42nd AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit

9 - 12 July 2006, Sacramento, California

AIAA 2006-5202

Hydrodynamic Modeling of Swirl Injectors with Multiple

Rows of Tangential Channels
Sergey I. Zakharov * , Renith Richardson and Stephen D. Heister
Purdue University, West Lafayette, IN, 47906

Analytic and nonlinear computations have been conducted to study the dynamics of swirl
injectors utilizing multiple rows of tangential inlet channels. The study is motivated by the
fact that the use of multiple rows of inlets provides a mechanism for wave cancellations
within the vortex chamber of the injector, thereby reducing injector response when
compared with a single row device. The existing linear inviscid theory has been amended to
account for the additional wave interactions arising from two rows of channels within the
chamber. In addition, a boundary element method technique has been utilized to study
nonlinear aspects of the problem. Results are presented for numerous frequencies in order
to characterize the overall response of this injector type.

Nomenclature
Bo
Di,j
G
P
q
Rv
Rin, ao
Rc
Rn, a
LT
Lx
Ln
Vin

Si,j

u
v

We

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

Bond number
doublet matrix element
Greens function
pressure
normal velocity
vortex chamber radius
inlet radius of tangential channel
gas core radius
nozzle radius
length of tangential channel
length between two inlets
length of nozzle
Inlet velocity
radial direction
distance along the surface
source matrix element
axial velocity
radial velocity
Weber number
axial direction
singular contribution in integral Laplace eq. (Eq. 1)
surface slope
circulation
surface curvature
velocity potential, phase shift
liquid surface tension
density
response function

Graduate Research Assistant, School of Aeronautics & Astronautics, 315 N. Grant St., Student Member
Graduate Research Assistant, School of Aeronautics & Astronautics, 315 N. Grant St., Student Member

Professor, School of Aeronautics & Astronautics, 315 N. Grant St., Associate Fellow.

1
American Institute of Aeronautics and Astronautics
Copyright 2006 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

= angular velocity
= equation of a surface wave
= initial amplitude of wave

Subscript

n
v, vc
T

k
a
b

sw
vw

=
=
=
=
=
=
=
=

nozzle
vortex chamber
tangential channel
head end
inlet a
inlet b
surface waves
vorticity waves

Superscript

'
^

= dimensional quantity
= perturbation quantity
= mean quantity

I.

Introduction

HE injector plays a critical role in defining the performance and stability limits of a liquid rocket engine and
remains as the one component that lacks predictive tools and design models that arent based largely/entirely on
the behavior of prior designs. The vast majority of prior instability problems have been solved chiefly by making
modifications to the injector; it is the component that designers can most readily use to affect the stability
characteristics of the engine. The state-of-the-art in injector design relies heavily on empirical data from past
designs and linear/empirical models of atomization, secondary atomization, and mixing processes.
The classical theory of the steady-state operation of swirl injectors has been published in numerous contexts1-7.
These theoretical treatments generally assume a linearized, inviscid flow and provide simple relationships for
injector exit conditions (film height, velocity, and cone angle) as a function of fluid parameters and injector design.
The results of these models have been compared to experiment and tend to agree well for low-viscosity fluids. Most
recently, a nonlinear treatment has been developed to provide additional insight into the steady operation of these
devices8. While there is still fundamental understanding lacking for viscous and non-Newtonian fluids, the steady
operation of the swirl injector is fairly well understood at present.
The dynamics of swirl injectors have been much less studied. In combustion systems, the injector can participate
as an active element and this has motivated the limited analyses that have been conducted. Simple drilled orifice
response has been characterized in the 1960s and 1970s9-11 using linear models. Because many of their liquid
rocket engines employed this injector type, the Russians were one of the early developers of theories related to the
dynamics of swirl injectors. Much of the published work stemming from these efforts is due to Dr. V. Bazarov and
his collaborators12-15. In these works, the waves formed in the vortex chamber due to unsteady flowrate interact with
the convergent section forming the nozzle and transmit pulsations of massflow as a result of either forced excitation
or passive excitation via an unsteady downstream pressure. Recently, a patent appeared for the injector style used in
many of the Russian booster engines16. This patent revealed a design implementing two rows of tangential inlet
channels feeding an open nozzle design injector in which the vortex chamber has the same internal diameter as the
nozzle section. The design permits wave cancellations at selected frequencies where the wavespeed and row
spacing provide a destructive interference of waves generated from the two rows of inlets. This design approach can
be attractive to reduce injector response at known acoustic frequencies for a given combustor design.
The motivation of the current study is to amend the linear theory to account for this design feature and to
construct a nonlinear model as a basis for comparison with the linear results. Both tools are then applied to a
candidate injector design to evaluate its performance over a range of frequencies. The following sections provide
descriptions of the computational model, the analytic tool, and results from exercising these tools for a given injector
design.
2
American Institute of Aeronautics and Astronautics

II.

