Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

https://doi.org/10.1007/s40430-018-1205-6 (0123456789().,-volV)(0123456789().,-volV)

TECHNICAL PAPER

Study of internal flow of a bipropellant swirl injector of a rocket

engine
Julio R. Ronceros Rivas1 • Amı́lcar Porto Pimenta2 • Saulo Gómez Salcedo2 • Gustavo Adolfo Ronceros Rivas3 •

Marie C. Girón Suazo4

Received: 25 March 2017 / Accepted: 20 April 2018 / Published online: 11 May 2018
Ó The Brazilian Society of Mechanical Sciences and Engineering 2018

Abstract
This work presents the study of the behavior of the internal flow in a swirl bipropellant injector, which is composed of an
open-end (without nozzle) and a closed injector (with nozzle). In this way, each of these injectors has a characteristic
behavior with respect to velocity distribution, pressure, and other main parameters. In this study, three methods are used,
which are: experimental, numerical, and analytical. For the numerical simulation was used a three-dimensional structured
mesh, capable of holding three important areas: the oxidizer swirl chamber (closed swirl injector), the fuel swirl chamber
(open-end swirl injector), and the area designed for the spray zone, which will include the phenomena caused by the
interaction of the flow of the oxidant and the fuel within the bipropellant injector. The simulation was carried out through
the commercial code CFD fluent in permanent regime, using the RNG k-epsilon turbulent model and the volume of fluid
multiphase model to locate the liquid–gas interface. In addition, experimental data and a mathematical model developed
based on theories of Abramovich and Kliachko are also presented .

Keywords Bipropellant swirl injector  Open-end swirl injector  Closed swirl injector  Abramovich theory 
Klia Kliachko theory

List of symbols
Technical Editor: Jader Barbosa Jr.
A Geometrical characteristics parameter of the
& Julio R. Ronceros Rivas tangential swirl injector
pcmajron@upc.edu.pe Ac Geometrical characteristics parameter of the
Amı́lcar Porto Pimenta conical swirl injector
amilcar@ita.br AE Equivalent geometrical characteristics parameter
Gustavo Adolfo Ronceros Rivas of the swirl injectors
gustavo_ronceros@hotmail.com Cd Discharge coefficient
Marie C. Girón Suazo Dh Hydraulic diameter
marie.giron@upc.edu.pe f Volumetric forces
fp Cross-sectional area of inlet port
1
Departamento de Ingenierı́a Mecatrónica, Universidad K Coefficient of loss due to liquid viscosity
Peruana de Ciencias Aplicadas (UPC), Campus Monterrico,
Prolongación Av. Primavera 2390, Surco, Lima, Peru Lmix Distance between nozzle injectors
2 n Number of inlet ports
Divisão de Engenharia Aeronáutica, Instituto Tecnológico de
Aeronáutica (ITA), Comando Geral de Tecnologia N Total number of phases
Aeroespacial - Praça Marechal Eduardo Gomes 50, Vila Das m_ Mass flow rate
Acacias, São José Dos Campos, São Paulo CEP 12228-900, P Pressure
Brazil DP Differential pressure
3
Curso: Engenharia de Energia, Universidade Federal de Rs Swirl chamber radius
Integração Latino-americana (UNILA), Av. Tancredo Neves Rinj Radius to axis inlet channel
3838 – Porto Belo, Foz do Iguaçu, PR CEP 85867-970, Brazil
ra Air core radius
4
Departamento de Ciencias, Universidad Peruana de Ciencias ro Outlet orifice radius
Aplicadas (UPC), Campus Monterrico, Prolongación Av.
Primavera 2390, Surco, Lima, Peru t Film thickness, ro–ra

123
289 Page 2 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

u Vectorial velocity the annular area occupied by the liquid to the total area of
Uin Inlet entrance velocity the outlet orifice, by the film flow coefficient area ‘‘u’’
U, W Velocities (Eq. 1), which will help us find many important parameters
of the swirl injector [15].
Greek symbols Due to the conservation of the angular momentum, the
a Half-spray angle liquid is pressed against the walls of the injector, being
b Swirl angle ejected in small droplets in a conical surface (spray), which
n Losses coefficient facilitates the exposure of a larger area of liquid, ensuring
u Film flow area coefficient complete and efficient combustion, minimizing the con-
c Tilt angle sumption of propellant.
k Resistance coefficient of Blasius A bipropellant injector is comprised of two single swirl
l Liquid absolute viscosity nozzles positioned concentrically, ensuring the mixing of
r Gradient operator the propellants in the exit orifice (Fig. 2); the bipropellant
w Helix angle injector used in this work is comprised of a closed injector
q Liquid density (which has a converged portion inside of it) and an open-
r Liquid surface tension end injector, which is completely cylindrical, being Lmix,
m Liquid kinematic viscosity the distance between the outlet orifices of the two nozzles.
v Volume fraction of fluid ra2
u¼1 ð1Þ
ro2
Subscripts
a Air core In the description of the flow behavior inside swirl
eq Equivalent parameter due to viscosity injector, over the last decades, various analytical models,
inj Parameters of inlet ports experimental and numerical methods have been utilized.
p Inlet ports For example, the model for flow inside swirl atomizers was
r Radial component published by Taylor [21] and a few years later by Giffen
s Swirl chamber and Muraszew [6]. In the Soviet Union, the works of
tot Total Glushko in 1932 and Abramovich [1] define a standard
h Tangential component methodology for swirl injector design. Chinn [4] revisited
z Axial component swirl injector theory with reference to the principle of
maximum flow well known as Abramovich solution
[1, 21, 6].
This study shows that the authors concluded basically in
1 Introduction the same inviscid formulation for pressure swirl atomizer
internal flow. The effect of swirler geometric parameters on
A centrifugal injector is characterized by a number of input spray characteristics has been experimentally investigated
channels distributed periodically around its swirl chamber by Kim et al. [10] and Chu et al. [5]. Although not all
(Fig. 1), and these channels are responsible for the fluid geometric parameters can be investigated, these
rotational movement within the injector which causes
inside the injector a cylindrical vacuum called ‘‘air core,’’
as can be seen in the right side of Fig. 1, which may relate

