Journal of Fluids and Structures 65 (2016) 217–237

Contents lists available at ScienceDirect

Journal of Fluids and Structures

journal homepage: www.elsevier.com/locate/jfs

Lift forces induced by phase lag between the vortex sheddings

from two tandem bluff bodies
Md. Mahbub Alam
Institute for Turbulence-Noise-Vibration Interaction and Control, Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen
518055, China

a r t i c l e i n f o abstract

Article history: Flow-induced forces on two tandem circular cylinders of identical diameter are numeri-
Received 17 November 2015 cally studied at a Reynolds number of 200. The cylinder center-to-center spacing ratio is
Received in revised form varied from 2 to 9. We focus on fluctuating (rms) lift coefficient of the upstream cylinder,
11 April 2016
vortex dynamics in the gap between cylinders, and phase lag between vortex shedding
Accepted 18 May 2016
Available online 16 June 2016
from the two cylinders. The phase lag was a linear function of the spacing ratio as known
in the literature; but it is, as proved here, indeed a nonlinear function of the spacing ratio,
Strouhal number and convection velocity of vortices in the gap between the cylinders. The
shedding from the two cylinders turns out to be inphase and antiphase alternately as the
spacing ratio increases. We unearth that both phase lag and spacing ratio influence the
fluctuating lift of the upstream cylinder. With an increase in the spacing ratio, while the
influence of the spacing ratio on fluctuating lift diminishes rapidly in an overdamped
manner, that of the phase lag makes the fluctuating lift variation underdamped sinusoidal.
The inphase and antiphase flows correspond to a local maximum and a local minimum,
respectively, in the fluctuating lift variation. An equation is deduced, showing the re-
lationship between the fluctuating lift, spacing ratio, and phase lag. The physics behind
the damped-sinusoidal variation in the fluctuating lift is discussed. The investigation di-
rects that the streamwise separation between two tandem wings of airplanes/submarines
should be taken into account or optimized. It would also be interesting to see whether fish
exploits phase-lag-induced lift to enhance its forward thrust.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Slender structures appear in groups in many engineering applications, for example, chimney stacks, tube bundles in heat
exchangers, cooling of electronic equipment, high-rise buildings, overhead power-line bundles, bridge piers, stays, and
chemical-reaction towers. Two cylinders are regarded as a simple model to study the fluid dynamics around multiple
cylinders. Flow interference between two cylinders is non-linear and very complex as Reynolds number Re ( = U∞ D/ν ) and
spacing ratio Ln ( ¼L/D) are changed, where U1 is the freestream velocity, D is the cylinder diameter, ν is the kinematic
viscosity of fluid, and L is the cylinder center-to-center spacing. The nonlinear interference-induced force on multiple cy-
linders, causing flow-induced vibrations, is a common problem in the engineering applications.
Zdravkovich (1977, 1987) reviewed the problem of flow interference between two cylinders in tandem, side-by-side and
staggered arrangements. Three flow regimes were identified for tandem arrangement: extended-body/overshoot regime

E-mail addresses: alamm28@yahoo.com, alam@hitsz.edu.cn

http://dx.doi.org/10.1016/j.jfluidstructs.2016.05.008
0889-9746/& 2016 Elsevier Ltd. All rights reserved.
218 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Nomenclature Vc convection velocity of vortices

U1 freestream velocity
D cylinder diameter ϕ phase lag between the vortex sheddings from
L spacing between the centers of the upstream the two cylinders
and downstream cylinders Lc critical spacing
(x, y) coordinate with the origin at the upstream Re Reynolds number (¼U1 D/ν)
cylinder center u local streamwise velocity
CD time-mean drag coefficient v local cross-stream velocity
CLf fluctuating (rms) lift coefficient ν kinematic viscosity of fluid
St Strouhal number (¼fD/U1) δ logarithmic decrement
f shedding frequency supn stands for normalization by U1 and/or D
Nn number of nodes Over- time-mean

(Ln o1.2–1.8, depending on Re), where the free shear layers separated from the upstream cylinder overshoot the down-
stream cylinder, and the flow in the gap between the cylinders is stagnant; reattachment regime (1.2–1.8 oLn o3.4–3.8),
where the free shear layers separated from the upstream cylinder reattach on the downstream one; and co-shedding regime
(Ln 43.4–4.0), where the shear layers do not reattach but roll into the gap between the cylinders. Alam et al. (2003) further
divided the reattachment regime into alternating reattachment (1.5 oLn r3.0) and steady reattachment (3.0 oLn r4.0)
regimes. The boundaries of the different regimes are highly dependent on Re (Alam 2014). The spacing differentiating the
n n
second and third regimes is known as critical spacing Lc . Alam (2014) observed Lc to be 3.7, 3.6, 3.7 and 4.0 at Re¼ 9.7  103,
1.6  104, 3.2  104, and 6.5  104, respectively. Oka et al. (1972) measured vortex shedding frequencies behind the upstream
and downstream cylinders for Re¼ 5.1  103–8.8  103. They did not observe vortex shedding from the upstream cylinder
until Ln ¼ 3.8. Strouhal number St (fD/U1, where f is the shedding frequency) of the downstream cylinder decreased
continuously at Ln ¼1–3.8 and experienced a sudden jump at Ln ¼ 3.8. Tanida et al. (1973) measuring St for 2 oLn o20 found
St jump occurring at Ln ¼5 for Re¼80 and Ln ¼3 for Re¼3.4  103. Igarashi (1981) continued the work for
n n
Re¼8.7  103  5.2  104 and pinpointed Lc ¼3.53. Okajima (1979) from St measurements failed to detect Lc at Re¼ 3.8  105,
n
as no obvious jump in St was detected. They, however, identified Lc ¼3.8 at Re¼ 1.7  10 and 2.5  105. Biermann and
5

Herrnstein (1933), Kostic and Oka (1972), Zdravkovich and Pridden (1977), Novak (1974), Igarashi (1981), Arie et al. (1983),
Kim et al. (2009) and Alam and Meyer (2013) have revealed considerable complexity in fluid dynamics as Ln is changed. The
works mainly focused on time-mean drag CD and St variations with Ln and/or Re. There have been very few scattered studies
involving fluctuating (rms) lift coefficient CLf and fluctuating drag coefficient CDf. Detailed measurements of CLf and CDf
including CD, St, and time-mean and fluctuating pressures at Re¼6.5  104 were done by Alam et al. (2003). They observed
n
that at Ln 4Lc the upstream-cylinder CLf variation with Ln is not monotonic but follows a damped sine curve with local
maxima and minima at different Ln. On the other hand, the downstream cylinder CLf decreases monotonically.
n
Two cylinders at a given Ln (4Lc ) do not shed vortices independently; there is a definite phase lag (ϕ) between the
vortex sheddings from the two cylinders (Sakamoto et al., 1987; Alam et al., 2003, 2006). Experimental data (e.g., Sakamoto
et al., 1987, Re¼5.6  104; Alam et al., 2003, 2006, Re ¼ 6.5  104) showed that ϕ varies almost linearly with Ln. Alam et al.
(2006) used a T-shape plate in the upstream of the upstream cylinder to reduce fluid forces on two cylinders. Their CLf data
for plain cylinders and controlled (T/D¼0.5) cylinders are shown in Fig. 1. They noted that CLf variation with Ln is a damped

