International Journal of Heat and Fluid Flow 25 (2004) 364–372

www.elsevier.com/locate/ijhff

Vortex shedding from two surface-mounted cubes in tandem

Robert J. Martinuzzi *, Brian Havel
Advanced Fluid Mechanics Research Group, Department of Mechanical and Materials Engineering, Faculty of Engineering,
University of Western Ontario, London, ON, Canada N6A 5B9

Abstract
Periodic vortex shedding from two surface-mounted cubes, of height H, in tandem arrangement placed in a thin boundary layer is
investigated for a spacing 2H using phase-averaged Laser Doppler Velocimetry. Tests were conducted for a Reynolds number of
22,000, based on H and the freestream velocity, and an approximately 0.07H thick laminar boundary layer. For obstacle separations
between 1.5H and 2.5H , the shedding frequency scales inversely with the obstacle spacing, S, such that the Strouhal number based
on S is constant or geometrically locked. In this locked regime, periodic shedding is triggered by the interference between a vertical
flow stream along the front face of the downstream obstacle and the vortex in the inter-obstacle cavity. This three-dimensional
mechanism is not observed for two-dimensional geometries and helps explain why a locked regime cannot be observed for square
cylinders in tandem arrangement. Furthermore, it is shown that the structure of the turbulent field in the cavity region differs
significantly from that in the base region of a two-dimensional obstacle.

2004 Elsevier Inc. All rights reserved.

Keywords: Vortex shedding; Tandem surface-mounted cubes; Phase-averaging; Turbulence; Experimental; Bluff-body interference

1. Introduction of the first obstacle over a critical obstacle spacing range

(cf. Zdravkovich, 1977; Albrecht et al., 1988). As the
The flow around multiple bluff bodies has been spacing between the obstacles is changed, up to four
extensively studied due to its practical significance in regimes have been identified for both circular (Zdrav-
many engineering applications, for example, in deter- kovich, 1977) and square cylinders (Hangan and Vic-
mining dynamic loading on neighboring structures or kery, 1999) in tandem arrangement. For very small
the influence on heat transfer and dispersion. An obstacle separation, the flow is similar to that for a
important mechanism for flow-structure interaction is single bluff body. For intermediate separations, there is
the coupled interference between the obstacles on the a region of bistable flow, for which two non-harmonic
vortex formation and shedding process. frequencies are reported. Each frequency is associated
Rockwell (1998) provides an encompassing review of with a different mode of motion and occurs intermit-
the studies conducted for two-dimensional bluff bodies. tently. Okajima (1982) reported a similar behavior for
When an obstacle undergoes forced periodic oscilla- two-dimensional, rectangular bluff bodies of sufficiently
tions, the shedding frequency can lock on to the obstacle large streamwise-to-cross-sectional length aspect ratios.
oscillation frequency, if the two are not too different. For larger separations, the shedding frequency from the
Gaydon and Rockwell (1999) have shown that, for a second obstacle locks on to that of the first. For very
fixed obstacle, incident vortices can dramatically shorten large separations, the obstacles shed independently.
the formation length in the cylinder wake and thus Three-dimensional geometries have received much
modify the vortex shedding frequency, which can also less attention, possibly because these often occur deep in
result in lock-on. Similarly, when two cylinders are boundary layers and periodic vortex shedding is sup-
placed in tandem in a uniform stream, the shedding pressed. However, if the boundary layer thickness is less
frequency from the second obstacle can lock on to that than approximately 70% of the body height, vortex
shedding is observed (cf. Castro and Robins, 1977; Sa-
*
Corresponding author. Tel.: +1-519-661-2111; fax: +1-519-661-
kamoto and Haniu, 1988). For surface-mounted rec-
3020. tangular bluff bodies in tandem arrangement, small
E-mail address: [email protected] (R.J. Martinuzzi). separation (intermittent), lock-on and large separation