Fig. 1 Left side: main parts of the simple swirl injector. Right side:
annular section in the orifice discharge Fig. 2 Diagram of a bipropellant swirl injector

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
experimental studies contribute to validate the importance
of most analytical models. Attempts were also made
numerically to explore the underlying mechanisms of fluid
injection and combustion. On the other hand, an interesting
investigation was conducted by Hinckel et al. [8].
The purpose of this work was to determine the accuracy
of the present mathematical model [16], with experimental
data and numerical simulation of the behavior of the
internal flow in a bipropellant swirl injector, similar to that
one used in the RD-0110 rocket engine (Fig. 2), being
presented. Thus, by comparing these three methods, we can
obtain better validations, representations, and understand-
ing about the flow behavior that may serve as a useful tool
in the preliminary design of a bipropellant swirl injector. In
the centrifugal bipropellant RD-0110, the nominal param-
eters of the oxidizer and fuel injector are shown in Table 1
[18]. The discharge coefficient was obtained by applying
Eq. (14).
For the experimental measurements, the injectors test
bench of the Aeronautical Engineering Laboratory from
Technological Institute of Aeronautics (ITA)was used, the
numerical simulation utilized a commercial CFD package,
and the turbulent model employed was RNG k-epsilon.
With respect to the mathematical model of Rivas et al.
[17], this has been adopted to use in a conical swirl
injector, which presents a more complex distribution of its
inlet channels with respect to the tangential channels of the
swirl injectors of the RD-0110 rocket engine. It is worth
noting that the mathematical model of Rivas et al. [17] was
motivated by the original works of Abramovich [1, 2] and
Kliachko [11].
2 Numerical simulation
To construct and model a mesh, capable of representing the
domain of a bipropellant centrifugal injector, are very
difficult because there are three large zones to be analyzed:
oxidizer injector zone, the fuel injector zone, and the dis-
charge zone of both nozzles (spray zone), what leads to the
consideration of a large number of elements (cells) and a
Table 1 Nominal parameters of bipropellant swirl injector RD-0110
Nominal parameters Oxidizer Fuel
Geometric parameter, A 2 24.5
Mass flow rate, m_ (kg/s) 0.1729 0.0648
Differential pressure, DP (kPa) 426 696
Number of inlet ports, n 6 6
Discharge coefficient, Cd 0.289 0.022
Spray angle 2a (°) 85 135
Page 3 of 16 289
high computational cost; however, despite these draw-
backs, this paper shows the behavior of the fluid inside of
this type of injector, through the use of a structured three-
dimensional mesh, taking into account rigorous care in
irregular geometries among the input channels and the
swirl chamber, preventing the formation of negative vol-
umes, not to affect the quality of the mesh. The structured
mesh of this work was created using the ICEM CFD
software and presents the ‘‘O grid’’ format in the 12 input
channels, two swirl chambers, and cylindrical zone adja-
cent to the exit of the bipropellant injector, facilitating the
refinement in the internal walls of the spray area (Fig. 3);
this computational domain comprises elements 2,149,825
and 2,205,408 nodes, and the distance used for Lmix in this
mesh is 1.5 mm (right side of Fig. 3).
The first part of the numerical simulation was performed
in steady state, being the velocity inlet, considered as a
condition of entry for both injectors. This was the most
advisable option to be used, as it facilitates the calculation
of the turbulent intensity (I = 0.16 Re-1/8), where Re is the
Reynolds number in the tangential inlet channel, which is
based on the hydraulic diameter (Table 2), obtaining in this
way, the main entry conditions for the turbulence in each
inlet channel. This steady-state simulation case comprises
seven input velocities to the oxidizer injector (7.5, 8, 9, 10,
11, 12, and 13 m/s) and eight cases of inlet velocities to the
fuel injector (14, 16, 18, 20, 22, 24, 26, and 28 m/s). Note,
in this first part, when the fuel injector is actuating, the
oxidizer injector is disabled and vice versa.
The second part of this simulation was performed in
transient state, where the flow of both injectors is injected
at the same time, under the entry conditions: DP = 150
kPa, for the fuel injector and DP = 200 kPa, for the oxidant
injector. With this example, we can see the phenomena
observed in both swirl chambers in more detail, as the
transient state simulation requires more computational
time; a single example was shown for different time
intervals.
Fluent CFD can process data with processors in parallel,
reducing computational time; in this case, the simulations
were done on a computer with 8 GB RAM and four pro-
cessors in parallel. For all cases, the turbulent model
employed was RNG k-epsilon; a no-slip condition was
adopted on the walls, and in the outlet of the bipropellant
injector, a gauge pressure was assumed to be equal to zero.
Being the incompressible flow which is one of the most
important and complex applications of fluid mechanics,
where the continuity equations (Eq. 2) and momentum
(Eq. 3) are independent, the algorithm correction of pres-
sure–velocity coupling was employed: SIMPLE ‘‘Semi-
Implicit Method for Pressure-Linked Equations’’ by
Patankar [13]:
123
289 Page 4 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

