Download as pdf or txt
Download as pdf or txt
You are on page 1of 52

Hypersonic Flow on a Blunted Cone-Flare and in the VKI-H3

Mach 6 wind tunnel

W. Dieudonne, H. L. Boerrigter, J. M. Charbonnier


von Karman Institute
Chaussee de Waterloo, 72
B-1640 Rhode-St-Genese
Belgium

Abstract
This note presents the results obtained in the VKI H3 Mach 6 wind tunnel on an axisym-
metric blunted cone- are for the surface pressure and heat transfer. A numerical study carried
out using the VKI MultiBlock code is compared to experimental results. Attention is focused
on grid convergence and it appears that the separation area where the experimental results
di er from the numerical ones, is not grid converged.
The fair agreement of the conical section of the blunted cone leads to the simulation of
the ow in the wind tunnel nozzle for both laminar and turbulent ows to asses the quality
of the design of the nozzle. The results in the test section show a good agreement with the
experimental data.
Finally, using the information obtained from the ow eld in the wind tunnel, the ow over
the blunted cone is re-computed. Although the separated region is still underestimated, a very
precise prediction of the blunted cone- are wall data is achieved with the numerical tools.
Contents
1 Introduction 1
2 Experimental Results 2
2.1 Wind tunnel : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 2
2.2 Surface static pressure acquisition : : : : : : : : : : : : : : : : : : : : : : : : : : 2
2.3 Infrared thermography: temperature acquisition : : : : : : : : : : : : : : : : : : 3
2.4 Repeatability, ow quality assessment and results : : : : : : : : : : : : : : : : : 5
3 Presentation of the code 7
3.1 Solver for the Navier-Stokes equations : : : : : : : : : : : : : : : : : : : : : : : 7
3.2 Solver for the turbulence model : : : : : : : : : : : : : : : : : : : : : : : : : : : 8
3.2.1 Algebraic model : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 8
3.2.2 One-equation model : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 9
4 The ow over the blunted cone- are 11
5 Wind tunnel nozzle ow computations 13
5.1 VKI H3 nozzle design : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 13
5.2 Computation of the ow eld in the H3 wind tunnel : : : : : : : : : : : : : : : 13
5.2.1 Laminar nozzle wall boundary layer : : : : : : : : : : : : : : : : : : : : 13
5.2.2 Turbulent nozzle wall boundary layer : : : : : : : : : : : : : : : : : : : 15
5.3 Rebuilding of the blunted cone- are in the VKI H3 wind tunnel : : : : : : : : : 16
6 Conclusions 18

i
List of Figures
1 Von Karman Institute H3 hypersonic wind tunnel : : : : : : : : : : : : : : : : : 20
2 Pressure model with the pressure tappings location : : : : : : : : : : : : : : : : 20
3 Uncertainty associated with the slopes for the E.P.S. : : : : : : : : : : : : : : : 21
4 Time response of the pressure scanner system : : : : : : : : : : : : : : : : : : : 21
5 Mach number isolines obtained from [10], from top to bottom: 10, 100 and 140
mm behind the nozzle exit : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 22
6 Pressure at low Re for the four rolls : : : : : : : : : : : : : : : : : : : : : : : : 23
7 E ects of Yaw and Pitch on separation and reattachement : : : : : : : : : : : : 23
8 Temperature isolines on the cone- are model showing ow axisymmetry : : : : 24
9 Temperature isolines on the cone- are model showing ow misalignment : : : : 24
10 Stanton number sensitivity to yaw (constant pitch) : : : : : : : : : : : : : : : : 25
11 Repeatability of consecutive runs : : : : : : : : : : : : : : : : : : : : : : : : : : 25
12 Repeatability on 2 days with model remounted : : : : : : : : : : : : : : : : : : 26
13 Pressure Coecient (x10;2) and Stanton number : : : : : : : : : : : : : : : : : 26
14 Pressure isolines : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 29
15 Mach number isolines : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 29
16 Velocity vectors in the recirculation bubble : : : : : : : : : : : : : : : : : : : : 30
17 Mesh for the medium level (201x41) with clustering : : : : : : : : : : : : : : : 30
18 Pressure coecient on the blunted cone : : : : : : : : : : : : : : : : : : : : : : 31
19 Stanton number on the blunted cone : : : : : : : : : : : : : : : : : : : : : : : : 31
20 Skin friction coecient on the blunted cone : : : : : : : : : : : : : : : : : : : : 32
21 Comparison of experimentally and numerically determined pressure coecient : 32
22 Comparison of experimentally and numerically determined Stanton number : : 33
23 H3 nozzle geometry : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 36
24 Example of a coarse grid for computing the nozzle ow : : : : : : : : : : : : : : 36
25 Boundary conditions for the outlet : : : : : : : : : : : : : : : : : : : : : : : : : 36
26 Wall pressure along the nozzle for the laminar case : : : : : : : : : : : : : : : : 37
27 Wall heat ux along the nozzle for the laminar case : : : : : : : : : : : : : : : : 37
28 Pitot pressure at 10mm behind the nozzle exit - laminar case : : : : : : : : : : 38
29 Mach number at 10mm behind the nozzle exit - laminar case : : : : : : : : : : 38
30 Mach number detail at 10mm behind the nozzle exit - laminar case : : : : : : : 39
31 Mach number along the centerline - laminar case : : : : : : : : : : : : : : : : : 39
32 Detail in the jet ow for the Mach number along the centerline - laminar case : 40
33 E ects of turbulence and transition location on centerline Mach number : : : : 40
34 E ects of turbulence and transition location on Mach number at 10mm behind
the nozzle exit : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 41
35 E ects of turbulence and transition location on normalised wall pressure : : : : 41
36 E ects of turbulence and transition location on wall heat ux : : : : : : : : : : 42
37 Comparison of the experimental and numerical Mach number on the centerline
in the test section : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 42
38 Comparison of the experimental and numerical Pitot pressure at various cross
sections in the jet ow : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 43
39 Mach number isolines in the throat region : : : : : : : : : : : : : : : : : : : : : 44
40 Mach number isolines in the nozzle region : : : : : : : : : : : : : : : : : : : : : 44
41 Mach number isolines in the jet ow : : : : : : : : : : : : : : : : : : : : : : : : 44
42 Boundaries of the wind tunnel and the blunted cone- are : : : : : : : : : : : : 44
43 Mach number distribution on the outer boundary of the blunted cone- are : : : 45
44 Velocity vector angle on the outer boundary of the blunted cone- are : : : : : : 45

ii
45 Iso Mach lines above the bow shock of the blunted cone- are in the wind tunnel 46
46 Wall pressure on the blunted cone- are for the uniform, non-uniform incoming
ow and experiments : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 47
47 Wall heat ux on the blunted cone- are for the uniform, non-uniform incoming
ow and experiments : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 47

iii
List of Tables
1 H3 ow elds conditions : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 2
2 Grids used for the blunted cone- are : : : : : : : : : : : : : : : : : : : : : : : : 11
3 Location of separation and reattachment points : : : : : : : : : : : : : : : : : : 11
4 Grid size for the H3 laminar computation : : : : : : : : : : : : : : : : : : : : : 14
5 Grid size for the H3 turbulent computation : : : : : : : : : : : : : : : : : : : : 16
6 Wall pressure data P [Pa] as a function of x [m] on the blunted cone- are : : : 27
7 Wall heat ux data q [W/m2] as a function of x [m] on the blunted cone- are : 28
8 Coordinates of the H3 nozzle contour (all data in mm) : : : : : : : : : : : : : : 35

iv
1 Introduction
The ow on a blunted cone- are possesses some of the major characteristics of the ow over
reentry vehicles. At hypersonic speeds, the bluntness causes a curved bow shock with the
associated entropy layer, and the presence of the are induces the ow to recirculate. The
recirculation, i.e. reversed ow motion, is going to be studied in more detail since it represents
the most dicult part of the simulation.
The investigation of the ow eld over a blunted cone- are is performed on two levels:
experimentally (H3 hypersonic wind tunnel of the VKI) and numerically. The two surface
data retained for this experiment are the pressure coecients obtained by means of electronic
pressure transducers, and the heat transfer coecients obtained through the use of infrared
thermography. The uncertainties are assessed and the e ects of ow misalignment are carefully
investigated.
The numerical study conducted with the VKI MultiBlock code are shown to give good
results except in the recirculation bubble where some problems still remain. The in uence of
grid re nement is presented and some aspects of the numerical method are discussed.
The ow eld in the H3 wind tunnel is then investigated by means of CFD simulations.
Both laminar and turbulent (one equation model) computations are carried out and the results
are compared to available experimental data.
Finally the ow on the blunted cone- are is recomputed taking into account the computed
test section ow eld using a non-uniform boundary condition. The new set of wall data are
compared to the rst computation and to the experimental data and the results appear to be
improved.

