Further studies on the statistics and structures of at-plate boundary layer

with passing wakes

Jean-Pierre Hickey, Scott Blakie, Callie Gray, Xiaohua Wu
*
Department of Mechanical Engineering, Royal Military College of Canada, PO Box 17000, Station Forces, Kingston, Ontario, Canada K7K 7B4
a r t i c l e i n f o
Article history:
Received 4 September 2009
Received in revised form 15 December 2009
Accepted 17 December 2009
Available online 8 February 2010
Keywords:
Wake perturbation
Zero-pressure gradient at-plate boundary
layer
Direct numerical simulation
Turbulent spot arrowhead
Transition mechanisms
Three-dimensional visualizations
a b s t r a c t
We present a numerical study of a at-plate boundary layer with free-stream periodically passing planar
wakes without external pressure gradients from Re
h
80 to fully turbulent. Two direct numerical simu-
lations were performed on the same ow with slightly different simulations parameters. In the con-
strained simulation design, the wakes are restricted to six boundary layer thicknesses above the wall
at the inlet, while in the other design the wakes are un-constrained. Boundary layer statistics before, dur-
ing and after transition are thoroughly evaluated through comparison with theoretical and experimental
data of the zero-pressure gradient at-plate boundary layer. Three-dimensional ow visualizations reveal
the instant and location of breakdown events. Negative streaks are not of the shape of a rectangular box,
rather, their wedged shape narrows towards the downstream passing wake and their strength diminishes
towards the wall and towards the upstream direction. Matured turbulent spots have arrowheads directed
upstream. Despite the intermittency of the passing wake disturbances, statistics in the early transitional
region exhibit characteristics in agreement with previous studies on boundary layer. Statistics in the tur-
bulent boundary layers differ from the canonical case only mildly near the outer edge.
2010 Elsevier Inc. All rights reserved.
1. Introduction
We consider an initially laminar incompressible at-plate
boundary layer under the perturbation of free-stream migrating
planar wakes at a passing frequency T without any imposed
streamwise pressure gradient. This setting abstracts from the tur-
bomachinery blade row interaction problem with geometrical
complexity and pressure gradients being removed. This ow also
bears resemblance to, but clearly differs from, the canonical incom-
pressible zero-pressure gradient at-plate boundary layer
(ZPGFPBL) under the perturbation of continuous free-stream iso-
tropic turbulence, although the degree of similarity remains to be
quantied.
Ten years ago, Wu, Jacobs, Hunt and Durbin carried out a simu-
lation on the passing wake induced boundary layer transition
problem, see Wu et al. (1999). Using two-dimensional ow visual-
izations they identied the occurrence of infant turbulent spots
and subsequent growth. These turbulent spots distinguish them-
selves from the type of manufactured turbulent spots studied in
earlier computational work in that the spots emerge naturally as
a result of the wake and boundary layer interaction process with-
out direct articial injection into the boundary layer. Wu et al. con-
cluded that longitudinal puff structures are formed inside the
laminar boundary layer as a result of inlet wake disturbance.
Breakdown of the longitudinal puff structure into infant turbulent
spot in wake induced boundary layer transition usually occurs in
the outer part of the boundary layer following the typical Kelvin
Helmholtz-type wavy motion in the velocity eld of puffs. We re-
gard puff as synonymous to the more commonly used descriptor
streak, which describes longitudinal structures inside perturbed
laminar boundary layers in the early transitional region. Wu
et al. made the argument that if breakdown occurs near the bound-
ary layer outer edge the resulting matured turbulent spot will ex-
hibit an arrowhead pointing towards the upstream direction.
Selected statistics were presented to demonstrate that the bound-
ary layer was carried in the simulation from laminar through tran-
sition to fully turbulent.
During the past ten years a large body of experimental and com-
putational work has been published on the problem of transitional
ow perturbed by passing wakes, although most of those studies
dealt with some kind of turbomachinery ow congurations. At
the same time there has also been notable progress in the numer-
ical simulation of transitional at-plate boundary layer interacting
with continuous free-stream isotropic turbulence (Durbin and Wu,
2007). The ndings of Wu et al. (1999) are upheld by some of these
more recent studies, however, in some other instances doubts were
raised on certain aspects of their study. In this paper, we report our
work on identifying and resolving several issues through modied
design, improved simulation and new results.
0142-727X/$ - see front matter 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.ijheatuidow.2009.12.012
* Corresponding author. Fax: +1 613 542 8612.
E-mail address: xiaohua.wu@rmc.ca (X. Wu).
International Journal of Heat and Fluid Flow 31 (2010) 315326
Contents lists available at ScienceDirect
International Journal of Heat and Fluid Flow
j our nal homepage: www. el sevi er. com/ l ocat e/ i j hf f
1.1. Recent studies on transitional ow with passing wakes
It is worthwhile mentioning the positive interaction between
computation and experiment on the problem of wake boundary
layer interaction. The simulation of Wu et al. (1999) was designed
by following the general idea of the earlier experiments of Liu and
Rodi (1991) conducted at the University of Karlsruhe. Liu and Rodi
measured some of the gross features of a at-plate boundary layer
transition under the perturbations of periodically migrating cylin-
der wakes in a free-stream. Following the papers of Wu et al. and
Wu and Durbin (2001), Wissink and Rodi (2006), and Rodi (2006)
reported a series of computational studies on transitional ow
with passing wakes in realistic turbomachinery geometries. They
used the same inlet planar wake velocity eld as in Wu et al. The
simulation results from the University of Karlsruhe group reinforce
the notion that the role of geometry generated adverse pressure
gradient is decisive in boundary layer transition with periodic
passing wakes. Their results are also broadly consistent with the
recent turbomachinery ow experiments performed at Cambridge
University, see Stieger (2002), Stieger and Hodson (2004), Hodson
