Research Article Vol. 29, No.

1 / 4 January 2021 / Optics Express 244

Tomographic imaging using multi-simultaneous

measurements (TIMes) for flame emission
reconstructions
C HEAU T YAN F OO, 1 A NDREAS U NTERBERGER , 1 J AN M ENSER , 2
AND K HADIJEH M OHRI 1,3,*
1 University of Duisburg-Essen, Institute for Combustion and Gas Dynamics - Tomography Group,
Carl-Benz-Str. 199, 47057 Duisburg, Germany
2 University of Duisburg-Essen, Institute for Combustion and Gas Dynamics - Reactive Fluids,
Carl-Benz-Str. 199, 47057 Duisburg, Germany
3 University of Duisburg-Essen, Institute for Combustion and Gas Dynamics - Chair of Fluid Dynamics,
Carl-Benz-Str. 199, 47057 Duisburg, Germany
* khadijeh.mohri@uni-due.de

Abstract: The method of tomographic imaging using multi-simultaneous measurements

(TIMes) for flame emission reconstructions is presented. Measurements of the peak natu-
ral CH* chemiluminescence in the flame and luminescence from different vaporised alkali metal
salts that were seeded in a multi-annulus burner were used. An array of 29 CCD cameras
around the Cambridge-Sandia burner was deployed, with 3 sets of cameras each measuring a
different colour channel using bandpass optical filters. The three-dimensional instantaneous and
time-averaged fields of the individual measured channels were reconstructed and superimposed
for two new operating conditions, with differing cold flow Reynolds numbers. The contour of
the reconstructed flame front followed the interface between the burnt side of the flame, where
the alkali salt luminescence appears, and the cold gas region. The increased mixing between
different reconstructed channels in the downstream direction that is promoted by the higher
levels of turbulence in the larger Reynolds number case was clearly demonstrated. The TIMes
method enabled combustion zones originating from different streams and the flame front to be
distinguished and their overlap regions to be identified, in the entire volume.

© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

1. Introduction
The inherent nature of practical flames calls for measurement techniques that can provide spatio-
temporal information, to enable in-depth understanding of different processes. In recent years,
three-dimensional (3D) tomography methods have seen a surge in development and application
in the field of engineering and natural sciences. Computed tomography (CT) has been combined
with a host of measurements such as X-ray, schlieren, shadowgraphy, emission, absorption and
so on [1–10]. The combustion community has contributed a considerable portion to the recent
literature for various investigations of different combusting flows. Tomographic techniques
have shown to provide a leap in better understanding real flames, and performing this using
multi-simultaneous measurements has been demonstrated by Kerl et al. [11] and later by Ebi and
Clemens [12] through novel approaches for simultaneous measurements of the 3D flame front
and velocity fields of premixed flames. Oil droplets were used for particle image velocimetry
(PIV) data collection from which the volumetric velocity field could be inferred tomographically,
and the preheat zone that is marked by vaporisation of the oil droplets was used to track the
flame front at the same time. This rich pool of simultaneous 3D tomographic data can be used to
analyse the turbulent flow-flame interactions. Wang et al. [13] presented simultaneous 3D particle
temperature, particle concentration and H2 O concentration distributions using multi-spectral

#412048 https://doi.org/10.1364/OE.412048
Journal © 2021 Received 13 Oct 2020; revised 22 Nov 2020; accepted 24 Nov 2020; published 22 Dec 2020
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 245

