Numerical Simulation of Inviscid Transient Flows

in Shock Tube and its Validations

Al-Falahi Amir, Yusoff M. Z, & Yusaf T
AbstractThe aim of this paper is to develop a new two
dimensional time accurate Euler solver for shock tube applications.
The solver was developed to study the performance of a newly built
short-duration hypersonic test facility at Universiti Tenaga Nasional
UNITEN in Malaysia. The facility has been designed, built, and
commissioned for different values of diaphragm pressure ratios in
order to get wide range of Mach number. The developed solver uses
second order accurate cell-vertex finite volume spatial discretization
and forth order accurate Runge-Kutta temporal integration and it is
designed to simulate the flow process for similar driver/driven gases
(e.g. air-air as working fluids). The solver is validated against
analytical solution and experimental measurements in the high speed
flow test facility. Further investigations were made on the flow
process inside the shock tube by using the solver. The shock wave
motion, reflection and interaction were investigated and their
influence on the performance of the shock tube was determined. The
results provide very good estimates for both shock speed and shock
pressure obtained after diaphragm rupture. Also detailed information
on the gasdynamic processes over the full length of the facility is
available. The agreements obtained have been reasonable.
Keywordsshock tunnel, shock tube, shock wave, CFD.
I. INTRODUCTION
HIS paper describes the procedure for performance
evaluation of a short-duration hypersonic test facility that
build at the College of Engineering, Universiti Tenaga
Nasional UNITEN in Malaysia. The facility is designed so
that it can be used as a shock tube, free piston compressor,
shock tunnel and gun tunnel. The facility has been designed,
constructed and commisioned for a wide range of diaphragm
pressure ratios and different driver/driven working gases to
get Mach number up to 6. The facility will allow various
researches to be done in the field of high speed supersonic
and hypersonic flows. The important application would be in
power plants where the working fluid is always in a very high
speed with high gas effects.
To verify and supplement some of the theoretical results, a
hypersonic test facility of a somewhat unconventional design
has been built. The bulk of the experimental investigations
undertaken to date have dealt with pressure studies using high
precision pressure transducers and an in house made fast
response thermocouple were used to predict the pressure
history and subsequently the shock wave strength P
2
/P
1
and
the surface temperature change profile during the facility
operation. Using two pressure transducers the shock wave
speed is measured experimentally for different driver/driven
gas combinations.
Al-Falahi Amir is with University Tenaga Nasional UNITEN, College of
Engineering, Jalan Kajang- Puchong, 43009 Kajang, Selangor Darul Ehsan,
Malaysia. e-mail: [email protected]
A two-dimensional time-accurate time-marching Navier-
Stokes solver for shock wave applications is described. It uses
the second-order accurate cell-vertex finite-volume spatial
discretization and fourth order accurate Runge-Kutta temporal
integration. Three simulations of particular conditions for a
short duration high speed flow test facility are then presented
and compared with experimental measurements. The
simulations provide very good estimates for both the shock
speed and shock strength obtained after diaphragm rapture
and also provides detailed information on the gas dynamic
processes over the full length of the facility. This detailed
information may be used to identify some of the causes for
observed variations in pressure and temperature. The
agreements obtained have been reasonable.
Construction of this facility is fundamentally important for
the development of advanced instrumentation (in this case,
fiber optic pressure sensors and fast response thermocouples),
and heat transfer/fluid mechanics studies that are relevant to
turbine investigations. The wind tunnel provides a convenient
and low cost experimental facility that can produce the flow
conditions (matched Mach and Reynolds numbers,
temperature ratios etc) necessary for experimental simulation
of turbulent flows. The advantage of developing
instrumentation and investigating relevant flow fields in the
wind tunnel environment is derived from the fact that the flow
duration is very short (less than 1 second). This is sufficient
time to establish the required flow fields and obtain the
required measurements, but the energy requirements
associated with operating the facility are relatively low.
Hence, the facility is a very cost effective way to
experimentally investigate critical heat and fluid flow
processes associated with turbine power plants.
II. PHYSICAL DESCRIPTION OF THE FACILTY
The detail components of the facility are described briefly and
shown in Figure 1. This project is a collaboration project
between the University of Southern Queensland (USQ) in
Australia and Universiti Tenaga Nasional (UNITEN) in
Malaysia to design and construct a 10 m short duration
hypersonic test facility. The facility consists of the following
significant items:
1. Driver section: a high-pressure section (driver), which will
contain the high pressure driver gas.
2. Discharge valve: to discharge the driver section after each
run.
3. Pressure gauge: to read the pressure inside the driver
section, this section is also provided with a static pressure
transducer to record the exact value of the driver pressure P
4
at which the diaphragm ruptures.
T
International Journal of Engineering and Natural Sciences 4:1 2010
52
Fig. 1 schematic diagram of UNITENs test facility
4. Vacuum pump: when the driver gas is not air (eg. Helium
or Hydrogen) then the driver section should be evacuated and
filled with the required driver gas.
5. The primary diaphragm: this is a thin aluminum membrane
to isolate the low-pressure test gas from the high-pressure
driver gas until the compression process is initiated.
6. Piston compression section: A piston is placed in the barrel
(driven tube) adjacent to the primary diaphragm so that when
the diaphragm ruptures, the piston is propelled through the
driven tube, compressing the gas ahead of it. This piston used
with gun tunnel tests only.
7. Discharge valve: to discharge the driven section after each
run.
8. Vacuum gauge: to set the pressure inside the barrel section
to low values (vacuum values) less than atmospheric value.
9. Barrel section: a shock tube section (smooth bore), to be
filled with the required test gas (air, nitrogen or carbon
dioxide).
10. Barrel extension: the last half meter of the barrel on which
the pressure transducers and thermocouples are to be attached
(see details A).
11. The secondary diaphragm: it is a light plastic diaphragm
to separate the low pressure test gas inside the barrel from the
test section and dump tank which are initially at a vacuum
prior to the run.
12. Test section: this section will expands the high
temperature test gas through a nozzle to the correct high
enthalpy conditions needed to simulate hypersonic flow. A
range of Mach numbers is available by changing the diameter
of the throat insert.
13. Vacuum vessel (dump tank): to be evacuated to about 0.1
mm Hg pressure before running. Prior to a run, the barrel, test
section, and dump tank are to be evacuated to a low-pressure
value.
It is intended that this CFD solver be a useful tool in the
design process of the test facility. The simulation will
complement the experimental data. That is any experimental
data which could be achieved from the facility need to be
verified using a numerical tool.
III. MATHEMATICAL MODEL
In this Section, the fluid flow governing equations are
presented in concert with the numerical scheme used to
compute the compressible flow within the shock tube.
A. Euler Equation
The two-dimensional continuity, x- and y-momentum and
energy equations describing the turbulent flow of a
compressible fluid expressed in strong conservation form in
the x-, y-Cartesian co-ordinate system may be written as
J
y
G
x
F
t
w
= + +
c
c
c
c
c
c
(1)
where w represents the conserved variables and F and G
are the overall fluxes in x-, y-directions respectively.
w can be expressed as:
|
|
|
|
|
.
|