Fig. 3 Upper left: oxidizer swirl injector, bottom left: fuel swirl injector and right side: vertical section of the total mesh of the bipropellant swirl
injector

Table 2 Hydraulic diameters of inlet channels of bipropellant swirl

injector
Hydraulic diameter, Dh (mesh) Oxidizer injector Fuel injector

Inlet channel (Dh = 2rinj) 1.7 mm 0.7 mm

oq
þ r  ðquÞ ¼ 0 ð2Þ
ot
 
o 1 1
ðuÞ þ rðuuÞ ¼ rp þ lrðr  uÞ þ lr2 u þ f
ot q 3 Fig. 4 Velocity distribution in turbulent flow near wall [19]
ð3Þ 1
tþ ¼ lnyþ þ B ð4Þ
where u represents the velocity vector; f is the vector of k
volumetric forces; l the absolute viscosity of the liquid; q where the constant B = 5, 2 and t? is a quantity for the
the liquid density; and p the pressure. tangential tt velocity defined by:
For the formulation in the boundary conditions wall;
tt
there are two possibilities to tackle this problem: the low tþ ¼ ð5Þ
Reynolds models (viscous layer) and the wall functions ut
approaches. In the present study wall function by means of y? denotes a normalized distance to the nearest wall point:
the logarithmic layer was used, the physical background of qu y
yþ ¼ lt p , where yp is the distance from the wall to the first
this approach is that in a fully developed turbulent flow a node of the mesh, and the wall shear stress velocity:
logarithmic wall law is valid, i.e., the velocity beyond the qffiffiffiffi
laminar layer logarithmically increase in the certain range U ¼ ut ¼ sqw . Finally, the logarithmic layer is valid
[19], see Fig. 4. approximately in the range 30 B y? B 300 [19].

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
Then, for wall mesh, was applied the methodology
logarithmic layer. It is worth emphasizing that y? was
obtained for both injectors, where the approximations
consist in the range 30 B y? B 300.
The atomization device has very small dimensions,
where the experimental measurements inside of it would be
difficult to obtain; however, with the VOF multiphase
model (HRIC ‘‘High-Resolution Interface Capturing’’), this
problem can be solved, which will give the liquid/gas
location, liquid film thickness, and spray angle. The VOF
‘‘Volume of fluid’’ model [7] found that the volume of a
phase cannot be occupied by another, thus arising the
concept of phase volume fraction, being the sum of these
fractions equivalent to the unit (Eq. 6). These details will
provide better information in the analysis of physical
phenomena at each point of the atomizer.
vliquid þ vgas ¼ 1; ð6Þ
where v represents the volume fraction of the fluid.
Finally, the results in steady state of the numerical
simulation are shown as a function of the inlet velocity, as
shown in Tables 3 and 4, from oxidizer swirl injector and
fuel swirl injector, respectively. The spray angle mea-
surements were obtained using the VOF method.
3 Experimental method
For the experimental measurements, the injectors test
bench of the Aeronautical Engineering Laboratory from
ITA was used and the layout of the test bench is shown in
Fig. 5. The bipropellant centrifugal injector is set internally
to the body injection (Fig. 6), where it will be subjected to
injection pressures (Fig. 7), and the liquid used to simulate
the propellant was water liquid.
SED brand rotameters were used to obtain the mass flow
rate; two metal cryo brand reservoirs with a capacity of
50 L and maximum pressure of 8.29 kgf/cm2 to the accu-
mulation of fluid, a line of compressed air from a com-
pressor external to the test bench were used to pressurize
the fluid reservoirs: The pressure measurement of both
reservoirs was made by ASTA analog gages in the
Table 3 Numerical main
Inlet velocity: Uin (m/s) Injection pressure: Dp (kPa)
parameters as a function of inlet
velocity for oxidizer swirl 7.5 145.14
injector [16]
8.0 165.33
9.0 209.63
10.0 259.48
11.0 314.47
12.0 375.20
13.0 440.64
Page 5 of 16 289
measuring range from 0 to 800 kPa, with uncertainty of
± 5 kPa.
The pressure ranges used in the tests were 150–450 kPa
for the injection of liquid into the oxidizer injector
(Table 5) and 150–600 kPa for the fuel injector (Table 6);
in total, five tests were carried out to obtain the average
mass flow rate, and consequently, the mean Cd for each
injection pressure (criterion that was used for both injec-
tors). Again see Tables 5 and 6, which show the standard
deviations.
The experimental discharge coefficient was obtained
using Eq. (14), being the diameter outlet orifice for oxi-
dizer injector: do = 5.1 mm and do = 10 mm for fuel
injector. The experimental data of discharge coefficient in
both injectors are very close to the nominal data in Table 1
(Cdox ¼ 0:289 and Cdfuel ¼ 0:022), and they have a quasi-
constant behavior with increasing pressure.
Finally, measurements of the spray angle were obtained
through photographs of a digital cam. The spray angle
tends to grow with increasing injection pressure. Never-
theless, for values higher than the nominal injection pres-
sure, the spray angle is fully developed and no longer
grows. The photographs were taken, taking into account
the position of the lens in a plane parallel and perpendic-
ular to the axis of the bipropellant injector, and the angles
were processed from the free software Meazure (viewers,
photographs, and graphics).
4 Mathematical model
In this first part of the mathematical model, the postulate
Abramovich is adopted [2], where the liquid is considered
ideal (incompressible and without friction, l = 0), and then
the Navier–Stokes equations are reduced to Euler equations
(Eq. 7) and then to the Bernoulli equation (Eq. 8), where
the flow is considered one-dimensional with permanent
movement along a streamline.
o 1
ðuÞ þ rðuuÞ ¼  rp þ f ð7Þ
ot q
Mass flow rate: m_ (kg/s) Spray angle, 2a (°)
0.0969 65.0
0.1034 65.5
0.1163 66.2
0.1293 67.0
0.1422 67.5
0.1551 68.0
0.1680 69.0
123
289 Page 6 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

