240, 040.1880, 120.1740, 250.0040, 300.2140, 300.6170.

1. Introduction of the reacting substances. These methods have been

Improving the efficiency and reducing the pollutant
emissions in the combustion process is an important
challenge for modern industrial processes. This fact
has motivated the design of many sensors and tech-
niques to carry out direct or indirect measurements
for monitoring and/or controlling the thermal perfor-
mance and pollutant emissions. Optical sensing
based on the flames spectrum is an important field
of development, and it is used not only to identify
the flame presence but also to prevent an incomplete
combustion. Thus spectroscopy methods can provide
information on flame instability, O2 concentration,
CO pollutant emissions, flame morphology, energy
distribution of the elements, identification of ele-
ments within the flame (molecules, atoms, radicals,
and ions), and concentration variation of these ele-
ments from temperature, pressure, and composition
0003-6935/08/320001-09$15.00/0
© 2008 Optical Society of America
developed by using photodetectors [1], photodiodes
[2], CCD cameras [3,4], radiometers [5–7], laser
diodes [5,8], photomultiplier tubes [2,9], and UV cells
[10].
The flame spectrum has been classified according
to its energy distribution [11]: the continuous spec-
tra and the discontinuous spectra. In the first one
the emitted energy is distributed in a continuous
nondiscrete spectrum over a wide region of wave-
length and usually exhibits a maximum intensity
over the wavelength region. The continuous spectra
are generally observed in the sooty region of hydro-
carbon flames, and in some cases this emission may
be due to processes such as recombination of ions or
associations of atoms and radicals. In the second one
the energy is mainly confined around certain narrow
wavelengths, which are attributed to isolated atoms
or molecules. The discontinuous spectra are ar-
ranged in groups; each group is called a band of free
radicals.

10 November 2008 / Vol. 47, No. 32 / APPLIED OPTICS 1


Fig. 1. Typical spectrum of gas and oil flames.
Typical spectra (visible and near-infrared ranges)
of hydrocarbon flames are shown in Fig. 1. It is pos-
sible to see that the main difference between oil
flame and gas flame spectra is the presence of a con-
tinuous background emission. It is known that this
continuous spectrum emission is due to the soot for-
mation, which is similar to a graybody but with a
more complex variation of emissivity with the wave-
length [11,12]. Also, in premixed flames of hydrocar-
bons, a CO continuum has been identified in the
range of 300 to 550 nm [12]. It has been shown that
the spectral emission of flames is a combination of its
monochromatic emissivity Eλ and the Planck’s law of
blackbody radiation. According to the approach
of Ngendakumana et al. [12], the spectral intensity
of flames I λ can be reduced to the following function:
 
