Studies of combustion characteristics of kerosene ethanol blends in an axi-symmetric combustor

Fuel 144 (2015) 205–213

Contents lists available at ScienceDirect

Fuel

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

Studies of combustion characteristics of kerosene ethanol blendsin an axi-symmetric combustor

http://dx.doi.org/10.1016/j.fuel.2014.12.0360016-2361/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +91 33 23355813; fax: +91 33 23357254.E-mail address: [email protected] (A. Datta).

Jitendra Patra a, Prakash Ghose a, Amitava Datta a,⇑, Mithun Das a, Ranjan Ganguly a, Swarnendu Sen b,Souvick Chatterjee b

a Department of Power Engineering, Jadavpur University, Salt Lake 2nd Campus, Kolkata 700 098, Indiab Department of Mechanical Engineering, Jadavpur University, Kolkata 700 032, India

h i g h l i g h t s

� Effects of ethanol–kerosene blends on characteristics of combustion analyzed experimentally.� Hollow cone pressure atomizer used for atomization.� High speed flame images analyzed for the brightness and flickering of the flames.� Temperature and select species concentrations compared to evaluate the performance.

a r t i c l e i n f o

Article history:Received 30 July 2014Received in revised form 14 October 2014Accepted 11 December 2014Available online 24 December 2014

Keywords:KeroseneEthanolFlame imagingFlicker frequencyCombustor performance

a b s t r a c t

Partial replacement of fossil fuels by biomass based alternative fuels in aviation applications is consid-ered as a possible option toward sustainability. The present work aims to study the changes in the flamecharacteristics and combustor performance in a cylindrical spray combustor when operated with kero-sene and kerosene–ethanol blends. A laboratory scale combustor with two concentric air entries alongwith a centrally fitted pressure swirl nozzle is used in the experiments. Three different fuels, viz. keroseneand two blends of ethanol (5% and 10% by volume) with kerosene have been used. Flame images are cap-tured using a high speed camera and analyzed for characterizing the flame behavior with different fuels.Temperatures of the combustor wall and exit gas have been measured by thermocouples while, the con-centrations of CO, CO2 and O2 in the exhaust gas are recorded using an online gas analyzer. The brightnessof the flame and combustor wall temperature are found to decrease considerably with 5% ethanol in theblend, although a further increase of ethanol (to 10%) decreases them marginally. However, the exit gastemperature reduces consistently with the addition of ethanol with kerosene. The standard deviation ofbrightness data is also much lower with the ethanol blended fuels. Flicker frequency of the flames, how-ever, remains almost unchanged with the blending of ethanol in kerosene. The concentrations of thethree species are found to be radially uniform at the exhaust indicating good mixing in the combustor.The CO concentration in the exhaust reduces with blending, indicating better combustion of the fuel.The CO2 concentration simultaneously decreases with ethanol blending representing less carbon loadingto the atmosphere. The reduction in carbon compounds at the combustor exit due to blending of ethanolis attributed to the lower C/H ratio of ethanol compared to kerosene.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Gas turbines are widely used as prime movers in transportationand power generation sectors. In the aviation sector, gas turbinesare operated with petroleum based liquid fuels, like gasoline, ker-osene or jet fuels, because of their high energy density, less storage

space requirement and thermal stabilities. Liquid fuel combustionin a gas turbine combustor is a complex phenomenon involving awide range of interacting and coupled physical and chemical pro-cesses, like fuel spray atomization and vaporization, two phase tur-bulent transport, radiation and finite rate chemical reaction. Thethermodynamic and physical properties of air and fuel, and alsothe characterizing parameters like Reynolds and Weber numbers,injection velocities, air–fuel ratio and swirl strength, influencethe combustion performance of liquid fuel spray in the combustor.

Nomenclature

C mass fractionD diameter of combustorF flicker frequencyG maximum gray valueM number of frequency components_m mass flow rate

N number of valuesp atmospheric pressurePk power density of the kth frequency componentR specific gas constantRc radius of the combustor

r radiusT temperatureV mean velocity of the gasX average brightness valuexi pixel value of ith pixelx1 pixel value of 1st pixelxN ith pixel value of Nth image�X brightness valueY volume fractionr standard deviationm kinematic viscosity of fuel

206 J. Patra et al. / Fuel 144 (2015) 205–213

Characteristics of the liquid fuel spray and the mixing of fuel andair inside the combustor play significant roles in controlling theoverall combustion performance.

Despite the continued advances in gas turbine combustiontechnology, the challenge in the design remains yet open toachieve further improvement in performance of the combustor.Engineers thrive to meet the desired classical requirements inthe combustor like high combustion efficiency, low emission, uni-form pattern factor, and increased combustor durability. In therecent years, emission of pollutants from hydrocarbon combustionhas been identified as a serious challenge toward environmentalsustainability [1]. The boom in the aviation sector has aggravatedthe situation and has also created a potential crisis of futurepetro-fuel resources. Research has been initiated recently towarddeveloping bio-derived renewable fuels for the aviation sector inorder to overcome the impending fuel crisis [2,3]. One importantissue in this regard is that the use of alternative fuels in gas turbinecombustor alters the combustion and emission characteristics [1].An improvement in the design methodology to meet the desiredrequirements calls for a physical understanding of the complexspray combustion process in a gas turbine combustion chamberwith new and alternative fuels.

