Biodiesel Airblast Atomization Optimization for Reducing Pollutant Emission in Small Scale Gas Turbine Engines

Transcription

1 ILASS Americas, th Annual Cnference n Liquid Atmizatin and Spray Systems, Chicag, IL, May 7 Bidiesel Airblast Atmizatin Optimizatin fr Reducing Pllutant Emissin in Small Scale Gas Turbine Engines C. D. Blsz and V. G. McDnell UCI Cmbustin Labratry University f Califrnia, Irvine, CA USA Abstract The present study assesses mdificatins t the fuel preparatin prcess (and subsequent emissins impact) in a cmmercial diesel fired gas turbine system fr peratin n sy bidiesel fuel (B99). The influence f liquid prperties and injectr peratin n the resulting drplet sizes and evapratin characteristics are assessed. A theretical analysis f plain-jet airblast atmizatin and subsequent spray prperties at atmspheric and engine cnditins is carried ut fr tw different plain jet airblast atmizer cnfiguratins. An adaptatin t an existing empirical Sauter mean diameter is presented which accunts fr the differing fuel prperties fund in B99 cmpared t DF. A theretical analysis f the influence f the fuel injectr peratin and liquid prperties n the fuel evapratin characteristics is als carried ut. In rder t assess the verall NOx emissins perfrmance attained, cmparisns are made with well stirred reactr results. The theretical and experimental findings indicate that, while it is pssible t imprve NOx emissins fr B99 thrugh ptimizatin f the fuel injectin prcess, ther factrs in additin t atmizatin cntribute t the generally higher NOx emissins bserved fr B99. Finally, discussin f an alternative apprach t imprving perfrmance is presented in which blends f ethanl and B99 are used t achieve fuel prperties that result in further ptimized fuel preparatin and therefre ptentially further imprved emissins.

2 ILASS Americas, th Annual Cnference n Liquid Atmizatin and Spray Systems, Chicag, IL, May 7 Intrductin Airblast atmizatin has been an attractive strategy fr liquid fuel breakup in gas turbines due t the large quantity f air available during engine peratin [1]. Here, the atmizer design aims t maximize the interface between the air and liquid, and the large vlume f high velcity air flw dictates the prductin f a well distributed spray pattern (creates a temperature prfile fairly insensitive t changes in fuel flw) which als prevents the depsit f liquid n the surrunding surfaces []. The large quantity f air als serves as an emissins reductin strategy f leaning ut the stichimetry in the cmbustr, resulting in lw st particulate, carbn mnxide (CO), xides f nitrgen (NOx) and radiative lsses due t lw flame luminsity []. When cnsidering the variables in cntinuus cmbustin systems fr pwer generatin, usually at high flw velcities and high pressures, a plain-jet airblast atmizer type is a viable ptin fr atmizing liquid fuels and is used fr the research herein. Alternative liquid fuels, specifically bidiesel and ethanl have an attractive ptential t help satisfy sciety s grwing appetite fr energy while striving t cnstrain carbn in the atmsphere. Due t cnstraints f available farm land, ther types f bi-derivative fuels like yellw grease bidiesel r cellulsic bidiesel and ethyl alchl derived frm waste material, can als be candidates in the attainment f energy resurces in the nt t distant future [3,4]. Questins regarding the rle f alternative liquid fuels n criteria pllutants, specifically NOx, have yet t be addressed and are in need f mre experimental data t prvide further insight int the cmbustin characteristics f these fuels [5,6,7]. Nmenclature ALR air t liquid mass rati D 3 Sauter mean diameter density surface tensin dynamic viscsity d fuel nzzle diameter Re Reynlds number We Weber number Oh Ohnesrge number L latent heat f vaprizatin LHV lwer heating value N-R nn recessed FBN fuel bund nitrgen Subscripts A air L liquid Objective and Apprach In the present study, a cmmercial gas turbine engine, designed fr lw emissins peratin n # diesel fuel (DF), is utilized t evaluate relative perfrmance with sy bidiesel with less than 1 percent diesel fuel, designated (B99) and pure ethanl (E1). The atmizatin and evapratin characteristics are assessed and related t the subsequent emissin behavir and varius engine lads. Experiment The gas turbine used is a cmmercial liquid fired engine designed fr peratin n DF (Capstne Mdel C3). This gas turbine is primarily used fr backup and remte pwer generatin. Past effrts by the authrs have assessed atmizatin quality at atmspheric and engine perating cnditins and quantified its impact n engine perfrmance [8,9]. Als the use f alternative liquid fuels (sy bidiesel and ethanl) in