Sie sind auf Seite 1von 24

Turbulence Modeling for Realistic Computation of Internal Flow in Liquid

Ejector Pumps

J. Masud1 and M. Imran2


Department of Mechanical & Aerospace Engineering, Institute of Avionics & Aeronautics,
Air University, Islamabad 44000, Pakistan

Abstract

In this paper flow analysis of a liquid fuel ejector pump is done by using
Computational Fluid Dynamics technique. The flow inside the ejector pump
is modeled as incompressible, steady and turbulent that is governed by the
RANS system of equations along with the two-equation K- family
(Standard, Realizable, Re-normalized Group etc.) of turbulence models for
closure. Due to the complex nature of flow inside the pump, the computed
results show that all the used turbulence models over-predict pump flow rate
in relation to the test data, with the Standard K- model showing the least
difference. This indicates that the default values of constants of the K-
family of turbulence models, specially the Standard K- model (C=0.09,
C1=1.44, C2=1.92, k=1.0 and =1.3), need adjustment in order to correlate
well with the experimental data. The values of these constants for the
standard K- model are then systematically adjusted for the unique flow field
1
2

Associate Prof, Dept of Mech. & Aero. Engg, IAA, Air University, Islamabad 44000, Pakistan.
Graduate Student, Dept of Mech. & Aero. Engg, IAA, Air University, Islamabad 44000, Pakistan.

inside the fuel ejector pump in order to computationally reproduce the test
data. The results of this study indicates that the computed effective viscosity,
that affects the pump flow rate, can either be controlled by varying the
constants C1 and C2 in a particular ratio, or by adjusting C independently.
Nomenclature
CFD =

computational fluid dynamics

Pin

pump inlet (nozzle) pressure

Pout =
P

pump outlet pressure


pressure

KW =

two equation K- turbulence model

MKE =

K-e turbulence model with modified constants

RNGKE two equation Re-normalized Group K- turbulence model


RKE =

two equation Realizable K- turbulence model

SKE =

two equation standard K- turbulence model

turbulence kinetic energy

turbulence kinetic energy dissipation rate

C1

constant in SKE turbulence model related to production of

C2

constant in SKE turbulence model related to dissipation of

constant in SKE turbulence model related to turbulent eddy viscosity

effective Prandtl number (constant) in SKE turbulence model for K

effective Prandtl number (constant) in SKE turbulence model for

Vy

radial velocity

Vx

axial velocity

Turbulent or eddy viscosity

axial coordinate

radial coordinate

y+

non-dimensional length scale associated with turbulence model


I.

Introduction

Ejector Pump (EP) is an important component of the fuel system of many modern aircraft and is
used for pressurizing / transferring fuel from various fuel tanks of the aircraft. The ejector pump
is extremely safe, reliable and has simple lightweight construction due to the fact that it has no
moving components. The overall efficiency of the fuel system of aircraft that employ ejector
pumps of various sizes is strongly dependent on the operating behaviour of these devices.
The fuel ejector pump, that is the subject of this study (referred to as the subject pump), has
been designed and tested as part of fuel system development of a multi-role fighter aircraft 9. The
geometry of the subject fuel ejector pump is shown in Figure 1. The basic operation principle of
the ejector pump is that high pressure fuel is supplied to the primary nozzle, which forms a high
speed jet inside the pump. This high speed jet entrains nearby fuel from the fuel tank via the
secondary nozzle. This fuel then enters the mixing chamber where momentum exchange occurs
resulting in increase in static pressure. Further increase in static pressure is accomplished in the
diffuser section of the pump after which the pressurized fuel leaves the pump. This pressure
variation is qualitatively shown in Fig. 1b for a generic liquid ejector pump.

Figure 1. Geometric layout of the fuel ejector pump.

The ejector phenomenon is not new in the field of aero / fluid dynamics 1,2,3. However, the
efficient application of this phenomenon to the fuel ejector pump offers a number of challenges
to the designer in the form of interrelated geometric (dimensions), fluid (density, viscosity etc)
and fluid dynamic parameters (flow rates, pressures, flow stability etc) that have to be analyzed
and matched in order to have efficient operation10. The use of Computational Fluid Dynamics
(CFD) analysis for a complex flow problem such as the fuel ejector pump first requires
verification against known test data. This test data (previously unpublished) for the present study
is available as part of the fuel system development process of an underdevelopment multi-role
fighter aircraft. Once this verification is done then subsequently CFD analysis can be used for
ejector pump modifications / improvements. The effect of turbulence and turbulence modeling in
devices involving ejector phenomenon has been the focus of significant studies 11-30. However, a
consensus for treatment of turbulence for initial CFD work is still elusive.