Fig. 1. (a) Variation in fluctuating (rms) lift coefficient CLf with Ln for two plain cylinders as sketched in Fig. (b) and for two cylinders with a T-shaped plate
at T/D ¼0.5 as sketched in Fig. (c). Alam et al. (2006). Re¼ 6.5  104.
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 219

sinusoidal (Fig. 1), and the relationship between ϕ and Ln is almost linear. They failed to dig out the physics behind the
damped sinusoidal variation in CLf.
Alam et al. (2006) pioneered that ϕ has a significant effect on CLf of the upstream cylinder. Local maxima and minima of
CLf occur for ϕ ¼2nπ (inphase) and ϕ ¼ (2n þ1)π (antiphase), respectively (where n ¼1, 2,. ). They provided an explanation
that when two cylinders are in an inphase mode, the shedding shear layer of the downstream cylinder accelerates the same-
side shear layer of the upstream cylinder. Similarly, in the next half cycle, the other side shear layer is also accelerated. The
CLf of the upstream cylinder is thus enhanced, showing a local maximum for the inphase mode. On the other hand, when
two cylinders are in an antiphase mode, the shear layer of the downstream cylinder accelerates the same-side shear layer of
the upstream cylinder in which the action of the shear layer that sheds from the opposite side of the upstream cylinder is
reduced; thus, the condition corresponds to a minimum CLf. Between these two extremities, the action changes gradually so
that CLf variation with Ln follows a sine curve with the amplitude decreasing. The explanation was perhaps a best guess
without substantiation. Later, Alam and Zhou (2007) developed a differential equation correlating ϕ, Ln, St and convection
velocity of the upstream cylinder vortices. Mahir and Altac (2008) simulating flow around two heated cylinders extracted
n
Nusselt number, CD, CLf, and St at Re ¼ 200 for Ln ¼2, 3, 4, 5, 7, and 10. With Lc lying between 3 and 4, CLf is obtained as 0.569,
0.517, 0.525 and 0.495 at L ¼ 4, 5, 7, and 10, respectively. The variation in CLf with Ln is not monotonic, CLf decreasing
n

between Ln ¼4 and 5, increasing between Ln ¼5 and 7 and then decreasing again. The authors neither pointed out the
periodic variation, not provided its explanation. Here we focus on these issues.
The objective of this work is to find the relationship between ϕ, CLf and Ln and to provide an insightful physics of the
relationship. A detailed 2-dimensional numerical simulation is conducted at Re¼200. CLf, St and ϕ are estimated and
presented as functions of Ln. Furthermore, to assimilate the insight into the relationship between ϕ, CLf and Ln, we extracted
velocity and pressure fields and information on vortex dynamics. For a single cylinder, vortex shedding happens at ReZ50
(Berger and Wille, 1972; Gresho et al., 1984). A Re¼200 is chosen to guarantee the vortex street behind each of the two
cylinders and to avoid three-dimensionality in the flow.

2. Problem descriptions and basic equations

The governing equations for an unsteady, viscous, laminar and incompressible fluid flow with constant properties are the
continuity and momentum equations expressed in Cartesian coordinate as
⁎
∇. u = 0,

and
⁎
∂u ⁎ ⁎ ⁎ 1 2 ⁎
⁎ + (u ⋅ ∇) u = ∇P + ∇ u,
∂t Re
where the freestream velocity U1 and the cylinder diameter D are used as the reference speed and reference length to
normalize the parameters. un ¼(un, vn) is the normalized velocity field, tn is the nondimensional time, and Pn is the nor-
malized static pressure. The differential equations are coupled and solved for the unknown P n, un and vn. The gravity force is
excluded. The computations are performed based on the finite volume method. While a standard scheme and a second-
order upwind scheme are used to discretize pressure and velocity, respectively, the first-order implicit formulation is
employed for time discretization. On the other hand, the coupling between the pressure and velocity fields is achieved using
the SIMPLE technique. Re ¼200 is adopted in the simulation.
The two tandem circular cylinders of the same diameter D, with cylinder center-to-center spacing L, are placed in a
uniform flow. The schematic diagram of the flow model is given in Fig. 2(a). The computational domain was 65D in the
streamwise direction and 30D in the cross-stream direction, which gives a blockage ratio of 3.3% only. The inlet boundary
was 15D away from the center of the upstream cylinder. An O-xy grid system near the cylinders and a rectangular grid
system away from the cylinders were used as shown in Fig. 2(b and c). The grid system was generated using Gambit. The
number of grids for the O-grid system was 200 in the transverse direction and 60 in the radial direction. Therefore, a total of
200 points were on the cylinder surface. The grid in the radial direction was denser near the cylinder surface with the
nearest grid being 0.005D away from the cylinder surface. While the no-slip boundary conditions are employed on the
surfaces of the cylinders, the free-slip boundary condition is adopted for the upper and lower boundary walls. That is,

un ¼vn ¼0 on the surfaces of the cylinders,

un ¼1, vn ¼0, and Pn ¼0 at the lower and upper walls,
un ¼1, vn ¼0, and Pn ¼ 0 at the inlet, xn ¼ x/D ¼ 15
∂un/∂xn ¼0 and ∂vn/∂xn ¼0 at the outlet, xn ¼50

In the computation domain, the initial flow velocities (at tn ¼0) are given as un ¼1, vn ¼0.
220 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Upper boundary

15D 50D
y
x

Inlet Outlet
15D

Lower boundary

Fig. 2. (a) Cylinder arrangement and definitions of symbols, (b) typical grid structure in the computational domain, and (c) zoom-in view of grids around a
cylinder.

Table 1
Results of grid and time-step independence test. Re ¼ 200.

Time step Parameters Number of nodes, Nn

33,000 38,000 56,000 66,700 89,000

0.05 CD 1.43 1.41 1.40 1.41 1.40

CLf 0.521 0.495 0.486 0.491 0.488
St 0.188 0.192 0.191 0.192 0.189

0.01 CD 1.41 1.40 1.40 1.40 1.41

CLf 0.505 0.488 0.487 0.486 0.487
St 0.189 0.195 0.194 0.194 0.193

0.005 CD 1.41 1.40 1.40 1.39 1.41

CLf 0.498 0.487 0.489 0.489 0.491
St 0.194 0.194 0.193 0.193 0.192
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 221

Table 2
Comparison of CD, CLf and St for a single cylinder at Re¼ 200.

CD CLf St

Present 1.4 0.488 0.195

Meneghini et al. (2001) 1.3 0.495 0.196
Zhao et al. 2005 1.43 0.496 0.196
Liu et al. (1998) 1.32 0.495 0.192
Farrant et al. (2001) 1.37 0.502 0.196
Ding et al. (2007) 1.35 0.467 0.196
Mahir and Altac (2008) 1.38 0.495 0.192
Singha and Sinhamahapatra (2010) 1.34 – 0.195
Dehkordi et al. (2011) 1.33 0.530 0.195
Koda and Lien (2013) 1.37 0.495 0.190

Experiments
Roshko (1954) – – 0.17–0.19
Williamson (1991) – – 0.196

3. Grid independence test and result validation

A grid independence test was done first. Five different meshes (nodes Nn ¼33000, 38000, 56000, 66700, 89000) for a
single cylinder are checked with the time step varied as 0.05, 0.01 and 0.005. Table 1 summarizes the test results, illustrating
the dependence of CD, CLf and St on the grid size and time step. Obviously, the forces and St results converged for Nn ¼38,000
and time step 0.01. At time step 0.01, the changes in CD, CLf and St are o3.1% between Nn ¼33,000 and 38,000 but smaller
(o1.5%) for Nn 438,000. On the other hand, a minor difference (o1.2%) in CD, CLf and St between time steps 0.01 and 0.005
is observed for Nn Z38,000. The time step and Nn are therefore chosen as 0.01 and 38,000, respectively. The Courant number
at this time step and Nn was less than 0.2 in the whole domain. Table 2 compares CD, CLf and St results obtained from the
present simulation with those from the literature. Present data display a good accord with the data by others. The departure
of the present results from those in the other literature is less than 4%, 3%, and 3% for CD, CLf and St, respectively.
Further validation of the simulation is done in Table 3 for two cylinders at Ln ¼3 and 4 corresponding to reattachment
and co-shedding flows, respectively. In the table, subscript ‘1’ and ‘2’ refer to the upstream and downstream cylinders,
respectively. Though CD2 by Mahir and Altac (2008) is overestimated, the present results overall show quite reasonable
agreement with those from the literature.

4. Results and discussion

4.1. Dependence of CD on Ln

Flow is simulated for a total of eighteen Ln (i.e., Ln ¼2.0, 2.5, 3.0, 3.5, 3.65, 3.75. 4.25, 4.5, 4.75, 5.0, 5.25, 5.5, 6.0, 6.5, 7.0,
7.5, 8.0, and 9.0), the first four nestling in the reattachment regime and the rest in the co-shedding regime. Since our goal is

Table 3
Validation of results for tandem cylinders at Ln ¼3 and 4 corresponding to reattachment and co-shedding regimes, respectively. Re¼200.