c1 · β c2
I λ ðα; β; TÞ ¼ αþ5 · exp − ; ð1Þ
λ λ·T
where λ is the wavelength, c1 ¼ 1:176 × 10−16 Wm2
and c2 ¼ 1:438 × 10−2 mK, T is the flame tempera-
ture, α ¼ 0:7, and β ¼ 1 × 10−9 .
Equation (1) represents the continuous spectral
emission of flames due to the soot formation. How-
ever, the visible and ultraviolet optical signature
in a gas flame spectrum typically consists of a com-
bination of molecular radiation (discontinuous
spectra) from electronically excited species, denoted
as chemiluminescence. Common wavelengths asso-
ciated to several radicals are shown in Table 1
[13]. Visible and ultraviolet wavelengths associated
with the chemiluminescence of several combustion
radicals are the OH
, CH
, and C2
at 306, 432,
and 516 nm, respectively. Radical OH
is attractive
because the reduced blackbody radiation (from par-
ticles and walls) in the ultraviolet, even in very lumi-
nous flames, increases its observability [2]. Radicals
CH
and C2
can provide easily observable signals, as
they are strong emitters, and their radiations are
2 APPLIED OPTICS / Vol. 47, No. 32 / 10 November 2008
Table 1. Flame Emission Wavelengths for
Common Combustion Radicals
Species Wavelength (nm)
OH 306–315
NH
336
CN
359–386
CH
390, 431
C2
469–473, 510–516
CH2 O
395, 423
CO2
broadband
spectrally resolved (i.e., spectrally narrow), though
their signals can be masked in oil flames due to
broadband CO and soot radiation that increases in
an incomplete combustion [3,11,12].
The massive use of radiometric techniques has
provided combustion state indexes, contributing to
a better understanding of combustion processes by
means of the effective detection of these radicals.
With such technology a combination of radical inten-
sity such as CH
=OH
, C2
=OH
, and C2
=CH
ratios
has been used in order to correlate them with the
different combustion parameters [1–3,7].
We present a new optical sensor for flame sensing
and combustion monitoring of hydrocarbon flames.
The sensor is validated by performing two sets of ex-
periments for both nonconfined and confined flames.
The main components of the proposed sensor are si-
licon photodiodes with interference optical filters
tuned for sensing the formation and behavior of ex-
cited CH
and C2
radicals in the combustion process.
These radicals can provide easily observable signals
to be monitored by the proposed sensor. According to
previous studies the formation of these radicals is
predominant in the reaction zone, emitting light at
430 and 514:5 nm center wavelengths. Along the
flame the emission of these radicals is masked,
and the axial distribution intensity on these bands
increases mainly by the soot radiation. On the other
hand, and according to the observations of Gaydon
and Wolfhard [14], the behavior of both radicals’
emission in the reaction zone exhibits a maxima at
rich mixtures of air-to-fuel, with a slight shift. There-
fore the C2
=CH
ratio exhibits a maximum at a
richer mixture than the maximum shown by the
respective radicals.
The cheaper construction cost, and the identifica-
tion of only two spectral bands, is a considerable im-
provement over commercial radiometers. The axial
emission distribution along the flame detected by
the sensor shows a high correlation with the mea-
surement results obtained by the USB2000 radio-
meter for a nonconfined oil flame. On the other
hand, an important similarity with the results and
observation given by Gaydon and Wolfhard [14]
was obtained in a confined gas flame, estimating
the maximum combustion efficiency near to the stoi-
chiometric conditions.
Section 2 provides the information concerning the
design of the photodiode-based sensor for flame
sensing, which uses the spectral detection of specific
radicals. The equations that describe the sensor
design are also presented in Section 2. Finally, in
Section 3, two common tasks for monitoring an oil
burner and a gas burner are presented, where
validation results are also discussed.
2. Monitoring Flames by Means of Photodiodes
Our goal is to demonstrate the use of a photodiode-
based sensor for flame sensing and combustion-
process monitoring by means of spectral detection.
This task was developed by measuring the formation
and behavior of two radicals’ intensity, CH
and C2
,
emitting at λ1 ¼ 430 nm and λ2 ¼ 514:5 nm. Two
silicon photodiodes assembled with optical filters
provided the required sensitivity to detect the
required spectral radical emissions.
A. Description of the Sensor System
The principal component in the built sensor is the
T-18 combined silicon photodiode/filter provided by
INTOR, Inc. The equation that describes the photo-
current generated by the photon incidence in this
device is
iph;λm ðλm Þ ¼ ℜðλm Þ · Popt;λm ;
function that describes the readout electronics for
both channels is given by the following:
V out;λm ðVÞ
¼ Gre ; m ¼ 1; 2; ð3Þ
iph;λm ðμAÞ
where V out;λm is the voltage measured in both
channels, iph;λm is the photocurrent generated by
the combined photodiode/filter (in μA) at λm , and
Gre is the readout electronic gain, fixed in 20.
Then, replacing Eq. (2) in Eq. (3), we obtain
V out;λm ¼ Gre · ℜðλm Þ · Popt;λm ; m ¼ 1; 2: ð4Þ
The T-18 devices are provided with the necessary
optical filters in order to sense the correspondent ra-
dicals. The center wavelength of the used filters are
λ1 ¼ 430  5 nm (for CH
radical sensing) and λ2 ¼
514:5  5 nm (for C2
radical sensing) with 45%
and 70% of transmission, respectively, as shown
in Fig. 2.
Because of the spectral difference in responsivity
ℜðλm Þ and filter transmission T x at both wave-
m ¼ 1; 2; ð2Þ lengths λ1 and λ2 observed in Fig. 2, the following
was used as a correction factor C:

where λm represents the center wavelength asso-

ciated to the optical filter used. ℜðλm Þ is the T-18 ℜðλ2 Þ T x ðλ2 Þ
C¼ · : ð5Þ
photodiode responsivity at λm , measured in am- ℜðλ1 Þ T x ðλ1 Þ
pere/watt, and it is shown in Fig. 2. iph;λm is the photo-
current generated by incidence optical power Popt;λm
emitted by the flame over the active area of the This factor was introduced to correlate the corre-
photodiode. spondent signals measured in the sensor channels
In order to drive these devices, two readout electro- with the real intensities of the emitted radicals in
nic channels were constructed. Both photodiodes the flame. This factor was applied by software.
were operated in the photoconductive mode; there- Finally, by using the Eqs. (4) and (5), the transfer
fore we obtained a linear response as expressed in function for each channel can be expressed by
Eq. (2) between the optical power and the photocur-
rent generated by the photodiodes. Thus the transfer
I s;λ1 ¼ C · V out;λ1 ; ð6Þ
0.8 80