\
|
=
0
e
v
u
w
p
p
p
p
(2)
F can be expressed as:
,
0
2
|
|
|
|
|
.
|

\
|
+
=
uh
uv
P u
u
F
p
p
p
p
(3)
and G can be expressed as:
,
0
2
|
|
|
|
|
.
|

\
|
+
=
vh
P v
uv
v
G
p
p
p
p
(4)
e
o
and h
o
are the specific total internal energy and specific
total enthalpy of the fluid respectively. In the usual notations:
2
2
V
e e
o
+ =
(5)
2
2
0
V
h h + = (6)
2 2 2
+ = v u V (7)
The body force J is zero vector since effect of gravity is
negligible for high speed flow.
|
|
|
|
|
.
|

\
|
=
0
0
0
0
J
IV. NUMERICAL SCHEMES
The numerical scheme was based on an earlier work by
Zamri [21], [22]. The earlier solver was developed for two-
dimensional transient flow of two-phase condensing steam in
low pressure turbine. Viscous effect was not taken into
account in the earlier program. In the current work
modifications were made so that the program can be applied
International Journal of Engineering and Natural Sciences 4:1 2010
53
for high speed flow in shock tube. The following sections will
discuss the numerical formulation of the modified flow
solver.
A. Cell-Vertex Finite-Volume Spatial Discretization
The flow domain is replaced by a finite number of control
volumes, which are generated algebraically by the current
solver. The mesh system is commonly known as H-mesh and
divides the physical domain into a set of discrete rectangular
control volumes. An example of H-mesh is shown in Figure 2.
A cell-vertex formulation is used in which the flow
variables are stored at cell vertices A, B, C and D as has been
shown in Figure 2. Cell-vertex formulation offers some
advantages over the cell-centered one in which cell-vertex
method offers higher accuracy on irregular grid.
(i+1, j)
(i+1, j+1)
(i, j+1) (i, j)
Cell (i, j)
Boundary S
Fixed Area
D C
B
A
Fig. 2 An illustration of a typical 2D H-mesh.
For a uniform mesh, there would be no difference between the
cell-centered and cell-vertex schemes; however, cell-vertex
scheme does not require extrapolation to the solid boundary to
obtain the wall static pressure, which is necessary in solving
the momentum equations for cells adjacent to the solid
boundary.
Starting from known values of primitive variables from the
previous time-step, the values of F and G are determined at
each node. Then the line integration is performed for each
control volume in turn for the four conserved variables.
Integrating Equation (1) with respect to volume
l l l
= + + 0 ) ( ) ( dV G dV F dV
t
w
C C
c
c
(8)
Using Gauss divergence theorem the whole equation can be
reduced to
( )
l

O
=
c
c
dx G dy F
t
w
C C
1
(9)
Applying the above equation to a typical control volume and
defining R
i, j
as the residual for the cell, the same discritized
equation for the cell is:
|
|
.
|

\
|
A A
O
=
c
c

x G y F
t
w
j i C j i C
j i
, , , ,
,
1
(10)

O
= |
.
|

\
|
c
c
=
j i C
j i j i
j i
R
t
w
w R
, ,
, ,
,
1
) (
(11)
where ( ) w R
ij
represents the residual for each cell and
) (
, , , , , ,

A A =

x G y F R
j i C j i C j i C
.
The calculated residuals apply to the values of properties
within the cell, whereas, the variables are actually stored at
the nodes. Consequently, they have to be redistributed to the
four surrounding nodes. This is done by sharing the changes
equally between the four nodes as shown in Figure 3, as
suggested in the second-order central differencing scheme.
Thus:
( ) ) ( 25 . 0
, 1 1 , 1 , 1 , j i j i j i j i A
R R R R w R

+ + + = (12)
(i, j)
A
- - - - - effective control volume for node A
0.25R(i-1, j)
0.25R(i-1, j-1)
0.25R(i, j-1)
0.25R(i, j)
Fig. 3 Distribution of cell residual to nodes.
For nodes at the wall, since two cells share a single node, the
residual obtained from Equation (12) is doubled. For the
nodes at the corners, since only one cell share the node, the
residual obtained is quadrupled. Thus, the equivalent
discretized equation for node (i, j) will be:
( ) w R
t
w
A
A
= |
.
|