Computational Model

A. BEM Method for a Swirling flow

Figure 1 provides a schematic of a classical simplex atomizer noting the fluid injection via tangential channels at the
head end of the vortex chamber. The gas core develops naturally as a function of the vortex chamber diameter, inlet
massflow and the degree of swirl imparted to the fluid. The contraction to the nozzle facilitates acceleration and
thinning of the film to provide a conical sheet that breaks up into a spray. Key dimensions and nomenclature are
noted in Fig. 1 for application to the analysis and subsequent discussion.
Yoon and Heister17 and Park18 provide a complete description of the basic model elements; only highlights will
be presented here in the interest of brevity. An inviscid, incompressible, axisymmetric flow is presumed such that
the flow dynamics are governed by Laplaces equation, 2 = 0 . The boundary element method utilizes an
integral representation of this equation to provide a connection between f values on the boundary, the local
geometry, and the local velocity normal to the boundary, q = / n , as follows:

( ri ) +

n qG ds = 0

(1)

where ( ri ) is the value of the potential at a point ri , S is the boundary of the domain, is the singular
contribution when the integral path passes over the base point, and G is the free space Greens function
corresponding to Laplaces equation. For an axisymmetric domain, the free space Greens function can be
expressed in terms of elliptic integrals of the first and second kinds and is a function solely of the instantaneous
surface geometry. For this reason, a discrete representation of Eq.(1) can be cast as a linear system of equations
relating local and q values. In the discretization, both and q are assumed to vary linearly along each element,
thereby providing formal second-order accuracy for the method. Since the resulting integrals do not have exact
solutions in this case, Gaussian quadrature is used to maintain high accuracy of integration and preserve secondorder accuracy overall.
While this governing equation is a linear, nonlinearities in these free surface problems enter through the

Figure 1. A classical swirl injector noting nomenclature used for design variables
boundary condition at the interface. The unsteady Bernoulli equation provides a connection between the local
velocity potential and the surface shape at any instant in time. Prior formulations17,18 have provided a derivation of
this result suitable for implementation in a Lagrangian surface tracking environment. For the swirling flow,
modifications are required to account for the centrifugal pressure gradient created by the swirl. The dimensionless
unsteady Bernoulli equation is as follows,

D 1 G 2 G G
Bo
z
= ut ut uv Pg
+
Dt 2
We We
3
American Institute of Aeronautics and Astronautics

(2)

where is the velocity potential and is the local surface curvature and the Weber number (We= U2a/ ) and
Bond number (Bo= ga2/) become the dimensionless parameters governing the problem. Physically, this result is
a Lagrangian form suitable for use for fluid elements moving with the local velocity of the free surface. The terms
on the RHS of the equation include the effect of dynamic pressure, local gas-phase pressure, capillary, and
hydrostatic pressure contributions respectively.
G
In Eq.(2), the total surface velocity, ut , can be computed via a superposition of the base axial flow in the
G
G
injector ( , u ) with a potential vortex ( v , uv ). Letting u,v,w represent axial, radial, and circumferential velocity
components respectively, we may write:

t = + v ut = u + uv vt = v + vv wt = w + wv

(3)

Superposition of a potential vortex can be achieved by starting with the complex potential:

F ( z) =

i
log( z )
2

(4)

where z is complex variable, is vortex strength, and F is the complex potential. The resulting velocity
components for this flow are as follows,

uv = 0, vv = 0, wv =

2 r

(5)

G
G
G
This vortex is irrotational as known since = u = 0 , except at r = 0 . Using Eq.(5), the total velocity in
Eq.(3) can be computed;
1
1 G 2 G G
1
= [u 2 + v 2 + wv 2 ] wv wv
ut ut uv = [(u + uv ) 2 + (v + vv ) 2 + ( w + wv ) 2 ]
2
2
2
(6)
1 2
1 2
2
[(u + uv ) uv + (v + vv ) vv + ( w + wv ) wv ] = [u + v ] wv
2
2
where
2