flame images to investigate axisymmetric sooting flames. Ren and Modest [14] recently utilised
machine learning to reconstruct the 3D temperature and species concentration simultaneously in
a laminar flame based on infrared emission spectral measurements.
Flame emission reconstructions have been published in numerous papers over the past
decade [8,15–26], presenting some excellent and novel ways in which chemiluminescence
(CL) measurements via multiple cameras were used for studying flames. Focus has been
on investigating the 3D flame structures either by measuring emissions from specific excited
molecules such as OH*, CH* and C2 * [23] or by measuring the entire emission spectrum [27],
and other phenomena such as the ignition process in complex supersonic flows [20], forced
and unforced combustion [17,18] and thermoacoustic instabilities [21,26]. However, there is a
lack in utilising multi-simultaneous measurements, and we would like to elaborate on this by
demonstrating that application of the TIMes method to flame emission reconstructions can be
relatively simple and straightforward, but can advance flame emission reconstructions to the next
Tomographic imaging using multi-simultaneous
level by adding valuable information about the interaction of different effects within the volume of
interest. To demonstrate this, we planned experiments that allow measurement of emission from
measurements (TIMes) for flame emission
separate streams of a dual-annulus burner individually. Based on literature, several methods can
reconstructions
be used to illuminate the streams. For example using the concept of particle-seeding as introduced
by Elsinga et al. [28] for PIV which requires volumetric illumination and has been implemented
tomographically [29,30], or oil droplet-seeding as introduced by Boyer [31] and implemented
C HEAU T YAN F OO , 1 , A NDREAS U NTERBERGER , 1 , J AN M ENSER , 3
by several researchers [11,12].1,2,* We chose to follow a less complex experimental approach by
AND K HADIJEH
measuring the flame M CLOHRI
simultaneously with the luminescence from vaporised alkali metal
1tracers
University
that of Duisburg-Essen,
were added to the Institute for Combustion
two different and Gasvia
burner streams, Dynamics
a simple- Tomography
adaptation toGroup,
our current
Carl-Benz-Str. 199, 47057flame
low-cost multi-camera Duisburg, Germany
imaging setup. Atomic emission spectroscopy (AES) has been
2deployed
Universitybyofmany
Duisburg-Essen,
researchersInstitute for Combustion
for flame diagnosticsand
suchGas
as Dynamics - Chair
thermometry andofequivalence
Fluid Dynamics,
ratio
Carl-Benz-Str.
determination,199,for 47057 Duisburg,
both open Germany
laboratory flames and harsh engine conditions [32–36]. Aspirating
3 University of Duisburg-Essen, Institute for Combustion and Gas Dynamics - Reactive Fluids,
alkali metal salt solutions into a heated region, such as the burnt gases of a combusting flow,
Carl-Benz-Str.
can cause the 199, 47057
alkali Duisburg,
metal atoms toGermany
achieve an excited energy state which has a limited lifetime
* khadijeh.mohri@uni-due.de
and will hence spontaneously relax to a lower energy state by emitting a photon as one possible
pathway. The wavelength of the emitted light will correspond to the energy difference between
the two energy states, and this is different for each alkali metal. The spectroscopic properties of
the alkali metal salts chosen for the current study are well-documented in the literature [37,38],
Figure 1
and the principle of AES and how it applies to the current work are depicted in Fig. 1.

Vapourisation/ -
Dissociation
M+ X
Desolvation
M+ M+ Metal ion
X-

+ -
M X

-
Gaseous
M+ X particles
Dense fog M+
Electronic
X-
of atomised excitation
M+
liquid particles M+ X - to higher Signal detected
energy states by CCD camera
with appropriate
bandpass filters
Emission of
+ - ++
light due to M+ Light
- - - electronic
+
Metallic de-excitation
Nebuliser
salt solution

Fig. 1. The principle of atomic emission spectroscopy as applicable to this work.

 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 246

We have seeded the two separated combusting flow streams of the Cambridge-Sandia burner
[39,40] with different vaporised salts that are known to emit light at distinctive wavelengths. The
simultaneous salt luminescence and flame front measurements were used to reconstruct each
measured channel separately. The flame emission spectra for each flame operating condition
were measured and combined with knowledge of the different salt luminescence wavelength
ranges to identify optically separated detection channels. The reconstructions presented here
were performed using computed tomography of chemiluminescence (CTC) [16,41] that is based
on the algebraic reconstruction technique to solve the inverse tomography problem.
The TIMes method for flame emission reconstructions can be of interest to a wide range of
applications. For example, 3D reconstructions based on simultaneous measurements of CL from
flame radicals and electronically excited material-specific molecules and atoms in flame pyrolysis
techniques for nanoparticle synthesis will allow the limits and spatial correlation of different
zones such as combustion, particle clouds and different fluid streams to be determined in the
entire volume. This enables better understanding of the synthesis process which allows us to
control it more effectively. Staged or sequential combustion technology which feature multiple
combustion zones, such as those used in power generation gas turbines [42], coal boilers, glass
melting furnaces and rich-quench-lean (RQL) concepts [43] for achieving pollution reduction or
tailor-made operation can also benefit from the TIMes method application that is presented in this
paper. Stratified combustion involves different premixed reactant streams with varied equivalence
ratios. TIMes can provide a new perspective for better understanding the stratification process
based on 3D experimental flame emission reconstructions which allow the interaction of different
streams with each other and with the flame front to be studied.

