You are on page 1of 10

36th AIAA Fluid Dynamics Conference and Exhibit AIAA 2006-3042

5 - 8 June 2006, San Francisco, California

Computational Simulation of Shock Oscillation around a


Supersonic Air-Intake

Uichi NISHIZAWA* and Masaharu KAMEDA†


Tokyo University of Agriculture and Technology, Tokyo, Japan

and

Yasushi WATANABE‡
Japan Aerospace Exploration Agency, Tokyo, Japan

Unsteady flow around an external-compression air intake was investigated numerically.


Our attention was focused to shock oscillation, which is known as ‘buzz’, in subcritical
operation. We used a two-dimensional model consisting of a wedge and a subsonic diffuser.
A slit was opened for natural bleed at the entrance of the subsonic duct. The wedge angle
was 10 degrees. The free stream Mach number was 1.64. Compressible viscous flow around
the model was calculated using UPACS (Unified Platform for Aerospace Computational
Simulation) developed at JAXA (Japan Aerospace Exploration Agency). We successfully
simulated a series of flow pattern (supercritical, critical, and subcritical operations) by
changing the intake mass flow rate. In the subcritical operation, multiple dominant
frequencies were found in the pressure fluctuation with the buzz. Low-frequency component
was caused by the acoustic resonance of the subsonic diffuser. On the other hand, high-
frequency component was generated by the periodic flow separation on the wedge. This
phenomenon was caused by the periodic suction of the separated air through the slit. The
high-frequency component disappeared when the diffuser length was long. The natural
bleed slightly suppressed the pressure fluctuation in the subsonic diffuser in a constant mass
flow rate through the diffuser.

Nomenclature
h = throat height
hc = capture height
he = exit height
lFP = length from the flow plug head to the rear end of the intake
lr = ramp length
M = Mach number
P0 = stagnation pressure
PRth = theoretical total pressure recovery
Re = Reynolds number
T0 = stagnation temperature

I. Introduction

N EW supersonic transports (SST) are under development now. Number of environmental and economical
problems is remained to accomplish the development. From mid ‘90s, several research projects on the new
SST have been carried out in Europe, U.S.A., Russia, and Japan. Computational fluid dynamics (CFD) is a key
technique for designing the new SST.

*
Graduate Student, Department of Mechanical Systems Engineering, Koganei, Tokyo 184-8588, Japan.

Associate Professor, Department of Mechanical Systems Engineering, Koganei, Tokyo 184-8588, Japan.

Researcher, Aviation Program Group, Mitaka, Tokyo 182-8522, Japan, Member AIAA.

1
American Institute of Aeronautics and Astronautics

Copyright © 2006 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
We investigated the flow around supersonic air intakes by a CFD technique. The intake is an important part of
supersonic aircraft engines. The intake has a subsonic duct with a supersonic compression wedge. We solved both
the internal and external flow of a compressible viscous fluid, which is a challenging topic on current CFD
techniques.
In the present paper, we focused our attention to an intake ‘buzz’ flow.1 All the supersonic intakes exhibit a
marked instability of flow in subcritical operation below some value of the flow ratio, in the form of oscillation of
the shock waves, which is colloquially known as ‘buzz’, generated at the supersonic compression ramp.
Two remarkable buzz flows was caused by (1) inflow of shear layer (Ferri type2), (2) separation of the flow from
the ramp surface (Dailey type3). Dailey also observed a high frequency and small-amplitude oscillation under low
mass flow rate case, which is caused by acoustic resonance of the subsonic duct.4, 5 Recently, computational fluid
dynamics was applied to simulate the buzz flow.6, 7 However, since the flow is complicated, the mechanism has not
been fully understood yet.
We calculated the shock oscillation of a two-dimensional external-compression supersonic air intake. Especially,
we paid our attention to influence of a slit at the entrance of the subsonic duct. The slit was commonly placed for
bleed of the air, which is effective to reduce the shock oscillation. In addition, we examined the length of the duct
using two models having different lengths.
We found that high-frequency pressure oscillation was generated by the periodic separation on the wedge surface
due to the suction from the slit. Low-frequency component, on the other hand, was caused by the acoustic resonance
of the subsonic duct. The high-frequency component disappeared when the diffuser length was long.

II. Intake Model


Figure 1 shows the supersonic air intake model used for the present numerical simulation. The model is a two-
dimensional external-compression intake. It consists of a subsonic diffuser with a wedge. The supersonic flow into
the intake is firstly compressed through shock waves generated by the wedge. Subsequently, the flow becomes
subsonic and is compressed in the diffuser. We call the diffuser surface aligned with the wedge by ‘ramp,’ while we
call another side of the diffuser by ‘cowl.’ A slit is opened on the ramp at the entrance of the subsonic diffuser. The
slit is used for natural bleed of air, which is often used to stabilize the position of shock waves. A flow plug is placed
behind subsonic diffuser to adjust the flow rate through the intake.