\
|
c
c
(13)
B. Treatment of the Inviscid Fluxes
All the inviscid fluxes are calculated by using central
differencing scheme, therefore the differencing scheme is
second order accurate, from equation (11);
) (
, , , , , ,

A A =

x G y F R
j i C j i C j i C
Applying to control volume shown in Figure 1
) (
) (
, , , ,
, , , , , ,


A + A + A + A
A + A + A + A =
x G x G x G x G
y F y F y F y F R
AD C BA C CB C DC C
AD C BA C CB C DC DC C j i C
(14)
where
) (
2
1
, , , C C D C DC C
F F F + = ,
) (
2
1
, , , B C C C CB C
F F F + =
,
International Journal of Engineering and Natural Sciences 4:1 2010
54
) (
2
1
, , , A C B C BA C
F F F + =
,
) (
2
1
, , , D C A C AD C
F F F + =
) (
2
1
, , , C C D C DC C
G G G + = ,
) (
2
1
, , , B C C C CB C
G G G + =
) (
2
1
, , , A C B C BA C
G G G + = ,
) (
2
1
, , , D C A C AD C
G G G + =
C. Artificial Dissipations
All second-order central-differencing schemes, even with a
stable time-step, suffer from certain tendencies to instability
due to the odd-even decoupling near a discontinuity. The
scheme can be stabilized by introducing a small amount of
artificial viscosity is suggested by Jameson et al. [11].
This artificial viscosity formulation is a blend of second and
fourth-order terms with a pressure switch to detect changes in
pressure gradient. The fourth-order terms, for any conserved
variables, w is:
) ( ) ( ) (
4 4
4
w D w D w D
yij xij ij
+ =
(15)
where:
| |
2 1 1 2
4 4
4 6 4 ) (
+ +
+ + =
ij ij ij ij ij xij xij
w w w w w w D
| |
j i j i ij j i j i yij yij
w w w w w w D
2 1 1 2
4 4
4 6 4 ) (
+ +
+ + =
The second-order dissipation terms are defined as:
) ( ) ( ) (
2 2 2
w D w D w D
yij xij ij
+ = (16)
where:
| |
1 1
2 2
2 ) (
+
+ =
ij ij ij xij xij
w w w w D
| |
j i ij j i yij yij
w w w w D
1 1
2 2
2 ) (
+
+ =

xij yij xij yij
2 2 4 4
, , and in the above equations are
coefficients, which are functions of the local pressure gradient
to be defined later.
The fourth-order terms will not affect the global accuracy
of the second-order scheme, but they eliminate the
background oscillations caused by the central-differencing
scheme, which is second-order accurate. However, near a
discontinuity, e.g. around a shockwave, they lead to
occurrence of oscillations and appearance of overshoots. On
the other hand, second-order terms are very dissipative and
can eliminate high oscillations near a discontinuity but
produce too much background oscillations. Following
Jameson et al. [11], the fourth-order terms are turned off near
discontinuities, by means of a pressure sensor, and only
second-order terms will be in operation in these regions. A
pressure sensor introduced to detect the steepness of the
pressure gradient and has the form:
1 1
1 1
2
2
2
+
+
+ +
+
=
ij ij ij
ij ij ij
xij
P P P
P P P
X SF q (17)
j i ij j i
j i ij j i
yij
P P P
P P P
Y SF q
1 1
1 1
2
2
2
+
+
+ +
+
= (18)
SF2X and SF2Y are second-order dissipation coefficients with
typical values of 0.10-0.50.
The pressure switch defined to turn off the fourth-order
and turn on the second-order terms near a discontinuity takes
the forms:

xij
4
= max (0, SF4X-
xij
2
) (19)

yij
4
= max (0, SF4Y-
yij
2
) (20)
where:

xij
2
= max (q
xij
, q
xij+1
)

yij
2
= max (q
yij
, q
yi+1j
)
SF4X and SF4Y are fourth-order dissipation coefficients with
typical values of 0.005-0.01.
It can be seen that near a discontinuity, q and
2
are of
order 1 and the second-order becomes the dominant
dissipative term. In the remainder of the domain the
background dissipation is provided by the fourth-order terms.
SF4X and SF4Y must not be too large since they will create
too much numerical viscosity in the flow domain and thereby
mask the physics of the flow.
The total dissipation will be:
( ) ( ) ( ) w D w D w D
yij xij ij
+ = (21)
where:
( ) ( ) ( ) w D w D w D
xij xij xij
4 2
=
( ) ( ) ( ) w D w D w D
yij yij yij
4 2
=
After the addition of the dissipation terms the descritized
equation for node A becomes:
( ) ( ) w D w R
t
w
A A
A
+ =
|
.
|