1 2 1
1 2 a0U
1 a0U
wv =

=
=
2
2 2 r
2 2 r
2 r

(7)

Choosing the ideal injection velocity (U), the orifice radius (a), and liquid density () as dimensions, the
dimensionless result can be written as,
2

Bo
1a 1
D 1 G 2
z o 2
= u Pg
+
Dt 2
We We
2a r

(8)

where the Weber and Bond numbers are defined as above. In addition, the [u 2 + v 2 ] of base flow is simplified as
G2
u . Since we have nondimensionalized against the tangential velocity, U, the Rossby number does not appear

explicitly in Eq.(8), but the last term on the RHS of the equation corresponds to the circumferential pressure
developed by the potential vortex. In this context, the radial location of the center of the tangential channel, ao,
defines the dimensionless strength of the vortical flow. The kinematic equation for motion of points on the free
surface can be expressed:

Dz
cos q sin
=
Dt s

Dr
= sin + q cos
Dt s

(9)

where is the local slope of the wave with respect to the horizontal direction. Equations (8) and (9) are
integrated in time using a 4th-order Runge-Kutta scheme to provide the evolution of the velocity potential and the
motion of the free surface.
For long integrations or resolution of highly distorted surfaces, points on the free surface will tend to bunch in
regions of higher curvature as a result of the free-surface motion. For this reason, the points on the free surface are
redistributed at each time step using a cubic spline fitting of the instantaneous shape. The Laplace equation is
solved to update velocities and the process is marched forward in time. Formally, the resolution of the scheme is
second-order in space and 4th-order in time, but surface curvature and capillary forces are resolved with 4th-order
4
American Institute of Aeronautics and Astronautics

accuracy given a set of points defining the instantaneous shape. More details regarding the numerical procedure can
be found in Yoon and Heister.17
B. . Oscillating Pressure Boundary Conditions and Domain Discretization
The computational model was upgraded to address the more realistic condition involving an oscillating chamber
pressure. This boundary condition will allow orifice massflows to adjust naturally to the instantaneous pressure
within the vortex chamber. In order to implement these physics into BEM code we need to know how inflow
responses to changes in chamber pressure. This can be done using a potential vortex approximation.
2

From the radial momentum equation, V = 1 p , there for the potential vortex: V =
2R
R R

Figure 2. Schematic of a classical swirl injector

After integration the dimensional pressure drop across the liquid film can be computed using the following
equation:

P1 =

2
8 2

1
1
2 2
RC
Rin

So that the nondimensional pressure drop can be written:

P1 =
Substituting nondimensional circulation
2

P1 =

Rin Vin2
2

2
8 2

1
1
2 2
Rin RC

= 2RinV = 2RinVin

1
1 V2
2 2 = in
Rin RC 2

we obtain:

Rin2

2 1
RC

(10)

Assuming a constant stagnation pressure in the manifold and neglecting viscosity losses, Bernoullis equation
gives us pressure drop across the inlet channel: P2 =

Vin2
2

So the total pressure drop across injector is the sum of pressure drops across the liquid film and inlet channel:

Vin2 Rin2
P = P1 + P2 =
2 RC2

(11)

Knowing this pressure drop inflow velocity can be computed as:

Vin = 2 P

RC
Rin

(12)

Where pressure drop is assumed to be a sinusoidal function of time P = PSteady + sin(t ) , where steady
state pressure drop is calculated from the initial conditions:
5
American Institute of Aeronautics and Astronautics

PSteady

2
qmean
Rin2
=
2 RC2 _ steady

(13)

Figure 3 shows a typical computational grid employed in the studies. Initial pressure drop is computed from Eq.
(12) based on steady state solution for the core radius using constant inflow rate. This pressure drop plus
perturbation is further converted to massflow on the inlet boundary using Eq. (13). This approach is used for both
channels so each of them sees different pressure drop depending on the local core radius. Nodes on solid walls are
subject to the flow tangency condition (q=0), and nodes of the free surface are subject to the Bernoulli condition
derived in Eq.(8). The node lying at the free surface junction with the head-end of the vortex chamber is treated as a
moving node such that the free surface remains perpendicular to the wall at this location. The grid spacing along the
head-end of the vortex chamber is stretched to accommodate movement of the corner node.
C. Solution at interior nodes
Outflow rates in the nozzle can be computed using velocities at two boundary points and assumption of linear
velocity profile. However this approach is not accurate and results in up to a 3% error based on numerical
experiments. In order to improve the accuracy we need to know velocity profile what requires to use an additional
approach in order to get interior velocities. As in the solution for conditions at nodes on the boundary, the solution
at interior nodes is made up of the following parts: the development of integral equation for the Greens function,
the integration and the calculation of the solution for unknown values of velocity potential and its derivatives.
To compute on the interior points equation (1) can be rearranged to give

i =
After discretization it becomes:

