Modelado Numerico y Experimental.pdf

Computers & Fluids 42 (2011) 37–43

Contents lists available at ScienceDirect

Computers & Fluids

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

Experimental and numerical modeling of the gas atomization nozzle for gas flowbehavior

Ozer Aydin ⇑, Rahmi UnalDumlupinar University, Department of Mechanical Engineering, Kutahya, Turkey

a r t i c l e i n f o a b s t r a c t

Article history:Received 3 December 2009Received in revised form 21 September2010Accepted 28 October 2010Available online 3 November 2010

Keywords:Gas atomizationCFDFlow separationMelt tip pressure

0045-7930/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.compfluid.2010.10.013

⇑ Corresponding author. Tel.: +90 274 2652031; faxE-mail address: [email protected] (O. Ay

Gas atomization is a widely used process for manufacturing of fine metal- and alloy-powder. To ensure astable process with high yields of metal powder, the negative pressure at the melt delivery tube tip baseand no flow separation conditions are necessary for a good atomization process. An important feature ofthese jets is that flow separation may occur over the outer surface of the liquid delivery tube for someconditions. Flow separation cause solidification and accumulation of metal, leading to a shape alterationof the liquid delivery tube in gas atomization process. Using computational fluid dynamics (CFD) soft-ware, a parametric study was conducted to determine the effects of atomizing gas pressure on the meltdelivery tube tip base pressure and flow separation. Atomization gas pressures of 1.0, 1.3, 1.7, 2.2, and2.7 MPa were used in the CFD model to initialize the pressure in gas inlet. CFD simulations were per-formed and the modeling results were compared with experimental data. These results showed thatthe CFD modeling can be used for the estimation of the melt tip base pressure of the nozzle. It is foundthat the flow separation formation is strongly dependent on the atomizing gas pressure.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction molten metal stream by impinging high-pressure gas jets [2]. It was

Atomization relies on a melt and disintegration of that melt intodroplets that freeze into particles. The use of air, nitrogen, helium, orargon as a gas for breaking up a molten stream is termed gas atom-ization. The idea is to transfer kinetic energy from a high velocity jet-gas expanded through a nozzle, to a stream of liquid metal, resultingin fragmentation and break up into metal droplets. Gas atomizationof liquid metal using close-coupled nozzle system is used to producemetal powders with characteristics that cannot be achieved withother powder production methods. It is often used to fabricate fine,highly spherical powders typically used in applications requiringdense packing and good flow characteristics. Alloy powders thatcannot be made with chemical techniques are readily made usinggas atomization. In addition, the rapid cooling rates inherent in theprocess show potential to produce amorphous powders with uniqueproperties. These capabilities ensure that gas atomization will con-tinue to be an important process in powder metallurgy. The atomiza-tion nozzle assemblies can be of two types: free-fall, or closecoupled. In free-fall atomizers, the stream of molten metal is allowedto fall unrestricted until it interacts with the gas jets. In close-cou-pled atomizers, the stream of molten metal is delivered by a ceramicconduit (named liquid-delivery-tube) to the interaction zone withthe gas jets [1]. Close-coupled gas atomization is a technique widelyused for the production of fine metal powders by the disruption of a

ll rights reserved.

: +90 274 2652066.din).

found that one of the most important feature of an efficient nozzle isthat the highest negative pressure at the melt delivery tube ensuresoptimum efficiency for a nozzle geometry [3]. For that reason it isimportant to determine the melt delivery tube tip base pressurefor a designed geometry by computational fluid dynamics.

The occurrence of boundary layer separation can essentially af-fect the efficiency of an entire system [4]. The occurrence of separa-tion, which is a function of atomization pressure and liquid deliverytube extension, has been suggested to cause liquid metal to be drawnfrom the end face of the liquid delivery tube into its outer surface,where it is exposed to the very cold expanding gas of the annular walljet. The extreme temperature difference between the metal and thegas promotes the solidification and accumulation of metal, leadingto a shape alteration of the liquid delivery tube. Typically, this se-quence of events induces a freeze off that ends the atomization pro-cess prematurely. Therefore, this separation is detrimental to theprocess of gas-metal atomization and should be avoided at all costs.