1
2 Experimental Results
2.1 Wind tunnel
The experiment is carried out in the H3 wind tunnel of the VKI (see gure 1). It is a hypersonic
blowdown wind tunnel with a Mach 6 free jet. The gas used is air, the stagnation pressure
ranges from 7 to 32 bars and the stagnation temperature ranges from 480 to 580 K. The
stagnation temperature is measured with a copper/constantan thermocouple (type K) located
in the settling chamber. Three operating conditions for the H3 wind tunnel have been de ned,
they correspond to three reservoir pressures:
 the Low Reynolds Case (P0=10 bars), Re=8.106 /m
 the Medium Reynolds Case (P0=20 bars), Re=14.106 /m
 the High Reynolds Case (P0=30 bars), Re=21.106 /m
The low Reynolds case has been chosen for this study because it guarantees that the ow on
the blunted cone- are is laminar even after reattachement.
The calibration of the thermocouples yields the following relation:

T0[K ] = 24:49ET 0[mV ] + 274:00 (1)


where ET 0 is the voltage output from the thermocouple. The error band is 0.18 K/mV.
The stagnation pressure is measured in the settling chamber with an absolute Statham (strain
gauge) pressure transducer. The pressure transducer is calibrated using a hydraulic balance
system. The resulting equation from the pressure transducer calibration is:

P0 [bar] = 3:7824EP 0[mV ] + 0:0598 (2)


where EP 0 is the voltage output of the transducer. The error band for the calibration slope is
0.0112 bar/mV giving an uncertainty of 0.30% with a con dence level of 10:1. The wind tunnel
side window can be adapted with a Germanium viewing window with a high transmissivity in
the infrared spectral range. The top window can also be mounted with a Germanium window
so that infrared testing can be carried out on two di erent planes.
The wind tunnel operating conditions have been recorded during the tests and they are
given for the low Reynolds case in table 1. It should be noted that the Mach number is based
on the design value, thus it may di er slightly and this point will be further investigated.

P0 1063650 Pa
T0 550 K
Mach 6

Table 1: H3 ow elds conditions

2.2 Surface static pressure acquisition


The model on which the static pressure is recorded is an axisymmetric blunted cone- are. A
schematic is presented in gure 2. The nose is detachable so that several nose radii can be
attached. For these experiments, the 3.5 mm rounded nose is chosen. The cone surface is
inclined at 7.5o with respect to the symmetry axis and the are de ection is 10o with respect
to the cone. The pressure model is made of aluminium to withstand the high temperatures

2
(especially at the nose). There are 19 pressure tappings aligned on the cone- are model. A set
of 6 supplementary pressure tappings are located at 90, 180 and 270o on the same radius as
the rst and the last pressure tappings, thus allowing to check for the ow eld axisymmetry.
An electronic pressure scanner connected to a personal computer is used to accurately
and quickly record the 25 pressures during the test. The electronic pressure scanner (E.P.S.)
uses two ESP-32RG scanners from Pressure Systems Inc., that is 64 channels are available
for recording. The manual with the complete hardware description can be found in [1]. As
mentioned by Zemsch [1], there is a linear relation between the number of digits and the
pressure. Each pressure scanner is thus represented by the two constants of the calibration
curve y=ax+b. The uncertainty in the EPS data is computed for each scanner (hence a total
of 64 uncertainty evaluations) and is plotted in gure 3. The uncertainty associated with the
slope decreases down to a constant level as the pressure increases. The high uncertainties for
the low pressures are due to the fact that the static error associated to the pressure reading is
relatively important, this does not a ect the pressure results during tests because the data is
in the high range.
The time response of the pressure measurement system is also computed. The time response
is the time it takes for the information (e.g. pressure) to be received without any deformation
and this is generally linked to piping (the length and the diameter). For the case considered,
it is the piping from the rst pressure port to the EPS system (longest distance). The time
responses are obtained by calculating a Fast Fourier Transform [2]. As shown in gure 4,
the worst case is for the low Reynolds because a perturbation takes longer to travel in a low
pressure (i.e. low density) medium, than in a high one. This analysis lead to the choice of a 3
second delay after model injection before recording the pressure data.
2.3 Infrared thermography: temperature acquisition
For the surface temperature measurement, a plexiglas model of the blunted cone- are is built.
It is coated with special paint to increase its emissivity and to decrease its re ectivity. The
nose of the model is removable and made of aluminium to resist the high temperatures in this
region. The infrared (IR) camera is an Agema 900 equipped with a 20x10o (HxV) eld of view
lens. The detector is cryogenically cooled with liquid nitrogen and it has a high detectivity in
the long wave spectral range (8 to 12 m) corresponding to the mean infrared. As speci ed by
the manufacturer, the infrared images are acquired at a frequency of 15 Hz, the sensitivity is
0.08 at 30oC and the accuracy is 1 C or 1% (absolute temperature measurements).
The signal output is given in intensity units, hence the camera has to be calibrated to
correlate the intensity to the true temperature. To generate a variable temperature source, a
metallic plate equipped with electric resistances is mounted on a polystyrene cushion. The plate
has the same coating as for the model. Using a high conductivity glue, calibrated thermocouples
are placed onto the plate. They are located in a region where the heating is the most uniform.
The heat source is then mounted into the wind tunnel at the location where the model will
be during tests. The data recorded is thus a voltage and an intensity level (from the Agema
System). The data shows a rather good linear distribution in the narrow temperature band
and a best t curve is:

I = 598:61  V [mV ] ; 499:74 (3)


Now using equation 1, the relation linking the intensity to the temperature is:

I = 25:79  T + 894:33 (4)

3
The error band for the calibration slope is  1.91 intensity unit/degree giving an uncertainty
of 7.40% with a con dence level of 20:3. This seems to be a rather high uncertainty, however
it is the best that could be achieved.
The procedure for acquiring the temperature data is described in [3]. The camera gives
an intensity eld which can be translated into a temperature eld. From the temperature,
one can obtain a heat transfer coecient (Stanton number). To extract the Stanton number
distribution from the IR tests, the semi-in nite slab technique with the constant heat transfer
coecient model is used. The equation for conduction is:

r2T = 1 @T
@t with k
= c (5)
where is the thermal di usivity coecient, k is the thermal conductivity of the wall,  the
wall density material and c the speci c heat of the wall material which are assumed constant
in the temperature range of the tests. The assumptions are that there is no heating of the
opposite wall and that there is no transverse conduction, hence equation 5 simpli es to:

@ 2T = 1 @T (6)
@z2 @t
The boundary conditions are:
 T prescribed at t=0 (initial condition) and at z=1 (boundary condition),
 on the model surface:

;k @T dt
@z jz=0 = ;k dz jz=0 = q_w (t) (7)
The heat transfer at the wall at time t is found using the constant heat transfer coecient
model (h=cst ):

qw (t) = h(Tr ; Tw (t)) at t=0: qw0 = h(Tr ; Tw0) (8)


so
 
qw (t) = qw0 1 ; TT;wT(t) (9)
r w0
therefore from equation 6:
p
Tw (t) 2 where = pck(T ;t T )
qw 0
T ; T = 1 ; e erf ( )
r w0 r w
(10)
and the error function is

Zt
erf ( ) = p2 0 e;x2 dx (11)

4
The coecient is found from a classic Newton-Raphson iterative method:

n+1 = n ; ff0(( )) with f ( ) = e 2 (1 ; erf ( )) ; (1 ; dT  I )


dI T0 ; Tw0
(12)
Then using equation 12 the heat transfer to the wall can be computed. From the heat
transfer, the Stanton number is easily found using the modi ed de nition:

St =  U C q(_ T ; T ) (13)
1 1 p 0 w

2.4 Repeatability, ow quality assessment and results


The model positioning is conditioned by several factors such as nozzle blockage. To choose the
best location for the model, one needs data describing the ow eld in the test section as found
in Boerrigter [4]. In gure 5, the results for the iso Mach lines in a cross plane show that the
ow is axisymmetric in the potential core region. Therefore the model is mounted such that
it is at the centerline and the nose is at 10mm behind the nozzle exit. It is seen however that
the Mach number varies in the streamwise direction: from M=5.98 at 10mm to 6.06 at 140mm
behind the nozzle exit.
One of the most dicult tasks for this experiment is to align the model with the incoming
ow. The results for the pressure coecient and for the Stanton number must be shown to
be axisymmetric in order to be reliable. The e ects of pitch and yaw have been investigated
and they lead to the conclusion that although the alignment with the incoming ow is critical
(1/10th of a degree creates large perturbations of the surface data, especially on the separated
area), it is necessary to achieve a relative 0o positioning.
The uncertainties for the pressure and the heat transfer data have been computed:
 St=11.5 %
 p = 7.3 %
The pressure model is mounted with a set of two circumferential pressure taps for the rst
and the last pressure tap. The data obtained from these readings give a rst indication on the
ow axisymmetry but are imprecise. To be more accurate, the pressure model is rotated along
its centerline at 0, 90, 180 and 270o and the model pitch and yaw are tuned until the four sets
of data coincide. The data showing that ow axisymmetry has been reached are presented in
gure 6. The e ects of yaw and pitch misalignments on the surface pressure distribution are
shown in gure 7.
The heat transfer data should also show that the ow is axisymmetric. Because the two
models do not have the same mounting, it is not possible to keep the pitch and yaw alignment
found from the pressure measurements. The procedure adopted is to make temperature mea-
surements with the camera aside of the wind tunnel and to tune the pitch angle. Then the
camera is set on top of the wind tunnel and the yaw angle is tuned. This has been repeated
twice because of the dependency between the yawing and pitching angles. The nal results for
the ow alignment is well emphasised by plotting the iso-temperature lines on the whole model
(see gure 8). It is observed in gure 9 where the model is misaligned with the incoming ow
by 0.25o that the ow over the model is extremely sensitive to the angle of attack.
The isolines corresponding to the separation and the reattachement are fairly straight show-
ing that the relative 0o angle has been reached. To illustrate the sensitivity of the heat transfer
to ow misalignment, the evolution of the Stanton number as a function of yaw for a constant
pitch is shown in gure 10. It can be seen that a slight ow misalignment produces appreciable
shifts in the heat transfer to a much larger extent as for the pressure. It can be noted that as
5
shown in [3] the ow eld sensitivity to misalignment is such that a 0.25o shift can change the
nature of the ow: from laminar to turbulent.
The repeatability of the pressure and heat transfer experiments has been checked. Two
types of repeatability are considered, the test is run twice without altering the set-up, and the
model is dismounted from the wind-tunnel and then placed back in position and the experiment
is repeated. These two types of repeatability tests are presented in gure 11 and gure 12 and
demonstrate that the experiments are repeatable.
The results for the pressure coecient and the Stanton number are shown in gure 13 and
tabulated in tables 6 and 7. From this graph, it seen that the locations of separation and
reattachement given by the pressure and the temperature measurements are the same. This
tends to prove the good quality of the results.