Figure 1b.

Qualitative pressure variation inside a generic liquid ejector pump 10.

Preliminary design studies generally involve the usage of simplified CFD strategies, such as
using 2D axisymmetric analysis instead of full 3D computations, such as our earlier work
involving ejector pumps31. In the present computational study, initially the K- family of
turbulence models (Standard, Realizable, Re-normalized Group etc.) 4 and the K- model is used
for steady 2-D axisymmetric analysis. After initial analysis, the standard K- is selected for
parametric refinement. The standard K- is a semi-empirical model based on transport equations
for the turbulence kinetic energy (K) and its dissipation rate (). For closure it uses five
constants4,5,6 i.e. C=0.09, C1=1.44, C2=1.92, k=1.0 and =1.3. The parameter C influences the
overall effective viscosity, while the parameters C1 and C2 appear in the dissipation rate
equation in the production of and dissipation of terms, respectively. The parameters k and
are the turbulent Prandtl numbers. The values these five constants are developed by
experimental data correlation5,6 and are not necessarily universal. Further discussion on the
standard K- turbulence model and its constants is done later in this paper.

In light of the above discussion, the aim of the present study is, therefore, to systematically
determine appropriate values of the constants of the standard K- turbulence model once 2D axisymmetric analysis is done to compute the turbulent flow-field inside of the fuel ejector pump.
This aim is accomplished by matching the computational results with experimental data by
varying the values of the constants (from their default values) and then by analyzing the flow
field in detail. For this purpose a commercial CFD code; ANSYS Fluent7 is used.
II.

Computational Setup

For CFD analysis, in addition to the ejector pump, the whole test setup was modeled as well.
This includs the fuel tank from which the ejector pump draws fuel, the ejector pump nozzle pipe,
the ejector pump outlet pipe, the adjustable restrictor valve in the outlet pipe and the ejector
pump itself. The adjustable restrictor valve was modeled computationally as an orifice restriction
that induces pressure loss. An appropriate loss coefficient value was specified which fulfilled the
initial test condition set point i.e. to achieve a particular value (0.05 MPa) of pump outlet
pressure (Pout) at pump nozzle inlet pressure (Pin) of 1.2 MPa. This was primarily done so that
the experimental / test parameters could be duplicated computationally and direct comparison
between CFD and test results could be made. The recreated geometry and the coordinate system
is shown in Fig. 2.

Figure 2. Computational domain for axisymmetric analysis of the fuel ejector pump.

The geometry of the subject pump was conducive to two-dimensional axisymmetric modeling
instead of three-dimensional; therefore 2-D axisymmetric (about x-axis) geometric model was
created. Quadrilateral elements were used to mesh the computational domain. Elements were
graded near the walls and the nozzle jet to resolve high shear areas in the vicinity of walls and
free jet boundary as shown in Fig. 3.

Figure 3. Partial view of the computational mesh inside the fuel ejector pump.

The CFD software ANSYS Fluent7,8 has been used in the present study. A number of popular
turbulence models, including the two-equation K- and its variants, with different wall
treatments is available. In the present work, incompressible RANS system of equations with
constant property JP-4 fuel (density = 0.7893 g/cm3 and dynamic viscosity = 0.011 dynes-sec /
cm2) was solved using the SIMPLER7 algorithm of with second order implicit formulation in
ANSYS Fluent7. Standard wall-function treatment near the pump walls was employed. No-slip
velocity boundary condition was enforced at the surfaces / walls of ejector pump (Fig. 1 and 2).
Pressure boundary conditions8 were used at nozzle inlet (six different settings), fuel tank inlet
(hydrostatic pressure corresponding to fuel height above the ejector pump) and outlet

downstream (same as fuel tank inlet) of the restrictor valve (Fig. 2). These pressure values
correspond to the available test data. Symmetry boundary condition8 was specified on the
symmetry axis. Since the thermal problem was not of importance in the present study, therefore
the energy equation was not solved.
Grid Sensitivity and Numerical Error Analysis

