S.M.A.Najafi,P.Mikaniki,H.Ghassemi

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

Keywords:Spray Heavy fuel oil Mazut Atomization Pulsed Pressure-Swirl Injector

ABSTRACT It is known that increasing the injection pressure reduces the breakup length and the droplet size.Adding pulses,on the other hand,helps to atomize the liquid into finer droplets,similar to air-assisted injectors but without altering the air-to-fuel concentration.To further reduce the droplet size and breakup length,a novel injector type,called ‘‘Pulsed Pressure-Swirl”(PPS),is introduced in this work,which is a combination of pressure-swirl and ultrasonic pulsed injectors.A pressure-swirl atomizer was designed and fabricated specifically for Mazut HFO (Heavy Fuel Oil).The droplet formation process and droplet size distribution have been studied experimentally(by shadowgraphy high speed imaging) and numerically (with the open-source Volume-of-Fluid code Gerris).Changing liquid injection pressure effect on the spray angle and film thickness has been quantified.These simulations have been used to study the primary breakup process and quantify the droplet size distributions,using different injection pulse frequencies and pressures.The numerical results have revealed that the new injector concept successfully produces finer droplets and results in a decrease in the breakup length,especially when applying high pulse frequencies,with no significant changes in the spray angle.

1.Introduction

Atomization plays a very important role in internal combustion engines,gas turbines,gasifiers and liquid fuel furnaces.It is essential to understand the atomization process in fuel injectors,as it strongly influences the evaporation process and heat transfer processes,and ultimately the combustion efficiency and amount of pollutant emissions.Therefore,this topic has been the subject of many experimental and modeling studies.However,there remain many topics and issues to be explored and improved.

Pressure-swirl atomizers are a very common type of spray nozzles.‘‘Simplex”atomizers are the simplest type of pressure-swirl atomizers [1].The liquid is fed into a chamber tangentially.Because of the tangential velocity component at the inlet a swirling flow prevails in the chamber which remains close to the injector internal walls.Air penetrates to the injector from the middle of the injector orifice and forms an air column in the middle,see Fig.1.At the end of orifice liquid exits as a hollow conical liquid sheet with axial and radial velocities.As the sheet expands,its thickness decreases and becomes unstable breaking into ligaments.The ligaments then disintegrate into droplets in the first stage (called primary breakup) and then each droplet breaks into smaller droplets due to the turbulent flow field in the second stage(secondary breakup).The atomization process can thus be divided into three main stages:film formation,sheet breakup and atomization.Despite the geometrical simplicity of the simplex pressureswirl atomizer,the hydrodynamic processes are highly complex[2].Fluid breakup happens due to the instabilities at the gasliquid interface,which are a result of the interplay between aerodynamic forces between liquid and gas,viscous,pressure and inertia forces,and surface tension at the interface.Thus,important parameters strongly affecting the atomization process,include gas and liquid properties,surface tension,liquid injection velocity and the injector diameter.

Many studies have been conducted on pressure-swirl atomizers.Rizk and Lefebvre [2] studied the effect of the pressure-swirl nozzle geometry on the film thickness.The analyses lead to the development of correlations for the discharge coefficient.Chin et al.[3,4] studied the influence of the downstream distance on the spray characteristics of pressure-swirl atomizers experimentally and analytically,over a wide range of conditions,including factors as ambient air pressure and velocity,fuel injection pressure,initial mean drop size,and initial drop size distribution.

Fig.1.Pressure swirl injector schematic

Rho et al.[5]examined the turbulence shear stress,mean velocity,turbulence intensity and mean droplet size distribution in a two-phase swirling jet using Phase Doppler Particle Analyzer systems.The instability in a liquid jet was studied,while neglecting liquid viscosity,both with and without rotation by Liao et al.[6].It was shown that the relative velocity between the gas and liquid phases,surface curvature and density ratio increase the interfacial aerodynamic instability.

Ibrahim et al.used linear and non-linear approaches for the analysis of viscous swirling flow in atomizers to predict the breakup length[7].It was shown that the linear theory cannot predict an annular sheet breakup and the developed nonlinear model is necessary to determine the breakup length accurately.

