Measurement of Soot Volume Fraction and Temperature for Oxygen-Enriched Ethylene Combustion Based on Flame Image Processing

A method for simultaneously visualizing the two-dimensional distributions of temperature and soot volume fraction in an ethylene flame was presented. A single-color charge-coupled device (CCD) camera was used to capture the flame image in the visible spectrum considering the broad-response spectrum of the R and G bands of the camera. The directional emissive power of the R and G bands were calibrated and used for measurement. Slightly increased temperatures and reduced soot concentration were predicted in the central flame without self-absorption effects considered, an iterative algorithm was used for eliminating the effect of self-absorption. Nine different cases were presented in the experiment to demonstrate the effects of fuel mass flow rate and oxygen concentration on temperature and soot concentration in three different atmospheres. For ethylene combustion in pure-air atmosphere, as the fuel mass flow rate increased, the maximum temperature slightly decreased, and the maximum soot volume fraction slightly increased. For oxygen fractions of 30%, 40%, and 50% combustion in O2/N2 oxygen-enhanced atmospheres, the maximum flame temperatures were 2276, 2451, and 2678 K, whereas combustion in O2/CO2 atmospheres were 1916, 2322, and 2535 K. The maximum soot volume fractions were 4.5, 7.0, and 9.5 ppm in oxygen-enriched O2/N2 atmosphere and 13.6, 15.3, and 14.8 ppm in oxygen-enriched O2/CO2 atmosphere. Compared with the O2/CO2 atmosphere, combustion in the oxygen-enriched O2/N2 atmosphere produced higher flame temperature and larger soot volume fraction. Preliminary results indicated that this technique is reliable and can be used for combustion diagnosis.


Introduction
Numerous thermal power plants in China use coal as fuel.When dealing with technologies related to coal combustion, particulate removal and CO 2 reduction have to be considered.A potential technique for CO 2 capture is oxy-combustion of coal.O 2 /CO 2 combustion is a process of burning coal in a mixture of oxygen and recycled flue gas, generating a CO 2 -concentrated sequestration-ready flue gas [1,2].Success in the implementation of O 2 /CO 2 combustion in coal-fired boilers depends on understanding the differences after the replacement of N 2 with CO 2 .The difference in the thermal properties of N 2 and CO 2 renders O 2 /CO 2 combustion significantly different from air combustion.Therefore, CO 2 participates in the chemical reaction and affects the combustion characteristics, altering the flame temperature, soot, and NO x formation [3][4][5][6].Knowledge of the temperatures and soot