Filter centred at
0.7 514.5nm 70
Responsivity
I s;λ2 ¼ V out;λ2 ; ð7Þ
0.6 Filter centred at 60
Filters transmission (%)

430nm
where I s;λm , m ¼ 1; 2, is the intensity sensor (in volts)
Responsivity (A/W)

0.5 50
measured by both channels at λ1 ¼ 430 nm
0.4 40 and λ2 ¼ 514:5 nm.
Equations (6) and (7) represent the stationary
0.3 30
state transfer function for both readout electronic
0.2 20
channels constructed for driving the photodiodes.
However, the linearity expressed in Eq. (4) between
0.1 10 V out;λm and Popt;λm can be affected by the total current
noise, given by shot noise isn and thermal (Johnson)
0
300 400 500 600 700 800 900 1000 1100
0
1200 noise iJn . Shot noise is the dominating noise source
Wavelength (nm) when the photodiode is operating in photoconductive
Fig. 2. Responsivity of the T-18 silicon photodiodes and transmis- mode, and the Johnson noise is the dominating one in
sion of the filters assembled with the photodiodes. a high-temperature medium, which is the case of this

10 November 2008 / Vol. 47, No. 32 / APPLIED OPTICS 3

application. In order to evaluate the incidence of the
shot and Johnson noise, the response linearity for
both channels was measured at an ambient tempera-
ture (22 °C) and a working temperature (40 °C). This
result is depicted in Figs. 3 and 4, where a linear
response is observed in both channels between
V out;λm and Popt;λm , which is not affected by changes
in the working temperature. The responsivity of both
photodiodes was calculated from Figs. 3(a) and 4(a),
resulting in ℜð430 nmÞ ¼ 0:11 ðA=WÞ and
ℜð514:5 nmÞ ¼ 0:205 ðA=WÞ. Replacing such results
in Eq. (5), the correction factor obtained is C ¼ 2:899.
The readout electronic channels were implemen-
ted by using high-performance operational ampli-
fiers and precision electronic devices, and it was
installed in a self-shielding box.

3. Experimental Testing
With the purpose of corroborating the presence of the
main combustion radicals, the spectral radiation for
oil and gas flames was measured in the reaction zone
by using the USB2000 radiometer (Ocean Optics
Inc.) as is shown in Fig. 5. Note that it is possible
to distinguish the main bands of CH
and C2
radi-
cals in both cases. One verifies that the main reacting
zone can be characterized by these specific emissions
[11,12,14], which are able to provide the observable
signals to be monitored by the photodiode array sen-
sor. The spectral response of the assembled filters is
displayed in Fig. 5 in order to corroborate its spectral
syntonization.
To validate the proposed sensor, two common tasks
for monitoring an oil burner and a gas burner were
used as experimental testing. The formation and be-
havior of excited CH
and C2
radicals were sensed at Fig. 3. Spectral band of the assembled filters and syntonization
different combustion conditions by changing the bur- in (a) the gas flame spectrum and (b) the oil flame spectrum.
ner air damper. In the first test the axial distribution
intensities in the spectral radical’s bands were mea-
sured along the oil flame by using constructed sensor where x ¼ I axial−rad;λm and y ¼ I axial−s;λm are the radio-
signal I axial−s;λm together with USB2000 radiometer meter and the sensor intensity result, which repre-
signal I axial−rad;λm for different air step (AS) changes. sents the axial distribution intensity in the
This test was carried out in a nonconfined environ- spectral radical’s bands along the flame.
ment (free space) as is shown in Fig. 6(a), where The second test is focused on the estimation of the
any change in the AS represents a different condition combustion efficiency and the CO emissions in a con-
of the combustion state. fined gas boiler. This task was developed by using the
The linearity of the radiometer response was also C2
=CH
ratio estimated by correspondent sensor in-
checked at λ1 ¼ 430 nm and λ2 ¼ 514:5 nm. This tensities I s;λm measured in the reaction zone. By
result is depicted in Figs. 3(a) and 4(a), where a non- using Eqs. (6) and (7), the ratio was calculated as
linearity for radiometer intensity I rad;λm was ob-
served at λ1 ¼ 430 nm in the range of 0 to 0:06 μW
[Fig. 3(a)]. C2
I s;λ2 V out;λ2
The oil flame was focalized on the photodiode array ¼ ¼ ; ð9Þ
and on the optical fiber (used by the radiometer) by CH
I s;λ1 C · V out;λ1
means of an optical system. In order to compare the
results obtained by both instruments, cross correla-
tion index r2x;y was calculated between both measure- at λ1 ¼ 430 nm and λ2 ¼ 514:5 nm.
ment results. This metric is defined as Sensor intensities I s;λm represent the behavior of
the radical’s intensity CH
and C2
at different com-
bustion conditions given by the change on air excess
covðx; yÞ
r2x;y ¼ ; ð8Þ lamdað1 þ eÞ. This variable represents the air-to-
varðxÞ · varðyÞ fuel ratio of the actual combustion processes over