6
3 Presentation of the code
3.1 Solver for the Navier-Stokes equations
The code used is the VKI Multiblock code, hereafter referred to as the VKI-MB. A very brief
description is given, more information can be found in [5], [6]. The VKI-MB is a multido-
main nite volume cell centered code based on an upwind numerical scheme. The spatial
discretization is based on quadrilaterals since the code is designed for structured grids.
The unsteady axisymmetric Navier-Stokes equations in conservative di erential form can
be found in Ho mann [7].
The Navier-Stokes equations in non-dimensionalized form are:
@Q + @E + @F + H = @Ev + @Fv + H (14)
@t @x @y @x @y v

is 0 for a two dimensional planar ow and is 1 for an axisymmetric ow. The quantities
Q, E, F, H, Ev , Fv and Hv are the ux vectors and axisymmetric source terms. They are
de ned as:
h iT
Q = [; u; v; et]T E = u; u2 + p; uv; (et + p)u

h iT h iT
F = v; uv; v2 + p; (et + p)u H = 1y v; uv; v 2; (et + p)v

Ev = [0; xyp; xy ; uxxp + vxy ; qx ]T Fv = [0; xy ; yyp ; uxy + vyyp ; qy ]T

2 3
0  
6 2 y @ v 7
166 xy ; 3 Re1 @x y  
7
7
Hv = 6
y 4 yyp ;  ; 23 Re1 vy ; Rey1 23 @y@  yv  
6 7
7
(15)
  5
uxy + vyyp ; qw ; 23 Re1 vy2 ; Rey1 @y@ 32  vy2 ; Rey1 @x@ 32  uvy
where the subscript v denotes viscous ux vectors, on the right-hand-side of equation 14.
The stress tensor components are:

xxp = Re 4 @u ; 2 @v  yyp = Re

4 @v ; 2 @u 
1 3 @x 3 @y 1 3 @y 3 @x
     
 @u @v
xy = Re @y ; @x  2 @u @v 4 v
 = Re ; 3 @x + @y + 3 y
1 1

qx = ; Re Pr(  ; 1)M 2 @T qy = ; Re Pr(  ; 1)M 2 @T (16)


1 1 @x 1 1 @y
The axisymmetric case which is three dimensional by nature can be reduced to a quasi two
dimensional problem thus avoiding the complexity of 3D solvers (e.g. mesh generation) and
extensive CPU resources.
7
The role of the limiting functions has been studied [3] and as shown by many authors, Roe's
scheme for hypersonic blunted case is not appropriate. Behind the bow shock the velocity
vectors show that the ow recirculates, this is called the Carbuncle phenomenon. To avoid this
numerical error, the limiting function has to be changed and that is why the hybrid scheme [8]
has been chosen for all the computations. Indeed, this recirculatory ow disappears with the
hybrid scheme.
To generate the viscous grids, a hyperbolic grid generator has been chosen [9]. A grid
sequencing technique has been adopted to check for grid convergence and to accelerate the
overall convergence. For this sequencing, a medium grid and a coarse grid are created by twice
eliminating one node out of two. Then the code is run on the coarse grid and the results are
interpolated onto the medium grid. This procedure is repeated for the ne grid.
3.2 Solver for the turbulence model
3.2.1 Algebraic model
A brief description of the Baldwin-Lomax algebraic model is given as it will be used for the
computation. A more detailed description can be found in [10].
The Baldwin-Lomax model has been chosen over the Michel algebraic model because it is
better adapted for the case of small recirculatory ows. It is a two layer model with the eddy
viscosity given by:
(
T = Ti if y  ym ; (17)
To if y > ym :
where ym is the minimum distance at which the inner eddy viscosity equals the outer one.
The inner layer eddy viscosity is derived from the Prandtl mixing length:
2 j! j
Ti = lmix (18)
where the mixing length writes:
h +i
lmix = y 1 ; e;y+ =A0 (19)
and ! is the magnitude of the vorticity vector.
In the outer layer, the eddy viscosity is de ned as:
To =  CcpFwake FKleb (y; ymax=CKleb ) (20)
where Fwake is given by:
2 =Fmax]
Fwake = min[ymaxFmax; Cwk ymaxUdif (21)
where Udiff is the di erence between the minimum and the maximum velocity in one
section. Fmax is the maximum value of Fy in the section and ymax is the corresponding wall
distance where the maximum is located. The function Fy writes:
+ !!
Fy = y j!j 1 ; exp ; y + (22)
A0
FKleb is Klebano 's intermittency function de ned as:
 6!;1
C
FKleb (y ) = 1 + 5:5 yKleb y (23)
max

8
The closure coecients for the Baldwin-Lomax model are:
A+0 = 26 Ccp = 1:6 CKleb = 0:3 (24)
Cwk = 0:25  = 0:4 = 0:0168
This turbulence model is not suited for ows with large recirculation zones but it can give
valuable approximation for the eddy viscosity magnitude. Actually, this model can be used
for de ning the initial conditions for more complex models which may be more adapted to the
case of separated ows.
3.2.2 One-equation model
The one-equation turbulence model implemented in the code is the Spalart-Allmaras model
[11], [12]. The choice of this model is supported by three reasons. First, it is a local model, i.e.
each point is computed independently of the rest of the ow (unlike algebraic models which are
based on velocity or vorticity pro les). Therefore it is well adapted to separated ows where
multiple shear layers are present. Moreover, it is more general than algebraic models and it
can be applied to a large number of ow situations without any changes. Finally, it is more
ecient than two equation models which have not shown very accurate predictions in separated
ows. Moreover it does not degrade the convergence as much as two equation based models
and it does not impose the same very tough constraints in terms of y+ distribution because
the transport quantity chosen behaves linearly near the wall.
The central quantity is the turbulent viscosity or eddy viscosity t . The eddy viscosity is
replaced by ~ following the relation:

3
t = ~fv1 fv1 = 3 + c3   ~ (25)
v1
The variable ~ is obtained through the transport equation:
D~ = c [1 ; f ] S~~
Dt b1 t2
1h
+  r:(( + ~)r~) + cb2 (r~)2
i

   2
w1 w
c
; c f ; f ~
b 1
2 t2 d
+ ft1 U 2
(26)
The four terms are the production term, the di usion term, the destruction term and the
source term (or trip term).
For the production term, the vorticity magnitude S is replaced by S~ de ned as:

S~  S + 2d2 fv2 (27)


where the function fv2 is:

fv2 = 1 ; 1 +f (28)


v1

9
and d is the distance to the closest wall.
The function ft2 is simply used to ensure that ~ is a stable solution:

ft2 = ct3exp(;ct42 ) (29)


No particular function appears in the di usion term; for the destruction term the function
fw is used. Its role is to accelerate the decay of the destruction term in the outer part of the
boundary layer, this is because it was observed that the predicted skin friction on a at plate
was too high. It writes as:
" 6 #1=6
1 + c
fw (r) = g g6 + cw63 g = r + cw2(r6 ; r) r  S2td2 (30)
w3
Finally, in the trip term U represents the velocity di erence from the eld to the trip
point. The function ft1 is computed as follows:
!
2
ft1 = ct1 gtexp ;ct2 !Ut 2 [d2 + gt2 + d2t ] (31)

where !t is the wall vorticity at the trip point and dt the distance from the eld to the trip
point. The function gt writes:
 
gt  min 0:1; xt !U (32)
t
The closure coecients used in this model are:
cb1 = 0:1355  = 2=3 cb2 = 0:622  = 0:41
cw1 = cb1= + (1 + cb2)= cw2 = 0:3 cw3 = 2 cv1 = 7:1 (33)
ct1 = 1 ct2 = 2 ct3 = 1:2 ct4 = 0:5

10
4 The ow over the blunted cone- are
The VKI-MB code is run with the stagnation conditions recorded during the experiments:
the total pressure is P0=1063650 Pa and the total temperature is T0 =550 K. The ow Mach
number is supposed to be uniform, parallel to the x-axis and equal to 6. For the grid sequencing
technique three grid levels are employed, the grid sizes are given in table 2.

