Sei sulla pagina 1di 10

Proceedings of ASME Turbo Expo 2015: Turbine Technical Conference and Exposition

GT2015
June 15 – 19, 2015, Montréal, Canada

GT2015-42250

Aeroelastic Analysis of Supersonic Intake under Downstream Perturbation

Cheng Yao, Zhansheng Liu, Jiajia Yan, Guanghui Zhang


School of Energy Science and Engineering
Harbin Institute of Technology
Harbin, China 150001

ABSTRACT design may overcome the trade-off in traditional intake variable


Smart flexible structure is used to reduce moving equipments geometry design [3], that is, large amounts of driving and control
of modern supersonic intake with variable geometry design. As equipments. However, fluid structure interaction problem is also
a result, aero-elasticity of flexible intake need to be considered introduced into engine intake.
in aircraft design. A fully coupled fluid structure interaction Due to soften structure stiffness, flexible wall deformation
(FSI) computation framework was developed in present work to and vibration of a supersonic intake with flexible structure play
simulate aero-elasticity of a two-dimensional supersonic intake a important role in the aerodynamic design process, which is
model with flexible cowl wall under downstream back pressure neglected in the traditional design with rigid wall through adding
perturbation. Transonic viscous flow in the intake was solved stiffen structures. Structure deformation increases cross sectional
using finite volume method. Finite element method was used to area under high static pressure loads from internal flow in a
model dynamic response of flexible wall through, consideration supersonic intake. Consequently, oblique shocks , terminal shock
of geometric non-linearity. Unsteady flow in the intake with rigid wave or shock train may move away from design condition. On
wall under back pressure perturbation shows a large amplitude the other hand, supersonic intake may vibrate under transient
oscillation of shock train, which includes a period of shock exciting loads from unsteady flow caused by shock/boundary
train appearing and disappearing. It is found in FSI results layer interaction [4], inlet buzz [5], or downstream perturbation
that flexible cowl wall vibrates with downstream perturbation [6], e.g. ramjet combustor.
frequency, although multiple frequency components excitation Downstream pressure perturbation of a supersonic intake
loads are found in spectrogram. Wall flexibility plays the role in ramjet propulsion system mainly comes from combustion
of ‘buffer zone’, which limits the formation of large amplitude instability, which covers wide range frequency components. Low
shock train oscillation during FSI process. Structural damping frequency or large amplitude pressure fluctuation is of most
could be used to enlarge ‘buffer zone’ effect. troublesome, moves shock upstream and leads to intake ‘unstart’
phenomena. Shock train oscillation tends to be stable under high
frequency pressure perturbation. Relevant studies can be found
1 INTRODUCTION in Ref. [7–9]. Shock oscillation generates moving excitation
Modern aircrafts designed to harness structural flexibility loads on flexible structure of intake. Small amplitude and long
enable higher maneuverability performance, and are of great period vibration occurs and may lead to structure fatigue even
interest. For example, Lockheed Martin Corp. puts much lighter failure of intake.
flexible wings on aircraft (e.g. X-56A) to develop active aero- Aero-elastic problem in an engine intake is rarely reported.
elastic control technologies for flutter suppression [1]. As key On one hand, traditional intake design uses plenty of stiffen
equipment in air breathing aircraft, engine intake tends to be structure to increase intake stiffness and suppress structure
designed with smart flexible material [2]. Smart flexible intake deformation. On the other hand, modern flexible intake design