\
|
c
c
(22)
The dissipation term ( ) D w
A
, is only evaluated in the first
stage and frozen for the next three stages of the Runge-Kutta
time stepping scheme, as follows.
D. The Multi-stage Runge-Kutta Time Stepping Scheme
Equation (22) is integrated with respect to time by means
of a four-stage Runge-Kutta time stepping scheme, as
proposed by Jameson et al. [11]:
,
0 n
w w =
( ),
0 0
1
0 1
D R t w w + A + = o
( ),
0 1
2
0 2
D R t w w + A + = o
International Journal of Engineering and Natural Sciences 4:1 2010
55
( ),
0 2
3
0 3
D R t w w + A + = o
( ),
0 3
4
0 4
D R t w w + A + = o
4 1
w w
n
=
+
(23)
where the superscripts, n and n+1 refer to the time intervals in
the main integration sequence, 1, 2, 3, 4 refer to the
intermediate time-steps in the Runge-Kutta scheme. The
coefficients o o o o
1 2 3 4
, , , are 0.250, 0.333, 0.500 and
1.000 respectively.
V. BOUNDARY CONDITIONS
A. Solid Boundary
Only the solid boundary condition is considered in the
current work since the flow is confined within the tube. At the
wall, no-slip boundary condition is imposed for the
momentum equations to enforce no mass fluxes can penetrate
through the solid boundary. For the energy equation, adiabatic
condition is assumed. Solid boundary nodes only have
contributions from two cells. However, the control volumes
associated with the solid boundary nodes are only half of that
for the internal nodes. So, prior to the temporal integration the
residuals are doubled.
At the solid boundary, it has been shown by Hirsch [23],
that for Euler equations only one characteristic enters the flow
domain and only a single physical boundary condition is to be
imposed. This condition is expressed by the vanishing normal
velocity:-
V
normal
= 0
As a consequence, all convective flux components through
the solid boundary will vanish. This means that in the
expression for the flux through a cell face on a solid
boundary, only the pressure remains, i.e.:-
F dy G dx P
dy
dx
C C
=

0
0
(24)
The pressure on the solid boundary face can be calculated
directly by taking the average values of the nodes at both ends
of the cell face. When the flow is assumed to be inviscid, the
velocity at the solid boundary is non-zero.
B. Treatment of Dissipation Terms near the Boundaries
To evaluate the dissipation terms, variables at two
neighbouring nodes on either side of the calculating points are
required. At the boundary, only variables on one side of the
node are known. The other two variables need to be
determined by extrapolation. Consistent with Gustafsson and
Sundstroms [24] recommendations, at the inlet and exit
boundaries, the extra variables are calculated by first-order
extrapolation. Only one extra variable is required as the
variables at the inlet and exit boundaries are fixed by the inlet
and exit boundary conditions respectively. For example, the
extra variable required to calculate the dissipative term at j =
2 is calculated by:-
w w w
i i i 0 1 2
2 = (25)
It has been observed by Pulliam [25], Swanson and Turkel
[26] and Caughey and Turkel [27] that the treatment of
dissipation terms, especially at the solid boundaries, can have
a strong effect on the accuracy and convergence rate of a
viscous or even inviscid flow computation. An important
conclusion that can be drawn from these studies is the
necessity to reduce the second-order dissipation terms at the
solid boundaries. The treatment used is that recommended by
Pulliam [25] and used by Bamkole [28] and Zamri [21,22].
Modifications are only needed for the pitchwise
components, ( ) D w
yij
. At i = 1, the fourth-order components
are replaced by the second-order one by using one-sided
difference, while the second-order terms are set to zero. So,

| |
D w w w w
y j y j j j j 1
4
1
4
1 2 3
2 ( ) = + (26)
At i = 2, the extra variable needed for the fourth-order
terms is calculated by means of a first-order extrapolation.
With reference to Figure (2), the resultant term will become: -

| |
D w w w w w
y j y j j j j j 2
4
2
4
1 2 3 4
2 5 4 ( ) = + + (27)
VI. INITIAL CONDITIONS
To start the iterations, initial flow field variables must be
specified at all calculating points. In the current work, the
pressure values are specified at both the driver and driven
sections, in accordance with the desired pressure ratio. The
flow is assumed to be in stagnant condition initially.
VII. STABILITY CRITERIA
Generally, explicit time-marching schemes suffer from
instability problem, particularly when the time step is larger
than that from the Courant Friedrichs Lewy (CFL) criterion.
In order to ensure numerical stability, the maximum allowable
time-step which can be used in the calculation is limited by:
( ) a V
x
CFL t
+
A
s A .
(28)
where
CFL = CFL number,
Ax = incremental distance,
V = speed,
a = speed of sound =(RT)
0.5
.
For the four-stage Runge-Kutta scheme applied in a 1D
problem to be specific, CFL =2 2 . However, this number
may be lower due to the non-linearity and multi-
dimensionality of the flow problem considered in the current
work. In most cases, strict convergence cannot be obtained
unless the time step is sufficiently small. Here, a time step
size ranging from 1.0 to 5.0 s is used, without any stability
problem.
International Journal of Engineering and Natural Sciences 4:1 2010
56
VIII SOLUTION PROCEDURE
The solution procedure of the solver can be summarized as
follows:-
Step 1: Generate the structured H-mesh. The details of the
mesh system will be explained in the following
sections.
Step 2: Initialise the flow variables at time = 0s.
Step 3: Initiate the 4-stage Runge-Kutta (RK) time integration
scheme. Here, the spatial integration of the governing
equations to determine the residual in Equation 11 is
calculated for the first RK stage. The cell residuals are
then redistributed back to the neighbouring vertices
using Equation 12. The solution vector is then time
marched (see Equation 23) using the residuals for
each vertex and the corresponding stage coefficient
().The flow variables are then updated accordingly.
Step 4: Step 3 is repeated until the maximum Runge-Kutta
stage (in this case is 4) is reached.
Step 5: Save the pressure value at the first station (refer to
Figure 4). Update the time level (t
n+1
=t
n
+t). Go to
Step 3 until the desired time level is reached.
Fig. 4 Location of station 1 and 2
IX. VALIDATION OF THE CFD CODE
To ensure the validity of the CFD code, in terms of the
ability to capture shocks and contact discontinuity and to
produce the correct pressure, density and speed profiles, the
code has been validated against an exact solution for Inviscid
Flow in Shock Tube. The situation considered is basically that
in a shock tube with a pressure difference applied across the
diaphragm. The initial conditions are shown in Figure 5.
100 kPa 1000 kPa
0.5 m 0.5 m
Solution domain
Diaphragm
x
Fig. 5 Solution domain
The direction x is chosen in the direction shown in Figure
6 because the air flow that develops after the diaphragm
ruptures will be from the high pressure section towards the
low pressure section, i.e. x is chosen to be in the direction of
the induced flow.
The Sod problem [29] is an essentially one-dimensional
flow discontinuity problem which provides a good test of a
compressible code's ability to capture shocks and contact
discontinuities with a small number of zones and to produce
the correct density profile in a rarefaction.
The problem spatial domain is 0 x 1. The initial
solution of the problem consists of two uniform states, termed
as left and right states, separated by a discontinuity at the
origin, x
o
= 0.5. The fluid is initially at rest on either side of
the interface, and the density and pressure jumps are chosen
so that all three types of flow discontinuity (shock, contact,
and rarefaction) develop. To the ``left'' and ``right'' of the
interface we have,