0142-727X/$ - see front matter
2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.ijheatfluidflow.2004.02.003
 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372 365

regimes similar to those for the two-dimensional based on cube height, H , and free-stream velocity, U,
geometries have been reported by Sakamoto and Haniu was 22,000. The on-coming boundary layer was laminar
(1988) and Havel et al. (2001). However, Martinuzzi and and matched the Blasius profile. At the mounting point
Havel (2000) have shown that the three-dimensional of the first cube, x ¼ 2:5H , the boundary layer thickness
geometry shows a critical range, between the intermit- was d=H
0:07.
tent and the lock-on regimes, where the shedding Surface pressure measurements were made at several
frequency scales inversely with the obstacle spacing. locations on the plate and cube surfaces. The pressure
Effectively, the shedding frequency is locked to the inter- taps were 0.7 mm in diameter. The reference was taken
obstacle cavity geometry. as the free-stream static pressure from the pitot static
In this work, the flow in the cavity between two tube. The frequency response of the system (pressure
surface-mounted cubes in the locked regime is investi- tap, tubing, and sensor) was flat to 200 Hz. The trans-
gated for the purpose of understanding why the shed- ducer resolution was ±0.5 Pa and data was sampled at
ding frequency scales with the obstacle gap and why a 2 kHz.
similar regime is not observed for two-dimensional Velocity data were acquired with a two-component,
obstacles in tandem arrangement. The cubes are TSI series 9000, fiber optic Laser Doppler Velocimetry
mounted two obstacle heights apart and placed in a thin (LDV) system. The optical setup consisted of a 0.453 m
boundary layer. Based on surface pressure and laser focal length lens with a 2.6 · beam expander resulting in
Doppler velocimetry measurements, mean and phase- an effective measurement volume of diameter 46 lm and
averaged representations of the flow field are used to length 340 lm. The LDV processor was a two channel
investigate the shedding process in greater detail and to TSI IFA 655 correlator operated in single-measurement-
discuss some aspects of the turbulence field. per-burst (SMB) mode. The coincidence window be-
tween the two velocity channels was set to 200 ls, based
on preliminary measurements at several locations
2. Experimental details showing the shear stress coefficient was insensitive to the
coincidence window setting in the range 50–500 ls. A
The flow was produced in an open, suction-type wind frequency shift of 5 MHz was applied to both mea-
tunnel with a honeycomb and several fine grids at the surement channels to eliminate directional ambiguity.
inlet. Air flowed through a 4:1 contraction into a The flow was seeded upstream of the honeycomb with
0.46 · 0.46 m2 test section. The inlet profile was uniform an atomized 10:1 water–glycerol mixture. The particle
within 1% of the free stream velocity, U ¼ 8.8 m/s, and number mean diameter was 4 lm with 90% of the par-
the free stream turbulence intensity was less than 1.5% ticles being less than 8 lm. Following Rudoff and
as measured with a hot-wire anemometer. The experi- Bachalo (1991), these particles are expected to follow the
mental set-up and nomenclature are shown schemati- flow.
cally in Fig. 1. Two cubes of dimension H ¼ 0:04 m were The phase-averaged velocity field was analyzed sub-
mounted on a thin (3 mm) flat plate along the stream- ject to the triple decomposition:
wise line of symmetry of the test-section at a distance of Ui ¼ ui þ uci þ u0i
2H from the leading edge. A plate end-flap was adjusted
to control separation at the plate leading edge and where: Ui is the instantaneous velocity, ui is the local
maintain a thin boundary layer. The Reynolds number long-time mean, uci is the fluctuating coherent (phase-
average) and u0i is the turbulent (incoherent) contribu-
tion. The subscript i indicates the physical velocity
component: u (streamwise), v (vertical), and w (span-
wise). The common reference event for all data points
was taken as the peak pressure during each shedding
period measured on the side face of the upstream cube.
Individual cycle periods were determined from the peak-
to-peak time differences. The velocity data were then
6H U redistributed over 20 equally spaced bins to reconstruct
a phase-averaged cycle. The phase-averaged operation is
H
denoted by h  i and represents the transit-time weigh-
δ
ted average within each bin for the first and second
y H moments:
x S
12H z 2H
H hui i ¼ hUi i ¼ ui þ uci and hu0i u0j i;

where the subscripts i and j indicate the component, as

Fig. 1. Schematic representation of the set-up and nomenclature. well as turbulence kinetic energy written explicitly as
366 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372

1 0.25 1.2
hki ¼ bhu02 i þ hv02 iþ < w02 >c:
2 locked 1
0.2
Mean and phase-averaged measurements of the three regime
velocity components were made in ten normal and 0.8
horizontal planes, each containing between 300 and 0.15