Numerically computed results change with the type and fineness of the mesh / grid used for
computations. For the present study, a mesh of about 0.045 million nodes (2D axi-symmetric)
gives mesh independent results, however a mesh of about 0.1 million nodes is used in all
computational results presented here. The specific values of important parameters at the
computational / experimental set point (discussed later) are shown in Table 1 below. It is evident
that both the meshes give consistent results. The value of computed turbulent law y + remains
between 30 and 90 all cases, which is well within the acceptable range for standard wall
treatment, and shows reasonable resolution of grid near the walls 5,8. The partial computational
mesh is shown in Fig. 3.
Table 1. Grid sensitivity and numerical error analysis for Standard K- turbulence model.
Secondary
Primary
Primary
Secondary
Pump
Pump outlet
Number
nozzle
nozzle
nozzle
nozzle
outlet
Mass flow
of mesh
mass flow
mass flow
static
static
Static
rate
nodes
rate
rate
pressure
pressure
Pressure
(Kg/s)
(Kg/s)
(Kg/s)
(Pa)
(Pa)
(Pa)
45117

6.6377

1.6232

1200308

1902.7

49961.7

8.2685

101264

6.6431

1.6254

1200236

1891.5

49991.2

8.2609

Extrapolated Infinite
grid solution

6.6444

1.6259

1200218

1888.7

49998.5

8.2590

% error

0.0201

0.0335

-0.0015

-0.1466

0.01461

-0.0228

Grid Convergence
Index (GCI)

0.0006

0.0010

0.00004

0.0044

0.0004

0.0007

The numerical error estimation for the present study is done by following the procedure
outlined in the literature32, 33 on the subject for a twogrid solution. The data presented in Tables 1
for the coarse mesh (45117 node, Table 1) and the fine mesh (101264 nodes, Table 1) is utilized
for this purpose. The extrapolated infinite grid solution and the Grid Convergence Index (GCI)
calculated from the two-grid solution is shown in the last two rows of Table 1. The numerical
error between the fine grid (101264 nodes, Table 1) solution and the extrapolated infinite grid
solution is a very small fraction 1% for all the computed parameters of interest (pressures and
flow rates). The corresponding GCI is also much less than one (GCI << 1) for all the parameters.
This numerical error analysis (Table 1) indicates good accuracy of the computed results with
respect to the discretization strategy used. The GCI in all cases is very small and the % error in
computed solution with respect to the extrapolated asymptotic estimate is also very small in all
cases. The error-bars associated with this very small error cannot be clearly shown in the relevant
figures presented in this work and therefore, are omitted.
III.

Ejector Pump Test Data

The designed ejector pump shown in Fig.1 was tested in a rig as part of aircraft 14 fuel system
development process. The test rig consisted of a fuel tank from which the ejector pump drew the
fuel to be pumped, a high pressure electric driven pump to supply the high pressure fuel to the
ejector pump primary nozzle, a downstream adjustable flow restrictor to control the ejector pump
flow rate (or output pressure), and multiple pressure and flow rate sensors to record the data. The
layout of the test rig is shown in Fig. 4.
The test process started with adjusting the primary nozzle pressure to 1.2 MPa and adjusting
the downstream flow restrictor such that ejector pump output pressure of 0.05 MPa is achieved.
This point is referred to as the Experimental Set Point in subsequent text and figures. After this

initial adjustment, the downstream flow restrictor was fixed for rest of the test, which now
included varying the primary nozzle pressure from 0.6 to 2.0 MPa in steps while measuring the
different pressures and flow rates (as shown in Fig. 4) for each step. The experimental data is
shown in subsequent figures of this paper, e.g. Figs. 5 and 6. The error associated with the
experimental data is less than 1% for the measured pressures and flow rates.

Figure 4. Layout of the fuel ejector pump test rig.

IV.

Results and Discussion

In the present analysis, computations were carried out at different pump nozzle inlet pressures
(Pin) in order to adequately cover the test (experimental) data range. All computations for a
particular turbulence model started by first duplicating the Experimental Set Point (i.e. P in=1.2
MPa and Pout=0.05 MPa) by adjusting the value of the modeled downstream flow restrictor. This
is referred to as the Computational Set Point in subsequent text and figures. Multiple cases
corresponding to different turbulence models, values of the constants in the standard K-
turbulence model, and for the six values of primary nozzle pressure (Pin) were computed.
A. Initial Analysis and Comparison with Experimental Data
The turbulent flow inside the ejector pump has a jet issuing from the nozzle, boundary layers
near the pump walls and strong pressure changes within the secondary nozzle and the mixing-