L
= 1
R
= 0.125
P
L
= 1 P
R
= 0.1
u
L
= 0 u
R
= 0
The ratio of specific heats is chosen to be 1.4 on both sides
of the interface.
A uniform grid spacing with the number N = 356 is used.
The boundary conditions of the problem are held fixed as a
short time span of the unsteady flow is considered. The wave
pattern of this problem consists of a rightward moving shock
wave, a leftward moving rarefaction wave and a contact
discontinuity separating the shock and rarefaction waves and
moving right rightward. Figure 6 shows comparisons between
the present results for the pressure, density and Mach number
at a time t = 0.2 ms and the exact solutions. It can be observed
that the present work is capable of capturing the different
types of discontinuities quite accurately.
Distance (m)
P
r
e
s
s
u
r
e
(
P
a
)
0 0.25 0.5 0.75 1
1.0E+05
2.0E+05
3.0E+05
4.0E+05
5.0E+05
6.0E+05
7.0E+05
8.0E+05
9.0E+05
1.0E+06
Exact
CFD
P4= 1000 kPa
P1= 100 kPa
N=356
Distance (m)
D
e
n
s
i
t
y
(
k
g
/
m
3
)
0 0.25 0.5 0.75 1
1
2
3
4
5
6
7
8
9
10
11
12
Exact
CFD
P4 = 1000 kPa
P1 = 100 kPa
N = 356
Shock wave
Contac surface
Distance (m)
M
a
c
h
0 0.25 0.5 0.75 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Exact
CFD
P4= 1000 Pa
P1= 100 kPa
N=356
Fig. 6 Results for the Sods shock tube problem at t = 0.2
ms
International Journal of Engineering and Natural Sciences 4:1 2010
57
The mesh size plays important role in determining the
accuracy of the numerical solution. In order to understand this
role the mesh size in x-direction has been investigated. Three
different cases with different mesh size (N = 256, N = 356,
and N = 456) and a diaphragm pressure ratio P
4
/P
1
=10 have
been used. Figure 7 shows the effect of using larger number
of grid points in x-direction and how the solution is affected
accordingly. From the Figure 8 it will be seen that the details
of the shock wave are well captured when a larger number of
grid points are used and the program becomes more stable and
less oscillation is produced. Also the numerical and exact
solutions are now very comparable when number of grid
points N = 456 is used. Consequently, N = 456 have been
used in all of the following runs.
Distance (m)
P
r
e
s
s
u
r
e
(
P
a
)
0 0.25 0.5 0.75 1
1.0E+05
2.0E+05
3.0E+05
4.0E+05
5.0E+05
6.0E+05
7.0E+05
8.0E+05
9.0E+05
1.0E+06
Exact
CFD
P4= 1000 kPa
P1= 100 kPa
N= 256
High oscillation
Distance (m)
P
r
e
s
s
u
r
e
(
P
a
)
0 0.25 0.5 0.75 1
1.0E+05
2.0E+05
3.0E+05
4.0E+05
5.0E+05
6.0E+05
7.0E+05
8.0E+05
9.0E+05
1.0E+06
Exact
CFD
P4= 1000 kPa
P1= 100 kPa
N=356
Less oscillation
Distance (m)
P
r
e
s
s
u
r
e
(
P
a
)
0 0.25 0.5 0.75 1
1.0E+05
2.0E+05
3.0E+05
4.0E+05
5.0E+05
6.0E+05
7.0E+05
8.0E+05
9.0E+05
1.0E+06
Exact
CFD
P4=1000 kPa
P1=100 kPa
N=456
Good stability
Fig. 7 Pressure History at Different Mesh size in X-Direction at t =
0.2 ms
The x-t diagram is one of the important tools which
give a good estimation for the maximum useful test time that
can be obtained after removal of the diaphragm. Figures 8 and
9 show the predicted x-t diagram for pressure and density
profiles obtained for an air-air gas combination run. The
diaphragm pressure ratio used for this simulation is (P
4
/P
1
)
=10. Figure 8 represents the x-t diagram for the pressure
history; the contact surface does not appear in this figure as it
follows the shock wave continuously as it apparent in Figure
6, however it is very clear in Figure 9 which shows the x-t
history for the density profile.
0
0.01
0.02
0.03
T
i
m
e
[
s
e
c
]
0
Distance [m] X
Y
Z
0.5 1
Fig. 8 x-t diagram for pressure history
Figure 9 represents the density history along the
whole length of the test facility. Both of the shock wave and
the contact surface are displayed in this figure. The shock
wave is followed by the contact surface until the reflected
shock wave interact with the contact surface and then the
wave is further reflects. The expansion waves are displayed in
the two figures and due to the sufficient length of the driver
section the shock wave and contact surface are intersected
before the reflected expansion waves reach the end of the
driven section.
0
0.01
0.02
0.03
T
i
m
e
[
s
e
c
]
0
Distance [m]
X
Y
Z
0.5 1
Fig. 9 x-t diagram for density history
X. INVISCID TRANSIENT FLOW IN SHOCK TUBE
In this study, a sensitivity study on the flow model used
has been performed as well as the time-step size, towards the
accuracy of the flow solution applied for a shot with similar
gases allocated in both the driver and driven sections (air-air).
CFD solution for inviscid simulation for a diaphragm pressure
ratio P
4
/P
1
of 10 will be discussed. The simulation has been
conducted using the actual dimensions of the test facility
shown in Figure 10.
In order to validate the numerical formulation for the
inviscid terms, solver was applied to transient shock wave
motion in real shock tube. In the actual shock tube, a bush
was used at the primary diaphragm section adjacent to the
diaphragm to facilitate rapture process. In order to represent
the bush, in the solver, an artificial wedge was included as