CD1 CD2 CLf1 CLf2 St

n
L 3
Present 1.018  0.13 0.022 0.212 0.125
Meneghini et al. (2001) 1.0  0.08 – – 0.125
Mahir and Altac (2008) 1.05  0.56 0.020 0.191 0.13
Dehkordi et al. (2011) 1.0  0.08 – – 0.129
Papaioannou et al. (2006) 1.02  0.12 – – 0.128
Koda and Lien (2013) 1.04  0.129 0.016 0.212 0.127

Ln 4
Present 1.255 0.36 0.551 1.117 0.175
Meneghini et al. (2001) 1.18 0.38 0.51 1.153 0.174
Mahir and Altac (2008) 1.34 0.558 0.569 1.407 0.181
Dehkordi et al. (2011) 1.16 0.52 0.575 1.059 0.179
Slaouti and Stansby (1992) 1.11 0.88 0.495 1.273 0.19
Koda and Lien (2013) 1.287 0.44 0.558 1.229 0.170
222 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Fig. 3. (a) Dependence on Ln of time-mean drag coefficient CD. (b, c) Vorticity patterns at Ln ¼ 2.5 and 4.0, respectively. The red and blue colors denote the
positive (anticlockwise) and negative (clockwise) vorticities. (For interpretation of the references to color in this figure legend, the reader is referred to the
web version of this article.)

n n
to pay attention to ϕ and forces at Ln 4Lc , a finer resolution in Ln was adopted near Lc . Fig. 3 shows CD distribution with
n
variation in Ln. In the literature, Lc nestles in Ln ¼3.5–5.0 depending on Re (Xu and Zhou, 2004; Alam, 2014). An obvious
n
jump in both CD is observed between Ln ¼3.5 and 3.65; the exact Lc will, therefore, be between the two Ln. For the sake of a
n
convenient discussion, we will consider Lc E3.65.
n
With Ln, CD of the upstream cylinder wanes in the reattachment regime (Ln oLc ) but augments, albeit slowly, in the co-
n n
shedding regime (L 4Lc ) to approach the single-cylinder drag CD0 ¼1.4. Mizushima and Suehiro (2005), Sharman et al.
n
(2005), Len et al. (2009) and Koda and Lien (2013) observed the same at Re¼100. At Ln oLc , the shear layers separating from
the upstream cylinder reattach onto the downstream one, making the flow between the two cylinders almost symmetric
(Fig. 3b). The flow behind the downstream cylinder is, however, asymmetric due to an alternating shedding from the
n
downstream cylinder (Fig. 3b). CD at Ln oLc declines for the upstream cylinder and improves for the downstream cylinder,
essentially governed by the shear layer reattachment position (Alam et al., 2003; Alam, 2014). The shear layer reattachment
n
and symmetry of the flow between the cylinders no longer persist at Ln 4Lc ; CD of the upstream cylinder again becomes
sensitive to L in a different mechanism. The upstream cylinder at a larger Ln gets a weaker feedback of the downstream
n

cylinder (Fig. 3c). The CD of the upstream cylinder thus approaches CD0. The downstream cylinder CD is much smaller than

Fig. 4. Variation in Strouhal number St with Ln.

 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 223

CD0, increasing with Ln. The smaller CD is attributed to the fact that the downstream cylinder is submerged in the wake of the
upstream cylinder, experiencing smaller local mean velocity and smaller mean pressure magnitude on both rear and front
n
surfaces. Since our objective is to understand the flow and CLf at Ln 4Lc , hereafter attention will be paid to the results at
n n
L 4Lc .

4.2. Relationship between ϕ and Ln

Fig. 4 shows St dependence on Ln. Both cylinders had the same St; one is therefore presented in the figure. The same St for
the two cylinders implies that the two vortex-shedding processes of the two cylinders, respectively, are highly correlated. St
n
is smaller at Ln o Lc , as the shear layers are retarded due to reattachment, taking a longer time to complete the vortex
n
shedding cycle. It jumps at Ln ¼Lc but is yet smaller than an isolated cylinder Strouhal number St0 ¼0.195, increasing with Ln
n n
to attain St0. For L 4Lc , St grows, which can be represented by a best fit curve equation, coefficients obtained using the least
square method, as follows.
⁎
−0.35L
St = 0.195 − 0.08e , (1a)

or
⁎
−0.35L
St = St0 − 0.08e . (1b)
n
The solid line in the figure represents the equation. The ϕ is related to the vortex shedding from both cylinders for Ln ZLc
only. The convective vortices from the upstream cylinder trigger the vortex shedding from the downstream cylinder (Sa-
kamoto et al., 1987; Alam and Sakamoto, 2005; Alam and Meyer, 2011). In other words, when a convective vortex from a side
of the upstream cylinder approaches the same side of the downstream cylinder, the vortex interacts with the shear layer and
results in a binary vortex shedding from the downstream cylinder (Papaioannou et al., 2008). The ϕ is therefore connected
to the time required for a vortex to travel from the upstream cylinder to the downstream one, which leads to the re-
lationship between ϕ and Ln as
⁎
2π fs L D
ϕ= ,
V¯c (2)
n
where L D represents the distance, the vortex travels, between the upstream and downstream cylinders, V¯c is the spatial-
n n
and time-mean velocity of the vortex between the cylinders. Since the equation is valid for Ln 4Lc only, and ϕ ¼2π at Ln ¼Lc
regardless of Re and body shape (Alam and Zhou, 2007), Eq. (2) can be rewritten as
2π St ⁎ ⁎ ⁎ ⁎
ϕ = 2π + (L − L c ) (L ≥ L c )
V¯c /U∞ (3a)

or
St ⁎ ⁎ ⁎ ⁎
ϕ = 2π { 1 + (L − L c )} (L ≥ L c )
V¯c /U∞ (3b)

In order to verify the equation, we estimated phase-averaged local velocity Vc of convective vortices between the two
cylinders for different Ln as shown in Fig. 5. Here Vc can therefore be regarded as the time-mean local velocity and was
estimated only from xn (¼x/D)¼1 to xn ¼Ln  0.5 where the upstream cylinder vortex core was recognized clearly. The Vc/U1
grows rapidly, and then dwindles, forming a hump between the cylinders for each Ln (Fig. 5a–e). The dwindling is due to the
presence of the downstream cylinder. The single cylinder Vc/U1 (Fig. 5f) on the other hand speeds up from 0.32 at xn ¼1 and
approaches Vc/U1 E0.82 at xn E5. Beyond xn E5, Vc/U1 is about 0.83, albeit increasing very slowly with xn. Tanaka and
Murata (1986, Re¼3.9  104), Tyler (1930, Re¼818) and Cantwell and Coles (1983, Re¼1.4  105) reported Vc/U1 E0.79 at
xn ¼ 4, and the plateau was reached with Vc/U1 E0.83  0.84 at xn E7. It seems that all graphs in Fig. 5, except that for the
single cylinder, are self-similar. The self-similarity can be explored if the horizontal axis is presented as xn ¼ 1.0 to xn ¼ Ln 
0.5, as shown in Fig. 6. For clarity, the positions of the two cylinders are given in the appropriate scale on the upper side of
the figure. The Vc/U1 at xn ¼1.0 is small, about 0.23–0.3, while that at xn ¼Ln  0.5 is about 0.5–0.57. Interestingly, it becomes
maximum at about the mid (xn E Ln/2) between the cylinders for all Ln, except for Ln ¼5.25 where the peak is slightly biased
left. The maximum Vc/U1 increases with Ln in general, except for a hold at Ln ¼5.25. The biased and hold at Ln ¼5.25 may be
connected to antiphase shedding occurring at this Ln, which will be shown later.
⁎
x ⁎ ⁎ ⁎
Then spatial mean V̄c of Vc was obtained as V¯c = ∫x⁎2 (Vc dx )/(x2 − x1 ) and presented in Fig. 7, where x1 = 1.0 and
1

x2 = L − 0.5. The V¯c /U∞ is much smaller, increasing from 0.45 at Ln ¼3.65 to 0.75 at Ln ¼9.0. The V̄c can be represented by an
⁎ ⁎

exponential equation as a function of Ln, viz.,

⁎
−0.35L
V¯c /U∞ = 0.83 − 1.3e . (4a)

Since the asymptotic value of V¯c /U∞ for a single cylinder is (V¯c /U∞ )0 ≈ 0.83 (Tanaka and Murata, 1986; Tyler, 1930; Cantwell
224 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Fig. 5. Vortex convection velocity Vc variation with xn at (a) Ln ¼3.65, (b) 4.25, (c) 5.25, (d) 6.5, (e) 7.5, and (f) 1 (single cylinder).

