You are on page 1of 11

KSCE Journal of Civil Engineering (0000) 00(0):1-11 Water Engineering

Copyright ⓒ2015 Korean Society of Civil Engineers


DOI 10.1007/s12205-015-0602-y pISSN 1226-7988, eISSN 1976-3808
www.springer.com/12205
TECHNICAL NOTE

Numerical Simulation of Flow Past Two Circular Cylinders in Tandem


and Side-by-side Arrangement at Low Reynolds Numbers
Huy Cong Vu*, Jungkyu Ahn**, and Jin Hwan Hwang***
Received October 10, 2014/Revised February 28, 2015/Accepted May 22, 2015/Published Online June 22, 2015

··································································································································································································································

Abstract

A numerical simulation was used to compare the characteristics of flows past two cylinders in tandem versus side-by-side
arrangements. Numerical experiments were performed at various Reynolds numbers and with different distances between the two
cylinders. Diverse characteristics of flow, such as drag force, vortex shedding, and pressure distributions were investigated here. In
tandem arrangement, at a certain distance, the drag coefficients of both cylinders change abruptly. However, at other distances, the
drag coefficients vary somewhat linearly with the distance between cylinders. In side-by-side deployment, when the distance is more
than three times the cylinders’ diameter, the higher Reynolds number has a smaller drag coefficient (as in the case of a single
cylinder). At points where the flow pattern, drag, and pressure coefficients change significantly, the critical spatial ratios between the
two cylinders were determined at different Reynolds numbers. In addition to the study of such changes in flow characteristics, the
effect of flow pattern on the pressure field and drag force coefficients, as well as their relationship, was also investigated.
Keywords: drag coefficient, circular cylinder, vortex shedding, wake, separation
··································································································································································································································

1. Introduction using either experimental or numerical methods (eg., Williamson,


1985; Huhe-Aode et al.,1985; Li et al., 1991; Surmas et al.,
Cylindrical structural bodies are easily observed in natural 2004; Mittal et al., 1997; Meneghini and Saltara, 2001; Sharman
streams and they significantly modify flow patterns and physical et al., 2005; Lijungkrona, et al., 1991; Zhou and Yiu, 2006).
characteristics, such as drag and lift forces, vortex streets, and These investigators found that flow characteristics depend on the
wake structures. The fundamentals of flow around an isolated Reynolds number (Re) and on the ratio of the distance between the
single circular cylinder have been well studied (eg., Li et al., centers of the two cylinders (L) and their combined diameter (D).
1991; Kumar and Mittal, 2006; Wissink and Rodi, 2008). In Flow pattern around cylinders undergoes a sudden transformation
natural stream, however, cylinder-like structures can be found at a critical spacing ratio, (L/D)c. The critical spacing ratio varies
not only singly, but also in groups. Groups are, in fact, more with the Reynolds number (Sumner, 2010). In the tandem
common and cause more complicated flow patterns. For arrangement, if the spacing ratio L/D is smaller than (L/D)c, there
example, the flows around cylindrical structures were analyzed is no vortex shedding from the upstream cylinder, while two
to better understand vegetated systems (Nepf, 1999). vortices could occur on both cylinders for other L/D values. This
The flow around two circular cylinders (which is the simplest was also found to be true for two cylinders in side-by-side
case of a group or multiple structure), has attracted many arrangements.
researchers (e.g., Li et al., 1991; Williamson, 1985; Mittal et al., Previous studies have established the relationship between L/D
1997; Meneghini and Saltara, 2001; Surmas et al., 2004; Sumner, and the variations of drag coefficients in side-by-side and tandem
2010; Zdravkovich, 1977). Zdravkvich (1977) suggested classifying arrangements (e.g., Mittal et al., 1997; Surmas et al., 2004).
two cylinder arrangements into three types: tandem (two cylinders However, many questions relating to flows past two cylinders
are arranged in-line and parallel to the flow direction); side-by- remain unstudied, as both the location of the cylinders as well as
side (two cylinders are arranged in-line and perpendicular to the the Reynolds number interact to affect the flow conditions in a
flow direction); and staggered (all arrangements other than the complex manner. As limited cases of L/Ds were examined in the
previous two cases). Studies on flow around two cylinders in previous studies (Mittal et al., 1997; Surmas et al., 2004;
tandem and side-by-side arrangements have been carried out Meneghini and Saltara, 2001), the sensitivity of flow characteristics,

*Ph.D., Candidate, Dept. of Civil & Environmental Engineering, Dongguk University, Seoul 100-715, Korea (E-mail: vhcongtltd@yahoo.com.vn)
**Member, Research Fellow, Dept. of Civil & Environmental Engineering, Seoul National University, Seoul 151-744, Korea (E-mail: ahnjk@snu.ac.kr)
***Member, Associate Professor, Dept. of Civil & Environmental Engineering, Seoul National University, Seoul 151-744, Korea (Corresponding Author, E-
mail: jinhwang@snu.ac.kr)