StH

StS
1100 measurement points. Typical data rates, N , ranged 0.6
from 200 Hz, in the base regions, to 5 kHz in the shear 0.1
layers. The average shedding frequency was approxi- 0.4
mately 23 Hz. At each point, data was collected for at 0.05
0.2
least 50,000 LDV-events or 1000 shedding cycles. Only
two perpendicular velocity components could be re-
0 0
corded simultaneously. It was thus necessary to measure
0 2 4 S/H 6
each plane twice to obtain the three velocity component
pairs u, v and u, w. The redundant data, usually the Fig. 2. Shedding frequency measured downstream of the second cube
streamwise component, u, were used to verify repeat- in terms of the non-dimensional Strouhal numbers based on H , StH
ability and uncertainty estimates. (closed symbols), and on S, StS (open symbols), as functions of the gap
The Taylor micro time scale, Tk , was estimated at size, S.
several stations in the flow field from the autocorrela-
tion function of the streamwise velocity. The effective quency asymptotically approaches the value for isolated
data density, NTk , was greater than 5 everywhere, cubes.
which according to Edwards (1987) ensures very low For the analogous, two-dimensional geometry, i.e.
velocity bias when using transit-time weighting. The square cross-section cylinders in tandem in a uniform
velocity uncertainty was estimated to be 1% in the stream, a cavity locked regime is not observed and there
mean and 5% for phase-averaged statistics in a 95% is an abrupt transition from intermittent to continuous
confidence interval (cf. Benedict and Gould, 1996). vortex shedding regimes occurring at S=H
2 (cf. Havel
These estimates were verified using the redundant u- et al., 2001).
velocity measurements. Further verification was ob- The differences in flow behavior can be conveniently
tained by re-processing selected velocity data using summarized from the changes in the drag coefficient, CD
the pressure reference signal on the side face of the based on the frontal area and free-stream velocity,
downstream cube. Again, these results agree within the measured for the two obstacles as a function of spacing
estimated uncertainty bounds. The measurement vol- as shown in Fig. 3. For the two-dimensional geometry,
ume was positioned using a three-axis traversing plat- CD for both obstacles shows an abrupt drop at S=H
2.
form accurate to 10 lm/m. This spacing coincides with the sudden transition from
intermittent to continuous periodic shedding. For larger
spacing, CD for the first obstacle is approximately that
3. Results for the single body in a uniform stream (CD
2:1). The
lower CD -value for the second body for S=H > 2 results
For low aspect ratio, square cross-section, surface from the reduced flow speed in the wake of the upstream
mounted obstacles arranged in tandem, Sakamoto and obstacle.
Haniu (1988), for obstacles 3H high, and Martinuzzi In contrast, for the three-dimensional geometries, the
and Havel (2000), for cubes, identified four distinct drag coefficient varies smoothly from the intermittent to
shedding regimes on the obstacle spacing, S. The aver- the continuous shedding regimes. For the upstream
age vortex shedding frequency, f , measured down- obstacle, CD initially decreases as S=H increases, reach-
stream of two cubes is shown in Fig. 2 as a function of ing a minimum approximately at the end of the cavity
S=H and is expressed in terms of the non-dimensional locked regime. For the downstream obstacle, the end of
Strouhal number based H , StH ¼ fH =U, and the spac- this regime is marked by an inflection in the CD vs. S=H
ing, StS ¼ fS=U. For the intermittent regime, S=H < 1:5, curve.
vortex shedding is interrupted by periods of random To elucidate why a cavity locked regime is observed
fluctuations. For the cavity locked regime, 1:5 < S= for three-, but not two-dimensional, tandem geometries
H < 2:5, vortex shedding is continuous and the fre- requires a detailed investigation of the velocity field.
quency scales inversely with S. For 4 < S=H < 6, shed- First, the mean velocity field will be described to high-
ding from the downstream cube locks on to those light some of the marking features in the locked regime.
vortices shed from the upstream cube, giving rise to a Subsequently, the shedding mechanism will be discussed
strong harmonic signal. For larger obstacle separations, in light of the velocity fields at different phases of the
the cubes shed independently and the shedding fre- shedding cycle.
 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372 367