1 G

qG d

4 n

2 i = Di , j j S i , j q j

(14)

(15)

Where i denotes interior nodes and j denotes nodes on the boundary. For any individual interior node D and S
become row vectors known from the solution on the boundary, so can be found from this equation.
In order to find velocities we need to compute derivatives

and
which can be evaluated by analytical
z
r

differentiation of equation (14) with respect to z and r. New integrals, resulting from the differentiation of the
Greens function in the z, r directions, result from this process. Ref. 19 and Ref. 20 provide a detailed discussion of
the methodology employed to discretize these
integrals.
After upgrading the code with this
methodology, the axial velocity profile at the
nozzle exit plane can be defined and integrated
to provide the overall massflow produced by
the device at a given instance in time.
Evaluation of this approach, using simple
trapezoidal quadrature to integrate the velocity
profile, shows that computed time averaged
inflow and outflow differ by less than 0.1%
using approximately the same grid spacing for
both exterior and interior nodes. Since we need
to place interior nodes only along a radial line
in the nozzle where liquid film is very thin, the
number of additional nodes was about 10-15 in
comparison to more than 300 surface nodes.
Along with the fact that matrix inversion is not
required on the interior, effect on computational
time was negligible but accuracy increased
more than on order in magnitude.
Figure 3. A grid system having two inlet channels
6
American Institute of Aeronautics and Astronautics

III.

Analytical Model

A. Single channel
The linear model of the pressure swirl atomizer consists of 3 components namely, tangential channels, vortex
chamber and nozzle. This analytical model was used to analyze the injector shown in Figure 1 and was developed by
Bazarov2.
The tangential channels are be modeled as a pressure atomizer whose length is much less than the wavelength of
oscillation.

T =

i L'T
V

VT
1
1 1 iShT
=
=
2
PT 2 i L' 2 1 + ShT2
T
1 +

V

Where T is defined as response of tangential channel and ShT =

(16)

L'T

as Strouhal number of tangential

channel.
There are two phenomena being considered in the vortex chamber. Firstly the oscillations in liquid flow rate in
'

l T , produces Surface Waves in the vortex chamber which propagates back and forth as it
tangential channels m
reflects from the entrance of the nozzle. Secondly we have what are called Vorticity Waves which refers to the
'
Vl T in the vortex chamber. The vorticity waves strongly
'
m
P vc is the vector sum
depends on radial velocity in the vortex chamber. The pressure drop in the vortex chamber

fluid regions swirling with different velocities due to

'

'

m
m
P v sw and vorticity waves
P v vw
of the pressure drop due to surface waves
The nozzle is assumed to be short and any losses (no pressure or viscous losses) in it are neglected. The
thickness of liquid through it is considered to be constant, so any effect due to vorticity waves is negligible. There
is only phase shift in the surface waves when it travels through the nozzle of length Ln given by n
'

L'n
Vwn'

Therefore the response of the nozzle would be

l 'n
m
n = ' = (1 ) e i n
l vn
m
'

(17)

'

l n & m
l vn are mass flow rate fluctuations in the nozzle and near the nozzle entrance respectively. is
Where m
the reflection coefficient i.e. the percentage of wave which reflects back from the contraction leading up to the
nozzle. Combining the individual responses of the components gives the response of the swirl injector (Eq.18). A
more thorough description can be found in Bazarov21.
l 'n
m
total =

'
m

m n'
P '

Rv2
T vn sw n
=
a 1 + 2 T ( k sw + v vw )

(18)

Here a is a geometric parameter and P is the pressure drop across the swirl injector.
The total response total is complex; hence both amplitude and phase can be deduced from it.
'

B. Modified model for dual channel injector

Consider a swirl injector with two rows of tangential injectors separated from each other by a distance Lx

(Figure 4). We assume that the response of the tangential channels and nozzle are unchanged (with respect to single
row of inlets) with the introduction of another row of inlet. Also the any modification needed due to the vorticity
7
American Institute of Aeronautics and Astronautics

effect in the vortex chamber is neglected as it is found to be negligible. However the interaction of the surface waves
in the vortex chamber has to be accounted for. See Richardson22 for a detailed analysis of a two channel swirl
injector.
The phase shift associated with the distance between the two inlet channels Lx is given by,

x =
Where angular velocity,

= 2 f

L'x

(19)

Vwv'

'

and Vwv is the wave velocity in the vortex chamber.