10

chamber of the pump (discussed in detail later in this paper). These features of the turbulent flow
field within the ejector pump suggests that a turbulence model applicability study is needed. For
this purpose the two-equation standard K-, the Realizable K-, renormalized group K- and the
standard K- were initially used to compute the flow-field inside the fuel ejector pump. The
computation process with all the turbulence models started with adjusting the value of the flow
restrictor to get the desired the Computational Set Point, as discussed earlier. The results in
terms of variation of pump outlet pressure with the primary nozzle inlet pressure is shown in Fig.
5. It is evident that all the tested turbulence models give results that are consistent with
experimental data. Thus the pressure input / output behavior of the ejector pump is easily
predicted by the adopted turbulence models.

Figure 5. Effect of turbulence model on pump outlet pressure.

The pump outlet flow rate, corresponding to these same computed points, is plotted in Fig. 6.
It is seen from Fig. 6 that once the data of Fig. 5 is plotted in terms of corresponding flow rates,
all the tested turbulence models over predict the pump outlet flow rate. If the pump outlet flow
rate is adjusted by to match the experimental data by adjusting the downstream restrictor value,
then the computed pump outlet pressures would disagree with the experimental data. Thus, all
the tested turbulence models show certain variation from the corresponding experimental data.

11

The results shown in Figs 5 and 6 also show that the standard K- turbulence model is the closest
to the corresponding experimental data in terms of the pump flow rates. Therefore the standard
K- model is selected for further study.

Figure 6. Effect of turbulence model on pump flow rate (Data points same as Fig. 5).

Before going into the details of turbulence model adjustemnt, the basic flow features of the
ejector pump, for different turbulence models, is discussed. These features are discussed at the
Computational Set point i.e. at Pin=1.2 MPa, and Pout=0.05 MPa, unless otherwise stated. The
basic flow field inside the ejector pump comprises high-speed flow from the primary nozzle
which entrains nearby fuel from the fuel tank via the secondary nozzle. This fuel then enters the
mixing chamber where momentum exchange occurs resulting in increase in static pressure.
Further increase in static pressure is accomplished in the diffuser section of the pump after which
the pressurized fuel leaves the pump (Fig. 1b). This flow field is shown as the stream function
plot in Fig. 7. From the streamlines (stream function) plot, small region of flow recirculation are
evident near the secondary nozzle. This region is smallest for SKE and largest for RKE
turbulence model, which is consistent with the flow rate behaviour shown in Fig. 5. The basic
flow field has similar pattern (qualitative) for all the computed cases i.e. for six nozzle inlet

12

pressures. Lack of strong recirculation / seperation regions indicates a good basic design of the
ejector pump.

Figure 7. Stream function (streamlines) plot for different turbulence models.

The absolute velocity (Vx2+Vy2) distribution plot inside the pump is shown in Fig. 8. From
this figure it can be seen that the maximum fluid velocity occurs at the ejector pump nozzle with
large radial and axial gradients in the vicinity of the nozzle. In the mixing chamber (Fig. 1 and 2)
of the pump, region of relatively uniform velocity is evident. This suggests good mixing of the
primary and secondary stream fluids as required for a good design. The three turbulence models
show barely perceptible variation in absolute velocity amongst each other.

Figure 8. Absolute velocity variation in the pump for different turbulence models.

The pressure distribution inside the ejector pump (not shown here) shows maximum pressure
change occuring in the primary nozzle, compared to which, the pressure change in the rest of the

13

pump cannot be resolved due small value (~0.05 MPa compared to 1.2 MPa). Additionally, as
discussed earlier, the pressure behaviour of the ejector pump is consistent with experimental data
as shown in Fig. 5.

Figure 9. Effective viscosity variation in the pump for different turbulence models.

Levels of effective viscosity can be quite high in turbulent flows. This phenomenon is evident
from Fig. 9 where the effective viscosity distribution inside the ejector pump is shown for
different turbulence models. About 3~4 orders-of-magnitude variation in viscosity is evident for
the figure (Fig. 9). In all cases the shear layer between the primary nozzle flow and the
secondary nozzle flow is the starting point of heavy turbulence generation and thus large
effective viscosity as well. In the mixing chamber (Figs. 1 and 2) significant axial (flow
direction) pressure and velocity variations (Fig. 8) again generate large levels of effective
viscosity. The variation in effective viscosity inside the pump amongst different turbulence
models is apparent from Fig. 9 and is manifested in computed variation in flow rate results (Fig.
6). Since the standard K- turbulence model is closest to the experiment flow rates (Fig. 6), the
systematic adjustment of its parameters inorder to computationally reproduce the experimental
data is discussed next.
B.