2.5 The flow field for two cubes in tandem at a spacing of

S=H ¼ 2 is representative for the cavity locked regime
2 (1:5 < S=H < 2:5). The mean velocity components in
the plane of symmetry, z=H ¼ 0, and in the horizontal
plane, y=H ¼ 0:375, are shown as vectors in Fig. 4. A
CD

1.5
similar mean flow topology is observed in the horizontal
planes 0:05 < y=H < 0:75. For clarity, the wall refers to
1
the solid boundary along y=H ¼ 0, the roof is the cube
locked
regime face along y=H ¼ 1, the upper and lower sides refer to
0.5 the cube faces along z=H ¼ 0:5 and 0.5, respectively.
(a) 0 2 4 S/H 6 8
From Fig. 4(a), the dividing streamline in the sepa-
rated shear layer, originating at the upstream cube
1 leading edge, impinges at the leading edge of the
downstream cube roof. The dividing streamline attaches
0.5 to the leading edge of the roof. There is a strong vertical
flow directed towards the wall along the front face, while
the flow over the roof is attached in the mean as well
CD

0 2D, Ricciardelli (1994)

2D, Hangan and Vickery (1999) through the shedding cycle. This flow pattern is similar
3D, Sakamoto and Haniu (1988) through the locked regime.
-0.5
3D, Present Study (Cubes)
In the horizontal plane, Fig. 4(b), the large, counter-
rotating recirculation vortices between the obstacles,
-1
(b) 0 2 4 S/H 6 8 with foci labelled V 1 and V 2, and the two saddle points,
S1 and S2, are the principal topological features of
Fig. 3. Drag coefficient, CD , vs. S=H for (a) upstream and (b) down- interest. As is discussed below, circulation is periodically
stream obstacles in tandem arrangements of square cross-section
advected downstream from these vortices, in alternating
obstacles. Data shown are those of: Ricciardelli (1994) and Hangan
and Vickery (1999) for tandem square cylinders; Sakamoto and Haniu fashion, to form a vortex street in the wake. The two
(1988) for surface-mounted prisms of height 3H ; and of the present saddle points mark the maximum mean downstream
study for surface-mounted cubes. extent of these vortices. These form due to the inter-
ference between the recirculation vortices and the

1.5
H/y

u/U=1
0.5

0
(a) 2 4 6 8
x/H

V1 S1
1.5

0.5
H/z

-0.5

-1

V2 S2
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
(b) x/H

Fig. 4. Mean flow representation in vector form for tandem cubes of spacing S=H ¼ 2: (a) plane z=H ¼ 0; (b) plane y=H ¼ 0:375.
368 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372

vertical flow along the windward face of the downstream The bottom shear layer now reattaches along the
obstacle. This three-dimensional effect cannot exist in lower face of the cylinder and a counter-clockwise re-
the case of the tandem square cylinder geometry. Also circulation vortex begins to grow upward along the
note that the flow along the side faces of the downstream leeward face. This new vortex displaces the clockwise
obstacle is attached. vortex into the base region. The clockwise vortex is still
The flow patterns for other S=H was investigated by fed by circulation from the upper shear layer and thus
Martinuzzi and Havel (2000). For S=H < 1:5, the continues to grow as well. However, the counter-clock-
dividing streamline of the shear layers generated at the wise vortex eventually extends over the entire leeward
upstream cube leading edges attach on the roof and side face and interferes with upper shear layer, thereby cut-
faces (downstream of the leading edges) of the down- ting off the flow of circulation to the clockwise vortex,
stream cube. Thus the downstream cube interferes with
the flow of circulation (from the upstream cube) into the
forming vortices in the cavity and disrupts the formation
process, resulting in irregularly shed vortices. For
S=H > 3, the separated shear layer attaches upstream of
the second cube, forming a new horseshoe vortex at its
base. A stagnation point appears on the front face of
this cube and the strong downward vertical stream is no
longer observed. This change results in higher pressures
on the front face of the cube, which are observed as an
increase in the drag coefficient, as shown in Fig. 3. The
flow patterns around the upstream cube, however,
change little and it is thus not surprising that the drag
coefficient is insensitive to S=H (see Fig. 3). The saddle
points S1 and S2 merge to form a single saddle point
along the plane of symmetry, which marks the end of the
formation region. The flow separates on the side faces of
the downstream obstacle. For 3 < S=H < 6, the shed-
ding of vortices from the base of the downstream
obstacle are synchronized with those shed from the first
cube, resulting in a strong harmonic peak (see Fig. 2).
For larger S=H , shedding from the two obstacles is no
longer synchronized. In the following section, it is pro-
posed that the triggering mechanism for vortex shedding
in the locked regime is three-dimensional in nature and
thus fundamentally different from the classical, two-
dimensional model. The latter is comprehensively dis-
cussed by Williamson (1996) and is here only briefly
summarized.
Consider the vortex formation process on a cylinder
placed in a uniform flow from left to right. The shedding
cycle, arbitrarily chosen to start with the formation of a
vortex on the upper face, can be described as follows.
The flow separates at the leading edge and reattaches on
the leeward face, giving rise to a clockwise rotating re-
circulation vortex. This vortex grows along the leeward
face, fed by circulation generated at the separation point
on the upper face and entrained into the base region.
The vortex grows until it covers the entire leeward face,
at which point the separation streamline interferes with
the opposing shear layer from the lower face. As a re-
sult, a saddle point forms along the lower shear layer
downstream of the obstacle and cuts off the flow of
circulation from the lower separation point to the Fig. 5. Phase-averaged vector representation of the hui  hwi velocity
counter-clockwise rotating vortex in the base region. components in the plane y=H ¼ 0:25 for phases (a) / ¼ 0, (b) / ¼ 54
This vortex subsequently sheds. and (c) / ¼ 162.
 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372 369

