Pouria Mikaniki,Seyed Mohammad Ali Naja fi,Hojat Ghassemi*

School of Mechanical Engineering,Iran University of Science and Technology,(IUST)P.O.B.16765-163,Tehran,Iran

Keywords:Atomization Pressure-swirl injector Mazut Size distribution Fuel Viscosity

ABSTRACT The available SMD(Sauter mean diameter)correlations on pressure-swirl injectors predict droplet sizing very different from each other,especially for heavy fuels.Also there was a lack in the literature for comparing available correlations.So an experimental study w as conducted on a heavy fuel oil(HFO)spray,Mazut 380.A pressure sw irl injector w as designed and fabricated.The experiments for Mazut at 40 °Cand 80 °Cw ere compared w ith the results for water,including spray half cone angle,breakup length and mean droplet diameter,at different injection pressures.Lower spray angle,higher breakup length and larger droplets were observed for low er injection pressures and higher liquid viscosity.SMD w as about 75 μm for w ater and about 87 μm for Mazut at 80°C.The results for droplet mean diameter were also compared with correlationsfrom previous studieson pressure swirl atomizers.The SMDresultsshow that for water spray,LISAmethod w asin good agreement,also Babu and Ballester correlations w ere successful w hen high viscosity fl uid w as injected.

1.Introduction

Pressure-sw irl injectors are w idely used in combustion chambers and various engineering applications.Determining the performance characteristicsof the injectorsand studying the spray behavior will provide valuable bene fi ts in optimizing the combustion system.The spray angle,breakup length and droplet diameter are the key factors for designing the size of the combustion chamber and evaporation process.

HFOs,such as Mazut,w ould not easily atomize because of the high viscosity.The use of thistype of fuel is dif fi cult because of their high viscosity and relatively high pollution(notably sulfur emissions),w hile these properties have made the heavy fuels considerably cheap,so they are the main supplier of marine engines,some factories and thermal pow er plants.Reducing the size of the atomized droplet diameter is a w ay to improve the ef fi ciency of the liquid fuel combustion[1].Improvement in fuel atomization processisan important issue in thecombustion process,w hich could be considered as shorter breakup length and smaller mean droplet size.

Goldsworthy[2]has examined the evaporation and combustion of HFOs in heavy duty diesel engines.In the article some parameters like ignition delay,velocity and spray space distribution have been measured and discussed.An experimental study w as performed by Jasuja[3]to de fi ne the difference betw een atomization of three different fuels(Kerosene,gas oil,and heavy residual fuel oil)by three fuel atomizer types.The result w as that the existing droplet size correlations are not suf fi ciently accurateto usefor high viscosity fuels.Someresearchers focused on spray models inside the combustion chambers of diesel engines.Fink et al.[4]obtained theexperimental resultsfor macroscopic and microscopic parameters of spray w ith tw o types of fuel(HFO and diesel fuel w ater emulsion,FWE 18%).The results led to development of spray simulation models.They concluded that the fuel type has a signi fi cant impact on the spray performance.Park et al.[5]studied the effect of fuel typeon theinjection of marinediesel engines.They observed that the penetration length decreases w ith increasing temperature.

Jedelsky et al.[6]studied the energy transfer betw een a liquid hollow cone spray and the surrounding air using both imaging and phase-Doppler techniques.The spray w as produced by a pressure sw irl atomizer discharging Jet A-1 fuel.A correlation betw een size and axial velocity of individual droplets was presented in the study.

A research on the frequency domain of a pressure sw irl injector discharging kerosene Jet A-1 has show n two oscillation modes at low frequency around 100 Hz and at high frequency around 1800 Hz.This issue is a key result as the spray oscillations affect the fl ame stability and the resulting combustion ef fi ciency because a change in the local air/fuel mixture ratio is made[7].

Rashad et al.[8]studied the effect of geometric parameters on spray characteristics of pressure sw irl atomizers,using w ater as the test liquid.Optimal values are determined for different geometrical ratios.

In the case of HFOatomization most of the researches are on the marine engines.Zhou et al.[9]investigated sensitivity analysis of HFO spray and combustion under low-speed marine engine-like conditions.

