Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

AIAA JOURNAL

Vol. 49, No. 11, November 2011

Multiple Injector Model for Primary Breakup


of a Liquid Jet in Crossflow

A. Mashayek∗
University of Toronto, Toronto, Ontario M5S 1A7, Canada
M. Behzad†
University of Toronto, Toronto, Ontario M5S 1A4, Canada
and
N. Ashgriz‡
University of Toronto, Toronto, Ontario M5S 3G8, Canada
DOI: 10.2514/1.J050623
A model for the atomization of a turbulent liquid jet injected into a subsonic gaseous crossflow is provided. The
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

trajectory of the jet before its breakup is predicted theoretically. The stripping of droplets from the jet is modeled by
defining multiple injectors along the predicted jet trajectory. The properties of the stripping droplets are prescribed
using empirical correlations specifically derived for jets in crossflow. The secondary breakup of the product droplets
and their subsequent motion in the gas flow are predicted using a Lagrangian–Eulerian numerical scheme.
Experiments are carried out to investigate the capability of the model in predicting spray penetration and droplet
size, and velocity distributions downstream of the nozzle.

Nomenclature divided into three main phases: 1) injection of a liquid jet perpen-
d = nozzle diameter dicular to a gaseous crossflow, 2) deflection and deformation of the
q = momentum ratio jet, and 3) disintegration of the jet into ligaments and droplets. The
ug = gas velocity atomization mechanism is believed to start by the growth of waves
uj = jet velocity that form on the surface of the jet after injection from the nozzle. The
up = droplet velocity relative velocity between the gas and the liquid phase amplifies the
We = gas Weber number surface waves up to the point where the jet breaks up into smaller
g = gas density droplets. Apart from the three phases mentioned above, more
j = liquid density complex small-scale physical processes are involved in the problem,
 = liquid surface tension such as particle stripping from the jet.
From the computational point of view, the complete numerical
simulation of such a problem, resolving most important flow scales
on the Eulerian frame, is still expensive, especially for industrial
I. Introduction applications. These issues signal the demand for some simpler yet

R ADIAL injection of liquid jets into a high-velocity and high-


temperature cross stream at elevated pressures has various
applications in fuel injection systems and advanced aircraft engines,
reliable models that can be used for industrial design purposes and, at
the same time, take into account the flow conditions and the physical
properties of the liquid and gas phases. In this paper, we review some
such as gas turbines, afterburners, augmenters, and various com- of the challenges involved in modeling of the JICF atomization and
bustors. Since this type of fuel injection can improve the fuel propose a model that provides realistic predictions of the jet atom-
atomization and vaporization characteristics, it is commonly used in ization in crossflows. Our focus will be turbulent jets. To validate the
turbojet augmenter sections and rich burn-quick quench-lean burn, model, we perform several experiments. We first validate our
lean premixed prevaporized, and ramjet and scramjet combustion experiments by showing that their resulting correlations are in good
systems. To date, several analytical, experimental, and numerical agreement with the available literature. Then, we use our experi-
studies have investigated various characteristics of the jet-in- mental results to validate our model results. We decided to perform
crossflow (JICF) atomization. Experimental studies have developed our own experiments rather than making comparisons to previous
several correlations that predict various features of the JICF studies for two main reasons. First, we needed spatial droplet size and
atomization. Because of the nature of the experimental studies, the velocity distributions downstream of the nozzle. Although there are
results and correlations proposed by each study are mainly applicable studies available that provide that information (such as Wu et al. [1]),
within the specific parameter ranges of that study. The JICF they mostly consider laminar jets, whereas we are focused on
atomization problem involves very complex physics, such as strong turbulent jets. Second, performing our own experiments provides us
vortical structures, small-scale wave formation, stripping of small with all the information we need to set up our simulations from initial
droplets from the jet surface, and formation of ligaments and droplets flow parameters to geometrical specifications.
with a wide range of sizes. The JICF atomization process can be A review of some of the experimental studies devoted to studying
various characteristics of the JICF problem can be found in
Received 24 April 2010; revision received 6 April 2011; accepted for Mashayek and Ashgriz [2]. We will present a brief review on some of
publication 8 April 2011. Copyright © 2011 by the American Institute of the experimental and theoretical literature more relevant to the
Aeronautics and Astronautics, Inc. All rights reserved. Copies of this paper purposes of this paper.
may be made for personal or internal use, on condition that the copier pay the One of the first models for atomization of a liquid phase injected
$10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood normal to a gas stream was that of Reitz [3]. He modeled the atom-
Drive, Danvers, MA 01923; include the code 0001-1452/11 and $10.00 in
correspondence with the CCC.
ization by estimating the wavelength and growth rate of the surface

Department of Physics, 60 St. George Street. waves and relating the breakup droplet sizes to the wavelengths. Liu

Department of Civil Engineering, 35 St. George Street. et al. [4] also modeled a liquid jet injected normal to a gas stream by
‡ successive injection of droplets into the gas phase. They did not
; Department of Mechanical and Industrial Engineering, King’s College
Road (Corresponding Author). consider the stripping of droplets from the surface of the jet and
2407
2408 MASHAYEK, BEHZAD, AND ASHGRIZ

studied the secondary breakup of the injected droplets by using the et al. [6], who actually calculated the trajectory theoretically) to
Taylor analogy breakup (TAB) model [5]. calculate the locations of droplet shedding from the jet. We calculate
Rachner et al. [6] applied a similar approach for modeling the JICF the jet trajectory using a theoretical model that incorporates the
atomization. In their study, however, the column breakup and droplet effects of jet deformation and mass stripping into effect. The studies
stripping from the jet surface were taken into account by using the listed in Fig. 1 do not consider the back effect of the jet on the gas
boundary-layer stripping (BLS) model of Ranger and Nicholls [7] stream. They all, however, note the importance of such a feedback for
along with experimental correlations obtained for single droplets development of flow structures in the jet’s wake and its effects on
subjected to a crossflow. They also studied the jet penetration and droplet dispersion. In this study, we include the jet feedback effect by
incorporated the effect of mass loss from the jet surface on the jet theoretically calculating the jet trajectory and by inserting the
deflection. Madabhushi [8] also modeled the jet as a series of injected calculated jet in the computational flowfield. We show the impor-
droplets and used the Kelvin–Helmholtz wave model for droplet tance of such a feedback on droplet dispersion. Furthermore, due to
stripping. The secondary breakup of both the droplets shed from the considerable similarities between single droplet breakup and the
jet and those formed at the column-fracture point was modeled by disintegration of a JICF (Wu et al. [1] and Mazallon et al. [12]),
employing several prescribed experimental criteria for the breakup of modeling of the primary breakup of the jet in the mentioned models
single droplets. In a follow-up study by Madabhushi [9], the char- (except for Madabhushi [9]) has been mostly based on experimental
acteristics of the droplets shed from the jet surface were modified by correlations conducted for individual droplets subjected to shock
using the correlations reported by Sallam et al. [10]. Khosla and waves (such as those presented by Chou et al. [13] or Hsiang and
Crocker [11] also modeled the JICF atomization by representing the Faeth [14]). In this study, we employ empirical correlations
jet with injected individual droplets. The onset and rate of the surface specifically derived for a turbulent jet in a crossflow. In the mentioned
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