A. Wedge
The angle of the wedge (ramp) is 10 degrees. The uniform stream (Mach number, M=1.64) passes an oblique
shock wave generated by the ramp. Theoretically, the flow is decelerated to M=1.295 and the recovery of total
pressure PRth=0.989. Subsequently, the flow is decelerated again to M=0.786 and PRth=0.979 by a normal shock
wave at the entrance of the subsonic diffuser (throat).

(a) Configuration
Figure 1. Two-dimensional model of an external-compression supersonic air-intake.

2
American Institute of Aeronautics and Astronautics
(b) Shock systems
Figure 1. Two-dimensional model of an external-compression supersonic air-intake.

Figure 2. Shock systems of air-intake.

B. Subsonic Diffuser
The characteristic length of subsonic diffuser is the height of the entrance (throat) h.
The length of subsonic diffuser is 10h. We designed the cross sectional area, which is corresponding to the
height in this two-dimensional model, along the streamline as follows.
The diffuser is separated into forward and rear parts, whose lengths are 3.5h and 6.5h, respectively. The forward
part is a bend. It has a constant deceleration rate; the Mach number gradient is dM/dh = -0.08. The turning angle is -
10 degree to align the direction of the flow parallel to the uniform external flow at the end of the forward part. The
rear part is a straight diffuser, in which the Mach number gradually decreases to a constant value (M=0.4).

3
American Institute of Aeronautics and Astronautics
C. Flow Plug
A triangle flow plug is used to control mass flow through the intake. The vertex angle of the flow plug is 20
degrees.
The cross sectional area of the diffuser exit can alter by insert length of the plug lFP. According to the insert
length, we can operate the intake as critical, subcritical and supercritical conditions, as shown in Fig. 2. The
supercritical condition is obtained when the insert length lFP is smaller than the length at the critical condition lFPC.
Oppositely, the subcritical condition is achieved when lFP is larger than lFPC. Remarkable shock oscillation is
observed in the subcritical condition.

III. Numerical Simulation

A. Flow Solver
We used UPACS (Unified Platform for Aerospace Computational Simulation) developed by JAXA (Japan
Aerospace Exploration Agency) 9, 10 as the flow solver. The governing equations were Navier-Stokes equations. The
Roe scheme was used for spatial difference. The Euler implicit method was used for the time advancement. No
turbulent flow model was incorporated in the present calculation.

B. Numerical Grid
A numerical grid around the intake model is shown in Fig. 3. It was a structured grid and was divided into four
or five blocks. The initial grids were generated by an algebraic method. Then an elliptic method was applied to
improve the diagonality of the final grid. The total numbers of the grid points are from 295,458 to 442,809 points,
which depend on the geometry of the intake model. The minimum non-dimensional spacing of the grid is 2.0×10-5,
which is smaller than the thickness of viscous layer (laminar sublayer). A multi-grid method is used to connect the
blocks.

Figure 3. Numerical grid.

4
American Institute of Aeronautics and Astronautics
IV. Numerical Results and Discussion

A. Calculation Conditions and Parameters


We defined the free-stream conditions as those in our wind tunnel experiments.8 The throat height h is 7.6 mm.
The uniform flow Mach number, M, was 1.64. The stagnation pressure, P0, was 300 kPa. The stagnation temperature,
T0, was 400 K. The free stream Reynolds number, Re, based on the capture height, hc, (10 mm) was 2.9×105. The
positions of the flow plug were changed to control the mass flow rate in the diffuser. The definition of the position is
displayed in Fig. 4.
We carried out the numerical simulation with three different intakes shown in Fig. 5. Figure 5(a) is our standard
model, whose geometry has been described in section 2. Figure 5(b) is the ‘long diffuser’ model, whose diffuser
length extends 10h from the standard model. Figure 5(c) is the ‘without slit’ model, whose geometry is the same as
the standard model except the slit.
Sequential data of flow parameters (pressure, Mach number etc.) was used to perform frequency analysis of the
buzz phenomena. Especially, we focused our attention to frequency spectra of pressure fluctuations at specific points.
The points are displayed in Fig. 4. We will show the pressure data close to the shock waves (b3 and b4), which are
located at 2/3 of the lamp length, lr, from the tip of the lamp. We will also show the data close to the end of subsonic
diffuser (z3 and z4), which are located in front of the flow plug.

Figure 4. Definition of flow plug position and pressure data points.

Figure 5. Three intake models.

5
American Institute of Aeronautics and Astronautics
(a) Numerical simulation (b) Experiment8

Figure 6. Schlieren image of shock systems.

