ILASS Americas, 20

th
Annual Conference on Liquid Atomization and Spray Systems, Chicago, IL, May 2007
_________________________________
*Corresponding author, Graduate Research Assistant
email: abani@wisc.edu
A Model to Predict Spray-tip Penetration for Time-varying Injection Profiles

Neerav Abani* and Rolf D. Reitz
Engine Research Center, University of Wisconsin-Madison

Abstract
A new model to predict spray-tip penetration for a time-varying injection profile has been formulated based on gas-
jet theory. The approach involves using an effective injection velocity for the spray tip based on a representative
spray response time. It is assumed that the instantaneous injection velocity affects the spray tip with an exponential
response function and that the response time is the particle residence time, consistent with the theory of translation
of jet vortex rings from Helmholtzs vortex motion analysis [11]. This Helmholtz theory is also shown to yield the
well-known velocity decay rate of turbulent gas jets. A Duhamel superposition integral is used to determine the
effective injection velocity for time-varying injection rates. The model is tested with different injection profiles and
different ambient densities. The results are also compared with numerical results from a CFD code that has been
calibrated for spray simulations. The comparisons agree very well and the new spray penetration model offers an
efficient method to predict penetrations. The model also can be used to predict equivalence ratio distributions for
combusting sprays and jets.

Keywords: Sprays, Droplet, Super-position integral, Vortex Motion, CFD

Introduction:

Spray models are extensively used in the modeling and
simulation of various combustion systems, such as
Internal Combustion engines and Gas Turbine engines.
Currently available zero-dimensional combustion
modeling codes use spray penetration models which are
based on a quasi-steady assumption [1, 2]. These spray
models are a direct extension of gas-jet theory based on
steady injections [3, 4, and 5]. Such gas-jet theory-
based models have been used extensively to determine
the spray-tip penetration for steady-state injections by
various researchers [3, 4]. The results have been found
to agree well with experiments and fine-mesh CFD
computations. However, realistic injection profiles are
usually time-varying and thus steady-state gas -jet
theory cannot be applied directly. CFD modeling of
sprays and jets provides good predictions of tip-
penetration, but as the mesh size is increased to be
compatible with practical engine computations, the
prediction accuracy becomes poorer. Hence, there is a
need for a better predictive model for spray penetration
that can be used with practical time-varying injection
velocity profiles and realistic engine ambient conditions.
In this work, we present a new method, which is based
on jet-theory and a superposition integral formulation to
determine an effective injection velocity and hence, the
spray-tip penetration.

There have been attempts to study non-stationary jets
and sprays. Measurements by Bore et al. [6], involved
a study of a sudden decrease in injection velocity, and
they proposed a self-similar result based on a temporal
scaling. They found that a time-scale of the form x/
U
inj,2
, where x is the position of the spray tip and U
inj,2
is
the suddenly decreased injection velocity, leads to a
self-similar result. However, their results and model
only apply to a sudden decrease in injection velocity
from an initial constant value to a constant lower value.
Previous works pertaining to simple one-dimensional
spray models include the packet penetration model of
Desantes et al. [7], where an injected spray particle
instantly travels with a different momentum once it is
overtaken by a speeding subsequently injected particle.
This approach is an improvement over other quasi-
steady state models in the literature, but their study did
not include a wide range of tested injection profiles.
Zhang et al. [8] studied the effects of flow acceleration
on turbulent jets with linear, quadratic and exponential
injection profiles using measurements. They found that
the temporal evolution of the spray front follows the
same form as the forcing function at the nozzle. Wan et
al. [9] also modeled spray penetration for evaporating
sprays by using scaling parameters. They showed their
model worked well for a linearly increasing injection
velocity profile. However, they did not consider other
injection velocity rate shapes. Breidenthal [10], in a
study of self-similar, turbulent jets postulated that if the
flow is non-stationary or non-steady a choice of self-
similarity function is the exponential function. This
self-similarity function can be thought of as the
response function of a particle in the jet to a change in
injection velocity. Crowe et al. [11] studied the
response function of a droplet to the surrounding gas
velocity. Their analysis also reveals an exponential
response function of time for the droplet to reach the
surrounding gas velocity.

In this work we propose a relatively simple explicit
model that can be used to predict spray tip penetration.

2
The model formulates the effective injection velocity
experienced at the spray tip at any given instant to
determine the spray tip penetration. The model extends
the isolated drop theory proposed by Crowe et al. [11]
that a droplet requires time to adjust to a change in
surrounding gas velocity. This change is based on a
response function which is exponential in time and the
response t ime is the ratio of a characteristic length scale
to the surrounding gas velocity. The same argument is
extended here to sprays where the characteristic length
scale is taken as the location of the particle from the
injector tip and velocity is the time-convolved injection
velocity. This is derived from the classical work of
Helmholtzs vortex motion theory [12-15], wherein in
the translational velocity of a vortex ring is related to
the circulation of the vortex. The present proposed
model is tested with different time -varying injection
profiles and varying ambient densit ies. The results
compare very well with CFD results. It is to be noted
that the choice of selecting CFD simulations as the
benchmark for comparison with the proposed analytical
model was based on the ability to test the model over
wide ranges of injection profiles, which experimentally
is a difficult and time -consuming task. The selected
injection profiles include smooth profiles, profiles
featuring a sudden decrease and sudden increase in
velocity, and their combinations.