In this study,an injector w as designed in the application of a HFO entrained fl ow gasi fi er and studied to obtain thereliable datafor further researches.The entrained fl ow gasi fi ers require a smaller fuel droplet size than the fl uid bed gasi fi er[10].Therefore in the present study,the important characteristics of a pressure injector w ere investigated.The main characteristics of an injector w hich are required in the design of a combustion chamber include discharge coef fi cient,breakup length,spray angle and the average diameter of the droplets.The discharge coef fi cient isimportant in thedetermination of injector mass fl ow rate,the spray angle is important in de fi ning the chamber crosssection area,the breakup length and mean diameter of droplets are very effective in designing the length of the chamber and the duration of the droplet presence in the chamber to evaporate.Therefore,in this study,these speci fi cations have been empirically investigated in order to evaluate the designed injector in the application of an entrained fl ow gasi fi er.Also the present study w as focused on comparing available pressureswirl injector SMD correlations(in terms of injection pressure,liquid density,viscosity and surface tension)for a HFO,Mazut.Water and Mazut,which have a high viscosity difference,were used asthe operating fl uids.

The collection of almost all w ell-know n SMD correlations for pressure swirl injector was presented to easily guide other researchers to fi nd the best injector for other applications.Also it w as observed that the available SMDcorrelations predict droplet sizing very different from each other and there w as a lack in the literature for comparing available correlations.

2.Injector Design

A pressure-sw irl injector w as designed by a procedure introduced by[11].The inputs of this design procedure include mass fl ow rate(˙m),spray angle(θ),pressure difference(Δp),density(ρl)and kinematic viscosity(ν).In order to investigate and study the fuel'sbehavior and examine the spray characteristics,a pressure-sw irl injector w as designed and fabricated.The pressure sw irl injector w as designed for Mazut w ith amass fl ow rate of 0.05 kg·s-1at 2 MPapressuredifference with a spray half angle of 30°.The injector hole is cylindrical type.The dimensions of the pressure-swirl injector are show n in Fig.1b.

3.Experimental

Fig.1a show s a schematic of the fuel spray test setup used in the present study.A stainless steel tank(4 in Fig.1a)is used to store the fuel at a particular pressure by meansof a compressed nitrogen gassupply(1 in Fig.1a)and a pressure regulator(2 in Fig.1a).A high pressure fuel fl ow line is used to carry the fuel to the atomizer(7 in Fig.1a)inlet.In the current study,fuel spray characteristics w ere examined for the injector with pressure difference,Δp up to 40 bar,injection pressure differenceand injection fl uid temperaturecan bemeasured by apressure gage(3 in Fig.1a)and a thermocouple(6 in Fig.1a).All of the experiments w ere carried out at atmospheric standard condition.

The characteristics of liquid sheet discharging from the atomizer w ere obtained by capturing images of liquid sheets using photographic techniques.A PCO 1200hs digital high-speed camera(resolution of 1280×1024 pixels)(9 in Fig.1a)w ith a diffused backlighting system(8 in Fig.1a)was used to take images of the liquid sheets.A macro lens(AFZoom Nikon 100 mm f/2.8D)was used in conjunction with the camera.The minimum measureable droplet diameter range is about 20μm,because of the camera resolution limit.Mass fl ow rate w as determined experimentally.Spray cone angle,breakup length and droplet size distribution were obtained by image processing(10 in Fig.1a).

Water and Mazut 380 are used asoperating fl uid.For Mazut at 40°C and 80°C,speci fi c density,kinematic viscosity and surface tension are reported in Table 1.More detailed speci fi cations are presented in[12].Also properties for w ater at 20°Cw ere included in the table for convenience of comparing w ith Mazut.

Table 1 Speci fi cations of test conditions and fl uid

Fig.1.a)A schematic of the experiment setup,b)schematic,geometry and dimensions of the injector(not depicted with real dimensions).

Fig.2.a)Mazut spray at 80 °Cand Δp=2 MPa,exposure time=9 μs,b)spray angle and breakup length,water spray.

Fig.2a shows an image of the spray for Mazut at 80°C(Mazut80)and pressure difference of 2 MPa.Spray conical shape,unstable w aves on the sheet,details of the sheet breakup and spray droplets w hich formed from the ligaments clearly can be seen.Detailed quantitative analysis of spray characteristics is presented in the follow ing sections.Mass fl ow rate,m˙and injector pressure difference,Δp are measured to fi nd the discharge coef fi cient,CDw hich can be expressed as[13]:

w here A0is the injector hole area.The injector discharge coef fi cient expresses the ratio of the actual discharge value to the theory value.

The fi lm velocity U[m·s-1]iscalculated from thefollowing equation[14]:

w hereΔp[Pa]is the injector pressure difference.The axial velocity component u[m·s-1]can be evaluated from u=U cosθ.The injection pressure difference,Δp and injection half cone angle,θw ere determined from the experimental data.