Grid level Size


coarse 101x21
medium 201x41
ne 401x81

Table 2: Grids used for the blunted cone- are


The isopressure lines and the iso Mach lines for the 401x81 grid are plotted in gures 14
and 15. From these graphs, the bow shock is clearly visible and the ow eld associated with
the separation bubble is well illustrated. Figure 16 shows a plot of the velocity vectors in the
separated ow region, with the separation and the reattachment of the boundary layer. Finally,
to illustrate the computational domain, the medium size grid is shown in gure 17 where the
grid clustering in the separated area is quite noticeable.
The computations for the three grid levels (coarse, medium and ne) are carried out and
the results for the pressure coecient are plotted in gure 18. It clearly appears from this
gure that the solution is still grid dependent especially in the recirculation area. The onset
of separation predicted using the medium grid varies from the value found with the ne grid.
The Stanton number on the model surface for the three grids is plotted in gure 19. It is
seen that grid convergence has been reached on the conical section but the separated region
is still changing. Figure 20 for the skin friction coecient shows well the location of the
separation and reattachement points. From the skin friction values, the location of separation
and reattachment is plotted in table 3.
From these results, the maximum grid level with 401 nodes on the surface does seem to
be ne enough. It should be noted that the surface nodes are not linearly distributed but the
number of points in the vicinity of the hinge line has been increased to obtain grid clustering.
The numerical results for the ne mesh are plotted against the experimental results. Both
gures 21 and 22, for the pressure coecient and the Stanton number respectively, show that
there is an appreciable shift in the location of the separation region. It should be noted that
the numerical results have a constant o set from the experimental data by about 4%. This
will be further investigated. The trend observed with the grid sequencing technique is that
the recirculation bubble starts close to the hinge line and monotonically expands. Therefore
the results look encouraging but the grid size needed for this computation shows the diculty
associated with the recirculation bubbles.
From the comparison between the numerical and the experimental data two e ects can be

Grid xsep n xhinge xreat n xhinge


Coarse 101x21 0.9496 1.0491
Medium 201x41 0.8939 1.1059
Fine 401x81 0.8579 1.1428

Table 3: Location of separation and reattachment points

11
noticed. First, the data on the conical part seem to have an o set; second, the size of the
recirculation zone (which is not grid converged) is not predicted correctly. It is right to wonder
to what extend the assumptions made for these computations, namely that the incoming Mach
number is constant, uniform and equal to 6 are valid. In the following section, the ow in the
H3 wind tunnel is investigated.

12
5 Wind tunnel nozzle ow computations
5.1 VKI H3 nozzle design
There exists two types of wind-tunnel nozzles, on the one hand contoured nozzles which have
been designed to produce a good ow quality (in terms of uniformity) at the design conditions.
On the other hand conical nozzles which are more straightforward to design and which yield
conical ows accelerating all along the nozzle and in the test section.
The VKI H3 hypersonic wind tunnel nozzle is contoured. It has been designed in 1970 by
the ONERA using the method of characteristics (like most other contoured nozzles) [13]. Two
problems immediately come to mind, to what extent the method of characteristics produces an
accurate ow eld prediction, and what happens if the tunnel is not run at design conditions.
The method of characteristic enables to do a rst approximation of the inviscid ow eld inside
the nozzle but breaks down because of the viscous part and thus it is coupled with a turbulent
boundary layer calculation. The VKI H3 wind tunnel has been designed for a total pressure of
P0 =25 atm and a total temperature of T0=623 K. This is another important point that could
largely a ect the quality of the ow and the magnitude of the incoming Mach number of the
model because the tunnel is run o -design.
The nature of the ow eld at the exit of the nozzle has been investigated [4] . It is found
that the ow is non uniform and that the outlet Mach number varies from its predicted value.
To get an insight into the ow eld generated in the VKI H3 wind tunnel, it is decided to
use the VKI Multiblock code to make computations as it has already been applied in the VKI
Longshot hypersonic nozzle [14], [15].
The nozzle contour is given by a set of points obtained from the engineering drawing and
therefore higher order interpolation techniques have been used to model the contour. Table 8
gives the set of points. The geometry of the domain is presented in gure 23. The nozzle is
followed by the test chamber which has, as seen later, an original treatment of the boundary
conditions.
The grids generated for the nozzle are created by a quasi-hyperbolic mesh generator. The
surface node distribution is non linear to allow clustering in the high gradient area (throat) and
for the same reasons, variable cell heights are used to satisfy y+ requirements (a value less than
one indicates that the thermal boundary layer is correctly captured). An iterative procedure
is employed to shoot the outer boundary onto the symmetry axis y=0. Finally, the j lines are
such that they are normal to the centerline. An example of the grids created is shown in gure
24.
5.2 Computation of the ow eld in the H3 wind tunnel
5.2.1 Laminar nozzle wall boundary layer
The whole domain is cut in two subdomains, the lower domain represents the VKI H3 nozzle
extended into the test section and the upper domain represents the test chamber. A grid
sequencing technique is employed to accelerate convergence rates and three grid levels are
used. Because of the length of the nozzle many points are necessary to accurately describe the
domain. The ne grid for the nozzle is discretized into 106941 points (829 in the i direction
and 129 in the j direction) for the lower domain and 11481 points (89 by 129) for the upper
domain. Table 4 gives the grid sizes for the two domains.
The treatment of the boundary conditions for the outlet is enabled by the nature of the
numerical scheme of the code. As mentioned by Barth [16], upwind codes allow to overspecify
information on the boundaries. Figure 25 shows the domain and the boundary conditions
needed. The same boundary conditions are imposed on most of the domain: a supersonic
inlet. The Mach number is 0 (it is only needed for supersonic inlet which is not present), so