this engine has been explred [1,11]. Spray size measurements [Sauter Mean Diameter (SMD r D 3 )] were carried ut using a Malvern RTS laser diffractin system. Alignment and calibratin f the instrument was checked using a phtmask reticle. Measurements were carried ut n an atmspheric spray rig. Details f the experimental setup and test prcedure have been prvided elsewhere [1]. In this study, the fuel injectr cnsidered is a plain-jet airblast atmizer as shwn in Figure 1. It is nted that it has very similar characteristics t the injectr used extensively by Rizk and Lefebvre in their develpment f drp size crrelatins fr plain-jet airblast atmizatin [1]. One distinctin between that injectr and the ne used in the present study is the degree t which the n-axis liquid fuel tube is inserted relative t the c-flwing atmizatin air. As shwn in Figure 1, a nn-recessed (N-R) liquid tube is illustrated. In the present wrk, a recessed versin was als utilized whereby the liquid injectin tube was pulled back slightly int the atmizing air plenum. The prperties f the liquids in this study are shwn in Table 1. All prperties except L were determined by measurement r frm cmpsitinal analysis using Mass Spectrscpy Gas Chrmatgraphy. L was determined frm fuels in the literature [13,14,15]. Table 1. Liquid fuels f interest and their prperties in current investigatin (labratry ambient cnditins). B99 E1 DF Mlecular Mass g/ml kg/m-s kg/s kg/m L J/kg LHV MJ/kg

3 Mixing Air Fuel Swirler Air Atmizing Air L premixer Figure 1. Fuel Injectr/Premixer Assembly. [9] Results Atmizatin Atmizer Perfrmance n Baseline Fuel Using the same argument set frth by Rizk and Lefebvre, the mechanisms behind the airblast atmizatin are primarily mmentum and viscus frces. In previus wrk, the riginal nn-recessed (N-R) plain-jet airblast atmizer prduces excellent agreement with the empirical equatin develped by Rizk and Lefebvre withut mdificatin t the equatins empirical cefficients r expnents [8]. This equatin is shwn belw, As mentined, a secnd ( recessed ) versin f the injectr was als used. Specifically, the fuel tube was recessed by abut.5 mm upstream f the airblast air passage exit plane. This small change results in a decreased D 3 fr a given air t liquid mass flw rati (ALR). This can be nted by measured drp sizes falling belw the predicted values f Equatin (1). A cmparisn f the measured spray size fr the tw nzzles is shwn in Figure 3. The measured D 3 fr a given ALR fr the N-R airblast atmizer is cmpared with the calculated drp size using Equatin (1). Nte that the divergence between the equatin and the measured data at lwer ALR s, shws a bundary where fully develped airblast atmizatin is nt achieved. The recessed plain-jet airblast atmizer was able t achieve even lwer values f D 3 fr a given ALR (slid circles) fr the same range f fuel flw rates as the N-R versin. The recessing f the fuel tube has shwn t decrease the drp size by 5 micrns at the higher ALR s (great atmizatin quality). The adjustment t the cefficients t establish agreement fr the recessed plain-jet airblast atmizer will be discussed shrtly. A divergence between a mdified Equatin (1) and the measured data still exists and is even mre prnunced by shwing a larger disagreement fr a given ALR and persists fr larger values f ALR (appraching ). 1 D d 3 8 AU Rd 1 1 ALR L.15 Ld ALR (1) The equatin has been validated at atmspheric cnditins using lw sulfur diesel fuel # (DF). The measured SMD (D 3 ) was cmpared with measurements by Rizk and Lefebvre and the predicted values using Equatin (1) shwed excellent agreement as shwn in Figure. It is nted that the scatter is greater at higher D 3 (prer atmizatin). Calculated SMD (m) Lw Viscsity Liquids DF (N-R) Measured SMD (m) Figure. Cmparisn f Current Measurements with Prir Data [1].and Predictins by Equatin (1). SMD (m) 8 6 DF Calculated SMD (N-R) DF Laser Diffractin SMD (N-R) DF Laser Diffractin SMD DF Calculated SMD ALR Figure 3. SMD vs. ALR fr recessed and N-R versin f the Airblast atmizer nzzle fueled with DF. Atmizer Perfrmance n Alternative Fuels Measurements f spray sizes prduced by the injectr perated n DF, B99 and E1 at engine flw cnditins were btained. Each fuel was tested using the same range f fuel and air mass flw in the recessed plain-jet airblast atmizer. The airflw (and therefre ALR) spanned the range f the engine ALR values. The fuel flw was fixed at the fuel flw crrespnding t the maximum pwer utput fr the gas turbine engine perated n DF. Drp size data was cllected and cmpared t Equatin (1). As shwn in Figure 4, a significant discrepancy ccurred between the predicted (Equatin (1)) and the measured data.