Adjustment of Standard K- Turbulence Model Parameters

14

The two-equation Standard K- turbulence model used in the present analysis is shown below
as equations 1 to 3. This model involves five constant parameters 4,5,6 i.e. C=0.09, C1=1.44,
C2=1.92, k=1.0 and =1.3. These values are best suited for attached boundary layer type of
turbulent flows5,6. The parameter C influences the overall turbulent / eddy viscosity, while the
parameters C1 and C2 appear in the dissipation rate equation (equation 2) in the production of
and dissipation of terms, respectively. The term with C1 is also influenced by the gradient
of mean velocity. The parameters k and are the turbulent Prandtl numbers for K and ,
respectively and appear in the respective diffusion terms of equations 1 and 2. The values these
five constants are developed by analytical / experimental data correlation 5,6 and are not universal
and may need to be modified for turbulent flows that involve strong jets, strong pressure
gradients, recirculation or wakes etc. Some of these features are present in the fuel ejector pump,
as discussed earlier.
Standard K- turbulence model:
Transport equation for production of turbulent energy (k):

u j
u i u i
k

x
x j x j
x i
j

D k

Dt
x j

(1)

Transport equation for dissipation of turbulent energy ():

D
t

Dt
x j

u j
2
u i u i

C C
t

1
2

x
x j x j
xi
k
k
j

(2)

The turbulent viscosity for this model is given as:


t C

k2

(3)

Where C=0.09, C1=1.44, C2=1.92, k=1.0 and =1.3 are model constants

15

A systematic computational study was done where the values of these constants were adjusted
and the corresponding result analyzed. The effect of turbulent Prandtl numbers k and is not
expected to be important for the present study since the velocities in the ejector pump are high
and thus the convective terms of K and (LHS total derivative term in equations 1 and 2) are
expected to be much larger than the diffusion terms, in which k and appear (equations 1 and
2). This was confirmed by varying the default values of k and by 20% and running
computational set point cases. The results of this initial study indicates that there is no noticeable
change in the pump computed pressure or flow rate behavior due to variation in k and/or .
Therefore, focus was then kept on varying and optimizing the values of C, C1 and C2. The
parameter adjustment strategy was such that first C 1 and C2 were varied and later C was also
adjusted. The results of the turbulence modeling study where C 1 and C2 were adjusted are
partially shown in Fig. 10. As seen earlier also, the pressure characteristics of the ejector pump is
not affected by the change in turbulence parameters (Modified K- or MKE) due to the
adjustment of the downstream restrictor valve for each combination of C 1 and C2 to correspond
to the Experimental and Computational Set Point discussed earlier. The values of C 1 and C2
were adjusted in a particular ratio, C1/C2=0.75 being the ratio at default values.

16

Figure 10.Effect of C1 and C2 constants ejector pump pressure behavior.

The pump flow rate characteristics, corresponding to the data of Fig. 10, are shown in Fig. 11.
As can be seen from this figure, the ratio of C1/C2=0.72 agrees reasonably well with experimental
data. The default ratio C1/C2= 0.75 over predicts the ejector pump computed flow rate while the
ratio C1/C2=0.69 under predicts the flow rate. This is related to the change in computed effective
viscosity in relation to the change in C1/C2 ratio. This effect is shown in Fig. 12, where it can be
seen that for C1/C2=0.75 the effective viscosity levels inside the ejector pump are lower
compared to C1/C2=0.72 and thus results in a bit higher flow rate for the same over all pressure
change (Figs 10 and 11). Thus changing C1 and C2 in a particular manner (ratio C1/C2) adjusts the
turbulence energy production and dissipation in a manner that effects the resultant effective
viscosity.

17

Figure 11.Effect of C1 and C2 constants on ejector pump flow rate behavior.

Figure 12.Variation of computed effective viscosity for different C1/C2.