Fig. 6. Self-similarity of the convection velocity Vc for different Ln.

and Coles, 1983), the above equation can be rewritten as

⁎
−0.35L
V¯c /U∞ = (V¯c /U∞ )0 − 1.3e . (4b)

Combining Eqs. (1) and (4) into Eq. (3), ϕ can be expressed as
⎧ −0.35L
⁎
⎫
⎪ 0.195 − 0.08e ⁎ ⁎ ⎪ ⁎ ⁎
ϕ = 2π ⎨
⎪
1 + −0.35L
⁎ (L − L c ) ⎬ (L ≥ L c ).
⎪
⎩ 0.83 − 1.3e ⎭ (5)

Now ϕ is the function of L only.

n
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 225

Fig. 7. Variation in spatial-average convection velocity Vc /U∞ with L*.

Fig. 8. Dependence on Ln of phase lag ϕ obtained from the equations developed and from the fluctuating lifts of the two cylinders.

Fig. 8 compares ϕ estimated from cross-correlation between fluctuating lifts of the two cylinders (open circle), from Eq.
(3) (open triangle) where St and V¯c values are provided from our simulations (i.e., Figs. 4 and 7) and from Eq. (5) (line). The
results from the three methods agree each other. The agreement confirms that Eq. (3) developed to find the relationship
between ϕ, St, V̄c and Ln is ingenious. It is conspicuous that ϕ variation with Ln is not exactly linear (Eq. (5)) but has a bend
with negative curvature particularly at Ln o6 (Fig. 8). The bend results from a higher value of d(V̄ c)/dLn than that of d(St)/dLn
(Figs. 4 and 7; Eqs. (1), (4) and (5). In the literature, assuming that ϕ varies linearly with Ln, a linear curve was fitted, though
a small kink was observed at smaller Ln (e.g., Sakata and Kiya, 1983; Sakamoto et al., 1987; Alam et al., 2002, 2003, 2006;
Alam and Zhou, 2007). The observation and Eqs. (3b) and (5) suggests that ϕ is a nonlinear function of Ln for Ln o7.5 and
almost linear for Ln 47.5. While the nonlinearity arises from the strong dependence of St and V̄c on Ln for Ln o7.5 (Figs. 4 and
7), the linearity results from the weak sensitivity of St and V̄c to Ln for Ln 47.5.

4.3. Relationship between CLf and ϕ

A concurrent plot of ϕ and CLf of the upstream cylinder with change in Ln is presented in Fig. 9(a). Like that in Fig. 1, CLf
n
firstly wanes rapidly (up to Ln E5.25) and then grows before declining again from Ln E 7.5. Given that, at Ln ¼ Lc , ϕ is
approximately 2π, irrespective of bluff body shape (Sakata and Kiya, 1983; Sakamoto et al., 1987; Alam et al., 2006; Alam and
n
Zhou, 2007). Here ϕ at Ln ¼3.65 is slightly higher than 2π, since Lc is not exactly at 3.65 but smaller, between 3.5 and 3.65, as
mentioned before. The CLf variation shows local maxima at L ¼3.65 and 7.5 and a local minimum at Ln E5.25. The former
n

two Ln correspond to ϕ E2π and 4π, respectively, inphase shedding (Fig. 9b, d). On the other hand, the latter corresponds to
ϕ ¼3π, antiphase shedding (Fig. 9c). That is, it can generally be said that the local maximum and minimum of CLf occur when
ϕ ¼2nπ and (2n þ1)π, respectively (where n ¼1, 2, 3 …). The CLf variation with Ln is very similar to a damped oscillation with
the magnitude of the local maximum or minimum diminishing with Ln, insinuating that the CLf variation with Ln is com-
posed of two, one associated with Ln only and the other associated with ϕ. Details of the two components will be provided
later after assimilating how ϕ influences the flow around the upstream cylinder.

4.4. Ln and ϕ effects on CLf

As understood, Ln and ϕ both influence CLf. The influence of Ln is perhaps more complicated than that of ϕ, since the
variation in ϕ is also connected to Ln, meaning that it is hard to extract the influence of Ln solely. As seen before, the CLf
profile of the upstream cylinder is like a sine curve with decaying in amplitude. In other words, the trend of CLf can be taken
226 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Fig. 9. (a) Relationship between ϕ, Ln and CLf of the upstream cylinder (Ln 4 3.65). (b, c, d) Vorticity patterns at Ln ¼ 3.65, 5.25 and 7.5, respectively. The red
and blue colors denote the positive (anticlockwise) and negative (clockwise) vorticities. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.)

as a combination of a decreasing variation and a damped sinusoidal variation with Ln. The former variation can be assumed
to be caused by Ln, while the latter is due to ϕ clearly. The ϕ changes with Ln which means the fluctuation of CLf with Ln is
also dependent on Ln. Indeed, the effect of Ln should be consistent with the decreasing trend. How Ln and ϕ influence CLf can
be explained explicitly by discussing the effect of Ln and ϕ on the intrinsic features of the flow structure.

4.4.1. Effect of Ln on flow structure

The Ln affects the mean parameters to a great extent. Contours of normalized time-mean streamwise velocity u¯ around
⁎

n ⁎
the upstream and downstream cylinders are shown in Fig. 10 at L ¼3.65, 4.25, 5.25, 6.5, and 7.5. The region enclosed by u¯ =0
is known as the recirculation region or wake bubble (Alam et al., 2011). With increase in Ln, (i) the wake bubble size of either
cylinder enlarges particularly in the streamwise direction, the extent is however greater for the downstream cylinder, (ii)
⁎ ⁎
the velocity gradient ∂u¯ /∂x behind the upstream cylinder along the wake centerline yn ¼ 0 becomes greater in the domain
⁎
u¯ >0, and (iii) the negative velocity in the wake bubble of either cylinder is enhanced (Alam et al., 2011). All these ob-
servations imply that a larger Ln enables a greater flow from the freestream sides into the wake behind the upstream
cylinder, which is consistent with the fact that V¯c is higher at a larger Ln (Fig. 7). While the shear layer thickness (say, lateral
width from the cylinder surface to maximum velocity) of the upstream cylinder is narrow, that of the downstream cylinder
is very wide. The narrow shear layer is also reflected from more densely arranged contours near the separation for the
⁎
upstream cylinder than the other. Another interesting feature is that the maximum streamwise velocity u¯ max in the upper or
n ⁎
lower side augments when L is increased. To make the observation more clear, u¯ max is plotted in Fig. 10(f). As seen, it has
almost the same trend as St. Here the curve fitting equation can be expressed as
⁎
⁎ −0.3L
u¯ max = 1.36 − 0.25e (6a)

or
⁎
⁎ ⁎ −0.3L
u¯ max = (u¯ max )0 − 0.25e , (6b)

very similar to the curve fitting equations of St and V¯c /U∞. Again, the first term in the right-hand side is very close to a single
⁎ ⁎
cylinder u¯ max ¼ 1.37 (not shown). Minimum velocity u¯ min in the recirculation bubble behind the downstream cylinder
displays an exponential decrease with Ln (Fig. 10g) with an interruption at Ln ¼ 5.25 where V¯c /U∞ was also interrupted. The
decrease in u¯ min with Ln results from the gradually enhanced flow between the cylinders. The interruption is therefore
⁎

connected to the antiphase shedding occurring at Ln ¼5.25. In other words, an inphase shedding can bring more fluid from
the freestream to the gap between the cylinders than the antiphase. Further evidence will be rendered later.
Fig. 11(a–e) shows time-mean pressure coefficient C¯P that is expected to be closely connected to u¯ . A larger Ln corre-
⁎

sponds to a negatively increased C¯P not only on the upper and lower sides of the cylinder but also in the wake. Minimum
pressure coefficient CP , min on the upper or lower side of the upstream cylinder (Fig. 11f) displays similar trend, magnitude
increasing with Ln, as that in u¯ max , following
⁎

⁎
−0.32L
CP, min = − 1.26 + 0.4e (7a)
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237
Fig. 10. Contours of time-mean streamwise velocity u¯ ⁎ around the upstream and downstream cylinders at (a) Ln ¼3.65, (b) 4.25, (c) 5.25, (d) 6.5, and (e) 7.5. (f) Maximum time-average streamwise velocity u¯max
⁎
,
⁎
extracted from (a–e), on the upper or lower side of the upstream cylinder. (g) Minimum velocity u¯min in the recirculation bubble behind the downstream cylinder, extracted from (a–e).