4 SMD (m) B99 Laser Diffractin SMD B99 Calculated SMD E1 Laser Diffractin SMD E1 Calculated SMD DF Laser Diffractin SMD DF Calculated SMD SMD (m secnd term in Equatin (1) (the viscus term) which is a functin f the Ohnesrge number and the ALR, was reduced by 47% (frm.15 t.8) while keeping the first term cnstant E1 Laser Diffractin SMD E1 Calculated SMD DF Laser Diffractin SMD DF Calculated SMD ALR Figure 4. SMD (measured and Equatin (1)) vs. ALR fr DF, B99 and E1 fr recessed Airblast atmizer. The discrepancy is largest fr B99, (5-7 micrn ver predictin) at well develped atmizing cnditins (ALR s > ). A critical ALR value seems t exist fr each fuel which is based n tw bservatins; (1) the substantial variatin in measured values (nted by the errr bars) at lwer ALRs and () by visual bservatin f the spray plume behavir with ALR. These transitinal ALRs were dcumented fr all f the three fuels tested and was discvered t signify a distinctin between atmizatin behavirs, as will be discussed n the fllwing page. The discrepancy between predicted and measured SMD fr DF and E1 is much smaller ( - 3 micrn ver predictin fr develped atmizatin). The larger ver predictin fr B99 is nt surprising since Equatin (1) was riginally derived t agree with what were cnsidered lw viscsity fuels (viscsities belw.3 kg/(ms)) [1]. Figure 4 als illustrates the pint made previusly regarding the imprvement in atmizatin achieved by recessing the fuel tube (nte the measured SMD values fall belw the SMDs predicted by Equatin (1)). It is interesting t nte that this mdificatin t spray hardware wrked t enhance the prductin f small drplets at higher ALRs, yet has resulted in prer atmizatin quality at sub ptimal ALRs. T imprve the agreement between the equatin and experiment, a statistical analysis was cnducted t establish ptimized cefficients and expnents that culd better predict the measured sizes fr the recessed injectr in the regin f well develped atmizatin. Mdifying the expnents did nt result in an verall better fit f the data, hwever changing the empirical cefficients in frnt f the tw terms in Equatin (1) demnstrated a shift the calculated values fr D 3 sufficiently t match the measured data (t within a micrn) fr all three fuels. A distinctin is als made here between the lw and high viscsity liquids. Fr E1 and DF (lw viscsity) the cefficient in frnt f the ALR Figure 5. SMD (Measured and Equatin ()) vs. ALR with mdified cefficients fr the Rizk & Lefebvre equatin fr DF and E1. The resultant mdificatin is shwn in Equatin (): D d 3 8 AU Rd 1 1 ALR L.8 Ld ALR () The imprved match fr E1 and DF is demnstrated in Figure 5. The results f the prir ptimizatin prcedure implied a need t mdify (increase) the cntributin f the first term fr bth these lw viscsity fuels. Fr the best agreement with B99, the cefficients in frnt f bth terms were mdified. Mdificatins t Equatin (1) t achieve gd agreement fr B99 included decreasing the value f the cefficient f the first term (a 71% decrease frm 8 t.14 fr the mmentum dminated term) while increasing the value f the secnd term (a 133% increase frm.15 t.35 fr the viscus term). D d 3.14 AU Rd 1 1 ALR L.35 Ld ALR (3) Very gd agreement is fund fr high ALRs (Figure 6), with predicted and measured SMD values diverging earlier and mre substantially in cmparisn t E1 and DF at mid t lwer ALR range tested. Earlier wrk explred a brute frce methd t imprve the predictin f spray (D 3 ) at different engine lads and flw rates (a very useful predicative tl frm an engineering standpint) [1]. This methd prvides a fairly gd agreement (D 3 < 8 micrns fr entire range f interest), yet it is nt funded n a fundamental physical arguments; merely an experimental curve fit. This result is shwn in Equatin (4):

5 SMD (m) Oh L D d L AU Rd ALR Ld.1 7 ALR B99 Laser Diffractin SMD B99 Calculated SMD ALR (4) Figure 6. SMD vs. ALR with mdified cefficients fr the Rizk & Lefebvre equatin fr B99. Delving further int the classificatin f the nzzle behavir, the aerdynamic Weber number (We Aer ), liquid Reynlds number (Re L ), and the liquid Ohnesrge number (Oh L ) were calculated frm experimental data t quantify the type f atmizatin respnsible fr drplet frmatin. T establish the liquid mechanisms respnsible fr the breakup f the fuel stream, the cnditins are pltted n an Ohnesrge Chart (Oh L vs. Re L ), Figure 7, t prvide further insight. DF data frm the N-R airblast is als included Sinuus, Wave Breakup Rayleigh, Varicse Breakup Wave-Like Breakup With Air Frictin B99 E1 DF DF (N-R) Secndary Atmizatin Re L Figure 7. Ohnesrge chart (Oh L vs. Re L ) fr breakup f B99, DF (bth recessed and N-R atmizers) and E1 liquid fuels [16]. Based n the assessment f the viscus frces, Figure 7 reveals that all three fuels and bth airblast atmizers (fr DF) maintain very similar behavir by their psitin very clse t the bundary between sinuus wave breakup and wave-like breakup with air frictin. Overall, it is expected that there shuld be n difference in the viscus breakup between the three fuels. Hwever B99 and DF N-R, had the farthest deviatin away frm the bundary (which is defined nt strictly. by near the bundary lines in Figure 7) and may demnstrate a liquid tube wiggling in a time varying sinusidal manner t causes liquid breakup, whereas DF and E1 may experience a slight enhancement due t interactin with the air frictin. The recessed airblast atmizer shws t have better liquid atmizatin perfrmance than the N-R versin. When analyzing the rle f the atmizatin air in the plain-jet airblast nzzle, the aerdynamic frces were quantified in rder t evaluate its breakup effect n the fuel stream. When pltting the available data fr all three fuels and bth atmizers, a significant develpment was achieved which explains the regin f deviatin with the calculated and measured data. This develpment is explained by cmparing the fuel s Re L t the atmizing air s We Aer, n a diagram prpsed by Lasheras and Hpfinger [17] in Figure 8. 