2. TIMes procedure
2.1. Experimental method
The Cambridge-Sandia burner which constitutes an inner and outer annulus, providing two
streams of premixed natural gas and air, a slow outer co-flow of air and a ceramic bluff body was
used. Figure 2 shows the cross section of the burner exit and details of the gas mixtures for each
flow stream. The equivalence ratio of the premixed gases in each annulus and the air co-flow
were regulated using Bronkhorst mass flow controllers. More detailed descriptions of the burner
are available from [39,40].
To visualise the streams separately, different vaporised metal salts (NaCl and Sr(NO3 )2 ) were
seeded into the inner and outer annuli, as depicted in Fig. 2, using a bubbler system for each
stream. Both salts are water-soluble and were prepared in the same manner at room temperature,
but with different final concentrations. The crystalline salts were first dissolved in a small
volume of distilled water before further dilution with ethanol. The final concentrations for NaCl
and Sr(NO3 )2 used in the experiments were 5g/L and 30g/L respectively. The NaCl solution
concentration was considerably lower, to avoid saturation of the measured signal due to the much
brighter light emission compared to the Sr(NO3 )2 stream. An ultrasonic nebuliser was placed
within the bubbler to vaporise part of the salt solution. An adequate amount of air flow was
introduced into the free space above the liquid inside the bubbler to push the vapour out, as shown
in Fig. 2. To prevent cross contamination between the bubbler flows, separate air in-flow streams
were used for each. The amount of ethanol in the outgoing streams was below the saturation
pressure in air since no mist was visible in the burner stream flows without ignition. With an
estimate of the mass fraction being 580 ppm for Na and about six times this for Sr, the amount of
salts in the streams is assumed to be too low to significantly alter the flame chemistry.
Two different flame conditions were considered, as detailed in Table 1, depending on the
volume flow rate for air V̇ air and natural gas V̇ nat. gas resulting in a different equivalence ratio ϕ,
and cold flow Reynolds number Re in each case. The co-flow constituted 366 slm of air and the
average bulk velocity was approximately 0.2 m/s for both cases. In the case of Flame I, the outer
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 247

Fig. 2. Schematic of the Cambridge-Sandia burner exit geometry cross-section, showing

the salt solutions that were seeded into each annulus.

gas mixture was slightly richer than the inner one. Flame II had the same equivalence ratio and
Reynolds number in both flow streams, with the Reynolds number being higher than for Flame I.

Table 1. The flame conditions used in the experiments. The subscripts i and o stand for the inner
and outer streams respectively, and texp is the camera exposure time.

Inner stream Outer stream

Cases V̇ (nat. gas)i V̇ (air)i Rei φi V̇ (nat. gas)o V̇ (air)o Reo φo texp
[slm] [slm] [slm] [slm] [µs]
Flame I 5.3 50.0 3723 1 5.0 40.0 2084 1.2 400
Flame II 8.9 84.4 6282 1 12.7 120.7 6282 1.0 600

Our generic multi-camera setup for CTC from previous work [22] was used here. In this case,
29 Basler acA-645-100gm monochrome CCD cameras, arranged equiangularly in one plane
within a total fan angle of 168◦ around the burner, were utilised. All cameras were positioned at a
fixed distance of 400mm from the burner centre, and were equipped with Kowa C-Mount lenses
′′
(focal length 12mm) set at the maximum aperture opening f/1.4. The cameras have a 12 Sony
ICX414 sensor with a resolution of 659 by 494 pixels, with each pixel being 9.9 µm by 9.9 µm
in size. The spectral response of the cameras at >10% is in the visible range, 400–775 nm. The
cameras were simultaneously triggered at 5 Hz from one source, and 200 instantaneous images
were captured per camera, each camera using an exposure time of texp = 400 µs and 600 µs
for Flame I and Flame II cases, respectively. The detected signals were maximised by using
the largest aperture opening and binning while imaging. The limits of the settings and blurring
effects due to exposure time for the cameras are discussed in [22], and were considered here to
find the best compromise. The laboratory lights were switched off and a black background was
placed behind the flame to improve the quality of the measurements. Images of the background
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 248

scene, captured without the flame by each camera, were subtracted from the flame images for
background correction. A picture of the setup, showing the flame operated with the vaporised
Figure 3 is presented in Fig. 3.
salt-seeding

x
y

Fig. 3. Experimental setup used for the method of TIMes for flame emission tomography.
The green, red and blue dots denote the cameras that detected the emitted light from the
inner (NaCl) and outer Sr(NO3 )2 streams and the flame front, respectively.
Figure 7
Optical bandpass filters were mounted in front of the lenses in the specific repetitive sequence
shown in Fig. 3, resulting in a total of 10 cameras for detecting the inner and outer streams each,
and 9 cameras for the flame front. The narrowband TECHSPEC filter, detecting wavelengths
in the range λ = 589 ± 5 nm, was used to measure the bright atomic emission of sodium at
λz/d==588.99
1.6 nm and λ = 589.59 nm [44] in the inner stream. The larger bandwidth TECHSPEC z/d = 1.6
filter, for wavelength detection in the range λ = 697 ± 37.5 nm, was used to cover the strontium
monohydroxide emission expected from the Sr(NO3 )2 [37] in the outer stream. Schott BG40
filters were added to the cameras for the outer stream measurements, to suppress detection of
the
z/d =thermally
0.4 excited H2 O emission [22]. The flame front was tracked using BrightLinez/dfilters
= 0.4
for the wavelength range λ = 433 ± 12 nm. Figure 4 presents the transmission bands for all the
filters used, superimposed onto the emission spectrum measured for both flames to illustrate our
procedure for optically separating the detected channels. Both the transmission and emission
data were normalised by the maximum value for each case. The emission measurements were
made after a NaCl seeding test, and hence the presented spectra exhibit the large atomic sodium
signal due to remnants within the burner.

