SCIENCE CHINA

Technological Sciences
May 2013 Vol.56 No.5: 1186–1197
doi: 10.1007/s11431-013-5208-3

Active control of circular cylinder flow by

affiliated rotating cylinders
WANG JianSheng, XU YuanXin* & TIAN YongSheng
Key Laboratory of Efficient Utilization of Low and Medium Grade Energy, MOE, School of Mechanical Engineering, Tianjin University,
Tianjin 300072, China

Received December 13, 2012; accepted March 20, 2013; published online April 7, 2013

This study puts forward an active control method for circular cylinder flow by placing two small affiliated rotating cylinders
adjacent to the main cylinder, and their effects on the drag and lift forces acting on the main cylinder as well as the heat trans-
fer effectiveness are numerically investigated. According to the diameter of the main cylinder the Reynolds number is chosen
as Re=200. The well-proven finite volume method is employed for the calculation. The code is validated by comparing the
present computed results of flow passing an isolated rotating cylinder with those available from the literature. To describe the
present control model, two parameters are defined: the rotation direction of the two small cylinders (including co-current rota-
tion and counter-current rotation) and the dimensionless rotation rate . In the simulation, the rotation rate  varies from 0 to
2.4. The results indicate that the optimum rotation direction of the subsidiary cylinders, which is beneficial to both drag reduc-
tion and heat transfer enhancement, is the co-current rotating (the upper affiliated cylinder spins clockwise and the lower affili-
ated cylinder spins counter-clockwise). We observe noticeable suppression of the vortex shedding and favorable reduction of
the fluid forces acting on the main cylinder as the rotation rate increases. Besides, the pressure and viscous components of the
drag force are analyzed. Energy balance between energy cost for activating the rotating cylinders and energy saving by the
momentum injection is considered. In addition, the influence of the affiliated rotating cylinders on heat transfer is also investi-
gated. The average Nusselt number is found to increase with the rotation rate.

active control, affiliated cylinders, vortex suppression, drag and lift forces, Nusselt number

Citation: Wang J S, Xu Y X, Tian Y S. Active control of circular cylinder flow by affiliated rotating cylinders. Sci China Tech Sci, 2013, 56: 11861197,
doi: 10.1007/s11431-013-5208-3

1 Introduction shedding which leads to considerable fluctuating fluid forc-

es. The wake and the separation region can be suppressed
by flow control which reduces unsteadiness and fluctuating
The unsteady viscous flow past a circular cylinder has been
forces on the bluff body. Therefore, understanding and con-
extensively studied by employing theoretical, numerical and
trolling wake is of great theoretical and practical importance.
experimental methods due to its simple geometry and rep-
An effective control method can save energy, increase the
resentative behavior of bluff body flow. Flow around chim-
propulsion efficiency, and reduce the induced vibrations of
neys, parallel suspension bridges, heat exchanger pipes and
the body. In general, the flow control techniques are classi-
high structures are examples of the bluff body flow in in-
fied into two categories. 1) Active flow control strategies in
dustrial application. A very important characteristic of this
which the flow is controlled by supplying additional power
flow is the shear layer separation and alternative vortex
and can be applied without any modifications to the geo-
metrical configuration. For the purpose of wake stabiliza-
*Corresponding author (email: [email protected]) tion, blowing and suction were used as a method of active

© Science China Press and Springer-Verlag Berlin Heidelberg 2013 tech.scichina.com www.springerlink.com
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1187

