DOM-Shock Wave Boundary
DOM-Shock Wave Boundary
DOM-Shock Wave Boundary
Our Channel
For GATE, ESE, AAI, PSU, ISRO, DRDO and Other Technical Exams
SCAN
OR
CLICK
YouTube Telegram
SCAN
OR
CLICK
YouTube Telegram
John K. Harvey is a Professor in Gas Dynamics at Imperial College and is a visiting pro-
fessor in the Department of Engineering at the University of Cambridge. He obtained
his PhD in 1960 at Imperial College for research into the roll stability of slender delta
wings, which was an integral part of the Concorde development program. In the early
1960s, he became involved in experimental research into rarefied hypersonic flows, ini-
tially with Professor Bogdonoff at Princeton University and subsequently back at Impe-
rial College in London. He has published widely on the use of the direct-simulation
Monte Carlo (DSMC) computational method to predict low-density flows, and he has
specialized in the development of suitable molecular collision models used in these
computations to represent reacting, ionized, and thermally radiating gases. He has also
been active in the experimental validation of this method. Through his association with
CUBRC, Inc., in the United States, he has been involved in the design and construc-
tion of three major national shock tunnel facilities and in the hypersonic aerodynamic
research programs associated with them. Professor Harvey has also maintained a strong
interest in low-speed experimental aerodynamics and is a recognized expert on the aero-
dynamics of F1 racing cars. Professor Harvey is a Fellow of the Royal Aeronautical Soci-
ety and an Associate Fellow of the American Institute of Aeronautics and Astronautics.
Cambridge Aerospace Series
Editors:
Wei Shyy and Michael J. Rycroft
Edited by
Holger Babinsky
University of Cambridge
John K. Harvey
Imperial College
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, São Paulo, Delhi, Tokyo, Mexico City
Cambridge University Press
32 Avenue of the Americas, New York, NY 10013-2473, USA
www.cambridge.org
Information on this title: www.cambridge.org/9780521848527
C Cambridge University Press 2011
A catalog record for this publication is available from the British Library.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . page 1
Holger Babinsky and John K. Harvey
2 Physical Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Jean Délery
vii
viii Brief Contents
Index 459
Contents
2 Physical Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Shock Wave–Boundary-Layer Interactions: Why They
Are Important 5
2.2 Discontinuities in Supersonic Flows 6
2.2.1 Shock Waves 6
2.2.2 The Shock-Polar Representation 7
2.2.3 Shock Intersections and the Edney Classification
of Shock-Shock Interferences 11
2.2.4 Shock Waves, Drag, and Efficiency: The Oswatitsch
Relationship 16
2.3 On the Structure of a Boundary-Layer Flow 19
2.3.1 Velocity Distribution through a Boundary Layer 19
2.3.2 The Multilayer Structure 24
2.3.3 The Boundary-Layer Response to a Rapid Pressure
Variation 25
2.4 Shock Waves and Boundary Layers: The Confrontation 26
2.4.1 The Basic SBLI in Two-Dimensional Flows 26
2.4.2 The Boundary-Layer–Shock-Pressure-Jump Competition 28
2.5 Interactions without Separation: Weakly Interacting Flows 31
2.5.1 The Incident-Reflecting Shock 31
Overall Flow Organisation 31
Shock Penetration in a Rotational Layer 33
2.5.2 Ramp-Induced Shock 35
2.5.3 Normal Shock and Transonic Interactions 36
2.5.4 Upstream Influence Scaling 38
2.6 Interaction Producing Boundary-Layer Separation: Strongly
Interacting Flows 39
ix
x Contents
structure under a variety of circumstances and also show how they impact on other
parts of their flowfield, influencing parameters such as the drag, the surface-flux dis-
tributions, and the overall body flow. Our intention is to develop an understanding
of which circumstances lead to their formation, how to estimate their effect, and
how to manage them if they do occur. We demonstrate how the present state of
our understanding has resulted through contributions from experiments, computa-
tional fluid dynamics (CFD), and analytical methods. Because of their significance
for many practical applications, SBLIs are the focus of numerous studies spanning
several decades. Hence, there is a considerable body of literature on the subject.
We do not attempt to review all of it in this book but we aim to distill from it the
information necessary to fulfill our aims.
a very personal perspective and describes those areas where he considered analyt-
ical methods could be applied effectively to the SBLI problem. We consider it a
fitting tribute to him that we are able to include his material in its entirety as, in
some measure, it summarizes a significant portion of his life’s work, that of apply-
ing asymptotic expansion methods to fluid mechanics. Building on the legacy of
Lighthill, Stewartson, and Neiland who developed the so-called triple deck method
in the late 1960s, George extended their concepts well beyond their early achieve-
ments to, for example, SBLIs including turbulent interactions; a section devoted to
this is included in his chapter as hitherto unpublished work. He was a great enthu-
siast for analytical methods and he firmly believed that they complemented and
were an essential adjunct to experiments and numerical methods, being an effective
means of enhancing our insight into the physics of a flow. While acknowledging that
for complex flows such as SBLIs these methods have proved to have had significant
limitations, they nevertheless continue to provide us with valuable interpretations
of observed phenomena and predict flow behavior over a wide range of conditions.
For this reason we were delighted to be able to have his contribution to this book.
Before entering academia, he had worked in the aerospace industry with McDonnell-Douglas, Bell
Aircraft, and the GM Research Laboratories. Over his career, he published extensively and became
a pioneer in the basic theory of high temperature chemically reacting gas flows and propulsion in
space.
2 Physical Introduction
Jean Délery
5
6 Physical Introduction
M 2, p 2 M 2, p2
M1, p1
M1, p1
σ
A A
a- attached shock wave (ϕ < limit deflection) b- detached shock wave (ϕ > limit deflection)
Figure 2.1. Attached and detached planar shock wave.
p
Normal shock p1 Strong shock
2
Subsonic
4
Supersonic
Maximum deflection
5
1
Weak shock
Vanishing shock p1
3
ϕ
Limit angle ϕ max
Figure 2.2. The shock-polar representation in term of flow deflection and pressure jump.
closed curves that are symmetrical with respect to the axis ϕ = 0 (if ϕ1 is assumed
to be equal to zero), as shown in the example plotted in Fig. 2.2. At the origin, the
polar has a double point corresponding to a vanishing shock (i.e., Mach wave). For
a given value of the deflection angle ϕ, there are two admissible solutions, (1) and
(2). (A third solution for which the pressure through the shock decreases is rejected
because it fails to satisfy the Second Law of Thermodynamics.) Solution (1), which
leads to the smaller pressure jump, is called the weak solution; the second is the
strong solution. For ϕ = 0, the strong solution is the normal shock – that is, point (4)
on the shock polar.
There is a maximum deflection ϕmax beyond which an attached shock at A is no
longer possible. If the deflection ϕ imparted by the ramp is greater than ϕmax , then
a detached shock is formed starting from the wall upstream of A (see Fig. 2.1b). In
this case, the flow downstream of the shock does not have a unique image point on
the polar but instead follows an arc extending from the normal shock image (i.e., for
the shock foot at the wall) to the image corresponding to the shock away from the
wall. Another particular point about the polar is the image of the shock for which
the downstream flow is sonic. This point is slightly below the maximum deflection
location and it separates shocks with supersonic downstream conditions from those
with subsonic downstream conditions. A shock polar exists for every upstream Mach
number; the shape of the curves for several examples is illustrated in Fig. 2.3. The
slope of shock polar (dp/dϕ)0 at the origin
√ passes through a minimum (dp/dϕ)min
for an√upstream Mach number M1 = 2 = 1.414,√with (dp/dϕ)0 > (dp/dϕ) √ min for
M1 > 2 and (dp/dϕ)0 > (dp/dϕ)min for M1 < 2. Thus, when M1 > 2, each
successive polar is above the previous one as the Mach number increases. The order
2.2 Discontinuities in Supersonic Flows 9
120
p M1 10
p1
100
9
80
8
60 7
40
6
5
20 4
3
2
-60 -40 -20 0 20 40 60
Figure 2.3. Shock polars for varying upstream flow Mach numbers (γ = 1.4).
√
reverses for M1 < 2, with the polar for a lower Mach number now above the pre-
vious polar (Fig. 2.4). This fact is significant when considering the shock penetration
of a boundary layer (see Section 2.5).
p9
8 M1 2.6
p1 7 2.4
6
2.2
5
2.0
4
1.8
3
1.6
2
M1 2
1.3
1.2
1. 1
()
1
-30 -20 -10 0 10 20 30
Figure 2.4. Relative positions of the shock polars (γ = 1.4).
10 Physical Introduction
5 45
p
p
p1 4.5 40
p1
4
M1 2 35 M1 6
3.5 Isentropic polar 30
Isentropic polar
3
25
20
2.5
15
2
10
1.5
5
1
-20 -10 0 10 20
-40 -20 0 20 40
a - M1 2.34 b - M1 2.34
Figure 2.5. Relative location of the shock polar and the isentropic polar (γ = 1.4).
Any flow of gas in equilibrium undergoing isentropic changes from known stag-
nation conditions is completely defined by two independent variables: pressure p
and direction ϕ. This flow has a unique image point in the plane [ϕ, p], which is
considered a hodographic plane. Passing through a shock wave entails a change in
entropy of the fluid, resulting in a jump of its image in the [ϕ, p] plane. The new
point must lie on the shock polar attached to the upstream state. An interesting
property of the hodographic representation [ϕ, p] is that two contiguous flows sep-
arated by a slip line have coincident images because – according to the Rankine-
Hugoniot equations – the condition for the flows to be compatible is that they have
the same pressure and direction. Similarly, a simple isentropic expansion or com-
pression can be represented by an isentropic polar in the [ϕ, p] plane. For a planar
two-dimensional flow of a calorically perfect gas, such a curve is defined by the fol-
lowing characteristic equation:
ω(M, γ ) ± ϕ = constant,
γ +1 γ −1 2
ω(M, γ ) = tan−1 (M − 1) − tan−1 M2 − 1
γ −1 γ +1
The polar representing an isentropic compression from the same Mach number
is plotted with the shock polar in Fig. 2.5. It can be demonstrated that at the origin,
the two curves have a third-order contact, so they remain very close until relatively
high deflection angles. At moderate Mach numbers, weak-type shock solutions can
be considered as almost isentropic. The isentropic polar is below the shock polar for
upstream Mach numbers that are less than about 2.34 and it passes above for higher
Mach numbers.
2.2 Discontinuities in Supersonic Flows 11
A2
ϕ2
(C2) 2 (C4) 4 (Σ)
M1 T
ϕ4 = ϕ 5
Slip line
1 ( C 1) 3 (C3)
5
ϕ1
A1
(a)
3
p 2
p1 4 5
10 2
C4 C3
2
1
10
1
C1
C2
1
()
0
10
-60 -40 -20 0 20 40 60
(b)
Figure 2.6 (a) Type I shock-shock interference. Physical plane. (b) Type I shock-shock inter-
ference. Plane of polars (M1 = 10, ϕ1 = 20◦ , ϕ2 = −30◦ ).
ϕ2
A2
p
3
p1 2
4 6 7
2
5
10 C4 C3
2
3
C5
101 1
C2 C1
1
()
0
10
-60 -40 -20 0 20 40 60
(b)
Figure 2.7 (a) Type II shock-shock interference. Physical plane. (b) Type II shock-shock
interference. Plane of polars (M1 = 10, ϕ1 = 30◦ , ϕ2 = −35◦ ).
flow is still supersonic downstream of (C4 ); and a supersonic jet bounded by two
slip lines, or jet boundaries, ( f1 ) and ( f2 ) are formed. The jet is surrounded by sub-
sonic flows in which the pressure is virtually constant. As in the previous case, the
flow contains two triple points, T1 and T2 . To maintain continuity in pressure when
shock (C4 ) impacts boundary ( f2 ), a centered expansion must form to offset the
pressure jump across (C4 ). This expansion is reflected by the opposite boundary ( f1 )
as a compression wave, which, in turn, is returned by ( f2 ) as an expansion wave, and
so on.
Type V interference occurs when incident shock (C1 ) crosses shock (C2 ) in a
region where (C2 ) is a strong-oblique shock; the interference involves two oblique
14 Physical Introduction
T2 7
6
C4 2
C5
C3 4 5
7
2 T2 6 2
C5
(a)
3
10
p
p1
2
3
2
10
4
C3
1
2
C2
1
10
C1
1 ()
0
10
-40 -20 0 20 40 60
(b)
Figure 2.8 (a) Type III shock-shock interference. Physical plane. (b) Type III shock-shock
interference. Plane of polars (M1 = 10, ϕ1 = 20◦ ).
shocks from the same family. As shown in Fig. 2.10, the resulting field adopts a
complex structure with two multiple points T1 and T2 similar to those associated
with Type II interference; however, a supersonic jet leaves from T2 instead of a
simple slip line. This complex structure also can be interpreted by considering a
2.2 Discontinuities in Supersonic Flows 15
Pressure and
C2 heat transfer peaks
1 M<1
3
M1
7 f2
T1 6
C3 4
C4 T2 8
C1 2 5
f1 M< 1
M 1 C5
(a)
103
3
boundary f 1
p 5 8
p1
6 2
3
3 1
102 7
4
2
boundary f 2
C3
C4
10
1 2
C2
C1
1 ()
0
10
-40 -20 0 20 40 60
(b)
Figure 2.9 (a) Type IV shock-shock interference. Physical plane. (b) Type IV shock-shock
interference. Plane of polars (M1 = 10, ϕ1 = 10◦ ).
C3 M 1
M1
2
6
T2 4
1
M =1
C1 C5 8 M 1
1
T1 7
2 M 1
C2 5
1 C4
which exists downstream of (C2 ), and state (3) are incompatible and intermediate
state (6) must be introduced, which is connected to state (4) by an expansion wave,
as shown in Figs. 2.11a and 2.11b. At moderately supersonic Mach numbers (i.e.,
maximum close to 2), the transition between states (4) and (6) may occur across a
shock wave, which is usually of very low intensity (Fig. 2.11c).
The quantities δs and δhst represent variations of specific entropy and total
enthalpy, respectively, on the surface relative to the uniform upstream state. The
aim of a propulsion system is to ‘inject’ total enthalpy into the flow to produce
thrust in order to compensate for the entropy term, thereby allowing the integral
to be equal to zero or to change its sign so that the thrust exceeds the drag. If we
concentrate on drag alone, then:
1
D= T∞ δsdqm ,
V∞ (s)
and it is immediately evident that drag and entropy production are closely linked.
Conversely, if the flow is adiabatic and the fluid is in an equilibrium state, then the
C3
Slip line
M1
3
T
6 Expansion wave
1
4
C1 C2
2
ϕ2
ϕ1
(a)
p 2
p1 10
3
4
6
3 C2
2 1 2
10
C3
Isentropic expansion polar
1
10
C1
1 ()
0
10
-40 -20 0 20 40 60
(b)
C3 Slip line
M1 T1 3
6
Shock wave
1
4
C2
C1
2
(c)
Figure 2.11 (a) Type VI shock-shock interference (case with expansion). Physical plane. (b)
Type VI shock-shock interference. Plane of polars (M1 = 10, ϕ1 = 20◦ , ϕ2 = 35◦ ). (c) Type VI
shock-shock interference (case with shock). Physical plane.
18 Physical Introduction
s = C p ln Tst − r ln pst ,
where Cp is the coefficient of specific heat and r is the gas constant. Because the
stagnation temperature is conserved in the flow, the following relationship exists
between the entropy rise and the stagnation pressure ratio through a shock wave:
pst2
s = s2 − s1 = r ln .
pst1
Hence:
pst2 s2 − s1
= exp − .
pst1 r
This demonstrates that efficiency losses also are directly connected to entropy pro-
duction (hence, viscosity) and that they are the internal aerodynamics equivalent
to the drag in external flows. Pressure recovery is not the only concern in air-intake
design; the quality of the flow entering the engine also is of utmost importance. Thus,
a good uniformity, or small distortion, is required, which implies the absence of
2.3 On the Structure of a Boundary-Layer Flow 19
1.2 1.2
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
1
2
u u w : Friction velocity
Laminar Blending w
u y : Distance to the wall
sublayer region
30 10 2
ue 5
10
Viscous sub layer
u u y
Wake
20 u
component
u uy
0.41ln 5
u
Logarithmic y 0.1 0.2 Outer region
10 region
u y
u y : Function of the Reynolds number Re
0
1 10 100 1000
Figure 2.13. Structure of a well-behaved flat-plate turbulent boundary layer.
=1+ ln − − 2−w
ūe k ūe δ δ k ūe δ
=1+ ln − − 2−w
ūe 0.41 2 δ δ 0.14 2 δ
Coles’s relationship can be used for compressible flows by replacing C f with the
‘incompressible’ skin friction C fi , which is related to C f by:
12
γ −1 2
C fi = C f 1+ Me
2
1
y
0.8
0.6
2.11
1.58
0.4 1.89
Hi 2.40 1.72 1.47
1.38
0.2
Hi 1.32
u
0
0 0.25 0.5 0.75
ue 1
Me 0
2
Hi
4
1.4 6
1.2
1
10 2 103 10 4 Me 0 R
Hi 2
1.4 4
1.2 6
1
10 4 105 106 107 R
(see Fig. 2.12). The distance from the wall of the sonic line in a turbulent velocity
profile as a function of the shape parameter is shown in Fig. 2.16 for two values of
the Mach numbers and different ratios of the wall temperature to the outer-flow
stagnation temperature. More precisely, the thermal condition at the wall should be
characterised by the ratio of the wall temperature Tw to the recovery temperature Tr
corresponding to adiabatic conditions at the wall; that is, the temperature acquired
by the fluid at the wall for zero heat transfer (the recovery temperature is inferior to
the outer-flow stagnation temperature but not very different from it). For flat-plate
conditions (Hi ≈ 1.3) and even for a moderate outer Mach number, the sonic point
is close to the surface. For cold-wall (Tw /Tr < 1) conditions, this point is even closer
because of the lower sound speed. However, the location of the sonic point in the
boundary layer rapidly moves away from the wall if it is heated (Tw /Tr > 1).
0.7 0.7
ys ys
Tw
0.6
δ Tst 0
=3 0.6
δ
2.7 2.4
2.1 1.8
0.5 0.5
1.5 1.2
0.4 0.4
0.9 0.6 Tw
0.3 0.3
=3 2.7
Tst 0
2.4 2.1
0.2 0.3 0.2
1.8 1.5
0.025
Cf
0.02
Me 0
1
2
0.015
3
4
0.01 5
6
0.005
Re x
0 3 4 5 6 7 8
10 10 10 10 10 10
0.025
Cf
0.02
Me 0
1
0.015 2
3
4
0.01 5
6
0.005
Re x
0 3 4 5 6 7 8
10 10 10 10 10 10
shear forces tend to oppose the retardation effect of the shock, we can infer that the
resistance of a laminar boundary layer to an imposed pressure gradient will diminish
with an increasing Reynolds number. For the turbulent boundary layer, the situation
is more subtle, as discussed subsequently.
1. An outer inviscid layer that usually is irrotational (i.e., isentropic) and there-
fore obeys the Euler equations or alternatives such as the potential equation.
However, there are exceptions in which this part of the flow is rotational – for
example, downstream of the curved shock formed ahead of a blunted leading
edge, where what is referred to as an entropy layer is formed. A similar rota-
tional layer can occur behind the near-normal but curved shock that forms on a
transonic aerofoil.
2. Closer to the surface and deeper within the boundary layer, first is an outer por-
tion where, over a streamwise distance of several boundary-layer thicknesses,
the flow can be considered as inviscid but rotational. In this part of the flow, vis-
cosity contributes to create entropy and, consequently, vorticity, in agreement
with Crocco’s equation connecting the gradient of entropy s with the rotational
vector in a nonviscous flow, as follows:
−−→ ×−→V
T grads = −V rot
Simply stated, this layer is a region of variable stagnation pressure and stagna-
tion temperature. Because the flow is considered inviscid, the stagnation con-
ditions are constant along streamlines because entropy is a transported quan-
tity. The static pressure is constant across the boundary layer; hence, the layer
behaves like an inviscid flow through which the velocity – and, hence, the Mach
number – decreases steadily from the outer value Me at the boundary-layer edge
(y = δ) towards zero at the wall.
3. The third layer is in contact with the wall and, to ensure the transition between
the previous region and the surface, viscosity again must play a role. This vis-
cous layer must be introduced to avoid inconsistencies because it is not possible
for a nonviscous flow to decrease its velocity without an increase in the static
2.3 On the Structure of a Boundary-Layer Flow 25
pressure; at the wall, the stagnation pressure is equal to the static pressure (i.e.,
the velocity is equal to zero because of the no-slip condition).
The structure described here was first suggested by Sir James Lighthill [7] in a
famous article published in 1953. A more formal justification was proposed in 1969
by Stewartson and Williams [8] for the case of a laminar boundary layer using an
asymptotic expansion approach. They introduced the triple-deck terminology to des-
ignate such a structure: The outer deck is the outer irrotational flow, the middle
deck is the inviscid rotational layer, and the inner deck is the viscous layer in con-
tact with the wall. This representation is valid only if the viscous forces have not
contributed to modify the entropy level of the boundary-layer streamlines (except
in the inner deck). This implies that the time scale of any phenomenon considered
with this approach is short compared to the time scale over which the viscous terms
take effect. This is the case for SBLIs, in which the shock imparts a sudden retarda-
tion to the flow. Such a model is also valid for a rapid acceleration, as in the centred
expansion wave that can occur at the base of a vehicle. In a turbulent boundary
layer, the middle deck represents the greatest part of the boundary layer, even at a
moderate Mach number, so that the behaviour of an interaction can be described
(for the most part) by considering a perfect-fluid model. However, for reasons cited
previously, such an inviscid model becomes inadequate close to the wall, where vis-
cosity must be considered.
adequate turbulence models for SBLIs, especially for the inception part of the
process.
The distinctive feature of a normal shock is that it decelerates the flow without
imparting deflection to the velocity vector – the Mach number behind the shock
is subsonic. However, in most practical cases, the shock is not perfectly normal
(i.e., the interferogram in Fig. 2.19c); rather, it is a strong oblique-shock solu-
tion to the Rankine-Hugoniot equations – even if the shock intensity is very
weak as in transonic flows! In these situations, the velocity deflection through
the shock is so small that the shock is said to be normal. Normal or near-normal
shocks are found in channel flows (e.g., turbomachine cascades, air intakes, and
supersonic diffusers), in shock tubes, and over transonic profiles in which a near-
normal shock terminates the supersonic pocket. Interactions in which the down-
stream flow is totally (or partially) subsonic lead to specific problems. These are
of special interest because of the possibility that downstream disturbances can
influence the shock and initiate an interactive process, which then can result in
large-scale unsteadiness involving the entire flow (e.g., in transonic buffeting or
air-intake buzz).
4. An oblique shock is produced if a supersonic flow encounters a change in pres-
sure (e.g., at the exit of an overexpanded nozzle). In this case, the pressure dis-
continuity induces a flow deflection; whereas in cases 1 and 2, the pressure dis-
continuity is induced by the deflection (Fig. 2.19d). This is the mirror problem
of the duality [deflection–pressure jump].
5. When a flow encounters an obstacle such as a forward-facing step, it separates
upstream of the step. The extent of the separation region is a direct function of
the step height. When the Mach number is supersonic, a shock wave forms at
the separation location.
28 Physical Introduction
Concerning the response of the boundary layer to the shock, there are no basic
differences between all of these situations – except perhaps case 4 in which the
interacting flow communicates with an atmosphere. Therefore, we do not dis-
tinguish among cases when discussing the viscous-flow behaviour in subsequent
sections. The major distinctions are between interactions with and without sepa-
ration.
∂u ∂u ∂ 2 ∂ dp ∂τ
ρu +ρv = ρu + (ρuv) = − + ,
∂x ∂y ∂x ∂y dx ∂y
where ρ is the density, u and v are the x-wise and y-wise velocity components (y
is normal to the wall), p is the pressure, and τ is the shear stress. The central part
of this equation expresses the streamwise derivative of the flow momentum. We see
that an adverse pressure gradient tends to make the momentum along the boundary-
layer streamlines decrease. In a flat-plate boundary layer, the shear stress is nearly
constant close to the wall and then steadily decreases towards the boundary-layer
edge. Thus, in the first part of an interaction process, the boundary-layer flow is
retarded under the combined action of the shock-induced pressure gradient and
the shear stress; the retardation is more important in the near-wall region because
of lower momentum. The result is a velocity profile presenting an inflection point
with an inner part in which the derivative ∂τ/∂ y is positive and an outer part in
which ∂τ/∂ y is negative. In this situation, illustrated in Fig. 2.20, the shear stress
in the boundary-layer inner part counteracts the action of the pressure gradient
by transferring momentum from the outer high-velocity to the inner low-velocity
regions.
2.4 Shock Waves and Boundary Layers: The Confrontation 29
Momentum
Adverse pressure
Shear forces
In this equation, the shear stress is neglected and the inner boundary is assumed to
be very close to the wall (δ << δi ). As shown herein, such a situation corresponds to
a turbulent boundary layer in which the shear stress has a minor role across most of
the boundary layer. This simplified relationship highlights the competition between
the flow momentum and the pressure gradient. Of course, this analysis is wrong if
the normal component v – or, more precisely, its derivative ∂v/∂ y – and the shear
stress are no longer negligible. This is the case in the inner part of a turbulent bound-
ary layer and over almost the entire thickness of a laminar boundary layer. Also, the
situation changes when the shock induces separation of the boundary layer, as dis-
cussed in the next section. Then, the viscous contribution is essential.
A major cause of the differences observed between laminar and turbulent flows
is in the velocity distribution in the boundary layer. During an interaction, the flow
behaviour is dictated by the resistance of the boundary layer to the pressure jump
imparted by the shock; therefore, it is clear that a turbulent boundary layer having a
‘full’ velocity distribution – and thus carrying more momentum – will react less than
a laminar boundary layer the profile of which is far less full. This is illustrated by the
velocity profiles plotted in Fig. 2.12, which are representative of flat-plate laminar
and turbulent boundary layers. The incompressible-shape parameter Hi (see Sec-
tion 2.3.1) characterises how full the boundary-layer profile is; a low value indicates
a fuller profile. For a flat-plate boundary layer, Hi is close to 2.5 in laminar flow and
ranges from 1.3 to 1.4 for turbulent flows. When a boundary layer is submitted to an
adverse pressure gradient, the retardation causes the shape parameter to increase;
this effect is more pronounced in the parts of the layer where the velocity is low. In
addition, the interaction mechanism depends on the thickness of the subsonic part
of the boundary layer. Thus, a laminar boundary layer the subsonic layer of which
is proportionally thicker will respond differently to a turbulent layer with a much
thinner subsonic layer for the same outer Mach number.
30 Physical Introduction
0.03
0.02
.015
0.01
0.005
Tw
0
1000 2000 3000 4000 5000
Figure 2.21. The variation of the sonic-point location with the wall temperature for a turbu-
lent boundary layer at Mach 10, stagnation temperature of 3,000K.
greatly depends on the way the shock is generated and the Mach number. If it
increases to hypersonic levels, distinctive phenomena occur due to the intensity of
the shocks and their small inclination relative to the velocity vector, as well as due to
the high enthalpy level of the outer flow. This is discussed briefly later in this chapter
and in greater depth in Chapter 6.
subsonic part of the boundary layer. This leads to a spreading of the wall-pressure
distribution over a distance on the order of the boundary-layer thickness, compared
with the purely inviscid-flow solution. As shown in Fig. 2.24, the pressure starts to
rise upstream of the inviscid pressure jump, after which it steadily increases and
tends towards the downstream inviscid level. In this case, the viscous (or real) solu-
tion does not depart far from the purely inviscid solution. Accounting for the viscous
effect would be a mere correction to a solution that is already close to reality. Such
behaviour is said to be a weak interaction process in the sense that the flow is weakly
affected by viscous effects. The dilatation of the boundary-layer subsonic region is
experienced by the outer supersonic flow, which constitutes the major part of the
boundary layer if the flow is turbulent. It acts like a ramp inducing compression
waves (η) that coalesce to form the reflected shock (C2 ). The thickness of the sub-
sonic layer depends on the velocity distribution; hence, a fuller profile – which has
a thinner subsonic channel – also has a shorter upstream-influence length. In addi-
tion, a boundary-layer profile with a small velocity deficit has a higher momentum
and, therefore, greater resistance to the retardation imparted by an adverse pressure
gradient.
C1 Reflected shock C2
Incident shock
1 2
y
Me waves
waves
waves
ondes
Sonic
ligne line
sonique
M 1 Subsonic layer
Viscous sublayer
Figure 2.23. A sketch of a turbulent shock reflection without boundary-layer separation.
2.5 Interactions without Separation: Weakly Interacting Flows 33
p
Interaction origin
p2
Viscous flow
p1
X
Upstream influence
Shock envelope
Incident shock
Subsonic layer
Subsonic la
(a ) (b ) c)
Figure 2.25. Shock-wave refraction into a rotational layer with reflected expansion waves: a,
wave pattern; b, corresponding shock-polar diagram; c, wave pattern for a fuller boundary-
layer profile (Henderson [9]).
34 Physical Introduction
Incident shock
Reflected shock
Fig. 2.26. Method of characteristic calculation of a shock reflection in a rotational layer. Wave
system and shocks. Turbulent boundary-layer profile (upstream Mach number 4, primary
deflection –6◦ )
both high accuracy (i.e., the shock is fitted) and a picture of the wave’s propagation
in the supersonic flows. Calculations were made for a turbulent velocity distribu-
tion represented by Coles’s analytical expression (see Section 2.3.1), with the outer
Mach number equal to 4. The part of the boundary layer of which the Mach number
is less than 1.8 was removed (i.e., this cut-off distance from the wall was chosen to
avoid singular shock reflection). The behaviour of the viscous sublayer is neglected,
which is justified for moderate shock strengths at high Mach numbers. The calcula-
tion corresponds to the reflection on a rectilinear wall of a shock producing a down-
ward deflection of –6 degrees in the outer irrotational stream. The characteristic
mesh represented in Fig. 2.26 shows the bending of the shock through the rotational
layer and the waves coming from the wall downstream of the reflection. The wall-
pressure distribution plotted in Fig. 2.27b shows that the pressure first jumps at the
impact point to an intermediate value and then progressively reaches the constant
level that corresponds to shock reflection in a Mach 4 uniform flow. This behaviour,
which is observed in high-Mach-number flows, thus can be interpreted by inviscid
arguments. At a lower Mach number, below 2.5, an overshoot is observed in the
wall-pressure distributions, which cannot be explained simply by rotational effects.
In these circumstances, the influence of the subsonic layer close to the wall as well
as the viscous inner layer can no longer be neglected, and a purely inviscid analy-
sis captures only part of the solution. The contours in Fig. 2.28 confirm that behind
the shock, there is a static-pressure decrease from the outer flow down to the wall.
The analysis proposed by Henderson and the method of characteristic calculations
are instructive because they describe the complex wave pattern that is generated
when a shock traverses a boundary layer considered as a rotational inviscid stream.
However, this scenario does not consider the upstream transmission through the
2.5 Interactions without Separation: Weakly Interacting Flows 35
0.06 0.02
p p
p st 0 p st 0
Rotational Rotational
0.05
characteristics 0.015 characteristics
X X
0.03
0 2 4 6 8 0.005
5 10 15
a – ramp induced shock (upstream Mach number b – shock reflection (upstream Mach number 4-
2.85 - wedge angle deflection 6°) primary deflection -6°)
Figure 2.27. Method of characteristic calculation of a shock reflection in a rotational layer.
Entropy gradient effect on the wall-pressure distribution.
subsonic part of the boundary layer with the subsequent generation of compression
waves, which coalesce to produce the reflected shock.
C1
y
Waves Figure 2.29. Ramp-induced shock with-
M
Mee
out boundary-layer separation.
Sonic line
ligne sonique
Subsonic layer
M 1
Viscous sublayer A
has remarkable agreement with the shadowgraph of the flow (Fig. 2.31). The wall-
pressure distribution in Fig. 2.27a and the static-pressure contours in Fig. 2.32
exhibit the same features as in the shock reflection.
M1 2.85
Figure 2.30. Method of characteristic calculation of a ramp-induced shock. Wave system and
shock. Turbulent boundary-layer profile (upstream Mach number 2.85, ramp deflection 8◦ ).
2.5 Interactions without Separation: Weakly Interacting Flows 37
Oblique shock
M1
Boundary layer edge
Figure 2.31. Shadowgraph visualisation of ramp-induced shock in a Mach 2.85 flow (Settles
[10]).
M1 2.85
Figure 2.32. Method of characteristic cal-
culation of a ramp-induced shock. Static
pressure contours. Turbulent boundary-
layer profile (upstream Mach number 2.85,
ramp deflection 8◦ ).
8
C1
M 1
Compression waves M 1
Me
Figure 2.34. Normal shock interaction
without separation. Sketch of flowfield. M 1
Sonic line
M 1
p p
Reattachment
X0 X0
X LS X
L0
a – upstream interaction length b – separation length
Figure 2.35. Characteristic lengths of a supersonic interaction.
normal shock. It then rises for the streamlines that have travelled through the lower
part of the interaction, where the compression is essentially isentropic; finally, it
rapidly decreases close to the surface on penetrating the boundary-layer inner part.
This typical stagnation-pressure (or pressure-loss) distribution has great importance
when considering SBLI control.
Transonic SBLIs have a specific character because of the partly subsonic nature
of the outer flow. The subsonic part causes the phenomenon to be dependent on the
downstream conditions; however, this is not the case for purely supersonic interac-
tions, which are ‘protected’ by the supersonic nature of the outer inviscid flow that
prevents the upstream propagation of disturbances. Transmission, of course, exists
through the subsonic inner channel of the boundary-layer flow close to the wall;
however, for the turbulent case, the upstream transmission length is very short.
If we presuppose that any typical streamwise length L0 will scale with the incom-
ing boundary-layer thickness or its displacement thickness, there remain the three
farther parameters – M0 , Reδ0 , and ϕ1 – that can influence the interaction. If we
focus our attention on the dimensionless length L/δ0 , then for a fixed value of
Reδ0 :
2.6 Interaction Producing Boundary-Layer Separation: Strongly Interacting Flows 39
For a turbulent boundary layer, the influence of the Reynolds number on L/δ0 is less
clear. However, it is agreed that L/δ0 increases when Reδ0 increases at a moderate
Reynolds number and decreases with Reδ0 at a higher Reynolds number, with the
changeover between the two tendencies occurring at around Reδ0 ≈ 105 . In fact, the
Reynolds number can be a misleading parameter in that it characterises two com-
peting effects. Considering a flat-plate turbulent boundary layer, a high Reynolds
number means reduced importance of the viscous forces. Hence, the boundary layer
has reduced resistance to an adverse pressure gradient; thus, the interaction length
tends to increase. Conversely, as illustrated in Figure 2.15, a high Reynolds number
means a fuller velocity profile (i.e., lower shape-parameter Hi ) and, hence, higher
momentum of the boundary-layer flow as well as a decrease in the subsonic-channel
thickness. Thus, the interaction length tends to contract (Settles [10]). In the first
case, the Reynolds number determines the character of the local flow; in the second
case, it characterises a history effect.
1 10
M1 1.95 Figure 2.36. Schlieren visualisation of an
incident-reflecting shock at Mach 1.95 (ONERA
document).
C1 C3
3
Reattachment shock
1
5 Expansion waves
1 H Slip line
C2
Separation shock C4 4
2
Compression waves 2 Compression waves
Sonic line
S
S R
Subsonic layer Separated bubble Dividing streamline
Figure 2.37. Sketch of the flow induced by a shock reflection with separation.
2.6 Interaction Producing Boundary-Layer Separation: Strongly Interacting Flows 41
S R
p p2
R
Second pressure rise
S
Inviscid solution First pressure rise
at separation
p1
X
Interaction origin
isobaric region at pressure p2 . This is isolated from the outer supersonic stream by a
slip line, which is now an isobaric (i.e., constant-pressure) boundary (Fig. 2.39). This
representation can be useful in improving our understanding of the strong coupling
between the separated boundary layer and the outer flow that leads to the formation
of specific shock patterns. If the separated region is sufficiently large, the pressure
in it is equivalent to the plateau pressure of the free-interaction theory (see Section
2.7.1). Under these conditions, the pressure rise at separation and the initial part of
the interaction do not depend on the downstream conditions, even for the transonic
case. The effect of the downstream constraints is to fix the location of separation,
not its nature.
Because plateau pressure p2 is higher than upstream pressure p1 , the isobaric
frontier ( f ) (i.e., the slip line) limiting the isobaric region starts at the ‘inviscid’ sepa-
ration point with an angle, inducing what is called separation shock (C2 ). This shock
intersects incident shock (C1 ) at point H, where (C1 ) undergoes a deflection (i.e.,
refraction) to become shock (C4 ); separation shock (C2 ) similarly becomes shock
(C3 ). Shock (C4 ) meets the isobaric frontier at point I. There, to ensure continuity
C3
C1 3 5
Expansion fan
1
Slip line
M1 H
Figure 2.39. The inviscid flow pattern
1
associated with shock reflection and C4 4
2
separation. I C5
C2 2 Isobaric boundary (f)
S R
42 Physical Introduction
10
p
3
p1 9 2
7
1
4 5 4
C4
3 C3 2
3
2
C2
C1 1
1
-30 -20 -10 0 10 20 30
Figure 2.40. Shock-pattern interpretation in the shock-polar diagram. upstream Mach num-
ber 2.5. Separation shock deflection 14◦ ; incident shock deflection –10◦ .
of pressure, the pressure rise produced by (C4 ) must be compensated for by a cen-
tred expansion emanating from I. This expansion provokes a deflection of isobaric
frontier ( f ), which is turned towards the wall such that the impact is at the ‘inviscid’
reattachment point R. There, a new deflection occurs with the formation of reat-
tachment shock (C5 ). In addition, a slip line emanates from intersection point H.
For this case, the two-shock system of the perfect-fluid oblique-shock reflection –
which comprises simply an incident plus reflected shock – is replaced by a pattern
involving five shock waves.
The pattern, made by shocks (C1 ), (C2 ), (C3 ), and (C4 ), is a Type I shock-shock
interference according to Edney’s classification (see Section 2.2.3), which can be
understood best by considering the shock-polar representation shown in Fig. 2.40.
The figure corresponds to an incoming uniform flow of Mach number 2.5. The
separation-shock deflection is given by a turbulent-separation criterion, which fixes
the Mach number behind the separation shock (see Section 2.7.2). This angle, which
is approximately 14 degrees, does not depend on the intensity of the shock that
caused the separation. Polar (1 ) is associated with upstream uniform state 1 and
represents any shock forming in state 1, specifically incident shock (C1 ). The image
of downstream flow 3 is point 3 on (1 ), and the deflection imparted by (C1 ) is neg-
ative (i.e., the velocity is deflected towards the wall). Separation shock (C2 ) is also
represented by (1 ) because the upstream state is 1. The image of downstream flow
2 is at point 2 on (1 ), and the deflection ϕ2 is upward. The situation downstream
of H is at the intersection of polars (3 ) and (2 ) attached to states 3 and 2, respec-
tively. Their intersection is the image of two states 4 and 5 with the same pressure
( p4 = p5 ) and the same direction (ϕ4 = ϕ5 ); hence, they are compatible with the
Rankine-Hugoniot equations. The set of successive shocks (C1 ) + (C3 ) is different
from the set (C2 ) + (C4 ); therefore, the flows that traversed each set have undergone
2.6 Interaction Producing Boundary-Layer Separation: Strongly Interacting Flows 43
3
C1 C3
T1
5
7 2
1 C5 Slip lines
Figure 2.41. Shock reflection with sin- 6
gular shock intersection or Mach phe- 1
different entropy increases. Thus, slip line () is formed separating flows 4 and 5,
which have different velocities, densities, temperatures, and Mach numbers (but
identical pressures). In a real flow, a shear layer develops along () thereby ensur-
ing a continuous variation of the flow properties between states 4 and 5. The fluid
that flows along a streamline passing under point H and belonging to the inviscid
part of the field crosses three shock waves: (C2 ) and (C4 ) plus reattachment shock
(C5 ). Thus, the final entropy level is lower than for the entirely inviscid case, in which
the fluid would have traversed only the incident plus reflected shocks. This is also
the case for an interaction without separation, which is close to the inviscid model at
some distance from the wall. The conclusion is that entropy production through the
shock system is smaller in a shock-induced, separated interaction than in an inter-
action without separation or in the limiting case of the inviscid model. This result is
exploited by control techniques that aim to reduce wing drag or efficiency losses in
internal flows.
If for a fixed upstream Mach number the strength of the incident shock is
increased, a situation is reached in which the two polars (2 ) and (3 ) do not inter-
sect. Then, an Edney Type II interference occurs at the crossing of shocks (C1 ) and
(C2 ) and a near-normal shock, or a Mach stem, is formed between the two triple
points T1 and T2 , as shown in Fig. 2.41. The singular shock interaction in Fig. 2.42 is
for an upstream Mach number of 2.5; the separation shock deflection of 14 degrees
results from the separation criterion. The Mach reflection is obtained by increasing
the incident-shock deflection. Downstream states 4 and 6 located at the intersec-
tion of polars (1 ) and (2 ) are separated in the physical plane by slip line (1 ),
whereas downstream states 5 and 7 at the intersection of (1 ) and (3 ) are separated
by slip line (2 ). The subsonic channel downstream of Mach stem (C5 ) is acceler-
ated under the influence of the contiguous supersonic flows such that a sonic throat
appears after which the flow is supersonic (Fig. 2.41). In this case, the interaction
produces a completely different outer-flow structure with the formation of a com-
plex shock pattern replacing the simple, purely inviscid-flow solution. The occur-
rence of a Mach phenomenon can be detrimental in hypersonic air-intakes because
the stagnation-pressure loss behind the normal shock is much greater than behind
the oblique shocks.
44 Physical Introduction
p
p1
5 2
3 7 6
5 4
4 1
C3
C4
3
C5
3
2
2
C1 C2
1
1
-20 -10 0 10 20
Figure 2.42. Shock reflection with singular shock intersection or Mach phenomenon: Situa-
tion in the shock-polar plane (M1 = 2, ϕ1 = −16◦ , ϕ2 = 14◦ ).
Slip line
Figure 2.43. Schlieren visualisation of
a ramp flow with laminar separation
Separation shock at Mach 5 (ramp angle 15◦ , Reynolds
number Re L = 1.5 × 105 (ONERA
document).
Reattachment shock
2.6 Interaction Producing Boundary-Layer Separation: Strongly Interacting Flows 45
C3 Slip line
3
1 T
6
4
Separation shock Reattachment shock
C2
C 2 2
S
R
Dividing streamline
S
Sonic line
Figure 2.44. Sketch of the flow induced by a ramp with separation.
separation point and a reattachment shock from the inviscid reattachment point. If
the Mach number is high enough, the two converging shocks intersect at a small
distance from the wall, giving a Type VI interference, as illustrated in Fig. 2.45. The
situation at point T where the two shocks (C1 ) and (C2 ) meet is represented in the
shock-polar diagrams in Fig. 2.46. The solution is made of shock (C3 ), which is seen
as the shock induced by the ramp at a great distance from the wall and by states 3 and
6, at the same pressure, and separated in the physical plane by slip line (). Polars
(1 ) and (2 ) are distinct; therefore, an intermediate state 4 must be introduced
between states 2 and 6. In the case shown in Fig. 2.44, polar (2 ) is above polar
(1 ), and downstream-compatible states 6 and 3 are found at the intersection of
(1 ) and polar (1 ), representing an isentropic expansion from state 4. In this case,
a centred expansion emanates from triple point T and propagates in the direction
of the wall, on which it is reflected as a new expansion wave. This situation is the
most common; however, at low supersonic Mach numbers (the limit is around 2;
see Section 2.2.3), the relative position of polars (1 ) and (2 ) change, the first one
now above the other in the region of interest. In this case, compatibility is achieved
through a fourth shock (C4 ), which is very weak. It emanates from T and propagates
towards the wall.
C3
3
1 T 6
Figure 2.45. The inviscid flow pattern
associated with a ramp flow with M1 4
C2
separation. 2
C1
f
140 p
p1
120
2
100
80
60
Expansion polar
40 1
4
3
20
C1 6
C2 C3
1 2
-40 -20 0 20 40 60 80
40
p
p1 2
36
4
32
6
28
Expansion polar
3
24
20
30 35 40
ramp shock
M1 2.96
shock reflection
25 X X1 mm
Figure 2.47. Shock- and ramp-induced separation of a turbulent boundary layer. Navier-
Stokes calculations (Shang et al. [12]).
Slip line
into the far-downstream part of the interaction. The sketch of this flow in Fig. 2.49
shows separation shock (C1 ), which is followed by a supersonic region terminated
by near-normal shock (C2 ). In contrast to the previous examples, it is not possible
to identify a shock linked to reattachment. The two shocks that do exist meet at a
point from which slip line () starts. For the outer supersonic stream, the upstream
part of the separated region behaves like a viscous wedge producing the oblique
shock after which the flow is still supersonic. Shock (C1 ) meets the normal shock
(C3 ) – which causes the separation – at point T and results in a pattern similar to the
Edney Type VI interference. The situation at point T in the shock-polar diagram is
shown in Fig. 2.50. Compatibility conditions downstream of T entail the formation
of ‘trailing’ shock (C2 ) represented on shock polar (2 ) attached to state 2. The two
compatible states 3 and 4 are separated by slip line (). At transonic velocity, where
the upstream Mach number M0 ≈ 1.4–1.5, the flow downstream of (C1 ) is weakly
supersonic (M2 ≈ 1.20–1.10). Shock (C2 ) satisfies the strong solution to the oblique-
shock equations but, in fact, the intensity is very weak. Downstream of (C2 ), the flow
may be subsonic or still supersonic with a Mach number close to unity. The remain-
der of the compression is nearly isentropic with a continuous transition to subsonic
C3
1
M 1
T 3 Slip line
M 1 4
2
C1 C2
separation shock
M 1 M 1
M 1
S
S
dividing streamline
R
2.6
p 2.4 4
p1 3
2.2
2 sonic points
(C 3 )
1.8
( 1)
1.6 ( 2) (C 2 )
1.4 2
1.2
(C1 )
1
1
0.8
-10 0 10
a – shock intersection at the triple point with subsonic downstream flow
(M1 = 1.5 ϕ1 = 6 °)
2.6
p ( 1)
2.4
p1
Sonic points
2.2
3
1.8 ( 2)
(C 2 )
2
1.6
1.4
(C 3 )
(C1 )
1.2
1
1
0.8
-10 0 10
b – shock intersection at the triple point with supersonic downstream flow
(M1 = 1.5 ϕ1 = 9°)
velocities, although shocklets often are observed in this region, which is sometimes
called the supersonic tongue. The extent of the supersonic domain behind (C2 )
depends on local and downstream boundary conditions. This structure, typical of
shock-induced separation in transonic flow, is termed a lambda shock pattern. The
shape of the downstream part of the isobaric-separated region is more conjectural
because the reattachment process depends on the coupling with an outer flow, which
primarily is subsonic.
Strong-normal shock interactions are encountered in internal flows such as air-
intakes and missile/space-launcher nozzles operating in overexpending conditions.
Such a flow is shown in Fig. 2.51 by a short-exposure-time shadowgraph of the flow
in a planar supersonic nozzle. For reasons not fully understood, the separated flow
adopts an asymmetric structure; in this case, the separation on the upper wall takes
place earlier than on the lower wall, although the geometry is symmetric (this point
is not discussed here). The separations taking place on each wall are supersonic in
nature, with the Mach number at the interaction origins close to 1.6. In the present
situation, the intersection of the two separation oblique shocks is singular, leading to
a Mach reflection. The large separation on the upper wall results in a lambda shock
pattern, with the flow behind the trailing shock (C3 ) still supersonic (Fig. 2.52). Sep-
aration on the lower wall causes a smaller lambda pattern associated with a reduced
separated region. Here, the intersection of separation shocks (C1 ) and (C2 ) is
Type II with the existence of a Mach reflection made of near-normal shock (C5 )
with the two triple points T1 and T2 . Slip lines (1 ) and (2 ) emanating from T1 and
T2 form a fluidic subsonic channel between two supersonic streams. Due to condi-
tions imposed by the adjacent supersonic flows, the flow in this channel accelerates
until it reaches the sonic speed at a throat (minimum of area). Thereafter, the cross
section of the fluidic channel increases and the expansion continues as a supersonic
flow. The penetration of (C3 ) into the separated shear layer generates a reflected
2.7 Separation in Supersonic-Flow and Free-Interaction Processes 51
R1
S1
M 1
(C1 ) (C 3 )
T1
M 1
(C 5 ) ( 1)
T2 M 1 M 1
(C 2 )
M 1
( 2)
S2
M 1 R2
expansion wave, which is reflected by slip line (1 ) into a wave made of converging
compression waves. In turn, these waves are reflected by the nearly isobaric sepa-
rated region as expansion waves and the pattern is repeated over a certain distance
several times. A similar pattern can be observed near the lower wall. The present
configuration is an example that demonstrates the coupling between the separated
regions and the inviscid part of the flow that generates complex structures because
of the confinement of the flow. Channel flows of this type can occur in supersonic
diffusers, compressor cascades, and propulsive nozzles.
∂u ∂u dp ∂τ
ρu +ρv =− + ,
∂x ∂y dx ∂y
where u and v are the velocity components along x and y, respectively; x and y are
the streamwise and normal coordinates with respect to the surface; p is the pres-
sure; and τ is the shear stress (including the turbulent contribution). The previous
equation written at the wall (y = u = v = 0) gives the following exact relationship
52 Physical Introduction
between the streamwise pressure gradient and the normal shear-stress gradient at
the wall:
dp ∂τ
=
dx ∂y w
An x-wise integration of the previous equation from the interaction origin x0
gives:
x
∂τ
p (x) − p (x0 ) = dx
x0 ∂y w
The physical variables are rendered dimensionless by introducing the following
appropriate scales:
r the dynamic pressure: qo for the pressure
r the wall shear stress at the interaction onset: τw0 for the shear stress
r the boundary-layer displacement thickness at x0 : δ0∗ for the ordinate
r a length L typical of the interaction streamwise extent for the abscissa
Fx
F x
Plateau value Plateau value
1.5 6
1 4 S
S
Value at separation
Value at separation
0.5 2
x x x0 / L x x x0 / L
0 1 2 3 0 1 2 3
a – laminar flow
b – turbulent flow
Figure 2.53. Free-interaction theory. Correlation functions for the pressure rise during
separation.
From the product of equations (2.1) and (2.2) and introducing the correlation
function:
F (x̄) = f1 (x̄) f2 (x̄),
we obtain the following expression for the pressure rise during the interaction:
p − p0 2C f0
= F(x̄) 1/2
q0 M −1
2
0
The dimensionless function F(x̄) exhibits an initial steep rise from the interaction
onset to the separation-point location (Fig. 2.53), followed by a more progressive
rise to reach an asymptotic-plateau value corresponding to the isobaric-separated
region.
The free-interaction theory is confirmed by experiment; the pressure rise at sep-
aration for various conditions is correlated by two functions, for the laminar and
turbulent cases (see Fig. 2.53). The following table gives the values of F(x̄) cor-
responding to the separation point and the plateau value reached in an extended
separation:
As shown, the values of F(x̄) are much higher in turbulent than in laminar flow,
which indicates the important susceptibility of laminar flows to separate. This anal-
ysis establishes that the pressure rise undergone by the boundary layer is of the
following form:
p − p0 −1/4
∝ (C f0 )1/2 M02 − 1
q0
−1/2 γ
In hypersonic flows (M02 − 1)−1/4 ≈ M0 and because q0 = 2
p0 M02 , we have:
p − p0 3/2
∝ (C f0 )1/2 M0
p0
It is usual to take as the streamwise scale the distance L = xs − x0 between the
interaction origin and the separation-point location. Dividing the equations [i.e.,
(2.1)/(2.2)] and considering the separation-point location where x̄ = 1, then:
C f0 M02 − 1 L2 f1 (1)
1= ∗ 2
2 δ f2 (1)
0
From this, we obtain a relationship for the interaction extent L of the following
form:
L ∝ δ0∗ (C f 0 )−1/2 (M02 − 1)−1/4
This equation suggests that the pressure rise pS at separation and the extent of
the first part of the interaction depend only on the flow properties at the interac-
tion onset and not on the downstream conditions, particularly the shock intensity.
During the first part of the interaction, the flow is a consequence of the reciprocal
and mutual influence, or coupling, between the local boundary layer and the invis-
cid contiguous stream, not the further development of the interaction – hence, the
portrayal of this phenomenon is as a free-interaction or free-separation process. This
important result, well verified by experiment, explains many features of interactions
with shock-induced separation.
A major consequence of the interaction is to split the pressure jump pT
imparted by the shock into an initial compression pS at separation, with associ-
ated shock (C1 ), and a second compression pR at reattachment, with the overall
pressure rise such that pS + pR = pT (Fig. 2.54). The extent of the separated
region is dictated by the ability of the shear layer that has its origin at separation
point S to overcome the pressure rise at reattachment. This ability is a function of
the momentum available at the start of the reattachment process. In this analysis,
the pressure rise up to separation does not depend on downstream conditions. Thus,
an increase in the overall pressure rise imparted to the boundary layer or a rise in
the incident-shock strength requires a greater pressure rise at reattachment. This
can be achieved only by an increase of the maximum velocity (Us )max attained on
the dividing streamline; hence, an increase in the shear-layer length is needed to
achieve a greater transfer of momentum from the outer flow. The length of the sep-
arated region therefore grows in proportion to the pressure rise at reattachment and
the separation point moves in the upstream direction.
Because the relative importance of the viscous forces decreases with increas-
ing Reynolds numbers, the free-interaction theory predicts an increase in the
2.7 Separation in Supersonic-Flow and Free-Interaction Processes 55
Reattachment pR
R
Figure 2.54. The splitting of the compres- pT
sion in an interaction with separation and
reattachment.
Separation plateau
S pS
X
interaction extent and a decrease in the overall pressure rise as the Reynolds num-
ber increases. A consequence for both laminar and turbulent regimes is that a
stronger shock is required to separate the boundary layer at lower Reynolds num-
bers than at higher values. This behaviour is confirmed by experiment for laminar
and turbulent flows as long as the local Reynolds number Reδ is less than about 105 .
Above this value, the tendency is reversed and the interaction domain contracts
when the Reynolds number is increased and the turbulent boundary layer becomes
more resistant to separation. This conflict is resolved when we recall that during the
interaction process, the boundary-layer behaviour is a consequence of the action of
both inertia and viscous forces, as explained in Section 2.4.2. The free-interaction
theory favours the viscous forces because it involves only the skin-friction coeffi-
cient. The predominance of viscosity is apparent in laminar flows or low-Reynolds-
number turbulent flows; however, at a high Reynolds number, the momentum
transported by the boundary layer becomes the dominant factor in the interac-
tion with the shock. Because the boundary-layer profile becomes fuller when the
Reynolds number increases, the resistance to the retarding influence of the shock
increases.
1. A criterion deduced from the free-interaction theory, which states that the pres-
sure p1 behind separation is such that:
p1 γ 2 2C f0
= 1 + 6 M0 1/2
p0 2 M −1
2
0
This criterion considers the influence of the Reynolds number through the skin-
friction coefficient.
2. The criterion proposed by Zhukoski [15], which is simply:
p1
= 1 + 0.5M0
p0
where there is no Reynolds number influence.
3. For separation in rocket-engine nozzles, the Schmucker [16] criterion is used
frequently:
p1
= (1.88M0 − 1)0.64
p0
For reasons already discussed, special attention is focussed on transonic inter-
action. The limit at which normal shock-induced separation can occur is of great
importance in airfoil design and in the mechanism of unsteady phenomena such as
buffeting and air-intake buzz (see Section 2.11). Incipient separation in turbulent
flow occurs when the Mach number upstream of the shock reaches a value close to
1.3. This is nearly independent of the boundary-layer incompressible shape param-
eter Hi , which appears to contradict the previous discussion. In reality, because a
higher Hi entails a spreading of the interaction domain and, hence, a weakening of
the adverse pressure gradient, separation is postponed compared with low Hi situ-
ations in which the pressure gradient is more intense. The two opposite tendencies
compensate such that the limit for shock-induced separation essentially depends
only on the upstream Mach number.
M0 2
p Cf
p
Cf
p st 0.2
flat plate
p st 0.002
Cf
0 Shock location 0 X(mm)
0 50 100
0 1 2 3 4 5 105 Rex
M0 2
p 0.4 0.00
Cf p Cf
p st Cf flat plate p st
0.2 0.00
Shock location
0 0
0 50 100 X(mm)
0 2 3 4 Rex
1 5 10
transition
M0
T T T
Separation Reattachment
p p p
a Fully laminar b c
X X X
T
T
p p
e
d
Fully turbulent
X X
Figure 2.56. Schematic representation of transition motion on a shock-separated flow (Gadd
et al. [18]).
less resistant to separation at the higher Reynolds number (see Section 2.7.1). As
transition moves farther forward and takes place approximately at the shock-impact
point, the flow can accommodate a steeper pressure rise at reattachment. The con-
sequences of this forward movement are twofold: The streamwise scale of the reat-
tachment domain shrinks and the associated pressure rise, along with the accompa-
nying pressure gradient, is amplified. Conversely, the overall extent of the separated
flow increases.
The situations encountered as this process takes place are illustrated in Fig. 2.56,
in which the effect of a Reynolds-number increase for a reflecting shock of con-
stant intensity is shown. Starting from a fully laminar interaction (a), a rise in the
Reynolds number Re L provokes a displacement of the transition in the upstream
direction until it reaches reattachment region (b). The peak pressure and heat trans-
fer then become much higher than in the fully laminar interaction. At the same time,
a reversal in the Reynolds-number dependence occurs and the separation extent
decreases with increasing Re L. Transition first stays in the reattachment region until
a limit value of Re L is reached, beyond which it suddenly moves to separation region
(d). With further increase, transition occurs upstream of the interaction, profoundly
affecting flow structure (e). For instance, the separated zone may disappear because
the shock strength is no longer sufficient to separate the boundary layer. During
the transitional phase of the interaction, the peak heat transfer at reattachment in
a hypersonic flow can be higher than in the fully turbulent case (c). Such an over-
shoot, also observed during boundary-layer transition over a flat plate, is due to the
existence of large and well-organised structures, denoting a pre-turbulence state and
enhancing transfer mechanisms. Such flow structures then are broken into smaller
eddies when the turbulent regime is established. Most of the observed so-called lam-
inar hypersonic interactions are, in fact, transitional because maintaining a laminar
regime throughout the interaction domain is difficult due to the extreme sensitivity
2.9 Specific Features of Hypersonic Interactions 59
of the separated shear layer to disturbances. This transition produces a mixed inter-
action in which separation has the feature of a laminar flow (i.e., decrease of the heat
transfer), whereas reattachment exhibits turbulent behaviour (i.e., higher pressure
and heat-transfer peaks).
1. When the wall temperature is well below the outer-stream stagnation temper-
ature, a cold-wall situation arises that may significantly affect the interaction
properties.
2. Heat-transfer processes take on dramatic importance, especially in separated
flows where the impact to the surface of reattaching shear layers leads to very
high heat fluxes.
3. Real-gas effects that result from the intense heating produced by the shocks
modify the thermodynamic and transport properties of the fluid (most often,
air) in a way that may influence the interaction.
We saw previously that if the approaching boundary layer is turbulent, the sonic
point is close to the wall. If, in addition, the incident Mach number is increased,
this point moves even closer; therefore, for hypersonic flow, the boundary-layer
subsonic channel is exceedingly thin (see Fig. 2.16). The consequence is that the
upstream propagation distance for any disturbance is very short. Thus, the com-
pression waves formed by the flow deflection in the vicinity of the separation and
reattachment points coalesce very rapidly to produce shock waves from so deep
within the boundary layer that they seem to originate almost from the wall. Also,
because these waves propagate at a small angle with respect to the streamlines, it is
possible for much of the shock pattern to be embedded within the boundary layer, as
shown in Fig. 2.57. The intersection of the separation and reattachment shocks usu-
ally produces a Type VI shock-shock interference pattern with a centred expansion
emanating from triple point T (see Section 2.6.2). The signature of this expansion
on the nearby wall is denoted by a sharp pressure decrease following the rise at reat-
tachment. This is shown in Fig. 2.58, in which the pressure distributions on the wall
are plotted for a ramp-induced interaction for increasing value of the ramp angle.
The Type VI interference also produces a jet of high-velocity fluid that moves in the
direction of the surface and enhances the heat flux downstream of reattachment.
The wall-pressure distributions plotted in Fig. 2.58 illustrate the behaviour
explained in Section 2.7.1. When the wedge angle is large enough to induce sepa-
ration, the pressure rise at separation is considered independent of the value of this
angle. When the intensity of the shock provoked by the wedge increases, the effect is
a displacement of the separation point in the upstream direction. At the same time,
60 Physical Introduction
C3
Slip line
Reattachment shock
T
C2
Separation shock C1
S R
S A
the pressure rise at reattachment increases to reach the downstream level resulting
from the wedge deflection (after an overshoot, as explained previously).
Separation Plateau
Inviscid
0
0 50 X L mm
Figure 2.58. Wall-pressure distribution in a hypersonic ramp flow for increasing ramp angle
(Elfstrom [19]).
2.9 Specific Features of Hypersonic Interactions 61
p
p1 Tw
Adiabatic 0.2
Tst
4
Non-viscous
Figure 2.59. Wall-temperature effect on a laminar 3
high Mach number interaction (Lewis et al. [20]).
1
50 100 X(mm)
vehicle releasing the heat stored during the earlier high-speed part of the reentry
trajectory.) The results show that wall heating results in a lengthening of the interac-
tion domain, which confirms a contrario observations made on a cooled wall (Délery
[22]). There is no unique explanation of the decrease in Ls /δ0 with the wall tempera-
ture (or its increase when Tw is raised). This tendency, however, is in agreement with
the free-interaction theory (see Section 2.7.1) because a lowering of the wall tem-
perature provokes an increase in the skin-friction coefficient and a decrease in the
boundary-layer displacement thickness (i.e., the density level in the boundary layer
is raised): hence, a contraction of Ls . However, the observed dependence on wall
temperature is beyond the scope of what can be predicted by the free-interaction
theory. The contraction of the interaction domain is also a consequence of the thin-
ning of the boundary-layer subsonic part (see Fig. 2.16) because the Mach number
in the inner part of the boundary layer is greater due to lower sound speed. The
same argument holds in reverse for a heated wall because then the sound velocity is
increased and the Mach number is reduced.
The wall temperature also has a more subtle influence through its effect on the
laminar or turbulent state of the incoming boundary layer. Surface-pressure dis-
tributions taken from ramp-flow experiments show that compared to the adiabatic
case, the separated region is more extensive when the model is cooled, with the
pressure plateau forming well ahead of the ramp origin (Délery and Coët [23]). At
the same time, the compression on the ramp is more spread out. At first sight, these
tendencies appear to contradict the previous conclusions. The present behaviour
must be attributed to the fact that in these flows, laminar-to-turbulent transition has
occurred within the interaction domain. Because wall cooling tends to delay transi-
tion, the boundary layer developing on the cooled model is ‘more laminar’ than on
an adiabatic model. For this reason, in the transitional flow, the separated zone is
more extensive when the boundary layer is cooled.
Ramp-induced shock
Boundary layer
Boundary layer
Shear layer
Separation shock
problem, which is crucial for the sizing of the thermal protection in parts of a vehicle
where such interactions are likely to occur, has been studied by many investigators
for both laminar and turbulent flows.
Shadowgraphs taken for two turbulent hypersonic interactions on a compres-
sion ramp at Mach 9.22 are shown in Fig. 2.60. In the lower illustration, the flow is
separated, and it is clear how the shear layer that develops on the upper edge of
the separated region impacts the ramp at a steep angle. Heat transfer is particularly
intense in the vicinity of reattachment, and a sharp rise is associated with the stag-
nation at R of the shear layer developing from the separation point (see Fig. 2.57).
In some respects, the situation at R is similar to a nose-stagnation point, the differ-
ence being that the flow that impinges on the ramp (or control surface) has been
compressed through a succession of oblique shocks at separation and reattachment
instead of a single normal shock. Consequently, the (average) stagnation pressure
is significantly higher, whereas the (average) stagnation temperature is comparable
to that of the outer flow so that the transfer processes are more efficient, causing
higher levels of heat transfer.
The surface-heat transfer most often is represented in a nondimensional form
by the Stanton number defined as:
qw
St = ,
ρ∞U∞ (hst ∞ − hw )
where qw is the wall-heat transfer (in W/m2 ), ρ∞ , U∞ is the density and velocity
of the upstream flow, hst ∞ is the upstream-flow stagnation enthalpy, and hw is the
2.9 Specific Features of Hypersonic Interactions 63
10-2
Stanton Reattachment
number
10-3
10-4
Separation
10-2
Stanton Reattachment
number
10-3
Separation
10-4 Laminar-turbulent
transition
flow enthalpy at the wall. The laminar heat-transfer distribution shown in Fig. 2.61a
was measured at M∞ = 10 in a two-dimensional ramp-induced separation (i.e., the
Reynolds number computed with the distance L from the model’s sharp leading
edge to wedge apex Re L = 2.3 × 106 ). A semi-logarithmic plot is used to emphasise
the phenomena in the first part of the interaction. The heat transfer decreases slowly
on the upstream part of the cylinder, in agreement with the hypersonic strong/weak
viscous-interaction theory (Hayes and Probstein [25]). A more rapid decrease starts
at a location coincident with the separation onset. This decrease is typical of shock-
induced separation in laminar flows. The heat transfer experiences a minimum in
the separated region, then rises during reattachment; the peak value is achieved
downstream of the reattachment point.
Results for a turbulent ramp-induced interaction at Mach 5 for Re L = 107 are
presented in Fig. 2.61b. The initial rise in heat transfer, followed by a slow decay,
64 Physical Introduction
Stagnation
pressure profile
Region of rapid
variation of shock
curvature
Figure 2.62. Entropy layer in a high-Mach-
number flow.
Region of rapid
variation of entropy
rise through the shock
is due to the transition from laminar to turbulent flow in the approaching boundary
layer ahead of the interaction. A second sharp rise occurs at the separation loca-
tion; this behaviour is opposite to that observed in laminar flow and is typical of
turbulent shock-induced separation. This can be explained by the amplification of
turbulence in the vicinity of the separation point and farther downstream. After this
point, the flow then ‘leaves’ the surface and large eddies develop, which promote
exchanges between the wall region and the outer high-enthalpy flow, leading to a
rise in heat transfer. Farther downstream, the heat transfer sharply increases during
reattachment to reach a peak value downstream of the reattachment point.
Cp
Sharp leading edge
0.2 Rounded r = 2.5mm
Rounded r = 5.0mm
0.1
X/L
0
0 0.4 0.8 1.2
Stanton
10 −2 number
Sharp leading edge
Rounded r = 2.5mm
Rounded r = 5.0mm
10 −3
10 −4
X/L
10 −5
0 0.4 0.8 1.2
that supports the ramp. The pressure on the plate upstream part is increased as
a consequence of the Mach number decrease, whereas the pressure level on the
ramp is reduced. With the blunt leading edge, there is a reduction by a factor of 10
of the peak heat transfer at reattachment. There also is a contraction of the sepa-
rated zone, the origin of which – denoted by a decrease in heat transfer – moves
in the downstream direction when the leading edge is rounded. This entropy-layer
effect is linked to the stagnation-pressure loss through the detached shock ahead
of the leading edge. The changes in stagnation pressure due to the rounding of the
leading edge lead to a decrease in the local Reynolds number and, consequently,
greater resistance of the flow to separation in agreement with free-interaction theory
(see Section 2.7.1). This compensates the opposite effect of the local-Mach-number
reduction. In addition, lowering the Reynolds number contributes to maintaining a
66 Physical Introduction
laminar regime throughout the interaction domain, whereas the interaction is tran-
sitional when the leading edge is sharp.
1. Because the thermodynamic properties are not the same, the structure of the
inviscid part of the flow will be modified compared to the constant γ case.