−1−
Huy Cong Vu, Jungkyu Ahn, and Jin Hwan Hwang

such as the mean drag coefficient and flow pattern, was not fully respectively.
investigated, along with small adjustments of L/D, especially The stress tensor τ is given by:
around the critical spacing ratio.
T 2
Previous studies have reported slightly different ranges for the τ = µ ( ∇v + ∇v ) – --- ( ∇ ⋅ vI) (3)
3
critical spacing ratio. An experimental study by Huhe-Aode et
al. (1985), reported that (L/D)c = 4.5~5 for Re =100; (L/D)c = where µ is the dynamic viscosity, I is the unit tensor, and the
3.5~4 for Re = 300; and (L/D)c = 3~3.5 for Re =1000. On the second term on the right hand side is the effect of volume
other hand, numerical studies have reported different ranges: (L/ dilation (ANSYS, 2010).
D)c = 3 at Re = 100 (Li et al., 1991); and (L/D)c = 3.75~4 at Re = The Shear Stress Transport (SST) k-omega model was applied
100 (Sharman et al., 2005). Moreover, these studies (numerical to simulate the turbulent flow. The reason for choosing the SST
and experimental) explored limited and small L/D values, and k-omega model will be discussed later. The Semi-Implicit Pressure
did not cover the range where the two cylinders become Linked Equation, SIMPLE, scheme was used to solve the coupled
independent from each other. Therefore, more cases should be equations of the pressure and velocity fields. The convective term
investigated to find the flow characteristics around two cylinders, was discretized by a second order upwind scheme.
especially with large spacing and with several different L/D Figure 1 shows schematic diagrams of the computational
ratios around the critical spacing. domains to simulate flows around single and two cylinders. The
Furthermore, only a few previous studies provided comparable computational domains were meshed with multi-block grid
measurements of the drag force coefficient, pressure coefficient, systems, which consist of body-fitted grids around the cylinders
and flow pattern in a single set of experiments; therefore, it is and rectangular grids for the rest of the domain. Finer cells were
difficult to see the overall trend from these measurements. The used near the cylinder surface, whereas coarser cells were
variations of pressure characteristics on cylinders with the adopted near the wall boundary. Fig. 2 shows the grid around a
respect to the spacing ratio and Reynolds numbers should also be cylinder. Depending on the distance between two cylinders,
investigated, to determine their relationship with flow patterns. the number of grid cells varies, with about 226,000~228,400
This study tried to obtain the flow characteristics around and 228,000~304,000 grid cells for side-by-side and tandem
cylinders for wide ranges of L/D in tandem and side-by-side arrangements, respectively.
arrangements. The sensitivities of drag force, pressure coefficient, The test for the grid independence of the flows around a single
and flow patterns were examined with respect to the Reynolds cylinder (Fig. 1(a)) was performed with three mesh resolutions.
numbers and L/D. Smaller adjustments of L/D were made The numbers of grid cells employed around the circumference of
compared to previous studies to find the critical spacing ratio, the cylinder were 160, 240, and 320. The Reynolds number, as
where drag force varies significantly, and larger spacing ratios defined below in Eq. (4), was set at 200 for this test.
were investigated to determine the limit of influence of one
ρU oD
cylinder on the other. Re = -------------
- (4)
µ
2. Numerical Model and Validation where Uo is the free stream velocity.
The computational domain and boundary conditions for a
Computational fluid dynamics, CFD, has been used by previous single cylinder were the same as that of two cylinders cases.
researchers to calculate the flow around single or multiple Velocity and pressure were uniformly introduced as the upstream
cylinders for a wide range of Reynolds numbers (e.g., Lam et al., and downstream boundary conditions, respectively. No-slip
2008). In the present work, a CFD software, Fluent, which is boundary conditions were imposed on the cylinder surface and
based on the finite volume method to solve the Navier-Stokes the periodic boundary was used at the side walls. The convergence
equation, was used. The equation for the conservation of mass, criterion for the inner iterations was 10−6 for the continuity,
or the continuity equation, is as follows: momentum, and energy equations.
The drag force is calculated by integrating the pressure and
∂ρ
------ + ∇ ( ρv ) = Sm (1) skin friction contribution:
∂t
2π 2π
where ρ is density of water, t is time, v is velocity, and Sm is sink Fd = ∫ Rpcosθ dθ + ∫ Rτw sinθdθ (5)
or source of mass (Hwang et al., 2006). In the present study, Sm 0 0

was zero.
The momentum equation is:

---- ( ρv ) + ∇ ⋅ ( ρvv ) = – ∇p + ∇ ⋅ ( τ ) + ρg + F (2)
∂t
where p is the static pressure, τ is the stress tensor, and ρg and Fig. 1. Computational Domain for: (a) Single Cylinder, (b) Two Cyl-
F are the gravitational body force and external body force, inders in Tandem, (c) Two Cylinders Side-by-side

−2− KSCE Journal of Civil Engineering


Numerical Simulation of Flow Past Two Circular Cylinders in Tandem and Side-by-side Arrangement at Low Reynolds Numbers

Table 1. The Calculated and Measured Results for the Flow Around
a Single Cylinder
Re Investigator Cd St
60 Tritton (1959) (Re=60,5) 1.47 -
Current work (Nc*=240, SST k-w) 1.468 0.137
100 Su and Kang (1999) 1.36 0.163
Lam et al. (2008) 1.36 0.16
Meneghini and Saltara (2001) 1.37 0.165
Mahir and Altac (2008) 1.368±0.029 0.172
Liu et al. (1998) 1.35±0.012 0.164
Ding et al. ( 2007) 1.356±0.01 0.166
Fig. 2. The Computational Mesh (every third grid point is dis- Current work (Nc*=240, SST k-w) 1.366 0.16
played) 200 Lam et al. (2008) 1.32 0.196
Meneghini and Saltara (2001) 1.3 0.196
Williamson (1996) - 0.183
where Fd is the drag force acting on the cylinder, R (=D/2) is the
Norberg (2001) 1.3-1.35 0.19-0.21
radius of the cylinder, τw is the local wall shear stress, and θ is the
Zhang and Dalton (1998) 1.32 0.198
angular displacement from the front stagnation point (θ = 0o).
Ding et al. (2007) 1.348±0.05 0.196
The drag force coefficient, Cd, is given by:
Liu et al. (1998) 1.31±0.049 0.192
2π 2π
2Fd
-=1 1 Current work (Nc*=160, SST k-w) 1.38 0.191
Cd = ------------- --- ∫ Cp cosθdθ + ---------2 ∫ τ w sinθ dθ (6)
ρUoD 2 0
2 Current work (Nc*=240, SST k-w) 1.33 0.186
ρUo 0
Current work (Nc*=320, SST k-w) 1.303 0.182
where Cp is pressure coefficient, as defined in Eq. (7). The first Current work (Nc*=240, K-ε) 1.2 -
term on the right hand side is the pressure drag coefficient (Cdp) 1000 Braza et al. (1986) 1.21 -
and the second term is the friction drag coefficient (Cdf). Current work (Nc*=240, SST k-w) 1.259 0.2
Nc: the number of grid cells on cylinder; SST k-w and K-ε are turbulent
C p = ( p – p∞ ) ⁄ ⎛ --- ρU0⎞
1 2
(7) models.
⎝2 ⎠
In addition to the drag force coefficients, the Strouhal number
is defined as: comparison. While the k-epsilon has been widely used in coastal
models (Jang et al., 2012), we chose SST k-omega model here.
St = fs D ⁄ Uo (8)
The Cd and St showed good agreement with the previous
where fs is the shedding frequency, evaluated by the time series numerical and experimental data for the SST k-omega model,
of the lift force coefficient using Fourier analysis (Meneghini while the k-epsilon model has a smaller drag coefficient
and Saltara, 2001; Surmas et al., 2004). For example, the power (compared with other results). This may be attributed to the
spectrum of the lift force coefficient for a single cylinder at Re = improvement of SST k-omega compared with k-epsilon models
200 is shown in Fig. 3. There is one dominant peak that reflects in predicting the flow separation under adverse pressure
the Strouhal number. gradients (Bardina et al., 1997). The separation is an important
For validation, the drag coefficients and Strouhal numbers for behavior of flow when it passes a circular cylinder. Therefore,
a single cylinder were summarized in Table 1. Two turbulent considering the accuracy of SST k-omega model results
models, SST k-omega and k-epsilon (Standard), were used for compared with the k-epsilon model results, an SST k-omega
model with 240 grid cells around the cylinder was used here.
The distribution of pressure coefficients around a single
cylinder, as defined by Eq. (7) at Reynolds numbers of 50, 100,
and 200 were calculated (shown in Fig. 4). The Cp distributions
around the cylinder were symmetric about the position of θ =
180o; hence this figure only shows Cp for a half of a cylinder. The
maximum pressure coefficient was observed at the frontal
position (θ = 0o); the pressure coefficient then deceases (in
general) as θ increase to 180o. Such trends agree well with the
results of Lixia et al. (2013).
The drag coefficient data for two cylinders in side-by-side and
tandem arrangements were compared with data obtained by
Fig. 3. Power Spectum of the lIft Coefficient Against the Strouhal Surmas et al. (2004), and Meneghini and Saltara (2001), for Re =
Number for Re = 200 200 (as shown in Fig. 5). Their data were limited to L/D = 1.5, 2,

Vol. 00, No. 0 / 000 0000 −3−


Huy Cong Vu, Jungkyu Ahn, and Jin Hwan Hwang

because they studied limited cases of L/D. For a side-by-side


arrangement, as in Fig. 5(b), the drag coefficient decreases
suddenly as L/D increases from 1.1 to 1.5, and then increases
again to a local maximum at around L/D = 2. After that, the drag
coefficient decreases gradually. There trend agrees well with that
observed by Surmas et al. (2004), except for L/D = 1.5. On the
other hand, Meneghini and Saltara (2001) reported a smaller
drag coefficient.
Overall, as the results from previous calculations and agree
well with data from the present experiments, the present results
are valid for appropriately simulating the flows around two
cylinders.