1 Copyright © 2015 by ASME


is still at exploring stage. Most researches [10–12] focus on b c
application possibility of flexible material to intake design. Yao
studied numerical vibration of a supersonic mixed compression
intake with a unidirectional FSI method [13]. And static wall
deformation’s effect on steady flow solution was then studied
with a bidirectional FSI method [14]. Liang studied an aero- f e d
elastic problem of flexible nozzle with oscillating shock [15].
h i j
However, viscous flow and boundary layer were not included in g
these aero-elastic researches, but are important in a supersonic
intake design. y
In present work, aero-elastic analysis of a supersonic intake
x k
under downstream pressure perturbation is studied using a fully a l
coupled bidirectional FSI method. Sajben diffuser flow case [7]
and supersonic panel flutter case [16] are used to validated the FIGURE 1. SUPERSONIC INTAKE GEOMETRY.
computational FSI framework. Large amplitude of shock train
motion is investigated in transient flow solution with rigid wall
For viscous flow, normal height of cells in near wall region
model. Fluid structure interaction analysis captures dynamic
must be refined enough to capture boundary flow. Therefore,
vibration process of flexible wall induced by moving shock train.
mesh cells are generated so that near-wall non-dimensional
Flexible wall vibration’s ‘buffer zone’ effect is found to limit
height y+ < 5.
large shock train oscillation during FSI process, and could be
Downstream pressure perturbation is applied on boundary j-
enlarged through large structural damping.
k. It is given through a time varied static pressure profile, and is
described using sinusoidal function as follow.
2 NUMERICAL MODELS
2.1 Flow Model pe (t) = [1 + Asin(2π fet)]Rp0 (2)
A two-dimensional supersonic intake is studied in the
presented work, as shown in Fig.1. Note that the dimension where perturbation frequency fe =250 Hz, amplitude A=5%,
ratio of horizontal to vertical direction is two in that figure. static pressure ratio R=12.
Transient viscous compressible flow is governed by Navier-
Stokes equations, written in conservation form as follow.
2.2 Structure Model
Most of the intake structure is considered to be rigid wall.
∂U ∂F ∂G
+ + =0 (1) Part of the cowl wall is modelled as flexible wall, indicated
∂t ∂x ∂y by blue solid region e-f -h-i in Fig. 1. The flexible cowl wall
is modelled as two-dimensional plate with two pinned ends.
Where, U, F and G are dependent vector variables, flux term Structure material is Aluminium alloy, with density ρ =2.7 × 103
matrix in x and y direction, respectively. Flow control equations kg/m, Poisson’s ratio ν =0.3, elastic modulus E=7.1×103 MPa.
are discretized by the Finite Volume Method code F LUENT. Transverse displacement w along y direction is controlled as
Turbulence models are chosen carefully to predict boundary follow.
layer flow near walls. Standard k-ε , standard k-ω and SST k-
ω turbulence models are compared with each other to predict a ∂ 2w ∂ 4w ∂ 2w
accurate solution. ρs h + D 4 − Nx 2 + p = 0 (3)
∂t 2 ∂x ∂x
Supersonic free flow with Ma0 =2.8 enters flow filed with
static pressure p0 =26629 Pa, static temperature T0 =234 K.
where L is plate length, h is plate height, bending stiffness
Boundary a-b-c are set with free flow parameter. For boundary
D=Eh3 /[12(1 − ν 2 )]. Nx is in-plane membrane force caused by
c-d, no parameter need to be specified, because supersonic flow
large deformation. p is external load from supersonic external
parameter at this section can been determined from in-field
flow on the top surface and transonic flow on the bottom surface.
solution. Static back pressure pe is applied on subsonic outflow
Deriving from Von Kármán theory, it is written as follow.
boundary j-k. Adiabatic and no-slip wall boundary condition is
used on boundary d-e-f -g-h-i-j and k-l. In FSI process, coupling
∫ L( )
interface e-f and h-i are set to moving walls, while others are kept Eh ∂w 2
Nx = dx (4)
as rigid walls. At boundary l-a, symmetry boundary is applied. 2L 0 ∂x