shear breakup and the rate of mass shedding at the column breakup models (except for Rachner et al. [6]), stripping droplets shed from
location were both considered to be functions of the local gas Weber the jet on the centerplane (i.e., the plane of symmetry of the spray),
number and the momentum flux ratio. Droplet shedding was whereas we introduce droplet stripping at the sides of the jet in
modeled by using a modified BLS model along with experimental accordance with the BLS model. This leads to a more realistic down-
correlations for single droplets subjected to shock waves. Secondary stream droplet distribution. Since our model considers the blockage
breakup of nonspherical fragments was modeled using the Rayleigh– effect of the jet, it is capable of investigating the effect of jet spreading
Taylor model. on the near-jet gas flow structures and droplet dispersion. This is a
Figure 1 presents a comparison between the available models new and very important capability of our model and makes it easy to
(including ours) along with their specific choices of submodels. In improve the model by considering more realistic jet shapes as our
short, in most of the mentioned modeling approaches, the liquid understanding of the jet deformation improves over time thanks to
column was represented as a set of discrete large droplets. Although high-resolution numerical simulations.
some of the models have either forced the jet-replacing droplets to To build a model with the characteristics mentioned above, we
follow a certain correlation-based trajectory or have modified the break the problem into three parts and try to address each part by
droplet drag correlation in such a way that those droplets would take proposing simplified submodels. This breakdown was originally
on a path similar to a jet in a crossflow, direct injection of droplets into proposed by Mashayek et al. [15]. The first part is the prediction of
the gas stream exposes the droplets to a strong gas stream leading to the trajectory of the jet and determining the location of the jet
rapid deflection of their trajectory and quick breakup. Most of the breakup. The second part is the prediction of the mass shedding from
mentioned models rely on trajectory correlations (except for Rachner the liquid jet, and the third part is the modeling of the secondary

Fig. 1 Summary of JICF atomization models (K–H denotes Kelvin–Helmholtz).


MASHAYEK, BEHZAD, AND ASHGRIZ 2409

breakup of the droplets stripped from the liquid column and those 1
produced at the location of the jet breakup. This approach to c1  j ab; c2  2j ab=2 
2
modeling the JICF atomization enables us to investigate the blockage  
3 c 1
effect of the jet on the gas stream, and subsequently on droplet c3    41  r2 a2   ; c4  g bug cos2
dispersion downstream of the nozzle. Moreover, we will study the 8 d 2
effect of the jet spreading (due to the aerodynamic force) on the spray
characteristics. and  is the distance from the center of mass of the cross section to the
The model presented in this paper is structured in three sections, center of mass of the half-cross section. The first, second, and third
each corresponding to one of the three submodels mentioned above. terms on the left-hand side of Eq. (1) correspond to mass times
A brief description of each of the three submodels is provided below. acceleration of the element, the viscous drag force, and the surface
Mashayek et al. [15] proposed a model for calculating the tension force, respectively. The term on the right-hand side
penetration of a liquid jet in a subsonic crossflow. Their model was in represents the aerodynamic force. The semimajor and semiminor
good agreement with the empirical correlation of Wu et al. [1]. In the axes of the elliptic cross section are denoted by a and b, respectively,
present model, we use the penetration model developed in [15] to while j and g correspond to the density of the jet and gas, j is the
predict the trajectory of the jet in the crossflow. We only consider the liquid viscosity,  is the deflection angle as defined in Fig. 2, and c
shear breakup regime (as defined by Mazallon et al. [12] and Sallam and d are defined by (see [15] for more details)
et al. [10]) since this regime corresponds to most of the practical  m 
a  r2m am m1=m
applications of the JICF atomization. Mashayek and Ashgriz [16] d2 (2)
developed a more complex model for calculating the jet trajectory by 2
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

predicting the deformation of the liquid jet into a sheetlike shape due
to the aerodynamic force. In this paper, we opt to use the simpler case c  2r2 4  am1  r2m am1  (3)
of [15] for the sake of simplicity and generality.
The deformation and breakup properties of turbulent and non-
turbulent liquid jets injected into gaseous crossflows or still gases p
The instantaneous equivalent radius req is defined by r  ab and
have been investigated experimentally by several works [1,10,17– m  33 .
26]. These studies provide invaluable information about the 40
The trajectory of the cylindrical element shown in Fig. 2 can be
properties of the fluid particles broken off the surface of the liquid jet,
calculated by integration of the velocities of the center of mass of the
the breakup of the jet as a whole, the size distribution of the droplets
element given by
produced due to the primary breakup, and the rate of mass stripping.
In this work, we use their information on turbulent jets in shear ux  vj sin; and uy  vj cos (4)
breakup regimes to model the primary breakup.
Once a liquid jet is broken into a number of droplets, one can use where vj is the jet velocity, and the deflection angle  is given by
conventional particle tracking methods to follow their secondary
breakups, their interactions, and their motions. One of the common d Faero  Fshear
 (5)
methods of modeling the secondary atomization is the TAB model dt j abhvj
[5]. There are various commercial packages that use the TAB model
for particle tracking. In this work, we use particle tracking and The aerodynamic force exerted on the element Faero and the
secondary breakup submodels of KIVA III [27]. KIVA uses an streamwise shear force Fshear are explained in [15] and can be
arbitrary Lagrangian–Eulerian methodology on a staggered grid and calculated using
discretizes space using the finite volume technique. The gas flowfield
is computed in the Eulerian frame, while the particles are treated in Faero  CD ahg ug cos2 ; and Fshear  abj vj h (6)
the Lagrangian frame. Other computational works are also available
that provide moe detailed modeling of JICF [28,29]. The drag coefficient introduced in Eq. (6) is calculated using the
This paper is organized as follows. In Secs. II and III, we discuss adaptive correlations developed in [15], which accounts for the cross-
the details of the model and explain the numerical implementations sectional deformation and the jet deflection. It should be mentioned
of the empirical correlations; in Sec. IV, we explain the experimental that the x-y trajectory obtained by integrating Eq. (4) is that of the
setups and procedures; and in Sec. V, we present our model results
and compare them with our experiments. We also analyze flow
structures found in the model results in Sec. V. We conclude in
Sec. VI with a summary and final remarks.

II. Model
In this section, we provide details of the model in the order of the
three submodels introduced in the introduction. Results and
discussion will follow in the next section.

A. Jet Trajectory
The model developed by Mashayek et al. [15] is used to determine
the jet trajectory. In this model, the effect of the jet deformation on the
jet trajectory is considered. The deformation of the jet is analyzed by
assuming a cylindrical element of infinitesimal thickness h from the
liquid column. Allowing for the jet cross section in the model to
deform only into elliptic shapes, using the work of Clark [30], the
deformation of the cross section over time is calculated using
 2   
d  d
c1  c 2  c3  c4 (1)
dt2 dt

where c1 through c4 are defined as Fig. 2 Schematic of the jet element movement along the trajectory.
2410 MASHAYEK, BEHZAD, AND ASHGRIZ