control on the cylinder surface in refs. [1–4]. In an up-to- test case, and we choose this Reynolds number to make
date study by Ma and Feng [5], flow control around a cir- validation of the numerical code by comparing with availa-
cular cylinder by synthetic jets was experimentally investi- ble results [25, 26] for flow past a rotating circular cylinder
gated. 2) In passive control strategies, flow is controlled by (see section 3.1). The aim of this work is to investigate the
modifying the shape of the bluff body or by appending ad- effectiveness of the active control method in drag reduction
ditive devices such as splitter plate, axial slat, control rod as well as heat transfer enhancement in a theoretical point of
and roughness elements on the body surface [6–10]. view. The physical mechanisms behind the active control
In recent years, flow around rotationally oscillating cyl- method are: 1) The co-current rotation of the attached cyl-
inder has been studied extensively by numerical methods inders retards the growth of the boundary layer by mini-
[11–15]. These studies cover a wide range of flow condi- mizing relative motion between the body surface and the
tions such as forcing frequency, oscillation amplitude, free stream; 2) the rotating cylinders bring extra momentum
Reynolds number, etc. The numerical results predict the to the boundary thus delay separation. Even though the nu-
qualitative nature of the flow around rotationally oscillating merical results in our work are all in laminar regime, the
cylinder, including the phenomenon of synchronization at drag reduction mechanism of this active control method is
low forcing frequency and drag reduction at high forcing supposed to be applicable to turbulent flows [16]. Gim et al.
frequency [4]. A large number of systematic experiments [27] presented an experimental study of flow past a cylinder
have been carried out by Modi and Mokhtarian [16, 17], with control rods at the rear stagnation point in a water
who have employed rotating cylinders as an integral part of channel by using PIV (particle image velocimetry) tech-
bluff bodies such as buildings and moving vehicles. This nique. They chose a Reynolds number range of 5000 to
control concept has been shown to be remarkably effective 20000. It was observed that a control rod with d/D=0.3 has
and successful in delaying separation. Since momentum is outstanding effect on reducing the turbulent intensity with
injected into the shear layers by the moving surface, in the vortex shedding strongly suppressed in the near wake. Kar-
literature it is also known as moving surface boundary layer abelas [28] computed a flow past a rotating cylinder at
control (MSBC), which functions in the following ways Re=140000 by using the Large Eddy Simulation with a ro-
[16]. 1) It retards the growth of the boundary layer by min- tation rate varying from 0 to 2. He found that as the spin
imizing relative motion between the body surface and the ratio increases, the mean drag decreases and the flow tends
free stream. 2) It injects momentum into the boundary layer. to be stable.
3) It engenders a region of high suction and thereby accel-
erates flow in its neighborhood and outside of the boundary
layer. 4) It postpones separation and interferes with the 2 Numerical model and solution methods
evolution of the wake.
In the past decades, many attempts of control algorithms The two-dimensional unsteady governing equations for an
have emerged in the non-linear dynamics and control com- incompressible flow can be written as follows:
munity such as chaos control, synchronization based control, The continuity equation:
pinning control, sporadic feedback control, etc. [18–21]. It u v
is of great importance to study the physical mechanism of   0. (1)
x  y
these control methods from a theoretical standpoint so that
they can further be applied to practical problems of engi- The momentum equation in the x and y directions:
neering interest. In the case of flow past a circular cylinder,
vortex shedding is the first global instability mode, which u u u 1 p  2 u 2 u 
u v    2  2  , (2)
needs to be predicted and controlled before attempting other  x y  x  x y 
modes that involve turbulence [22]. In addition, Strykowski
et al. [23] have found that the vortex shedding behind cir- v v v 1 p  2 v 2 v 
cular cylinder could be altered and suppressed altogether u v    2  2  . (3)
 x y  y  x y 
over a limited range of Reynolds numbers by a proper
placement of a second, smaller cylinder in the near wake of The energy equation:
the main cylinder. Therefore, in the present study, an active
flow control strategy is proposed to control vortex shedding T T T   2 T 2 T 
u v  a  2  2 , (4)
behind a circular cylinder by means of two small rotary at-  x y  x y 
tached cylinders. The Reynolds number based on the diam-
eter of the main cylinder is chosen as Re=200. This Reyn- where x and y are the transverse coordinate and longitudinal
olds number is chosen because it is in a laminar flow regime, coordinate in the Cartesian coordinates, respectively; u and
ensuring the boundary layer separation at the cylinder sur- v are the velocity components in the x direction and y direc-
face is laminar [24]. The laminar flow over a circular cylin- tion, respectively; p is the pressure,  is the fluid density, υ
der with Re=200 has been considered by many authors as a is the kinematic viscosity, T is the temperature, and a=
1188 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5

/(cp) is the thermal diffusion coefficient. The Reynolds grid independent tests, a meshing method with total ele-
number is calculated as Re=U∞D/υ with U∞ being the ments of 205488 is chosen. A well-proven stabilized finite
free-stream velocity and D the diameter of the main circular volume method based on SIMPLE scheme [30] is adopted
cylinder. A Reynolds number of 200 is used for all the sim- to solve the incompressible Navier-Stokes equations. The
ulations. second order semi-implicit scheme is used for the time inte-
It is essential to make the inlet, far-field and outlet gration. The spatial discretization is performed on a stand-
boundaries far away enough from the cylindrical zone under ard collocated grid using the finite volume method.
consideration so that the boundary conditions will not lead
to undesirable effects. The computational model is shown in
Figure 1. The diameters of both the two attached cylinders 3 Results and discussions
are d=0.1D. The gap between the main cylinder and the
attached cylinders is g=0.01D. The location of attached 3.1 Validation of the code
cylinders is suitable, as it is an approximate place to enable
tangential suction from the favorable pressure gradient re- The drag and lift coefficients are calculated as
gion of the main cylinder and injection into the adverse FD FL
pressure gradient zone. Price et al. [29] investigated the CD  , CL  . (5)
1 1
effect of changing the gap (h) between wall and the main U  D
2
U  D
2