and Howell (2005), and Gostelow et al. (2009). These Whittle Lab-
oratory experimental studies focused on the interaction between
passing wakes and laminar separation bubble. We note that those
studies are more directly relevant to engineering applications than
the abstracted fundamental problem considered in the present
work.
Sengupta et al. (2002) simplied the wake induced boundary
layer transition problem with a model consisting of an innite
row of vortices convecting over a laminar at-plate boundary
layer. Their numerical results suggest that the convecting discrete
vortices can induce a local instability on the underlying boundary
layer with very high growth rate. Ovchinnikov et al. (2006) studied
at-plate boundary layer transition under the disturbance of con-
tinuous free-stream cylinder wakes. They noted the similarity
and difference between the wake induced boundary layer transi-
tion and the canonical problem of boundary layer transition under
continuous free-stream turbulence. Their boundary layer did not
develop into a canonical ZPGFPBL downstream of transition. Since
their cylinder was located quite close to the boundary layer, the ef-
fect of streamwise pressure gradient associated with the cylinder
geometry on their ow may be appreciable.
Anthony et al. (2005) measured unsteady surface heat ux in a
transitional at-plate boundary layer under varying levels of free-
stream turbulence. Their results show that under moderate free-
stream turbulence enhanced heat ux streaks rst appear elon-
gated in the streamwise direction. These are followed by a distrib-
uted breakdown into streaky turbulent spots. No evidence of
backwards arrowhead of turbulent spots was seen in the surface
heat ux data from these experiments. Anthony et al. did not mea-
sure the shape of the spots above the wall, but they commented
that the turbulent surface heat ux almost always appears pointed
in the downstream direction. They also reported that symmetrical
arrowhead spots are rarely seen under moderate free-stream tur-
bulence. In these conditions, many turbulent streaks quickly ap-
pear and coalesce before discrete patches are allowed to develop
into the classical arrowhead shape.
1.2. Identied issues and scope
Although many turbomachinery papers have been published re-
cently on transitional ow with passing wakes, the congurations
used in those studies often deal with certain types of turbine cas-
cade or airfoil, e.g., Jeon et al. (2005). While useful from an engi-
neering point of view, the specic geometries and associated
pressure gradients severely limit their connection with the canon-
ical ZPGFPBL transition problem thereby reducing their appeal to
the fundamental uid mechanics community. Stieger (2002) mea-
sured some properties of a transitional at-plate boundary layer
with periodically passing wakes, but he imposed external stream-
wise pressure gradients to simulate the engine environment. The
importance of the setting considered in the present study rests
on the preservation of the essential ingredients of the turboma-
chinery blade row interaction problem and at the same time main-
taining a good resemblance to the fundamental ZPGFPBL transition
problem. Removal of the geometric complexities and pressure gra-
dients isolates the effect of passing wakes on boundary layer tran-
sition. More importantly, such an idealization permits a
meaningful comparison of the collected statistics before break-
down with Blasius and after transition with canonical turbulent
ZPGFPBL data. Without a thorough quantitative evaluation of the
boundary layer statistics before and after transition the value and
applicability of any qualitative visualization for transition mecha-
nism is doubtful. This is because the transition mechanism in a
poorly controlled process with nite level of perturbations is not
necessarily the same as the transition mechanism in a well-con-
trolled process taking the Blasius layer to the canonical turbulent
ZPGFPBL. We also note that the subject of the present study, at-
plate boundary layer with passing wakes in the absence of external
pressure gradients, has not been further explored by recent direct
numerical simulation efforts since the work of Wu et al. (1999).
Wu et al. (1999) reported limited comparison of their second-
order statistics after transition with the DNS data of Spalart
(1988) and obtained reasonable agreement. What can be drawn
from that limited evaluation at most is that transition was carried
to completion in the simulations and the boundary layer state
downstream of transition mimics to a certain degree the canonical
turbulent ZPGFPBL. Deviation from Blasius solution in the early
transitional region was not quantied. The degree of similarity or
difference between the turbulent at-plate boundary layer transi-
tioned due to passing wakes and the classical turbulent ZPGFPBL
was not assessed in detail. For instance, it is unclear to what extent
the free-stream wakes affect the underneath fully turbulent
boundary layer. In the rst part of this paper, we report the ow
statistics of two at-plate boundary layers with passing wakes to
address the issues as to how far the boundary layers deviate from
Blasius prior to breakdown, and how close the boundary layers
resemble the canonical turbulent ZPGFPBL after transition. We as-
sess the quality of the simulation through a detailed comparison
with theoretical results before breakdown and with classical
experimental and DNS data after breakdown. We also present fre-
quency spectrum results in all the stages. These efforts on quanti-
fying the accuracy and reliability of the present simulations are
also critical in establishing the creditability of the systematic ow
structural investigations reported in the second part of the paper.
Wu et al. (1999) concluded that breakdown into infant turbu-
lent spot in wake induced boundary layer transition usually occurs
in the outer part of the boundary layer following the typical Kel-
vinHelmholtz-type wavy motion in the velocity eld of the longi-
tudinal structures. In light of the work by Wu et al. and the more
recent investigations done on wake perturbed boundary layer
ows a series of questions remain un-answered. Streaks due to
the passing wakes were visualized in selected planes parallel to
the at-plate: do they share similar characteristics in terms of spa-
tial variations of geometry and strength as those widely reported
in ZPGFPBL with free-stream isotropic turbulence? Previous stud-
ies were often not precise on the identication of space-time coor-