Measurement Principle
The measurement of temperature and soot volume fraction is based on the radiative emission from a soot particle, and the line-of-sight radiative intensity is determined by soot, which emits and absorbs radiation: where I b,λ (l) = 2πhc 2 /λ 5 e hc/λkT(l) − 1 is the monochromatic blackbody radiative intensity, c is the speed of light, h is Planck's constant, k is the Boltzmann constant.κ λ is the absorption coefficient (m −1 ).κ λ (l)I b,λ (l) is defined as the spectral emission source term, and exp[− l f l κ λ (l )dl ] is the self-absorption term.A color CCD camera is used to obtain flame images in visible spectrum.The R, G, and B data of the flame images represent the relative radiation intensities of the flame.A blackbody furnace is used to calibrate the relationship between the absolute radiation intensity and the raw R, G, and B data of the image obtained by the camera [18], which will be described in a later section.With consideration of the spectral response of the camera, the absolute radiation intensity obtained by the camera from the flame can be expressed as follows: where i = (R, G, B), and η i,λ is the relative spectral response efficiency of the camera, and the method used to obtain the spectral characterization of the camera is described in a subsequent section.
The self-absorption effect is disregarded during the initial calculation; thus, Equation (3) can be written as: In the current study, Tikhonov regularization similar to that in the literature [15,22,23] is used to reconstruct H R (l) and H G (l).After H R (l) and H G (l) are obtained, the local flame temperatures and the soot absorption coefficients can be derived from the ratio of H R (l) toH G (l), as shown below: • 2πhc 2 /λ 5 e hc/λkT(l) − 1 dλ According to the Rayleigh approximation, the soot volume fraction with its spectral absorption coefficient is estimated as follows [24,25]: where E(m) is a function of the real and imaginary parts of the refractive index m, expressed as: E(m) is typically regarded as a constant independent of the wavelength [18,24,25] with a magnitude of 0.26 in the visible region.
Considering Equation ( 6) and l)dλ, the emission source terms H R (l) and H R (l) can be rewritten as: Thus, Equation ( 5) can be expressed as follows: Equation ( 9) contains only one unknown T(l), which can be solved using the Newton-type iterative algorithm.After the temperatures T(l) are obtained, the soot volume fractions f v (l) can be calculated from Equation (8), and the absorption coefficient κ λ (l) can be expressed using Equation (6).
Notably, κ λ (l) and T(l) are intermediate computational results; the self-absorption term exp − l f l κ λ (l )dl is disregarded.In the current study, the self-absorption effect is corrected using the method similar to that in Ref. [24] in multi-wavelength emission tomography for flame temperature Energies 2017, 10, 750 4 of 16 and soot volume fraction measurements.The self-absorption term exp − l f l κ λ (l )dl can be obtained after κ λ (l) is determined, and Equation (3) can be expressed as a matrix equation: where ∆l is the length of the path, H self is the self-absorption term, ∆l is the updated length of the path that the self-absorption term considers.The reconstructed emission source terms H R and H G can be updated using another instance of Tikhonov regularization, and the updated T(l) and κ λ (l) can be recalculated by solving Equations ( 6), (8), and ( 9).After several iterations, the absorption coefficient κ λ (l) and the temperature T(l) are finally calculated to reach convergence so that the self-absorption effect is considered.The whole solution procedure to simultaneously reconstruct 2-D distributions of temperature and soot volume fraction is shown in Figure 1.
In the current study, the self-absorption effect is corrected using the method similar to that in Ref. [24] in multi-wavelength emission tomography for flame temperature and soot volume fraction measurements.The self-absorption term can be obtained after ( ) λ κ l is determined, and Equation (3) can be expressed as a matrix equation: where Δl is the length of the path, self T l are finally calculated to reach convergence so that the self-absorption effect is considered.The whole solution procedure to simultaneously reconstruct 2-D distributions of temperature and soot volume fraction is shown in Figure 1.

Measurement Algorithms Validated by Simulation
To validate the performance of the proposed measurement method, some measurement simulations of a fictitious flame with a given temperature and a soot volume fraction were conducted.The temperature of flame and soot volume fraction was measurement by flame emission spectrum taken from Ref. [24].The relative spectral response of the CCD camera was simply expressed by a Gaussian function during simulation, and by using Equation (2), the absolute radiation intensity of the flames E R and E G could be obtained, as shown in Figure 2.
simulations of a fictitious flame with a given temperature and a soot volume fraction were conducted.The temperature of flame and soot volume fraction was measurement by flame emission spectrum taken from Ref. [24].The relative spectral response of the CCD camera was simply expressed by a Gaussian function during simulation, and by using Equation (2), the absolute radiation intensity of the flames R E and G E could be obtained, as shown in Figure 2.
(a) (b)   The reconstructed emission source terms H R and H G could be calculated using E R and E G , respectively, on the basis of Tikhonov regularization.The initial values of T(l) and f v (l) could be obtained in accordance with Equations ( 6), (8), and (9), and those intermediate results of T(l) and f v (l) could be used for iterative calculation later.The solution reached convergence as iterations increased, and the convergence value was the final result.In the current study, the convergence was evaluated using |∆ f v |, and the iterations were considered to be completed if |∆ f v | ≤ 0.001.These changes in the value of the soot volume fraction at a height of 30 mm during iterative calculation are shown in Figure 3.As shown in Figure 3, the calculation value of the soot volume fraction quickly approximates the real value after only one iteration, and the difference between the soot volume fraction calculated using the iterative algorithm and the soot volume fraction in the real-life scenario is not discernible after five iterations.Figure 4 shows the results of the comparison when self-absorption effects are considered and disregarded.
As shown in Figure 3, the calculation value of the soot volume fraction quickly approximates the real value after only one iteration, and the difference between the soot volume fraction calculated using the iterative algorithm and the soot volume fraction in the real-life scenario is not discernible after five iterations.Figure 4 shows the results of the comparison when self-absorption effects are considered and disregarded.As shown in Figure 4, both the temperature and the soot volume fraction can be reconstructed accurately when the self-absorption term is considered.The calculation errors of the temperature and the soot volume fraction increase gradually near the center of the flame when the self-absorption term is disregarded.Without the self-absorption term considered, the temperature decreases but the soot volume fraction increases in the central flame.At the flame height of 30 mm, the relative errors in the calculation of the temperature and the soot volume fraction were 3% and 23%, respectively.Thus, the reconstructed results were affected by the self-absorption term and must thus be considered.With the aforementioned results, the performance of the proposed method of flame measurement was proven.