Fig. 4. Pressure Coefficient Distribution Around A Single Cylinder


3. Results and Discussion
at Different Reynolds Numbers
3.1 Tandem Arrangement
In this experiment, the wakes developing behind the upstream
3, and 4. Fig. 5(a) shows that the drag coefficients of both cylinder affect the flow around the downstream cylinder. Below,
cylinders in tandem arrangement agree well with results obtained we discuss the details of flow patterns, drag coefficients, and
by both Surmas et al. (2004) and Meneghini and Saltara (2001). pressure coefficient distributions around two cylinders with
The drag coefficients of upstream, Cd1, and downstream respect to Re and L/D.
cylinders, Cd2, gradually change as L/D increases, until around L/D
= 4. At L/D = 4, where the critical spacing exists, and there is a 3.1.1 Flow Pattern
discontinuous jump of drag coefficients for both upstream and Depending on the distance between the cylinders and the
downstream cylinders. Before this jump, especially, Cd2 can have background flow speed, flows around each cylinder may show
negative value (as will be explained in detail in sections 3.1.2 different patterns (Zdravkovic, 1987; Lijungkrona et al., 1991;
and 3.1.3). After this jump, the drag coefficients gradually Zhou and Yiu, 2006). Therefore, flow patterns were examined by
decrease as L/D increases. The sudden increase of drag changing L/D and Reynolds number. Flow patterns can be
coefficient was clarified by Alam et al. (2003a) with physical classified into four types based on the location and the wake
experiments for a high Reynolds number, Re = 6.5 × 104 (see patterns that developed behind each cylinder (similar to Zhou
Fig. 5(a)). and Yiu (2006)): (i) no vortex shedding, “NV”, at small L/D and
In engineering problems, estimating drag force values is low Reynolds numbers; (ii) single body regime, “SB”, named
always important and such sudden increase or decrease in drag the extended-body regime by Zhou and Yiu (2006), at small L/D
forces can lead to major unexpected problems; therefore, they where two cylinders are close and behave like a single body; (iii)
need to be studied extensively. However, there is a lack information reattachment regime, “RG”, at intermediate L/D, where the free
on the jumps in drag coefficient at the critical spacing at low shear layer from the upstream cylinder reattaches to the downstream
Reynolds numbers. It should be noted that Surmas et al. (2004) cylinder, (iv) vortex shedding regime, “VS”, named the co-
and Meneghini and Saltara (2001) did not discuss in detail either shedding regime by Zhou and Yiu (2006), at large L/D, where
this discontinuity or the trends of Cd when L/D increases over 4, two vortices occur on both cylinders.

Fig. 5. Mean Drag Coefficient Compared with Previous Results at Re = 200 (except for Alam et al. (2003)) for: (a) Tandem, (b) Side-by-
side Arrangement

−4− KSCE Journal of Civil Engineering


Numerical Simulation of Flow Past Two Circular Cylinders in Tandem and Side-by-side Arrangement at Low Reynolds Numbers

Figure 6 shows flow patterns around cylinders under various cylinder’s surface (Li et al., 1991; Alam et al., 2003a). These
conditions. At Re = 60 and 100, there is no vortex shedding attachment points will be discussed in section 3.1.3.
when L/Ds are less than 5.5 and 4.2, respectively, as seen in Figs. At higher values of L/D (L/D > 6, 4.2, 4, and 3.2 for Re = 60,
6(a), (b), and (c). In these cases, the role of the downstream 100, 200, and 1000, respectively), the downstream cylinder is
cylinder was similar to the splitter plate, which prevents the located outside the vortex region formed by the upstream
appearance of vortex shedding, as indicated by Anderson and cylinder. Thus, the separated shear layer formed by the upstream
Szewczyk (1997). Ding et al. (2007) also found that vortex cylinder does not reattach to the downstream cylinder. Vortex
shedding disappeared at Re = 100 with L/D = 2.5. Vortex shedding occurs from both upstream and downstream cylinders
shedding occurs even with a small L/D (of 1.5), for Re = 200 and as shown in Figs. 6(g) - (i) for Re = 60, (f) - (k) for Re = 100; (w)
1000, as seen in see Figs. 6(m) and (n). The free shear layer from - (y) for Re = 200 and (s) - (z) for Re = 1000. This type of flow
the upstream cylinder covers the downstream one completely. pattern is classified as the vortex shedding flow pattern (VS).
So, the two cylinders behave as a single body and there is only The “VS” regime, found at L/D = 4 for Re = 200, is consistent
one vortex street behind the cylinders. Zhou and Yiu (2006) with the results obtained by both Surmas et al. (2004) and
investigated flow patterns and reported similar results at Meneghini and Saltara (2001). The vortex shedding of the
Reynolds number of 7000. According to their studies, the single upstream cylinder becomes obvious as L/D increases. The free
vortex shedding regime occurs at 1 < L/D < 2. shear layer from the upstream cylinder rolls up into vortices
For intermediate L/D (L/D = 5.5 for Re = 60, L/D = 3 ~ 3.9 for before striking the downstream cylinder. It is clear that at low Re,
Re = 200, and L/D = 3 for Re = 1000), the downstream cylinder the wakes behind the upstream cylinder are more elongated
is far enough not to be covered by the shear layer. The shear (compared with cases of higher Re), as shown in Figs. 6(i) and
layer reattaches onto the surface of the downstream cylinder, as (k). This suggests that the vortex at low Re requires a larger L/D
shown in Fig. 6(e) for Re = 60; (p), (r), and (t) for Re = 200; and to roll up completely.
(q) for Re = 1000. The attachment of the shear layer to the The flow patterns around two tandem cylinders with respect to
downstream cylinder is marked by the increase of the pressure L/D and Reynolds number are summarized in Fig. 7. Clearly, the
coefficient at the reattachment point on the downstream flow patterns around two tandem cylinders can be classified into
four categories depending on L/Ds and the Reynolds numbers.
At the lower Reynolds numbers of 60 and 100, with the smaller
L/D, no vortex shedding was observed behind either upstream or
downstream cylinders. On the other hand, the single body regime
with one vortex behind both cylinders was found at a small L/D
(=1.5), at the higher Reynolds numbers of 200 and 1000. The
higher Reynolds number and the smaller L/D (such as L/D > 5.5
~6 for Re = 60 and L/D > 3~3.2 for Re = 1000) are necessary for
the occurrence of the vortex shedding regime.
The critical spacing ratio, (L/D)c, marks the transformation of
the flow pattern into the vortex shedding regime. The (L/D)c is in
the range of: 5.5~6 for Re = 60, Figs. 6(e) and (g); 4~4.2 for Re =
100, Figs. 6(b) and (d); 3.9~4 for Re = 200, Figs. 6(t) and (w);
and 3~3.2 for Re = 1000, Figs. 6(q) and (s). At critical spacing
ratios, the force and pressure coefficients also show a significant
change, which will be discussed in section 3.1.2 and 3.1.3. For
Re = 200 and 1000, the critical spacing ratios agree well with the