center of the jet. To calculate the upper or lower boundaries of the momentum of turbulent fluctuations can overcome the surface
liquid jet, one has to take into account the local value of the tension force. Incorporating the effects of turbulence, Lee et al. [24]
semiminor axis of the cross section along the trajectory, as shown in proposed the following correlation for the height of the location
Eqs. (45) and (46) of [15] for the upper boundary. where the smallest droplets begin to shed from the jet:
Another important factor in the calculation of the jet trajectory is
the amount of mass reduced from the liquid column in the form of the zonset =1  0:07zonset =0:44 WeL
0:22
G =L u1 =v 0j 2 0:9
stripped liquid particles. The rate of mass stripping at different 0:6
 1890WeL (10)
heights of the liquid column strongly depends on the gas flow con-
ditions and the breakup regime. The following are used to calculate where  is the integral length scale and has a value of d=8 for
the mass shed from the liquid jet [15]: turbulent pipe flow, WeL is the Weber number based on the integral
3 t length scale and is defined by WeL  L v2j =, and zonset is the
Mshed  d1:5 j onset GHug RM t (7) vertical position of the location of the onset of the primary breakup
4 t
and can be calculated using
where
s zonset  vj tonset (11)
 1=3  1=3
g g 8j 3h
G ; H ; RM  (8) where vj is the jet velocity, tonset is the time of onset of primary
j j 3Gg 4r breakup, and v0j is the rms fluctuation of the jet velocity along the gas-
streamwise direction. Lee et al. [24] compiled their experimental
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

t is the time elapsed from the onset of the mass stripping, and t is results for turbulent jets in crossflow with those of some earlier
the characteristic time defined by studies on turbulent jets in still gases (Sallam et al. [22] and Wu and
p Faeth [18]) and proposed the following relation for the size
d0 j =g
t  (9) distribution of the droplets formed from ligaments due to Rayleigh-
ug like breakup:
Equation (7) can be used to reduce the mass of the cylindrical SMD =  0:56z=We0:5 0:5
L  (12)
element of our model along its path from the nozzle to the column
breakup location. The reduction is performed by scaling down the where SMD is the Sauter mean diameter of the shedding droplets.
area of the elliptic cross section of the jet. The details of the Lee et al. [24] showed that the size distribution of the primary-
assumptions behind the formulas presented in this subsection along breakup-induced droplets was not affected much by the crossflow
with the analysis of the effects of some of the assumptions on the jet and was similar to those of a jet in still gas. The velocities of the
trajectory are presented in [15]. The trajectory of the liquid jet in our droplets after the primary breakup for the turbulent case were found
model is calculated using the method explained here. The resulted to be nearly independent of the drop diameter, according to Lee et al.
deflected liquid column is then inserted in the computational domain [24], and are given by
to capture the effect of the presence of the jet on the surrounding gas
flow. This is done by representing the jet as a series of point forces wp =vj  0:75 (13)
closely located besides each other along the jet trajectory. In the next
subsection, we explain the scheme we use to model the droplets
stripped from the liquid column. up =uL  up L =G 0:5 =u1  4:82 (14)

where up and wp are the x-direction and z-direction droplet


B. Droplet Stripping
velocities, respectively. The location of the point of the column
Once the trajectory of the liquid jet is predicted and the jet is breakup can be calculated from
inserted into the computational domain, we have to deal with
primary-breakup processes. Figure 3 shows a schematic of a jet in the xb =dj  Cxb (15)
crossflow. As the figure suggests, the next part of the modeling
process is to account for the droplets and ligaments stripping from the with the constant being Cxb  5:20 for turbulent jets. This constant is
liquid jet along its trajectory as well as the relatively large liquid smaller than the corresponding value for nonturbulent jets (see [10]),
droplets formed at the location of the column breakup. meaning that the breakup length is smaller for turbulent jets because
Sallam et al. [10] and Lee et al. [24] compiled experimental results (jet) turbulence speeds up the process of column breakup and leads to
of earlier studies on jets in still gases and in crossflows with their own faster disintegration of the liquid column as one would expect. The
experimental results and provided correlations for various JICF average removal rate over the downstream side of the jet is param-
characteristics. Sallam et al. [10] studied nonturbulent liquid jets eterized in exactly the same way as was done in the jet trajectory
injected into a crossflow, whereas Lee et al. [24] focused on turbulent calculation (see [15]).
jets. In the following section, we describe how the above relations are
For the case of a turbulent flow, the onset of primary breakup is implemented in a numerical model to study the secondary breakup
controlled by the turbulence level, and breakup occurs once the and size and velocity distributions downstream of the nozzle.

III. Numerical Implementation


The implementation of the materials discussed in the previous
section into the numerical model can be divided into three parts. First,
we calculate the jet trajectory and insert the jet into the computational
domain; second, we calculate the location, rate, and size distribution
of the droplets that strip from the jet periphery or form at the column
breakup; and third, we calculate the secondary breakup of the
stripped droplets.
Figure 3 shows a schematic of the rectangular computational
domain. The left and right faces of the domain have inflow and
Fig. 3 Schematic view of the primary and secondary breakup outflow boundary conditions. The other four faces of the domain
mechanisms and the formation of vortices around the jet. The box all have no-slip boundary condition imposed. Since the results
around the jet shows the computational domain. presented in the next section are compared with experimental
MASHAYEK, BEHZAD, AND ASHGRIZ 2411

observations, the size of the domain for the simulations are set to that known. A more proper way of distributing the droplets along the
of the experimental test chamber. The turbulence level of the input liquid jet can be developed once better experimental or numerical
gas flow is also determined based on the experiments. In general, the information are available on mass stripping mechanisms.
turbulence level of a gas flow is a function of the gas flow Reynolds The last step in the construction of the JICF in our computational
number and the geometry of the flow channel. Figure 4 shows a domain is to provide the necessary information for the relatively
sample result in which the main body of the jet can be recognized as larger droplets that form as the consequence of liquid column
the dark deflected column. The trajectory of this section is calculated breakup at the column-fracture location. These droplets are also
using the methods described in Sec. II.A. The location of the column modeled as injectors at the fracture location similar to the primary-
breakup (which is pointed out by the large square in the figure) is breakup nozzles. The mass flow rate of the sum of this group of
calculated using Eq. (15). It should be noted that, unlike some other droplets can be easily calculated by subtracting the total rate of mass
studies that tried to solve for the deformation of the jet along with the loss (due to stripping) up to the fracture point from the initial jet flow
breakup processes, the jet shape is prescribed from the outset in our rate. Again, the number of this group of droplets, their injection
simulations. frequencies, and their sizes are not known and need to be further
The next step is to introduce the droplet stripping from the liquid studied. Unfortunately, these missing information affect the down-
column. As Fig. 4 shows, the sizes of the stripped droplets increase stream size distribution considerably. Theoretical studies of the
with distance from the nozzle according to Eq. (12). The stripped breakup mechanisms that lead to the column breakup along with
droplets at each specific height are introduced in the computational experimental imaging techniques and computational fluid dynamics
domain in the form of double nozzles located at the sides of the jet simulations can all contribute to providing estimates for the missing
slightly shifted toward the downstream side (in agreement with the information at the tip of the column-fracture location. For the present
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