Design procedure of a hollow cone pressure swirl atomizer was suggested by Lacava et al.[8].They examined the procedure by comparing the results for SMD (Sauter Mean Diameter) and spray angle.Nonnenmacher and Piesche [9] presented a calculation model to predict the droplet size depending on the hollow cone injector’s geometry and the volume flow for atomization of Newtonian fluids.The method was successful in various experimental studies while changing the geometry and the transport properties of the fluids,especially the viscosity.

Some researchers focused on modifying the injector geometry.Mohammadi et al.proposed a new injector by adding spiraling rifling-like guides to a conical nozzle in an internal combustion engine.Their numerical results show that the liquid spray from the newly proposed nozzle has a surprisingly larger cone angle,which led to improved mixing of air and fuel droplets and consequently to less pollutant emissions [10].

Spray of heavy fuels is mostly investigated for the application of marine diesel engines.Kyriakides et al.examined the effect of fuel properties on spray atomization by performing simulations using the KIVA CFD code.The results showed that the heavy fuel spray is characterized by comparable values of penetration length,but larger droplet size in comparison to a diesel spray [11,12].

In the previous mentioned articles scientists changed the injector geometry,air and fuel temperature and pressure in order to achieve finer droplets and better spray characteristics for heavy fuel oil (HFO) atomization.Later,some researchers examined the new idea of atomization of an emulsified HFO to improve the spray characteristics.Kim et al.measured the macro-and microscopic spray parameters of HFO emulsified in water.Results showed that emulsification and flashing enhanced the spray characteristics[13].

In this study,a new strategy is presented for the atomization of HFO,by combining two available mechanisms,viz.pressure-swirl and pulsed jet atomization.It is anticipated that adding pulsed instabilities to a pressure-swirl atomizer improve the spray characteristics decreasing the mean droplet diameter.To achieve very small droplets,air-assisted injectors are often a better choice in comparison to pressure-swirl atomizers and they are used widely in entrained flow combustors.Fan et al.combined pressure-swirl with air-assist to obtain a smaller mean droplet diameter [14].However,the assist-air disturbs the flow field near the injector.Moreover,when it is required that to control the oxygen concentration,like in a gasification process,air-assisted injectors could pose difficulties.It is investigated whether pulsed pressure-swirl(PPS) injectors could be an interesting alternative for this kind of applications.

In this paper,experimental results on the droplet size distribution for a pressure-swirl injector for a specific HFO(Mazut)will be presented and discussed.The results are used to set up the numerical simulations to improve the understanding of the influence of pressure and pulsation frequency on the spray characteristics of PPS atomizers,using the open-source Volume-of-Fluid (VOF) code Gerris.Especially the droplet size distribution for different operation modes are of interest,because this may benefit larger-scale models that simulate the entire combustion process.Because of the largely differing length scales involved at the different stages of the atomization process,the embedded adaptive mesh refinement allows the simulation of the entire atomization process with acceptable CPU time and costs.In this article the first stage of the atomization (film formation) is calculated via analytical calculations using data from the presented experimental measurements as input,whereas the next two stages (primary and secondary breakup) are simulated by direct numerical simulations.Information about the injector droplet size distribution can be obtained,which can eliminate modeling of the atomization process in many CFD-DEM (Computational Fluid Dynamics-Discrete Element Method) simulations by using a specific droplet size distribution instead.It can save on the computation time significantly.

The first fluctuation induction in an injection setup was for control of combustion instabilities.In order to reduce the oscillations associated with combustion instabilities [15] the technique has been presented by dephasing heat release rate oscillations,acoustic pressure and stabilizing lifted flames [16],providing practical solutions to pollution problems [17] and high destruction rate incineration fluctuations are the advantages of the active control.

Active instability control used by Hantschk et al.[18] on a model liquid-fired burner,by adding a rapid servo-valve upstream of the fuel injector used to produce the desired flow rate oscillations.Yu et al.[19] used automotive injector solenoid valves to time liquid fuel injections with controlled coherent vortex roll-up in order to control instability.