227
 228
Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237
Fig. 11. Contours of time-mean pressure coefficient C̄P at different Ln. (a) Ln ¼ 3.65, (b) 4.25, (c) 5.25, (d) 6.5, and (e) 7.5. (f) Variation in minimum pressure coefficient C̄Pf , min , extracted from (a–e), on the upstream-
cylinder side surface, with increase in Ln.
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 229

or
⁎
−0.32L
CP, min = − (CP, min )0 + 0.4e . (7b)
⁎
Both u¯ and C¯P fields confirm that the flow around the upstream cylinder and in the upstream-cylinder wake is enhanced
as Ln increases. The enhanced flow causes CD and St to swell with Ln (Figs. 3 and 4). It can be summarized that the
downstream cylinder at smaller Ln obstructs the flow around and behind the upstream cylinder and the obstruction gets
weaker with an increase in Ln. Interestingly, the index coefficient in the second term on the right-hand sides of Eqs. (1), (4),
(6) and (7) is 0.35, 0.35, 0.3 and 0.32, respectively, very close to each other. The index perhaps would be a constant
(¼  0.32) for all these mean parameters.

4.4.2. Effect of Ln on coherent flow

In order to obtain a clearer understanding of the coherent flow, contours of fluctuating (rms) pressure CPf around the upstream
and downstream cylinders are presented in Fig. 12(a–e) for Ln ¼3.65, 4.25, 5.25, 6.5 and 7.5, where Ln ¼3.65, 4.25, and 5.25
correspond to CLf decreasing from a maximum to minimum and Ln ¼5.25, 6.5 and 7.5 correspond to CLf increasing from a
minimum to maximum. When Ln increases from 3.65 to 5.25 (i.e., when the flow structure changes from inphase to antiphase), CPf
on the upper and lower sides of the upstream cylinder weakens rapidly. On the other hand, as Ln changes from 5.25 to 7.5 (i.e,
when the flow structure transmutes from antiphase to inphase), CPf on both sides rises albeit slowly. A higher value is found for
inphase condition Ln ¼3.65 and 7.5, while a relatively small value for antiphase condition Ln ¼5.25. In order to comprehend the
trend more conspicuously, maximum fluctuating pressure CPf,max on the upstream cylinder surface is plotted as in Fig. 12(f). Now
the difference is obvious, CPf,max waning at 3.65oLn o5.25 and boosting at 5.25oLn o7.5. The change is smaller at larger Ln,
consistent with the variation in CLf. The overall trend is similar to that of CLf. It is worthwhile to point out that the recovery of
CPf,max between Ln ¼5.25 and 7.5 is not the same as much as CPf,max loses between Ln ¼ 3.65 and 5.25. That is, though the flow is
inphase at both Ln ¼3.65 and 7.5, CPf,max is not the same at the two Ln. The deficit is obviously connected to the effect of Ln.
The CPf,max appears on the side surface (near the boundary-layer separation) for the upstream cylinder but on the front
surface (at θ E50° measured from front stagnation point) for the downstream cylinder. The difference in CPf,max location
results from the mechanism of producing CPf,max that is largely contributed by alternating vortex shedding for the former
case and by the upstream-cylinder-generated vortices alternately impinging on the front surface of the downstream cy-
linder for the latter case. The CPf,max for the downstream cylinder diminishes with Ln (Fig. 12g). The diminishing rate is again
interrupted at Ln 45.25 where an antiphase shedding turns in to be an inphase.

4.5. Effect of ϕ on the flow structure

As already discussed, ϕ is functions of Ln, St and V¯c , and it mainly influences the fluctuating component of the flow. Since
L mainly influences the mean flow field, the effect of ϕ on the flow field or a parameter can be understood from the
n

instantaneous fluctuating field that can be obtained as the instantaneous field minus mean field, for instance, instantaneous
fluctuating pressure coefficient C˜P = CP − CP , where CP is the instantaneous pressure coefficient and CP is the time-mean
pressure coefficient. In other words, C̃P is the pressure where the datum is the mean pressure.
During Karman vortex shedding from a cylinder, the difference between the flow on the upper and lower sides is the largest
when CL is maximum or minimum. Since our focus is to understand how ϕ influences the flow on the upper and lower sides of
the upstream cylinder, C̃P at the instant corresponding to the minimum CL of the upstream cylinder will be compared at
different Ln. Fig. 13(a–e) illustrates C̃P field at different Ln where the CP field corresponds to the minimum CL (solid circle) as
presented in the CL histories at the top of each Ln. At the top, the solid and dashed lines denote the lift histories of the upstream
and downstream cylinders, respectively. The C̃P for the upstream cylinder is positive on the upper side and negative on the
lower side, which indicates the shedding happening on the lower side of the upstream cylinder. The shedding (negative
pressure on the lower surface) from the downstream cylinder occurs from the lower side at Ln ¼3.65 (inphase), upper side at
Ln ¼5.25 (antiphase), and lower side at Ln ¼7.5 (inphase). It is obvious that C̃P magnitude on either side of the upstream cylinder
dwindles for Ln ¼ 3.65–5.25 where ϕ transforms from inphase to antiphase and escalates for Ln ¼5.25–7.5 where ϕ modifies
from antiphase to inphase; the C̃P trend resembles CLf. A more fragrant topography can be viewed from a plot of maximum and
minimum C̃P (i.e., C̃P,max , C̃P,min ) as a function of Ln (Fig. 13f). In the range Ln ¼3.65–5.25, C̃P, max and C̃P, min wane and augment,
respectively, indicating that the flow over the upper side of the upstream cylinder accelerates with Ln and that over the lower
side decelerates, when ϕ transforms from inphase to antiphase. The opposite phenomenon prevails at Ln ¼ 5.25–7.5.
⁎ ⁎ ⁎ ⁎
The remark is again reflected in u˜ = u − u¯ fields in Fig. 14, where u˜ is presented around the upstream cylinder only.
Negative and positive u˜ on the upper and lower sides, respectively, are the signatures of a smaller un on the upper side and
⁎

a higher un on the lower side, consistent with the positive and negative C̃P , respectively. Here one should not be confused
with negative (reverse flow) velocity on the upper side. As the vortex shedding occurs from the two sides in alternating
⁎ ⁎
fashion, the flow velocity with reference to time-mean flow will be oscillating back and forth. Once more, u˜ max and u˜ min
displayed in Fig. 14(f) reflect the same view, velocity increasing on the upper surface and decreasing on the lower surface
when ϕ changes from inphase to antiphase at Ln ¼ 3.65  5.25, and vice versa at Ln ¼5.25–7.5. It is worth mentioning that
C̃P , max , C̃P , min , u˜ max and u˜ min each at Ln ¼3.65 and 7.5 is not the same, but of smaller in magnitude at Ln ¼ 7.5, though ϕ is
⁎ ⁎

inphase at both L . Now it is clear that the effect of ϕ is dependent on Ln as well, feeble at a larger Ln.
n
 230
Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237
Fig. 12. Contours of fluctuating (rms) pressure coefficient CPf at (a) Ln ¼3.65, (b) 4.25, (c) 5.25, (d) 6.5, and (e) 7.5. (f, g) Variation in maximum fluctuating pressure coefficient C̄Pf , max , extracted from (a–e), on the
upstream and downstream cylinders, respectively, with Ln.
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237
Fig. 13. Contours of C˜P = CP − C¯P around the two cylinders when the upstream cylinder is subjected to the minimum lift (downward) at (a) Ln ¼ 3.65, (b) 4.25, (c) 5.25, (d) 6.5, and (e) 7.5. (f) Dependence on Ln of

231
maximum and minimum C̃P (i.e., C̃P ,max and C̃P ,min ), extracted from (a–e), on the upper and lower sides of the upstream cylinder. The solid and dashed lines in the rectangular box at the top of C̃P contours denote
the time-histories of lifts of the upstream and downstream cylinders, respectively, where the horizontal and vertical axes represent time and lift, respectively.
232 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Fig. 14. Contours of u˜ ⁎ = u⁎ − u¯ ⁎ around the upstream cylinder when the upstream cylinder is subjected to the minimum lift (downward) at (a) Ln ¼ 3.65,
(b) 4.25, (c) 5.25, (d) 6.5, and (e) 7.5. (f) Dependence on Ln of maximum and minimum u˜ ⁎ (i.e., u˜max
⁎ ⁎
and u˜min ), extracted from (a–e), on the upper and lower
sides of the cylinder.