We can use the principle of superposition to examine the influence of reflected waves in the vortex chamber of
any arbitrary length. Let us consider an initial wave in the vortex chamber at inlet a given by

vh = a eit

(20)

Where a is the initial height of the wave due to the perturbation in the inlet a. Near the nozzle this wave
'

travels a distance of Lv (length of Vortex Chamber) and becomes,

vn = a ei(t )
v

(21)

Where

v the shift angle is associated with traveling L'v distance and 'v ' is the reduction in the amplitude due

to viscosity . A part of this wave gets reflected which is given by,

'

'

(r )vn

i t v ) v

= a e (

(22)

Where is the reflection coefficient. This reflected wave changes in the following manner when it travels a
'

length Lv to reach the closed end of the Vortex Chamber

vh = a ei(t 2 ) 2
v

(23)

This wave is reflected completely from the closed end and reaches the nozzle,

vn = a ei(t 3 ) 3
v

The reflected wave will be,

(r )vn

= 2 1e (

i t 3v ) 3v

(24)
(25)

And so the waves reflect back and forth until their amplitudes become negligibly small.
Using the principle of superposition we have the following wave near the closed end of the vortex chamber,

Lx
a

Lv

Figure 4. A classical swirl injector with two inlets at locations a & b separated by a distance Lx

8
American Institute of Aeronautics and Astronautics

vh = a n ei(t 2 n ) 2 n
v

(26)

n=0

Near the nozzle,

vn = a n e (

i t ( 2 n +1)v ) ( 2 n +1)v

(27)

n =0

Similarly we can derive equations for surface waves in the vortex chamber due to the perturbations in the inlet
channel b.
Then, using the principle of superimposition we can compute the wave near the head end of vortex chamber as
follows

vh = a n ei(t 2 n ) 2 n + b n ei(t 2 n + ) 2 n +
v

n=0

(28)

n =0

Similarly near the nozzle inlet we have

vn = a n ei(t (2 n +1) ) 2 n + b n ei(t (2 n +1) + ) (2 n +1) +

v

n=0

(29)

n=0

The above two equations are used in computing the response of the dual channel injector in conjunction with
Eq.(18).

IV.

Results

A. Single channel
A single channel swirl injector was analyzed using the linear model and compared with computational results
generated by BEM code described in section II using oscillation pressure boundary conditions. A general but
realistic injector was chosen whose design parameters are given in Table 1. A time step of 0.0003 along with a nodal
spacing of 0.032 was used. Each computation was allowed to run for 15 sec which took approximately 4 days of
CPU time. The nodal spacing and time step are optimized for reducing run time without affecting the output. See
Richardson22 for details.

Inlet
channel

radius

of

tangential

Rin' = 0.1270 in

Radius of nozzle

Rn' = 0.085 in

Radius of vortex chamber

Rv' = 0.1667 in

Radius of tangential channel

RT' = 0.0253 in

Length of tangential channel

L'T = 0.0597 in

Length of nozzle

L'n = 0.059 in

Length of vortex chamber

L'v = 0.179 in
n=4
VT' = 690 in/s

No of inlet channels
Inlet velocity

Table1 1. Design parameters of single channel swirl injector

9
American Institute of Aeronautics and Astronautics

200

Theory
BEM

150
Phase Shift in Degrees

2.5

Amplitude

2
1.5
1
0.5
0
0

Theory
BEM

100
50
0
-50
-100
-150

2000

4000
6000
Frequency in Hz

8000

10000

-200
0

2000

4000
6000
Frequency in Hz

8000

10000

Figure 6. Phase angle vs Frequency

Figure 5. Amplitude vs Frequency

The amplitude predicted by linear theory is in close agreement with computational results for frequency range <5
kHz (Figure 5.) However at higher frequencies the computations predict a much more stable injector. Possible
causes could be the time taken to turn the flow from the tangential channel to vortex chamber (90o turn) and
associated phase shift for this. From Figure 6 we see that the change in phase between theoretical and BEM code
results are higher after ~5 kHz. Another probable cause could be non-linear effects playing more significant role in
the stability of the injector at higher frequencies of operation.
B. Dual Channel results
A dual channel swirl injector was analyzed using the modified linear model and compared with computational
results generated by BEM code described in section II using oscillation pressure boundary conditions. An injector
currently being tested by Miller23 at Purdue University was used as basis for this study (Table 2).