The goal of the computer simulations is to determine if the de-sired flow features of an efficient gas atomization process can bepredicted by CFD. Using a CFD software, numerical modeling of agas atomization process is studied widely. The Reynolds’s averagedNavier Stokes equations coupled with different turbulence modelswere solved using the finite volume code The Navier Stokes MultiBlock (NSMB). Numerical modeling is believed to be an effectiveapproach to examine the underlying thermal-physics of powderatomization and numerous mathematical models have been devel-oped for powder atomization over last decade [5–8]. Nowadays,

Nomenclature

c speed of soundE total energy~F external body forcesH enthalpyM mach numberMw molecular weightp static pressurepop operating pressurer radial coordinateR universal gas constant~q heat flux vectorS effective temperatureSm mass added to the continuous phase from the dispersed

second phase

T static temperaturevx axial gas velocityvr radial gas velocityvz swirl velocityx axial coordinate

Greek symbolsq gas densityq~g gravitational body force��s stress tensorl molecular viscosityc ratio of specific heats

38 O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43

due to the rapid development of computer hardware, directnumerical simulation is becoming more and more applicable inthe field of multi-fluid flow research [9]. The flow of compressibleatomizing gas jets has been investigated by many researchers byusing numerical modeling and simulation of the gas-only case,i.e. ignoring, for now, the interaction with the liquid metal [10].

In this study, gas-only flow CFD simulation was modeled in aclose-coupled gas atomizer. The gas field was generated with acommercial CFD code, FLUENT 6.3. CFD simulations were per-formed and the modeling results were compared with experimen-tal data. A parametric study was conducted to determine theeffects of atomizing gas pressure on flow separation.

2. Models and numerical formulation

A modeling study of the effects of atomization gas pressure onthe gas-only flow in the atomization system has been performedusing the CFD code FLUENT 6.3. FLUENT is a state-of-the-art com-puter program for modeling fluid flow and heat transfer in complexgeometries [11]. The commercial CFD software package, FLUENT6.3, which is based on the finite volume approach, was used forsolving the set of governing equations. The mass conservation orcontinuity equation, the momentum conservation or Navier–Stokes transport equations and the energy conservation equationare numerically solved. For 2D axisymmetric geometries, the con-tinuity equation is given by [11]

@q@tþ @

@xðqvxÞ þ

@

@rðqv rÞ þ

qv r

r¼ Sm ð1Þ

where q is the gas density, x is the axial coordinate, r is the radialcoordinate, vx is the axial gas velocity and vr is the radial gas veloc-ity. The source Sm is the mass added to the continuous phase fromthe dispersed second phase (e.g., due to vaporization of liquid drop-lets) and any user-defined sources. Eq. (1) is the general form of themass conservation equation and is valid for incompressible as wellas compressible flows.

For 2D axisymmetric geometries, the axial and radial momen-tum conservation equations are given by

@

@tðqvxÞ þ

1r@

@xðrqvxvxÞ þ

1r@

@rðrqv rvxÞ

¼ � @p@xþ 1

r@

@xrl 2

@vx

@x� 2

3ðr �~vÞ

� �� �þ 1

r

� @

@rrl @vx

@rþ @v r

@x

� �� �þ Fx ð2Þ

and

@

@tðqv rÞ þ

1r@

@xðrqvxv rÞ þ

1r@

@rðrqv rv rÞ

¼ � @p@rþ 1

r@

@xrl

@v r

@xþ @vx

@r

� �� �þ 1

r@

@rrl 2

@v r

@r� 2

3ðr �~vÞ

� �� �

� 2lv r

r2 þ23

lrðr �~vÞ þ q

v2z

rþ Fr ð3Þ

where

r �~v ¼ @vx

@xþ @v@rþ v r

rð4Þ

where p is the static pressure, ��s is the stress tensor, l is the molec-ular viscosity, vz is the swirl velocity and q~g and ~F are the gravita-tional body force and external body forces (e.g., that arise frominteraction with the dispersed phase), respectively.