Experimental studies in gas turbine can combustors werereported by many researchers in the last century, like Banhawyet al. [4], Jones et al. [5], Heitor et al. [6], Bicen et al. [7] andCameron et al. [8]. The characteristics of the liquid fuel spray,namely the initial mean drop diameter and spray cone angle arethe important input parameters affecting the combustion behaviorin a liquid fueled gas turbine combustor [4–8]. Lefebvre [9] con-cluded that the combustion efficiency in a gas turbine combustorshould increase with the finer initial droplets, as it increased theevaporation rate of the fuel. However, Odgers et al. [10] found thatsometimes the larger droplets help in improving the combustionefficiency due to their higher penetration and better mixing. Dattaand Som [11] found from a numerical study that an optimum sizeof the spray droplets is required to result the best combustionefficiency in a gas turbine combustor.

Flame stability in the combustor is another important consider-ation, particularly in the design of aero gas turbine engines. Burg-uburu et al. [12] investigated the effect of hydrogen enrichment ofkerosene in a spray combustion experiment in aero-gas turbinecombustor at constant power condition. Flame stability was foundto be strongly affected by injection of hydrogen and the lean blowout limit was found to be reduced. A small amount of hydrogeninjection was found to reduce the CO emission considerably dueto the enhancement of OH radicals. Yoon et al. [13] performedexperiments in a gas turbine combustor to examine the relation-ship between combustion instability and flame structure. It wasconcluded that at an unstable flame condition, the modulation in

combustor pressure changes the air and fuel mixture flow ratescausing a large variation in flame root size. Thus pressure fluctua-tion of combustion causes deformation of flame structure andvariation of flame has a strong effect on combustion instability.

The use of alternate liquid fuels in combustion has receivedattention in different sectors. One such liquid fuel is ethanol, whichcan also be used as blends with kerosene in the gas turbine com-bustor. It may be emphasized here that ethanol is blended withdiesel in countries like Brazil, USA Germany and France as vehicu-lar fuel [14]. Asfar and Hamed [15] studied the combustion perfor-mance of diesel–ethanol blends in a combustor. They found thatblending ethanol reduces the soot in flame and emission of pollu-tants like CO, CO2, NOx and HC from the combustor. However,increase of ethanol beyond 10% in the blend does not reduce thesoot and pollutants any further. Sequera et al. [16] studied emis-sions from flames of diesel, biodiesel, emulsified bio-oil, and die-sel–biodiesel blends in gas turbine combustor. They showed thatthe CO emissions from the blended fuel flames were comparablylower than those from diesel flames. Mendez et al. [17] used pureJet A fuel blended with ethanol by different volume fractions in agas turbine engine to study the performance and emission charac-teristics. They found lower operational thrust when ethanol isblended, but at the same time also observed lower emissions ofCO and NOx. Later on, they also reported similar observation byblending butanol with Jet-A in the same engine [18]. Khalil et al.[19] investigated distributed combustion using gaseous (methane,diluted methane, hydrogen enriched methane and propane) andliquid fuels, including both kerosene and ethanol with emphasison pollutants like CO and NO emissions and combustor perfor-mance with each fuel. They obtained low emissions in case ofthe alternative fuels. It demonstrates the outlines of combustorability for fuel flexibility without any modifications to the injec-tors, while maintaining high performance. Chiariello et al. [20]reported the experimental results concerning similar level of emis-sions (CO and NOx) from 30 kWe commercial micro gas turbineengine fed with fossil fuel and its blends with four differentstraight vegetable oils.

In addition to emission, it is also important to focus on issueslike characteristics and stability of the flame and the health ofthe combustor when alternative fuels are to be employed. Digitalimaging and image processing techniques are capable of measur-ing a range of physical parameters to ascertain the flame condi-tions, like brightness, flicker and temperature. These parametersprovide instantaneous and quantitative information about thephysical characteristics of the flame and consequently the perfor-mance of the combustion process. Yan et al. [21] presented thegeometrical and luminous characteristics of pulverized coalflame using optical measurements. Lu et al. [22] developed thekey aspects of the design, implementation, and experimental

Table 1Different test conditions of combustor experiment.