4 APPLIED OPTICS / Vol. 47, No. 32 / 10 November 2008


Fig. 4. Linear response at 430 nm of (a) the photodiode sensor
and USB2000 radiometer at 22 °C and of (b) the photodiode sensor
at 40 °C.
the corresponding stoichiometric air-to-fuel ratio,
defined as
ðair=fuelÞactual
lambdað1 þ eÞ ¼ : ð10Þ
ðair=fuelÞstoichiometric
The experimental setup is shown in Fig. 6(b). The
flame area is sensed by the photodiodes from the rear
part of the burner where the photodiodes array is
cooled by means of the fresh gas and air flux. The
tests were performed in a central heating boiler of
150 kW equipped with a Riello GS20 gas burner
and instrumented with gas and water flow meters,
thermocouples, and a gas analyzer (at exit of chim-
ney) to obtain the online combustion parameters
and the boiler efficiency. In order to verify the spec-
tral detection of the proposed sensor, online signals
I s;λm are depicted in Fig. 7 for several air excess va-
lues (lamdað1 þ eÞ). Note that the changes on
Fig. 5. Linear response at 514:5 nm of (a) the photodiode sensor
and USB2000 radiometer at 22 °C and of (b) the photodiode sensor
at 40 °C.
lamdað1 þ eÞ are tracked by the photodiode sensor
signals. In addition, I s;514:5 nm (for the C2
radical
sensing) is stronger than I s;430 nm (for the CH
radical
sensing), according to previous studies [11,12,14].
A. Measuring Axial Radiation
This test was carried out in a dark nonconfined en-
vironment as shown in Fig. 6(a). The axial distribu-
tion intensities of flames I axial−s;λm and I axial−rad;λm
were measured versus the axial position from 0 to
80 cm at λ1 ¼ 430  5 nm and λ2 ¼ 514:5  5 nm.
The main experimental results are depicted in Figs. 8
and 9. The AS was increased from AS ¼ 1 to 8, where
AS ¼ 8 represents the maximum air quantity
allowed by the burner air damper.
From the experimental results it can be seen that
the intensities in the spectral radical’s CH
and C2

bands, measured with both instruments, followed
similar patterns. A maximum peak is found along
the flame where the intensity starts to decrease
and finally disappears at the flame end. Note that
10 November 2008 / Vol. 47, No. 32 / APPLIED OPTICS 5
Fig. 6. Test bench for (a) nonconfined oil fuel and (b) confined gas
fuel tests.

I axial−s;430 nm ranges between 0 and 0:373 V and

I axial−s;514:5 nm ranges between 0 and 4:79 V. Note also
that by using Eqs. (4) and (6), the optical power range Fig. 8. Sensing the CH
radical formation along the flame versus
calculated at λ1 ¼ 430 nm is 0 to 0:058 μW, and by axial position by using (a) radiometer intensity I rad;430 nm and
using Eqs. (4) and (7), the optical power range at λ2 ¼ (b) intensity sensor I s;430 nm.
514:5 nm is 0 to 1:17 μW.
Axial distribution intensities I axial−s;λm and
I axial−rad;λm were stored in a personal computer by In order to compare the experimental results, the
using a data acquisition board at 100 kHz and were cross-correlation index given by Eq. (8) between
processed by LabView (National Instruments). I axial−rad;λm and I axial−s;λm was calculated and summar-
ized in Table 2. Note that for the CH
(at
0.37 5 λ1 ¼ 430 nm) sensing, it is possible to identify a smal-
0.36 4.83
ler cross correlation at AS ¼ 6, 7, and 8. This is be-
cause at that condition, the radiometer is operating
0.35 4.66 in the nonlinear range.
C2* radical signal lambda(1+e)
0.34 4.5
1.6 B. Measuring Combustion Efficiency
lambda(1+e)
(V)
(V)