This paper is organized as follows. A detailed analysis
including the derivation of the new model is provided
first. A short description of the CFD and spray model
[16] and details of the computational domain follows.
In the same section we also discuss modifications to
previous penetration correlations [1] as applied to non-
stationary and non-steady jets and sprays. Analytical
expressions for an injection profile with rising and
falling rates are then derived based on an assumed
average response-time. The analytical expression is also
extended to consider the case of a sinusoidal injection
profile. Next , we present 13 different injection cases to
validate the proposed model. Finally, discussion about
the use of the model for analyzing vaporizing sprays
and the usefulness of the model is presented.


Theory:

The steady-state solution for the velocity profile within
a turbulent gas jet can be obtained a using similarity
analysis of the 2D Navier-Stokes equations as [3, 4]:

1
1
1
1
1
1
1
]
1

1
1
]
1

,
_

2
2 2
t
2 2
eq
2
inj
t
2
eq
2
inj
inj
x 256
r d U 3
1 x 32
d U 3
, U min ) r , x ( U

(1)
where U
inj
is the injection velocity at the injector exit . x
is the axial distance of a particle from the injector tip
and r is the radial distance of the parcel from the spray
symmetry axis. d
eq
is the equivalent or effective
diameter of the gas jet defined as [3]:


g
l
noz eq
d d

(2)
where d
noz
is the nozzle diameter and
l
and
g
are the
injected and surrounding fluid densities, or in present
study, the liquid and gas -phase densities, respectively.?
In the case of a turbulent jet
t
is the turbulent viscosity
given by [3]:
2 /
5 . 0
eq inj t t
d U C (3)

C
t
is a constant, as reported by Abraham [3] and
Schlichting[17] who used C
t
= 0.0161. For evaluating
spray tip penetration, the centerline tip position can be
evaluated by setting r=0 in Eq. (1) and generalizing the
velocity decay rate at the centerline as:

( )
0 eq
eq inj
x K d 3 x
x K
d U 3
dt
dx
) x ( U
(4)
where K is the entrainment constant and is equal to
16
0.5
C
t
. Schlichtings choice of C
t
=0.0161 gives
K=0.457. If we consider x to be the spray tip position at
all times, then integrating Eq. (4) provides spray tip
penetration as a function of time. For a constant
injection velocity the analytical solution of spray tip
position as a function of time is given as:

( )
0
2 / 1
2 / 1
eq inj
2 / 1
x x for t d U
K
6
) t ( x
,
_

(5)
where the entrainment constant, K is assumed to be 0.5
in this study to give favorable agreement with
experimental data. For cases with time-varying
injection velocity, an analytical expression for the spray
tip penetration is derived below.

A model of an isolated droplet responding to a given
surrounding gas velocity was given by Crowe et al. [11].
The equation of motion for a particle in a gas can be
expressed by the balance of drag force to the change of
momentum of the particle as:


3
( ) v u v u
4
D
C
2
1
dt
dv
m
g
2
D

(6)
where m is the mass of the particle, v the particle
velocity, C
D
the drag coefficient, D the particle size,
g

the surrounding gas density and u the surrounding gas
velocity. The particle Renolyds number is:

]

v u D
Re
g
P

(7)
where is the dynamic viscosity of the surrounding gas.
Crowe et al. [11] assumed a low Reynolds number flow,
where the factor 24/C
D
Re
P
approaches unity. We retain
the same assumption to analyze the physics of jet
momentum and response times. Using Eq. (6) and Eq.
(7) the particle momentum equation can be written as:

( ) v u
1
dt
dv
v

(8)
where
v
is the momentum response time, which
signifies the time required for a particle released from
rest to achieve 63% of the free stream velocity as:
p D
2
g
v
Re C
24
18
D

(9)
Assuming the momentum response time is constant, the
solution for constant u is:

( )
v
/ t
e 1 u v

(10)

The above analysis reveals that a particle responds to a
change in surrounding gas velocity exponentially.
Consistent with the suggestion of Breidenthal [7], we
also extend this response to sprays where the injection
velocity changes with time. We assume that the spray
tip responds to the change in injection velocity with an
exponential response function with a characteristic
momentum response time that depends on the particle
residence time in the jet. The momentum response time
for the spray tip is modeled using the ratio of the
particle distance from the injector location to the
instantaneous injection velocity, U
inj
, as depicted
schematically in Figure 1. The schematic depicts the
head vortices at the jet tip. For a given spray angle, the
eddy size at the tip position is assumed to be related to
the tip position and the tangent of the half spray angle .
Therefore, L
eddy
= x tan().