The Reynoldsnumber and aerodynamic Weber number,discharging from the simplex pressure-swirl atomizer are expressed as[13]:

whereρgisthedensity of ambient gas(1.2 kg·m-3),d0istheinjector hole diameter,shown in Fig.1b(d0=1.5 mm in this study),ρlis the fl uid density,μlis the fl uid dynamic viscosity and σ isthe fl uid surface tension.

4.Results and Discussion

4.1.Mass fl ow rate

Water at 20 °C,Mazut at 40 °C(Mazut40)and Mazut at 80 °C(Mazut80)are representing fl uids with three different viscosities.The results are show n in Fig.3a.Larger mass fl ow rates w ere observed for low-viscosity fl uid and higher injection pressures.

4.2.Discharge coef fi cient,spray angle and breakup length

The injector discharge coef fi cient is essentially dependent on the injector geometry[15,16],moreover some experimental studies show that the injection pressure also plays an important role[17,18]w hich is approved by the results of the present study.By increasing the injection pressure(Reynolds number),CDexperiences an increasingdecreasing variation(and then remains at a constant value for water),w hich isdepicted in Fig.4,thisisalso reported by previous studies[19].

Fig.3.a)Mass fl ow rate in terms of injection pressure,b)normalized breakup length.

Fig.4.Spray angle and discharge coef fi cient.

Fig.5.a,b)Spray global Probability Density Function and the global mean droplet sizes.c,d)Cumulative droplet volume and Rosin-Rammler distribution and experiment data for w ater.e,f)Drop-size distribution as a function of Cumulative Volume Fraction for relative number count of droplets,Δp=2 MPa.

In low er injection pressures(lower Reynolds numbers),the central air core at the injector is still not fully developed and the air pressure fl uctuates,at higher injection pressures that the air core is completely formed,the discharge coef fi cient remains almost constant[19](also observed for spray angleand breakup length Figs.4c and 3b).Thisconstant CDw asmeasured about 0.62 for w ater.Resultsindicate that by increasing viscosity,CDwasdecreased.Half cone angle also increases by an increase in Reynolds number and stops increasing at higher Reynolds numbers.The half cone angle is plotted in Fig.4.Half cone angle w as increased by an increase in injection pressure;variations are much sharper in low er injection pressures.

Some previous studies investigated the in fl uence of pressure difference on spray angle.No evident effect on cone angle was observed for the nozzleΔp by Dodge and Biaglow[20].De Corso[21]also observed no in fl uence on the angle at the atomizer tip and concluded that an increase inΔp causes spray contraction by removing the nozzle tip.The measurements of Chen[22]and Rizk and Lefebvre[23]noticed an increase of the angle w ith pressure difference.Lefebvre[24]show ed a maximum for cone angle by an increase in the injection pressure difference.Among all previous studies various trends w ere observed,therefore a unique description for the effect ofΔp on spray cone angle cannot be proposed.The trend might depend on some other parameters,such as operating fl uid density,viscosity and surface tension or the atomizer's exact geometry and dimensions.Generally increasing Δp transfers the fl uid fl ow from a simple round jet to a hollow cone w hile momentum,velocity and air core diameter are increased and led to an increase in sw irl velocity and spray cone angle.As the air core appears,a sharp increase happens in spray angle and at higher pressures w ith a developed air core,spray angle variations are limited and converge to a maximum that w as con fi rmed by[24].

Another major parameter in the design of the combustion chamber is breakup length.The outlet fl uid from theinjector isnot atomizing immediately after theexit.The length of the un-atomized part isknown as the breakup length,and its variation in different Weber numbers is show n in Fig.3b.Breakup length w as normalized by injector outlet diameter(1.5 mm).It can be found in this fi gure that in the same condition low viscosity fl uids can be atomized in shorter breakup length.By increasing the injection pressure,breakup length decreased but thevariationsare much larger at lower Weber numbers.

During breakup process,the momentum force overcomes the viscousforce;therefore by increasing injection pressure difference,momentum force increases and can overcome the viscous force in shorter distance(breakup length).

4.3.Droplet sizing

The results for global drop size,PDF,for the designed injector at the injection pressureΔp=2 MPa,are depicted in Fig.5a,b.Also in this fi gure,the global SMD(D32)is represented by dash line,the formulation w as presented in[24].

The opaque nature of the HFO,which is associated with limitations w ith respect to the applicability of optical measurement techniques[25]such as PDA,guided the use of shadowgraphy technique and image processing to determine SMD and PDF.In order to measure the droplet diameter,by image processing,the pixel area is determined by calibration,and the size of each droplet is determined by the number of pixels that occupied by each droplet.