Compressibility effects are encountered in gas flows at highvelocity and/or in which there are large pressure variations. Whenthe flow velocity approaches or exceeds the speed of sound of thegas or when the pressure change in the system (Dp = p) is large, thevariation of the gas density with pressure has a significant impacton the flow velocity, pressure, and temperature. For compressibleflows, the ideal gas law is written in the following form:

q ¼pop þ p

RMw

Tð5Þ

where pop is the operating pressure defined in the operating condi-tions panel, p is the local static pressure relative to the operatingpressure, R is the universal gas constant, and Mw is the molecularweight. The temperature, T, will be computed from the energyequation.

Energy conservation equations can be written as:

q@E@tþ qr � ðH ~vÞ ¼ rð��s ~vÞ � r �~q ð6Þ

where E is the total energy, H is the enthalpy and ~q is the heat fluxvector.

Compressible flows can be characterized by the value of theMach number:

M � v=c ð7Þ

Here, c is the speed of sound in the gas:

c ¼ffiffiffiffiffiffiffiffifficRT

pð8Þ

and c is the ratio of specific heats (cp/cv).

Fig. 1. Schematics of the annular type nozzle and set-up for measuring the meltdelivery tube tip pressure.

Fig. 3. The grid structure of the concergent–divergent region of the nozzle inaxisymmetric 2D flow field.

O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43 39

Sutherland’s viscosity law resulted from a kinetic theory bySutherland using an idealized intermolecular-force potential. Theformula is specified using two or three coefficients. In this study,Sutherland’s law with three coefficients was used.

Sutherland’s law with three coefficients has the form:

l ¼ l0TT0

� �3=2 T0 þ ST þ S

ð9Þ

where l is the viscosity in kg/ms, T is the static temperature in K, l0

is a reference value in kg/ms, T0 is a reference temperature in K, andS is an effective temperature in K, called the Sutherland constant,which is characteristic of the gas. For air at moderate temperaturesand pressures, l0 = 1.7894 � 0�5 kg/ms, T0 = 273.11 K, andS = 110.56 K.

Schematics of the annular type nozzle and set-up for measuringthe melt delivery tube tip pressure are shown in Fig. 1. The nozzlehas a throat area of 8.5 mm2, protrusion length of the melt deliverytube is 5 mm and apex angle of this nozzle is 26�. The atomizationchamber and the enclosed flow can be treated with rotational sym-metry. Although the atomizer geometry has axial symmetry,resulting turbulent flow pattern could not have complete symme-try. One could use a 2D axis-symmetric model instead of the full3D flow simulation, with huge savings in both computing timeand resources. For that reason only half of an axial section of theatomizer could be used in a 2D simulation. The calculation wasperformed on an axisymmetric 2D field as the atomization nozzlesystem has a symmetrical shape.

Prior to the CFD calculations, the geometry was defined and agrid was generated using GAMBIT 2.4.6. GAMBIT is the preproces-sor for geometry modeling and mesh generation [12]. Due to rota-

Fig. 2. Schematic drawing of the computational fi

tional symmetry of the nozzle geometry, a 2D simulation isperformed on one half of an axial section of the atomizer chamber.The Computational domain and grid structure of the convergent-divergent region of the nozzle in axisymmetric 2D flow field isshown in Fig. 2. Because the gap at the throat of the nozzle is0.2 mm, very fine mesh grid structure is applied in the conver-gent-divergent section of the nozzle. The simulation starts with acoarse grid which has a cell number of 344196 (mesh 1). Then fur-ther refinement is made with a cell number of 487917 (mesh 2) tomake sure that the grid is fine enough to capture the high-pressuregas dynamics. The final grid has a cell number of 629222 (mesh 3)which demonstrates that extensive refinement is achieved in thezones of high pressure gradients and near the wall. More gridswere used in regions of large property gradients and in the vicinityof the walls (Fig. 3). For all cases studied in this paper, triangularelements were generated. The influence of the mesh refinementon the distribution of the melt tip pressure along line 1, line 2and line 3 was tested and it was defined that the solution wasnot affected by the mesh number above 629222 (Fig. 4a–c). Thestructured grid is able to capture all the flow characteristics andthe flow predictions are consistent with the compressible flow the-ories and literature [13]. Additional grid independent study con-firms that further increasing the grid density has not given moreaccurate results or any change of flow patterns. To make sure theaccuracy of the results and to avoid any problems encounteredfor solution convergence, very small time step (10e-6) is deployedwhich has significantly increased the computational time for eachsimulation.