dinates of the breakdown events and the visualizations were often
fragmented: is it possible to obtain a set of systematic three-
dimensional ow images which clearly addresses the relevant
transition mechanism questions? In Wu et al. the wake distur-
bances entered the computational domain through the free-stream
as well as through the inner boundary layer: will the breakdown
316 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326
process be different if the wake disturbances enter the computa-
tional domain through the free-stream only? Additionally, the sur-
face heat ux measurements of Anthony et al. (2005) casted doubt
on the proposition of Wu et al. that if breakdown occurs near
boundary layer outer edge the resulting matured turbulent spot
will exhibit an arrowhead directing upstream. Therefore, in order
to reconcile the differences it is also useful to revisit the issue of
turbulent spot arrowhead direction in a wake induced boundary
layer transition.
2. Details of the present computation
Two incompressible direct numerical simulations were per-
formed in the present study, see Fig. 1. In the constrained inlet
wake ow design, the inlet planar wakes are restricted to be six
boundary layer thicknesses away from the at-plate. In other
words, the inlet planar wakes are not permitted to reach any posi-
tion that is less than y 0:024L at the inlet station. The reason for
this design is so that the Blasius boundary layer at the inlet re-
mains undistorted by the imposed perturbations and therefore
avoids the arbitrariness in prescribing perturbed velocity prole
inside the layer as in the un-constrained wake case. This design
has the added benet of ensuring that disturbances inside the
boundary layer originate solely from the free-stream. In the un-
constrained wake ow, the inlet wakes are allowed to penetrate
into the boundary layer and touch the at-plate. This is the same
as the case studied by Wu et al. (1999) but with improved stream-
wise resolution. In both designs, the mean wake decit is 14% of
the free-stream velocity and the mean wake half-width is 10% of
unit length.
The passing period T was set to be 1.67 in the present simula-
tion. Thus, at t 1:67 or at any integer multiple of T there is an-
other isotropic turbulence box entering the inlet plane from the
top according to the Eqs. (6) and (7) in Wu et al. (1999). All the
other inow conditions, boundary conditions, geometrical param-
eters and numerical method are identical as those described in the
Section 2 of Wu et al. (1999). Specially, the computational domain
is dened as 0:1 6 x=L 6 3:5, 0 6 y=L 6 0:8 and 0 6 z=L 6 0:2,
where L is the unit length scale. The inlet momentum thickness
Reynolds number is Re
h
80 and the Reynolds number based on
free-stream velocity U
1
and L is Re 1:5 10
5
. The nite-differ-
ence grid size is 2048 400 128 along the x; y and z directions,
respectively. The number of grid points along the streamwise
direction is twice of that used in Wu et al. (1999). The simulation
was performed using 1024 processors on 128 IBM 8-way P655+
nodes. For the constrained wake simulation, the computational
time step was xed at Dt T =2000 0:000835. For the un-con-
strained wake simulation, the computational time step was xed
at Dt T =1000 0:00167. In both cases, the initial velocity elds
were prescribed as the Blasius prole. The ows were allowed to
evolve from the initial unrealistic velocity prole for 20 passing
period T to reach statistically steady state. Statistics were then col-
lected for another 20T . In addition to averaging in time, the statis-
tical sample was enhanced by averaging in the homogeneous
spanwise direction. Velocity signals were saved at each time step
for a duration of 20T at an array of selected locations for the pur-
pose of frequency spectrum computation. At each point the total
spectrum sample sequence was divided into overlapping seg-
ments, each containing 1024 points. All frequency power spectra
presented in this paper were computed using the Hanning window
following the procedures and notations in Choi and Moin (1990).
The numerical method and the geometrical parameters used
this work are substantially similar to those in Wu et al. (1999).
But this is to be expected given the canonical and representative
nature of the subject of interest. One only needs to realize that
the turbulent ZPGFPBL has been investigated countless times using
essentially the same hot-wire technique over a time span of several
decades, and the fully-developed turbulent channel ow has also
been computed with the same DNS technique countless times for
a range of purposes. We note that the work described here is only
the second DNS investigation on the problem of at-plate bound-
ary layer perturbed by passing wakes without external pressure
gradients.
Fig. 1 shows images of the boundary layer developments in both
designs using contours of streamwise turbulent uctuating
velocity component u
0
over one xy-plane. It is evident that at
t 39:75T that the constrained wake near the inlet region does
Fig. 1. Side-view image of u
0
over one random xy-plane. Upper: the boundary layer with constrained passing wakes at 39:75T ; lower: the boundary layer with un-
constrained passing wakes at 30:85T . White color represents higher positive values of u
0
.
J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326 317
not directly contact the boundary layer. Further downstream a
perturbed region inside the boundary layer trails the free-stream
convecting wake and tapers off gradually with increasing distance
from the wake. In both cases it can be judged qualitatively that the
boundary layers attain a fully turbulent state and transition seems
to have completed by x 3:0.
3. Overall boundary layer properties
Engineering turbomachinery boundary layers are subjected to
complicated streamwise and spanwise pressure gradients whose
effects are overwhelming and often mask the interaction between
laminar boundary layer and passing wakes. One of the rationales
for avoiding either intentionally or un-intentionally imposed
streamwise pressure gradient in the current study is analogous
to the reason of studying the canonical ZPGFPBL in aeronautical
applications: even though real world airfoils are subjected to
pressure gradients, the laboratory canonical turbulent ZPGFPBL
however provides an idealized constant reference frame and
therefore is more important from a research point of view. Fig. 2
presents, for the boundary layer with constrained inlet wake, mean
wall static pressure gradient C
pw
p
y0;x
p
y0;x0:1
=
1
2
qU
2
1