1 Re L Shear Breakup Risk & Lefebvre Eq. Hlds Membrane Breakup B99 E1 DF DF (N-R) Fiber Type and Recirculating Gas Cavity We Aer Figure 8. We Aer vs. Re L fr breakup f B99, DF (bth atmizers) and E1 liquid fuels. In Figure 8, the data fr each fuel (and atmizer) crsses a bundary which separates the membrane frm fiber type breakup. The divisin islates data pints which clsely crrelate with the Rizk and Lefebvre Equatins (1-3) (see Figures 3,5,6) t thse data pints which substantially deviate frm calculated values fr D 3 versus ALR. This finding implies that the Rizk and Lefebvre equatin hlds fr atmizatin cnditins which lye t the right f the bundary. The fiber type breakup is what is believed t lead t gd atmizatin quality by the breakup f liquid sheets t ligaments (fibers) and with an increase in air velcity the earlier the ligaments frm, until they are frmed at the lip [18]. Again, the bundary is nt strictly defined by the line prvided, but resides in a narrw regin in the general vicinity f the line. The ALRs used

6 in the experimental study by Rizk and Lefebvre t derive this equatin ranged frm t 8 [1]. The ALRs utilized fr the present investigatin range frm ~.5 t.65, significantly lwer than the derivatinal study. The deviatin in curve fit prvided the existence f a bundary where the equatin may be applied and als that ALR s lwer than can prvide a superir fit. The mdificatins made prir t the Rizk and Lefebvre equatin nly adjusted the magnitudes f each terms t crrelate mre clsely with experiment, while the physics were nt tampered with. It is nt surprising that the regins which prvided the best agreement and n significant deviatin (at higher ALRs) all begin and fall t the right f the bundary. Ntice hw B99 s data lcatin is dminated the membrane breakup regin, and hence shws pints with the mst deviatin in trend, as nted earlier. This develpment prvides insight int cmparing drplet breakup behavir fr each fuel and the ptimizing necessary t attain better atmizatin quality. Specifically, DF and E1 behave very similarly when cmparing atmizatin behavir upn review f Figure 7 and Figure 8. Hwever, B99 appears t require additinal aerdynamic breakup (t vercme the wave and ligament frmatin suppressin due t higher viscsity) in rder t achieve the same type f airblast atmizatin quality [18]. This was the initiating step fr a bidiesel atmizatin ptimizatin study in the C3 gas turbine, which has been reprted elsewhere [11]. Evapratin A mdified classic drplet evapratin mdel with refinements made fr cnvective effects and film thery was applied t assess the relative vaprizatin time f the three liquid fuels f interest [1,19]. A Lewis number f unity, a single cmpnent fuel apprximatin, and a ne dimensinal gas flw were assumed fr the heat up and steady state prtins f the mdel. Fuel prperties fr B99 and DF were determined frm labratry measurements, mass spectrscpy gas chrmatgraphy analysis and varius trusted published surces [1,13,14,], while the prperties fr E1 are readily available frm chemistry handbks, e.g. ref. [1]. A mre specific descriptin f the single drplet islated mdel is explained elsewhere [1]. Using an ALR determined frm the gas turbine engine at full lad cnditins, a D 3 value f micrns was determined fr the gas turbine perated at n a designated fuel, DF. Engine cnditins were measured with pressure transducers and thermcuples near the exit f the airblast atmizatin nzzle. A mre specific descriptin f the measurement apprach was explained in prir wrk by the authrs [8]. The first drplet cmparisn mdel was cnducted in a quiescent envirnment withut the effect f frced gas cnvectin. The results are shwn in Figure 9. SMD (micrns) Heat Up B99 Steady State B99 Heat Up DF Steady State DF Heat Up E1 Steady State E Time (ms) Figure 9. Drplet evapratin mdel at engine cnditins in quiescent envirnment fr B99, DF and E1 at an initial drplet diameter f micrns. The results shw that bth B99 and E1 require mre time t evaprate than DF (B99 ms and E1.6 ms). Fr E1, nte the quick heat up perid prir t drplet evapratin (due t a higher vlatility), yet the significantly larger latent heat f evapratin (refer t Table 1) which results in an verall slwer evapratin rate in cmparisn t DF and B99. B99 s heat up perid ccupies a large prtin f its vaprizatin histry. SMD (micrns) Heat Up B99 Steady State B99 Heat Up DF Steady State DF Heat Up E1 Steady State E Time (ms) Figure 1. Drplet evapratin mdel at engine cnditins with inclusin f cnvective effects fr B99, DF and E1 at an initial drplet diameter f micrns. In the secnd scenari (als at the same initial drplet diameter f micrns), the cnvective effects f the cmbustin air and drplet trajectry were incrprated (an average relative drplet injectr velcity was used fr simplicity). The cnvective air velcity (when changing the drplet t an Eulerian frame f reference) was fuel dependent and ranged frm 15. t 16.3 m/s. The results are shwn abve in Figure 1.