0.33 1.71
4.32 This experiment was performed for the following two
s,514.5nm

lambda(1+e)
s,430nm

1.82
0.32 lambda(1+e)
4.16
firing rates: 130 kW [low firing rate (LFR)] and
1.88
150 kW [high firing rate (HFR)], with lamdað1 þ eÞ
I

lambda(1+e)
0.31 1.95 4 decreasing by closing the burner air damper from
0.3
CH* radical signal
3.83
2.11 to 1.40 and from 1.87 to 1.38, respectively.
The real-time response of I s;430 nm (for the CH
radical
0.29 3.66 sensing behavior), I s;514:5 nm (for the C2
radical sen-
3.5 sing behavior), and the gas analyzer signals were
0 200 400 600 800 1000 1200
Time (s) stored in a personal computer by using a data acqui-
sition board at 100 kHz and were processed by
Fig. 7. Online sensor signals measured in a confined gas flame. LabView.

6 APPLIED OPTICS / Vol. 47, No. 32 / 10 November 2008

 abruptly and the boiler efficiency [Fig. 10(a)] began
to decrease at low levels of lamdað1 þ eÞ.
Sensor intensities I s;λm that represent the behavior
of the radical’s intensity CH
and C2
have a similar
dynamic as is shown in Fig. 10(b). Note also that sig-
nal peaks are achieved at excess air values higher
than the critical one defined by the CO increasing.
In addition, a slight shift of the maximal levels for
CH
and C2
radicals is observed according to the
observations of Gaydon and Wolfhard [14]. However,
the trends of the C2
=CH
intensity ratio [Fig. 10(c)]
show a maximum value when lamdað1 þ eÞ decreases
to near the critical point of incomplete combustion
and also decreases when CO increases abruptly.

4. Conclusions
A photodiode-based sensor has been designed for
sensing the formation and behavior of CH
and
C2
radical emission of oil and gas flames in the com-
bustion process. The sensor has been tested by mea-
suring the flame axial distribution intensities in the
spectral radical’s bands along a nonconfined oil flame
and the combustion efficiency by using the C2
=CH

ratio estimated by the corresponding sensor inten-
sity signals in the reaction zone of a confined gas
flame. In the first test the sensor was validated by
using a radiometer, where a high correlation has
been achieved between the measurements of both in-
struments for different AS changes. The results ob-
tained in the second test for a confined gas flame
agree with the results and observations given by
Gaydon and Wolfhard [14]. It was experimentally de-
monstrated that the sensor is able to detect and track
the air flux changes in the combustion process in
spite of the power combustion changes. The results
Fig. 9. Sensing the C2
radical formation along the flame versus have shown that the peak levels of CH
and C2
ra-
axial position by using (a) radiometer intensity Irad;514:5 nm and dical intensities were located at different air excess
(b) intensity sensor I s;514:5 nm. lamdað1 þ eÞ and that the C2
=CH
intensity ratio es-
timated by the sensor signals is a good indicator of
the combustion state, since it can predict the point
The results of the boiler performance and the of optimal air excess condition leading to a maximal
radical intensities at LFR and HFR versus the air efficiency of the equipment near to the stoichiometric
excess (lamdað1 þ eÞ) are shown in Fig. 10. Note that conditions and at a low level of pollutant emissions
(CO). Such a result means that it is possible to envi-
the optimal operation condition is clearly given by
sage an optimal strategy for combustion control. The
critical excess air when CO emission increased
low construction cost, the fast response, the nonin-
trusive character, and the identification of only two
spectral bands are considerable improvements over
Table 2. Cross Correlation between the Measurement commercial radiometers providing high potential
Obtained by the Radiometer and the Proposed Sensor for in advanced-control strategies aiming at the optimi-
Different Air Steps (Different Combustion Conditions) zation of the combustion process.
AS r2 x;y ; 430 nm r2 x;y ; 514:5 nm The authors thank the Thermofluid Laboratory,
1 0.9888 0.9981 University of Concepcion, Chile, and the Thermody-
2 0.9914 0.9978 namics Laboratory, University of Liege, Belgium.
3 0.9882 0.9972 This research was partially supported by the Fund
4 0.9800 0.9963 for the Promotion of Scientific and Technological
5 0.9672 0.9928 Development (FONDEF) grant D03I1106, Alma-
6 0.9532 0.9918 Conicyt 30105006, Grant Milenion ICM P02-049,
7 0.9514 0.9879
and the collaboration of Santander Santiago
8 0.9527 0.9902
Fellowship.