Fig. 6. Flow Pattern Shown by Vorticity Distribution Around Cylin-


ders at Different Spacing Ratios, and at Re of: (a) 60, (b)
100, (c) 200, (d) 1000. NV: No-vortex; SB: Single Body;
RG: Reattachment Regime; VS: Vortex Shedding Regime Fig. 7. Classification of Flow Patterns for Two Tandem Cylinders

Vol. 00, No. 0 / 000 0000 −5−


Huy Cong Vu, Jungkyu Ahn, and Jin Hwan Hwang

experimental data reported by Huhe-Aode et al. (1985), who case. Alam et al. (2003a) observed a similar trend for a Reynolds
reported (L/D)c = 3.5~4 for Re = 300, and (L/D)c = 3~3.5 for Re number of 6.5 × 104. When L/D is greater than the critical (L/D)c,
= 1000. However, for Re = 100, (L/D)c = 4~4.2 (the computed the Cd1/Cdo increases for Re = 60, 100, and 200; but decreases
value from this study), is smaller than (L/D)c = 4.5~5 for Re = gradually for Re = 1000 as L/D increases.
100, as reported by Huhe-Aode et al. (1985); and larger than (L/D)c For the downstream cylinder, when two cylinders are close
= 3.75~4, as reported by Sharman et al. (2005), as well as (L/D)c together, Cd2/Cdo increases gradually, even becoming negative at
= 3 as calculated by Li et al. (1991), both for Re = 100. This Re = 100, 200, and 1000 (as shown in Fig. 8). As for the reason
disagreement may be due to the coarse grid cells (190 elements) why the Cd2/Cdo is negative, it may be because the downstream
used in their simulations, in contrast to the present work, which cylinder is immersed in a low pressure region formed by the
used more than 226,000 elements. upstream cylinder. The distribution of Cp around cylinder
corroborates this explanation (see Figs. 9(b) and (d)). The static
3.1.2 Force Coefficient pressure coefficients are smaller at the front cylinder than that at
The drag coefficient ratios Cd1/Cdo (upstream) and Cd2/Cdo, the rear cylinder. Sumner et al. (2005) noted that at these locations,
(downstream) in the various conditions are shown in Fig. 8, the downstream cylinder is completely enclosed by the shear
where Cdo is the drag coefficient for a single cylinder. Within a layers that are generated by the upstream cylinder, and a thrust
single body’s reattachment regime (L/D < 5.5, 3.9, and 3 for Re = force pulls the downstream cylinder close to the wake region and
60, 200, and 1000, respectively), the drag coefficients of the toward the upstream cylinder. Similar to what occurs at the
upstream cylinder are smaller than Cdo, because the downstream upstream cylinder, Cd2/Cdo also increases suddenly at the
cylinder affects the upstream flow conditions. downstream cylinder, when the flow pattern transitions to two
The placement of a cylinder downstream results in an increase vortex shedding regimes at L/D = 5.5~6, 4~4.2, 3.9~4, and 3~
in the static pressure behind the upstream cylinder (compared 3.2 for Re = 60, 100, 200, and 1000, respectively (as shown in
with the static pressure for a single cylinder). For Re = 200, Cp at Fig. 8). At the critical spacing ratio, (L/D)c, the Cd2/Cdo increases
the rear of a cylinder with L/D = 3 and 3.9 is larger than Cp at the more significantly for the downstream cylinder than for the
rear of a single cylinder. This leads to the decrease of the drag upstream cylinder. In addition, there is a sharper increase of Cd2/
coefficient of the upstream cylinder. Surmas et al. (2004), and Cdo at higher Reynolds number.
Meneghini and Saltara (2001) also noted the decrease of the For Re = 1000 and L/D = 3.2~8, the variation of Cd2/Cdo with
upstream cylinder’s drag coefficients at L/D = 1.5. In fact, Cd1/ L/D has a concave characteristic. In this range, Cd2/Cdo decreases
Cdo gradually decreases with L/D until the end of reattachment from L/D = 3.2 to a minimum at around L/D = 5 and then rises
regime, for this range of L/D (L/D = 5.5, 3.9, and 3 for Re = 60, again. This feature also exists at Re = 100 and 200. Previous
200, and 1000, respectively). It is interesting that this transition studies by Alam et al. (2003a) at Re = 6.5 × 104 and Sharman et
of flow pattern accompanies the sudden change of drag al. (2005) at Re = 100 also described similar features. In addition,
coefficient. At the critical spacing (L/D)c, where flow patterns
change to a vortex shedding regime (at L/D = 5.5~6, 4~4.2, 3.9~
4, and 3~3.2 for Re = 60, 100, 200, and 1000, respectively), the
drag coefficient ratio, Cd1/Cdo, jumps. After this sudden increase,
Cd1/Cdo becomes larger than 1 for Re = 1000, but remains smaller
than 1 for Re = 60, 100, and 200. This indicates that in a tandem
arrangement, the drag force coefficient of the upstream cylinder
does not always decrease, compared with the single cylinder

Fig. 9. Mean Pressure Coefficient Distribution Around Two Tan-


dem Cylinders with Different Spacing Ratios (L/D) and Re:
(a) Upstream Cylinder at Re = 200; (b) Downstream Cylin-
der at Re = 200, (c) Upstream Cylinder at Re = 1000, (d)
Downstream Cylinder at Re = 1000. RG: Reattachment
Fig. 8. Ratio of Drag Coefficient Versus L/D for Two Tandem Cylin- Regime; VS: Vortex Shedding Regime. Note: Y-axes Scales
ders at Re = 60, 100, 200, and 1000 differ