2. Dissociation and chemical phenomena affect the transport properties (i.e., vis-
cosity, heat conduction, and diffusion coefficients), which have repercussions on
the viscous part of the flow.
F
D
S R S R
Y
Z
Outer streamline
Ve Outer velocity
Cross-wise profile
Streamwise profile
τx
τz
βlim X
Z
behaviour of the skin-friction lines in the vicinity of a point where the skin friction
vanishes. This point, termed a critical point, can be of the node, saddle-point, or focus
type according to the behaviour of the skin-friction lines in its vicinity. These points
have the following properties:
r All skin-friction lines pass through a node (Fig. 2.66a) that can be of the separa-
tion or attachment type according to flow direction. An attachment node occurs
at the origin of the skin-friction lines on an object and a separation node is
found where they end. At a node, all of the skin-friction lines except one have
a common tangent.
r An isotropic node, at which the skin-friction lines have distinct tangents (Fig.
2.66b), corresponds to an axisymmetric attachment/separation point.
r Only two skin-friction lines run through saddle point S; all of the other lines
avoid S by taking a hyperbolic shape (Fig. 2.66c). These special skin-friction
lines, called separators, are of the separation or attachment type according to
the flow behaviour in their vicinity.
r All of the skin-friction lines spiral around the focus where they eventually ter-
minate (Fig. 2.66d). A focus is the surface trace of a tornado-like vortex and is
a key feature of separated flow in three dimensions.
r If the flow is two-dimensional or axisymmetric, the focus degenerates into a
centre (Fig. 2.66e).
Critical-point theory, which also can be applied to the velocity field, is a power-
ful tool for rationally describing the organisation, or topology, of three-dimensional
flows. Within this framework, a flow is said to be separated if the skin-friction-line
pattern contains at least one saddle point through which a separation line passes.
The flow in the vicinity of a separation saddle point is illustrated in Fig. 2.67a. The
skin-friction lines run towards the saddle point, where they separate into two fami-
lies flowing along what is called the separation line. On approaching the separation
line, the streamlines close to the surface tend to lift off, flowing in the innermost
2.10 A Brief Consideration of Three-Dimensional Interacting Flows 69
part of the boundary layer and moving up into the outer stream. The separation line
is the trace on the surface of a stream surface (i.e., the separation surface), which
rolls up to form a vortical structure typical of three-dimensional separated flows. In
the attachment process, the direction of the skin-friction lines is reversed; they first
follow what is now called an attachment line and then flow away from the saddle
point (Fig. 2.67b). At the same time, the outer flow dives towards the surface. The
attachment line is the trace of an attachment surface.
70 Physical Introduction
Streamlines
h
Separation line
(l 2 ) (l1 )
Separation saddle point
nn
h
n
Skin friction lines
(f 2 ) (f1 )
Streamlines
(l1 ) (l2 ) h
Separation line
n
Separation saddle point n
h Skin friction lines
(f1 ) (f )
2
Y
α
X λ
Z
M0
3
1
T
(C2 )
(C1 ) 2
4
N2
N1 S2
S1
S3
(S3)
(A2)
(S1) (A1) (S2)
layer emanates before striking the fin leading edge. Information on the interaction
topology is provided by the surface-flow visualisation showing the skin-friction-line
pattern (Figs. 2.70a and 2.70b). Separation saddle point S1 occurs in front of the
obstacle, through which the primary separation line (S1 ) passes. This separates the
skin-friction lines originating upstream from those originating at attachment node
N1 situated behind S1 . A second saddle point S2 exists downstream of S1 with a sec-
ondary separation line (S2 ). For topological reasons, attachment line (A1 ) originates
at node N1 and separates the skin-friction lines flowing towards (S1 ) from those
flowing towards (S2 ). A third separation line (S3 ) is present close to the fin. The flow
topology in the plane of symmetry is illustrated in Fig. 2.71a (for clarity, the verti-
cal scale is greatly increased). The main outer-flow separation line (S1 ) springs from
separation point S1 and spirals around focus F1 . Attachment line ( A1 ) ends at the
half-saddle point coincident with node N1 in the surface. The streamlines flowing
between (S1 ) and (A1 ) disappear into focus F1 . Two other similar structures exist
associated with separation lines (S2 ) and (S3 ). The three foci F1 , F2, and F3 are the
2.10 A Brief Consideration of Three-Dimensional Interacting Flows 73
(S1)
(A1)
(A2)
S1 N1 F2 S2 F1 N2 S3 F3
(Σ3)
(Σ1)
(S1) (Σ2)
traces in the symmetry plane of three horseshoe vortices surrounding the obstacle,
as illustrated in Fig. 2.71b. These vortices can be identified by the spiralling sepa-
ration surfaces (1 ), (2 ), and (3 ), which are associated with the three separation
lines that lie along the surface. The topology of the flow in the planes almost normal
to (S1 ) is similar, the difference being that the lines in question are projected and
not actual streamlines.
(S) (A)
S1 S2
Side wall
b – topology of the skin friction line pattern
the figure. For axisymmetric configurations, the flows are less prone to show three-
dimensional effects. If they do, the periodicity condition imposes an organisation in
which a finite number of critical points of the node and saddle types is distributed
in succession on the reattachment line (Figs. 2.73a and 2.73b). This pattern can be
interpreted as the trace on the surface of Görtler-type vortices, the origin of which
is unclear. These are intensified by the concave curvature effect resulting from the
reattachment of the shear layer. It is probable that these ‘microstructures’ – which
are scaled by the incoming boundary-layer thickness – have a weak influence on the
overall flow organisation. Therefore, in reality, the flows that exhibit these struc-
tures are not far from the axisymmetric idealisation; this is generally not the case for
planar, supposedly two-dimensional configurations.
Saddle point
Node
(S) (A)
High frequencies:
Higher frequencies: Boundary
layer
Figure 2.74. Instant shadowgraph of a separated flow induced by shock reflection. Incoming
Mach number 2.3, incident-shock deflection 8◦ (Dupont et al. [32]).
unsteady configurations (see Section 2.11). The upstream influence of the shock
and the resistance of a turbulent boundary layer to separation depend mainly on
the momentum (see Section 2.4). Thus, one way to limit the shock’s effect is to
increase the boundary-layer momentum prior to interaction with the shock, which
can be done by appropriate boundary-layer manipulation techniques, such as the
following:
r mass injection (or boundary-layer blowing) through one or several slots located
upstream of the shock origin or impact point
r distributed suction applied over a specified boundary-layer length upstream of
the interaction, which lowers the shape parameter, thereby producing a fuller
velocity profile
r removal of the low-speed part of the boundary layer by applying strong suction
through a slot located within or slightly upstream of the interaction region
r use of vortex generators upstream of the shock, which transfer momentum from
the outer high-speed flow thereby reenergising the boundary-layer low-velocity
parts; this transfer enhances resistance to an adverse pressure gradient
r localised suction also can be applied within or in the immediate vicinity of the
interaction
A key factor in controlling the interaction is to determine the velocity that is
achieved on the separating streamline (S). Any action changing its magnitude influ-
ences the interaction and modifies the shear-layer reattachment (see Section 2.6).
If some fluid is sucked through the wall, topological considerations lead to a flow
structure like that illustrated in Fig. 2.77b. The streamline (S2 ), which stagnates at
reattachment point R, originates at a greater distance from the wall, and the velocity
(US )max is greater than in the basic case shown in Fig. 2.77a. This enhances the abil-
ity of the flow to withstand more significant compression and causes a subsequent
contraction of the interaction domain. For the case of fluid injection at low veloc-
ity, illustrated in Fig. 2.77c, the velocity on (S2 ) is reduced because (S2 ) reaches a
lower altitude on the velocity profile. The effect is to lengthen the separated bubble.
However, if the injected-mass flow is increased, there is a reversal of this effect. The
velocity on the lower part of the profile (in particular, (US )max ) is increased if the
mass flow fed into the separated region exceeds a certain threshold.
We also can contract the interaction domain by cooling the wall on which the
interaction occurs because the boundary layer is more resistant to separation on a
cold wall (see Section 2.3). This control technique would be practical for vehicles
using cryogenic fuels (e.g., hypersonic planes and space launchers).
(Us)max
(S)
S R
(Us)max
(Us)max
(S1) (S2)
S1 S2 R2
because, here again, the efficiency loss is the result of entropy production through
the compression shocks and boundary layer.
Active Control
By removing the low-energy part of the boundary layer, the upstream propagation
mechanism is inhibited and the thickening of the boundary layer is reduced. The
flow behaviour for transonic flows then tends towards the perfect-fluid solution with
a near-normal shock extending down close to the surface. Thus, any action that ener-
gises the boundary layer also tends to strengthen the shock because any spreading
caused by the interaction is reduced. This effect is illustrated in Figs. 2.78a and 2.78b,
which show a comparison between a transonic interaction first without and then with
control through a suction slot. In the latter case, the entropy production through the
shock is increased; thus, the wave drag is higher. Conversely, because the down-
stream boundary-layer profile is fuller, the momentum loss in the boundary layer is
reduced.
Passive Control
As described in Section 2.6, when separation occurs, the smearing of the shock sys-
tem and the splitting up of the compression process that is achieved by the inter-
action reduce the wave drag and efficiency loss due to the shock. However, the
momentum loss in the separated boundary layer is far greater than for an attached
boundary layer. Thus, the separation can result in an increase in drag or an over-
all efficiency loss. However, because separation has a favourable effect on the
wave drag, we can envisage replacing a strong but unseparated interaction with
a separated-flow organisation (or one that mimics the separated flow) to gain the
advantages of the low-wave drag. Passive-control schemes have been devised to
exploit these concepts, which combine the two effects by spreading the shock sys-
tem while minimising (in theory) the boundary-layer thickening. The most com-
mon form of passive control involves replacing a part of the surface with a perfo-
rated plate installed over a closed cavity. The plate is positioned so that it is in the
shock region of the interaction. A natural circulation occurs – via the cavity – from
the downstream high-pressure part of the interaction to the upstream low-pressure
part. The upstream transpiration provokes growth of the boundary-layer displace-
ment thickness, which causes an oblique shock (referred to previously as (C1 )). The
situation is similar to the case of the natural shock-induced separation considered
in Section 2.6.3, with the strong-normal shock being replaced by an oblique two-
shock system in the vicinity of the surface; hence, the wave drag is reduced. The
detrimental effect caused by thickening the boundary layer is limited by the suc-
tion operated in the downstream part of the perforated plate. The effect of passive
control is illustrated in Fig. 2.78c, in which the smearing of the transonic shock and
the thickening of the boundary layer are visible. Passive control can be effective
in stabilising a shock, although the advantage in terms of drag reduction is ques-
tionable. This concept can be improved by adding a suction slot downstream of
the passive-control cavity. The device combines the advantages of passive control,
which reduces the wave drag and the effectiveness of suction to lower the friction
losses.
82 Physical Introduction
Wall Contouring
Because friction-drag production in passive control is generally unacceptable and
because fluid suction requires an energy supply that can compromise the econom-
ical benefit of control, mimicking the separated flow structure with a local defor-
mation of the surface can be considered. For example, a bump with a double-wedge
shape could reproduce the flow characteristics of the viscous separated fluid induced
by a shock reflection (see Section 2.6.1). In transonic flow, bumps with a more
progressive upstream concave contour that achieve a nearly isentropic compres-
sion are effective because they weaken the normal shock forming at their location
(Fig. 2.78d). This only slightly affects the boundary layer while substantially reduc-
ing the wave drag.
outer flowfield tends to isolate the interaction domain from downstream perturba-
tions; such interactions are mostly free of large-scale instabilities.
The physics of perfect-gas SBLI in two-dimensional flows is considered to be
well understood. However, there is no room for complacency because nearly all of
the practical situations are three-dimensional. The obstacle is then the difficulty in
comprehending the structure of three-dimensional flows so as to arrive at a con-
sistent topological description of the flowfield organisation. At the same time, the
definition of separation in three-dimensional flow is far more subtle than in two-
dimensional flows, where it is associated with the skin-friction coefficient being
reduced to zero. The description and study of three-dimensional SBLIs first must
consider this basic question, which necessitates calling on the critical-point theory.
Predicting turbulent interactions is still limited because theoreticians are con-
fronted with the frustrating problem of turbulence modelling. The usual two-
transport equation models perform poorly as soon as a noticeable separated region
forms. Turbulence in interacting flows involves many aspects: compressibility terms
in the time-averaged equations, shock/turbulence interaction, history effects, flow
unsteadiness, strong anisotropy, and transfer processes, to name only the more
important – to which turbulence/chemistry coupling should be added for hypersonic
flows. Substantial improvements result from using Reynolds Stress Equation mod-
els transporting the full Reynolds stress tensor or models using nonlinear explicit
expansion of the Reynolds stress tensor in terms of the strain and vorticity ten-
sors (the Boussinesq law is the first term of this expansion). Also, modelling based
on the capture of the unsteady travelling big eddies, such as Large Eddy Simula-
tion (LES) and Detached Eddy Simulation (DES), results in substantial improve-
ment by providing a more faithful representation of the turbulent behaviour in the
shock-separated region. However, we must be cautious when drawing overly pes-
simistic conclusions from a poor agreement between theory and experiment. Many
of the existing results are instructive for the physical understanding of interactions;
although they appear unsatisfactory from code-validation assessments, the discrep-
ancies are frequently attributable to poorly identified flow conditions, undesirable
side effects, or unwanted transition occurring in the course of the interaction.
n] = 0
[ρ V.
84 Physical Introduction
Momentum equation:
V.
[V]ρ n + [ p]n = 0
Energy equation:
n=0
[hst ]ρ V.
The above equations are satisfied by several types of discontinuity and not exclu-
sively shock waves. We consider the various possibilities:
·n
1. If, V = 0 then the velocity component normal to () is zero. The component
of the momentum equation in a plane tangent to () gives:
V
[V]ρ ·n
= [Vt ] × 0 = 0
This shows that there is no condition on the velocity component tangent to the
surface, which can be discontinuous. The normal projection of the momentum
equation gives:
0 + [ p] = 0
which means that the pressure must be the same on each side of ().
Similarly, the energy equation shows that:
0 × [hst ] = 0
Thus [hst ] is arbitrary, as are also the temperature and the density (hence
entropy). Such a surface is a vortex sheet or, in two-dimensional flow, a slip
line.
2. If Vt = 0 one obtains the special case a contact surface as in one-dimensional
unsteady flow.
n
= 0, the normal velocity component is non-zero and there is a mass flux
3. If V.
through the surface which is of the shock wave type of discontinuity. In this case,
one has the following relations:
[Vt ] = 0: the tangential velocity component does not change through the shock.
[Vn ] ρVn + [ p] = 0 for the normal component of the momentum equation.
The energy equations now gives [hst ] = 0, which establishes that the stagnation
enthalpy remains constant when crossing a shock wave.
s2 − s1 ≥ 0
Such a global condition avoids invoking the Navier-Stokes equations because the
dissipative terms, including viscosity and thermal conductivity, impose the correct
sense to the variation in entropy. Shock-structure analysis (applied to a shock nor-
mal to the direction x) leads to the following expression for the change in the
References 85
entropy flux through the shock region, which is a zone of rapid but continuous vari-
ation of the flow properties (Délery [38]):
+∞ 2
1 4 du dT 2
m (s2 − s1 ) = 2
μT +λ dx,
−∞ T 3 dx dx
where s is the specific entropy, m = ρu is the constant mass-flow rate (per unit sec-
tion), μ is the molecular viscosity, λ is the thermal conductivity, x is the streamwise
coordinate, u is the x-wise velocity component, and T is the temperature. Because
μ and λ are positive (i.e., the condition imposed to satisfy the Second Law of Ther-
modynamics), the integrated quantities are strictly positive. Thus, the Navier-Stokes
equations predict that an entropy rises through the shock consistent with the law.
REFERENCES
87
88 Transonic Shock Wave–Boundary-Layer Interactions
Wing shock
Compression waves
Sonic line M = 1
Expansion waves
M>1
Shock wave
M <1
M<1
M>1
M>1
p0
Figure 3.4. The link between stagnation-pressure losses and wave drag.
ahead of the shock wave. For this reason, the severity of the SBLI on a typical
airfoil depends on the airfoil shape, the free-stream Mach number, and the angle of
attack. Thicker and more highly curved airfoils, higher free-stream Mach numbers,
and greater angles of attack all cause stronger shock waves. On aircraft wings, the
strength of the shock is kept relatively low by design, in the range of M = 1.1–1.2.
More severe shocks can be encountered under off-design conditions.
Figure 3.5. Pearcey’s models of transonic airfoil separation. Left: Model A, where separation
first originates at the shock. Right: Model B, where separation originates at the trailing edge.
M>1 to engine
Figure 3.6. External compression inlet. M<1
Transonic SBLI
surface pressures over the rear of the airfoil. Similarly, separations (whether shock-
induced or otherwise) also change the pressure distribution over an airfoil, thereby
contributing to pressure drag. Skin-friction drag, however, is affected by the state of
the boundary layer, which can be reduced by maintaining laminar flow over more of
the airfoil surface.
Modern supercritical airfoils are designed carefully to limit the shock strength to
avoid the detrimental effects of SBLIs as well as wave drag. The problems associated
with both of these effects increase rapidly with Mach number; for this reason, shock-
related problems are the major constraint on the cruise-Mach-number or cruise-
lift coefficient. For civil-transonic aircraft, it is also necessary to retain sufficient
margin before any severe adverse effects occur (e.g., shock stall or buffet); typical
design shock Mach numbers for transonic cruise range from M = 1.1–1.2. A better
understanding of SBLIs and control of their adverse effects is the key to improving
transonic-airfoil performance.
One method by which a reduction in shock strength can be achieved is to uti-
lize the compression waves that reflect from the sonic line. These waves decrease
the Mach number in the supersonic region, thereby reducing the shock Mach num-
ber. This process often is referred to as isentropic compression to distinguish it from
shock compression. In principle, it is possible to achieve a shock-free isentropic air-
foil [4] in which the compression at the end of the supersonic region is achieved
entirely by compression waves. In practice, however, this is difficult because the
compression waves are prone to merging into a shock wave.
Diffuser (subsonic)
M>1 to engine
Terminal shock
Transonic SBLI
Oblique shock waves
Figure 3.7. Internal compression inlet.
Separated flow
Figure 3.8. Separation originating from the terminal shock wave in an internal compression
inlet.
3.3 Normal SBLIs in Detail 95
Flow spillage
Finally, any flow separations are also likely to introduce considerable unsteadi-
ness into the flow, which can lead to unacceptable dynamic loads on the engine.
If the terminal (near-normal) shock oscillation is so extreme that it reaches the
converging part of the inlet geometry, it becomes unstable. At this point, it moves
rapidly upstream, making more of the flow inside the inlet subsonic, until it is even-
tually expelled from the intake causing unstart (or buzz, if this phenomenon is peri-
odic), as illustrated in Fig. 3.9. This is comparable to shock stall or shock-induced
buffet on transonic wings; such a violent event is extremely damaging to the engine.
To avoid the problems associated with strong transonic SBLIs in inlets, design-
ers make use of flow control to enable the boundary layer to stay attached even
when the shock waves have considerable strength. The most popular control method
is boundary-layer suction, or bleed. Flow control for transonic SBLIs is discussed in
more detail later in this chapter, but we emphasize here that modern inlets cannot
function successfully without some form of SBLI control.
M>1
M<1
Figure 3.10. Transonic SBLI in a diffuser flow.
96 Transonic Shock Wave–Boundary-Layer Interactions
M>1
M<1 Figure 3.11. Transonic SBLI in a nozzle.
on cambered airfoils or turbine blades, strong surface curvature in the shock region
generally is avoided by design. For this reason, it is convenient to concentrate on
the idealized case of a normal shock wave interacting with a boundary layer on a
flat surface. The effects of surface curvature and three-dimensionality or sweep are
discussed in Section 3.3.4.
In most practical situations, transonic SBLIs tend to occur where the bound-
ary layer is turbulent. In principle, the flow structure is similar for laminar and
turbulent transonic interactions. However, because turbulent boundary-layer pro-
files are much fuller, they also exhibit a smaller upstream influence and interaction
size compared to laminar interactions. Conversely, laminar boundary layers cannot
withstand adverse-pressure gradients very well; therefore, it is rare to see attached,
transonic interactions for laminar flow.
M1 > 1 M2 < 1
Inviscid layer
Interaction region
Viscous layer
3.3 Normal SBLIs in Detail 97
M>1 M<1
Compression waves
Upstream influence
In the free stream, the shock wave constitutes a sudden pressure jump (ignoring
the small thickness of the shock, which is on the order of a few molecular-mean-free
paths); thus, the adverse-pressure gradient is infinitely strong. Because boundary-
layer separation depends primarily on the gradient of adverse pressure rather than
the magnitude of the pressure rise, any boundary layer would separate if the shock
extended unchanged into the boundary layer. In practice, however, this is not the
case due to the way the boundary layer and the shock interact.
Inside the boundary layer, there is a “sonic line” that separates the supersonic
from the subsonic portion of the boundary layer. Below the sonic line, pressure
information can travel upstream; this spreads the pressure increase ahead of the
shock location. In response, the boundary-layer profile changes shape according to
the adverse-pressure gradient experienced. This results in an increase in shape fac-
tor and thus an increase in displacement thickness. Consequently, the streamlines
in the (supersonic) flow above the sonic line are deflected away from the surface,
generating compression waves. The compression waves act to increase the local
boundary-layer-edge pressure, which in turn affects the boundary-layer develop-
ment. Conversely, the compression waves also decrease the Mach number ahead
of the shock, thereby weakening its strength. As a result of this interplay, the flow
settles into an equilibrium position, shown schematically in Fig. 3.13. The sketch
can be compared to a schlieren image of a normal shock–turbulent boundary-layer
interaction at M = 1.3.
The compression waves form in the supersonic portion of the incoming bound-
ary layer from where they propagate into the outer flow. In this region of the bound-
ary layer, the flow behaves almost like an inviscid supersonic flow with a nonuni-
form (and rotational) inflow. Viscosity has little bearing on the flow development
because the interaction process is taking place over a short distance (comparable
to a boundary-layer thickness), which allows viscous forces little time to act. Only
when very close to the wall do viscous forces remain significant – and, of course,
they cause the nonuniformity of the inflow in the first place.
The fact that the shock wave is “smeared” into a series of compression waves
near the boundary-layer edge has important consequences. First, the shock-induced
pressure rise along the surface is spread over a streamwise distance. This reduces
98 Transonic Shock Wave–Boundary-Layer Interactions
0.5
p/p0
0.4
0.3
-60 -40 -20 0 20 40 60
x [mm]
the pressure gradient experienced by the boundary layer. Second, the flow passing
through the smeared shock foot undergoes near-isentropic compression with little if
any shock compression. In transonic interactions, it is commonly observed that the
main shock deteriorates to almost a sonic wave at the boundary-layer edge. Third,
the streamlines ahead of the main shock are deflected away from the surface (in
line with the increase in displacement thickness). The schlieren image in Fig. 3.13
reflects this in the shape of the main shock, which is tilted slightly backwards as the
boundary-layer edge is approached (remaining approximately normal to the local
flow direction).
As more of the shock wave is replaced by compression waves in the shock foot,
the pressure jump through the remaining shock reduces as the wall is approached.
At the boundary-layer edge, the shock is almost completely reduced to a Mach wave
and the surface pressure is close to the critical pressure at sonic conditions ( p/ p0 =
0.528). The remainder of the overall shock-induced pressure rise imposed by the
outer flow occurs in the subsonic region following the shock. Here, flow changes are
gentler due to the nature of subsonic flow and the absence of waves. As a result, the
initial surface-pressure increase is more rapid in the supersonic portion of the inter-
action than in the later part; this is reflected in a break or kink in the surface-pressure
distribution, as shown in Fig. 3.14. Particularly in confined flows (i.e., channel flows),
it can take a long time before the inviscid pressure jump is reached at the wall.
3.3 Normal SBLIs in Detail 99
O 2 1.03
M = 1.3 1.9 A
O 1.02
Normal shock wave 1.8 M= 0.81034 B
Normal Shock
Compression waves
1.7 M= 0.85921 C
1.01
A 1.6 M= 0.92356
p02/p0
p2/p1
D M= 1.2308 M= 1.1469 M= 1.0783
B 1.5 1
M= 0.92912
M= 0.87719
C
1.4
1.3 M= 1.0851 0.99 M= 0.82362 B C
D
1.2 Isentropic
A
M= 1.1734 0.98
1.1 compression
M= 1.2541 M = 1.3 M = 1.3
1 0.97
−8 −6 −4 −2 0 2 4 6 8 0 1 2 3 4 5 6 7
Deflection angle [deg] Deflection angle [deg]
a) flowfield
b) pressure change across isentropic compression followed by normal shock wave
c) stagnation pressure change (isentropic compression followed by normal shock)
Figure 3.15. Static- and stagnation-pressure changes through the smeared shock foot.
The smearing of the shock foot causes the pressure immediately behind the
shock to vary with wall distance. This is illustrated in Fig. 3.15, which shows the
pressure-deflection diagram for an isentropic compression followed by a normal
shock. Here, we consider a Mach number of 1.3, which is at the high end for an
unseparated turbulent interaction.
Far from the wall, the flow experiences a normal shock wave and the associ-
ated pressure jump (marked “O”). In position “A,” there are only a few compres-
sion waves before the main shock, which has reduced minimally in strength. With
decreasing wall distance (moving to “B” and “C”), more of the shock is replaced
by compression waves; this has considerable impact on the overall pressure jump as
well as the local flow direction (which becomes more inclined away from the sur-
face). A compression by isentropic waves alone is close to the actual situation at the
boundary-layer edge (marked “D”). The pressure is only slightly more than half of
the inviscid pressure jump across the main shock (“O”). This variation on postshock
pressures generates a strong pressure gradient in wall-normal direction, which must
be supported by convex streamline curvature in this region and is shown in Fig. 3.16.
Behind the smeared shock foot, other flow parameters such as Mach number
and stagnation pressure also vary, as indicated in Fig. 3.15. The effect is seen clearly
in measurements such as those shown in Fig. 3.17, where the stagnation pressure
behind the smeared-shock system is greater than the free-stream postshock value.
At the boundary-layer edge, the stagnation pressure is almost identical to the inflow
value, which confirms that the compression is achieved almost isentropically.
Although the effect is felt in the inviscid flow, the reduction in stagnation-
pressure losses through the smeared-shock foot is a consequence of the viscous
interaction. This highlights the fact that the viscous forces experienced inside the
boundary layer are only part of the viscous effects in SBLIs. Inviscid-prediction
methods that cannot correctly account for the SBLI therefore are likely to over-
predict shock losses (or wave drag on airfoils).
The near-sonic velocity at the boundary-layer edge makes the development
of secondary supersonic regions behind the shock foot probable. As discussed
previously, this is a region in which convex streamline curvature is expected to
support the vertical pressure gradient as well as because the flow direction has
100 Transonic Shock Wave–Boundary-Layer Interactions
Pressure increasing
M>1
Curved streamlines
Figure 3.16. Pressure gradient and
Compression waves
streamline curvature behind the shock
foot.
Upstream influence
been deflected away from the wall by the interaction and needs to return to the
horizontal. Similar to a transonic airfoil, the combination of convex curvature and
near-sonic velocities can lead to the development of supersonic regions (i.e., “super-
sonic tongues”) in which expansion waves are generated at the surface that reflect
as compression waves from the sonic line. Similar to supersonic regions on transonic
airfoils, supersonic tongues often are terminated by weak secondary shocks. Their
size and magnitude depend on the degree of flow curvature after the shock (which
may be enhanced by confinement effects), the curvature of the boundary-layer dis-
placement surface, and the extent to which the main shock is replaced by compres-
sion waves (which relates to the growth rate of the boundary layer ahead of the
shock). If the supersonic region is relatively short, then the terminating secondary
shock is likely to be in a region where there is continued streamline curvature and
other (but smaller) supersonic regions with additional (and even weaker) shocks
can be observed. This gives the appearance of a string of ever-weaker “shocklets”
just downstream of the smeared shock foot. An example is shown in Fig. 3.18, which
includes a contrast-enhanced schlieren image to emphasize these features.
M<1
Supersonic tongue
M>1
Secondary shock
Secondary shocks
M=1
Region 1 Region 2
Figure 3.18. Unseparated normal SBLI with supersonic tongue and secondary shocklets.
Secondary shocks aside, the main effect of the supersonic tongue on the bound-
ary layer is to delay the pressure increase downstream of the shock so that it can
travel a fairly large distance before the inviscid postshock pressure is reached. This
generally goes hand in hand with a period of relaxation in which the boundary-layer
profile settles down to return to typical flat-plate conditions. Figure 3.19 shows the
variation of key boundary-layer parameters through the M = 1.3 normal SBLI intro-
duced in the previous images. These results were obtained with detailed nonintru-
sive Laser Doppler Velocimetry (LDV) measurements along a series of wall-normal
traverses. Although the LDV technique can provide excellent spatial resolution, it
requires the flow to be steady; thus, it is not completely able to resolve all fine-scale
details. For example, although the supersonic tongue is detected in general, individ-
ual shocklets cannot be observed.
As indicated in Fig. 3.18, the interaction region as a whole can be divided into
two domains: (1) the upstream supersonic region, and (2) the postshock, subsonic
(except for supersonic tongues) region.
3.5
3.0
2.5
δ /δ 0
*
2.0
*
1.5
1.0
0.5
0.0
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
x/δ
b) Incompressible displacement thickness
3.0
2.5
2.0
θ/ θ0
1.5
1.0
0.5
0.0
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
x/δ
c) Incompressible momentum thickness
1.8
1.7
1.6
1.5
Hi
1.4
1.3
1.2
1.1
1.0
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
x/δ
theory (see Chapter 2). Typically, Region I starts a few boundary-layer thicknesses
upstream of the main shock (i.e., the upstream-influence length) and ends where the
boundary-layer-edge velocity and surface pressure indicate M = 1.
Inflow–Shape-Factor Effects
Because of the dependency of the flow in Region I on upstream-flow properties
only, it is possible to make general observations on the size of the interaction.
Whereas the streamwise extent of the interaction is determined by the upstream
influence, the height of the smeared shock foot above the boundary layer is deter-
mined by this length and the Mach angle of the inflow. Therefore, the upstream-
influence length is the critical dimension of the interaction.
There is a clear dependence of the upstream influence on the shape factor of the
incoming boundary layer. A fuller boundary-layer profile features larger velocities
close to the surface, a thinner subsonic layer, and reduced upstream influence. In
contrast, a less-full incoming boundary layer has a thicker subsonic layer and gener-
ally exhibits a larger upstream influence.
If we concentrate on Region I, we can define the upstream interaction length
L* as the distance between the origin of the interaction (i.e., the point where the
pressure at the wall starts to rise) and the streamwise station, where the local pres-
sure is equal to the critical value p* corresponding to a Mach number equal to unity.
The data plotted in Fig. 3.20 show the variation of the scaled interaction length
L∗ /δ0∗ with the upstream Mach number, Reynolds number (computed using the
boundary-layer displacement thickness δ0∗ ), and shape parameter Hi0 . It shows that
for a given value of Hi0 , the incoming boundary-layer displacement thickness is the
correct scaling for the interaction length. Moreover, the ratio L∗ /δ0∗ is insensitive
to the upstream Mach number. The scatter when M0 approaches 1.3 indicates that
the flow comes close to separation (see Section 3.3.2). The invariability of L∗ /δ0∗
with respect to M0 can be understood from the following argument: Raising the
upstream Mach number increases the strength of the perturbation and therefore
its tendency to propagate farther upstream. At the same time, the subsonic part
of the boundary layer, which determines the upstream influence, becomes thinner.
104 Transonic Shock Wave–Boundary-Layer Interactions
M0 M0 STANEWSKY
Hi0 ROODE
δ0 GOBERT et al.
1.4
Separated flow
INGER theory
1.2
Unseparated flow
average experimental limit
1
1.2 1.3 1.4 1.5 Hi 0
7 6 5 4
10 10 10 10 Rδ0 (flat plate)
the pressure gradients “felt” by the boundary layer. On flat surfaces in nominally
two-dimensional flow, shock-induced separation of turbulent boundary layers is
generally observed once the shock Mach number reaches the range of Ms = 1.3–
1.35. There is no single exact Mach number at which separation occurs because this
depends on a number of parameters.
Although we would intuitively argue that less-full boundary-layer profiles, such
as those having experienced roughness or previous adverse-pressure gradients,
would separate more easily, it is those same flow cases that also exhibit the largest
upstream influence and, therefore, the greatest degree of shock-smearing. This
causes a reduced adverse-pressure gradient that, to some extent, offsets the greater
sensitivity of the flow to the pressure rise. In the case of very full boundary layers,
the greater resistance to separation is offset almost exactly by the greater pressure
gradients experienced due to a shortened interaction length. As a result, the onset
of separation is relatively insensitive to changes in shape factor. It appears that the
interaction adjusts to generate an adverse pressure gradient that suits the incoming
boundary layer. Compared to other flow separations – for example, in incompress-
ible flow – the sensitivity to separation in transonic interactions to boundary-layer
shape is significantly reduced.
Much research has been undertaken to predict the onset of incipient separation
in general terms. Most of it is based on the free-interaction theory outlined in Chap-
ter 2, which states that separation occurs when the pressure rise across the shock
wave exceeds the plateau value predicted by the following equation:
p Cf
= 1 + k M∞ 2
1/4 (3.3)
p∞ M2 − 1
∞
Here, the parameter k relates to the dimensionless factor F used in the original free-
interaction theory, as follows:
γ
k = F (x) √ (3.4)
2
For practical purposes, this makes k equivalent to F. Various researchers proposed
values for k or F at separation. In extensive studies of a wide database, Zheltovodov
et al. [6,7] concluded that k ranges from 6 to 7.4. They argued that the onset of
intermittent separation can be predicted by comparing the theoretical pressure jump
across a normal shock wave with the free-interaction theory estimate of plateau
pressure. This is illustrated in Fig. 3.22, where it is shown that the intersection of the
theoretical normal-shock pressure jump (ξ ) with the pressure rise predicted by free-
interaction theory for k values between 6 and 7.4 marks the onset of experimentally
observed intermittent separation (i.e., the shock Mach numbers are around 1.25).
Conversely, the actual wall pressure across a shock wave is always smaller than
the theoretical pressure jump across a normal shock due to interaction effects. The
experimentally observed wall pressures immediately behind the shock for a range
of shock Mach numbers are included in Fig. 3.22. Using the actual wall pressures,
it is demonstrated that fully developed separation is observed when they reach the
pressure rise predicted by free-interaction theory (for k = 6).
In practice, it is desirable to be able to predict the onset of incipient sepa-
ration without having to refer to actual wall-pressure measurements. This can be
106 Transonic Shock Wave–Boundary-Layer Interactions
Pp
P∞
M∞
ξ
2.0 K = 7.4
– 4
– 5 6.0
– 6
– 7
1.5
2 3
γu =
1 γu = 0.008 Ms 0.55
Mi
1.0
1.0 1.2 1.4 M∞
Figure 3.22. Experimentally observed pressure ratio across normal shocks: Region
1 –
unseparated flow; 2 – intermittent separation; 3 – fully developed separated flow
(adapted from Chapter 4, Fig. 4.4).
achieved by noting that the boundary-layer-edge flow just behind the shock foot is
close to sonic conditions, as described previously. Figure 3.23 compares the post-
shock wall pressures of Fig. 3.22 with the pressure calculated for sonic conditions
(p = 0.528p0 ), where the stagnation pressure p0 is assumed to be unchanged from
the free-stream value ahead of the shock wave. Because the main shock is smeared
into isentropic compression waves, this assumption is reasonably valid and it can
be seen that the predicted pressure values follow the experimental data very well.
There is some departure at small Mach numbers because the assumption of a com-
pletely smeared shock foot is no longer justified and at larger shock Mach numbers,
where the presence of significant separation makes this simple model invalid. Nev-
ertheless, for shock strengths near separation (e.g., for turbulent boundary layers),
it is possible to determine the separation onset by equating the theoretical pressure
rise to sonic conditions with the plateau pressure from free-interaction theory. For
M<1
Triple point Slip surface
M>1
Separation Reattachment
M<1
Slip surface
M>1
Supersonic tongue
M=1
a) Sketch of flowfield
b) Schlieren photograph
0.5
0.4
p/p0
0.3
0.2
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
x/ δ
d) Surface pressures
Figure 3.25. Mildly separated transonic SBLI (Ms = 1.5).
110 Transonic Shock Wave–Boundary-Layer Interactions
significantly below the theoretical normal shock value, which is reached asymptoti-
cally only after a long streamwise distance.
Because of the continued pressure increase in the subsonic flow following the
shock system, streamtubes expand, which in turn aids the reattachment process.
If the separation bubble is relatively long, a more rapid pressure increase often is
observed around the reattachment point. Therefore, a separated shock boundary-
layer interaction often is typified by a two-step pressure increase with a rapid
pressure rise until separation, a near-constant pressure underneath the separation
bubble, and a second pressure rise (but not as steep as in the supersonic region)
starting around the reattachment region. This “two-step” pressure rise distinguishes
the wall-pressure profile from that of attached interactions and sometimes allows the
identification of separation from pressure measurements alone. However, when the
separation bubble is small, these features are close together and quite smeared such
that it is often difficult to distinguish between a separated and an attached pressure
distribution.
The presence of a λ-shock structure in separated transonic SBLIs has a notice-
able effect on the stagnation-pressure losses, as shown in Fig. 3.26. Because multi-
ple but weaker shocks in the shock foot incur reduced losses, there is a noticeable
region of increased stagnation pressure between the boundary-layer edge and the
slip surface. In practice, this region is small – typically only a few boundary-layer
thicknesses in height; therefore, it has limited implications on the overall flowfield.
However, the principle of achieving a reduction in losses due to a λ-shock structure
forms the basis of shock control, which is discussed in Section 3.4.
initial boundary-layer thickness. This roughly coincides with the region of steepest
streamwise pressure gradient where there is a general retardation of the flow. Far-
ther downstream, the normal stress contribution becomes negligible.
A second typical feature of turbulence behavior in strong interactions is the lag
that is established between the mean and turbulent flowfields, leading to significant
nonequilibrium. This can be demonstrated by considering a “phase plane” of the
following two variables through the interaction and downstream recovery region:
r the square root √Cτ of the turbulent shear-stress coefficient defined as:
2 (ρu v )max
Cτ = − (3.9)
ρe u2e
1
J =1− (3.10)
Hi
which has the merit of tending to a finite value when the vertical extent of
the separation region increases (then Hi tends to infinity). Empirical corre-
lation laws [10] show that the following variation of the well-known Clauser
parameter:
(Hi − 1) J
G= √ =√ (3.11)
Hi 0.5Cτ 0.5Cτ
remains constant and equates to 6.55 for most equilibrium turbulent boundary-
layer flows, the value of the constant being that of a “well-behaved” flat-plate
boundary layer (see Chapter 2). The following equation:
1
1− = 6.55 0.5Cτ (3.12)
Hi
1
(1− H1 ) = 6.55 Cτ
i 2
−ρ(u′v′)max 2 Reattachment
Cτ = 1 ρ − 2
e ue
2
0.2
rehabilitation
Flow
0.1 1
starting point
2
3
4
Separation
equilibrium boundary - layer
1
0 1− H
0 0.2 0.4 0.6 0.8 1 i
r Thereafter, especially for separated flows, the shear stress and the shape param-
eter continue to increase together in a manner typical for a developing free-
shear layer. This ‘mixing-like’ behavior could continue until J = 1 (i.e., Hi →
∞). During the rapid-interaction process, the separating boundary layer under-
goes such an overwhelming perturbation that the development of the free-shear
layer is not significantly influenced by its initial conditions – that is, the initial
boundary-layer properties (provided it is fully turbulent). Here, turbulence pro-
duction continues in proportion to the growth of the large-scale structures form-
ing near the separation location.
r At some location, the shape parameter J peaks and starts to decrease. This
reversal signifies the onset of the reattachment process for separated flows. Dur-
ing this phase of the interaction, the shear-stress level continues to grow until it
reaches a maximum at a streamwise station that almost coincides with the reat-
tachment point. In the course of the reattachment process, the trajectories in
the phase plane bend and cross the equilibrium locus at a point whose distance
from the origin increases with the size of the separated region.
r Downstream of reattachment (and in the absence of any additional destabilizing
influences), the trajectories return towards a final point located on the equilib-
rium line. During this relaxation process, the flow lies above the equilibrium
line; at some point, a second maximum departure from equilibrium is reached.
Farther downstream, the trajectories collapse onto a common trajectory tending
towards the equilibrium locus.
M>1 Secondary
M<1 shock
M>1
Secondary shock
M>1 Supersonic tongue
Secondary shocklet
section) is necessarily straight (ignoring asymmetric flows for the moment), the
variation in streamline curvature between the boundary-layer edge and the cen-
tre is more rapid than in unconfined flows. This enhances any wall-normal flow
variations and increases the size of any supersonic tongues. Furthermore, the core
flow between the boundary-layer edge and the channel centre experiences a sig-
nificant contraction due to the fact that the boundary layer grows rapidly through
the interaction. This accelerates the subsonic flow behind the shock, which also
serves to make secondary supersonic flow more likely. This is shown schematically in
Fig. 3.30b.
If the blockage becomes too severe, it is possible that the supersonic tongues
from all sides merge into a secondary supersonic flow region spanning the channel,
as shown in Fig. 3.30c. Such a secondary supersonic flow generally features a termi-
nating shock wave, which can be followed by further supersonic regions and shocks.
When there is this close succession of supersonic regions and secondary shocks, it
is called shock train (Fig. 3.31), which can be the cause of significant losses in tran-
sonic internal flows. These losses occur because, on the one hand, the core flow
experiences a succession of (albeit weak) shock waves, each incurring a stagnation-
pressure decrease; on the other hand, the boundary layer is subjected to a succession
of shock waves that cause it to grow rapidly. However, the continued reacceleration
to supersonic conditions reduces the pressure gradient along the wall and it can take
a long streamwise distance for the pressure to reach the asymptotic value. Often
more important than the losses in the flow is the nonuniform and unsteady nature
of shock trains, which can have harmful consequences – for example, if the flow is to
enter an engine. Therefore, it is advisable to avoid this phenomenon when possible.
Generally, the occurrence of shock trains depends on the ratio of boundary-
layer displacement thickness to duct height or displacement area to duct area. Typ-
ically, whenever this ratio is greater than a few percentage points (based on dis-
placement thickness), multiple shocks are likely. It is difficult to establish a general
116 Transonic Shock Wave–Boundary-Layer Interactions
rule of thumb because the critical confinement parameter depends on the oncom-
ing Mach number as well as the Reynolds number. For a good description of such
flowfields at Mach numbers of around 1.6, readers are referred to the experimental
studies by Om and Childs [11] and Carroll and Dutton [12] in circular and rectan-
gular ducts, respectively. Both studies observed shock trains for confinement ratios
δ * /H between 4 and 6 percent, where H is the channel half-height or radius. Con-
versely, when the confinement ratio drops to less than 2 percent, only a single shock
wave was present. Generally, it has been observed that the confinement necessary
to cause shock trains reduces with increasing Mach number because the boundary-
layer growth is accelerated for stronger shock waves.
In strong shock interactions in internal flows, it is possible to observe the
interactions on either side of the channel interfering with one another. Strong
normal-shock interactions are encountered in internal flows such as air-intakes or
missile/space-launcher nozzles operating in overexpanded conditions. An example
is shown in Fig. 3.32, which is a short exposure shadowgraph of strong interactions
in a planar supersonic nozzle [13]. The Mach number at the interaction origins is
close to 1.6, which is strong enough to cause significant separation. The λ regions of
the separations on each nozzle wall are large enough to interfere with one another
and, for reasons not yet fully understood, the flow adopts an asymmetric structure –
despite the fact that the nozzle geometry is symmetrical. Here, the separation on
(C 3 )
(C1 )
T1
(C 5 ) M<1
(Σ 1 )
M=1 M>1
T2
(C2 ) (Σ 2 )
(C 4 )
the upper wall occurs earlier than on the lower wall, and the intersection of the two
oblique separation shocks gives rise to a Mach reflection in the centre of the channel.
The large separation on the upper wall gives rise first to a λ-shock structure, with
the flow behind the rear shock leg (C3 ) still supersonic (Fig. 3.32). On the lower wall,
a smaller λ structure and a reduced separated region are observed. The intersection
of separation shocks (C1 ) and (C2 ) forms an Edney Type II interference (i.e., Mach
reflection) with the Mach stem being a nearly normal shock (C5 ) bounded by two
triple points T1 and T2 . Slip lines (1 ) and (2 ) emanating from T1 and T2 form a flu-
idic subsonic channel between two supersonic streams. Due to conditions imposed
by the adjacent supersonic flows, the flow in this channel accelerates until its reaches
sonic speed at a throat. Thereafter, the fluidic channel cross section increases, with
a further supersonic expansion. Where the rear legs of λ structures (C3 , C4 ) impinge
on the separated shear layers, expansion waves are reflected, which in turn reflect
from slip lines ( 1 , 2 ) as converging compression waves. In turn, these waves are
reflected by the nearly isobaric separated region as expansion waves, and the pattern
is repeated over some distance.
This configuration demonstrates the complex flow structures caused by the cou-
pling between the separated regions and the inviscid part of the flow in confined
channels. Similar flows can be observed in other channel-type flows, such as super-
sonic diffusers, compressor cascades, and propulsive nozzles.
Surface-Curvature Effects
On transonic airfoils and wings, the surface underneath the interaction is not flat
but rather generally features some form of convex curvature. The free-stream Mach
number ahead of the shock also is not uniform but rather decreases with distance
from the surface. However, apart from cases in which very large separations are
present, the interaction region is relatively small compared to the size of the airfoil;
the surface curvature in the shock region is also relatively mild. For this reason,
most of the features described previously are as valid for attached and separated
interactions on airfoils.
However, there are important consequences of surface curvature. The first
effect is that the shock strength required for separation generally is found to be
greater than on a flat surface. This was demonstrated theoretically [14,15,16] and
in experiments [3,17]. Pearcey [3] suggested that this was due to the fact that con-
vex curvature alone would cause a streamwise decrease in pressure that reduces the
adverse-pressure gradient experienced by the boundary layer. Bohning and Zierep
[16] offered an alternative explanation, arguing that convex surface curvatures lead
to greater postshock expansions and that this effect can feed upstream through the
subsonic portion of the boundary layer, thereby reducing the shock-induced pres-
sure rise. Nevertheless, the curvature effect on incipient separation is relatively
weak.
A more significant effect of surface curvature on the SBLI flowfield often can be
observed just downstream of the interaction because the flow there is close to sonic
conditions and therefore sensitive to geometric factors. Convex surface curvature
in this region is likely to accelerate the flow, making secondary supersonic regions
more likely or considerably increasing their size. This is generally an undesirable
effect because of the increasing likelihood of secondary shocklets.
118 Transonic Shock Wave–Boundary-Layer Interactions
Sweep Effects
In transonic flows, the shock wave causing the SBLI is generally a normal or near-
normal shock and the sweep angles are small. The most common example of a tran-
sonic SBLI with considerable sweep is on transonic aircraft wings where the shock
follows the wing sweep, which can reach values of up to 30 degrees. Even in this
case, experience seems to suggest that sweep effects are negligible as long as the flow
remains attached. The main effect of sweep is on the onset of shock-induced sepa-
ration. Research into the interaction of swept shock waves with flat-plate boundary
layers suggests that these separate once the shock normal Mach number increases
beyond 1.2 [19]. From this, we can conclude that cross-flow effects may bring about
separation somewhat earlier than observed in the equivalent two-dimensional flow
(in the absence of cross flow). Once shock-induced separation is observed, the flow
becomes more three-dimensional; however, unless the flow downstream of the main
shock is supersonic, the flowfield in a shock-normal plane is similar to the flow pat-
terns shown previously.
A A A
B B B
C C C
D D D
Figure 3.34. Schlieren images of normal shock interaction subjected to oscillating back pres-
sure at three different frequencies, M = 1.4.
to the fact that the flow downstream of the shock wave is subsonic. This allows pres-
sure fluctuations originating downstream to travel back towards the shock wave and
affect its strength and position. Typically, an increase in the back pressure causes
shock waves to move upstream and vice versa.
Figure 3.34 shows a series of schlieren images from an experimental study of a
normal shock wave in a constant-area channel subjected to sinusoidal back-pressure
oscillations at various frequencies [20]. In response, the shock wave oscillates in a
periodical motion around the steady-equilibrium position. Four images are shown
per cycle: at the extremes of the shock location and halfway through each motion
when the shock is near the equilibrium location but moving at maximum speed.
The pressure waves originating downstream and traveling back towards the shock
wave can be seen clearly in several of the images. With increasing frequency, the
amplitude of the shock motion reduces (the pressure oscillation amplitude has been
kept constant). The reduction of shock-wave travel is a result of the observation
that the maximum shock velocity is determined by the amplitude of the pressure
disturbance; because this was kept constant, a higher frequency allows the shock
less time to move away from the mean position. In nonparallel ducts, the shock-
motion amplitude due to pressure variations also is affected by the geometry. This
is illustrated by Fig. 3.35, which compares the relationship between shock-oscillation
amplitude and frequency (for a fixed-pressure amplitude) for parallel and divergent
ducts. At high frequencies – where the shock oscillation amplitudes become small –
there is little effect of divergence angle; however, at low frequencies, the geometry
clearly has an effect.
Similar amplitude-frequency behaviour has been observed in the response of
shock waves to disturbances originating from the upstream inflow or from inside
the incoming boundary layer. This also explains why the response of shock waves to
the relatively high frequency disturbances present in a turbulent boundary layer is
120 Transonic Shock Wave–Boundary-Layer Interactions
100
0
0 20 40 60 80 100
Frequency (Hz)
relatively small in amplitude, causing shock ‘rippling’ rather than large-scale depar-
tures of shock location from the mean. For many technical applications in which
shock oscillations are of concern, it is therefore particularly important to minimize
low-frequency pressure fluctuations.
Pressure fluctuations not only cause shock motion but also affect the shock
strength. A shock moving into the oncoming flowfield faces an inflow of greater
relative Mach number (i.e., the free stream plus the shock motion); the opposite is
true for a shock moving downstream. As a result, the pressure jump across the shock
wave is increased during upstream motion and decreased when moving downstream.
This is illustrated by the experimental results shown in Fig. 3.36.
0.6
Predicted pressure rise
(upstream motion)
0.55
Steady interaction
0.4 pressure rise
0.35
Downstream shock motion
(Relative Mach number = 1.379)
0.3
−60 −40 −20 0 20 40 60
x (mm)
Figure 3.36. Instantaneous surface-pressure distribution for forced shock oscillation at M =
1.4 (f = 43Hz).
3.3 Normal SBLIs in Detail 121
Even when the shock velocities are small, the differences in the shock-induced
pressure rise can be considerable (e.g., a ± 1.5 percent variation in shock Mach
number causes a 10 percent difference in postshock pressures), highlighting the non-
linear nature of SBLIs in general. This effect is particularly noticeable at transonic
Mach numbers because the relative variations are significant compared to the inci-
dent flow. Once the oncoming Mach numbers become large, however, this effect
diminishes.
The sensitivity of transonic SBLIs to disturbances originating from the down-
stream subsonic flow can have serious implications when the unsteady nature of
the interaction causes global flowfield changes. This can be the case when there is
considerable shock-induced separation. A pressure increase imposed from down-
stream would cause the shock to run upstream (and increase in strength), and both
may cause the separation to enlarge and change position. Depending on the overall
flow geometry, this in turn can alter the downstream pressure; a dangerous feed-
back mechanism may result, which greatly amplifies the pressure disturbances. An
example of such an event sometimes is observed on transonic wings when large-
amplitude shock oscillations occur along the upper surface (sometimes linked to a
corresponding shock oscillation on the lower surface). This is one of the forms of
buffet (there also are incompressible forms of wing buffet) that puts a severe strain
on the wing structure and can destroy aircraft.
In this phenomenon, the normal shock moves over a significant portion of
the airfoil-chord length, with the separated bubble disappearing and reappearing
in a periodic manner. A scenario for such a periodic unsteadiness is proposed in
Fig. 3.37, obtained from Navier-Stokes calculations. In the first frame, the shock
occupies a downstream position but is moving upstream in response to a pressure
disturbance. Thus, the flow velocity relative to the shock is increased. If the rela-
tive Mach number is large enough, separation occurs and a large separation bubble
forms due to the combined effect of shock-induced pressure rise and compression
along the profile towards the rear of the airfoil. This separation has the effect of
reducing the effective aerodynamic incidence so that the shock weakens as it con-
tinues to move upstream until it stops; simultaneously, the separation bubble starts
to shrink after having reached its greatest size (frame 3). This leads to an increase in
the effective aerodynamic incidence, with the shock now moving downstream. The
relative shock strength is then less (because the downstream movement reduces its
effective Mach number) and separation at the shock foot is suppressed (frame 4).
When the shock slows down, separation occurs and is then amplified when the shock
starts to move upstream again and the cycle repeats.
Another example in which the flowfield downstream of an SBLI is sensitive to
changes in the interaction is a supersonic engine inlet, which can suffer from severe
shock unsteadiness. Inlet unstart (in which an unsteady shock is spilled out of the
inlet causing engine shutdown) was previously discussed, but there also is a severe
form of quasiperiodic shock unsteadiness, sometimes referred to as ‘intake buzz’.
The prime purpose of an air intake is to slow down the oncoming supersonic
flow before the engine to subsonic speeds through a number of shock waves. Pos-
sible shock configurations are discussed in Section 3.22. For internal or mixed-
compression engine inlets, the terminating near-normal shock wave is usually quite
strong and flow separation is a considerable problem. When a sudden pressure
122 Transonic Shock Wave–Boundary-Layer Interactions
disturbance causes the shock to move upstream (and thus strengthen), it can
increase the separated flow region and cause the downstream flow to ‘choke’,
thereby generating an increased pressure pulse that drives the shock farther
upstream. The inlet-unstart scenario is illustrated in Fig. 3.38.
A scenario similar to transonic buffet is at work in the buzz of supersonic
air-intakes. In this case, a periodic large-amplitude motion of the shock system
3.4 Control of Transonic SBLIs 123
a) b) c)
d) e) f)
Figure 3.38. (a)–(f) Time sequence of inlet unstart developing in response to a pressure
disturbance originating from downstream (RANS CFD solution provided by B. Anderson,
NASA Glenn).
is triggered by unsteady SBLIs. There are two possible origins of air-intake buzz:
(1) shock-induced separation, the so-called Dailey’s scenario; and (2) the so-called
Ferri’s scenario, the swallowing of the slip line resulting from an Edney Type IV
interference between the cowl shock and a shock formed by a compression ramp.
Overexpanded propulsive nozzles are affected by unsteady and asymmetric shock-
induced separation, which can be the source of high side loads during the start-up
transient.
Large-amplitude oscillations also occur within rotating machines, such as com-
pressors, turbines, and helicopter rotors. In these cases, the shock oscillation is
forced by the device itself, which potentially can cause coupling between the body
motion and the SBLI fluctuation frequencies, including those due to turbulence. A
similar coupling mechanism is found in the aeroelastic response of a structure, such
as compressor blades. In this case, the deformation of the structure induces shock
displacements and a subsequent change in the pressure load, which can lead to a
divergent process or flutter.
2. The stagnation pressure losses incurred by the shock wave introduce drag or
efficiency reductions. Strictly speaking, this loss is unconnected to the shock’s
interaction with the boundary layer – in fact, the interaction actually helps to
reduce the stagnation-pressure loss by smearing the shock wave near the sur-
face. Nevertheless, the shock losses often are considered part of SBLI losses
and changes to the SBLI structure affect their magnitude.
3. Where considerable separations occur, they lead to significant changes in the
shock structure and overall flow. In some cases, this can introduce large-scale
flowfield unsteadiness (i.e., buffet, inlet buzz, or engine unstart).
SBLI control aims to alleviate one or more of these effects. Depending on which
type of physical mechanism is addressed, we speak of either shock control (which
aims to change the shock-system structure to reduce stagnation-pressure losses) or
boundary-layer control (which aims to reduce viscous losses and separation). Due
to the complex nature of SBLI flows, this separation is not strict. For example, any
modification of the shock wave has an effect on the boundary layer and vice versa.
Depending on the exact circumstances, both methods have been shown to reduce
shock-wave unsteadiness.
Control
p/p0
For both Mach numbers shown, compression to subsonic speeds can be achieved
with significantly reduced stagnation-pressure losses. At M = 1.3, an optimum lies
around a deflection angle of about 5 degrees, giving an overall loss of 0.3 percent,
whereas at M = 1.4, the optimum deflection angle is around 7 degrees with an over-
all loss of 0.7 percent. In both cases, the stagnation-pressure losses across the shock
system are reduced by around 85 percent. However, as indicated in the flowfield
sketch, the deflection angle also changes the location of the triple point and, thus,
0.99
M = 1.3 M
0.97
0 1 2 3 4 5 6 7
deflection
deflection angle angle
normal shock
1.00
weak oblique shock
0.99
0.98
0.97
Figure 3.40. Stagnation-pressure loss across a simplified λ-shock system (oblique shock fol-
lowed by normal shock wave) as a function of initial oblique-shock deflection angle for
M∞ = 1.3 (left) and M∞ = 1.4 (right).
126 Transonic Shock Wave–Boundary-Layer Interactions
Control
Figure 3.42. Surface pressures in a controlled SBLI.
With Control
No Control
X
3.4 Control of Transonic SBLIs 127
-region
them to a minimum. This means that the boundary-layer thickness and shape factor
downstream of a controlled interaction should be similar to that observed behind
an uncontrolled (i.e., attached) interaction. However, most control mechanisms do
not fulfil this aim, and the overall benefit of shock control is a balance between the
detrimental effects on the boundary layer and the advantageous effects on the shock
structure. A reduction in wave drag often is counteracted by an increase in viscous
drag; it is the overall balance between the two that determines whether a control
can be considered effective.
An assessment of the success of a control method is complicated further by the
fact that the flow downstream of the interaction is often subject to more adverse-
pressure gradients. A shock control that has modified the boundary-layer develop-
ment also is likely to lead to changes in the flow downstream. Such changes can be
adverse (i.e., earlier separation) or beneficial (i.e., delayed separation or improved
boundary-layer health) and the final balance depends on the actual geometry of the
flow. Therefore, it is often difficult to fully assess the costs and benefits of a con-
trol strategy unless the complete flowfield is considered. This also means that it is
difficult to take lessons learned from one situation and apply them to a different
environment.
Possible secondary
supersonic region
Strong second
shock leg
No beneficial
λ-shock region Additional
viscous drag M increasing
(a) (b)
Figure 3.44. Contoured-surface bump under off-design conditions: (a) shock too far
upstream; (b) shock too far downstream.
original normal shock. The maximum bump height is reached after the rear shock
leg. From this point onwards, the bump height reduces and the flow is returned to
the airfoil surface. This area of the bump must be shaped with care to avoid sepa-
ration and reacceleration to supersonic flow (which can cause unwanted additional
shock waves).
Many researchers [23] successfully demonstrated the shock-control-bump con-
cept for transonic wings, in both computations and experiments on airfoil sections as
well as on fully three-dimensional wings. A well-designed bump is capable of gener-
ating a sizeable λ-shock region while incurring only modest viscous flow penalties. In
such a successful design, the boundary-layer health behind the bump is comparable
to that observed without a control present.
One major difficulty of bump control is that the benefits are sensitive to the
position of the shock wave relative to the bump. Under off-design conditions, with
the shock wave either too far upstream or downstream from the ideal, penalties are
incurred, as shown schematically in Fig. 3.44.
When the shock wave is too far upstream (Fig. 3.44a), there is virtually no ben-
eficial smearing of the shock foot, and the additional wetted surface of the control
bump incurs a viscous drag penalty. In extreme cases (i.e., for large bump heights),
it is possible that the curvature of the bump causes a secondary supersonic region
with a second shock that also would increase the wave drag.
When the shock wave is too far downstream (i.e., beyond the bump crest), the
concave shape of the bump causes expansion waves to emerge inside the λ-shock
region (Fig. 3.44b). This locally increases the flow Mach number ahead of the rear
shock leg. As a result, the rear leg strengthens significantly, which incurs consider-
able additional wave drag. To avoid off-design penalties when using shock-control
bumps, it is necessary either to use movable or deployable bumps or to employ
airfoil-section shapes designed to have relatively little movement of shock location
over the range of cruise Mach numbers encountered in flight.
Surface bumps also have been widely reported as capable of delaying shock-
induced buffet on transonic wings [23]. This generally is observed for bumps located
downstream from the optimum location for wave-drag reduction under design con-
ditions. The exact mechanism for buffet alleviation through contour bumps is not
yet well understood, but the following factors are likely to be at work:
3.4 Control of Transonic SBLIs 129
r Buffet generally occurs when a shock wave has moved far back on a wing and
grown in strength to cause significant boundary-layer separation. A contour
bump placed well aft of the shock location at the design point would only come
into effect under such conditions. The generation of a λ structure when a shock
has moved into ‘buffet territory’ weakens the adverse-pressure gradients expe-
rienced by the boundary layer, which is likely to reduce the danger of buffet.
r Shock-induced buffet is driven by the large oscillating pressure loads experi-
enced when a shock wave moves back and forth along the wing surface. When
a shock moves similarly on a contour bump, the leading leg of the shock struc-
ture remains fixed to the leading edge of the bump. Only the rear leg oscillates,
which reduces the overall pressure fluctuations.
‘Passive’ Control
Although in many ways the surface bump is a highly successful form of shock con-
trol, many researchers have looked for ways to overcome the penalties incurred
when a bump is under off-design conditions. At present, active shock-bump systems
(in which the bump can be retracted or moved) are considered too complex and
heavy for practical applications.
An alternative concept that has been studied extensively [23] is ‘passive control’.
In this control method, a cavity covered with a porous surface is placed underneath
the shock location, as shown in Fig. 3.45.
The pressure inside the cavity self-adjusts to a value between the preshock and
postshock levels. Downstream of the shock wave, therefore, is a pressure above
the porous surface that is larger than inside the cavity and vice versa ahead of the
shock. As a result, boundary-layer flow is driven into the cavity behind the shock
and fluid is blown out of the cavity ahead of the shock. The flow blowing out of the
cavity displaces the boundary layer similar to the action of a contour bump and lead-
ing oblique compression waves are formed (often merging quickly into an oblique
shock). The overall flow is similar to that observed over contour bumps and the
achievable savings in wave drag are comparable. Figure 3.45 compares a sketch of
this flowfield with a schlieren image of passive control applied to a normal shock
wave at Ms = 1.3.
130 Transonic Shock Wave–Boundary-Layer Interactions
The ‘viscous bump’ formed by the recirculating flow across the porous surface
self- adjusts for a wide range of shock locations. As long as the shock wave is posi-
tioned above the control, a λ structure can be observed, with an optimum when the
main shock is located slightly downstream of the centre of the cavity (i.e., around
two-thirds length). When the shock is outside the control region, the recirculat-
ing flow stops and the cavity has almost no aerodynamic effect, other than a slight
increase in viscous drag due to the roughness of the surface. This control requires
no external action or power supply – hence, the name “passive control.”
Unfortunately, it has been found that introducing recirculation of boundary-
layer flow incurs significant viscous-drag penalties, mainly due to the insertion of
low-momentum fluid upstream of the shock wave. Although the suction region at
the rear of the control improves the boundary-layer flow to some extent, it is not
sufficient to avoid significant additional thickening of the boundary layer through
the interaction and associated momentum loss. In most cases, it has been observed
that the viscous-drag penalties incurred by passive control are equal to or greater
than the achievable savings in wave drag.
control locations and thus are likely to be much smaller than those observed on a
comparable two-dimensional control. There also is evidence that three-dimensional
controls may incur smaller off-design penalties. Furthermore, it is considerably eas-
ier to fit individual three-dimensional devices to an existing wing or even retrofit a
small number of devices near ‘trouble spots’ where an unduly strong shock wave is
observed. Finally, individual three-dimensional devices also may be easier to make
‘active’ (i.e., deployable or moveable).
In principle, almost all forms of two-dimensional shock control have a three-
dimensional counterpart, and examples are shown in Fig. 3.46. Contour bumps have
been studied in their three-dimensional ‘incarnations’. Streamwise slots are similar
to passive control, and various other combinations of recirculating control incorpo-
rating suction and blowing are easily imaginable. The basic flow features observed
for two types of three-dimensional shock controls are illustrated in Fig. 3.47.
-shock
structure
Main shock
Front
shock leg
dge
er e
ar y-lay
Bo und
Rear shock leg
3D rounded
bump
(a) (b)
Figure 3.47. Basic features of a three-dimensional controlled normal SBLI: (a) slot control;
(b) three-dimensional bump control.
132 Transonic Shock Wave–Boundary-Layer Interactions
(a) (b)
It can be seen that an array of slots (Fig. 3.47a) generates a relatively two-
dimensional λ-shock structure, whereas the boundary-layer flow is highly three-
dimensional (thus, the viscous penalties are highly localized). In the case of the
single three-dimensional shock-control bump (Fig. 3.47b), the λ-shock structure
slowly reverts back towards a single shock, but there is considerable spanwise dis-
tance in which a significant λ structure remains. The boundary layer behind the
three-dimensional bump is slightly thickened; however, it again is confined to the
immediate wake behind the device – elsewhere, the boundary-layer profile essen-
tially is unchanged. The subject of three-dimensional shock control is still under
investigation but early results show promise.
Streamwise vortex
Streamwise vortex
Vortex generator
Figure 3.49. Vane-type vortex generators. Top: sketch of flowfield. Bottom: rear view.
One suggestion to reduce the parasite drag incurred by VGs is the use of sub-
boundary-layer vortex generators (SBVGs), also referred to as micro-vortex gen-
erators. Such devices measure less than a boundary-layer thickness in height (some
are as small as one displacement thickness) and they incur much reduced viscous
drag due to the small wetted area and protrusion into much slower portions of
the flow. Nevertheless, they have been shown to be surprisingly effective in reduc-
ing shock-induced separation [24, 25]. For example, Fig. 3.50 shows how pairs of
counter-rotating vortices generated by micro-vanes (h/δ ≈ 0.2) can eradicate shock-
induced separation at Ms = 1.5. Current research suggests that the mechanism of
flow control through SBVGs is slightly different from that of more traditional VGs.
Whereas both introduce streamwise vortices into the flow, traditional VGs entrain
high-momentum fluid from outside the boundary layer. Typically, these devices are
placed well ahead of any region of adverse-pressure gradient to allow the entrained
momentum to spread throughout the boundary layer, giving a fuller velocity profile
that is more resistant to separation. Conversely, micro-VGs introduce streamwise
vortices that are embedded inside the boundary layer and redistribute momentum
internally. Such vortices are less strong, but they are placed closer to the surface
where they can be more effective. However, the location close to the wall can lead
to an earlier dissipation of the vortices and it is therefore likely that such devices
must be closer to the adverse-pressure gradient than traditional VGs.
Figure 3.50. Sub-boundary layer vortex generators ahead of a normal SBLI at Ms = 1.5.
134 Transonic Shock Wave–Boundary-Layer Interactions
Compression corner
separation
Separation
Primary vortex pair
Figure 3.51. Micro-ramp vortex generator: (a) surface oil-flow visualization; (b) velocity
change (relative to baseline) measured downstream of device.
normal shock-induced separation, even at much higher shock Mach numbers (i.e.,
approaching M = 2). Fig. 3.52 shows suction holes in the inlet of a supersonic fighter
aircraft.
Surface suction, or bleed, is a highly effective method of preventing (or delay-
ing) shock-induced separation. It is particularly attractive in inlets because it is rel-
atively easy to implement in practice and, other than preventing separation, it also
makes the boundary layer downstream of the SBLI thinner, which reduces flow dis-
tortion. However, there is a performance penalty because the mass flow reaching the
engine is reduced by the suction-mass-flow rate (which may be considerable); there-
fore, the inlet area must be increased, which introduces additional drag. A subtler
effect of bleed is that the fuller (and thinner) boundary layer reaching the SBLI
causes a considerable reduction in interaction length, which leads to stronger or less
smeared shocks with an associated (slight) increase in stagnation-pressure loss.