z/d = 2.8 z/d = 2.8

z/d = 0.8 z/d = 0.8

Fig. 4. The transmission curve of the optical filters used in the experiment and the emission
spectrum of both flames investigated.
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 249

As shown in Fig. 4, the wavelength ranges detected for each salt-seeded stream do not coincide
with each other or with the intense spectral lines of the CL emission from CH*. Additionally, the
separate detection channels were tested by seeding the salts one at a time and checking that only
the intended cameras measured a signal for the camera settings used in the experiments. It must
be noted that all bandpass filtered signals contain the broadband CL emission from CO*2 which
was not corrected for in this work. Since we do not aim to infer any quantified information about
a particular flame property at this stage, we believe that the correction for CO*2 is not paramount
here. When necessary for particular investigations, corrections can be applied and some methods
have been proposed [45].

2.2. Reconstruction method

Details of the CTC algorithm used in this project, and previous applications of it for experimental
reconstructions are documented in [16,22,27], and hence only a brief description of the fundamen-
tals is provided here. The CTC technique involves measuring two-dimensional (2D) projections
of the CL emitted by a combusting gas from unique perspectives. The radiative transfer equation
(RTE) describes the propagation and interaction of light within the domain of interest, and is the
basis of the measurement model that is used. In the CTC used here, subtracting the background
signal (the scene, flame off) and neglecting scattering and re-absorption simplifies the RTE.
These assumptions are reasonable for the flames that are used here, and unlike the CL signal from
OH*, the bandpass-filtered CL emission that is measured in our experiments is due to CH* and
CO2* on a large part and suffers from minimal reabsorption for atmospheric flames [46,47]. The
intensity of the signal detected on each pixel of the camera sensor is considered to be the integral
or sum of the emitted light (from CL or luminescence) that passes along a ray to that pixel. The
simplified RTE is approximated by discretising the reconstruction domain into a finite number of
cubic voxels, which are assumed to contain a uniform distribution of chemiluminescence. An
inversion process is applied to the measurement model to estimate the 3D distribution of CL in
the domain of interest, using information from all the 2D projections.
The direct 3D reconstructions were made with a domain containing 130 by 130 by 140 voxels
in the x-, y-, z-directions (the coordinates are shown in Fig. 3). The spatial resolution in the
reconstruction domain was 0.56 mm per voxel. All the instantaneous shots (about 200) for each
flame case were reconstructed, and the results were used to calculate the averaged reconstruction.
A spatial Gaussian filter of kernel size 3 by 3 by 3 voxels and standard deviation of σ = 0.7 was
applied to the reconstructed fields for smoothing. The filter size is below the expected spatial
resolution of the reconstructed field based on previous work using the same CTC method [16].

3. Results and discussion

The measurements and reconstructions confirm that the detectable vaporised salt luminescence
occurred on the burnt side of the flame. For example, in Fig. 5 the salt illumination from the inner
stream contains two distinct dark peaks that correspond to the shape illustrated by the flame front
image. Additionally, the superimposed reconstructed fields shown in Fig. 7 (that are discussed
later) illustrate that the salt illuminations appear from the downstream side of the flame front.
The outer stream images for Flame II were the noisiest, which was due to the increased camera
gain that had to be used. For a fixed amount of salt-seeding into the streams and camera settings,
the measured luminescence signals are higher for the richer flames. This effect was visible even
for the slightly higher global equivalence ratio of Flame I which exhibited higher signals. Since
a portion of the signal from the outer stream was cut by the available filters that were used to
avoid any contamination from thermally excited H2 O emission, the equivalence ratio in the outer
stream was chosen to be slightly higher than in the inner stream for Flame I. Furthermore, the
lower Reynolds number chosen for this flame favours less mixing between the streams and the
flame due to lower turbulence levels, and hence higher signals would be maintained due to less
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 250

dilution effect. As a consequence of the discussion given above, a lower camera exposure time
could be used for Flame I, given in Table 1, resulting in less motion blur.

Fig. 5. Exemplary instantaneous flame images obtained from one camera for each measured
channel. The signals are normalised to a 0 - 1 range.

Figure 6 presents horizontal slices at different heights above the burner, from an exemplary
instantaneous reconstruction for each flame. For Flame II, the outer stream was only visible
from around z/d = 2.4 (based on analysing the averaged distribution), as indicated in Fig. 5.
Therefore, the outer stream for Flame II was only reconstructed for the streamwise distance
of z/d ≥ 2.4. The horizontal slices demonstrate the distinct regions that are occupied by the
two streams, which are encapsulated by each other and by the flame front. The line artefacts
exhibited in the horizontal slices are not unexpected considering the number of cameras dedicated
to each measurement channel, and are in agreement with the trend seen in our previous CTC
reconstructions of a turbulent swirl-stabilised flame [22], where the effect of number of cameras
on the reconstruction was systematically studied. In one experiment for Flame I, all the 29
cameras were used to measure the flame front using a camera exposure time of texp = 200 µs,
and the resulting reconstructions are shown in Fig. 6. The horizontal and vertical slices for the
29-camera reconstruction show the expected improved quality in terms of reduced line artefacts
compared to those shown for the same horizontal slice from the 9-camera reconstruction, and
confirm the multiple flame fronts for Flame I. Also, since the 29-camera reconstruction used
images that were captured using half the exposure time, finer structures are more clearly visible
due to less motion blur.