−6− KSCE Journal of Civil Engineering


Numerical Simulation of Flow Past Two Circular Cylinders in Tandem and Side-by-side Arrangement at Low Reynolds Numbers

the minimum values in the concavities and the corresponding L/D that the presence of the downstream cylinder does not affect the
values, both, vary with the Reynolds number. The minimum was pressure in front of the upstream cylinder. Hence, the variation of
found at L/D = 6 for Re = 100 in this study, which agrees well drag coefficient on the upstream cylinder (as discussed earlier) is
with the results of Sharman et al. (2005). caused by the change in pressure behind it. In the region behind
The drag coefficient variation may exhibit concave characteristics the cylinder, as L/D increases, the “VS” pressure curves become
because of the changes in the magnitude of vortex shedding smaller for Re = 200 (Fig. 9(a)), but larger for Re = 1000 (Fig.
between the two cylinders as L/D increases. At high Reynolds 9(b)).
numbers (such as Re = 1000), the vortex shedding behind the Based on the above discussion and according to Eq. (6), an
upstream cylinder appears right in front of the downstream increase of the pressure coefficients behind the cylinder results in
cylinder, induces a strong back flow in front of the downstream the decrease of drag coefficients (and vice versa). These features
one and reduces Cd2/Cdo. When L/D > 5, the position of the completely coincide with the variation of Cd1/Cdo, in which Cd1/
appearance of vortex shedding is far from the downstream Cdo increases for Re = 200 and decreases for Re = 1000, for flow
cylinder. following a “VS” pattern (as shown in Fig. 8). Around the
In other words, this vortex shedding becomes weaker when it critical spacing ratio (L/D)c, where the flow pattern transfers
moves toward front of the downstream cylinder, thus, it is not from reattachment (“RG”, L/D = 3.9 for Re = 200 and L/D = 3
enough to reduce Cd2/Cdo. Similarly, at low Reynolds number, Re for Re = 1000) to the vortex shedding regime (“VS”, L/D = 4 for
= 60, although the vortex shedding appears in front of the Re = 200 and L/D = 3.2 for Re = 1000), there is a significant
downstream cylinder and produces a back flow, this is weaker difference between the pressure coefficient curves even for a
and not enough to reduce Cd2/Cdo. After exhibiting the concave slight change of L/D, as shown in Figs. 9(a) and (c).
characteristic, Cd2/Cdo increases gradually with L/D at all As Re becomes larger, differences increase between the pressure
Reynolds numbers. It is interesting to note that the upstream curves. For example, at θ = 180o, the pressure coefficient varies
cylinder becomes an isolated single cylinder (Cd1/Cdo = 1) at L/D from -0.5 to -0.75 when the L/D changes from 3.9 to 4 at Re =
= 14, whereas the downstream cylinder needs a longer distance 200 (Fig. 9(a)), and from -0.65 to -1.1 when the L/D changes
to become independent. This trend agrees well with results from from 3 to 3.2 at Re = 1000 (Fig. 9(c)). These variations in
Sharman et al. (2005), for Re = 100 and L/D < 10. pressure coefficient curves are accompanied by variations in
drag coefficients. A sudden pressure coefficient drop behind the
3.1.3 Pressure Coefficients cylinder when L/D changes from 3.9 to 4, coincides with a
The pressure coefficients, Cp, on the surface of the upstream sudden increase of Cd1/Cdo from 0.762 to 0.969 for Re = 200.
and downstream cylinders are shown in Fig. 9. It is well known Eventually, when L/Ds increase to 14, the pressure coefficient
that a cylinder’s drag coefficient consists of two components: distributions gradually converge to that of a single cylinder case.
form and skin drags. When flow separates from a cylinder, a low Because of the flow disturbances generated by the upstream
pressure region occurs in the wake behind it. The difference in cylinder, the distributions of the pressure coefficients are
pressure in front of and behind the cylinder results in the form complicated around the downstream cylinder as shown in Figs.
drag. 9(b) and (d) for Re = 200 and 1000, respectively. If the downstream
For cylindrical bodies in flow fields, Braza et al. (1986) showed cylinder is submerged in the low pressure region generated by
that form drag dominates over skin drag. Hence, they suggested the upstream cylinder, a low drag force is observed. This drag
that the variation of drag coefficient was mainly due to the force may even be negative if the pressure at the front is smaller
variation in pressure. The form drag, Cdp, can be calculated by than at the rear.
integrating the pressure coefficient (Cp) around cylinder, When two cylinders are close together, the flow patterns are in
according to Eq. (6). Therefore, the pressure distribution on the the reattachment regime or in transition from reattachment to
cylinder determines the drag coefficient characteristics. In other vortex shedding regimes, and the pressure coefficients are all
word, the changes of pressure in Fig. 9 will be consistent with the negative over the whole surface of the cylinder. In this case, the
change of drag coefficient in Fig. 8, as shown in previous part. downstream cylinder is entirely located in the low pressure
In general, the pressure in the wake behind the upstream region caused by the upstream cylinder. Furthermore, at Re =
cylinder (θ = 90~270o), is nearly constant and lower than the 200 for L/D = 3 and L/D = 3.9 and at Re = 1000 for L/D = 3, the
pressure in front (θ = 0~90o), as shown in Figs. 9(a) and (c). The pressure coefficients at the front (θ < 90o) are smaller than those
larger the difference between the pressure in front of and behind at the rear (θ > 90o), as shown in Fig. 9(b) and (d), respectively.
the upstream cylinder, the higher the drag coefficient. The drag When two cylinders are closely placed, the downstream
coefficient of the upstream cylinder is larger than that of the cylinder is shielded from the incoming flow by the upstream
downstream cylinder. cylinder, and thus there is a region of very low pressure in front
Regarding the effect of L/D, the differences in Cps were found of the downstream cylinder. Consequently, the drag coefficient
behind the upstream cylinder (θ = 90~270o), as seen in Figs. 9(a) becomes negative for the downstream cylinder (as shown in Fig.
and (c). In contrast, the region in front of the upstream cylinder 8). When the flow pattern is in the reattachment regime (L/D = 3
was almost independent from the variation of L/D, suggesting and 3.9 for Re = 200, in Fig. 9(b); L/D = 3 for Re = 1000, in Fig.

Vol. 00, No. 0 / 000 0000 −7−


Huy Cong Vu, Jungkyu Ahn, and Jin Hwan Hwang

9(d)), the local maximum pressure occurs at the location of the


reattachment of the shear layer that separated from the upstream
cylinder (Li et al., 1991, and Alam et al., 2003a). As L/D
increases, this location moves forward. For example, as shown in
Fig. 9(b), the maximum pressure changes its location from 72o to
60o when L/D increases from 3 to 3.9, for Re = 200. At L/D = 14,
the local maximum of the pressure coefficient moves to the front
(θ = 0o), which corresponds to the pressure stagnation point.
Similar to the upstream cylinder, the pressure coefficients are
very different on the downstream cylinder when the flow pattern
changes to the vortex shedding regime. As shown in Figs. 9(b)
and (d), when L/D varies from 3.9 to 4 for Re = 200, the pressure
coefficient changes significantly by approximately 0.5 at θ =
180°. This trend is also found for Re = 1000 (Fig. 9(d)). Changes
in pressure coefficient around the critical spacing were found
consistently with changes in drag coefficients, as shown in Fig.
8. Therefore, it can be seen that the change in flow pattern is
accompanied by the variation in the pressure coefficient and the
drag force on both cylinders.