13
Grid H3 nozzle and Outlet Chamber
level test section
Coarse 208x33 23x33
Medium 415x65 45x65
Fine 829x129 89x129

Table 4: Grid size for the H3 laminar computation

the ow angle is unspeci ed and may adopt any value. The pressure imposed is the static
pressure which has been recorded during wind tunnel tests and the temperature is the wall
temperature which is taken as being 300 K. For the supersonic outlet, no information is needed.
The subsonic outlet only requires the static pressure (given), and the subsonic inlet requires
the pressure, the temperature and the ow angle (not given since M=0).
The rst case run is for the fully laminar nozzle wall boundary layer. The coarse grid is
run rst for an Euler case. It is important to note that unless a reasonable initialisation for the
nozzle ow is made (e.g. 1D relations), convergence can not be attained. Then a rst order
Navier-Stokes computation is carried out, followed by a second order accurate computation.
The solution is then interpolated onto a ner grid where a second order Navier-Stokes compu-
tation is done. Another observation made during the computation is that the shear layer in
the jet ow tends to get very unphysical if the convergence level is not sucient.
For the wind tunnel computation, the four sets of data analysed are: the wall pressure pw
and the wall heat ux qw in the nozzle, the Mach number and the Pitot pressure in a cross
section 10 mm behind the nozzle exit in the jet ow. The Mach number is also presented on
the centerline as a function of the distance to the nozzle exit. The Pitot pressure (which is
actually the pressure measured by a Pitot probe behind a shock) writes:
! ; 1   ;1 1
Pt2 = ( + 1)M 2 +1 (34)
P0 ( ; 1)M 2 + 2 2 M 2 ; ( ; 1)
Experiments have been carried out for the low Reynolds case condition in the H3 wind
tunnel to assess the ow eld [4]. Pressure taps mounted on a at plate in the test section are
used to obtain the distribution of the Mach number.
Experimental data on the nozzle wall are not available, thin lms could be located up to
the throat because the temperature levels are within acceptable range. The experimental heat
ux data would give a precise insight into the nature of the boundary layer, one could hope to
get information on the transition area.
The values of Pw and qw are plotted in gures 26 and 27 for the three grid levels. The wall
pressure normalised by the reservoir pressure shows a very smooth variation. In the reservoir
section the ratio is 1, the ow is stagnant, and at the nozzle exit (x=0.758m) the ratio is 5.10;4
which means that the wall pressure is 532 Pa. It is seen that the variation observed on the wall
heat ux between the medium and the ne grid indicate that the solution is grid converged.
The wall heat ux curve is characterised by a kink right after the throat, it corresponds to the
location at which non-uniformities are generated in the ow.
In the jet ow, gure 28 for the Pt2 for the three grid levels also indicates grid convergence.
From this graph, the overall Pt2 looks uniform in the potential core region. The boundary
layer thickness is rather thin: xedge =0.071 m. The Mach number distribution in the jet ow at
10 mm behind the ow exit ( gure 29) can be compared to the experimental values. It has an
overall constant value of 6.1 but a more detailed view of the Mach number in the core region
14
shows variations as seen in gure 30. The Mach number is also plotted on the centerline as a
function of the distance from the nozzle exit ( gure 31). The Mach number smoothly increases
from 0 to 6.3 at 0.4m behind the throat. It then decreases a bit and levels out to a constant
value of 6.1. A close look at the Mach number on the centerline in the jet ow is presented in
gure 32 and it indicates that the ow accelerates slightly in the test section from Mach 6.1 to
6.15.
Comparisons with experimental data indicates that the numerical Mach number in the test
section overpredicts the experimental data. This is consistent with the fact that the nature
of the nozzle wall boundary layer is most certainly turbulent from some location behind the
throat. The laminar boundary layer is thin and therefore the ow expands more; this means
that as observed the Mach number increases. The ow non uniformities both in the nozzle and
in the test section will be emphasised in the next section.
The simulation of the ow in the H3 wind tunnel nozzle and in the test section assuming a
laminar nozzle wall boundary layer has shown that the ow is not uniform in the test section.
There is a streamwise and a cross-wise variation of the Mach number, however these variations
are acceptable. Comparison with experimental data tends to show that the boundary layer is
actually turbulent and therefore, this calls for further simulations.
5.2.2 Turbulent nozzle wall boundary layer
Modeling turbulent ow in hypersonic still represents a challenge in CFD. For the Mach 6
hypersonic wind tunnel, the transport phenomena associated with the ow are not negligible
and therefore it is not correct to apply a simple algebraic model such as the Baldwin-Lomax
model. The VKI-MB code has been extended [12] to deal with turbulent ows using the one
equation model of Spalart and Allmaras [11]. During the turbulent computation, the eddy
viscosity is increased at the trip point until it reaches a peak after which the eddy viscosity
is convected and di used in the rest of the ow. This process is very slow and even with a
subiteration technique (i.e. for one iteration on the conservative variables, several iterations on
the turbulent variables are carried out), it is time consuming. Because it is an algebraic model,
the Baldwin-Lomax immediately gives the eddy viscosity for each grid point and therefore it
is used as a starter to set the initial condition for the one equation model. Equation 25 which
gives the eddy viscosity as a function of the turbulent variable: ~ can be inverted and therefore
a proper initialisation for the ~ variable can be set. This technique brings a large gain in time
and also accelerates the convergence rate on the turbulent variable and therefore on the overall
convergence. Initialising the turbulent eld is just carried out for the coarse grid and then the
results are interpolated for the ner grid levels.
One of the major problems in turbulent ows is to set the transition location. Since there
are no experimental data on the nozzle wall, it is not possible to get the transition location
and therefore guesses have to be made. Moreover, data near the nozzle exit in the boundary
layer are not available so the boundary layer thickness is unknown. Three computations have
been carried out for the transition onset location (xt ):
(a) xt =-0.01m (i.e. at the throat)
(b) xt =0.02m
(c) xt =0.04m
New computational domains satisfying the constraints imposed by the turbulent quantities
have been de ned. The height of the rst cell and the stretching factor have been adapted for
the turbulent computation. The grid sizes are given in table 5.
It is by comparing the numerical results for coarse grids to the experimental data that it
has been decided that case (c) was the most likely transition location. Then the computation
for the fully turbulent case (transition set at the throat) and for the transition starting at
xt =0.04m are computed for the three grid levels.
15
Grid H3 nozzle and Outlet Chamber
level test section
Coarse 231x33 23x33
Medium 461x65 45x65
Fine 921x129 89x129

Table 5: Grid size for the H3 turbulent computation

The e ect of turbulence and transition location on the ow eld is seen in gure 33 for the
detailed variation of the Mach number on the centerline in the nozzle and the jet ow area.
Figure 34 for a cross cut of the Mach number in the jet ow at 10mm behind the nozzle exit
also shows the di erence between the laminar and the turbulent computations and the e ect
of transition positioning. It is seen that for the turbulent ow, the Mach number in the jet is
lower than for a fully laminar nozzle wall boundary layer. The boundary layer is thicker for
turbulent ows but it decreases when the transition location is moved downstream.
The results comparing the wall data for the laminar and the turbulent computations (case
c) are shown in gure 35 for the wall pressure and gure 36 for the wall heat ux. It is seen
that the wall heat uxes are largely a ected by the change in the nature of the boundary layer
and the transition location produces a local peak heating to reach the turbulent heat ux level.
The Mach number for the new turbulent case on the centerline in the test section is presented
in gure 37 together with the experimental data. The Pitot pressure which has been measured
at di erent cross-wise locations in the jet ow is compared to the numerical results for the nest
grid with the transition onset set at xt =0.04m. As seen in gure 38 the ow eld in the core
region is well predicted by the computation for all stations in the jet ow. Therefore we are
con dent that the ow eld in the test section can be rebuilt numerically. It is interesting to
note that the experimental Pt2 data indicate that ow closure occurs for the two last stations
(x=0.858m and x=0.898m). The numerical data indicate that the ow actually expands at
x=0.898m. This means that the pressure imposed in the outlet test chamber in the computation
is lower than actual pressure. It can be explained by the fact that the ow over Pitot probe
rake is de ected into the test chamber and thus builds up the chamber pressure.
Finally, the ow eld non uniformities are discussed. First, downstream of the throat region
the boundary layer growth is very rapid and the transition onset acts as a wedge. As shown in
gure 39 a compression wave is generated at the wall at x=0.04m, it is then re ected onto the
symmetry axis and it is transported into the nozzle ow. Figure 40 shows that the ow is not
quite uniform in the nozzle especially near the nozzle exit. Finally, the ow eld in the test
section is shown in gure 41 for which the scale accentuates the non uniformities.
An accurate description of the ow eld in the test section is now known. It can be used to
investigate the ow over the blunted cone are model taking into account the non-uniformities.
5.3 Rebuilding of the blunted cone- are in the VKI H3 wind tunnel
The blunted cone- are computational grid can be superimposed on the H3 wind tunnel mul-
tidomain grids, gure 42 shows the boundary limit of the three domains. The outer boundary
of the blunted cone- are grid lies in a region of supersonic ow in the test section of the H3
tunnel. It is therefore possible to decouple the two computations and the inlet boundary con-
ditions for the blunted cone- are can be obtained by doing an interpolation of the solution
obtained from the computation of the H3 tunnel. Figure 43 shows the Mach number on the
boundary of the cone- are as a function of the distance from the nozzle exit, and gure 44
shows the angle of the velocity vector on the boundary of the computational domain around
16
the cone- are. From these gures, it is seen that on the most part of the blunted cone- are the
incoming ow seen by the boundary is rather uniform: Mach 5.8 and 0.2o inclination upward.
The end of the boundary is more disturbed and this is due to the jet closure which was shown
in gure 41.
The computations on three grid levels are carried out for the blunted cone- are with the
non-uniform boundary conditions. It should be noted that the inlet boundary condition on the
coarser grids are obtained from the data on the boundary condition of the nest grid. In other
words, the nest grid for the wind tunnel is always used for de ning the boundary conditions
on the blunted cone- are.
To illustrate the e ects of the non-uniform boundary condition of the ow, gure 45 shows
a detailed view of the Mach number isolines between the bow shock and the outer boundary
condition.
From the lower incoming Mach number, it is expected to see the heat ux decreasing and
the pressure increasing. It is not obvious to de ne reference conditions for the computation
with the non-uniform incoming ow, that is why it is decided to use dimensional values: Pw
(in Pa) and qw (in W/m2). The results for the uniform and non-uniform ow computations
are plotted against the experimental results in gure 46 for the wall pressure and in gure 47
for the wall heat ux. It can be seen that indeed the new computations are showing a very
good agreement with the experimental data for both the wall pressure and the wall heat ux.
The separated region is still not resolved correctly, it is not grid converged and it is known
that a more re ned grid in the direction normal to the wall is needed to be able to show grid
convergence.
The ow non uniformities seen over the model both for the ow inclination and the Mach
number magnitude is actually not in uencing the wall data in a large extend. The streamlines
path indicates that the ow over the blunted cone are is coming from a narrow band on the
centerline just ahead of the bow shock, therefore the driving factor in the numerical simulation
is the magnitude of the Mach number. It is found from the numerical simulation of the H3
wind tunnel running at the low Reynolds conditions that the incoming Mach number at a point
on the centerline just ahead of the bow shock is M=5.90. The good agreements observed in
the wall pressure and the wall heat ux indicate that the reference Mach number used for the
pressure coecients and the Stanton numbers should not be 6 but 5.90.

17
6 Conclusions
The experimental study on the blunted cone- are has been successfully made and the produced
results have been shown to be repeatable and reliable for both the pressure coecient and the
heat transfer. The e ects of ow misalignment has been shown to be important especially on
heat transfer. A set of tabulated data are given so that further comparisons can be carried out
by other researchers.
The numerical study assuming a uniform incoming Mach number of 6 has shown that
although good results were obtained outside of the separation bubble, grid convergence was
not reached in this area. The grid which had already been clustered in the recirculation area
is not ne enough and more e orts should be made to obtain grid converged results. An o set
of the wall data between the numerical and the experimental results can be observed on both
the conical and are section.
The simulation of the ow in the wind tunnel nozzle and in the test chamber has demon-
strated that the nozzle wall boundary layer of the H3 wind tunnel starts to transition at .05m
behind the nozzle throat. A good agreement between the numerical and the experimental data
for the ow in the test section shows that the ow in the H3 wind tunnel can be accurately
rebuilt numerically. Some minor ow non uniformities are present in the test section but more
important is the fact that the Mach number is substantially smaller than the design Mach
number for 10 bars.
A study of the ow over the blunted cone- are taking into account the ow properties
found from the numerical simulation of the wind tunnel has shown a very accurate prediction
of the wall data on the blunted cone- are. Although the length of the separated region did not
improve, the new results are very satisfying.