Fig. 6. Horizontal slices from the instantaneous reconstructions of Flame I and Flame II,
at different heights z above the burner (d is the bluff body diameter). Also shown are
slices (horizontal and vertical) from the instantaneous flame front reconstruction of Flame I,
based on bandpass filtered CH* measurements using all the cameras with exposure time
texp = 200 µs.

Exemplary horizontal and vertical slices for all three reconstructed channels are superimposed
and presented in Fig. 7. For illustration purposes, the data for the inner and outer streams and the
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 251

flame front are colour coded in green, red and blue, respectively. The resulting superimposed
representation shows the global spatial distribution of each measured channel, and provides a
subtle illustration of the overlap regions between different channels. For better illustration of
the individual regions and where the different colour channels co-exist, i.e. mixing regions,
slices from the reconstructions are presented as contour plots in Fig. 8. The plots were made
by normalising the intensities of the individual channels by their maximum. The same green,
red and blue colour coding for the inner, outer and flame front channels is used in this figure.
Regions where mixing between the different colour channels exists are defined by the resulting
colour that appears from the combination of the primary colours used for the individual channels,
as illustrated by the key on the top of Fig. 8. For example, all areas where both the inner (green)
and outer (red) streams co-exist will have a yellow colour, and white areas indicate the presence
of signal from both the streams and the flame front. A distinctly lower level of overall mixing
between the flame front and both streams is exhibited for Flame I than for Flame II in the slices
presented in this figure. This is expected since for Flame II the flow Reynolds number is higher
(in both inner and outer streams). This leads to additional turbulence in the shear layer between
the slow co-flow and the outer stream. As the gradients between the co-flow and outer stream
flow are convected downstream, further fluctuations will be produced and this engulfs some
of the inner stream flow which also affects the region where the flame front exists. Hence,
mixing between all three channels at further downstream locations is promoted, and this is clearly
indicated by the large white segment in the averaged horizontal slice at z/d = 2.8 for Flame II. In
contrast, most of the mixing for Flame I is restricted to between neighbouring channels, indicated
by the cyan, magenta and yellow regions in Fig. 8.
The data shown in Fig. 8 provides a visualisation of the spatial distribution of the separate
reconstructed channels and their combinations in the same region at distinct horizontal and
vertical planes. To track the same information quantitatively in the streamwise direction, the
fraction of each segmented region, represented by the same seven colours used in Fig. 8, relative
to the total area of all segments for each slice was calculated and plotted as a function of height
above the burner in Fig. 9, for the averaged reconstructions. The fluctuations in the curves are
most likely due to noise, from imaging and the inversion process. Since the detected signals
for Flame I are the strongest closer to the burner exit, the reconstructions are less noisy in this
region and become increasingly noisy with height above the burner. Hence the curves for Flame I
are relatively smooth up to about z/d ≈ 2.7. With overall noisier images that were captured for
Flame II (due to lower signals), the reconstructions are also generally more noisy compared to
Flame I, and this in turn is reflected by the fluctuations in the curves of Fig. 9.
For both flames, nearest to the burner exit, the blue line representing the relative contribution
of the flame front signal has the largest peak as expected, and decreases thereafter as the
contribution from the luminescence in the inner and outer streams on the burnt side of the flame
and combinations of different mixed regions appear further downstream. Near the burner exit, at
z/d < 2.8, Flame I exhibits a large contribution from the outer stream signal (red) with a peak at
z/d ≈ 1.6 whilst the inner stream (green) is slimmer and hence shows a lower contribution but it
is already moderately mixed with its neighbours, these being the flame front and outer stream
signals as indicated by the cyan and yellow curves, respectively. The outer stream signal which
covers the largest segment in this height above the burner region for Flame I also shows mixing
with the flame front signal from the branch of the flame front that is directly neighbouring it, as
indicated by the magenta curve. In contrast, Flame II in the same upstream region, for z/d < 2.8,
exhibits contributions only from the inner stream and the flame front signals, and a relatively
large mixed region between them (cyan) which is almost double in size compared to Flame I.
Further downstream, for z/d > 2.8, for Flame I the inner stream signal (green) gradually
and continually increases due to its increasing thickness, whilst the outer stream signal (red)
decreases as the hot gas zone from the outer steam tapers off in the streamwise direction. Mixing
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 252

Figure 7

z/d = 1.6 z/d = 1.6

z/d = 0.4 z/d = 0.4

z/d = 2.8 z/d = 2.8

z/d = 0.8 z/d = 0.8

Fig. 7. Horizontal slices at different heights above the burner z/d and vertical slices at
the flame centreline x/d = 0 from the instantaneous and averaged reconstructions. The
approximate burner exit geometry is also shown. The flame front, inner and outer streams
are colour coded in blue, green and red, respectively, for illustration purposes.