It is shown in the present paper that the above analysis
can also be derived from Helmholtz' vortex motion
theory [12 (see also Tait [13], Taylor [14] and Lamb
[15]). If a jet is issued from a circular hole as depicted
in Fig. 1, the counter rotating vortices are due to the
vortex ring at the leading tip of the jet. If is the total
circulation around the core, a mean jet radius is L
eddy
,
and the cross-sectional radius of the vortex ring is r
a
,
then the velocity of the translation of the vortex ring,
U
V
is expressed as [12]:

eddy a
eddy
eddy
V
L
~
4
1
r
L 8
ln
L 4
U

1
1
]
1

,
_

(11)

According to Helmholtzs second vortex theorem [13],
circulation is the same for all cross sections of the
vortex tube (here vortex ring) and is independent of
time. If we assume that the vortex ring expands only in
its cross-sectional area, then the circulation around
the core of the vortex at time t is equal to the circulation
when the ring was generated at the injector tip at time
t=0, and is given [13, 14] as:
noz inj
noz inj
INJ
eddy
d U ~
d U 4 L ) x ( U 4
) x (

(12)
where the circulation (x) is the circulation of the ring
at location x and
INJ
is the circulation of the vortex at
the injector location, and it is written as a function of
the injection velocity U
inj
and the nozzle diameter d
noz
.
If we assume that the translational velocity of the
vortex is given by a ratio of a characteristic length scale
(
~
L ) and time-scale (
~
T ), then from Eqs. (11) and (12)
we have:
( )

tan x
d U
~
L
~
T
L
~ U
noz inj
eddy
~
~
V
(13)

Reitz and Bracco [18] predicted the variation of spray-
half angle and concluded that tan() ~ (
g
/
l
)/
1/2
.
Using the correlation of tan() and the effective
diameter d
eq
, Eq. (13) can thus be re-written as:

,
_

inj
eq eq inj
~
~
V
U
x
d
x
d U
~
T
L
~ U
(14)
Note that this result, derived independently from
Helmholtz vortex theory, is identical in form to the
equation for the gas jet velocity decay of Eq.(4), when
the vortex translational velocity equals the jet tip
velocity. The only time scale that appears in the above
equation is (x/U
inj
) and hence it is reasonable to assume
that the response time of the eddy to a particular
injection velocity at the nozzle exit is of the order of the
convective time-scale or flow residence time scale,
F =

(x/U
inj
). This is also valid in the case of an injection
velocity that is changing with time. Crowe et al. [11]
argued in the case of droplets, that the response time is
also proportional to the flow-time scale and the two
time scales were related by the Stokes number.

4
Introducing the Stokes number, St =
v
/
F
, where
F
is a
characteristic time of the flow, which in the spray case
is x/U
inj
, following Crowes model [11], the response
time of the particle,
v
is given as:

,
_

inj
v
U
x
St
(15)
This form of the time response is also consistent with
that introduced in the experimental study of Bore et al.
[6]. In the present study the Stokes number is taken an
experimental constant assumed to be St=3 and the
effective injection profile of the spray tip corresponding
to an injection velocity change from U
1
to U
2
is thus:

( )

,
_

,
_

+
v
birth
1 2 1 eff , INJ
t t (
exp 1 U U U U

(16)

where t
birth
is the time at which the injection velocity
changes from U
1
to U
2
. This effective injection velocity
is then used in Eq. (4) or (14) to determine the spray tip
penetration. For a time -varying injection where U
inj

changes n times, the Duhamel superposition principle
[19] provides the effective injection velocity for the
spray tip position as a function of time as:
( )
( )
k
n
1 k
k eff , INJ
k
k
n
1 k
k eff , INJ
U ) t t , x ( A ) 0 ( U ) t , x ( U
t
t
U
) t t , x ( A ) 0 ( U ) t , x ( U

,
_

+
(17)
where x is the spray tip position, U(0) is the initial
injection velocity at time t=0, t
k
is the birth time of the
new injection velocity U(k), and (U)
k
= U
k
-U
k-1
, i.e.,
the change in injection velocity at t=t
k
. The function
A(x,t-t
k
) accounts for the exponential response and is
given as:

,
_

k , v
k
k
t
exp 1 ) t , x ( A

(18)
where the
v,k
is the response time associated with the
eddy at the spray tip at time t=
k
and location x and is
given as:
k
k , v
U
x
St
(19)
Note that the summation in Eq. (17) can be replaced by
integrals in the case of a continuous function U
inj
(t).
The spray tip position is determined using the effective
injection velocity in Eq. (4) as:

( )
0
eq eff , INJ
x x
x
d ) t , x ( U
K
3
dt
dx

(20)
Ricou and Spalding [20] found in their experiments that
the entrainment constant, K, increases as the injection
Reynolds number is reduced. However, for simplicity,
we assume a constant value of K=0.5. The above model
is tested below with different injection profiles. At each
step during the integration, the effective injection
velocity at the spray tip position is evaluated from Eq.
(17) and the new spray tip position is evaluated by
integrating Eq. (20) with fourth-order Runge-Kutta
integration.