Fig. 15. Contours of v˜ ⁎ = v⁎ − v¯ ⁎ when the upstream cylinder is subjected to the minimum lift (downward) at (a) Ln ¼3.65, (b) 4.25, (c) 5.25, (d) 6.5, and
⁎
(e) 7.5. (f) Variation in maximum and minimum v˜ ⁎ (i.e., v˜max ⁎
and v˜min ), extracted from (a–e), in the near wake with increase in Ln.

⁎ ⁎ ⁎
Similarly, instantaneous fluctuating cross-stream velocity v˜ = v − v¯ is presented in Fig. 15(a-e). Two peaks are observed
along the wake centerline at xn E1.5 and 2.6, respectively. The first peak is positive, caused by the shedding from the lower
side. On the contrary, the second peak is negative, attributed to a separated vortex from the upper side. A small peak at the
⁎
base of the cylinder is induced by the anticlockwise rolling of the vortex from the lower side. v˜ values at the two dominant
⁎ ⁎ n ⁎
peaks (v˜max and v˜min ) are plotted in Fig. 15(f) as a function of L . As noted before, u˜ on the lower-side shear layer decreases
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 233

when the flow changes from inphase to antiphase between Ln ¼3.65 and 5.25. So does v˜max that is associated with the lower
⁎

shear layer. The v˜max recovers at Ln ¼5.25–7.5 where u˜ max does the same. When the velocity in the upper shear layer
⁎ ⁎

n ⁎
increases between L ¼3.65 and 5.25, the v peak magnitude associated with the shear layer decays, as the growing shear
˜
layer on the downstream cylinder tends to pull the downward flow upward. On the contrary, between Ln ¼5.25 and 7.5, the
lower shear layer (which is growing) of the downstream cylinder tends to accelerate the downward flow of the upper shear
layer. That is, an inphase shedding can bring more fluid from the freestream into the gap between the cylinders than the
antiphase.

4.6. Decomposition of the effect of Ln and ϕ on CLf

The effect of Ln and ϕ on CLf can also be extracted from a curve fitting of CLf, assuming CLf is functions of Ln and ϕ only.
Based on the above results and discussion, CLf can be written as
−αL
⁎ ⎛ π⎞
CLf = Ae + η sin ⎜ ϕ + ⎟ + C .
⎝ 2⎠ (8a)

Even for a given ϕ, the influence of ϕ at different L would be different, weaker at a larger L . The η may, therefore, be
n n

again in an exponential form.

Eq. (8a) can then be written as
−αL
⁎
−βL
⁎ ⎛ π⎞
CLf = Ae + Be sin ⎜ ϕ + ⎟ + C .
⎝ 2⎠ (8b)

The constants A, α, B, β and C can be obtained from the CLf vs Ln curve.

CLf data obtained (Fig. 9) are fitted following Eq. (8b) as
⁎
−1.43L
⁎
−0.2L ⎛ π⎞
CLf = 13.8e + 0.036e sin ⎜ ϕ + ⎟ + 0.488.
⎝ 2⎠ (9a)

The last term is equal to a single cylinder CLf, denoted by CLf0. The equation can, therefore, be rewritten as
−1.43L
⁎
−0.2L
⁎ ⎛ π⎞
CLf − CLf 0 = 13.8
 e  + 0.036e sin ⎜ ϕ + ⎟ .
⎝ 2⎠
⁎
L − induced C
  
Lf
ϕ − induced CLf (9b)
n
The equation predicts that CLf ¼ CLf0 at L ¼ 1, consistent with our intuition.
While the first term in the right-hand side is associated with the effect of Ln only, the second term is connected to ϕ,
reflecting the influence of ϕ at different Ln. We will, therefore, call the first and second terms in the right-hand side as the
contributions of Ln and ϕ, respectively.
Original and curve fitting (equation) data are presented in Fig. 16(a), showing the closeness between them. In order to get
the information on the individual contributions of Ln and ϕ to CLf, the values of the first and second terms are given in Fig. 16
(b). As seen, the contribution of Ln to CLf declines rapidly, negligible at Ln 47.5, while that of ϕ persists up to the Ln examined
with the amplitude of the wave (coefficient of the sine term) diminishing slowly. It would be interesting to see the decay of
the wave up to a long Ln. The second term in Eq. (9b) is thus plotted for Ln up to 35 in Fig. 16(c). It is now conspicuous that
the influence of ϕ persists up to Ln ¼25 that is 4 times larger than that for Ln influence. Note that the peak-to-peak period of
the wave is found to be ΔLn ¼3.8, 4.46, 4.38, 4.28 and 4.27 for the first, second, third, fourth and fifth periods, respectively.
The period is smaller in the first period (Ln ¼3.65–7.5) and decreases from the second period, reaching its asymptotic value
at the fifth period at Ln ¼25. The observation again suggests that the relationship between ϕ and Ln is non-linear in the first
period, i.e., up to Ln E7.5.
The degree of the CLf-amplitude decay with Ln at the n-th period can be expressed in a form of logarithmic decrement,
viz.,
(ACLf )n
δ n = ln ,
(ACLf )n + 1 (10a)

where ACLf is the amplitude of CLf.

Obtaining ACLf from Eq. (9b) results in
⁎
−0.2L n
0.036e ⁎
0.2 (L n⁎ + 1− L n ) ⁎
δ n = ln −0.2L n⁎ + 1
= ln e = 0.2ΔL .
0.036e (10b)

Following Eq. (10b) and Fig. 16(c), δ is estimated to be δ1 ¼0.75, δ2 ¼0.89, δ3 ¼0.87, δ4 ¼0.86 and δ5 ¼0.85. The δ is almost
constant after the first period. As we know, by definition, logarithmic decrement is the number of oscillations during which
amplitude is reduced by a factor of e.
As we mentioned, the influence of ϕ weakens as Ln becomes longer. How the ϕ influence decays with Ln can now be
234 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

Fig. 16. (a) Upstream cylinder CLf obtained from simulation and curve fit equation. (b) Decomposed CLf induced by Ln and ϕ. (c) ϕ -induced CLf from Eq. (9)
showing influence of ϕ for a longer range of Ln.

specified by damping ratio ζ =

1
2
. Assuming δ is a constant ( E0.85), we can get ζ ¼0.13, much smaller than 1.0,
⎛ 2π ⎞
1+⎜ ⎟
⎝ δ ⎠

underdamped. That is why, ϕ can influence CLf for a longer Ln than Ln itself. Note that the decay of CLf induced by Ln alone is
exponential without a wave, meaning that ζ Z1.0, critical or overdamped. Now it can be summarized that the Ln effect on
flow interaction is critical or overdamped with an increase in Ln, and the ϕ effect is underdamped. The corollary suggests
that when two bodies are in tandem, ϕ plays a significant role on the lift of the upstream cylinder and its role prevails for a
longer Ln.
It is worth pointing out that airplanes/submarines and many fishes have wings in tandem. The CLf generated on the front
wing may generate unwanted wing vibration and/or noise, reducing the fatigue life or service lift of the wing. The
streamwise separation between two wings might be taken into account and optimized in engineering. An investigation on
two elastic tandem airfoils would be interesting to see the flow-induced responses of the airfoils at inphase and antiphase
conditions depending on the streamwise spacing between them. On the other hand, fish generates forward thrust from
oscillations of their wings. The fish may need less power for the oscillation in the case of the inphase shedding. It would be
interesting to know whether fish exploit this technique as most creatures are naturally optimized. Note that the wings for
either case may not be exactly tandem, but still the mechanism can succeed.

