Flow Turbulence Combust (2011) 87:165–187

DOI 10.1007/s10494-011-9336-1

Development of an Intermittency Equation

for the Modeling of the Supersonic/Hypersonic
Boundary Layer Flow Transition

L. Wang · Song Fu

Received: 12 February 2010 / Accepted: 4 February 2011 / Published online: 1 March 2011
© Springer Science+Business Media B.V. 2011

Abstract An intermittency transport equation is developed in this study to model

the laminar-turbulence boundary layer transition at supersonic and hypersonic
conditions. The model takes into account the effects of different instability modes
associated with the variations in Mach numbers. The model equation is based on
the intermittency factor γ concept and couples with the well-known SST k–ω eddy-
viscosity model in the solution procedures. The particular features of the present
model approach are that: (1) the fluctuating kinetic energy k includes the non-
turbulent, as well as turbulent fluctuations; (2) the proposed transport equation for
the intermittency factor γ triggers the transition onset through a source term; (3)
through the introduction of a new length scale normal to wall, the present model
employs the local variables only avoiding the use of the integral parameters, like
the boundary layer thickness δ, which are often cost-ineffective with the modern
CFD methods; (4) in the fully turbulent region, the model retreats to SST model.
This model is validated with a number of available experiments on boundary layer
transition including the incompressible, supersonic and hypersonic flows past flat
plates, straight/flared cones at zero incidences, etc. It is demonstrated that the present
model can be successfully applied to the engineering calculations of a variety of
aerodynamic flow transition with a reasonably wide range of Mach numbers.

Keywords Supersonic boundary layer · Transition · Turbulence model ·

Intermittency factor

L. Wang
ISTA, Technische Universitaet Berlin, Mueller-Breslau-Str. 8, 10623,
Berlin, Germany

S. Fu (B)
School of Aerospace, Tsinghua University, Beijing, 100084, China
e-mail: [email protected]
166 Flow Turbulence Combust (2011) 87:165–187

Nomenclature

a local sound speed

Cf skin friction
cr phase velocity of disturbances
d wall distance, m
FSTI freestream turbulence intensity, FSTI= (2k/3)0.5 /U ∞
k turbulent kinetic energy
Mae Mach number at the boundary-layer edge
n turbulent spot formation rate
Pk production term of k equation
Pr Prandtl number
Rex Reynolds number based on arc length from the leading edge and local
conditions at the boundary-layer edge
Sij mean strain tensor
Te edge temperature
xt transition onset location, m
Ue edge velocity
γ intermittency factor
ζ length scale in transition flows
μef f effective viscosity
σ spot propagation parameter
τnt characteristic timescale in the flow transition

1 Introduction

Wall-bounded laminar-turbulent transition prediction plays an important role in

the design of high-speed flight vehicles, especially for the thermal protection and
propulsion systems, as in the transitional region at the vehicle surface the heat flux
and skin friction increase rapidly and attain the values even higher than those in the
fully turbulent region. With a low background noise level at low speed conditions,
e.g., incompressible boundary layer flows, such transition process, called natural
transition, is first triggered by the Tollmien–Schlichting (T–S) instability when the
flow Reynolds number Re reaches certain value. The exponential growth of unstable
waves then causes the flow undergoes a non-linear breakdown to turbulence [1].
In high-speed flows, like the compressible boundary layer flows, transition takes
place not only as a function of the Reynolds number, the compressibility or the
Mach number Ma also plays an important role to the flow transition. Here, the
high-frequency Mack modes appear as the extra solutions of the stability equation
because the phase velocity of the corresponding disturbances is supersonic with
respect to the near-wall region. These disturbances are not T–S waves (the first-mode
disturbances) by character or behavior and represent acoustical waves that reflect
inviscidly between the solid wall and the relative sonic line in the boundary layer [2].
Moreover, the mechanisms are more complex for natural transition in 3-D boundary-
layer flows, where exits the crossflow instability caused by the crossflow velocity
Flow Turbulence Combust (2011) 87:165–187 167

perpendicular to the local inviscid streamline that is generated by the combination of

curvilinear inviscid streamlines and the viscous no-slip condition at the wall [35]. The
different modes of instability obviously lead to different paths to the flow transition
to turbulence.
In the numerical prediction of the aerodynamic transitional flows, the approaches
based on Direct Numerical Simulation (DNS), Large-Eddy Simulation (LES) or sta-
bility theories, e.g., Parabolized Stability Equations (PSE) are becoming increasingly
popular due to the rapid development of computer techniques. However, model
predictions, such as the e N method based on the linear stability theory (LST) of
parallel flows and the Reynolds-averaged Navier–Stokes (RANS) model, are still
the effective approaches for engineering applications.
The semi-empirical e N method, in which e N represents the ratio between the
disturbance amplification at the transition onset and the neutral point, is perhaps the
most frequently used for the transition prediction in aerospace industry. However,
this method has difficulties with the non-linear transition to turbulence process
and the N factors for the natural transition and cross-flow transition are often
different. For this reason, a two N factor e N method coupled to RANS solvers
was developed recently [3] to calculate the flow transition on an inclined prolate
spheroid. This model can give satisfactory results but its iterative computation cycle
is too complicated to be compatible with the current computational fluid dynamics
(CFD) code. Moreover, the e N computations require highly accurate mean-flow
computations complete with accurate second derivatives, which have difficulties in
the high-speed flow cases [44]. On the other hand, the RANS approach in modeling
the flow transition has the advantage of being compatible with the modern advanced
CFD methods with low cost of computation. In fact, modeling the transitional flow
with the RANS approach has received significant research attention in recent years.
It is now even available in commercial software [4].
It is known that the low Reynolds number models, with the aid of damping
functions to characterize near-wall viscous effects on turbulence, have certain ability
to predict the transition, especially bypass transition that is forced by high freestream
turbulence level or large isolated roughness elements, as often exits in turbomachin-
ery flows. Essentially, this ability is the consequence of the mathematical properties
of the turbulence model equations, as analyzed by Wilcox [5] with a k–ω eddy
viscosity turbulence model. Such model ability is not based on the intrinsic physics
related to transition as in the development and calibration of the turbulence models,
as the laminar flow behavior was not taken into account.
Savill [6] reviewed the early development of low Reynolds number turbulence
models for the simulation of bypass transition. A recent study by Lardeau et al.
[7] validated the non-linear eddy viscosity models for bypass transition prediction.
Reynolds stress models were also developed to simulate the bypass and separation-
induced transitional flows [8, 9]. These later models provided better results than
the linear ones in incompressible flows. But reliable transition prediction for var-
ious combinations of Reynolds number, free stream turbulence level and pressure
gradient is still a difficult task, not to mention the effects of Mach number. The
TRANSPRETURB European network realized that a turbulence model without
making use of the intermittency proves to be very delicate and often extremely
unreliable in the prediction of transition (http://transition.imse.unige.it/).
168 Flow Turbulence Combust (2011) 87:165–187