Inlet
channel

radius

of

tangential

Rin' = 0.454 in

Radius of nozzle

Rn' = 0.454 in

Radius of vortex chamber

Rv' = 0.454 in

Radius of tangential channel

RT' = 0.0175 in

Length of tangential channel

L'T = 0.359 in

Length of nozzle

L'n = 0.0 in

Length of vortex chamber

L'v = 0.4767 in

Length between two channels

L'x = 0.0999 in
n=8
VT' = 913 in/s

No of inlet channels
Inlet velocity

Table 2. Design parameters of a dual channel swirl injector

Since the injector to be analyzed is an open injector that is the radius of vortex chamber and nozzle being equal
further modification where required to the existing model. The effect due to reflection coefficient was eliminated
(set to a very small value) and the nozzle length was set to an insignificant value. The flow was shared equally
among both the inlet channels.

10
American Institute of Aeronautics and Astronautics

0.4
BEM
Single channel
Two channels

0.35

Amplitude

0.3
0.25
0.2
0.15
0.1
0.05
0
0

2000

4000
6000
Frequency in Hz

8000

10000

Figure 7. Amplitude vs Frequency for dual channel swirl injector

The linear model shows a local minima around 1344 Hz (Figure 7,8). This frequency corresponds to the distance
between the two inlets Lx = 0.1 in. The amplitude of response is never greater than unity; this is a feature of the
'

open injector which makes it more attractive option over the classical swirl injector. The dual channel injector is
200
dual channel
BEM

150

Phase Shift in Degrees

100
50
0
-50
-100
-150
-200

500

1000

1500

2000
2500
3000
Frequency in Hz

3500

4000

4500

5000

Figure 8. Phase shift vs frequency for dual channel injector

much more stable than the single inlet design due to the destructive interference of the surface waves in the vortex
chamber. The computational results seem bounded by the single and dual channel analytical analysis for the same
design parameters.
At low frequencies (< 1000 Hz) the interference of surface waves is more prominent owing to longer
wavelengths and fewer waves in the vortex chamber. However at higher frequencies the computational model shows
a diminishing effect of the two channel inlet probably due to decreasing wavelengths of the surface waves and
probably the simple dual channel analysis seems inadequate to capture all the phenomena in the vortex chamber.
Also the non-linear effects are not being accounted for in the theoretical model.
11
American Institute of Aeronautics and Astronautics

0.35
inlet a=25%,b=75%
inlet a=50%,b=50%
inlet a=75%,b=25%

0.3

Amplitude

0.25
0.2
0.15
0.1
0.05
0
0

1000

2000
3000
Frequency in Hz

4000

5000

Figure 9. Amplitude vs frequency for different combinations of flow through the two sets of inlets

The amount of mass flow rate (Fig. 9) was varied among the two sets of inlets. A 50-50 split shows to the most
stable combination with clear local minima at 1344 Hz. Other combinations are a little less stable than the 50-50
split and also dont show any clear local minima.
0.156
0.154

Amplitude

0.152
0.15
0.148
0.146
0.144
0.142
0.14
10

20

40

60

80

Perturbation in chamber pressure (%)

Figure 10. Amplitude vs perturbation

Fig. 10 shows BEM results for amplitude amplification factor (ratio of outflow to inflow perturbation
magnitudes) as a function of chamber pressure perturbations for the frequency of 2000 Hz. As we can see even 80%
perturbation causes nonlinearity of just above 5%.

IV.

Conclusion

The dynamics of swirl injectors utilizing multiple rows of inlet channels have been addressed using linear theory
and nonlinear axisymmetric boundary element method calculations. The use of multiple inlet channels is believed to
provide a mechanism to cancel waves and hence injector unsteadiness at specific frequencies. Computations have
been conducted for a candidate design over the relevant frequency range to assess the overall phase and amplitude
12
American Institute of Aeronautics and Astronautics

response of these injectors. For designs making use of a single row on inlet channels, the linear theory and the
nonlinear BEM results agree quite well at low frequencies, with substantial departure in both amplitude and phase at
higher frequencies. This departure is theorized to be due to flow turning within the vortex chamber that is not
presently considered in the linear model. The two-channel results show much greater disparity as the complex wave
interactions are not all considered in the modified linear model. In addition, the disparity in results may also be
attributed to flow turning effects that alter the phasing of the waves emanating from the two injection regions.
Additional study is required to further illuminate this issue. Finally, a series of BEM computations were conducted
to assess the effect of perturbation amplitude on the response of the device, i.e. nonlinear effects. The results show
that high amplitude forcing is required to cause a substantial departure from the linear result.