However these injectors produce poor quality droplets(around 100 μm in diameter).Piezo-electric actuators with high-frequency oscillations (≈10 kHz) have also been employed to control spray formation by exciting the instability modes of liquid sheets [20].This atomizers has been based on the mechanism of highamplitude velocity-modulation,which has been firstly introduced by Dressler [21–23] and combined with pressure-swirl injector by[20,24].However,as the chamber acoustics are a dominant feature in combustion control applications,these frequency oscillations (≈10 kHz) are rarely encountered [24].

Haile[24]have developed and characterized a pulsed fuel injector for active combustion control applications to modulate the fuel flow rate at acoustic frequencies without degrading the droplet size distribution or the distribution of combustible mass in the spray.Longitudinal oscillation of a liquid sheet by parallel air flows was investigated by [25].They showed that liquid viscosity does not affect longitudinal oscillation frequency and wavelength of the liquid sheet in a wide range of liquid Reynolds number.

The main purpose of this research is investigating the impact of velocity modulation on droplet size,made by pressure-swirl injector using heavy fuel oils.In comparison to the previous articles[20,24],in this article the working fluid is not water,a heavy fuel oil (Mazut) was used.Because of the high viscosity of heavy oils,fine droplets cannot be achieved with pressure-swirl injectors.The question is how much droplet size could be reduced by this technique.

The following section outlines the used experimental methods,and analytical and numerical techniques.Following that,we present the experimentally measured injection cone angles and how it influences the injection velocities and film thickness,followed by a characterization of the droplet size predicted by direct numerical simulations.The paper then wraps up with the main conclusions.

2.Methods and Governing Equations

2.1.Experimental measurements

Experimental measurements were conducted to examine the influence of pressure on the macroscopic specifications of HFO atomization.A pressure-swirl atomizer (nonpulsating) was designed according to the HFO (Mazut) characteristics,which are summarized in Table 1,more details on [26].The injector was designed for a mass flow rate of 50 g﹒s-1,a spray angle of 30 degrees and an injection pressure of 2.0 MPa.The design method was obtained from[27]for the design of pressure-swirl atomizers.The dimensions of the pressure-swirl injector are specified in Fig.2.

Mazut is very viscous at ambient temperatures,but the viscosity can be strongly reduced by increasing the temperature (as evident from Table 1),which improves the atomization process.Therefore,the liquid was heated to 110 °C (kinematic viscosity,10 mm2﹒s-1).The setup consists of:(1) Nitrogen gas cylinder,(2)Gas regulator (5.0 MPa),(3) Pressure gauge (10.0 MPa),(4) Liquid input and relief valve,(5) High pressure cylinder,(6) Valve,(7)Temperature gauge,(8)Injector,(9)Spray,(10)High speed camera,computer and image processing code and (11) Backlight and light diffuser.The setup configuration is schematically depicted in Fig.3.The liquid fuel was pressurized in the high pressure cylinder with high pressure nitrogen.The pressure and temperature were measured at the injector entrance.A high-speed PCO camera was used to capture images using an exposure time of 9 μs and Nikon macro lens (100 mm focal length).

Table 1 Mazut properties

2.2.Analytical calculations

The direct numerical simulations that will be described later on requires a set of boundary conditions for the inlet of the HFO into the computational domain,notably the (angular) velocity and liquid film thickness at the tip of the pressure-swirl atomizer.Inner part of the injector could be also simulated in conjunction with the outer part.But it would lead to extensive simulations.So in this study,experimental measurements (injection pressure difference and mass flow rate) and an analytical model used to calculate the required parameters at inner part of the injector.The calculated parameters are the injector velocity and liquid film thickness at the tip of the pressure-swirl atomizer.

Analytical calculations were performed using the LISA (Linearized Instability Sheet Atomization)model[28].Due to the swirling flow inside the pressure-swirl injector,a liquid film along the injector walls is formed.According to the model,the film thickness h0can be computed by:

where .m [kg﹒s-1] is the mass flow rate through the injector,ρl[kg﹒m-3] is the liquid density,and d0[m] is the injector diameter.Moreover,u [m﹒s-1] is the axial velocity component of the film at the injector exit and can be calculated with the injection half cone angle θ,and the total velocity U.The total velocity U[m﹒s-1]is then calculated from the following equations [28]:

Fig.2.a)The left part is the swirl chamber and the right side is the main injector b)Schematic of geometry and dimensions of the injector(geometry is not depicted with real dimensions).

where Δp[Pa]is the pressure difference across the injector and kv is the discharge coefficient:

The axial and radial velocity components,u and w can be evaluated from the following equations:

where it has also been assumed that the tangential velocity component equals the radial velocity component [29].