observations) and in the spirit of the boundary stripping model of model, observations based on shadowgraph images of the column
Ranger and Nicholls [7]. This is shown schematically in Fig. 4. breakup location from various authors suggest that the column-
Although the size distribution of injected droplets from both nozzles fracture droplets have diameters of the order of the jet diameter at the
at each height obey that of Eq. (12), they have different instantaneous disintegration point. Several test cases with different number of
values. We make this choice to achieve a more chaotic nonsymmetric column-fracture droplets used in our model led to choosing 8–10
mass stripping. The mean velocities of the stripped droplets are nozzles at the fracture point for shear breakup regimes of turbulent
governed by Eqs. (13) and (14). At this stage of the modeling, we face jets. This number is much smaller than the 39 fragments considered
two major challenges: by Rachner et al. [6]. The velocities of these droplets are set equal to
1) How many nozzles shall we define along the jet trajectory from the jet velocity at the disintegration point.
the location of the onset of the primary breakup (small square in Once the jet and primary droplets are defined in the computational
Fig. 4) to the location of the column breakup (large square in Fig. 4)? domain, KIVA solver is used to study the secondary breakup and the
2) What is the frequency of injection from each of the primary- atomization processes. In Sec. II, we discuss the results of the model
breakup nozzles? for various test cases and investigate the accuracy of the results by
The correlations for the mass rate and size distribution of the making comparisons to similar experimental cases.
droplet stripping have been discussed, in the previous section, but no
information is available to answer the two difficulties mentioned
above. Physically, there is no fixed number for the locations of IV. Experiments
droplet stripping along the jet. However, we need to assign a number Experiments were conducted in a high-pressure chamber with a
to it for our modeling purposes. We assume that the frequency of the square cross section of 5 cm high and 4 cm wide. Figure 5a shows the
injection of the stripping droplets is the same as the frequency by test chamber with the side walls removed. Injection is done through a
which one would need to inject droplets of diameter d from the main nozzle with a diameter of 0.5 mm embedded in an injection strut.
jet nozzle to obtain a flow rate similar to that of the main liquid jet.
This way, the rate of stripping is directly proportional to the jet flow
rate. This assumption, however, is not well justified and is subject to
further investigation. This assumption puts a restriction on the
number of primary-breakup nozzles we define along the liquid jet
since the size distribution and the total mass stripping rates are

Fig. 4 Sample test case used for illustration of the representation of the Fig. 5 Test section: a) schematic view of the test chamber with a vertical
main jet and the stripping particles in the model. The large square shows laser sheet passing through the top and bottom windows for PIV
the column-fracture point, and the small square shows the location of the measurements; and b) the experimental chamber. The dimensions of the
start of droplet stripping. side windows are 12 by 4.94 cm.
2412 MASHAYEK, BEHZAD, AND ASHGRIZ

Table 1 Test condition for the experimental test cases


Case no. Gauge pressure, bar Jet velocity, m=s Gas velocity, m=s Momentum ratio q Gas Weber Weg
1 1 6.7 51.4 7.3 47.6
2 1 9.6 53.3 13.8 51.2
3 1 12.4 53.3 22.9 51.2
4 1 15.1 53.3 34.2 51.2
5 2.75 6.7 40.8 6.1 57.2
6 2.75 9.6 40.8 12.3 57.2
7 2.75 12.4 40.8 20.4 57.2
8 2.75 15.1 40.8 30.6 57.2

Water is used as the injection liquid in all the experiments. The test time-averaged spray. The solid and dashed lines in Fig. 6 represent
chamber consists of two optical windows on each side with proper higher and lower pressure cases, respectively. Table 1 shows that the
quartz windows. The upstream windows are installed to enable cases can be divided into four pairs (cases 1 and 5; cases 2 and 6;
visualization of the gas field before the nozzle. The second window cases 3 and 7; and cases 4 and 8). The two cases of each pair have
on each side is installed to allow for measurements at the injection almost similar momentum ratios but different Weber numbers (due to
zone and at the downstream of the nozzle up to a distance of 240 change in air pressure and velocity). A comparison between the
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

times the nozzle diameter. Care has been taken to control the curves for each pair in Fig. 6 reveals that the low-pressure (1 bar)
turbulence level of the incoming subsonic gas flow. The test chamber cases penetrate slightly more with respect to their corresponding
is capable of withstanding pressures of up to 8 bar and is designed to high-pressure (2.75 bar) cases. It was expected, based on previous
facilitate particle image velocimetry (PIV) and phase Doppler studies, that the spray penetration be reduced with an increase in the
particle analyzer (PDPA) measurements. Windows are installed on Weber number (at a fixed q) [15]. Measuring the spray penetration at
the top and bottom faces of the pressure chamber to allow for laser various downstream distances for each case and performing a
transmission of vertical laser sheets or jets for PIV and PDPA regression analysis, we find the following correlation for the spray
measurements. Figure 5a shows (schematically) a vertical laser sheet trajectory:
passing through the centerplane of the spray. PIV measurements are
made from the side windows using a Dantec Dynamics commercial  0:39
z x
PIV system along with a New Wave double-pulse YAG laser. The  2:6q0:36 (16)
d d
PIV field of view is shown in Fig. 5b. Details of the experimental
facility and procedure can be found in Mashayek [31].
Table 1 contains the details of the PIV test cases. The mea- Figure 7 shows the extent of the agreement of Eq. (16) with the
surements are done at two different pressures and at various gas and experimental results. The exponent of q in Eq. (16) is in agreement
jet velocities to investigate the effects of variations in momentum with the 0.33 value obtained by Wu et al. [1] and the 0.36 value
ratio and Weber number on the jet penetration and on the size and obtained by Inamura et al. [32]. The exponent of x=d is also in good
velocity distributions of the resulting sprays. As explained by agreement with 0.33 obtained by Wu et al. [1]. The gas Weber
Mashayek et al. [15], spray characteristics are dependent on various number is not incorporated in Eq. (16) since the results showed little
test conditions (other than q and Weg ), among which are the nozzle sensitivity to it. Based on Eq. (16), we conclude that our experiments
properties, turbulence level of the liquid and gas phases, and are in qualitative agreement with some published studies and deemed
geometry of the test section. One of the contributors to the dis- valid to be used for examining the accuracy of our model results.
crepancies that exist among the jet trajectories obtained by various The PIV measurements were made at the centerplane of the jet.
correlations is the different measurement techniques used to obtain Before injection, the incoming gas flow was characterized using
them. Time-averaged penetration measurements provide a smooth upstream seeding of small droplets. The turbulence level of the flow
trajectory curve that better represents the spray penetration. Hence, was, at most, 8% of the air velocities for all cases. Figure 8 plots the
we use time-averaged pulsed-laser shadowgraph images in this study centerplane droplet velocity profiles at different downstream
for spray trajectory measurements. About 100 images are recorded locations for cases 4, 6, and 8 of Table 1. Focusing on Fig. 8a, the
for each test case. The images are then averaged using image- velocity profiles become more uniform with an increase in the
processing techniques to represent a time-averaged jet trajectory for streamwise distance. The profiles suggest that the droplet velocities
each case. With 40 ms time delay between images, the trajectories are
averaged over 4 s for each case. Figure 6 plots the measured spray
penetrations. For each case shown in the figure, a time-averaged
image was created by superimposing several PIV images on
each other. A curve was then fitted to the top boundary of the

Fig. 7 Distribution of the measured spray penetrations around the fit


Fig. 6 Jet trajectories measured for test cases of Table 1. presented in Eq. (16).
MASHAYEK, BEHZAD, AND ASHGRIZ 2413
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Fig. 8 Centerplane droplet velocity profiles at different locations along the streamwise direction obtained using PIV measurements: a) case 4, b) case 6,
and c) case 8. Dashed lines show experimental data.