18
References
[1] Zemsch, S.: 'Development of an Electronic Pressure Scanner System' von Karman Institute
TM, January 1995
[2] Boerrigter, H.L. : 'Calculating the frequency response and step response of a pressure
measurement system' von Karman Institute, VKI Technical Memorandum 51, January
1996
[3] Dieudonne, W. : 'Hypersonic Flow over Axisymmetric Bodies'. von Karman Institute
Project Report 1995-18, June 1994
[4] Boerrigter H.L. : 'Calibration of the H3 wind tunnel using Pitot probes'. VKI Internal
Note 94, December 1993
[5] Mensink, C. : 'A 2-D Parallel MultiBlock Method for Viscous and Inviscid Compressible
Flow'. PhD Thesis VKI/U. Twente, December 1992
[6] Broglia, R.; Manna, M.; Deconinck, H.; Degrez, G. : 'Development and Validation of an
Axisymmetric Navier-Stokes Solver for Hypersonic Flows' von Karman Institute Technical
Note 188, May 1995
[7] Ho mann, K.A. : 'Computational Fluid Dynamics for Engineers' Engineering Education
System, 1989
[8] Coquel, F.; Liou, M. : 'Field by Field Hybrid Upwind Splitting Methods'. AIAA, Vol.
93-3302-CP, 1993
[9] Banks, D.W. : 'Hybrid Structured / Unstructured grid generation, an oriented approach'
von Karman Institute, Project Report 1995-11, June 1995
[10] Wilcox, D.C.: 'Turbulence Modeling for CFD'.
[11] Spalart, P.R.; Allmaras S.R.: 'A one-equation turbulence model for aerodynamic ows'.
La Recherche Aerospatiale, 1994 no 1, 5-21.
[12] Paciorri, R.: 'Implementation and validation of the Spalart-Allmaras turbulence model
for application in hypersonic ows', Technical Note HT-TN-E34-531-VKIN, von Karman
Institute, February 1996.
[13] Gautier, B.; Carriere, P.: 'Communication between ONERA and VKI'. 1970.
[14] Dieudonne, W. ; Charbonnier, J.M. ; Deconinck, H. : 'Nozzle Flow Sensitivity Analysis
and Wind Tunnel Operating Envelope; Part I: Laminar Case.', Technical Note HT-TN-
E34-511-VKIN, von Karman Institute, March 1996.
[15] Dieudonne, W.; Charbonnier, J.M.; Deconinck, H.: 'Nozzle Flow Sensitivity Analysis and
Wind Tunnel Operating Envelope; Part II: Turbulent Case', Technical Note HT-TN-E34-
541-VKIN, von Karman Institute.
[16] Barth, T.J. : 'Parallel CFD algorithms on unstructured meshes', AGARD Report R-807,
Oct. 1995

19
Figure 1: Von Karman Institute H3 hypersonic wind tunnel

142.42

0.8
69
50
Ø10.53 ±25 5
40
20
R=3.5 Ø6 h6 Ø10.53 10
M5
M50 x 1

7 ±Ø59
9
40 7.5°
16.69 18 Ø75
34.91 25
Ø6 H7 M5

17.5º
Pressure taps
at 90, 80, 270º 15

73.42

Figure 2: Pressure model with the pressure tappings location

20
10

Uncertainty [%]

−5

−10
0 50 100 150 200
Pressure [mm H2O]

Figure 3: Uncertainty associated with the slopes for the E.P.S.

First tap at low Re


First tap at medium Re
First tap at high Re
Last tap at low Re
Last tap at medium/high Re

100
% of transmitted signal

50

0
0.00 1.00 2.00 3.00
Time [s]

Figure 4: Time response of the pressure scanner system

21
Figure 5: Mach number isolines obtained from [10], from top to bottom: 10, 100 and 140 mm
behind the nozzle exit

22
4.0
o o
yaw=+0.10 , R=0
o o
yaw=+0.10 , R=180
o o
yaw=+0.10 , R=90
o o
yaw=+0.10 , R=270

3.0

Pwall/Pfirst tap

2.0

1.0

0.0
40 60 80 100 120 140
Distance from the nose [mm]

Figure 6: Pressure at low Re for the four rolls

0.0025
o o
pitch=0. , Y=0.
o o
pitch=−0.25 , yaw=+0.25
o o
pitch=−0.25 , yaw=0.
o o
pitch=−0.25 , yaw=−0.25

0.0020

0.0015
St

0.0010

0.0005

0.0000
0 50 100 150
Distance from the nose [mm]

Figure 7: E ects of Yaw and Pitch on separation and reattachement

23
Figure 8: Temperature isolines on the cone- are model showing ow axisymmetry

Figure 9: Temperature isolines on the cone- are model showing ow misalignment

24
Figure 10: Stanton number sensitivity to yaw (constant pitch)

500
10 bars − run 1
10 bars − run 2

400
Pressure [mm H2O]

300

200

100

0
40 60 80 100 120 140
X−location [mm]

Figure 11: Repeatability of consecutive runs

25
500
10 bars − day 1
10 bars − day 2

400

Pressure [mm H2O]


300

200

100

0
40 60 80 100 120 140
X−location [mm]

Figure 12: Repeatability on 2 days with model remounted

0.0030
St
−2
Cp x 10

0.0025

0.0020
St, Cp

0.0015

0.0010

0.0005

0.0000
0.00 0.05 0.10
distance from the nose [m]

Figure 13: Pressure Coecient (x10;2) and Stanton number

26
x [m] P [Pa]
0.046032 1209
0.050989 1224.8
0.055946 1220.63
0.060904 1227.69
0.065861 1300.23
0.070818 1613.16
0.075775 1792.74
0.080733 1870.54
0.08569 1901.75
0.090459 2014.23
0.095228 2493.23
0.099996 3091.86
0.104765 3531.4
0.109534 3893.13
0.114303 4145.83
0.119072 4283.87
0.123841 4306.72
0.128609 4358.11
0.133378 4427.27

Table 6: Wall pressure data P [Pa] as a function of x [m] on the blunted cone- are

27
x q x q x q x q
0.1423 12328.2 0.1416 12743.7 0.1409 12587.9 0.1401 12336.8
0.1387 12778.4 0.1380 12683.1 0.1372 12172.3 0.1365 12258.9
0.1351 12700.4 0.1343 12397.4 0.1336 12856.3 0.1329 12856.3
0.1314 13237.2 0.1307 13462.3 0.1300 13237.2 0.1293 13081.4
0.1278 13237.2 0.1271 13522.9 0.1263 13522.9 0.1256 13436.3
0.1242 13756.6 0.1234 13774.0 0.1227 13990.4 0.1220 13748.0
0.1205 13730.7 0.1198 14016.4 0.1191 14077.0 0.1184 14068.3
0.1169 14033.7 0.1162 14102.9 0.1155 14293.4 0.1147 14137.6
0.1133 14423.3 0.1125 14206.8 0.1118 14198.2 0.1111 14258.8
0.1096 14102.9 0.1089 14128.9 0.1082 14094.3 0.1075 14154.9
0.1060 13644.1 0.1053 13609.5 0.1046 13479.6 0.1038 13358.4
0.1024 12873.6 0.1017 12657.2 0.1009 12518.6 0.1002 12025.2
0.0988 11670.2 0.0980 11116.1 0.0973 10665.9 0.0966 10189.8
0.0951 9817.5 0.0944 9220.2 0.0937 8683.4 0.0929 8248.8
0.0915 7318.1 0.0908 6777.9 0.0900 6244.6 0.0893 5794.4
0.0879 4747.7 0.0871 4126.1 0.0864 3777.2 0.0857 3088.1
0.0842 2486.4 0.0835 2150.5 0.0828 2071.7 0.0820 1994.7
0.0806 2045.8 0.0799 2070.9 0.0791 2070.0 0.0784 2225.0
0.0770 2381.7 0.0762 2407.6 0.0755 2641.4 0.0748 2719.3
0.0733 3192.0 0.0726 3378.1 0.0719 3804.1 0.0712 4257.7
0.0697 5042.9 0.0690 5537.3 0.0683 5680.1 0.0675 6090.5
0.0661 6513.8 0.0653 6513.8 0.0646 6487.9 0.0639 6626.4
0.0624 6736.3 0.0617 6682.7 0.0610 6794.4 0.0603 6822.1
0.0588 6767.5 0.0581 6794.4 0.0574 6764.9 0.0566 6908.6
0.0552 6991.7 0.0545 6851.5 0.0537 6849.8 0.0530 7075.7
0.0515 7187.4 0.0508 7047.1 0.0501 7132.0 0.0494 6988.3
0.0479 7245.4 0.0472 7388.2 0.0465 7190.9 0.0457 7159.7
0.0443 7191.7 0.0436 7279.2 0.0428 7417.7 0.0421 7495.6
0.0407 7527.6 0.0399 7335.4 0.0392 7507.7 0.0385 7503.4
0.0370 7419.4 0.0363 7783.0 0.0356 7985.6 0.0348 7843.6
0.0334 7868.7 0.0327 8040.1 0.0319 8182.1 0.0312 8040.1
0.0298 8303.3 0.0290 8271.3 0.0283 8249.7 0.0276 8300.7
0.0261 8560.5 0.0254 8588.2 0.0247 8761.3 0.0240 8735.3
0.0225 8674.7 0.0218 8960.4 0.0210 8943.1 0.0203 9402.0
0.0189 9185.5 0.0181 9505.9 0.0174 9566.5 0.0167 9280.8

Table 7: Wall heat ux data q [W/m2] as a function of x [m] on the blunted cone- are

28
Linear Contour Plot.
Stat. Temperature lines Step = 10.0000

0.050

0.025

0.000
0.000 0.050 0.100 0.150
x

Figure 14: Pressure isolines

Linear Contour Plot.