The injector exit diameter d0is known from the nozzle geometry,shown in Fig.4 (d0=1.5 mm in this study),also the injection pressure difference Δp and injection half cone angle θ should be determined from the experimentally obtained data.With these inputs one can calculate the three components of velocity and the film thickness at the tip of nozzle.In this work these are used as input for the boundary conditions for the numerical calculations.

This is a simplification which was made because of the large iterative computation,in order to avoid from simulating the inner part of the injector.The simplification is not unrealistic and also was examined successfully in other researches [29].Moreover using analytical calculations feeding with real experimental data could be even much accurate than pure numerical calculations.Also because of that the pulsed pressure swirl injector is only a concept in this stage,the detailed characteristics and the geometry of the inner parts of the injector are not specified to be simulated numerically.

For the simulations the domain size is set at 67d0×67d0-×100d0,where the injector is located at the middle of the smallest side of the cube,see Fig.4.All boundary conditions are BoundaryOutflow conditions,i.e.a fixed value condition for the pressure(0 Pa)and a zero gradient condition for other parameters.

2.3.Numerical simulation

For the numerical calculations the open source code Gerris was used [30–32].This code uses the Volume of Fluid (VOF) approach(with geometric interface reconstruction) to account for the twophase flow.The model solves the dynamic two-phase flow using the incompressible 3D Navier-Stokes equations:

g is gravity acceleration and u=(u,v,w)is the fluid velocity,ρ≡ρ(x,t) is the fluid density and μ≡μ(x,t) is the dynamic viscosity which are computed in each computational cell based on the volume fraction c.

In these equations the index refers to the two fields.S is the deformation tensor,defined as S=,where σ is t he surface tension,κ and n are the curvature and normal to the interface,respectively.The Dirac δ-function of δsn concentrates the normal forcing term directly on the interface.The Volume of Fluid advection scheme,balanced-force surface tension discretization and height function curvature estimation were used to accurately estimate the surface tension forces.The Adaptive Mesh Refinement method is used to efficiently reduce the computational cost.More details the interested reader is referred to [30–32],in addition,Baltussen et al.compared three different surface tension models for the Volume of Fluid method [33].

3.Results and discussion

3.1.Experimental measurements

Details of the measurements are presented in Table 2.Mass flow rate and injection pressure difference were measured directly with the maximum relative uncertainty of 2.5% and 1.25% respectively.The experiment was repeated many times to reduce the statistical uncertainty and to look for possible influence of instabilities.The uncertainty of the multiple measurements was calculated by standard deviation method.The experimental images were taken by a shadowgraphy technique:the hollow cone injection is shown in the picture as if it is a solid.If the injection were illuminated by a light sheet,it would show as a hollow cone.The half angle was measured by analyzing the raw images which are shown in Fig.5 (using ImageJ free software).The breakup length can be measured by knowing the pixel width and length from the image calibration and measuring breakup length in terms of pixels.The uncertainty for each test was calculated dividing a pixel length by the breakup length.Breakup length uncertainty is calculated to be in the range of 0.15%-0.4%.Similar procedure used to measure half cone angle (applying trigonometry by knowing the dimensions in pixels)and the half cone angle uncertainty is calculated to be in the range of 0.5%-1%.