slightly between planes, indicating that the rotation axis

of the flow structures is slightly tilted relative to a ver-
tical axis. Note that the results are presented in an
absolute frame of reference and the sectional stream-
traces are included as visual aids.
Significant differences between the two- and three-
dimensional shedding processes are seen by inspection
of Fig. 5. Unlike the two-dimensional case, well-devel-
oped recirculation vortices are observed in the base re-
gion of the upstream obstacle for all phases. Phase 1
(/ ¼ 0, Fig. 5a) is arbitrarily defined as the phase of the
shedding cycle immediately after the upper vortex has
shed. A saddle point, S1, is located directly downstream
of the upper base vortex, V 1. Note that there is a strong
vertical down flow along the front face of the down-
stream cube (Fig. 6(a)). The dividing streamtrace along
the lower shear layer impinges on the cube corner. The
flow is attached along the lower side of the cube. Cir-
culation is fed to the counter-clockwise base vortex V 2
and both V 1 and V 2 continues to grow as shown in Fig.
5(b), / ¼ 54. In this phase, the spanwise flow along the
front face is directed towards the upper edge (hwi < 0)
and the core of the region of strong (downward, hvi < 0)
vertical flow, shown in Fig. 6(b), reaches its maximum z-
excursion and maximum hvi velocities are located near
the upper cube corner.
The core of the vortex V 2 and the saddle point S1
move downstream until the location of S1 coincides with
the upper leading edge corner of the downstream cube.
There is a rapid switch in the direction of the spanwise
flow, within a shedding phase difference of D/ ¼ 20,
towards the lower leading edge and the vertical stream
moves towards the lower leading edge, as in Fig. 6(c) for
/ ¼ 162, to interfere with the bottom shear layer flow
causing the appearance of a saddle point, and a counter-
clockwise vortex sheds. At this phase of the shedding
cycle, the dividing streamline along the upper shear layer
now impinges on the leading corner of the downstream
Fig. 6. Phase-averaged contour representation of the vertical, hvi, cube, while the lower shear layer no longer attaches on
velocity component in the plane y=H ¼ 0:25 for phases (a) / ¼ 0, (b) the obstacle lower face. In Fig. 5(c), / ¼ 162, the saddle
/ ¼ 54 and (c) / ¼ 162. hvi is normalized by U: (- - -) negative values, point S2 is clearly visible, the vortex V 2 is reduced in size
(––) positive values; contour interval 0.01. and the top vortex, V 1, penetrates deeply into the inter-
obstacle cavity. The half-cycle is then repeated, but 180
out of phase, to lead to the shedding of V 1.
which then sheds. The upper shear layer reattaches on Contrary to the two-dimensional case, the vortices V 1
the upper face and the cycle is repeated. and V 2 are not completely shed. Rather, only part of the
The shedding process for the tandem, surface- circulation is convected downstream in the shed vorti-
mounted cube geometry is shown in Figs. 5 and 6 for ces. Immediately after the upper shedding event, the
selected phases of the shedding cycle. The phase aver- remnant core for V 1 is still visible in the upper half of
aged streamwise, hui, and spanwise, hwi, velocity com- the flow plane (Fig. 5(a), / ¼ 0). Similarly, the remnant
ponents in the plane y=H ¼ 0:25 are shown as vectors in core for V 2 is also observed for / ¼ 162 (Fig. 5(c)).
Fig. 5, while the contours representing the vertical The mechanism triggering shedding is different for the
velocity component, hvi, in Fig. 6. The flow field two- and three-dimensional geometries. In the classical
topology is found to be similar for all horizontal planes (two-dimensional) model, the vortex growing along the
0:05 < y=H < 0:75. However, it is observed that the cylinder face in the base region and eventually extends
location of the nodes and saddles for a given phase differ across the leeward face to interfere with the opposing
370 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372