2 Copyright © 2015 by ASME


Structure control equations are discretized in the Finite 3.1 Transonic Flow Benchmark
Element Method code A BAQUS. Rayleigh damping is included Transonic flow within a convergent and divergent diffuser
in the structural dynamic calculation. Damping matrix in finite is used to validate flow model. To explore flow unsteadiness
element method is assumed to be the combination of mass matrix in ramjet engine, it has been well studied during the past thirty
and stiffness matrix as follow. years numerically [17, 18] and experimentally [19, 20], and has
become a standard validation case for NASA’s CFD code WIND.
Diffuser geometry is shown in Fig. 2. Throat height hth =44
C = αM + β K (5) mm, and is located at x=0 mm. Detailed top wall coordinate
can be found in Sajben’s work [7]. Air enter the diffuser with
Mass proportional damping coefficient α is neglected in subsonic velocity, and is accelerate to sonic velocity at the throat.
present work. Only stiffness proportional damping coefficient Downstream of throat, air become supersonic. A normal shock
is set to β =2 × 10−5 . wave occurs at some position of divergent portion. Flow velocity
at exit section is subsonic. For a pressure ratio R=0.72, defined
by outflow static pressure pe to inflow total pressurept0 , flow
2.3 FSI Coupling Scheme separation occurs just downstream of the normal shock wave.
FSI coupling scheme must be carefully implemented so that Experiment time mean pressure and velocity data from Sajben’s
numerical coupling process is stable and accurate. Generally, work are also available in the validation case web page of NASA
two kinds of coupling scheme can be found in literature, i.e. WIND.
monolithic scheme and partitioned scheme. In the monolithic The strong shock case (i.e. R=0.72) is studied in present
scheme, governing equations for flow model and structure model validation case. Pressure inflow type of boundary condition is
are discretized together in one global matrix equation. It is time implemented at inflow section, where total pressure pt0 =135000
consuming and numerically unstable in solving such a global Pa, static pressure p0 =116355 Pa, static temperature T0 =273 K.
matrix equation for most comprehensive FSI problem. Therefore At the outflow boundary, only static pressure is specified, i.e.
it is rarely used in actual aero-elastic study of aircraft. The pe =97200 Pa. Top wall and bottom wall inflow boundary layer
partitioned scheme connects any existed CFD code and CSD thickness is about 9% and 4.5% of throat height. Two inflow
code by data transfer at the coupling interface. It is easy to turbulence parameter, turbulence intensity I and turbulence
implement, but may cause numerical instability through adding length scale l, are determined through following empirical
additional energy at coupling interface. equation, which would be corrected depending on turbulence
Partitioned coupling scheme is used in present work to solve models.
the intake FSI problem. Coupling time step is controlled by CFD
code, since time step in unsteady flow is much smaller than that I = 0.16(ReD )1/8 (6)
in structure dynamics. Transient flow solution is updated firstly
at every coupling step. Static pressure near flexible wall is then l = 0.4δ99 (7)
mapped to coupling interface. Pressure load is interpolated to
structure boundary. Structure’s dynamic response is determined
Top wall and bottom wall are set to adiabatic no-slip
after CSD updating. New interface position is transfer back to
wall type of boundary condition. To simulate boundary flow
CFD code from structure solution. A spring based smoothing
accurately, y+ near top and bottom walls is not larger than 5.
type of dynamic mesh generation scheme is used to regenerate
Three types of turbulence models (i.e. standard k-ε , standard k-
CFD mesh, and absorb interface motion in CFD domain. This
ω and SST k-ω ) are implemented to compare with experiment
mesh generation scheme is time saving and is suitable for
data. All steady state flow solution is solved in commercial CFD
interface deformation without large rotational displacement.
software FLUENT.

3 RESULTS AND DISCUSSION


Numerical procedure for a FSI problem is sensitive to flow
model, structure model and coupling scheme. To obtain accurate
FSI solution of the studied intake stalely, two benchmark cases
are implemented to validate previous numerical models, i.e. a
transonic diffuser flow case [7] and a nonlinear supersonic panel
flutter problem [16]. Transient flow in rigid wall model is also
used to study excitation load characteristic on cowl wall. FIGURE 2. TRANSONIC FLOW DIFFUSER GEOMETRY.

3 Copyright © 2015 by ASME


Pressure distributions on top and bottom wall are shown in k-ε model is 1.418, which is larger than 1.369 and 1.335 with
Fig. 3. For solution with standard k-ω turbulence model, I and l SST and standard k-ω model respectively. The maximum Mach
is corrected from Eqn. (6) and Eqn. (7) so that the normal wave number in experiment data is 1.353 , which is close to that with
locates at the position in experiment data. And the other two standard k-ω model. Downstream of the shock, flow separation
solutions utilize value from the empirical equations. Solution zone is predicted in all models. Due to high ability in reverse flow
with standard k-ε turbulence model is not affected by inflow prediction, standard k-ω model describes the largest separation
turbulence parameters’ modification, because it is not sensitive zone among all solution, while the standard k-ε model gives the
to the upstream shear flow. All numerical solutions agree well smallest one. As a result, Mach number contour downstream of
with experiment data in the region upstream the normal shock. the shock in all models are different from each other. From static
Solutions with SST k-ω and standard k-ω turbulence model pressure distribution and Mσ , it is derived that Mach number
accurately predict the normal shock wave position, while that in contour with standard k-ω model is close to experiment results.
standard k-ε turbulence model is downstream of them. However,
only standard k-ω model solution agrees well with experiment
data in the region downstream of the normal shock. Overall,
standard k-ω turbulence model provides better agreement with
the experiment pressure distribution than other two models.
0.000 0.284 0.567 0.851 1.134 1.418
a) Standard k-ε
1.00
Experment
0.86 Standard k-ε
SST
0.72 Standard k-ω
0.000 0.274 0.548 0.821 1.095 1.369

0.58
b) SST k-ω
0.44

0.30
-4.2 -1.6 1.0 3.6 6.2 8.8
x̄ 0.000 0.267 0.534 0.801 1.068 1.335
a) Top wall c) Standard k-ω
1.00 FIGURE 4. MACH NUMBER CONTOUR.
Experment
0.86 Standard k-ε
SST
Standard k-ω
Overall, standard k-ω turbulence model shows better
0.72
agreement to experiment data than standard k-ε and SST k-ω