2 2
cylinder, ranging from h/D=0 to h/D=2, where D is the di-
ameter of the main cylinder. It was found that the flow The pressure and viscous components of the fluid forces
around the cylinder with control has little difference with are calculated as
the flow around an isolated circular cylinder for h/D>1.0. In
2( p  pref ) 2 w
the present simulation, the free stream velocity boundary Cp  , Cf  . (6)
condition is applied to the inlet: u=U∞ and v=0. On the sur- U 2
U 2 
face of the main and attached cylinders, no-slip boundary The rotation rate  is defined as
condition is imposed. The boundary conditions, u x  0
D
and v x  0 are applied to the exit boundary. Symmetry  ,
2U 
boundary conditions, u y  0 and v=0, are applied to
the far field boundaries. The two attached cylinders are where FD and FL are the total drag and lift forces, respec-
placed at angles of ±120° respectively to the upstream vec- tively; w is the shear stress,  is the angular velocity of the
tor direction, right behind the main cylinder (shown in Fig- attached cylinder. In addition, symbol “+” represents
ure 1). co-current rotation, and symbol “–” represents coun-
To simplify description, the rotating direction of the two ter-current rotation.
small subsidiary cylinders is defined as follows. The finite volume method, which has been applied to
The case that the upper attached cylinder rotates coun- various flow problems, is utilized to solve the incompressi-
ter-clockwise and the lower attached cylinder rotates ble Navier-Stokes equations in the primitive variables for-
clockwise is defined as counter-current rotation. The case mulation. The numerical validations are mainly focused on
that the upper attached cylinder rotates clockwise and the the fluid forces acting on an isolated rotating cylinder with
lower attached cylinder rotates counter-clockwise is defined different rotation rates of . Figure 2 displays the compari-
as co-current rotation (see Figure 1). son of the mean drag coefficient obtained by the present
To facilitate meshing and speed up the simulation, the numerical calculation with available results in the literature.
whole computational domain is divided into two parts. Finer The computed results are in good agreement with the results
mesh is employed in the cylindrical zone where the gradi- obtained by Kang et al. [25] and Stojkovic et al. [26]. The
ents of variables are likely to be very high. After a series of maximum deviation is kept within 5%, which ensures the
reliability and accuracy of the present computational model
and methods.

3.2 Effects of rotation direction

3.2.1 Analysis of the vortex shedding in the wake region

For a flow past an isolated circular cylinder, a pair of steady
and symmetric vortices can be found at a very low Reynolds
number. As the Reynolds number increases, the upstream-
downstream symmetric vortices disappear and two-attached
Figure 1 Computational model. eddies appear in the wake behind the cylinder. When the
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1189

Figure 2 Comparison of mean drag coefficient with existing results of

flow past a rotating cylinder for different rotation rates.

Reynolds number exceeds a critical number, alternative

vortex shedding can be found to form the famous Karman
vortex street. Generally speaking, three characteristic pa-
rameters are used to describe the property of the vortex:
main direction (the link direction of the two biggest distance
summits on the vorticity isoline), center of vorticity (the
location of maximum vorticity value), and vorticity isoline
shape rate (each vorticity isoline can be transformed into a
standard curve).
Figure 3 displays the comparison of vortex shedding
structures for counter-current rotation and co-current rota-
tion. It could be observed that in the case of counter-current
rotation, the vorticity strengthens gradually as the rotation
rate increases, while in the case of co-current rotation, the
vorticity weakens with the increasing of the rotation rate. In
the former case, as the rotation rate  increases, the vortici- Figure 3 Instantaneous plots of vorticity contours for flow at different
ty center value of the first shedding vortex behind the main control cases. (a) =0; (b) =+0.5, (c) =+1.0; (d ) =0.5, (e) =1.0.
cylinder amplifies and the generated vortices behind the
main cylinder become larger and stronger. Moreover, the tion trends could also be found in the plots of lift coefficient;
width of the wake is magnified. Whereas for co-current ro- see Figure 4(b). The reduction of the lift force fluctuation
tation, the formation of the near wall vortex is delayed and agrees well with the results of vortex shedding. The fluctu-
the value of the first shedding vortex weakens. This is due ating lift force is due mainly to the fluctuating pressure act-
to the fact that additional momentum is injected into the ing on the surface of the cylinder caused by the alternative
flow separation region by the two co-current rotating cylin- vortex shedding. Table 1 displays the comparison of the
ders, which postpones the shear layer separation. On the mean drag coefficient CDM and amplitude of lift coefficient
other hand, the attached cylinders hinder the mainstream
fluid (flow far from the main cylinder) in the case of coun- CLA for different rotation directions.
ter-current rotation. Compared with =0, the mean drag coefficient increases
by 22.73% for counter-current rotation, yet decreases by
3.2.2 Analysis of drag and lift forces 21.8% for co-current rotation at =0.5 and the amplitude of
The time variations of the drag and lift forces acting on the lift fluctuation increases by 39.15%, and decreases by
main cylinder show significant distinction for different rota- 42.5%, accordingly. For =1.0, compared to =0, the
tion directions. Comparing to the case of a main cylinder mean drag value and amplitude of lift coefficient increases
with stationary subsidiary cylinders (=0), the mean drag by 42.6% and 71.83% for counter-current rotation, yet de-
coefficient becomes larger in the case of counter-current creases by 45.2% and 85.14% for co-current rotation, re-
rotation, while for co-current rotation, significant reduction spectively.
is observed and the amplitude of the drag force fluctuation The influence of rotation direction on wake vortex is in
is greatly reduced, as shown in Figure 4(a). Similar varia- accordance with the impact on fluid forces. The co-current
1190 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5