Experimental Setup
The schematics for burners and experimental equipment is shown in Figure 5.The co-flow burner is similar to the one designed by the National Research Council Canada.The burner consists of a fuel tube with an inner diameter of 10.9 mm, centered in an oxidizer nozzle with an inner As shown in Figure 4, both the temperature and the soot volume fraction can be reconstructed accurately when the self-absorption term is considered.The calculation errors of the temperature and the soot volume fraction increase gradually near the center of the flame when the self-absorption term is disregarded.Without the self-absorption term considered, the temperature decreases but the soot volume fraction increases in the central flame.At the flame height of 30 mm, the relative errors in the calculation of the temperature and the soot volume fraction were 3% and 23%, respectively.Thus, the reconstructed results were affected by the self-absorption term and must thus be considered.With the aforementioned results, the performance of the proposed method of flame measurement was proven.

Experimental Setup
The schematics for burners and experimental equipment is shown in Figure 5.The co-flow burner is similar to the one designed by the National Research Council Canada.The burner consists of a fuel tube with an inner diameter of 10.9 mm, centered in an oxidizer nozzle with an inner diameter of 88 mm and an outer diameter of 100 mm.Electronic mass flow controllers (Type: Sevenstar CS230, Sevenstar Electronics Mass Flow Meter Branch, Beijing, China) with a long-term accuracy of ±1.0%S.P. (≥35% F.S.) governed the flow rates of all gases, and these gases are delivered at room temperature and atmospheric pressure (294 K, 1 atm).Axisymmetric laminar diffusion ethylene flames in pure air, O 2 /N 2 , and O 2 /CO 2 oxygen-enhanced atmospheres were generated by the burner.Figure 6

presents
Energies 2017, 10, 750 7 of 16 a photo of the experimental setup.Table 1 summarizes the flow rate of gases in the nine different cases.The adiabatic flame temperatures in Table 1 were calculated using CHEMKIN PRO (ANSYS, Inc., Canonsburg, PA, USA).
The visible flames were captured using a digital camera (Type: Manta G-504, Allied Vision Technologies, Stadtroda, Germany).The camera has a Sony ICX655 sensor (Sony Corporation, Tokyo, Japan) to receive R, G, and B data.The size of the sensor is 2/3 inch with about one million effective pixels (1226 × 1028).Flame images were saved in a 12-bit lossless compressed raw format.To improve the camera resolution, a special telecentric lens (Type: Computar TEC-55, Computar, Tokyo, Japan) with an effective aperture of 33.0 mm was used.The distance between the camera lens and the nozzle center of the burner was 14 cm.In this situation, nearly only the parallel incident light was able to enter the lens, and the barrel distortion of the image was relatively weak.The resolution of the flame image was 64 pixels/mm.To capture the whole flame, the camera was placed on the platform lift, 20 frame flame images were captured at each flame height, and the whole flame image could be obtained by image processing techniques.Figure 7 shows five flame images with different heights and a single composite image.
Energies 2017, 10, 750 7 of 17 diameter of 88 mm and an outer diameter of 100 mm.Electronic mass flow controllers (Type: Sevenstar CS230, Sevenstar Electronics Mass Flow Meter Branch, Beijing, China) with a long-term accuracy of ±1.0%S.P. (≥35% F.S.) governed the flow rates of all gases, and these gases are delivered at room temperature and atmospheric pressure (294 K, 1 atm).Axisymmetric laminar diffusion ethylene flames in pure air, O2/N2, and O2/CO2 oxygen-enhanced atmospheres were generated by the burner.Figure 6 presents a photo of the experimental setup.Table 1 summarizes the flow rate of gases in the nine different cases.The adiabatic flame temperatures in Table 1 were calculated using CHEMKIN PRO (ANSYS, Inc., Canonsburg, PA, USA).The visible flames were captured using a digital camera (Type: Manta G-504, Allied Vision Technologies, Stadtroda, Germany).The camera has a Sony ICX655 sensor (Sony Corporation, Tokyo, Japan) to receive R, G, and B data.The size of the sensor is 2/3 inch with about one million effective pixels (1226 × 1028).Flame images were saved in a 12-bit lossless compressed raw format.To improve the camera resolution, a special telecentric lens (Type: Computar TEC-55, Computar, Tokyo, Japan) with an effective aperture of 33.0 mm was used.The distance between the camera lens and the nozzle center of the burner was 14 cm.In this situation, nearly only the parallel incident light was able to enter the lens, and the barrel distortion of the image was relatively weak.The resolution of the flame image was 64 pixels/mm.To capture the whole flame, the camera was placed on the platform lift, 20 frame flame images were captured at each flame height, and the whole flame image could be obtained by image processing techniques.Figure 7 shows five flame images with different heights and a single composite image.