Despite the drawbacks, the need for flow control to prevent separation in super-
sonic inlets is so great that no current device in service operates without some
form of boundary-layer bleed. Similar considerations apply to other internal flows
(e.g., inside turbomachines), where bleed is also popular. For external transonic-
flow applications, however, bleed is not a widely used method of boundary-
layer control for SBLIs (which is not to be confused with bleed for laminar-flow
control).
The area of boundary-layer control for SBLI applications is an active topic of
current research. Various novel control methods have been proposed, among them
zero-mass flux-pulsed jets (sometimes referred to as virtual jets) and plasma actua-
tors, which continue to be in the early experimental phase; readers are referred to
current research publications for more information.
REFERENCES
[1] A. H. Shapiro. Dynamics and Thermodynamics of Compressible Fluid Flow (New York:
Ronald Press Co., 1954).
[2] R. C. Lock. The prediction of the drag of aerofoils and wings at high subsonic speeds.
RAE TM Aero, 2044 (1985).
[3] H. H. Pearcey. “Shock-Induced Separation and Its Prevention by Design and Boundary
Layer Control.” In Boundary Layer and Flow Control, ed. G. V. Lachmann (Oxford:
Pergamon Press, 1961), pp. 1166–344.
[4] H. Sobieczky, N. J. Yu, K. Y. Fung, and A. R. Seebass. New method for designing shock-
free transonic configurations. AIAA Journal, 17(7) (1979), 722–9.
[5] I. Alber, J. Bacon, B. Masson, and D. Collins. An experimental investigation of turbu-
lent transonic viscous-inviscid interactions. AIAA Journal, 5(11) (1973), 620–7.
[6] A. Zheltovodov, R. Dvorak, and P. Safarik. Shock wave/turbulent boundary layer inter-
action properties at transonic and supersonic speeds conditions. Izvestiya of SO AN
SSSR, 6 (1990), 31–42 (in Russian).
[7] A. Zheltovodov and V. Yakovlev. Stages of development, gas dynamic structures and
turbulence characteristics of turbulent compressible separated flows in the vicinity of 2-d
obstacles. Preprint No. 27–86, Institute of Theoretical and Applied Mechanics, Russian
Academy of Sciences, Novosibirsk (1986) (in Russian).
[8] E. R. Van Driest. Turbulent boundary layer in compressible fluids. Journal of Aeronau-
tical Sciences, 18(3) (1951), 145–216.
[9] J. E. Green. Interactions between shock waves and turbulent boundary layers. Progress
in Aerospace Sciences, 11 (1970), 235–339.
136 Transonic Shock Wave–Boundary-Layer Interactions
4.1 Introduction
Effective design of modern supersonic and hypersonic vehicles requires an under-
standing of the physical flowfield structure of shock wave–boundary layer interac-
tions (SBLIs) and efficient simulation methods for their description (Fig. 4.1). The
focus of this chapter is two-dimensional supersonic shock wave–turbulent bound-
ary layer interactions (STBLIs); however, even in nominally two-dimensional/
axisymmetric flows, the mean flow statistics may be three-dimensional. The discus-
sion is restricted to ideal, homogeneous gas flow wherein the upstream free-stream
conditions are mainly supersonic (1.1 ≤ M∞ ≤ 5.5). Computational fluid dynam-
ics (CFD) simulations of two-dimensional STBLIs are evaluated in parallel with
considerations of flowfield structures and physical properties obtained from both
experimental data and numerical calculations.
137
138 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
∂ρ ∂ρu j
+ =0
∂t ∂xj
∂ρui ∂ρui u j ∂p ∂τi j
+ =− +
∂t ∂xj ∂ xi ∂xj
∂ρe ∂(ρe + p) ∂
+ = (ui τi j − q j )
∂t ∂xj ∂xj
p = ρ RT (4.1)
4.1 Introduction 139
where xj are the Cartesian coordinates, ui are the corresponding velocity compo-
nents, ρ is the density, p is the static pressure, e is the total energy per unit mass,
1
e = cv T + u j ui (4.2)
2
R is the gas constant, τ ij is the shear stress, and qi is the heat flux. For a Newtonian
fluid:
∂uk ∂ui ∂u j
τi j = λ δi j + μ + (4.3)
∂ xk ∂xj ∂ xi
where μ is the dynamic molecular viscosity and, from Fourier’s Law:
∂T
qi = −k̂ (4.4)
∂ xi
where k̂ is the thermal conductivity. These equations are valid for both laminar and
turbulent flows. In the latter case, the solution of the equations is denoted as DNS,
which is computationally impractical for flight-scale engineering configurations due
to large values of the Reynolds number. However, DNS at lower Reynolds numbers
achievable in experiments are important for a better understanding of the physical
processes in STBLIs. For simulation of flight-scale engineering configurations, the
compressible Navier-Stokes equations are averaged and statistical mean values are
computed. In this context, two different approaches were developed. The first is the
RANS equations, which are obtained by introducing the Favre ensemble average of
a function f˜ defined by the following equation:
v=n
1 1
f˜ = lim (ρ f )(v)
ρ̄ n→∞ n
v=1
ρf
=
ρ̄
f = f + f
˜ (4.5)
in which the overbar indicates an ensemble average. The resultant equations are as
follows:
∂ ρ̄ ∂ ρ̄ ũ j
+ =0
∂t ∂xj
∂ ρ̄ ũi ∂ ρ̄ ũi ũ j ∂ p̄ ∂ Ti j
+ =− +
∂t ∂xj ∂ xi ∂xj
∂ ρ̄ ẽ ∂(ρ̄ ẽ + p̄)ũ j ∂
+ = (ũi Ti j − Q j )
∂t ∂xj ∂xj
p̄ = ρ̄ RT̃ (4.6)
where the total stress-tensor and heat-transfer vectors are as follows:
Ti j = −ρui uj + τi j
Q j = c p ρT uj + q̄ j (4.7)
and where ( ) represents the velocity fluctuations, and the first and second terms
are the turbulent and mean laminar contributions, respectively.
140 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
eddies that define the size of the turbulent boundary layer. LES equations are as
follows:
∂ ρ̄ ∂ ρ̄ ũk
+ =0
∂t ∂ xk
∂ ρ̄ ũi ∂ ρ̄ ũi ũk ∂ p̄ ∂ Tik
+ =− +
∂t ∂ xk ∂ xi ∂ xk
∂ ρ̄ ẽ ∂ ∂
+ ( p̄ẽ + p̄)ũk = (Qk + Tik ũi )
∂t ∂ xk ∂ xk
p̄ = ρ RT̃ (4.11)
Tik = −ρ̄(u
i uk − ũi ũk ) + τ̄ik
k − T̃ ũk ) − q̄k
Qk = −ρ̄c p (Tu (4.12)
The term −ρ̄(u i uk − ũi ũk ) represents the Subgrid Scale (SGS) stress and the
k − T̃ ũk ) represents the SGS heat flux. Two different approaches
term −ρ̄c p (Tu
were developed to model these terms. In the first approach, explicit models for the
SGS stress and heat flux are posited [23, 24, 25]. In the second approach, implicit
models for the SGS stress and heat flux are posited based on the concept of mono-
tone integrated large eddy simulation (MILES) [26].
Development of hybrid LES/RANS simulations of unsteady STBLIs recently
appeared as a less-computationally expensive (compared to LES) approach for high
Reynolds-number flows [27, 28]. The flow-dependent blending functions (similar
in construction to those used in Menter’s k-ω/k-ε model [29]) shift the turbulence
modeling from the two-equation model near a solid surface to an LES subgrid clo-
sure away from the solid surface and in free-shear regions around a separated zone.
1 For oblique SBLIs, an equivalent specification of the shock strength is the incident inviscid flow-
deflection angle α.
2 Equivalently, the Reynolds number based on the compressible displacement thickness δ * or
momentum thickness θ may be used because the incoming turbulent boundary layer is assumed
to be in equilibrium.
142 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
(C1) (C4)
10 Δv1 = a
: supersonic Expansion fan
M = 0.71 Δv2 slip line
Y M0 = 1.47 unshaded: subsonic
δ0 normal shock M0
M∞
M∞
δ0
δ0 R X
h X h
α
R
S 0
S 0
a. M = 1.11
b. M = 1.31
c. M = 1.45
Figure 4.3. Normal shock wave turbulent boundary layer interaction: Schlieren photographs
(left column) and surface-flow patterns (right column) (Zheltovodov et al. [30]).
waves induced by the separation forward of the normal shock (Figs. 4.3c and 4.2a).
The foot of the normal shock is transformed into compression waves or a lambda
shock configuration, with the curved part of the shock in the external flow associated
with the strong oblique-shock solution of the Rankine-Hugoniot equations. Conse-
quently, the observed surface-pressure level p2 /p1 is lower than the values predicted
144 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
P2 P Inviscid 1 5
P∞ 2 6
P∞ P2 3 7
2.0 4
x Pp P∞ at:
Lint
Pp Reθ = 103
ξinv 4.103
P∞
ξi
1.5
Ps P∞
Ps III
P∞ II
Pmax
Pson
I γru = 0.08 γru = 0.55
1.0
1.0 1.2 Mi Mef 1.4 Ms M∞
Figure 4.4. Pressure ratio across normal shock and interactions regimes: I – unseparated flow;
II – intermittent separation; III – developed separation (Zheltovodov et al. [30]).
in equation (4.13). Additionally, the subsonic nature of the outer flow downstream
of the normal shock renders the flow globally dependent on downstream conditions.
The wave structure of NSTBLIs is discussed in detail in Chapter 3.
Figure 4.4 displays the measured surface-pressure ratio p2 /p1 (i.e., symbol 1
[30]) in the isobaric region downstream of the shock (see sketch) as a function
of the free-stream Mach number M∞ = M1 , where p1 = p∞ . The normal shock-
static-pressure ratio satisfying equation (4.13) is shown as a dashed line (ξ inv ). At
unseparated-flow conditions (M∞ < 1.25, regime I), the data point (1) corresponds
to the maximum shock-deflection pressure curve pmax . A sudden reduction in the
growth rate of (dp/dx)max = (p2 – p1 )/Lint is observed at M∞ = Mi = 1.25 ± 0.02.
This point corresponds to the “true” incipient-separation conditions, although the
flowfield is not yet greatly altered by the existence of the small separated bubble.
This conclusion is supported by the fact that the measured surface-static-pressure
ratio at Mi = 1.25 coincides with the bare minimum value (p2 /p1 )min = 1.4, which
is required to cause true incipient separation on a transonic airfoil at the shock foot
in accordance with Pearcey’s [31] experiments. The external “inviscid” shock-wave
strength ξ inv = f(M∞ ) at this stage is in good agreement (i.e., within the limits of
accuracy for M∞ = Mi = 1.25 ± 0.02), with the lower critical value ξ *i = 1.61 corre-
sponding to incipient separation in the vicinity of normal shocks in accordance with
an experimental correlation by Grodzovskyi [32] (i.e., line 6).
The appearance of bifurcations in the surface-flow pattern at M∞ = Mef = 1.3
± 0.01 (Fig. 4.3b) corresponds to the beginning of “significant” (or “effective”) sep-
aration, wherein a noticeable separation zone emerges in the foot of the shock. The
data for p2 /p1 confidently reach the sonic-pressure curve pson at this stage (Fig. 4.4),
similar to the classic experiments of Pearcey [31], and also begin to correspond to
the surface-static-pressure ratio at the separation point ahead of the forward-facing
step predicted by the empirical correlation [33] ps /p1 = 0.365 M∞ + 1 (line ps /p∞ ).
Thus, the condition of equivalence between the sonic pressure and separation-point
pressure (pson = ps ) corresponding to Pearcey’s [31] criterion reasonably predicts
4.2 Two-Dimensional Turbulent Interactions 145
the critical Mach number value Mef ≈ 1.3 at the onset of effective incipient separa-
tion. The surface-pressure ratio p2 /p1 in regime II at M∞ ≥ 1.3 begins to correspond
to the plateau-pressure dependence pp /p1 (i.e., line 2) [33] observed for separation
zones in front of forward-facing steps at low Mach numbers. LDV measurements
[34] in the vicinity of the normal shock foot at M∞ = 1.34 and the value p2 /p1 shown
by symbol 3 reveal the intermittent nature of a small reversed-flow zone that exists
for a short fraction of time (i.e., 8 percent). This is in accordance with the inter-
mittency coefficient γ ru = 0.08, which corresponds to an “intermittent-detachment”
stage [35].
The well-defined separation line forming at M∞ = Ms = 1.43 ± 0.02 (see
Fig. 4.3c) corresponds to conditions when the measured p2 /p1 level in the Reynolds-
number range Reθ = (1 − 4) × 103 reaches the following plateau-pressure ratio:
pp 2 −1/4 1/2
= k M∞
2
M∞ − 1 cf + 1 (4.14)
p∞
for developed steady-separation zones predicted by the classic Free Interaction The-
ory (FIT) of Chapman et al. [36] (Fig. 4.4, solid line 5). The experimentally estab-
lished [3] constant k = 5.94 for turbulent flow was used for this prediction [1, 30] with
the skin-friction coefficient cf calculated for a flat-plate turbulent boundary layer on
the basis of Kutateladze and Leont’ev’s theory [37]. In accordance with Fig. 4.4, the
condition pson = pp can be used analogously with Pearcey’s [31] criterion for predic-
tion of the critical Mach number M∞ = Ms corresponding to incipience of developed
separation. As shown in experiments by Morriss et al. [38], at M∞ = 1.48 (indicated
by the symbol 4 for the level p2 /p1 ), the reversed flow exists more than 50 percent
of the time (γ ru = 0.55) with a distinct separation line demonstrated by the oil-flow
visualization. Thus, a developed “quasistationary” separation (or “full detachment”
[35]) is realized in regime III at M∞ > Ms .
Experimental and theoretical analyses of incipient-separation conditions for
NSTBLI are the subject of many studies [1, 3, 8, 39]. The Mi data for true incip-
ient separation on a flat-plate surface by Zheltovodov et al. [30] and Grodzovskyi
[32] (Fig. 4.5, symbols 3 and 4) are in good agreement with the theoretical prediction
by Inger [40, 41] (i.e., line 1) and close to the empirical correlation (i.e., dotted line
2), which generalizes the experiments (i.e., symbol 2) at ONERA for transonic flows
past curved surfaces (e.g., airfoils and bumps on a wind-tunnel wall) in accordance
with Délery and Marvin [3] and Haines [39]. Stanewsky [42] defined the limits of
effective incipient separation on an airfoil surface by plotting the variation of the
boundary-layer displacement-thickness δ* kink in the vicinity of the normal shock
(i.e., symbols and line 5). He also extrapolated measured separation-bubble lengths
back to zero length (i.e., the solid line) and considered Pearcey’s [31] criterion
(ps = pson , symbol 6). The data (Mef ) for effective incipient separation on a flat-
plate surface (i.e., symbols 7, 8) are in good agreement with these empirical corre-
lations. A tendency to a small decrease in the levels of Mi and Mef for the flat-plate
surface (i.e., symbols 3, 4 and 7, 8), compared to experiments for slowly curved sur-
faces (lines 2, 5), is in agreement with a theoretical prediction [43]. All examined
correlations and data demonstrate a slight increase in Mi and Mef with a decreasing
value of the incompressible-shape parameter Hio = δ*i /θ i in the incoming bound-
ary layer, where δ*i and θ i are the incompressible-displacement and momentum
146 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
MO MO Significant Separation
Hio Experiments:
-5, - 6 (DFVLR, Stanewsky, 1981)
δO - 7 (Grodzovskyi, 1961)
Separated
1.5 - 8 (Zheltovodov et al., 1990)
-9
“Zero” bubble
δ∗- kink
Mef ps = psonic length
Mef Mef
1.3 Mi
Mi
True Incipient Separation
Unseparated - 1: Theory (Inger, 1980)
1.1 - 2: Experiments (ONERA)
- 3: Experiments (Grodzovskyi, 1961)
- 4: Experiments (Zheltovodov et al., 1990)
1.2 1.4 1.6 Hi0
thicknesses [3, 39] (i.e., the increasing Reδo ). A lower value of Hio means a fuller
velocity profile and, accordingly, a greater resistance of the developed (“well-
behaved”) turbulent boundary layer to separation. The lack of significant variation
in Mi and Mef is explained [3, 42] on the premise that an increase in Hio increased
the subsonic part of the boundary layer and, consequently, the likelihood of its sep-
aration. Conversely, it also increased the length of the upstream influence, thereby
alleviating the adverse pressure gradient and the tendency to separation. Therefore,
the two factors tend to negate one another. Nevertheless, against a background of
limited accuracy of the data presented in Fig. 4.5, they seem to demonstrate a ten-
dency to an apparent variation with a decrease in the Reynolds numbers (at Hio <
1.4) shown by band 9, the probable reason of which is discussed next.
A collection of numerous experimental data [30, 32, 34] for nominally two-
dimensional transonic interactions with boundary-layer separation (i.e., closed sym-
bols) and without separation (i.e., open symbols) is shown in Fig. 4.6. Mainly, the
flows with no transverse curvature and with no (or, at most, mild) streamwise curva-
ture are included. In accordance with [34], from this set of studies only the data by
Schofield (1983) and the transonic diffuser experiments of Bogar et al. and Salman
et al. (1983), Sajben et al. (1991) [34], and Morris et al. (1992) [38] address NSTBLIs
coupled with a subsequent adverse-pressure gradient.
Considering that the empirical correlation [32] for the critical shock-wave static-
pressure ratio ξ *i = (p2 /p∞ )* , at which an appearance of separation has been fixed
at transonic Mach numbers (see Fig. 4.4, line 6), corresponds to the tendency for
pp /p∞ predicted by equation (4.14) (i.e., solid line 5), the FIT can be applied to
describe the conditions for incipient intermittent separation [1, 30]. Equating the
plateau pressure (see equation 4.14) to the normal shock strength (see equation
4.13) yields the criterion for true incipient separation, as follows:
2
[2γ M∞ − (γ − 1)] 2 −1/4 1/2
= k M∞
2
M∞ − 1 cf + 1 (4.15)
γ +1
4.2 Two-Dimensional Turbulent Interactions 147
Sawyer (1982)
Ackeret (1947) Bogar et al., Salmon et al. (1983)
Hakkinen (1954) Schofield (1983)
Seddon (1960) Delery (1983)
Alber, Bacon (1973) γ = 0.08, Sajben et al. (1991)
Vidal, Vitliff (1973)
Kooi (1975, 1978) γ = 0.55, Morris et al. (1992)
Abbis (1976) Grodzovskyi (1961)
Vidal, Catlin (1977)
Padova (1979) Zheltovodov et al. (1990)
1.6
Developed separation
1.5
Ms
1.4 MS
Ms
MO Mef 3 44
1.3
Mef
2
Mi
1.2 Mi 1
1.0
3
10 104 105
Reθ
Figure 4.6. The normal shock interaction data for regimes with separation (black symbols)
and without separation (opened symbols) and prediction of “true” incipient-separation con-
ditions: 1 – the FIT (equation 4.15), 2 – theory of Inger [40, 41]; incipience of developed
separation at pson = pp : 3 – from FIT, 4 – from experimental correlations by Grodzovskyi
[32], Zukoski [33], and Zheltovodov [46].
0
1.1
0.7
0.9
25
y (mm)
M∞ = 1.48
50
75
−50 −25 0 25 50 75
x (mm)
(a)
Figure 4.7. Computed Mach contours,
0.7 0.8 M = 0.1, in two-dimensional diffuser (a),
comparison of computations (solid lines)
and experiment (symbols) for top wall
0.6 0.6
pressure and computed surface skin friction.
(b) (Blosch et al. [47]) (the total pressure pt
wall pressure Pw /p1
0.4 0.2
0.3 0
0.2 −0.2
−50 0 50 100 150
x (mm)
(b)
Reθ ≤ 104 using the plateau pressure pp calculated on a basis of FIT (see equation
4.14) are shown by solid line 3. The lower boundary of the scattered data at the
separation-flow regime (i.e., black symbols) corresponds to line 3 up to this value of
the Reynolds number. This behavior is in agreement with a decrease in the fullness
of velocity profile [3, 39, 45] and, respectively, increasing Hi in a relaxing turbulent
boundary layer in the immediate region downstream of the laminar-turbulent tran-
sition region as the Reynolds number is increased up to Reθ ≈ 104 . Above these
values, the tendency is reversed and the boundary-layer profile becomes fuller and
Hi decreases with an increasing Reynolds number. The corresponding boundary
for Ms at Reθ > 104 is shown by shaded band 4. The scatter of experimental data
and the difference in the empirical correlations [32, 33, 46], which were used for
the plateau pressure at the high Reynolds numbers, are characterized by the width
of this band. As shown, this boundary is qualitatively similar to theory [40, 41] for
Mi (Reθ ) shown by line 2. The demonstrated boundaries specify the regions of exis-
tence of unseparated, intermittent-separation, and developed-separation regimes in
NSTBLI conditions.
xo x
2.5 Shock generator
M∞ δo Shock
2
M∞ = 1.44,
Reδo = 5 × 105
P/P∞
1.5 Computations
(Viegas, Horstman, 1978)
- 0-Eqn., Cebeci-Smith
- 2-Eqn., Wilcox-Rubesin
1
- Experiment
(Matteer, Brosh, Viegas, 1964)
5
−10 −5 0 5 10 15 20 25
(x − xo)/δo
2
1
Cf × 10−3
−1
−10 −5 0 5 10 15 20 25
(x − xo)/δo
Figure 4.8. Comparison of computations and experiment for surface-pressure and skin-
friction distributions in conditions of transonic normal shock interaction in axisymmetric flow
(Viegas and Horstman [48]).
shock in the vicinity of the top wall (Fig. 4.7a) corresponds to the regime of devel-
oped boundary-layer separation and is in qualitative agreement with the experi-
ment. The computed wall-pressure profile on the top surface (Fig. 4.7b, solid line)
underestimates the upstream influence of the interaction, which is typical for the
standard zero- and two-equation turbulence models. The computed skin-friction
coefficient shows a significant region of reversed flow. The inflection point in the
surface-pressure distribution is shown in the calculations and experiments in the
vicinity of the minimal skin-friction region in the separation zone. The experimen-
tal and calculated static-pressure ratio in this point p/p∞ ≈ 1.7 (see Fig. 4.4, symbol
7) is within the limits of the generalized experimental data for the plateau-pressure
value pp /p1 and predicted by the FIT. The computed and experimental velocity pro-
files at several locations upstream and downstream of the interaction are in close
agreement [47].
A RANS simulation of the normal-shock interaction inside a cylindrical test sec-
tion [48] is shown in Fig. 4.8. The surface pressure is accurately predicted by both
algebraic (Cebeci-Smith) and two-equation (Wilcox-Rubesin) models (Fig. 4.8a).
Nevertheless, the two-equation model demonstrates better agreement with the
experiment in the skin-friction distribution (Fig. 4.8b) and velocity-profile develop-
ment [48] downstream of the interaction. The two-equation model predicts the stage
150 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
1.5
LES
k-w (original)
1.0 k-w (SST)
0.5
−Cp
0.0
Experiment
(Harris, 1981)
−0.5
NASA-0012
M1
−1.0 α
M∞
C
−1.5
0.0 0.2 0.4 0.6 0.8 1.0
x/c
(a)
0.004
laminar-turbulent
transition suction side
0.003
0.002
Cf
0.001
pressure side
0.000
separation
zone
−0.001
0.0 0.2 0.4 0.6 0.8 1.0
x/c
(b)
Figure 4.9. LES (Nakamori and Ikohagi [49]) and RANS (Wilcox [52]) computations of a
normal shock-wave turbulent boundary layer interaction on an NACA 0012 airfoil at M∞ =
0.8, M1 = 1.38, Rec = 9 × 106 , α = 2.26 degrees.
close to incipient separation (cf = 0); however, the one-equation model demon-
strates a region of reversed flow. The computed and experimental profiles of tur-
bulence kinetic energy are generally similar [48]. Nevertheless, the experimental
results show significant free-stream turbulence downstream of the shock, which is
not evident in the calculations. The inclusion of this effect tends to increase the cal-
culated skin-friction downstream of reattachment relative to the values shown in
Fig. 4.8b.
The capability of LES [49] to predict an NSTBLI on a NACA 0012 airfoil at
Mach 0.8, angle of attack α = 2.26 degrees, and Reynolds number Rec = 9 · 106
based on the airfoil chord is illustrated in Fig. 4.9. The Scale Similarity Model of
Bardina et al. [50] was used. The normal shock appears on the upper (i.e., suc-
tion) side of the airfoil at about 60 percent of the chord length, with Mach number
4.2 Two-Dimensional Turbulent Interactions 151
y/h
a h = 15 mm
2 o−δ
A 8 x/h
4 6
0 2
0 Experiment
y/h
b
2
A 8 x/h
4
0
Computations: k − ω
0
2 Figure 4.10. Comparison between
c h = 15 mm
P experiment and RANS with the stan-
6 mm
P∞
1
dard k-ω (Borisov et al. [53]) and the
k-ε turbulence models (Horstman and
Cf k−ω
d k−ε Zheltovodov [54]) for compression/
Cf0
decompression ramp flow. M∞ = 2.9,
1 α = 8◦ , Tw /Taw ≈ 1, Reδo = 1.22 × 105
at h = 15 mm (h/δ o = 4.4) and Reδo
0,5 = 6.36 × 104 at h = 6 mm, h/δ o = 2.6
−4 0 4 A 8 12 x/h (from Zheltovodov [1], Borisov et al.
[53]).
y,
mm e k − ω, k-ε
8
6.66 8.47 9.87
2.33
x/h = −0.73
4
0
0,5 0,5 0,5 0,5 0,5 1,0
U/U∞
corresponding rise in the mean surface pressure (Fig. 4.11c), coalescing into a shock
wave3 (i.e., the separation shock). A “plateau” in the mean surface pressure forms
in the region of reversed flow between points S and R, as shown in the velocity
profiles (Fig. 4.11e). The surface-skin friction cf = 0 at both points (Fig. 4.11d). A
second compression-wave system forms in the vicinity of mean-reattachment point
R as the flow is deflected by the corner surface and the external flow compresses to
coalesce into a shock (i.e., the reattachment shock). The two shock waves intersect
to form a λ-shock with a slip line (see Fig. 4.11a) and a secondary expansion fan, as
shown in Fig. 4.2c, or a weak shock extending from a triple point at a low M∞ . Such
3 The shock wave may form either within or outside the boundary layer depending on M∞ , Reδo ,
and α.
4.2 Two-Dimensional Turbulent Interactions 153
y/h
a h = 15 mm
2
1
2
3 A 4 6
2 x/h
0 R Experiment
−2 S 0
y/h
b
2
A 4 8
0 x/h
R Computations: k − ω
−2 S 0
P
Figure 4.11. Comparison between P∞ h = 15 mm
c
experiment and RANS with the stan- 4 6 mm
dard k-ω (Borisov et al. [53]) and the k−ω
k−ε
k-ε turbulence models (Horstman and
Zheltovodov [54]) for compression- 2
decompression ramp flow. M∞ = 2.9,
Cf
α = 25◦ , Tw /Taw ≈ 1, Reδo = 1.48 × 105 d
Cf0
at h = 15 mm (h/δ o = 3.7), and Reδo =
6.36 × 104 at h = 6 mm (h/δ o = 2.6): 1
1 – sonic line (M = 1), 2 – zero velocity S R
line, 3 – line of maximum reversed 0
flow velocity (from Zheltovodov [1],
Borisov et al. [53]). −1
−4 0 A 4 8 x/h
y,
5.23
mm e k − ω, k−ε
0.6 0.73
8
x/h = −2.2
−0.8
0
0,5 0 0,5 0 0 0 0,5 1,0
U/U∞
the decompression ramp. Limited experimental skin-friction data are available for
h = 15 mm and they are supplemented with additional experimental data for
h = 6 mm.
For configurations with a short compression-surface length, the flow proper-
ties begin to depend on the length (or the ratio h/δ o ) because the influence of the
expansion of the reattaching subcritical boundary layer on the back tip of the step
is transmitted in the upstream direction [55, 56]. If the reattachment point reaches
the shoulder, the flow is basically the same as that ahead of a normal step, and the
separation point reaches a distance approximately 4.2 step heights upstream of the
shoulder [56]. These flow regimes are demonstrated in Figs. 4.12 and 4.13. The flows
also exhibit a small, strongly unsteady separation zone immediately downstream of
the step’s top A (see Figs. 4.12a and 4.13a) and secondary separation (S ) and reat-
tachment (R ) lines initiated by a localized vortex emerging in the bottom corner
ahead of the step [46, 53, 57] (see Fig. 4.13a,d).
The computed flowfields (see Figs. 4.12b and 4.13b) correctly describe the
shock-wave and expansion-wave structures, and they accurately predict the loca-
tion of separation and reattachment as well as behavior of the sonic line (1), dividing
streamline (2), zero-velocity line u = 0 (3), and maximum reversed-flow velocity line
(3). Nevertheless, the computations do not exhibit a small, strongly unsteady sepa-
ration zone near the top of the steps. The computed and experimental surface pres-
sures (see Figs. 4.12c and 4.13c) demonstrate good agreement except in the vicinity
of the decompression corner at α = 45 degrees. The computed and experimental
skin-friction coefficient is displayed in Figs. 4.12d and 4.13d. Limited experimental
skin-friction data are available for h = 15 mm, and additional experimental data for
h = 6 mm are shown. The agreement between computation and experiment is good;
similar agreement is demonstrated in Figs. 4.12e and 4.13e for the mean-velocity
profiles in the separation zone with the reversed flow.
A series of RANS computations [58] of the compression corner at Mach 2.96
for α = 15 to 25 degrees at Reδ = 1.5 × 105 is demonstrated in Fig. 4.14. The
zero-equation turbulence model of Cebeci and Smith was used with a relaxation
model for the turbulent-eddy viscosity μt = μt∞ + (μtequil − μt∞ )[1 − e−(x−x∞ )/λ ],
where μt∞ is the eddy viscosity evaluated at location x∞ immediately upstream
of the compression-corner interaction and μtequil is the standard (i.e., equilibrium)
Cebeci–Smith eddy viscosity. The assumed relaxation length is λ = 10δ o , where δ o
is the boundary-layer thickness immediately upstream of the interaction. The com-
puted and experimental mean-surface pressure is displayed in Fig. 4.14a, where L =
1 ft is the length of the upstream plate and S is the distance along the surface. The
overall agreement is good; in particular, the position of upstream influence (i.e., the
location of the initial mean-pressure rise) is accurately predicted as a function of the
compression angle. The mean-streamwise computed and experimental velocity pro-
files at and downstream of the mean-separation location are shown in Fig. 4.14b.
Good agreement is observed; nonetheless, the relaxation length λ = 10δ o is not
a universal length scale; a value twice as large is necessary to obtain close agree-
ment with experiments for the incident-oblique shock configuration at the same
free-stream Mach number and comparable Reynolds number (see the following
discussion).
4.2 Two-Dimensional Turbulent Interactions 155
y/h h = 15 mm
1 a
2 2
3
4
A 2 4
R x/h
0
−2 0 Experiment
S
y/h
b
2
A 2 4
R
0 x/h
−4 S −2 0 Computations: k − ω
−8 −4 0 A 4 8 x/h
16
−0.81
e −1.33
y,
mm k−ω −2.07
0
0,5 1 0 0 0 0,5 1,0
U/U∞
y/h a Experiment
1
2
2
3
4 A R 2
x/h
S′
0 h = 15 mm
S −4 −2 R′ 0
y/h
b Computations: k − ω
R 2
x/h
S′
0
S −4 −2 R′ 0
P Figure 4.13. Comparison between exper-
h = 15 mm
P∞ c 6 mm iment and RANS for compression/
3 k − ω: decompression ramp flow at M∞ = 2.9,
h = 15 mm α = 90◦ , Tw /Taw ≈ 1, Reδo = 1.44 × 105
6 mm at h = 15 mm (h/δ = 3.8) and Reδo =
1 6.36 × 104 at h = 6 mm (h/δ o = 2.6):
2 1 – sonic line (M = 1), 2 – dividing
d streamline, 3 – zero velocity line, 4 – line
Cf
Cf0 S′ of maximum reversed-flow velocity (from
S R′ R
0 Zheltovodov [1], Borisov et al. [53]).
−8 −4 0 A 4 8 x/h
20
−1.23
y, e −1.77
mm k−ω −1.87
10
x/h = −6.0 2.0
0
0,5 1 0 0,5 0 0 0,5 1,0
U/U∞
conclusion is supported by recent computations [52] (Fig. 4.15c,d) with a new ver-
sion of the k-ω model both with and without the stress limiter. Most important, in the
case with a stress limiter (i.e., solid lines), the computed initial pressure rise matches
the measured rise and the predicted pressure-plateau level in the separation bub-
ble is close to the measurements. The computed cf is also in close agreement with
the data, but a discrepancy with measurements downstream of the reattachment
indicates a lower rate of recovery of the disturbed boundary layer to equilibrium
conditions.
Despite intensive experimental and computational study, several aspects of rel-
evant physics involved in these flows remain poorly understood and some of the
4.2 Two-Dimensional Turbulent Interactions 157
(a)
(b)
Figure 4.14. Comparison between RANS computations on a basis of relaxation version of
Cebeci-Smith model with experimental data for compression-ramp flow (Shang and Hankey
[58]).
1.5
y,cm
2
EXPERIMENT (Settles. 1975)
0 − EQ MODEL 1.0
1 − EQ MODEL
1 2 − EQ J-L MODEL
2 − EQ W-R MODEL 0.5
0
−3 −2 −1 0 1 2 3 4 5 0
0.5 0 0.5 0 0 0 0 0.5 1.0
x /δo u/u∞
(a) (b)
Pw / P∞ 103 cf
5 4
k – ω model:
without limiter 3
4
with limiter
2
3
1
2
0
0 −2
−4 −2 0 2 4 6 −4 −2 0 2 4 6
x/ δo x/ δo
(c) (d)
Figure 4.15. Comparison of RANS computations with experimental data for compression-
ramp flow at α = 24◦ , M∞ = 2.8, Reδ o = 1.33 × 106 , Tw /Twa = 0.88: a, b – Viegas and Horstman
[48]; c, d – Wilcox [52].
2 3
M∞
1 4
S R R
S
6 5
P
P
P
R
Figure 4.16. Specific physical features of flow over compression-decompression ramp config-
uration (Zheltovodov [1, 2, 80].
4.2 Two-Dimensional Turbulent Interactions 159
(a) (b)
Figure 4.17. Separation length as function of the ramp angle at M∞ = 3.5, Tw /Taw = 1.0
(Appels and Richards [60]): a) Reδo = 2.07 × 105 , b) Reδo = 9.34 × 104 , 2.07 × 105 , 3.23 × 105 ,
4.37 × 105 , 5.38 × 105 , and 6.42 × 105 .
35
SETTLES, BOGDONOFF
a & VAS [62,63]:
SHOCK WAVE PHOTOS
30 SEC 1 OIL FLOW
KU SEC 2 OIL FLOW
EH
N SEC 3 OIL FLOW
M∞ = 5.4 PRESSURE “KINK”
M∞ = 3.5
α∗S, KE Tw/Taw = 0.72
SS
αiL LE αS∗ , APPELS & RICHARDS [60]
R
1.0 MKE
KUE & THO
HN ROSHKO
20 M∞ = 2.9 LAW
SPAID &
FRISHETT
[61]
15 αS∗ ,
Tw/Taw < 1.0 ELFSTROM
HOLDEN [64]
=1.0 [65]
10
APPELS & RICHARDS [60]
Figure 4.18. Summary of data for critical compression-ramp angles α i and α S * corresponding,
respectively, to incipience of “small” and “large-scale” separation: 1 – δo = 8.38 ÷ 8.0 mm
(Spaid and Frishett [61]); 2 – 9.96 ÷ 8.08 mm (Spaid and Frishett [61]); 3 – 30 ÷ 10 mm
(Settles [63]); 4 – 3.4 mm (Zheltovodov et al. [66]); 5 – 8.27 mm (Kuntz et al. [67]; 6 – 6.5 ÷
5.9 mm (Appels and Richards [60]); 7 – 19.8 ÷ 18.2 mm (Appels and Richards [60]).
shows good agreement with experiments except for a doubtful tendency for increas-
ing α s* at higher Reδo .
The experimental data for α i (Fig. 4.18b) represent the best efforts in the
exceedingly difficult task of measuring the onset of small separation (see, e.g.,
Appels and Richards [60], Spaid and Frishett [61], and Settles et al. [62, 63]). It
is clearly evident, however, that α i is smaller than α s* for a given M∞ and Reδo . The
α i values for the adiabatic wall at M∞ = 2.9 (symbols 1, 3 − 5) range from 6.5 to 12
degrees for 104 < Reδo ≤ 107 and increase with Reδo for Reδo > 105 . This behavior
is consistent with the decreasing fullness of velocity profiles (i.e., increasing Hi ) in
a relaxing boundary layer immediately downstream of the laminar-turbulent transi-
tion up to Reδo ≈ 105 [39, 45], which becomes less resistant to separation. Also, the
increased thickness of the interacting viscous-wall layer and its subsonic portion at
4.2 Two-Dimensional Turbulent Interactions 161
low Reδo promotes an earlier onset of small separation. The increasing fullness of
velocity profiles (i.e., decreasing Hi ) with reduced shear stress at Reδo > 105 makes
the boundary layer more resistant to separation. Also, the movement of the sonic
line closer to the wall and the decrease in viscous-layer thickness with increasing
Reδo reduces the thickness of the interacting wall layer, which defines the scale of
the interaction and causes a rise in α i . It has been noted (see, e.g., Spaid and Frishett
[61]) that wall-cooling suppresses incipient separation (Fig. 4.18b, symbol 2).
The five stages of turbulent-separation development for adiabatic-wall CR and
CDR (Fig. 4.19) are I: unseparated flow; II: intermittent separation; III: develop-
ing small-scale separation; IV: large-scale separation in CR; and V: maximum-scale
separation in front of the forward-facing step. The features of separation devel-
opment in the vicinity of a normal shock explain the stages of the separated-
flow development near a two-dimensional CR. Following Elfstrom’s [65] concept,
162 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 4.20. Comparison of pressure ratio in the “plateau” region pp /p∞ (symbols 1–3) and
in separation point ps /p∞ (4–6) with critical shock-wave strength ξ s in compression ramp: 1,
4 – Zukoski [33]; 2, 5 – Zheltovodov [46]; 3, 6 – Zheltovodov and Schülein [73].
Zheltovodov [1] proved that incipient small separation and the early stages of its
development at low to moderate Reynolds numbers (Reδo ≤ 105 ) are characterized
by the CR critical angles α i , α ef , and α s (Fig. 4.19), corresponding to detachment
shock-wave conditions realized at different distances from the surface where the
local critical Mach numbers achieve the values Mi , Mef , and Ms , respectively. These
critical Mach numbers characterize the stages of true incipient separation, beginning
effective intermittent and developed separation in the vicinity of the normal shock,
respectively (see Fig. 4.4). The data [60, 61, 62, 63, 66, 67] for incipient-separation
conditions in CRs with adiabatic walls (see Fig. 4.18b) are concentrated close to
these boundaries (Fig. 4.19, symbols 1, 3, 6, 8, and 9) with additional data [68, 69, 70,
71] (i.e., symbols 10–13, 16) and support the proposed considerations. Decreasing
the external-flow Mach number increases the thickness of the transonic near-wall
portion of the boundary layer and produces earlier onset of resolved small separa-
tion, even at the higher Reynolds number. In this connection, the additional data
[68] at M∞ = 1.5, Reδ = 1.79 · 105 (Reθo = 17,500) are indicative because a weak
intermittent separation was found at these conditions at α = 6 degrees (i.e., sym-
bol 10) and more prominently at α = 9 and 12 degrees (i.e., symbols 11 and 12,
respectively).
The strength (measured by the pressure ratio) of the oblique shock wave for the
CR ξ inv = ξ s corresponding to α = α s (see Fig. 4.19) for a range of Mach numbers
M∞ ≈ 2 − 5 practically coincides with experimental data and an empirical corre-
lation by Zukoski [33] for the static-pressure ratio in the separation point ps /p∞
(Fig. 4.20). The following simple expression:
ξs = ps / p∞ = 0.365M∞ + 1 (4.17)
ξ inv with the plateau-pressure ratio ξ p = pp /p∞ realized in the separation region
[33]:
ξinv = ξ p = pp / p∞ = 0.5M∞ + 1 (4.18)
another value of the critical CR α = α s max can be predicted. As shown in Fig. 4.19,
this prediction (i.e., line α s max ) is in good agreement with the upper limit of the CR
values revealed in experiments at high Reynolds numbers Reδo > (1 − 2) · 105 .
The next boundary marked as α s* in Fig. 4.19 is represented by band 14 cor-
responding to the incipient large separation [72] at Reδo > 105 . The tendency for
increasing α s* with Mach number was predicted by Elfstrom [65] (i.e., band 15
between two dashed lines) for 2 · 105 < Reδo ≤ 107 . The experimental data for α s*
(i.e., symbols 2, 5, 7) are also in good agreement with this correlation. Thus, area
III in Fig. 4.19 corresponds to developing small-scale separation and area IV char-
acterizes large-scale separation when the attached oblique shock should form for
CRs in inviscid flow. The last boundary marked as α d corresponds to the appear-
ance of a detached shock ahead of the step with an inclined face of limited length
at inviscid-flow conditions. The maximum-scale separation zones are realized above
this boundary in region V, in fact, in NSTBLI conditions ahead of the forward-facing
steps.
The stages and features of separation development explain the different behav-
iors of the upstream influence distance Lu /δ o (i.e., the location of the rise in surface
pressure, measured from the corner) and the separation-line location Ls /δ o (i.e.,
measured from the corner) with increasing α for adiabatic and cold walls [61] at
M∞ ≈ 2.9 and at low Reynolds numbers Reδo < 105 (Fig. 4.21). In the case of an
adiabatic wall (Fig. 4.21a), the linear behavior of Lu /δ o corresponds to the unsepa-
rated flow stage I, and its sudden reduction or constant value occurs at intermittent-
separation conditions (α ef ≤ α ≤ α s , stage II). The development of a small-scale
separation in the near-wall portion of the boundary layer at α s ≤ α ≤ α s* (i.e., stage
III) is accompanied by an increase in Lu /δ o . The rate of increase is significantly
greater in the next stage of large-scale separation development (i.e., stage IV) in
the external portion of the boundary layer. The separation-line position Ls /δ o is
practically independent of the Reynolds number in conditions of intermittent and
small-scale separation (i.e., stages II and III), but a tendency to increase with rising
Reδo = (3.63 − 5.92) · 104 emerges on the stage of large-scale separation develop-
ment (i.e., stage IV). This can be explained by a decreasing mean velocity-profile
fullness in the external portion at such conditions. It is remarkable that the rise
of Lu /δ o at α s < α ≤ α s* (i.e., stages III and IV) is caused mainly by increasing
separation distance Ls /δ o . The wall-cooling promotes the degeneration of subsonic
and viscous near-wall portions of the interacting boundary layer and the suppres-
sion of separation (Fig. 4.21c). As a result, an obvious reduction of the upstream-
influence distance Lu /δ o with decreasing wall-temperature ratio Tw /Taw , especially
pronounced at Tw /Taw = 0.775 – 0.474, occurs at all stages I–IV (Fig. 4.21b) com-
pared to the adiabatic-wall case. The value α s* ≈ 16 − 17 degrees determined from
Figs. 4.21a,c is approximately the same for adiabatic and cooling walls and inde-
pendent of the Reynolds number in the considered range (see the band of data by
Spaid and Frishett [61] for Tw /Taw = 1.0 and Tw /Taw < 1.0; Fig. 4.18a) as in the data
by Settles et al. [62] (i.e., open symbols) at Reδo > 105 . This value corresponds to
164 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Reδo increases
Lu/δ0 , 1.8
2.89 4.10 0.775 8.69
1.6 2.93 3.04 0.604 9.45
Ls/δ0 2.97 2.18 0.474 9.96 IV
1.4 Ls
δ0
1.2
1.0 a
2.2
0.8
2.0 III
0.6
1.8
0.4 II α∗S Lu
1.6
δ0
0.2
1.4
I α, deg
0
Experimental
range
50
M∞ RANS:
45 δ -1 - 13
-2 - 14
LSR -3 - 15
40 DNS -4 - 16
LES RANS -5 - 17
35 -6 - 18
-7 - 19
-8 - 20
30 -9 - 21
- 10 - 22
Lsep
25 - 11 - 23
- 12 - 24
DNS: - 25
20
- 26
- 27
15 - 28
LES: - 29
- 30
10
- 31
- 32
5 - 33
- 34
0
4
10 105 106 107 108 109
Reδ0
Figure 4.22. Comparison of separation length for RANS, DNS, and LES with experiment
(from Knight et al. [7]). RANS: 1 – 3 – β = 25◦ (Shang and Hankey [58]); 4 – 7 – β = 20◦
(Horstman and Hung [74]); 8 – 11 – β = 24◦ (Shang and Hankey [58]; 12 – 15 – β = 16◦ , 16
– 19 – β = 20◦ , 19 – 20 – β = 24◦ (Visbal and Knight [75]); 21, 22 – β = 16◦ , 23 – (Ong and
Knight [76]); 24 – β = 25◦ , k–ε model (Horstman and Zheltovodov [54]); 25 – β = 25◦ , k-ω
model (Borisov et al. [53]). DNS: M∞ = 3, β = 18◦ : 26 – (Adams [98]), 27 – (Rizzetta and
Visbal [111]); 28 – M∞ = 2.9, β = 24◦ (Ringuetta, Wu, and Martin [112]). LES: M∞ =2.95: 29
– β = 25◦ (Loginov et al. [103, 105]), 30 – β = 25◦ (Yan et al. [100]), 31, 32, 33 – β = 18◦ , 20◦
and 24◦ (Rizzetta and Visbal [111]), 34 – β = 18◦ (El-Askary [96]).
An empirical scaling law [73] for the total extension of large-scale separa-
tion in CRs in regime IV (see Fig. 4.19) at adiabatic-wall conditions L* sep =
LSR /Lc = f(Reδo ) (Fig. 4.22; the region between solid lines) incorporates the
effects of Reynolds number Reδo , Mach number M∞ , compression angle α (through
the inviscid-pressure rise p2 /p∞ ), and boundary-layer thickness δ o . Here, LSR
is the shortest distance between separation S and reattachment R points, Lc =
(δ o /M∞ 3 )(p2 /pp )3.1 , and the plateau pressure pp /p∞ is described by the empirical
formula (4.18) [33]. There is a distinct change in trend at Reδ ∼ = 105 that corresponds
to the variation of the fullness of the mean-velocity profile with Rδo noted previ-
ously. Fig. 4.22 is a summary [2, 7] of several RANS predictions of separation length
for two-dimensional CR configurations. Results of different researchers (i.e., sym-
bols 1–25) who used various algebraic turbulence models (including the Baldwin-
Lomax model and its modification) [58, 74, 75, 76], the k-ε and k-ω models [53,
54], and various ad hoc modifications correspond to the range β = 16–25 degrees,
M∞ = 1.96–3.0 at Reδo ≥ 105 . As shown, in general, the simulations do not predict
accurately the separation length in comparison with the experimental correlation
166 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Lsep Lsep
h = 5.0 RANS
h
5
Lsep
= 4.2
h
3 -1 -6
M∞ -2 -7
δO
-3 -8
R h -4 -9
S α
Lsep -5 - 10
1
0 45° 90°
α
Figure 4.23. Correlation for extension of the maximum-scale separated zone ahead CDR in
regime V (at h/δ o > 1.5). α = 25◦ , 45◦ and 90◦ : 1 – M∞ = 2.0, 2 – 2.25, 3 – 3.0, 4 – 4.0 (Zhel-
tovodov [46]; Dem’yanenko and Zheltovodov [57]; Zheltovodov, Schülein, and Yakovlev
[66]); 5 – α = 90◦ , M∞ = 5.0 (Schülein and Zheltovodov [77]); α = 24◦ : 6 – M∞ = 1.6, 7 –
2.5 (Coe,, from [33]); 8 – α = 70◦ , M∞ = 3.0 (Hahn [78]); 9 – scatter of data for α = 90◦ at M∞
= 2.1– 6.5 (Zukoski [33]); 10 – RANS with the k-ω model at M∞ = 2.9 (Borisov et al. [53]).
(i.e., solid lines). The behavior of calculations on the basis of LES and DNS is dis-
cussed later in this chapter.
Another simple empirical correlation Lsep /h ∼ = 4.2 can be used to predict the
extent of the maximum-scale separation in front of CDR (Fig. 4.23) in regime V (see
Fig. 4.19) corresponding to the situation with a detached normal shock in inviscid-
flow conditions. This correlation was demonstrated by Zukoski [33] for normal
(α = 90 degrees) steps at M∞ = 2.1 − 6.5 and can be used at h/δ o > 1. Shown in
Fig. 4.23 are additional data [46,57] for α = 90 degrees at M∞ = 3.0 and 4.0 (i.e.,
symbols 3 and 4) and at M∞ = 5.0 (i.e., symbol 5 [77]) correspond to this correlation
and are located in the limits of scatter band 9 of the data [33]. Experimental data
by Zheltovodov et al. [66] for α = 25 degrees at M∞ = 2.0 and 2.25 (i.e., symbols 1
and 2), α = 45 degrees at M∞ = 2.25, 3.0, and 4.0 (i.e., symbols 2, 3, and 4), and by
Coe (i.e., symbol 7, see [33]) at M∞ = 2.5, as well as by Hahn [78] at α = 70 degrees
at M∞ = 3.0 (i.e., symbol 8) are also in good agreement with this correlation. The
RANS computations [53] with the k-ω model for α = 45 and 90 degrees at M∞ =
2.9 demonstrate good agreement with the experiments (i.e., line 10) (see also Figs.
4.12 and 4.13), although a small overprediction of Lsep /h is seen at α = 90 degrees.
The data [46, 57] for α = 90 degrees at M∞ = 2.0 and 2.25 (i.e., symbols 1 and 2) and
by Coe at α = 45 degrees at M∞ = 1.6 (i.e., symbol 6) demonstrate the increase of
Lsep /h up to the value Lsep /h ∼
= 5.0. This tendency corresponds to the data consid-
ered by Zukoski at 1.4 ≤ M∞ < 2.1. It can be explained by a significant rise of the
normal-shock detachment distance ahead of the step in inviscid-flow conditions at
low Mach numbers (see, e.g., [79]).
a
7 M∞ = 4 M∞
b R A h x
P/P∞
S 0 α
5 3
R Ch / Cho
c
3 S 2.2
3
S M∞ = 4
1
−4 −2 0 2 A 4 x/h
2
3
2
−12 −8 −4 0 A 4 8 x/h
Figure 4.24. Surface pressure (b) and heat-transfer coefficient (c) distributions in the vicinity
of CDR configuration (a) at α = 25◦ : b – h/δ 0 = 3.3–3.9, Reδo = 1.25 × 105 , 1.48 × 105 , and
1.76 × 105 , respectively, at M∞ = 2.2, 3.0 and 4.0, Tw /Taw ≈ 1; с – h/δ 0 = 2.3 – 2.4, Reδo =
8.06 × 104 , 9.3 × 104 , and 1.16 × 105 , respectively, at M∞ = 2.0, 3.0, and 4.0, Tw /Taw = 1.04 –
1.05 (from Zheltovodov [80] and Zheltovodov et al. [81]).
configurations [80, 81]. Figure 4.24 shows the variation of the surface-pressure p/p∞
(Fig. 4.24b) and the heat-transfer coefficient Ch /Cho distributions (Fig. 4.24c) for
a CDR at α = 25 degrees and different Mach numbers M∞ , where Cho is the
heat-transfer coefficient value immediately upstream of the initial surface-pressure-
rising region. Some of the observed variations in the surface-heat transfer correlate
qualitatively with the surface-pressure distributions, including (1) the increase in
length of the separation zone upstream of the CR with a decreasing Mach number;
and (2) the appearance of a localized region of positive surface-pressure and heat-
transfer gradients immediately downstream of the top of A at M∞ = 2.2 and 2.0 (Fig.
4.24b,c), with the advent of an additional small separation zone (Fig. 4.24a) at α ≥
α d in regime V (see Fig. 4.19). However, there are notable differences between the
surface-heat transfer and pressure distributions, including (1) the maximum heat
transfer is the same for M∞ = 3.0 and 4.0 upstream of the top of A (Fig. 4.23c),
whereas the maximum pressure increases with M∞ (Fig. 4.24b); (2) the significant
rise of the heat transfer on the step’s top surface downstream of point A with an
increasing M∞ value despite the same pressure levels at least at x/h > 3.5; and (3)
the decrease in the heat-transfer level in the separation zone on the CR with an
increasing Mach number opposite the rising surface-pressure level between separa-
tion S and reattachment R lines.
168 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 4.25. Longitudinal mass-flux fluctuations in the external flow (a) and along a line of
maximal value in the external portion of boundary layer (b): M∞ = 2.95, Reδo = 6.36 × 104 ,
Tw /Taw ≈ 1, h/δ o = 2.5 – 2.64 (from Zheltovodov [80]).
producing a false turbulence peak in the measured profiles. It is evident that the
SBLI in the CR significantly enhances the mass-flow fluctuations in the disturbed
external flow (Fig. 4.25a) and in the outer region of the boundary layer on the com-
pression surface (Fig. 4.25b). The turbulence amplification increases with the rising
CR inclination angle α and, despite partial suppression by the expansion waves in
the vicinity of the decompression ramp, the heightened level remains downstream in
the external flow over a long distance, whereas a more rapid relaxation of the max-
imum fluctuations to their initial value is observed in the boundary layer. Predom-
inant acoustic fluctuations were revealed in the external flow [84, 85] that are asso-
ciated with moving weak shocklets generated by large-scale vortices in the external
part of the disturbed boundary layer.
Figure 4.26 compares RANS computations with different turbulence models
and experimental data for surface-heat-transfer coefficient distributions supporting
previous conclusions regarding the limitations of the RANS eddy-viscosity mod-
els. As shown, Horstman’s computations [86] with the standard Jones-Launder k–
ε turbulence model and its two-layer modification by Rodi [86] did not obtain an
acceptable agreement with experimental data by Zheltovodov et al. [81, 83]. (The
scatter bars in Fig. 4.26a indicate the level of spanwise heat-transfer variation in the
experimental data in accordance with additional measurements by Trofimov and
Shtrekalkin [87] caused by Görtler-type vortices [see Fig. 4.16, sketch 5].) The two-
layer k-ε model by Rodi shows improvement in the prediction of wall-heat transfer
in the CR vicinity but underestimates the level downstream of the expansion corner
similar to the first model. Later attempts [88] were directed toward improving the
prediction of different parameters in the separation zones ahead of CR and CDR
test configurations by controlling the balance between the turbulence production
and dissipation in the framework of the Wilcox k–ω model. Significant improve-
ments were achieved by limitating the specific turbulent kinetic dissipation rate
ωe ∼ ωw 10 − 3 , where ωw is the maximum value on the wall. These computations
improved the heat-transfer prediction in the CR vicinity (Fig. 4.26a), as well as
surface-pressure, skin-friction, and velocity profiles by modeling the phenomenon
of the reverse-flow relaminarization in the separation zone in accordance with Zhel-
tovodov’s [46] experiments (see also [89]). However, as shown in Fig. 4.26a, signif-
icant underestimation of surface-heat transfer remained on the top surface down-
stream of the expansion fan, as in other turbulence models.
separation shock is unsteady and exhibits both low-frequency (i.e., small compared
to the frequency U∞ /δ of the energy-containing eddies in the incoming turbulent
boundary layer) and high-frequency motions. This is illustrated by experimental
results [68] for a CR at the low Mach number M∞ = 1.5. Figure 4.27 presents the
variation in rms wall-pressure fluctuation intensity σ p (in the 0–65 kHz bandwidth)
normalized by the mean wall pressure upstream of the interaction for angles α = 6,
9, and 12 degrees. The fluctuation intensity increases with the length of the inter-
action region upstream of the ramp (x/δ 0 < 0), with rising α (i.e., inviscid oblique
shock strength ξ ) under the influence of both shear-layer flapping and decays down-
stream of the interaction (x/δ 0 > 0) with turbulent stress relaxation. The positions
4.2 Two-Dimensional Turbulent Interactions 171
of the low-frequency peaks (in the 0–4 kHz bandwidth) – analyzed in detail herein –
are indicated by arrows. The appearance of a small intermittent-separation bubble
at α = 6 degrees in the experiments at M∞ = 1.5 (see Fig. 4.19, symbol 10) is in
agreement with the predicted critical value α = α ef corresponding to the beginning
of the intermittent-separation regime. More prominent unsteady separation zones
were observed at α = 9 and 12 degrees (i.e., symbols 11 and 12).
The power spectra [68] for the wall-pressure fluctuations p’ presented in the
terms of f · G( f ) versus ln( f ), where G( f ) is the power spectral density and f is
a temporal frequency, are characterized by many common features; the 12-degree
case is presented as an example in Fig. 4.28. Two characteristic regimes are evi-
dent. The fluctuations with frequency on the order of the peak value fp ≈ U∞ /δ
≈ 30 kHz correspond to the characteristic frequency of turbulence in the incoming
boundary-layer flow. These fluctuations exhibit continuous amplification through
the interaction region, and a decrease in the spectral peak fp value is caused by a
reduction in convection velocity during the interaction process. The development
of a secondary central peak centered near a much lower characteristic frequency
on the order fsh ≈ 0.1U∞ /δ is indicated. These low frequencies are characteristic of
shock oscillations and also were observed in several CR experiments at higher Mach
numbers M∞ = 2–5 (see Chapter 9). The ratio of rms pressure fluctuations in the
0–4 kHz bandwidth to the total fluctuation intensity (in the entire 0–65 kHz band) is
shown in Fig. 4.29. The streamwise variation of this ratio exhibits two well-defined
peaks upstream and downstream of the CR apex that are associated with local-
ized separation-shock oscillations and separation-bubble motion (i.e., unsteady flow
reattachment), respectively. The locations of these peaks are indicated by arrows in
Fig. 4.27. The existing upstream peak in the low-frequency wall-pressure fluctua-
tion intensity (within the 0–4 kHz bandwidth) corresponds to the location of the
172 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
separation extent from α. At the same time, significant influence of the step height
h (or the ratio h/δ*) is demonstrated in regime V, which does not exist in regime IV
[90, 91]. Similarly, the influence of the Reynolds number is different in regimes IV
and V. In regime IV, an increase in the Reynolds number causes a broadening of
the range of pressure fluctuations at high frequencies and a decrease in the intensity
at low frequencies associated with the formation of free-separation zones (with free
separation and reattachment points). However, in conditions of fixed reattachment
(i.e., regime V), an increase in the Reynolds number causes an increase in the wall-
pressure fluctuation intensity throughout the entire spectrum. The influence of the
174 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Dolling [94] also demonstrated that the low-frequency unsteadiness of the separa-
tion region/shock foot can be explained by a turbulent mechanism because of the
presence of elongated (i.e., lengths greater than 40δ) spanwise coherent structures
(i.e., “strips”) of uniform momentum in the upstream supersonic boundary layer.
Thus, as suggested by the authors, a physical model that includes the presence of
such elongated strips probably can coexist with Plotkin’s [93] mathematical model
by reevaluation of the constant that defines the restoring mechanism to account for
the presence of large-scale turbulence structures.
LES and DNS enable the investigation of unsteady effects in parallel with
experimental research. Figure 4.31a shows the flowfield structure in a CR at α =
8 degrees, M∞ = 2.9 Reδo = 2 × 104 in accordance with LES by Urbin et al. [95]
using a static Smagorinsky model. An instantaneous image of the shock wave is
visualized by the isosurface p/p∞ = 1.25, and large-scale spanwise turbulent struc-
tures are seen clearly in the undisturbed boundary layer in the horizontal cut of
the streamwise velocity field at a distance from a wall y + = 10 (in wall units). The
oblique shock wave is significantly disturbed due to the interaction with these struc-
tures, and the individual localized intermittent-separation zones with instantaneous
negative velocity appearing in the shock-foot vicinity correspond to intermittent-
separation conditions (see Fig. 4.19, regime II). The computed timewise and span-
wise averaged mean wall-pressure distribution indicated by line 4 (Fig. 4.31b), as
well as a similar prediction by El-Askary [96] using MILES (i.e., line 5), are in good
agreement with the experiment [61] at the low Reynolds number Reδo = 2.18 · 104
(i.e., symbol 1) – which in this regime is close to the data [53, 66, 84, 97] at the higher
Reδo values (i.e., symbols 2 and 3). RANS calculations performed for the higher
Reynolds number using the Jones-Launder k-ε turbulence model [54] (i.e., line 6)
and Wilcox k-ω model [53, 97] (i.e., line 7) also show good agreement with exper-
imental data. However, the LES predicts the mean wall-skin-friction coefficient
Cf ≈ 0 (Fig. 4.31c) corresponding to intermittent-separation conditions (Fig. 4.31a)
in contrast with RANS, which predict Cf > 0. At that time, a small incipient-
separation zone with the size Ls ≈ (0.2–0.3)δ o = (0.44–0.66) mm was revealed by
the oil-flow visualization (see Fig. 4.19, symbol 8) at the higher Reδo , which was not
realized in RANS. As also shown in Fig. 4.31c, various LES give higher skin-friction
levels downstream of the CR (x/δ o > 0) compared to the experiment and RANS
predictions at the higher Reynolds numbers. At the same time, predictions with the
k-ω model (i.e., symbol 7) correspond better to experimental data than computa-
tions with the k-ε model (i.e., symbol 6).
DNS of a supersonic CR flow [98] for α = 18 degrees at Mach 3 and Reθo = 1,685
(Reδo = 2.1 · 104 ) provides insight to the flowfield structure and turbulence behavior
(Fig. 4.32) at conditions corresponding to the boundary between regimes III and IV
(see Fig. 4.19, band 14). The predicted mean wall-skin-friction coefficient distribu-
tion (Fig. 4.32b) displays a small quasistationary separation zone. A time trace of the
spanwise-averaged locations of separation and reattachment is shown in Fig. 4.32c.
The computed shock motion is limited to less than about 10 percent of the mean
boundary-layer thickness δ o and it oscillates slightly around the mean location with
a frequency of similar magnitude to the bursting frequency of the incoming bound-
ary layer f = U∞ /δ o . Schlieren visualization of the instantaneous density-gradient
magnitude averaged in the spanwise direction demonstrates the emergence of
176 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
2.2
Experiment: M∞ = 2.9
b
2.0 - 1, Reδo= 2.18 · 104
- 2, 6.5 · 104 inviscid theory
1.8 - 3, 1.1 · 105
LES, Reδo= 2.0 · 104
-4 K–ω
P/P∞
1.6
-5
K–ε
1.4 RANS, Reδo= 1.1 · 10
5
-6
-7
1.2
LES
1.0
−4 −2 0 2 4
X/δo
C
1.5
K–ε
Cf/CfO
K–ω
1 Experiment:
- 2, Reδo= 6.5 · 104
- 3, 1.1 · 105
0.5 LES LES, Reδo = 2 · 104:
-4
-5
RANS, Reδo = 1.1 · 105
0 -6
-7
−6 −4 −2 0 2 4 6
◦
Figure 4.31. LES and RANS of interaction in CR at α = 8 , M∞ = 2.9, Tw /Twa ≈ 1 (regime II,
from Zheltovododov [2]), a – instantaneous flowfield visualization [95], b – mean wall-surface
pressure, c – mean wall skin-friction coefficient distribution. Experimental data: 1 – [61], 2 –
[53, 66, 84], 3 – [53, 97]; LES: 4 – [95], 5 – [96]; RANS: 6 – k-ε model [54], 7 – k-ω model
[53, 97].
4.2 Two-Dimensional Turbulent Interactions 177
Figure 4.32. DNS of interaction in CR at α = 8◦ , M∞ = 2.9, Reθo = 1685 (Reδo = 2.3 × 104 ),
Tw /Twa = 1 (the boundary between regimes III and IV): a – flowfield Schlieren imitation; b –
mean wall skin-friction coefficient distribution; c – time trace of spanwise averaged separation
and reattachment line (Adams [98]).
distinct coherent vortex structures in the boundary layer downstream of the inter-
action region (Fig. 4.32a) and compression waves emanating from its external edge
that stimulate undulations of the oblique shock front.
Figure 4.33 presents numerical results by Urbin et al. [99] and Yan et al. [100] for
the flow over a CR at α = 25 degrees, M∞ = 2.9, Reδo = 2 × 104 using MILES. The
flow conditions correspond to regime IV (see Fig. 4.19). An instantaneous image of
the λ-shock (specifically, the isosurface p/p∞ = 1.4 and p/p∞ = 2.0) and stream-
wise cut of the velocity (at y + = 20) are shown in Fig. 4.33a. The interaction of the
178 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
shock wave with the streaks in the undisturbed boundary layer causes intensive rip-
pling of the separation shock and undulation of the unsteady separation line close
to the foot, as also observed in the experiments [94]. The separation-shock rippling
penetrates along the front of the external oblique shock wave. The modulus of the
vorticity (Fig. 4.33c) indicates streaks in the boundary layer close to the wall (at
y + = 15) and their behavior during the interaction with the shock wave is reminis-
cent of the vortex-breakdown phenomenon [101]. Downstream of the interaction,
the size of the eddies has increased but small-scale vortices penetrate the wall in
the reversed flow between lines R and S. A quasisteady separation zone appears
with an intermittency region upstream of the mean separation as typical interaction
4.2 Two-Dimensional Turbulent Interactions 179
Laser sheet at
y/δ = 0.7
y δ Separation
region
x
Laser sheet at
y/δ = 0.2 Intermittent
region
(b)
Figure 4.34. Schematic of the compression-ramp interaction (a) α = 20◦ , M∞ = 2.0, Reδo ≈
4.86·105 (regime IV), and planar laser scattering image (b) at y/δ = 0.2. The ramp corner is
at x/δ = 0 (Ganapathisubramani et al. [94]).
Figure 4.35. Computed (a, b) and experimental (c, d) schlieren images at two different
instants of time of interaction in CR at α = 25◦ , M∞ = 2.9, Reδo = 6.3 × 104 , Tw /Twa = 1
(regime IV). (The computed and experimental instants are not related.) 1 – incoming bound-
ary layer, 2 – forward (separation) shock, 3 – reverse-flow region, 4 – detached shear layer, 5 –
traveling compression waves (shocklets), 6 – unsteady rearward shock waves in reattachment
region (Loginov et al. [105]).
shock system. The forward-separated shock 2 (Fig. 4.35a), reverse flow region 3,
detached shear layer 4, moving shock waves (shocklets) 5 above the detached shear
layer, and rear compression shock waves 6 in the reattachment region are evident.
Along with high-frequency fluctuations of the shock system, a large-scale shock
motion was confirmed by the simulation, and the streamwise length of the separa-
tion shock excursion was estimated as 1.3δ o . Compression waves are formed behind
the forward shock around the coherent vortex structures and propagate downstream
together with them at velocities ranging from 0.1 to 0.4U∞ . The higher-speed waves
form shocklets, thereby accounting for the high level of turbulence in the exter-
nal flow between the separation shock and the detached shear layer. The rearward
shocks in the reattachment region are highly unsteady, as in the experiments, and
become invisible at irregular time intervals.
The CR flow was realized in experiments [66, 84, 97, 107] using a CDR test con-
figuration (Fig. 4.36b) and LES of the flow over the entire test configuration (Fig.
4.36a) was performed by Loginov et al. [106]. In these conditions, the boundary-
layer turbulence interacts downstream of reattachment with the expansion fan 7,
and compression waves arise immediately downstream of the expansion due to over-
expansion of the near-wall portion of the boundary layer around the decompression
ramp (see Figs. 4.11a and 4.36b, arrow 8). (The flowfield features indicated by sym-
bols 1–7 correspond to those described in Fig. 4.35.) The computed and experimen-
tal mean surface-pressure and mean-skin-friction coefficient are displayed, respec-
tively, in Fig. 4.36c,d. As shown, the LES accurately predicts the mean surface
pressure and skin friction.