3.2 Side by Side Arrangement

3.2.1 Flow pattern


In the side-by-side arrangement, three types of flow patterns
exist depending on the distance between the two cylinders: (i)
single body, “SB”, when two cylinders are closely placed and act
as a single body; (ii) biased flow pattern, “BF”, at intermediate L/
D, and (iii) flow pattern with synchronized vortex shedding,
“VS”, at large L/D. These terms for flow patterns are similar to
the terminology defined by Sumner et al. (1999) and Zdravkovich
(1987).
Figures 10(a), (b), (m), and (n) show the flow pattern around
two cylinders at L/D = 1.1. In the single body regime, only one
vortex street appears in the wake, because the cylinder walls are
too close to each other for eddies to form in the gap between
them. Thus, only one eddy is shed from the outside of the two Fig. 10. Flow Pattern Shown by Vorticity Distribution Around Cylin-
ders at Different Spacing Ratios and Re = 60, 100, 200,
cylinders and a single vortex street develops behind them, as
1000. SB: Single Body; BF: Biased Flow, and VS: Vortex
explained by Xu et al. (2003). Shedding Regime
In the intermediate range of L/D, the biased flow pattern
(“BF”) was observed. When L/D is in the range of 2 to 3 for Re =
60, the flow shows a biased flow pattern in Figs. 10(c), (e), and and the two vortices behave more independently.
(g). “BF” patterns were also observed when L/D is between 2 The sensitivity of the Reynolds number to the change of flow
and 2.2 for Re = 100 and Re = 200 (Figs. 10(d), (f), (p), and (r)), pattern was also studied. The “VS” regime appears initially
and around L/D = 2 for Re = 1000 (Fig. 10(q)). In the “BF” regime, when L/D becomes 2.2, 3, 3, and 4, respectively, for Re = 1000,
two vortex-sheddings occur behind both cylinders and they 200, 100, and 60. In the “VS” regime, two vortices exist; they
affect each other. The flows in the gap between the cylinders are may be either out of phase or in-phase (Alam et al., 2003b, and
biased to one side, either forming a narrower wake behind the Williamson, 1985). Alam et al. (2003b) pointed out that vortex
upper cylinder and a wider wake behind the lower cylinder (as shedding (out of phase and in phase) appears independent of the
shown in Figs. 10(d) and (f)) or the reverse (Fig. 10(c)). Xu et al. Reynolds numbers. According to Williamson (1985), the out of
(2003) mentioned similar features with L/D = 1.3 for Re = 450 phase vortex shedding should dominate flow patterns. As
and 1000; and Alam et al. (2003b) reported L/D = 1.4 for Re = suggested by that previous study, in the present simulation, out of
5.5 × 104. When L/D becomes larger, two vortex-sheddings phase vortex shedding was observed in most cases (Figs. 10(s),
appear on both cylinders as shown in Figs. 10(i), (k), (t), (u), (x), (t), (u), and (y)). Only one exceptional in phase vortex shedding
and (y). As L/D increases, the two cylinders affect each other less was observed (Fig. 10(x)).

−8− KSCE Journal of Civil Engineering


Numerical Simulation of Flow Past Two Circular Cylinders in Tandem and Side-by-side Arrangement at Low Reynolds Numbers

Cds/Cdo increased with an increase in the Reynolds number (Fig.


12), and the Cds/Cdo = 1.44, 1.61, 1.72, and 1.85 for Re = 60, 100,
200, and 1000, respectively. Around L/D = 1.1~2, the drag
coefficient suddenly drops but in all L/D ranges, the ratio
increases with the Reynolds numbers. At L/D = 14, the drag
coefficients are 97.6, 98.8, 99.2, and 99.6% of the single cylinder
case for Re = 60, 100, 200, and 1000, respectively. In these cases,
cylinders became independent. Surmas et al. (2004) investigated
the effect of L/D on drag coefficient with L/D = 1.5, 2, 3, and 4
and found that the Cds/Cdo were 1.17, 1.19, 1.16, and 1.11,
respectively. Their values agree well with the present study’s
values of Cds/Cdo = 1.1, 1.2, 1.15, and 1.1. Additionally, it was
found that when the distance becomes 14 times larger than
Fig. 11. Classification of Flow Pattern for Two Side-by-side Cylin- diameter (L/D = 14), each cylinder (in the side-by-side arrangement)
ders
behaves independently, like a single cylinder.

Figure 11 summarizes the variation of flow pattern for side-by- 3.2.3 Pressure Coefficient
side arrangements, with respect to L/D and Re. The “SB” regime A quantitative analysis of the features of pressure distributions
occurs at small L/D ratios, regardless of the Reynolds number. will improve the understanding of the effect of the flow
On the other hand, “VS” regimes can develop with a smaller L/D characteristics (such as drag coefficient) on cylinders. In fact, the
for higher Reynolds numbers. These results are in accordance pressure drag coefficient is a direct result of the pressure
with the experimental observations reported by Sumner et al. distribution on the surface of the cylinder. In this section, we
(1999), especially for Re = 1000. They conducted experiments present the numerical results for the distribution of pressure
for Re = 500~3000 and reported single body behavior for 1 < L/ coefficient along the surface of two cylinders in the side-by-side
D < 1.1 ~ 1.2; biased flow regimes for 1.1 ~ 1.2 < L/D < 2 ~ 2.2; arrangement.
and parallel vortex streets for L/D > 2.2. Figure 13 shows the time averaged pressure coefficients (Cp)
of the upper cylinder. When flows are time averaged, upper and
3.2.2 Force Coefficient lower cylinders have symmetric flow patterns and characteristics.
The mean drag coefficient versus the gap spacing ratio is The angle (θ) at the front of the cylinders is in the range of 0~90o
shown in Fig. 12. The simulated results are presented in terms of and 270~360° and the rear angle in the range of 90~270°. In
drag coefficient reduction, Cds/Cdo, defined as: general, from the stagnant point (θ = 0o) to the separation point,
Cps show positive values; and except in that range, the values are
Cds ( Cd1 + Cd2 )
------- = ------------------------- (9) negative. When L/D is large and the two cylinders are independent
C do 2C do
from each other, Cps are symmetric upto θ = 180o; but as the
where Cds is the mean drag coefficient of the systems, and distance ratio of L/D decreases, the asymmetry increases. The
subscripts “1” and “2” refer to the upper and the lower cylinder, asymmetric behavior of the pressure distribution indicates that
respectively. there are differences in flow structure between the upper and
When two cylinders were closely placed with L/D = 1.1, the lower sides of each cylinder, and this is caused by the effect of
the neighboring cylinder. For example, when L/Ds are 1.1 and
1.5, Cps show higher values (in the range of 300~340o, rather
than 0~45o). When the L/D becomes 14, Cps are almost
symmetric upto θ = 180 o.
Due to turbulence at higher Reynolds numbers, the magnitude
of Cps behind the cylinders (θ = 90~270o) decreases, and is
accompanied by strong vortex shedding (Lijungkrona et al.,
1991). At L/D = 1.1, the flow for two cylinders looks like that of
a single elongated body normal to the stream, with the high
pressure region of Cp at the front (θ = 300~340o), as seen in Fig.
13. This high Cp region suggests an explanation for why the drag
coefficient is largest at L/D = 1.1 (Fig. 12). At L/D = 1.1, the
stagnation points located by the maximum Cps are at 322o, 325o,
328o, and 334o for Re = 60, 100, 200, and 1000, respectively
Fig. 12. Ratio of Drag Coefficient (Cds/Cdo) Versus L/D for Two (Fig. 13). As the Reynolds number increases, the stagnation
Side-by-side Cylinders at Re = 60, 100, 200, and 1000 points move further upstream, as expected. Meanwhile, at L/D =