0.58
turbulence model. It is recommended to use the standard k-ω
turbulence model in supersonic intake flow modeling.
0.44
3.2 Panel Flutter Benchmark
0.30
-4.2 -1.6 1.0 3.6 6.2 8.8 A supersonic nonlinear panel flutter case is implemented to
x̄ validate dynamic characteristic of the CSD model and the FSI
b) Bottom wall coupling scheme. It has been studied by several scholars, such
as Dowell with linear potential aerodynamics theory [16], Davis
FIGURE 3. PRESSURE DISTRIBUTION ON TOP WALL. with Euler equation [21]. It is also used in Gordnier’s work to
validate a three-dimensional viscous aero-elastic solver [22]. The
nonlinear panel geometry is shown in Fig. 5, which is simply
Mach number contours are shown in Fig. 4. At convergent supported at its two ends. Free supersonic idea gas flow pass over
region, all models give same subsonic Mach number contour. the top of panel structure. Free stream Mach number M∞ and
Larger shock position causes larger local Mach number upstream static temperature T∞ are 1.2 and 293K, respectively. Free stream
of normal shock. The maximum Mach number Mσ with standard static pressure p∞ depends on the control parameter: dynamic

4 Copyright © 2015 by ASME


pressure λ . Uniform static pressure applied on panel’s bottom dynamic pressure increases, panel flutter amplitude increases.
surface is equal to the free stream static pressure p∞ . Inviscid gas Larger vibration amplitude causes larger nonlinear panel
flow solution is determined by solving two-dimensional Euler stiffness. Therefore panel flutter frequency also increases with
equation. Density ratio of structure and gas is 5000. Panel dynamic pressure.
height h and length L is 0.002 m and 1.0 m, respectively.
Considering Von Kámán type of geometric nonlinearity, panel
structure is modelled with two dimensional plane stress finite 1.8
element cells. For simply supported boundary condition on the
two ends, rotational degrees of freedom (DOF) is free, and two
1.2
translational DOFs are set to zero. Previous studies show that

w/h
panel flutter characteristic is function of dynamic pressure λ , Dowell
damping , temperature and other parameters. Dynamic pressure 0.6 Gordnier
λ is varied in present validation. Flutter solution is calculated Present
with five different λ of 15, 25, 50, 75 and 100. To speed up initial
numerical process to the limited cycle status, an initial velocity 0.0
profile is applied on all finite element nodes with the function of 0 40 80 120
ẇ(x̄) = ẇ0 sin(π x̄). λ
a) Flutter amplitude
3.0
Free Stream R = 20L
p∞
T∞ ωλ /ω0 2.0
M∞
h/L=0.002
(ρ∞ L)/(ρs h)=0.1 1.0
Gordnier
Present
L
0.0
ρs E ν h 0 40 80 120
y λ
x p∞ b) Flutter frequency
FIGURE 6. SUPERSONIC PANEL FLUTTER RESULTS.
FIGURE 5. SUPERSONIC PANEL FLUTTER GEOMETRY.

Presented flutter results agree well with Dowell and


Gordnier’s results. Flutter amplitude and frequency results are 3.3 Solution with Rigid Wall
shown in Fig. 6. Panel flutter starting point locates between Transonic flow in intake with rigid wall should be
λ =15 and 25. For case with λ =15, panel vibration caused by determined before FSI calculation, including steady state flow
initial velocity profile vanishes with time, and panel flutter does solution and transient flow solution. Steady state flow solution
not occur. Structural damping is not included in panel structure provides basic intake flow patterns, e.g. oblique shock waves,
model, so energy dissipation is caused only by the effective terminal shock, shock train and flow separation zone, which
‘positive’ damping from inviscid supersonic flow. Panel flutter could also be used to initialize flow field in transient calculation.
occurs in the other four cases, where supersonic flow plays roles Transient flow solution includes excitation load parameters on
of ‘negative’ and ‘positive’ effective damping. Energy transfer walls, such as excitation load frequency components and its
between supersonic flow and structure vibration keeps panel corresponding amplitude.
vibration from decaying . For a linear structure model, panel For transient flow calculation, flow initialization greatly
vibration amplitude keeps increasing with time which is affects flow developing process. Good initial flow field may
corresponding to the ‘linear flutter’ phenomena. By considering accelerate numerical procedure. In present work, steady state
Von Kármán geometric nonlinearity, the out-plane deformation transonic flow solution with R=pe /p0 =12 is chosen to initialize
is limited by nonlinear stiffness caused by in-plane membrane transient flow field. Static pressure and Mach number contour
force. So panel’s vibration stays at a limited cycle status. As are shown in Fig 7. Supersonic flow causes two oblique shock