3.3 Investigating the co-current rotation of attached

cylinders

3.3.1 Analysis of flow field

The computed results show that the two subsidiary rotating
cylinders have altered the cylinder wake significantly. The
vorticity isoline of wake region, contoured velocity and
streamline patterns are shown for different co-current rota-
tion rates of  in Figures 5 and 6. In Figure 5(a)–(f), the
simulated results reveal that the center value of vorticity
decreases gradually as  increases, the intensity of eddy
weakens and the vortex shedding is completely suppressed
at =2.4. In Figure 5(g)–(j), when  is relatively low, ed-
dies still exist in the cylinder wake, yet the width of wake
stream is reduced. With the increase of , the periodical
vortex shedding is suppressed due to the weakening of the
instability in the near wake zone, making it hard to form
circulation into the shear layer. However, the convective
instability still persists in the far wake region. The stream-
line patterns show that the sine-shaped fluctuation of wake
streamlines tends to become mild as  increases, as shown
in Figure 6. The streamline becomes almost straight at =
2.4, revealing the wake is stable.

3.3.2 Analysis of drag and lift coefficients

It is important to note that the total fluid forces are made up
of two components, namely, the viscous (or skin friction)
component and pressure component. The total drag coeffi-
Figure 4 Comparisons of drag (a) and lift (b) coefficient histories in the cient CD is largely dependent on the pressure component
cases of =0 and =±1.0. due to flow separation and this separation could be delayed
by a moving (rotating) wall. As proposed by Homescul et al.
[31], separation is completely eliminated on one side of the
Table 1 Comparison of the mean drags and amplitudes of lift coefficients
for different rotation directions
cylinder where the moving wall and the free stream move in
the same direction and separation only develops incom-
Co-current Counter-current pletely on the other side of the cylinder. In the present study,
=0
=0.5 =1.0 =0.5 =1.0 for co-current rotation of the attached cylinders, the flow
C M
D 1.500 1.173 0.823 1.841 2.139 separation can be postponed by reducing the velocity dif-
C A
0.774 0.445 0.115 1.077 1.330
ference between the cylindrical zone and the free stream.
L
Consequently, the drag force acting on the main cylinder
decreases greatly. Figure 7 shows the variation of the drag
and lift coefficients with the rotation rate . It can be found
rotation of the attached cylinders can achieve the purpose of in Figure 7 that the amplitudes of lift and drag coefficients
suppressing vortex shedding and reducing the drag and lift decline gradually with the increasing  and the fluctuation
forces. The reason is that additional momentum is injected of lift and drag coefficients tends to disappear due to the
into the shear layers by the placement of the two small ro- increase of stability in the near wake region. Figure 8 pre-
tating cylinders, which delays the formation of the vortex. sents the variations of the mean drag coefficient and ampli-
On the contrary, the counter-current rotation of the attached tude of the lift coefficient fluctuations at different rotation
cylinders fosters the development of vortex and increases rates. Compared with =0, the mean drag coefficient for
the drag and lift forces. It is because the direction of the =2.4 is decreased by 78.67%. The declining speed of the
rotating surfaces is opposite to the mainstream fluid, which lift coefficient is faster than that of the drag coefficient
results in the increase of the drag force. However, for the when the rotation rate is relatively low (<1.0), while in the
co-current rotation, it retards the growth of the boundary case of a higher , the lift fluctuation tends to disappear.
layer by minimizing relative motion between the cylinder This tendency corresponds well to the flow field of vortex
surface and the mainstream fluid. shedding, in other words, the lift forces disappear in the
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1191

Figure 5 Vorticity isolines for different rotation rates: (a) =0, (b) =+0.2, (c) =+1.0, (d) =+1.5, (e) =+2.0, (f) =+2.4. Contoured velocities for
different rotation rates: (g) =+1.0, (h) =+1.5, (i) =+2.0, (j) =+2.4.