In order to measure the delivery tube tip pressure, three lineswere used with a one mm interval along the base of the melt deliv-

eld showing the geometry of the boundaries.

The points on Line 10 1 2 3 4 5 6 7 8 9 10 11

0 1 2 3 4 5 6 7 8 9 10 11

0 1 2 3 4 5 6 7 8 9 10 11

Pres

sure

(Pa

)

1.00e+5

1.01e+5

1.02e+5

1.03e+5

1.04e+5

1.05e+5

1.06e+5

1.07e+5

1.08e+5

1.09e+5

1.10e+5Mesh 1Mesh 2Mesh 3

Atmospheric Pres.

The points on Line 2

Pres

sure

(Pa

)

1.00e+5

1.01e+5

1.02e+5

1.03e+5

1.04e+5

1.05e+5

1.06e+5

1.07e+5

1.08e+5

1.09e+5

1.10e+5

Mesh 1Mesh 2Mesh 3

Atmospheric Pres.

The points on Line 3

Pres

sure

(Pa

)

1.00e+5

1.01e+5

1.02e+5

1.03e+5

1.04e+5

1.05e+5

1.06e+5

1.07e+5

1.08e+5

1.09e+5

1.10e+5

1.11e+5

Atmospheric Pres.

Mesh 1Mesh 2Mesh 3

a

b

c

Fig. 4. The influence of the mesh size on the distribution of the melt tip pressure onthree different lines. (a) Line 1, (b) Line 2, (c) Line 3.

Fig. 5. The position of the pressure lines at the melt tip base of the nozzle for thedetermination of melt tip base pressure by CFD.

Table 1Nitrogen properties.

C (J/kg K) 1040.67Thermal conductivity (W/m K) 0.0242Viscosity (kg/ms) Sutherland lawMolecular weight (kg/kg mol) 28.0134

Fig. 6. Comparison of the melt tip base pressure for experimental and CFD solution.

40 O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43

ery tube as shown in Fig. 5. These lines were 1.5 mm long from thecenter line of the delivery tube along the base surface. There wereten measuring points on the each line. The average of the all pres-sure data obtained from the measuring points is used as the deliv-ery tube tip pressure. The melt delivery tube inner diameter was3 mm in the experiments. In order to obtain the pressure valuesin the simulation the line lengths were selected 1.5 mm in y axisdirection. Three pressure lines shifted 1 mm from each other wereused to improve the accuracy of the obtained pressure values fromthe simulation. The average of predicted pressure values of allpoints was called as melt tip pressure and was compared withexperimental value.

CFD simulations were performed using a pressure based steady-state segregated implicit solver. Flow turbulence was simulatedusing the realizable k–e model with enhanced wall treatment,the latter being one of the available tools in FLUENT to modelthe near-wall region. Default values for the model constants wereapplied. Some researchers have used k–e model to stimulate thenozzle flow behavior by CFD. Those studies can give some informa-tion about the designed nozzle for the improvements of the de-signs. By treating the turbulent flow with the k–e model, Xuet al. numerically investigated the effects of the protrusion of theliquid delivery tube on the compressible gas flow [14]. The fluidis nitrogen, modeled as compressible gas with ideal gas law fordensity and the gas constants given in Table 1. According to NIST

Fig. 7. Velocity of the gas according to the atomizing gas pressure.