Fig.4.Injector and sheet breakup schematic and dimensions

Table 2 Detailed data for the pressure-swirl injector at different injection pressures (1 bar=105 Pa)

The experiment was also conducted with water with the same pressure-swirl injector as the one used for the experiments with Mazut.Various stages of the spray development can be discerned as a result of the water injection pressure as shown in Fig.6.At lower pressures (Δp≤0.1 MPa,Wel≤240) a cone forms at the orifice,which contracts due to surface tension forces into a closed bubble.In other words,the spray angle decreases with the distance from the injector (Onion stage).By increasing the water pressure,the fluid momentum force increases and overcomes the surface tension forces,so the bubble opens into a hollow tulip,ultimately breaking up into fairly large drops(Tulip stage).Further increasing the liquid pressure leads to forming a fully conical sheet which disintegrates to ligaments and then fine droplets (developed stage).

Surface tension resists the formation of new surface area,basically opposing the key purpose of the atomization process.Whether or not a droplet breaks up further due to surface instabilities can be identified using the gas Weber number (d0/σ,where ρa(bǔ)ir[kg﹒m-3]is the air density,Vrelthe relative velocity between droplet and air [m﹒s-1],d0[m] the injector diameter,and σ surface tension of the droplet [N﹒m-1]).Some previous studies have determined different regimes for atomization of pressure swirl injectors [34].For Mazut in this case according to Table 2 at Weg<19.1 (Δp<0.3 MPa),sheet breakup regime and no fine atomization was observed [Fig.5(a)-(c)],which is not suitable for using in a combustion chamber.Droplets are not small enough to create a flammable environment in this Weber number range.For 19.1100 at the shear breakup regime.

Fig.5.Mazut atomization images at various pressure differences.(Pressure differences indicate the difference between the ambient pressure and the fuel pressure,1 bar=105 Pa)

Fig.6.Water atomization images at increasing pressure differences (1 bar=105 Pa).

3.2.Numerical simulation

For the numerical code,a comprehensive explanation of the numerical schemes and discretization methods,conservation errors and validation against the linear instability theory can be found in [30–32].A grid convergence study was conducted,increasing the maximum grid refinement to increasingly higher levels.Table 3 is showing the corresponding results.Using this table a judgment can be achieved about the suitable mesh sizes.Also the CPU time consumption for each grid system was presented.The data was presented for the case of Δp=4.0 MPa,where finer droplets are produced and consequently finer mesh was needed.Error was calculated by comparing SMD for each mesh results with the finest mesh results.The results show that the level of refinement which creates the minimum mesh size of 12 μm is the best to gain reliable results with the minimum amount of CPU time.

For the initial conditions,a hollow cylinder was assumed as the liquid phase with the diameter and the length equal to d0and the thickness equal to the film thickness determined from the measurement data listed Table 2.The initial refinement applied at the border between the two phases was set at level 8.

Table 3 Grid study details (for Δp=4.0 MPa)

3.3.Validation

The code was validated using experimental data from[35],diesel fuel.A comparison between experimental data and numerical simulation data is shown in Fig.7.The experimental data was also fitted with a chi-square distribution of order 15 with a maximum occurrence diameter of 13 μm.According to the level of refinement(13),the smallest droplet size was about 10 μm,and consequentlysmaller droplets sizes could not be captured by these VOF simulations,and we thus have to discard the left-most part of the graph.Improving the level of refinement did not have a significant impact on the results.Overall,a good agreement was obtained between the numerical simulation and experimental data,but a small discrepancy between the measurements and simulations appears between 20 μm and 40 μm.This may be attributed to the fact that the temperature of the liquid (and thus the viscosity and surface tension)are considered to be constant in the simulations,whereas in the experiments especially the smaller droplets could exchange heat with the environment and experience these properties dynamically.

3.4.Pressure swirl simulations

Fig.8 shows snapshots of the results of the numerical simulations for different injection pressure differences.For Δp=0.1 MPa,the swirling velocity is very small,and it is observed that the liquid spray regime is very similar to a liquid jet[Fig.5(a)].This is also confirmed with the analytical calculations listed in Table 2 for Δp=0.1 MPa,since the film thickness is equal to the nozzle radius,which indicates that the liquid develops as a round jet.

Fig.7.Comparison of droplet size distributions obtained from numerical simulations and experimental data,diesel fuel,pressure difference 2.0 MPa and 0.25 mm nozzle diameter [35]

Fig.8.Snapshots from the numerical simulations for different injection pressure differences (1 bar=105 Pa).