Modification to Spray-tip Penetration Correlations:

Practical spray-tip penetration models in zero-
dimensional or Multi-Zone Combustion models [1, 2]
use correlations such as that of Hiroyasu and Arai [21]:
( )
( )
5 . 0
g
noz l
b b
5 . 0
noz
25 . 0
g
b
5 . 0
l
b
P
d
65 . 28 t : , t ,
t d
P
95 . 2 S : t t 0 ,
t
P 2
39 . 0 S : t t 0 ,

1
1
]
1

< <
1
]
1

< <
is time breakup the where
breakup After
breakup Before
(21)
where P is the injection pressure difference.
Substituting the Bernoulli result , P =0.5
l
U
2
inj
, and
considering the penetration after breakup, i.e., with
S>x
0
in Eq. (4) yields:

( )
2 / 1
2 / 1
eq inj
t d U 48 . 2 S (22)

Eq. (22) is similar to Eq. (5), except that the constant
2.48 is different, viz.,
46 . 3 ~
5 . 0
6
K
6

. Thus,
the Hiroyasu penetration correlation predicts
quantitatively a slightly lower spray tip penetration than
the present model. The breakup time can also be
compared to the ratio of x
0
/U
inj
and is given as :

( ) ( )
5 . 0
g
noz l
5 . 0
g
noz l
inj
eq
inj
0
b
P
d
3
P
d
K 2
3
U
d
K
3
U
x
t

(23)
This constant in the modified correlation is 3 as
compared to the constant of 28.65 in the Hiroyasus
correlation. This indicates that the apparent origin of
the jet occurs further downstream of the nozzle in the
case of sprays, reflecting the fact that the jet breakup
process takes time to occur.



5
Approximate Analytical Solutions for Simple
Injection Profiles:

The model presented above evaluates spray-tip
penetration bas ed on Eqs. (17) to (20). Approximate
analytical expressions can be obtained by assuming
quasi-steady average response times to evaluate the
velocity field. Consider an injection velocity profile of
the form U
inj
(t)=(a+mt), where a is the initial injection
velocity and m is the slope of injection velocity rate
shape. For simplicity it is assumed that the response
time,
V,av
is constant. Simplifying Eq. (17) in integral
form gives:

( ) ( )
( )
( )

d U
t
exp ) t ( U ) t , x ( U
d
d
dU t
exp 1 ) 0 ( U ) t , x ( U
t
0 av , V
eff , INJ
t
0 av , V
eff , INJ

,
_

,
_

,
_

,
_

(24)
where U denotes the time differential and
v,av
is an
averaged response time modeled as:

( )

T
0
inj noz l
T
0
2
inj noz l
AV
AV
eq A
av , V
dt U A
dt U A
U ;
U
d St
t

(25)
where T is the injection duration and U
AV
is the mass-
averaged injection velocity. St
A
is viewed as an
adjustable constant chosen here as 50. This high value
is consistent with the constant used in the Hiroyasu
correlation Eq. (23). From Eq. (15), the length scale x is
the distance from the particle location to the injector
nozzle, which is dynamically calculated over the time.
Since a-priori information about the particle position in
the approximate analytical approach is unavailable, the
length scale chosen is the effective nozzle diameter.

Using U= m, Eq. (24) gives an effective injection
velocity for a linear injection profile as:
( )

,
_

,
_

+
av , V
av , V eff , INJ
t
exp 1 m t m a ) t , x ( U

(26)
Substituting Eq. (26) in Eq. (20) and solving for x gives
the spray-tip penetration as:
( )
2 / 1
2
av , V
av , V
2
av , V
2
av , V
eq
m 2
t
exp m 2
mt m a t 2
K
D 3
) t ( x

,
_

,
_

,
_

(27)
where for a particular linear-injection profile,
V,av
is
given fromEq. (25) as:
( )
( )
1
1
1
1
]
1

,
_

3 3
eq A av , V
a mT a
mT
2
1
a mT 3
d St T , m , a

(28)
Note that Eq. (27) reduces to Eq. (4) if m=0 and a=U
inj

(top hat injection profile).

Similar approximate solutions can also be found for
other profiles such as for a sinusoidal injection profile
of the form U
inj
(t)=C sin(t), where =2/ , and is
the total duration of injection. Using U()= cos()
and Eq. (24), the effective injection velocity is
predicted to be:
( )
( ) ( )

,
_

,
_

+
+

av , V
av , V 2
av , V
2
av , r
eff , INJ
t
exp t sin t cos
1
C
t sin C ) t , x ( U

(29)
where for a particular sinusoidal-injection profile,
V,av

is given by Eq. (25) as:
( )
( ) ( )
( )
1
1
1
1
]
1

,
_

T 2 sin
2
1
T C
1 T cos 2
d St T , , C
eq A av , V

(30)
The spray-tip penetration for a sinusoidal injection
profile is then obtained by substituting the effective
injection velocity obtained above in Eq. (20) and
solving for x, yielding:


( )
( )
( )
( )
2 / 1
2
av , V
2
av , V
av , V
2
av , V
2
eq
2
av , V
3
1
t sin
t
exp
t cos
K
D . C 6
1
t x
1
1
1
1
1
1
1
]
1

,
_

,
_

+
+

(31)

Fuel-air distribution in Evaporating Sprays:

An additional application of the present model is to
consider spray combustion. For steady-injections under
evaporating conditions, the equivalence ratio at each
axial location in the spray is defined as the ratio of the
air mass flux to the fuel mass flux at that location [5] to
the stoichiometric air/fuel ratio. For unsteady injections
the effective injection velocity can be used to determine
the equivalence ratio distribution. For example,
assuming a stoichiometric air/fuel ratio of 15, the local
equivalence ratio at position x is:

6

( ) ( )
( )
( )
0
2
eq
eff , inj inj l
g
x x
) ( tan x
K d 5
t , x U A
t , x U t , x A
15
) t , x (

,
_

(32)
where A(x,t)=(xtan)
2
is the local cross-sectional area
of the spray, A
inj
is the nozzle hole area, and U
inj,eff
is
the effective injection velocity. Substituting x from Eq.
(20) into Eq. (32) gives the equivalence ratio
distribution as a function of time and space for time-
varying injection profiles. The results for the cases of
linear and sinusoidal injection profiles, Eqs. (27) and
(31) are useful to give approximate analytical
expressions for the equivalence ratio distributions for
those injection profiles.