Table 4 Numerical main

Inlet velocity: Uin (m/s) Injection pressure: Dp (kPa) Mass flow rate: m_ (kg/s) Spray angle, 2a (°)
parameters as a function of
differential pressure for fuel 14 160.83 0.0322 115.3
swirl injector [16]
16 206.53 0.0368 115.4
18 257.77 0.0414 115.5
20 314.78 0.0460 115.6
22 377.68 0.0506 115.7
24 445.39 0.0552 115.9
26 521.63 0.0598 116.0
28 603.08 0.0644 116.2

Fig. 5 Experimental setup

schematic

Fig. 6 Bipropellant swirl injector inserted into the body injection

Fig. 7 Injection body set on the test bench

u2 p m_
þ þ gz ¼ cte: ð8Þ Uin ¼ ð9Þ
2 q qnfp
In Fig. 8a, b, we can display geometric differences Besides, in the conical swirl injector (Fig. 8a), the
between a conical swirl injector and a swirl injector with relationship between inlet velocity tangential with and b
tangential inlets, where the main difference lies in the helix angles into swirl chamber is:
angle ‘‘w’’ [14], which is contained in a vertical plane
parallel to plane xz and the swirl angle ‘‘b’’, which can be Whin ¼ Uin cosw  sinb ð10Þ
seen better in top view on both injectors. Assuming the hypotheses of a constant angular moment
Then, was applied the equation continuity for obtaining and constant axial velocity at the cross-section area of the
the inlet velocity in the injector channels (Eq. 9). exit orifice (radial position ‘‘ra’’ in Fig. 1), we have the

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289 Page 7 of 16 289

Table 5 Experimental main parameters as a function of differential pressure for oxidizer swirl injector [16]
Differential pressure, DP (kPa) Mass flow rate, m_ (kg/s) _
SD (m) Discharge coefficient (Cd) SD (Cd) Spray angle, 2a (°)

150 0.0971 0.0037 0.274428 0.0106 82.0

200 0.1121 0.0029 0.274376 0.0071 82.5
250 0.1253 0.0017 0.274306 0.0038 83.2
300 0.1371 0.0023 0.273988 0.0046 83.8
350 0.1480 0.0018 0.273831 0.0034 84.6
400 0.1581 0.0011 0.273626 0.0020 85.3
450 0.1676 0.0012 0.273478 0.0020 85.8

Table 6 Experimental main parameters as a function of differential pressure for fuel swirl injector [16]
Differential pressure, DP (kPa) Mass flow rate, m_ (kg/s) _
SD (m) Discharge coefficient (Cd) SD (Cd) Spray angle, 2a (°)

150 0.0317 0.0014 0.023303 0.0010 –

200 0.0366 0.0010 0.023300 0.0007 124.5
250 0.0409 0.0023 0.023289 0.0013 –
300 0.0448 0.0021 0.023287 0.0011 127.0
350 0.0484 0.0029 0.023292 0.0014 –
400 0.0517 0.0023 0.023273 0.0010 129.2
450 0.0549 0.0028 0.023300 0.0012 –
500 0.0578 0.0053 0.023272 0.0010 132.3
550 0.0605 0.0052 0.023226 0.0011 –
600 0.0631 0.0053 0.023193 0.0012 134.2

Fig. 8 a Main parameters of a conical swirl injector and b main parameters of a tangential swirl injector

tangential and axial velocities in Eqs. (11) and (12), In Bernoulli equation (Eq. 13), at the air liquid interface
respectively: (radial position ‘‘ra’’), the radial component is not consid-
Uhin  Rinj ered and the axial velocity is considered constant [12]. The
Whra ¼ ð11Þ
ra relation between mass flow rate and total pressure drops is
m_ expressed through the coefficient discharge in Eq. (14).
Wzra ¼   ð12Þ
qp  ra2
ro2 Wz2ra Wh2ra
DP ¼ q þq ð13Þ
2 2

123
289 Page 8 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

m_ through the channel walls and is obtained in function of the

Cd ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð14Þ
pro2 2q  DP coefficient of friction of Blasius, by Eq. (23) [20].
sffiffiffiffiffiffiffiffiffi
According to the maximum mass flow rate principle, is 2Dp qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
obtained the maximum amount of the discharge rate of the Utotal ¼ ¼ Wz2a þ Wh2a þ ntot Uin2 ð19Þ
q
injector for an optimal amount of u, dCd/du. From this
relationship are got the principal geometric characteristics 1
Cdeq ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi ð20Þ
of the swirl atomizer and the film flow area coefficient u. ðAK Þ2
2
1
u2eq þ þ ntot RArinjo
The geometrical parameter of the conical swirl injector 1ueq