shear layer, thereby cutting off the flow of circulation to -2

the opposite-rotating vortex in the base region, causing
it to shed completely. For the present three-dimensional 0.01
0.02
0.04
case, however, it is the stream of strong vertical flow -1 V1 0.02
0.03
0.01
along the face of the second cube, which moves towards 0 S1
0.01 0.01
Upper side
the bottom leading edge. The vertical stream thus ap- 0 0 0.030.02
Downstream

z/H
pears to split the vortex, which is partly shed, while the 0 -0.02
0
Cube
-0.01
remainder begins to grow again. 0
-0.01 Lower side -0.03 -0.02
-0.03
V2
-0.04
-0.03
-0.03 -0.03
1 -0.02 -0.03 -0.02
-0.01
4. Discussion 0
-0.01

Some important differences between the random 2

(a) 3 4 5
x/H
6 7
(turbulent) contributions to the fluctuations are
observed between results for the shedding of the three- -2
dimensional, tandem geometry and the classical two-
dimensional (square) cylinder of Lyn et al. (1995). The 0.01 0.02 0.03
turbulent kinetic energy, hki, and the Reynolds shear -1 0.03
0.04
0.02
stress, hu0 w0 i, in the plane y=H ¼ 0:25 are shown in Figs. V1 S1 0.01
0
0.01

7 and 8, respectively, for the phases / ¼ 0 and / ¼ 54. 0

Upper side 0.02

Immediately after the shedding of the upper vortex Downstream

z/H
0 Cube
(/ ¼ 0, Fig. 7(a)), large values of hki are concentrated -0.01
Lower side
-0.03
0 -0.04
0.01
-0.05
-0.02
along the windward face of the downstream cube and V2
-0.03
along the shear layer, extending from the vortex V 2 1 -0.03 -0.01
-0.01 -0.02

0
-2
2
(b) 3 4 5
x/H
6 7
0.04
0.06 0.08
-1 0.1 Fig. 8. Contours of the random contribution to the turbulent shear
V1 stress hu0 w0 i for the phases (a) / ¼ 0 and (b) / ¼ 54. hu0 w0 i is nor-
S1 0.06
0.1 Upper side malized by U2 : (- - -) negative values, (––) positive values; contour
0.08
0.1 0.04
Downstream 0.06 interval 0.01.
z/H

0 Cube
0.04
0.06
Lower side
V2 0.08
0.06
downstream past the cube. The hki values are large
0.08
1 0.08 along the upper shear layer in the vicinity of the remnant
0.04
0.08
0.06 core, V 1, while about the saddle point S1, hki is signif-
icantly smaller. The hu0 w0 i-distribution for / ¼ 0 is
2 shown in Fig. 8(a). Again, large values occur along the
(a) 3 4 5
x/H
6 7
shear layers. However, unlike for hki, large (absolute)
values are also observed in the vicinity of the saddle
-2
points.
0.04 For the single obstacle, two-dimensional geometry at
-1
0.04
0.06 0.08 0.06 an equivalent phase in the shedding cycle, the turbulent
0.1 0.1
0.06V1 S1 field distribution in the formation region shows some
0.08
Upper side similarity: the maximum magnitude for hki and hu0 w0 i is
0.06 Downstream similar and concentrated along the shear layers. Close to
z/H