Mach number lines Step = 0.2500

0.050

0.025

0.000
0.000 0.050 0.100 0.150
x

Figure 15: Mach number isolines

29
Vector Plot.

0.01550

0.01500

0.01450

0.01400

0.08450 0.08500 0.08550 0.08600 0.08650


x

Figure 16: Velocity vectors in the recirculation bubble

Grid Plot.

0.050

0.025

0.000
0.000 0.050 0.100 0.150
x

Figure 17: Mesh for the medium level (201x41) with clustering

30
Coarse 101x21
Medium 201x41
Fine 401x81

0.25

0.20
Cp

0.15

0.10

0.05

0.00
0.00 0.05 0.10
distance from the nose [m]

Figure 18: Pressure coecient on the blunted cone

2.00e−03
Coarse 101x21
Medium 201x41
Fine 401x81

1.50e−03
St

1.00e−03

5.00e−04

0.00e+00
0.00 0.05 0.10
Distance from the nose [m]

Figure 19: Stanton number on the blunted cone

31
0.006
Coarse 10x21
Medium 201x41
Fine 401x81

0.004
Cf

0.002

0.000

−0.002
0.00 0.05 0.10
Distance from the nose [m]

Figure 20: Skin friction coecient on the blunted cone

Experiment
Fine grid (401x81)

0.20
Cp

0.10

0.00
0.00 0.05 0.10 0.15
Distance from the nose [m]

Figure 21: Comparison of experimentally and numerically determined pressure coecient

32
0.0020
Experiment
Fine grid (401x81)

0.0015
St

0.0010

0.0005

0.0000
0.00 0.05 0.10
Distance from the nose [m]

Figure 22: Comparison of experimentally and numerically determined Stanton number

33
x y x y x y x y x y
-200.000 75.000 -187.274 75.000 -180.301 74.696 -173.382 73.785 -166.568 72.274
-159.912 70.175 -153.464 67.505 -147.274 64.282 -141.387 60.532 -136.000 56.409
-134.000 54.722 -132.000 53.081 -130.000 51.484 -128.000 49.929 -126.000 48.415
-124.000 46.941 -122.000 45.506 -120.000 44.108 -118.000 42.748 -116.000 41.424
-114.000 40.134 -112.000 38.879 -110.000 37.657 -108.000 36.468 -106.000 35.311
-104.000 34.185 -102.000 33.090 -100.000 32.026 -98.000 30.990 -96.000 29.984
-94.000 29.007 -92.000 28.058 -90.000 27.136 -88.000 26.241 -86.000 25.374
-84.000 24.533 -82.000 23.717 -80.000 22.928 -78.000 22.164 -76.000 21.425
-74.000 20.711 -72.000 20.021 -70.000 19.355 -68.000 18.714 -66.000 18.096
-64.000 17.501 -62.000 16.930 -60.000 16.382 -58.000 15.856 -56.000 15.354
-54.000 14.873 -52.000 14.415 -50.000 13.979 -48.000 13.565 -46.000 13.173
-44.000 12.802 -42.000 12.453 -40.000 12.126 -38.000 11.820 -36.000 11.535
-34.000 11.271 -32.000 11.028 -30.000 10.807 -28.000 10.606 -26.000 10.426
-24.000 10.266 -22.000 10.128 -20.000 10.010 -18.000 9.913 -16.000 9.836
-14.000 9.780 -12.000 9.745 -10.000 9.730 -9.554 9.730 -8.000 9.736
-6.000 9.762 -4.000 9.809 -2.000 9.876 0.000 9.964 2.000 10.073
4.000 10.203 6.000 10.354 8.000 10.526 10.000 10.719 12.000 10.933
14.000 11.168 16.000 11.424 18.000 11.702 20.000 12.001 22.000 12.321
24.000 12.662 26.000 13.024 28.000 13.406 30.000 13.807 32.000 14.225
34.000 14.658 36.000 15.104 38.000 15.561 40.000 16.027 42.000 16.500
44.000 16.978 46.000 17.459 48.000 17.941 50.000 18.423 52.000 18.905
54.000 19.387 56.000 19.868 58.000 20.348 60.000 20.827 62.000 21.305
64.000 21.782 66.000 22.258 68.000 22.733 70.000 23.207 72.000 23.679
74.000 24.149 76.000 24.617 78.000 25.083 80.000 25.547 82.000 26.008
84.000 26.467 86.000 26.924 88.000 27.378 90.000 27.830 92.000 28.280
94.000 28.727 96.000 29.171 98.000 29.613 100.000 30.052 102.000 30.489
104.000 30.923 106.000 31.354 108.000 31.783 110.000 32.209 112.000 32.632
114.000 33.052 116.000 33.469 118.000 33.883 120.000 34.294 122.000 34.702
124.000 35.107 126.000 35.509 128.000 35.908 130.000 36.304 132.000 36.696
134.000 37.085 136.000 37.471 138.000 37.854 140.000 38.234 142.000 38.611
144.000 38.985 146.000 39.355 148.000 39.722 150.000 40.086 152.000 40.446
154.000 40.803 156.000 41.157 158.000 41.507 160.000 41.854 162.000 42.198
164.000 42.538 166.000 42.875 168.000 43.209 170.000 43.539 172.000 43.866
174.000 44.190 176.000 44.510 178.000 44.827 180.000 45.141 182.000 45.452
184.000 45.759 186.000 46.063 188.000 46.364 190.000 46.662 192.000 46.957
194.000 47.249 196.000 47.538 198.000 47.824 200.000 48.107 202.000 48.388
204.000 48.666 206.000 48.942 208.000 49.215 210.000 49.486 212.000 49.754
214.000 50.020 216.000 50.283 218.000 50.544 220.000 50.802 222.000 51.058
224.000 51.311 226.000 51.562 228.000 51.811 230.000 52.057 232.000 52.301
234.000 52.542 236.000 52.783 238.000 53.020 240.000 53.255 242.000 53.488
244.000 53.719 246.000 53.948 248.000 54.175 250.000 54.400 252.000 54.623
254.000 54.844 256.000 55.063 258.000 55.280 260.000 55.495 262.000 55.708
264.000 55.919 266.000 56.129 268.000 56.337 270.000 56.543 272.000 56.747
274.000 56.949 276.000 57.149 278.000 57.348 280.000 57.545 282.000 57.741
284.000 57.935 286.000 58.127 288.000 58.318 290.000 58.507 292.000 58.695
294.000 58.881 296.000 59.066 298.000 59.249 300.000 59.431 302.000 59.611
304.000 59.789 306.000 59.966 308.000 60.141 310.000 60.315 312.000 60.488