In accordance with surface oil-flow visualization, specific three-dimensional
effects reminiscent of periodic Görtler-like vortices were observed in experiments
(Fig. 4.37a). Such periodic vortices usually are associated with a boundary-layer
instability due to the influence of centrifugal forces that try to remove the gas from
the external part of the boundary layer to a concave wall. The signs of these vor-
tex structures are seen especially well in the vicinity of the reattachment region,
where distinct saddle (Sd) and nodal (N) points are located periodically along diver-
gence line R; periodic longitudinal convergence and divergence lines develop down-
stream from these points (Fig. 4.37b). These longitudinal convergence and diver-
gence lines also tend to spread with the reversed flow to separation line S; however,
by then, they become rather blurred. Undulation of upstream separation line S also
is observed, indicating the probable existence of periodic saddle and nodal points
along the line as on reattachment line R. The periodic spanwise disturbances can
be amplified significantly by initiating periodic vortex perturbations upstream of the
CR using thin foils in the form of a zigzag band [109], which stimulates a more dis-
tinct flow topology corresponding to the plots in Fig. 4.37b. The three-dimensional
flowfield structure, as well as the periodic spanwise wall-skin-friction and heat-
transfer variation caused by artificially stimulated intensive periodic vortexes, was
predicted in the framework of RANS with the Spalart-Allmaras turbulence model
[126].
LES results [103, 104, 105, 106] exhibit Görtler-like vortices formed naturally
at the SBLI and demonstrate agreement with the experiments [66, 84]. The pre-
dicted mean-wall-skin-friction coefficient panorama (Fig. 4.38a) corresponds to an
experimental oil-flow image (see Fig. 4.37a). In accordance with the computations,
182 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 4.36. LES of interaction in CDR at α = 25◦ , M∞ = 2.9, Reδo = 6.36 × 104 , Tw /Twa =
1 (regime IV): a – LES schlieren-type visualization, b – schlieren photograph [66], c – wall-
pressure distribution, d – wall skin-friction coefficient distribution. Computations: solid line –
LES averaged in time and over spanwise direction; dotted lines – LES averaged in time, min
and max values over the spanwise direction (Loginov et al. [106]).
undulating solid convergence and divergence lines S and R correspond to the mean-
flow separation and reattachment at cf = 0. The darker color corresponds to the
lower cf value. Two pairs of flow convergence and divergence lines are observed
downstream of reattachment on the CR surface and their period is about 2δ o in
the computations, as in the experiments. This period corresponds to the distance
between the streaks observed in the undisturbed boundary layer [94] and between
the maxima and minima appearing on the predicted undulating separation line S.
This indicates that the interaction of the streaks with successive separation and reat-
tachment shock waves and the baroclinic-vorticity generation can be considered the
predominant mechanism that stimulates the appearance of Görtler-like vortices.
4.2 Two-Dimensional Turbulent Interactions 183
N Sd
R
S
N Sd
(a) (b)
Figure 4.37. Oil-flow visualization pattern (a) and its interpretation (b) on CR surface for the
configuration shown in Fig. 4.36 (Zheltovodov et al. [66]).
(a) (b)
Figure 4.38. Predicted by LES the mean wall skin-friction coefficient panorama on CR sur-
face (a) and Cf distributions in different CDR spanwise sections (b): averaged in time – solid
lines, averaged in time and over the spanwise direction, z – dotted lines (Loginov et al. [103,
105, 106]).
respective values at initial station E1 in the undisturbed flow. Also shown are pre-
liminary results [106] from processing additional statistical data for the flow fluctua-
tions downstream of decompression ramp A. Peaks I and II (Fig. 4.39a) correspond
to unsteady separation and reattachment shock system (see Figs. 4.35a and 4.36a),
and additional peak III is caused by an unsteady shock emerging downstream of
the expansion fan due to the flow overexpansion around the decompression ramp.
Traveling shocklets above the shear layer downstream of the separation line sup-
port the higher turbulence level, which persists for a long distance after the step top
A. The predicted distribution of surface-pressure fluctuations (Fig. 4.39b) correlates
with the turbulence behaviour and demonstrates three similar peaks (i.e., I, II, and
III), as well as the higher level in the separation zone ahead of the CR and on the
upper surface of the step downstream of top A. The wall-pressure-fluctuation dis-
tribution on the CR (up to point A) is in good agreement with experimental data at
the higher Reynolds numbers (see [2, 103, 104, 105]).
4.2 Two-Dimensional Turbulent Interactions 185
Figure 4.40. Variation of separation length (a) beneath an incident oblique shock (Green
[113]) and separation shock-wave detachment distance (b) (Petrov et al. [115]) at different
interaction regimes.
The similarity between the IOSTBLI and the CR interactions [3, 113]
also is confirmed by good agreement of experimental surface-pressure distri-
butions and their numerical (i.e., RANS) predictions [117] at M∞ = 2.96,
Reδ ≈ 1.5 · 105 (Rex = (1.0 – 1.2) × 107 ) (Fig. 4.41c). The incident shock was induced
by a 12.7-degree shock-wave generator (Fig. 4.41a) so that the inviscid pressure rise
is identical to the 25-degree CR (Fig. 4.41b). RANS calculations were performed
with the zero-equation turbulence model of Cebeci and Smith with a relaxation
model for the turbulent-eddy viscosity that was used successfully to predict CR
interactions [58] (see Fig. 4.14). The predicted density-contour plots (Fig. 4.41a,b)
Figure 4.41. Predicted density contour graph of OSWTBLI generated by a wedge with α w
= 12.27◦ (a) and compression ramp α = 25◦ flow (b) (regime IV); comparison of surface-
pressure distribution (c) (Shang et al. [117]).
188 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
800
60000
160000
Exp Exp Exp
50000 Pr1 Variable 140000 Pr1 Variable
600 Pr1 Variable
Pr1= 0.09 Pr1= 0.09 120000
Pr1= 0.09
40000
qw (W/m2)
400 100000
Pw (Pa)
τw (Pa)
30000
80000
200
20000 60000
40000
10000
0
20000
0
250 300 350 400 300 350 450
x (mm) 250 300 350 450
x (mm) x (mm)
are in general agreement with the holographic interferogram images for shock gen-
erators varying from α = 7.93 to 12.17 degrees. However, a parametric study of the
relaxation-length scale λ [117] revealed that this parameter is not universal, and the
value of λ = 20δ o is superior to the value λ = 10δ o previously used for CR flows
to achieve better agreement with experiments in the flowfield structure, surface-
pressure distribution, and locations of separation and attachment points.
Figure 4.42 is a comparison of RANS calculations [118] with experimental data
[119, 120] at α = 14 degrees, M∞ = 2.9 (i.e., regime IV). A fresh approach was
used to calculate the turbulent Prandtl number Prt as part of the solution to ensure
the incorporation of relevant physics at higher Mach numbers into the model equa-
tions. The approach is based on a two-equation model for the enthalpy variance
and its dissipation rate [118]. As indicated in Fig. 4.42a,b, the prediction of surface-
pressure and wall-shear stress is almost identical for both constant and variable Prt .
Both calculations underpredict the pressure in the separation zone, but the extent
of separation is well predicted. However, the wall-shear stress is overpredicted in
the relaxing flow downstream of reattachment. As shown in Fig. 4.42c, although the
variable Prt results demonstrate improvement in the prediction of the wall-heat flux
over constant Prt in the relaxing-flow downstream of reattachment including peak
heating, the discrepancy with experiments remains significant. Moreover, the vari-
able Prt results underpredict wall-heat flux in the separation zone as compared to a
constant Prt .
These significant errors in RANS predictions of wall pressure and heat flux
with different modern turbulence models are illustrated by many examples [59]
for a wide range of Mach numbers and oblique shock-wave strengths; there-
fore, an understanding of unsteadiness effects is important. Unique measurements
of fluctuating heat fluxes in IOSTBLIs were carried out [121] at M∞ = 3.85
( ± 0.04), Tw /Taw ≈ 0.6, Reδo ≈ 1.76 × 105 . Two representative oblique shock angles
α sw = 18.5 and 22.3 degrees corresponding to flow-deflection angles α ≈ 4.15 and
8.35 degrees, respectively, were selected to analyze unseparated and separated
flow cases. In accordance with the interaction regimes shown in Fig. 4.19 for CR
interactions at adiabatic-wall conditions, incipient intermittent separation (i.e.,
regime II) is expected at the double CR angle value 2α = 8.3 degrees and the stage
4.2 Two-Dimensional Turbulent Interactions 189
Figure 4.43. Wall pressure (a), mean heat-transfer distribution (b), rms heat-transfer fluctua-
tions (c), and signals of the heat transfer in different sections for IOSTBLI (d), at α ≈ 4.15◦ ,
M∞ = 3.85, δ o ≈ 7 mm ( ± 0.04) (unseparated flow regime) (Hayashi et al. [121]).
near the separation point (Fig. 4.44c) due to the large oscillatory motion of the sep-
aration shock that is similar to that of the pressure fluctuations. A second small
peak in the separation zone is associated with the incident shock wave, and a third
peak corresponds to the reattachment point. These additional three peaks also are
observed in the wall-pressure fluctuation measurements [121]. The heat-flux fluctu-
ations are amplified through the boundary-layer separation and reattachment, and
the values of σ qw /qw in the relaxing flow downstream of reattachment point R are
higher than those upstream for (x – ximp ) > 10δ o (Fig. 4.44c). The heat flux fluctu-
ates intermittently and more frequently up to x – ximp ≈ − 25.8 mm (Fig. 4.44d),
and no intermittency is observed downstream of this section. The wall pressure also
demonstrates comparable intermittency in the same region with nearly the same
frequency at the same locations [121]. This analogy of the behavior of wall-heat-flux
and pressure fluctuations also is observed in the flow over the CDR configuration.
In particular, the higher level of the predicted surface-pressure fluctuations not only
in the separation zone ahead of the CR but also on a surface downstream of top
A of the decompression ramp (see Fig. 4.39b) corresponds to the observed higher
4.2 Two-Dimensional Turbulent Interactions 191
level of heat transfer (see Fig. 4.26). Thus, the unsteady nature of the complex flows
is a serious limitation for their adequate modeling on the basis of RANS turbulence
models.
DNS of an impinging oblique shock-wave interaction [122] at α sw = 33.2
degrees, α ≈ 8 degrees, M∞ = 2.25, Reθo = 3,725 (Reδo = 5.15 × 104 ), Tw /Taw = 1
demonstrates the principal unsteady effects and explains fundamental physical
mechanisms. The flow corresponds to the boundary between the regimes of small-
scale separation (II) and large-scale separation (III) (see Fig. 4.19, 2α ≈ 16 degrees).
In accordance with the computed instantaneous density field shown in a vertical
plane (Fig. 4.45a) and pressure and velocity fields (not shown), the existence of a
complex organized motion in the outer part of the boundary layer is revealed. It
is characterized by the occurrence of turbulent three-dimensional bulging coher-
ent structures inclined at an acute angle with respect to the wall and also observed
in experiments. This figure confirms the enhancement of mixing, which is typical
for STBLI. The velocity isocontours reveal a small separation zone of reverse flow
confined between the upstream influence point x0 and oblique shock-impingement
point xI , which exhibit a highly intermittent nature. The computed mean-wall-
pressure and average-skin-friction coefficient show rather good agreement with the
experiments of Deleuze [123] and Laurent [124] (Fig. 4.45b,с). The faster relax-
ation of the calculated skin friction to an equilibrium state was explained by the
higher-momentum thickness in the experiments Reθo = 4,808, which also caused a
20 percent difference in the skin friction upstream of interaction. The existence of
elongated, intermittent, streaky vortical structures upstream of the interaction and
the appearance of larger, coherent structures in the mixing layer above the separa-
tion zone (Fig. 4.46a) are qualitatively similar to CR interactions, as discussed previ-
ously. The flow in the separation zone is highly three-dimensional and characterized
by scattered spots of flow reversal in the interaction zone (Fig. 4.46b). However,
the existence of the Görtler-like vortices was not observed in the flow, although
they have been observed distinctly in Brazhko’s [125] experiments by optical (i.e.,
PLS) visualization and surface-heat-transfer measurements at M∞ = 5 and 6, α = 15
degrees (i.e., regime IV). The average separation and reattachment points exhibited
an rms displacement (normalized by δ o ) on the order of 5 and 16 percent, respec-
tively. As concluded on the basis of analysis of the calculated mean-velocity profiles,
the mean-flow relaxation distance after the interaction is close to 10δ o ; however, in
accordance with turbulence statistics, complete equilibrium is attained in the inner
part of the boundary layer, whereas in the outer region, the relaxation process is
incomplete. This conclusion is in agreement with the behavior of wall-heat-flux fluc-
tuations shown in Fig. 4.44c.
Figure 4.47 illustrates several important dynamic features of IOSTBLI flow as
shown by the flowfield structure at a time interval of 0.22δ o /u∞ . The complex orga-
nized motion in the outer part of the boundary layer is characterized by the occur-
rence of large coherent vortical structures that are shed close to the average separa-
tion point, resulting in a mixing layer that grows in the streamwise direction and is
considered a major cause of large-scale unsteadiness. The turbulence-amplification
mechanism is associated primarily with the formation of the mixing layer. The figure
reveals the intense flapping motion of the reflected shock (and, to a lesser extent,
the incident shock) and its branching. The foot of the incident shock experiences an
192 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 4.45. Instantaneous density isocountour lines (a) and comparison of the DNS with
experimental measurements by Deleuze [123] and Laurent [124] for average wall pressure
(b) and skin-friction coefficient (c) (Pirozzoli and Grasso [122]).
oscillatory motion that is related to the passage of the coherent structures through
the tip. As concluded on the basis of a detailed analysis of calculated wall-pressure
fluctuations of power-spectral density, the large-scale, low-frequency unsteadiness is
sustained by an acoustic-resonance mechanism established in the interaction region
[122]. The power-spectral density at various locations exhibits a nearly constant
4.3 Summary 193
Figure 4.46. Projection of vortex structures (a) and distribution of the instantaneous “stream-
wise” skin-friction coefficient (b) on flat plate surface (Pirozzoli and Grasso [122]).
4.3 Summary
Our knowledge and understanding of two-dimensional STBLIs is virtually com-
plete. Extensive experimental investigations revealed accurate and useful correla-
tions for determining the conditions for fully attached, incipiently separated, and
fully separated flows in STBLIs for canonical geometries including normal shock,
CR, compression-expansion ramp, and oblique-shock interactions. The correlations
are based on fundamental physical knowledge of gas dynamics and turbulence struc-
ture, and they provide useful data for engineering design. The inherent similarities
in two-dimensional SBLIs also are evident and provide a better understanding of
flowfield physics. It is now entirely feasible to predict not only peak aerothermo-
dynamic loadings associated with two-dimensional STBLIs but also details of the
194 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 4.47. Sequence of x-y flowfield visualization at time intervals of 0.22δ o /u∞ . Solid lines
depict isocontour levels of the pressure-gradient modulus, 20 ≤ |∇p| ≤ 400; gray patches cor-
respond to vortex structures. The arrow indicates the instantaneous direction of the incident
shock foot motion, and the circle tracks the evolution of specific vortex (Pirozzoli and Grasso
[122]).
requirement for resolution of the viscous sublayer, wherein the dynamics of tur-
bulence production are of paramount importance. Further research is needed in
the development of approximate models of the viscous-sublayer dynamics to enable
LES at higher Reynolds numbers. DES models combining RANS approaches in
the near-wall region (to avoid the Reynolds-number limitation of LES) with LES in
the outer region represent an area of current research interest. Further progress in
this methodology can be expected, principally in the assimilation of accurate RANS
models in the near-wall region to LES in the outer portion of the flow. DNS pro-
vides accurate predictions of SBLIs without introduction of modeling parameters;
however, it is limited to low Reynolds numbers for the same reason as LES. Addi-
tional research is needed in the development of faster algorithms for DNS to raise
(or, it is hoped, remove) the Reynolds-number barrier.
REFERENCES
[18] J. Marvin. Turbulence modeling for computational aerodynamics. AIAA J., 21 (1983),
7, 941–55.
[19] D. D. Knight. “Numerical simulation of compressible turbulent flows using the
Reynolds-averaged Navier-Stokes equations: Turbulence in compressible fluids.
AGARD Report 819, pp. 5–1 to 5–52 (1997).
[20] H. S. Zhang, R. M. C. So, T. B. Gatski, and C. G. Speziale. “A Near-Wall 2nd-Order
Closure for Compressible Turbulent Flows.” In Near-Wall Turbulent Flows (1993), pp.
209–218.
[21] M. Gnedin and D. Knight. A Reynolds stress turbulence model for compressible flows.
Part I: Flat-plate boundary layer. AIAA Paper 95–0869 (1995).
[22] G. Zha and D. Knight. Three-dimensional shock boundary layer interaction using
Reynolds stress equation turbulence model. AIAA J., 34 (1996), 7, 1313–20.
[23] G. Erlebacher, M. Hussaini, C. Speziale, and T. Zang. Toward the large eddy simula-
tion of compressible turbulent flows. J. Flui. Mech., 238 (1992), 1550–85.
[24] M. Lesieur and P. Comte. Large eddy simulations of compressible turbulent flows: Tur-
bulence in compressible fluids. AGARD Report 819 (1997), 4-1–4-39.
[25] J. A. Domaradzki, T. Dubois, and A. Honein. A subgrid-scale estimation model
applied to large eddy simulations of compressible turbulence. In CTR Proceedings,
1998 Summer Program, Center for Turbulence Research, Stanford University and
NASA Ames Research Center, Stanford, CA (1998).
[26] F. Grinstein, L. Margolin, and W. Rider (eds.) Implicit Large Eddy Simulation (New
York: Cambridge University Press, 2007).
[27] X. Xiao, J. R. Edwards, H. A. Hassan, and R. A. Baurle. Inflow boundary condi-
tions for hybrid large eddy–Reynolds averaged Navier-Stokes simulation. AIAA J., 41
(2003), 8, 1418–89.
[28] J. R. Edwards, J.-L. Choi, and J. A. Boles. Large-eddy/Reynolds-averaged Navier-
corner interaction. AIAA J., 46 (2008), 4, 977–91.
[29] F. R. Menter. Two-equation eddy-viscosity turbulence model for engineering applica-
tions. AIAA J., 32 (1994), No. 8, 1598–605.
[30] A. Zheltovodov, R. Dvorak, and P. Safarik. Shock waves/turbulent boundary layer
interaction properties at transonic and supersonic speeds conditions. Izvestiya SO AN
SSSR, Seriya Tekhnicheskih Nauk, 6, (1990), 31–42 (in Russian).
[31] H. H. Pearcey. Some effects of shock-induced separation of turbulent boundary layers
in transonic flow past airfoils. ACR R&M, No. 3108 (1959).
[32] L. G. Grodzovskyi. Experimental research of shock waves/boundary layer interaction
at the Mach number range M = 1.0 – 1.8. Izvestiya AN SSR, Energetika i Avtomatika,
No. 4 (1961), 20–31 (in Russian).
[33] E. E. Zukoski. Turbulent boundary-layer separation in front of a forward-facing step.
AIAA J., 5 (1967), 10, 1746–53.
[34] M. Sajben, M. J. Morris, T. J. Bogar, and J. C. Kroutil. Confined normal–
shock/turbulent–boundary-layer interaction followed by an adverse pressure gradient.
AIAA J., 29 (1991), 22, 2115–23.
[35] S. J. Kline, J. L. Bardina, and R. C. Strawn. Correlation of the detachment of two-
dimensional turbulent boundary layers. AIAA J., 21 (1983), 1, 68–73.
[36] D. Chapman, D. Kuehn, and H. Larson. Investigation of separated flows in supersonic
and subsonic streams with emphasis on the effect of transition. NACA Report 1356
(1958).
[37] S. S. Kutateladze and A. I. Leont’ev. Turbulent Boundary Layer of Compressible Gas
(Novosibirsk: SO AN SSSR, 1962) (in Russian).
[38] M. J. Morris, M. Sajben, and J. C. Kroutil. Experimental investigation of normal–
shock/turbulent–boundary-layer interaction with and without mass removal. AIAA J.,
30 (1992), 2, 359–66.
[39] A. B. Haines. 27th Lanchester Memorial Lecture: Scale effects in transonic flow.
Aeronauutical J., 91 (1987), 907, 291–313.
[40] G. R. Inger. Upstream influence and skin friction in non-separating shock/turbulent
boundary-layer interactions. AIAA Paper 80–1411 (1980).
References 197
[63] G. S. Settles. “An experimental study of compressible turbulent boundary layer sepa-
ration at high Reynolds numbers.” Ph.D. Dissertation (Princeton, NJ: Aerospace and
Mechanical Sciences Department, Princeton University, 1975).
[64] M. S. Holden. Shock wave-turbulent boundary layer interaction in hypersonic flow.
AIAA Paper 72–74 (1972).
[65] G. M. Elfstrom. Turbulent hypersonic flow at a wedge-compression corner. J. Flui.
Mech., 53 (1972), Pt. 1, 113–27.
[66] A. A. Zheltovodov, E. Kh. Schülein, and V. N. Yakovlev. Development of turbulent
boundary layer at the conditions of mixed interaction with shock waves and expansion
fans. Preprint No. 28–83, ITAM USSR Academy of Sciences, Novosibirsk, (1983) (in
Russian).
[67] D. W. Kuntz, V. A. Amatucci, and A. L. Addy. Turbulent boundary-layer properties
downstream of the shock-wave/boundary-layer interaction. AIAA J., 25 (1987), 5, 668–
75.
[68] F. O. Thomas, C. M. Putnam, and H. C. Chu. On the mechanism of unsteady shock
oscillation in shock wave/turbulent boundary layer interactions. Experiments in Fluids,
18 (1994), 1/2, 69–81.
[69] P. Ardonceau, D. H. Lee, T. Alziary de Roquefort, and R. Goethals. Turbulence
behavior in a shock wave/boundary layer interaction. In AGARD CP-271, 8-1–8-14
(1999).
[70] M. A. Goldfeld and V. N. Dolgov. Investigations of turbulent boundary layer on test
model of compression surfaces of supersonic diffuser. In Aerophysical Research, ITAM
SB USSR Acad. Sci., Novosibirsk, 2, 98–9 (1972) (in Russian).
[71] M. A. Goldfeld and V. N. Dolgov. Experimental research of turbulent boundary layer
on delta plate with wedge. Izv. SO AN SSSR, Ser. Tekhn. Nauk 8 (1973), 2, 16–22 (in
Russian).
[72] R. H. Korkegi. Comparison of shock-induced two- and three-dimensional incipient tur-
bulent separation. AIAA J., 13 (1975), 4, 534–5.
[73] A. Zheltovodov and E. Schülein. Peculiarities of turbulent separation development in
disturbed boundary layers. Modelirovanie v Mekhanike (Modeling in Mechanics), 2, 1,
53–8, 1988 (in Russian).
[74] C. Horstman and C. Hung. Reynolds number effects on shock-wave turbulent
boundary-layer interaction: A comparison of numerical and experimental results.
AIAA Paper 77–42 (1977).
[75] M. Visbal and D. Knight. The Baldwin–Lomax turbulence model for two-dimensional
shock-wave/boundary-layer interactions. AIAA J., 22 (1984), 7, 921–8.
[76] C. Ong and D. Knight. Hybrid MacCormack and implicit Beam-Warming algorithms
for a supersonic compression corner. AIAA J., 25 (1987), 3, 401–7.
[77] E. Schülein and A. A. Zheltovodov. Development of experimental methods for the
hypersonic flows studies in Ludwieg tube. International Conference on the Methods of
Aerophysical Research, Part 1, Russia, Novosibirsk, 191–9 (1998).
[78] J. S. Hahn. Experimental investigation of turbulent step-induced boundary-layer sepa-
ration at Mach numbers 2.5, 3 and 4. AEDC-TR-69–1 (1969), 31.
[79] H. Liepman and A. Roshko. Elements of Gasdynamics (New York: John Wiley & Sons,
1957).
[80] A. A. Zheltovodov. Peculiarities of development and modeling possibilities of super-
sonic turbulent separated flows. Separated Flows and Jets IUTAM Symposium Novosi-
birsk, USSR, eds. V. V. Kozlov and A. V. Dovgal (Berlin Heilderberg: Springer-Verlag
Berlin, 1991), 225–36.
[81] A. A. Zheltovodov, E. G. Zaulichnyi, and V. M. Trofimov. Development of models for
calculations of heat transfer at supersonic turbulent separated flows conditions. Zhur-
nal Prikladnoi Mechaniki i Tekhnicheskoi Fiziki (J. of Applied Mechanics and Techni-
cal Physics), No. 4, 96–104 (1990) (in Russian).
[82] M. S. Holden. Shock wave-turbulent boundary layer interaction in hypersonic flow.
AIAA Paper 77–45 (1977).
References 199
[83] A. Zheltovodov, E. Zaulichniy, V. Trofimov, and V. Yakovlev. The study of heat trans-
fer and turbulence in compressible separated flows. Preprint No. 22–87, ITAM, USSR
Academy of Sciences, Siberian Branch, Novosibirsk (1987), 48 (in Russian).
[84] A. A. Zheltovodov and V. N. Yakovlev. Stages of development, gas dynamic structure
and turbulence characteristics of turbulent compressible separated flows in the vicinity
of 2-D obstacles. Preprint No. 27–86, ITAM, USSR Academy of Sciences, Novosibirsk,
55 (1986) (in Russian).
[85] A. A. Zheltovodov, V. A. Lebiga, and V. N. Yakovlev. Measurement of turbulence
parameters in compressible boundary layers in the vicinity of separation zones. Zhur-
nal Prikladnoi Mechaniki i Tekhnicheskoi Fiziki (J. of Applied Mechanics and Techni-
cal Physics), No. 3, 108–13 (1989) (in Russian).
[86] A. Zheltovodov, A. Borisov, D. Knight, C. Horstman, and G. Settles. The possibilities
of numerical simulation of shock waves/boundary layer interaction in supersonic and
hypersonic flows. International Conference on the Methods of Aerophysical Research,
Part 1, Novosibirsk, Russia, 164–70 (1992).
[87] V. Trofimov and S. Shtrekalkin. Longitudinal vortices and heat transfer in reattached
shear layers. Separated Flows and Jets, IUTAM Symposium, Novosibirsk, Russia, July
9–13, 1990, eds. V. V. Kozlov and A. V. Dovgal (Berlin Heilderberg: Springer-Verlag,
1991), 417–20.
[88] I. Bedarev, A. Zheltovodov, and N. Fedorova. Supersonic turbulent separated flows
numerical model verification. International Conference on the Methods of Aerophysical
Research – Part 1, Novosibirsk, Russia, 30–5 (1998).
[89] A. V. Borisov, A. A. Zheltovodov, A. I. Maksimov, N. N. Fedorova, and S. I. Shpak.
Verification of turbulence models and computational methods of supersonic sepa-
rated flows. International Conference on the Methods of Aerophysical Research, Part
1, Novosibirsk, Russia, pp. 54–61 (1996).
[90] B. M. Efimtsov and V. B. Kuznetsov. Spectrums of surface pressure pulsations at the
supersonic flow over forward-facing step. Uchenie Zapiski TSAGI (Scientific Notes of
TSAGI), 20 (1989), 3, 111–15 (in Russian).
[91] V. N. Bibko, B. M. Efimtsov, and V. B. Kuznetsov. Spectrums of surface pressure pul-
sations ahead of inside corners. Uchenie Zapiski TSAGI (Scientific Notes of TSAGI),
20 (1989), 4, 112–17 (in Russian).
[92] V. N. Bibko, B. M. Efimtsov, V. G. Korkach, and V. B. Kuznetsov. About oscillations
of shock wave induced by boundary layer separation. Mekhanika Zhidkosti i Gaza
(Fluid Dynamics), No. 4, 168–70 (1990) (in Russian).
[93] K. J. Plotkin. Shock wave oscillation driven by turbulent boundary layer fluctuations.
AIAA J., 13 (1975), 8, 1036–40.
[94] B. Ganapathisubramani, N. T. Clemens, and D. S. Dolling. Effects of upstream bound-
ary layer on the unsteadiness of shock-induced separation. J. Flui. Mech., 585 (2007),
369–94.
[95] G. Urbin, D. Knight, and A. Zheltovodov. Compressible large eddy simulation using
unstructured grid: supersonic boundary layer in compression corner. AIAA Paper 99–
0427 (1999).
[96] W. El-Askary. Large eddy simulation of subsonic and supersonic wall-bounded flows.
Abhundlungen aus dem Aerodynamishen Institut, der Rhein.-Westf. Technischen
Hoschule Aachen (Proceedings), Heft 34. Institute of Aerodynamics Aachen Univer-
sity, ed. Prof. Dr. W. Schröder, Aachen, 12–27 (2003).
[97] A. V. Borisov, A. A. Zheltovodov, A. I. Maksimov, N. N. Fedorova, and S. I. Shpak.
Verification of turbulence models and computational methods of supersonic separated
flows. International Conference on the Methods of Aerophysical Research, Part 1, Rus-
sia, Novosibirsk, 54–61 (1996).
[98] N. A. Adams. Direct simulation of the turbulent boundary layer along a compression
ramp at M = 3 and Reo = 1,685. J. Flui. Mech., 420 (2000), 47–83.
[99] G. Urbin, D. Knight, and A. Zheltovodov. Large eddy simulation of a supersonic com-
pression corner: Part I. AIAA Paper 2000–0398 (2000).
200 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
[100] H. Yan, D. Knight, and A. Zheltovodov. Large eddy simulation of supersonic compres-
sion corner using ENO scheme. Third AFOSR International Conference on DNS and
LES, August 5–9, 2001. Arlington: University of Texas, 381–8 (2001).
[101] A. A. Zheltovodov, E. A. Pimonov, and D. D. Knight. Numerical modeling of vor-
tex/shock wave interaction and its transformation by localized energy deposition.
Shock Wave, No. 17, 273–90 (2007).
[102] E. Schülein, A. A. Zheltovodov, E. A. Pimonov, and M. S. Loginov. Study of the bow
shock interaction with laser-pulse-heated air bubbles. AIAA Paper 2009–3568 (2009).
[103] M. Loginov, N. Adams, and A. Zheltovodov. Large-eddy simulation of turbulent
boundary layer interaction with successive shock and expansion waves. International
Conference on the Methods of Aerophysical Research, Part I. Publishing House Non-
parel, Novosibirsk, 149–57 (2004).
[104] M. Loginov, N. Adams, and A. Zheltovodov. LES of shock wave/turbulent boundary
layer interaction. Proc. High Performance Computing in Science and Engineering’04,
eds. E. Krause, W. Jäger, and M. Resch (Berlin Heilderberg: Springer-Verlag, 2005),
177–88.
[105] M. Loginov, N. Adams, and A. Zheltovodov. Large-eddy simulation of shock-
wave/turbulent-boundary-layer interaction. J. Flui. Mech, 565 (2006), 135–69.
[106] M. Loginov, N. A. Adams, and A. A. Zheltovodov. Shock-wave system analysis for
compression-decompression ramp flow. Fifth International Symposium on Turbulence
and Shear Flow Phenomenon. Eds. R. Friedrich, N. A. Adams, J. K. Eaton, J. A. C.
Humprey, N. Kasagi, and M. A. Leschziner. TU München, Garching, Germany, 27–29
August 2007, I, 87–92.
[107] A. Zheltovodov, V. Trofimov, E. Schülein, and V. Yakovlev. An experimental doc-
umentation of supersonic turbulent flows in the vicinity of forward- and backward-
facing ramps. Report No. 2030, Institute of Theoretical and Applied Mechanics, USSR
Academy of Sciences, Novosibirsk (1990).
[108] S. Stolz and N. A. Adams. An approximate deconvolution procedure for large-eddy
simulation. Phys. Fluids, 11 (1999), 1699–1701.
[109] H. Lüdeke, R. Radespiel, and E. Schülein. Simulation of streamwise vortices at the
flaps of reentry vechicles. AIAA Paper 2004–0915 (2004).
[110] R. G. Adelgren, H. Yan, G. S. Elliott, D. D. Knight, T. J. Beutner, and A. A. Zhel-
tovodov. Control of Edney IV interaction by pulsed laser energy deposition. AIAA J.,
43 (2005), 2, 256–69.
[111] D. Rizzetta and M. Visbal. Large eddy simulation of supersonic compression ramp
flows. AIAA Paper 2001–2858 (2001).
[112] M. Ringuette, M. Wu, and M. P. Martin. Low Reynolds number effects in a Mach 3
shock/turbulent-boundary-layer interaction. AIAA J., 46 (2008), 7, 1883–6.
[113] J. Green. Interactions between shock waves and turbulent boundary layers. Prog.
Aerospace Sci., 11, 235–340 (1970).
[114] L. Landau and E. Lifshitz. Fluid Mechanics (Oxford: Pergammon Press, 1959).
[115] G. I. Petrov, V. Ya. Likhushin, I. P. Nekrasov, and L. I. Sorkin. Influence of viscosity
on a supersonic flow with shock waves. Trudi CIAM (Proceedings of CIAM), No. 224,
28 (1952) (in Russian).
[116] S. Bogdonoff and C. Kepler. Separation of a supersonic turbulent boundary layer. J.
Aero. Sci., 22 (1955), 414–30.
[117] J. S. Shang, W. L. Hankey Jr., and C. H. Law. Numerical simulation of shock wave–
turbulent boundary-layer interaction. AIAA J., 14 (1976), 10, 1451–7.
[118] X. Xiao, H. A. Hassan, J. R. Edwards, and R. L. Gaffney Jr. Role of turbulent Prandtl
numbers on heat flux at hypersonic Mach numbers. AIAA J., 45 (2007), 4, 806–13.
[119] E. Schülein. Optical skin-friction measurements in short-duration facilities. AIAA
Paper 2004–2115 (2004).
[120] E. Schülein, P. Krogmann, and E. Stanewsky. Documentation of two-dimensional
impinging shock/turbulent boundary layer interaction flows. DLR, German Aerospace
Research Center, Paper 1, B 223–96 A 49 (1996).
References 201
[121] M. Hayashi, S. Aso, and A. Tan. Fluctuation of heat transfer in shock wave/turbulent
boundary-layer interaction. AIAA J., 27 (1989), 4, 399–404.
[122] S. Pirozzoli and F. Grasso. Direct numerical simulation of impinging shock wave/
turbulent boundary layer interaction at M = 2.25. Phy. Flui., 18 (2006), 6, Art. 065113,
1–17.
[123] J. Deleuze. Structure d’une couche limite turbulente soumise á une onde de choc inci-
dente. Ph.D. Thesis, Université Aix-Marseille II (1995).
[124] H. Laurent. Turbulence d’une interaction onde de choc/couche limite sur une paroi plane
adiabatique ou chaufée. Ph.D. Thesis, Université Aix-Marseille II (1996).
[125] V. N. Brazhko. Periodic flowfield and heat transfer structure in the region of attach-
ment of supersonic flows. Uchenie Zapiski TSAGI (Scientific Notes of TSAGI), X
(1979), 2, 113–18 (in Russian).
[126] P. R. Spalart and S. R. Allmaras. A One-Equation Turbulence Model for Aerodynamic
Flows. AIAA Paper 92-0439 (1992).
[127] G. S. Glushko. Turbulent Boundary Layer on a Flat Plate in an Incompressible Fluid.
Bulletin of Academic Sciences USSR, Mechanical Series, No. 4 (1965), 13–23.
5 Ideal-Gas Shock Wave–Turbulent
Boundary-Layer Interactions in Supersonic
Flows and Their Modeling:
Three-Dimensional Interactions
Alexander A. Zheltovodov and Doyle D. Knight
5.1 Introduction
This chapter continues the description of supersonic turbulent shock wave–
boundary layer interactions (STBLIs) by examining the flowfield structure of three-
dimensional interactions. The capability of modern computational methods to
predict the observed details of these flowfields is discussed for several canonical
configurations, and the relationships between them and two-dimensional interac-
tions (see Chapter 4) are explored.
202
5.3 Three-Dimensional Nature of Separated Flows 203
α χ α α
δ δ
a. Sharp unswept fin (SUF) b. Sharp swept fin (SSF) c. Semicone (SC)
Y α1 α2
δ δ α
δ X
χ
Z
d. Swept compression ramp (SCR) e. Blunt fin (BF) f. Double Sharp fin (DSF)
Figure 5.1. Swept shock wave generators.
5.3.1 Introduction
In accordance with the fundamental work of Prandtl [2], a two-dimensional steady
flow separates from a no-slip boundary at the point where the wall shear vanishes
and exhibits a negative gradient. The concept of two-dimensional separation was
useful in the early stages of development of theoretical approaches for various
engineering applications. However, in reality, the two-dimensional STBLI exam-
ples considered in Chapter 4 frequently demonstrate an obvious three-dimensional
structure that is especially pronounced at conditions of incipient and developed
separation. For example, the two-dimensional compression corner (see Fig. 4.37)
displays Görtler-like vortices and a set of singular (i.e., critical) node–saddle-point
combinations located along the convergence (i.e., separation) and divergence (i.e.,
attachment) lines. Consequently, the two-dimensional separation concept cannot
be strictly applied to achieve an adequate understanding of the organization (i.e.,
topology) of complex separated flows and to obtain a correct interpretation of their
physics.
The Critical Point Theory (CPT) – developed by Legendre [3, 4, 5] to analyze
wall shear lines in a steady flow in topological terms based on Poincare’s theory of
two-dimensional vector fields – is the appropriate tool for interpreting experimental
and computational results and for understanding the topology of three-dimensional
separated flows (see, e.g., Délery [6, 7]). The trajectories of the wall shear vec-
tor field almost coincide with the limiting streamlines (or surface streamlines)
that can be observed in experiments using the oil-soot-film visualization technique
(i.e., oil-flow visualization). CPT can be used to interpret the surface streamlines
204 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.2. Classifications critical points in the [p, q] plane (Délery [6, 7]).
divergence lines on the surface, which present the boundaries of secondary flows in
the boundary layer or separation zones, are determined by the location of such sin-
gular points that significantly affect the three-dimensional flowfield structure. The
separation and attachment surfaces are attached to these lines, which emanate from
the saddle points and end at a node or focus [6]. The following relationship exists
between the number of nodes/foci and saddle points on a closed-surface obstacle
[6, 8]
associated with the places where the boundary layer can separate or attach. Figure
5.5a illustrates the boundaries of the interaction regimes for STBLI in the vicinity
of a vertical unswept fin depending on the deflection-angle α and the inviscid static-
pressure ratio in the swept shock (ξ = p2 /p1 ), as defined by changes in the limiting
streamline patterns [10, 11, 12, 13] (Fig. 5.5b). The experimental data characterize
the stages of formation of the primary convergence (S1 ) and divergence (R1 ) lines
Figure 5.5. The boundaries of STBLI regimes (a) and corresponding surface-flow patterns (b)
for SUF. 1 – M∞ = 2 – 4, Reθo = (1.1 – 2.3) · 104 , Tw /Taw ≈ 1 (Zheltovodov [11], Zheltovodov
et al. [12]); 2 – M∞ = 5, Reθo = (6.2 – 7.1) · 103 , Tw /Taw ≈ 0.8 (Schülein and Zheltovodov
[13]).
5.3 Three-Dimensional Nature of Separated Flows 207
as well as the secondary lines (S2 , R2 ) that are associated, respectively, with primary
and secondary separation and attachment.
We identified six specific regimes [11, 12] depending on the deflection-angle
α with increasing shock-wave strength. In weak unseparated interaction regime I
(Fig. 5.5b), the limiting streamlines veer underneath the inviscid shock wave SW
without forming a convergence line. At an increased angle α (i.e., regime II), a
zone with approximately parallel limiting streamlines moves outward of the inviscid
shock-wave trace. At an intermediate stage (i.e., I–II) corresponding to the bound-
ary between regimes I and II (Fig. 5.5a), the limiting streamlines on either side of the
SW trace are almost parallel to it. Stanbrook [14] arbitrarily considered this stage to
be incipient swept separation of the flow and Korkegi [15, 16], following the argu-
ments of McCabe [17], proposed a practical criterion for incipient swept separation
of the following form:
M∞ · αi = 0.3 (5.2)
for M∞ > 1.6 and γ = cp /cv = 1.4, where the fin deflection angle at incipient-
separation α i is measured in radians. The boundary between regimes I and II is
shown in Fig. 5.5a in accordance with this relationship.
As shown in Fig. 5.5b (stages IIIa and IIIb), a further increase in α stimulates the
appearance of a convergence zone with asymptotic-convergence line S1 upstream of
the shock-wave trace as well as primary divergence line R1 near the fin–plate junc-
tion. Nevertheless, asymptotic-convergence line S1 exists on a plate even at stage I
and lies well inside the zone bounded by the fin and the inviscid shock [18, 19]. This
was demonstrated by placing surface tracer material only on the flat plate upstream
of the fin prior to the experiment. Such an asymptotic-convergence line spreads from
a saddle point that always exists on a plate close to the sharp-fin leading edge. In this
connection, it was concluded that the flow, in fact, is always separated; Stanbrook’s
regime can be considered a stage of the emergence of separation but not a strict cri-
terion of incipient separation. At the same time, an explicit separation line S1 begins
to form with increasing shock strength and intensification of the convergence of the
limiting streamlines (see stage IIIb). Another important feature of regime III is the
appearance of secondary convergence line S2 . With increasing α, this line extends
farther toward the fin leading edge, as seen in stages IIIa and IIIb. Regimes IV and
V are characterized, respectively, by a gradual suppression of line S2 and eventual
disappearance (apart from a limited region in the vicinity of the fin leading edge).
In regime VI, secondary convergence line (S2 ) reappears accompanied by a diver-
gence line (R2 ) considerably closer to the fin (approximately in the vicinity of the
inviscid shock-wave trace) than in previous regimes. The following empirical corre-
lation [20]:
M∞ · αi2 = 0.6, (5.3)
where α i2 is the specific fin deflection angle measured in radians, can be used to
determine the appearance of secondary convergence line S2 – hence, the boundary
between regimes II and III (Fig. 5.5a).
In accordance with Lighthill’s [9] arguments, the convergence of the limiting
streamlines in the vicinity of separation line S1 (Fig. 5.5b; see stages IIIa, IIIb, and
IV) is an important factor resulting in the abrupt departure from the surface (see
208 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.6. Flow in the vicinity of a separation line in accordance with Lighthill [9] (from
Délery [6]).
also Délery [6] and Tobak and Peake [8]). Consider an infinitely small streamtube
with a rectangular cross section located close to the wall between the two adjacent
limiting streamlines l1 and l2 and two skin-friction lines f1 and f2 (Fig. 5.6). From the
conservation of mass, the height of streamtube h = C (μ/nρτ w )1/2 . Consequently,
the height can increase greatly not only in the vicinity of a singular point where
τ w becomes very small but also in the vicinity of the convergence (i.e., separation)
line where n significantly decreases. The latter describes the emergence of three-
dimensional separation far from the singular point where the limiting streamlines
start to leave the surface with a significant vertical-inclination angle. The symmet-
rical attachment process can be defined, mutatis mutandis, in identical terms, with
the difference that the skin-friction lines in such a situation diverge from the separa-
tor, which becomes an attachment (or reattachment) line [5, 6]. The velocity along
the attachment line tends to increase with increasing distance from the saddle point
where the attachment line originates. The distance n between diverging skin-friction
lines increases, and distance h of the outer-flow streamlines from the surface can
decrease in the attachment region.
An understanding of the factors that determine the type of flow symmetry in the
vicinity of swept shock waves is important to gain insight into the three-dimensional
flowfield structure and to develop scaling laws. Many investigators concluded that
the swept-shock interaction is quasiconical, except for an initial region in the imme-
diate vicinity of the juncture of the fin leading edge and the flat plate (see, e.g., [1,
18, 21, 22, 23, 24, 25, 26, 27, 28, 29]). Consequently, the spherical polar-coordinate
system is a proper frame for analysis of these flows. The flowfield features (Fig.
5.7a) and surface streamlines (Fig. 5.7b), outside of an inception region of length Li ,
can be considered in two dimensions by projection from the vertex of the Virtual
Conical Origin (VCO) onto the surface of the sphere. A commonly used simplifi-
cation is the replacement of such a spherical segment with a plane tangent to the
sphere and normal to the inviscid shock wave. In this case, the pertinent specific
Mach number describing the interaction depends on the shock-angle β 0 and the
component of M∞ in this plane according to Mn = M∞ sin β 0, rather than M∞ . The
5.3 Three-Dimensional Nature of Separated Flows 209
Figure 5.7. Projection of quasiconical interaction flowfield onto spherical polar coordinate
surface (a) and sharp-fin interaction footprint (b) (Settles [18]).
flow in the cross section normal to a swept shock wave is similar to the equivalent
two-dimensional normal shock wave–turbulent boundary layer interaction (NST-
BLI) case. Fig. 5.8 demonstrates (a) a typical surface oil flow pattern, (b) a photo-
graph of flow visualization by Planar Laser Scattering (PLS) in the vertical section
normal to a swept shock wave, (c) a sketch of the flowfield in this cross section,
with (d) a qualitative surface pressure distribution. Flow conditions correspond to
the boundary between regimes I and II (i.e., Stanbrook’s flow regime) at M∞ = 3,
α = 5 degrees (see Fig. 5.5a). As shown in Fig. 5.8b–d, there is gradual compres-
sion between the upstream-influence line U (corresponding to beginning deflection
(a) y (b)
SW
P
L
S
PLS section
(c)
Mn
U
α
Mn z
p/p∞
z
(d)
ξ
M∞ 1
0 U z
Figure 5.8. SUF interaction at M∞ = 3, α = 5 degrees, Reθ = 2 · 104 (the boundary between
regimes I and II) (Zheltovodov [21], Zheltovodov et al. [23]).
210 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.9. SUF interaction at M∞ = 3, α = 19 deg., Reθ = 2 · 104 (regime IV) (Zheltovodov
et al. [23]).
of limiting streamlines in the interaction region) and surface trace of the deformed
shock wave; the pressure “plateau” region forms in this region where the limiting
streamlines are almost parallel. An additional pressure rise is observed after the
shock wave in the direction of the fin. The flowfield structure (Fig. 5.8b,c) is quali-
tatively similar to two-dimensional NSTBLI (see Fig. 4.3a,b). At greater deflection-
angles α, a λ-shock foot is observed (Fig. 5.9b,c) as in the separated flow regime of
a two-dimensional NSTBLI (see Figs. 4.2a and 4.3c) and a distinct surface pressure
plateau is formed downstream of explicit coalescence line S1 (Fig. 5.9a,c,d). How-
ever, in contrast to the two-dimensional interaction, the three-dimensional flow is
capable of developing transverse-velocity components, allowing fluid to escape lat-
erally, and the streamlines exhibit a helix structure (Fig. 5.9c). Thus, the flow shown
in a cut (i.e., PLS) plane characterizes the behavior of “pseudo-streamlines” appear-
ing as a result of projection of the velocity-vector field in this plane. In these condi-
tions, the streamline emanating from separation line S1 is distinct from the stream-
line penetrating reattachment line R1 , and the separation zone is opened for the
external flow penetrating it.
Within the framework of the two-dimensional analogy, the plateau pressure
data pp /p∞ = f(Mn ) for a fin-generated three-dimensional STBLI (Fig. 5.10a, sym-
bols 5–8) are in good agreement with empirical correlations for two-dimensional
interactions [24, 30] (i.e., lines 2 and 3). The Free Interaction Theory (FIT)
[31] expression (4.14) in Chapter 4 (i.e., solid line 1) agrees well with experi-
mental data (i.e., symbols 5–7) for an adiabatic wall and the Reynolds numbers
5.3 Three-Dimensional Nature of Separated Flows 211
Reθ = (1.4 − 2.3) × 104 with k = 7.4 recommended for these conditions [10, 32].
The value of cf was determined by Kutateladze and Leont’ev’s [33] method. Similar
to two-dimensional normal-shock-wave interactions (see, e.g., Fig. 4.4), the plateau
pressures tend to lie below the FIT predictions for small normal Mach numbers,
although here, the spread is somewhat larger. The scatter in the collected two- and
three-dimensional data [24, 25] is within the limits of the experimental results shown
by band 4 (Fig. 5.10a). This correspondence of two- and three-dimensional data
follows from the validity of the FIC within the framework of the two-dimensional
analogy for different three-dimensional STBLI flows with cylindrical and conical
symmetry [34, 35]. In accordance with other recommendations [28, 29], the normal
component of the Mach number MnS1 = M∞ sin β S1 to separation line S1 (see Fig.
5.7b) also can be used to generalize the plateau pressure data of three-dimensional
interactions. However, the Mach-number component Mn normal to the swept-shock
trace, rather than MnS1 , provides the best quantitative agreement with existing two-
dimensional correlations [34]. This is shown clearly in Fig. 5.10b, where MnS1 values
are used to generalize the data.
In the framework of the two-dimensional analogy by assuming that M∞ = Mn
and ξ i is equal to the plateau pressure ratio for separated two-dimensional normal
212 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
shock-wave interactions, equation (4.15) can be used to predict the critical Mach
number Mni and corresponding pressure rise ξ i required to produce fin-generated
incipient swept separation at Reθ = (1.4 − 2.3) × 104 , Tw /Taw = 1. The predicted
values Mni = 1.24 – 1.26 and ξ i = 1.62 – 1.69 (Fig. 5.10a) correspond to Stanbrook’s
interaction regime on the boundary between the regimes I and II (see Fig. 5.5b,
case I–II and Fig. 5.8a) and agree well with conditions of emerging “true” incip-
ient separation for two-dimensional NSTBLIs (see Fig. 4.4). Fig. 5.11a shows the
predicted values ξ i (Reθ ) using k = 7.4 (following recommendations [10, 32]) and
k = 5.94 (as suggested by Charwat [36]), demonstrating the tendency to decrease
as Reθ increases from 103 to 1.5 × 104 . Experimental data gathered by a number of
researchers [12, 24, 26, 27, 28, 29, 37, 38, 39] (i.e., symbols 1–8) are in agreement with
5.3 Three-Dimensional Nature of Separated Flows 213
β 0S (see Fig. 5.10a). A distinct λ-shock foot is formed around the separation zone
at α ≥ α S (see Fig. 5.9b,c). Analogous to the regimes of two-dimensional NSTBLIs
(see Fig. 4.4), the respective regimes of unseparated flow (I), intermittent separa-
tion (II), and developed large-scale separation (III) can be identified for the three-
dimensional swept-shock-wave interaction (see Fig. 5.10a). In accordance with the
indicated critical Mach numbers Mn = MS , the third regime in the three-dimensional
case is realized at the higher pressure ratio in shock wave (ξ S ) with an increase in
the Mach number M∞ . The shock-pressure-ratio dependency ξ S versus Mach num-
ber M∞ for conditions of fully developed, large-scale swept separation emerging at
α = α S or Mn = MS (see Fig. 5.11b, band 3) almost coincides with Mach numbers
M∞ ≈ 2.2 – 3.6, with the correlation ξ S * (M∞ ) (i.e., band 6) proposed by Korkegi
[15] for the onset of the large-scale separation stage in two-dimensional compres-
sion ramp (CR) flows corresponding to α = α S * (see band 14 in Fig. 4.19). However,
within the framework of a two-dimensional analogy, the swept-shock-pressure ratio
ξ S variation must be considered versus the normal Mach number Mn (see Fig. 5.11b,
band 4) in the same way as the data for the plateau pressure (see Fig. 5.10a). With
decreasing Mn , the dependence 4 logically tends to the value ξ S (i.e., symbol 7) for
two-dimensional NSTBLIs in accordance with values of MS being considered for
this case (see Fig. 4.6) at Reδ ≥ 105 (or Reθ ≥ 2 × 103 ). Comparison of the gener-
alized data for the critical shock-wave strengths ξ i and ξ S at the stages of incipient
true and large-scale separation in conditions of two- and three-dimensional SWTB-
LIs demonstrate their close correspondence at the supersonic Mach numbers for the
respective interaction regimes. This is opposite to the conclusion by Korkegi [15] –
namely, that three-dimensional swept-shock-wave–turbulent boundary-layer inter-
action always leads first to separation and possible flow breakdown in rectangular
diffusers and inlets and the critical shock-pressure ratios are well below the incipient
values for two-dimensional cases. This conclusion, however, cannot be recognized
as totally reliable because it was based on the comparison of empirical dependen-
cies 1 (ξ i = 1.5) and 6 (see Fig. 5.11b) that characterize different three- and two-
dimensional interaction regimes.
5.3 Three-Dimensional Nature of Separated Flows 215
Figure 5.14. Transformation of the flowfield structure with increase of fin-inclination angle at
M∞ = 3.95: a, b – α = 16◦ (regime IV), с – α = 20◦ (the boundary between regimes IV and
V), d – α = 25◦ (regime VI) (Settles [18] and Alvi and Settles [44, 45]).
yet formed (see Fig. 5.5b) and there is only little liftoff (z/δ 0 ∼ 0.2) of the stream-
lines. Similar flow behavior was observed in experiments [37] in conditions of the
same interaction regime II at M∞ = 2.95, α = 10 degrees. In accordance with the
experiments, large horizontal and small vertical inclinations of the streamlines are
typical for the formation of secondary flow in the near-wall portion of the boundary
layer. The third case, shown in Fig. 5.13c (M∞ = 4, α = 16 degrees), corresponds to
the band of critical values α = α S (shown in the limits of regime IV in Fig. 5.5a) at
which the large-scale separation appears, and the streamlines converge and lift off
the surface to reach a significant height of z/δ 0 = 0.55.
Figure 5.14 depicts the transformation of the flowfield structure for an interac-
tion case studied experimentally by Alvi and Settles [18, 44, 45]. The flowfield fea-
tures in different interaction regimes were obtained by optical visualization using
conical shadowgram and interferogram as well as PLS methods. In conditions of
regime IV at M∞ = 3.95, α = 16 degrees (Fig. 5.14a) corresponding to RANS cal-
culations shown in Fig. 5.13c, the separated free-shear-layer is deflected away from
the wall by the separation shock and then returns toward it near the rear leg of the
λ-shock. The separated free-shear-layer rolls into a tight vortex with a distinct core.
Aft of the separation bubble, the flow impinges on the surface, forming attachment
line R1 , where some flow is directed back upstream. This reverse flow then encoun-
ters secondary-coalescence line S2 . The impinging jet structure is bounded by the
slip line originating from the triple point of the λ-shock (i.e., the outer boundary)
and the separation bubble (i.e., the inner boundary). The turning of this jet is accom-
plished via a Prandtl-Meyer expansion fan that reflects from the slip line as a com-
pression fan. The compression waves occasionally coalesce and form “shocklets,”
which are drawn as solid lines. The structure of this jet impingement [18, 44, 45] is
5.3 Three-Dimensional Nature of Separated Flows 217
Figure 5.15. Surface-flow visualization (a, b) and flowfield structure (c) at α = 30.5◦ , M∞ = 4
(regime VI); Zheltovodov [11] and Zheltovodov et al. [12]).
reminiscent of the Edney type-IV leading-edge shock impingement [46] and thus
explains the reason for the peak pressure, skin friction (see Fig. 5.14b), and heating
in the rear part of the fin interaction. The surface skin friction is finite (i.e., nonzero)
in the vicinity of primary- and secondary-coalescence lines S1 and S2 in the section
that is distant from the singular point located ahead of the fin leading edge.
As the interaction strength increases further (Fig. 5.14c,d), the shocklets in the
impinging jet become more intensive; eventually, the final shocklet becomes a “nor-
mal shock” (i.e., to the flow in the conical projection) that terminates supersonic
flow in the jet prior to impingement. Secondary-coalescence lines S1 and S2 are
located at a distance upstream from the inviscid shock-wave trace during regimes
III and IV up to regime V (see, e.g., Fig. 5.14a,c). The conical shadowgrams reveal
an obvious “bulge” in the reversed flow in the vicinity of line S2 , but the authors
concluded that it is not a clear indication of secondary separation. For even greater
interaction strengths (starting from regime V), the “reverse” flow inside the separa-
tion zone accelerates to supersonic speeds (see also [24, 28]), which can lead to the
formation of a shocklet located approximately underneath the inviscid shock wave
(Fig. 5.14d).
The surface-flow-pattern visualization with the corresponding qualitative flow-
field structure and surface pressure distribution at strong interaction conditions (i.e.,
regime VI, α = 30.5 degrees, M∞ = 4) [11, 12] are shown in Fig. 5.15. Here, the
flow is characterized by the appearance not only of primary-coalescence and diver-
gence lines S1 and R1 but also of distinct secondary lines (i.e., S2 and R2 ), which are
218 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
14
all sections
12 Y
x = 124 mm Long fin
10 plate
X
β
8 Z
Pw /P 1
Flow
6
0
0.4 0.6 0.8 1.0 1.2 1.4
(a)
0.025
0.020
S1
UI
Skin friction coefficient
0.015
R1 S2R2
0.010
0.005
Fin
0.000
R1 S2R2 S1 UI
Shock
–0.005
0.4 0.6 0.8 1.0 1.2 1.4
z/x
(b)
Figure 5.17. Surface pressure (Schülein and Zheltovodov [26]) (a) and skin-friction distribu-
tions (Schülein [47]) (b) at x/δ 0 = 32.6, α = 23◦ , M∞ = 5, Reδ 0 = 1.4 × 105 (regime VI).
Fig. 5.15c) – may be caused by similar effects. However, these secondary lines are
of limited extent because they are shifted quickly by the “reversed” flow to primary
line S1 in contrast to similar lines S2 and R2 emerging downstream close to the invis-
cid shock-wave trace.
RANS simulations are capable of resolving important details of the flow struc-
tures described herein. Fig. 5.19 shows the flowfield predicted by calculations [50]
using the Baldwin-Lomax turbulence model for Mach M∞ = 4, α = 20 degrees at
220 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.18. 3-D separation in the vicinity of vertical cylinder (Voitenko et al. [48, 49]).
Figure 5.19. Predicted by RANS calculations flowfield model for α = 20◦ , M∞ = 4 (regime
V) (Knight et al. [50]).
5.3 Three-Dimensional Nature of Separated Flows 221
like B and C may take more than 100 upstream boundary-layer thicknesses in the
spanwise direction to complete a single revolution of the vortical structure. Stream-
line A, on the surface, deflects at the upstream-influence line and flows along the
surface, parallel to the line of convergence. Streamline B, slightly above streamline
A, passes over the convergence line, turns to the approximate inviscid shock-wave
direction, and flows along Surface No. 2. Streamline C, just below Surface No. 1,
turns and rises initially but then descends, continuing to turn slowly. Streamline D,
just above Surface No. 2, follows a similar path but, as it descends and approaches
the floor, it continues downstream as the surface streamline for the downstream
flow. Bogdonoff [51] concluded that the primary feature of this type of flowfield is
a flattened vortical flow. Nevertheless, if the rotational velocities or static pressures
are analyzed, a fully developed vortex cannot be detected and the structure is pri-
marily a supersonic flow along the shock direction with a slow rotational component.
Figure 5.20. Comparison of RANS computations with surface pressure data (a, b) and
primary-separation-line angle variation vs. fin-inclination angle (c) (Zheltovodov [10]).
shows the closest agreement with experiments for the weaker interaction at α <
20 degrees (see, e.g., [10, 52, 53]). As the interaction strength increases, differences
between the computations and experiments become more severe with a comparable
level of accuracy for different turbulence models (Fig. 5.20a,b). None of the com-
putations predicted secondary separation in the regimes in which it was observed
experimentally. A similar tendency is seen when comparing primary-separation-
line-angle β S1 versus fin-angle α (Fig. 5.20c). The integrated-to-the-wall (ITW)
results are in better agreement than the wall-function (WF) results. For α < 20
degrees, these computations are definitely within the uncertainties of the measure-
ments; however, for α > 20 degrees, the computations and data diverge.
Figure 5.21 shows a RANS computation [54] for conditions close to the bound-
ary between regimes III and IV (see Fig. 5.5a), using a modified Baldwin-Lomax
turbulence model. Several cross-flow cuts and skin-friction lines are presented in
Fig. 5.21b in accordance with their positions marked in Fig. 5.21a. It can be seen
that this calculation can capture the secondary separation. It also can be observed
that the inner layers of the undisturbed boundary layer, where the eddy viscosity
is high, wind around the core of the vortex. However, the outer layers (which have
low turbulence) move above the vortex and penetrate the separation bubble at the
reattachment region, forming a low-turbulence “tongue” close to the wall under-
neath the vortex (Fig. 5.21b, sections VI and VIII). Because the intermittency of the
flow inside this tongue and the outer layers of the vortex is small, the flow is almost
laminar in this region. At the initial stage of development, the conical separation
vortex is completely composed of turbulent fluid (section II); however, as it grows in
downstream direction, the low-turbulence tongue forms (sections IV, VI, and VIII).
Panaras [53, 54] observed that an increase of the interaction strength causes more of
the low-turbulence outer flow to be entrained in the vortex. Conversely, in a weak
interaction, a low-turbulence tongue is not formed. Figure 5.22 demonstrates that an
5.3 Three-Dimensional Nature of Separated Flows 223
(a) (b)
Figure 5.21. Skin-friction line patterns and cross sections of the vortices at M∞ = 4, α = 16◦ ,
Reδ o = 2.04 × 105 (the boundary between regimes III and IV) (Panaras [54]).
Figure 5.23. Surface pressure (a) and surface-streamlines prediction for α = 30.6o and
M∞ = 4, Reδ o = 1.6 × 105 (regime VI); (b) weakly nonlinear k-omega model (WD + ); (c)
linear Wilcox’s k-ω model (Thivet et al. [58]).
5.3 Three-Dimensional Nature of Separated Flows 225
Figure 5.24. Turbulence kinetic energy in cross section (x = 122.5 mm): a – weakly nonlinear
k-ω model (WD + ), c – linear Wilcox’s k-ω model (Thivet et al. [58]).
Y M∞ = 3.0, α = 20°
Swept fin
(a) 4.2
x α χ = 0°
X 15°
PL 30°
S 45°
βο 60°
3.4
Z
M∞
2.8
1.8
(b)
1
0 10 20 30 β deg.
◦ ◦
Figure 5.25. PLS flow visualization around swept fin (a, b) at χ = 60 , α = 20 , M∞ = 3, Reθ =
2 · 104 and surface pressure distributions (b) at different χ values (Zheltovodov and Schülein
[61]).
attachment of separated flow occurs closer to the fin with increasing leading-edge
sweep-angle χ . However, the plateau pressure ratio pp /p∞ versus Mn (see Fig.
5.10a), the critical normal Mach number Mn = Mi , and shock-wave critical strength
ξ = ξ i corresponding to Stanbrook’s interaction regime – as well as the second crit-
ical Mach number Mn = MS and shock-wave critical value ξ = ξ S corresponding to
the emergence of developed large-scale separation – are the same for unswept- and
swept-fin interactions (Fig. 5.26).
Fig. 5.27a demonstrates the evolution of the surface-flow pattern near a semi-
cone for interaction regimes I–VI that is similar in some details to the regimes
observed for unswept and swept sharp fins. The corresponding cross-sectional flow
structures are shown in Fig. 5.27b with increasing half-angle α and shock-strength
ξ . The sketches are based on observations made from PLS and surface-flow visu-
alizations as well as surface pressure measurements [62]. The boundaries of these
regimes (see Fig. 5.26, dotted lines) correspond to extrapolated boundaries for
swept fins. The experiments with the semi-cone half-angles α = 8, 10, 15, 20, 25, and
30 degrees are indicated by open symbols. Closed symbols correspond to exper-
iments with asymmetric flows when the cone axis on the plate was turned by an
additional angle α 0 relative to the undisturbed-flow direction. The flowfields in
5.3 Three-Dimensional Nature of Separated Flows 227
Figure 5.26. The boundaries of STBLI regimes for sharp unswept and swept fins and semi-
cone at Reδ 0 = 1.45 – 2.76) × 105 , Tw /Tad = 1: open symbols – symmetric flow over semi-conе
(α 0 = 0), closed symbols – asymmetric flow (α 0 > 0) (Zheltovodov et al. [12]).
regimes I–II (Fig. 5.27) are characterized by an increasing penetration of the semi-
cone surface boundary layer onto the plate with increasing α and ξ . This stimulates
the appearance of the distinct secondary-convergence line S2 on the plate surface in
regime III. A localized vortex appears close to the junction of the cone surface and
the plate between lines S3 and R3 . Similar flow features were observed in the early
stages of interactions in the vicinity of an unswept sharp fin [27].
The flow in regime IV is characterized by the disappearance of secondary-
convergence line S2 at a downstream distance from the semi-cone tip and by the
appearance of a distinct primary-convergence line S1 that is associated with a stage
when the fluid starts actively to leave the plate surface. The compression-wave
system above the plate has changed into a λ-shock configuration. The onset of
developed boundary-layer separation on the plate appears once the shock strength
exceeds a second critical value ξ = ξ S (indicated by black bands in Fig. 5.26a–c).
As shown in this figure, ξ S is constant within experimental uncertainty throughout
the range of χ for Mach numbers M∞ = 2.27, 3, and 4. At higher Mach numbers,
however, larger ξ S values are necessary to cause fully developed three-dimensional
separation.
As shown in Fig. 5.27b (i.e., regimes IV–VI), divergence line R1 appears on the
semi-cone surface for ξ > ξ S , which is associated with flow attachment. A surface-
pressure maximum appears on this line as in the reattachment point below the crest
of a two-dimensional forward-facing step (see Fig. 4.13a,b) and on the reattach-
ment line below the top of a swept, forward-facing step [20, 62]. The appearance
of secondary separation and reattachment (S2 , R2 ) underneath the reversed flow is
characteristic for regime VI. Regime VII, observed at M∞ = 3 and 4 (see Fig. 5.26),
corresponds to conditions when secondary lines S2 and R2 emerging at regime VI
and located closer to the semi-cone coexist with secondary-convergence line S2 aris-
ing in the vicinity of its top at regime III.
228 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
a b
1/3 1/3
ln /δ Reδ /Mn = f l S /δReδ (5.4)
U U
5.3 Three-Dimensional Nature of Separated Flows 229
(a) (b)
Figure 5.28. Scaling of the spanwise extent of the upstream influence line: 1 – M = 2.94, α =
12 – 24◦ , χ = 0 (Wang and Bogdonoff [65]); 2–5 – fins, 6–8 – semi-cones (Zheltovodov et al.
[12]).
or
Here, the scaling (1/δ)Reδ 1/3 is applied uniformly in two orthogonal directions
(Fig. 5.28а) (i.e., along the inviscid shock-wave trace ls and in the normal direc-
tion ln ). Mn is the Mach-number component normal to the shock and Reδ is the
Reynolds number based on the boundary-layer thickness δ at the fin leading edge.
Because sharp-fin interactions have no other characteristic length, equations (5.4)
and (5.5) apply only to the inception length. Downstream of this inception zone, the
flowfield is invariant to any scale transformation. Thus, these equations simplify to
the following:
1/3
(Li /δ)U /Mn ∝ 1/Reδ (5.6)
where β U1 , β S1i , and β R1i are the inclination angles of these lines corresponding to
incipient-intermittent separation (ξ = ξ i and, respectively, β 0 = β 0i ). They can be
predicted by the following approximate relationships:
βUi = 1.22β0i + 3.4 (5.10)
β S1i = β0i (5.11)
β R1i = 0.55β0i (5.12)
The angle of the secondary-separation line for subsonic reversed flow (i.e., regimes
III–IV) can be described by the following equation:
β S2 − βS2i = 1.6(β0 − β02i ) (5.13)
and for supersonic reversed flow (i.e., regime VI) by the following equation:
β S2 = β0 (5.14)
The value β S2i in equation (5.13) corresponds to the first appearance of the
secondary-separation line at α = α i2 and, respectively, β 0 = β 02i in accordance with
condition (5.3) on the boundary between regimes II and III (see Fig. 5.5a). These
relationships proposed on a basis of generalization of wide experimental data at
M∞ = 2 – 4 also provide good agreement with experimental data [13, 26] for sharp-
unswept fins at M∞ = 5 up to fin-inclination angles α = 27 degrees, as shown in
Fig. 5.30a. The empirical correlations also can be used with some success for flow
over semi-cones [12, 62], as shown in Fig. 5.30b–d.