Computational Fluid Dynamics Model:

Spray simulations were carried out using the KIVA-3V
CFD code [16] to assess the validity of the present jet
models . The geometry chosen was an axi -symmetric
chamber of 40 mm radius and 100 mm height.
Computations were made using a 0.5 sector mesh with
periodic boundary conditions, as shown in Figure 2.
The mesh was chosen to be very fine and uniform with
dimensions of 0.25mm in the axial direction and 0.25
mm in the radial direction, respectively (400 cells in the
axial and 160 cells in the radial directions, respectively).
Such mesh resolution has been shown to be adequate by
Abraham et al. [22]. Drop collisions, coalescences ,
evaporation and breakup were not considered in the
present study so as to just isolate the two-way
momentum coupling effect of swarms of particles in a
spray. The nozzle diameter was 254 m and the
injected particle size was the same as the nozzle
diameter. The droplet momentum equation is Eq. (6),
with the drag coefficient C
D
expressed as a function of
particle (or droplet) Reynolds number [16]. The
chamber gas was Nitrogen at 451K with densit ies of
60.6 kg/m
3
and 30.2 kg/m
3
. The liquid diesel fuel had a
density of 700 kg/m
3
. The spray tip penetration was
defined as the location with 95% of the total liquid
mass at any given instant from the injector.

Details of Test Cases:

A total of 13 different injection profiles were chosen
and the model was tested at the two different ambient
gas densities. The injection profiles were selected as
combinations of smooth functions and profiles with
sudden increases or decreases in injection velocity.
Figure 3 shows the injection profiles with a descriptive
title in each case for ease of discussion. For all cases the
injection velocities were the same for both ambient
densities.

Results and Discussion:

The spray-tip penetration results from the present zero-
dimensional model in Eq. (20) are compared with the
results from the corresponding CFD simulations. The
comparisons are presented in Figs. 4 to 29. Each figure
shows a comparison of spray tip penetration predicted
from the zero -dimensional model (solid line) and the
CFD simulation (circles). The plots also show the
actual injection velocity profile (dashed line) and the
effective injection velocity profile as experienced by
the spray tip (dotted line). The plots also show the error
in the penetration as a function of time. The error is
calculated from:

( )
CFD
2
CFD D 0
S
S S
100 Error

(33)

where S
0D
and S
CFD
are the spray tip penetration from
the 0-D model and CFD simulations, respectively. The
relative error can be high initially when the absolute
penetration is low. However as the spray progresses, the
error reduces with time. Thus, the error is evaluated
starting only after 0.2 ms from the beginning of
injection to avoid the anomalous near nozzle region.
Tables 1 and 2 show the mean percentage error for all
the cases at the two different ambient densities of 30.3
and 60.6 kg/m
3
, respectively. It can be noted that the
error is low for most of the profiles and in most cases is
less than 5%.

Figures 4, 5 and 6 show the flat injection velocity
profiles for the chamber density of 60.6 kg/m
3
. It can be
noted that the error for intermediate and high injection
velocities is larger than for the low injection velocity,
although the mean error is less 5%. Also, the effective
injection profile at the tip coincides with the actual
injection profile, (see Eq. (17)). Figure 7 corresponds to
the case with a peak in the injection profile . Note that
the zero-dimensional model follows the trend of the
spray tip penetration during the change in injection
velocity. The injection velocity does not increase
suddenly, as would occur with a quasi-steady state
model, but instead gradually increases following the
rise in injection velocity. A similar trend is observed
during the decreasing portion of the velocity profile.
Figure 8 shows the sinusoidal injection profile and the
penetration appears to be very close to the numerical
result with reduced error as time progresses. Figures 9,
10 and 11 show the results from the STEP, STEP3 and
STEPR3 injection profiles. The penetration is seen
again to compare well with the numerical results, but

7
with slight differences towards the end. It can be noted
that the slopes of the penetration curves are similar to
those obtained from the CFD results.

Figure 12 shows case PEAK2, where the peak exists for
a shorter duration but with a higher injection velocity
during the rise as compared to case PEAK. As can be
seen, a more severe high peak also agrees well with the
CFD predicted penetration. Figures 13, 14 and 15
represent pulsed injection cases (PULSE1, PULSE2
and PULSE3), respectively. In the case of multiple
pulses of equal pulse duration and equal pulse height
(viz., PULSE1 and PULSE3), the effective injection
velocity oscillates about a particular velocity. Figure 14
is a one pulse case. The tip penetration from the new
model compares well with the spray-tip penetration
from the CFD simulation and demonstrates an
exponentially reducing the effective injection velocity
at the spray-tip. Finally, the triangular injection profile
case is shown in Figure 16. The penetration again
compares well with the CFD prediction.