Figure 6 Streamline patterns for different rotation rates: (a) =+0.2, (b) =+1.5, (c) +=2.0, (d) =+2.4.

cases where the vortex is suppressed. On the other hand, the which induces the fluctuating forces on the cylinder surface
mean drag force declines all along with the increasing . is gradually suppressed, resulting in the remarkable reduc-
With the increase of the rotation rate, vortex shedding tion of the drag and lift forces. When the flow wake
1192 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5

Figure 7 Comparison of drag (a) and lift (b) coefficient histories for different rotation rates.

Figure 8 Variations of mean drag coefficient and amplitude of lift coef-

ficients for different rotation rates.

Figure 9 Percentages of pressure and viscous components for different

rotation rates and directions.
becomes stable at a higher , the lift coefficient is almost
negligible.
rotation, the percentage of pressure drag reduces gradually
3.4 Discussion of the pressure and friction components as  augments, while the percentage of friction drag rises.
of drag force For counter-current rotation, comparing to co-current rota-
tion, the variation trends of friction and pressure drag per-
As mentioned in the above analysis, the total drag force is centage are opposite. Because of the dominating role of the
made up of skin friction component Cf and pressure com- pressure component, it determines the variation trend of the
ponent Cp, which is often expressed as total drag acting on the main cylinder. Figure 10 demon-
CD  C p  C f . strates the drag force comparison between a cylinder with-
out control (single cylinder) and the controlled cases at dif-
The pressure component of the fluid force is obtained by ferent rotation rates, by displaying the pressure and friction
integrating the pressure acting on the body faces in the x coefficients periodically.
and y directions, and depends on the shape of the immersed As is shown in Figure 10, the pressure and friction drag
body. Similarly, the viscous component is calculated by are greater than those of the single cylinder for relative low
integrating the shearing forces acting on the body faces in rotation rates. However, both the pressure and friction drag
the x and y directions [10]. The pressure component which experience favorable reduction when the rotation rate is
is also described as the pressure difference between the high high enough. In addition, the periods of pressure change are
pressure in the front stagnation region and the low pressure different for the single cylinder and cylinder with control,
in the rear separated zone contributes largely to the total which should not be neglected. Figure 11 displays the com-
drag force. In Figure 9, the histograms about the percent- parison of pressure coefficient and viscous coefficient vari-
ages of friction and pressure drag show the leading role of ations in one period for different rotation rates. It can be
the pressure component in the drag force. For co-current seen that the frequency of vortex shedding amplifies at the
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1193

Figure 10 (a) Comparison of pressure coefficient (Cp) variations in a period for different rotation rates; (b) comparison of viscous coefficient (Cf) varia-
tions in two periods for different rotation rates.

especially for =1.0.

Figure 13 shows the distribution of local pressure coeffi-
cient on the main cylinder. For =0, the pressure at the rear
of the main cylinder is lower than that of the single cylinder,
so the pressure difference between front stagnation region
and the rear zone amplifies. For co-current rotation of the
attached cylinders, the rear pressure of the main cylinder
increases remarkably as  increases. As a result, the pres-
sure difference between the front and rear zones decreases
significantly. The pressure difference also decreases be-
tween the upper and lower faces of the main cylinder be-
cause of the weakening of the wake oscillation. Whereas in
cases of counter-current rotation, the rear pressure of the
main cylinder drops as  increases, consequently the am-
Figure 11 Comparison of pressure coefficient (Cp) and viscous coeffi- plification of the pressure difference between the front and
cient (Cf) variations in one period for different rotation rates.
rear walls (see Figure 13(b)).

relative low rotation rate (+). As for the case of counter- 3.5 Cost function analysis
stream rotation (), the pressure and friction coefficient
enlarge increasingly along with the increasing rotation rate. According to the definition, active control generally re-
Moreover, phase position is inconsistent with the periodical quires external energy to activate the control devices. It is
sine fluctuation between the pressure coefficient and the then essential to ensure the energy balance (difference be-
friction coefficient. The changing speed of friction drag is tween gain and loss of energy owing to the control) to be
slightly slower than that of the pressure. positive. Based on the experimental studies of Beaudoin et
Figure 12 displays the velocity distribution along the al. [32, 33], the electric power cost in the measurements is a
positive x-axis behind the main cylinder in different rotation quadratic function of the rotation speed (). And they ob-
directions and at different rates. The velocity of rear region served that the electric power does not depend on the flow
close to the main cylinder reduces gradually along the axial velocity, given by the following equation:
direction with the increasing co-current rotation rate (+),
Pe =22, (7)
and notable reduction is found at =1.0 (Figure 12(a)).
However, the velocity increases at the location far away where the constant 2 represents the presence of two small
from the cylinder zone. For counter-stream rotation, the control cylinders placed in the simulated model. The pa-
velocity distribution along x-axis in the near wake is com- rameter  is obtained by fitting and regression of the ex-
pletely different from in the co-current rotation (Figure perimental data. According to the experimental studies of
12(b)). The velocity in the wake region close to main cylin- Beaudoin et al. [32], we can obtain a value of =0.0008 W/
der amplifies as  becomes larger. Whereas the velocity (rad s1)2 by adopting a similar procedure (the value chosen
decreases in the region far away from the main cylinder, through a similar experiment).
1194 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5

Figure 12 Velocity distributions along x-axis behind the main cylinder. (a) Co-current rotation; (b) counter-current rotation.