are smaller at lower z values. The droplets at the lower heights are momentum exchange that occurs between the column and the gas
mainly formed by mass stripping from the liquid jet and have flow. As z increases, two factors affect the downstream droplet
relatively smaller droplet sizes. These droplets have smaller velocities. The first one is that, once the height becomes large
velocities due to the presence of the liquid column and the enough, the effect of the liquid column on the downstream flow is
2414 MASHAYEK, BEHZAD, AND ASHGRIZ

diminished (because the jet disintegrates into droplets). This leads to


a more direct momentum exchange between the droplets and the gas
and a larger acceleration for droplets. The second important factor is
that the sizes of the droplets shed from the liquid jet increase with
height, as reported by Sallam et al. [10] and Lee et al. [24]. This leads
to an increase with height in the downstream droplet sizes, and so a
decrease (with height) of the droplet velocities. The interplay
between the two factors leads to an initial increase (with height) in the
downstream droplet velocities followed by a slight decrease. The
effect of the upstream disintegration of the liquid jet is usually
observed in the form of a peak in the downstream velocity curves.
The droplets traveling near the top boundary of the spray are mainly
those produced as a result of the column breakup and are the largest
droplets separated directly from the parent jet. So, their downstream
velocities are small, as shown by the peaks in the curves in the figure.
Figure 8 also plots the centerplane droplet velocity profiles for
cases 6 and 8, which both have pressures of 2.75 bar and gas Weber
numbers of 57.2, but momentum ratios of 12.3 and 30.6. Velocity
trends similar to case 4 can be observed for these cases. A comparison
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

between Figs. 8a and 8c shows that the velocity curves are very Fig. 9 Sample model result that shows the blocking effect of the jet on
similar because the two cases have very close momentum ratios. As q the incoming gas stream by plotting the gas flow streamlines on a slice of
is the main player in setting the penetration of the parent jet, its the domain.
deflection and disintegration height play key roles in the velocity
distribution of the secondary droplets downstream of the jet.
However, a comparison between the two cases shows that, for the spray. The penetration of the spray into the gas phase increases from
low-pressure case 4, the droplets reach almost 90% of the gas case 6 with the momentum ratio of q  12 to case 8 with q  30. The
velocity (ug  53 m=s) by the time they reach x=d  215, whereas figure shows that the spray plume penetrates into the gas stream
for high-pressure case 8, the droplets reach nearly 96% of ug slightly less compared with the experiments. It seems that the spray in
(ug  40:8 m=s). This indicates that, for a larger gas velocity, it takes the model lacks the momentum necessary to penetrate into the gas
longer distances for the droplet velocities to merge to the gas stream field as much as the experiments. This can be due to few reasons. The
velocity. A comparison between Figs. 8b and 8c shows that, for a first one is that the ratio of the vertical velocity of the stripped droplets
fixed Weber number, the spray resulted from the atomization of the to the jet velocity is considered to be constant in the model. Although
jet with a higher q (case 8) penetrates more into the gas flow. Also this comes from the empirical correlations, on the fine scales at which
included in the figure for case 6 are the size measurements obtained we are defining the stripping properties, more information is needed
using a PDPA system. The smallest droplets are at lower heights, as than that provided by the correlations. We expect some z variations in
mentioned earlier. Moving up the size curves (i.e., increasing z), up and vp not being captured by the empirical correlations. More
there is a smooth increase in droplet sizes due to the increase in the important, the number, sizes, and velocities of the droplets formed
sizes of upstream droplets and the ligaments separated from the jet. A due to the disintegration of the jet at the column breakup location are
steeper increase is observed in the upper parts of the droplet size not properly defined. In fact, it is very hard for experiments to provide
curves, which is associated with the upstream liquid jet disintegration detailed information on the size and velocities of droplets produced at
and formation of large droplets. In general, the size and velocity the column-fracture point, and one needs to resort to high-resolution
curves show that the larger the droplets, the slower they move down- numerical simulations to gain more insight into the jet disintegration.
stream. It should be mentioned that the near-nozzle behaviors of the From the modeling perspective, the column-fracture point is
size distribution and velocities of the droplets are quite different from probably the part that lacks the most amount of information. To work
the downstream results shown here. This is mainly due to the local around this obstacle, one can tune the droplet number, sizes, and
impact of the jet on the flowfield around it. Near-nozzle size and velocities at the column-fracture point in such a way that the spray
velocity distributions were studied in detail by Elshamy et al. [33]. trajectory in the model would match the experimentally observed
trajectory. This tuning is easy, as the measurement of the spray
V. Model Results
In this section, we present the results of the numerical model. We
then verify the results through comparison with the experimental
cases discussed in the previous section. Afterward, we use the
numerical results to discuss the structures of the flowfields and the
effects of the liquid jet on the surrounding gas and the droplet
dispersion downstream of the nozzle. As the first step, we investigate
the effect of the prescribed jet on the gas flow in the computational
domain.
Figure 9 shows a three-dimensional view of a sample jet along
with the airflow streamlines on a cross-gas-streamwise plane. The
streamline pattern clearly shows the formation of double vortices in
the wake of the jet, which form due to the high momentum ratio.
Since the double vortices play an important role in the droplet
dynamics and their dispersion downstream of the jet, obtaining them
was one of the main goals of the present model, which was achieved
by defining the jetlike obstacle in the flowfield.
The results of the model are compared with data from cases 6 and 8
from Table 1. They have similar test conditions except for the jet
velocity (and hence the momentum ratio). Figure 10 shows the
calculated sprays for the two cases. The solid lines imposed on each Fig. 10 Side views of the model sprays with the experimental trajectory
of the figures show the experimentally measured trajectory of the superimposed for a) case 6 and b) case 8.
MASHAYEK, BEHZAD, AND ASHGRIZ 2415
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Fig. 11 Model results for case 6. Dashed lines correspond to droplet size, and solid lines correspond to droplet velocity.

trajectory is straightforward. The simulation tuned to the experiment similar average values, which implies that the difference between the
in this fashion can then be more reliably used to study other experiment and the simulation decreases as droplets reach their
characteristics of the spray, which are hard to study experimentally. terminal sizes and velocities. Apart from the underprediction of the
Once more information on the properties of the droplets produced at size profiles, they seem to capture the real trend well.
the column-fracture point are available through experimental or A similar comparison between the experiment and simulation for
numerical studies, this missing part of the model can be easily case 8 shows similar results to case 6. So, we just plot the velocity
improved. profiles for the experiment and the simulation of case 8 in one figure
Next, we compare the size and velocity distributions for the two to facilitate direct comparison. The results are shown in Fig. 12. The
cases to experimental results. Figure 11 shows the velocity and size dashed and solid curves correspond to experimental and model
profiles downstream of the jet obtained from the model. This figure results, respectively. Similar to case 6, the velocity profiles are
should be compared with its corresponding experimental counterpart slightly overpredicted because the sizes are underpredicted by the
in Fig. 8b. Similar to the experiments, the model results are model. The agreement between the model and the experiment,
smoothened by averaging over an adequate number of time frames to however, seems to be better for case 8, which has a higher momentum
filter out the instantaneous fluctuations. Comparisons between the ratio. The figure also shows that the model-predicted velocity profiles
two figures show that the particle sizes are underpredicted by the become relatively uniform faster (i.e., at smaller X=d) in case 8
model by about 10% at the top boundary of the jet and by up to 40% at compared with case 6.
the lower boundary of the jet (close to the bottom wall). At At this point, we conclude that our model captures the trends in
midheights, however, the sizes are in agreement with the experi- variations of the droplet size and velocity distributions correctly (as
ments. The slight underprediction of the sizes near the top boundary compared with the experiments), and we move on to analyze the flow
of the spray is mainly due to the lack of information at the column structures and the effects of the jet on the gas flow and droplet
disintegration location. Moreover, the lack of z dependence in the dispersion in more detail. As the strength and extent of the vortical
initial velocities of droplets separated from the jet plays a role in the structures formed in the wake of the jet depend on the relative gas-jet
overall profile of the droplet sizes downstream of the jet. This is more velocities, we will consider cases with higher gas velocities to help us
obvious in lower parts of the spray. It seems that, at lower parts, the better demonstrate the blockage effects of the spray. In the analyses
larger droplets travel up in the z direction more than they should, that follow, we will consider cases with a gas velocity of 100 m=s
leaving only very small droplets behind, which leads to the large corresponding to Weg  123.
underprediction in size. It also seems that the velocity and size To analyze the effect of the spray on the gas flow in more detail,
profiles in the last frames (X=d  215) of Figs. 8 and 11 have nearly we consider a case with Weg  123 and q  22 once more.