(a) (b)
(c) (d)
Figure 5.30. Upstream influence (β U ), primary convergence (β S1 ), secondary convergence
(β S2 ), and primary divergence (β R1 ) lines scaling for unswept fin (a) (Schülein and Zhel-
tovodov [26]) and semi-cones (b–d) (Zheltovodov and Maksimov [62]).
conical regimes is determined by the inviscid shock detachment stage from a SCR
(Fig. 5.32a, solid line); this boundary is invariant with δ 0 and Reδ 0 . Thus, cylindrical
and conical symmetries correspond to attached and detached shocks, respectively.
Different regimes of separation with cylindrical (1) and conical (2, 3) symmetry with
reattaching flow on the ramp or the test (plate) surface can be realized in these con-
ditions depending on the values of χ and α (see sketches 1, 2, and 3, respectively, in
Fig. 5.32b). The inviscid shock shapes for different test configurations can be corre-
lated by a detachment-similarity parameter [35], as follows:
a b
Figure 5.31. Schematic of surface streak lines for asymptotic cylindrical (a) and conical (b)
symmetry of flow in the vicinity of swept compression ramp (Settles and Teng [68]).
(a) (b)
Figure 5.32. Swept compression ramp flow regimes (a) (Settles and Kimmel [35]) and incep-
tion length similarity plot (b) (Settles [69]) at M∞ = 3. Cylindrical flow symmetry: 1 – sep-
aration with reattachment on the ramp surface; conical flow symmetry: 2 – separation with
reattachment on the ramp surface; 3 – separation with reattachment on the wall surface; 4 –
inaccessible due to tunnel blockage.
5.3 Three-Dimensional Nature of Separated Flows 233
(a) (b)
Figure 5.33. Free-interaction surface pressure comparison (a) and normalized interaction
response (β U – β U )/Mn vs. detachment similarity parameter ζ (b) for quasiconical interac-
tions (Settles and Kimmel [35]).
form (Li /δ0 )Reaδo = f (ζ ), with the empirical constant α = 1/3, for quasiconical fins
and SCR interactions. A similarity plot (Fig. 5.32b) compares all three members
of the fin/swept-corner shock-interaction family in accordance with symmetry type
and position of the primary-reattachment line. Transition from a cylindrical (1) to
a conical (2) flow regime corresponds to ζ = 43.6; above a critical value ζ ≈ 100,
it was found that both fin and swept-corner generators produced the same generic
flow topology with reattachment on the wall surface.
The same surface-flow patterns for SCRs, fins, and semi-cones are produced
when the inviscid shock and upstream-influence-line angles β 0 and β U have simi-
lar values at these conditions [35]. The similarity of surface topography is a con-
sequence of close agreement in surface pressure distributions for these different
cases (Fig. 5.33a), which can be considered to illustrate the classic FIC for three-
dimensional STBLI cases. This similarity is in agreement with the conclusions of
Zheltovodov and Kharitonov [34], who demonstrated the validity of FIC for var-
ious three-dimensional STBLIs with cylindrical and conical symmetry by gener-
alizing experimental data for surface pressure distributions in the framework of
the two-dimensional analogy in the form (p – p∞ )/(pp – p∞ ) = F (x – xS1 )/(xS1 –
xR1 ) at different Mach numbers M∞ and their comparison with similar generalized
data for different two-dimensional separated flows. Based on experimental data at
M∞ = 2.95, Settles and Kimmel [35] proposed the following empirical relationships
for quasiconical interactions:
βU = 1.59β0 − 10.0 (swept and unswept fins and swept corners, ζ 100) (5.17)
The similarity among the flows produced by different shock generators is demon-
strated in Fig. 5.33b, which also reveals the functional dependency f in equation
(5.18).
234 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
(see Fig. 5.11). In the region of cylindrical symmetry at χ > 0, the calculated values
ξ S max also coincide with the experiment (1) (Fig. 5.34, band 4). The calculated val-
ues ξ inv = ξ S max at the higher Reynolds numbers Reθ > 2.3 · 104 with using equation
(4.18) with empirical correlation by Zukoski [30] for the “plateau” pressure ratio
pp /p∞ = 0.5M + 1 (line 5) correspond to the data (1) and predictions (4) on a basis
of FIT. At χ = 0, the top experimental point 1 corresponds to two-dimensional,
large-scale incipient-separation conditions at α = α S * (i.e., point 7; see Fig. 4.19),
which agrees with Korkegi’s predictions. Also shown in Fig. 5.34 is an additional
point 2 of Settles et al. [71, 72] that corresponds to the incipient, steady small-scale
separation, which is in good agreement with the predicted ξ S max value.
Figure 5.35 demonstrates a comparison of experimental (a) and RANS-
predicted surface flow patterns [73] with the algebraic (b) and the k-ε turbulence
models in a SCR vicinity in conditions of conical-interaction regime 2 (i.e., α = 24
degrees, χ = 40 degrees, M∞ = 2.9; see Fig. 5.32) when separated flow attaches to
the ramp surface. The calculations with both turbulence models are in good agree-
ment with the experiments. Flowfield surveys were conducted for two δ 0 values dif-
fering by 3:1 to check the accuracy of the following Reynolds-number similarity law
proposed by Settles and Bogdonoff [63]:
(xw /δ0 )Reaδ0 = f (zw /δ0 )Reaδ0 (α, λ, M∞ fixed) (5.19)
(a) (b)
Figure 5.36. Reynolds number scaling (Settles et al. [73]): a – of floor and plate interaction
structure, b – of measured and computed upstream influence line. 1 – δ 0 = 1.494 cm, Reδo =
9.39 × 105 , 2 – δ 0 = 0.429 cm, Reδo = 2.7 × 105 .
(a) (b)
Figure 5.37. Computed and experimental surface pressure for α = 24◦ at Reδo = 9.39 × 105 ,
Tw /Twa ≈ 1: a – χ = 40◦ , b – χ = 60◦ , M∞ = 2.95, Reδo ≈ 9 × 105 : showing Cebeci-Smith
turbulence model, and Jones-Launder model (Knight et al. [74]) respectively.
5.3 Three-Dimensional Nature of Separated Flows 237
direction. The line of divergence defines the intersection of the second surface (2)
with the SCR. The fluid contained between the wall and surface 2 is entrained in
the vortical flow. The fluid above surface 2 passes over the vortical structure and
continues up the CR. As shown, the general features are similar to those observed
for the sharp fin (see Fig. 5.19).
Figure 5.39. Oil-flow visualization (a) (Zheltovodov et al. [75]), and computed skin-friction
lines (b) −k − ε model (Zheltovodov et al. [76, 77]) and (c) −k − ω model (Thivet et al. [78,
79, 80] for 7o × 7o case at M∞ = 4, Reδ∞ = 3.1 × 105 .
Figure 5.40 compares the computed distributions of surface pressure and heat-
transfer coefficients using the Wilcox k-ω model with the experimental data along
the axis of symmetry (i.e., TML) and in three cross sections denoted, respectively,
as I, II, and III (see Fig. 5.39a). The heat-transfer measurements were carried out
at Tw /Taw = 1.04 – 1.05. Again, the computations on different grids are in good
agreement with the measurements.
In accordance with the experiment at α 1 × α 2 = 11 × 11 degrees (Fig. 5.41a),
the throat formed between lines S1 and S2 becomes narrower. The flow expanding
behind this throat and directed along the centerline is terminated above the point
of the first intersection of the inviscid shocks (denoted by dotted lines), with a sub-
sequent formation of central-longitudinal-divergence line R3 of finite length along
the flow. This divergence line gradually degenerates to a dividing streamline in the
region of transverse-flow intersection. Secondary-convergence lines S3 and S4 prop-
agate more upstream toward the fin vertices than in the previous case. The flow
5.3 Three-Dimensional Nature of Separated Flows 239
Figure 5.40. Surface pressure (a–d) and heat-transfer coefficient distributions (e–h) for
7o × 7o case and their prediction with the k-ω model at M∞ = 4, Reδ∞ = 3.1 × 105 (Thivet
et al. [78, 79, 80]).
240 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
I II III
S1 S3 R1 R3 S7
U
(a)
S8
S2 S4 R2
S3 R3
S1
(b)
S2 S4
P/P∞
6
Experiment
4 Computation
(c)
2
0
0 10 20 30 40 50 X
P/P∞
6
I II
4 III
(d)
2
0
5 0 5 5 0 5 5 0 5 Z
Figure 5.41. Oil flow visualization (a) (Zheltovodov et al. [75]) and computed with the k-ε
model (Zheltovodov et al. [76, 77)] skin-friction lines (b) and surface pressure distributions
(c, d) for 11o × 11o case at M∞ = 4, Reδ∞ = 3.1 × 105 .
distribution along the channel centerline (Fig. 5.41c) and in cross sections I, II, and
III (Fig. 5.41d) shows good agreement for X = x/δ 0 < 24. For X > 24, however, the
calculations tend to overestimate the value of the pressure level near the centerline
at 24 < X< 40. The calculated and measured pressure distributions in cross sections
I, II, and III (Fig. 5.41d) are in good agreement.
242 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.44. Comparison of experiment and computations for surface pressure and heat-
transfer coefficient along TMP for 15o × 15o case at M∞ = 4, Tw /Tadw = 1.05 (Thivet
et al. [79]).
Figure 5.45. Prediction of double-fin interaction for 15◦ × 15◦ case at M∞ = 3.98, Reδo =
2.66 × 105 (Panaras [53]).
5.3 Three-Dimensional Nature of Separated Flows 245
skin-friction coefficient distribution (d) along the centerline as compared to the stan-
dard B-L model (see Fig. 5.45c). The model provides a mixed-type separation flow
in the vicinity of every isolated fin: turbulent in the region of separation lines (S1
and S2 ; Fig. 5.45b) and almost laminar between reattachment lines (R1 and R2 )
and corresponding secondary-separation lines (S3 and S4 ). This yields significant
improvement in predicting crossing shock-wave-interaction flow between two fins.
However, despite the improvement, the unsteady nature of the crossing SBLI flow
is not captured by steady RANS computations and may be an additional reason
for the observed discrepancies in the predicted flow topology, surface pressure, and
heat-transfer level compared to the experiments. In accordance with the experimen-
tal research of Batcho et al. [82], in conditions of a weak 7 × 7–degree interaction
at M∞ = 2.95, the rms surface pressure fluctuations increase along the centerline,
reach a constant value before the theoretical shock-wave crossing point, and are
relatively constant downstream of this point in the remainder of the investigated
region. In contrast, for the 11 × 11–degree case, after the initial rise, the large rms
surface pressure fluctuation peak lies behind the theoretical shock-wave crossing
point and is followed by a decrease and then another increase. The unique measure-
ments of surface temperature fluctuations in the vicinity of the large rms pressure-
fluctuations peak support the assumption that the pressure fluctuations are con-
nected to surface temperature and heat-transfer fluctuations. The unsteady nature
of these interactions is supported by additional surface pressure-fluctuation mea-
surements performed by Poddar and Bogdonoff [83] at M∞ = 2.95 and by Davis
and Hingst [84] at M∞ = 2.95. Thus, LES and Direct Numerical Simulation (DNS)
computations of crossing shock-wave interaction are important for the analysis of
unsteadiness effects at such conditions on surface heat-transfer, skin-friction, and
pressure-distribution predictions.
Figure 5.46 is a three-dimensional perspective of the complex crossing shock-
wave structure realized in the vicinity of a 15 × 15–degree double-fin configuration
at M∞ = 3.83 in accordance with the experimental research of Garrison and Settles
[85] using PLS images. Considering the inherent symmetry of the crossing shock-
wave interaction, only half of the flowfield structure is shown. Initial cross section
I represents a flow structure typical of the single-fin interaction case. The incident-
separation (1), rear (2), and inviscid (3) shocks with a slip line (4) penetrating from
the triple point to a fin-side/plate cross line, as well as a separation vortex (5) under
the bifurcated shock system, are visible in this cross section. To understand the inter-
section of the two separate single-fin interactions, the vertical plane of symmetry is
considered an inviscid-reflection plane. For the symmetrical crossing shock-wave
interaction, shock waves that intersect this plane must reflect from it to satisfy con-
tinuity. As shown, in cross section II, the incident-separation shock-wave reflection
from the symmetry plane is an irregular (i.e., Mach) reflection. The Mach reflec-
tion results in a straight shock-wave segment, the Mach stem (7), which spans the
interaction centerline, a reflected portion of the separation shock (6), and the newly
formed triple point (10). In accordance with the perspective view and cross section
III, the entire incident λ-shock structure reflects from the center plane in an irreg-
ular manner and remains intact (although somewhat distorted), propagating away
from the center toward the fin surface. As shown in cross section III, two additional
shock-wave segments and two triple points are observed to form as a result of this
246 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
Figure 5.46. Double-fin flow structure at α 1 × α 2 = 15◦ × 15◦ , M∞ = 3.83 (Garrison and
Settles [85]).
crossing. One centerline segment (12) spans the interaction centerline between the
reflected inviscid shock waves (9) and the second “bridge” segment (13), and then
connects the reflected inviscid shock (9) to the reflected separation (6) and rear (8)
shocks. A “mushroom-shaped” separation structure (5) is observed in the centerline
vicinity in the last-shown cross section of the twin, single-fin separations that begin
the crossing shock-wave interaction.
Elements of the effect of interaction strength on the flowfield structure are illus-
trated in Fig. 5.47a–d in accordance with different interaction regimes (see Figs.
5.39b, 5.41b, and 5.42b) predicted at M∞ = 4 on the basis of the k-ε model [76, 77,
86, 87, 88]. Select streamlines shown in the symmetry plane normal to the plate may
be interpreted as intersections of stream surfaces with this plane. In the case of a
weak interaction (i.e., 7 × 7–degree, Fig. 5.47a), the fluid in the incoming thicken-
ing boundary layer moves away from the surface under the action of the opposite
pressure gradient, which occurs without any sign of separation and reverse flow.
With increasing interaction strength (11 × 11–degree, Fig. 5.47b), line S2 is strength-
ened. However, near the plane of symmetry, it turns partially parallel to a line of
5.3 Three-Dimensional Nature of Separated Flows 247
Figure 5.47. Evolution of the computed separation topology with increase in interaction
strength (Zheltovodov et al. [77]; Schmisseur et al.[87]).
symmetry and all upstream lines are squeezed into the narrow channel between line
S2 and the line of symmetry. On the downstream side of Line S2 , each line is turned
streamwise prior to reaching the symmetry plane under the influence of the reflected
separation shock (RSS). These lines then form secondary-separation line S4 . On the
symmetry plane, the pattern shows the lifting of the incoming boundary layer and
the formation of a shear layer. Beneath this forms a line of divergence/bifurcation
(LB), which can be viewed as denoting the approach of fluid from the sides toward
the symmetry plane. Fluid below the LB attaches to the plate at the line of symmetry
at R3 and forms the longitudinal centerline vortices.
Increased interaction strength produces the first pair of critical points on the
line of symmetry with zero shear stress on the surface (Fig. 5.47c). The primary line
of separation S2 terminates at node N1 on the line of symmetry. Saddle-point C1 is
formed downstream. All lines of separation and attachment have numerical desig-
nators subscripted, critical points have designators superscripted, and critical points
on the symmetry plane are designated by lowercase letters. On the symmetry plane,
critical point N1 corresponds to saddle point c1 , whereas saddle-point C1 is associ-
ated with node n1 , which acts as a source of all fluid beneath the separated boundary
248 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
layer. The LB now emanates from this node and again can be interpreted as denot-
ing the approach of fluid from the sides toward the symmetry plane. The bifurcation
corresponding to the 15 × 15–degree case is shown in Fig. 5.47d. Four critical points
arranged in a diamond-shaped pattern are observed on the plate. Downstream-
saddle-point C1 observed in the previous regime bifurcates into saddle-point C2 off
the line of symmetry; its mirror image C1 on the other side under the influence
of intensive reversed flow penetrates from node N2 to node N1 . This reversed flow
replaces node n1 on the symmetry plane, which forms an interior stagnation point as
a focus/node above the centerline. Correspondingly, on the symmetry plane, point
n1 is replaced by saddle-point c2 .
Schülein and Zheltovodov [13, 26] performed experiments with crossing shock-
wave interactions at M∞ = 5 to identify important details of the topology evolution
in the real flow. A very liquid mixture of mineral oil with lampblack and oil paint
was used for this visualization. The movement of the mixture particles along the lim-
iting streamlines was recorded using a video camera, which provided the possibility
of analyzing the dynamics of the surface flow topology development. Increased frag-
ments of the surface flow pattern and their corresponding schemes in the vicinity of
the central separated zone “apex” specify the stages of the flow-topology develop-
ment at different double-fin inclination angles (Fig. 5.48). The formation of two sym-
metric primary-separation lines S1 and S2 , two secondary-coalescence lines S3 and
S4 , and the centerline divergence line R3 was fixed at the 16 × 16–degree case (Fig.
5.48a). Centerline saddle-point C1 exists downstream of the liquid throat between
lines S1 and S2 in the region shaded in the sketch, with node N1 located downstream
(not shown), as in the 15 × 15–degree case at M∞ = 4 (see Fig. 5.42a). The appear-
ance of central–cross-separation line S0 closing the liquid throat and located down-
stream of the small central separation zone was registered in the 17 × 17–degree
case (Fig. 5.48b). This local zone is limited both upstream and downstream by two
centerline saddle-points, C01 and C02 . Two additional symmetric nodes, N01 and N2 0 ,
are located off the centerline, and the rule (i.e., equation 5.1) governing critical
points requiring equality of nodes and saddle points is satisfied under this config-
uration considering downstream node N1 . The size of the central-separation zone
increases in the 17.5 × 17.5–degree interaction case, but the general surface flow
topology does not change (Fig. 5.48c).
The next photograph and scheme for the 18 × 18–degree crossing-shock-
wave interaction case (Fig. 5.48d) displays the appearance of reversed flow from
downstream-located centerline-node N1 to saddle-point C01 in the center of cross-
separation line S0 . To show all significant features including nodal point N1 , the scale
of this figure is decreased approximately twice compared to previous figures. Cen-
terline saddle-point C1 bifurcates into a pair of symmetric saddle points C1 and C2
under the influence of such active reversed flow directed along central-convergence
line Sc . In a similar way, centerline saddle-point C2 0 bifurcates into a pair of symmet-
ric saddle points, C02 and C03 . These points are drawn into the separation zone and
located between node N02 and focus N2 as well as between node N01 and focus N3 ,
respectively. Similar large-scale topological structures are observed in experiments
with crossing SWTBLIs in conditions of flow around two parallel, conically sharp-
ened bodies of revolution located above a plate [89] (Fig. 5.49). The figure specifies
details of the surface flow pattern in the vicinity of saddle-points C02 and C03 . Some
5.3 Three-Dimensional Nature of Separated Flows 249
S1 C30 C1 N1
N10
N3
S0
N20
S2
C20 C2
250 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
notation is changed in the figure from that used in [89] in accordance with Fig. 5.48d.
Topological rule (i.e., equation 5.1) is satisfied again under this new configuration.
A comparison between the experiments of Schülein and Zheltovodov [13, 26]
and the RANS computations of Gaitonde and Schmisseur using the k-ε model (see
[76, 87, 88]), as well as with the computations of Panaras with his modification of
the B-L model [55], is shown in Fig. 5.50 for strong 18 × 18–degree CSTBLIs at
M∞ = 5. The predicted surface flow patterns (Fig. 5.50b,c) are in general good
agreement with the experimental portrait (Fig. 5.50a). However, the first center-
line nodal point N01 is predicted using the k-ε model in the apex of the separation
zone (Fig. 50b) instead of the central saddle-point C01 and two symmetric nodes N01
and N02 on cross-separation line S0 observed in the experiment (see Figs. 5.48d and
5.50a), as well as symmetric saddle-points C02 , C03 and nodes N2 , N3 emerging in the
confluence of small forward and large-scale downstream separation zones. This dif-
ference is associated with an overpredicted turbulence level by this model, which
stimulates more active compression of the central separation zone by the contrary-
directed near-wall flows that spread from primary-reattachment lines R1 and R2 to
the centerline. Computations using the modified B-L model demonstrate better pre-
diction of the flow details observed in the experiments (Fig. 5.50c). Both models
predict symmetric saddle-points C1 and C2 in agreement with the experiments (see
Fig. 5.48d). The secondary-separation lines are not observed in experiments and in
computations in the flows directed to the centerline between primary-reattachment
(R1 , R2 ) and separation (S1 , S2 ) lines in agreement with regime V (see Fig. 5.5) for
the isolated fin. Computations using the k-ε model [76, 87, 88] and the modified
B-L model of Panaras [55] predict well the surface pressure distributions in dif-
ferent cross sections. However, they overpredict the level in the centerline vicinity
(Fig. 5.51), although the computations using the modified B-L model are again in
better agreement with the experiments (Fig. 5.51b).
The comparison between the experiments and RANS computations is shown in
Fig. 5.52 for a strong 23 × 23–degree CSTBLI at M∞ = 5. The surface-oil-flow pho-
tograph (Fig. 52a) demonstrates a significant growth of the central separation zone
and increased width of the cross separation line at the apex with a middle saddle-
point C01 . The flow properties downstream of this line are qualitatively similar to
the 18 × 18–degree case (see Fig. 5.50a) and the RANS computations using the k-ε
5.3 Three-Dimensional Nature of Separated Flows 251
(a) (b)
Figure 5.51. Surface pressure centerline distribution for 18◦ × 18◦ interaction at M∞ = 5.
(a) (b)
Figure 5.53. Flow topology [76, 87] (a) and ribbon plot [87, 88] (b) prediction on a basis of
the k-ε model, for 23◦ × 23◦ CSW interaction at M∞ = 5.
primary-separation line and does not reattach. The vortex-interaction region con-
tains the fluid that attaches near the fin and sweeps spanwise to fill the void left
by separated boundary-layer fluid. The boundary-layer and vortex-interaction flows
separate on either side of the primary-separation line. Two vortex filaments (only
one is shown in the figure) leave the surface at the two foci on either side of the
central cross-separation line. Fluid originating near the fin leading edge, which is
swept spanwise to the symmetry plane before turning downstream, comprises the
central vortex region. The high-energy entrainment flow from the inviscid region
near the fin leading edge attaches downstream and sweeps toward the centerline
before separating. Separation identified with line S5 (Fig. 5.53a) is associated with a
shock-wave feature that can be traced to the system originating at the opposite fin.
The entrained flow penetrating as a wall-jet-like structure to the symmetry plane
is separated by this shock wave. Secondary-separation line S2 terminates at down-
stream foci N4 , where a vortex filament leaves the surface. A similar second vortex
filament forms on the opposite side of the interaction region.
As shown in Fig. 5.54 and similar to previous examples, surface pressure along
the TML again is overpredicted by calculations of the 23 × 23–degree interaction
case. Although some improvement can be achieved using the modified B-L tur-
bulence model, the discrepancy apparently is not accidental but rather the result
of limitations of the RANS approach. As in other types of SBLIs, the unsteadi-
ness inherent in the flow may be of considerable importance. Such unsteadiness was
observed experimentally, but it cannot be reproduced by RANS computations. For
example, computations using the k-ε model demonstrated the regular interaction of
external crossing shock waves [87, 88], whereas in experiments, a strongly unsteady
interaction regime was observed [76].
Experimental research and RANS computations of the asymmetric (α 1
=
α 1 ) double-fin interactions demonstrated more complex and manifold flowfield
structures and topology. Interested readers are directed to papers in which these
interactions are analyzed in more detail [10, 52, 57, 78, 79, 90, 91, 92, 93, 94].
5.4 Summary 253
(a) (b)
Figure 5.54. Centerline surface pressure distribution for 23◦ × 23◦ interaction at M∞ = 5.
5.4 Summary
Our knowledge and insight of three-dimensional STBLIs has progressed dramat-
ically in the past three decades; nevertheless, it is far from complete and further
research is needed to provide a comprehensive understanding. Numerous experi-
mental investigations yielded detailed descriptions of the flowfield structure for sev-
eral canonical three-dimensional configurations, including the single unswept/swept
fin, SCR corner, and double-fin configurations. Correlations for determining con-
ditions for fully attached, incipiently separated, and fully separated flows for sev-
eral simple canonical three-dimensional configurations were developed analogous
to and with knowledge gleaned from two-dimensional STBLIs. Despite the knowl-
edge gained from the canonical configurations, their usefulness in predicting criti-
cal aerothermodynamic loads (i.e., surface pressure and heat transfer) and flowfield
structure (in particular, the location and extent of separation) for realistic air-vehicle
configurations is limited by the sheer complexity of typical shock-wave interactions
in three dimensions.
The potential to predict complex flows using computational methods advanced
dramatically in recent years. Concerning three-dimensional STBLIs, the RANS
models (the standard for engineering applications) demonstrated a general capa-
bility to predict the mean aerothermodynamic loads for weak three-dimensional
interactions. Furthermore, these models defined the flowfield structure for the
canonical configurations discussed in this chapter. However, they failed to pre-
dict peak aerothermodynamic loading for strong three-dimensional STBLIs except
when configuration-specific (i.e., nonuniversal) modifications to the models are uti-
lized. To date, it has not been possible to apply the more sophisticated LES and
Detached Eddy Simulation (DES) numerical methods to even the specially chosen
canonical configurations due to high computational costs. The application of the
even more demanding DNS to these flows remains a distant hope. Progress in using
LES, DES, and (ultimately) DNS methods for the three-dimensional STBLIs will be
possible only with dramatic improvements in the efficiency of numerical algorithms
and reduced-order modeling of near-wall turbulence dynamics.
254 Ideal-Gas Shock Wave–Turbulent Boundary-Layer Interactions
REFERENCES
6.1 Introduction
Some of the most serious and challenging problems encountered by the designers of
hypersonic vehicles arise because of the severity of the heating loads and the steep-
ness of the flow gradients that are generated in shock wave–boundary layer interac-
tion (SBLI) regions. The characteristics of these flows are difficult to predict accu-
rately due in no small measure to the significant complexity caused by shear-layer
transition, which occurs at very low Reynolds numbers and can lead to enhanced
heating loads and large-scale unsteadiness. Even for completely laminar flows, vis-
cous interaction can degrade appreciably the performance of control and propulsion
systems. It is interesting that both of the two major problems encountered with the
U.S. Space Shuttle program were associated with SBLI. The first was the so-called
Shuttle Flap Anomaly that nearly resulted in disaster on the craft’s maiden flight
due to a failure in the design phases to account correctly for the influence of real-
gas effects on the shock-interaction regions over the control surfaces. During the
flight, a significantly larger flap deflection was required to stabilize the vehicle than
had been determined from ground tests in cold-flow facilities. Miraculously, it was
possible to achieve the necessary control, and disaster was narrowly averted. The
second problem was the leading-edge structural failure caused by the impact of foam
that had been fractured and released from the shuttle tank as a result of the dynamic
loads caused by a shock interaction. Figure 6.1 is an example of the shock structures
that are generated among the shuttle, the main tank, and the solid reusable boosters.
The contour plot illustrates the corresponding computer-predicted pressure distri-
bution. Aerothermal loads generated by shock waves in the region of the bipod
that supports the shuttle nose caused the foam glove to be fractured and released.
Unfortunately, the damage this caused resulted in a tragic accident.
Even on the Apollo capsule, which is the simplest of hypersonic reentry vehi-
cles, significant problems arose due to SBLIs that occurred in the base region as a
consequence of the interplay between the reaction control system jets and the near-
wake. As in previous examples, the occurrence of transition in the separated flow,
coupled in this case with the complexity of combustion in the recirculated region,
makes these flows – including shock-shock interactions as well as SBLIs – extremely
259
260 Experimental Studies of Shock Wave–Boundary-Layer Interactions
(a) (b)
Figure 6.2. (a) Type IV interaction; (b) X15 pylon.
6.1 Introduction 261
experimental studies of real-gas effects, which are discussed in this chapter and else-
where in this book.
The effect on the control forces, acoustic loads, and aero-optic phenomena asso-
ciated with jet interaction and the accompanying SBLIs are of major concern to
designers of hypersonic vehicles that are maneuvered with large thrusters. These
problems are particularly challenging at high altitudes where jet interaction pro-
duces a separated region (Fig. 6.3), which can extend to envelop the vehicle almost
to the nose tip. Infrared images of this flowfield confirm that burning gases from
the thrusters are entrained in the recirculating flows, potentially blinding the sen-
sors placed behind windows even close to the nose tip. When laminar-boundary
layer separation occurs with transition in the free-shear layer, large-scale instabil-
ities occur in the recirculating flow upstream of the transverse jet. The fluctuating
pressures generated by the unsteady interaction region can create serious acoustic
loads with major consequences on the performance of the optical sensors. Here, the
combined effects of shear-layer transition, air and combustion chemistry, and large-
scale unsteadiness render these flows virtually impossible to predict with any degree
of accuracy.
All of these shock-interaction phenomena and others confront designers of
ram/scramjet-powered boost and cruise vehicles. In scramjet-powered hypersonic
vehicles such as the X-43, the flow in the inlet and isolator sections of the engine
contain laminar, transitional, and turbulent regions of viscous interaction and shock-
shock interaction. The interaction regions generated by multiple reflected and swept
shock waves control the efficiency of the inlet and the characteristics of the flow
entering the injector/combustor section of the engine. Although the multiple shock
interactions in the engine (Fig. 6.4) promote mixing and can act as sites for igni-
tion and flame-holding, they also are responsible for pressure losses and drag penal-
ties, which can significantly reduce engine performance. Accurately predicting shock
interaction in the presence of transition, compressible nonequilibrium turbulence,
and air and combustor chemistry remains a significant task.
6.2.1 Introduction
The intense aerothermal heating loads that arise within regions of shock-wave inter-
action with turbulent-boundary layers are regarded generally as presenting the
most significant problems for hypersonic flight. However, it may well be the lam-
inar viscous–inviscid interaction and flow separations that are associated with these
flows at greater altitudes that pose the greatest fundamental limitations on the per-
formance of maneuvering and air-breathing hypersonic vehicles. The effectiveness
of intakes and flap-control systems, as well as the performance of vehicles using
jet interaction, may be compromised seriously by the occurrence of shock-induced
laminar separation. For the intakes and flaps, compression surfaces essentially can
be “faired in,” and the separated regions formed in front of the transverse jet can
change the force or moment characteristics of the vehicle or – more seriously for
some applications – result in the obscuration of optical-seeker devices. The range
in performance of scramjet engines on a single stage to orbit vehicles also may be
limited basically by the occurrence of laminar-flow separation on the sidewall and
cowls of the engine, which eventually can result in “engine unstart.” The use of
boundary-layer controls to alleviate these problems is difficult in hypersonic flow
because the major portion of the mass and momentum in the layer over a cooled
wall is contained at its outer edge.
In the past twenty years, there has been a massive increase in computational
capabilities with which to make a direct assault on using the Navier-Stokes equa-
tions to solve flows that include SBLIs with embedded recirculating regions. These
methods have had great success in describing the flows for hypersonic Mach num-
bers, and certain outstanding questions for cases with large embedded separated
regions were resolved by three-dimensional computation.
(a) (c)
(b) (d)
Figure 6.5. Wedge- and shock-induced laminar interaction regions in hypersonic flow.
M∞ = 11.7 and ReL = 2.5 × 105 . (a) attached laminar interaction region; (b) separated lam-
inar interaction region; (c) attached laminar interaction region; (d) separated laminar inter-
action region.
wedge surface downstream of the corner (Fig. 6.5a); it is only when a large separated
region is created (Fig. 6.5b) that the interaction region moves upstream. In these
flows, the strong curvature in the reattachment region results in a significant normal-
pressure gradient, which must be modeled correctly to obtain accurate predictions.
A similar situation occurs for interaction regions generated by an external shock
(Fig. 6.5c,d). Here, we observe that for a less-intense shock, most of the interaction
region occurs downstream of the shock-impingement point. Only when extensive
separation occurs does the separation shock appear ahead of the incident shock
(Fig. 6.5d). The interaction takes place in only four or five boundary-layer thick-
nesses, and there is significant flow curvature during the reattachment-compression
process.
Typical skin-friction distributions measured in wedge-induced interaction
regions for attached and separated flows are shown in Fig. 6.6a [1]. Incipient separa-
tion occurs in these flows for wedge angles of just over 10 degrees. Measurements of
the pressure rise needed to induce incipient separation, and the value of the plateau
pressure from a number of sources for both wedge- and shock-induced interaction
regions are compared in Fig. 6.6b [1]. The figure shows that the measurements cor-
relate remarkably well when plotted in terms of the following viscous-interaction
parameter:
√ μ Tr
χ̄ = M3 C Re L where C = . (6.1)
μr T
(a)
102
PRESENT STUDY
HOLDEN 1969
HOLDEN 1967
HOLDEN 1964
NEEDHAM & STOLLERY
HAKKINEN et al 1959
1 HOLDEN 1968
10 HARVEY 1968
ANDERS 1970
MILLER et al 1964
PATE 1964
CHAPMAN et al 1956
PO
Po SYMBOL)
PP - P O (OPEN
PO SYMBOL)
10–1
10–3 10–2 10–1 100 101 102
xL
(b)
Figure 6.6. (a) Skin-friction measurements to define incipient separation and correlations
of incipient separation for wedge- and shock-induced flows [1]. M∞ = 11.7 and Re/m =
5.2 × 104 (Re/ft = 1.7 × 105 ); (b) Correlation of pressure rise to induce incipient separation
and the plateau pressure in shock- and wedge-induced interactions taken from a variety of
sources [1].
Holden 1978
10.0 Holden 1967
Holden 1964
NFEDHAM
MILLER ET AL
RIETHMULLER & GINOUX
BLOY & GEORGEFF
BOEING HOTSHOT TUNNEL
M∞θincip
HARVEY
1.0
0.1
0.01 0.1 1.0 10.0 102
xL
(a)
102
SYM M∞
18.9
15.6
11.7
(PMAX/PO)0.85
OPEN SYM – WEDGE INDUCED
SOLID SYM – SHOCK INDUCED
qMAX/qO
(P /P )0.7
.
MAX O
10
.
(PMAX/PO)0.5
q Pmax (1 + 2βχL) ½
=
qO PO (1 + 2βχ )+(1 + βχ )•兰 P θχ
L L P (L − L )
1
100 101 102
PMAX/PO
(b)
Figure 6.7. Correlations of separation and peak heating in wedge- and shock-induced lami-
nar shock interaction regions: (a) correlation of incipient separation conditions for wedge-
induced interactions; (b) comparison between simple correlation and prediction method.
WAVES
COMPRESSION
M∞− WAVES
M∞+
NECK
δ
SEPARATION DIVIDING STREAMLINE REATTACHMENT
Pe
P∞−
x x x
SEP. SHOCK REATT.
(a)
12
10 β = –.025
β = –0.05
8
β = –0.10
β = –0.15
η 6
β = –0.18
4
0
0 .5 1.0 0 .5 1.0 0 .5 1.0 0 .5 1.0 0 .5 1.0
u/ue
(b)
Figure 6.8. “Free interaction” flow model used to describe shock-induced separated regions
with momentum integral techniques: (a) schematic of flowfield and pressure distribution in
shock-induced laminar separated interaction; (b) reverse-flow profiles used in integral meth-
ods to predict separated flow.
and above the reverse-flow region. Figure 6.9a is an example of the good agree-
ment between the Lees and Reeves prediction and the measurements of Chapman,
Kuehn, and Larson [8] in laminar regions of SBLI. A similar approach by Nielsen
et al. [9] that used the power-law profiles to describe the velocity distribution across
the interaction regions is compared with measurements of Lewis, Kubota, and Lees
[10] and the Lees and Reeves predictions for wedge-induced separated flow shown
in Fig. 6.9b. Again, there is reasonable agreement between predictions and mea-
surement.
The momentum-integral analysis, although adequate for supersonic flows over
adiabatic walls, begins to break down in hypersonic flows where regions of viscous–
inviscid interaction occur over fewer boundary-layer thicknesses (see Fig. 6.5).
One of the last sets of analysis to use integral techniques to describe the heat-
transfer, pressure, and skin-friction distribution in SBLI regions over cold walls was
developed by Holden, who added the energy equation [11] and then incorporated
268 Experimental Studies of Shock Wave–Boundary-Layer Interactions
2.2
Rex,o=1.08x104 L= 0.9
EXPERIMENT (CHAPMAN, KUEHN, AND LARSON)
2.0 PRESENT THEORY R
1.8
SHOCK
Pe
1.6
Po
1.4
1.2 S
1.0
0 0.5 1.0 1.5 2.0 2.5
Figure 6.9. Comparison between
X/L
experiments in supersonic flow and
(a) momentum integral predictions by Lees
and Reeves, and Nielsen, Goodwin and
3.8 Kuhn for adiabatic-wall conditions:
DATA OF LEWIS, KUBOTA, AND LEES
Rexc = 0.15 × 106
(a) Lees and Reeves prediction for
3.4
M∞ = 6.06 LEES-REEVES adiabatic wall; (b) Nielsen, Goodwin,
TW /TtO = ADIABATIC
3.0 and Kuhn predictions for adiabatic
THEORY wall.
RO/xO = 8.25 × 105/ft
2.6 MO = 5.72
pi /po TW /TtO = 1.0 R
2.2 xO = 0.091 ft
PRESENT THEORY
1.8
S
1.4
10.25°
1.0
.06 .10 .14 .18 .22 .26 .30 .34 .38 .42
x,ft
(b)
a normal-pressure gradient [12] in integral solutions for the first- and second-order
boundary-layer equations. Heat-transfer and pressure predictions made in laminar-
wedge-induced separated regions are compared with Holden’s predictions [11] in
Fig. 6.10, which shows that although the length of the interaction and the pressure
distribution are calculated with reasonable accuracy, the method overpredicts heat
transfer in the separated region.
Comparisons with measurements in higher-Mach-number flows with the the-
ory incorporating normal-pressure gradient (Fig. 6.11) [13] show good agreement
between theory and experiment for the pressure and skin-friction distributions.
However, heat transfer in the separated region is overpredicted.
q/qo
2.0
p/po
MAXIMUM FLOW
DISPLACEMENT
4.0 THICKNESS (NECK) 1.5
3.0 1.0
2.0 0.5
MAXIMUM DISPLACEMENT THICKNESS
1.0 0
–1.0 –0.5 0 0.5 1.0 1.5 –1.0 –0.5 0 0.5 1.0 1.5
X (INCHES) X (INCHES)
Figure 6.10. Comparison between pressure heat and transfer measurements in wedge-
induced separated flows and predictions by Holden ([11]). M∞ = 10 and ReL = 1.4 × 105 .
REATTACHMENT
SHOCK
REATTACHMENT
SEPARATION POINT
DIVIDING
SHOCK STREAMLINE
SEPARATION
POINT
(a) (b)
WEDGE ANGLE = 18°
6
CH/CHo Cf /Cf
CP/CPo 4
100 CF
2
0
–2
10–1
5 7 9 11 13 15 17 19 21 23 25 27 29
DISTANCE FROM LEADING EDGE ∼ INCHES
(c)
Figure 6.11. Comparison among heat transfer, pressure, and skin friction in wedge-induced
separated flows and predictions by Holden incorporating normal-pressure gradient. M∞ = 16
and ReL = χ L = 19.8.
270 Experimental Studies of Shock Wave–Boundary-Layer Interactions
6
(a) 10
(SKIN FRICTION)
(HEAT TRANSFER)
5
∝ = 18°
∝ = 18° 4
3
1.0
2
0.1 –1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
X/L X/L
100 10
(b)
(HEAT TRANSFER) (SKIN FRICTION) ∝ = 24°
∝ = 24° 8
6
10
1.0
0
–2
0.1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
X/L X/L
3 5
Pressure contours – downstream plane Experiment
log (102——)
3
Cp
CFL3D
2
101 × 10 × 25 grid
1 3 24° (3–D)
Pressure 1
18° (2–D)
contours – plane of symmetry
−1 1
15° (2–D)
−1
3 Experiment
log (103Ch )
1 3
24° (3–D)
1
18° (2–D)
−1 1
15° (2–D)
−1
−1
0 .5 1.0 1.5 2.0
X
L
(c) Surface heat transfer
Figure 6.13. Comparisons between measurements and Navier-Stokes solutions with the CFL
3D code for wedge-induced separated flows.
10−2 6 10−3 6
5 5
4 4
2 2
10−3 10−4
0 6 12 18 24 30 0 6 12 18 24 30
x, in. x, ln.
(b) θ = 6.45˚ (c) θ = 6.45˚
6
103 C
f
CFL 3D
0
USA-PG2
Experiment
−2
0 6 12 18 24 30
x, ln.
(d) θ = 6.45˚ (e) Pressure
Figure 6.14. Comparisons between measurements and Navier-Stokes solutions with the CFL
3D code for shock-induced separated flows.
272 Experimental Studies of Shock Wave–Boundary-Layer Interactions
Stanton
Cp
0.4
0.3
10−2
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 10−3
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
X/L X/L
Hollow Cylinder Flare
sharp leading edge
30 deg
15 deg
20 deg
φ 0.065 m
φ 0.045 m
φ 0.115 m
L = 0.1017 m
0.145 m 0.025 m
Figure 6.15. Comparison between Navier-Stokes predictions and pressure and heat-transfer
measurements on hollow-cylinder–flare configuration.
10−1
FAY and RIDDEL
0.030
Y (m)
0.000 10−4
0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070 0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070
X (m) X (m)
Figure 6.16. Navier-Stokes predictions for the Mach number contours and heat transfer coef-
ficient compared with measurements on a hyperboloid flare configurations [18].
conclusions were reached for these comparisons; again, differences associated with
numerical dissipation in the different schemes were considered to be important.
1.60 0.060
Pressure
1.40 Gnoffo Cp
Heat Transfer 0.050
Cp
1.20 Gnoffo St St
0.040
1.00
0.80 0.030
0.60
0.020
0.40
0.010
0.20
0.00 0.000
0 0.5 1 1.5 2 2.5
X/L
1.4 0.1
Cp Candler
1.2 Cp Candler (Slip)
Cp (Run 192) 0.08
1 St Candler
St Candler (Slip)
0.8 St (Run 192) 0.06
Cp
St
0.6 0.04
0.4
0.02
0.2
0 0
0 1 2 3 4 5 6 7 8 9 10
Distance (inches)
Figure 6.18. Comparisons between measurements with Navier-Stokes solutions incorporat-
ing surface slip for pressure coefficient and Stanton number.
Cp
St
St
0.1 0.1
1.5 1.5
0.08 0.08
1 0.06 1 0.06
0.04 0.04
0.5 0.5
0.02 0.02
0 0 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance (inches) Distance (inches)
(a) (b)
Figure 6.20. Comparison between measured and computed heat transfer and pressure in low-
density flows over the cone/cone body using (a) Navier-Stokes, and (b) DSMC codes.
and shock-shock interaction and it was more challenging to compute than the rel-
atively benign hollow-cylinder–flare flow. The new studies were conducted under
high-enthalpy conditions using nitrogen as the test gas to eliminate the more com-
plex nonequilibrium and chemistry effects present in air flows. Low-Reynolds-
number conditions were selected for all of the studies to ensure that the flow
remained fully laminar at all times, and low-density cases were added to enable
DSMC methods to be used.
Typical results from this study are shown in Fig. 6.20. For denser flows, Navier-
Stokes solutions were obtained by Candler and Gnoffo [21] and were in excel-
lent agreement with one another and the experiment. As the density is reduced,
boundary-layer slip effects begin to be significant; however, it was demonstrated
that by accounting for them, the Navier-Stokes method could still be used to good
effect. Figure 6.20a is a typical comparison between prediction and the measured
heat-transfer and pressure distributions on the double cone; again, it exhibits excel-
lent agreement between prediction and experiments. Recent improvements by Bird
[22] and others to the DSMC codes also resulted in computations that are in excel-
lent agreement with the measurements over a large range of Reynolds numbers.
An example of such a comparison is shown in Fig. 6.20b. The clear conclusion from
these and other comparative studies is that when appropriately and expertly applied,
both the Navier-Stokes and DSMC methods can describe accurately the flowfield
and the pressure and heat-transfer distributions in these complex flows, proving that
they are fully laminar and free from flow chemistry or real-gas effects. It is apparent
that within these limitations, both computational methods are capable of predict-
ing even the most complicated interaction regions likely to develop over hypersonic
vehicles.
6.3.1 Introduction
Predicting the size and distribution of flow properties through a region of shock
wave–turbulent boundary-layer interaction for supersonic and hypersonic Mach
276 Experimental Studies of Shock Wave–Boundary-Layer Interactions
numbers remains a major challenge for numerical simulation, which, thus far,
has had little success. The major problem encountered is the difficulty in finding
adequate ways to model the turbulence through the separated-interaction regions
in which reverse flow, turbulent nonequilibrium, compressibility, and shock–
turbulence interaction effects must be accurately reproduced. Intuitively, it is
expected that modeling the macroscopic and major unsteadiness of turbulent-
separated regions necessitates the use of the more complicated and time-consuming
Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) techniques.
However, so far, they also have been relatively unsuccessful in determining the posi-
tion of separation, possibly principally due to the difficulties associated with the
“wall-layer” modeling. One saving feature for vehicle designers is that turbulent-
boundary layers are difficult to separate in hypersonic flows and, when separation
occurs, the major relevant characteristics (i.e., peak heating and pressure) can be
estimated with simple methods combined with correlations from the experiments.
As yet, no definitive methods have been developed to describe the length of the
separated interaction region or the distribution of skin friction and heat transfer in
the separated region. Earlier detailed reviews of shock-induced separated-turbulent
flows and the methods to predict their characteristics by Greene [23], Stollery [24],
Délery and Marvin [25], and Knight and Degrez [26] concentrated principally in the
supersonic-flow regime for adiabatic walls. The large Reynolds numbers required to
generate fully turbulent regions of SBLI in hypersonic flows are difficult to attain;
hence, there are significantly fewer data from experimental studies available as well
as validated models to characterize the size and detailed structure of turbulent-
separated flows induced by SBLIs.
Figure 6.21. Turbulent separation over flat-plate (b) WEDGE ANGLE = 30°
wedge compression surfaces at Mach 8.
shock does not occur in the Mach 8 flow until the turning angle exceeds 17 degrees
(Fig. 6.22d). In well-separated flows, there is a streamwise fluctuation of the sepa-
ration point (determined from skin-friction measurements) over approximately two
thirds of the boundary-layer thickness (Fig. 6.23). As previously noted, in hyper-
sonic flow, it is relatively difficult to separate a turbulent-boundary layer in a com-
pression corner or with an incident shock. This occurs for turning angles below
14 degrees in supersonic flows; however, at hypersonic speeds, the flow remains
attached on compression surfaces with turning angles as high as 30 degrees or with
shock-generator angles of 15 degrees. In fact, the exact determination of incipient
separation in these flows is more difficult because of their unsteady nature as illus-
trated in Fig. 6.23, in which the streamwise position is shown to fluctuate. For this
reason, computations based on time-averaged formulations are unlikely to prove
satisfactory. Further discussion about the unsteady aspects of SBLIs can be found in
chapter 9.
It is surprising that the conditions to induce incipient separation and the salient
characteristics of wedge- and shock-induced hypersonic flows can be correlated rea-
sonably well in terms of the free-stream Mach number (M), the overall pressure rise
ratio ( pinc − p0 )/ p0 , and the skin friction coefficient (C f ) to the wall immediately
upstream of the interaction. This is illustrated in Fig. 6.24, in which the conditions
to promote incipient separation in wedge- and shock-induced interaction regions is
shown to correlate well over a wide range of conditions with a combination of these
variables.
278 Experimental Studies of Shock Wave–Boundary-Layer Interactions
(b) SHOCK GENERATOR ANGLE = 15° Figure 6.22. Turbulent separation induced by incident
shock at Mach 8.
UPSTREAM
OF
SEPARATION
DOWNSTREAM
OF
Figure 6.23. Flow separation and unsteady charac-
SEPARATION teristics of the heat transfer recorded during the run
time of a short-duration shock tunnel.
6.3 SBLI in Turbulent and Transitional Flows 279
2
10
ELFSTROM
COLEMAN
34
STOLLERY M0 = HOLDEN (1972)
9 13.0 HOLDEN 1997
30 8.6 11.3 M=8
KUEHN 7.9
Mo = ROSHKO &
(pINC– p0)/p0
26 Mo =
6.5 THOMKE
αi. DEGREES
4.0 SYM
4.92 10
1
PRESENT STUDY
22 3.93
3.0 2.95
18
ELFSTROM
2.0 STERRETT & EMERY
14 KUEHN
SPAID AND FRISHETT 1.95
ROSHKO & THOMKE (1969)
Mo = 2.9
10 SOLID SYM – SHOCK INDUCED INTERACTION
OPEN SYM – WEDGE INDUCED INTERACTION
6 0
10
104 104 104 104 −2 −1 3 0 1
Reδ 10 10 M Cfo 10 10
Figure 6.24. Correlation of incipient separation in wedge- and shock-induced turbulent flows.
Also, for well-separated flows, the plateau pressure can be correlated easily, as
shown in Fig. 6.25. Most important, it is possible to correlate the maximum heating
levels generated in the reattachment region – relative to the undisturbed upstream
value – in terms of a simple power-law relationship with the pressure rise across the
interaction region qmax /q0 = ( pmax / p0 )0.85 (Fig. 6.26).
SYM M∞
20 3.5
7.5
RESHOTKO & THOMKE
WATSON ET AL
9–9.2 EFSTROM
Y
OR
HE
ST
10
VE
9
EE
8
&R
6
SC
5
TO
1
2 4 6 8 10 12 14
M∞
280 Experimental Studies of Shock Wave–Boundary-Layer Interactions
SYM M∞
10
0.85
qMAX pMAX
qo
= po
1
100 101 102
pMAX/po
100
M∞ = 11.33
10 θ = 15°
pW/p∞ EXPERIMENT (HOLDEN)
COMPUTATIONS (HORSTMAN)
4
k– ε
k– ε (MOD A)
k– ε (MOD B)
k– ε (MOD C)
1 CEBECI–SMITH
BALDWIN–LOMAX
4
–8 –6 –4 –2 0 2 4 6 8 10
x, cm
100
M∞ = 11.33
10 θ = 15°
qW/q∞ EXP EXPERIMENT (HOLDEN)
4 COMPUTATIONS (HORSTMAN)
k– ε
k– ε (MOD A)
k– ε (MOD B)
k– ε (MOD C)
1 CEBECI–SMITH
BALDWIN–LOMAX
4
–8 –6 –4 –2 0 2 4 6 8 10
x, cm
Figure 6.27. Comparisons between Navier-Stokes predictions for pressure and heat transfer
with a range of turbulence models and measurements in a Mach 11 well-separated region of
SBLI.
various turbulence models used in the calculations produce different answers for
the separation length and the pressure rise in the separation process. Although most
schemes are capable of predicting the peak heating and pressure at the end of the
recompression process, they are unable to predict the distribution of heating during
separation and in the separated region. It is difficult to blame this mismatch solely
on the use of the time-averaged Navier-Stokes equations rather than their unsteady
counterpart. However, the gross modeling of the turbulence in a manner that is
unable to reflect the changes in its physical characteristics through these regions
is certainly in question. There is a clear need for combined numerical and experi-
mental studies to provide detailed flowfield and surface information with which to
construct accurate models for development of turbulence in these flows in both the
supersonic- and hypersonic-flow regime.
ROTATION - REF.
TOTAL TEMPERATURE, PITOT PRESSURE,
AND HEAT TRANSFER RAKE ASSEMBLY
LIMIT DIA-SCHLIEREN
TUNNEL
EXTENSION SECT.
CONE MODEL
CONE/FLARE MODEL
INSTALLED IN THE 96" TUNNEL
Figure 6.28. Large-cone–flare configuration used in studies of high Reynolds number shock
wave–turbulent boundary layer interaction [27]. M∞ 10.97 to 13.1; Re/m 1.12 × 106 to
1.39 × 106 . Flare cone angles 36 and 42 degrees.
ensure that the flow is not transitional. One wind-tunnel experiment in which the
necessary conditions were met was conducted by Holden [29] on a large cone–
cylinder-flare configuration. This was an integral part of a program to investigate
the flow on axisymmetric bodies aimed at providing boundary-layer transition data
and measurements in regions of shock wave–turbulent boundary-layer interaction
for code-validation studies in the high-Mach-number/high-Reynolds-number range.
The model, shown in Fig. 6.28a,b, was tested at a free-stream Mach number of 11.
This size was chosen so that transition occurred well upstream of the interaction
region (i.e., more than one thousand boundary-layer thicknesses ahead) to eliminate
concerns that the unsteadiness generated during the transition process would signif-
icantly influence the unsteady characteristics of the separated region. The mech-
anisms that cause this in turbulent separated flow have long been a concern of
researchers; in particular, Dolling [30] investigated such effects in detail. In many
studies of hypersonic flow, unsteadiness in the separated region was observed. How-
ever, it was not possible to rule out upstream transition as a trigger for these insta-
bilities because there were extensive regions of boundary-layer transition upstream
of the separated interaction region. In a more recent study conducted at a Reynolds
numbers of 108 and a Mach number of 4.5, it was observed that the separated region
ahead of a spherical dome was completely stable such that the propagation of opti-
cal beams through the shear layer ahead of the dome showed minimal distortion
[31]. Measurements in the separated regions over the large-cone configuration also
exhibited few large-scale instabilities, although it was observed that shock waves
penetrating the boundary layer were distorted as a result of the interaction with
boundary-layer turbulence. It appears from these observations that hypersonic-
turbulent shock–boundary-layer interactions do not inherently exhibit large-scale
unsteadiness but rather that they are susceptible to upstream disturbances. In
the hypersonic-regime transition, the establishment of an equilibrium-turbulent
boundary layer is a lengthy process and the residual disturbances are likely to
6.3 SBLI in Turbulent and Transitional Flows 283
perturb the SBLI if it is not located well downstream, thereby leading to apparent
instabilities.
In a further phase of the combined study referred to at the beginning of Section
6.3.4, a slightly blunted cone followed by a cylindrical section and then a flare was
selected for the initial ground-test studies of the HIFIRE vehicle configuration, in
which a SBLI occurs at the flare junction (Fig. 6.29). High-frequency pressure and
platinum thin-film instrumentation were used to define the characteristics of the
transitional flow over the cone and cone–flare junction. These tests illustrate the
problems encountered in properly predicting turbulent interactions.
Figure 6.30a,b shows the surface-pressure and heat-transfer measurements over
two configurations of this model compared with code results. For the results pre-
sented here, the fore-cone half angle is 6 degrees and the flare angles are 36 and
42 degrees, respectively. For the computations, the selection of the compressibil-
ity model can significantly influence the heating level, even to the conical surface
CH
CH
(a) Schlieren photograph of flow over (b) Gaitonde pre-test prediction using AVUS,
cylinder/37° flare configuration for M= 7.16 and unstructured NS-Solver
Rel/m = 0.98E+6.
Figure 6.31. Comparison between measured and predicted flow pattern over cylinder–37-
degree flare configuration: (a) Schlieren photograph of flow over cylinder–37-degree flare
configuration for M = 7.16 and Rel /m = 0.98E + 6; (b) Gaitonde pretest prediction using
AVUS, unstructured NS-Solver.
(a) (b)
(c)
Figure 6.32. Comparison between predictions of size of separated flow and measurements
on HIFIRE Flare: (a) DPLR solution using SST Model; (b) DPLR solution using Spalart-
Allmaras model; (c) Schlieren photograph of separated region over 33-degree flare.
PEAK CH /CH
O
HEATING
FIN
UPSTREAM PLATEAU
INFLUENCE PRESSURE
SHOCK FLAT PLATE/FIN MODEL
INPINGEMENT x/δO
LINE
SURFACE PRESSURE
DISTRIBUTION
Figure 6.33. Typical surface heating and surface pressure distributions through swept shock–
turbulent boundary layer interaction.
SYM SOURCE
THEORY
NEUMANN AND TOKEN 1974
(McCABE)
12 LAW 1975
STANBROOK 1961
FOR INCIPIENT SEPARATION, αi
McCABE1966
SHOCK GENERATOR ANGLE
LOWRIE 1965
10 STALKER 1957
GOLDBERG 1973
(HOLDEN) 1983
4 KORKEGI
CRITERION
0
0 2 4 6 12
M∞
Figure 6.34. Variation of shock generator angle to induce incipient separation with Mach
number [40].
6.3 SBLI in Turbulent and Transitional Flows 287
4
SYM M
3.75
NORMALIZED PRESSURE (P/P∞)
0
0 1.0 1.5 2.0 2.5 3.0
COMPONENT OF FREESTREAM MACH NUMBER
NORMAL TO SHOCK WAVE (M∞ sin θ)
6.36. Fig. 6.36 shows that the maximum pressure rise through the interaction region
can be calculated with good accuracy from inviscid-flow relationships.
Another approach to exploring flow separation in regions of three-dimensional
SBLI is to begin with a two-dimensional or axisymmetric interaction and sweep this
interaction (or, for the axisymmetric case, introduce angle of attack) to progres-
sively establish a crossflow in the interaction region. Experimental studies of this
type were conducted by Settles et al. [33], who studied the interaction over swept
and unswept flat-plate configurations over an adiabatic wind-tunnel wall in Mach 3
airflow. Considerable effort was expended in this study to determine the Reynolds-
number scaling and the length from the upstream tip of the wedge for the flow to
SYM M∞
12 2.00
2.95
3.01
10 3.75
NORMALIZED PRESSURE (P/P∞)
3.93
4.50
11.33
8
P2 /P∞
1.167 (M∞ sinθ)2.2 − 0.167
Figure 6.36. Correlation of maximum 6
pressures recorded in swept-shock
interaction regions [40].
4
0
0 1.0 1.5 2.0 2.5 3.0
COMPONENT OF FREESTREAM MACH NUMBER
NORMAL TO SHOCK WAVE (M∞ sin θ)
288 Experimental Studies of Shock Wave–Boundary-Layer Interactions
100 RUN 11
S.G. = 15°
SWEEP ANGLE = 30°
Figure 6.37. Effect of skewed shock on size and
properties of interaction region.
10
.1
–10 –8 –6 –4 –2 0 2 4
DISTANCE FROM PT. OF SHOCK IMPINGEMENT – inches
4
CURRENT
STUDIES
Figure 6.38. Variation of interaction
3
length with shock skew. SETTLES ET. AL.
XINT.
δo
2 θ = 15°
θ = 12.5°
1
0
0 10 20 30 40 50
ψ degrees
that is opposite to that observed by Settles et al. [33] in the Mach 3 flow of a wall-
mounted swept wedge. It is also of interest that as in two-dimensional flows, the
maximum heat transfer generated in the recompression regions of both swept and
skewed interaction can be correlated with the standard power-law equation shown
in Fig. 6.39.
100
SYM M∞ SHOCK
11.3 SWEPT
11.3 SKEWED
qMAX
qo
10
Figure 6.39. Correlation of peak-
heating rates in skewed- and swept-
shock interaction regions. 0.85
qMAX PMAX
=
qo Po
1 0
10 101 102
PMAX
Po
290 Experimental Studies of Shock Wave–Boundary-Layer Interactions
Figure 6.40. Replicas of the NRV and RTE nose tips recovered from flight tests and the
installation of NRV model in 96-inch shock tunnel.
140
130
120
110
100
MAXIMUM HEATING LEVEL
80
70
60
TIME AVERAGED VALUE
50
40
MINIMUM HEATING LEVEL
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
S/Rn
Figure 6.43. First Navier-stokes computation of a separated shock interacting flow over
indented nose shape. Right: Heat-transfer comparison between theory and experiment for
the Widhopf 1 configuration (1-10-10).
equations for laminar flows over the two idealized nose shapes shown in Fig. 6.43,
designated as the Widhopf 1 and 2 shapes. These pioneering numerical computa-
tions also obtained unsteady solutions to the flow over the Widhopf 2 configura-
tion, predicting flow instabilities not dissimilar to those observed earlier over spiked
bodies. In fact, their calculations also suggested that the flow over the Widhopf 1
configuration was unsteady, with heating levels varying between the maximum and
minimum values plotted in Fig. 6.43c. Also shown are measurements of heat trans-
fer obtained in experimental studies by Holden [58], which did not indicate gross
unsteadiness in the flowfield. The surface and flowfield measurements over this con-
figuration demonstrated that regions of SBLI and shock-shock interaction were gen-
erated, resulting in large heating rates in the recompression region near the shoulder
of the nose tip.
6.4.1 Introduction
The heating and pressure loads generated by regions of shock-shock boundary-layer
interaction are among the largest that can be imposed on thermal-protection sys-
tems of vehicles traveling at hypersonic speeds. Typical heating loads can exceed
more than thirty times the undisturbed stagnation-heating value and, in many cases,
the flows are transitional, unsteady, and therefore extremely difficult to predict.
As previously noted, shock-shock heating was responsible for catastrophic failures,
including the loss of a National Aeronautics and Space Administration (NASA)
scramjet engine when the supporting pylon was destroyed by shock-impingement
6.4 Characteristics of Regions of Shock-Shock Boundary-Layer Interaction 293
heating (see Fig. 6.2b) in an X-15 program [59]. Edney [60] classified a series of flow
situations in terms of the relative position of the incident shock on the bow shock
over a blunt body (see Section 2.2.3). The Type IV interaction (see Chapter 2 and
Fig. 6.9a,b) generates a jet-like flow that impinges on the surface, forming a local
stagnation region that can result in severe heating rates even if the flow remains
laminar. However, for Type III interactions, in which a free shear layer is formed,
transition in this layer can lead to even higher heating levels. Because this occurs
at relatively low Reynolds numbers and therefore at high altitudes, in most practi-
cal situations, transitional shear layers are formed for both Types III and IV inter-
actions, making the heating loads developed in these regions large and extremely
difficult to predict correctly.
16
14
12
10
Q/Q0,FR
0
–70 –60 –50 –40 –30 –20 –10 0
Angular position (degrees)
Figure 6.46. Distributions of heat transfer in fully laminar regions of shock-shock
interactions.
flows with shear-layer Reynolds numbers higher than 5,000, transition occurs within
this layer and, under these conditions, heating-enhancement factors of 15 and higher
occur. Further increasing of the Reynolds number moves the transition point in the
shear layer closer to the shock-interaction point, resulting in heating-enhancement
factors for Type IV interactions higher than 20.
Similar variations in the peak heating with shock-impingement point are seen
for transitional and fully turbulent shear layers [62]. The variation of the highest
heating rate with Reynolds number for these flows is shown in Fig. 6.47. In the
25.00
5D 3"
20.00 6C 3"
5C 3"
5A 3"
15.00 6AA
6C 3/4"
Q/QoFR
6B 3"
6B 3/4"
Figure 6.47. Variation of peak heating with shear layer Reynolds Number for both Types III
and IV interactions.
6.4 Characteristics of Regions of Shock-Shock Boundary-Layer Interaction 295
20 35
18
30
16
14 25
Amplification Ratio
Amplification Ratio
12
20 Turbulent Corelation Values
10 C = 1.05
Laminar Corelation Values
C = 1.5 15 n = 1.3
8
n = 1.04
6 Heat Transfer, 5A 10 Heat Transfer, 5D
n Pressure, 5A n Pressure, 5D
4 qpk ppk qpk ppk
≈C Heat Transfer Curve Fit 5 ≈C Heat Transfer Curve Fit
2 qstag,∞ pstag,∞ qstag,∞ pstag,∞
Pressure Curve Fit Pressure Curve Fit
0 0
–50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 –50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0
Position (degrees) Position (degrees)
(a) (b)
Figure 6.48. Correlation of heat-transfer measurements for nominally laminar and transi-
tional and turbulent flows [61]: (a) nominally laminar; (b) transitional and turbulent.
former case, the shear-layer structure is such that the peak heating occurs for a
Type III rather than a Type IV interaction, as is the case for a fully turbulent interac-
tion. Because directly predicting transitional and turbulent heating using the Navier-
Stokes code is particularly difficult for these flows, the most tractable approach may
be to obtain peak-pressure estimates using Navier-Stokes codes (which is relatively
simple with acceptable precision compared to obtaining the heat transfer) and then
calculate a peak heating by using correlations such as those shown in Fig. 6.48a,b.
8000
ONERA Exp Data
Moss DSMC Data
7000 D’Ambrosio N-S Avg’d Data
D’Ambrosio N-S Data
Marini N-S Fine Res. Data
6000
5000
Figure 6.49. Comparison between pressure
P, Pa
53 mm
4000 25 φ
distribution in shock-interaction region from
ONERA studies and Navier-Stokes and 3000 110 mm
DSMC calculations.
2000
1000
0
–50 0 50
Angle, deg
296 Experimental Studies of Shock Wave–Boundary-Layer Interactions
14
CUBRC Exp Data
12 Candler N-S Data
Boyd DSMC Data
10
q/qc,s
0
–100 –50 0 50
Angle (deg)
shown in the previous section. For these tests, the peak pressure and heating levels
are significantly overpredicted by the numerical solutions. However, the pressure
distribution is generally in good agreement.
Shown in Fig. 6.50 are comparisons between Navier-Stokes and DSMC solu-
tions obtained by experienced computationalists and the measurements obtained
in the CUBRC wind tunnels for a fully laminar interaction region. Typically, the
calculations significantly overpredict the peak-heating and pressure levels, although
general distributions are well characterized. Exactly how DSMC or Navier-Stokes
predictions should be used to obtain peak-heating rates in laminar flows of this type
remains to be evaluated.
6.5.1 Introduction
“Back face” vehicle-cooling methods, in which heat is extracted from beneath the
skin, are frequently incapable of handling the large and dynamic heating loads
developed in shock-interaction regions. Film- and transpiration-cooling are two
alternative techniques that were proposed for hypersonic vehicles; the advantage is
that they also reduce the skin friction to the walls. Although the relative merits must
be evaluated against the complexities associated with the mechanical design of such
systems, a more important consideration is their effectiveness in the severe envi-
ronment generated by unsteady shock interactions. Film-cooling techniques using
low-molecular-weight, high-specific-heat gases were used successfully to reduce the
aerothermal loads on the optical windows of missile-seeker heads. Film-cooling
is particularly attractive because of the flexibility of this technique for a range
of free-stream conditions; however, recent studies show that the gaseous layers
used for film-cooling may be separated readily and dispersed by incident shocks.
Transpiration-cooling techniques also were shown to be highly successful in reduc-
ing heat-transfer and skin-friction levels on nose tips and the conical frustra of
6.5 SBLI Over Film- and Transpiration-Cooled Surfaces 297
(a)
Θsg = 5°
NOZZLE
INCIDENT SHOCK SHOCKS
RE-COMPRESSION
SHOCK
BOUNDARY
LAYER
COOLANT FILM
(b)
Θsg = 10°
INCIDENT SHOCK
OCK
NOZZLE SH
SHOCK
SEPA TION
RA
CK
ESSION SHO
RE-COMPR
UPSTREAM
BOUNDARY LAYER
COOLANT FILM
hypersonic reentry vehicles. They also were effective in the combusting flows down-
stream of injectors in scramjet engines.
5
SYM λ
4 0
.102
.235
3
Figure 6.52. Heat-Transfer distribution
CH/CHFP 2
in regions of incident-shock/wall-jet
1
interaction (sg = 5.5◦ , slot height =
3 mm [0.120 inch]).
0
–1
0 3 6 9 12 15 18
DISTANCE FROM SLOT EXIT (IN.)
reduction in peak-heating. For the configuration shown in Fig. 6.53, in which the
flow was separated by the stronger incident shock, the coolant film was dispersed
and the heating levels returned rapidly to those in the absence of the film. These
studies suggest that in regions of strong shock interaction, the film-cooling technique
is not robust and significantly less effective than transpiration-cooling (discussed in
the following section). Predicting whether the film remains intact when impacted by
the shock is a task that cannot be accomplished reliably.
8
SYM λ
0
.102
.237
5
Figure 6.53. Heat-transfer distribution
CH/CHFP
in regions of incident shock/wall-jet
2
interaction (sg = 8.0◦ , slot height =
3 mm [0.120 inch]).