shown in Figure 10. The exact shape of the bush cannot be
used in the solver since it can only handle H-type mesh and
therefore could not handle abrupt change in geometry.
International Journal of Engineering and Natural Sciences 4:1 2010
58
Fig. 10 Mesh spacing allocated for each section
The pressure, temperature, density and Mach number of
the flow were stored in two stations, the first station is at the x
= 6183 mm from the left hand side end of the driver section,
the second station is with an axial separation of 342 mm from
the first station as shown in Figure 11.
Station 2 Station 1
wall
end
Flow direction
Barrel
40 342
Fig. 11 The two stations at the end of shock tube
The pressure history for the above mentioned shot is
depicted in Figure 12 from which one can follow the physics
of the flow inside the shock tube. The first jump represents
the shock wave, for which the pressure inside the barrel
increases from 100 kPa to around 220 kPa. As the shock wave
proceeds to the end of the tube it will hits the wall and reflects
moving in the opposite direction increasing the pressure to
about 450 kPa. The shock wave will then interact with the
contact surface which is following the shock wave, and due to
this interaction between the shock wave and the contact
surface the pressure will be increased until it reaches its peak
pressure value of 530 kPa.
.
Time [sec]
P
r
e
s
s
u
r
e
[
P
a
]
0 0.005 0.01 0.015 0.02 0.025
1.0E+05
1.5E+05
2.0E+05
2.5E+05
3.0E+05
3.5E+05
4.0E+05
4.5E+05
5.0E+05
5.5E+05
Station 1
Station 2
Shock wave
Reflected shock
wave
Fig. 12 Pressure history for inviscid flow (air-air, P
4
/P
1
=10)
The shock wave speed can be determined from the CFD data
obtained from this simulation. As the distance between the
two stations is known (0.342 m) and the time of shock travels
from station 1 to station 2 can be obtained from the pressure
history graph, as show in Figure 13, the shock wave speed is
determined for this shot is 518 m/s. comparing to the
theoretical value for this pressure ratio (550 m/s) the
percentage error is around 6% which is reasonable
Time [sec]
P
r
e
s
s
u
r
e
[
P
a
]
0.0066 0.0068 0.007 0.0072 0.0074
1.0E+05
1.5E+05
2.0E+05
Station 1
Station 2
t = 0.0066 sec A
Fig. 13 Shock wave speed (inviscid flow)
Using the same procedure, the reflected shock wave speed can
be determined as the wave reflects from the tube end and
moves in the opposite direction (left direction), due to impact
with the end wall the wave will lose some of its kinetic energy
and consequently its speed decreases to about 342 m/s, as
shown in Figure 14 which represents a close view for the
reflection region in the pressure history graph. The time
period from station 2 to station when is around 0.001 sec
which is longer than the time period when the shock travels
from station 1 to station 2 due to the energy lost after
reflection.
Time [sec]
P
r
e
s
s
u
r
e
[
P
a
]
0.0075 0.008 0.0085
2.5E+05
3.0E+05
3.5E+05
4.0E+05
4.5E+05
Station 1
Station 2
t = 0.001 sec A
Fig. 14 Reflected sock wave speed
The same trend can be noted when the temperature
history is investigated as shown in Figure 15. The first jump
in the temperature profile represents the shock wave and the
second jump is due to the reflected shock wave. The
temperature is increased from the initial value 300 K to about
380 K due to shock wave affect and when the shock reflects
from the tube end the temperature rises to 475 and after
interaction between reflected shock wave and the contact
surface the flow temperature becomes about 490 K.
Time [sec]
T
e
m
p
e
r
a
t
u
r
e
[
K
]
0.005 0.01 0.015 0.02 0.025
300
325
350
375
400
425
450
475
500
Station 1
Station 2
Fig. 15 Temperature history inside the shock tube (inviscid flow)
International Journal of Engineering and Natural Sciences 4:1 2010
59
The x-t diagram is one of the important tools which
give a good estimation for the maximum useful test time that
can be obtained after removal of the diaphragm. Figures 16
and 17 show the predicted x-t diagram for pressure and
density profiles obtained for an air-air gas combination run.
The diaphragm pressure ratio used for this simulation is
(P
4
/P
1
) =10.
0
0.005
0.01
0.015
0.02
0.025
0.03
T
i
m
e
[
s
e
c
]
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
Distance [m]
X
Y
Z
Fig. 17 x-t diagram for the pressure profile (inviscid flow)
Figure 18 represents the density history along the
whole length of the test facility. Both of the shock wave and
the contact surface are displayed in this figure. The shock
wave is followed by the contact surface until the reflected
shock wave interact with the contact surface and then the
wave is further reflects. The expansion waves are displayed in
the two figures and due to the sufficient length of the driver
section the shock wave and contact surface are intersected
before the reflected expansion waves reach the end of the
driven section.
0
0.005
0.01
0.015
0.02
0.025
0.03
T
i
m
e
[
s
e
c
]
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
Distance [m]
X
Y
Z