B. Shock Systems of Critical, Supercritical and Subcritical Operations


Three forms of the shock system obtained from the numerical simulation are shown in Fig. 6. The schlieren
images of numerical results are compared with those obtained by a wind tunnel experiment.8 The intake model is our
standard one. The position of flow plug, lFP/hc, for the subcritical, critical and supercritical operations are 2.55, 2.05
and 1.75 in the numerical simulation. Those are 2.30, 2.10 and 1.95 in the wind tunnel experiment.
We successfully capture three principal forms of shock system by the present numerical simulation. The
numerical results agree quantitatively well with the experimental ones.

C. Sequential Images of Shock Oscillation


Next, we move to the shock oscillation in the subcritical operation. Sequential images of the numerical results
are shown in Fig. 7 as schlieren images of density gradients. The position of flow plug for the standard type (Fig.
7(a)) is lFP/hc=2.55, for the ‘long diffuser’ type (Fig. 7(b)) is lFP/hc=2.45, and for the ‘without slit’ type (Fig. 7(c)) is
lFP/hc=2.35. The averaged mass flow ratio is listed in Table 1. The reference is the mass flow rate of the free-stream
through the capture height h. We fixed the mass flow ratio in the subsonic diffuser close to 0.7. In this case, the mass
flow ratio of the suction through the slit is about 0.1.
The oscillation of shock waves close to the wedge is clearly captured. From comparison between the standard
and ‘long diffuser’ (Figs. 7(a) and 7(b)), we do not find any remarkable differences in the shock system except the
amplitude of the shock oscillation. The standard model has larger amplitude than the ‘long diffuser’ one.

6
American Institute of Aeronautics and Astronautics
Table 1. Averaged mass flow ratio.

Type Inflow/Inflow Outflow/Inflow Slit/Inflow

Standard 1 0.681 0.112

Long diffuser 1 0.713 0.101

Without slit 1 0.701 *******

On the other hand, substantial differences are found between the standard and ‘without slit’. The averaged shock
position in the ‘without slit’ model is closer to the tip of the wedge than that of the standard one. This is primary
caused by the difference of total mass flow rate captured by the intake, which is the sum of the mass flows through
the subsonic duct and the slit. The total mass flow of ‘without slit’ model is much lower than that of the standard
model.
The dominant frequency of the shock oscillation was obtained from these sequential images: It is 4.2 kHz in Fig.
7(a), 4.6 kHz in Fig. 7(b), and 600 Hz in Fig. 7(c). We can find a large discrepancy between the standard and the
‘without slit’ models. The cause of this discrepancy is discussed in next subsection.

(a) Standard (b) Long diffuser (c) Without slit

Figure 7. Shock oscillations (one cycle).

D. Frequency Spectra of Pressure Fluctuation


The frequency spectra of pressure fluctuations at four locations (b3, b4, z3, and z4 in Fig. 4) are shown in Fig. 8.
In the standard model (Fig. 8(a)), the dominant frequency is 4.1 kHz at upstream side (b3 and b4) and 2.0 or 4.0
kHz at downstream side (z3 and z4). In the ‘long diffuser’ model (Fig. 8(b)), the dominant frequency is 3.9 kHz at
upstream side and 110 Hz at downstream side. In the ‘without slit’ model (Fig. 8(c)), the dominant frequency is 640
Hz regardless of the positions.
Notice that the amplitude of the ‘without slit’ model is slightly larger than those of the two models.

7
American Institute of Aeronautics and Astronautics
0 0
10 10
pressure data point: pressure data point:
4084[Hz] :b3 :z3
:b4 :z4
–2 –2
10 10
Normalized amplitude

Normalized amplitude
–4 –4
10 10

–6 –6
10 10

–8 –8
10 10
Standard type Standard type

0 2000 4000 6000 8000 0 2000 4000 6000 8000


Frequency [Hz] Frequency [Hz]
(a) Standard
0 0
10 10
pressure data point: pressure data point:
111[Hz] :b3 :z3
3902[Hz] :b4 :z4
–2 –2
10 10
Normalized amplitude

Normalized amplitude

–4 –4
10 10

–6 –6
10 10

–8 –8
10 10
Long diffuser type Long diffuser type

0 2000 4000 6000 8000 0 2000 4000 6000 8000


Frequency [Hz] Frequency [Hz]
(b) Long diffuser
0 0
10 10
pressure data point: pressure data point:
:b3 :z3
:b4 :z4
–2 –2
10 10
Normalized amplitude

Normalized amplitude

639[Hz]
–4 –4
10 10

–6 –6
10 10

–8 –8
10 10
Non–slit type Non–slit type

0 2000 4000 6000 8000 0 2000 4000 6000 8000


Frequency [Hz] Frequency [Hz]
(c) Without slit
Figure 8. Frequency spectra of pressure fluctuation.