‘‘Ac’’ has an important role in the design procedure [17],

ntot ¼ ninj þ nk ð21Þ
according to Eq. (15)

p  ro  Rinj linj
Ac ¼ cosw  sinb; ð15Þ c ¼ arctan ð22Þ
n  fp Rs
linj
which can also be applied in centrifugal injectors with nk ¼ k ð23Þ
2rinj
tangential inlets, where the angles ‘‘w’’ and ‘‘b’’ take values
of remarkable angles (w = 0°, b = 90°), being reduced to k is the coefficient of friction of Blasius (k = 0.3164
Abramovich number, ‘‘A,’’ in Eq. (16). Re-0.25), and Re is the Reynolds number at the inlet of the
pffiffiffi injector, which is calculated using Eq. (24), where m is the
p  ro  Rinj 2ð1  uÞ
A¼ ¼ pffiffiffiffi ð16Þ kinematic viscosity of the working fluid:
n  fp u u
Uin Dh
Constructive geometric parameters shown in Eqs. (15) Re ¼ ð24Þ
m
and (16) are based on the coefficient ‘‘u,’’ which facilitates
the obtaining of other geometric parameters such as the Deriving the discharge coefficient of Eq. (20) with
discharge coefficient and spray half angle using Eqs. (17) respect to the coefficient of annular equivalent section
and (18), respectively [9]: ‘‘ueq’’ and applying the principle of Abramovich maximum
sffiffiffiffiffiffiffiffiffiffiffiffi flow, Eq. (25) is obtained, the equivalent of constructive
u3 geometric parameter Kliachko ‘‘Aeq’’ [11]; accordingly, by
Cd ¼ ð17Þ
2u inserting Eq. (25) into Eq. (20), we have the equivalent
pffiffiffi discharge coefficient as ‘‘ueq’’ function (Eq. 26):
2 2ð 1  u Þ pffiffiffi 
sina ffi pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi ð18Þ 2 1  ueq
ð1 þ 1  u Þ 2  u Aeq ¼ AK ¼ pffiffiffiffiffiffiffi ð25Þ
ueq ueq
Then, these main parameters can relate to the function
‘‘u’’ through Fig. 9 (see Eqs. 16, 17, and 18), where the 1
Cdeq ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ð26Þ
liquid is assumed to be perfect, being able to appreciate 2ueq 2
Aro
u3eq þ ntot Rinj
that when the coefficient ‘‘u’’ is equal to unity, there is no
rotational movement, so ‘‘air core’’ does not exist and the Finally, we get the half angle of spray (Eq. 27), con-
centrifugal injector behaves like a ‘‘jet’’ with discharge sidering the losses due to the viscosity of the liquid [3]:
coefficient being equal to a unity. The constructive geo-
metric parameters of the injector of oxidant and fuel are: 2Cdeq Aeq
sin aeq ffi  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð27Þ
Aox = 2 and Afuel = 24.5, respectively (Table 1). pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2
1 þ 1  ueq 1  ntot Cd2eq RArinjo
In a real flow, is considered the liquid viscosity, the
angular momentum losses ‘‘K’’ and hydraulic losses ‘‘ntot’’
so considering these losses in Bernoulli’s equation, shown
in Eq. (19). The ‘‘K’’ coefficient directly influences the
5 Results
velocity tangential component and hydraulic losses ntot in
the input channels, as can be seen in the actual rate of
5.1 Oxidizer swirl injector (closed injector)
discharge or equivalent (Eq. 20), and such hydraulic losses
are related by the sum of ninj and nk (Eq. 21), being ninj the
In Figs. 11 and 12, the contours of total pressure of the
coefficient of losses due to the geometry of the inlet channel
oxidizer injector in the steady state from minimum value
and it is a function of tilt angle c and inlet channel length
are shown: Uin = 7.5 m/s (DP = 1.45 9 105 Pa) and
‘‘linj’’(Eq. 22) and is obtained from empirical data shown in
maximum value: Uin = 13 m/s (DP = 4.41 9 105 Pa),
Fig. 10. nk is the coefficient of losses due to the flow friction

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289 Page 9 of 16 289

Fig. 9 Discharge coefficient,

spray angle, and film flow area
coefficient ‘‘u’’ in terms of
constructive geometric
parameter A

Fig. 10 Coefficient ‘‘ninj’’ as a

function of the tilt angle ‘‘c’’

respectively. In Fig. 11, the existence of pressure is seen in

the spray zone, because the liquid is still present as a
conical surface, which shows that there is no complete
atomization in this zone. The provenience of vortices in the
swirl chamber is due to the geometry of the oxidizer swirl
injector (closed injector); arising pressure losses in this
region, the effects of viscosity and friction on the wall are
significant. Figure 12 shows the total pressure contour of
the oxidizer injector in horizontal view, where the radial
pressure behavior is due to the equation of angular motion .
In Fig. 13 is shown the contour of total velocity (left
side) and tangential velocity (right side) in the oxidizer
stage from intermediate value: Uin = 11 m/s
(DP = 3.14 9 105 Pa). In the left side, one can see that
there is low pressure in the region of the ‘‘air core,’’ con-
ducting and axially accelerating the air particles into the
Fig. 11 Contours of total pressure (Pa) of the oxidizer injector in
injector, while in the vicinity of the walls of the conver- vertical plane from Uin = 7.5 m/s
gence zone, the higher velocity module is concentrated,
due to centrifugal forces acting on the fluid in this zone. In