Similar results were obtained with a lower chamber
density of 30.2 kg/m
3
, and the comparison of the spray
tip penetration with the zero-dimensional model and the
CFD simulations agreed well for most of the cases.
Only cases PEAK2 and SINE shown in Figs. 17 and 18
had significant errors in penetration. The SINE case
shows that the penetration is under predicted from
about 0.7 ms after the start of injection. This show that
choice of the St=3, after 0.7 ms of injection gives a
relatively slower response-time and hence, a relatively
lower effective injection velocity in the first half which
also effect the penetration in the second half of
injection. In case of PEAK2, the injection velocity
changes from about 280 m/s to 690 m/s. The
penetration compares well with the CFD result during
the duration when injection velocity is increasing.
However during the period when the injection velocity
is decreasing the penetration is under-predicted which
is due to a relatively lower effective injection velocity.
This show that the choice of St=3 results in a relatively
faster time response during this period of injection.
Hence, these two cases show that for some particular
injection cases, a variable St depending upon the
magnitude of the injection velocity change could help
in improving result. The agreement could also have
been improved by considering a variable entrainment
constant , as suggested by Ricou and Spalding [13], or
with a variable Stokes number for the droplet
momentum response. Such tuning of the model could
be considered as a future improvement for wider ranges
of injection conditions, such as for the case of the
sudden large increase in injection velocity of Fig. 12.

It is also of interest to explore results obtained with the
analytical spray tip penetrations derived for the linear
and sinusoidal injection profiles in Eqs. (27) and (31).
Spray-tip penetration plots and the effective injection
velocity profile for the line-shape injection cases are
shown in Figs. 19 and 20 for positive and negative
slopes U
inj
=a +mt, where a is the injection velocity at
t=0 and m is rate of change of injection velocity,
respectively.

Figure 19 shows the case with positive slopes. Figure
19(a) shows the comparison of penetration predictions
from the zero-dimensional model, CFD model and the
analytical model. The penetration predictions from
analytical model agree well with both the CFD model
and the zero-dimensional model. Figure 19(b) compares
the prediction of the effective injection velocity profile
at the spray tip from the zero-dimensional model and
analytical model with that of the actual injection
velocity profile. The effective velocity profile from the
analytical model is observed to be greater than that of
the zero-dimensional model. However its effect on the
penetration prediction is less, as observed from Fig.
19(a). Similar plots are shown for the case with
negative slope of the linear injection profile in Figs.
20(a) and 20(b). From Fig. 20 (b) it can be observed
that the prediction of the effective injection velocity at
the tip from the analytical model matches well with that
of the CFD and zero-dimensional models. The
penetrations also match very well as shown in Fig.
20(a). Figures 21(a) and 21(b) show similar plots for
the sinusoidal case and show a good match with the
CFD results and the zero-dimensional model. Again it
can be observed from Fig. 21(b) that the effective
injection velocity at the spray tip from the analytical
model is under-predicted compared to that of the zero-
dimensional model, but the penetrations agree well, as
shown in Fig. 21(a). Hence, it be concluded that for the
linear and sinusoidal injection profiles, the approximate
model to predict the effective injection velocity profile
can provide good predictions of spray tip penetration.

Finally, Fig. 22 shows a plot of the variation of
equivalence ratio at the spray tip with time for four
different cases, viz., the linear injection profile with
positive and negative slopes, and the sinusoidal cases as
discussed above analytically, plus a constant injection
velocity case having an average velocity approximately
equivalent to the former three cases (385 m/s). The
equivalence ratio is analytically calculated from Eq.
(32). Figure 22 shows the comparison of the four cases.
It can be observed that the constant injection case and
the linearly decreasing case exhibit similar trends of
variation of . The plot also shows a change in slope in
the near nozzle area, which is due to the change in slope
of the penetration, x in the near nozzle area. This is

8
consistent with the observations of Naber and Siebers
[5] and with Eqs. (21) where in the near nozzle area,
penetration is proportional to t and in the far-field
proportional to t
1/2
. This trend is not visible for the other
two cases of the linearly increasing profile and the
sinusoidal profile. This is because these profiles start
with a very low injection velocity, which is not
significant enough to reveal the changes in the gradient
in the near-field. The plot also shows the time when the
equivalence ratio at the spray tip reaches unity. For the
linear profile with negative and positive slopes , the time
to reach = 1.0 is
m-
=0.05 ms and
m+
=0.26 ms,
respectively. Similarly for constant injection and
sinusoidal injection profiles, the time to reach =1 is

const
= 0.07 ms and
sine
=0.37 ms, respectively. Thus,
m-
~
const
<
m+
<
sine
. Hence, it can be deduced that for a
similar mass average injection velocity, the spray tip
reaches a combustible equivalence ratio early for
constant injection cases and for linearly decreasing
cases, and later for linearly increasing cases and even
later for a sinusoidal rate shape. This suggests that
linearly increasing or the sinusoidal injection profiles
could potentially be a choice for late combustion
regimes and linearly decreasing and constant injection
velocity profiles are best suited for early combustion
regimes for diesel engines.