–1
0 3 6 9 12 15 18
DISTANCE FROM SLOT EXIT (IN.)
6.5 SBLI Over Film- and Transpiration-Cooled Surfaces 299
Figure 6.54. Shock generator supported above the transpiration-cooled flat plate.
3.0 0.20
h n
without 0.18 without
2.5 with 0.16 with
2.0 0.14
0.12
Ch
Cp
1.5 0.10
0.08
1.0 0.06
0.5 0.04
0.02
0.0 0.00
0 4 8 12 16 20 24 28 32 36 40 0 4 8 12 16 20 24 28 32 36 40 44
DISTANCE (inches) FROM LEADING EDGE DISTANCE (inches) FROM LEADING EDGE
(a) (b)
Figure 6.55. (a) Heat-transfer and (b) pressure measurements at Mach 6 with and without
helium transpiration cooling for shock interaction from 5.3◦ shock generator [62].
300 Experimental Studies of Shock Wave–Boundary-Layer Interactions
1.2
Helium
Nitrogen
1.0
0.8
Ch/Che
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12 14 16 18 20 22
m /( ρsUsChs)(Cp,inj /Cp,fs ) 0.7 (Mfs / Minj ) 0.5
Figure 6.56. Correlation of heating reduction ratio with modified blowing parameter for
shock-generator angles of 5◦ , 7.5◦ , and 10◦ and both nitrogen and helium coolants [61].
6.6.1 Introduction
During the major part of the hypersonic-flight regime, the dissociation thresholds of
oxygen and nitrogen are exceeded (Fig. 6.61) and the air in the shock and bound-
ary layers enveloping a vehicle are in thermal and chemical nonequilibrium. At
high altitudes, the flow also is likely to be ionized. The changes that these real-gas
A
B
35
REGION 8
30
25
MORRIS and KEYES
Q/Qo (F − R)
TYPE IV
Figure 6.58. Heat-transfer distribu- 20
tion in shock-shock-interaction regions
Run 37
induced by a 10-degree shock generator 15 Mach = 12.160
Re/Ft = 3.507 x 105
over a transpiration-cooled hemisphere REGION 7 λ = 0.00
0
−80 −60 −40 −20 0 20 40
θ, Angular Position Relative to Stagnation Point (degrees)
18
×
16 ×
×
14
×
12 ×
×
Q/Qo(F-R)
30
25
33
20 23 372231
24 28 42
Qpeak/Qfr
25
15 30 36
34
41
10 38 Sym λ
Smooth Rough
5 46 0
0.20
0 0.31
−50 −40 −30 −20 −10 0
θ, Angular Position Relative to Stagnation Point (degrees)
Figure 6.60. Variations of peak heating with angular position of the interaction region for
various blowing parameters λ at Mach 12, Re/m = 1.07 × 106 .
302 Experimental Studies of Shock Wave–Boundary-Layer Interactions
90 300
NASP Mars Far
Altitude × 10–3, ft
75 250 return solar
system
km
45 150
30 100
O2 N2 Ionization
15 50 Dissociation Peak heating
0 0
10 20 30 40 50 60
Velocity × 10–3, ft/sec (approximate Mach Number)
0 3 6 9 12 15 18
km/sec
Figure 6.61. Velocity altitude trajectories for past and current reentry systems.
FLIGHT
Real gas/plasma effects on
re-entry communication
Real gas, bluntness/3D flow
effects on boundary layer
transition
CFD
20
Body Flap
Deflection
15 Flight
(deg) 10
Expected
5
.01
Pitching Moment
0
Coefficient
Flight
–.01
–.02 Predicted*
–.03
5 10 15 20 25
Mach Number
Figure 6.63. Comparison of predicted and measured flap deflection and pitching-moment
coefficient from STS-1 flight.
+0.02
Viscous Effect
+0.01 Fligh
tC
M (Ba
sic
B ody
)
0
CM Mach
–0.01 Number
Tunnel B Effect
Figure 6.64. Postulated real gas effect to M∞ = 8 (ref. 16)
explain pitch trim anomaly. Real-Gas
–0.02 Effect
ADOB M∞ = 20
Preflight Prediction
M∞ = 23
–0.03 (ref. 3)
Altitude = 73.1 km (240,000 ft)
δe = 0 = δBF
–0.04 CG = 0.650 L
20 25 30 35 40
a, deg
Nevertheless, with the exception of tests in the Mach 5 CF4 2 wind tunnel at NASA
and preliminary studies conducted on the Hallis configuration in European high-
enthalpy facilities, there are no definitive experimental measurements to validate
the calculations on which the design was based. The exact scale of the influence of
real-gas effects and chemistry on SBLI phenomena relative to flap controls has yet
to be determined (Fig. 6.65).
Section 6.3 concludes that in the absence of real-gas effects, reliable solu-
tions to the Navier-Stokes equations can be obtained for laminar separated flows
and those involving SBLIs. Using methods such as those employed by Candler
and Gnoffo [21], accurate predictions are possible for the nitrogen flow over the
hollow-cylinder–flares, compression surfaces, and complex regions of shock-shock
and SBLIs with imbedded separated flows over double-cone configurations. How-
ever, as shown in Fig. 6.66, this is not the case for flows when the nitrogen is replaced
by air for velocities of 4.25 km/s (14,000 ft/s) and above, in which the fluid in the
shock layer and separated regions are in a state of chemical nonequilibrium. The
measured separated region is significantly larger than that predicted by the codes.
2 The test medium in this facility is tetrafluoromethane. The density ratios across shock waves in
this gas are heightened and, consequently, flows in the hypersonic regime with the Mach numbers
roughly twice that in the test flow can be simulated in some measure.
304 Experimental Studies of Shock Wave–Boundary-Layer Interactions
0.72
DATA BOO
0.70
XCP/LB
The experimental heat transfer and pressure levels in the reattachment region also
differ significantly from computations made by Candler [66] (see Fig. 6.66). A simi-
lar result was obtained independently by Gnoffo [67] with the Langley Aerothermo-
dynamic Upwind Relaxation Algorithm (LAURA) code for the double-cone body
at 4.25 km/s.
This inability to predict flows involving SBLIs with Navier-Stokes codes is
in contrast to the experience for other situations in which real-gas effects have
an important role. For example, in America and Europe, studies examined their
0.16
0.14
0.12
0.1
St
0.08
Run 42 St
0.06 Candler St Run 42
Run 43 St
0.04 Candler St Run 43
0.02
0
0 1 2 3 4 5 6
Distance (inches)
3
Cp
Run 42 Cp
2 Candler Cp Run 42
Run 43 Cp
Candler Cp Run 43
1
0
0 1 2 3 4 5 6
Distance (inches)
Figure 6.66. Comparisons between experiment and prediction by Candler for flows of air and
nitrogen for the double-cone body at 4.25 km/s.
6.6 Real-Gas Effects on Viscous Interactions Phenomena 305
3
20
2
0.55
10
1
0 0 0.50
–3 –2 –1 0 1 2 3 0 5 10 15 20
radial distance, in total enthalpy, MJ/kg
(a) (b)
Figure 6.67. Comparisons between prediction and measurements on the Apollo capsule
demonstrating reduced vehicle stability resulting from changes in the pressure distribution
with real-gas effects: (a) comparison between prediction and measurement; (b) computation
of flowfield.
High Frequency
Pressure Instrumentatior
(a) (b)
Figure 6.68. Models and instrumentation for studies of real-gas effects on control-surface per-
formance and boundary-layer transition: (a) photograph of shuttle configuration in LENS I
tunnel for real-gas studies; (b) instrumentation locations for high-enthalpy real-gas studies.
2
Real Gas Chemistry Decreased the Size of
1.8 the Separated Region for Ho = 10 MJ/kg
1.6
1.4
1.2
0.8
0.6 N2
larger pressures are generated. The size of the separated-interaction region also is
influenced by the Reynolds number, as shown in Fig. 6.69. The fact that the inter-
action increases in size with Reynolds number is another strong indication that the
separated-interaction region remained laminar.
The effects of real-gas chemistry over the surfaces adjacent to a flap also were
investigated by conducting tests at a fixed Reynolds number while varying the total
enthalpy of the flow from 3 to 10 MJ/kg to change the level of oxygen dissociation
1.2
0.6
0.4
0.2
[Ho = 10 MJ/kg]
0
21 22 23 24 25 26 27
Distance from Nose (inches)
Figure 6.70. Measurements on shuttle wing demonstrating a reduction in pressure as a result
of real-gas effects.
308 Experimental Studies of Shock Wave–Boundary-Layer Interactions
Skin friction
1.8
PRESSURE COEFFICIENT (Y/B = 0.00) 0.05 STANTON NUMBER (Y/B = 0.00)
1.6
0.04
computed
1.4 computed
mesures
mesures 0.03
1.2
0.02
1.0
0.8 0.01
0.6 0.00
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
x / Lref x / Lref
Figure 6.71. Comparison between Navier-Stokes predictions and pressure and heat-transfer
measurements on Hallis shuttle configuration.
in the shock layer over the wing. The measurements obtained in this series of runs
are in Fig. 6.70, nondimensionalized by a reference pressure ahead of the curved
section of the wing to demonstrate that nonequilibrium-air chemistry acts to reduce
the relative pressure in the expansion process on the curved aft section of the wing.
Although pressure differences are relatively small, they act over a large area at the
trailing edge of the wing, thereby resulting in a significant decrease in the pitch-
ing moment. Thus, measurements in these studies indicate that contrary to popu-
lar belief, the flap is relatively more effective as a result of nonequilibrium effects.
However, as anticipated, the major contributor to the reduced stability of the shuttle
relative to measurements made in low-enthalpy flows results from a decrease in the
pitching moment caused by reduced pressures on the rearward part of the curved
wing.
Since the studies surrounding the investigation of the shuttle stability problem
were undertaken, little has been done to experimentally validate the numerical pre-
dictions used to explain these phenomena. However, more recent measurements
taken on the Hallis shuttle configuration in Europe were in general agreement with
numerical predictions (Fig. 6.71). This is in contrast to the unsatisfactory outcome
shown in Fig. 6.66 and, although it provides encouragement, it is clear that further
research is needed to develop and validate the computational models that reproduce
the chemistry within the codes.
REFERENCES
[1] M. S. Holden. Study of flow separation in regions of shock wave–boundary layer inter-
action in hypersonic flow. AIAA Paper 78–1169 (1978).
[2] M. J. Lighthill. On boundary layers and upstream influence. Part II: Supersonic flows
without separation. Proceedings of the Royal Society, London. Series A, 217 (1953), 1131,
217, and 478.
[3] K. Oswatitsch and K. Weighardt. Theoretical analysis of stationary flows and boundary
layers at high speed. German Wartime Report (1941). Translated as NACA TM 1189.
[4] G. E. Gadd. A theoretical investigation of laminar separation in supersonic flow. Journal
Aeronautical Sciences, 24 (1957), 10, 759–71.
[5] H. S. Glick. Modified Crocco-Lees mixing theory for supersonic separated and reattach-
ing flows. Journal Aeronautical Sciences, 29 (1962), 10, 1238–44.
[6] M. Honda. A theoretical investigation of the interaction between shock waves and
boundary layers. Journal Aeronautical and Space Science, 25 (1958), 11, pp. 667–8; see
also Tokyo University, Japan. Report Institute High Speed Mechanics, 8 (1957), pp. 109–
30.
[7] L. Lees and B. L. Reeves. Supersonic separated and reattaching laminar flows: I. Gen-
eral theory and application to adiabatic boundary layer–shock wave interactions. AIAA
Journal, 2 (1964), 1907–20.
[8] D. R. Chapman, D. M. Kuehn, and H. G. Larson. The investigation of separated flows
in supersonic and subsonic streams with emphasis on the effects of transition. NACA
Report 1356 (1958).
[9] J. N. Nielsen, F. K. Goodwin, and G. D. Kuhn. Review of the method of integral rela-
tions applied to viscous interaction problems including separation. NEAR Paper No. 7,
(presented at Symposium on Viscous Interaction Phenomena in Supersonic and Hyper-
sonic Flow, May 7–8, 1969).
[10] J. E. Lewis, T. Kubota, and L. Lees. Experimental investigation of supersonic lami-
nar, two-dimensional boundary layer separation in compression corner with and without
cooling. AIAA Paper 67–191 (1968).
[11] M. S. Holden. Theoretical and experimental studies of separated flows induced by shock
wave-boundary layer interaction. AGARD Specialists Meeting (Separated Flows, Brus-
sels, Belgium, 10–13 May 1966); see also AGARD Conference Proceedings No. 4, Part I,
Separated Flows, pp. 147–80 (1966).
[12] M. S. Holden. Theoretical and experimental studies of the shock wave–boundary layer
interaction on curved compression surfaces. ARL Symposium on Viscous Interac-
tion Phenomena in Supersonic and Hypersonic Flow, WPAFB, OH, USA (May 7–8,
1969).
[13] M. S. Holden. Shock wave–turbulent boundary layer interaction in hypersonic flow.
AIAA Paper 72–74 (1972).
[14] R..W. MacCormack and B. S. Baldwin. A numerical method for solving the Navier-
Stokes equations with application to shock-boundary layer interactions. AIAA Paper
75–1 (1975).
[15] C. M. Hung and R. W. MacCormack. Numerical solutions of supersonic and hypersonic
laminar flows over a two-dimensional compression corner. AIAA Paper 75–2 (1975).
[16] D. H. Rudy, J. L. Thomas, A. Kumar, P. A. Gnoffo, and S. R. Chakravarthy. A val-
idation study of four Navier-Stokes codes for high-speed flows. AIAA Paper 89–1838
(1989).
[17] T. Pot, B. Chanetz, M. Lefebvre, and P. Bouchardy. Fundamental study of shock-shock
interference in low-density flow: Flowfield measurements by DLCARS. In Rarefied Gas
Dynamics (Toulouse: Cépadués-Éditions, 1998), Part II-545.
[18] Proceedings of the Aerothermodynamics Workshop Reentry Aerothermodynamics and
Ground-to-Flight Extrapolation: Contributions to Hyperboloid Flare Test Cases, and
Hallis in F4 Conditions Computation of case 1.D. Eds. S. Menne and G. Hartmann,
ESTEC, Noordwijk, the Netherlands (European Space Agency, 1996).
References 311
[64] M. S. Holden and S. J. Sweet. Studies of transpiration cooling with shock interaction in
hypersonic flow. AIAA Paper 94–2475 (1994).
[65] M. S. Holden, K. M. Rodriguez, and R. J. Nowak. Studies of shock-shock interaction
on smooth and transpiration-cooled hemispherical nose tips in hypersonic flow. AIAA
Paper 91–1765 (1991).
[66] I. Nompelis, G. Candler, M. Holden, and M. MacLean. Investigation of hypersonic
double-cone flow experiments at high enthalpy in the LENS facility. AIAA Paper 2007–
203 (2007).
[67] P. Gnoffo. Private communication, 2006.
[68] M. S. Holden, T. P. Wadhams, M. MacLean, E. Mundy, and R. Parker. Experimental
studies in LENS I and LENS X to evaluate real-gas effects on hypervelocity vehicle
performance. AIAA Paper 2007–204, (2007).
[69] J. R. Maus, B. J. Griffith, K. Y. Szema, and J. T. Best. Hypersonic Mach number and
real-gas effects on space shuttle orbiter aerodynamics. AIAA Paper 83–0343 (1983).
[70] P. Gnoffo. CFD validation studies for hypersonic flow prediction. AIAA Paper 2001–
1025 (2001).
[71] M. Marini and A. Schettino, MSTP Workshop 1996 – Reentry aerothermodynamics and
ground-to-flight extrapolation contributions to hyperboloid flare test cases I.c and III.b,
CIRA, Capua, Italy (1996) Proceedings of the Aerothermodynamic Workshop held at
Estec, Noordwijk, Netherlands, 1996.
7 Numerical Simulation of Hypersonic Shock
Wave–Boundary-Layer Interactions
Graham V. Candler
7.1 Introduction
Hypersonic flows are synonymous with high-Mach number flows and therefore are
characterized by very strong shock waves. Every hypersonic vehicle has a bow shock
wave in front of it, which bounds the flow around the vehicle. On the windward side
of a vehicle, the bow shock usually is aligned closely with the vehicle surface, and
the distance between the surface and the shock wave is usually small relative to the
characteristic dimension of the vehicle. Thus, this shock-layer region is usually quite
thin. Hypersonic vehicles tend to fly at high altitudes so that convective heating lev-
els can be managed. Thus, the characteristic Reynolds numbers tend to be low and
boundary layers are usually thick. In addition, shear heating in hypersonic boundary
layers increases the temperature and viscosity, which also increases the thickness.
The low Reynolds number and the relative stability of hypersonic boundary layers
mean that many practical hypersonic flows are laminar or transitional. If the flow
is turbulent, it is often only marginally turbulent. Therefore, hypersonic flows are
particularly susceptible to shock wave–boundary-layer interactions (SBLIs).
The design of a hypersonic vehicle – whether a planetary-entry capsule, a high-
lift reentry vehicle such as the Space Shuttle, or a scramjet-powered aircraft – is
dominated by aerodynamic heating. Blunt bodies such as planetary-entry capsules
are protected from large heat fluxes by the thick boundary layer that insulates the
vehicle from the high-temperature stagnation-region flow. Anything that interrupts
this insulating layer (e.g., an impinging shock wave) can be catastrophic to a vehicle.
Shock interactions on wing leading edges and other vehicle components can be dev-
astating to a vehicle, particularly at a high Mach number in which heat-transfer rates
are especially severe. SBLIs can cause flow separation and change the effectiveness
of control surfaces on hypersonic vehicles. The flow inside the isolator of a scramjet
engine is dominated by interactions between shock waves and turbulent-boundary
layers, and these interactions are responsible for raising the pressure and prevent-
ing engine unstart due to fluctuations in engine-operating characteristics. Thus,
SBLIs can either destroy a hypersonic vehicle or be responsible for its effective
operation.
The development of future hypersonic vehicles depends on high-fidelity simula-
tion tools for their design and optimization. Current computational fluid dynamics
314
7.2 Hypersonic SBLI Physics 315
(CFD) methods can predict many aspects of practical hypersonic flows. However,
the accurate simulation of SBLI flows remains a severe challenge for the best avail-
able CFD methods. This chapter discusses the physics of hypersonic SBLI flows,
with emphasis on how the physics affects the numerical simulation of the flows.
Three general types of SBLI flows are considered: compression-corner interactions,
leading-edge shock interactions, and interactions on double-cone geometry at high-
enthalpy-flow conditions.
Many examples in the literature show that hypersonic SBLI flows are difficult to
simulate accurately. Even fully laminar flows have a huge range of length scales and
often are unsteady and three-dimensional. At higher Reynolds numbers, shear lay-
ers become transitional and reattachment is from transitional to turbulent. Turbu-
lent SBLI flows are notoriously difficult to predict because of unsteady interactions
and the inability of Reynolds-averaged Navier-Stokes (RANS) turbulence models
to represent these flows. Shock interactions on leading edges tend to be inherently
unsteady for strong interactions. All of these issues result in extreme sensitivity to
the grid quality, design, and alignment with the shock waves and other strong gradi-
ents. Simulations are also very sensitive to the numerical methods, and these flows
provide a stringent test of CFD methods.
Experiments on hypersonic SBLI flows also are difficult for several reasons.
Extreme instrument density is required to resolve strong peaks. Many flows are
unsteady, and care must be taken to minimize the free-stream fluctuations that may
drive unsteadiness and cause premature transition in shear layers. The flows are sen-
sitive to any nonideal behavior in the wind-tunnel flow; therefore, the tunnel-flow
state must be characterized with added care. Shock generators tend to contribute
to fluctuations, especially at a high Mach number because the shock angles are very
oblique, which tends to amplify disturbances.
Figure 7.1. Temperature (top) and synthetic schlieren (bottom) in the flowfield of a Mach 6
boundary layer and a 10-degree wedge-shock generator.
runs far upstream. In reality, the long shear layer at the edge of the separation zone
likely undergoes transition to turbulent flow, which would reduce the separation-
zone length. The CFD simulation cannot represent this process, and a fully laminar
flow is assumed. The CFD also produces large-scale vortical structures in the sepa-
ration zone, which generate waves that can be seen in both plots. These vortices are
transient and therefore produce unsteadiness in the interaction. Turbulent SBLIs
of this type are also known to be highly unsteady and driven by the same type of
unsteadiness in the upstream separation zone.
Figrue 7.2 plots the computed pressure and heat flux on the surface. Notice
the dramatic increase in the pressure downstream of the shock; the ideal inviscid
postinteraction pressure ratio is 10.43. Here, the increased pressure is caused by
the additional flow-turning due to the boundary layer and the complex system of
shock waves generated by the interaction. The compression of the boundary layer
in the region of shock impingement results in a large increase in the heat flux at the
interaction point, followed by a gradual reduction as the boundary layer expands
downstream of the interaction.
Figure 7.3. Mach number (top) and schlieren (bottom) in the flowfield of a 15-degree com-
pression corner at Mach 6 and unit Reynolds number of 106 per meter.
At the Mach 12 condition (Fig. 7.6), the separation shock is strong enough to pro-
duce a significant secondary SBLI, which further raises the pressure and changes
how the boundary layer adjusts to the pressure gradient.
As in the simple Mach 6 boundary-layer interaction shown in Fig. 7.1, the
separation zones predicted in these simulations are excessively large because the
flow is assumed to be laminar. However, a hypersonic transitional or turbulent
compression-corner flow would have the same basic structure as shown in the figure.
Figrues 7.7 through 7.9 show the surface-pressure and heat-transfer-rate coeffi-
cients for 10-, 15-, and 20-degree compression-corner flows at Mach numbers from
6 to 12. Interesting conclusions can be drawn from these figures. For the 10-degree
corner, the nondimensional pressure increase is largest for Mach 6, decreasing with
Figure 7.4. Mach number (top) and schlieren (bottom) in the flowfield of a 15-degree com-
pression corner at Mach 8 and unit Reynolds number of 106 per meter.
7.2 Hypersonic SBLI Physics 319
Figure 7.5. Mach number (top) and schlieren (bottom) in the flowfield of a 15-degree com-
pression corner at Mach 10 and unit Reynolds number of 106 per meter.
Mach number. However, for the 15- and 20-degree turning angles, the Mach 12 flow
has the largest pressure rise. This is a result of the separation-zone shock wave inter-
acting with the primary shock, causing an additional pressure rise. This transmitted
shock then reflects between the surface and the primary shock, which produces the
pressure variations shown in Figs. 7.8 and 7.9.
In all cases, the pressure increase on the wedge is significantly larger than pre-
dicted by inviscid theory. Newtonian aerodynamics provide pressure coefficients of
0.060, 0.134, and 0.234 for the 10-, 15-, and 20-degree cases, respectively. The larger
computed pressure increase is a result of the additional displacement of flow due
to the boundary layer and, as in the flat-plate interaction, the more complex system
of shock waves that the interaction causes. Far downstream from the corner, the
Figure 7.6. Mach number (top) and schlieren (bottom) in the flowfield of a 15-degree com-
pression corner at Mach 12 and unit Reynolds number of 106 per meter.
320 Numerical Simulation of Hypersonic Shock Wave–Boundary-Layer Interactions
Figure 7.7. Pressure and heat-transfer coefficients on a 10-degree compression corner for
laminar flow at a unit Reynolds number of 106 per meter.
Figure 7.8. Pressure and heat-transfer coefficients on a 15-degree compression corner for
laminar flow at a unit Reynolds number of 106 per meter.
Figure 7.9. Pressure and heat-transfer coefficients on a 20-degree compression corner for
laminar flow at a unit Reynolds number of 106 per meter.
7.2 Hypersonic SBLI Physics 321
pressure approaches the lower theoretical values. The pressure coefficient scales
approximately with the corner angle squared, as predicted by inviscid theory.
The separation zone increases with Mach number for a given turning angle and
with corner angle for a given Mach number. These increases are expected because
the dimensional pressure rise increases with both Mach number and corner angle,
resulting in a stronger adverse pressure gradient and a stronger separation.
The heat-transfer coefficient shows a simple variation with Mach number and
turning angle; in all cases, it increases as the strength of the interaction increases.
The largest heat-transfer rate occurs in the region of strongest boundary-layer com-
pression, where the foot of the corner shock is located. The peak value has a strong
Mach number and turning-angle scaling, particularly because the heat-transfer coef-
ficient already has an M3 scaling.
Figure 7.10. Mach number contours for a hypersonic shock interaction on a cylinder; shock-
impingement locations vary for each image.
is compressed so much that it is essentially invisible at the plot scale. Naturally, this
produces a huge increase in the heat flux at the interaction location. The second
image shows a slightly different condition, in which the jet impinges obliquely on
the cylinder surface. It is interesting that this condition produces a higher-peak heat
flux than the condition with the surface-normal jet.
Figure 7.12 plots the variation of surface pressure and heat flux for the cases
discussed herein. The normalization is with the stagnation-point pressure and heat
16 24
20
12
16
p/p0
q/q0
8 12
0 0
–90 –60 –30 0 30 60 90 –90 –60 –30 0 30 60 90
θ (deg) θ (deg)
Figure 7.12. Surface pressure and heat flux for the shock-interaction flow shown in Fig.
7.10; each line corresponds to a different shock-impingement location. Normalization is with
respect to the undisturbed stagnation-point pressure and heat flux.
324 Numerical Simulation of Hypersonic Shock Wave–Boundary-Layer Interactions
flux for the undisturbed case. The shock interaction gradually strengthens as the
impingement point is lowered (i.e., moved toward negative zero). First there is a
strong interaction (shown at the top of Fig. 7.11), then a reduction in maximum pres-
sure and heat flux, until the strongest interaction occurs (shown at the bottom of
Fig. 7.11). This worst case gives a peak pressure of about sixteen times the undis-
turbed stagnation-point pressure and twenty-two times the stagnation-point heat
flux. The peaks are narrow, with a width similar to the length scale of the terminal
shock in the impinging jet. The spreading of the peak is due mostly to the finite
thickness of the thermal boundary layer on the cylinder; as such, it is expected
that higher-Reynolds-number conditions have even narrower pressure and heat-
flux peaks. Clearly, these types of interactions must be designed against to avoid
thermal-protection failure.
Comparisons with experimental data are not presented here, but interested
readers are referred to the work of D’Ambrosio for details [9]. In general, the
comparisons are not especially good; the computational simulations overpredict
the peak pressure and heat-transfer rates by about 30 percent. D’Ambrosio makes
a careful study of how the finite dimension of the pressure and heat-flux gauges
and the varying surface temperatures affect the comparisons. Accounting for these
nonideal effects improves the agreement and illustrates the difficulty of conducting
shock-interaction experiments. The additional complication of flowfield unsteadi-
ness was not addressed in D’Ambrosio’s study, which would be difficult because the
amplitude and frequency of the possible flowfield variations are not known.
r The method must have a low level of numerical dissipation in the inviscid
numerical flux function. The method must be at least second-order accurate,
and the slope limiters and other means for controlling the solution must be cho-
sen with care.
7.3 Numerical Methods for Hypersonic Shock–Boundary-Layer Interaction Flows 325
5.75
5.50
L, cm
of axial-direction grid points for a
Roe + minmod
double-cone flow at Mach 12; various Roe + van Albada
widely used upwind methods are shown 5.00 Roe + van Leer
with a range of slope limiters for the Roe + superbee
LF + minmod
numerical fluxes [23]. HLLC + minmod
4.75 HLL + minmod
4.50
256 512 1024 1536 2048
Nx
r The grid used must be of high quality with smoothly varying grid spacing
and high resolution in the boundary layer and shock triple-point vicinity. The
separation and reattachment points must be well resolved in both the stream-
wise and wall-normal directions so that the separation point is predicted cor-
rectly. Careful grid-convergence studies must be conducted to show that the
grid resolution is sufficient.
r Simulations must be run for many flow times to produce a steady-state result.
The separation zone takes a long time to establish, and care must be taken to
verify that the solution has reached a steady state. If the flow is separated, the
separation-zone size changes very slowly near the end of convergence; it must
be monitored as a convergence metric. In addition, experimental data must be
kept at arm’s length to reduce the possibility of running the CFD solution until
the separation zone matches that of the experiment. In some cases, there is a
good reason why the simulated separation may be larger than the experiment
(e.g., unmodeled physics in the simulation).
r Care must be taken when using local time-stepping methods or approximate
implicit methods for the time integration. Due to the extreme range of length
scales in the grid, these approximations in the time domain may cause aphysical
solutions to be produced with CFD.
r Turbulent shock–boundary-layer interactions are notoriously difficult to sim-
ulate, and several turbulence models with appropriate realizability constraints
should be used to compare the sensitivity of the predictions to these modeling
choices.
r The separation-zone size and peak heating rates are strong functions of the
numerical method and grid resolution. It is important to fully investigate this
sensitivity before drawing conclusions about the accuracy of a CFD result.
next section) for various popular numerical methods and slope limiters. The exact
details of the methods are not particularly important for this discussion; instead,
there can be extreme variation of the separation-zone size depending on the numer-
ical method used. Also, dissipative methods (i.e., those that tend to smooth gradi-
ents over large numbers of grid cells) may require massive numbers of grid points to
obtain the correct asymptotic solution. It is interesting that this study shows that the
choice of slope limiter used to control numerical errors near strong gradients can
have a major effect on the accuracy of a method for predicting SBLIs. Therefore,
it is mandatory that careful studies are conducted to verify the accuracy of CFD
simulations of shock-interaction flows.
An additional numerical issue concerns the use of local time-stepping meth-
ods in the simulation of separated flows. It is not widely recognized in the litera-
ture that local time-stepping methods can cause artificial periodic shedding of vor-
tices and other spurious numerical artifacts. Gnoffo [10] clearly illustrates this effect
for a concave geometry and for the separated flow on the double-cone SBLI flow.
Furthermore, he shows that using a time-accurate method eliminates this behavior
and allows CFD to obtain the steady-state experimentally observed solution. There-
fore, great care must be taken in how the equations are integrated in the temporal
domain.
The numerical simulation of hypersonic shock interactions with turbulent flows
remains an outstanding challenge to the CFD community. It is only recently that
low-Reynolds-number turbulent shock interactions have been simulated with Direct
Numerical Simulations (DNS) and Large-Eddy Simulations (LES) [11, 12, 13]. For
realistic Reynolds numbers, we still must rely on RANS simulations for interactions
with turbulent boundary layers. All RANS models use turbulent kinetic energy
(TKE) production terms that depend on the local strain rate; in shock waves, the
production becomes extremely large, often reaching aphysical values. Furthermore,
it can be shown [14, 15] that the unsteady interaction of turbulence with shock waves
actually suppresses TKE production. This effect is not represented in widely used
RANS models. MacLean et al. [16] show the importance of this effect in compar-
isons of Mach 7.16 turbulent flow over a 33-degree flare. Figure 7.14 plots the exper-
imental schlieren image of the turbulent shock interaction and results of two simula-
tions. In the first simulation using the baseline Spalart-Allmaras (SA) one-equation
RANS model [17], no separation is predicted. Then, using the strain-adaptive lin-
ear SA model of Rung et al. [18], the separation zone is in much better agreement
with the experiment. This model modification limits the production of TKE in the
high strain-rate regions of the flow. This approach improves the pressure distribu-
tion in the region of the separation; unfortunately, it does not significantly improve
the heat-flux prediction [16]. Clearly, there is great need for improved turbulence
modeling for these flows.
A related problem occurs in the isolator of scramjet engines. The isolator is
usually a constant-area section of the propulsion system located between the exit
plane of the inlet and the combustor section. The purpose of the isolator is to sepa-
rate fluctuations in the inlet and combustor conditions so as to reduce the potential
for engine unstart. Also, the isolator is responsible for providing additional pres-
sure rise upstream of the combustor. The boundary layers on the isolator surface
are usually turbulent (because of either natural transition or they have been tripped
7.4 Example: Double-Cone Flow for CFD Code Validation 327
to turbulence); thus, shock interactions in the isolator occur with thick turbulent-
boundary layers. The source of the shock waves is either from the imperfect can-
celing of shock waves by the inlet, due to off-design operation of the inlet, from
the inlet cowl, or from the varying displacement of the turbulent-boundary layer
itself. In any case, high-Mach-number isolator flows are characterized by shock
wave–turbulent boundary-layer interactions, and these interactions are responsible
for effective operation of the isolator. The resulting shock train can absorb pres-
sure increases that occur in the combustor without allowing the pressure to travel
through the isolator boundary layer, potentially resulting in a catastrophic unstart
of the inlet.
Little is known about how to predict the isolator flowfield. The classic exper-
imental work of Waltrup and Billig [19] with associated empirical expressions for
the pressure rise in cylindrical ducts is probably state-of-the-art in this area. This
problem is particularly difficult to simulate with CFD because of the importance
of the turbulence model in turbulent interactions. As discussed previously, a small
adjustment in the turbulence model can change a compression-corner flow from
attached to separated. This dependence is magnified in an isolator flow because of
the numerous shock interactions that can occur in the shock train. Any error result-
ing from missing physics in the turbulence model is amplified every time a shock
wave interacts with the wall boundary layer.
M>>1
M<1
Contact surface
Contact surface
Supersonic jet
Separation shock
Oblique shock Transmitted shock
Separation zone
The double-cone flow was used for code-validation studies related to a NATO
Research and Technology Organization (RTO) Working Group Study; as a result,
several research groups simulated this flowfield. The experiments were performed
at well-characterized hypersonic conditions in the CUBRC, Inc. (formerly the
Calspan–University at Buffalo Research Center) Large Energy National Shock
Tunnel (LENS) [24, 25]. These experiments used a large model with many surface-
mounted heat-transfer-rate and pressure transducers. Nitrogen was used for the
test gas to minimize the effects of chemical reactions, and the experiments were
conducted at low pressure to ensure laminar-boundary layers and shear layers. In
general, the comparisons between simulation and experiment were good [26]; how-
ever, there were several important differences. It is interesting that the simulations
performed with high-quality numerical methods on the finest grids slightly over-
predicted the size of the separation zone, and all simulations predicted excessive
heating in the attached region prior to separation. This is shown in Fig. 7.16, which
presents typical results for two double-cone cases. The error in heat-transfer rate
on the first cone is as much as 20 percent, which is particularly puzzling because
the pressure is accurately predicted in this region. Many attempts were made to
explain this difference by running CFD cases with finite nose-tip bluntness, model
misalignment, and uncertainties in reaction rates. None of these effects explained
the differences shown in Fig. 7.16.
The specification of the free-stream conditions in a hypersonic shock tunnel can
be difficult because these facilities can be subject to nonideal effects in the nozzle –
namely, a reflected shock wave is used to heat and compress the test gas, which
results in vibrational excitation and chemical reaction. It is then rapidly expanded
through the nozzle, and the thermochemical state may not fully de-excite during
expansion. As a result, the gas flowing over the model may be in a nonideal thermo-
chemical state. During design of these experiments, we were aware of the potential
for this problem and chose nitrogen as the test gas to minimize chemical reactions.
We also kept the enthalpy of the tests relatively low (i.e., ho < 4 MJ/kg) to fur-
ther reduce this effect. As a result, there is essentially no chemical reaction of the
test gas prior to expansion through the nozzle. Nitrogen vibrational modes relax
very slowly; for these test conditions, this results in elevated vibrational energy in
the wind-tunnel test section. A vibrational finite-rate simulation of the nozzle flow
shows that the vibrational-energy modes are frozen near the throat temperature
(i.e., Tv = 2,560 K). This has two major effects: (1) the kinetic-energy flux is reduced
by about 10 percent; and (2) because nitrogen vibrational-energy modes are ineffi-
cient at accommodating to most metallic surfaces, they do not transfer their energy
to the model. These two effects reduce the heat flux by about 20 percent and sig-
nificantly improve the comparison between CFD and experiment (Fig. 7.17). More
details about the double-cone code validation study are available in [20].
Consider the effects of high-enthalpy-flow conditions on the double-cone prob-
lem – specifically, how finite-rate chemical reactions affect shock interactions in this
flowfield. As a follow-on to the code-validation study discussed previously, a series
of high-enthalpy nitrogen and air experiments was performed on the double-cone
geometry [24]. Results of this study are summarized in Fig. 7.18 in the form of heat-
transfer rates to the double-cone for three air cases [25, 27, 28]. These runs were at
approximately the same free-stream Reynolds number and Mach number, and the
total enthalpy was increased from 4.5 to 15.2 MJ/kg (at 4.5 MJ/kg, Re = 3.1 × 105 /m
per meter; at 10.4 MJ/kg, Re = 2.9 × 105 /m and at 15.2 MJ/kg, Re = 2.3 × 105 /m).
330 Numerical Simulation of Hypersonic Shock Wave–Boundary-Layer Interactions
Figure 7.18. Measured and predicted heat flux to the double-cone model for high-enthalpy
air conditions (4.5, 10.4, and 15.2 MJ/kg total enthalpy).
First, we focus on the computed results shown with the lines. As the free-stream
enthalpy is increased, the simulations predict that the separation zone decreases in
size (visualized by the extent of low heat-flux upstream of the x = 10 cm point). This
is because the increasing energy of the flow increases the level of chemical reaction
in the flow; in particular, there is an increased level of reaction in the separation
zone. In air, the reactions are endothermic, which has the effect of decreasing the
temperature and molecular weight of the mixture (i.e., diatomic molecules (O2 in
particular) dissociate to form atoms).
Why does the increasing level of reaction decrease the size of the separation
zone? For a normal shock wave in air, the main effect of chemical reactions is a
decrease in the postshock temperature and an increase in the postshock density
relative to a perfect-gas shock wave. Second, the molecular weight of the mix-
ture decreases. The combination of these changes results in approximately the
same postshock pressure because the pressure is set by the momentum-flux balance
across the shock wave. The result of this change in the postshock conditions is to
decrease the shock standoff distance on blunt bodies. (The shock standoff distance
7.4 Example: Double-Cone Flow for CFD Code Validation 331
is inversely proportional to the density rise across the bow shock wave.) Thus, the
same effect occurs in the separated-flow region: increasing chemical reaction causes
a larger density increase in the separation zone and a smaller separation zone.
This effect is predicted by CFD simulations of the high-enthalpy air flows shown in
Fig. 7.18. The CFD predicts a reduction in the heat flux caused by the chemical reac-
tions; this occurs because the smaller separation zone produces more oblique shock
waves and a weaker shock interaction between the separation shock and the bow
shock on the second cone. Thus, the higher-enthalpy interactions are less clearly
defined and result in a lower peak-heat-transfer rate (when nondimensionalized by
the velocity cubed or by ho3/2 ).
Consider the experimental data points shown in Fig. 7.18. At the lowest
enthalpy condition, the comparison between simulation and experiment is good but
not perfect. CFD predicts a smaller separation zone, which results in a stronger
and more upstream interaction. For this case, the heat flux to the first cone is well
predicted, indicating that the nozzle energy flux is well characterized for this con-
dition. At the higher-enthalpy conditions, the agreement worsens significantly. At
10.4 MJ/kg, the heat flux is underpredicted and the separation zone is much smaller
than measured. At the highest enthalpy, the discrepancy is even worse, with the
measured separation zone twice as large as predicted by CFD. Furthermore, it is
interesting that the experimental trend is different than the CFD results: the small-
est measured separation zone is at the intermediate enthalpy (10.4 MJ/kg), whereas
CFD predicts a decreasing separation-zone size. The discrepancy in predicted heat-
ing on the cone forebody for the two higher-enthalpy cases indicates that the free-
stream conditions used for these simulations are likely wrong [28].
The discrepancy between CFD and the experimental measurements currently
is not understood. However, there is no reason to believe that there is a funda-
mental problem with the simulations, especially because the lower-enthalpy double-
cone flows in nitrogen were accurately predicted (once all the relevant physics were
modeled). Clearly, there is more to be learned about simulating these flows and,
most important, how to model expansion of the high-enthalpy flow in a reflected
shock-tunnel test facility. More details about modeling uncertainties and attempts
to understand the differences discussed here are in [27] and [28].
In related work by Holden, Wadhams, and MacLean [25], the control-surface
effectiveness problem on the Space Shuttle Orbiter is examined in the LENS facil-
ity. This work revisited the problem identified during the first reentry of the Orbiter
during STS-1 [29, 30], in which the predicted aerodynamic performance was signif-
icantly different than encountered during flight. Maus et al. [31] showed that the
differences were caused primarily by Mach number and real-gas effects. Weilmeun-
ster et al. [32] used more sophisticated CFD methods to show that real-gas effects
decrease the size of the separation zone on the body flap, changing the pressure
distribution and pitching moment. The new experiments of Holden et al. [25] fur-
ther confirm this finding, showing that at the same Reynolds number and total
enthalpy, the separation zone in air is smaller than in nonreacting nitrogen. This
effect increases with increasing enthalpy, as expected. In addition to changes in the
SBLI on the body flap, air chemistry (primarily oxygen dissociation) decreases the
pressure in the expansion over the curved aft section of the Orbiter wing, further
changing the pitching moment.
332 Numerical Simulation of Hypersonic Shock Wave–Boundary-Layer Interactions
From this discussion, several conclusions are drawn about high-enthalpy effects
on hypersonic shock interactions. First, because the separation zone is a slowly mov-
ing region, the gas tends to reach thermal and chemical equilibrium in the separa-
tion zone. This high level of chemical reaction absorbs energy into the internal- and
chemical-energy modes, which affects the gas dynamics of the flow through changes
in temperature and gas constants. This changes the size of the separation zone, just
as the shock standoff distance is reduced by real-gas effects. It is interesting that the
experimental data shown in Fig. 7.18 do not correspond to the expected trend; the
CFD results follow this trend. Of course, there are other subtle effects that are dif-
ficult to classify: the chemical reactions change the viscosity of the flow; the reduced
temperature increases the density, which affects the rate of entrainment across the
shear layer; and the shock angles are changed due to the real-gas effects. These
processes also affect the size of the separation zone and the strength of the interac-
tion between the separation shock and the detached shock on the second cone. This
changes the peak-heating level on the second cone. Again, CFD predicts a reduction
in the interaction heating rate due to the reduced separation-zone size.
7.5 Conclusions
Hypersonic shock wave–boundary-layer interactions have many features in com-
mon with interactions at transonic and supersonic Mach numbers. In all condi-
tions, the interactions result in complex flowfields characterized by a huge range of
length scales. However, hypersonic interactions have features that distinguish them
from their lower-Mach-number counterparts. Most important, hypersonic interac-
tions can produce huge increases in the surface pressure and heat flux because they
can interrupt the thick insulating boundary layer on hypersonic vehicles, allowing
the high-energy external flow to directly attack the surface. Thus, to a large extent,
strong shock interactions (e.g., Edney Types IV and V interactions) must be pre-
vented at all costs.
Another interesting aspect of hypersonic interactions is that the impinging
shock wave tends to be at a highly oblique angle to the boundary layer. However,
even a shallow hypersonic-shock wave can produce a large adverse-pressure gra-
dient, with resulting flow separation. For a laminar-boundary layer, the separation
zone tends to run far upstream, producing a long shear layer at the edge of the sepa-
ration zone. This free shear layer is unsteady and prone to transition, except at small
Reynolds numbers. Thus, laminar separations tend to become turbulent in the shear
layer, resulting in transitional or turbulent heating levels at reattachment. Current
conventional CFD methods cannot capture the complex transitional-flow physics.
Hypersonic-shock interactions on compression corners can be computed accu-
rately if the flow remains laminar, the free-stream conditions are well characterized,
and there is little or no chemical reaction in the flowfield. As shown for the hyper-
sonic double-cone code-validation studies, CFD can predict this rather narrow class
of flow if the CFD method is not excessively dissipative and the grid is quite large.
These flows are difficult to simulate because any error in the physical modeling or
solution of the governing equations is amplified by the separation zone. A too-small
separation zone results in a weaker interaction and smaller pressure rise, which
reduces the adverse-pressure gradient and further reduces the separation-zone size.
References 333
Thus, there is a feedback loop that accentuates error in the numerical simulation
or in the specification of free-stream conditions. Endothermic chemical reactions
and other internal-energy excitation processes decrease the temperature rise and
increase the density rise across shock waves. In general, this results in thinner shock
layers and, as a result, smaller separation zones. However, it is difficult to perform
experiments at relevant high-enthalpy conditions such that the flow remains lami-
nar and the free-stream conditions are well understood. Current experimental data
suffer from a degree of uncertainty, and CFD cannot replicate these flows.
Shock wave–turbulent boundary-layer interactions at hypersonic conditions
remain a major difficulty for numerical simulations. Most RANS turbulence models
are built on results from the incompressible literature. Simple compressibility cor-
rections are added to account for nonuniform densities in compressible flows. It is
rather optimistic (and unrealistic) to hope that such corrections can account for the
huge density variations that occur in hypersonic flows. In practice, RANS models
may be tuned to obtain reasonable agreement with experimental data for a narrow
class of interactions, but a truly predictive capability currently does not exist. Shock
interactions with turbulent boundary layers are particularly important in scramjet
isolators because a shock train undergoes a series of boundary-layer interactions.
Therefore, turbulence-modeling errors accumulate and the pressure rise through
the isolator is poorly predicted with current CFD methods.
Acknowledgments
Most of the work presented herein was supported by the Air Force Office of
Scientific Research (AFOSR) under Grants FA9550–04-1-0114 and FA9550–04-1-
0341. The author acknowledges Dr. John Schmisseur for his support of the code-
validation studies over an extended period. The views and conclusions in this chap-
ter are those of the author and should not be interpreted as necessarily representing
the official policies or endorsements – either expressed or implied – of the AFOSR
or the U.S. Government.
REFERENCES
[1] B. E. Edney. Anomalous heat transfer and pressure distributions on blunt bodies at
hypersonic speeds in the presence of an impinging shock. FFA Report 115, Aeronautical
Research Institute of Sweden, Stockholm (1968).
[2] B. E. Edney. Effects of shock impingement on the heat transfer around blunt bodies.
AIAA Journal, 6 (1968), 1, 15–21.
[3] S. R. Sanderson. Shock wave interaction in hypervelocity flow. Ph.D. thesis, California
Institute of Technology, USA (1995).
[4] S. R. Sanderson, H. G. Hornung, and B. Sturtevant. Aspects of planar, oblique and
interacting shock waves in an ideal dissociating gas. Physics of Fluids, 15 (2003), 5, 1638.
[5] S. R. Sanderson, H. G. Hornung, and B. Sturtevant. The influence of non-equilibrium
dissociation on the flow produced by shock impingement on a blunt body. Journal Fluids
Mechanics, 516 (2004), 1.
[6] M. S. Holden, J. Moselle, and S. Martin. A database for aerothermal measurements in
hypersonic flow for CFD validation, AIAA Paper 1996–4597 (1996).
[7] M. S. Holden. Real gas effects on regions of viscous-inviscid interaction in hypervelocity
flows. AIAA Paper 1997–2056 (1997).
334 Numerical Simulation of Hypersonic Shock Wave–Boundary-Layer Interactions
8.1 Introduction
Many hypersonic vehicles are designed to follow trajectories that extend well into
the upper atmosphere where the density is extremely low. Despite this, aerodynamic
heating is still a critical issue because of the very high flight velocity. The U.S. Space
Shuttle Orbiter, for instance, experienced peak heating at a height of about 74 km
even though ambient density at that altitude is not much more than one millionth
of sea-level density. Shock wave–boundary-layer interactions (SBLIs) that occur
within these flows are nearly always sites of intense localized heating; thus, it is
essential to predict the level correctly to avoid vehicle structural failure or incur-
ring unnecessary weight penalties by carrying excessive thermal protection.
Along vehicle trajectories in the upper atmosphere, with increasing altitude,
the ambient density drops more rapidly than the velocity rises. For this reason, the
Reynolds numbers drop to relatively low values and most (if not all) of the flow is
likely to be laminar. An SBLI is then more likely to involve separation; neverthe-
less, predicting the flow should be relatively straightforward because uncertainties
associated with deciding the location of transition and choosing an appropriate tur-
bulence model are avoided. However, it is observed that above 50 to 60 km, discrep-
ancies begin to appear between the measured flowfields and flow solutions obtained
using Navier-Stokes methods. Initially, as the altitude increases, these differences
are small and can be avoided by using modified surface-boundary conditions to
account for velocity slip and temperature jump. However, at greater altitudes where
the density drops farther, the discrepancies are more significant and cannot be elim-
inated using this approach. This genuine departure from the Navier-Stokes predic-
tions is attributable to so-called ‘rarefaction effects’ and it is due to the underlying
foundations on which these equations are based being no longer strictly applicable.
Rarefaction leads to fundamental changes in the way the flowfield is structured and
hence it can significantly influence the way that SBLIs develop. Generally, rarefac-
tion reduces the severity of the heat transfer; however, in other respects, it can be
detrimental – for example, by reducing the effectiveness of control surfaces. Hence,
it is important to quantify accurately the impact.
Before reviewing specific examples that illustrate how rarefaction influences
SBLIs, this chapter outlines general characteristics of rarefied flows and discusses
336
8.2 Prediction of Rarefied Flows 337
a few methods available to predict them. Readers familiar with this background
material may proceed to Section 8.4, in which implications for SBLI flows are con-
sidered.
Kn = λ/L (8.1)
and it is the key parameter that determines the degree of rarefaction. If it exceeds 10,
an insignificant number of particle collisions occur within the gas near the body and
the flow is then said to be collisionless, or free molecular (FM). These flows are rel-
atively easy to analyze and appropriate methods to solve them are well documented
[1]. The solutions are useful because they define a limiting behavior of rarefied flows
at the low-density extreme. At the other extreme, the continuum limit corresponds
strictly to vanishingly small values of Kn but, for hypersonic flows, it safely can be
assumed to apply if the value based on the free-stream mean-free path is less than
roughly 10 − 3 . For vehicles flying in the upper atmosphere,2 this value is exceeded
frequently; hence, it is necessary to acknowledge that the fluid is not a continuum
and therefore rarefaction effects must be considered.
Even if the density is very low, it is not possible to solve these flows deterministi-
cally by following the motion of each particle because of the exceedingly large num-
bers involved. Fortunately, alternatives solution methods are available; the princi-
pal options come within what is termed the classical Kinetic Theory of Gases and
the associated numerical methods. Vincenti and Kruger [2], Chapman and Cowl-
ing [3], and Cercignani [4] provide excellent explanations of this subject and their
works are recommended for further reading. Analytic approaches to kinetic theory,
for the most part, have been unsuccessful as a means for solving practical aerody-
namic problems, and certainly nothing nearly as complex as a flow involving an
SBLI has been tackled successfully. Nevertheless, the importance of this theory
should not be underestimated because it provides the foundation for understand-
ing the physics of rarefied flows and is the basis on which successful numerical
particle-simulation methods have been devised. Of these methods, the powerful
Direct Simulation Monte Carlo (DSMC) method devised by Bird [1] is the most suc-
cessful tool for predicting practical rarefied flows. For readers unfamiliar with the
1 In some situations (e.g., when studying the structure of shock waves), a more appropriate length
scale to use is one based on local gradients within the flow, such as ρ/(dρ/ds), where ρ is the
density and s is the scalar distance.
2 The mean free path is about 0.005 m at 81 km altitude in the earth’s atmosphere.
338 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
Kinetic Theory of Gases and the DSMC method, the relevant aspects are outlined in
Appendix A.
Most of the SBLI discussion in this chapter is based on DSMC computations,
although experimental results also are presented. However, there is a dearth of
reliable measured data because it is inherently difficult to perform wind-tunnel
experiments on rarefied flows. Signal levels from instruments are low, the flows
are sensitive to probe interference, and unique measurement errors arise that are
attributable to rarefaction effects. Furthermore, providing a suitable environment
in which to conduct the tests is a challenging task and it is difficult to obtain properly
characterized and gradient-free wind-tunnel flows. Conventional flow-visualization
techniques such as schlieren and shadowgraph are usually insufficiently sensitive;
however, the low density allows alternative nonintrusive optical diagnostic meth-
ods such as the widely used electron-beam fluorescence technique to be used. This
technique readily yields information on flow density, and the translational, vibra-
tional, and rotational temperatures also can be inferred using spectrometric meth-
ods. This technique was used to obtain measurements shown in Appendix Figs. A.1
and A.3. Conventional methods of measuring heat transfer (e.g., thin-film resistance
gauges) are generally satisfactory except for problems in addressing low signal lev-
els. Static-pressure measurements are more problematic and sizeable errors arise if
the conventional technique of using small holes (or “tappings”) in the model sur-
face is used.3 As Kn is increased, these errors become larger and more difficult to
quantify because they are dependent on the unknown velocity distribution of the
gas particles in the flow just above the tapping. The error can be avoided by using
pressure transducers with sensing diaphragms set flush to the surface of the body.
Small instruments with adequate sensitivity are not widely available but were used
in the experiment discussed in Section 8.4.
the translational energy. If the balance of energy between these modes varies, the
reaction probabilities will change, which will bear directly on the species concentra-
tions within the gas. The reduced collision frequency also can lead to the population
of the quantum levels within any specific mode being perturbed, which can cause
further deviation in the reaction rates from those of continuum (i.e., Arrhenius)
chemistry. For these reasons, determining thermodynamic changes and chemical
reactions within rarefied flows is difficult and only with the advent of powerful
particle-simulation methods, that even extend to dealing with the chemical reactions
at a quantum physics level, has any significant progress been made in realistically
solving these problems.
ceases to satisfy the Rankine-Hugoniot conditions even though the inclination vir-
tually is unchanged. Merging occurs in other situations, and the existence of shock
waves that do not satisfy the Rankine-Hugoniot conditions is a unique feature that
distinguishes rarefied flows from continuum low-Reynolds-number flows. The merg-
ing and weakening of the shock wave reduces the pressure and heat transfer on the
flat plate nearby below the levels predicted by strong-interaction theory. This first
occurs when Kn is roughly 0.005; upstream of this point, kinetic-theory or particle-
simulation methods are the only viable ways to analyze the flow correctly. Merging
and the consequential weakening of the shock waves are major contributors leading
to the reduced intensity of SBLIs in the rarefied flow regime.
Close to the leading edge, Kn increases toward unity. The flow then can be
best visualized as a cloud of gas molecules surrounding the tip of the plate that
experiences relatively few collisions to change their velocity. Within this so-called
‘kinetic region’, the shock wave completely loses its identity.
The changes in the flow character toward the leading edge create a significant
departure in the pressure and heat-transfer distributions from the continuum values.
Fig. 8.2a shows a typical pressure distribution obtained using the DSMC method
compared with results from the strong-interaction theory. The data are plotted
against the inverse Kn; this corresponds to the physical streamwise dimension. The
heat-transfer distribution to a flat plate for a higher-Mach-number example is plot-
ted in Fig. 8.2b and shows more detail of the flow near the leading edge. Both the
pressure and heat transfer rise to peaks near the tip and then drop within the kinetic
region when 1/Kn is less than about 10.
The second rarefied flow example exhibits merging but it also is a reminder
that at low densities, shock waves have significant thickness and therefore occupy a
finite space within the flowfield. They typically have a thickness of ten to twenty
times the incident mean-free path; therefore they can no longer be regarded as
discontinuities. The flow considered here is of a flat-ended circular cylinder with
the axis aligned to the approaching free stream. Figrue 8.3 shows three sets of
DSMC computations4 in which the degree of rarefaction is increased from near
4 The validity of the computed results shown here is supported by evidence from similar computations
for the same flow shown in Fig. 3.A in the Appendix of this chapter, where predictions compare
favorably with experimental data.
8.3 Characteristics of Rarefied Flows 341
(b)
continuum toward FM flow. The figure shows the density variation along the stag-
nation streamline as a fraction of the free-stream value, where x is the coordinate in
the free-stream direction measured from the front face of the cylinder and n is the
particle number density. Kn is based on the cylinder’s radius R, which is 0.01m.5
For the densest case for which Kn = 0.0083, the shock wave – indicated by a
sharp rise in density – forms at around x/R = –0.45. Although this Kn places the flow
close to the continuum limit, the finite thickness and internal structure of the shock
wave are evident. Downstream of the shock wave, a plateau in density indicates that
this wave is separated from the viscous flow on the front face of the cylinder. Apart
5 Other flow conditions in this example are T∞ = 40K, n∞ = 3.0 × 1021 /m3 , T∞ /Twall = 0.135, n is
the particle number density, and the gas is nitrogen.
342 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
from the shock-wave thickening, the flow structure otherwise appears to be similar
to the continuum counterpart for a cold-walled body.
At the intermediate density, for which Kn = 0.027, a true plateau downstream
of the shock wave no longer can be identified. This is due to the shock wave merging
with the viscous flow on the front face of the cylinder, and it is not possible to state
exactly where the shock wave ends. Nevertheless, it is clear from the structure that
the shock wave is considerably thicker than in the first case. In the final example, Kn
is increased to 0.083, which is only ten times larger than in the first flow. However,
it is apparent that over this relatively small Kn range, the structure of the flow has
altered completely. There is no trace of a shock wave; rather, a smooth monotonic
rise in density occurs as the flow approaches the surface. If the density is reduced
further, much the same pattern persists, but the thickness of the perturbed layer on
the front face of the body decreases with rising Kn. This new flow structure can be
understood by first visualizing the equivalent “near FM” flow in which the incident
stream of particles will not experience a significant number of collisions until they
almost reach the body’s face, where they congregate in a dense layer on the surface
before flowing radially outward. A similar process is evident in the lowest density
example (i.e., Kn = 0.083), although the flow cannot be classed as truly FM because
collisions occur within the gas well ahead of the body. Nevertheless, for this flow,
the pressure and heat-transfer coefficients on the front face approach closely the FM
values, which are higher than the corresponding continuum values. For SBLIs in rar-
efied conditions, at least the primary shock wave over the body can take on similar
characteristics to those noted herein, thus affecting the interaction. Again, attention
is drawn to the rapidity in which the changes in the flow structure occurred: once
within the rarefied regime, they occur within a one-order-of-magnitude increase
in Kn.
Throughout most continuum hypersonic flows, the vibrational temperature is
frozen, whereas the rotational component, which can be changed by relatively fewer
molecular collisions, usually follows closely the variations in the translational value.
However, in rarefied flows, this is not generally the case; in much of the flow the
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 343
(a) (b)
30 30
100 100
Translational n/n Translational n/n
T/T Temperature T/T Temperature
20 20
Rotational
Temperature Rotational
Temperature
50 50
Number 10 10
Density Number
Density
0 0 0 0
-1 -0.5 0 -1.5 -1 -0.5 0
x/R x/R
Figure 8.4. DSMC-computed profiles of number density and temperature along the axis of
a blunt-ended cylinder: (a) Kn = 0.027; (b) Kn = 0.083 with the density and temperature
normalized with respect to the free-stream values.
rotational component also lags behind changes in the translational value as a con-
sequence of the reduced collision frequency. It is most noticeable in shock waves in
which the changes are rapid. This is illustrated clearly in the blunt-ended cylinder
flow. The two plots shown in Fig. 8.4 correspond to the intermediate- and lowest-
density examples in Fig. 8.3; it is shown in both cases that these two temperature
components come into equilibrium only near the surface, where the flow is denser
and not subjected to rapid changes. In the Kn = 0.027 flow, this occurs at about
x = –0.25 but nearer the surface in the more rarefied example. The vibrational
temperature, which requires many more collisions to affect a change, is completely
frozen in both examples except for very close to the surface, and hence not shown.
Also note that the translational temperature rises well upstream of any change in
the density, which is a consequence of a small number of particles being reflected
from the denser flow within the shock layer and penetrating upstream into the low-
density-approaching flow. The high relative velocity between the two sets of parti-
cles leads to intense collisions that elevate the translational temperature. This effect
is also evident in the results shown in Appendix Fig. A.4a for a blunted cone, and
can lead to chemical changes in the gas upstream of the main body of the shock
wave in high-enthalpy flows.
8.4.1 Introduction
The previous section concluded that as Kn increases, rarefied flows experience
structural changes that distinguish them from their continuum counterparts. These
changes eventually lead to the transition to FM flow as density falls. They are due
in part to an increase in the relative importance and extent of viscous phenom-
ena that can be thought of as a low-Reynolds-number effect. However, there are
other phenomena – such as velocity slip, temperature jump, shock-wave merging,
and weakening – that are strictly rarefied-flow effects, which (when they occur) lead
to fundamental physical changes in overall flow structure. Also notice that a degree
344 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
6 The outer radius (R) of the cylinder was 32.5 mm (1.25 inches) and the length (L) before the flare
was 101.7 mm (4.0 inches).
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 345
Figure 8.5. Schematic of the hollow-cylinder–flare body flow under continuum conditions.
(The axis is shown as the dashed line.)
then separates on approaching the corner (if the flare angle is sufficiently large) and
subsequently reattaches on the flare. A secondary shock originates at the separa-
tion point, which then impinges on the bow shock wave, thereby intensifying and
steepening it. This secondary shock compresses the flow between the primary shock
and the separation bubble; for a free-stream Mach number of 12 and a 30-degree
flare, the Mach number is typically reduced to about 4.5 after the second shock. The
size of the recirculation region varies considerably with the Reynolds number, Mach
number, and flare angle; however, similar flow structures nevertheless are produced
unless separation is suppressed. If the ambient density is reduced, velocity slip and
temperature jump occur on the cylinder, which delay separation, thereby reducing
the size of the bubble.
Reattachment occurs on the flare if it is long enough. The resultant flow deflec-
tion produces a succession of compression waves that quickly coalesce into a third
shock wave that impinges on the bow shock wave, causing it to be deflected far-
ther away from the body and significantly strengthened. The form of the resultant
shock-shock interaction depends on the flare angle and flow conditions. For small
angles, two shock waves and a shear layer are formed; however, if the interaction
is more intense, two triple points occur separated by a short planar shock wave. In
either case, a shock wave from the interaction is directed toward the body where
it impinges on the flare-boundary layer. By this time, the fluid near the surface has
passed through four shock waves or sets of compression waves; thus, the density
and temperature will have risen sharply. This process thins the boundary layer on
the flare and creates localized and intense peaks in the pressure and heat-transfer
distributions downstream of reattachment. In flight applications, this intense heating
can seriously endanger the structure of a vehicle.
Figures 8.6 and 8.7 are examples of the computed and measured distributions of
pressure coefficient (Cp ) and Stanton number (St) for the lowest-density conditions
included in the RTO study. For this, Kn based on L equals 0.00087; hence, this
flow is at the dense extreme of the rarefied flow regime. The measured results are
taken from the CUBRC experiment [7] and the DSMC results were computed by
Markelov et al. [10] using code they developed. The two coefficients are plotted
346 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
1.2
Cubdat
0.8 Markelov et al 11c
Markelov et al 11b
Cp 0.6
0.4
0.2
0
0 2 4 6 8 10
distance (ins)
Figure 8.6. Pressure-coefficient (Cp) distribution for the hollow-cylinder–flare body: M =
12.4; Kn = 0.00087; n∞ = 1.197E22; T∞ = 95.6 K; Twall = 93K; nitrogen. Experimental results
are from CUBDAT [7] and the DSMC data are from Markelov et al. [10]. The cone intersec-
tion is at 4.0 inches (101.6 mm).
against the distance s along the external surface of the body. In Fig. 8.6, a small rise
in pressure is shown at about s = 3.5 inches (90 mm: x/L = 0.874), which indicates
separation. In the flow upstream, strong viscous interaction occurs that causes the
rise in pressure near the leading edge. After separation, a small plateau in pressure
is observed; however, once on the flare (i.e., s > 4 inches = 101.6 mm), the pressure
0.06
0.05
Cubdat
Markelov et al 11c
0.04
Markelov et al 11b
St
0.03
0.02
0.01
0.00
0 2 4 6 8 10
distance (ins)
Figure 8.7. The corresponding heat-transfer distribution for the same hollow-cylinder–flare
flow shown in Figure 7.6.
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 347
and heat transfer both rise increasingly rapidly toward the intense peaks located
downstream of reattachment (for reasons explained previously). From these plots,
it is not possible to discern precisely where reattachment occurs. Downstream of
the peaks, the pressure and heat transfer quickly recover toward the laminar values
for a 30-degree cone, thereby confirming that transition to turbulent flow had not
occurred in the experiments.
An impressive correlation between the computed and measured data is evident
especially for the heat transfer. Not shown but equally good results for this flow were
obtained using Bird’s DSMC code [11] by Moss and Bird [12]. This is a complex
flow involving separation, reattachment, and shock-shock interactions and SBLIs,
and the agreement between experiment and computation is a remarkable outcome.
Achieving the correct solution is crucially dependent on accurately predicting the
positions of separation and reattachment, something that has proven historically to
be a struggle to achieve consistently well using CFD.
Great care had to be exercised in performing the experiments for this study
because the absolute levels of pressure and heat flux were very low. Potential errors
due to rarefaction effects were avoided by using flush-mounted pressure transduc-
ers. That being said, it is believed that the two heat-transfer measurements in the
region of the peak of the curve in Fig. 8.7, where there is some discrepancy, are
anomalous due to experimental error.
Figure 8.8 shows the effect on the computed pressure distributions of pro-
gressively increasing Kn, in this case using Bird’s DSMC code. The increase was
achieved by reducing the density while retaining the other flow parameters constant.
Thus, the Reynolds number also drops. The densest case is for Kn = 0.0017, which is
twice the value in the previous example shown in Figs. 8.6 and 8.7. This change has
the effect of moving the separation point downstream, almost to the beginning of the
flare. The increase in Kn thickens the shock waves and there also is an increase in the
level of velocity slip on the cylinder. The intense peak in the pressure distribution
is still present, although the magnitude is reduced slightly. However, the peak is at
348 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
almost the same position as it was for the lower Kn flow, despite the separation bub-
ble being much smaller and with reattachment occurring closer to the corner. The
peak in the pressure is linked to the reflection of the reattachment shock toward the
body from the shock-shock interaction; therefore, it is expected that the peak would
have moved upstream as the size of the separation bubble is reduced. (Even when
separation is completely suppressed, a compression wave is formed at the junction
with the cone that interacts with the primary shock.) However, the increase in Kn
creates a thickening of the shock waves and results in the flow taking on a generally
more viscous and benign character, which has the effect of reducing the sharpness
of the gradients within the flow especially in the primary shock because the density
upstream of it is the lowest in the flow. With increasing Kn, this “smearing” of flow
structures attenuates the peaks and moves them downstream, eventually leading to
a complete merging of the shocks with the viscous flow on the body. Thus, two fac-
tors influence the position of the peak; for the increase of Kn to 0.0017, they roughly
balance one another. With further increases in Kn, the peak moves downstream and
quickly decays. For the most rarefied example shown in the figure, for which Kn is
0.128, there is no discernable peak in the pressure distribution. Instead, the flow
begins to resemble closely the FM pattern for which there would be a constant-
pressure coefficient on the flare equal to 0.650 solely dependent on the inclination
of the surface. What is striking about the results shown in Fig. 8.8 is how quickly the
flow structure changes from the continuum pattern to one resembling FM flow. The
transition occurs principally in the Kn range from 10 − 3 to 10 − 2 .
Figure 8.9 shows how the absolute value of the aerodynamic heating for this
flow varies with increasing rarefaction. The overall levels drop due to the reduction
in density, but it is also clear that the sharpness of the peak is diminished, which
echoes the trend seen in the pressure. As expected, the pressure and heat-transfer
peaks coincide at the same point on the flare for each flow condition. The heat-flux
data are replotted in Fig. 8.10 in terms of the Stanton number; it is shown here that as
the density is reduced – that is, as Kn is increased and the Reynolds number falls –
the overall levels of this coefficient rise. This follows the Reynolds-number trend
observed in laminar-continuum flows. From this plot, we can observe more clearly
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 349
how the peak moves back along the flare as Kn increases from 0.0017 to 0.0169. In
the most rarefied case, for which Kn = 0.128, there is no longer a discernable down-
stream peak produced by the reflected shock wave, which is now very diffuse and
impinging on the flare. Although this flow appears from the pressure distribution to
have taken on the character of a FM flow, it is not as evident in the heat-transfer dis-
tribution; the FM value on the flare would have been constant and equal to 0.28 but
instead it is shown falling away from this value with increasing x. Thus, the computed
heat-transfer distribution still retains some of the continuum character, illustrating
the complex nature of these transitional flows within this Kn range.
to the surface. We propose that close to the surface but for y a little greater than
the local mean-free path λ, the velocity gradient is assumed to be constant such that
du/dy = k, which depends on the shear level in the boundary layer above the region
being considered. We concentrate on the region up to one mean free path from the
surface. The flow within it consists of particles that have a mean velocity plus a
random component due to their thermal motion. Therefore, some move toward the
surface (i.e., the incident stream); others that have collided with the surface move
away from it into the flow (i.e., the reflected stream). Consider now an incident
particle that is moving toward the surface as a consequence of its random thermal
motion. Once it is within a mean-free path of the wall, it does not (on average) expe-
rience a further collision to change the velocity before reaching the surface. Thus,
it carries with it the momentum acquired from the last collision within the gas; it
arrives at the wall with a positive finite velocity in the x direction, the magnitude
of which depends on the values of k and λ. Each particle that reaches the surface
Figure 8.12. A schematic impression of the velocity profile close to the surface in a boundary
layer on a plane wall.