123
289 Page 10 of 16
Fig. 12 Contours of total pressure (Pa) of the oxidizer injector in top
view from Uin = 13 m/s
the right side of Fig. 13, the tangential velocity component
tends to maximum value in the convergent zone of the
swirl chamber; downstream of this position, the tangential
velocity decreases due to the increase of the axial com-
ponent within the nozzle.
In Fig. 14 is shown the contour of axial velocity (left
side) and radial velocity (right side) in the oxidizer stage
from Uin = 11 m/s. On the left side of Fig. 14 is shown the
axial component of velocity, which can be observed that
the axial velocity of the liquid tends to maximum value and
it is considered constant in the section of the annular area
filled by the liquid in the nozzle exit. On the right side of
Fig. 14, the radial velocity component in the internal flow
is approximately zero; this consideration facilitates the
calculation of parameters of interest in the mathematical
Fig. 13 Contour of velocity magnitude (left) and tangential velocity (right) (m/s), in the oxidizer injector, from Uin = 11 m/s
123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
model and is supported with the theories proposed by
Kessaev and Kupatenkov [9] and Lefebvre [12]. However,
in the spray zone, the velocity radial component of liquid is
greater with respect to radial component of the internal
flow, due to the powerful centrifugal forces which are
expelled from the orifice discharge .
Then, the results corresponding to the mathematical
model (assuming viscosity losses), numerical simulation
and experimental data will be compared. In Fig. 15 is
shown the mass flow depending on the pressure differential
in the oxidizer nozzle (Eq. 14), where there is a good
approximation of the mathematical model with respect to
the experimental data; however, the numerical results show
an allowable deviation with the increase of the injection
pressure.
Figure 16 shows the results of the discharge coefficient
(Cd) of the oxidant stage as a function of the differential
pressure. The results of the Cd of the mathematical model
are satisfactory in relation to the experimental data (the
results were obtained using Eq. 26). On the other hand, the
results obtained numerically have a maximum error of
1.5% (at low injection pressures: 150 kPa), with respect to
data obtained experimentally (as shown in Table 5).
The radius of the ‘‘air core’’ increases as the injection
pressure increases because the centrifugal forces acting on
the liquid inside the nozzle, and consequently, the spray
angle tend to increase; this phenomenon can be observed
through the use of the VOF method, in Fig. 17a, for Uin-
= 7.5 m/s and Fig. 17b, for Uin = 13 m/s. In both cases, it
can be seen that the spray does not touch the walls of the
fuel injector, which ensures that the mixture of propellants
is external, showing that the method VOF is useful in the
liquid/gas interface prediction and the measurements of the
liquid film thickness.
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289 Page 11 of 16 289

Fig. 14 Contour of axial velocity (left) and radial velocity (right) in the oxidizer injector (m/s), from Uin = 11 m/s

Fig. 16 Comparison of the discharge coefficient as a function of the

differential pressure of the oxidizer injector
Fig. 15 Comparison of the mass flow rate as a function of the
differential pressure from the oxidizer injector
In Fig. 19, the measures corresponding to the spray
As follows, we can see the experimental measurement
angles for the oxidizer injector are displayed; we can see
of the spray angles of two cases of pressure injection:
that the spray angle obtained with the mathematical model
DP = 150 kPa (Fig. 18a) and 450 kPa (Fig. 18b), using
has good approximation (the results were obtained using
water as the working fluid; in the case DP = 150 kPa, it is
Eq. 27) regarding the experimental data. Nevertheless, the
seen that there is no complete atomization (no rupture of
results of numerical simulation show considerable devia-
the liquid cone surface), and for the case of increased
tion, perhaps because these measurements were made
injection pressure (DP = 450 kPa), there is the presence of
within a small region, region that had a limited presence of
disturbance waves causing the collapse of the liquid sur-
nodes, making it difficult to visualize the interface between
face and ligaments formation, leading to the atomization in
the water/air fluids; therefore, such limitations are related
distant areas of the outlet orifice, resulting in very large
to the mesh, which only has a distance of 9 mm below the
drop sizes. At this stage, the atomization is completely
outlet orifice to display spray (grid of Fig. 3); this distance
developed and the spray angle no longer presents any
does not predict the actual measurement of the developed
significant variation [22].

123
289 Page 12 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

Fig. 17 Contours of volumetric fraction from: a Uin = 7.5 m/s and b Uin = 13 m/s, of the oxidizer swirl injector

Fig. 18 Pictures of the spray

angle: a DP = 150 kPa and
b DP = 450 kPa of the oxidizer
injector

5.2 Fuel swirl injector (open-end injector)

In Figs. 20 and 21 are shown the contours of the total

Fig. 19 Spray angles as a function of differential pressure from
oxidizer injector
spray angle. In photographs, the length downstream of the
injector outlet is 100 mm minimum; this space allows for
angle measures fully developed spray.
pressure of the fuel injector in steady state for Uin = 14 m/s
(DP = 1.61 9 105 Pa) and Uin = 28 m/s
(DP = 6.03 9 105 Pa), respectively; according to Khavkin
[11], the ‘‘K’’ constant of loss due to decreasing amount of
angular momentum is equal to the value maximum (K = 1)
in open-end swirl injectors, where Aeq = A (Eq. 25), and it
can be seen in the fuel swirl chamber, in which a minimum
number of vortices are recorded, and the effects of vis-
cosity and pressure losses can be neglected. Furthermore,
because the channel length of fuel injector is greater than
the oxidant injector, there are more losses due to the fric-
tion in the inlet channels of fuel injector.
In Fig. 22 are compared the contour of the total velocity
(left side) and the contour of the tangential velocity com-
ponent (right side) from Uin = 26 m/s (DP = 5.22 9 105
Pa), where one can appreciate that the magnitude of the
tangential component of liquid is just less than total
velocity; this examination indicates that the tangential
component is greater than the axial and radial components,