7 SMD (micrns) The vaprizatin histry shws a substantial fuel influence when cnsidering effects f the surrunding envirnment in a gas turbine. Mst ntably, E1 s vaprizatin time has decreased significantly (a 45% decrease frm 9.1 t 5 ms). The cnvective effects affect the steady (actually quasi steady) state vaprizatin, which is dminant in E1 s drplet lifetime. The evapratin rates fr B99 and DF imprved as well, yet t a much shrter extent (1.6 t 1. ms, a 4.% reductin and 8.6 t 7.8 ms, a 9.3% reductin). The findings present a scenari f practical imprtance, yet this rder f magnitude analysis is best utilized as a means t establish a direct cmparisn between fuels. The third and final scenari cnsiders the initial drp sizes which wuld be expected by transitining fuels withut mdificatin t the airblast atmizatin as well as cnvective effects in the injectr/cmbustr envirnment. Due t the similar atmizatin prperties between DF and E1, the atmized drplet size shuld be rughly the same (less than a 5 micrn difference) and therefre are bth set t micrns. Hwever, fr B99, a difference f at least 1 micrns (determined frm a sensitivity analysis n equatins (1) and (3) shwed a 15% increase in drp size using data frm engine testing) is shwn t ccur at engine flw rates (see Figure 4), and an initial drp size f 5 micrns was chsen Heat Up B99 Steady State B99 Heat Up DF Steady State DF Heat Up E1 Steady State E Time (ms) Figure 11. Drplet evapratin mdel at engine cnditins with inclusin f cnvective effects fr B99, DF and E1 at expected drp sizes fr transitining fuels withut mdifying injectr air flw. The results f the mdel shw that fr B99, a fuel with higher viscsity (a larger initial drplet size) crrespnds t a much larger evapratin time (an increase frm 1. t 15.9 ms, a 56% increase fr nly a 1 micrns difference). The findings stress the imprtance f atmizatin quality when seeking t achieve superir fuel preparatin, cmplete cmbustin and lw exhaustin f pllutant emissins. Prperties f Bidiesel and Ethanl Blends Anther interesting idea is the ability t cmbine the alternative fuels used in this study t help tailr the atmizatin and vaprizatin prperties t achieve better atmizatin quality as well as t mre clsely simulate DF by prducing a desired mixture between these tw fuels. The additin f E1 t B99 significantly reduces the fuels viscsity, surface tensin, and density t values which apprach that f DF. Past wrk has addressed similar strategies fr the ptential f a biderivative liquid fuel in cmbustin devices [1,11]. The atmizatin prperties fr blends f bidiesel and ethanl are discussed in this sectin. Blending f E1 int B99 can greatly affect its verall fuel atmizatin prperties. In fcusing n the change in fluid prperties f interest, this idea can serve t design bidiesel t better suit cmbustin systems which are intended t perate slely n petrleum derivative fuels. The fuel blending s effect n viscsity frm labratry measures using a ball drp viscmeter can be bserved in Figure 1 at atmspheric cnditins fr 1/, 95/5, 9/1, 8/, 6/) mixtures f sy bidiesel/ethanl. A steep decay is nticed as the percent f E1 is increased in B99. By bserving the verall trend in blending, a mixture clse t a /6 blend f B99 and E1 shws t have the same viscsity as the pure DF, typically used fr this gas turbine generatr. The decrease in viscsity culd prvide fr a very attractive ptin t imprve the fuels atmizatin quality as well as the perfrmance f the fuel preparatin system (i.e., by making it mre efficient, due t a decrease in the required pumping pwer). (kg/m-s B99 & E1 DF Additin f Ethanl in Bidiesel (%) Figure 1. Change in viscsity fr 95/5, 9/1, 8/ and 6/ percent blends f B99 and E1. Mdificatin t B99 s surface tensin was als measured, using a stalagmeter, t understand E1 s blending affect in cmparisn t the designated DF. The values are shwn in Figure 13 and als shw an