Fig. 8. Plot of segmented regions from the instantaneous and averaged reconstructions of
Flame I and Flame II. Horizontal slices at different heights above the burner z/d, vertical
slices at the flame centreline x/d = 0, and the approximate burner exit geometry are shown.
The same colour coding for the inner and outer streams and flame front (green, red and blue,
respectively) is used.
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 253

Fig. 9. Curves showing the fraction of the area for each segment relative to the total area
of all segments at each height above the burner z/d, corresponding to the averaged data
presented for Flame I and Flame II in Fig. 8.

of the inner and outer stream signals (yellow) is maintained with height above the burner. The
outer stream signal for Flame II only begins in this downstream region and is immediately mixed
with the flame front signal, sharing a significant mixed region (magenta) for a distance of up to
z/d ≈ 5 which coincides with the distance that the flame front signal extends in the downstream
direction. At further heights above the burner the outer stream signal continues increasing and
shares a mixed region (yellow) with the inner stream signal of a similar proportion. The inner
stream signal contribution for Flame II continuously decreases with downstream distance as most
of it is either mixed with the flame front or outer stream signals. The white curve, which tracks
the segments where signal from the flame front and both streams co-exist, shows that the relative
contribution of the regions where all channels are mixed is generally higher for Flame II than for
Flame I, as discussed earlier.

4. Conclusions
Tomographic imaging using multi-simultaneous measurements (TIMes) for flame emission
reconstructions has been proposed and tested with the Cambridge-Sandia burner using two new
flame conditions that were chosen particularly for this project, to demonstrate the technique.
The diagnostic method allows for identification of the flame and the surrounding burnt gas
streams, based on seeding vaporised alkali metal salts into the streams of interest. Simultaneous
measurements of the resulting luminescence from the salts in the streams and the flame front
were used to reconstruct the individual 3D fields, which were also superimposed to unveil regions
of mixing and interaction.
The results allowed us to interpret the two flame cases studied in terms of the spatial extent
of each measured channel separately and together, within the entire 3D volume. Knowledge
of the emission properties for the chosen salts, combined with the correct choice of optical
filters and measurements of the flame emission was key to the successful implementation of
this technique. The implementation was not complicated and required relatively straightforward
modifications to our existing multi-camera setup for flame emission measurements. This novel
application of TIMes for flame emission reconstructions will allow for a better understanding of
many phenomena in complex multi-stream combustion processes, as it can distinguish the hot gas
regions originating from different streams. The excited species chemiluminescence measurements
can reveal the location of the flame front whilst the luminescence from the salt-seeded streams
identifies the burnt side of the flame from the unburnt side. Furthermore, the data can also serve
as a useful validation source for numerical simulations of complex flows and flames.

Funding
Ministerium für Kultur und Wissenschaft des Landes Nordrhein-Westfalen.
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 254

Acknowledgements
The authors gratefully acknowledge the financial support of the Ministerium für Innovation,
Wissenschaft und Forschung des Landes Nordrhein-Westfalen. We thank Andreas Kempf and
Christof Schulz at the IVG for sparking the first ideas that lead to this project and useful
discussions. We are also grateful to Simone Hochgreb for lending the Cambridge-Sandia burner
to us. We acknowledge support by the Open Access Publication Fund of the University of
Duisburg-Essen.

Disclosures
The authors declare that there are no conflicts of interest related to this article.