Primary:secondary air split 40:60Primary air vol flow rate (lpm) 608.34Primary air mass flow rate (kg/s) 0.012Primary air velocity (m/s) 6.15Secondary air vol flow rate (lpm) 912.51Secondary air mass flow rate (kg/s) 0.018Secondary air velocity (m/s) 3.03

J. Patra et al. / Fuel 144 (2015) 205–213 207

evaluation of a digital imaging based multifunctional instrumenta-tion system for the on-line monitoring of fossil fuel fired flames.Later, Sun et al. [23] applied a similar system for the stabilitymonitoring and characterization of combustion flames in a heavyoil burner. Cencerrado et al. [24] worked on a method for theanalysis of flame images based on two-dimensional distributionsand zonal analysis of significant flame features in a semi-industrialscale swirl burner of pulverized fuel. While the above methodologyof using digital image processing has shown promise incharacterizing flame stability, such method has not been, to thebest of our knowledge, tested kerosene ethanol blends in gasturbine applications.

In the present work, we have studied the flame characteristicsand combustor performance of kerosene and ethanol blends exper-imentally in a laboratory combustor. Ethanol is added in smallquantities (5% and 10% by volume) to the parent kerosene fuel toobserve the effects on flame brightness, standard deviation inbrightness and flickering. The variations in combustor wall tem-perature and exit pattern factor are measured for different fuelblends. The pattern of mixing in the combustor and the combus-tion performance are also judged by measuring the distributionof species like CO, CO2 and O2 at the exit plane of the combustor.

Total air volume flow rate (lpm) 1520.85Total air mass flow rate (kg/s) 0.030Fuel consumption (lph) 1.229Velocity of fuel jet (m/s) 6.955Air–fuel ratio (by volume) 110Primary Re (Vprimary ⁄ D/m) 38,760

Cases A B C

Kerosene (% by volume) 100% 95% 90%Fuel mass flow rate (kg/s) 2.765 � 10�4 2.762 � 10�4 2.758 � 10�4

Total heat release rate (kW) 12.775 12.531 12.287

2. Experimental

2.1. Set up and methodology

A test rig, as described in Fig. 1, has been fabricated for studyingthe spray combustion of liquid fuel in the combustor. It has a cylin-drical stainless steel combustor with two concentric air entries forthe primary and secondary air streams. The primary air enters

Fig. 1. Experimental set up of the combustor te

centrally through a 60� constant vane angle swirler to impart a tan-gential motion to the flow. The secondary air is admitted throughan annular passage around the primary swirler. This helps to coolthe combustor wall as well as bring down the temperature of thecombustion products. A set of wire mesh in the annular passagestraightens the secondary air flow. The air flow rates are measuredusing their respective orifice-meters and are listed in Table 1.

The liquid fuel is pumped through a nozzle into the combustorand the flow rate of the fuel is measured using a calibrated rotame-ter. Three different fuel cases are considered, viz., pure kerosene,kerosene blended with 5% ethanol and with 10% ethanol. A Lechlermake hollow cone pressure nozzle (No. 212.085.17.AC) is used in

st rig (a) schematic and (b) pictorial view.

208 J. Patra et al. / Fuel 144 (2015) 205–213

the experiments for injecting the fuel. The fuel is ignited by insert-ing a pilot flame through an ignition port placed at one side of thecombustor. The combustor is equipped with nine temperaturesensors (K-type thermocouple, wire diameter = 0.7 mm, beaddiameter = 1.7 mm, SS sheathed) on its outer wall to study the var-iation in wall temperature along its length. The wall-mounted tem-perature sensors are placed in their respective pots with thermalsilicon-paste (thermal conductivity = 2.3 W/m K) and positionedwithin 1 mm from the inner surface of the combustor. The distanceof the first thermocouple is 50 mm from the entrance plane and thesubsequent thermocouples are placed at 50 mm intervals. Atraversing thermocouple (K-type) placed near the exit of thecombustor is used to measure the radial variation of the exit gastemperature. All the temperature readings are recorded in digitaltemperature indicators fitted on a panel. The exit gas temperaturevalues are radiation corrected at the thermocouple tip consideringthe emissivity of stainless steel sheath. A view port of quartz glasson the side wall opposite to the ignition port is kept for imaging theflame.

Exit gas from the combustor is analyzed online using AVL DIGAS444 G gas analyzer, which measures the volume percentage of CO,CO2 and O2 present in the exhaust gas. The probe of the gas ana-lyzer is traversed on the exit plane to record the radial distributionof the measured species concentration.

2.2. High speed flame imaging

Luminosity is the most noticeable characteristic of a hydrocar-bon fuel flame. Visualisation techniques and image processingtechniques are used to visualise the flame zone and consequentlycharacterize the flame quantitatively. The characteristic parame-ters like brightness, standard deviation of brightness and flickerfrequency are derived directly from the flame images usingMATLAB software package. This information, together withcombustor data such as the fuel and air flow rates and emissions,can then be used to assess the quality of the flame.