10 November 2008 / Vol. 47, No. 32 / APPLIED OPTICS 7

Fig. 10. Experimental results at LFR and HFR for confined gas flame. (a) Thermal efficiency of the boiler versus air excess (lamdað1 þ eÞ).
(b) CH
and C2
radical-behavior sensing by using intensity sensor Is;430 nm and I s;514:5 nm versus air excess. (c) C2
=CH
intensity ratio
versus air excess obtained by using the photodiode sensor intensities.

References 3. C. Martins, J. Carvalho, and M. Ferreira, “Ch and C2 radicals

characterization in natural gas turbulent diffusion flames,” J.
1. J. L. Pau, J. Anduaga, C. Rivera, A. Navarro, I. Alava, M.
Braz. Soc. Mech. Sci. Eng. 27, 110–118 (2005).
Redondo, and E. Munoz, “Optical sensors based on iii-nitride
4. D. Sbarbaro and O. Farias, “Real time monitoring and
photodetectors for flame sensing and combustion monitoring,”
characterization of flames by principal-components analysis,”
Proc. IEEE 45, 7498–7503 (2006).
Combust. Flame 132, 591–595 (2003).
2. J. Ballester, A. Sanz, R. Hernandez, and A. Smolarz,
5. N. Docquier and S. Candel, “Combustion control and
“Detection and analysis of emitted radiation for advanced
sensors: a review,” Prog. Energy Combust. Sci. 28, 107–150
monitoring and control of combustors,” Proc. SPIE 5948,
(2002).
594824.1–594824.11 (2005).

8 APPLIED OPTICS / Vol. 47, No. 32 / 10 November 2008

 6. J. B. Michel, C. Champinot, and J. Dugue, “State of the art on
emerging combustion control sensors,” presented at the 6th
International Conference on Technologies and Combustion
for a Clean Environment, 2001.
7. C. Romero, X. Li, S. Keyvan, and R. Rossow, “Spectrometer-
based combustion monitoring for flame stoichiometry and tem-
perature control,” Appl. Thermal Engineer. 25, 659–676 (2005).
8. A. Arnold, H. Becker, R. Hemberger, W. Hentschel, W.
Katterle, M. Kollner, W. Meienburg, P. Monkhouse, H. Neckel,
M. Schafer, K. P. Shindler, V. Sick, R. Suntz, and J. Wolfrum,
“Laser in situ monitoring of combustion processes,” Appl. Opt.
29, 4860–4872 (1990).
9. M. Khesin, “Demonstration of new frequency-based flame
monitoring system,” Proceedings of the 59th Annual American
Power Conference, Vol. 58-II (1997), pp. 1010–1013.
10. J. Michel, O. Chetelat, N. Weber, and O. Sari, “Flame signa-
ture as a low-cost flame control method,” Proceedings of the
View publication stats
5th International Conference on Technologies and Combustion
for a Clean Environment (Combustion Institute, 1999),
pp. 1251–1253.
11. G. Zizak, “Flame emission spectroscopy: fundamentals and
applications,” Tech. Rep. (Instituto per la Tecnologia dei
Materiali e dei Processi Energitici, 2000).
12. P. Ngendakumana, B. Zuo, and E. Winandy, “A spectro-
scopic study of flames for a pollutant formation regulation
in a real oil boiler,” presented at the 2nd International
Conference on Technologies and Combustion for a Clean
Environment, 1993.
13. M. Morrel, J. Seitzman, M. Wilensky, E. Lubarsky, J. Lee, and
B. Zinn, “Interpretation of optical emissions for sensors in
liquid fueled combustors,” AIAA J. 787, 1–12 (2001).
14. A. G. Gaydon and H. G. Wolfhard, The Spectroscopy of Flames,
1st ed. (Chapman and Hall, 1957), Chap. 7.
10 November 2008 / Vol. 47, No. 32 / APPLIED OPTICS 9