Figure 13 Distributions of pressure coefficient on the main cylinder surface. (a) Co-current rotation; (b) counter-current rotation.

Figure 14 illustrates the variations of mean drag coeffi-

cient ( CDM ) and net drag savings ( CDM  CDM,Single  CDM ) at
different rotation rates of . Figure 15 shows the electric
power consumption and net drag savings as  increases. It
is displayed in Figure 16 that the proportion of the net drag

Figure 15 Comparison of electric power consumption (Pe) and net drag

savings ( CDM ).

saving CDM to the electric power expenditure Pe and the

efficiency of the input power in drag reduction could be
observed. Therefore, considering the energy cost, an opti-
mum point of =0.5 could be found where we obtain the
Figure 14 Variations of mean drag coefficient and net drag savings. maximum drag reduction for per Watt of Pe. Moreover, it is
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1195

cylinders, the average Nusselt number increases remarkably.

In addition, it is worth noting that the local Nusselt number
has two sharp peaks in the circumferential distribution in
comparison with the single cylinder case (see Figure 17(b)).
The two peaks of Nusselt number locate in the adjacent
zone of the two subsidiary control cylinders (=±120°).
The main reason is that the existence of the two small rotat-
ing cylinders enhances largely the flow instability around
them, thus facilitates heat transfer. The two peaks make a
paramount contribution to the augment of the average
Nusselt number. It is observed in the local Nusselt number
distribution that the heat transfer is slightly weakened in
Figure 16 Proportion of CDM to Pe at different rotating rates. comparison with the single cylinder case in the front and
rear zones of the cylinder. However, the local heat transfer
enhancement near the two rotating cylinders far outweighs
necessary to consider the whole efficiency of the control the little deterioration and can compensate for it. Therefore,
method by heat transfer analysis in the following section. the active control facilitates the total heat transfer even
though the local heat exchanging process is slightly weak-
3.6 Heat transfer results ened. In general, for the purpose of both drag reduction and
heat transfer enhancement, a co-current rotation of the sub-
It is presented in this section the heat transfer results of the sidiary cylinders with a higher rotating speed is favorable.
main cylinder with control for different rotation rates of  In the simulation, for >0.5, the increment of the total heat
at Re=200 and Pr=7.02. transfer (average Nusselt number) for =1.0 and 1.5 are
Firstly, the accuracies of the present heat transfer model 5.26% and 15.7%, respectively if they are compared with
and methods are validated. A number of observers have that single cylinder case.
measured the mean heat transfer coefficient for flow over a Figure 18 shows the instantaneous isotherm patterns
single cylinder. A correlation for a cylinder at uniform around the cylinder with and without control for different
temperature (Ts) in cross-flow applied to liquids and gases co-current rotation rates. As observed from these profiles,
has been proposed by Zukauskas [34] as follows: the isotherm patterns in the flow wake are greatly influ-
m 0.25 enced by the rotation speed. As previously observed, the
havg D U D   Pr 
Nuavg   C    Pr n   . (8) instability in wake region is suppressed with the increasing
     Prs  rotation rate.
In the present study, C=0.51, m=0.5, and n=0.37.
In addition, the local Nusselt number and average 4 Conclusions
Nusselt number are defined as follows:
hx D In the present study, the unsteady laminar flow past a circu-
N uloc  , Nuavg   Nu x dx. (9) lar cylinder with attached control cylinders is numerical

studied. Simulations are carried out at a Reynolds number
The predicted average Nusselt number of the present of 200. The influences of different rotation directions and
computational model is compared with that gained by eq. rates of the two rotating cylinders on fluid forces, flow field
(8). The deviation of Nuavg is 4.88%, indicating that the structures as well as heat transfer characteristics are com-
model and numerical methods applied are reliable and ca- prehensively presented. On the basis of the numerical re-
pable of modeling the heat transfer procedure. sults, the following conclusions can be made.
Figure 17 presents the comparison of the average Nusselt 1) The rotation direction of attached cylinders has great
number and local Nusselt number on the main cylinder for influence on the flow structure as well as the fluid forces of
different rotation rates. For counter-current rotation, the cylinder flow. The co-current rotation can help suppress
average Nusselt number increases (see Figure 17(a)). How- vortex shedding and reduce the drag and lift forces. The
ever, according to the above-mentioned results, the fluid vortex formation is delayed and weakened because of the
forces on the main cylinder also increase obviously for momentum injection generated by the two rotating cylinders.
counter-current rotation. As for the case of co-current rota- For =2.4, the vortex shedding is completely suppressed
tion, the average Nusselt number of the main cylinder is and the drag acting on the main cylinder is reduced by
slightly lower than that of the isolated cylinder when the 77.63%, as compared with single cylinder, and the corre-
rotation rate is relatively small, i.e. =0.5 (see Figure 17(a)). sponding lift force is close to 0. Whereas for counter-stream
When higher rotation speeds are imposed on the control rotation, both the lift and drag forces increase.
1196 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5