Acknowledgments
The authors gratefully acknowledge the support of AFOSR Contract F49620-03-1-0025 under program manager,
Dr. Mitat Birkan for their support of this work.

References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.

Yule, A. J., and Chinn, J. J., Swirl Atomizer Flow: Classical Inviscid Theory Revisited, ICLASS-94,
Rouen, France, July 1994
Taylor, G. I., The Mechanics of Swirl Atomizers, 7th Int. Congress of Applied Mechanics, Vol. 2., Sept
1948
Giffen, E., and Muraszew, A., Atomization of Liquid Fuels, Chapman and Hall London, 1953
Lefebvre, A. H., Atomization and Spray, Hemisphere, Washington 1989
Bayvel, L., and Orzechowski, A., Liquid Atomization, Taylor and Francis, 1993
Khavkin, Y. I., Theory and Practice of Swirl Atomizers, Taylor and Francis, 1989
Doumas, M., and Laster, R., Liquid-Film Properties for Centrifugal Spray Nozzles, Chemical
Engineering Progress, Vol. 49, No. 10, October, 1953.
Park, H., and Heister, S. D., Nonlinear Simulation of Free Surfaces and Atomization in Pressure Swirl
Atomizers, to appear, Physics of Fluids, 2006.
Reba, I., Brosilow, C., Combustion Instability: Liquid Stream and Droplet Behavior. Part III: The
Response of Liquid Jets to Large Amplitude Sonic Oscillations, WADC Technical Report 59-720, Wright
Air Development Center, United States Air Force, 1960.
Harrje, D., Reardon, F. Liquid Propellant Rocket Combustion Instability, NASA SP-194, 1972.
Liquid Rocket Engine Injectors, NASA SP-8089, 1976.
Bazarov, V., Yang, V., Liquid-Propellant Rocket Engine Injector Dynamics, Journal of Propulsion and
Power, Vol. 14, No. 5, SeptemberOctober 1998.
V. G. Bazarov and L.A. Lyul'ka, ``Nonlinear Interactions in Liquid Propellant Rocket Engine Injectors'',
AIAA 98-4039, 1998.
Bazarov, V. G., Liquid-Propellant Rocket Engine Injector Dynamics, Journal of Propulsion And Power
Vol. 14, No. 5, Sept-Oct 1998
Bazarov, V. G., Fluid Injectors Dynamics, Mashinostroenic Publication, Inc., Moscow, Russia, 1979
Vasin, A., et. al., United States Patent, US 6,244,041 B1, 12 June, 2001.
Yoon, S. S., and Heister, S. D., A Fully Nonlinear Model for Atomization of High-Speed Jets,
Engineering Analysis with Boundary Elements, V28, pp 345-357, 2004.

18. Park, H. B., Flow characteristics of viscous high-speed jets in axial/swirl injectors, PhD Thesis Dept. of
Aeronautics and Astronautics, Purdue University, 2005
19. Rump, K. M., and Heister, S. D., Modeling the Effect of Unsteady Chamber Conditions on Atomization
Processes, Journal of Propulsion and Power, V. 14, pp. 576-578, 1998.
20. Heister, S. D., Rutz, M., and Hilbing, J., Effect of Acoustic Perturbations on Liquid Jet Atomization,
Journal of Propulsion and Power, V. 13, No. 1, pp. 82-88, 1997.
21. Bazarov, V. G., Fluid Injectors Dynamics, Mashinostroenic Publication, Inc., Moscow, Russia, 1979
22. Richardson, R., Linear and Non Linear Dynamics of Swirl Injectors PhD Thesis, Department of
Aeronautics and Astronautics, Purdue University, August 2006
23. Miller J, K., Experimental Study of Longitudinal Instabilities in a Single Element Rocket Combustor,
MS Thesis, Department of Aeronautics and Astronautics, Purdue University, May 2005

13
American Institute of Aeronautics and Astronautics