Fig. 12 Droplet velocity curves for case 8. Dashed lines correspond to experiments and solid lines to the computations.
2416 MASHAYEK, BEHZAD, AND ASHGRIZ

Figure 13a shows the isosurfaces of the magnitude of the xz investigate the extent to which the simple circular jet used in this
vorticity defined by model imposes blockage on the gas flow and affects the droplet
q transport downstream of the nozzle, we consider the same case of
j!xz j  !2x  !2z (17) Fig. 13 and repeat the calculations for two other scenarios. In the first
scenario, we do not insert a circular liquid jet in the domain, and in the
second scenario, we insert a flattened jet (a jet with width of 3d) into
where !  r  ug is the vorticity vector. The plot shows two the domain. We then compare the two cases with that of Fig. 13. It is
contour levels at 0.2 and 0:5 1=s. The level 0.5 surface sits inside important to note that, for all three cases, the jet flow rate is the same
the level 0.2 surface, which indicates that the vortical structures are and is equal to the total flow rate of the droplets stripped from the jet
stronger closer to the jet and diminish as the gas flows forward. The or formed at the column break location. Hence, we are merely
vortices shown in the figure play an important role in the dispersion investigating the effect of the presence of the jet and its shape on the
of droplets in the wake of the jet. So, one can expect the spanwise gas field and the droplet dispersion. Figure 14 shows the results of the
extend of the spray plume to be affected by the structure of these comparisons. The left, middle, and right columns in the figure
vortices. Figure 13b plots the streamlines of the perturbation gas correspond to the no-jet, circular-jet, and flattened-jet scenarios,
flow around the spray to demonstrate the effect of the spray on the respectively. Figure 13a shows the blocking effect of the jet on the
surrounding gas in its close vicinity. To obtain the streamlines, we incoming gas flow. Clearly, as the jet gets larger from left to right in
have subtracted the mean gas velocity ug from the x-direction the figure, the blockage becomes larger. Since in real flows the jet
velocity. This allows us to visualize only the perturbations to the deforms from a circular shape at the nozzle to a sheetlike shape very
flowfield due to the presence of the jet. The streamlines show the rapidly (see [15]), the comparison of the three panels shows that
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

gas moving around the spray from the sides and the top. At the top using just a circular jet leads to underestimation of the jet blocking
boundary, the streamlines bend upward due to the blockage effect. Figure 13b compares the strength of the vortical structures
imposed on the gas by the spray. In the downstream of the spray, the downstream of the nozzle by plotting the x-vorticity contours of the
streamlines move closer to each other, and ultimately become gas phase on a plane normal to the gas flow. Overlaid on the contours
parallel lines once the spray has equilibrated, and the droplets move are the gas phase streamlines on that plane. The left column shows
in the x direction. that the vortical structures form downstream of the jet, even without
Once a jet is injected normal to a high-speed gas flow, it deforms the insertion of the circular jet. This implies that the droplets getting
into complex elongated (almost like a cobra head) shape. To stripped from the jet and those formed at the column breakup location
also impose a large blockage and contribute greatly to the formation
of double vortices. Comparing the left panel to the middle and right
panels shows that, as the jet is added, and is flattened, the vortical
structures naturally grow larger in both the spanwise and vertical
directions.
To investigate the effect of the changes in the jet shape (and
subsequently in the gas flow) on droplet dispersion, in the first row of
Fig. 15, we plot a z-y view of the spray plumes (of the three cases
under consideration) as observed from a distance of 140d down-
stream of the jets. This figure shows the spray plume as observed
from a distance of 140d downstream of the nozzle (top row), SMD
distribution obtained by interpolating the spray droplet sizes on the
x  140d plane (middle row), and spanwise droplet velocity (v) map
obtained by interpolating the spray droplet spanwise velocity on the
x  140d plane (bottom row). All figures are instantaneous and not
averaged. The interpolated data are not smoothened in favor of
accuracy, and so the contour lines are not smooth. In the bottom row,
the darker and brighter regions correspond to negative and positive
spanwise velocities, respectively. omparing the left and middle
panels shows that the presence of the jet leads to accumulation of
large droplets near the center top of the spray plume, and it enhances
the spanwise dispersion of the droplets, although very slightly.
Compared with experimental studies such as Wu et al. [1], the spray
width is still underpredicted for the circular jet in the middle panel.
This is because, in real life, the jet spreads out into a sheetlike shape,
and so the spanwise distance between the droplets stripping from the
sides of the jet is larger than that of the middle panel leading to a
larger spray width downstream of the jet. This can be seen by
comparing the flattened jet in the right panel with the circular jet in
the middle panel. The spray produced by the flattened jet has a
rounder profile with a large volume-flux core more similar to those
observed in the experiments. As the flattening of the jet can be more
intense than that of our flattened jet, we can expect an even wider
spray if we insert a more realistic jet shape into the domain. The
reason behind the change in droplet distribution from the left panel to
the right is simple: for the no-jet case, the large droplets formed at the
top parts of the jet due to stripping and column breakup come to an
intense contact with the incoming gas flow, and hence lose their
vertical velocity rapidly and follow a straight path in the x direction.
Once the jet is added in the middle panel, its blocking effect (as
shown in Figs. 13b and 14a) suppresses the aerodynamic force on the
Fig. 13 Droplet transport downstream of the nozzle: a) isosurfaces of large droplets, and so they lose their vertical velocity later and follow
!xz for two levels of 0:5 1=s (inner surface) and 0:2 1=s (outer surface), a path similar to top streamlines shown in Fig. 13b. The jet’s blocking
and b) streamlines of the perturbation flowfield around the jet. effect is even more pronounced for the flattened jet leading to more
MASHAYEK, BEHZAD, AND ASHGRIZ 2417
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Fig. 14 Gas phase velocities: a) gas-streamwise velocity contours on a parallel plane, and b) streamlines overlaid on x-vorticity contours on planes
normal to the gas freestream.