Calibrations of the Digital Color Camera
Consumer digital cameras are not specifically designed to be used as scientific detectors.To further characterize the cameras, calibration was required.In a previous study on image processing techniques, the R, G, and B data of a pixel in the color digital camera are not proportional to the monochromatic radiative intensities because of the broad spectral response spectra of the R, G, and B bands of the color camera [16][17][18][19].To characterize the relative spectral response of the camera, an experiment similar to Ref. [26] was performed.As shown in Figure 8a, the light source is a 1000 W xenon lamp (Type: OBB Tunable KiloArc, HORIBA Group, Kyoto, Japan), the monochromatic light is obtained with a monochromator, and the light from the monochromator is coupled through a diffuser for the light to be as uniform as possible.The monochromatic light was imaged onto the camera in the visible spectrum of 380-800 nm (negligibly sensitive to non-visible wavelengths for the camera used).The bandwidths of every wavelength are set to 2 nm, although this is not sufficiently narrow for an optical bandwidth.However, this width can sufficiently characterize the relative spectral response of the camera.To obtain sufficiently detailed relative spectral response sensitivity curves, 85 images were taken at each wavelength in steps of 5 nm. Figure 8b shows the typical monochromatic radiation image.Meanwhile, a detector (Trap M), in addition to the camera, whose relative spectral response curves was known, was used to capture monochromatic images controlled

Calibrations of the Digital Color Camera
Consumer digital cameras are not specifically designed to be used as scientific detectors.To further characterize the cameras, calibration was required.In a previous study on image processing techniques, the R, G, and B data of a pixel in the color digital camera are not proportional to the monochromatic radiative intensities because of the broad spectral response spectra of the R, G, and B bands of the color camera [16][17][18][19].To characterize the relative spectral response of the camera, an experiment similar to Ref. [26] was performed.As shown in Figure 8a, the light source is a 1000 W xenon lamp (Type: OBB Tunable KiloArc, HORIBA Group, Kyoto, Japan), the monochromatic light is obtained with a monochromator, and the light from the monochromator is coupled through a diffuser for the light to be as uniform as possible.The monochromatic light was imaged onto the camera in the visible spectrum of 380-800 nm (negligibly sensitive to non-visible wavelengths for the camera used).The bandwidths of every wavelength are set to 2 nm, although this is not sufficiently narrow for an optical bandwidth.However, this width can sufficiently characterize the relative spectral response of the camera.To obtain sufficiently detailed relative spectral response sensitivity curves, 85 images were taken at each wavelength in steps of 5 nm. Figure 8b shows the typical monochromatic radiation image.Meanwhile, a detector (Trap M), in addition to the camera, whose relative spectral response curves was known, was used to capture monochromatic images controlled by the computer.By using the method described in Ref. [26], the relative spectral response of the camera was obtained.Figure 9 shows the results for the camera (Type: Manta G-504) with lens (Type: Computer TEC-55, Computar, Tokyo, Japan).by the computer.By using the method described in Ref. [26], the relative spectral response of the camera was obtained.Figure 9 shows the results for the camera (Type: Manta G-504) with lens (Type: Computer TEC-55, Computar, Tokyo, Japan).by the computer.By using the method described in Ref. [26], the relative spectral response of the camera was obtained.Figure 9 shows the results for the camera (Type: Manta G-504) with lens (Type: Computer TEC-55, Computar, Tokyo, Japan).