34
314.000 60.659 316.000 60.829 318.000 60.997 320.000 61.164 322.000 61.330
324.000 61.494 326.000 61.657 328.000 61.819 330.000 61.980 332.000 62.139
334.000 62.297 336.000 62.454 338.000 62.609 340.000 62.763 342.000 62.916
344.000 63.068 346.000 63.219 348.000 63.368 350.000 63.516 352.000 63.663
354.000 63.809 356.000 63.954 358.000 64.098 360.000 64.241 362.000 64.383
364.000 64.524 366.000 64.664 368.000 64.802 370.000 64.939 372.000 65.075
374.000 65.210 376.000 65.344 378.000 65.477 380.000 65.609 382.000 65.740
384.000 65.870 386.000 65.999 388.000 66.125 390.000 66.252 392.000 66.377
394.000 66.501 396.000 66.624 398.000 66.746 400.000 66.868 402.000 66.989
404.000 67.109 406.000 67.228 408.000 67.346 410.000 67.463 412.000 67.579
414.000 67.694 416.000 67.808 418.000 67.921 420.000 68.033 422.000 68.144
424.000 68.254 426.000 68.363 428.000 68.471 430.000 68.578 432.000 68.684
434.000 68.789 436.000 68.893 438.000 68.997 440.000 69.100 442.000 69.202
444.000 69.303 446.000 69.404 448.000 69.504 450.000 69.603 452.000 69.701
454.000 69.799 456.000 69.896 458.000 69.992 460.000 70.087 462.000 70.182
464.000 70.276 466.000 70.369 468.000 70.461 470.000 70.552 472.000 70.642
474.000 70.732 476.000 70.821 478.000 70.909 480.000 70.997 482.000 71.084
484.000 71.170 486.000 71.255 488.000 71.339 490.000 71.422 492.000 71.505
494.000 71.587 496.000 71.668 498.000 71.748 500.000 71.827 502.000 71.905
504.000 71.983 506.000 72.060 508.000 72.136 510.000 72.211 512.000 72.285
514.000 72.358 516.000 72.431 518.000 72.503 520.000 72.574 522.000 72.645
524.000 72.715 526.000 72.784 528.000 72.852 530.000 72.920 532.000 72.987
534.000 73.053 536.000 73.119 538.000 73.184 540.000 73.249 542.000 73.313
544.000 73.376 546.000 73.439 548.000 73.501 550.000 73.562 552.000 73.622
554.000 73.682 556.000 73.741 558.000 73.799 560.000 73.857 562.000 73.914
564.000 73.971 566.000 74.027 568.000 74.083 570.000 74.138 572.000 74.193
574.000 74.247 576.000 74.300 578.000 74.352 580.000 74.404 582.000 74.455
584.000 74.506 586.000 74.556 588.000 74.606 590.000 74.655 592.000 74.704
594.000 74.752 596.000 74.800 598.000 74.847 600.000 74.893 602.000 74.939
604.000 74.984 606.000 75.029 608.000 75.073 610.000 75.117 612.000 75.160
614.000 75.203 616.000 75.245 618.000 75.287 620.000 75.329 622.000 75.370
624.000 75.411 626.000 75.452 628.000 75.492 630.000 75.532 632.000 75.571
634.000 75.610 636.000 75.649 638.000 75.687 640.000 75.725 642.000 75.762
644.000 75.790 646.000 75.835 648.000 75.871 650.000 75.907 652.000 75.942
654.000 75.977 656.000 76.011 658.000 76.045 660.000 76.078 662.000 76.111
664.000 76.143 666.000 76.175 668.000 76.207 670.000 76.238 672.000 76.269
674.000 76.299 676.000 76.329 678.000 76.358 680.000 76.387 682.000 76.416
684.000 76.444 686.000 76.472 688.000 76.500 690.000 76.527 692.000 76.554
694.000 76.581 696.000 76.607 698.000 76.633 700.000 76.659 702.000 76.684
704.000 76.709 706.000 76.734 708.000 76.759 710.000 76.783 712.000 76.807
714.000 76.831 716.000 76.855 718.000 76.878 720.000 76.901 722.000 76.924
724.000 76.947 726.000 76.969 728.000 76.991 730.000 77.013 732.000 77.035
734.000 77.057 736.000 77.078 738.000 77.099 740.000 77.120 742.000 77.141
744.000 77.162 746.000 77.183 748.000 77.204 750.000 77.225 752.000 77.246
754.000 77.267 756.000 77.288 758.000 77.309 0.000 0.000 0.000 0.000

Table 8: Coordinates of the H3 nozzle contour (all data in mm)

35
Figure 23: H3 nozzle geometry

Figure 24: Example of a coarse grid for computing the nozzle ow

Subsonic inlet

Subsonic
inlet
Solid
wall
Subsonic
outlet

Solid Multi
wall
domain
Supersonic Supersonic
inlet outlet
Axisymmetry

Figure 25: Boundary conditions for the outlet

36
Fine grid − 829x129
1e+00 Medium grid − 415x65
Coarse grid − 208x33

1e−01
Pw/P0

1e−02

1e−03

1e−04
−0.5 0.0 0.5 1.0
X [m] distance from throat

Figure 26: Wall pressure along the nozzle for the laminar case

1e+06

H3 − Laminar

1e+05
qw [W/m2]

1e+04

1e+03

1e+02
−0.5 0.0 0.5 1.0
X [m] distance from throat

Figure 27: Wall heat ux along the nozzle for the laminar case

37
1e−01
H3 − Laminar − Coarse (231x33)
H3 − Laminar − Medium (461x65)
H3 − Laminar − Fine (921x129)

Pt2/P0 1e−02

1e−03

1e−04
0.00 0.02 0.04 0.06 0.08
Y [m] distance from center line

Figure 28: Pitot pressure at 10mm behind the nozzle exit - laminar case

8
H3 − Laminar − Coarse (231x33)
H3 − Laminar − Medium (461x65)
H3 − Laminar − Fine (921x129)

6
Mach

0
0.00 0.02 0.04 0.06 0.08
Y [m] distance from center line

Figure 29: Mach number at 10mm behind the nozzle exit - laminar case

38
6.4
H3 − Laminar − Coarse (231x33)
H3 − Laminar − Medium (461x65)
H3 − Laminar − Fine (921x129)

6.2
Mach

6.0

5.8
0.00 0.02 0.04 0.06 0.08
Y [m] distance from center line

Figure 30: Mach number detail at 10mm behind the nozzle exit - laminar case

8.0
H3 − Laminar − Coarse (231x65)
H3 − Laminar − Medium (461x65 )
H3 − Laminar − Fine (921x129)

6.0
Mach

4.0

2.0

0.0
−0.5 0.0 0.5 1.0
X [m] distance from throat

Figure 31: Mach number along the centerline - laminar case

39
6.4
H3 - Laminar - Coarse (231x65)
H3 - Laminar - Medium (461x129)
H3 - Laminar - Fine (921x129)
Experiment

6.2

Mach

6.0

5.8
0.80 0.85 0.90 0.95
X [m] distance from throat

Figure 32: Detail in the jet ow for the Mach number along the centerline - laminar case

6.4
H3 − Laminar
H3 − Turbulent − xt=throat
H3 − Turbulent − xt=0.04m

6.2
Mach number

6.0

5.8

5.6
0.25 0.50 0.75 1.00
X [m] distance from the throat

Figure 33: E ects of turbulence and transition location on centerline Mach number

40
6.20
H3 − Laminar
H3 − Turbulent − xt=throat
H3 − Turbulent − xt=0.04m

6.10

Mach number

6.00

5.90

5.80
0.00 0.02 0.04 0.06 0.08
Y[m] distance from the center line

Figure 34: E ects of turbulence and transition location on Mach number at 10mm behind the
nozzle exit

H3 − Laminar
1e+00 H3 − Turbulent − xt=throat
H3 − Turbulent − xt=0.04m

1e−01
Pw/P0

1e−02

1e−03

1e−04
−0.20 0.05 0.30 0.55
X [m] distance from the throat

Figure 35: E ects of turbulence and transition location on normalised wall pressure

41
1e+06
H3 − Laminar
H3 − Turbulent − xt=throat
H3 − Turbulent − xt=0.04m

1e+05

qw [W/m2]

1e+04

1e+03

1e+02
−0.20 0.05 0.30 0.55
X [m] distance from the throat

Figure 36: E ects of turbulence and transition location on wall heat ux

Experiment
H3 − Turbulent − xt=0.04m

6.10
Mach number

5.90

5.70

5.50
0.75 0.85 0.95
X [m] distance from the throat

Figure 37: Comparison of the experimental and numerical Mach number on the centerline in
the test section

42
0.040

0.030
Pt2/P0

0.020

exp − x=0.768m
exp − x=0.813m
exp − x=0.858m
exp − x=0.898m
num − x=0.768m
0.010 num − x=0.813m
num − x=0.858m
num − x=0.898m

0.000
0.00 0.02 0.04 0.06 0.08
Y [m] distance from the center line

Figure 38: Comparison of the experimental and numerical Pitot pressure at various cross
sections in the jet ow

43
Figure 39: Mach number isolines in the throat region

Figure 40: Mach number isolines in the nozzle region

Figure 41: Mach number isolines in the jet ow

Figure 42: Boundaries of the wind tunnel and the blunted cone- are

44
6.5

6.0

Mach number

5.5

5.0

4.5
0.75 0.80 0.85 0.90
X [m] distance from the throat

Figure 43: Mach number distribution on the outer boundary of the blunted cone- are

4.0

3.0
Vector angle (deg)

2.0

1.0

0.0
0.75 0.80 0.85 0.90
X [m] distance from the throat

Figure 44: Velocity vector angle on the outer boundary of the blunted cone- are

45
Figure 45: Iso Mach lines above the bow shock of the blunted cone- are in the wind tunnel

46
5000
Experiment
Computation: uniform flow
Computation: varying flow

4000

Pressure [Pa]

3000

2000

1000
0.00 0.05 0.10
X [m] distance from the nose

Figure 46: Wall pressure on the blunted cone- are for the uniform, non-uniform incoming ow
and experiments

2.0e+04
Experiment
Computation: uniform flow
Computation: variable flow

1.5e+04
Qw [W/m2]

1.0e+04

5.0e+03

0.0e+00
0.00 0.05 0.10
X [m] distance from the nose

Figure 47: Wall heat ux on the blunted cone- are for the uniform, non-uniform incoming ow
and experiments

47

You might also like