0 0.04
Cube
0.04
Lower side
0.1 0.08 the plane of symmetry, z=H ¼ 0, and downstream of the
0.06 0.08
V2 0.08 0.06 saddle point, S1, hwi is small while hki is large.
1 0.1 0.06
0.06
For phase / ¼ 54 (Figs. 7(b), 8(b)), the lower vortex
0.04
0.06
penetrates deeply in the cavity. High hki values are
0.04
concentrated about the vortex cores and close to the
2 upper leading corner of the downstream cube, where the
(b) 3 4 5
x/H
6 7
vertical stream interferes with the upper shear layer.
Fig. 7. Contours of the random contribution to the turbulent kinetic Along the front face of the obstacle, hki is generally
energy, hki, for phases (a) / ¼ 0 and (b) / ¼ 54. hki is normalized by lower, while hu0 w0 i is much larger than for / ¼ 0. In
U2 : contour interval 0.02. this phase, hwi is large. For the equivalent phase for the
 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372 371

two-dimensional geometry, Lyn et al. (1995) observe -2

that about z=H ¼ 0, both hki and hu0 w0 i are low in
magnitude. They argue that the large hwi component 0 .0 3
V1 0.05
convects ‘‘free stream’’ fluid into the base region, -1 0.19 0.03
resulting in lower turbulence levels. For the three- S1

dimensional geometry, hki does not decrease as signifi- Upper side

z/H
Downstream
cantly and hu0 w0 i increases in magnitude, being close to 0 Cube
0.03
an absolute maximum as hwi reaches its maximum. Lower side
V2 0.13 0.03
Lyn et al. (1995) analysed the hki-production term:
  1 0.09
0.03 0.09 0.03
02 ohui 02 ohwi 0 0 ohui ohwi
hP i ¼ hu i  hw i  hu w i þ ;
ox oz oz ox
2
for the two-dimensional square cylinder geometry in the (a)
3 4 5
x/H
6 7
base region. An excerpt of their results is shown in Fig.
9, where hP i-contours are shown for a phase corre- -2
sponding approximately to phase / ¼ 0 in this study.
Note that the symbols + and · denote the location of 0.03 0.03

vortex cores and vorticity saddles, respectively. The hP i- -1 0.13

0.09
0.17
contours form a neck from one side of the cylinder to
V1
the other between forming vortices. These features are S1 Upper side

related to large differences in hu02 i and hw02 i and large

z/H
Downstream
0 Cube 0.03
ohui=ox. These regions of high hP i are associated with Lower side
0.15
V2
bursts of large hki observed between shed vortices. Peak 0.03

0.15 0.11
production terms are also observed near, but distinctly 1
0.03
away, from the location of saddle points. These peaks
result mainly from shear contribution to the production
term and coincide with peaks in hu0 w0 i. (b) 3
2
4 5 6 7
x/H
For the present geometry, hP i-contours are shown in
Fig. 10 for / ¼ 0 and 54. Similarly to the two- Fig. 10. Contours of the normalized hki-production terms, hP i, in the
dimensional case, concentrations of large hP i values are plane y=H ¼ 0:25: (a) / ¼ 0, (b) / ¼ 54. hP i is normalized by U3 /H;
observed along the shear layers. However, in the cavity contour interval 0.02.
(i.e. formation or base) region, major differences to the
two-dimensional case are observed. For example, the and across the cavity. In Fig. 7(b) for / ¼ 54, the
contours do not form a neck across the base region and concentration of large hki seen at the upper corner is
production levels in the cavity region remain generally convected downstream at the same time as the lower
low throughout the shedding cycle. This difference is vortex is shed, thus creating a high hki-burst in-phase
probably due to the presence of the downstream cube, with the passage of the vortex. In contrast, for the two-
which limits the magnitude of ohui=ox. Thus the large dimensional case, the hki-burst is generated locally be-
values of hki observed in the cavity region are princi- tween the vortices.
pally generated in the shear layers and convected into Unlike the two-dimensional case, peak production
rates do not coincide with peak hu0 w0 i in the cavity.
These differences are expected to result from three-
dimensional effects, since the major differences are ob-
served upstream of the second cube, where vertical
gradients and hvi are largest.