The intermittency factor γ was introduced initially to measure the probability of

the time the flow is turbulent in a given spatial point in the transitional region. It
was observed that the behavior of the intermittency factor was rather universal in
an attached boundary layer within a large range of free-stream Reynolds number
and Mach number [10]. By treating the global flow as a linear combination of a
laminar flow and a turbulent one according to the respective weights 1 − γ and γ ,
the parameter γ was later implemented in a number of correlation-based models to
inflict the transition process. Then a more elaborate approach, called the conditional
averaging technique which led to a set of turbulent equations and a set of non-
turbulent ones for mass, momentum and energy conservation, was proposed by
Libby [11] and further refined by Steelant and Dick [12, 13]. However, this type of
model is too complex from an applied point of view.
Since the increase of heat flux and skin friction in the transitional region is closely
linked to the variation in the stress–strain relationship, it is appropriate to introduce
intermittency dependence in the Reynolds-stress constitutive relations. For the eddy-
viscosity models this can be achieved through the modification on the effective eddy
viscosity μe f f as the Reynolds stress tensor is presumed to be 2μef f Sij, where Sij is the
mean strain tensor (e.g. Rodi [14], Byggstoyl and Köllmann [15], and Cho and Chung
[16]). For simplification, it was often accepted to treat μef f as the multiplication of
the intermittency factor γ and the eddy viscosity μt [17], i.e. γ μt . This however led
to a heavy loss of flow physics prior to transition, the model applicability was thus
limited to a narrow range. To avoid this shortcoming, a series proposals by Hassan
and coworkers (e.g. Warren and Hassan [18], McDaniel et al. [19]) and further
developed by Papp et al. [20], took into account of the contribution of non-turbulent
fluctuation to μef f . Consequently, these models were able to be extended to calculate
the transition in supersonic or highly 3D flows. Despite the success it should be
pointed out that these models are appropriate only for the situations where the first
mode instability is dominant. This is however not the case when Ma is significantly
large.
The earliest widely accepted empirical correlation to describe the stream-wise
evolution of γ in the transition region was proposed by Dhawan and Narasimha [10].
Then, more complex algebraic equations accounting for physical phenomena such as
the spreading of turbulent spots and the relaminarization (e.g. Schulte and Hodson
[21]) were deduced. Further developments were dynamic equations for intermittency
so that γ was not prescribed as a constant over a stream-wise profile, for example,
Menter et al. [4, 22, 23], Langtry et al. [24], Suzen et al. [25], Pecnik et al. [26], Steelant
and Dick [12, 13], Suzen and Huang [27], etc. This technique has also been extended
to simulate the unsteady flow transition [28]. These models require a separate
transition onset criteria which typically correlate the transitional Reynolds number
to local free-stream conditions such as turbulence intensity, pressure gradient and
so on. Though these correlations are not difficult to calibrate and can be modulated
according to different transition mechanisms, they contained non-local variables, like
boundary layer thickness. The use of non-local variables is somehow undesirable to
general CFD methods based on unstructured grids for massive parallel execution.
The models based on local variables are thus much preferred for the purpose of
application.
A successful example was the work of Menter et al. [4], which had now been
incorporated into a commercial software package. In this model, the value of the
Flow Turbulence Combust (2011) 87:165–187 169

transitional momentum thickness Reynolds number, which was used to determine

the transition onset with an algebraic criterion, was now obtained with a new trans-
port equation instead of being calculated from the integration over the boundary
layer. A different type of local formulation related to triggering transition was based
on the concept of laminar fluctuation energy, k L , introduced by Mayle and Schulz
[29]. A one-equation or two-equation model was used to describe such ‘turbulent-
like’ fluctuations that were very different from true turbulent fluctuations illustrated
both in experiment [30] and analysis [31] in the pre-transitional region. Recent
examples of such models were established by Walters and Leylek [32], Lardeau et al.
[7], Volino [33], etc.
It has been demonstrated that the existing local-variable-based models can pro-
vide transition prediction in subsonic and two-dimensional boundary layer flows
reasonably satisfactory, including some more complex applications. However, their
performance in supersonic flows or cross-flow transition require significant further
modeling efforts. One reason is that there is a lack of the fundamental physical
phenomena responsible for the compressibility effect in the transition process in
these models. Although easily modified for different transition mechanisms from a
numerical standpoint, they are limited to a narrow range of application. For example,
it is well known that the flow instability modes can be rather different in supersonic
boundary layers from that in incompressible flows, which were not dealt in different
ways in those models.
The purpose of this investigation is to develop an improved model applicable
to natural transition in subsonic as well as supersonic flows. In the present model,
different modes of instability are considered with the aid of flow stability the-
ory. Two independent statistic parameters, the non-turbulent kinetic energy and
the intermittency factor, are introduced here and their transport equations are
established. The non-turbulent and turbulent kinetic energy constitute the whole
fluctuating kinetic energy k. The transport equation for k here is based on that in
the well-known SST model [34] but modified to cover the physics related to the non-
turbulent fluctuation. Respective modification is also made in the transport equation
for the specific dissipation ω in the SST model. Thus, a nominal k–ω–γ model is
developed here which converts to the standard SST model in the fully turbulent flow
region. Moreover, by introducing a new length scale normal to the wall, this model
becomes strictly based on local variables which are compatible with the modern CFD
methods.

2 Mathematical Basis for the Transitional Flows

The present approach is based on the framework of Reynolds-averaged Navier–

Stokes (RANS) equations. In particular, the Favre-averaging approach is adopted
to account for the density variations. It is important to be clear that flow transition,
as well as flow instability, takes place not in a random manner. It occurs when the
Reynolds number Re reaches a certain threshold value. The best example goes back
to the experiment of Osborne Reynolds’ who observed that transition takes place
in a pipe when Re reach roughly the value of 2300. Transition is thus a statistically
deterministic behavior, hence it can be modeled with the RANS approach.
170 Flow Turbulence Combust (2011) 87:165–187

In the Favre RANS approach, the continuity, momentum and energy conservation
equations are thus written as
∂ ρ̄ ∂ ρ̄ Ũ i
+ =0 (1)
∂t ∂ xi

∂ ρ̄ Ũ i ∂ ρ̄ Ũ i Ũ j ∂P̄ ∂ σ̄ij ∂ ρ̄ u
 
i uj
+ =− + − (2)
∂t ∂ xj ∂ xi ∂ xj ∂ xj
  
∂   ∂  P̄ ∂

ρ̄ E + Ũ jρ̄ E + = σ̄ijŨ i + σ̄ijui + ui σij
∂t ∂ xj ρ̄ ∂ xj
⎛ ⎞
∂ ⎝ ρ̄ 
u   
u u
q̄ j + ρ̄c p u ⎠
  i i j
− i T + ρ̄τijŨ i +
∂ xj 2

(3)
The viscous stress tensor is
   
1 1 ∂ Ũ i ∂ Ũ j
σ̄ij  2μ̄ S̃ij − S̃kk δij , S̃ij = + (4)
3 2 ∂ xj ∂ xi

and the molecular heat flux is represented as

C p μ̄ ∂ T̃
q̄ j  − (5)
Pr ∂ xj
where C p stands for the specific heat at constant pressure and Pr is the molecular
Prandtl number. The density, pressure and temperature are of course related by the
equation of state that gives the additional relation needed for the closure of the above
equation set.
It should be noted here that in the above Eqs. 1, 2 and 3, a physical quantity,
say the velocity component U i , had decomposed into a Favre-averaging part Ũ i ,
represented by the wavy overbar, and a Favre-fluctuating part ui , represented
by the double prime,. The fluctuation contains non-turbulent as well as turbulent
components. The velocity correlation u  
i u j is still called Reynolds stress as in the
modeling of the full turbulent flows although laminar or non-turbulent fluctuations
are also considered here.
  
The Eqs. 2 and 3 contain a number of unknowns, such as u      
i u j , ui , ui σij , ui ui u j

and u
       
i T , which require closure models. While the terms ui , ui σij and ui ui u j are not
expected to play any major roles in flow transition, the knowledge of the second-
order correlations u   
 
i u j and ui T is important to bring out turbulent and non-
turbulent, including instability modes, effects in the respective transport equations.
In the present work, the Boussinesq hypothesis on the second-order correlations is
adopted in the same manner as in the conventional eddy-viscosity models. Thus, for
the Reynolds stress, the following constitutive relation holds:
 
   1 2
ui u j = −2νe f f S̃ij − S̃kk δij + δijk (6)
3 3
Flow Turbulence Combust (2011) 87:165–187 171


Here, k = 0.5u  
i ui represents the total fluctuating kinetic energy, and νef f = μef f ρ̄
is the effective dynamic viscosity representing the increased viscous effect due to
the presence of the non-turbulent and turbulent fluctuations. The success of the
transition prediction with the RANS approach thus relies on the quality of μef f
model which will be discussed in detail in the next section.
By analogy to the Reynolds stress relation, the correlation u  
i T takes a similar
diffusivity model, i.e.