References
1. F. Stritzke, S. van der Kley, A. Feiling, A. Dreizler, and S. Wagner, “Ammonia concentration distribution measurements
in the exhaust of a heavy duty diesel engine based on limited data absorption tomography,” Opt. Express 25(7),
8180–8191 (2017).
2. J. Klinner and C. Willert, “Tomographic shadowgraphy for three-dimensional reconstruction of instantaneous spray
distributions,” Exp. Fluids 53(2), 531–543 (2012).
3. D. Bauer, H. Chaves, and C. Arcoumanis, “Measurements of void fraction distribution in cavitating pipe flow using
x-ray CT,” Meas. Sci. Technol. 23(5), 055302 (2012).
4. Y. Wu, W. Xu, Q. Lei, and L. Ma, “Single-shot volumetric laser induced fluorescence (vlif) measurements in turbulent
flows seeded with iodine,” Opt. Express 23(26), 33408–33418 (2015).
5. Q. Wang, T. Yu, H. Liu, J. Huang, and W. Cai, “Optimization of camera arrangement for volumetric tomography
with constrained optical access,” J. Opt. Soc. Am. B 37(4), 1231–1239 (2020).
6. T. Li, J. Pareja, F. Fuest, M. Schütte, Y. Zhou, A. Dreizler, and B. Böhm, “Tomographic imaging of OH laser-induced
fluorescence in laminar and turbulent jet flames,” Meas. Sci. Technol. 29(1), 015206 (2018).
7. L. Ma, Q. Lei, J. Ikeda, W. Xu, Y. Wu, and C. Carter, “Single-shot 3D flame diagnostic based on volumetric laser
induced fluorescence (VLIF),” Proc. Combust. Inst. 36(3), 4575–4583 (2017).
8. A. Unterberger, A. Kempf, and K. Mohri, “3D evolutionary reconstruction of scalar fields in the gas-phase,” Energies
12(11), 2075 (2019).
9. S. J. Grauer, A. Unterberger, A. Rittler, K. J. Daun, A. M. Kempf, and K. Mohri, “Instantaneous 3D flame imaging by
background-orientated schlieren tomography,” Combust. Flame 196, 284–299 (2018).
10. E. Boigné, N. R. Bennett, A. Wang, K. Mohri, and M. Ihme, “Simultaneous in-situ measurements of gas temperature
and pyrolysis of biomass smoldering via X-ray computed tomography,” Proc. Combust. Inst. (2020).
11. J. Kerl, C. Lawn, and F. Beyrau, “Three-dimensional flame displacement speed and flame front curvature measurements
using quad-plane PIV,” Combust. Flame 160(12), 2757–2769 (2013).
12. D. Ebi and N. T. Clemens, “Simultaneous high-speed 3D flame front detection and tomographic PIV,” Meas. Sci.
Technol. 27(3), 035303 (2016).
13. F. Wang, Z. Xie, J. Yan, and K. Cen, “Simultaneous measurement of three-dimensional particle temperature, particle
concentration, and H2O concentration distributions using multispectral flame images,” Combust. Sci. Technol.
189(11), 1891–1906 (2017).
14. T. Ren and M. F. Modest, “Reconstruction of three-dimensional temperature and concentration fields of a laminar
flame by machine learning,” in Proceedings of the 9th International Symposium on Radiative Transfer, RAD-19,
(SAE International, 2019).
15. N. Anikin, R. Suntz, and H. Bockhorn, “Tomographic reconstruction of the OH*-chemiluminescence distribution in
premixed and diffusion flames,” Appl. Phys. B 100(3), 675–694 (2010).
16. J. Floyd, P. Geipel, and A. M. Kempf, “Computed Tomography of Chemiluminescence (CTC): Instantaneous 3D
measurements and Phantom studies of a turbulent opposed jet flame,” Combust. Flame 158(2), 376–391 (2011).
17. J. Samarasinghe, S. Peluso, M. Szedlmayer, A. De Rosa, B. Quay, and D. Santavicca, “Three-dimensional
chemiluminescence imaging of unforced and forced swirl-stabilized flames in a lean premixed multi-nozzle can
combustor,” J. Eng. Gas Turbines Power 135(10), 101503 (2013).
18. N. A. Worth and J. R. Dawson, “Tomographic reconstruction of OH* chemiluminescence in two interacting turbulent
flames,” Meas. Sci. Technol. 24(2), 024013 (2013).
19. J. Weinkauff, J. Koeser, D. Michaelis, B. Peterson, A. Dreizler, and B. Böhm, “Volumetric flame measurements in a
lifted turbulent jet flame using tomographic reconstruction of chemiluminescence,” in 17th International Symposium
on Application of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, (2014), p. 11.
20. L. Ma, Q. Lei, Y. Wu, W. Xu, T. M. Ombrello, and C. D. Carter, “From ignition to stable combustion in a cavity
flameholder studied via 3D tomographic chemiluminescence at 20 kHz,” Combust. Flame 165, 1–10 (2016).
 Research Article Vol. 29, No. 1 / 4 January 2021 / Optics Express 255