,
together with two other mean streamwise pressure gradient
parameters b d

=s
w
d p=dx and P

m=qu
3
s
d p=dx. Very mild
streamwise pressure gradients exist in the computational domain
for the boundary layer of the constrained wake design. The value
of C
pw
in the present ow varies between 0 and 0.025, compared
to the much larger range of variation of 3:6 6 C
pw
6 1 on the
suction side of the T106 turbine blade, which was imposed by Stie-
ger (2002) on his transitional at-plate boundary layer experiment
with passing wakes. Weak favorable pressure gradient exists from
the inlet to Re
h
150 due to the introduced free-stream passing
wakes, and also exists near the exit due to the imposed outlet
boundary condition. Over a signicant portion of the computa-
tional domain from Re
h
150 to 900 the streamwise pressure gra-
dient for the boundary layer with constrained passing wakes is
even much weaker and can therefore be considered as nominally
zero. The pressure gradient near the inlet of the boundary layer
with un-constrained passing wakes is slightly larger than that
shown in Fig. 2 due to the impingement of the passing wakes di-
rectly onto the at-plate at Re
h
80, see Fig. 32 in Wu et al. (1999).
Skin-friction coefcient C
f
is shown in Fig. 3. The region in
which C
f
strictly follows the Blasius prole is signicant, and the
location at which C
f
attains a minimum is relevant. By the word
strictly we mean the deviation is no more than one or two per-
cent. We consider the streamwise range in which C
f
strictly follows
the Blasius solution as one in which disturbance level inside the
boundary layer may be regarded as innitesimal, which in turn
permits the application of linear stability analysis. We also con-
sider the minimum C
f
station as the statistical averaged location
of breakdown events, which in turn divides the transitional region
into the early and late stages. It is evident that perturbations inside
the boundary layer produced by the constrained inlet wakes up-
stream of Re
h
180 can be considered as innitesimal, but the
same can not be said about those in the un-constrained wake de-
sign. C
f
in the boundary layer with the un-constrained inlet wakes
deviates from Blasius right from the inlet, and disturbance level in-
side that boundary layer should be considered as nite. The aver-
aged breakdown locations as indicated by the x-coordinate
where C
f
attains a minimum value are found at Re
h
250 and
260 for the un-constrained and constrained boundary layers,
respectively. In both boundary layers at Re
h
800, skin friction
coefcient has achieved its fully turbulent level as indicated by
good comparison with the experimental data of Murlis et al.
(1982) and Adrian et al. (2000).
Boundary layer integral parameters are presented in Fig. 4. The
onset of the fully turbulent region Re
h
800 is found at x 3:05
0 200 400 600 800 1000
-0.03
-0.02
-0.01
0
0.01
0.02
Fig. 2. Mean streamwise pressure gradient as a function of Reh in the boundary
layer with constrained passing wakes. Solid line: wall static pressure coefcient
Cpw; dotted line: b
y0:1
; dashed line: 20P

y0:1
.
0 200 400 600 800 1000
0
0.001
0.002
0.003
0.004
0.005
0.006
Fig. 3. Mean skin friction coefcient C
f
as a function of Reh. Solid line: boundary
layer with constrained passing wakes; dotted line: boundary layer with un-
constrained passing wakes; dashed line: Blasius solution; solid circle: Murlis et al.
(1982); open circle: Purtell et al. (1981); diamond: Adrian et al. (2000).
0 200 400 600 800 1000
0
1
2
3
4
5
6
Fig. 4. Mean boundary layer integral parameters as a function of Reh. Solid line:
boundary layer with un-constrained passing wakes; dashed line: boundary layer
with constrained passing wakes; symbols: corresponding Blasius solutions. open
circle: boundary layer thickness d multiplied by 100; solid circle: displacement
thickness d

multiplied by 400; diamond: streamwise distance x.