Fig. 8. The schematic of the gas flow separation by vectoral presentation.

O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43 41

database [15], the compressibility factor for Nitrogen at 11 atmo-spheric pressure and temperature of 300 K is 0.998, in that caseit is reasonably accurate to use the ideal gas law in this simulation.The boundary conditions are illustrated in Fig. 1. Atomization gaspressures of 10, 13, 17, 22, and 27 bars were used in the CFD modelto initialize the pressure in gas inlet. The gas inlet temperature wastaken as 300 K. The gas outlet was defined as pressure outlet andpressure was defined as atmospheric pressure. The temperaturesof all the walls were taken 300 K.

Because of the nonlinearity of the equation set being solved byFLUENT, it is necessary to control the change of /. This is typicallyachieved by under-relaxation, which reduces the change of / pro-

Fig. 9. The x component of the wall shear stress on the oute

duced during each iteration. In a simple form, the new value of thevariable / within a cell depends upon the old value, / old, the com-puted change in /, D/, and the under-relaxation factor, a, as follows:

/ ¼ /old þ aD/ ð10Þ

To avoid a divergent solution, under-relaxation parameters forpressure, momentum, k and e is set to 0.2, 0.5, 0.5 and 0.5 respec-tively, for compressible flow calculations [11]. The difficulties asso-ciated with solving compressible flows are a result of the highdegree of coupling between the flow velocity, density, pressure,and energy. This coupling may lead to instabilities in the solutionprocess and, therefore, may require special solution techniques inorder to obtain a converged solution. For that reason, the energyequation for the first 50 iterations was turned on and the energyunder-relaxation at 1.0 was used. The pressure under-relaxationand momentum under-relaxation were selected as 0.4 and 0.3,respectively. After the solution stabilized, the energy equationwas turned on and pressure under-relaxation increased to 0.7.

A standard discretisation scheme was used for the continuityequation while a first-order upwind scheme was used for both theturbulence kinetic energy equation and the turbulence dissipationrate equation. To reduce numerical diffusion, a second-order upwindscheme was selected for the discretisation of the momentum and en-ergy equations. The relationship between velocity and pressure cor-rections was calculated using the SIMPLE algorithm. The SIMPLEalgorithm uses a relationship between velocity and pressure correc-tions to enforce mass conservation and to obtain the pressure field.In this approach, higher-order accuracy is achieved at cell facesthrough a Taylor series expansion of the cell-centered solution aboutthe cell centroid. For triangular and tetrahedral grids, since the flowis never aligned with the grid, generally more accurate results can beobtained by using the second-order discretisation. The governingequations for flow, turbulence and energy were solved iterativelyuntil convergence was obtained. The convergence of the calculationwas judged by the mass flow rate difference between inlet and outletboundaries. Moreover a solution was considered converged whenthe scaled residuals have been dropped to six orders of magnitudefor the energy equation.

3. Results

In order to validate the computational procedure, comparison ofthe present calculations with the experimental data of Unal

r surface of nozzle for different atomization pressures.

Fig. 10. Pressure contours (Pa) occur on the outer surface of nozzle for differentatomization pressures. (a) 1.0 MPa, (b) 1.3 MPa, (c) 1.7 MPa, (d) 2.2 MPa, (e) 2.7 MPa.

Fig. 11. Velocity vectors (m/s) show the recirculation zone at the melt tip base ofthe nozzle.