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 351
is returned to the flow after adjusting to the wall conditions. For “practical engi-
neering surfaces,” it is justifiable to assume that these particles are returned fully
accommodated and diffusely re-emitted into the flow. This is equivalent to them
being returned as if they were coming from an imaginary stationary gas beneath the
surface at the wall temperature and pressure.7 Thus, the re-emitted particles have
zero-mean tangential velocity. The flow velocity at the wall is obtained by combin-
ing the properties of the incident and reflected streams, and there is thus a resultant
finite velocity in the x direction at the wall, referred to as the slip velocity. This is
in contravention to the usual no-slip boundary condition assumed in continuum-
fluid mechanics. The region in which this happens is approximately one mean-free-
path thick and is referred to as a Knudsen layer. Velocity slip is accompanied by
a corresponding temperature-jump effect, in which a difference occurs between
the temperature of the wall and that of fluid adjacent to it. The velocity-slip and
temperature-jump values depend on the degree of rarefaction and the level of shear
and temperature gradient in the boundary layer, respectively. They always have the
effect of reducing the shear stress and usually the heat transfer.
These effects are present in all boundary layers but, in most circumstances, the
mean-free path is so small that the scale renders them completely insignificant. They
generally can be ignored and the no-slip condition can be applied legitimately if Kn
is less than 10 − 4 . Above this value, continuum-flow calculations should be adapted
by modifying the wall boundary conditions, which can be accomplished using semi-
empirical expressions to determine the velocity-slip and temperature-jump effects.
This procedure is acceptable for a small Kn; however, as the value increases, the
approximation becomes less reliable and the real influence can be accounted for
only by analyzing the flow using kinetic theory or particle simulation – at least, for
the flow near the surface. Although the changes in the boundary-layer profiles due
to velocity slip and temperature jump will be experienced mostly near a leading
edge or tip of a body where Kn is highest, they can materially affect the develop-
ment of the subsequent boundary-layer flow, thereby influencing SBLIs that occur
downstream.
The velocity-slip effect has been the subject of numerous kinetic-theory stud-
ies that properly consider any particle collisions that occur within Knudsen layer.
It has proved to be a difficult problem to solve; however, a successful result using
a linearized approximation that disregards temperature gradients was provided by
Cercignani [4]. He gives the velocity profile near the surface as follows:
π −1/2 λ y
u(y) = k x + ζ − I (8.3)
2 λ
where ζ is a so-called slip coefficient equal to 1.01615λ and k is (as before) the gradi-
ent of mean velocity in the boundary layer just beyond Knudsen layer. The function
I(y/λ) is plotted by Cercignani and rises close to the surface, but it is virtually zero
outside Knudsen layer. A velocity profile corresponding to this expression is illus-
trated in Fig. 8.12. Although the layer is only about one mean-free-path thick, some
collisions between particles do occur that lead to the curvature of the profile. Other
7 The re-emitted particles therefore have velocities corresponding to an equilibrium distribution for
a stationary gas f0 (c, Twall ), for v > 0 where v is the outward y component of velocity and Twall is
the wall temperature.
352 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
attempts were made to refine the kinetic-theory analysis; however, for the most part,
they failed to provide results that agree well with measured data. Recent numeri-
cal solutions using the DSMC method, however, confirm Cercignani’s results [13].
From the figure, it is evident that the true slip velocity at the surface is not the most
appropriate value to use as the extrapolated boundary condition for Navier-Stokes
calculations of the flow beyond Knudsen layer. A higher value, indicated as u∗ in
the figure, is more suitable. When appropriate, this value can be incorporated into
continuum calculations as the wall boundary condition using an expression such as
the following:
1/2
∗ μ 2k̂T ∂u
u = σp (8.4)
p m ∂ y e
where u∗ is the effective tangential velocity of the gas at the solid surface, μ is the
gas viscosity, p and T are the local pressure and temperature (respectively), m is
the molecular mass of the gas, and k̂ is the Boltzmann constant. The degree of
velocity slip is proportional to the shear stress evaluated at the outer edge (i.e.,
subscript e) of Knudsen layer. The constant σ p is an alternative velocity-slip coef-
ficient first introduced by Maxwell in 1879 [32], but his value was inaccurate. For
practical application of this theory, the extensive reviews by Sharipov et al. [14,15]
can be consulted for appropriate suggestions for the coefficients covering various
situations including gas mixtures. Corresponding expressions for the temperature
jump also are discussed by these authors.
The velocity profile in Fig. 8.12 is consistent with the viscosity, effectively
decreasing in Knudsen layer as the wall is approached because there is no mecha-
nism for the shear stress to vary greatly within the short distance. A DSMC solution
by Torczynski et al. [16] confirmed this and demonstrated that the effective viscos-
ity and thermal conductivity fall sharply by as much as 50 percent very close to the
wall when compared with the Chapman-Enskog [33] (i.e., the continuum) values.
This again underlines the care with which the Navier-Stokes methods should be
used for flows exhibiting rarefaction effects because these changes in the transport
coefficients cannot be incorporated in any way.
The kinetic-theory solution from which equation 8.3 is derived assumes that
the gas just outside Knudsen layer is in local molecular equilibrium. This normally
would be a valid assumption in continuum flows; however, for low-density hyper-
sonic situations, some degree of nonequilibrium is likely even in the lower strata of
the boundary layer. This can occur in situations where it otherwise would be legiti-
mate to use Navier-Stokes methods with slip-boundary conditions. For example, it
could arise due to reactions taking place in the flow. In developing the velocity-slip
theory, it also is normally assumed that the gas re-emitted from the surface is in
full equilibrium at the wall temperature. This may be an unsound assumption for
two reasons. First, the accommodation of the internal-energy modes on surfaces is
acknowledged to be frequently incomplete; for example, an accommodation coeffi-
cient of only 0.35 was predicted [17] for the rotational mode for nitrogen in certain
circumstances (e.g., a value of 1.0 indicates full accommodation). Furthermore, it is
conjectured that the vibrational mode is even less likely to be fully accommodated to
wall conditions. Second, surfaces are important in chemically reacting flows: If they
are not fully catalytic, the re-emitted gas will not be in full chemical equilibrium.
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 353
Figure 8.13. Schematic of the flow over the biconic body showing flow topology.
The degree to which these factors influence the velocity-slip and temperature-jump
phenomena has not been investigated and cannot be quantified easily.
Kn = 0.0024
y/L
0.8 1 1.2
x/L
thereby strengthening and deflecting it away from the body. The path of the sep-
aration shock can be inferred from the kinks in the streamline and the change
in slope of the primary shock at about x/L = 0.95. A complex interaction, simi-
lar to that observed on the hollow-cylinder–flare body, occurs near reattachment;
however, in comparison, a more intense flow is established here due to the greater
flow deflections imposed by this body. From the shape of the streamline defining
the edge of the separation bubble, it is apparent that immediately upstream of reat-
tachment point (R), the shear layer is concave. This produces a set of focusing com-
pression waves that radiate away from the surface and interact with the bow shock
wave to produce the shock-shock interaction. These waves usually (but not always)
focus to form a shock wave before interacting with the primary shock. The inter-
action is similar to those categorized by Edney [19] for flows in which an oblique
shock wave impinges on the bow shock wave ahead of a blunt leading edge. Flows
analogous to his Type III and, more important, Type IV interactions are possible
on the biconic body. The structure of an Edney interaction for a similar degree of
rarefaction is illustrated in Fig. 8.15a, in which DSMC results for an oblique shock
impinging on a circular body obtained by Moss et al. [20] are shown. Mach-number
contours are plotted from which the shock-wave pattern can be identified clearly.
Two triple points (TP) are produced by the interacting shock waves between which
a near-planar oblique shock wave forms. Downstream, a supersonic jet forms that
directs fluid toward the surface and creates intense localized peaks of heat transfer
and pressure on the leading edge. Examples of this occurring in hypersonic flows
are discussed in Chapter 7. The effect of reducing the density on a Edney type
IV interaction similar to that computed by Moss et al. is illustrated in Fig. 8.15b.
As in previously considered examples, increasing the degree of rarefaction reduces
the intensity of the interaction. The thickening of the shock waves smears the flow
structure and the two triple points that create the high-velocity jet in Fig. 8.5a can-
not be identified. The intense peaks in heat transfer and pressure on the surface
correspondingly are greatly attenuated. The biconic body does not exhibit exactly
the same flow structure as the Edney interaction; nevertheless, a very-high-velocity
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 355
(b)
jet is produced that is clearly discerned in Fig. 8.14 as the dark strip running close to
the rear-cone surface. The velocity within this jet reaches a value of about 75 per-
cent of the incident-flow value. Instead of being directed toward the surface as in the
Edney interaction, this jet travels considerable distance parallel to the cone. If we
trace back the streamlines bounding this jet, it is evident that the fluid within it pre-
viously passed through several oblique shock waves: the primary shock, the separa-
tion shock, and then the shock or compression waves emanating from the reattach-
ment region. As in a multishock supersonic intake, high pressures can be achieved
this way because the process tends toward an isentropic compression; indeed,
Fig. 8.16, which shows the pressure contours, confirms this. A pressure peak is gener-
ated immediately downstream of the shock-shock interaction that is more than 650
times the free-stream value. The primary shock immediately upstream of this region
356 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
0.8
n = 0.378E22
Kn = 0.0024
CUBDAT #7
R p/p ∞
corner of the biconic body. CUBDAT
650
test case No. 7, Kn = 0.0024. M = 10,
550
450 Twall = 293 K; T∞ = 94 K; nitrogen.
350
250
150
50
0.4 S
S
0.8 1 1.2
x/L
is strongly deflected and it can be seen that it becomes essentially a normal shock
downstream of which the flow is subsonic. The extent of the low-velocity region can
be gauged from Fig. 8.14. The strong compression combined with the high shear
produced by the jet causes a thinning of the boundary layer on the aft cone and, for
this Kn, very sharp and intense peaks in the pressure and heat-transfer distributions
downstream of reattachment (Figs. 8.17 and 8.18). Disturbances from the interac-
tion create oscillations in the pressure distributions along the cone after the first
peak, which – to a lesser extent – also are evident in the heat transfer. The curves
shown in this example correspond to Case No. 7 in the CUBDAT dataset and,
although the experimental points are not shown, there is good agreement between
the two.
Computations also were made for this body for three higher Kn at the same
Mach number and flow temperature. With increasing rarefaction, separation is
3
Cp Kn = 0.0024
CUBDAT #7
0.0095
0.029
2
Figure 8.17. The pressure distribution on a
sharp biconic body for varying degrees of
rarefaction. M = 10, Twall = 293 K; T∞ =
94 K; nitrogen.
1 0.090
0
0 0.5 1 1.5
x/L
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 357
0.4
St
Kn = 0.090
0.3
0.1 0.0095
0.0024
CUBDAT #7
0
0 0.5 1 1.5
x/L
delayed due to the reduction in Reynolds number and the influence of increased
velocity slip on the fore cone (although the latter is less significant than on the
hollow-cylinder–flare body because the density levels on the surface are higher).
Examination of the streamline patterns confirms that up to Kn = 0.029, separation
still occurs, but then it is almost at the cone intersection; by Kn = 0.0899, it is sup-
pressed. Also, with increasing Kn, the intensity of the SBLIs diminishes and, as in
the hollow-cylinder–flare flow, thickening of the shock waves, merging, velocity slip,
and the increased viscous nature of the flow all contribute to this process. Peaks in
the surface pressure and absolute heat-transfer distributions decay with rising Kn
and pressure oscillations are suppressed.
Figure 8.19 shows the velocity contours and streamlines near the corner for the
Kn = 0.029 flow. The interaction is more dispersed than in the previous example and
the primary shock wave, far from being a discontinuity, shows a structure similar
Kn = 0.029
y/L
0.75
Figure 8.19. A DSMC solution for the
flow near the corner for Kn = 0.029 V/V inf
showing streamlines and velocity. 0.9
0.7
0.5
0.5
0.3
0.1
0.25
0.5 0.75 1 1.25
x/L
358 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
Kn = 0.029
y/L
1.
p/p∞
0.75
350
300
250
200
0.5 150
100
50
10
0
0.25
0.5 0.75 1 1.25
x/L
Figure 8.20. Contour of normalized pressure in the region of the corner of the biconic body.
Kn = 0.029; M = 10, Twall = 293 K; T∞ = 94 K; nitrogen.
to the compression wave seen ahead of the blunt-ended cylinder at similar values
of Kn (see Figs. 8.3 and 8.4). Elsewhere, a general smearing of the flow structure
occurred, which leads to an attenuation of the peaks in the surface fluxes and the
intense shock-shock interaction does not occur. An extra, darker band of contours
was added, centered at approximately 30 percent of the free-stream velocity. This
indicates where the boundary layer is; from its position above the surface of the
aft cone, it can be inferred that the high-velocity jet generated by the shock-shock
interaction process, observed in Fig. 8.14 for the lower Kn, is no longer present.
Consequently, the peak in the surface pressure that occurs downstream of reattach-
ment is attenuated because the flow structure is more benign. Figure 8.17 shows that
the maximum value drops by a factor of almost 2 when Kn changes from 0.0024 to
0.0899. The effect of increasing Kn on the heat transfer is shown in Fig. 8.18. As in
the hollow-cylinder–flare body, overall levels of the coefficients increase due to the
falling Reynolds number, but the peaks in the actual heat flux that occur on the aft
cone are weakened and broaden with increasing rarefaction (for reasons previously
explained). However, because the pressure levels are higher on the cone than in the
corresponding cylinder–flare flows, even for the most rarefied case (i.e., Kn = 0.09),
the profiles do not approximate to the constant FM values on either cone.
Figure 8.20 shows contours of p/p∞ for Kn = 0.029. These contours can be
compared directly to Fig. 8.14, in which Kn is one tenth the value. Again, it is
apparent that all of the intense compression regions were eliminated. The maximum
p/p∞ levels in this flow are about half of those in the denser example. Neverthe-
less, the compression of the flow at the corner is still strong enough to require the
primary shock to be deflected into a near-normal wave upstream of the reattach-
ment region.
Figure 8.21, the final figure in the sequence, indicates the levels of molecular
nonequilibrium that can be expected in this type of flow. Because this flow is pure
8.4 Examples of SBLIs in Rarefied Hypersonic Flows 359
10 Kn= 0.029
wall
0.6 0.8 1
x/L
nitrogen and the vibrational mode is frozen, non-equilibrium can only be manifested
as a difference between the translational and rotational temperatures. Values of
these are plotted along the line at y/L = 0.62, which is a little farther outboard
from the axis of the body than the point where the reattachment shock intersects
the primary shock. The line goes through the low-velocity patch shown in Fig. 8.19.
Curves are shown for Kn equal to 0.0024 and 0.029. The contrast is obvious: For the
denser flow, the sharp rise at x/L = 1 indicates the compact nature of the shock wave
in which the rotational temperature lags minimally behind the translational value.
The flow comes quickly into equilibrium after this and remains so up to the surface.
With only one order-of-magnitude increase in Kn, the shock changes completely in
character and becomes very diffuse. The curve now closely resembles profiles shown
for the blunt-ended cylinder in Fig. 8.4b. Throughout, the rotational temperature
lags far below the translational value and only comes into equilibrium in the denser
flow close to the surface.
1
y/L
7.5
0.8
Figure 8.22. Contours of NO fraction (as %)
5 downstream of the cone intersection at Kn of
0.0015. Stagnation enthalpy: 10 MJ/kg. The solid
line depicts the cone surface.
0.6 2.5
0.4
0.8 1 1.2 1.4
x/L
y/L
7.5
0.8
Figure 8.23. Corresponding contours of NO frac-
5
tion (as %) for Kn of 0.048 for the same stagna-
tion enthalpy.
0.6 2.5
0.4
0.8 1 1.2 1.4
x/L
would have occurred in the absence of reactions. However, as the fluid approaches
the surface, the temperature falls and reverse reactions occur that return the heat
to the flow or the surface. The latter is fully catalytic and complete recombination
of nitrogen and oxygen would occur at the wall. However, it appears that virtually
all of this recombination occurred before the fluid reaches the body.8 Figure 8.22, in
which Kn = 0.0015, shows that the NO generated almost immediately downstream
of the shock-shock interaction persists in the flow between the outer shock and the
surface of the cone above the jet. In this flow, dissociation of NO is relatively slow
and the concentrations remain roughly the same along the length of the rear cone.
Figure 8.23 shows the effect on the NO concentration of lowering the density
while maintaining the same stagnation enthalpy. Kn for this flow is 0.048. The same
contour levels were used as those in the previous figure, and it is immediately evi-
dent from the lower concentrations that the formation of NO is both delayed and
attenuated. There is now a physical separation between the shock-shock interaction
(which is more spread out and in which dissociation is initiated) and the place where
the reactants first appear in the flow. This is a direct consequence of the reduced col-
lision frequency.
The effects of changing the stagnation enthalpy on the surface-flow properties
for the reacting flow are illustrated in Figs. 8.24 through 8.28. First, we consider the
distributions of pressure and heat transfer near the reattachment region for a flow
at Kn = 0.0015. This is close to the continuum limit, although some influence of
rarefaction can be expected. Figure 8.24 shows the pressure coefficient for three dif-
ferent enthalpies; as expected, increasing the energy of the incident flow raises the
peak-pressure level. It also delays separation, reducing the extent of the recircula-
tion region; consequently, the peak moves closer to the junction between the two
cones at x/L = 1. A result for a nonreacting nitrogen flow (i.e., H0 = 2.1 MJ/kg) also
is shown, and it is evident from comparisons with the similar 3-MJ/kg air flow that
changing from inert nitrogen to a reacting gas results in changes in the flow struc-
ture. Most notably, the size of the recirculating region increases with the separation
point moving forward, causing the peak in Cp to move rearward.
8 Results obtained with surface recombination turned off show virtually identical heat-transfer distri-
bution in the vicinity of the peak, confirming that this occurs before the fluid reaches the surface.
362 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
5
Stagnation Enthalpy
CP 15
10
4 3
Nitrogen
3
Figure 8.24. Pressure distribution in the
reattachment region for Kn = 0.0015 in
air.
2
0
0.8 1 1.2 1.4
x/L
3 Stagnation Enthalpy
15
CP 10
3
2
Figure 8.25. Pressure distribution in the
reattachment region for Kn = 0.015.
1
0
0.8 1 1.2 1.4
x/L
8.5 Concluding Remarks 363
0.75
Stagnation Enthalpy
St
15
10
3
0.5
Nitrogen
0
0.8 1 1.2 1.4
x/L
the stagnation-enthalpy level; however, for the heat transfer, there is an approxi-
mate 35 percent increase in the peak value when increasing the enthalpy from 3 to
15 MJ/kg. Recall that the pressure was almost completely unaffected by this change.
The direct effects of chemical activity now are largely displaced downstream from
the SBLI region due to the reduced collision frequency and the more benign shock-
shock interaction.
The influence of rarefaction on the heat transfer for the reacting flow is summa-
rized in Fig. 8.28, in which data are replotted covering two orders-of-magnitude vari-
ation of Kn. Increasing this is equivalent to reducing the Reynolds number; hence,
the overall levels of the Stanton number rise – a trend that is most obvious on the
fore cone (x/L < 1). Results for Kn ≈ 0.14 are added, which is a condition approach-
ing FM flow. The pressure for this case was not shown and is almost constant on each
cone and very close to the FM values, which are determined simply by the surface
inclination. However, this is not the case for the heat transfer and there is still a
noticeable peak near the corner with a value in excess of the FM on the aft cone,
echoing the trend in the nonreacting flow (see Fig. 8.18).
St 15
10
3
Figure 8.27. Heat-transfer distribution 0.2
in the reattachment region for Kn =
0.015.
0
0.8 1 1.2 1.4
x/L
364 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
0.75
15 MJ Kn
0.0015
St 0.015
0.15
3 MJ
0.5 0.0015
0.016
0.14 Figure 8.28. Heat-transfer distribution
in the reattachment region for a range
of Knudsen numbers for the biconic
body in air.
0.25
0
0.8 1 1.2 1.4
x/L
where this requirement is not met due to the reduction in collision frequency. In
these circumstances, the use of these coefficients is unsound and explains one rea-
son why the Navier-Stokes equations are unable to predict satisfactorily rarefied
flows.
Figure A.1. DSMC computed (left) and measured (right) density contours for a 70◦ spheri-
cally blunted at zero angle of attack. M = 20; Kn = 0.0045; nitrogen.
cone – typical of a planetary-probe reentry capsule [24] – and they show calculated
density contours compared with measurements taken in the ONERA R5Ch wind
tunnel using the nonintrusive electron-beam fluorescence technique. The same con-
tour intervals are used in both images but, for clarity, values above seven times the
free-stream value were omitted in plotting the measured results.
These results reveal the structure of the flowfield, thereby providing a more sen-
sitive assessment of the code’s ability to capture the detailed physics of the flow than
would be possible, for example, from surface-pressure or heat-transfer information.
The agreement between the two is impressive and it is clear that most of the impor-
tant features of the flow were captured precisely. Although not initially appreciated,
it soon became evident that Bird’s DSMC method was potentially extremely versa-
tile and, in time, successful ways were devised to adapt it to include dissociation,
chemical reactions, ionization, and high degrees of molecular nonequilibrium. Sev-
eral sophisticated codes currently exist that are capable of addressing multispecies
reacting flows and geometrically complex shapes. Only an outline of the DSMC
method is presented here because a full description is provided in [21], in which
Bird comprehensively and practically explains his own code and the underlying the-
ory. Further explanation, more within the context of kinetic theory, is in the work
of Cercignani [4].
DSMC computations advance in real time and track the progress of an array
of simulator particles using time-steps t that must be smaller than the local mean-
collision time. The simulators have three components of velocity even if the flow
that is being computed is not three-dimensional. A defining and innovative feature
of Bird’s scheme is that he decouples the movement of the particles from their colli-
sions by alternately simulating the two processes, depicted schematically in Fig. A.2.
The Boltzmann equation [3] for a dilute gas is expressed as follows:
∂f 1
+ v · ∇x f = Q( f ) (8.A.4)
∂t ε
where v is the velocity and Q is the collision operator that is a complex function
to determine how f changes as a consequence of intermolecular collisions. Over
the small time increment [nt, (n + 1)t], first the left-hand side of the equation
368 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
Surface collision
Surface
(a) (b)
Figure A.2. A schematic representation of the two steps in a two-dimensional DSMC com-
putation: (a) the move phase, and (b) the collision phase.
(i.e., ∂ f/∂t + n∇x f = 0) is replicated by allowing the free motion of all of the simu-
lators using the following approximation:
xi ((n + 1)t) = xi (nt) + vi (nt)tvi vi ((n + 1)t) = vi (nt). (8.A.5)
This is depicted in Fig. A.2(a). If a simulator encounters a solid surface during
this phase, it is returned to the flow with new velocity components that are deter-
mined using an assumed surface-scattering kernel. During this phase, new simulator
particles are introduced at inflow boundaries that are selected randomly from distri-
butions commensurate with the state and velocity of the fluid entering the domain.
The number entering at each time-step is set by the ratio chosen between the density
of the simulators and that of the real flow.
In the next time-step (see Fig. A.2(b)), the right-hand side of the Boltzmann
equation, ∂ f/∂t = Q( f ), is simulated. An appropriate number of “collision part-
ners” are selected from within a small localized partition x of the spatial domain.
These “cells” must be smaller than the local mean-free path. If the selection of part-
ners within a cell is random, Q( f ) is effectively replaced by the spatially smoothed
operator Qx ( f ). Alternatively, nearest-neighbor collision pairs can be identified
to improve on this smoothing. Efficient numerical schemes were devised to imple-
ment this method and they have provided significant improvements in the accu-
racy of the results, especially for denser flows. After each simulated collision, new
properties of the two particles are determined using an appropriate representa-
tion of the interaction. Only ten to twenty simulators are required in each cell at
any time. The alternate steps are repeated many times and the flow properties are
determined from information collected about the population of simulators as they
pass through small sampling regions of the flow. Bird’s time-step–splitting procedure
was an inspired move that greatly increased the numerical efficiency of the DSMC
method over alternative particle-simulation methods, such as molecular-dynamics
computations.10 Although Fig. A.2 depicts a two-dimensional flow, the method can
be extended readily to three dimensions.
The computations advance in real time from an impulsive start; for problems
for which the boundary conditions are such that a steady-state solution exists, an
10 The DSMC method is N dependent, whereas the molecular-dynamics methods are N2 , where N is
the number of particles in the simulation.
Appendix A: Kinetic Theory and the DSMC Method 369
T(rot)/T(inf)
Figure A.3. The rotational temperature along
the stagnation stream line of a blunt-ended
0.5
cylinder. M = 20.6; T∞ = 52K; Kn = 0.02;
nitrogen. DSMC solution; experimental
results.
0
1 0.5 0
x/R
initial transient settling period must be allowed before sampling begins. Typically,
well-resolved computations are performed using on the order of 107 simulators.
Hence, the particle-collision routines must be called many times during a compu-
tation, which limits the level of sophistication that can be tolerated. Usually, phe-
nomenological models are used in which the postcollision properties are determined
by randomly sampling from appropriate predetermined distributions. The number
of collisions allowed during each time-step can be determined in different ways, but
the no-time-counter method is commonly used [21].
The behavior of molecules during the simulated collisions depends on the
choice of the intermolecular-force potential. Models used range in complexity from
simple hard spheres and inverse-power representations to the Morse and Lennard-
Jones potentials [25], which include the realistic long-range attraction between
molecules as well as the short-range repulsion. It is accepted that the latter two
models impose too high a computational overhead for their use to be justifiable in
an engineering context. Simpler alternatives that proved to be very good yet remain
computationally efficient were devised; the most commonly used is the Koura and
Matsumoto [26] Variable Soft Sphere model. For polyatomic particles, the transfer
of energy to and from the internal modes also must be considered. The most popular
is a stochastic model devised by Borgnakke and Larsen [27]. It was proven success-
ful in describing the macroscopic distribution of energy in the gas even though the
treatment of individual collisions is not rigorous. Uncertainties about the real dis-
tributions that prevail led them to adopt an implicit local equilibrium assumption,
insofar as the postcollision properties are sampled from a varying but equilibrium
distribution for each collision. Empirical evidence indicates that this model works
well for nonreacting flows;11 however, the validity of using it for situations in which
chemical reactions occur is not obvious because the local equilibrium assumption
for the products is not always appropriate. This has not been properly tested.
An example of results showing the success of the Borgnakke and Larsen model
for a low-density nonreacting flow is shown in Fig. A.3. The body is a blunt-ended
circular cylinder with the axis aligned to the incident stream, which in this case is
11 This is probably because in these interactions, all the available energy states tend to be populated
without any specific preference. This leads to rapid energy equilibration within just a few collisions.
However, this is not generally the case for collisions in which chemical reactions occur.
370 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
(a) (b)
Figure A.4. (a) DSMC [30] and Navier-Stokes [31] results showing the temperature compo-
nents along the stagnation streamline for a spherically blunted 70◦ cone in real air; (b) Species
concentrations along the stagnation streamline. (Note that the flow is from right to left.)
a Mach 25 nitrogen flow. Results for this flow also were cited in Figs. 8.3 and 8.4.
Figure A.3 shows the rotational temperature profile along the stagnation line. The
computed result is compared with experimental data obtained using the electron-
beam fluorescence technique. From the close agreement, it is clear that the transfer
of energy between the translational and rotational internal mode was reproduced
with good precision using this model.
The requirement for all of the computational cells to be smaller than the local
mean-free path while also retaining approximately the same number of particles
per cell limits the density for which DSMC computations can be made with the
computer resources currently available. Nevertheless, it is possible to predict flows
that overlap the range in which the Navier-Stokes methods also are valid. This has
allowed informative validation exercises to be undertaken, in which results from the
two methods were compared with one another and with experimental data. In one
study [28] that included flows involving SBLIs, it was concluded that the DSMC
method is capable of accurately predicting flows in the low-density end of the con-
tinuum regime.
A feature that sets the DSMC technique apart from the continuum-CFD meth-
ods is the level of sophistication that can be incorporated in modeling dissociation,
chemical reactions, and ionization. In a complex way, these processes are depen-
dent on energy components of individual particles involved in each collision. The
continuum codes generally use global Arrhenius equilibrium-reaction rates com-
bined with a simplistic dependence on the translational and vibrational temper-
atures (e.g., Park’s model [29]) to compute changes due to chemical processes.
Particle-simulation codes offer the possibility of using more realistic modeling of
the reactions, including those occurring on the surface by considering the energy
states of individual particles before each collision. This is especially significant for
flows in which there is a substantial level of molecular nonequilibrium in the gas.
Sample results [30] that demonstrate the use of this method when reactions
occur are presented in Fig. A.4. The gas is air and the full spectrum of possible
References 371
REFERENCES
[1] G. A. Bird. Molecular Gas Dynamics and the Direct Simulation of Gas Flows (Oxford:
Clarendon Press, 1994).
[2] W. G. Vincenti and C. H. Kruger. Introduction to Physical Gas Dynamics (New York:
Wiley, 1967).
[3] S. Chapman and T. G. Cowling. The Mathematical Theory of Non-Uniform Gases, 3rd
edition (London: Cambridge University Press, 1970).
[4] C. Cercignani. Rarefied Gas Dynamics (London: Cambridge University Press, 2000).
[5] V. V. Mikhailov, V. YA. Neiland, and V. V. Sychev. The theory of viscous hypersonic
flow. Annual Review in Fluid Mechanics, 3 (1971), 371–96.
[6] M. S. Holden and T. P. Wadhams. Code validation studies of laminar shock–boundary
layer and shock-shock interaction in hypersonic flow. Part A: Experimental measure-
ments. AIAA Paper 2001-1031 (2001).
[7] M. S Holden and T. P. Wadhams. A database of aerothermal measurements in hyper-
sonic flows in “building-block” experiments for CFD validation. AIAA Paper 2003-1137
(2003).
[8] B. Chanetz, R. Benay, J.-M. Bousquet, R. Bur, T. Pot, F. Grasso, and J. Moss. Exper-
imental and numerical study of the laminar separation in hypersonic flow. Aerospace
Science and Technology, 2 (1998), 3, 205–18.
[9] B. Chanetz and M.-C. Coet. Shock wave–boundary layer interaction analyzed in the
R5Ch wind tunnel. Aerospace Research, 5 (1993), 43–56.
[10] G. N. Markelov, M. S. Ivanov, S. F. Gimelshein, and D. A. Levin. Statistical simulation
of near-continuum flows with separation. Rarefied Gas Dynamics: 23rd International
Symposium, AIP Conference Proceedings, 663 (2003), 457–64.
[11] Available online at www.aeromech.usyd.edu.au/dsmc gab/.
[12] J. N. Moss and G. A. Bird. DSMC simulation of hypersonic flow with shock interactions
and validation with experiments. AIAA Paper 2004–2585 (2004).
[13] C. Cercignani. “Knudsen layer theory and experiment.” In Recent Developments in The-
oretical and Experimental Fluid Mechanics, eds. U. Müller, K.G. Rösner, and B. Schmidt
(Berlin: Springer-Verlag, 1979), pp. 187–95.
[14] F. Sharipov and V. Seleznev. Data on internal rarefied gas flows. Journal of Physical
Chemistry Ref. Data, 27 (1998), 657.
372 Shock Wave–Boundary-Layer Interactions Occurring in Hypersonic Flows
[15] F. Sharipov and D. Kalempa. Velocity slip and temperature jump coefficients for
gaseous mixtures. I: Velocity slip coefficients. Physics Fluids, 15 (2003), 6, 1800–6; see
also Velocity slip and temperature jump coefficients for gaseous mixtures. II: Thermal
slip coefficients. Physics of Fluids 16 (2004), 3, 759–64.
[16] J. R. Torczynski, M. A. Gallis, and D. J. Rader. DSMC simulations of Fourier and Cou-
ette flow: Chapman–Enskog behaviour and departure therefrom. Rarefied Gas Dynam-
ics (AIP Conference Proceedings #762, 2005), pp. 620–5.
[17] H. Takeuchi, K. Yamamoto, and T. Hyakutake. Behaviour of reflected molecules of a
diatomic gas at a solid surface. Rarefied Gas Dynamics (AIP Conference Proceedings
#762, 2005), pp. 987–92.
[18] M. S. Holden, T. P. Wadhams, J. K. Harvey, and G. V. Candler. Comparisons between
measurements in regions of laminar shock wave boundary layer interaction in hyper-
sonic flows with Navier-Stokes and DSMC solutions. Technologies for Propelled Hyper-
sonic Flight, RTO-TR-AVT-007-V3, pp. 4-1–56 (Research and Technology Organisation
[NATO], 2006).
[19] B. Edney. Anomalous heat transfer and pressure distributions on blunt bodies at hyper-
sonic speeds in the presence of an impinging shock. Aeronautical Research Institute of
Sweden, FFA Report 115, Stockholm (1968).
[20] J. N. Moss, T. Pot, B. Chanetz, and M. Lefebvre. DSMC simulation of shock-shock
interactions: Emphasis on Type IV interactions. Proceedings of the 22nd International
Symposium on Shock Waves, London (1999).
[21] G. A. Bird. Direct simulation of the Boltzmann equation. Physics of Fluids, 13 (1970),
11, 2676–82.
[22] W. Wagner. Monte Carlo methods and numerical solutions. Rarefied Gas Dynamics,
ed. M Capitelli (AIP Conference Proceedings, 762 2005), pp. 459–66.
[23] J. K. Harvey and M. A. Gallis. Review of code validation studies in high-speed low-
density flows. Journal of Spacecraft and Rockets, 37 (2000), 1, 8–20.
[24] F. Coron, J. K. Harvey, and H. Legge. “Synthesis of the rarefied flow test cases.” In
Hypersonic Flows for Reentry Problems Part III, eds. R. Abgrall, J-A. Desideri, R.
Glowinski, M. Mallet, and J. Periaux (Berlin: Springer Verlag, 1992).
[25] M. N. Macrossan. Diatomic collision models used in the direct simulation Monte Carlo
method applied to rarefied hypersonic flows. Ph.D. Thesis, University of London (1983).
[26] K. Koura and H. Matsumoto. Variable soft sphere molecular model for inverse-power-
law or Lennard Jones Potential. Physics of Fluids A, 3 (1991), 10, 2459–65.
[27] C. Borgnakke and P. S. Larsen. Statistical collision model for Monte Carlo simulation
of polyatomic gas mixtures. Journal of Computational Physics, 18 (1975), 405–20.
[28] J. K. Harvey, M. S. Holden, and T. P. Wadhams. Code validation studies of laminar
shock–boundary layer and shock-shock interaction in hypersonic flow. Part B: Compar-
ison with Navier-Stokes and DSMC solutions. AIAA Paper 2001–1031, (2001).
[29] C. Park. Nonequilibrium Hypersonic Aerothermodynamics (New York: Wiley, 1990),
p. 114.
[30] M. A. Gallis and J. K. Harvey. The modelling of chemical reactions and thermochemi-
cal non-equilibrium in particle simulation computations. Physics of Fluids, 10 (1998), 6,
1344–58.
[31] R. B. Greendyke, P. A. Gnoffo, and R. W. Lawrence. Calculated electron number den-
sity profiles for the aero-assisted flight experiment. Journal of Spacecraft and Rockets,
29 (1992), 621.
[32] J. C. Maxwell. On stresses arising in rarefied gases from inequalities in temperature.
Phil. Trans. Roy. Soc. 170 (1879), 231–56.
[33] S. Chapman and T. G. Cowling. The Mathematical Theory of Non-Uniform Gases
(London: Cambridge University Press, 1940).
9 Shock-Wave Unsteadiness in Turbulent
Shock Boundary-Layer Interactions
P. Dupont, J. F. Debiève, and J. P. Dussauge
9.1 Introduction
If the shock wave associated with a shock wave–boundary-layer interaction (SBLI)
is intense enough to cause separation, flow unsteadiness appears to be the almost-
inevitable outcome.1 This often leads to strong flow oscillations that are experienced
far downstream of the interaction and can be so severe in some instances as to inflict
damage on an airframe or an engine. This is generally referred to as “breathing” or,
simply, “unsteadiness” because it involves very low frequencies, typically at least
two orders of magnitude below the energetic eddies in the incoming boundary layer.
The existence of these oscillations raises two questions: “What is their cause?” and
“Is there a general way in which they can be understood?”
There are several distinct types of SBLIs, depending on the geometry and
whether the flow separates, and it is possible that these create fundamentally dif-
ferent types of unsteadiness. An interpretation was proposed by Dussauge [1] and
Dussauge and Piponniau [2] using the diagram reproduced in Fig. 9.1. The organiza-
tion of the diagram requires comment: In the upper branch, unseparated flows are
depicted; those that separate are restricted to the lower branch. In both cases, the
shock wave divides the flow into two half spaces: the upstream and the downstream
layers. Hence, the shock wave can be considered an interface between the two con-
ditions and its position and motion vary accordingly. With these various elements in
mind, the shock motion can be analyzed from the perspective of the upstream and
downstream conditions. The discussion in this chapter is a commentary about flow
organization and other phenomena related to the two branches of the diagram.
1 This observation does not include laminar flows, and instances of very high Reynolds-number
hypersonic ramp-induced separated flows in which no associated unsteadiness is detectable have
been reported.
373
374 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
Distorted
turbulence
Shock wave
Incoming Downstream Figure 9.1. A diagrammatic representa-
turbulence boundary tion of SBLIs.
layer
Separated
zone
creates new initial conditions for the downstream flow and, as the perturbed bound-
ary layer relaxes, it evolves into a new equilibrium state. It is necessary to consider
the distortion of the boundary layer as it passes through the shock if separation
occurs because this is known to be sensitive to the turbulence level. For both sep-
arated and unseparated flows, the change in characteristics of the upstream turbu-
lence can be described adequately by local theories (i.e., those that consider only
what happens at a single point).
In general, shock waves are not unstable. Nevertheless, they may have a par-
ticular frequency response that depends on the shape of the shock and the flow
around it (Culick and Rogers [3]; Robinet and Casalis [4]). In particular cases (e.g.,
the transonic experiments of Sajben and Kroutil [5]), shock waves were shown to
be frequency-selective; however, in general, the transfer function is not known and
depends on the flow. If shock waves are in any way frequency-selective, the appar-
ent overall trend is that they behave rather like low-pass filters; therefore, they are
expected to be more sensitive to low frequencies and to reject the higher ones.
This behavior is shown in experiments in which, away from the boundaries, the
shocks seem not to move despite the foot being imbedded in a turbulent boundary
layer. Similarly, in numerical simulations of shocks in which the foot is subjected
to unsteadiness by being located within a turbulent boundary layer, perturbations
are not found in the outer flow. This is because they are damped as they propagate
outward along the oblique shock in the direction of the tangential velocity (see, e.g.,
Garnier [6] and Wu and Martin [7]).
The action of shocks on the mean flow and turbulent field is summarized as fol-
lows: For mean flow, the overall effect of the shock is to alter the average velocity
distribution of the incoming boundary layer so as to produce profiles more linear or
less energetic near the wall. Turbulence subjected to a shock wave is distorted and
generally amplified in passing through it; the anisotropy is modified in this process.
If the distortion occurs fast enough, the nonlinear effects can be neglected and the
evolution of the turbulence can be described by a (linear) theory of rapid distor-
tion. There are several versions of such theory and a common feature is that they
all consider a linearized set of equations. The differences result from the degree
of complexity of the formulations. The simpler formulations assume that the shock
is steady but, even so, they enable the level of the downstream turbulence to be
determined with considerable precision (see, e.g., Ribner and Tucker [8]; Debiève,
Gouin, and Gaviglio [9]; Jacquin, Cambon, and Blin [10]). Other approaches
(e.g., Ribner [11]) account for adaptation of the shock geometry to the incoming
turbulence and, hence, provide an idea of the shock motion. Such theories are
9.2 The Upper Branch: Unseparated Flows 375
9.3.1 Introduction
This section considers situations in which the shock wave is strong enough to cause
the boundary layer to separate. The diagram in Fig. 9.1 shows that the lower branch
is constructed with double-ended arrows, which indicate that the downstream flow
potentially can control the motion of the shock. Therefore, couplings between the
downstream layer and the shock wave may exist.
0.4
0.35
0.3
0.25
arbitrary scale
Figure 9.3. Spectra at the foot of the
unsteady shock and in the external flow;
oblique-shock reflection. Solid line: wall 0.2
pressure, flow deviation 9.5◦ ; dashed
line: wall pressure, flow deviation 8◦ ;
0.15
dots: hot-wire anemometer, external
flow, flow deviation 5.5◦ .
0.1
0.05
0
1 2 3 4
10 10 10 10
f (Hz)
airfoil. This frequency is compared with other characteristic scales observed with
SBLIs in Section 9.4.
0.12
0.1
0.08
SL
0.06
0.04
0.02
0
0 1 2 3 4 5 6
M
Figure 9.4. Dimensionless frequency (SL ) of the shock oscillation in various configuration:
() subsonic separation, (Kiya & Sasaki [27]; () compression ramp cases; (∗) IUSTI reflec-
tion cases; ( + ) overexpanded nozzle (RSS, () blunt fin (Touber & Sandham [25]); (◦) esti-
mated superstructures upstream influence for the 8◦ IUSTI case. Adapted from Dussauge
et al. [28].
r the extrapolation to the wall of the incident shock, for reflection cases
r the corner or, if known, the location of reattachment for compression ramps
r the beginning of the obstacle, for blunt fins
The collapse of the data versus Mach number for the different flows (i.e., shock
reflections, compression ramps, blunt bodies, and channel flows) is not excellent;
the scatter is about 20 percent. Moreover, no particular trend with Mach number is
observed, except for the compression-ramp flow of Thomas et al. [30]. Higher fre-
quencies were found than in other data; however, because the Mach number was 1.5,
transonic effects may have been present. This collapse, even if only partial, is sur-
prising because the dominant frequency reasonably could be expected to depend on
the flow geometry. It suggests that the different flow cases share common features
for the origin of the unsteadiness. However, the scatter indicates that some details
are not well represented. Nevertheless, it is evident that most of the experiments
have a Strouhal number around 0.03–0.04. As for well-developed interactions, L
is much larger than the boundary-layer thickness δ, and it appears unambiguously
that the unsteadiness is several orders of magnitude below the frequencies U∞ /δ
produced by the energetic eddies of the incoming layer of typical size δ.
The amplitude of the shock oscillations is now considered. As shown by Dus-
sauge et al. [28] and recalled by Dussauge and Piponniau [2], in many SBLIs, the
shock can move at a low frequency over a longitudinal distance Lex , which is on the
order of a significant fraction of the interaction length (typically, Lex /L ≈ 0.3). Dif-
ferent types of interactions – compression ramps, incident-shock waves, and blunt
fins – produce similar results (Fig. 9.5).
380 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
1
Incident shock wave
compression ramp
0.9 blunt fin
overexpended nozzle (RSS)
0.8
0.7
0.6
Lex/L
0.5
0.4
0.3
0.2
0.1
0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
M
Figure 9.5. Amplitude of the oscillation of the separation shock. From Dussauge et al. [28].
A shock velocity Us can be derived from this excursion length and from the
following frequency scale:
Because SL << 1, this implies that Us << U∞ . In supersonic conditions, this sug-
gests that the shock wave behaves quasistatically, the additional velocity Us being
insufficient to produce a significant added dynamic effect (i.e., the shock intensity
can be regarded as independent of velocity). This is in contrast to transonic situa-
tions in which the shock intensity depends generally on the velocity (see, e.g., Bruce
and Babinsky [23]).
Other properties of these interactions can be recalled. First, we examine fea-
tures shared by different flows at the beginning of separation. In this region, high
levels of velocity fluctuations are observed with the maxima in the turbulence inten-
sity being located away from the wall, as shown in Fig. 9.6 for the shock-reflection
case.
Smits and Dussauge [31] assessed the amplification of turbulence levels through
the shock using the “rapid-distortion approximation” for the case of a 24-degree
compression-ramp flow, assuming that the dilatation effects are predominant as
well as by using Debiève’s formulation (Debiève et al. [9]). (Incidentally, this is an
appropriate way to assess the levels of turbulence in the 8-degree shock-reflection
case illustrated in Fig. 9.6.) Smits and Dussauge concluded that the resulting levels
were not consistent with linear-amplification mechanisms. Instead, they observed
that they are consistent with the level of turbulence in mixing layers, u /U ∼10
percent, for example. The spatial organization of the flow indicated in Fig. 9.6 is sim-
ilar to what is observed in subsonic separated flows (Kiya and Sasaki [27]; Cherry
et al. [32]). It is related to the development of a mixing layer starting at the sepa-
ration line: From this point, the mean velocity profiles are strongly deformed with
9.3 The Lower Branch: Separated Flows 381
Figure 9.6. A schlieren image of an oblique-shock reflection interaction at M = 2.3; flow devi-
ation 8◦ below which, to the same scale, is a plot of isocontours of the RMS values of vertical
velocity fluctuations.
inflectional shapes, as observed in plane mixing layers. The high level of turbulence
is related to the development of large coherent vortical structures of the Kelvin-
Helmholtz type [56]. In the second part of the separated region, subsonic studies
showed that these large vortices are shed into the reattached boundary layer and
that they can be observed over long distances downstream of reattachment.
The same type of organization is found in shock-induced separated flows. PIV
measurements (Dupont et al. [33]) for the oblique-shock reflection at M = 2.3 (see
Fig. 9.6) revealed the development of large vortical structures in the first part of the
interaction. In the second part of the interaction, these large eddies are shed down-
stream, as observed in subsonic separated flows. They contribute to high-turbulence
intensities well away from the wall in the downstream boundary layer and persist
over several interaction lengths. This is shown in Fig. 9.7, in which the vorticity
detector proposed by Graftieaux et al. [34] was applied to the PIV measurements.
The development of a mixing layer is detected along with vortex shedding into the
reattached layer.
Therefore, to summarize different characteristics of the interactions, it is pos-
sible to list the following five distinct types of fluctuations that may have different
frequency content and levels:
1. Turbulence of the upstream boundary layer. This involves mainly high frequen-
cies ( f ∼U∞ /δ), although very large scales (or superstructures) on the order of
30δ also were noted (Adrian et al. [35] and Ganapathisubramani et al. [36])
382 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
Figure 9.7. A contour plot of vorticity obtained from PIV measurements in the region near
the wall superimposed on a schlieren image of an oblique-shock reflection, M = 2.3; flow
deviation 8◦ .
knowledge of the mean structure of the interactions. In turn, this depends on know-
ing the correct separation-point location, which is difficult to ascertain precisely.
Examples are provided in Chapters 4 and 5 of this book as well as in several pub-
lished experimental compilations. The results often are flow-sensitive, for example,
whereas in nozzle flows at large Reynolds numbers, the onset of separation seems to
be relatively independent of this number (see, e.g., Zukoski [38] and Reijasse [39]),
and it appears that for compression-ramp and oblique-impinging shock flows, the
onset of separation at moderate Reynolds numbers is sensitive to it. To date, the
numerical prediction of separation and reattachment remains a challenging task for
many turbulence models. Although the precise extent and magnitude of the reverse
flow is generally difficult to estimate, when the interaction is strong enough to cause
the flow to separate, general observations can be made.
Drawing on classical considerations (e.g., Chapman’s [57] free-interaction the-
ory), one observation concerns the intensity of the leading shock formed upstream
of the separated region. As noted by Délery and Marvin [40] and by Haddad
[41], the strength of this shock does not depend on the particular downstream
conditions – that is, the ramp angle in compression-ramp flows or the incident-shock
angle in oblique-shock reflections. It also is observed that wall pressure starts to
increase upstream of the mean separation-shock location. In the past, this initial
rise was referred to as the “upstream influence” but, as noted by Dolling [42], it is
likely due to the intermittency caused by the shock traveling backward and forward.
This does not appear to depend on the specific conditions creating the interaction
but rather is much more general; it is tempting to conclude that this is a universal
feature that can be applied to explain the dominant frequency of the shock beating.
The frequency content of the different types of fluctuations is now discussed.
First, we consider the influence of the incoming turbulence in more detail. As shown
in Fig. 9.1, it is evident that upstream turbulence can influence the shock motion. As
noted previously, there is evidence of very long structures within boundary layers
(see Adrian et al. [35] and Ganapathisubramani et al. [43] for subsonic boundary
layers; see Ganapathisubramani et al. [43] for supersonic flows). It is possible to
determine the frequency scale generated by these superstructures in the upstream
flow and check whether it is consistent with the observed Strouhal numbers. Such
structures are supposed to be 30δ long, with a convection speed of 0.75U∞ . The
resulting frequency is therefore f = 0.75U30δ
∞
and the corresponding Strouhal number
will be SL = 0.025 δ . It turns out that such structures can generate fluctuations with
L
the right Strouhal number for interactions so that L/δ ∼ 1. For compression ramps
and oblique-shock reflections, such low ratios generally correspond to nonseparated
interactions or incipient separations but not to cases in which a separated zone is
well developed. Interactions like the oblique-shock reflections studied by Dupont
et al. [45 and 46] have ratios L/δ about 5 to 7 for Strouhal numbers about 0.03–
0.04. For separated compression ramps with M > 2, typical values of 0.02–0.05 are
obtained (Dussauge et al. [28]).
Further insight was obtained from tomographic visualizations of the oblique-
shock reflection by Piponniau et al. [47]. These authors recorded photographs of this
interaction for different values of the flow deviation. For small values of the deflec-
tion angle (typically, 4 or 5 degrees), the flows were not separated and the reflected
shock did not move much. However, on increasing the angle, the amplitude of the
384 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
1.8
1.6
1.4
1.2
1/St=U1/FL
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
X*=(X−Xo)/L
Figure 9.8. Longitudinal evolution of the inverse Strouhal numbers in subsonic and super-
sonic separations, ( ) subsonic separated flow Cherry et al. [32]; (· · ·) M = 1.5
compression-ramp flow Thomas et al. [30]; oblique-shock reflection M = 2.3 ( × ) θ = 7◦ ,
( + ) θ = 8◦ , (◦) θ = 8.8◦ , and () θ = 9.5◦ .
r 0 < X* < 0.5 corresponds to the region of development of the mixing layer
r 0.5 < X* < 0.8 corresponds to the shedding of large structures
r 0.8 < X* < 1.2: For the shock-reflection case, an increase in the frequency asso-
ciated with the structures shed downstream is observed in this region where the
flow is turned down to return parallel to the wall. This zone depends explicitly
on the geometry of the interaction and does not appear in compression-ramp
flows or the subsonic case.
r X* > 1.2: The frequency is associated with eddies that persist in the reattached
boundary layer over large distances after the shedding process, whatever the
configuration.
Figure 9.8 clearly shows that the longitudinal evolution of the Strouhal number
associated with large structures is globally similar for the different cases. Never-
theless, regarding the actual numbers, significant differences are observed between
supersonic and subsonic or transonic cases: For the M = 2.3 shock-reflection case,
the dimensionless shedding frequency is reduced by about 40 percent compared
with the incompressible value. Because mixing layers are known to be particularly
sensitive to compressibility effects, an attempt to relate this behavior to the char-
acteristic Mach number of the flow was made (Dupont et al. [46]), leading to the
following results. The Strouhal number can be expressed as follows:
fL Uc L f δω Uc L Uc Sδ ∗−1
SL = = =
Sδ ≈ X (9.3)
U∞ U∞ δω Uc U∞ δ x U∞ δ
Uc Sδ
SL, shed = 2 (9.4)
U∞ δ
Moreover, the spreading rate of the canonical plane mixing layer can be expressed
as follows (Papamoschou and Roshko [48]):
U
δ = δref
(Mc ) (9.5)
Uc
where U is the velocity difference across the mixing layer and δ ref is the spreading
rate of the subsonic mixing layer, with constant density and U = 0 on the low velocity
side. Based on the convection velocity of the large coherent scales Uc and the exter-
nal velocities U1 and U2 (Papamoschou and Roshko [48]), (Mc ) is a decreasing
function of the convective Mach number. The function (Mc ) expresses the drastic
reduction of the mixing-layer spreading rate and, consequently, of the entrainment
process when the convective Mach number increases.
386 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
In incompressible flows, the convection velocity can be estimated from the fol-
lowing equations:
√ √
Uc 1+r s U (1 − r ) 1 + s
= √ or = √ (9.6)
U∞ 1+ s Uc 1+r s
fshed L Uc L λshed −1 Uc 1 δ 1
SL, shed = = thus = SL, = ≈ (9.7)
U∞ U∞ λshed L shed
U∞ 2 Sδ 2
9.3 The Lower Branch: Separated Flows 387
Figure 9.9. Average distance between large eddies for several shock-reflection interactions
at M = 2.3.
Therefore, because the variation of the shedding Strouhal number is related to the
ratio Uc /U∞ , relation 9.7 indicates that typical wavelengths of about L/2 could be
expected, whatever the Mach number. This value is indicated by a dotted line in
Fig. 9.9, which agrees well with measurements taken in the rear half of the separa-
tion region (i.e., 0.5 < X* < 1.0).
From these considerations, the dominant frequencies in the separated region
are indicated. These flows have many features in common with their subsonic coun-
terparts; in particular, the recirculating zones do not generate a single frequency but
rather a range of frequencies that evolve with longitudinal distance. Experiments
with subsonic flows revealed that very-low-frequency perturbations are present
in the separated bubble. However, significant differences can be expected when
the speed is high enough to generate convective Mach numbers larger than 0.5.
Fig. 9.8 shows that the associated Strouhal numbers are much larger than the shock
Strouhal numbers (see Fig. 9.4). Their frequency is an order of magnitude lower
than the unsteadiness associated with the large coherent structures formed in the
mixing layer. Such low frequencies are associated with the flapping of the mixing
layer or with the bubble “breathing,” with a typical Strouhal number of 0.12.
Erengil and Dolling [55] in an M = 5 compression-ramp experiment were prob-
ably the first to remark that the subsonic low-frequency flapping (SL = 0.12) could
not be compared directly to the low-frequency shock motions (SL = 0.03). System-
atic compilations performed by Dussauge et al. [28] (see Fig. 9.4) confirmed that for
any geometry, similar conclusions can be drawn. For shock-induced separations, if
the external Mach number is larger than two, the shock-motion frequency is at least
four times smaller than their subsonic counterpart. Moreover, the same reduced
frequencies are observed for both the low-frequency bubble “breathing” and the
388 Shock-Wave Unsteadiness in Turbulent Shock Boundary-Layer Interactions
Shock
h Dividing
stream line
S R
L1
shock motions (SL ≈ 0.03). Although not the most energetic band, these low fre-
quencies were found in the wall-pressure spectra in the separated zone (Dupont
et al. [33]). Furthermore, the coherence between wall-pressure fluctuations at the
foot of the shock and in the recirculation region are high, typically 0.8 (Dupont
et al. [46]). Subsequently, Piponniau et al. [24] took conditional measurements of
the size of the separated bubble and found that it is strongly intermittent, with a few
events during which intense backward flow penetrates the separation pocket. Their
measurements showed clearly that these events generate large vertical-amplitude
oscillations of the mixing layer, which are linked to significantly large longitudinal
excursions of the shock. In their analysis of the origin of the unsteadiness, Pipon-
niau et al. reasoned that the large shock pulsations are closely related to the flap-
ping of the mixing layer formed at the edge of the separated bubble, as illustrated
in Fig. 9.10. Therefore, they proposed an explanation based on considerations of air
entrainment by this mixing layer.
Their objective was to find the parametric dependence of the shock-motion fre-
quency rather than a complete theoretical description. Piponniau et al. suggested
that the low-frequency flapping of the shear layer must be related to an unbal-
anced mass budget in the separated region. The basic mechanism is illustrated in
Fig. 9.10. Large vortical structures develop in the initial part of the interaction (X* <
0.5), entraining mass from the reverse flow. Then, in the second part of the interac-
tion, the structures are shed into the downstream layer, causing a mass defect in the
separated region. Considering these elements, they evaluated the amount of mass
contained in the bubble and the rate of mass entrainment. The ratio of these two
quantities provides a time scale, which represents the interval necessary to drain a
significant amount of mass from the separated zone. The inverse of this time pro-
vides an order-of-magnitude estimate of the frequency scale at which new air is
injected into the recirculating zone. Piponniau et al. assumed that the dependence
of the spreading rate of the mixing layer on the density and velocity ratios and on
the convective Mach number is the same as in canonical mixing layers. It appears
9.4 Conclusions: A Tentative Classification of Unsteadiness and Related Frequencies 389
that a length scale introduced naturally by this model is the height h of the separated
bubble. The analysis provides a Strouhal number Sh = fh/U∞ based on this height
of the following form:
in which (Mc ) is again the normalized spreading rate of the mixing layer. The func-
tion g is from the contributions of density and velocity ratios (i.e., r and s, respec-
tively) to the spreading rate and the rate of mass entrainment. It is a weak function
of r and s and varies minimally for the usual values found in boundary-layer separa-
tions over adiabatic plates. An expression was proposed in which it is assumed that
the velocity profiles in the mixing layer follow an error function, as follows:
√
δref (1 − r )(1 + s) r
g (r, s) = √ (1 − r ) C + , with C ≈ 0.14 (9.9)
2 1+r s 2
L
SL = (Mc ) g (r, s) (9.9)
h
from which it appears that SL is proportional to the aspect ratio L/h of the sepa-
rated bubble. If the aspect ratio of the bubble is taken as constant, this implies that
the Strouhal number varies with the convective Mach number like the spreading
rate of the mixing layer. This hypothesis is supported by the existing data shown in
Fig. 9.11.
In most of the interactions considered in this chapter, the convective Mach num-
ber of the large structures in the mixing layer is close to 1, which corresponds to a
value of (Mc ) of about 0.2. This implies that the aspect ratio of the separation
bubbles is about 6, which is consistent with what is known about their geometry.
Therefore, it appears that this simple model provides a more general representation
of the unsteadiness. Of course, this scheme is limited to two-dimensional situations
in which a reattachment point exists – for example, in the so-called RSS in nozzle
flows. Moreover, it can be inferred from the previous analysis that SL depends on
the geometry of separated zones and thus will have different values if the shape of
the separation bubble varies considerably. The model, however, indicates the lead-
ing elements for analyzing other situations and how to control them.
10
7
−1
S .[ Φ(M ).g(r,s)]
5
c
4
L
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
M
Figure 9.11. Strouhal number corrected for compressibility effects, according to Piponniau
et al. [24].
range of the unsteadiness depends on the nature of the interaction. Three typical sit-
uations considered in previous sections are used for illustration: the transonic inter-
action with acoustic coupling, the supersonic-separated case, and the supersonic-
nonseparated case. The transonic case is exemplified by a case of acoustic coupling
between the trailing edge and the shock wave on a wing profile. As noted in Sec-
tion 9.3.2, this leads to the following frequency:
fbu f ≈ (a2 − u2 ) (9.10)
Normalized Order of
Flow Phenomenon Frequency frequency magnitude
δ
Transonic Acoustic coupling (a2 − u2 )/ (M∞ − 1) <10 − 2
interaction
U∞ δ
Separated, Mass conservation (Mc ) g(r, s) (Mc ) g(r, s) <10 − 1
h h
supersonic
Nonseparated Eddy convection U∞ /δ 1 1
M∞ M2 ≈ 1 (9.13)
REFERENCES
[46] P. Dupont, C. Haddad, and J. F. Debiève. Space and time organization in a shock-
induced separated boundary layer. Journal of Fluid Mechanics, 559 (2006), 255–77.
[47] S. Piponniau, P. Dupont, J. F. Debiève, and A. Sidorenko. Unpublished work, IUSTI
(2007).
[48] D. Papamoschou and A. Roshko. The compressible turbulent shear layer: An experi-
mental study. Journal of Fluid Mechanics, 197 (1988), 453–77.
[49] W. Blumen. Shear layer instability of an inviscid compressible fluid. Journal of Fluid
Mechanics, 40 (1970), 769–81.
[50] P. Muscat. Structures à grandes échelles dans une couche de mélange supersonique. Anal-
yse de Fourier et analyse en ondelettes. Thèse de Doctorat, Université de la Méditerranée
(Aix-Marseille II), Marseille, France (1998).
[51] D. Papamoschou. Structure of the Compressible Turbulent Shear Layer. AIAA Journal,
29, 5 (1991).
[52] S. Barre, P. Dupont, and J. P. Dussauge. Estimation de la vitesse de convection des struc-
tures turbulentes à grande échelle dans les couches de mélange supersonique. Aerospace
Science and Technology, 1 (1997), 4, 355–66.
[53] C. K. W. Tam and F. Q. Hu. The instability and acoustic wave modes of supersonic
mixing layers inside a rectangular channel. Journal of Fluid Mechanics, 203 (1989), 51–
76.
[54] J. A. Greenough, J. J. Riley, M. Soetrisno, and D. S. Eberhardt. The effect of walls on a
compressible mixing layer. AIAA Paper 89–0372 (1989).
[55] M. E. Erengil and D. S. Dolling. Unsteady wave structure near separation in a Mach 5
compression ramp interaction. AIAA Journal, 29 (1991), 5, 728–35.
[56] Jie-Zhi Wu, Hui-Yang Ma, Ming-De Zhou, Vorticity and Vortex Dynamics (Berlin Hei-
delberg: Springer Verlag, 2006).
[57] D. Chapman, D. Huehn, H. Larson, Investigation of separated flows in supersonic and
subsonic streams with emphasis on the effect of transition. NACA TM 3869, 1957.
10 Analytical Treatment of Shock
Wave–Boundary-Layer Interactions
George V. Inger
10.1 Introduction
395
396 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
These attributes argue strongly for the inclusion (whenever possible) of a vigorous
analytical component to fluid-mechanics research in general and certainly to stud-
ies of SBLI in particular. A focused combination of analytical, computational, and
experimental approaches wherein each is used to complement the other remains the
most powerful strategy for the investigation of the complex physics involved.
10.1.3 Content
Section 10.2 is a qualitative overview of the interaction process, the primary triple-
deck structure, and the essential physical distinctions between laminar and turbu-
lent boundary-layer responses to either externally impinging or wall-compression-
corner-generated shocks. Section 10.3 is a detailed, step-by-step analysis of either
laminar or turbulent nonadiabatic-disturbance flow within the triple-deck layers and
how they are matched, followed by a discussion of the conditions imposed by the
various Mach-number regimes of the interactive-inviscid outer deck. Section 10.3
concludes by combining all of these elements into sets of final, canonical, nondi-
mensional triple-deck formulations, along with determination of the corresponding
interactive flow-scaling factors. To provide background for the “uninitiated,” Sec-
tion 10.3 offers more intermediate detail and explanation on the triple-deck theory
than typically found in the literature.
Section 10.4 addresses application of the previously mentioned theory to lam-
inar interactions involving supersonic flow past adiabatic walls, hypersonic nona-
diabatic flows, and transonic flows. The mechanisms of free interaction, upstream
10.2 Qualitative Features of SBLIs 397
Flow
M∞
Lam. B. L.
Shock
Lam. B.L.
The prominent expansion feature (2), which effectively eliminates the external
shock jump at the boundary-layer edge, is due to the previously mentioned fact
that the sonic line in a laminar boundary-layer profile occurs near the outer edge,
thereby causing the incident shock to locally reflect as if from a constant-pressure
surface (i.e., as an expansion). The combined effect of all of these features from the
standpoint of a theoretical model is that the wave system of a laminar interaction
is dispersed so smoothly both streamwise and laterally that the resolution of the
incident-shock–diffraction process across the thin, upper, supersonic region near the
boundary-layer edge is of negligible importance.
Examination of the typical pattern of a nonseparating turbulent interaction,
conversely, discloses a different physical situation in two respects (see Figs. 2.22
and 2.23): (1) the upstream-influence region is much shorter (u ∼ δ0 ), the outgo-
ing curved compression waves emanating from deep within the boundary layer; and
(2) the incident shock penetrates far down into the boundary layer because of the
deeply buried sonic line, with the resulting dispersed expansion reflections from
this sonic line exiting the boundary layer to externally interact with the upstream-
generated compression waves of (1). In contrast to the laminar case, these features –
with the much tighter streamwise focus of the interaction – make the resolution of
the shock diffraction within the boundary layer a more prominent aspect of the the-
oretical problem.
1. An outer region of inviscid flow above the boundary layer, which contains the
inviscid shock and interactive-wave systems
2. An intermediate deck of negligible shear-stress-perturbation rotational invis-
cid disturbance flow occupying the outer 90 percent or more of the incoming
boundary-layer thickness
3. An inner shear-stress disturbance sublayer consisting of both laminar and tur-
bulent eddy (i.e., Reynolds) stress contributions, which contains the interactive
skin-friction perturbations (and, hence, any possible incipient separation) and
the origin of the upstream influence of the interaction
The inviscid shock associated with the “forcing function” of the problem is
impressed by the outer deck on the boundary layer. The middle deck couples this to
the response of the inner deck but can modify the disturbance field to some extent,
whereas the slow flow in the thin inner deck reacts strongly to the pressure-gradient
disturbances imposed by these overlying decks. Qualitatively, this triple-deck struc-
ture also was established for nonasymptotic Reynolds numbers Re ∼ 105 –108 by
Gadd et al. [19], Lighthill [2], and Honda [20] and is supported by a large body of
experimental evidence and numerical studies with the full Navier-Stokes equations
[21, 22]
Triple-deck theory consists of applying a systematic analysis of the disturbance
flow in each deck (to the leading approximation in an appropriate basic small param-
eter) and then matching them to form a mutually interactive, self-consistent set of
equations explicitly linked to the properly scaled forcing function involved. The
solution to such a formulation then provides a description of important interaction
10.3 Detailed Analytical Features of the Triple Deck 401
Shock
Me
Profile
(shape factor H)
Rotational,
Middle
Inviscid, h
δo disturbance flow
Deck
LAMINAR TURBULENT
h/L δ0 /L δ0 /L
(δ0 /L)3/2
xUP /L (δ0 /L)3/4
δ0 /L ReL−1/2 Cf0
y
CURVED SHOCK
M o→ 1
δo
INTERMEDIATE REGION
INNER REGION
δi
x
y
Mo(y) ∼ 1 δi SHOCK FOCUSING
SONIC LINE
M= 1
l
x
UPSTREAM
INFLUENCE
Figure 10.4. Further smaller-scale regions needed to resolve the local shock structure in a
turbulent interaction. (Enlarged view of inner region shown in lower figure.)
Thus, the governing equations of the middle deck may be considered as those of a
particle-isentropic rotational inviscid flow.