Experimental Results and Analysis
The flame images of ethylene combustion in nine different cases are shown in Figure 12.Notably, these images represent the unsaturated flame images, which were used for reconstructing the flame temperature and soot volume fraction.The exposure times were as follows: case1 to case3, 65, 000 s μ ; case4 to case9, 28, 000 s μ , 12, 000 s μ , 8500 s μ , 60, 000 s μ , 15, 000 s μ , 8500 s μ , respectively.

Experimental Results and Analysis
The flame images of ethylene combustion in nine different cases are shown in Figure 12.Notably, these images represent the unsaturated flame images, which were used for reconstructing the flame temperature and soot volume fraction.The exposure times were as follows: case1 to case3, 65,000 µs; case4 to case9, 28,000 µs, 12,000 µs, 8500 µs, 60,000 µs, 15,000 µs, 8500 µs, respectively.

Pure Air Atmosphere
Case1 to case3 were flame combustion in different fuel flux in pure air atmosphere.The maximum temperatures and soot volume fraction from case1 to case9 are summarized in Table 2.The condition of case3 was similar to that in the literature [24,27].The C2H4 fuel flow rate was set to the smoke point, 194 mL/min (21 °C, 1 atm), and the air co-flow was set to 284 L/min.The maximum soot volume fraction of soot was about 7.8 ppm, and the maximum flame temperature was about 2058 K.The profiles of the temperature and soot volume fraction in the present study were in good agreement with Ref. [24,27], as shown in Figure 13.The detailed 2-D distribution of temperature and soot volume fraction for case1 is shown in Figure 14.The measurement results for case1 to case3 are shown in Figure 15.Theoretical calculations as those in Table 2 revealed that the adiabatic flame temperature of the three tested cases was equal to 2369.0 K; however, the actual flame temperature was determined by both the chemical reaction heat and the radiant heat.The flame length increased with fuel flow, and soot formation was enhanced because of a lack of oxygen, causing an increase in

Pure Air Atmosphere
Case1 to case3 were flame combustion in different fuel flux in pure air atmosphere.The maximum temperatures and soot volume fraction from case1 to case9 are summarized in Table 2.The condition of case3 was similar to that in the literature [24,27] with Ref. [24,27], as shown in Figure 13.The detailed 2-D distribution of temperature and soot volume fraction for case1 is shown in Figure 14.The measurement results for case1 to case3 are shown in Figure 15.Theoretical calculations as those in Table 2 revealed that the adiabatic flame temperature of the three tested cases was equal to 2369.0 K; however, the actual flame temperature was determined by both the chemical reaction heat and the radiant heat.The flame length increased with fuel flow, and soot formation was enhanced because of a lack of oxygen, causing an increase in radiant heat loss.Therefore, the maximum temperature of case1 was higher than that in case2 and case3.Table 2 also shows that the maximum soot volume fractions of case1 and case2 are less than that of case3 because case3 was a smoke point; the incomplete burning is enhanced as fuel flow increases owing to lack of oxygen.

O 2 /N 2 Oxygen-Enhanced Atmosphere
As shown in Figure 12, the luminous flame heights are depicted with increasing oxygen concentration in O 2 /N 2 oxygen-enhanced atmosphere.This trend agrees with the Roper model.According to Roper's approach [29], a jet diffusion flame ends on the symmetry axis at the point where Energies 2017, 10, 750 13 of 16 oxygen and fuel meet at the stoichiometric ratio, and the point declines along the centerline owing to increases in diffusive oxygen exchange with increasing oxygen concentration.
In Figure 16, the flame temperature contours and soot volume fractions as the oxygen fraction changes from 30% to 50% are plotted.As shown in Figure 16, as the oxygen mole fraction increases, the high-temperature region transitions from the relatively lower wings of the flame to the tip of the flame along the centerline.Meanwhile, the high concentration of the soot region is located in the wings of the flame.The use of the adiabatic flame temperature as the characteristic flame temperature was verified [28,30].The evidence for the correlation between the adiabatic flame temperatures and the measured maximum temperatures are illustrated in Table 2.As indicated in the table, both the maximum flame temperature and the adiabatic flame temperature increase with increasing oxygen.
Table 2 also suggests that the maximum soot volume fraction initially increases and then decreases with increasing oxygen, and the maximum soot volume fractions are 13.6, 15.3, and 14.8 ppm.This occurrence was attributed to the competition between early nucleation, surface growth, and oxidation [31,32].The increase in oxygen concentration leads to an increase in peak flame temperature.Such an increase promotes the pyrolysis of the fuel, resulting in the production of hydrocarbon radicals and H atoms, which enhance soot formation.Increased oxygen concentrations also increase the flame temperature, and soot surface growth rates are increased.However, the increased presence of oxygen promotes oxidative mechanisms because of the attack of soot particles by OH and O radicals, tending to diminish soot concentrations.Table 2 also suggests that the maximum soot volume fraction initially increases and then decreases with increasing oxygen, and the maximum soot volume fractions are 13.6, 15.3, and 14.8 ppm.This occurrence was attributed to the competition between early nucleation, surface growth, and oxidation [31,32].The increase in oxygen concentration leads to an increase in peak flame temperature.Such an increase promotes the pyrolysis of the fuel, resulting in the production of hydrocarbon radicals and H atoms, which enhance soot formation.Increased oxygen concentrations also increase the flame temperature, and soot surface growth rates are increased.However, the increased presence of oxygen promotes oxidative mechanisms because of the attack of soot particles by OH and O radicals, tending to diminish soot concentrations.