8 expnential decay with increasing percentage f E1. The surface tensin reaches that f DF at an 8/ blend. Thus far, bth changes t fuel prperties due t fuel blending have been nn linear B99 & E1 DF prperties appears t be a strategy t simulate DF s fuel prperties. Hwever, the energy decrease due t blending must als be cnsidered (as well as the cst f these fuels) when selecting the ptimum blend between these fuels. Lastly, the increase in the amunt f E1 in B99 is increased, the LHV decreases and this change is tabulated in Table. A minr (yet significant) decrease is bserved with the blending f the tw liquid alternative fuels. (kg/s ) Table. Decrease in lwer heating value fr 95/5, 9/1, 8/ and percent blends f B99 and E1. Blend B99 B95E5 B9E1 B8E B6E LHV (kj/kg) 37, 36,87 36,3 35,8 33,16 % decrease - 1.4%.8% 5.7% 11.3% Additin f Ethanl in Bidiesel (%) Figure 13. Surface tensin fr 95/5, 9/1, 8/ and 6/ percent blends f B99 and E1. (kg/m^ B99 & E1 DF Additin f Ethanl in Bidiesel (%) Figure 14. Change in density fr 95/5, 9/1, 8/ and 6/ percent blends f B99 and E1. The change in the fuel s density was als measured fr the mixtures f interest as well. The results are shwn in Figure 14. The change in density is shwn t behave quite linearly with the change in the percentage f ethanl. This prperty can influence the fuel system in a psitive manner in rder t imprve fuel preparatin fr the system, yet plays an adverse rle in airblast atmizatin (eq. (1) r (3)) [18]. The density f bidiesel is higher than DF s until apprximately a 75% additin f ethanl t the fuel. This can nce again be a ptential fuel design ptin t help prepare the fuel fr a system f interest. A cmprmise between the liquid Emissins Reductin (ppmvd) Emissins Behavir In the atmizatin ptimizatin study by the authrs, reprted elsewhere, the specific analysis due t the link between atmizatin and emissins was left t be explained in cntent f the analysis f this manuscript [11]. The C3 gas turbine perated n B99 was cnfigured t allw its atmizatin air flw t be increased (increasing ALR) fr a cnstant fuel flw. Fr engine lad set pints f 5, 3 and 1 kw, an increase in atmizatin airflw rate was perfrmed in an attempt t reduce NOx emissins. The results revealed that fr each f the lads tested the minimum emissins levels were achieved fr the same ALR value f.85 and experienced the same change in ALR fr each lad setpint due t the peratinal ALR being the same fr each f these lads, as shwn in Figure C 3 design ALR CO decreases ALR fr ptimized emissins due t atmizatin ALR ( ) Figure 15. Change in Airblast atmizer ALR and cnsequent emissins reductin during peratin f Capstne C3 [11]. NO 5 kw NO 3 kw NO 1kW CO 5 kw CO 3 kw CO 1kW Behavir nticed frm DF test ("verly ptimized") T achieve minimum emissins levels, the ALR was increased by 17%. A sensitivity analysis f the

9 Rizk & Lefebvre equatin using the change in fuel prperties frm DF t B99 suggested the change in empirical cefficients frm equatin (1) t (3) t accunt fr the change in relative velcity and fuel prperties. The increase in dynamic viscsity f B99 (the dminating liquid prperty) and increase in ALR required fr minimum emissins, were f the same rder f magnitude. Befre the ALR was increased, the analysis suggested mdificatin f the empirical cefficients was sufficient t accunt fr the change in fluid prperties as shwn in Equatin (3) (within 5% errr). Hwever, when the ALR is increased, the subsequent effect n relative velcity dminates the change in drp size (by a factr f 4.5 larger than the ALR change). The drp size decreases frm 3 t 4 micrns fr B99 at the ptimized ALR, which is significantly smaller than the micrn SMD fr DF at the baseline ALR. Thus B99, with ptimized ALR features a significantly shrter drplet evapratin time than DF by cmparisn f ptential evapratin histries. This finding shws that the atmizatin quality with B99 at ptimized ALRs shuld be mre than sufficient t achieve the desired fuel/air preparatin, yet emissins levels are still higher than they are fr DF at the baseline cnditins. This result indicates that, perhaps nt surprisingly, ther factrs in additin t atmizatin quality need t be cnsidered when ptimizing emissins. Fr example, visual inspectin f the injectrs fllwing engine tests n B99 withut ALR ptimizatin indicated wall impingement in the C3 [11]. Visually, the recessed airblast plain-jet atmizer prduced a larger spray cne angle (8 degrees wider at atmspheric cnditins) which in turn likely plays a rle in the spatial distributin f fuel and likely emissins [1]. Als, increasing ALR results in a substantially higher pressure drp, which may nt be acceptable frm a design