The images of flame are taken by a BASLER make color highspeed camera (Model No. acA2000-340kc) having a full pixel reso-lution of 2046 � 1086 at a frame rate of 340 fps (frames per sec-ond). The flame images are taken through the viewing port of thecombustor. A telephoto lens and a macro lens (Navitar zoom7000, focal length = 18–108 mm, F2.5) are used to take closeimages of the flame through the port. The flame videos arerecorded for 3500 frames at 784 fps by changing the resolutionof the camera and then split into still images by an image process-ing software (X-CAP 3.7 by Epix Inc., USA). Instantaneous flameimages for three different cases taken from the combustor areshown in Fig. 2. The instantaneous images are than analyzed byMATLAB software for the predictions of different optical parame-ters of the flame.

Fig. 2. Instantaneous flame image taken for (a) C

Brightness: Given a group of pixel gray values (x1 . . . xi . . . xN) at apixel point in N different images, the normalized brightness valueð�XÞ at that point is defined [22] as the averaged pixel gray valuecalculated as percentage with respect to the maximum level.

�X ¼ 1N

XN

i¼1

xi �100

Gð1Þ

where N is the number of values and G is the maximum gray value(255 in this case). Brightness can be related to the size of the flame,and the radiation level from the flame on the chamber walls [25].

Standard deviation ðrÞ: Standard deviation in brightness is ameasure of the variability or dispersion in the brightness data ofthe flame. Following the previous notation, the standard deviationis given by the following equation [22]:

r ¼ 1N

XN

i¼1

ðxi � XÞ2 ð2Þ

where X is the average brightness value given as �X � G=100. The var-iability in the brightness value can be attributed to the turbulencein the combustion chamber.

Flicker frequency (F): Flicker frequency is defined as theweighted average frequency over the frequency range. To evaluatethe flicker frequency, the average pixel gray values of each frameare plotted over time. The data are converted from the timedomain to the frequency domain by applying fast Fourier transfor-mation. Flicker frequency is then calculated as the weighted aver-aged frequency over the measuring range, according to followingequation [24]:

F ¼PM�1

K¼0 Pkj jf kPM�1K¼0 Pkj j

ð3Þ

where Pk is the power density of the kth frequency component andM is the number of frequency components considered. We haveconsidered frequency range up to 180 Hz to calculate the flicker fre-quency in this present study.

Flicker is associated to the turbulent mixing and the multitudeof resulting eddies during the combustion process. Flicker fre-quency is considered as an indicator of flame stability. High flickerlevel means a more stable flame, while low flicker is associated topoor combustion efficiency [22].

3. Results and discussions

The instantaneous flame images for the three cases of fuel injec-tion (as given in Table 1), taken with high speed camera from afixed position, are shown in Fig. 2(a)–(c). In these figures the flamestructures and orientation indicate toward very high turbulence inthe combustor. The turbulence is created by the high inlet velocityof the air streams, swirl imparted to the primary air stream and

ase A, (b) Case B and (c) Case C of Table 1.

Table 2Flicker frequency and average percentage brightness values of three different cases.

Sl. no. Case Average percentageBrightness value (%)

Flicker frequency (Hz)

1 A 12.1 67.62 B 2.2 66.73 C 1.7 64.4

J. Patra et al. / Fuel 144 (2015) 205–213 209

also by the interaction between the primary and secondarystreams. It increases the rate of evaporation of the fuel and accel-erates the processes of transport in combustion. The keroseneflame in Fig. 2(a) is found to be extremely luminous due to thepresence of soot particles in the flame. The luminosity of the flamedecreases with the increasing content of ethanol in fuel. This canbe attributed to the presence of oxygen in ethanol, which helpsin decreasing soot in the flame. On the other hand, it is also empha-sized in Table 1 that the heat release rate and the resulting com-bustion intensity somewhat reduce with the replacement ofkerosene by ethanol; this is attributed to the lower heating valueof the latter fuel type.

Fig. 3(a1)–(c1) shows the spatial distribution of average bright-ness percentage ð�XÞ of flames in Cases A, B and C (of Table 1),respectively, evaluated using Eq. (1). The axis in the figures repre-sent the dimensions of the registered area of the images in terms ofpixels. The abscissa corresponds to the longitudinal dimensionwith the fuel injector located on the left side of the images while,the ordinate represents the radial dimension. In all the cases, anasymmetry in the flame with respect to the chamber axis has beenobserved. In case of 100%, 95% and 90% kerosene in the fuel, themaximum brightness values obtained are 37%, 9% and 7%, respec-tively while, the average brightness values evaluated from the fig-ures are 12.1%, 2.2% and 1.7% (Table 2). Brightness value of a flameimage is the measure of temperature and emissivity of the flame,since it represents the radiation spectrum emitted from the flameand received by the camera. The flame emissivity is stronglydependent on the concentration of soot in flame as the soot parti-cles are far more emissive compared to the participating gaseouscomponents, like carbon dioxide, water vapor, etc. The suddendecrease in brightness by the addition of 5% ethanol, as observedin Fig. 3(b1) in comparison to Fig. 3(a1), can be attributed to the

Fig. 3. Average brightness contours of (a1) Case A, (b1) Case B, and (c1) Case C; and correspTable 1 for the cases.

decrease of soot in flame due to the addition of oxygen in the fuel.Further 5% ethanol addition has reduced the brightness only a little(Fig. 3(c1)).