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
Fig. 20 Contour of total pressure (Pa) of the fuel injector in vertical
plane from Uin = 14 m/s
Fig. 21 Contour of total pressure (Pa) of the fuel injector in
horizontal view from Uin = 28 m/s
as analyzed theoretically, due to the constant radius that
presents the swirl chamber of an open-end injector.
In Fig. 23 is shown the contour of the axial component
of velocity (left side) from Uin = 26 m/s, which can be
observed that the axial velocity in the liquid film adjacent
to the walls of the swirl chamber has a uniform behavior,
which indicates that the axial component is also considered
constant in the annular area formed by the liquid in the
outlet section; furthermore, it can be noted the axial
direction acceleration of air particles between the internal
wall of the swirl chamber and the external nozzle wall of
the oxidizer injector, due to the low pressure registered
between both walls.
Page 13 of 16 289
On the right side of Fig. 23, we can view the low
modules of the radial component that presents the fuel
injector which was appreciated in the internal flow of the
oxidizer injector; for this reason, in the mathematical
model, the radial component is not considered, in order to
eliminate variables in the mathematical model. On the
other hand, the greater radial component of liquid is reg-
istered in the spray zone.
In Fig. 24 is shown the mass flow as a function of dif-
ferential pressure to the fuel injector, where we see that the
three techniques agree up to 350 kPa; after this point, the
numerical simulation and mathematical model acquire a
similar behavior deviating slightly from the experimental
curve.
In Fig. 25 are shown the discrepancies of the results of
the ‘‘Cd’’ mathematical model with respect to the experi-
mental data, presenting a maximum error of 14.72% at the
point of the injection pressure 600 kPa. In the case of the
results obtained numerically, a good approximation can be
observed regarding the experimental tests, presenting a
maximum error of 2.59% at the point of the injection
pressure 600 kPa. (The numerical result at this point is
within the margin of experimental error 0.02319 and SD
0.0012.) The three curves of the Cd tend to have a behavior
almost constant as the pressure increases.
In Figs. 26 and 27, the behavior of the internal flow in
the fuel injector is shown using the VoF method in steady
state for Uin = 14 m/s (DP = 1.61 9 105 Pa) and Uin-
= 28 m/s (DP = 6.03 9 105 Pa), respectively; the open-
end injector has a cylindrical form (Rs = ro), which makes
the liquid film very thin and constant along the walls of the
injector; the thickness of the liquid film is designed so there
may be a space capable of keeping the oxidizer nozzle into
swirl chamber fuel; these calculations are related to the
film flow area coefficient ‘‘u,’’ which is an important
parameter in the design of the bipropellant swirl injectors.
In both cases, it may be seen that the rupture of the conical
surface of the liquid occurs very near the discharge orifice,
which shows that in this zone, the atomization is imminent
at low pressures.
In Fig. 28 are shown photographs of the spray angles of
the fuel injector, for DP = 200 kPa (left side) and DP =
600 kPa (right side); in both cases, the spray angle is
greater than 100°; this promotes atomization because the
liquid sheet exhibits a high velocity of rotation, shaping
and amplifying the oscillations which cause the rupture of
the liquid sheet at a distance even closer from the outlet
[12]. In addition, compared to the oxidizer injector, the
rupture of the fuel liquid sheet is closer to the discharge
orifice, because the film thickness is thinner than that of the
oxidizer injector (in Table 1 and oxidant mass flow rate is
2.7 times the fuel mass flow rate).
123
289 Page 14 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289

Fig. 22 Contour of velocity magnitude (left) and tangential velocity (right) in the fuel injector (m/s), from Uin = 26 m/s

Fig. 23 Contour of axial velocity (left) and radial velocity (right) in the fuel injector (m/s), from Uin = 26 m/s

In Fig. 29, the measures corresponding to the spray
angles for the fuel injector are displayed; we see that the
spray angle obtained with the mathematical model has
reasonable approximation regarding the experimental data,
taking into account the range of the angles measured in
degrees and the ‘‘restriction’’ of the mathematical model to
fuel injector (open-end injector). However, the results
obtained for spray angle with the mathematical model
corresponding to the spray angles for the oxidizer injector
were more accurate (see Fig. 19). On the other hand, the
results of numerical simulation show a considerable devi-
ation with respect to experimental data (17% approxi-
mately); this is due to limitations of the mesh in the
interface, exposed and described in the case of the spray
angles obtained in the numerical simulation for the oxidizer
swirl injector, Fig. 19.
6 Conclusions
The mathematical model developed in this work is rooted
in the Euler equations (Navier–Stokes equations without
considering the viscosity), which shows satisfaction and
confidence in their use; where the factor ‘‘coswsinb’’ as a
correction or adjustment is of importance [17], for use
generally in the various kinds of centrifugal injectors, being