To represent droplet sizing,some distribution functions are used,the most well-known distribution is Rosin-Rammler,asintroduced in Eq.(4):

where,d representsthe droplet diameter(μm)and Ydthe volume fraction of the droplets with a diameter greater than d.From the experiment data forΔp=2 MPa,and n w ere found for Mazut80 equal to 113.8μm and 2.97,respectively w hich is show n in Fig.5c and for w ater equal to 95.12μm and 2.60 w hich is shown in Fig.5d.Due to the high viscosity of Mazut at 40°C,the liquid w as not w ell atomized and therefore the corresponding results have not been presented.

Fig.6.Spray images at different injection pressures,Mazut80.

Table 2 Correlations on SMD for pressure swirl atomizers

The histogram plots of droplet size are shown in Fig.5e,f.It represents the relative number count of droplet versus the droplet diameter,w hile the curves represent the cumulative volume versus the droplet diameter for pressure difference 2 MPa.For the case of w ater and Mazut,histogram number count peak occurs at about 25 μm and 40 μm,respectively(Fig.5e,f).This also indicates that more large droplets are created for the case of Mazut(more viscous fl uid).

Fig.6 show s the images of Mazut80 spray at different injection pressures.Increase in spray angle and decrease in breakup length are evident by an increase in injection pressure.

There are a lot of correlations predicting SMD for pressure sw irl atomizers.Athorough survey hasbeen made in theliteratureand almost all popular correlations on SMD prediction of pressure sw irl injectors werefound,w hich aresummarized and presented in Table2.Each correlation isvalid in acertain range of fl uid density,surfacetension,viscosity,and injection pressure difference,w hich are noted in Table 2.So,if a researcher w ants to choose,use or design a pressure-sw irl injector,Table 2 helps a lot.Surprisingly the available correlations predict SMD very far from each other and for a heavy viscous fl uid like Mazut,the difference ismuch more.We tried to evaluatethe correlations by designing and examining a pressure sw irl injector for Mazut and fi nd the best correlation for predicting SMD of a pressure swirl injector.

Fig.7 show s a comparison of the present research results and the correlations w hich were described in Table 2.It reveals that for water,any of the correlations cannot predict droplet mean diameter for injector pressures below design injection pressure(2 MPa)and for higher injection pressures the LISA method is the best.For Mazut,Babu correlation[27]could predict SMD w ith about 10%overestimation;also the results are in mild agreement w ith Ballester correlation[30].Each correlation w as depicted in the validΔp range w hich w as mentioned in Table 2.

5.Conclusions

In this study,a pressure-swirl injector was designed and examined for HFO.Water and Mazut,w hich have a high viscosity difference,w ere used.Injection pressure,mass fl ow,discharge coef fi cient,spray angle,breakup length and droplet diameter w ere the characteristics that w ere measured or obtained,and their variations w ere expressed as a function of the Reynolds and Weber numbers or the injection pressure.

Low er spray angle,higher breakup length and larger droplets w ere observed for low er injection pressures and higher liquid viscosity.SMD w as about 75 μm for w ater and about 87 μm for Mazut at 80 °C andΔp=2 MPa.The available SMD correlations on pressure-sw irl injectorspredict droplet sizing very different from each other,especially for HFOs.So they w ere presented and compared w ith the present study in order to choose thebest correlation for each w ater or Mazut.The SMD results show that for w ater spray,the LISA method w as in good agreement for injection pressure equal and higher than design pressure also Babu and Ballester correlations w ere successful in predicting SMD when the viscous fl uid wasinjected.

Fig.7.The comparison of SMDvariation for different injection pressures,measured data and the correlations on pressure swirl atomiz er.

Nomenclature

A0ori fi ce area,m2

Apthe total area of holes of sw irl chamber,m2

CDdischarge coef fi cient

D diameter,m

d droplet diameter,μm

d0ori fi ce diameter,m

d size constant,μm

dpsw irl chamber inlet diameter,m

FN fl ow number,m2

Lbbreakup length,cm

l ori fi ce length,m

lPsw irl chamber inlet length,m

lssw irl chamber length,m

m˙mass fl ow rate,kg·s-1

PDF probability density function,μm-1

Re Reynolds number

SMD Sauter mean diameter,μm

tffi lm thickness,m

U fi lm velocity,m·s-1

u axial velocity components,m·s-1

We Weber number

Ydvolume fraction of the droplets with a diameter greater than d

Δp pressure difference,Pa

θ spray half angle,(°)

μ dynamic viscosity,Pa·s

ν kinematic viscosity,m2·s-1

ρ density,kg·m-3

σ surface tension,N·m-1

Subscripts

g air

i class i

l liquid

n size distribution parameter

s swirl chamber