4.7. Discussion

The above interpretation proves that ϕ and Ln have significant effects on the flow around the upstream cylinder. An
increase in Ln speeds up the flow around and behind the upstream cylinder, inducing more fluid from the freestream into
the gap between the cylinders. Global mean parameters, such as St and CD, thus rise with Ln. On the other hand, fluctuating
parameters, such as CLf, CPf, etc., weaken in general, having a wavy variation with Ln, essentially due to the influence of ϕ.
A shear layer shedding or growing is accompanied by a higher velocity and a lower pressure (Braza et al., 1986; Sumner
et al., 1999). When a shear layer from the downstream cylinder grows/accelerates, it, because of the low pressure, pulls the
fluid at the same side of the upstream cylinder and tends to accelerate the flow/shear-layer on the side. A sketch is provided
in Fig. 17, reflecting the overall picture of ϕ influence on the flow. For inphase sheddings of the two cylinders (Fig. 17a), the
growing shear layer from the lower side of the downstream cylinder accelerates the same side shear layer on the upstream
cylinder. In the next half cycle, the other side shear layer will be accelerated. The alternating acceleration of the two shear
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 235

Fig. 17. Sketches showing the effect of shedding phase lag on the flow around and behind the upstream cylinder. (a) Inphase flow, and (b) antiphase flow.

layers of the upstream cylinder results in an enhanced CLf. When the sheddings change from inphase to antiphase, the
velocity in the growing shear layer of the upstream cylinder reduces and that in other side increases, which leads to a decay
in CLf with Ln. At antiphase (Fig. 17b), the growing shear layers of the two cylinders are on opposite sides. The growing shear
layer of the downstream cylinder, therefore, accelerates the flow in the non-growing shear layer of the upstream cylinder,
which results in a minimum CLf. The larger the Ln, the smaller the influence of ϕ or pulling effect, as a larger Ln corresponds
to a greater amount (mass) of fluid between the same-side shear layers of the cylinders. The ϕ effect on the downstream
cylinder CLf is negligible or very small because the downstream cylinder is submerged in the upstream cylinder wake and
has very wide shear layers. In other words, the momentum fluctuation in the upstream-cylinder shear layer is redistributed
over the wide shear layer of the downstream cylinder; the fluctuation in momentum around the downstream cylinder is
thus very less or negligible.

5. Conclusions

A 2-dimensional numerical simulation is conducted of flow around on two tandem circular cylinders of identical dia-
meter at Re¼200. The Ln is systematically varied from 2 and 9. CD, CLf, St, and ϕ are extracted from the simulation results,
then how Ln and/or ϕ influences CD, St and CLf of the upstream cylinder is discussed in light of velocity and pressure fields
and vortex dynamics. The investigation leads to the following conclusions.
Mean flow parameters including St, V¯c and u¯ max all increase exponentially with Ln. The downstream cylinder at a smaller
⁎

L obstructs the flow around and behind the upstream cylinder. A larger Ln thus enhances the flow around the upstream
n

cylinder, enables a greater flow from the freestream sides into the wake, and boosts the recirculation in the wake bubble. All
these contribute to the increase in St, V¯c and u¯ max with Ln. The shear layer thickness is narrow for the upstream cylinder and
⁎

very wide for the downstream cylinder. The ϕ effect on the downstream cylinder CLf is negligible or very small because the
downstream cylinder is submerged in the upstream-cylinder wake and has very wide shear layers.
In coshedding flow regime, the upstream cylinder CLf variation with Ln follows a sine curve with decaying in amplitude,
influenced by Ln and ϕ. The Ln effect causes an exponential decrease in CLf with Ln while the ϕ effect makes the CLf variation
sinusoidal. The CLf variation with Ln is therefore composed of two, one associated with Ln and other associated with ϕ. Local
maximum and minimum of CLf occur when the shear layer shedding of the two cylinders is inphase (ϕ ¼2nπ, n ¼1, 2, 3 …)
and antiphase (ϕ ¼(2nþ1)π), respectively. For inphase flow where the same side shear layer of the two cylinders grow
simultaneously, the growing shear layer of the downstream cylinder, having a low pressure, pulls the same-side growing
shear layer of the upstream cylinder. The velocity in the same-side shear layer of the upstream cylinder thus increases. In the
next-half cycle of the vortex shedding period, the velocity in the other-side shear layer is enhanced. The CLf therefore
236 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237

becomes maximum for inphase flow. For antiphase flow where shear layers grow from the opposite sides of the two cy-
linders, the growing the shear layer of the downstream cylinder pulls the non-growing shear layer of the upstream cylinder,
reducing the lift induced by the growing shear layer of the upstream cylinder. A minimum CLf thus prevails for antiphase
flow. The pulling effect weakens at a larger Ln, as a greater mass of fluid lies between the two same-side shear layers. The
amplitude of the sinusoidal variation of CLf thus diminishes with Ln.
In the literature, it was known that ϕ is a linear function of Ln. The present investigation identifies that it is non-linear at
least Ln ¼7.5, following ϕ = 2π {1 + St
V¯c /U∞
(L
⁎ ⁎
}
− Lc ) . The ϕ is a function of Ln, St and V¯c , influencing the fluctuating com-
ponent of the flow around and behind the upstream cylinder significantly. When ϕ changes from inphase to antiphase with
a variation in Ln, the streamwise velocity increases in the upper shear layer of the upstream cylinder (Figs. 14f and 17), and
the downward flow by the upper shear layer reduces (Figs. 5f and 17). The opposite phenomenon is afoot when the anti-
phase shedding evolves into inphase. The effects of Ln and ϕ on CLf are also extracted from a curve fitting of CLf. An equation
⁎ ⁎
−αL −βL
of CLf is deduced as CLf − CLf 0 = Ae + Be (
sin ϕ +
π
2 ).
ϕ influence persists up to L ¼25 that is four times larger than that for Ln influence. The decay of CLf induced by Ln
The n

alone is exponential without a wave, corresponding to a critical or overdamped interaction. On the other hand, the ϕ results
in a sinusoidal variation in CLf, yielding an underdamped interaction.
The streamwise separation between two wings could be considered or optimized in engineering after a further in-
vestigation on flow-induced response of tandem airfoils at inphase and antiphase conditions. It would also be interesting to
see whether fish exploit this technique to enhance its forward thrust.

Acknowledgment

The author wishes to acknowledge supports given to him from the Research Grant Council of Shenzhen Government
through Grants KQCX2014052114423867 and JCYJ20120613145300404. The contributions of Ma Zhe and Xu Guoqing to the
simulation and preparing figures, respectively, are also gratefully acknowledged.