42 O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43

[3,16,17] was performed. The comparison of the theoretical dataobtained by CFD simulation with the experimental data for the

melt delivery tube base pressure was given in Fig. 6. The trendsof the melt tip base pressure graphs are the same for the theoret-ical CFD solution and experimental data. The difference betweenthem was nearly same for all the nozzle gas pressures. Experimen-tal values are only 11–15% smaller than the theoretical CFD values.Espina and Piomelli found in their study that the numerical calcu-lations generally miss the prediction of the aspiration pressure by10–20% (a result consistent with similar numerical data obtainedfrom supersonic base flow simulations) [2]. Aspiration pressurevalues were observed at the melt tip base pressure in the experi-mental measurements. Pressure measurement at the tip of themelt delivery tube during the atomization process is not possible.For that reason, before atomizing each melt, the gas only aspirationprofile of the atomizer nozzle was measured with a digital pressuretransducer to determine the pressure of melt delivery tube tip [17].Aspiration pressure value was observed at the simulation by CFD.All the simulation melt tip base pressure values are greater thanthe experimental pressures. For that reason, aspiration pressurevalues are not observed in the simulations. However, the simula-tions capture accurately the trends observed in the experimentaldata better than previous studies by Espina and Piomelli, as shownin Fig. 6. This result showed that the CFD modeling can be used forthe estimation of the melt tip base pressure of the nozzle.

The theoretical gas velocity of the nozzle is given in Fig. 7. Thegas velocity has a maximum value of 663 m/s at 2.7 MPa gas pres-sure and minimum value of 631 m/s at 1.0 MPa gas pressure. Thisshows that the gas velocity could not be increased in the order ofpressure increase. For that reason, for an efficient gas atomizationprocess the geometry can give the maximum gas velocity for thesame mass flow rate of the gas. Hence, efficiency of a gas atomiza-tion nozzle greatly depends on the geometry of the nozzle.

An important feature of these jets is that flow separation mayoccur over the outer surface of the liquid delivery tube for someconditions (see Fig. 8). The occurrence of separation, which is afunction of atomization pressure and liquid delivery tube exten-sion, has been suggested to cause liquid metal to be drawn fromthe end face of the liquid delivery tube into its outer surface, whereit is exposed to the very cold expanding gas of the annular wall jet.The extreme temperature difference between the metal and thegas promotes the solidification and accumulation of metal, leadingto a shape alteration of the liquid delivery tube. Typically, this se-quence of events induces a freeze off that ends the atomizationprocess prematurely. Therefore, this separation is detrimental tothe process of gas-metal atomization and should be avoided atall costs.

As shown in Fig. 9, the large, adverse pressure gradient inducedby the shock causes the boundary layer to separate. Flow reversalis indicated here by negative values of the x component of the wallshear stress. Pressure contours occured at the outer surface of noz-

Fig. 12. The flow separation formation at the atomization of tin: (a) before the atomization, (b) flow separation during the atomization at 2.2 MPa gas pressure, (c) no flowseparation during the atomization at 1.3 MPa gas pressure.

O. Aydin, R. Unal / Computers & Fluids 42 (2011) 37–43 43

zle are seen for the different atomization pressures in Fig. 10. It isseen clearly that flow separation is more efficient with the pressureincrease. At the 1.0 MPa gas pressure, flow separation is not ob-served. If the pressure is increased, flow separation is determinedmore efficiently beyond the 1.7 MPa pressure. The liquid metal isdrawn into the recirculation zone by the aspiration. Where, lateralspreading occurs at the tip of the melt delivery tube. In the recircu-lation zone downstream of the melt tip, the pressure is usuallyhigher at the centerline. This will force the metal outwards intothe part of the gas stream where it is most energetic (Fig. 11). Flowseparation was observed experimentally as shown in Fig. 12. Be-fore the atomization started the melt tip base is shown inFig. 12a, and the flow separation was accomplished during theatomization of tin as shown in Fig. 12b. On the other hand, no flowseparation was observed during the atomization of tin at 1.3 MPaas shown in Fig. 12c. The freeze off was not observed during theatomization of tin due to the sufficient overheating of the melt.

4. Conclusions

Using computational fluid dynamics (CFD) software, a paramet-ric study was conducted to determine the effects of atomizing gaspressure on the melt delivery tube base pressure and flow separa-tion. The melt delivery tube base pressure CFD simulations captureaccurately the trends observed in the experimental data. This re-sult showed that the CFD modeling can be used for the estimationof the melt delivery tube base pressure of the nozzle.