The standard deviation (r) in the brightness of flames with100% (Case A), 95% (Case B) and 90% (Case C) kerosene, respec-tively, are shown in Fig. 3(a2)–(c2). It is seen that the standard devi-ation for the kerosene flame (Fig. 3(a2)) is more than that for theblended fuels with ethanol (Fig. 3(b2) and (c2)). In case of 100%,95% and 90% kerosene the maximum values of standard deviationare 46, 20 and 16, respectively. The peak value of the standarddeviation in Fig. 3(a2) for the kerosene flame is observed to occurin the same region where the brightness peak of the correspondingflame (Fig. 3(a1)) is found. On the other hand, the peak of standarddeviation in Fig. 3(b2) and (c2) are found to occur at different loca-tions of the flame compared to that in Fig. 3(a2). Standard deviationindicates the degree of fluctuation of the brightness value of theflame from the mean brightness. These fluctuations can be attrib-uted to the turbulence structures present in the flow. Due to therandom fluctuations in the flow, the fuel droplets injected intothe combustor take random trajectories and the combustion frontalso becomes random. As a result, there will be a spatio-temporalvariation in the heat release in the combustor causing the variationin the brightness value of the flame. Additionally, with the random

onding standard deviation contours of (a2) Case A, (b2) Case B, and (c2) Case C. Refer

210 J. Patra et al. / Fuel 144 (2015) 205–213

fluctuations in the flow, the position of the soot laden zone alsofluctuates randomly. It is interesting to note that although the inletair flow rates and flow conditions are same for the three flamecases of Fig. 3(a2)–(c2), a significant difference in standard devia-tion is observed in their characteristics. It indicates that the fluctu-ation of heat release pattern decreases as the kerosene is replacedby ethanol in the fuel. The change in the region of peak standarddeviation further indicates that a variation in the spray patternstakes place in the three different cases.

In order to measure the flicker frequency of the flames, the tem-poral variation of the average pixel gray values of the individualframes are first plotted using the flame images captured. Fig. 4(a)shows such variation for the kerosene flame. A fast Fourier trans-form (FFT) of the temporal variation of the average pixel gray valueof the images converts the data from time domain to frequencydomain, showing power density at different frequency values(Fig. 4(b)). The power densities obtained from the FFT are foundto have many frequency peaks varying over a wide range. Thisclearly indicates that no regular flickering (with any dominant fre-quency and harmonics) of the flame exists. This also agrees withthe observed behavior of the flame, which is randomly movingwithin the combustor guided by the turbulence in the flow. Aweighted flicker frequency has been computed using Eq. (3) fromthe frequency domain fluctuation (as in Fig. 4(b)) for the threeflames and are listed in Table 2. The flicker frequency has beendescribed as a measure of stability of the flame [24]. Lu et al.[22] showed that a flame having higher flicker frequency is morestable in its character. It is found from the present results that

Fig. 4. Variation of (a) average pixel gray value with number of frames and (b) power dlegend, the reader is referred to the web version of this article.)

Fig. 5. (a) Image showing the vertical slicing of the flame and (b) quant

the addition of ethanol in the fuel decreases the flicker frequencyonly slightly indicating, that there is not much deviation in the sta-bility of the flame with the blending.

We have further computed the zone-wise flicker frequency ofthe flame by dividing the flame image into five equal regions ofinterest (ROI, see Fig. 5(a) for the kerosene flame as in Case A). Aweighted flicker frequency of each zone is calculated by takingthe data of that ROI from the different recorded images. Fig. 5(b)shows the flicker frequency values of each ROI for the keroseneflame. It is seen that the frequency of the different ROIs of theflame remain almost the same in the combustor. This result is dif-ferent from that of Lu et al. [22] who observed a decrease in flickerfrequency in the middle region of their flame compared to that inthe root region.

In order to analyze the overall performance of the combustorfurther experimental data comprising of combustor wall tempera-ture, exit gas temperature and selected species distributions arerecorded for kerosene and the blends of kerosene and ethanol asfuels. Fig. 6 shows the variation of wall temperature with axial dis-tance along the combustor. The temperature results in the figureindicate the mean values obtained from different runs (3–5 runsfor each case) and the data in the figure fall within an uncertaintyrange of ±5% at 95% confidence level. The wall temperature isfound to increase initially at a small axial distance (from the burnerplane) and then to reduce toward the exit. The wall temperature ishigher near the flame region due to increased radiation heat trans-fer from the flame. This is also corroborated by the higher bright-ness values recorded near the flame region (see Fig. 3). The

ensity with frequency. (For interpretation of the references to colour in this figure

itative flicker frequency in the vertical sections of image for Case A.