The effective viscosity levels for the standard K- turbulence model can also be independently
increased or decreased by adjusting the value of C parameter, as discussed earlier. Therefore, a
number of computations were done where the value of C was adjusted while C1 and C2 were left
at their default values. The computed pressure characteristics of ejector pump is consistent with
experimental data, similar to all earlier cases (Figs 5 and 10) and is not shown again. The pump
flow rate for adjusted value of C is shown in Fig. 13, earlier results for C 1/C2=0.72 is also
superimposed for reference. As can be seen from Fig. 13, independently adjusting C can also
adjust the pump flow rate for same pressure characteristics by adjust the levels of computed

18

effective viscosity. This effect can be seen from Fig. 14 where the effective viscosity variation
inside the pump is plotted for C=0.12 (default value 0.09), earlier cases for C 1/C2=0.72 are also
repeated for comparison. It is evident from this figure (Fig. 14) that independently adjusting
C=0.12 or C1/C2=0.72 results in similar effective viscosity variation inside the pump thus
resulting in similar flow rates that are consistent with experimental data (Fig.13). This
modification in turbulence model constants is the most optimal for the present ejector pump
analysis and shows that for the complex turbulent flow inside the pump, the effective viscosity
needs to be adjusted by either by controlling the production and dissipation of by the ratio
C1/C2 or directly by C. This is an overall adjustment to the standard K- model to match
experimental results for the understudy ejector pump. More detailed 3D LES analysis is planned
for future to get greater insight into the nature of turbulent flow inside the subject ejector pump.

Figure 13.Effect of C on ejector pump flow rate behavior.

19

Figure 14.Effect of C=0.12 and C1/C2=0.72 on computed effective viscosity.

V.

Conclusion

In this paper the values of constants of standard K- turbulence model have been
systematically adjusted to computationally reproduce experimental data for a fuel ejector pump.
For this case, it is found the computed effective viscosity level inside the ejector pump needs to
be adjusted. This can be achieved by either adjusting the ratio of C1/C2=0.72 or adjusting
C=0.12 from their respective default values.
References
1

Gullia, A., Laskaridis, P. and Ramsden, K.W. Ejector Pump Theory Applied to Gas Turbine
Engine Performance inside Indoor Sea-Level Test Cell-Analytical and CFD Study AIAA
2006-3152, 25th AIAA Aerodynamic Measurement Technology and Ground Testing
Conference, 5 - 8 June 2006, San Francisco, California, USA.

Mary, J. L. ExperimentaI Investigation of an Ejector-Powered Free-jet Facility AlAA-923569, 28th Joint Propulsion Conference cosponsored by the AIAA, ASME, SAE, and ASEE,
July 6-8, 1992, Nashville, Tennessee, USA.

20

Vermeulen, P. J., Ramesh., V., Meng, G. C., Miller, D. N. and Domel, N. Air Ejector Pumping
Enhancement Through Pulsing Primary Flow AIAA 2004-2621, 2nd AIAA Flow Control
Conference 28 June - 1 July 2004, Portland, Oregon, USA.

Lauder, B. E., and Spalding, D. B., Lectures in Mathematical Models of Turbulence, Academic
Press, London, England, 1972

White, F. M., Viscous Fluid Flow, 2nd ed., McGraw Hill, New York, 1991, Chap. 6.

Chung, T. M, Computational Fluid Dynamics, Cambridge University Press, Cambridge, UK,


2002, Chap. 21.

ANSYS FLUENT, Computational Fluid Dynamics Software Package User Guide, Ver. 12.0.

GAMBIT, Geometry and Mesh Generation Software Package, Ver. 2.2.30.

Jennings, G JF-17 Production Commences Janes Defense Weekly, Jan 2008.

10

ESDU 85032 Ejectors and Jet Pumps Design and Performance for Incompressible Liquid
Flow

12

Bi, R., Ma, L., Cheng, H., Tan, X., Zheng, S. Multi-scale simulation of turbulent mixing in
ejector (2009) Huagong Xuebao/CIESC Journal, 61 (1), pp. 157-162.

13

Lin, K.-L., Bi, R.-S., Tan, X.-S., Zheng, S.-Q. Using PLIF technology to study the turbulent
mixing characteristic in a liquid-liquid ejector (2009) Gao Xiao Hua Xue Gong Cheng Xue
Bao/Journal of Chemical Engineering of Chinese Universities, 23 (1), pp. 28-33.

14

Gao, H., Lin, W., Du, Z. Numerical flow and performance analysis of a water-jet axial flow
pump (2008) Ocean Engineering, 35 (16), pp. 1604-1614.