5 Copyright © 2015 by ASME


waves at the first and the second ramp wall. A curved oblique train is regenerated again. Therefore, there exists a ‘life period’
shock is formed in the external flow region near cowl, causing for shock train in the unsteady flow solution, which contains
pressure increase at its downstream region. At the cowl lip, two shock train appearing, developing, disappearing process. In the
oblique shock waves from ramp walls are reflected and form isolator region, subsonic flow is mainly influenced by back
one oblique shock, which enters the supersonic intake. After pressure perturbation.
several oblique shock reflection in the intake, a ‘X’ shape oblique
shock is formed at the downstream end of throat region. Flow
separation bubbles are generated downstream of the terminal
shock at top wall and bottom wall. Near bottom wall, a large flow ⇑

separation zone is predicted in the divergent region and isolator.


Due to shock/boundary layer interaction, shock train is found ⇑

downstream of the ‘X’ shape shock near top wall. Therefore,


several local maximum pressure region are generated in shock ⇑
train region. Pressure in one of them is even close to the outflow
pressure. Shock train causes complicated loads on intake walls, ⇑
which includes a series of pressure increasing and decreasing
processes along stream-wise direction. Boundary layer control

techniques could used to remove the shock train structure, e.g.
wall bleeding or vortex generator. However, to simulate the
serious transient excitation load on intake wall, these flow control ⇑

techniques are not included in present work.


1.00 3.21 5.42 7.63 9.84 12.05


a) Pressure contour
FIGURE 8. TEMPORAL PRESSURE CONTOUR IN ONE CYCLE.

Unsteady flow causes transient excitation load on intake


walls. Pressure load spectrogram of points ( or finite volume
cell) on bottom surface of cowl wall can be generated by FFT
b) Mach number contour calculation using open source code FFTW3. Figure 9 shows
FIGURE 7. STEADY STATE FLOW SOLUTION. pressure power spectrogram of points on cowl wall. Due to
shock motion, excitation load on intake walls varies along flow
path. In supersonic flow region, no excitation load exists, since
Unsteady flow in the supersonic intake is driven by supersonic flow is not influenced. In the region where shock train
downstream pressure perturbation. Temporal pressure contours locates, there exist multiple frequency components, including
during one perturbation cycle are shown in Fig. 8, where black base frequency component 250 Hz and its high order components
arrows indicate shock train position. Large amplitude of shock (e.g. two or three times of the base frequency).
motion is found in the flow field. Since pressure disturbance The base excitation frequency component 250 Hz is equal
translates upstream and downstream with the form of weak to downstream perturbation frequency, which indicates shock
Mach wave, only shock train and its downstream flow region are transient flow is dominated by external disturbance. Amplitude
influenced by pressure perturbation. Solution in the supersonic of the base frequency component depends on position. There
flow region ahead of ‘X’ shape shock keeps invariant, and is not exist three local maximum amplitude values along stream-wise
included in this serial of subfigures. Shock train structure varies direction x̄ in spectrogram figure. They are caused by the ‘X’
during perturbation cycle. At the most downstream position, the shape shock motion, shock train ‘life period’ and back pressure
‘X’ shape shock is replaced by a ‘λ ’ shape shock, and the shock oscillation, respectively. Since the maximum pressure in the
train disappears. When terminal shock moves upstream , shock whole intake locates at the ‘X’ shape shock, as shown in Fig. 8,

6 Copyright © 2015 by ASME


For the first vibration mode, there is no stationary point between
0.32 two ends of the structure. Stationary points can be found in
the other mode shapes, and are indicated by blue region. The
0.24 base excitation load frequency 250 Hz lies between the first and
second order structure frequency (i.e. 195 Hz and 536 Hz), and
(p̄)2

0.16 is much closer to first one. Under base frequency excitation


load, flexible wall may vibrate easily with the first order mode
0.08 shape. High order excitation load frequencies are close to the
1.00 other order natural frequencies. High order vibration mode may
0.75 be excited with proper excitation load distribution and amplitude.
0.00
0 0.50 For example, the fourth order excitation load frequency 1000
500 x̄
1000 0.25 Hz is near the 3rd natural frequency 1048 Hz. However, high
f 1500
2000 0.00 order mode shape is hard to be excited, because high frequency
vibration can be easily reduced through structural damping or
effective flow damping.
FIGURE 9. PRESSURE SPECTROGRAM ON COWL WALL.