Figure 17 Comparison of Nusselt numbers on the main cylinder for

different rotation rates. (a) Time histories of average Nusselt number on the
main cylinder; (b) circumferential distributions of local Nusselt number.

2) The percentage of pressure component of drag force

decreases gradually with the increasing co-current rotation
rate (+), while the percentage of friction drag increases. Figure 18 Isotherm profiles of cylinder with and without control for
For counter-stream rotation (), the changing tendency of different co-current rotation rates. (a) Single cylinder; (b)  = 0.5; (c)  =
pressure and friction components is quite the contrary. The 1.0; (d)  = 1.5; (e)  = 2.0.
pressure differences on the main cylinder surface reduce
obviously between the front and rear, and upper and lower cylinder. The main reason is that the existence of the two
cylinder walls, especially for a larger co-current rotation small rotating cylinders enhances largely the flow instability
rate. around them and thus facilitates heat transfer.
3) It is found that the total cost function has an optimum
which depends on the balance between energy saved by
Nomenclature
momentum injection and energy expenditure in activating
the control cylinders. a thermal diffusion coefficient (=/(cp)), m2/s;
4) The average Nusselt number on the main cylinder with CD drag coefficient, dimensionless;
control increases with the rotation rate and better heat CDM mean value of drag coefficient, dimensionless;
transfer performance is observed for a relatively higher ro-
Cf friction coefficient, dimensionless;
tation rate (>0.5), as compared with an isolated cylinder.
CL lift coefficient, dimensionless;
In addition, two peaks which locate in the adjacent zone of
the two subsidiary control cylinders (= ±120°) are found in CLA amplitude of lift coefficient, dimensionless;
the circumferential distribution of local Nusselt number, Cp pressure coefficient, dimensionless;
which contribute most to the increase of the average Nusselt cp specific heat capacity, J/(kg K);
number and compensate for the little deterioration of local D diameter of the main circular cylinder, m;
Nusselt number in the front and rear zones of the main d diameter of the two small cylinders, m;
 Wang J S, et al. Sci China Tech Sci May (2013) Vol.56 No.5 1197