O2/CO2 Oxygen-Enhanced Atmosphere
Similar to ethylene combustion in O2/N2 oxygen-enriched atmosphere, the height of the flame decreased, along with increasing oxygen concentration.A comparison of case7 to case9 in Table 2   2 suggests that partial replacement of N 2 with the same concentration of CO 2 caused a slight reduction in flame height.
Figure 17 present the measured distributions of temperature and soot volume fraction for case7 to case9.Results in Figure 17 indicate that the high-temperature region transitioned from the wings of the flame to the tip of the flame along the centerline.This occurrence was similar to combustion in O 2 /N 2 Energies 2017, 10, 750 oxygen-enriched atmosphere.However, the maximum soot volume fraction steadily increased as the oxygen fraction increased, compared with the O 2 /N 2 oxygen-enriched atmosphere.Table 2 shows that the maximum temperatures are 1916, 2322, and 2535 K for case7, case8, and case9, respectively.The maximum temperatures are 2276, 2451, and 2678 K for case4, case5, and case6.The reason is that the specific heat capacity of CO 2 is larger than that of N 2 , which results in a significant decrease in flame temperature.
Compared with combustion in O 2 /N 2 oxygen-enriched atmosphere, the maximum soot volume fractions also decreased with the same mole fraction of O 2 .Table 2 reveals that the maximum soot volume fractions are 4.5, 7.0, and 9.5 ppm for case7, case8, and case9.Meanwhile, the maximum soot volume fractions are 13.6, 15.3, and 14.8 ppm for case4, case5, and case6.The decrease in peak temperature results from weakened fuel oxidation.The decrease in flame temperature causes a decrease in the fuel pyrolysis, which leads to a further decline in C 2 H 2 and H concentrations, which inhibit soot formation [32].According to [3,33,34] CO 2 addition exerts chemical effects, in addition to thermal effects, on the reduction in soot formation.The chemical mechanism of CO 2 addition might be to promote the concentrations of oxygen atom and hydroxyl in order to increase the oxidation of soot precursors in soot-forming regions.In a real-life scenario, these effects occur simultaneously and are intimately coupled.Numerical simulation results [3] suggest that the presence of CO 2 in the combustion environment can directly influence the chemical reduction of soot and PAH production tendencies.By promoting the main reaction CO 2 + H → CO + OH, CO 2 enhances the presence of OH radicals in the post-flame zone, which can be responsible for the oxidation of PAH and soot and most of the chemical destruction of their gaseous precursors.
Energies 2017, 10, 750 15 of 17 inhibit soot formation [32].According to [3,33,34] CO2 addition exerts chemical effects, in addition to thermal effects, on the reduction in soot formation.The chemical mechanism of CO2 addition might be to promote the concentrations of oxygen atom and hydroxyl in order to increase the oxidation of soot precursors in soot-forming regions.In a real-life scenario, these effects occur simultaneously and are intimately coupled.Numerical simulation results [3] suggest that the presence of CO2 in the combustion environment can directly influence the chemical reduction of soot and PAH production tendencies.By promoting the main reaction CO2 + H → CO + OH, CO2 enhances the presence of OH radicals in the post-flame zone, which can be responsible for the oxidation of PAH and soot and most of the chemical destruction of their gaseous precursors.