large amplitude motion of ‘X’ shape shock generates the global


maximum amplitude of base frequency component on cowl a) 1st order mode shape
wall. It is also the maximum amplitude among all frequency
components’ amplitude. The second local maximum amplitude
caused by shock train ‘life period’ is smaller than the first one. b) 2nd order mode shape
But it is larger than that induced by weak Mach wave propagation
near intake outlet. c) 3rd order mode shape
High order excitation frequency components mainly come
from the shock train motion. Suppose that point i locates at
‘X’ shape shock’s moving region. Supersonic flow and static d) 4th order mode shape
pressure remain constant when shock train locates downstream 0.0 0.2 0.4 0.6 0.8 1.0
of point i, as shown in Fig 8. When the ‘X’ shape oblique
shock moves upstream from this point, pressure changes greatly FIGURE 10. FIRST FOUR MODE SHAPE.
because of ‘X’ shape shock motion and shock train’s ‘life
period’. After performing FFT calculation, multiple frequency
components excitation are found at point i. In the downstream
region of shock train or isolator, these high frequency excitation TABLE 1. FLEXIBLE WALL NATURAL FREQUENCIES.
components vanish. Subsonic flow solution in that region is
mainly influenced by back pressure oscillation frequency 250
Hz. Order Frequency (Hz) Order Frequency (Hz)
As shown in the above discussion, complicated excitation 1st 195 3rd 1048
loads act on intake walls under downstream perturbation. Two
types of loads could be found, i.e. moving shock excitation 2nd 536 4th 1727
load near throat region and harmonic excitation load in isolator.
Moving shock load contains several frequency components,
which could induce multiple frequency wall vibration depending
on structural damping. Harmonic excitation load would cause Temporal deformations of bottom surface of the flexible
harmonic wall vibration. wall in one back pressure perturbation cycle are shown in
Fig. 11. Blue arrows indicate vibration direction, while black
solid squares show positions with maximum deformation value
3.4 FSI Solution at every time frame. Due to high static pressure load, maximum
Mode results, including natural frequency and mode shape, static deformation is about 70% of wall thickness, and flexible
are always used to evaluate dynamic response of structure. With wall vibrates at one side of its initial configuration. Since the
mode analysis, the first four order natural frequencies are shown maximum deformation is larger than half of wall thickness,
in Table 1. Corresponding mode shapes are shown in Fig 10. geometric non-linearity must be included in FEM model.

7 Copyright © 2015 by ASME


No stationary point is found in all transient deformation static pressure load, flow path is enlarged in region of flexible
curves, which are similar to the first order mode shape, as shown wall. As a result, the equilibrium position of shock train moves
in Fig. 10-a). It means that flexible wall vibrates mainly with first upstream from those in rigid wall model. Wall vibration yield a
order mode shape. This is mainly caused by large static pressure adaptive flow path, which acts as a ‘buffer zone’ to shock train
load. Although there are multiple frequency excitation loads motion.
on the flexible wall, Only one single vibration frequency 250 Attributing to ‘buffer zone’ effect, shock train structure in
Hz is found in vibration spectrogram. The vibration frequency FSI solution is different from that in the rigid wall model. The
is equal to base load excitation frequency or the perturbation ‘life period’ phenomena could be found in shock train motion
frequency. High frequency vibration components are dissipated of the‘ rigid wall model, as shown in Fig. 8. However, shock
by structural damping and effective flow damping. Therefore, train structure does not vary with time in flexible wall solution,
flexible wall’s vibration is mainly controlled by base frequency as shown in Fig. 12. And the ‘X’ shape oblique shock stays at
excitation load. the starting position of shock train structure during oscillating
However, flexible wall vibration shape is not purely the process. Therefore, shock train oscillation under downstream
first order mode shape, a harmonic vibration response as shown perturbation is stabilized through flexible wall vibration’s ‘buffer
in Fig. 10-a). Under harmonic excitation load, the maximum zone’ effect.
displacement points stay at the same x coordinate during
dynamic process with a first order mode shape. In Fig. 11 the
maximum displacement point varies along x coordinate within
one perturbation cycle. And the obit of black solid squares is ⇑

approximately like the Arabic number ‘8’. This is the result


of moving shock load, attributing to oscillating motion of shock ⇑
train.