νe f f ∂ T̃
u
 
iT =− (7)
Prt ∂ xj

With the relations (6) and (7) the momentum and energy equations can now take the
forms:
  
∂ ρ̄ Ũ i ∂ ρ̄ Ũ i Ũ j ∂P̄ ∂   1
+ =− + 2 μ̄ + μef f S̃ij − S̃kk δij (8)
∂t ∂ xj ∂ xi ∂ xj 3

     
∂   ∂  + P̄ ∂   1
ρ̄ E + Ũ jρ̄ E = 2Ũ i μ̄ + μef f S̃ij − S̃kk δij
∂t ∂ xj ρ̄ ∂ xj 3
  
μ̄ μef f ∂ T̃
+C p + (9)
Pr Prt ∂ xj

Nominally, these equations take the same forms as in the conventional eddy-viscosity
models that offer the advantage of readiness of adaptation of the present transition
model to the existing CFD codes.

3 Modeling of the Transport Equations

3.1 Modeling of the effective viscosity

As the effective viscosity μef f plays a dominant role in predicting the flow transition
with the above RANS modeled momentum and energy transport equations, it is
important that μe f f correctly reflects the delicate variations in the physics of flow
transition. The most apparent behaviour that μef f should have is the switching to the
conventional eddy viscosity μt when the flow goes fully turbulent after the transition.
Before the transition, μef f should reflect the effective viscosity caused by the laminar
or non-turbulent fluctuations in the flow. A simple but effective relation serving these
purposes is [18]:

μe f f = (1 − γ ) μnt + γ μt (10)

where the subscript nt denotes to the non-turbulent part in the effective viscosity and
γ is the intermittency factor bridging the non-turbulent and turbulent contributions.
Modeling of the turbulent eddy viscosity μt can now be seen as well established,
for the present work adopts the SST model. But other eddy-viscosity models are
readily acceptable to the present framework. Here, attention will focus on the
172 Flow Turbulence Combust (2011) 87:165–187

modeling of the non-turbulent fluctuation. In fact, the practice of Warren and Hassan
[18] is adapted that reads

μnt = Cμ ρ̄kτnt (11)

where Cμ is the model coefficient, k the total fluctuating kinetic energy and τ nt
represents the characteristic timescale in the flow transition, which is strongly
associated with the instability modes [2].

3.2 Modeling of the total fluctuating kinetic energy equation

Since the present approach treats the full turbulent flows with the two-equation
SST k–ω model, the transport equations for k and ω are thus modified to cover
the transitional region. Considering the total fluctuating kinetic energy first, it now
contains two parts: one relating to the non-turbulent fluctuation k L and one to the
full turbulence kT , i.e., k = (1 − γ )k L + γ kT . In the existing study, a separate
transport equation for non-turbulent kinetic energy, k L , had been proposed [29]
though in the same manner as the conventional form of the transport equation for
the turbulence energy, kT . In the present work, the employment of two separate
equations for kT and k L is found not necessary. A more cost-effective approach is to
develop a single equation for the total fluctuating kinetic energy k.
In fact, without invoking the nature of fluctuations, the transport equation for the
fluctuating kinetic energy, whether turbulent or not, has the following analytic form:
 
∂ (ρk) ∂ ρujk   ∂ Ũ i
+ = Dk −u i uj −ε (12)
∂t ∂ xj ∂ xj
  
Pk

where Dk , Pk and ε represent the energy diffusion, production and dissipation terms,
respectively. With the constitutive stress–strain relations for the Reynolds-stress u 
i uj
(6), the energy production term Pk is thus expressed as
 
1 2
Pk = 2μef f S̃ij S̃ij − S̃kk S̃ll + S̃kk ρ̄k (13)
3 3
while the dissipation rate will be calculated through an additional transport equation
but with the SST model approach, the diffusion term of k equation is expressed as
  
∂ μe f f ∂k
Dk = μ+ (14)
∂ xj σk ∂ xj
where the Prandtl number σ k takes the value of 1.

3.3 Modeling of the timescales in the transition

3.3.1 The proposal of a length scale for transition

With the total fluctuating kinetic energy known in the expression (11), model-
ing of the timescale becomes the crucial element in the non-turbulent viscosity
coefficient μnt . According to experimental correlations and theoretical analysis [35],
the timescale mentioned above would involve the boundary layer thickness which,
Flow Turbulence Combust (2011) 87:165–187 173

regarded as non-local, is calculated through the integration over the boundary

layer. In modern CFD methods the unstructured grids and parallel computations
are commonly applied, the use of the non-local parameters is highly undesirable
and very inefficient. There is a tendency in recently transition research to develop
length scales based on local properties. In this study a length scale ζ normal to wall,
with

ζ = d2  (2Eu )0.5 (15)

is defined to characterize the transition. Here, d is the distance to wall,  the absolute
value of the mean vorticity, and Eu (= 0.5*|U|2 ) stands for the kinetic energy of
the mean flow related to the wall. Derivation of this length scale comes from the
relationship in Blasius boundary layer, which is firstly used by Wilcox [36] as
 2   
ρd  ρUθ
Reν,max = = 2.193 = 2.193Reθ (16)
μ max μ
where U stands for the local mean velocity, μ is the molecular viscosity, and θ
represents the moment thickness of boundary layer. Menter et al. [4] also employed
Eq. 16 for their local-variable-based model.
The turbulence length scale l T takes the conventional definition as k1.5 /ε or
k0.5 /(β ∗ ω). And the bound of length scale, l B = k0.5 / (Cμ |S|), is used to avoid the
stagnation point anomaly [37]. The effective length scale for transition, ζ e f f , is then
set as the minimum value among ζ , l T and C1 l B , where C1 is a model constant.

3.3.2 The time scales of instability modes

From the linear stability theory (LST) and experimental observations it is known
that at low Mach number the so-called first-mode disturbance, i.e. two-dimensional
Tollmien–Schlichting waves in incompressible flows, is the primary cause of insta-
bility, the effect of the second-mode disturbances becomes prominent at high-Ma
flows however [2]. This mode variation to the flow instability, related to the effect
of compressibility, suggests that the transition in a boundary layer exhibits different
mechanisms. In the present model the effect of the compressibility on the instability
modes is accommodated inherently in the modeling through the introduction of the
local relative Mach number Mrel which is employed to determine the nature of the
instability modes. Mrel is defined as

Mrel = (U − cr ) a (17)

where a is the local sound speed, and cr represents the phase velocity of disturbances
with the same value for all Mack modes. The timescale τ nt is determined with
reference to Mrel , that is

τnt1 , |Mrel | ≤ 1;
τnt = (18)
τnt1 + τnt2 , |Mrel | > 1.
where τ nt1 and τ nt2 represent the timescales of the first-mode and second-mode
disturbances, respectively.
There had been a number of studies on the transition timescales relating to
the instability modes employing non-local parameters (e.g. Warren and Hassan
[18]). According to the experimental correlation of frequency of the first-mode
174 Flow Turbulence Combust (2011) 87:165–187

disturbances with the maximum amplification rate [38], its characteristic timescale,
τ nt1 , can thus be set as
 0.5
τnt1 = C2 · ζef1.5f (2Eu )0.5 ν (19)

where ν is the molecular kinetic viscosity. Also, it was observed that the wavelength
of the second-mode fluctuation is approximately twice as the boundary layer thick-
ness, and their specified phase velocity cr is equal to the local mean velocity at the
generalized inflection point ys [2], i.e., where
 
d dU
ρ =0 (20)
dy dy
Thus the timescale for the second-mode dominated transition at high-Ma flow, τ nt2 ,
is chosen as

τnt2 = C3 · 2ζef f U(ys ) (21)

It will be seen soon that the use of the present timescales is effective in reflecting
mode effect in the compressible boundary layer transition.