Imagine that the flow consists of disturbances proportional to an appropriate
small parameter ε (which may be Reynolds-number dependent) about the undis-
turbed state of the incoming boundary layer (denoted by subscript zero) and so we
write the following:
u = U0 (x, y) + εu (x, y) (10.3.1)
∂u ∂U0 1 ∂ p
U0 + v − < x-momentum > (10.3.8)
∂x ∂y ρ0 ∂ x
∂v 1 ∂ p
U0 − < normal momentum > (10.3.9)
∂x ρ0 ∂ y
ρ T
p / p0e + <equation of state> (10.3.10)
ρ0 T0
∂ H dH0
U0 + v 0 <steady adiabatic energy> (10.3.11)
∂x dy
where H = (γ RT )/(γ − 1) + U0 u . An important immediate consequence of these
equations can be realized by combining equations (10.3.7), (10.3.8), (10.3.10), and
(10.3.11) to obtain the following vertical-velocity relationship:
∂ (v / U0 ) 1 − M02 ∂ ( p / p0e )
(10.3.12)
∂y γ M02 ∂x
where M02 (y) = ρ0 (y) U02 (y)/γ p0e is the undisturbed Mach-number distribution
across the incoming layer. Integrating equation (10.3.12) across the boundary layer
from the effective edge y = yi of the inner deck to the edge of the incoming bound-
ary layer y = δ 0 , and assuming ∂ p /∂ y 0, we obtain:
ve v ∂ p /∂ x δ0 1 − M02
= (yi ) + dy (10.3.13)
U0e U0 γ p0e yi M02
which describes the streamline slope change across the middle deck (the so-called
streamline-divergence effect). Equation (10.3.13) is important because it connects
the streamline deflection of the outer deck to the vertical displacement of the inner
deck. At this point, we might further integrate the normal-momentum equation
(10.3.9) to obtain the lateral pressure variation across the deck, as follows:
δ0 2
pe − pw M0 ∂(v /U0 )
dy (10.3.14)
ρe U0e yi M0e ∂x
which is seen as very small except where the streamwise (x) scale of the interac-
tion is much less than the already-small lateral y-scale (δ 0 ). Thus, the influence of
this variation compared to that of the rapid streamline pressure changes along the
interaction can be neglected unless the finer details of the wave structure within the
boundary layer are of interest [32]. Hereafter, we regard p as a function of x only.
Equation (10.3.13) combined with the streamwise-momentum equation
(10.3.8) – as well as the energy equation if the heat-transfer aspects are of inter-
est – constitute the key relationships needed to link the bottom of the middle deck
10.3 Detailed Analytical Features of the Triple Deck 405
(i.e., the top of the inner deck) to the outer-deck disturbance flow. Further insight
into the behavior near the body now requires a more detailed consideration of the
velocity profile and the scaling properties of the flow, both of which depend strongly
on whether the flow is laminar or turbulent.
(10.3.15)
" #
∂ H ve dH0 dp /dx dH0
lim =− + Im (y) (10.3.16)
y→yw ∂x U0e dy y→yw γ p0e dy y→yw
from which the pressure disturbance p (x) can be found once the inner-deck height
and a pressure deflection-angle relationship for the outer deck are specified (dis-
cussed herein). Furthermore, because U0e − U0 , ρ0e − ρ0 , dU0 /dy and dH0 /dy in
turbulent flow are each proportional to εT in the defect region occupied by the mid-
dle deck [12], the leading inner approximations of equations (10.3.8) and (10.3.11)
to first order in εT become:
∂u 1 dp /dx
lim − (10.3.19)
y→yw ∂x (ρ0e + · · ·) (U0e + · · ·)
∂ H
lim 0 (10.3.20)
y→yw ∂x
because the terms involving vi are negligibly small. These equations may be rec-
ognized as those governing adiabatic inviscid small-perturbation flows in a parallel
stream.
The turbulent middle deck downstream of the shock is significantly different
because the interaction there must be dominated by the imposed forcing function
(e.g., an incident-shock pressure jump) that is necessarily included in the ve /U0e
term. Consequently, although the vertical-velocity effect is still small, it cannot be
dismissed; rather, it must be retained in the momentum equation (10.3.8) by again
using equations (10.3.15) and (10.3.16) but now with high-Reynolds-number turbu-
lent values for the incoming boundary-layer properties. Thus, we have the inner
conditions for x > 0 that:
∂u dp /dx ve dU0 dp /dx dU0
lim =− − lim + lim Im
y→yw ∂x ρ0e U0e U0e y→yw dy γ p0e y→yw dy
(10.3.21)
∂H ve dH0 dp /dx dH0
lim =− lim + lim Im (10.3.22)
y→yw ∂x U0e y→yw dy γ p0e y→yw dy
where the first two terms on the right-hand side of equation (10.3.21) are dominant.
This downstream situation significantly interacts with the inner deck, unlike the dis-
turbance flow upstream. If the incoming boundary layer is nonadiabatic, so is this
downstream interaction.
In formulating the governing flow equations, we assume that the inner deck
lies within the Law of the Wall region of the incoming turbulent boundary layer
(which appreciably overlaps the inner part of the outer-defect region) and treat
the turbulent Reynolds stress and heat transfer therein by the well-known Van
Driest eddy-viscosity model appropriate for nonseparating flows [12]. Furthermore,
we neglect the effect of streamwise variations of the undisturbed flow over the short
extent of nonseparating interactions, consistent with the same approximation in the
main-deck analysis.
To allow for the presence of a possible local wall-shape distortion yw (x) rela-
tive to the incoming upstream flow, such as a “bump” or compression corner, it is
convenient to formulate the inner-deck equations in terms of the following shifted
variables:
z ≡ y − yw (10.3.23)
which are the vertical distance relative to the distorted surface and the vertical veloc-
ity relative to that associated with an inviscid streamline along yw , respectively.
Application of the Prandtl Transposition Theorem [39] shows that the resulting
continuity, momentum, and energy equations are invariant under such a transfor-
mation. In addition, it is convenient to absorb much of the compressibility effect
in the problem by further introducing the following Howarth-Dorodnitzn transfor-
mations [40]:
⎫
U(X, Z) ≡ u(x, z) ⎪
⎪
⎪
⎪
∂Z ⎪
⎪
W(X, Z) ≡ U + ρw(x, z)/ρ0w ⎬
∂X (10.3.25)
dZ ≡ ρdz/ρ0w ⎪
⎪
⎪
⎪
⎪
⎪
X= x ⎭
which render the continuity equation into the following incompressible form:
∂U ∂W
+ =0 (10.3.26)
∂X ∂Z
When the Chapman-Rubensin [40] approximation ρμ = const. = ρ0w μ0w is further
applied to the laminar portion of the shear stress, the compressible momentum and
energy equations resulting from the transformations (i.e., equations [10.3.23] and
[10.3.25]) assume the following forms when expressed in terms of the perturbations
u = u − U0 (y) and H = H − H0 (y):
∂u dU0 ∂u 1 dp
(U0 + u ) +W + +
∂X dZ ∂Z ρ0 dX
∂ ∂u ρ0 μT0 ∂u
= v0w +2 (10.3.27)
∂Z ∂Z ρ0w ρ0w ∂ Z
∂ H dH0 ∂ H
(U0 + u ) + W +
∂X dZ ∂Z
∂ v0w ∂ H μT0 ρ0 μT0 ∂ H dH0 ∂u
+ + (10.3.28)
∂ Z PR ∂ Z PR ρ0w ρ0w ∂ Z dU0 ∂ Z
408 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
where the undisturbed flow was subtracted and the small effect of a term
(1 − PR)/∂/∂ y[∂(u2 /2)/∂ y] was neglected on the right-hand side of equation
(10.3.28). The undisturbed background flow is governed by the Law of the Wall
relationships [12]:
ρ0 μT0 dU0 τ0w
v0w + = ≡ Uτ2 (10.3.29)
ρ0w ρ0w dZ ρ0w
v0w 1 ρ0 μT0 dH0 −q̇w0
+ = (10.3.30)
PR PRT ρ0w ρ0w dZ ρ0w
whereas the perturbed- and basic-flow eddy viscosities, respectively, are:
∂u
εT (dU0 /dZ) μT0 (10.3.31a)
∂z
and
μT0 ρ0 z dU0
= kD Z (10.3.31b)
ρ0w ρ0w Z dZ
with D, the Van Driest damping factor:
These equations are solved subject to either the no-slip conditions that W , u ,
and H vanish along the wall Z = 0 for all X or equivalent conditions outside
the laminar sublayer (discussed herein), as well as outer boundary conditions on
u and H as Z → ∞ that enforce matching with the previously mentioned inner
behavior of the overlying middle deck. In conjunction with the pressure gradient in
momentum equations (10.3) through (10.27), near the surface, the variable-density
coefficient ρ0−1 = RT0 / p0 (T0 /Tw )ρ0w −1
can be expressed as the following Taylor
series expansion in the distance Z across the inner deck after noting that U0w = 0:
ρ0w T0w + ∂∂TZ0 w Z + · · · 1 ∂ H0
= 1+ Z (10.3.33)
ρ0 T0w H0w ∂ Z w
This relationship indicates that ρ 0 may be taken as a constant across the thin inner
deck when the incoming boundary layer is adiabatic or when (∂ H0 /∂ Z)w Z/H0w is
negligible compared to unity.
Although we know at least the vertical scale of the middle deck (ym ∼ δ0 ), at
this point we have no a priori knowledge of the scales of the inner deck and the
disturbance flow. To determine them, the “hated” nondimensional variables x̂ ≡
X/ XR, ẑ ≡ Z/ZR, û ≡ u/u R, ŵ ≡ W /WR, ρ̂ ≡ ρ/ρ R, p̂ ≡ p /ρ0e U0e 2π R are intro-
duced with Ĥ ≡ H/HR, where the R-subscripted reference parameters are scaling
factors to be found. These variables are substituted into the previous governing
equations and they require that the UR , WR , and so on be chosen such that the
resulting nondimensional terms are all of equal order-unity. Thus, for example, the
continuity equation (10.3.26) reduces to:
if we set:
whatever the value of the individual parameters. Addressed next are the momentum
and energy equations and a detailed turbulence model is introduced.
where Ĥ0 ≡ H0 /HR and the evaluation of d Ĥ0 /dÛ 0 (i.e., dividing equation [10.3.30]
by equation [10.3.29]) indicates that it is rendered of order unity if HR is chosen as
follows:
Equations (10.3.38) and (10.3.39) are to be solved subject to the wall no-slip condi-
tions uw = 0, T = Tw along the body ẑ = 0 as well as the outer boundary-conditions
provided by the appropriate inner-middle-deck matching relationships.
yet contains the integrated effect of the true wall no-slip condition across the lami-
nar sublayer. The classical logarithmic Law of the Wall for turbulent boundary lay-
ers, which appreciably overlaps the defect region (see Fig. 2.13), in fact provides a
good engineering account of these features over a wide range of conditions, even in
the presence of moderately adverse pressure gradients if significant separation does
not occur. Thus, we may apply it locally along the interaction both at the height
zc 60v0w /Uτ just outside the laminar sublayer at the bottom of the defect region,
and along the undisturbed boundary-layer edge, where dU0 /dy and dH0 /dy van-
ish. As shown in Appendix 10.A, this procedure yields a pair of local perturbation
relationships linking the desired slip velocity along zc and the corresponding skin-
friction disturbance to the overlying interactive disturbance field. After introducing
scaled variables, they are as follows:
û (x̂, ẑc ) = − (Bi /B0 ) p̂(x̂) (10.3.51)
τw (x̂) ∼
= −2τ0w p̂B0−1 (10.3.52)
where Bi and B0 are Law of the Wall parameters for the incoming boundary layer
defined in Appendix A. An analogous procedure applied to the Temperature Law
of the Wall yields a corresponding pair of relationships for the local total-enthalpy
slip and perturbation heat transfer, respectively (Appendix A):
Ĥ (x̂, zc )/C p TR = −(Bt − kt−1 ) p̂B0−1 − Bi (q̇w /q̇w0 ) (10.3.53)
H (X, Z → 0) = H0 (Z → 0) + H (x, Z)
H0w + (dH0 /dZ)w · Z + H (x, Z) (10.3.56)
and substituting the values of u and H obtained by streamwise integration of the
middle-deck relations, (equations (10.3.15) and (10.3.16)) the principle yields the
412 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
where, in obtaining the last term of equation (10.3–70), we used ZR/δ0 = kεT from
equation (10.3.47) and where:
kεT−1
−1
f (ẑ) ≡ (k ẑ) 1 − M02 (ẑ) /M02 (ẑ) d ẑ (10.3.71)
ẑi
general, the pressure-perturbation field p (x) consists of the pressure jump ps due
to the imposed inviscid shock (whether from an externally imposed wave or due to
a wall-compression corner), as well as a local viscous displacement-induced interac-
tive component pINTER as follows [42, 43]:
p (X) = J ps + pinter
(10.3.72)
where J = 0 for X < 0 and J = 1 for X ≥ 0 introduces the step-function driving-
pressure jump* imposed at X = 0 (see Fig. 2.24), whereas pinter is proportional to
the local interactive streamline deflection along the outer edge of the middle deck
and vanishes both far upstream and far downstream of the interaction zone. Thus,
an appropriate expression relating pinter to ve /U0e is needed, and it is here that the
particular Mach-number regime of the outer flow primarily enters the problem.
When the external flow is supersonic and outside the transonic range (Me ≥
1.2 or so), the tangent-wedge approximation [8, 9] as well as Van Dyke’s com-
bined Supersonic–Hypersonic Similarity Rule [44] are satisfactory for engineering
purposes throughout the entire supersonic–hypersonic regime unless the far-field
aspects of the interaction well above the boundary layer are of interest. This gives
the following:
−1 2
pinter /ρ0e U0e (ve /U0e ) β
2
1 + κ H (ve /U0e ) + κ H (ve /U0e )
2 (10.3.73)
* For example, for a compression corner of small angle θw at moderately supersonic Mach num-
bers, ps ∼
= ρe Ue2 θw /β, whereas for a weak impinging oblique shock wave of incident-pressure-
jump strength, pi , ps ∼= 2pi .
10.3 Detailed Analytical Features of the Triple Deck 415
βπ R = ZR/ XR (10.3.79)
which then yields from equations (10.3.61), (10.3.62), and (10.3.72) the following
final form of the outer matching conditions:
x̂
[ p̂ − J p̃] d x̂
lim [ûinner (x̂, ẑ)] = ẑ − 1/2
ẑ→∞ −∞ [1 + χ H ( p̂ − J p̃)]
δ0 p̃ M0e
2
+ lim [Im (zi /δ0 )] (10.3.80)
β XR ẑ→0
lim Ĥinner (x̂, ẑ) A = lim [ûinner (x̂, ẑ)] (10.3.81)
ẑ→∞ ẑ→0
Bodonyi and Kluwick [47] showed that this most conveniently is scaled by introduc-
ing the independent laminar-transonic interaction parameter:
χT ≡ β 2 /(γ + 1) M0e
4
πR (10.3.83)
and choosing:
4 3/2 1/2
ZR /XR = (γ + 1) M0e π R χT = βπ R , (10.3.84)
Equation (10.3.82) applied to equation (10.3.61) then yields the following form of
the outer boundary condition:
x̂ "
3/2 #
2
lim [ûinner (x̂, ẑ)] = ẑ − χT 1 − 1 − p̂χT−1 d x̄ (10.3.85)
ẑ→∞ 3 −∞
where we dropped the resulting last term (δ0 /χ R) p̂ in equation (10.3.61) because it
8/5
is of negligible higher order (εL ) in transonic flow [47].
When mixed-transonic external flow with a near-normal shock is involved, this
still applies to the upstream region (the linearized supersonic version usually is suffi-
ciently accurate in practice), whereas downstream equation (10.3.77) is appropriate.
After using equations (10.3.83) and (10.3.84), we then obtain that:
XR ve 1 − M22 1 ∞ p̂e (x) − J p̂
= dξ̂ (10.3.86)
ZR U0e Me2 − 1 π 0 x̂ − ξ̂
where equation (10.3.61), without the last term, yields the downstream boundary
condition in the form of equation (10.3.85) with the right-hand side replaced by:
1 − M22 x̂ " ∞ p̂ − J p̂s #
= ẑ − dξ̂ d x̂ (10.3.87)
πβ 0 0 x̂ − ξ̂
lim Ĥinner (x̂ ≥ 0, ẑ) = p̂ + lim [ûinner (x̂, ẑ)] (10.3.91)
ẑ→∞ ẑ→0
In the transonic regime with purely supersonic-inviscid flow, the ve /εT U0e term
from equation (10.3.76) can be expressed as follows:
ve /εT U0e = (2/3) βχTi 1 − [1 − ( p̂/χTi )]3/2 (10.3.92a)
latter in this chapter, based on Eckert’s empirical result (further supported by fun-
damental compressible boundary-layer theory [48]) for air:
TR /T∞ ∼
= 0.50 + 0.039M∞
2
+ 0.05Tw /T∞ (10.3.93)
vw XR = UR Z2R (10.3.96)
ZR/ XR = βπ R (10.3.98)
When we supply appropriate relationships for the two wall gradients (dU0 /dZ)w
and (dH0 /dZ)w , these relationships then can be solved to express the scalings in
−1/8
terms of the basic laminar triple-deck small parameter ε L ≡ Re L and the prop-
erties of the undisturbed incoming flow. Introducing the laminar compressible flat-
plate boundary-layer values [49]:
−1/2
(dU0 /dZ)w = τw0 /μw0 = (U0e λ/L) CR (Te /TR) ε−4
L (10.3.100)
−1/2
PR (μe /μ R) λε −4
1/3
(dH0 /dZ)w = (HWAD − H0w ) CR L /L (10.3.101)
with λ = 0.332, μ = μ R (T/TR)ω and C R ≡ μ RTe /μe TR, equations (10.3.94) through
(10.3.99) yield the following values:
π R = Cπ ε2L (10.3.104)
UR/U0e = Cu ε L (10.3.105)
HR = C H εL (10.3.106)
from which it is apparent that χ H is actually only of second-order importance ∼ ε 2L
even in moderately strong hypersonic flows. Thus, as in the contemporary triple-
deck literature, we neglect the effect on the previously mentioned boundary condi-
tions and thereby appreciably simplify the theory. The other parameter of possible
importance at high Mach number – the coefficient of p̂ in the last term of equation
(10.3.80) – takes the following form after applying the above scalings and the value
1/2
δ0 /L = Cδ ε4L with Cδ = 5.2CR (TR/Te ):
δ0
M2 Im = ε L SIm (10.3.109)
β XR e
where S = Cδ Me2 /Cx β. Although formally of order ε L and hence negligible at ordi-
nary supersonic speeds, it is observed that this streamline-divergence term can be
significant at hypersonic speeds [50] because then the coefficient S ∼ M0e and hence
it can be large, especially if the influence of heat transfer (which enters via the inte-
gral Im ; see Appendix B) is of interest [51]. Finally, when the values of equations
(10.3.102) through (10.3.107) are applied to the heat-transfer factor on the right-
hand side of equation (10.3.33), this factor becomes:
24/5
ZR/L = kzε L (10.3.112)
8/5
π R = kπ ε L (10.3.113)
4/5
UR/U0e = ku ε L (10.3.114)
16/5
VR/U0e = kv ε L (10.3.115)
where the five coefficients kx , kz, kπ , ku , and kv are also given in Appendix C. The
attendant laminar-interaction parameter χT (equation [10.3.83]) assumes the form:
2
χT = M0e − 1 / (γ + 1) M0e 4
kπ ε−8/5 (10.3.116)
which clearly implies that this (prescribed) value is asymptotically large. The hyper-
8/5
sonic terms involving χ H and δ0 Me2 p̂/ XR are both of negligible higher order ε L
under this scaling.
420 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
where:
x̂
Iu (x̂) ≡ [ p̂ − J p̃] d x̂ + ε L SIm p̂ Supersonic/Hypersonic (10.3.121)
−∞
" , Transonic : at all x̂ for M > 1 -
2
x̂
3/2 # 2
−1
= χT 1 − 1 − p̃χT d x̂ or at x̂ < 0 only for M2 < 1
3 −∞
(10.3.122)
, Mixed Transonic Flow, -
1 − M22 1 x̂ ∞
p̂ − p̂
= dξ̂ d x̂ M2 < 1 : at x̂ > 0
β π 0 x̂ − ẑ
0
(10.3.123)
The energy-equation aspects are ignored in the transonic case, whereas in the mod-
erate supersonic case, the streamline-divergence term ε L SIm p̂ can be consistently
neglected.
UR
= εT (10.3.127)
u0e
HR = εT (dH0 /dU0 )w U0e = εT PR0.6 [(HW,AD − Hw ) /(Te /Tw )] (10.3.128)
Because these
relations directly involve the basic small-perturbation parameter
εT ≡ Uτ /U0e = (ρe /ρw ) C f0 /2, no further development is needed.
10.3 Detailed Analytical Features of the Triple Deck 421
where we reiterate that this x̂ > 0 formulation does not resolve the detailed shock-
diffraction effects in the region 0 ≤ x̂ 1. When the external flow is transonic with
Me > 1, the boundary conditions assuming adiabatic flow are instead those given by
equation (10.3.92a).
422 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
2.0
θˆ B = 2.00
p̂
1.0 (a)
θˆ B = 1.00
.5
0
−10 −5 0 5 10 15 xˆ 20
1.0
θˆ B = 1.00
θˆ B = 1.57 = θ i.s.
τˆw θˆ B = 2.00
.5
(b)
−.5
−10 −5 0 5 10 15 x̂ 20
recovery downstream of the ramp. Careful comparisons of such results with experi-
ments and the predictions of nonasymptotic interacting boundary-layer theory over
a wide range of Reynolds numbers [52, 53] showed that the error involved in the
first-order asymptotic approximation becomes significant at Re L ≤ 5 × 105 , primar-
ily due to neglect of the streamline-divergence effect (Fig. 10.6). This situation is not
improved by including second-order terms in the asymptotic theory [54].
Surface bump-generated interactions in supersonic flow also were studied [55].
They involve a compression on the front face followed by a downstream overexpan-
sion and then ultimate recompression in the wake. Very small bumps much thin-
ner than the incoming boundary layer permit a linearized treatment of the triple-
deck equations that yields analytical solutions for both the upstream free-interaction
zone and the algebraic-disturbance decay in the far downstream wake (see [55] for
details).
4 α = 3.5 1/4
reattachment α scaled angle α∼a R
P
3
1 α=1 1.5
2.5
τw 2 2.5
3
2 2
separation 3.5
1.5
0
1
0 −1
−20 0 20 X −20 0 20 X
a - Wall pressure distribution b - Wall shear-stress distribution
4 RL= 6.8 × 104
P
P∞ Ma= 4
α = 2.5 c - Comparison with experiment
2
(a)
experiment (Lewis et. 1967)
1
α = 10 deg. X
0 1 2 L
– Application of the triple-deck theory to a laminar corner
flow by Rizzetta, Burggraf and Jenson (1978).
P 2
interacting B.L. model
τw 4
Re = 10
2 6
10
8
limit solution 10
1
(b)
1 interacting B.L. model
4
Re = 10
6
10 0
8
10 limit solution
0
−10 0 X 10 −10 0 X 10
a - Wall pressure distribution b - Wall shear-stress distribution
1.4
separation
P M∞ = 2.7
P∞
RL = 3.9 × 104
1.2 (c)
theory
1
0.2 0.4 0.6
X
The Stewartson and Williams theory of laminar self induced L
separation (Stewartson and Williams, 1969).
Figure 10.6. Comparisons of laminar triple-deck results with experiment and interacting
boundary-layer theory.
10.4 Application to Laminar-Flow Interactions 425
along with the requirement that the vorticity ∂ û /∂ ẑ and, hence, the disturbance
shear stress vanish as ẑ → ∞. Inspection suggests separation of variables with an
x-wise exponential behavior, so we set:
∂ û
(x̂, ẑ) = aeκ x̂ S (ẑ) (10.4.5)
∂ ẑ
d2 s . /
zs = z ≡ κ 1/3 ẑ (10.4.7)
dz2
whose solution that vanishes at z → ∞ is:
s = C Ai (z) (10.4.8)
where C is a constant. Next, we integrate equation (10.4.5) subject to the zero wall-
slip condition to obtain:
z
κ x̂
û (x, z) = ae (C/κ )
1/3
Ai (z) dz (10.4.9)
0
This result, along with equation (10.4.6), now can be applied to the matching
requirement equation (10.4.4) and the so-called wall-compatibility condition that
the momentum equation (10.4.2) be satisfied at z = 0. This yields the following two
equations in C and κ:
0 ∞
−2/3
C = −κ Ai (z) dz (10.4.10)
0
and
where we note that enforcing these relationships serves to couple interactively the
outer and inner decks independently of the disturbance amplitude or cause. Elim-
inating C between equations (10.4.10) and (10.4.11)( and applying the known Airy
∞
derivative function properties Ai (0) = −0.259 and 0 Ai (z) dz = 1/3, we find:
1.6
P̂
1.2
1.0
0.8 ˆτw
Figure 10.7. Typical properties of a numerical
free-interaction solution of the nonlinear laminar
triple-deck equations.
0.4
−8 −6 −4 −2 0 2 4
x
laminar flow:
ω
u u
ˆ (TR/Te )3/2 (Tw /TR)1+ 2
5/4 · (10.4.14a)
L λ 3/8
β 3/4 Re L
1 3
! !
u u
ˆ (TR/Te ) 2 [1− 4 (1−ω)] TR/Te
5/4 (10.4.14b)
L λ 1/8
β 1/4 Re L 2 − 1) Re ]
[(M∞ L
1/4
1 23 41 23 4
A B
where λ = 0.332 for flat-plate boundary layers. Here, we have written the basic
triple-deck result equation (10.4.14a) in the alternative factored form equation
(10.4.14b) to show that only the factor B – which is based on Chapman’s original,
less sophisticated free-interaction analysis assuming an x-scale ∼0 (δ0 ): (see Section
2.7.1) – is needed to successfully correlate laminar-interaction data [59]. The factor
A is evidently a correction for the fact that the interactive scale is slightly longer
by a factor 0(1/ε L). Conversely, because use of the factor B already agrees well
with experiment, we may alternatively view factor A as representing the weakly
Mach- and Reynolds-number-dependent error incurred when the asymptotic the-
ory for ε L → 0 is applied at ordinary Reynolds numbers (Stewartson and Williams
[3]). In any case, equation (10.4.14) predicts that u decreases with increasing Mach
−1/2
number; noting that δ0 /L ∼ Re L , the equation also predicts that the ratio u/δ0
increases slowly with increasing Reynolds number.
where p̂ (x̂) is the universal curve shown in Fig. 10.5. Equation (10.4.15) clearly
suggests that either theoretical results or experimental data for laminar-interaction
pressure distributions should fall on a single universal curve when plotted in the
1/4 1/4
nondimensional form ( p − p∞ ) β 1/2 Re L /CR ρe U0e 2
versus x̂ when x̂ is based on the
factor B of equation (10.4.14b), as discussed previously. As shown in Section 2.7.1,
this indeed is found to be the case over a wide range of Mach and Reynolds numbers
(see Fig. 10.5). Thus, asymptotic theory confirms the similarity laws established by
Chapman’s free-interaction concept.
Numerical solutions of the full nonlinear triple-deck equations show that the
wall shear vanishes in the nonlinear region just upstream of x̂ = 0 (Fig. 10.5) at a
value p̂is = 1.03 (in the case of ramp-generated interactions, this corresponds to a
scaled ramp-angle value of θ̂w = 1.57). Thus, equation (10.4.15) predicts that the
pressure coefficient at incipient-separation scales as:
1/4 −1/4 −1/4
(C p )is = 2.06 CR 2
M∞ −1 Re L (10.4.16)
2
Figure 10.8. Streamline divergence
effect on an adiabatic hypersonic- p
laminar interaction α = 2.5; −, σ = 0;
− − −, σ = 1; −−−, σ = 2; , separation
1
point; , re-attachment point.
Present theory
2.0
1.6
Pw
( pi.s. − p∞ ) /γ p∞ ∼
= C4 χ̄ (10.4.19a)
or
2 −1/4
(C p )is ∼ (C R/Re L)1/4 M∞ −1 (10.4.19b)
( pi.s. − p∞ ) /γ p∞ ∼
= 0.5 C42 (γ + 1) χ̄ (10.4.20)
10
Slope = 1
pi.s. − p∞
p∞
Figure 10.10. Comparison of triple-
deck predictions and experiment
for incipient separation pressure in
hypersonic flow. 1
Slope = 1/2
Pˆw
.1
.01 .1 – 1 10
χ
2.0
ˆ = 2.00
ΘB
1.5
ˆ = 1.57 = Θ
ΘB i.s.
pˆ w
1.0
ˆ =1.00
Θ B
.5
(a) Wall Pressure
0
–10 –5 0 5 20
Figure 10.11. Local interactive heat- x̂ 10 15
transfer predictions for laminar nonadi-
1.00
abatic ramp-generated interactions.
.75
1.00
qw∞
qw
.50 1.57
2.0
ˆ = 2.5
Θ B
.25
corner
ΘB (DEG)
ˆ = .87)
4.6 (Θ
ˆ = 1.44)
7.6 (Θ
qw∞
qw
1 ˆ = 1.87)
10.0 (Θ
87
1.44
M∞ = 9.7 ˆ =1.87
Θ B
Experiment Theory
Figure 10.12. Comparison of experiment and triple-deck theory for hypersonic laminar
interactions.
viscous-interaction effect that occurs downstream of the corner) clearly show how
the adverse pressure-induced reduction in heating grows in direct proportion to
the strength of the interaction, with a sharply peaked minimum at the shock foot
(i.e., the corner) as long as the flow remains unseparated. For the stronger interac-
tions that promote local separation, however, this peaking quickly disappears into a
broader smoothed minimum.
Direct comparison of the theoretical predictions with experimental data may
be made if we account for the added influence of the larger-scale viscous-inviscid
effect caused by the thickening boundary layer downstream of the corner when
M∞ is large (the appropriate theory was given by Stollery [67]). This is shown
in Fig. 10.12, where the local heat-transfer distributions measured by Needham
[68] in a Mach 9.7 interacting corner flow of varying angles are compared with
the theoretical predictions. Regarding the small-scale, triple-deck aspect of the
interaction upstream and immediately at the corner, it is seen that the predicted
deepening of the heat-transfer reduction with increasing interactive strength and
the associated sharply peaked minimum at the shock foot are both clearly con-
firmed by the data. Moreover, the abrupt disappearance of the sharp minimum
peak after separation indicated by the theory is also experimentally corroborated;
indeed, Needham suggested that this feature, in fact, could serve as an incipient-
separation criterion: Triple-deck theory provides firm support for this empirical
observation.
Concerning the interaction zone downstream of the corner, it is shown in
Fig. 10.12 that the locally developing large-scale viscous-interaction effect at this
truly hypersonic Mach number of 9.7 completely overwhelms the wake of the small-
scale, triple-deck structure, producing a very rapid pressure and heat-transfer rise
in rough agreement with experimental data. This behavior is consistent with basic
10.4 Application to Laminar-Flow Interactions 433
1.0
2.0
0.8
1.6 k 0 = 1.180
k 0 = 2.357 (4) 0.6 2.357
1180 (2.3)
1.2 7.410 (10) τw
P [ 17.623 (.20)
∞ 0.4
17.623
∞ ]
0.8
0.4 0.2
0.0 0.0
–8 –4 0 4 8 12 16 20 24 –8 –6 –4 –2 0
x – xs x – xs
(a) Pressure Distribution for (b) Wall Shear Stress
Selected Values of
Θ
Figure 10.13. Effect on the transonic interactions parameter on laminar nonlinear free inter-
actions. (a) Pressure distributions. (b) Wall shear-stress distributions.
2 −1/5 2 −1/5
k0 = M∞ − 1 ε−8/5 = λ−2/5 CR M∞ − 1 Re L (10.4.21)
θw
θw
(a)
x
−10 0 10
0.5
3
0.4
0.3 2
p∼ – p∼∞ p
p∼∞
0.2 (b)
0.1
0 0
−80 −50 0 50 100 120
x∼ (mm)
Figure 10.15. (a) Flared-cylinder configuration. (b) Comparison of laminar triple-deck theory
and experiment (+) for the flared-cylinder problem in supersonic flow.
to ultimately vanish in contrast to the finite inviscid ramp shock value observed in
two-dimensional flow. Allowing for finite Reynolds-number effects and the use of
a linearized axisymmetric flow model of the outer-inviscid deck, this theory pre-
dicts a pressure distribution in reasonably good agreement with experimental data
of LeBlanc and Ginoux [74] (Fig. 10.15b). A subsidiary issue regarding the effect of
the body-radius to boundary-layer-thickness ratio (i.e., transverse-curvature effect)
on the upstream free-interaction behavior was also studied in detail [75].
where the constant Cτ depends only weakly on the Reynolds number. However, in
the upstream-influence region governed by equation (10.5.1), the square-bracketed
term in equation (10.5.4a) is exactly unity; after introducing the present rescaled
variables, we thus obtain:
τw /τw0 = Cτ β p̂ (x̂) (10.5.4b)
in qualitative agreement with equation (10.3.52). Equation
(10.5.4b)
2 further predicts
for incipient separation (τw = −τw0 ) that pis / p0e ∼ C f0 /Cτ M0e /β 1/2 . Because
10.5 Application to Turbulent Interactions 437
1/12 −1/12
typically C f 0 ∼ Reδ∗ [12], pis / p0e thus is nearly constant over a wide range
of practical Reynolds numbers 105 ≤ Reδ∗ ≤ 108 while increasing moderately with
increasing M∞ ; these trends are in agreement with both experiments [76] and simil-
itude analyses [59–61], as discussed in Section 2.7. When alternatively expressed in
terms of the pressure coefficient, the foregoing gives:
1/2 2 1/4
(C p )is ∼ C f0 Me − 1 (10.5.5a)
−1/5
which on use of the classical turbulent boundary-layer value C f0 ∼ Re L yields
−0.25 −1/10
(C p )is ∼ Me2 − 1 Re L (10.5.5b)
This result is in fairly good agreement with Erdos and Pallone’s proposed semi-
empirical correlation [59]:
−0.31 −1/10
(C p )is ∼ Me2 − 1 Re L (10.5.5c)
The available experimental data on very-high-Reynolds-number turbulent
compression-corner interactions [27] suggest that, in fact, this weak Reynolds num-
ber dependence disappears when even small amounts of separation occur.
This discussion strongly suggests that turbulent interaction pressure-
distribution data may be correlated according to free-interaction concepts in
a manner analogous to the laminar case discussed previously. In fact, this was
successfully accomplished by several researchers [59,77], resulting in a univer-
sal nondimensional pressure-distribution curve similar in form to the laminar
distribution (see Fig. 2.53).
The companion inner-deck energy equation (10.3.132) for the upstream region
can be rewritten in a more revealing form by subtracting equation (10.3.131) and
using the fact that ∂ p̂/∂ ẑ ∼
= 0 ; this yields:
" #
∂ Ĥ − (û + p̂) −1 ∂ ∂
∂ û
= Pr T ẑ Ĥ − (û + p̂) + 2 (1 − Prt ) ẑ (10.5.6)
∂ x̂ ∂ ẑ ∂ ẑ ∂ ẑ
which is to be solved subject to the outer-boundary condition that [ Ĥ − (û + p̂)]
vanish as z → ∞ and the inner slip-boundary conditions equations (10.3.51) and
(10.3.53). As in the wall-shear stress, the local heat transfer is known from equation
(10.3.54), so it remains to solve equation (10.5.6) for the enthalpy profile in a manner
similar to that applied to the momentum equation. An interesting special solution
can be realized when Prt = 1, in which case equation (10.5.6) yields the Crocco-type
integral:
Ĥ (x̂, ŷ) = C1 + [û (x̂, ŷ)] + p̂ (10.5.7)
where the outer-boundary condition requires the constant C1 = 0. It is shown from
the inner-boundary conditions that this solution, in fact, pertains to an adiabatic
wall.
pressure-integral term in equation (10.3.136). Thus, with the εT Me2 term neglected,
we have the outer-boundary condition on the inner-deck solution that:
x̂
−2 −1
lim [ûINNER (x̂, ẑ)] = − p̂ − βk ẑ ( p̂ − J p̂) d x̂ (10.5.8)
ẑ→∞ 0+
which implies that the corresponding disturbance vorticity ζ̂ ≡ ∂ û /∂ ẑ decays alge-
braically as follows:
x̂
−2 −2
lim ζ̂ (x̂, ẑ) = −βk ẑ ( p̂ − J p̂) d x̂ (10.5.9)
ẑ→∞ 0+
pω 1.5 (a)
1
−2 0 2 4 6
Figure 10.16. Comparison of asymptotic
∼ ∼
turbulent interaction theory predictions x/δ
with experiment (shown as discrete points)
downstream of a ramp. 1.6
1.2
(b)
τω 0.8
0.4
0
0 2 4 6
∼ ∼
x/δ
(1
where the constant DA ≡ 0+ [( p̂ − p̂s ) / p̃s ] d x̂ is a known integral property of
the small-scale analysis. Because the last limit term vanishes for any finite value of
x̂ > 1, this result reduces to:
lim [û (x̂ > 1, ẑ)] ≡ ûe (x̂) = − p̂s 1 + C A x̂ −1 (10.5.13b)
ẑ→∞
which describes the downstream decay of the inviscid-disturbance velocity ûe toward
its ultimate post-shock value associated with p̂s . Beneath such external behavior,
we then postulate the corresponding inner-deck solution for x̂ > 1 that
û (x̂, ẑ) − p̂s 1 + C A x̂ −1 + ûi (x̂, ẑ) p̂s (10.5.14)
subject to the outer condition that ûi (x̂, ẑ → ∞) = 0. If desired, the solution of
equation (10.5.15) provides the detailed velocity profile across the inner-deck sub-
ject to the x-dependent inner-boundary condition on ẑ = ẑc given by equations
(10.3.51) and (10.5.14). The corresponding wall-shear-stress disturbance is given
independently by equations (10.3.52) and (10.5.12) as:
τw /τw0 −2 p̂s B0−1 1 + C A x̂−1 (x̂ 1) (10.5.16)
440 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
which describe the downstream approach of τw /τw0 to its final negative value asso-
ciated with the full shock-pressure jump. Although strictly valid only for x̂ 1, the
far-field results were found [78] to give good results down to x̂ˆ ≥ 2 ; they are also
in qualitative agreement with a similar downstream analysis by Kluwick and Stross
[32]. An analogous treatment of the energy relations can be conducted to obtain the
x̂ > 1 behavior of the local heat-transfer disturbance along a nonadiabatic wall.
2
χTi ≡ β 2 /(γ + 1) Me4 εT ∼ M0e − 1 / C f0 (10.5.17)
10.5 Application to Turbulent Interactions 441
SHOCK WAVE
SHOCK WAVE
M∞
INNER NONLINEAR
TRANSONIC REGION
WALL LAYER
SONIC LINE
The limiting case χTi 1 corresponds to linearized supersonic flow with the expo-
nential solution (10.5.2). Otherwise, (10.3.92b) is a nonlinear equation for p̂ x̂ /ˆ u
involving the single parameter χT that possesses solutions for p̂ > χTi .
Regarding the downstream region x > 0, we have a more interactive inner deck
subject to boundary condition equation (10.3.92c) that:
!
2 1 − [1 − p̂/χT ]3/2
lim [û (x̂, ẑ)] = − p̂ − βχT (10.5.18)
ẑ→∞ 3 κ 2 ẑ
DIMENSIONLESS INTERACTIONS
y
α α α
x λ
z
δ λ
α
δ
δ
SEMICONE CONE IN CIRCULAR CHANNEL CYLINDER/OBLIQUE SHOCK
α1
λ α2
the shock) or conical (i.e., properties constant along rays from the origin) as one
moves radially outward from the corner [91] (Fig. 10.19). This issue of spanwise-
disturbance propagation is important in the design, interpretation, and correlation
of experimental data; in particular, it is important to understand the far-field behav-
ior of such swept flows and establish the so-called inception distance Li required
to reach an asymptotic conical state (Fig. 10.20a). Inger [90] performed a detailed
analysis of the three-dimensional interactive flow equations and determined that
Li /δ0 ∼ tan in agreement with Settles’s experiments [91] over a wide range of
attached inviscid shock strengths and boundary-layer states (Fig. 10.20b).
in fact, the boundary layer is usually fully turbulent. Conversely, when applied
to lower Reynolds numbers in which the flow is likely to be laminar, the leading
asymptotic approximation of triple-deck theory suffers from significant quantitative
444 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
1. Attached flow
1. Attached flow
Shock
Free-stream Shock
surface flow Free-stream
surface flow
External Shock
flow generator
2. Incipient separation
2. Separated flow Separation
Separation
line line
Free-stream Free-stream Shock
Shock
surface flow surface flow
External
Shock
flow
generator
A
Virtual conical origin
ΔZ
A x
Li
Li
2
5 2 5
i
L mi Lm
4
1
1
Z z 4
3 3
R
C R M S C
M S
the results in fact proved less accurate than the first-order results, and it is better in
practice dealing directly with a numerical solution based on interacting boundary-
layer theory. Nevertheless, we reemphasize that the asymptotic approach remains
of very real value in revealing the correct local velocity and coordinate scales in
the interaction zone to guide purely numerical work and in providing the correct
limiting behavior of the flow-structure in the Re L → ∞ limit.
Viewed in light of contemporary CFD capabilities, another practical difficulty
with even the first-order triple-deck theory is the required numerical solution of
the generally nonlinear partial-differential equations involved: Due to their implicit
elliptic nature and the consequent need to satisfy proper downstream conditions,
such solutions are no less challenging (and perhaps more expensive) than the direct
use of a state-of-the-art Navier-Stokes code while still having the Reynolds-number
limitations. Indeed, once separation occurs, the asymptotic approach requires the
addition of more decks and even smaller local streamwise rescaled regions near
separation and reattachment; without these, instabilities appear in the triple-deck
solutions when significantly reversed flow occurs.
Another practical limitation on the triple-deck solutions is their extended
downstream-wake behavior at x̂ 1 as governed by slow algebraic-fractional
power-law decays [55] in the interactive pressure and skin-friction disturbances.
Although this feature is useful in efficiently implementing the downstream condi-
tions in triple-deck numerical solutions, it proves awkward to accommodate when
seeking to insert a triple-deck solution as a local interactive module in a larger scale
global flowfield prediction code. Moreover, at x̂ ≥ 2, we may well question the par-
allel shear-flow assumption [U0 (y) independent of x] underlying the triple-deck
theory because, in reality, the streamwise development of the baseline flow then
becomes influential. Indeed, at hypersonic speeds, the viscous-interaction effect
associated with the displacement-thickness growth of the baseline flow completely
overwhelms the asymptotic “tail” of the triple-deck solution (see Fig. 10.12).
Finally, a limitation on the far-field–behavior aspect of the triple-deck theory is
noted. As pointed out by Kluwick [33], the upper-deck inviscid-flow equations are
linear if the flow is not transonic or strongly hypersonic; consequently, triple-deck
solutions are not uniformly valid at long distances from the interaction region if the
flow outside the boundary layer is supersonic. Linear theory predicts that the dis-
turbances generated by the interaction process propagate along the unperturbed
Mach lines; however, it is well known that even small disturbances may lead to
significant distortions of Mach lines at long distances from their origin. Thus, to
obtain uniformly valid results, it is necessary to account for nonlinear far-field
effects when seeking to compare the calculated and observed wave patterns in this
field.
except that the numerical solutions of the resulting turbulent triple-deck equations
are obviously even more demanding, especially if we seek the flow details in the
ultra-small-scale (x, y δ0 ) shock vicinity within the boundary layer.
The issue of the accuracy of the εT → 0 assumption underlying the asymptotic-
theory approach arises again in the turbulent case when applying the results to finite
Reynolds numbers 106 ≤ Re L < 108 . Notwithstanding the value of such rigorous
asymptotic analysis in establishing the correct limiting behavior as Re L → ∞ and
the proper local scalings of the interaction field (including those of the very finely
scaled events of the shock penetration deep within the boundary layer), the fact
remains that the use of such asymptotic results at practical Reynolds numbers is an
approximation that yields engineering results often less accurate than those based
on a carefully constructed (albeit formally “nonrational”) nonasymptotic small-
perturbation theory [76]. Excluding hypersonic flows where the near-field inviscid-
disturbance flow cannot be linearized, a nonasymptotic approach offers a valuable
practical alternative in addressing viscous-inviscid interaction problems at realistic
Reynolds numbers.
where the constant B 5 includes the integrated effect of the underlying laminar
sublayer and the no-slip condition. Subtracting the undisturbed flow component
then yields the perturbation relationship between the slip velocity and skin-friction
disturbance that:
u (x, zc ) ∼
= Bi uτ (x) (10.A.2)
where Bi = B + k −1 ln 60. However, from the definition τw ≡ ρw u2τ , we also have the
small-perturbation relationship:
τw (x) ∼
= 2 [uτ (x) /uτ0 ] τw0 (10.A.3)
u (x, zc ) ∼
= Bi [τw (x) /τw0 ] uτ0 /2 (10.A.4)
Appendix A The Wall-Slip Boundary Conditions 447
To provide a second relationship, we apply the Law of the Wall but this time along
the edge of the local boundary layer ze (x) ∼
= δ0 , where dU0 /dy = 0. Neglecting the
perturbation of the wake component [16] (which proves to be a second-order effect)
and using the fact that δ ∼ εT , we obtain:
δ /δ0 (uτ /Uτ0 ) − ue /U0e (10.A.5)
and, hence, from the Log Law expression along y = δ, the first-order perturbation
relationship:
ue (x, ze ) B0 uτ (x) (10.A.6)
where B0 ≡ B + k −1 ln(uτ0 δ0 /υw ) and where a second-order term ∼ (εr /κ) ue /U0e
was neglected. Then, taking ue (x, δ0 ) ∼
= − p (x) /ρ0e u0e from equation (10.3.8) eval-
uated at y = δ with dU0 /dy = 0, equations (10.A.3) and (10.A.6) yield the following
important interactive skin-friction relationship:
τw (x) ∼ 2 p /ρ0e U0e
2
=− (10.A.7)
τw0 B0 εT
Hence, by using equation (10.A.4), the desired effective inner-slip velocity:
u (x, zc ) −U0e (Bi /B0 ) p /ρ0e U0e
2
(10.A.8)
An analogous procedure may be used to obtain the corresponding total-enthalpy
slip needed as an inner-boundary condition on the solution of the energy equation.
We thus conduct a perturbation analysis of the Temperature Law of the Wall given
by:
q̇w 1 zuτ
Tw − T̄ = ln + Bt (10.A.9)
Pr C p uτ kt vR
where kt and Bt are the thermal counterparts [92] to the parameters in equation
(10.A.1). In relating the enthalpy to the temperature, we note that along the inner
cutoff height zc , the velocities are so low that kinetic-energy contribution u2 /2 can
be neglected such that H C p T (this approximation, in fact, is consistent with
the present treatment of the inner-deck compressibility effect for turbulent flow).
Conversely, along the outer edge of the boundary layer, where dH0 /dy = 0 and
the disturbance flow is adiabatic, we have by equation (10.3.11) that H = 0 and so
C p T −U0 u . Examining first the temperature-disturbance field along the inner
boundary z = zc including the attendant heat-transfer disturbance q̇w , equation
(10.A.9) yields the following first-order temperature-perturbation relationship:
T (x, zc ) u q̇
Bti − κt−1 τ − Bti w (10.A.10)
TR uτ0 qw0
where TR ≡ q̇w0 /Pr C p ut0 is a Law of the Wall reference temperature, Bti ≡
Bt + κt−1 ln 60, and uτ /uτ0 is known from equation (10.A.3). Reapplying equation
(10.A.9) to the temperature disturbance Te along the outer edge z = δ0 + δ where
dU0 /dz0 and dH0 /dz0 both vanish, we obtain after using equation (10.A.5) the sec-
ond temperature-perturbation equation that:
Te (x, δ0 ) u u q̇
= Bt0 − 2κt−1 t + κt−1 e − Bt0 w (10.A.11)
TR ut0 u0e qw0
448 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
where Bt0 ≡ Bt + κt−1 ln (uγ 0 δ0 /vw ) and ue /u0e is given by equation (10.A.6). Accom-
panying this is the auxiliary relationship for adiabatic inviscid disturbance flow that:
Te (x, δ0 ) u20e uτ 0 2
=− B0 − (10.A.12)
TR C p TR u0e κ
Inspection shows that equations (10.A.10) and (10.A.11) as well as (10.A.12) pro-
vide a pair of relationships in the two desired unknowns T (x, z) and q̇w ; solving for
them yields the following heat-transfer perturbation:
q̇ Bt0 − 2κt−1 uτ −1 ue u20e uτ0 B0 − 2κ −1
+ κt B0 +
qw 0 Bt0 ut0 ue0 C p TR u0e Bt0
(10.A.13a))
= −λq1 p /ρ0e u20e εT B0 + λq2 εT u20e /C p TR (10.A.13b))
where
κτ Bτ 0 − 2 B0
λq1 ≡ + εT (10.A.13c))
κτ Bτ 0 κτ Bτ 0
κ B0 − 2
λq2 ≡ (10.A.13d))
κ Bτ 0
The corresponding enthalpy slip near the wall is found to be
H (x, zc ) /c p TR = − Bt − κt−1 p /ρ0e u20e εT B0 − Bti (q̇w /qw0 ) (10.A.14)
where r ∼
1/3
= Pr is the recovery factor. Writing M02 = U02 /γ RT02 = Me2 (u0 /ue )2
(T0 /Te ) and using equation (10.B.2) in the integrand, we can thus express Im (zi )
in terms of three positive nondimensional velocity-profile integrals, as follows:
1 1
Tw /Te dη Tw,AD − Tw dη
Im (zi ) = +
Me2 zi /δ0 [u0 /ue (η)]
2 T M
e e
2
zi /δ0 [u0 /ue (η)]
γ −1 zi
− 1+r 1+ (10.B.3)
2 δ0
where η ≡ z/δ0 and Tw,AD /Te = 1 + (γ − 1) r Me2 /2 is the adiabatic wall tempera-
ture. The lower limit of the integrals is very small and understood to lie within the
linear inner part of the velocity profile. In particular, this limit is accurately approx-
imated by Lighthill’s analysis [2] of the viscous-displacement effect in laminar inter-
actions, which gives zi as follows:
1/4
yw,eff ∼ (0.78)3 vw Me2
zi = = 2 = εCL (10.B.4a)
δ0 βU0 (0) M0 (0) δ04
where M0 (0) = (dM0 /dz)w and
⎛ ⎞
ρe Tw μw 4
0.830 μ R ⎜ ⎟
⎜ ρw Te μ R ⎟ C −1/2 Te
CL = ⎝ ⎠ R TR (10.B.4b)
5.2 μe 3/2
βC R δ03
by any convenient Law of the Wall expression for u∗0 (z) outside the laminar sub-
layer, such as a crude but simple power law u∗ = (ue /uτ )(vw z/uτ δ0 )1/N with N 5
to 7, or the more sophisticated Spalding–Kleinstein expression given in White [12].
In the case of very-high-Reynolds-number turbulent interactions in which the
boundary-layer profile is much fuller than the laminar case, the exact value of the
inner limit has little influence on the integral Im ; for example, use of the fractional
power-law velocity profile in equation (10.B.5) shows that the contribution of the
inner limit to the resulting integrals is rather small.
κπ = CR λ−2/5 λ−1/4
1/5
(10.C.9)
−1/8
κu = CR λ−1/5 (Tw /T∞ )1/2 k0
3/10
(10.C.10)
2/5 1/8
κv = CR λ4/5 (Tw /T∞ )3/2 k0 (10.C.11)
where
−1/5 −8/5
k0 = λ−2/5 CR 2
M∞ − 1 εL (γ − 1) M0e
4
(10.C.12)
Appendix D Nomenclature 451
Appendix D Nomenclature
a amplitude factor in triple-deck solution; (equations
(10.4.8), (10.4.9))
At Airy function; (equations (10.4.8), (10.4.9))
Bi , B0 , Bt Law of the Wall parameters
Cf ≡ 2τw /ρ0e U0e2
, skin-friction coefficient
Cp ≡ 2 p /ρ0e U0e , pressure coefficient
2
−1/8
εL = Re L , basic laminar-interaction small asymptotic
parameter
ετ = Uτ0 /U0e , basic turbulent-interaction asymptotic para-
meter
εT perturbation eddy viscosity; equation (10.3.31a))
γ ideal-gas specific-heat ratio
sweepback angle
κ universal nondimensional influence factor; (equations
(10.4.5) and (10.4.6))
λ Blasius flat-plate boundary-layer factor 0.332; (equation
(10.3.100))
μ laminar coefficient of viscosity
μT turbulent eddy-viscosity; (equations (10.3.27) through
(10.3.31b))
v kinematic viscosity μ/ρ
ω laminar viscosity-temperature dependence exponent
(μ ∼ T w )
πR interactive pressure-scaling factor p̂ = p /ρ0e U0e
2
πR
ρ density
θ body-deflection angle
τw wall-shear stress
D.1 Subscripts
AD adiabatic wall conditions
AV averaged value across boundary layer
e boundary-layer edge (y = δ)
0 undisturbed incoming flow
is incipient separation
R reference-temperature–based value; (equation (10.3.93))
S impinging-shock value
w conditions at body surface (“wall”)
REFERENCES
[28] T. C. Adamson and A. Feo. Interaction between a shock wave and a turbulent layer in
transonic flow. SIAM Journal of Applied Mathematics, 29 (1975), 7, 121–44.
[29] R. I. Sykes. An asymptotic theory of incompressible turbulent boundary-layer flow over
a small hump. Journal of Fluid Mechanics, 101 (1980), 647–70.
[30] T. C. Adamson Jr. and A. F. Messiter. Analysis of two-dimensional interactions between
shock waves and boundary layers. Annual Review of Fluid Mechanics, 12 (1980), 103–38.
[31] S. Agrawal and A. F. Messiter. Turbulent boundary layer interaction with a shock wave
at a compression corner. Journal of Fluid Mechanics, 143 (1984), 23–46.
[32] A. Kluwick and N. Stross. Interaction between a weak oblique shock wave and a turbu-
lent boundary layer in purely supersonic flow. Acta Mechanica, 53 (1984), 37–56.
[33] A. Kluwick. “Interacting Laminar and Turbulent Boundary Layers.” In Recent
Advances in Boundary Layer Theory (ed. A. Kluwick) (Springer XXIII, 1998), pp. 232–
330.
[34] R. B. Melnick, R. L. Cusic, and M. J. Siclari. “An Asymptotic Theory of Supersonic
Turbulent Interactions in a Compression Corner.” In Proceedings of IUTAM Sympo-
sium on Turbulent Shear Layer–Shock Wave Interaction (New York: Springer, 1986),
pp. 150–62.
[35] W. C. Rose and D. A. Johnson. Turbulence in shock wave-boundary layer interactions.
AIAA Journal, 13 (1975), 7, 884–9.
[36] R. E. Davis. Perturbed turbulent flow, eddy viscosity and the generation of turbulent
stresses. Journal of Fluid Mechanics, 63 (1974), 4, 674–93.
[37] R. G. Deissler. Evolution of a moderately-short turbulent boundary layer in a severe
pressure gradient. Journal of Fluid Mechanics, 64 (1974), pp. 763–74.
[38] K. Stewartson. “Multistructured Boundary Layers on Flat Plates and Related Bod-
ies.” In Advances in Applied Mechanics, 14. (New York: Academic Press, Inc., 1974),
pp. 145–239.
[39] L. Rosenhead. Laminar Boundary Layers. (Oxford: Clarendon Press, 1963).
[40] F. K. Moore (ed.). Theory of Laminar Flows. (Princeton, NJ: Princeton University.
Press, 1964), pp. 214–22.
[41] M. D. Van Dyke. Perturbation Methods in Fluid Mechanics. (New York: Academic
Press, 1975).
[42] T. Davis and M. J. Werle. “Numerical Methods for Interacting Boundary Layers.” In
Proceedings of the 1976 Heat Transfer and Fluid Mechanics Institute (Stanford, CA:
Stanford University Press, 1976), pp. 317–39.
[43] O. R. Burggraf and P. W. Duck. “Spectral Computation of Triple-Deck Flows.” In
Numerical and Physical Aspects of Aerodynamic Flows (ed. T. Cebeci), (Springer-
Verlag, 1982), pp. 145–58.
[44] M. Van Dyke. The combined supersonic–hypersonic similarity rule. Journal of the Aero-
nautical Sciences, 18 (1957), pp. 499–500.
[45] L. Sirovich and C. Huo. Simple waves and the transonic similarity parameter. AIAA
Journal, 14 (1976), 8, 1125–7.
[46] G. R. Inger. Theory of local heat transfer in shock–laminar boundary layer interactions.
Journal of Thermophysics and Heat Transfer, 12 (1998), 3, 336–42.
[47] R. J. Bodonyi and A. Kluwick. Freely interacting transonic boundary-layer. Physics of
Fluids, 20 (1979), 9, 1432–7.
[48] W. H. Dorrance. In Viscous Hypersonic Flow. (New York: McGraw-Hill, 1968),
pp. 134–9.
[49] K. Stewartson. Theory of Laminar Boundary Layers in Compressible Fluids. (Oxford:
Oxford University Press, 1964).
[50] D. P. Rizzetta, O. R. Burggraf, and R. Jenson. Triple-deck solutions for viscous super-
sonic and hypersonic flow past corners. Journal of Fluid Mechanics, 89 (1978), 535–52.
[51] R. M. Kerimberkov, A. I. Ruban, and I. D. A. Walker. Hypersonic boundary layer sep-
aration on a cold wall. Journal of Fluid Mechanics, 274 (1994), 163–95.
[52] O. R. Burggraf, D. Rizzetta, M. J. Werle, and V. N. Vatsa. Effect of Reynolds number
on laminar separation of a supersonic stream. AIAA Journal, 17 (1979), 4, 336–45.
456 Analytical Treatment of Shock Wave–Boundary-Layer Interactions
[77] P. Carriere, M. Sirieix, and J.-L. Solignac. Propriétés de similitude des phènomènes de
décollement laminaires ou turbulents en écoulement supersonique non uniforme. In Pro-
ceedings of 12th International Congress of Applied Mechanics, (Stanford University,
August 1968) and ONER TP No. 659F (1968).
[78] A. F. Messiter and T. C. Adamson Jr. “A Study of the Interaction of a Normal Shock-
Wave with a Turbulent Boundary-Layer at Transonic Speeds.” In NASA Langley
Research Center Advanced Technology Airfoil Research, 1 (1978), 1, 271–9.
[79] R. E. Melnik. Turbulent interactions on airfoils at transonic speeds: Recent develop-
ments. AGARD CP 291, Paper 10 (1981).
[80] R. E. Melnick and B. Grossman. “Further Developments in an Analysis of the Interac-
tion of a Weak Normal Shock Wave with a Turbulent Boundary Layer.” In Proceedings
of Symposium Transonicum II (Springer-Verlag, 1975), pp. 262–72.
[81] T. C. Adamson Jr., M. S. Liou, and A. F. Messiter. Interaction between a normal shock-
wave and a turbulent boundary-layer at high transonic speeds. NASA-CR-3194 (1980).
[82] A. F. Messiter. Interaction between a normal shock wave and a turbulent boundary layer
at high transonic speeds. Part I: Pressure distribution. Journal of Applied Mathematics
and Physics (ZAMP), 31 (1980), 2, 204–26; see also (with appendices) NASA CR 3194
(1980).
[83] T. C. Adamson and M. S. Liou. Interaction between a normal shock wave and a tur-
bulent boundary layer at high transonic speeds. Part II: Wall shear stress. Journal of
Applied Mathematics and Physics (ZAMP), 31 (1980), 2, 227–46; see also NASA CR
3194 (1980).
[84] G. R. Inger and W. H. Mason. Analytical theory of transonic normal shock–turbulent
boundary-layer interaction. AIAA Journal, 14 (1976), 9, 1266–72.
[85] G. S. Settles and D. S. Dolling. “Swept Shock Wave Boundary Layer Interactions.” In
Tactical Missile Aerodynamics (New York: AIAA Progress in Astronautics and Astro-
nautics, 104, 1986), ch. 8, pp. 297–379.
[86] E. Migotsky and M. V. Morkovin. Three-dimensional shock-wave reflection. Journal
Aeronautical Sciences, 18 (1951), 7, 484–9.
[87] S. I. Gai and S. L. Teh. Interaction between a conical shock wave and a plane turbulent
boundary layer. AIAA Journal, 28 (2000), 7, 804–11.
[88] R. J. Stalker. Sweepback effects in turbulent boundary layer shock wave interaction.
Journal Aeronautical Sciences, 8 (1960), 5, 348–56.
[89] H. Kubota and J. L. Stollery. An experimental study of the interaction between a glanc-
ing shock wave and a turbulent boundary layer. Journal of Fluid Mechanics, 116 (1982),
431–58.
[90] G. R. Inger. Spanwise propagation of upstream influence in conical swept shock–
boundary layer interactions. AIAA Journal, 25 (1987), 2, 287–93.
[91] G. S. Settles. “On the Inception Lengths of Swept Shock Wave–Turbulent Boundary
Layer Interactions.” In Proceedings of IUTAM Symposium on Turbulent Shear Layer–
Shock Wave Interactions (New York: Springer, 1986), pp. 203–15.
[92] J. A. Schetz. Boundary Layer Analysis (Englewood Cliffs, NJ: Prentice Hall, 1993),
p. 433.
Index
Acoustic coupling, 376, 390–391 chalenges to, 82–83, 137–141, 160–161, 167–173,
Active cooling, 261–262 194–195, 242, 245–252, 258, 332–333
Aerodynamic heating, 259–262, 263, 336 Direct Numerical Simulation (DNS), 139,
compression corner, 61–64, 169 159–165, 171, 175–177, 185, 189–191, 195
cone/flare, 283–284, 373 Direct Simulation Monte Carlo method,
cylinder/flare, 272 DSMC, 337, 366–371
cylindrical leading edge, 323–324 hybrid LES/RANS, 141, 185, 191
double cone flow, 274–275, 303–304, 330–331, Large eddy simulation (LES), 140–141, 151,
355–356. See also Double Cone Body 170–173, 175–185, 191, 194–195
double fin, 242–245 local time steppping, 326
hypersonic shock impingement, 317 numerical dissipation, 325–326
oblique shock reflection, 153–156, 189 Reynolds-averaged Navier-Stokes (RANS)
rarefied flat plate, 340 simulations, 139–140, 148–151, 154–156,
rarefied hollow cylinder/flare, 345–349 159–160, 161, 165–166, 168–169, 172,
shock induced separation, 270, 272 173–191, 194–195, 215–225, 235–237, 245,
transpiration cooled surface, 298–300 250–252, 254, 258, 262, 284, 326, 332–333
wege induced separated flow, 267–268, 269 turbulence models, 139–140, 149–151, 154–160,
Analytical methods, 395–453 169, 173–225, 237–245, 250–252, 258
Apollo capsule, 259–260, 304–305 turbulent flow, 326
Cone/flare. See Hypersonic Flow
Bi-conic body. See Double cone body Confinement effect, 114
Bird, Prof. G.A., 337, 366 Critical point theory, 67–74, 203–205, 218–219,
Blasius, 19 247–250, 251, 258
Boundary condition, 416–417, 421–422, 437–438, Crocco’s equation, 24
446–448 Cylinder/flare body, 272–274, 283–284, 303–304,
Boundary layer equation, 28–29, 51 344–349
Boundary layer shape factor, 21–22, 29–30, 56, 97,
101–104, 105, 111, 145–146, 148–149, 397 Diffuser, 95
Boundary layer transition. See Transition Direct Simulation Monte Carlo method. See
Boundary layer velocity profile, 19–24, 29–30, 105 Computational Fluid Dynamics (cfd)
Bow shock wave, 314 Divergent duct, 119. See also Diffuser
Breathing, 373, 387–388 Double cone body, 273–275, 303–304, 327–332,
Buffet, 376–377. See also Transonic airfoils 353–359
Double fin interaction, 204, 237–253
Chemically reacting flow, 329–330, 331, 332, 333, Downstream influence, 376–377
359–363 DSMC. See Computational Fluid Dynamics (cfd)
Coherent vortical structures, 381, 382. See also
Vortical structures Edney classification, 11–16, 42, 43, 48, 59, 117,
Coles, D. E., 20–21 123, 321, 322, 354–356
Compression corner, 1, 26, 35–36, 44–47, 151–185, Entropy, 7, 16–18, 108
191, 263, 317–321, 399–400 Entropy layer, 64–66
Computational fluid dynamics (CFD), 314–315 Equilibrium shape parameter, 113–114
459
460 Index
Experiments, hypersonic, 259, 270–272, 273–274, LENS. See Large Energy National Shock Tunnel
275, 281–283, 290, 293, 298–300, 315, Facility
337–338 Lighthill, H. J., 207, 218, 220, 258
Lock, R. C., 91
Film cooling, 296–298 Logarithmic law, 20
Flare, flow over, 326
Flat plate. See Hypersonic flow Mean free path, 337
Flow control, 77–82, 123–135 Mixing layer, 385–386
boundary layer control, 132–135 Multi-layer structure, 24–25. See also Triple deck
passive control, 79–82, 129–130 theory
suction, 79–81, 129–130, 134–135
surface bump, 81–82, 127–129 Navier-Stokes equations, 138, 139
three-dimensional shock control methods, Non-equilibrium, 338–339
130–132 chemical, 360–363
vortex generator, 79, 132–134 molecular and thermodynamic, 342–343, 359
Flow oscillations, 373 rotational, 338, 342–343, 352–353, 358–359,
Free interaction theory, 51–55, 61, 103, 106, 107, 369–370
186, 192, 210–212, 221, 233, 234, 258, 263, vibrational, 329
264–266, 267, 395, 423–425 Normalised frequency, 390–391
Free molecular flow, 337 Normal shock wave, 26–27, 36, 47, 48, 50, 95–123,
Frequency response, 374, 382–391 142–151. See also Transonic interactions
supersonic tongue, 100, 101, 115
Heat transfer. See Aerodynamic heating Numerical simulation, 314, 324–327. See also
High enthalpy flow, 329–332 Computational Fluid Dynamics (CFD)
High Reynolds number behaviour, 397–398
Hollow cylinder/flare, 272, 345–349 Oblique shock, 7, 27
Hypersonic flow, 59–67, 259–310, 314–333, Oblique shock reflection, 26, 31–35, 39–44, 166,
336–337, 338–349, 353–363 185–193, 398–399
characteristics, 314 Oswatitsch, K., 16, 91
cone/flare, 281–282, 283, 284
flat plate, 339–340 Particle simulation methods, 338, 366
Peak heating rates, 279
Incipient separation, 55–56, 104–107, 144, Pearcey, H. H., 91–92, 144, 145
145–148, 149, 161–166, 169, 186, 207, 234, Pressure distribution, 340
264, 265, 276–277, 377–378, 391, 395, compression corner, 35, 47, 59, 61, 64, 318–321
427–428, 430 cylindrical leading edge, 323–324
Indented nose, hypersonic, 289–292 double cone flow, 356
Intake buzz, 77 chemically reacting flow, 361–362
Intermittency, 377, 388 double fin, 242–245
hollow cylinder flare, 270–275, 345–348
Jet interaction, 260–261 hypersonic shock impingment, 317
normal shock, 98–99, 108, 149, 150–151
Kinetic theory of gases, 337, 365–366 oblique shock reflection, 32, 40, 147, 187–188,
Chapman-Enskog expansion, 365 195
molecular velocity distribution function, 365 rarefied flat plate, 340
Knudsen number, Kn, 337 semi-cone, 226–227
Korkegi, R. H., 207, 213, 214, 234, 235 sharp fin, 209, 217–218, 225–226, 227, 258
Lambda shock, 47–51, 108–110, 117, 124–126, 148, Ramp. See Compression corner
150, 210, 214, 216, 217, 218, 225, 226, 227, Rankine-Hugoniot Equations, 7–8, 83–85, 142,
245, 247 143
Laminar interaction, 262–275, 315–317 Rapid distortion approximation, 380–381
Large eddy simulation (LES), 83, 137, 138, 141, Rarefaction effects, 338–343
170, 194, 253, 254 on chemisrty, 360–363
Large Energy National Shock Tunnel facility, 305, on double cone body, 356–359
306, 328, 331, 332 on hollow cylinder/flare, 347–349
Law of the wall, 19 on shock wave structure, 340–342
Leading edge interaction, 259–260, 292–295, 296, Real gas effects, 66–67, 259, 300–308, 331–332
321–324, 354–355 Recovery temperature, 22, 30, 60
Index 461
Follow Us On
For Suggestion & Query
Email : [email protected]