1.20

0.90
w/h


0.60

0.30 ⇑

0.00
0.00 0.07 0.15 0.22 0.30 ⇑
x/L
a) nT < t < (n + 12 )T ⇑

1.20

0.90
1.00 3.21 5.42 7.63 9.84 12.05
w/h

0.60
FIGURE 12. PRESSURE CONTOUR IN ONE CYCLE.
0.30

0.00
0.00 0.07 0.15 0.22 0.30 Shock train oscillation amplitude is limited by the ‘buffer
x/L zone’. Starting points of shock train on ramp wall are used to
b) (n + 12 )T < t < (n + 1)T indicate shock train motion, which are shown with arrows in
Fig. 8 and Fig. 12. The root mean square amplitude of shock train
FIGURE 11. PANEL DEFORMATION IN ONE CYCLE. motion in flexible wall model is determined as 0.0253 m, about
half of that in rigid wall model, i.e. 0.0572 m. To quantitative
describe shock train oscillation amplitude reduction by ‘buffer
Unsteady flow solution in the intake is influenced by zone’, parameter Φ is introduced and defined by x f /xr , where x f
flexible wall vibration. Temporal pressure contours during one and xr are shock train oscillation amplitude in flexible wall model
perturbation cycle are shown in Fig. 12. Due to large internal and rigid wall model. For current case, amplitude reduction

8 Copyright © 2015 by ASME


parameter Φ = 44.2%. Unsteady transonic flow solution is obtained with standard
The basic mechanism of flexible wall vibration’s ‘buffer k-ω model in the supersonic intake with rigid wall model.
zone’ effect is unknown. A reasonable explanation may be the Under back pressure perturbation, shock train maintains a large
energy transfer phenomena between unsteady flow solution and amplitude motion with a ‘life period’, including appearing and
dynamic vibration of flexible wall. Suppose that the external disappearing process. Two types of exciting loads are found
energy e0 is introduced into intake flow field through downstream on intake structure. In regions of throat and divergent portion,
pressure perturbation. So shock train is forced to oscillate with a moving shock load acts on the wall, while harmonic exciting
large amplitude x0 in the rigid wall model. During FSI process, load in isolator wall. Through FFT calculation, several several
part of the external energy ew is transferred to structure model, frequency components are investigated in the exciting load
which causes flexible wall vibration and energy dissipation spectrogram. Base frequency equals to perturbation frequency,
through structural damping. Driven by reduced external energy while the other components are its multiple results.
e0 -ew , shock train motion amplitude decreases to x1 from x0 . In FSI solution shows that flexible wall obtains forced vibration
this point of view, one could enlarge ‘buffer zone’ effect through status. Wall vibration frequency is controlled by downstream
increasing energy transfer at the FSI interface. A simple way perturbation frequency, while high frequency components vanish
is to change structural damping. Structural damping’s effect on through structural damping and effective flow damping. Moving
shock train oscillation amplitude reduction is shown in Table 2. shock load cause the maximum flexible wall deformation to
As structural damping increase, flexible wall’s ‘buffer zone’ vary along stream-wise direction during one perturbation cycle.
Flexible wall vibration’s ‘buffer zone’ effect is found to stabilize
unsteady flow in the supersonic intake, which would limit shock
TABLE 2. SHOCK TRAIN OSCILLATION AMPLITUDE WITH train motion amplitude to one half of that in the rigid wall model,
VARIOUS STRUCTURAL DAMPING. depending on structural damping.

β Φ Amplitude(m) ACKNOWLEDGEMENT
0 45.6% 0.0261 The financial supports provided by the Funds of the National
Natural Science Foundation of China (Grant NO. 51106035) and
2 × 10−5 44.2% 0.0253 China Postdoctoral Science Foundation funded project (Grant
5 × 10−5 43.9% 0.0251 NO. 2012M510088) are gratefully acknowledged.

1 × 10−4 43.7% 0.0250


REFERENCES
[1] Ryan, J. J., Bosworth, J. T., and J, B. J., 2014. “Current
and future research in active control of lightweight, flexible
effect enlarges, and shock train oscillation amplitude decreases. structures using the x-56 aircraft”. In 52nd Aerospace
However, the limit of structural damping’s effect on shock train Sciences Meeting, AIAA, AIAA 2014-0597.
amplitude reduction tends to be 43%, as shown in Table 2. [2] Hopkins, M. A., Dunne, J. P., and Baumann, E. W.,
Other parameters affecting ‘buffer zone’ effect include flexible 1999. “Adaptive fighter engine inlet”. In 40th
wall stiffness, shock train position, downstream perturbation Structures, Structural Dynamics, and Materials Conference
frequency and amplitude, which need be explored in future and Exhibit, AIAA, AIAA-99-1512.
analysis. [3] Sbester, A., 2007. “Tradeoffs in jet inlet design: A
historical perspective”. Journal of Aircraft, 44(3), pp. 705–
717.
4 CONCLUSIONS [4] Carroll, B. F., and Button, J. C., 1990. “Characteristics of
Aero-elastic study on a supersonic intake is implemented multiple shock wave/turbulent boundary-layer interactions
with a fully coupled FSI computational framework. Three types in rectangular ducts”. Journal of Propulsion and Power,
of turbulence models are compared with each other to simulate 6(2), pp. 186–193.
shock/boundary layer interaction in Sajben transonic diffuser. [5] Trapier, S., Duveau, P., and Deck, S., 2006. “Experimental
With proper inflow condition, standard k-ω model predicts better study of supersonic inlet buzz”. AIAA Journal, 44(10),
agreement with experiment data than standard k-ε model and pp. 2354–2365.
SST k-ω model, and is recommended in present supersonic flow [6] Yu, K. H., Trouve, A., and Daily, J. W., 1991.
simulation. “Low-frequency pressure oscillations in a model ramjet