8
American Institute of Aeronautics and Astronautics
E. Discussion
Frequency spectra indicate that the high-frequency oscillation (4 kHz) is observed in the standard model as well as
the ‘long diffuser’ one. This oscillation is caused by the periodic suction through the slit of the air in the flow
separation on the wedge surface.
Figure 9 shows iso-Mach contour obtained from the numerical simulation. Figure 9(a) corresponds to the
standard model. In Fig. 9(a), a large separation of flow is observed on the surface of the wedge. The air in the
separation region is sucked into the slit due to the pressure difference between inside and outside of the intake. Just
after the suction, the mass flow into the subsonic diffuser becomes maximum value so that the normal shock is
placed its rear limit position. Then the separation is gradually developed behind the normal shock, which leads to
decrease of the mass flow into the subsonic diffuser. The position of normal shock is also gradually close to the tip
of the wedge due to the decrease of the mass flow into the subsonic diffuser. When the position reaches the front
limit, the air in the flow separation is sucked again.
Figure 9(b) is the iso-Mach contour of the ‘without slit’ model. The separation region of the ‘without slit’ model
is much larger than that of the standard model. In this case, periodic phenomena may be caused by the acoustic
resonance of the subsonic duct. The lowest resonant frequency is defined as
( )
f = 1 − M 2 c 4L (1)
where L, M, and c denote the length, averaged Mach number and speed of the sound in the diffuser, respectively.6
In the present calculation, f = 642 Hz with L = 76 mm, M = 0.7 and c = 380 m/s. This value is close to the dominant
frequency shown in Fig. 8(c).

V. Concluding remarks
The shock oscillation around supersonic air intakes was examined. Especially, we focused our attention to the
effect of the natural bleed through the slit. The principal conclusions are as follows.

1) Natural bleed causes high-frequency shock oscillation. This is due to the periodic suction of the air in flow
separation on the surface of the supersonic ramp (wedge).
2) The high-frequency component disappears in the subsonic diffuser in the case of ‘long diffuser’ model.
3) Natural bleed suppresses the pressure fluctuation at the end of the subsonic diffuser in the case of constant
mass flow ratio of the subsonic diffuser (0.7).

(a) Standard (b) Without slit


Figure 9. Iso-Mach contour.

References
1
Seddon, J., and Goldsmith, E. L., Intake Aerodynamics, 2nd ed., Blackwell Science, 1999, pp. 245-265.
2
Ferri, A., and Nucci, L.M., “The Origin of Aerodynamic Instability of Supersonic Inlets at Subcritical Conditions, ” NACA
RM-L50K30, 1951.
3
Dailey, C. L., “Supersonic Diffuser Instability,” Journal of the Aeronautical Sciences, Vol.22, No.11, 1955, pp.733-749.
4
Sterbentz, W. H., and Evvard, J. C., “Criterions for Prediction and Control of Ram-jet Flow Pulsations,” NACA TN-3506,
1955.

9
American Institute of Aeronautics and Astronautics
5
Trimpi, R. L., “An Analysis of Buzzing in Supersonic Ram Jets by a Modified One-dimensional Nonstationary Wave
Theory,” NACA TN-3695, 1956.
6
Newsome, R. W., “Numerical Simulation of Near-critical and Unsteady Subcritical Inlet Flow Fields,” AIAA-Journal, Vol.
22, No. 10, 1984, pp. 1375-1379.
7
Lu, P. J., and Jain, L. T., “Numerical Investigation of Inlet Buzz Flow,” Journal of Propulsion and Power, Vol.14, 1998,
No.1, pp.90-100.
8
Nishizawa, U., Yamamoto, S., Kameda, M., and Watanabe, Y., “Shock Oscillation of an External-compression Supersonic
Air Intake,” Proceedings of 2005 JSASS-KSAS Joint International Symposium on Aerospace Engineering, Nagoya, Japan, 2005,
pp. 109-112.
9
Yamazaki, H., Enomoto, S., and Yamamoto, K., “A Common CFD Platform UPACS,” High Performance Computing: 3rd
International Symposium, ISHPC 2000, edited by M. Valero, K. Joe, M. Kitsuregawa, H. Tanaka, Lecture Notes in Computer
Science, Springer, Berlin, 2000, pp.182-190.
10
Takaki, R., Yamamoto, K., Yamane, T., Enomoto, S., and Mukai, J., “The Development of the UPACS CFD
Environment,” High Performance Computing: 5th International Symposium, ISHPC 2003, edited by A. Veidenbaum, K. Joe, H.
Amano, and H. Aiso, Lecture Notes in Computer Science, Springer, Berlin, 2003, pp.307-319.

10
American Institute of Aeronautics and Astronautics

You might also like