3.4 Transport equation for intermittency factor γ

Proposing a transport equation for the intermittency factor γ has now been accepted
in the RANS transition modeling groups [39]. While the existing models have been
reasonably successful in the low-Ma transitional flows, it is necessary to propose a
new one as part of the model package in the present work. The transport equation
for γ is given here as
    
∂ (ργ ) ∂ ρujγ ∂ μef f ∂γ
+ = μ+ + Pγ − εγ (22)
∂t ∂ xj ∂ xj σγ ∂ xj
where Pγ and ε γ represent the intermittency production and dissipation term,
respectively. These two terms require further modeling study.
Following the work of Dhawan and Narasimha [10], the streamwise development
of γ in the transition region appears to be quite universal in an attached boundary
layer within a large range of free-stream Reynolds number and Mach number and
can be described with the following well-known empirical correlation
 
γ = 1 − exp − (x − xt )2 nσ/U e (x > xt ) (23)

where the subscript t stands for the transition onset, n is the turbulent spot formation
rate and σ represents the spot propagation parameter. In an attempt to propose the
generation term Pγ , the practice of Lodefier et al. [28] is worthwhile taking that leads
to the derivation of Eq. 23 as

dγ nσ  0.5
=2 (1 − γ ) − ln (1 − γ ) (24)
dx Ue
The dimensionless spot formation rate n̂ = nνe2 /U e3 is then introduced. According to
the empirical correlation from the flatplate experimental data [40], in the transition
region, n̂σ is affected by the freestream turbulence intensity (FSTI), i.e. (2k/3)0.5 /U ∞ .
Flow Turbulence Combust (2011) 87:165–187 175

Table 1 Model constants Cμ C1 C2 C3 C4 C5 C6 σγ

0.09 0.7 0.35 0.005 8e-5 0.07 1.2 1.0

Furthermore, based on the dimension analysis, (Pγ − εγ ) ∼ U(dγ /dx). Pγ and ε γ are
thus modeled as follows:
 0.5
 d
Pγ = C4 ρ Fonset − ln (1 − γ ) 1 + C5 2Ek u |∇ Eu | , εγ = γ Pγ (25)
ν
where the function Fonset is set to trigger the onset of transition and is given by
 
ζef f k0.5 |∇k|
Fonset = 1.0 − exp −C6 (26)
ν |∇ Eu |

3.5 Final package of the present model

The present model package includes the transport equations for the fluctuating
kinetic energy k, the specific dissipation rate ω and the intermittency factor γ . They
can now be summarized as follows:
    
∂ (ρk) ∂ ρujk ∂ μef f ∂k  
+ = μ+ + Pk μe f f − ε
∂t ∂ xj ∂ xj σk ∂ xj
    
∂ (ρω) ∂ ρujω ∂ μef f ∂ω
+ = μ+ + Pω − Dω + Cdω
∂t ∂ xj ∂ xj σω ∂ xj
    
∂ (ργ ) ∂ ρujγ ∂ μef f ∂γ
+ = μ+ + Pγ (Fonset ) − εγ (27)
∂t ∂ xj ∂ xj σγ ∂ xj
These equations convert to the standard SST model [34] in the fully turbulent region
(γ = 1), with the compressibility corrections for the k equation [41, 42], as detailed
in the Appendix.
In summary, in the present model γ is set as a weighting function between the non-
turbulent and the turbulent components of the effective viscosity μe f f , i.e. Eq. 10, in
Pk and Pω , the production terms of equations for k and ω, respectively. In the pre-
transitional region, μef f ≈ μnt , different instability modes dominate the development
of k. The developments of k and the mean flow then determine the value of the
transition-onset trigger Fonset , i.e. Eq. 26, in the production term of γ equation. Fonset
rapidly goes from zero to unity after the onset point. In the transitional region, since
μnt << μt , flow develops according to the distribution of γ . In the fully turbulent
region, μef f = μt , the model retreats to SST model. All the model constant values
are shown in Table 1.

4 Characteristics of the Present Model in Pre-transition Region

The above proposed intermittency transport equation, and indeed the k–ω transport
Eq. 27, are solved in the entire flow field covering turbulent as well as laminar flow
regions. The model is thus required to first properly resolve the flow in the laminar
region before transition. The capability of the transition onset prediction is also a
crucial indicator of the model quality. Here, the effects of flow instabilities, which
176 Flow Turbulence Combust (2011) 87:165–187

Table 2 Parameters for m −0.05 0 0.1 0.5 1

selected Falkner–Skan profiles
β 0.105 0 0.182 0.667 1
ηδ 3.8 3.48 3.14 2.59 2.38

have great influence in the pre-transitional region, are examined with the aid of
the self-similar solutions of the corresponding laminar flow. The numerical methods
employed to obtain those solutions use a Chebyshev collocation technique for the
spatial discretization, which is included in the Matlab linear algebra package [43].

4.1 Incompressible flow over a flat plate with external pressure gradients

This flow is a good case for illustrating the properties of the length scale proposed
in Eq. 15. The self-similar Falkner–Skan boundary layer can be obtained when the
external potential flow is assumed to vary proportionally to xm , where x is the stream-
wise coordinate, and no-slip conditions are imposed at the wall. Solutions for the
non-dimensional streamfunction f (η) corresponding to this situation is given by the
nonlinear ordinary differential equation as
2m  
f  + f f  + 1 − f 2 = 0 (28)
m+1
where the primes indicate derivatives with respect to the similarity variable

m + 1 Ue
η=y (29)
2 νe x
where the subscript e denotes boundary layer edge value and the boundary condi-
tions are
η = 0 : f  = f = 0 and η → ∞ : f  = 1 (30)
  m
The mean flow thus are described by U = f U e ∼ f x . Different m values con-
sidered here are shown in Table 2, where the Hartree parameter β = 2m/(m + 1).
Essentially, the present situations are related to the attached boundary layer flow in
the stagnation region of a two-dimension wedge of opening angle βπ . β (or m) is
a measure of the acceleration or deceleration of the outer flow. Figure 1 gives the
profile of f  for several m values. It indicates that a point of inflection of velocity

Fig. 1 Variation of 1
Falkner–Skan profiles with
different imposed pressure 0.8
gradients, i.e. m values.
 0.5
η = y (m + 1) U e / (2 ve x) 0.6 m = -0.05
U /U e

m=0
m = 0.1
0.4
m = 0.5
m=1
0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
η
Flow Turbulence Combust (2011) 87:165–187 177

profile exists in the decelerated flow described by m < 0 but not in the Blasius
boundary layer recovered for m = 0 and the accelerated flow obtained for m > 0.
Profiles of the variation of the ratio of the length scale ζ over the boundary layer
thickness δ at different m values across the boundary layer are shown in Fig. 2(a). It
is seen that with increasing m value the profiles approach to a single curve at m = 1.
In fact, when m > 0.5 the variation in the profiles is small. The maximum level of the
ratio also decreases as the m value increase. This is consistent with the flow physics
for the favorable pressure gradient causes reduction in boundary layer thickness.
Figure 2(b) shows the effect of the imposed pressure gradient on the distribution
of the dimensionless group in Falkner–Skan boundary layer as
 
  ∗ m + 1 −0.75 ν 0.75 x0.25   
Pk k nt1 × ηδ1.5 × = e 1.75 Pk k nt1 (31)
2 ρe U e
where the subscript e denotes variables at the outer edge of the boundary-layer, the
subscript nt1 represents the first-mode disturbance, the superscript * denotes non-
dimensionalized variables based on a characteristic length (ν e x/ U e )0.5 and velocity
U e , Pk defined in Eq. 13 is the production term of k equation, and ηδ shown in Table 2
is the η value at the boundary-layer edge. The ratio (Pk /k) represents the generation
rate of the non-turbulent kinetic energy. It illustrates that with the decrease in the
imposed pressure gradient, the maximum of the profile moves outward from the wall
while its average value increases considerably at constant streamwise, x, position,
which enhances the development of k and consequently results in a more upstream
transition onset defined here as xt . Such numerical characteristic agrees qualitatively
with the physical phenomenon that the unit transitional Reynolds number, i.e.
U e xt /ν e , increases in the accelerated flow while decreases in the decelerated flow [44].
It is demonstrated that the present model has basic ability to represent the physical
phenomenon mentioned above.