Fig. 6. Variation of wall temperature with axial distance along the combustor.

J. Patra et al. / Fuel 144 (2015) 205–213 211

subsequent reduction in the temperature along the wall takes placedue to heat transfer from the wall to the secondary air flowing alongthe wall. The wall temperature decreases considerably when 5%ethanol is blended with kerosene. The observed change in walltemperature is attributed to the fact that, increase in ethanolpercentage in the fuel blend leads to a decreases in the fuel heatingvalue, increase in fuel volatility and a decrease of the flameemissivity due to reduction in soot (as shown in Fig. 2). This alsoagrees with the observation made in case of brightness contouras in Fig. 3. However, increasing the share of ethanol to 10% inthe blend does not change the wall temperature any significantly.Such observation was also reported by Asfar and Hamed [15] intheir experiments with diesel–ethanol blend. The maximum walltemperature recorded for Cases A, B and C are 485, 429 and412 K, respectively. The decrease in the peak wall temperaturewith increased ethanol blending arises as a secondary advantageof using ethanol-blended fuel since it would result in a longer lifeof the combustor components.

Fig. 7 shows the variation of exit gas temperature with radialdistance in the combustor with three different fuel blends. Theuncertainty for the results lies within ±6.2% at 95% confidencelevel. The exhaust gas temperature in case of a gas turbine combus-tor is an indication of the energy content in the gas going to theturbine. Moreover, the temperature distribution in the gas givesthe pattern factor, which has important implication on the life ofthe gas turbine blades. In the present experiments, the tempera-ture is observed to be maximum at the center and constantlydecreasing toward wall due to the mixing of secondary air. The exitgas temperatures with ethanol blended fuels are less than thatwith kerosene fuel due to the lower heating value of ethanol andalso because of a smaller presence of radiating species in the flame.Similar observations were made by Asfar and Hamed [15] whosuggested that the optimum alcohol content in the blend is around10%.

Fig. 7. Variation of exit gas temperature with radial distance.

For further assessment of combustion performance and mixingwithin the combustor, the concentrations of selected species in theexhaust gas are measured for kerosene fuel and 5% ethanol blend.Fig. 8 shows the variation of (a) CO, (b) CO2 and (c) O2 concentra-tions (volume percentages) with the radial distance of the combus-tor exit. The uncertainty in the concentration measurements of allthe species lies within ±7.5%, with the highest uncertainty found inCO measurement. The percentages of CO, CO2 and O2 are found tobe almost constant due to good mixing induced by high turbulencewithin the combustor. Blending of ethanol in the fuel reduces theCO level in the exhaust gas (Fig. 8a). This may be attributed to morecomplete combustion owing to the presence of oxygen in the fuelitself. However, it is also observed that the concentration of CO2 isalso lower with the fuel blend (Fig. 8(b)). The C/H ratio in ethanol(0.33) is less than that in kerosene (0.51) [26]. Therefore, thereplacement of kerosene by ethanol in fuel results in the reductionof CO2 emission in the exhaust gas. In earlier work, Canakci et al.[27] studied the emission characteristics of S.I. engine with gaso-line fuel blended with ethanol. They also found substantial reduc-tion in the CO2 content in the exhaust gas when 5% ethanol isblended with gasoline and attributed the reduction to the reducedC/H ratio of ethanol. This clearly indicates that ethanol additionwith hydrocarbon fuel can reduce the carbon loading in the atmo-sphere due to combustion of liquid fuel. The variation of O2 con-centrations for the two cases in Fig. 8(c) shows almost uniformdistribution. The O2 percentage in the gas is rather high and notmuch different from that in air. This is because of the very highoperating air–fuel ratio in the experiments. The stoichiometricair–fuel ratio by mass for kerosene is about 15 while, the experi-mental air fuel ratio is 110.

Finally, in order to compare the effect of ethanol blending onthe specific CO2 emission index (EICO2 , which is the mass of CO2

emitted per unit energy released in the combustor), we have madea simple calculation based on the measured CO2 concentrationdata and assuming azimuthal symmetry in the field variables atthe combustor exit. The velocity profile at the combustor exit isassumed to be uniform as the flow is highly turbulent and the inletswirl is expected to die down due to the viscous action before theexit plane. Moreover, as the overall air–fuel ratio is very high(�110), the thermo-physical properties of the exhaust gas mixtureis assumed not to differ significantly from that of air. The meanvelocity (V) of the gas at the combustor exit is evaluated by balanc-ing the mass flow rate as,

_mgas ¼ _mair þ _mfuel ¼ VZ Rc

02prqðrÞdr ð4Þ

where Rc is the radius of the combustor. The density of the gas canbe related to its temperature as qðrÞ ¼ p=RTðrÞ following the idealgas law. The pressure (p) in the combustor is atmospheric, the spe-cific gas constant (R) is considered as that of air and the tempera-ture at different radii are obtained from the measured data points,shown in Fig. 7. Curve fits through computed density values at dif-ferent radii, as in Fig. 9(a) and (b) (with R2 value of 0.996 and 0.985for kerosene and kerosene–ethanol blend, respectively), show theradial variation of density for evaluating velocity using Eq. (4).