Conclusions
A method for simultaneously visualizing 2-D distributions of temperature and soot volume fraction in an ethylene soot flame was presented.The radiation visible flame images were used for the measurements.Compared with previous studies, the effect of self-absorption was specifically considered, and the distributions of 2-D temperature and soot volume fraction were obtained.High-resolution images of the flame (1 mm/65 pixels) were obtained using image processing techniques, and those images were used for measurements by solving the radiative transfer equation with iterative reconstruction.
This study evaluated the effects of fuel flow rate and oxygen concentration on the temperature and soot concentration in three different atmospheres.The experiment results indicates that for combustion in a pure air atmosphere, the maximum temperature slightly declined, and the maximum soot volume fraction slightly increased as the fuel mass flow rate increased.For the same oxygen fraction, combustion in O2/N2 atmosphere produced both a higher flame temperature and

Conclusions
A method for simultaneously visualizing 2-D distributions of temperature and soot volume fraction in an ethylene soot flame was presented.The radiation visible flame images were used for the measurements.Compared with previous studies, the effect of self-absorption was specifically considered, and the distributions of 2-D temperature and soot volume fraction were obtained.High-resolution images of the flame (1 mm/65 pixels) were obtained using image processing techniques, and those images were used for measurements by solving the radiative transfer equation with iterative reconstruction.
This study evaluated the effects of fuel flow rate and oxygen concentration on the temperature and soot concentration in three different atmospheres.The experiment results indicates that for combustion in a pure air atmosphere, the maximum temperature slightly declined, and the maximum soot volume fraction slightly increased as the fuel mass flow rate increased.For the same oxygen fraction, combustion in O 2 /N 2 atmosphere produced both a higher flame temperature and soot volume fraction compared with combustion in O 2 /CO 2 oxygen-enriched atmosphere.The results showed that this technique is reliable and can be used for combustion diagnosis.

H
is the self-absorption term, ′ Δ l is the updated length of the path that the self-absorption term considers.The reconstructed emission source terms R H and G H can be updated using another instance of Tikhonov regularization, and the updated ( ) T l and ( ) λ κ l can be recalculated by solving Equations (6), (8), and (9).After several iterations, the absorption coefficient ( ) λ κ l and the temperature ( )

Figure 1 .
Figure 1.Flow diagram of iterative reconstruction.Figure 1. Flow diagram of iterative reconstruction.

Figure 1 .
Figure 1.Flow diagram of iterative reconstruction.Figure 1. Flow diagram of iterative reconstruction.

Figure 2 ..
Figure 2. Measurement simulation of the absolute radiation intensity of the flame (a) at 30 mm high; and (b) at 50 mm high.

Figure 3 .
Figure 3. Updated trend value of the soot volume fraction at a height of 30 mm during iterations.

Figure 2 .
Figure 2. Measurement simulation of the absolute radiation intensity of the flame (a) at 30 mm high; and (b) at 50 mm high.

Figure 2 ..
Figure 2. Measurement simulation of the absolute radiation intensity of the flame (a) at 30 mm high; and (b) at 50 mm high.

Figure 3 .
Figure 3. Updated trend value of the soot volume fraction at a height of 30 mm during iterations.Figure 3. Updated trend value of the soot volume fraction at a height of 30 mm during iterations.

Figure 3 .
Figure 3. Updated trend value of the soot volume fraction at a height of 30 mm during iterations.Figure 3. Updated trend value of the soot volume fraction at a height of 30 mm during iterations.

Figure 4 .
Figure 4. Iteration results with (a) temperature at 30 mm; (b) soot volume fraction at 30 mm; (c) temperature at 50 mm; and (d) soot volume fraction at 50 mm.

Figure 4 .
Figure 4. Iteration results with (a) temperature at 30 mm; (b) soot volume fraction at 30 mm; (c) temperature at 50 mm; and (d) soot volume fraction at 50 mm.

Figure 5 .
Figure 5. Schematic of the experimental equipment and the burner.

Figure 6 .
Figure 6.Photo of the digital flame images acquisition system.

Figure 5 .
Figure 5. Schematic of the experimental equipment and the burner.

Figure 5 .
Figure 5. Schematic of the experimental equipment and the burner.

Figure 6 .
Figure 6.Photo of the digital flame images acquisition system.

Figure 6 .
Figure 6.Photo of the digital flame images acquisition system.