318 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326
0 200 400 600 800 1000
1
1.5
2
2.5
3
Fig. 5. Shape factor H d

=h as a function of Reh. Solid line: boundary layer with

constrained passing wakes; dotted line: boundary layer with un-constrained
passing wakes; solid circle: Murlis et al. (1982); box: Adrian et al. (2000); chain-
dotted line: Blasius solution.
0 2 4 6 8 10 12 14 16 18 20
0
0.01
0.02
0.03
0.04
Fig. 6. u
0
rms
as a function of n yU1=mx
1=2
in the early transitional region of the
boundary layer with constrained passing wakes. circle: Reh 100; box: Reh 120;
triangle: Reh 140; plus: Reh 160; diamond: Reh 180.
1000 100 10 1
10
-11
10
-10
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
Fig. 7. Frequency spectrum of the streamwise velocity uctuations in the early
transitional region of the boundary layer with constrained passing wakes at
Reh 144. Solid circle: y=d 0:11; diamond: 0.22; plus: 0.55; triangle: 1.1.
1000 100 10 1
0
5
10
15
20
25
30
35
Fig. 8. Mean velocity u

as a function of y

in the late transitional region of the

boundary layer with constrained passing wakes. Dotted line: Reh 200; dashed
line: Reh 300; long dashed line: Reh 400; chain-dotted line: Reh 500; chain-
dotted-dotted line: Reh 600; solid line: Reh 800; diamond: Spalart (1988) at
Reh 670.
0 20 40 60 80 100 120 140 160 180 200
0
0.5
1
1.5
2
2.5
3
Fig. 9. u
0
rms
as a function of y

in the late transitional region of the boundary layer

with constrained passing wakes. Dotted line: Reh 200; dashed line: Reh 300;
long dashed line: Reh 400; chain-dotted line: Reh 500; chain-dotted-dotted
line: Reh 600; solid line: Reh 800; diamond: Spalart (1988) at Reh 670.
1000 100 10
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
Fig. 10. Frequency spectrum of the streamwise velocity uctuations in the late
transitional region of the boundary layer with constrained passing wakes at
Reh 421. Circle: y 0:027d; diamond: y 0:13d; plus: y 0:54d; triangle:
y 1:07d. Solid lines from left to right: 1, 5/3 and 7 slopes, respectively.
J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326 319
for the boundary layer with constrained wakes and at x 2:75 for
the boundary layer with un-constrained wakes. Compared to other
integral parameters the shape factor H d

=h shown in Fig. 5 is an
excellent indicator of boundary layer transitional state. It is seen
that H in the boundary layer with un-constrained inlet wakes de-
parts very soon after the inlet from the Blasius solution at around
Re
h
170, where as the constrained setup follows the solution
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0
0.2
0.4
0.6
0.8
1
Fig. 11. Reynolds shear stresses as a function of y=d at Reh 900. Solid line:
boundary layer with constrained passing wakes; dotted line: boundary layer with
un-constrained passing wakes; dashed line: ZPGFPBL of Wu and Moin (2008);
circle: Honkan and Andreopoulos (1997) at Reh 2790; diamond: DeGraaff and
Eaton (2000) at Reh 1430; plus: Spalart (1988) at Reh 670.
0 20 40 60 80 100
0
0.2
0.4
0.6
0.8
1
Fig. 12. Reynolds shear stresses as a function of y

at Reh 900; Solid line:

boundary layer with constrained passing wakes; dotted line: boundary layer with
un-constrained passing wakes; solid circle: Spalart (1988) at Reh 670.
Fig. 13. Iso-surfaces of u in the boundary layer with constrained passing wakes at t 39:2T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
320 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326
until about Re
h
210. This would seem to suggest that the un-con-
strained setup is under a nite rather than a innite disturbance.
After the completion of transition shape factors in both boundary
layers agree well with existing experimental data for the canonical
turbulent ZPGFPBL.
4. Detailed boundary layer statistics
It is well-known from early temporal inviscid analysis that tur-
bulence in the presence of homogeneous shear will exhibit linear
growth with time in the streamwise intensity and negligible
growth in the other two intensity components, e.g., see Phillips
(1969). Application of the temporal inviscid result to a perturbed
spatially developing laminar boundary layer would suggest the
also well-known linear growth of u
0
rms
with x, ignoring all the nec-
essary approximations. The u
0
rms
proles for the boundary layer
with constrained inlet wakes shown in Fig. 6 are consistent with
the theory. The peak location of u
0
rms
is located consistently at
0:4d, very similar to the results reported by Westin et al. (1994)
in a laminar boundary layer perturbed by free-stream isotropic
turbulence.
Fig. 7 shows frequency spectra of the streamwise turbulent nor-
mal stress in the boundary layer with constrained inlet wakes at
Re
h
144 for four wall-normal positions, y 0:11d; 0:22d; 0:55d
and 1:1d. There is a cross-over between the spectrum at
y 0:55d and that at y 1:1d, indicating that in the middle of
the layer the streamwise uctuations have stronger low frequency
content but much weaker high frequency content compared to
those immediately outside the boundary layer edge. There is a fur-
ther reduction in magnitude from y 0:55d to 0:22d over all the
frequency ranges but the decrease in the low to intermediate fre-
quency range is more substantial. In the near-wall region from
y 0:22d to 0:11d the high-frequency content of u
02
remains un-
changed but there is noticeable further decrease in the low fre-
quency content.
Global integral parameters and statistics discussed in the last
section have indicated that the constrained passing wakes in the
free-stream produce rather small level of perturbations to the
underneath laminar boundary layers in the early transition region.
For instance, deviations of skin friction and shape factor from the
corresponding Blasius proles are negligible prior to breakdown.
This is one of the reasons for adopting this particular simulation
design: quality of the computational results in the early transi-
tional region can be evaluated by their differences from the Blasius
solutions. The second reason is that, given the fact that the
imposed wake disturbances are sufciently mild and there are no
extra imposed pressure gradients, the statistics in the late transi-
tional region should converge towards the canonical low-Reynolds
number turbulent ZPGFPBL proles, at least in the near-wall re-
gion. If the statistics deviate noticeably from the turbulent
Fig. 14. Iso-surfaces of u in the boundary layer with constrained passing wakes at t 39:45T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326 321
ZPGFPBL proles, the quality of the computational results would
become questionable. In this manner the accuracy of the computa-
tional results downstreamof breakdown is quantied. We consider
quantication of the quality of computational results as a neces-
sary condition for the structural ow visualizations to be discussed
in the next two sections.
Mean and second-order statistics in the late transitional region
of the boundary layer with constrained inlet passing wakes at six
streamwise stations from Re
h
200 to 800 are documented in Figs.
8 and 9. They demonstrate a systematic convergence with increas-
ing x towards the ZPGFPBL DNS of Spalart (1988) in the near-wall
region. The comparison between the present results in wall units at
the end of the late transitional region, Re
h
800, with those of
Spalart at Re
h
670 is remarkably good. This suggests that in the
boundary layer with constrained wakes the inner region after tran-
sition can be considered as substantially close to that of a typical
ZPGFPBL, as expected. Frequency spectrum results in the late tran-
sitional region at Re
h
421 for the boundary layer with con-
strained passing wakes are presented in Fig. 10. Compared to
Fig. 7 for the early transitional region, content over the high fre-
quency range of the streamwise uctuations has now been lled.
There exists a set of secondary peaks in the spectra, especially near
the region of xd=u
s
30. Similar secondary peaks are also clearly
visible in the spectra of the other two components.
It has been shown for the boundary layer with constrained
passing wakes near-wall statistics converge towards those of Spal-
art (1988) in the late transitional region. Here we present statistics
in the fully turbulent region at Re
h
900 for the two boundary lay-
ers with constrained and with un-constrained wakes. The results
are compared with low-Reynolds number turbulent ZPGFPBL data.
Such an evaluation will help to address the issue as to the similar-
ity between the turbulent boundary layer with constrained inlet
wakes and the turbulent boundary layer with un-constrained inlet
wakes. It also helps to answer the question as to the similarity be-
tween the present transitioned turbulent boundary layers and the
canonical turbulent ZPGFPBL. Reynolds shear stress proles are
compared to existing ZPGFPBL data in Fig. 11 using outer coordi-
nate and in Fig. 12 using inner coordinate, respectively. u
0
v
0
in
the present two boundary layers with passing wakes attain nega-
tive values beyond y=d 1:2 while those for the ZPGFPBL remain
positive. Overall statistics in the fully turbulent region of the two
boundary layer transitioned from passing wakes show good agree-
ment with classical ZPGFPBL data.
5. Identication of the breakdown events
We have demonstrated in Sections 3 and 4 that our boundary
layer statistics deviate very slightly from Blasius in the early