Vol. 00, No. 0 / 000 0000 −9−


Huy Cong Vu, Jungkyu Ahn, and Jin Hwan Hwang

bers; (ii) the single body regime, where two cylinders are so
closely placed that there is only one vortex shedding behind
them; (iii) the reattachment regime, where the free shear
layer from the upstream cylinder reattaches onto the down-
stream cylinder; and (iv) the vortex shedding regime, where
two vortices occur behind the cylinders.
2. Three types of the flow patterns were identified for the side-
by-side arrangement: (i) the single body regime; (ii) the vor-
tex shedding regime (with similar characteristics at both
small and large L/Ds); and (iii) the biased flow pattern
regime, observed at intermediate L/D.
3. The calculated Cd, Cp, and critical spacing ratio, (L/D)c, for
flow pattern transformations agree well with previous inves-
tigations. Such flow patterns determine the effect of cylin-
ders on Cd and Cp. There were significant increases (for the
Fig. 13. Mean Pressure Coefficient Distribution Around Upper Cyl- tandem arrangement) and decrease (for the side-by-side
inder in Side-by-side Arrangement with Spacing Ratio and arrangement) of Cd and Cp when any flow pattern transfor-
Re: (a) Re = 60, (b) Re = 100, (c) Re = 200, (d) Re = 1000. mation occured.
SB: Single Body; BF: Biased Flow, and VS: Vortex Shed- 4. Based on the sudden changes in the flow patterns and Cds, in
ding Regime tandem arrangements, a different critical distance ratio was
found at different Reynolds numbers: (L/D)c was in the
1.1, the minimum Cp exists below the cylinders (around θ = 270o range of 5.5~6, 4~4.2, 3.9~4, and 3~3.2 for Re = 60, 100,
at Re = 1000), indicating a gap-flow of very high speed between 200, and 1000, respectively. For side-by-side arrangements,
the cylinders (Fig. 13(d)). As L/D increases, the region of high Cp (L/D)c depended less on Reynolds numbers, and was around
(θ = 300~340o) becomes narrower, and the stagnation point 1.1~1.5 for all cases.
moves to the front (θ = 0o). For example, at Re = 200, the stagnation 5. Cd and Cp at (L/D)c vary with the Reynolds numbers and are
point moves from θ = 345o to 352o when L/D increases from 1.5 related to each other, as expected. The distribution of Cp on
to 2. the cylinder surface is correlated with the variation of Cd.
The transition of the flow pattern from a single body regime to This effect was more obvious at the downstream cylinder
the biased flow regime occurs along with a sudden change in the than that at the upstream cylinder for tandem arrangements.
pressure coefficient, from L/D = 1.1 to 1.5, for all Reynolds However, even in tandem arrangement, the upstream cylin-
numbers. A big single vortex shedding with low pressure occurs der can be affected by the backwater of the downstream cyl-
at L/D = 1.1; but at L/D = 1.5, this is replaced by two smaller inder.
vortices, shedding behind the two cylinders with higher pressure.
This trend in the change of pressure coefficient at the rear from Acknowledgements
L/D = 1.1 to 1.5 coincides with the flow pattern change, from
“SB” to “BF” (as shown in Fig. 11), and with a sudden drop in This research was a part of the project entitled “Integrated
the drag coefficient (as seen in Fig. 12). Beyond the critical management of marine environment and ecosystem around
spacing values, as L/D increases the pressure coefficient curve Saemangeum” and “Development of integrated Keum river
gradually becomes symmetric (at θ = 180°), because the effects estuarine management system” funded by the Ministry of Ocean
between the two cylinders decease. At L/D = 14, the pressure and Fisheries, Korea.
coefficient curve moves close to that of the single cylinder case;
i.e., the two cylinders become independent. References

4. Conclusions Alam, M. M., Moriya, M., Takai, K., and Sakamoto, H. (2003a).
“Fluctuating fluid forces acting on two circular cylinders in tandem
The present study simulated the flows past two cylinders arrangement at a subcritical Reynolds number.” Journal of Wind
Engineering and Industrial Aerodynamic, Vol. 91, Nos. 1-2, pp.
deployed in tandem and side-by-side arrangements at the various
139-154. DOI: 10.1016/S0167-6105(02)00341-0.
distances from one another and for various Reynolds numbers. Alam, M. M., Moriya, M., and Sakamoto, H. (2003b). “Aerodynamic
The flow pattern, drag, and pressure coefficients were investigated characteristic of two side by side circular cylinders and application
and the results are summarized as follows: of wavelet analysis on the switching phenomenon.” Journal of
1. Four types of flow patterns were identified for two cylinders Fluids and Structure, Vol. 19, Nos. 3-4, pp. 325-346
in the tandem arrangement: (i) the no-vortex shedding Anderson, E. A. and Szewczyk, A. A. (1997). “Effects of a splitter plate
regime, which occurs at small L/Ds and low Reynolds num- on the near wake of a circular cylinder in 2 and 3-dinmensional flow

− 10 − KSCE Journal of Civil Engineering


Numerical Simulation of Flow Past Two Circular Cylinders in Tandem and Side-by-side Arrangement at Low Reynolds Numbers