The radial variation of CO2 mass fraction ½CCO2 ðrÞ� in the exhaustgas is obtained from the measured radial variation of CO2 molefraction (i.e. volume fraction) ½YCO2 ðrÞ� as,

CCO2 ðrÞ ¼YCO2 ðrÞMWCO2

MWairð5Þ

where MWCO2 and MWair represent the molecular weights of CO2

and air, respectively. The distribution of YCO2 ðrÞ is again obtainedby curve fitting through experimental data points given in

Fig. 8. Radial variation of species concentrations in the exit gas (a) CO, (b) CO2 and (c) O2.

Fig. 9. Variation of exit gas density q(r) with radius for (a) 0% ethanol and (b) 5% ethanol blend.

212 J. Patra et al. / Fuel 144 (2015) 205–213

Fig. 10(a) and (b) (with R2 value of 0.987 and 0.984 for kerosene andkerosene–ethanol blend, respectively).

The mass flow rate of CO2, _mCO2 in the exhaust gas may be eval-uated as,

_mCO2 ¼ VZ R

02prqðrÞCCO2 ðrÞdr ð6Þ

Accounting that the amount of CO and unburned hydrocarbon(data measured, but not presented here) are only very little atthe combustor exit, we consider complete heat release due toburning of the fuels. The specific emission index of CO2 is thereforecalculated as,

Fig. 10. Variation of CO2 mass fraction CCO2 ðrÞ with ra

EICO2 ¼_mCO2

_mfuelHHVð7Þ

where _mfuel is the mass flow rate of fuel supplied to the combustorand HHV is the higher heating value. In case of blended fuel, theheating value of the blend is used in Eq. (7). The calculation inthe above line shows that the emission index ðEICO2 Þ with kerosenefuel is 0.038 g/kJ of CO2. On the other hand, with 5% ethanol blendedwith kerosene the value decreases to 0.03 g/kJ of CO2. This shows inabsolute term the reduction in CO2 emission that has occurred dueto blending of ethanol in the fuel. It is further to be mentioned thatwhen the ethanol is bio-derived, the CO2 produced due to its

dius for (a) 0% ethanol and (b) 5% ethanol blend.

J. Patra et al. / Fuel 144 (2015) 205–213 213

burning is not considered to cause any carbon loading to the atmo-sphere. This is due to the carbon–neutral attribute of biomass fuel.Therefore, the effective reduction in CO2 emission will be even morethan that obtained from the calculation shown above.

4. Conclusion

An experimental study has been made in a cylindrical combus-tor to observe the effects of blending ethanol (5% and 10%) withkerosene on flame characteristic and combustor performance. Highspeed images of the flame have been analyzed to find out varia-tions in brightness, standard deviation of brightness and flicker fre-quency of the flames. The combustor wall temperatures as well asthe variations of temperature and selected species concentrationsin the exit gas have been recorded.

It has been observed that blending 5% ethanol with kerosenereduces the flame luminosity and brightness value considerably.Further increase of ethanol to 10% decreases these characteristicsfurther by a small amount. The decrease in flame brightness dueto blending ethanol in fuel occurs due to the reduction of soot inflame. The standard deviation of brightness is also lower in theblended fuel flame compared to the flames of kerosene. However,flame flicker frequency remains almost unaltered in the threeflames, which indicates that the stability of the flame is notaffected by the fuel blending.

The peak wall temperature of the combustor decreases bynearly 11.5% and 4%, respectively, with 5% and 10% blending of eth-anol. This is due to the reduction of radiative heat flux from the lessluminous flame of blended fuel. The exit gas temperature with theblended fuel is less than that with kerosene because of the lowerheating value of ethanol. The concentration distributions of CO,CO2 and O2 at the combustor exit plane are found to be fairlyuniform. Combustion of the fuel is more complete with ethanolblending which is reflected in the lower CO concentration in theexhaust. Moreover, less CO2 in the exhaust with the ethanol blend-ing would favor reduced carbon loading in the environment.

Acknowledgments

This work is supported by a grant from the Gas TurbineResearch Establishment, DRDO, Govt. of India under the GATETScheme (Grant No. GTRE/GATET/CA07/1012/026/11/001). All theauthors are grateful for the financial support.

References

[1] Lefebvre AH, Ballal DR. Gas turbine combustion: alternative fuels andemissions. Taylor & Francis Group; 2010.

[2] Blakey S, Rye L, Wilson CW. Aviation gas turbine alternative fuels: a review.Proc Combust Inst 2011;33:2863–85.