Fig. 15. Iso-surfaces of u in the boundary layer with constrained passing wakes at t 39:55T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
322 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326
transitional region, and converge nicely to the DNS of Spalart
(1988) near the wall in the late transitional and turbulent regions.
Having established the quality of the computational results here
we are ready to address the identied structural issues in at-plate
boundary layer transition with free-stream passing wakes through
systematic three-dimensional ow visualization. We consider a
breakdown event as the switch in any one streaky structure devel-
oped in the early transitional region from semi-regular to locally
chaotic. Within any particular wake passing period there can be
multiple breakdown events. Precise location of the space-time
coordinates of the breakdown events is crucial for the determina-
tion of breakdown mechanism because processes that occur after
breakdown will not be helpful in determining the mechanism that
has caused that particular breakdown. We are of the opinion that
identication of breakdown events must include documentation
of the relevant development prior to the occurrence of chaotic
scales in the streaks. Presentation of a small turbulent spot does
not in itself identify a breakdown event.
Figs. 1317 present iso-surfaces of instantaneous streamwise
velocity u=U
1
from t=T 39:2 to 39.8 for the boundary layer with
constrained passing wakes. The particular passing period of
t=T 39 was randomly chosen without any screening. At each in-
stant four iso-surfaces are shown in row-major order correspond-
ing to u=U
1
0:35; 0:7; 0:85 and 0.98, respectively. The idea is to
reduce the difculty in interpreting the three-dimensional charac-
teristics of the streaks found in previous visualization studies. A
bulge in the iso-surface of u=U
1
represents a region of negative
u
0
. It is seen that the negative streaks in the early transitional re-
gion are not of the shape of a regular rectangular box which is uni-
form along the x-direction and uniform in y-direction. Instead,
many of the negative streaky structures display a wedged shape.
Along the streamwise x-direction, they are stronger in strength
and thinner in width near the downstream end where the tails of
the free-stream wakes are found. In the wall-normal y-direction,
the negative streaks are strongest near the upper boundary layer
edge, as indicated by the iso-surface of u=U
1
0:85. Signatures
of the negative streaky structures near the wall as indicated by
the bulges in the iso-surfaces of u=U
1
0:35, are merely faint
foot-prints of the upper layer streaks. Initially the streaks are rela-
tively short. In Fig. 13 at t=T 39:2 the streamwise length of the
negative streaks is approximately 0.1, roughly equivalent to ten lo-
cal boundary layer thicknesses. By the time of t=T 39:55 in
Fig. 15 the negative streaks have lengthened to approximately
0.25, equivalent to approximately twenty local boundary layer
thicknesses. The sequence of visualization images from Figs. 13
16 demonstrates convincingly that one breakdown event occurs
at t=T 39:65. We also point out that if one only concentrates
on the near-wall region in the search for signatures of breakdown
Fig. 16. Iso-surfaces of u in the boundary layer with constrained passing wakes at t 39:65T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326 323
it would be difcult to declare that breakdown has occurred to the
streaky structure shown in Fig. 16.
6. Turbulent spot arrowhead direction
Wu et al. (1999) proposed that direction of matured turbulent
spot depends on the relative wall-normal location of the break-
down event in the boundary layer. If breakdown is initiated from
the boundary layer edge, the resulting spot arrowhead will point
towards the upstream direction. Their reasoning is quite logical
and analogous to the explanation given to the phenomenon of
Mach cone in compressible ow. More recent authors paid little
attention to the issue except for Anthony et al. (2005), who re-
ported that they found no evidence in support of the idea of Wu
et al. Here we present visualization images to show that the rea-
soning of Wu et al. is indeed sound.
For the boundary layer with constrained passing wakes, after
breakdown of the negative streak at t=T 39:65 a turbulent spot
develops from the shape of a narrow stripe into a wedged shape
with a distinct arrowhead directing towards the upstream in
Fig. 17 at t 39:8T . For the boundary layer with un-constrained
passing wakes, the turbulent spot shown in the central region of
Fig. 19 at t 30:45T is a result of the simultaneous breakdown
of the two neighboring negative streaks near the center plane,
e.g., see Fig. 18 at t 30:1T . This causes the loss of a clear arrow-
head in the merged turbulent patch. We emphasize that the
instantaneous ow visualizations presented in this paper were ex-
tracted from velocity elds saved during the computations from
randomly picked passing periods without any pre-screening or
post-screening.
7. Concluding remarks
Flat-plate boundary layer under the disturbance of free-stream
periodically passing planar wakes in the absence of imposed pres-
sure gradients is a well-dened, idealized laboratory ow with suf-
cient scientic generality and good engineering relevance. Its
close resemblance to the canonical ZPGFPBL allows one to quanti-
tatively evaluate the simulation results. The present investigation
is the second DNS study on this ow, and has resolved a number
of structural and statistical questions identied through an exam-
ination of the earlier work of Wu et al. (1999) in light of several re-
cent relevant papers on transition in at-plate boundary layers
with free-stream isotropic turbulence.
In the early transitional region, statistics such as skin friction
and shape factor of the boundary layer with constrained inlet
Fig. 17. Iso-surfaces of u in the boundary layer with constrained passing wakes at t 39:8T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
324 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326
wakes deviate from Blasius solution very mildly and the pertur-
bations inside that boundary layer can be considered as inni-
tesimal. A similar comparison of the statistics for the