perspective. As a result, injectr redesign may be necessary t achieve the desired ALR and avids high pressure drps. Further, the rle f cmbustin has nt been cnsidered. It is likely that heat release will influence drplet evapratin and cnsequent emissins. Lastly, develpments in chemical kinetics as well as insight in engine design fr such fuels at B99 are needed t help explain bserved NOx emissins behavir. Hence, NO x emissins, thugh reduced thrugh ptimizatin f atmizatin, are als impacted by additinal factrs. Nnetheless, the results shw the ptential fr emissins reductin thrugh ALR ptimizatin. Figure 16 cmpares ptimized airblast atmizatin NO x emissins with emissins fr a straight fuel change. Thrugh ALR mdificatin, a 1 t 5% reductin in NO x emissins was achieve with CO remaining unchanged. NO & CO 15 % O ) B99 NO B99 CO B99 NO Optimized B99 CO Optimized DF NO DF CO Pwer Output (kw) Lads where Atmizatin was Optimized fr B99 Figure 16. Emissins reductins due t atmizatin ptimizatin strategy in the C3 gas turbine engine [11]. In light f the varius factrs affecting emissins as described abve, it is difficult t assess if the levels attained are the lwest pssible levels. As a result, the questin f what is the emissins reductin ptential fr bidiesel, warrants sme discussin. Experimental well stirred reactr wrk prefrmed at the University f Washingtn has shed light n this subject fr varius fuels []. This wrk was perfrmed at temperatures (clse t 179 K), residence times (n the rder f 1 t 5 ms) which are similar thse bserved in the Capstne C3. Figure 17 shws that the NOx emissins are well crrelated with carbn t hydrgen rati. Results fr B99 and DF fr the present study have been pltted here as well (C:H ratis were determined frm mass spectrscpy gas chrmatgraphy). Bth fuels fall within a trend regin f alkanes, methyl alchl, and diesel fuels. Sy methyl ester can be thught f as a semi straight chain alkane, but with duble bnds and with a methyl alchl head n the mlecule. The influence f this structure n the emissins is nt yet well understd, hwever, due t its hydrcarbn cmpsitin, shuld reside in the vicinity f the trend line and prduce NO x within these limits. Overall, due t the very clse carbn t hydrgen rati between the tw fuels, it is expected that the minimum NO x emissins shuld by very clse (between 7.35 t 9.5 ppm at 15% O ). Nte that methanl falls belw the trendline and B99 resides lwer n the emissins levels than the dirtier diesels and sme lw sulfur diesels in Figure 17. Cnclusins Cnclusins drawn frm this wrk include: In terms f atmizatin, B99 prperties require mdificatin t the injectr t attain emissins perfrmance similar t DF. A rapid decay in atmizatin quality with a decrease in ALR is due t aerdynamic frces.

10 The Rizk & Lefebvre equatin prvides excellent agreement in the fiber type breakup regime and at ALR s belw. The cefficients f the Rizk & Lefebvre equatin were mdified t prvide better agreement with measured data. A brute frce methd which prvided adequate results was derived in rder t establish better predictability fr the range f flws f interest. Based n analysis f drplet evapratin, the criticality f including cnvective effects is demnstrated. A cmparisn f the three fuels shwed that B99 has a significant hurdle t vercme with its large evapratin times linked t its large heat up perid. Thugh atmizatin imprvements were shwn t imprve emissins perfrmance fr B99, it is apparent that ther factrs als play an imprtant rle in the emissins. By cmparing NO x entitlement behavir fr DF and B99, the cntributin f these ther factrs can be estimated. T vercme emissins and peratinal issues with B99, the cncept f blending mixtures f B99 and E1 appears t be an attractive ptin. This can imprve fuel prperties fr B99 and mre clsely simulate the fuel fr which cmbustin systems have been designed fr. Acknwledgements The authrs wuld like t thank Prfessr Sctt Samuelsen, Rich Hack, Bryan Hayden and Mike Crespin f the UCI Cmbustin Labratry fr their supprt and guidance. A special thanks t the UCI Cmbustin Labratry undergraduate researchers Adrian Narvaez and Tavis Werts fr their help in data analysis and aid in experimentatin. Capstne Turbine Crpratin fr the prvisin f the gas turbine used fr the research presented herein.