References

Alam, M.M., Sakamoto, H., 2005. Investigation of Strouhal frequencies of two staggered bluff bodies and detection of multistable flow by wavelets. J. Fluids
Struct. 20 (3), 425–449.
Alam, M.M., Zhou, Y., 2007. Phase lag between vortex shedding from two tandem bluff bodies. J. Fluids Struct. 23, 339–347.
Alam, M.M., 2014. The aerodynamics of a cylinder submerged in the wake of another. J. Fluids Struct. 51, 393–400.
Alam, M.M., Moriya, M., Takai, K., Sakamoto, H., 2003. Fluctuating fluid forces acting on two circular cylinders in a tandem arrangement at a subcritical
Reynolds number. J. Wind Eng. Ind. Aerodyn. 91, 139–154.
Alam, M.M., Moriya, M., Takai, K., Sakamoto, H., 2002. Suppression of fluid forces acting on two square cylinders in a tandem arrangement by passive
control of flow. J. Fluids Struct. 16, 1073–1092.
Alam, M.M., Sakamoto, H., Zhou, Y., 2006. Effect of a T-shaped plate on reduction in fluid force on two tandem cylinders in a cross-flow. J. Wind Eng. Ind.
Aerodyn. 94, 525–551.
Alam, M.M., Meyer, J.P., 2011. Two interacting cylinders in cross flow. Phys. Rev. E 84, 056304–056316.
Alam, M.M., Zhou, Y., Wang, X.W., 2011. The wake of two side-by-side square cylinders. J. Fluid Mech. 669, 432–471.
Alam, M.M., Meyer, J.P., 2013. Global aerodynamic instability of twin cylinders in cross flow. J. Fluids Struct. 41, 135–145.
Arie, M., Kiya, M., Moriya, M., Mori, H., 1983. Pressure fluctuations on the surface of two cylinders in tandem arrangement. ASME J. Fluids Eng. 105, 161–167.
Berger, E., Wille, R., 1972. Periodic flow phenomena. Annu. Rev. Fluid Mech. 4, 313–340.
Biermann, D., Herrnstein Jr., W.H., 1933. The Interference Between Struts in Various Combinations. National Advisory Committee for Aeronautics, United
States468.
Braza, M., Chassaing, P., Haminh, H., 1986. Numerical study and physical analysis of the pressure and velocity fields in the near wake of a circular cylinder. J.
Fluid Mech. 165, 79–130.
Cantwell, B.J., Coles, D., 1983. An experimental study of entrainment and transport in the turbulent near-wake of a circular cylinder. J. Fluid Mech. 136,
321–374.
Dehkordi, B.G., Moghaddam, H.S., Jafari, H.H., 2011. Numerical simulation of flow over two circular cylinders in tandem arrangement. Comput. Fluids 23 (1),
114–126.
Ding, H., Shu, C., Yeo, Y.O., Xu, D., 2007. Numerical simulation of flows around two circular cylinders by mesh-free least square-based finite difference
methods. Int. J. Numer. Methods Fluids 53, 305–332.
Farrant, T., Tan, M., Price, W.G., 2001. A cell boundary element method applied to laminar vortex shedding from circular cylinders. Comput. Fluids 30,
211–236.
Gresho, P.M., Chan, S.T., Lee, R.L., Upson, C.D., 1984. A modified finite element method for solving the time-dependent incompressible Navier–Stokes
equations. Part 2 applications. Int. J. Numer. Methods Fluids 4, 619.
Igarashi, T., 1981. Characteristics of the flow around two circular cylinders arranged in tandem (first report). Bull. JSME 24, 323–331. (1981).
Kim, S., Alam, M.M., Sakamoto, H., Zhou, Y., 2009. Flow-induced vibrations of two circular cylinders in tandem arrangement, Part 1: characteristics of
vibration. J. Wind Eng. Ind. Aerodyn. 97, 304–311.
Koda, Y., Lien, F.S., 2013. Aerodynamic effects of the early three-dimensional instabilities in the flow over one and two circular cylinders in tandem
predicted by lattice Boltzmann method. Comput. Fluids 74, 32–43.
Kostic, Z.G., Oka, S.N., 1972. Fluid flow and heat transfer with two circular cylinders in cross flow. Int. J. Heat Mass Transf. 15, 279–299.
Len, K., Yang, K.S., Yoon, D.H., 2009. Flow-induced forces on two circular cylinders in proximity. Comput. Fluids 38, 111–120.
Liu, C., Zheng, X., Sung, C.H., 1998. Preconditioned multigrid methods for unsteady incompressible flows. J. Comput. Phys. 139, 35–37.
Mahir, N., Altac, Z., 2008. Numerical investigation of convection heat transfer in unsteady flow past two cylinders in tandem arrangements. Int. J. Heat Fluid
 Md.M. Alam / Journal of Fluids and Structures 65 (2016) 217–237 237

Flow 29, 1309–1318.

Meneghini, J.R., Saltara, F., Siqueira, C.L.R., Ferrari Jr., J.A., 2001. Numerical simulation of flow interference between two circular cylinders in tandem and
side-by-side arrangements. J. Fluids Struct. 15, 327–350.
Mizushima, J., Suehiro, N., 2005. Instability and transition of flow past two tandem circular cylinders. Phys. Fluids 17, 104107.
Novak, J., 1974. Strouhal number of a quadrangular prism, angle iron and two circular cylinders arranged in tandem. Acta Technica CSAV 19 (3), 361–373.
Oka, S., Kostic, Z.G., Sikmanovic, S., 1972. Investigation of the heat transfer processes in tube banks in cross flow. In: International Seminar on Recent
Development in Heat Exchangers, Trogir, Yugoslavia, 1972 (preprint).
Okajima, A., 1979. Flow around two tandem circular cylinders at very high Reynolds numbers. Bull. JSME 22 (166), 504–511.
Papaioannou, G.V., Yue, D.K.P., Triantafyllou, M.S., Karniadakis, G.E., 2008. On the effect of spacing on the vortex-induced vibrations of two tandem cy-
linders. J. Fluids Struct. 24, 833–854.
Papaioannou, G.V., Yue, D.K.P., Triantafyllou, M.S., Karniadakis, G.E., 2006. Three-dimensionality effects in flow around two tandem cylinders. J. Fluid Mech.
558, 387–413.
Roshko, A., 1954. On the drag and shedding frequency of two-dimensional bluff bodies. National Advisory Committee for Aeronautics (NACA), Washington
(Technical Note 3169).
Sakamoto, H., Haniu, H., Obata, Y., 1987. Fluctuating forces acting on two square prisms in a tandem arrangement. J. Wind Eng. Ind. Aerodyn. 26, 85–103.
Sakata, I., Kiya, M., 1983. Fluctuations acting on two circular cylinders in tandem arrangement. Trans. JSME 49, 2618–2623. (in Japanese).
Sharman, B., Lien, F., Davidson, L., Norberg, C., 2005. Numerical predictions of low Reynolds number flows over two tandem circular cylinders. Int. J. Numer.
Methods Fluids 47 (5), 423–447.
Singha, S., Sinhamahapatra, K.P., 2010. High-resolution numerical simulation of low Reynolds number incompressible flow about two cylinders in tandem.
ASME J. Fluids Eng. 132 (1), 011101.
Slaouti, A., Stansby, P.K., 1992. Flow around two circular cylinders by the random-vortex method. J. Fluids Struct. 6, 641–670.
Sumner, D., Price, S.J., Paidoussis, M.P., 1999. Tandem cylinders in impulsively started flow. J. Fluids Struct. 13, 955–965.
Tanaka, S., Murata, S., 1986. An investigation of the wake structure of a circular cylinder using a computer aided flow visualization. Bull. JSME 29,
1446–1459.
Tanida, Y., Okajima, A., Watanabe, Y., 1973. Stability of a circular cylinder oscillating in uniform flow or in a wake. J. Fluid Mech. 61, 769–784.
Tyler, E., 1930. A hot-wire method for measurement of the distribution of vertices behind obstacles. Philosophical Magazine 7th Series 9, pp. 1113–1130.
Williamson, C.H.K., 1991. 2-D and 3-D aspects of the wake of a cylinder, and their relation to wake computations, In Vortex Dynamics and Vortex Methods
(eds C. R. Anderson and C. Greengard, Lectures in Applied Mathematics, Vol. 28, American Mathematical Society, pp. 719–751.
Xu, G., Zhou, Y., 2004. Strouhal numbers in the wake of two inline cylinders. Exp. Fluids 37, 248–256.
Zdravkovich, M.M., Pridden, D.L., 1977. Interference between two circular cylinders; series of unexpected discontinuities. J. Ind. Aerodyn. 2, 255–270.
Zdravkovich, M.M., 1977. Review of flow interference between two circular cylinders in various arrangements. ASME J. Fluids Eng. 99, 618–633.
Zdravkovich, M.M., 1987. The effects of interference between circular cylinders in cross flow. J. Fluids Struct. 1, 239–261.
Zhao, M., Liang, C., Teng, B., Liang, D., 2005. Numerical simulation of viscous flow past two circular cylinders of different diameters. Appl. Ocean Res. 27,
39–55.