5. Concluding remarks

In the two-dimensional case, shedding is triggered by

the interference between the forming vortex and the
opposite shear layer. For three-dimensional geometries
Fig. 9. Contour plots of hP i at phase ð9; 19Þ downstream of a square in the locked regime, the vortex sheds as a result of the
cylinder as adapted from Lyn et al. (1995). The shedding phase cor-
responds approximately to phase / ¼ 0 in this study. Symbols þ and
interference between the vertical stream along the front
 denote vortex cores and vorticity saddle points, respectively. face of the downstream obstacle and the side shear
Contour intervals for x < 3:5 are 0.04 and for x > 3:5, 0.008. layers. The opposite vortex induces the spanwise flow
372 R.J. Martinuzzi, B. Havel / Int. J. Heat and Fluid Flow 25 (2004) 364–372

across the face of the downstream cube, which provides Gaydon, M., Rockwell, D., 1999. Vortices incident upon oscillating
the opposing shear layer coupling necessary for stable, cylinder: Flow structure and loading. J. Fluids Struct. 13, 709–
722.
periodic shedding. It thus appears reasonable to expect Hangan, H., Vickery, B.J., 1999. Buffeting of 2D (sharp-edged) bluff
that the shedding frequency will scale linearly with the bodies. J. Wind Eng. Ind. Aero. 82, 173–187.
obstacle spacing, since the shedding of the vortex is Havel, B., Hangan, H., Martinuzzi, R., 2001. Buffeting for 2D and 3D
triggered by the vertical stream and that the location of sharp-edged bluff bodies. J. Wind Eng. Ind. Aero. 89, 1369–1381.
this stream is determined by the location of the second Lyn, D.A., Einav, S., Rodi, W., Park, J.H., 1995. A laser-Doppler
velocimetry study of ensemble-averaged characteristics of the
obstacle. When the obstacle spacing is increased suffi- turbulent near wake of a square cylinder. J. Fluid Mech. 304,
ciently, the top shear layer attaches upstream of the 285–319.
downstream obstacle, eliminating the vertical stream Martinuzzi, R., Havel, B., 2000. Turbulent flow around two interfering
and a locked regime is no longer possible. The vertical surface-mounted cubic obstacles in tandem arrangement. Trans.
stream is a three-dimensional effect and does not exist ASME, J. Fluids Eng. 122, 24–31.
Okajima, A., 1982. Strouhal numbers of rectangular cylinders. J. Fluid
for two-dimensional tandem arrangements, which may Mech. 123, 379–398.
explain why a locked regime has not been observed in Ricciardelli, F., 1994. Aerodynamics of a pair of square cylinders,
these cases. M.E.Sc. Thesis. The University of Western Ontario, London,
Ontario, Canada.
Rockwell, D., 1998. Vortex body interactions. Ann. Rev. Fluid Mech.
References 30, 199–229.
Rudoff, R.C., Bachalo, W.D., 1991. Seed Particle Response and Size
Albrecht, T., Barnes, F.H., Baxendale, A.J., Grant, I., 1988. Vortex Characterization in High Speed Flows. Proc. ASME Fluids Eng.
shedding from two cylinders in tandem. J. Wind Eng. Ind. Aero. Conf., Symp. on Laser Anem., Adv., Appl. 2, 443–447.
28, 201–208. Sakamoto, H., Haniu, H., 1988. Aerodynamic forces acting on two
Benedict, L.J., Gould, R.D., 1996. Towards better uncertainty for square prisms placed vertically in a turbulent boundary layer.
turbulent statistics. Exp. Fluids 107, 161–164. J. Wind Eng. Ind. Aero. 31, 41–66.
Castro, I.P., Robins, A.G., 1977. The flow around a surface-mounted Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake.
cube in uniform and turbulent streams. J. Fluid Mech. 79Z, 307– Ann. Rev. Fluid Mech. 28, 477–539, Ann. Rev. in Fluid Mech. 30,
335. 199–229.
Edwards, R.V., 1987. Report of the special panel on statistical particle Zdravkovich, M.M., 1977. Review of flow interference between two
bias problems in laser anemometry. Trans. ASME, J. Fluids Eng. circular cylinders in various arrangements. Trans. ASME, J. Fluids
109, 89–93. Eng. 99, 618–633.