11 1 NOx Emissins 15 O ) Bidiesel (B99) UCICL-Lw Sulfur Diesel Methanl Methane L Methane S Ethane L Ethane S Prpane L Prpane S Ddecane L Hexadecane L Hexadecane S Heptane L Hexane L Hexane S Pentane L K-Light Naphtha S K-Light Naphtha L C-Lw Sulfur Diesel USOR-Lw Sulfur Diesel L USOR-Light Naphtha L C-Lw Sulfur Diesel L (N FBN) USOR-Lw Sulfur Diesel L (N FBN) T-Lw Sulfur Diesel L (N FBN) T-High Sulfur Diesel L (N FBN) T-High Sulfur Diesel S (N FBN) Fuel Carbn t Hydrgen Rati Figure 17. Entitlement NOx emissins fr Bidiesel (B99) in cmparisn t Diesel Fuel #, where S and L represent shrt and lng residence times f 18 and 5 ms. [adapted frm ] References 1. Lefebvre, A. H., Atmizatin and Sprays, Hemisphere Publishing, 1989, p. 7, 1, Ch. 8.. Lefebvre, A. H., Gas Turbine Cmbustin, McGraw-Hill, 1983, p. 19, Demirbas, A., Bidiesel Prductin frm Vegetable Oils via Catalytic and Nn-Catalytic Supercritical Methanl Transesterificatin Methds. Prg. Energy & Cmb. Sci., Vl. 31, Sept. 3, 5, pp Tat, M. E., and Van Gerpen, J.H., Fuel Prperty Effects n Bidiesel. Paper 364, ASAE Meeting, Las Vegas, Nevada, July, Envirnmental Prtectin Agency (), A Cmprehensive Analysis f Bidiesel Impacts n Exhaust Emissins, USA Octber. 6. McCrmick, R. L., Williams, A., Ireland, J., Brimhall, M., and Hays, R. R., Effects f Bidiesel Blends n Vehicle Emissins. Milestne Reprt NREL/MP-5-554, Octber, Lupandin, V., Niklayev, A., and Thamburaj, R., Test Results f the OGT5 Gas Turbine Engine Running n Alternative Fuels: Biil, Ethanl, Bidiesel and Crude Oil, Paper GT , ASME Turb Exp, Ren- Tahe, Nevada, USA, June Blsz, C. D., Mauzey, J. L., McDnell V. G., and Nakamura, S., Experimental Investigatin f Liquid Fuel Atmizatin, Mixing and Pllutant Emissins fr a 3 kw Gas Turbine Engine. Prceedings, 18 th ILASS- Americas Cnference, Irvine, CA, May Nakamura, S., McDnell, V.G., and Samuelsen G.S., The Effect f Liquid-Fuel Preparatin n Gas Turbine Emissins, 6. Paper GT-6-973, accepted fr publicatin in ASME J. f Engineering fr Gas Turbines and Pwer. 1. Blsz, C. D., McDnell, V. G., and Samuelsen, G. S., Impact f Bidiesel n Fuel Preparatin and Emissins fr a Liquid Fired Gas Turbine Engine. ASME Turb Exp 7, Mntreal, Canada, May , 7.

12 11. Blsz, C. D., and McDnell, V. G., Optimizatin f Fuel Preparatin fr Emissins Reductin frm a Small Scale Gas Turbine Engine Operated n Bidiesel. 5 th meeting f the Cmbustin Institute, San Dieg, CA, march 7 1. Rizk, N. K. and Lefebvre, A. H., Spray Characteristics f Plain-Jet Airblast Atmizers, J. Engr. Gas Turbines and Pwer, Vl. 16, pp , Schumacher, L., Chellappa, A., Wetherell, W., and Russell, M. D., The Physical and Chemical Characterizatin f Bidiesel Lw sulfur Diesel Fuel Blends, Natinal Bidiesel Bard. Jan. 5, Tysn, K. S., Bidiesel Handling and Use Guidelines, NREL Technical Reprt N: NREL/TP-58-34, USA, September., Mnyem, A., Van Gerpen, J.H., The Effect f Bidiesel Oxidatin n Engine Perfrmance and Emissins. J. Bimass & Energy, Vl., Dec. 18,, pp Orhesrge, W., Frmatin f Drps by Nzzles and the Breakup f Liquid Jets, Z. Angew. Math. Mech., Vl. 16, 1936, pp Lasheras, J. C.,and Hpfinger, E. J., Liquid Jet Instability and Atmizatin in Caxial Gas Stream, Annu. Rev. Fluid Mech., Vl. 3, pp Lefebvre, A. H., Airblast Atmizatin. Prg. Ener. & Cmb. Sc., Vl. 6, Jan. 7, 198, pp Abramzn, B. and Sirignan, W. A., Drplet vaprizatin mdel fr spray cmbustin calculatins, Int. J. Heat Mass Transfer Vl. 3, pp , Yaun, W., Hansen, A. C., and Zhang, Q., Predicting the Physical Prperties f Bidiesel fr Cmbustin Mdeling, Transactins f the ASAE, Vl. 46(6):pp , Lide, D. R., ed. CRC Handbk f Chemistry and Physics, 87 th Ed. Taylr and Francis, Lee, J.C.Y., Malte, P.C., and Benjamin, M.A., Lw NOx Cmbustin fr Liquid Fuels, Atmspheric Pressure Experiments Using a Staged Prevaprizer-Premixer, Paper 1-GT-81, ASME Turb EXPO, New Orleans, June 1.