9 Copyright © 2015 by ASME


combustor”. Journal of Fluid Mechanics, 232, Nov., “Characteristic frequencies of transonic diffuser flow
pp. 47–72. oscillations”. AIAA Journal, 21(9), pp. 1232–1240.
[7] Sajben, M., BOGAR, T. J., and KROUTIL, J. C., 1984. [21] Davis, G., and Bendiksen, O., 1993. “Transonic panel
“Forced oscillation experiments in supercritical diffuser flutter”. In 34th Structures, Structural Dynamics and
flows”. AIAA Journal, 22(4), pp. 465–474. Materials Conference, AIAA, AlAA 93-1476.
[8] Bogar, T. J., Sajben, M., and Kroutil, J. C., 1985. [22] Gordnier, R., and Visbal, M., 2002. “Development of a
“Response of a supersonic inlet to downstream three-dimensional viscous aeroelastic solver for nonlinear
perturbations”. Journal of Propulsion and Power, panel flutter”. Journal of Fluids and Structures, 16(4),
1(2), Mar., pp. 118–125. May, pp. 497–527.
[9] Nori, V., Lerma, N., Gustavsson, J., Segal, C., and
Fernandez, R., 2006. “Forced oscillations in a mixed-
compression inlet at mach 3.5 for pulse detonation engine
systems”. Journal of Fluids Engineering, 128(3), p. 494.
[10] Pitt, D. M., Halleux, J. P., and White, E. V., 2001.
“SAMPSON smart inlet SMA powered adaptive lip design
and static test”. In 42nd AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials
Conference, AIAA, AIAA 2001-1359.
[11] Pitt, D. M., Dunne, J. P., and White, E. V., 2002. “Design
and test of a SMA powered adaptive aircraft inlet internal
wall”. In 43rd AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, AIAA,
AIAA 2002-1356.
[12] Song, G., Ma, N., and Penney, N., 2004. Design and
control of a proof-of-concept active jet engine intake using
shape memory alloy actuators. Tech. Rep. NASA/TM-
2004-213124, NASA.
[13] Yao, C., Liu, Z.-s., Wei, J.-q., Qu, S.-y., and Zhang, G.-
h., 2014. “Effect of wall deformation on aerodynamic
performance for mixed compression intake”. In 44th AIAA
Fluid Dynamics Conference, AIAA, AIAA 2014-2445.
[14] Cheng, Y., Liu, Z., Ma, R., and Zhang, G., 2013.
“Numerical vibration analysis of supersonic mixed-
compression intake”. In ASME Turbo Expo 2013: Turbine
Technical Conference and Exposition, Vol. 2, ASME,
GT2013-95444.
[15] Liang, S.-M., Tsai, C. J., and Yu, F.-M., 1995. “Shock
oscillation in a two-dimensional, flexible-wall nozzle”.
AIAA Journal, 33(8), pp. 1538–1541.
[16] Dowell, E. H., 1967. “Nonlinear oscillations of a fluttering
plate. II.”. AIAA Journal, 5(10), pp. 1856–1862.
[17] Hsieh, T., Bogar, T. J., and Coakley, T. J., 1987. “Numerical
simulation and comparison with experiment for self-excited
oscillations in a diffuser flow”. AIAA Journal, 25(7),
pp. 936–943.
[18] Liou, M.-S., and Coakley, T. J., 1984. “Numerical
simulations of unsteady transonic flow in diffusers”. AIAA
Journal, 22(8), pp. 1139–1145.
[19] Chen, C. P., Sajben, M., and Kroutil, J. C., 1979. “Shock-
wave oscillations in a transonic diffuser flow”. AIAA
Journal, 17(10), pp. 1076–1083.
[20] Bogar, T. J., Sajben, M., and Kroutil, J. C., 1983.

10 Copyright © 2015 by ASME

Potrebbero piacerti anche