123
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
Fig. 24 Comparison of the mass flow rate as function of the
differential pressure from the fuel injector
Fig. 25 Comparison of the discharge coefficient as a function of the
differential pressure of the fuel injector
Fig. 26 View of the internal flow and spray angle for Uin = 14 m/s
Page 15 of 16 289
Fig. 27 View of internal flow and spray angle for Uin = 28 m/s
shown that this mathematical model is applicable in the
design of open-end and closed swirl injectors of tangential
inlet channels.
Understanding the flow behavior inside and outside the
centrifugal injector bipropellant is achieved through the use
of multiphase model VOF, the resolution of the liquid/gas
interface, and the viscosity model k-epsilon RNG, which is
the most suitable for flows with swirl effect, obtaining very
approximate results to the experimental data; we can also
see that the behavior of the flow inside a closed swirl
injector is different than an open-end swirl injector,
because each of these injectors has a different format in its
swirl chamber.
The mesh created has the advantage of working with
several options, depending on user requirements, for
example, can work with a single injector flow through
certain input conditions (velocity, pressure, or mass flow
inlet), disable or close the other injector using the input
condition: wall; the other option was also used in this work
in transient state, which consists of the operation of open-
end and closed swirl injectors at the same time, being the
option that demands greater computational cost by working
with almost all the domain bipropellant injector.
The structured mesh used contains 2,205,408 nodes and
is considered a robust mesh processor type used in this
work; for this reason, the spray area was not expanded
because this would increase the computational cost; due to
this, are the discrepancies between the results obtained
spray angle should by numerical simulation, with respect to
the experimental data (in the discharge zone, there is less
number of nodes compared to the amount lodged and
refined inside the mesh ).
123
289 Page 16 of 16 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:289
Fig. 28 Pictures of the spray
angle for DP = 200 kPa (a) and
DP = 600 kPa (b) of the fuel
injector
Fig. 29 Spray angles as a function of differential pressure from fuel
injector
In this present study, comparing the mathematical
model, experimental data, and numerical simulation of the
behavior of the internal flow in a swirl bipropellant ato-
mizer, similar to that one used in the RD-0110 rocket
engine, we can obtain better validations, representations,
and understanding about the flow behavior that may serve
as a useful tool in the preliminary design of a bipropellant
swirl injector.
Finally, expanding the area of spray can be suggested
for future work, since it has the required computational
resources and thereby obtains a better viewing angle spray
fully developed. Also, is essential the complement of
chemical reaction tests for further approximation of the
actual phenomena that exist in a combustion chamber.
Acknowledgements The authors would like to thank the following
institutions which have contributed to the development of this work:
Technological Institute of Aeronautics, ITA, and its Propulsion and
Power Laboratory, as well as the Coordenação de Aperfeiçoamento
de Pessoal de Nı́vel Superior, CAPES, for the financial support
through all these years of study.
References
1. Abramovich GN (1944) The theory of swirl atomizers. Industrial
Aerodynamics, BNT ZAGI, Moscow
2. Abramovich GN (1976) Applied gas dynamics. Nauka, Moscow
123
3. Bazarov V, Yang V (2004) Design and dynamics of jet and swirl
injectors. Prog Astronaut Aeronaut 200:19–103
4. Chinn JJ (2009) An appraisal of swirl atomizer inviscid flow
analysis, part 1: the principle of maximum flow for a swirl ato-
mizer and its use in the exposition and comparison of early flow
analyses. Atom Sprays 19:263–282
5. Chu C-C, Chou S-F, Lin H-I, Liann Y-H (2008) An experimental
investigation of swirl atomizer sprays. Heat Mass Transf
45:11–22
6. Giffen E, Muraszew A (1953) Atomization of liquid fuels.
Chapman and Hall, London
7. Hirt CW, Nichols BD (1981) Volume of fluid (VOF) method for
the dynamics of free boundaries. J Comput Phys 39:201–225
8. Hinckel JN, Villa Nova HF, Bazarov VG (2008) CFD analysis of
swirl atomizers. In: 44th AIAA/ASME/SAE/ASEE joint propul-
sion conference and exhibit
9. Kessaev K, Kupatenkov VD (1997) Injectors design for liquid
rocket engines. Book of Fundamental Course in Engine Design,
SP-São José dos Campos, CTA/IAE/ASA-P, pp 31–49
10. Kim D, Im J-H, Koh H, Yoon Y (2007) Effect of ambient gas
density on spray characteristics of swirling liquid sheets. J Propul
Power 23(3):603–611
11. Khavkin YI (2004) The theory and practice of swirl atomizers.
CRC Press, New York
12. Lefebvre AH (1989) Atomization and sprays. Hemisphere, New
York
13. Patankar SV (1980) Numerical heat transfer and mass transfer.
Hemisphere, New York
14. Reddy KU, Mishra DP (2008) Studies on spray behavior of
pressure swirl atomizer in transition regime. J Propul Power
24:74–80
15. Rivas JRR (2009) Estudo e simulação numérica do escoamento
no interior de um injetor centrı́fugo cônico. MSc thesis, Instituto
Tecnológico de Aeronáutica, Brazil
16. Rivas JRR (2015) Modelo matemático e simulação numérica da
atomização de lı́quidos em injetores centrı́fugos de uso aeroes-
pacial. PhD thesis, Instituto Tecnológico de Aeronáutica, Brazil
17. Rivas JR, Pimenta AP, Rivas GR (2014) Development of a
mathematical model and 3D numerical simulation of the internal
flow in a conical swirl atomizer. J At Sprays 24(2):97–114
18. Rubinsky VR (1994) Combustion instability in the RD-0110
engine. AIAA, Washington, pp 89–112
19. Schafer M (2006) Computational engineering—introduction to
numerical methods. Springer, Berlin
20. Souza JRP (2001) Estudo de um injetor centrı́fugo bipropelente
utilizado em motor foguete a propelente lı́quido. MSc thesis,
Instituto Tecnológico de Aeronáutica, Brazil
21. Taylor G (1948) The mechanism of swirl atomizers. In: Pro-
ceedings of the 7th international congress for applied mechanics,
London, vol 2
22. Wehmann CF (2010) Caracterização do funcionamento de um
injetor centrı́fugo bipropelente. PhD thesis, Instituto Tecnológico
de Aeronáutica, São José Dos Campos