When operating with a larger pressure difference Δp=0.2 MPa,the tangential velocity is increased which causes the jet to twist,resulting in a distorted ribbon near the nozzle,see Fig.8(a).The waves that appear on the sheet surface,which can also be seen on the edges of the jet in Fig.5(b),finally break the sheet into horizontal ligaments around one wave length in width,which can be observed on the edges of the jet shown in Fig.8(a).Moreover,the increase in pressure and consequent increase in tangential velocity leads to a wider spray angle.This distorted ribbon spray pattern remains up to Δp=0.5 MPa [Fig.8(b)],clearly showing the sinusoidal waves on the edges of the sheets [Fig.5(b)-(d)].

At even higher pressure differences[Fig.8(c):1.0 MPa;Fig.8(d):2.0 MPa) the changes in the spray angle become even more pronounced.The angle is wider adjacent to the nozzle and contracts at some distance (similar to Tulip stage for water atomization,Fig.6(b).But the onion stage [Fig.6(a)] was not fully occurred in Mazut atomization and that is because of the higher viscosity of heavy fuel in comparison to water.The formation of the Onion stage is because of the presence of surface tension force.In fact,after that the fluid leaves the nozzle and forms a thin sheet,the spray angle decreases due to the surface tension force.The angle reduction continues as far as the sheet is again collected at a point[Fig.6(a)].Inertia and viscose forces are in opposition to the surface tension force in atomization process.As the pressure difference and viscosity increases,the effect of the surface tension force diminishes.So in the case of HFO,due to the fluid high viscosity,the Onion stage may not occur.

The increase in injection pressure leads to a higher mass flow rate,a smaller sheet thickness(Table 2)and a smaller surface wave length (Fig.5).

For higher pressures the Onion and Tulip stages are no longer visible due to high momentum force among viscosity and surface tension forces[Fig.8(d),(f)]similar to water atomization[Fig.6(c)].

It is shown that the injector performance strongly depends on the injection pressure.Since each injector was designed for a specific injection pressure that is most often used,some detailed information is only presented for the design injection pressure,such as the droplet size distribution.Information about the injector droplet size distribution can eliminate modeling of the atomization process in many CFD-DEM simulations,and allow these simulations to use a specific droplet size distribution instead,saving significantly on the computation time.Fig.9 shows the computed droplet size distribution for the designed pressure-swirl atomizer at Δp=2.0 MPa.The figure also shows a Rosin-Rammler distribution,given by Eq.(9),that was fitted to the numerical results:

where,d represents the droplet diameter (μm) and Ydthe volume fraction of the droplets with a diameter greater than d.From the simulation data for Δp=2.0 MPa,d and n were found equal to 77.1 μm and 3.2,respectively.Also the computed droplet size distribution could be fitted well by a modification in Rosin-Rammler as part of a Bell curve Gausian equation:

where a,b and c are constants and equal to 1.041,21.73 and 37.56,respectively (RMS=0.99,deviation from numerical results).

Fig.9.Droplet diameter distribution (blue bars) and the fitted Rosin-Rammler distribution (green curve) for Δp=2.0 MPa.

In addition to the droplet size distribution,the Sauter Mean droplet Diameter(SMD)can also indicate the injector’s performance to produce a fine spray.The simulations show that the SMD decreases when operating at higher injection pressures.The decrease in SMD is much more pronounced at lower pressure differences in comparison to higher pressure differences,as shown in Fig.10.The results are in agreement with previous studies on pressure swirl atomizer which indicated that:SMD ∝(Δp)-cte.[36].Effective spray half angle variation with injection pressure differences is also presented in Fig.10.Overall,a good agreement was obtained between the numerical simulation and experimental data.The SMD results were compared with the correlations of experimental results conducted by [37],and a good agreement was observed especially at the higher injection pressures(the uncertainty for SMD correlation was calculated less than 1.8%,according to the procedure of [38]).

Fig.10.The pressure-swirl injector droplet Sauter Mean Diameter(SMD)and spray half angle for different injection pressure differences (1 bar=105 Pa).