4.2 Compressible flow over a flat plate with an adiabatic wall

The self-similar compressible boundary layer on an adiabatic flat plate can be ob-
tained in the Illingworth transformed plane [45]. Solutions for the non-dimensional
-0.75

0.4 m = -0.05 m= -0.05

× (0.5m +0.5)

m=0 0.8 m= 0
m = 0.1 m= 0.1
ζ (= d Ω/ | U | ) / δ

m = 0.5 m= 0.5
0.3 0.6
m=1 m= 1
1. 5

0.2 0.4
2

(P k /k )nt 1 × ηδ
*

0.1 0.2

0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.2 0.4 0.6 0.8
d/δ d/δ

(a) (b)
Fig. 2 Effects of the imposed pressure gradient, m value, on distributions of (a) ratio of the length
scale ζ and the boundary layer thickness δ, and (b) dimensionless group involving specific production
rate of k, in Falkner–Skan boundary layer. The subscript nt1 denotes the component of the first-mode
disturbances. ηδ is the η value at the boundary-layer edge
178 Flow Turbulence Combust (2011) 87:165–187

streamfunction f (ξ ) and temperature g(ξ ) are given by the nonlinear ordinary

differential equations as follows
 
ρμ   1 
f + ff = 0 (32)
ρe μe 2

 
ρk  1 ρμ 2
g + Pr fg = −Pr (γ − 1) Ma2e f (33)
ρe ke 2 ρe μe
where Pr is the Prandtl number, the subscript e denotes variables at the boundary-
layer edge, and the primes denote derivatives with respect to the similarity variable

  1/2 y   
ξ = ρe U e μe x ρ ρe dy (34)
0

The boundary conditions for the above ordinary differential equations are
ξ = 0 : f  = f = g = 0 ; ξ → ∞ : f  = g = 1 (35)
The mean flow thus are described by U = f  U e and T = gTe , where U e and Te are
the edge velocity and temperature, respectively. Effects of Mach number on the
profiles of f  and g respectively, are shown in Fig. 3(a) and (b), where the horizontal
coordinate, η, represents the dimensionless distance from the wall with reference
to (U e /ν e x)0.5 . The Mach number is in the range of 1 to 5.91. It indicates that for
large Mach numbers the velocity distribution is approximately linear over much of
the whole boundary-layer thickness, and the increase in the wall temperature causes
significant the increase in the thickness of the boundary layer.
Again, it is interesting and important to know the compressibility effect on the
length scale proposed in Eq. 15. Figure 4(a) shows the variations of the ratio of
the length scale ζ over the boundary layer thickness δ at different freestream Mach
numbers. It is seen that with the increase of Mae , the maximum of the profile
increases and moves outward to the external flow due to the increasing flatness of
shape of the velocity profile. It should be noted that the length scale ζ remains less
than the boundary layer thickness δ for Ma up to about 6. The effects of Mach

1
Mae = 1
6 Mae = 2
0.8 Mae = 3.5
5 Mae = 5
Mae = 5.91
0.6
U /U e

T / Te

4
Mae = 1 Te = 56.2 K
0.4 Mae = 2 3
Mae = 3.5
Mae = 5 2
0.2 Mae = 5.91
T e = 56.2 K
1

5 10 15 20 5 10 15 20
η η

(a) (b)
Fig. 3 Profiles of dimensionless (a) velocity and (b) temperature in compressible laminar boundary
layer on adiabatic flat plate, U e , Te and Mae are the edge velocity, temperature and Mach number,
respectively,Te = 56.2 K, Pr = 0.72, η = y (U e /ν e x)0.5
Flow Turbulence Combust (2011) 87:165–187 179

Mae = 1
Mae = 2
Mae = 3.5 0.8 Te = 56.2 K
Mae = 5
0.6
Mae = 5.91
0.6
ζ /δ

0.4
T e = 56.2 K 0.4
( d / δ )crit
0.2 cr / U e
0.2

0 0
0.5 1 2 4 6 8
d/δ Mae

(a) (b)
Fig. 4 Effects of edge Mach numbers on (a) distribution of ratio of the length scale ζ over
the boundary layer thickness δ in the laminar, adiabatic flat-plate boundary layer and (b) the
dimensionless phase velocity of the two-dimensional disturbance, cr , and the critical position of the
boundary-layer profile, where the absolute value of the local relative Mach number, Mrel , equals 1

number on the enlargement of area of the effective region for the second-mode
disturbances, i.e. the region where the absolute value of the local relative Mach
number defined in Eq. 17 is larger than 1, is illustrated in Fig. 4(b). It reveals that
the enlargement slow down shapely when Mae is up to 5, corresponding to the phase
velocity of the disturbances up to 0.9 times U e , as shown in Fig. 4(b) either.
Figure 5(a) and (b) show the effects of Mach number on the distribution of the
components of the dimensionless group defined as follows

  ∗∗   ∗ L0.5 ηδ1.5 νe0.75 L0.5   

Pk k nt1 Re−0.25 = Pk k nt1 = Pk k nt1 (36)
x
x0.5 ρe U e1.75 x0.25

  ∗∗   ∗ Lηδ Lνe0.5   

Pk k nt2 Re−0.5 = Pk k nt2 = Pk k nt2 (37)
x
x ρe U e1.5 x0.5

Mae = 1
0.12 0.8
Mae = 2
Mae = 3.5 Mae = 3.5
0.1
( P k / k )nt 2 ** × Rex- 0.5

Mae = 5
- 0.25

0.6 Mae = 5.91

× Rex

0.08
T e = 56.2 K
**

0.06 0.4 T e = 56.2 K

(P k / k )nt 1

0.04
0.2
0.02

0 0
0 0.2 0.4 0.6 0.8 1 1.2 0.2 0.4 0.6 0.8 1
d/δ d/δ

(a) (b)
Fig. 5 Profiles of components of the multiplication of dimensionless specific production rate of k,
and local Reynolds number, Rex , in the laminar, adiabatic flat-plate boundary layer for several edge
Mach number. The subscripts nt1 (a) and nt2 (b) of longitudinal coordinates denote the components
of first-mode and second-mode disturbances, respectively. δ stands for the boundary layer thickness
180 Flow Turbulence Combust (2011) 87:165–187