15

Park, B.H., Lim, J.H., Yoon, W. Fluid dynamics in starting and terminating transients of zerosecondary flow ejector (2008) International Journal of Heat and Fluid Flow, 29 (1), pp. 327339.

21

16

McBean, S.F., Birk, A.M. CFD (Realizable K-) and cold-flow testing for the design of air-air
ejectors with triangular tabbed driving nozzles (2007) Proceedings of the ASME Turbo Expo,
6 PART A, pp. 17-25.

17

Chen, Q., Birk, A.M. Experimental and CFD study of an exhaust ejector with round entraining
diffuser (2007) Proceedings of the ASME Turbo Expo, 6 PART A, pp. 27-35.

18

Wang, J., Zhang, H.-F. Simulation of the optimum length of mixing room in liquid-liquid gas
ejector (2007) Petrochemical Equipment, 36 (3), pp. 4-7.

19

Lee, J.-H., Kumar, V.R.S., Kim, H.D., Bo-Gyu, C., Kim, K.H., Setoguchi, T. Studies on flow
characteristics of variable ejector systems (2006) Collection of Technical Papers AIAA/ASME/SAE/ASEE 42nd Joint Propulsion Conference, 7, pp. 5459-5465.

20

Timouchev, S., Karpyshev, A., Aksenov, A., Bogdanov, V. Computational study of ejectorpumps (2005) Proceedings of the American Society of Mechanical Engineers Fluids
Engineering Division Summer Conference, 1 PART B, art. no. FEDSM2005-77355, pp. 14251430.

21

Balasubramanyam, M.S., Landrum, D.B., Chen, C.P., Lineberry, D. Numerical investigation


of cold flow non-axisymmetric ejectors (2005) 43rd AIAA Aerospace Sciences Meeting and
Exhibit - Meeting Papers, pp. 10613-10624.

22

Al-Ansary, H.A.M., Jeter, S.M. Numerical and experimental analysis of single-phase and twophase flow in ejectors (2004) HVAC and R Research, 10 (4), pp. 521-538.

23

Wilson, W.M., Imber, R.D. CFD analyses of compact, high aspect ratio ejectors (2003)
Proceedings of the ASME Fluids Engineering Division Summer Meeting, 1, pp. 7-13.

22

24

Chen, Q.-G., Xu, Z., Zhang, Y.-J. Application of two versions of a RNG based k- model to
numerical simulations of turbulent impinging jet flow (2003) Journal of Hydrodynamics, 15
(2), pp. 71-76.

25

Rebaine, A., Soulaimani, A. Numerical simulation of two-dimensional compressible turbulent


flows in ejectors (2001) Transactions of the Canadian Society for Mechanical Engineering, 25
(2), pp. 227-251.

26

Zhou, B., Fleck, B.A., Bouak, F., Gauthier, J.E.D. Comparison of swirling effects on ejector
performance using four turbulence models (2000) Canadian Aeronautics and Space Journal,
46 (4), pp. 178-182.

27

Djebedjian, B., Abdalla, S., Rayan, M.A. Parametric investigation of boost jet pump
performance (2000) American Society of Mechanical Engineers, Fluids Engineering Division
(Publication) FED, 251, pp. 541-548.

28

Yule, A.J., Damou, M., Kostopoulos, D. Modeling confined jet flow (1993) International
Journal of Heat and Fluid Flow, 14 (1), pp. 10-17.

29

Yule, A.J., Damou, M. Investigations of ducted jets (1991) Experimental Thermal and Fluid
Science, 4 (4), pp. 469-490.

30

Dejoode, A.D., Patankart, S.V. Prediction of Three-Dimensional Turbulent Mixing in an


Ejector (1978) AIAA Journal, 16 (2), pp. 145-150.

31

Masud, J., and Javed, A., Modification of Standard K- Turbulence Model for Analysis and
Design Refinement of Fuel Ejector Pump, Paper # AIAA-2007-530, 45th AIAA Aerospace
Sciences Meeting at Reno, Nevada, USA (2007)

23

32

T. L. Phillips, C. J. Roy Evaluations of Extrapolation-Based Discretization Error and


Uncertainity Estimators paper AIAA-2011-215, 49th AIAA Aerospace Sciences Meeting, 4-7
Jan 2011, Orlando FL.

33

P. J. Roache Prespective: A Method for Uniform Reporting of Grid Refinement Studies


Journal of Fluids Engineering, Vol. 116, pp. 405-113.

24

Das könnte Ihnen auch gefallen