[3] Kallio P, Paesztor A, Akhtar MK, Jones PP. Renewable jet fuels. Curr OpinBiotechnol 2014;26:50–5.

[4] Banhawy YE, Whitelaw JH. Experimental study of the interaction between afuel spray and surrounding combustion air. Combust Flame 1981;42:253–75.

[5] Jones WP, Toral H. Temperature and composition measurement in a researchgas-turbine combustion chamber. Combust Sci Technol 1983;31:249–75.

[6] Heitor MV, Whitelaw JH. Velocity, temperature and species characteristics ofthe flow in a gas turbine combustor. Combust Flame 1986;64:1–32.

[7] Bicen AF, Senda M, Whitelaw JH. Scalar characteristics of combusting flow in amodel annular combustor. ASME J Eng Gas Turb Power 1989;111:90–6.

[8] Cameron CD, Brouwer J, Wood CP, Samuelson GS. A detailed characterizationof the velocity and thermal fields in a model can combustor with wall jetinjection. ASME J Eng Gas Turb Power 1989;111:31–5.

[9] Lefebvre AH. Fuel effects on gas turbine combustion–ignition, stability andcombustion efficiency. J Eng Gas Turbines Power 1985;107(1):24–37.

[10] Odgers J, Kretschmer D, Pearce GF. The combustion of droplets within gasturbine combustors: some recent observations on combustion efficiency. J EngGas Turbines Power 1993;115(3):522–32.

[11] Datta A, Som SK. Combustion and emission characteristics in a gas turbinecombustor at different pressure and swirl conditions. Appl Therm Eng1999;19:949–67.

[12] Burguburu J, Cabot G, Renou B, Boukhalfa AM, Cazalens M. Effects of H2

enrichment on flame stability and pollutant emissions for a kerosene/airswirled flame with an aeronautical fuel injector. Proc Combust Inst2011;33:2927–35.

[13] Yoon J, Kim MK, Hwang J, Lee J, Yoon Y. Effect of fuel–air mixture velocity oncombustion instability of a model gas turbine combustor. Appl Therm Eng2013;54(1):92–101.

[14] Ronghou L, Nan L, Qingyu L, Shuzhi L, Anlin G. Alcohol and cotton oil asalternative fuels for internal combustion engines. In: Nan L, Best G, CarvalhoNeto CCde, editors. Integrated energy systems in China: the cold northeasternregion experience, Rome, Italy: FAO; 1994. p. 111–151.

[15] Asfar KR, Hamed H. Combustion of fuel blends. Energy Convers Manage1998;39(10):1081–93.

[16] Sequera D, Agrawal AK, Spear SK, Daly DT. Combustion performance of liquidbiofuels in a swirl-stabilized burner. J Eng Gas Turbines Power2008;130:032810–32818.

[17] Mendez CJ, Parthasarathy RN, Gollahalli SR. Performance and emissioncharacteristics of a small scale gas turbine engine fueled with ethanol/Jet Ablends. In: 50th Aerospace Sciences Meeting, AIAA, Nashville, Tennessee 2012;Paper No. AIAA 2012-0522.

[18] Mendez CJ, Parthasarathy RN, Gollahalli SR. Performance and emissioncharacteristics of butanol/Jet A blends in a gas turbine engine. Appl Energy2014;118:135–40.

[19] Khalil AEE, Gupta AK. Fuel flexible distributed combustion for efficient andclean gas turbine engines. Appl Energy 2013;109:267–74.

[20] Chiariello F, Allouis C, Reale F, Massoli P. Gaseous and particulate emissions ofa micro gas turbine fueled by straight vegetable oil–kerosene blends. ExpThermal Fluid Sci 2014;56:16–22.

[21] Yan Y, Lu G, Colechin M. Monitoring and characterization of pulverized coalflames using digital imaging techniques. Fuel 2002;81:647–56.

[22] Lu G, Yan Y, Colechin M. A digital imaging based multifunctional flamemonitoring system. IEEE Trans Instrum Meas 2004;53(4).

[23] Sun D, Lu G, Zhou H, Yan Y. Flame stability monitoring and characterizationthrough digital imaging and spectral analysis. Meas Sci Technol2011;22:114007–15.

[24] Cencerrado AG, Pena B, Gil A. Coal flame characterization by means of digitalimage processing in a semi-industrial scale PF swirl burner. Appl Energy2012;94:375–84.

[25] Huang Y, Yan Y, Lu G, Reed A. On-line flicker measurement of gaseous flamesby image processing and spectral analysis. Meas Sci Technol 1999;10:726–33.

[26] Wen Z, Yun S, Thomson MJ, Lightstone MF. Modeling soot formation inturbulent kerosene/air jet diffusion flame. Combust Flame 2003;135:323–40.

[27] Canakci M, Ozsezen AN, Alptekin E, Eyidogan M. Impact of alcohol–gasolinefuel blends on the exhaust emission of an SI engine. Renew Energy2013;52:111–7.