accumulation of droplets into the core of the spray. Moreover, as illustrated in Fig. 14. For the flattened jet (right panel in the bottom
shown in Fig. 14, the vortical structures become larger from the no-jet row), spanwise droplet velocities have a magnitude larger than
case to the circular-jet case and from the circular jet to the flattened 5 m=s for smaller droplets closer to the lower boundary and around
jet. This improves the spanwise dispersion of the droplets, which is 2 m=s for the larger droplets in the core of the spray. Considering
clearly seen in the first row of Fig. 15. To better illustrate the that the freestream gas velocity is about 100 m=s, this means
discussions just mentioned, the middle and bottom rows of Fig. 15, that by the time the droplets reach a distance of 10 cm downstream of
respectively, show the contour of the droplet size distribution and the the nozzle, the spray width increases by
1–2 cm. Therefore, to
contour of the droplet spanwise velocity on the x  140d plane. The predict a realistic spray width, one needs to account for the jet
middle row shows that, as the jet blocking becomes stronger, the flattening.
droplet sizes become smaller downstream of the jet. Although one It should be mentioned that, most probably, there exists another
might expect the opposite to happen (due to decrease in the group of droplets (not included in our model) that get pinched off the
freestream-droplet direct impact with increase in the blockage), it leeward side of the flattened jet due to the action of the vortical
seems that the stronger and larger vortical structures formed due to structures on the jet surface. This, however, is hard to verify, as this
the larger blockage enhance the secondary breakup of the droplets group is extremely hard to detect by experimental methods, and so
leading to smaller droplet sizes as we move from the left panel to the far, not much information has been provided for it from the numerical
middle and to the right. The bottom row of the figure shows the map simulations (to our knowledge). Moreover, although our results show
of the spanwise velocity of the droplets interpolated on the x  140d the formation of a core of high-volume flux in the spray once we
plane. It should be noted that these maps correspond to the liquid account for the jet spreading, the vertical height of that core is of
phase velocities, while those in Fig. 14 represent the gas velocities. importance too. This height greatly depends on (and is very sensitive
Comparing the maps of the spanwise velocity for the liquid to) the initial velocities of the stripping droplets upon their separation
droplets (not the gas) in the bottom row of the figure shows that the from the jet. Although we have used an empirical correlation that
vortical structures formed in the wake of the jet indeed play an gives a constant ratio of vertical droplet velocities to jet velocity
important role in the spanwise dispersion of the droplets. It also [Eq. (13)], there should certainly be a z dependence to the vertical
shows that this role becomes more pronounced as we consider a more droplet velocities upon separation. This is expected because the jet
spread jet. Droplets move to the positive or negative y direction deflects and deforms greatly along its path, and the gas flow shows
(depending on which half they are located in) due to the gas flow significant variations with height along the jet. So, it is almost
2418 MASHAYEK, BEHZAD, AND ASHGRIZ
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Fig. 15 Spray plume (top row), SMD distribution (middle row), and spanwise droplet velocity (v) map (bottom row) at 140d downstream of the nozzle.

impossible for the ratio of vertical stripping droplet velocities to be a for all the three cases are the same as those of Fig. 14. For our
constant factor of the initial jet velocity. simulations, we chose grid spacing of 1 mm in all three spatial
Finally, we need to comment on the grid resolution that we used in directions, but the figure suggests that, had we chosen larger grid
the simulations discussed throughout our analyses. The grid spacing sizes (up to 5 mm), the result would not have changed by much. A
used in our simulations was selected to be of the same order of the further decrease in the grid size below 1 mm leads to very little
nozzle diameter. A grid dependency study revealed that the grid size change in the droplet sizes and velocity distributions. It is worth
does not significantly affect the size distribution of droplets down- mentioning that very small or very large grid spacings might result in
stream of the nozzle. To show this, Fig. 16 shows plots of the ineffectiveness of the collision submodel of the standard KIVA
centerline SMD profiles at the distance of 200d downstream of the model, as the potential for droplet collisions or coalescence within a
nozzle and for three different grid resolutions. The flow conditions cell depends strongly on the grid size [34]. But, in the case of JICF,
MASHAYEK, BEHZAD, AND ASHGRIZ 2419

References
[1] Wu, P. K., Kirkendall, K. A., Fuller, R. P., and Nejad, A. S., “Breakup
Processes of Liquid Jets in Subsonic Crossflow,” Journal of Propulsion
and Power, Vol. 13, No. 1, 1997, pp. 64–73.
doi:10.2514/2.5151
[2] Mashayek, A., and Ashgriz, N., “Atomization of a Liquid Jet in a
Crossflow,” Handbook of Atomization and Sprays, Springer, New York,
March 2011, pp. 657–683.
[3] Reitz, R. D., “Modeling Atomization Processes in High-Pressure
Vaporizing Sprays,” Atomization and Spray Technology, Vol. 3, No. 4,
1987, pp. 309–337.
[4] Liu, F., Smallwood, G. J., and Gülder, Ö. L., “Numerical Study of
Breakup Processes of Water Jet Injected into a Cross Air Flow,”
Proceedings of the 8th Annual Conference on Liquid Atomization and
Spray Systems, Inst. for Liquid Atomization and Spray Systems,
Pasadena, CA, July 2000, pp. 67–74.
[5] O’Rourke, P. J., and Amsden, A. A., “The TAB Method for Numerical
Calculation of Spray Droplet Breakup,” Soc. of Automotive Engineers
Paper 872089, Warrendale, PA, 1987.
[6] Rachner, M., Becker, J., Hassa, C., and Doerr, T., “Modeling of the
Atomization of a Plain Liquid Fuel Jet in Crossflow at Gas Turbine
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Conditions,” Aerospace Science and Technology, Vol. 6, No. 7, 2002,