CONCLUSION:

A new zero-dimensional model has been formulated to
evaluate spray-tip penetration for time-varying injection
velocity profiles based on gas-jet theory that is
consistent with Helmholtzs vortex-model.. The new
model formulates an effective injection velocity for
particles located at the spray tip and exhibits an
exponential response function to changes in the
injection velocity. A Duhamel superposition integral is
used to determine the effective injection velocity at the
spray tip for time-varying cases. The model is tested
with a wide variety of injection profiles and different
ambient densities. The results compare very well with
numerical results from CFD simulations. The new spray
model can be used to improve current zero-dimensional
engine codes and other applications where accurate
spray tip penetration prediction is required.

Analytical results for linear and sinusoidal injection
velocity profiles are also derived. The analytical results
compare well with the CFD results and the zero-
dimensional model. The model can also be applied to
predict the fuel-air distribution within evaporating
sprays. The equivalence ratio distributions for four
different injection profiles are compared with the
analytical model. The results show significant influence
of the choice of the injection profiles on the time for the
spray tip region to reach an equivalence ratio of unity.
For late combustion regimes, linearly increasing and
sinusoidal profiles are helpful to delay the mixing while
for early combustion, constant and linearly decreasing
injection profiles are best suited. The new spray model
thus offers a predictive tool for the selection of
injection profiles based on the mixing criteria in this
example application.

Acknowledgements:
Authors would like to acknowledge funding
support from the DOE Sandia Labs.

References:

1. Yoshizaki, T., Nishida, K., and Hiroyasu, H.,
Approach to Low NOx and Smoke Emission
Engines by Using Phenomenological Simulation.,
SAE Paper 930612, 1993.
2. Jung, D., and Assanis D. N., Multi-Zone DI
Diesel Spray Combustion Model for Cycle
Simulation Studies of Engine Performance and
Emissions, SAE Paper 2001-01-1246, 2001.
3. Abraham, J., Entrainment characteristics of
transient jets, Numerical Heat Transfer, Part -A 30,
347-364, 1996.
4. Peterson B. R., and Ghandhi J. B., Transient
High-Pressure Hydrogen Jet Measurements, SAE
Paper 2006-01-0652, 2006.
5. Naber, J.D. , and Siebers, D. L., SAE technical
paper 960034, 1996.
6. Bore J., Atassi N., Charnay G., and Taubert L.,
Measurements and Image Analysis of the
Turbulent Field in an Axisymmetric jet Subject to a
Sudden Velocity Decrease, Experimenteal
Thermal and Fluid Science, 14, 45-51, 1997.
7. Desantes, J. M., Payri, R., Salvador, F.J. , and
Gimeno, J., Prediction of Spray Penetration by
Means of Spray Momentum Flux, SAE Paper
2006-01-1387, 2006.
8. Zhang, Q., and Johari, H., Effect of acceleration
on turbulent jets, Phys. Fluids. 8 (8), 2185-2195,
1996.
9. Wan, Y., and Peters, N., Scaling of Spray
Penetration with Evaporation, Atomisation and
Sprays, Vol. 9, pp. 111-132, 1999.
10. Breidenthal R., The Turbulent Exponential Jet,
Phys. Fluids, 29(8), 2346-2347, 1986.
11. Crowe, C., Sommerfeld, M., and Tsuji, Y.,
Multiphase Flows with Droplets and Particles.,
CRC Press, Chapter-2, 1998.
12. Helmholtz, H. ,ber die Integrale der
hydromechanischen Gleichung, Crelles J., 55, 25,
1858.
13. Tait P. G., On Integrals of the Hydrodynamical
Equations which express Vortex-motion. By H.
Helmholtz. The London, Edinburgh and Dublin

9
Philosophical magazine and Journal of Science,
Vol. 33, Series 4, 485-512, 1867.
14. Taylor G. I., Formation of a Vortex Ring by
Giving an Impulse to a Circular Disk and then
Dissolving it Away, Journal of Applied Physics,
Vol. 24, Number1, 1953.
15. Lamb, Hydrodynamics, Cambridge University
Press, Cambridge, 6
th
Edition, Sec 163, p. 241.
16. Amsden,A.A., KIVA-3V: A block structured
KIVA program for engines with vertical or canted
valves, Technical report No. LA-13313-MS, Los
Alamos National Laboratory, July, 1997.
17. Schlichting, H., Boundary Layer Theory, McGraw-
Hill, New York, 1976.
18. Reitz , R.D., and Bracco, F. B., On the
Dependence of Spray Angle and Other Spray
Parameters on Nozzle Design and Operating
Conditions, SAE Paper 790494, 1979.
19. Hilderband, F. B., Advanced Calculus for
Applications, Prentice-Hall Inc., pp. 451-453,
1962.
20. Ricou. F. P. , and Spalding. D. B., Measurements
of entrainment by axisymmetrical turbulent jets, J
Fluid Mech., 11, 21-32, 1961.
21. Hiroyasu, H., and Arai, M., Fuel Spray
Penetration and Spray Angle of Diesel Engines,
Trans. Of JSAE, Vol. 21, pp.5-11, 1980. SAE-
930612.
22. Abraham, J., 1997, "What is Adequate Resolution
in the Numerical Computations of Transient Jets?"
SAE Trans., 106, No. 3, pp. 141155.
