21. B. D. Geraedts, C. M. Arndt, and A. M. Steinberg, “Rayleigh index fields in helically perturbed swirl-stabilized flames
using doubly phase conditioned OH* chemiluminescence tomography,” Flow, Turbul. Combust. 96(4), 1023–1038
(2016).
22. K. Mohri, S. Görs, J. Schöler, A. Rittler, T. Dreier, C. Schulz, and A. Kempf, “Instantaneous 3D imaging of highly
turbulent flames using computed tomography of chemiluminescence,” Appl. Opt. 56(26), 7385–7395 (2017).
23. Y. Jin, Y. Song, X. Qu, Z. Li, Y. Ji, and A. He, “Three-dimensional dynamic measurements of ch* and c2*
concentrations in flame using simultaneous chemiluminescence tomography,” Opt. Express 25(5), 4640–4654 (2017).
24. S. M. Wiseman, M. J. Brear, R. L. Gordon, and I. Marusic, “Measurements from flame chemiluminescence
tomography of forced laminar premixed propane flames,” Combust. Flame 183, 1–14 (2017).
25. J. Menser, A. Unterberger, A. Kempf, and K. Mohri, “Instantaneous 3d imaging of turbulent stratified methane/air
flames using computed tomography of chemiluminescence,” in 5th International Conference on Experimental Fluid
Mechanics, Munich, Germany, (2018), pp. 766–770.
26. C. Ruan, T. Yu, F. Chen, S. Wang, W. Cai, and X. Lu, “Experimental characterization of the spatiotemporal dynamics
of a turbulent flame in a gas turbine model combustor using computed tomography of chemiluminescence,” Energy
170, 744–751 (2019).
27. A. Unterberger, M. Röder, A. Giese, A. Al-Halbouni, A. Kempf, and K. Mohri, “3D instantaneous reconstruction of
turbulent industrial flames using Computed Tomography of Chemiluminescence (CTC),” J. Combust. 2018, 1–6
(2018).
28. G. E. Elsinga, F. Scarano, B. Wieneke, and B. W. van Oudheusden, “Tomographic particle image velocimetry,” Exp.
Fluids 41(6), 933–947 (2006).
29. F. Scarano, “Tomographic PIV: principles and practice,” Meas. Sci. Technol. 24(1), 012001 (2013).
30. J. Weinkauff, D. Michaelis, A. Dreizler, and B. Böhm, “Tomographic piv measurements in a turbulent lifted jet
flame,” Exp. Fluids 54(12), 1624 (2013).
31. L. Boyer, “Laser tomographic method for flame front movement studies,” Combust. Flame 39(3), 321–323 (1980).
32. L. Withrow and G. M. Rassweiler, “Studying engine combustion by physical methods a review,” J. Appl. Phys. 9(6),
362–372 (1938).
33. J. Reissing, J. M. Kech, K. Mayer, J. Gindele, H. Kubach, and U. Spicher, “Optical investigations of a gasoline direct
injection engine,” in SAE Technical Paper, (SAE International, 1999).
34. K. W. Beck, T. Heidenreich, S. Busch, U. Spicher, T. Gegg, and A. Kölmel, “Spectroscopic measurements in small
two-stroke si engines,” Tech. rep., SAE Technical Paper (2009).
35. P. R. Medwell, A. R. Masri, P. X. Pham, B. B. Dally, and G. J. Nathan, “Temperature imaging of turbulent dilute
spray flames using two-line atomic fluorescence,” Exp. Fluids 55(11), 1840 (2014).
36. M. Mosburger, V. Sick, and M. C. Drake, “Quantitative high-speed imaging of burned gas temperature and equivalence
ratio in internal combustion engines using alkali metal fluorescence,” Int. J. Engine Res. 15(3), 282–297 (2014).
37. A. G. Gaydon, The Spectroscopy of Flames (Chapman and Hall Ltd, 1974).
38. A. Kramida, Yu. Ralchenko, J. Reader, and NIST ASD Team, NIST Atomic Spectra Database (ver. 5.7.1), [Online].
Available: https://physics.nist.gov/asd [2017, April 9]. National Institute of Standards and Technology, Gaithersburg,
MD. (2019).
39. M. S. Sweeney, S. Hochgreb, M. J. Dunn, and R. S. Barlow, “The structure of turbulent stratified and premixed
methane/air flames I: Non-swirling flows,” Combust. Flame 159(9), 2896–2911 (2012).
40. R. Zhou, S. Balusamy, M. S. Sweeney, R. S. Barlow, and S. Hochgreb, “Flow field measurements of a series of
turbulent premixed and stratified methane/air flames,” Combust. Flame 160(10), 2017–2028 (2013).
41. J. Floyd and A. M. Kempf, “Computed Tomography of Chemiluminescence (CTC): High resolution and instantaneous
3D measurements of a matrix burner,” Proc. Combust. Inst. 33(1), 751–758 (2011).
42. ULN System for the New SGT5-8000H Gas Turbine: Design and High Pressure Rig Test Results, vol. Volume 3:
Combustion, Fuels and Emissions, Parts A and B of ASME Turbo Expo: Power for Land, Sea, and Air.
43. C. Peterson, W. Sowa, and G. S. Samuelsen, “Performance of a model rich burn-quick mix-lean burn combustor at
elevated temperature and pressure,” Tech. rep., NASA Technical Reports Server: NTRS (2002).
44. P. A. Tipler, Physics for scientists and engineers, fourth edition (W. H. Freeman and Company, New York, USA,
1999).
45. M. Lauer and T. Sattelmayer, “On the adequacy of chemiluminescence as a measure for heat release in turbulent
flames with mixture gradients,” J. Eng. Gas Turbines Power 132(6), 061502 (2010).
46. R. R. John and M. Summerfield, “Effect of turbulence on radiation intensity from propane-air flames,” J. Jet Propuls.
27(2), 169–175 (1957).
47. M. R. W. Lauer, “Determination of the heat release distribution in turbulent flames by chemiluminescence imaging,”
Ph.D. thesis, Technical University of Munich (2011).