Figure 7 .
Figure 7. Five different parts of a flame (a,b) the composite image.

Figure 7 .
Figure 7. Five different parts of a flame (a,b) the composite image.

Figure 8 .
Figure 8.(a) Camera calibration of relative spectral response setup; and (b) Typical monochromatic radiation images.

Figure 9 .
Figure 9. Spectral response curves of the R, G, and B bands of the color digital camera (Type: Manta G-504).A blackbody furnace (Type: Mikron Model M330) with a temperature range of 300 °C to 1700 °C (temperature uncertainty 0.25% of reading ±1 °C) was used to calibrate the directional emissive powers and the raw data of the R and G bands of the camera, as shown in Figure 10.The variations of log( ) R E and log( ) G E

Figure 8 .
Figure 8.(a) Camera calibration of relative spectral response setup; and (b) Typical monochromatic radiation images.

Figure 8 .
Figure 8.(a) Camera calibration of relative spectral response setup; and (b) Typical monochromatic radiation images.

Figure 9 .
Figure 9. Spectral response curves of the R, G, and B bands of the color digital camera (Type: Manta G-504).A blackbody furnace (Type: Mikron Model M330) with a temperature range of 300 °C to 1700 °C (temperature uncertainty 0.25% of reading ±1 °C) was used to calibrate the directional emissive powers and the raw data of the R and G bands of the camera, as shown in Figure 10.The variations of log( ) R E and log( ) G E

Figure 9 . 17 Figure 10 .
Figure 9. Spectral response curves of the R, G, and B bands of the color digital camera (Type: Manta G-504).

Figure 10 .
Figure 10.Schematic of the camera blackbody calibration.

Figure 11 .
Figure 11.Variations of (a) log(E R ) and (b) log(E G ) with log(R/exposure time) and log(G/exposure time), respectively.

Figure 12 .
Figure 12.Color images of ethylene flames in pure air atmosphere, O 2 /N 2 , and O 2 /CO 2 oxygen-enhanced atmospheres.
. The C 2 H 4 fuel flow rate was set to the smoke point, 194 mL/min (21 • C, 1 atm), and the air co-flow was set to 284 L/min.The maximum soot volume fraction of soot was about 7.8 ppm, and the maximum flame temperature was about 2058 K.The profiles of the temperature and soot volume fraction in the present study were in good agreement Energies 2017, 10, 750 11 of 16

Figure 13 .
Figure 13.Comparison of temperature and soot volume fraction profiles at flame height z = 30 mm of case 3 with the results in Ref. [24,28].Figure 13.Comparison of temperature and soot volume fraction profiles at flame height z = 30 mm of case 3 with the results in Ref. [24,28].

Figure 13 .
Figure 13.Comparison of temperature and soot volume fraction profiles at flame height z = 30 mm of case 3 with the results in Ref. [24,28].Figure 13.Comparison of temperature and soot volume fraction profiles at flame height z = 30 mm of case 3 with the results in Ref. [24,28].

Figure 13 .
Figure 13.Comparison of temperature and soot volume fraction profiles at flame height z = 30 mm of case 3 with the results in Ref. [24,28].

Figure 15 .
Figure 15.2-D distributions of temperature T (K) and soot volume fraction (10 −6 ) for C2H4 combustion in pure air atmospheres.

Figure 15 .
Figure 15.2-D distributions of temperature T (K) and soot volume fraction (10 −6 ) for C 2 H 4 combustion in pure air atmospheres.

5. 3 .
O 2 /CO 2 Oxygen-Enhanced Atmosphere Similar to ethylene combustion in O 2 /N 2 oxygen-enriched atmosphere, the height of the flame decreased, along with increasing oxygen concentration.A comparison of case7 to case9 in Table
Table 1 summarizes the flow rate of gases in the nine different cases.The adiabatic flame temperatures in Table 1 were calculated using CHEMKIN PRO (ANSYS, Inc., Canonsburg, PA, USA).

Table 1 .
Experimental conditions.T ad represents the adiabatic flame temperature, calculated using CHEMKIN PRO.
Note: T represents the adiabatic flame temperature, calculated using CHEMKIN PRO.

Table 2 .
Maximum temperatures and soot volume fractions of the nine cases.

Table 2 .
Maximum temperatures and soot volume fractions of the nine cases.
f L (K) ad T max (K) T ( )