Table 3 Parameters for profiles of compressible laminar boundary layer on adiabatic flat plate
Mae 1 2 3.5 5 5.91
ηδ 5.39 6.79 10.5 15.6 19.4
Cr /U e 0.15 0.46 0.85 0.9 0.93
(d/δ)crit 0 0 0.324 0.45 0.505
ηδ is the value of η at the boundary-layer edge. The subscript crit denotes the critical position in
which the absolute value of the local relative Mach number, Mrel , equals 1

where the ratio (Pk /k)nt1 represents the generation rate of the non-turbulent kinetic
energy corresponding to the first-mode disturbances for the given total fluctuating
energy while (Pk /k)nt2 represents the contribution of the second-mode ones, the su-
perscripts * and ** denote the non-dimensional variables based on the characteristic
velocity U e and the different characteristic lengths (ν e x/ U e ) 0.5 and L(scale of body),
respectively. For the component of the first-mode disturbance, in order to keep ρ e
constant, the calculated x(m) coordinates are selected as 0.2, 0.1 and 0.057 corre-
sponding to Mach numbers 1, 2 and 3.5, respectively. Figure 5 illustrates that with
the increase in Mae (constant of local Reynolds number), the maximum of the profile
moves forward to the external flow while its average value decreases, especially in
the near wall region. For the component of the second-mode disturbance, only high
freestream Mach number case is studied as there is no region that the second-mode
instability can make influence at low Mae , as shown in Table 3. The calculated x(m)
coordinates are selected as 0.1, 0.07 and 0.059 corresponding to the Mae numbers 3.5,
5 and 5.91, respectively. It reveals that the level of maximum (Pk /k)nt1 and (Pk /k)nt2
reduces, possibly to an asymptotic value, as Mae increases.
Effects of Mach number on the ratio of (Pk /k)nt1 over (Pk /k)nt2 are shown in Fig. 6.
The calculated x(m) coordinates are also selected as 0.1, 0.07 and 0.059 corresponding
to the Mae numbers 3.5, 5 and 5.91, respectively. It is seen that this ratio is very small
in the effective region for the second-mode instability but it increases rapidly towards
the outer boundary layer.
In summary, three Mach number regions with different numerical characteristics
of the present model can be distinguished for the adiabatic flat-plate boundary
layer. In the first region with Mae up to about 2.5, only the first-mode disturbances
are of important. The absolute value of Mrel is basically less than 1, thus, the

Fig. 6 Effects of edge Mach 0.7

numbers on distributions of Mae = 3.5

0.6
ratio of (Pk /k) components of Mae = 5
the first-mode and 0.5 Mae = 5.91
(Pk / k)nt 1/ nt 2

second-mode disturbance in
0.4
the compressible, laminar Te = 56.2 K
boundary layer on adiabatic 0.3

flat plate. δ stands for the 0.2

Re x = 5.536E5

boundary layer thickness

0.1

0
0.2 0.4 0.6 0.8 1 1.2
d/δ
Flow Turbulence Combust (2011) 87:165–187 181

second-mode
 disturbances
∗∗ could not be influential. The reduction in the maximum
values of Pk k nt1 as the Mae increases suggests that the transition is hampered to a
further downstream
  ∗∗ location. In the second region with  Mae =∗∗2.5 to 5, the effective
region for Pk k nt2 , whose value is much larger than Pk k nt1 , covers much of the
boundary layer. It leads the freestream transitional Reynolds number, U e xt /ν e , to
decrease sharply, i.e. the transition takes place earlier. In the third region with Mae
beyond 5, the enlargement of  area of the effective region for the second-mode dis-
 ∗∗
turbances slow down while Pk k nt2 decreases gradually. Therefore, the transition
onset moves slowly downstream. Such numerical characteristics match qualitatively
with the results from LST and experimental observations [44]. It is demonstrated
that the present model has the basic ability to represent the physical phenomenon
mentioned above. However, it is noted that the decrease unit transitional Reynolds
number, at Mae 2.5 is actually due to the interaction of three-dimensional first-mode
disturbances which is not considered in the present work. The effect is considered in
the contribution of the second-mode.

5 Results and Discussion

The present model proposal is calibrated and validated for a reasonably wide range
of transitional cases involving the incompressible flows past a flat plate and a wing,
supersonic and hypersonic flows over a flat plate, straight/flared cones at zero
incidences. Except for the last case (see Fig. 13), all the data for comparison come
from the quiet nozzle or the DNS solution.
In the numerical process, the RANS is solved on the non-staggered H-type grid
system. The convection and diffusion terms in RANS are discretized with AUSM+
[46] and central difference scheme respectively. To solve the pressure field, SIMPLE
algorithm is used and in order to eliminate the pressure fluctuation associated with
the use of non-staggered grid system, momentum-interpolation technique is utilized
to calculate the velocity on the face of finite volume.
Figure 7 gives the skin friction for Schubauer and Klebanoff [47] flat-plate test
case with zero pressure gradients. It is seen that the flow transition profile are
well captured with the present and the existing models. The present model predicts
transition onset and length marginally better. Moreover, the calculated mean velocity

Fig. 7 Skin friction coefficient 0.005

(C f ) for the Schubauer and Experiment

Menter's model (2006)
Klebanoff test case 0.004
Present model

0.003
Schubauer & Klebanoff
Cf

flat plate (1956)

0.002
FSTI = 0.03%

0.001

1E+06 2E+06 3E+06 4E+06

Rex
182 Flow Turbulence Combust (2011) 87:165–187

(a) (b) (c) (d)

Fig. 8 Representative mean velocity profiles for the Schubauer and Klebanoff test case

profiles in laminar, transitional and turbulent region agree well with the experiment,
as shown in Fig. 8.
Figure 9 compares the skin friction calculated by the present approach and the e N
method with the experiment data for incompressible flow over a supercritical airfoil
[48], whose configuration results in a fairly constant favorable pressure gradient on
the upper surface. The present model predicts the transition-onset location much
better than does the e N method although it gives slightly higher peak value in the
skin friction coefficient and longer transition region.
Figure 10 compares the measured and computed recovery factor distribution. The
cone’s half-angle is 5◦ and the angle of attack is zero [49]. The free-stream Mach
number is 3.5. There the present model shows very satisfactory agreement with the
experiment. The transition onset and length are all well predicted. However, the peak
value of the calculation is slightly less than the experimental data. The McDaniel–
Nance–Hassan model [19] also exhibits good transition behavior though there is an
overshoot in the recovery factor profile at the end of the transition.
The Mach 4.5 adiabatic flat-plate boundary-layer flow is a good test case for the
model validation because the transition is initiated with both the first-mode and the
second-mode instability. As shown in Fig. 11 the present model predicts an earlier

Fig. 9 Skin friction coefficient Present calculations

(C f ) for top surface of a 0.008 eN method (Mateer et al. 1996)
supercritical airfoil Experimental data

Re c, ∞ = 2.0E6 FSTI = 0.09%

0.006
angle of attack: -0.5o
Cf

0.004

Airfoil experiment of
0.002 Mateer et al. (1996)

0
0.2 0.4 0.6 0.8
x/c
Flow Turbulence Combust (2011) 87:165–187 183

Fig. 10 Comparison of 0.9

Ma∞ = 3.5 o
5 half-angle cone
computed and measured Re∞ = 5.89E7 m
-1
adiabatic wall
recovery factor (r) for an T∞ = 92.3 K 0 angle of attack

Recovery factor
adiabatic straight cone 0.88 FSTI = 0.1%

0.86

Hassan's Model (2000)

0.84 Experiment (chen et al. 1989)
Present model

0.05 0.1 0.15 0.2 0.25 0.3 0.35

Surface distance (m)

start of the transition than the DNS results. The overall variation of the skin-friction
coefficient curve along the streamwise direction is however satisfactory. The reason
for the transition taking place earlier than the DNS result can be attributed to the
over-emphasizing of the second-mode disturbance in this model, as discussed in
Section 4.2.
The hypersonic flow over an adiabatic flared cone with 5◦ half-angle [50] is
the situation where the second-mode disturbance becomes the primary cause of
instability. Using the flared-cone configuration in the experiment is because that
without the pressure gradient, transition would not have been attained within the
limited quiet flow Reynolds number range of the facility, and the flare curvature
also maintains an approximately constant boundary layer thickness over the flare
region, which allows second-mode disturbances of nearly constant frequency to grow.
Figure 12(a) gives the comparison of the computed and measured wall temperature
distribution. The present model gives accurate transition onset but misses the peak
value. The McDaniel–Nance–Hassan model somehow gives too low temperature
level before the transition although the onset location of the transition seems not bad.
It is noted that the present model results seem to miss the peak in the late transition
as seen from the previous figures. It is speculated here that the lack of the interaction
between turbulent spots and surrounding laminar flows in the present model may be
the main source of error.
Since the flared-cone configuration cannot be extended in the calculation, the
results of the fully turbulent region are obtained from the calculation performed
on the straight-cone boundary layer (long enough) with the same inlet condition.