In many applications such as reacting chambers,finer droplets improve the performance,especially when the injected liquid is relatively viscous and resists atomization.A new idea was studied in order to achieve finer droplets by adding additional disturbances to the liquid flow before the injection.

The pressure-swirl injector is investigated by applying pulses of 5 kHz,10 kHz,20 kHz and 40 kHz.Sinusoidal pulses were applied with 10 percent amplitude in the injection velocity.Numerical results show that adding pulses to the pressure-swirl atomizer does not change the spray angle,as can be discerned from the snapshots shown in Fig.11,while for higher pulse frequencies shorter breakup lengths are observed.Moreover,a smaller mean droplet size was found at higher frequencies.Comparing the results of the cases without pulses and 5 kHz reveals a 33 and 9 percent decrease in breakup length and SMD,respectively.Increasing the pulse frequency to 40 kHz leads to respectively a 78 and 17 percent decrease in breakup length and SMD in comparison to the case without pulses,as presented in Fig.12,and consequently a significant improvement of the evaporation process.

Fig.12.Pressure-swirl injector breakup length(white bars)and droplet SMD(gray bars) for different pulse frequencies,Δp=2.0 MPa

Droplet size is a key parameter in designing entrained flow reactors.In the application of a chemical reactor,a good injector should produce fine droplets that can evaporate fast in the hot medium prevailing in the reactor.HFO is very viscous at ambient temperature which renders the atomization rather difficult.The results of the pressure-swirl atomizer showed that adding momentum by swirl velocity helps atomizing the fuel.In some applications (like gasification of HFO),air-assist or twin fluid injectors could be used,but this type of injectors would also impact the air to fuel concentration ratio,which needs to be controlled carefully in gasification applications.Adding pulses to the pressureswirl atomizer helps atomizing HFO to finer droplets,like in airassist injectors,but without the problems associated with the disturbance of the air to fuel concentration.Therefore,the application of the pulsed pressure-swirl atomizer in entrained flow gasifiers looks very promising.

Fig.11.Snapshots of the simulations for the pressure-swirl,pulsed pressure-swirl and pulsed jet atomization for different pulse frequencies with a pressure difference Δp=2.0 MPa

4.Conclusions

A pressure-swirl injector was designed for an experimental study on the atomization of very viscous (Mazut) HFO.The setup has been used to perform optical measurements of pressureswirl atomization of HFO.The Gerris Volume of Fluid code was used to investigate the atomization process for different pressure differences.Moreover,the benefits of operating the pressure-swirl injector with pulses has been investigated numerically and the influence of the pulse frequency has been quantified.The numerical simulations were set up (viz.boundary conditions) using a standard atomization model for pressure-swirl injectors combined with data obtained from the experiments.The numerical simulations have been used to analyze the droplet size distributions and breakup lengths in detail.A Rosin-Rammler droplet size distribution was fitted for the injector (d-=77.1 μm and n=3.2) at given design conditions.For a standard pressure-swirl injector,finer droplets,shorter breakup length and a wider spray angle was found for operation at larger injection pressure differences,where the influence of the injection pressure was largest at lower pressures.

The breakup length decreases at higher injection pressures,the addition of pulses was investigated to further decrease the breakup length and mean droplet size,thereby introducing an innovative PPS atomization device.This study introduces a new method for dispersing HFOs,combining the pulsed jet and pressure-swirl methods.With numerical simulations it is shown that the new injector can atomize HFO to smaller droplets with shorter breakup lengths.Moreover,even smaller droplet sizes and shorter breakup lengths were predicted for higher pulse frequencies.Using pulses with a frequency up to 40 kHz about a 78 and 17 percent decrease in the breakup length and SMD was found respectively,in comparison to the case without pulses,while no significant changes were observed in the spray angle.The addition of pulses to the pressureswirl atomizer thus helps atomizing HFO to finer droplets,like in air-assist injectors,but without affecting the air to fuel concentration.Therefore,the application of pulsed pressure-swirl atomizers looks particularly promising for entrained flow gasifiers.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We thank Prof.Dirk Roekaerts,Prof.Niels Deen,Prof.Martin van Sint Annaland and Dr.Ivo Roghair for the useful discussions.