pp. 495–506.
Fig. 16 SMD distributions for three different simulations with
doi:10.1016/S1270-9638(01)01135-X
different grid spacings but for same flow conditions (similar to those of
[7] Ranger, A. A., and Nicholls, J. A., “Aerodynamic Shattering of Liquid
the cases in Fig. 14). The SMD profiles are for a distance of 200d
Drops,” AIAA Journal, Vol. 7, No. 2, 1969, pp. 285–290.
downstream of the nozzle and on the centerline (y  0) plane.
doi:10.2514/3.5087
[8] Madabhushi, R. K., “A Model For Numerical Simulation of Breakup of
a Liquid Jet in Crossflow,” Atomization and Sprays, Vol. 13, No. 4,
and as opposed to diesel sprays, the breakup mechanism is the most 2003, pp. 413–424.
dominant process rather than the collision mechanism; therefore, the doi:10.1615/AtomizSpr.v13.i4.50
grid spacing does not affect the downstream droplet characteristics as [9] Madabhushi, R. K., “Simulation of Breakup of a Liquid Jet in Crossflow
long as the spacing is fine enough to resolve for the gas flowfield at Atmospheric Conditions,” Proceeding of ASME Turbo Expo, Vienna,
properly. American Soc. of Mechanical Engineers, Fairfield, NJ, June 2004.
[10] Sallam, K. A., Aalburg, C., and Faeth, G. M., “Breakup of Round
Nonturbulent Liquid Jets in Gaseous Crossflow,” AIAA Journal,
Vol. 42, No. 12, 2004, pp. 2529–2540.
VI. Conclusions doi:10.2514/1.3749
A new model for atomization of a turbulent liquid jet in a subsonic [11] Khosla, S., and Crocker, D. S., “CFD Modeling of the Atomization of
crossflow is developed. This model is based on using multiple Plain Liquid Jet in Crossflow for Gas Turbine Applications,” American
injectors along the jet trajectory [hence, multiple injector model Soc. of Mechanical Engineers Paper GT2004-54269, Fairfield, NJ,
(MIM)]. It also uses earlier theoretical models and empirical June 2004.
[12] Mazallon, J., Dai, Z., and Faeth, G. M., “Primary Breakup of
correlations specific to jet primary breakup rather than single droplet
Nonturbulent Round Liquid Jets in Gas Crossflows,” Atomization and
breakup. Unlike previous models, MIM includes the feedback of the Sprays, Vol. 9, No. 3, 1999, pp. 291–312.
jet on its surrounding gas. This allows for the investigation of the flow [13] Chou, W.-H., Hsiang, L. P., and Faeth, G. M., “Temporal Properties of
blockage and the effect of jet spreading on droplet dispersion. To Drop Breakup in Shear Breakup Regime,” International Journal of
investigate the accuracy of the model, experiments are performed Multiphase Flow, Vol. 23, No. 4, 1997, pp. 651–699.
using PIV and PDPA techniques. Comparison between the model doi:10.1016/S0301-9322(97)00006-2
results and the experiments confirmed that the model is capable of [14] Hsiang, L.-P., and Faeth, G. M., “Near-Limit Drop Deformation and
qualitatively capturing various spray characteristics. The model can Secondary Breakup,” International Journal of Multiphase Flow,
be improved if more information is provided on the processes leading Vol. 18, No. 5, 1992, pp. 635–652.
doi:10.1016/0301-9322(92)90036-G
to jet disintegration: droplet and ligament separation from the
[15] Mashayek, A., Jafari, A., and Ashgriz, N., “Improved Model for the
leeward side of the jet, and magnitude and direction of stripping Penetration of Liquid Jets in Subsonic Crossflows,” AIAA Journal,
droplets’ velocities as they leave the parent jet. Vol. 46, No. 11, 2008, pp. 2674–2686.
The flow structures in the wake of the jet are also analyzed, and the doi:10.2514/1.28254
effects of the jet shape on the gas blockage, the strength of the vortical [16] Mashayek, A., and Ashgriz, N., “Model for Deformation of Drops and
structures formed in the wake of the jet, droplet dispersion in the Liquid Jets in Gaseous Crossflows,” AIAA Journal, Vol. 47, No. 2, 2009,
wake of the jet, and downstream droplet size and velocity distri- pp. 303–313.
butions are discussed. It is found that, apart from the jet body itself, doi:10.2514/1.36148
the droplets stripping from it and those formed at the column-fracture [17] Wu, P.-K., Tseng, L.-K., and Faeth, G. M., “Primary Breakup in Gas/
Liquid Mixing Layers for Turbulent Liquids,” Atomization and Sprays,
point contribute greatly to the formation of the vortical structures in
Vol. 2, No. 3, 1992, pp. 295–317.
the wake of the jet. It is also shown that the spreading of the jet into a [18] Wu, P.-K., and Faeth, G. M., “Aerodynamic Effects on Primary Breakup
sheetlike shape enhances the strength and the extent of the vortical of Turbulent Liquids,” Atomization and Sprays, Vol. 3, No. 3, 1993,
structures in the wake of the jet, and therefore affect the droplet pp. 265–289.
dynamics downstream of the jet. This enhancement also enhances the [19] Wu, P.-K., Miranda, R. F., and Faeth, G. M., “Effects of Initial Flow
secondary breakup processes leading to a decrease in the overall Conditions on Primary Breakup of Nonturbulent and Turbulent Round
droplet sizes in the spray downstream of the column breakup Liquid Jets,” Atomization and Sprays, Vol. 5, No. 2, 1995, pp. 175–196.
location. [20] Wu, P.-K., and Faeth, G. M., “Onset and End of Drop Formation Along
the Surface of Turbulent Liquid Jets in Still Gases,” Physics of Fluids A,
Vol. 7, No. 11, 1995, pp. 2915–2917.
doi:10.1063/1.868667
Acknowledgment [21] Sallam, K. A., Dai, Z., and Faeth, G. M., “Drop Formation at the Surface
The authors would like to thank Amirreza Amighi from the of Plane Turbulent Liquid Jets in Still Gases,” International Journal of
Multiphase Flow and Spray Systems Laboratory for his invaluable Multiphase Flow, Vol. 25, Nos. 6–7, Sept. 1999, pp. 1161–1180.
doi:10.1016/S0301-9322(99)00042-7
help in carrying out the experiments.
2420 MASHAYEK, BEHZAD, AND ASHGRIZ

[22] Sallam, K. A., Dai, Z., and Faeth, G. M., “Liquid Breakup at the Surface [29] Pai, M. G., Pitsch, H., and Desjardins, O., “Detailed Numerical
of Turbulent Round Liquid Jets in Still Gases,” International Journal of Simulations of Primary Atomization of Liquid Jets in Crossflow,” 47th
Multiphase Flow, Vol. 28, No. 3, 2002, pp. 427–449. AIAA Aerospace Sciences Meetings, Orlando, FL, AIAA Paper 2009-
doi:10.1016/S0301-9322(01)00067-2 0373, Jan. 2009.
[23] Sallam, K. A., and Faeth, G. M., “Surface Properties During Primary [30] Clark, M. M., “Drop Breakup in a Turbulent Flow, I: Conceptual and
Breakup of Turbulent Liquid Jets in Still Air,” AIAA Journal, Vol. 41, Modeling Considerations,” Chemical Engineering Science, Vol. 43,
No. 8, 2003, pp. 1514–1524. No. 3, 1988, pp. 671–679.
doi:10.2514/2.2102 doi:10.1016/0009-2509(88)87025-8
[24] Lee, K., Aalburg, C., Diez, F. J., Faeth, G. M., and Sallam., K. A., [31] Mashayek, A., “Experimental and Numerical Study of Liquid Jets in
“Primary Breakup of Turbulent Round Liquid Jets in Uniform Crossflow,” M.S Thesis, Univ. of Toronto, Toronto, 2006.
Crossflows,” AIAA Journal, Vol. 45, No. 8, 2007, pp. 1907–1916. [32] Inamura, T., Nagai, N., Watanabe, T., and Yatsuyanagi, N.,
doi:10.2514/1.19397 “Disintegration of Liquid and Slurry Jets Traversing Subsonic
[25] Amsden, A. A., O’Rouke, P. J., and Butler, T. D., “KIVA II Manual, A Airstreams,” Proceedings of the 3rd World Conference on
Computer Program for Chemically Reactive Flows with Sprays,” Los Experimental Heat Transfer, Fluid Mechanics and Thermodynamics,
Alamos National Lab., Los Alamos, NM. Honolulu, HI, Elsevier, Amesterdam, 1993, pp. 1522–1529.
[26] Becker, J., and Hassa, C., “Breakup and Atomization of a Kerosene Jet [33] Elshamy, O. M., Tambe, S. B., Cai, J., and Jeng, S. M., “PIV and LDV
in Crossflow at Elevated Pressure,” Atomization and Sprays, Vol. 12, Measurements for Liquid Jets in Crossflow,” 45th AIAA Aerospace
Nos. 1–3, 2002, pp. 49–67. Sciences Meeting and Exhibit, Reno, NV, AIAA Paper 2007-1336,
doi:10.1615/AtomizSpr.v12.i123.30 Jan. 2007.
[27] Linne, M. A., Paciaroni, M., Gord, J. R., and Meyer, T. R., “Ballistic [34] Abani, N., Munnannur, A., and Reitz, R. D., “Reduction of Numerical
Imaging of the Liquid Core for a Steady Jet in Crossflow,” Applied Parameter Dependencies in Diesel Spray Models,” Journal of
Downloaded by Universite de Sherbrooke on November 24, 2012 | http://arc.aiaa.org | DOI: 10.2514/1.J050623

Optics, Vol. 44, No. 31, 2005, pp. 6627–6634. Engineering for Gas Turbines and Power, Vol. 130, No. 3, 2008,
doi:10.1364/AO.44.006627 Paper 032809.
[28] Herrmann, M., “Detailed Numerical Simulations of the Primary doi:10.1115/1.2830867
Atomization of a Turbulent Liquid Jet in Crossflow,” Journal of
Engineering for Gas Turbines and Power, Vol. 132, No. 6, 2010,
Paper 061506. R. Lucht
doi:10.1115/1.4000148 Associate Editor

You might also like