It is also found that the geometry is very important for an effi-cient nozzle in order to get the highest velocity at the nozzle exit.The gas velocity has a maximum value of 663 m/s at 2.7 MPa gaspressure and minimum value of 631 m/s at 1.0 MPa gas pressure.This shows that the gas velocity could not be increased in the orderof pressure increase. For that reason, for an efficient gas atomiza-tion process the geometry could give the maximum gas velocityfor the same mass flow rate of the gas.

The occurrence of flow separation is affected by atomizationpressure and liquid delivery tube extension parameters. This flowseparation was observed experimentally during the atomizationof tin. For a good nozzle design the flow separation could beavoided in order to prevent the melt freeze off. It is found thatthe flow separation is strongly dependent on the atomizing gaspressure. As a result of this study, it is concluded that CFD simula-

tions can be used as a good tool for predicting atomization effi-ciency and design improvement.

Acknowledgement

The author would like to gratefully acknowledge support of theTUBITAK – The Scientific and Technological Research Council ofTurkey through Grant No. 107M189.

References

[1] Espina PI, Piomelli U. Numerical simulation of the gas flow in gas-metalatomizers. In: Proceedings of FEDSM’98, 1998 ASME fluids engineering,division summer meeting, Washington, DC, USA; June 21–25, 1998.

[2] Mullis AM, Adkins NJ, Huang Z, Cochrane RF. Quantitative high frame rateanalysis of the spray cone geometry during close-coupled gas atomization. In:Third international conference on spray deposition and melt atomizationSDMA2006, Bremen, Germany; 4–6th September 2006.

[3] Unal R. Investigation of the metal powder production efficiency of a newconvergent-divergent nozzle in close-coupled gas atomization. Powder Metall2007;50:302–6.

[4] Lang M, Rist U, Wagner S. Investigations on controlled transition developmentin a laminar separation bubble by means of LDA and PIV. Exp Fluids2004;36:43–52.

[5] Tinoco J, Widell B, Fredriksson H, Fuchs L. Modeling the in-flight events duringmetal spray forming. Mater Sci Eng A 2004;365:302–10.

[6] Ting J, Anderson IE. A computational fluid dynamics (CFD) investigation of thewake closure phenomenon. Mater Sci Eng A 2004;379:264–76.

[7] Zeoli N, Gu S, Kamnis S. Numerical modelling of metal droplet cooling andsolidification. Int J Heat Mass Transfer 2008;51:4121–31.

[8] Zeoli N, Gu S. Numerical modelling of droplet break-up for gas atomization.Comput Mater Sci 2006;38:282–92.

[9] Tong M, Browne DJ. Direct numerical simulation of melt–gas hydrodynamicinteractions during the early stage of atomization of liquid intermetallic. JMater Process Technol 2008;202:419–27.

[10] Tong M, Browne DJ. Modelling compressible gas flow near the nozzle of a gasatomiser using a new unified model. Comput Fluids 2009;38:1183–90.

[11] FLUENT 6.1 User’s Guide, Fluent Inc., Centerra Resource Park, 10 CavendishCourt, Lebanon, NH 03766, USA; 2003.

[12] Fluent Inc., GAMBIT User’s Guide, Fluent Inc., Lebanon; 2002.[13] White FM. Fluid mechanics. 2nd ed. McGraw-Hill Book Co; 1988.[14] Xu Q, Cheng D, Trapaga G, Yang N, Lavernia EJ. Numerical analysis of fluid

dynamics of an atomization configuration. J Mater Res 2002;17:156–66.[15] NIST reference fluid thermodynamic and transport properties database

(REFPROP) version 7.0, National Institute of Standards and Technology,Boulder, CO 80305-3328, USA.

[16] Unal R. The influence of the pressure formation at the tip of the melt deliverytube on tin powder size and gas/melt ratio in gas atomization method. J MaterProcess Technol 2006;180:291–5.

[17] Unal R. Improvements to close coupled gas atomization nozzle for fine powderproduction. Powder Metall 2007;50(1):66–71.