Fig. 18 x-t diagram for the density profile (inviscid flow)
Finally, to investigate the flow properties in 2D scheme,
the results of the numerical simulations have been displayed
in contour plots. The contour plot of the pressure history
along the facility is shown in Figure 19. Time step used in this
simulation is 0.000005 sec and the total number of iterations
is 6000. The solver is programmed so that it stores the output
data after each one hundred iterations, subsequently there will
be 60 output files. Each file represents the data after 0.0005
sec. As shown in this Figure, driver pressure P
4
=100 kPa and
pressure in the driven section P
1
is 10 kPa.
Fig. 19 Contour plot for pressure history at t = 0
After diaphragm rupture, shock wave travels right through the
barrel while the expansion wave travels left through the driver
section. These two waves are captured after 0.005 sec and
shown in Figure 20.
Fig. 20 Shock and expansion waves
These two waves continue their journey towards the tube
ends, at time 0.0081 s the shock wave hits the barrel end on
the right hand side while the expansion wave reaches the left
hand side of the facility as shown in Figure 21.
Fig. 21Shock and expansion waves at the facility ends
The two waves then will reflect from the end of the tube as
shown in Figure 23.
Fig. 22 Shock and Expansion wave reflection
The reflected shock wave now moves to the left towards the
contact surface while the reflected expansion wave moves to
the right towards the contact surface and shock wave as
shown in Figure 23.
Fig. 23 Reflected waves move towards the contact surface
International Journal of Engineering and Natural Sciences 4:1 2010
60
The shock wave interacts with the discontinuity surface and
reflects again. This process continues for several times until
getting pressure balance along the whole facility as shown in
Figure 24.
Fig. 24 Interaction between shock wave and contact surface
Figure 25 shows velocity contour at t= 0.0029 sec after
diaphragm rapture, at this ime the shock did not reflects yet,
consequently the flow is symmetry and uniform in y-
direction.
Fig. 25 Velocity contour before shock reflection
As the shock wave reflects from the tube end it will move to
the left and interact with the discontinuity surface and the
flow no longer symmetry as shown in Figure 26.
Fig. 26 Velocity contour after shock reflection
XI. CONCLUSIONS
The paper described the formulation of the 2D-CFD solver
designed for simulation of flow in shock tube. The program
has been applied to a standard case of inviscid flow in shock
tube. The agreement with the analyzed solution is very good
which proved the validity of the basic numerical scheme
developed.
The present code showed good capability to provide the x-t
diagram successfully. From this diagram one can determine
the useful duration (for this case it is about 10 ms), which is
quite comparable compared to other facilities. It can be
concluded, based on the agreement with the analytical results,
that the numerical formulation for the inviscid part of the
solver is valid.
The results presented in this paper show that two-
dimensional modeling of the hypersonic test facility is an
effective way to obtain facility performance data. Although
this paper focused on the HTF facility, the CFD code is
generic and may be applied to other facilities.
The code could be further improved if a cylindrical
coordinate system is used for mesh generation instead of the
Cartesian coordinate system currently used. The simulations
had successfully indicated that the flow is symmetry before
shock wave reflection off the tube end and the flow is
disturbed after shock reflection and interaction with contact
discontinuity.
REFERENCES
[1] Caleffi, V., Valiani, A. and Zanni, A. (2003), Finite Volume Method
for Simulating Extreme Flood Events in Natural Channels, Journal of
Hydraulic Research, Vol. 41, No. 2, pp. 167-177.
[2] Valiani, A., Caleffi, V. and Zanni, A. (2002), Case Study: Malpasset
Dam-Break Simulation using a 2D Finite Volume Method, Journal of
Hydraulics Engineering, Vol. 128, No. 5, pp. 460-472.
[3] McDonald, P.W. [1971] 'The Computation of Transonic Flow
Through 2D Gas Turbine Cascades ASME Paper No. 71-GT-89
[4] Couston, M. [1976] 'Time-Marching Finite Area Method' V.K.I.
Lecture Series 84
[5] Thiaville, J.M. [1977] 'Comparison of a Finite Difference Method with
a Time-Marching Method for Blade to Blade Transonic Flow
Calculations' . in Transonic Flow Problems in Turbomachinery, (ed.
Adamson, T.C & Platser, M.F.)
[6] Farn, C.L. and Whirlow, D.K. [1977] 'Application of Time-Dependent
Finite Volume to Transonic Flow in Large Turbine' in Transonic
Flow Problems in Turbomachinery, (ed. Adamson, T.C & Platser,
M.F.)
[7] Denton, J.D. [1975] 'A Time Marching Method for Two- and Three-
Dimensional Blade to Blade Flows' ARC R. & M. No. 3775
[8] Lerat, A. and Sides, J. [1982] 'A New Finite Volume Method for the
Euler Equations with Applications to Transonic Flows' in Numerical
Methods in Aeronautical Fluid Dynamics, (Ed. Roe, P.L), Academic
Press, pp. 245-288