configurations.” Experiments in Fluids, Vol. 23, No. 2, pp. 161-174, 11, pp. 1315-1344, DOI: 10.1002/(SICI)1097-0363(19971215)
DOI: 10.1007/s003480050098. 25:11<1315::AID-FLD617>3.0.CO;2-P.
ANSYS (2010). ANSYS Fluent theory guide. Nepf, H. M. (1999). “Drag, turbulence, and diffusion in flow through
Bardina, J. E., Huang, P. G., and Coakley, T. J. (1997). Turbulence emergent vegetation.” Water Resources Research, Vol. 35, No. 2, pp.
modeling validation, testing, and development, NASA Technical 479-489, DOI: 10.1029/1998WR900069.
Memorandum 110446. Norberg, C. (2001). “Flow around a circular cylinder: Aspects of fluctuating
Braza, M., Chassing, P., and Ha Minh, M. (1986). “Numerical study and lift.” Journal of Fluids and Structures, Vol. 15, Nos. 3-4, pp. 459-69,
physical analysis of the pressure and velocity fields in the near wake DOI: 10.1006/jfls.2000.0367.
of a circular cylinder.” Journal of Fluid Mechanics, Vol. 165, pp. 79- Sharman, B., Lien, F. S., Davidson, L., and Norberg, C. (2005). “Numerical
130, DOI:10.1017/S0022112086003014. predictions of low Reynolds number flows over two tandem circular
Ding, H., Shu, C., Yeo, K. S., and Xu, D. (2007). “Numerical simulation cylinders.” International Journal for Numerical Methods in Fluids,
of flows around two circular cylinders by mesh-free least square- Vol. 47, No. 5, pp. 423-447, DOI: 10.1002/fld.812.
based finite difference methods.” International Journal for numerical Su, M. and Kang, Q. (1999). “Large eddy simulation of the turbulent
method in Fluids, Vol. 53, No. 2, pp. 305-332, DOI: 10.1002/ flow around a circular cylinder at sub-critical Reynolds numbers.”
fld.1281. ACTA Mechanica Sinica, Vol. 31, pp. 100-105.
Huhe-Aode, Tasuno, M., and Taneda, S. (1985). “Visual studies on wake Sumner, D. (2010). “Two circular cylinders in cross-flow: A review.”
structure behind two cylinders in tandem arrangement.” Reports of Journal of Fluids and Structures, Vol. 26, No. 6, pp. 849-899, DOI:
Research Institute for Applied Mechanics, Vol. XXXII(99). 10.1016/j.jfluidstructs.2010.07.001.
Hwang, J. H., Yamazaki, H., and Rehmann, C. R. (2006). “Buoyancy Sumner, D., Richards, M. D., and Akosile, O. O. (2005). “Two staggered
generated turbulence in stably stratified flow with shear.” Vol. 18, circular cylinders of equal diameter in cross-flow.” Journal of Fluids
No. 4, pp. 045104, DOI: 10.1063/1.2193472. and Structures, Vol. 20, No. 2, pp. 255-275, DOI: 10.1016/
Jang, D., Hwang, J.H., Park, Y., and Park S. (2012). “A Study on salt j.jfluidstructs.2004.10.006.
wedge and river plume in the Seom-Jin river and estuary.” Journal Sumner, D., Wong, S. S. T., Price, S. J., and Paidoussis, M. P. (1999).
of Civil Engineering, Vol. 16, No. 4, pp. 676-688, DOI 10.1007/ “Fluid behavior of side-by-side circular cylinders in steady cross-
s12205-012-1521-9. flow.” Journal of Fluids and Structures, Vol. 13, No. 3, pp. 309-338,
Kumar, B. and Mittal, S. (2006). “Prediction of the critical Reynolds DOI: 10.1006/jfls.1999.0205.
number for flow past a circular cylinder.” Computer Methods in Surmas, R., Stantos Luis, O. E., and Philippi, P. C. (2004). “Lattice
Applied Mechanics and Engineering, Vol. 195, Nos. 44-47, pp. Boltzmann simulation of the flow interference in bluff body wakes.”
6046-6058, DOI: 10.1016/j.cma.2005.10.009. Future Generation Computer System, Vol. 20, No. 6, pp. 951-958,
Lam, K., Gong, W. Q., and So, R. M. C. (2008). “Numerical simulation DOI: 10.1016/j.future.2003.12.007.
of cross-flow around four cylinders in-line square configuration.” Tritton, D. J. (1959). “Experiments on the flow past a circular at low
Journal of Fluids and Structures, Vol. 24, No. 1, pp. 34-57, DOI: Reynolds numbers.” Journal of Fluid Mechanics, Vol. 6, No. 4, pp.
10.1016/j.jfluidstructs.2007.06.003. 547-567, DOI: 10.1017/S0022112059000829.
Li, J., Chambarel, A., Doneaud, M., and Martin, R. (1991). “Numerical Williamson, C. H. K. (1985). “Evolution of a single wake behind a pair
study of laminar flow past one and two circular cylinders.” Computers of bluff bodies.” Journal of Fluid Mechanics, Vol. 159, pp. 1-18,
and Fluids, Vol. 19, No. 2, pp. 155-170, DOI: 10.1016/0045-7930 DOI: 10.1017/S002211208500307X.
(91)90031-C. Williamson, C. H. K. (1996). “Three-dimensional wake transition.”
Lijungkrona, L., Norberg, C., and Sunden, B. (1991). “Free-stream Journal of Fluid Mechanics, Vol. 328, pp. 345-407, DOI: 10.1007/
turbulence and tube spacing effects on surface pressure fluctuations for 978-94-009-0297-8_113.
two tubes in an in-line arrangement.” Journal of Fluids and Structure, Wissink, J. G. and Rodi, W. (2008). “Numerical study of the near wake
Vol. 5, No. 6, pp. 701-727, DOI: 10.1016/0889-9746(91) 90364-U. of a circular cylinder.” International Journal of Heat and Fluid flow,
Liu, C., Zheng, X., and Sung, C. H. (1998). “Preconditioned multigrid Vol. 29, pp. 1060-1070, DOI: 10.1016/j.ijheatfluidflow.2008.04.001.
methods for unsteady incompressible flows.” Journal of Computational Xu, S. J., Zhou, Y., and So, R. M. C. (2003). “Reynolds number effects
Physics, Vol. 139, No. 1, pp. 35-57. DOI: 10.1006/jcph.1997.5859. on the flow structure behind two side-by-side cylinders.” Physics of
Lixia, Q., Norberg, C., Davidson, L., Peng, S. H., and Wang, F. (2013). Fluids, Vol.15, No. 5, pp. 1214-1219.
“Quantitative numerical analysis of flow past a circular cylinder at Zdravkovich, M. M. (1977). “Review of flow interference between two
Reynolds number between 50 and 200.” Journal of Fluids and circular cylinders in various arrangements.” Journal of Fluids
Structures, Vol. 39, pp. 347-370, DOI: 10.1016/j.jfluidstructs.2013.02. Engineering, Vol. 99, No. 4, pp. 618-633, DOI: 10.1115/1.3448871.
007. Zdravkovich, M. M. (1987). “The effects of interference between
Mahir, N. and Altac, Z. (2008). “Numerical investigation of convective circular cylinders in cross flow.” Journal of Fluids and strutures,
heat transfer in unsteady flow past two cylinders in tandem Vol. 1, No. 2, pp. 239-261, DOI: 10.1016/S0889-9746(87)90355-0.
arrangements.” International Journal of Heat and Fluid flow, Vol. 29, Zhang, J. and Dalton, C. (1998). “A three-dimensional simulation of a
No. 5, pp. 1309-1318, DOI: 10.1016/j.ijheatfluidflow.2008. 05.001 steady approach flow past a circular cylinder at low Reynolds number.”
Meneghini, J. R. and Saltara, F. (2001). “Numerical simulation of flow International Journal of Numerical Methods in Fluids, Vol. 26, No.
interference between two circular cylinders in tandem and side-by- 9, pp. 1003-1022, DOI:10.1002/(SICI)1097-0363(19980515) 26:9
side arrangements.” Journal of Fluids and Structures, Vol. 15, No. 2, <1003::AID-FLD611>3.0.CO;2-W.
pp. 327-350, DOI: 10.1006/jfls.2000.0343. Zhou, Y. and Yiu, M. W. (2006). “Flow structure, momentum and heat
Mittal, S., Kumar V., and Raghuvanshi, A. (1997). “Unsteady incompressible transport in a two-tandem-cylinder wake.” Journal of Fluid Mechanics,
flows past two cylinders in tandem and staggered arrangement.” Vol. 548, pp. 17-48, DOI: 10.1017/S002211200500738X.
International Journal for Numerical Methods in Fluids, Vol. 25, No.

Vol. 00, No. 0 / 000 0000 − 11 −

You might also like