Fig. 11 Comparison of the

skin friction coefficient (C f ) DNS result (Jiang et al. 2006)
Present calculation
from DNS solutions and the 0.002
present model computation for adiabatic flat plate
an adiabatic flat plate
0.0015 Ma ∞ = 4.5 T∞ = 61.1 K
Cf

FSTI=0.1%
0.001

0.0005

2E+06 4E+06 6E+06

Rex
184 Flow Turbulence Combust (2011) 87:165–187

Hassan's Model (2000)

U+ = Y+
0.92 Experiment (Lachowicz et al. 1996)
Incompressible
Present model 15
Compressible
o
5 half-angle flared cone adiabatic wall Present calculations
0.9
0 angle of attack
T / T0 ∞

10
Re∞ = 9.348E6 m -1

U+
Ma ∞ = 5.91
Ma∞ = 5.91 Re∞ = 9.348E6
0.88 T ∞ = 56.2 K FSTI = 0.1%
T∞ = 56.2 K FSTI=0.1% o
5 half-angle straight cone
5 adiabatic wall
0.86
Law of wall
0 angle of attack

0
0.1 0.2 0.3 0.4 0.5
100 101 102
x (m) Y+

(a) (b)
Fig. 12 Comparison of (a) computed and measured wall temperature distribution, and (b) computed
near-wall profiles of the mean velocity with the law of the wall, for an adiabatic flared cone

Figure 12(b) shows that the present calculation agrees well with the law of the wall
for compressible flows [1], which demonstrates that the compressibility modification
in the present model is effective.
Figure 13 shows the variation of calculated transition-onset location on an adia-
batic flat plate with the freestream Mach number in the range from subsonic to hy-
personic. All the corresponding experiments [51–53] are based on the same Reynolds
number and performed in similar facilities. It is verified that the present approach
can reproduce the three characteristic regions of Mach number, as mentioned in
Section 4.2. The main reason of the distinct difference between the calculation
and the experiment is probably because that the experimental data all come from
the conventional high-speed wind tunnels, where there exist high level of acoustic
disturbances that can seriously affect the transition process in the test section.
However, the different trends of variation of the calculation and of the experimental
data suggests that the present model over-predict the transition onset in the Mach
region [4, 4.5], which is also shown in Fig. 11.

Fig. 13 Variation of Expriment (Kendoll 1974)

12
transition-onset location with Expriment (Coles 1954)
the freestream Mach number Expriment (Deem&Murphy 1965)
10 Present calculations (FSTI = 0.1%)

8 Re∞ = 1.18E7 m -1
Retr × 10 - 6

4 adiabatic flat plate

0
0 2 4 6 8
Ma ∞
Flow Turbulence Combust (2011) 87:165–187 185

6 Conclusion

In conclusion, an intermittency transport γ equation is developed in this work for

the modelling of boundary layer transitional flows at freestream Mach numbers up
to about 6. The model equation couples with the modified SST k–ω eddy-viscosity
model in the solution procedures. The modification affects the transitional region
only. In the fully turbulent region, the model reverts back to the standard SST
k–ω model. In effect, this work proposes a k–ω–γ three-equation eddy-viscosity
transition/turbulence model considering the significance of the different instability
modes associated with the incoming freestream Mach numbers. This model is based
on local variables and is able to trigger the onset of transition automatically with
the function in the source term of γ equation, which has been successfully used to
simulate natural transition and provides better results than those of the other models
and can be applied to engineering calculations with wide Mach range.
The capability of transition onset prediction of the present model, which is
designed to describe the effects of flow instabilities in the pre-transitional region,
is validated with the self-similar solutions of Falkner–Skan boundary layer and the
compressible adiabatic flat-plate boundary layer. Both numerical characteristics of
the present model match qualitatively with the results from stability analysis and
experimental observations. It is noted that although stability codes are not used in
the calculations, results from stability theory plays an important role in determining
expressions for the effective viscosity resulting from the presence of non-turbulent
fluctuations. The key to the current approach is the recognition of the flow physics
underlying the first-mode and the second-mode instabilities that are significantly
different. Therefore, corresponding stress–strain laws governing the flowfield are
different.

Acknowledgements This work was supported by the National Natural Science Foundation of
China (Grant No. 90505005, 10932005).

Appendix

The Favre-averaged equation for the turbulent kinetic energy k can be written as

Dk
ρ̄ = ρ̄ Pk − ρ̄εs − ρ̄εd + Tk + p d + M
Dt

where the terms on the right-hand side are the production term of turbulent kinetic
energy ρ̄ P̃, the solenoidal (incompressible) dissipation ρ̄εs , the dilatation (com-
pressible) dissipation ρ̄εd , the turbulent transport term Tk , the pressure–dilatation
term p d , and the mass flux variation M. The dilatation–dissipation and pressure–
dilatation terms appear explicitly in k equation and directly affect the turbulence
energetics. Therefore, we adopt the dilatation–dissipation model proposed by Sarkar
et al. (1991) as

εd = 0.6Mt2 εs
186 Flow Turbulence Combust (2011) 87:165–187

Here, Mt is the turbulent Mach number defined by Mt = (2k)0.5 /a, a being the speed
of sound. And we adopt the pressure–dilatation model proposed by Sarkar (1995) as
p d = 0.15ρ̄ Mt2 εs − 0.2ρ̄ Mt Pk