[9] Rizzi, A. and Eriksson, L.E. [1981] Transfinite Mesh Generation and
Damped Euler Equations Algorithm for Transonic Flow Around
Wing-Body Configurations . AIAA 5th Computational Fluid
Dynamics Conference, Paulo Alto, pp. 43-68
[10] Rizzi, A. [1982] Damped Euler-Equations Method to Compute
Transonic Flow Around Wing-Body Combinations AIAA Journal,
Vol. 20
[11] Jameson, A., Schmidt, W. and Turkel, E. [1981] 'Numerical Solutions
of the Euler Equations by Finite Volume Methods using Runge-Kutta
Time Stepping Schemes' AIAA Paper No. 81-1259
[12] Jameson, A. [1982] 'Transonic Aerofoil Calculations Using the Euler
Equations' Numerical Methods in Aeronautical Fluid Dynamics (P.L.
Roe ed.), Academic Press, pp. 289-308
[13] Jameson, A. and Baker, T.J. [1983] 'Solution of the Euler Equations
for Complex Configurations' Proc. of the AIAA 6th Computational
Fluid Dynamics Conference, AIAA, New York, pp. 293-302
[14] Buratynski, E.K. and Caughey, D.A. [1986] 'An Implicit LU scheme
for the Euler Equations Applied to Arbitrary Cascades' AIAA Journal,
24 (1): 39-46
[15] Dawes, W.N. [1987] 'Application of a Three-Dimensional Viscous
Compressible Flow Solver to a High-Speed Centrifugal Compressor
Rotor - Secondary Flow and Loss Generation' Proc. Inst. Mech.
Engrs., Conf. Turbomachinery Efficiency, Prediction and
Improvement, C261-87, pp. 53-61
[16] Van Leer, B. [1979] Towards the Ultimate Conservative Difference
Schemes. V. A Second Order Sequel to Gudunovs Method J. Comp.
Physics, 32 : 101-36
[17] Harten, A. [1983] High Resolution Schemes for Hyperbolic
Conservation Laws Journal of Computational Physics, 49, 357-93
[18] Harten, A. [1984] On a Class of High Resolution Total Variation
Stable Finite Difference Schemes SIAM Journal of Numerical
Analysis, 21, 1-23
[19] Osher, S. [1984] Riemann Solvers, the Entropy Condition and
Difference Approximation SIAM Journal Numerical Analysis, 21, pp.
217-35
[20] Osher, S. and Chakravarthy, S.R. [1984] High Resolution Schemes
and the Entropy Condition SIAM Journal of Numerical Analysis, 21,
pp. 955-84
International Journal of Engineering and Natural Sciences 4:1 2010
61
[21] Yusoff M. Z. [1997] An improved treatment of two-dimensional
Two-Phase Flow of Steam by a Runge-Kutta Method,Ph.D. Thesis,
University of Birmingham
[22] Yusoff M. Z. A Two-Dimensional Time-Accurate Euler Solver for
Turbo machinery Applications Journal-Institution of Engineers,
Malaysia Vol. 5 No. 3 1998
[23] Hirsch, C. [1990] Numerical Computation of Internal and External
Flows
Volume 1(Fundamentals of Numerical Discretization) and Volume 2
(Computational Methods for Inviscid and Viscous Flows), John Wiley
& Sons
[24] Gustafsson, B. and Sundstrom, A. [1978] Incompletely Parabolic
Problem in Fluid Dynamics SIAM Journal of Applied Mathematics,
35 (2): 343-357
[25] Pulliam, T.H. [1986] Artificial Dissipation Model for the Euler
Equations. AIAA Journal , 24: 1931-1940
[26] Swanson, R.C. and Turkel, E. [1987] Artificial Dissipation and
Central Difference Schemes for the Euler and Navier Stokes
Equation. AIAA Paper No. 87-1101, Proc. AIAA 8th Computational
Fluid Dynamics Conference, pp. 55-69
[27] Caughey, D.A. and Turkel, E. [1988] Effects of Numerical
Dissipation of Finite Volume Solutions of Compressible Flow
Problems AIAA Paper No. 88-0621, AIAA 26th Aerospace Sciences
Meeting
[28] Bamkole, B.O. [1991] A Two Step Method for Solving the Euler
Equations Internal Report, Manuf. & Mech. Eng. Dept., University of
Birmingham
[29] G. A. Sod, A survey of several finite difference methods for systems of
nonlinear hyperbolic conservation laws, J. Comput. Phys. 43 (1978)
1-31
NOMENCLATURE
a Speed of sound
D Artificial dissipation component
e Specific internal energy
e
o
Total internal energy
F Axial component of the inviscid flux vector
FT Time step factor
G Tangential component of the inviscid flux
vector
h
o
Stagnation enthalpy
h Specific enthalpy
P Static pressure
P
o
Stagnation pressure
P
b
Downstream static pressure
R Flux residual
SF2X Second order smoothing factor in axial
direction
SF2Y Second order smoothing factor in tangential
direction
SF4X Fourth order smoothing factor in axial
direction
SF24Y Fourth order smoothing factor in tangential
direction
T Temperature
t Time
At Time step for main calculation
V Overall velocity
V
x
Velocity component in axial direction
V
y
Velocity component in tangential direction
w Conserved variable vector
x Axial distance
y Tangential distance
Greek Symbols
o Integration constant in Runge-Kutta time
stepping
Enthalpy damping coefficient or coefficient
in scaling factor equation
c Pressure gradient sensor
p Density
u Implicit residual averaging coefficient
O Volume of element
Artificial dissipation coefficient component
Subscript
o Stagnation condition
x Cartesian co-ordinates
y Cartesian co-ordinates
1 Initial state in driven section
2 Flow conditions between shock wave and
contact surface
3 Flow conditions between rarefaction wave
and contact surface
4 Initial state in driver section
International Journal of Engineering and Natural Sciences 4:1 2010
62