FD drag force, N; stagnation points. Sci China Tech Sci, 2013, 56: 158–170
6 Kwon K, Choi H. Control of laminar vortex shedding behind a circu-
FL lift force, N;
lar cylinder using splitter plates. Phys Fluids, 1996, 8: 479–486
g gap between main cylinder and attached cylinders, m; 7 Darekar R M, Sherwin S J. Flow past a bluff body with a wavy stag-
h distance between up (down) boundary and the cyl- nation face. J Fluid Struct, 2001, 15: 587–596
inder center line, m; 8 Lee S J, Lee S I, Park C W. Reducing the drag on a circular cylinder
by upstream installation of a small control rod. Fluid Dyn Res, 2004,
L1 distance between main cylinder center and inlet 34(4): 233–250
boundary, m; 9 Hwang J, Yang K. Drag reduction on a circular cylinder using dual
L2 distance between main cylinder center to outlet detached splitter plates. J Wind Eng Ind Aerody, 2007, 95: 551–564
boundary, m; 10 Malekzadeh S, Sohankar A. Reduction of fluid forces and heat trans-
fer on a square cylinder in a laminar flow regime using a control plate.
L distance from the cylinder center, m; Int J Heat Fluid Fl, 2012, 34: 15–27
Nu Nusselt number, dimensionless; 11 Williamson C H K, Roshko A. Vortex formation in the wake of an
Nuavg average Nusselt number, dimensionless; oscillating cylinder. J Fluids Struct, 1998, 2: 355–381
12 Lu X Y, Sato J. A numerical study of flow past a rotationally oscil-
Nuloc local Nusselt number, dimensionless;
lating circular cylinder. J Fluid Struct, 1996, 10: 829–849
Pr Prandtl number, dimensionless; 13 Baek S J, Sung H J. Numerical simulation of the flow behind a rotary
p pressure, Pa; oscillating circular cylinder. Phys Fluids, 1998, 10: 869–876
pref reference pressure, Pa; 14 Cheng M, Chew Y T, Luo S C. Numerical investigation of a rota-
tionally oscillating cylinder in mean flow. J Fluid Struct, 2001, 15:
Pe electric power consumption, W; 981–1007
Re Reynolds number (= U∞D/υ), dimensionless; 15 Srinivas K, Fujisawa N. Effect of rotational oscillation upon fluid
T temperature, K; forces about a circular cylinder. J Wind Eng Ind Aerody, 2003, 91:
t time, s; 637–652
16 Modi V J. Moving surface boundary-layer control: A review. J Fluid
U∞ free-stream velocity, m/s; Struct, 1997, 11: 627–663
u velocity component in x direction, m/s; 17 Mokhtarian F, Modi V J. Fluid dynamics of airfoils with moving
v velocity component in y direction, m/s; surface boundary-layer control. J Aircr, 1988, 25: 163–169
18 Ott E, Grebogi C, Yorke J A. Controlling chaos. Phys Rev Lett, 1990,
x transverse coordinate; 64(11): 1196–1199
y longitudinal coordinate. 19 Wei G W. Synchronization of single-side locally averaged adaptive
coupling and its application to shock capturing. Phys Rev Lett, 2001,
86(16): 3542–3545
Greek symbols 20 Tang G, Guan S, Hu G. Controlling flow turbulence with moving
controllers. Eur Phys J B, 2005, 48: 259–264
 rotation rate (   D / 2U  ), dimensionless; 21 Tang G, Hu G. Controlling flow turbulence using local pinning feed-
back. Chin Phys Lett, 2006, 23(6): 1523–1526
 proportional parameter, W/ (rad s ) ; 1 2
22 Muddada S, Patnaik B S V. An active flow control strategy for the
 angle to the upstream vector direction; suppression of vortex structures behind a circular cylinder. Eur J
 increment; Mech B-Fluid, 2010, 29: 93–104
23 Strykowski P J, Sreenivasan K R. On the formation and suppression
 density, kg/m3; of vortex shedding at low Reynolds numbers. J Fluid Mech, 1990,
w shear stress, N/m; 218: 71–107
 heat conductivity, W/(m K); 24 Lienhard J. Synopsis of lift, drag and vortex frequency data for rigid
circular cylinder. Washington State University, College of Engineer-
 kinematic viscosity, m2/s; ing, Research Division Bulletin 300, 1966
 angular velocity, rad/s. 25 Kang S, Choi H, Lee S. Laminar flow past a rotating circular cylinder.
Phys Fluids, 1999, 11: 3312–3321
26 Stojkovic D, Breuer M, Durst F. Effect of high rotation rates on the
Self-defined symbols laminar flow around a circular cylinder. Phys Fluids, 2002, 14(9):
3160–3178
+ co-current rotation; 27 Gim O S, Kim S H, Lee G W. Flow control behind a circular cylinder
– counter-current rotation. by control rods in uniform stream. Oc Engrg, 2011, 38(17-18):
2171–2184
28 Karabelas S J. Large eddy simulation of high-Reynolds past a rotat-
ing cylinder. Int J Heat Fluid Fl, 2010, 31: 518–527
1 Park D S, Ladd D M, Hendricks E W. Feedback control of von Kar- 29 Price S J, Sumner D, Smith J G, et al. Flow visualization around a
man vortex shedding behind a circular cylinder at low Reynolds circular cylinder near to a plane wall. J Fluid Struct, 2002, 16(2):
numbers. Phys Fluids, 1994, 6: 2390–2405 175–191
2 Gunzburger M D, Lee H C. Feedack control of vortex shedding. 30 Fluent 6.3 User’s Guide. Fluent Inc., 2006
Trans ASME J Appl Mech, 1996, 63: 828–835 31 Homescu C, Navon I M, Li Z. Suppression of vortex shedding for
3 Fransson J H M, Konieczny P, Alfredsson P H. Flow around a porous flow around a circular cylinder using optimal control. Int J Numer
cylinder subject to continuous suction or blowing. J Fluid Struct, Meth Fluid, 2002, 38: 43–69
2004, 19: 1031–1048 32 Beaudoin J F, Cadot O, Aider J L, et al. Bluff body drag reduction by
4 Fujisawa N, Asano Y, Arakawa C, et al. Computational and experi- extremum-seeking control. J Fluids Struct, 2006, 22: 973–978
mental study on flow around a rotationally oscillating circular cylin- 33 Beaudoin J F, Cadot O, Aider J L, et al. Drag reduction of a bluff
der in a uniform flow. J Wind Eng Ind Aerody, 2005, 93: 137–153 body using adaptive control methods. Phys Fluids, 2006, 18 (8): 085107
5 Ma L Q, Feng L H. Experimental investigation on control of vortex 34 Zukauskas A A. Heat transfer from tubes in cross flow. Adv Heat
shedding mode of a circular cylinder using synthetic jets placed at Trans, 1972, 8: 93–106