References

1. White, F.M.: Viscous Fluid Flow, 2nd ed. McGraw-Hill, New York, Chap 1 and 7 (1991)
2. Mack, L.M.: Boundary-layer linear stability theory. AGARD Rept.709 (1986)
3. Stock, H.W.: e(N) Transition prediction in three-dimensional boundary layers on inclined prolate
spheroids. AIAA J. 44, 108–118 (2006)
4. Menter, F.R., Langtry, R., Völker, S.: Transition modelling for general purpose CFD codes. Flow
Turbul. Combust., 77, 277–303 (2006b)
5. Wilcox, D.: Turbulence Modelling for CFD. La Canada, CA, DCW Industries, Chap 3 (1992)
6. Savill, A.M.: One-point closures applied to transition. In: Hallbäck, M. (ed.) Turbulence and
Transition Modeling, pp. 233–268. Kluwer (1996)
7. Lardeau, S., Leschziner, M.A., Li, N.: Modelling bypass transition with low-Reynolds-number
nonlinear eddy-viscosity closure. Flow Turbul. Combust. 73, 49–76 (2004)
8. Westin, K.J.A., Henkes, R.A.W.M.: Application of turbulence models to bypass transition. J.
Fluids Eng. 119, 859–866 (1997)
9. Hadžić, I., Hanjalić, K.: Separation-induced transition to turbulence, second-moment closure
modeling. Flow Turbul. Combust. 63, 153–173 (2000)
10. Dhawan, S., Narasimha, R.: Some properties of boundary-layer flow during transition from
laminar to turbulent motion. J. Fluid M. 3, 414–436 (1958)
11. Libby, P.A.: On the prediction of intermittent turbulent flows. J. Fluid Mech. 68, 273–295 (1975)
12. Steelant, J., Dick, E.: Modelling of bypass transition with conditioned Navier–Stokes equations
coupled to an intermittency transport equation. Int. J. Numer. Meth. Fluids 23, 193–220 (1996)
13. Steelant, J., Dick, E.: Modelling of laminar-turbulent transition for high freestream turbulence.
J. Fluids Eng. 123, 22–30 (2001)
14. Rodi, W.: A review of experimental data of uniform density free turbulent boundary layers. In:
Launder, B.E. (ed.) Studies in Convection, vol. 1, pp. 79–165. Academic Press, San Diego (1975)
15. Byggstoyl, S., Köllmann, W.: Closure model for intermittent turbulent flows. Int. J. Heat Mass
Trans. 24, 1811–1823 (1981)
16. Cho, J.R., Chung, M.K.: A k–ε–γ equation turbulence model. J. Fluid Mech. 237, 301–322 (1992)
17. Simon, F.F., Stephens, C.A.: Modeling of the heat transfer in bypass transitional boundary-layer
flows. NASA Technical Paper 3170 (1991)
18. Warren, E.S., Hassan, H.A.: Transition closure model for predicting transition onset. J. Aircraft
35, 769–775 (1998)
19. McDaniel, R.D., Nance, R.P., Hassan, H.A.: Transition onset prediction for high-speed flow. J
Spacecr. Rockets 37, 304–309 (2000)
20. Papp, J.L., Kenzakowski, D.C., Dash, S.M.: Extensions of a rapid engineering approach to
modelling hypersonic laminar to turbulent transitional flows. AIAA Paper 2005-0892 (2005)
21. Schulte, V., Hodson, H.P.: Prediction of the becalmed region for LP turbine profile design. J.
Turbomach. (Trans. ASME) 120, 839–845 (1998)
22. Menter, F.R., Esch, T., Kubacki, S.: Transition modelling based on local variables. In: Rodi,
W., Fueyo, N. (eds.) 5th International Symposium on Turbulence Modelling and Measurements.
Elsevier (2002)
23. Menter, F.R., Langtry, R., Likki, S.R., Suzen, Y.B., Huang, P.G., Völker, S.: A correlation based
transition model using local variables part 1—model formulation. J. Turbomach. (Trans. ASME)
128, 413–422 (2006)
24. Langtry, R., Menter, F.R., Likki, S.R., Suzen, Y.B., Huang, P.G., Völker, S.: A correlation
based transition model using local variables part 2—test cases and industrial applications. J.
Turbomach. (Trans. ASME) 128, 423–434 (2006)
25. Suzen, Y.B., Huang, P.G., Hultgren, L.S., Ashpis, D.E.: Predictions of separated and transitional
boundary layers under low-pressure turbine airfoil conditions using an intermittency transport
equation. J. Turbomach. (Trans. ASME) 125, 455–464 (2003)
26. Pecnik, R., Sanz, W., Gehrer, A., Woisetschläger, J.: Transition modeling using two different
intermittency transport equations. Flow Turbul. Combust. 70, 299–323 (2003)
Flow Turbulence Combust (2011) 87:165–187 187

27. Suzen, Y.B., Huang, P.G.: An intermittency transport equation for modelling flow transition.
AIAA paper 2000-0287 (2000)
28. Lodefier, K., Merci, B., De Langhe, C., Dick, E.: Intermittency based RANS bypass transition
Modelling. Prog. Comput. Fluid Dyn. 6, 68–78 (2006)
29. Mayle, R.E., Schulz, A.: The path to predicting bypass transition. J. Turbomach. (Trans. ASME)
119, 405–411 (1997)
30. Volino, R.J., Simon, T.W.: Boundary layer transition under high free-stream turbulence and
strong acceleration conditions: part 2—turbulent transport results. J. Heat Transfer 119, 427–432
(1997)
31. Leib, S.J., Wundrow, D.W., Goldstein, M.E.: Effect of free-stream turbulence and other vortical
disturbances on a laminar boundary layer. J. Fluid Mech. 380, 169–203 (1999)
32. Walters, D.K., Leylek, J.H.: A new model for boundary layer transition using a single-point
RANS approach. J. Turbomach. (Trans. ASME), 126, 193–202 (2004)
33. Volino, R.J.: Separated flow transition under simulated low-pressure turbine airfoil conditions—
part 1, mean flow and turbulence statistics. J. Turbomach. (Trans. ASME), 124, 645–655 (2002)
34. Menter, F.R.: Two-equation eddy-viscosity turbulence models for engineering applications.
AIAA J. 32, 1598–1605 (1994)
35. Arnal, D., Casalis, G.: Laminar-turbulent transition prediction in three-dimensional flows. Prog.
Aerosp. Sci. 36, 173–191 (2000)
36. Wilcox, D.: Turbulence Modelling for CFD. La Canada, CA, DCW Industries, Chap 3 (1992)
37. Medic, G., Durbin, P.A.: Toward improved prediction of heat transfer on turbine blades. J.
Turbomach. (Trans. ASME) 124, 187–192 (2002)
38. Walker, G.J.: Transition flow an axial turbomachine blading. AIAA J. 27, 595–602 (1989)
39. Savill, A.M.: New strategies in modelling by-pass transition. In: Launder, B.E., Sandham, N.D.
(eds.) Closure Strategies for Turbulent and Transitional Flows, pp. 492–521. Cambridge Univer-
sity Press (2002)
40. Mayle, R.E.: The role of laminar-turbulent transition in gas turbine engines. J. Turbomach.
(Trans. ASME) 113, 509–537 (1991)
41. Sarkar, S., Erlebacher, G., Hussaini, M.Y.: The analysis and modelling of dilatational terms in
compressible turbulence. J. Fluid Mech. 227, 473–493 (1991)
42. Sarkar, S.: The stabilizing effect of compressibility in turbulent shear flow. J. Fluid Mech. 282,
163–186 (1995)
43. The MathWorks, Natick, M.A.: Using MATLAB version 5, revised for MATLAB 5.1 edition
(1997)
44. Schlichting, H.: Boundary Layer Theory, 7th ed. McGraw-Hill, New York, Chap 15 (1979)
45. Harris, J.E.: Numerical solution of flow equations for laminar, transitional, and turbulent com-
pressible boundary layers for planar or axisymmetric flows. NASA-TR-R-368 (1971)
46. Liou, M.S.: A Sequel to AUSM: AUSM+. J. Comput. Phys. 129, 364–382 (1996)
47. Schubauer, G.B., Klebanoff, P.S.: Contribution on the mechanics of boundary layer transition.
NACA TN 3489 (1956)
48. Mateer, G.G., Monson, D.J., Menter, F.R.: Skin-friction measurements and calculations on a
lifting airfoil. AIAA J. 34, 231–236 (1996)
49. Chen, F.J., Malik, M.R., Beckwith, I.E.: Comparison of boundary-layer transition on a cone and
flat plate at Mach 3.5. AIAA J. 27, 687–693 (1989)
50. Lachowicz, J.T., Chokani, N., Wilkinson, S.P.: Boundary-layer stability measurements in a hy-
personic quiet tunnel. AIAA J. 34, 2496–2500 (1996)
51. Kendall, J.M.: Wind tunnel experiments relating to supersonic and hypersonic boundary-layer
transition. AIAA J. 13, 290–299 (1975)
52. Coles, D.: Measurements of turbulent friction on a smooth flat plate in supersonic flow. J. Aero.
Sei. 21, 433–446 (1954)
53. Deem, R.E., Murphy, J.S.: Flat plate boundary layer transition at hypersonic speeds. AIM Paper
65-128 (1965)