boundary layer with un-constrained passing wakes indicates a
large level of deviation from Blasius and the disturbance magni-
tude inside that boundary layer should be considered as nite.
After the completion of transition, statistics in both of the two
turbulent boundary layers compare very well with the classical
experimental data of Purtell et al. (1981) and Murlis et al.
(1985) for low-Reynolds number turbulent ZPGFPBL as well as
with the DNS of Spalart (1988). Shear stress prole and shear
stress frequency spectra deviate from the turbulent ZPGFPBL re-
sults of Wu and Moin (2008) only beyond the outer 10% of the
boundary layer.
Our critical evaluation of the boundary layer statistical mea-
sures lends creditability to the systematic ow visualization
study, which begins with a precise determination of the space-
time coordinates of breakdown events. In boundary layer with
free-stream passing wakes but without external pressure gradi-
ents, negative streaks form and trail behind the wakes. The neg-
ative streaks are initially short and wedge shaped with its strong
narrow end being located close to the downstream traveling
wake. The negative streaks are not of the shape of a rectangular
box with uniform strength in either the wall-normal direction or
the streamwise direction, rather, they are much stronger in the
outer portion of the boundary layer. The near-wall signatures
of the negative streaks are merely faint foot-prints of the stron-
ger outer layer streak structures. Along the streamwise direction,
the streaks are stronger near the downstream wakes, and their
strength taper offs towards the upstream. These characteristics
are consistent with a process in which the current negative
streaks are a result of the turbulent eddy diffusion from the
downstream passing wakes coupled with the shearing distortion
by the boundary layer. Matured turbulent spots have arrowheads
directing upstream. But if a turbulent patch originates from the
simultaneous breakdown of multiple nearby streaks its arrow-
head conguration will be less sharp.
Acknowledgments
The computer program used in this study was developed by the
late Dr. Charles D. Pierce of the Center for Turbulence Research at
Stanford. This work was supported by NSERC Discovery Grant,
Department of Defense Academic Research Program, and the Can-
ada Research Chair Program.
Fig. 18. Iso-surfaces of u in the boundary layer with un-constrained passing wakes at t 30:1T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right:
0.98.
J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326 325
References
Adrian, R.J., Meinhart, C.D., Tomkins, C.D., 2000. Vortex organization in the outer
region of the turbulent boundary layer. J. Fluid Mech. 422, 154.
Anthony, R.J., Jones, T.V., LaGraff, J.E., 2005. High frequency surface heat ux
imaging of bypass transition. ASME J. Turbomach. 127, 241250.
Choi, H., Moin, P., 1990. On the space-time characteristics of wall-pressure
uctuations. Phys. Fluids 2, 14501460.
DeGraaff, D.B., Eaton, J.K., 2000. Reynolds-number scaling of the at-plate turbulent
boundary layer. J. Fluid Mech. 422, 319346.
Durbin, P.A., Wu, X., 2007. Transition beneath vortical disturbances. Annu. Rev.
Fluid Mech. 39, 107128.
Gostelow, J.P., Thomas, R.L., Adebayo, D.S., 2009. the Response of a strongly diffusing
ow to propagating wakes. ASME J. Turbomach. 131, 021006-1021006-8.
Hodson, H.P., Howell, R.J., 2005. Bladerow interactions, transition and high-lift
aerofoils in low-pressure turbines. Annu. Rev. Fluid Mech. 37, 7198.
Honkan, A., Andreopoulos, Y., 1997. Vorticity, strain-rate and dissipation
characteristics in the near-wall region of turbulent boundary layers. J. Fluid
Mech. 350, 2996.
Jeon, W.P., Park, T.C., Kang, S.H., 2005. Experimental study of boundary-layer
transition on an airfoil induced by periodically passing wake. Exp. Fluids 32,
229241.
Liu, X., Rodi, W., 1991. Experiments on transitional boundary layers with wake-
induced unsteadiness. J. Fluid Mech. 231, 229256.
Murlis, J., Tsai, H.M., Bradshaw, P., 1982. The structure of turbulent boundary layers
at low Reynolds numbers. J. Fluid Mech. 122, 1356.
Ovchinnikov, V., Piomelli, U., Choudhari, M.M., 2006. Numerical simulations of
boundary-layer transition induced by a cylinder wake. J. Fluid Mech. 547, 413
441.
Phillips, O.M., 1969. Shear-ow turbulence. Annu. Rev. Fluid Mech. 1, 245264.
Purtell, L.P., Klebanoff, P.S., Buckley, F.T., 1981. Turbulent boundary layer at low
Reynolds number. Phys. Fluids 24, 802811.
Rodi, W., 2006. DNS and LES of some engineering ows. Fluid Dyna. Res. 38, 145173.
Sengupta, T.K., Chattopadhyay, M., Wang, Z.Y., Yeo, K.S., 2002. Bypass mechanism of
transition to turbulence. J. Fluids Struct. 16, 1529.
Spalart, P.R., 1988. Direct simulation of a turbulent boundary layer up to Rh = 1410.
J. Fluid Mech. 187, 6198.
Stieger, R.D., 2002. The effects of wakes on separating boundary layers in low
pressure turbines. Ph.D. Thesis, University of Cambridge, Cambridge, UK.
Stieger, R.D., Hodson, H.P., 2004. The transition mechanism of highly loaded low-
pressure turbine blades. ASME J. Turbomach. 126, 536543.
Westin, K, J.A., Boiko, A.V., Klingmann, B.G.B., Kozlov, V.V., Alfredsson, P.H., 1994.
Experiments in a boundary layer subjected to free stream turbulence. Part 1.
Boundary layer structure and receptivity. J. Fluid Mech. 281, 193218.
Wissink, J.G., Rodi, W., 2006. Direct numerical simulation of ow and heat transfer
in a turbine cascade with incoming wakes. J. Fluid Mech. 569, 209247.
Wu, X., Jacobs, R., Hunt, J.C.R., Durbin, P.A., 1999. Simulation of boundary layer
transition induced by periodically passing wakes. J. Fluid Mech. 398, 109153.
Wu, X., Durbin, P.A., 2001. Evidence of longitudinal vortices evolved from distorted
wakes in a turbine passage. J. Fluid Mech. 446, 199228.
Wu, X., Moin, 2008. Direct Numerical Simulation of Turbulence in a Nominally Zero-
Pressure-Gradient Flat-Plate Boundary Layer. Center for Turbulence Research
Manuscript 193. Stanford University, Berlin.
Fig. 19. Iso-surfaces of u in the boundary layer with un-constrained passing wakes at t 30:45T . Upper left: u=U1 0:35; upper right: 0.7; lower left: 0.85; lower right: 0.98.
326 J.-P. Hickey et al. / International Journal of Heat and Fluid Flow 31 (2010) 315326