Table 1: Error in predicted spray tip penetration
for
g
=60.6 kg/m
3
.




















Table 2: Error in predicted spray tip penetration

g
=30.6 kg/m
3
.













Std. Deviation Mean
LOWFLAT 0.7 1.1
FLAT 2.7 6.3
HIGHFLAT 2.7 3.3
PEAK 1.9 2.5
SINE 2.1 3.5
STEP 2.2 2.4
STEP3 2.1 3.3
STEPR3 1.8 2.8
PEAK2 3.3 4.2
PULSE1 2.1 3.5
PULSE2 0.9 4.2
PULSE3 2.4 8.1
TRIANGULAR 5.0 6.4
% Error (
g
= 60.6 kg/m
3
)
Std. Deviation Mean
LOWFLAT 1.6 1.6
FLAT 1.6 1.6
HIGHFLAT 2.4 4.2
PEAK 2.1 2.1
SINE 7.6 8.6
STEP 1.9 2.2
STEP3 3.3 6.1
STEPR3 1.2 0.8
PEAK2 5.6 4.9
PULSE1 3.3 4.6
PULSE2 1.7 7.0
PULSE3 1.9 9.2
TRIANGULAR 6.9 7.3
% Error (
g
= 30.2 kg/m
3
)

10
















Figure 1: Schematic describing the momentum response time for the spray-tip eddy to adjust to the
injection velocity
























Figure 2: Details of computational domain and mesh size for CFD simulations.













Injector Location
=0.5
100 mm; Nx=400
40 mm; Nr=160
Injector Location
=0.5
100 mm; Nx=400
40 mm; Nr=160
L
eddy
= X tan()
X
Injector
location
Jet showing counter rotating vortices
at the spray tip and Vortex Ring

U
inj

v
= X/Uinj
Radius of cross
section = r
a
L
eddy
= X tan()
X
Injector
location
Jet showing counter rotating vortices
at the spray tip and Vortex Ring

U
inj

v
= X/Uinj
Radius of cross
section = r
a

11





















































Figure 3: Nomenclature of the 13 injection profiles tested.


1. LOWFLAT 2. FLAT
3. HIGHFLAT
4. PEAK
5. SINE 6. STEP
7. STEP3 8. STEPR3 9. PEAK2
1. LOWFLAT 2. FLAT
3. HIGHFLAT
4. PEAK
5. SINE 6. STEP
7. STEP3 8. STEPR3 9. PEAK2
10. PULSE1 11. PULSE2 12. PULSE3
13. TRIANGULAR
10. PULSE1 11. PULSE2 12. PULSE3
13. TRIANGULAR

12


















Figure 4: Spray Tip Penetration for LOWFLAT
case,
g
~60.6 kg/m
3
.


























Figure 5: Spray Tip Penetration for FLAT case,

g
~60.6 kg/m
3
.


























Figure 6: Spray Tip Penetration for
HIGHFLAT case,
g
~60.6 kg/m
3
.


























Figure 7: Spray Tip Penetration for PEAK
case,
g
~60.6 kg/m
3
.









13



















Figure 8: Spray Tip Penetration for PEAK case,

g
~60.6 kg/m
3
.
























Figure 9: Spray Tip Penetration for STEP case,

g
~60.6 kg/m
3
.




























Figure 10: Spray Tip Penetration for STEP3
case,
g
~60.6 kg/m
3
.
























Figure 11: Spray Tip Penetration for STEPR3
case,
g
~60.6 kg/m
3
.










14


















Figure 12: Spray Tip Penetration for PEAK2
case,
g
~60.6 kg/m
3
.
























Figure 13: Spray Tip Penetration for PULSE1
case,
g
~60.6 kg/m
3
.




























Figure 14: Spray Tip Penetration for PULSE2
case,
g
~60.6 kg/m
3
.
























Figure 15: Spray Tip Penetration for PULSE3
case,
g
~60.6 kg/m
3
.











15



















Figure 16: Spray Tip Penetration for
TRIANGULAR case,
g
~60.6 kg/m
3
.



























Figure 17: Spray Tip Penetration for SINE case,

g
~30.2 kg/m
3
.

























Figure 18: Spray Tip Penetration for PEAK2 case,

g
~30.2 kg/m
3
.


























(a) (b)
Figure 19: Spray Tip Penetration for line-shaped
case (positive slope),
g
~60.6 kg/m
3
.








16



















(a) (b)
Figure 20: Spray Tip Penetration for line-
shaped case (negative slope),
g
~60.6 kg/m
3
.
























(a) (b)
Figure 21: Spray Tip Penetration for sinusoidal-
shaped case,
g
~60.6 kg/m
3
.



























Figure 22: Equivalence Ratio at Spray-tip position
for linear and sinusoidal-shaped injection rate cases,

g
~60.6 kg/m
3
.

m-

m+

const

sine

m-

m+

const

sine