Sie sind auf Seite 1von 11

403

2008,20(4):403-413

LARGE-EDDY SIMULATION OF FLOW AROUND CYLINDER ARRAYS


AT A SUBCRITICAL REYNOLDS NUMBER*

ZOU Lin
School of Mechanical and Electronic Engineering, Wuhan University of Technology, Wuhan 430070, China,
E-mail: l.zou@163.com
LIN Yu-feng, LAM Kit
Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hong Kong, China

(Received November 12, 2007, Revised January 31, 2008)

Abstract: The complex three-dimensional turbulent flows around a cylinder array with four cylinders in an in-line square
configuration at a subcritical Reynolds number of 1.5 ×104 with the spacing ratio at L / D = 1.5 and 3.5 were investigated using the
Large Eddy Simulation (LES). The full field vorticity and velocity distributions as well as turbulent quantities were calculated in
detail and the near wake structures were presented. The results show that the bi-stable flow nature was observed at L / D = 1.5 and
distinct vortex shedding of the upstream cylinders occurred at L / D = 3.5 at Re = 1.5 u 104 . The techniques of Laser Doppler
Anemometry (LDA) and Digital Particle Image Velocimetry (DPIV) are also employed to validate the present LES method. The
results show that the numerical predictions are in excellent agreement with the experimental measurements. Therefore, the full field
instantaneous and mean quantities of the flow field, velocity field and vorticity field can be extracted from the LES results for further
study of the complex flow characteristics.

Key words: four-cylinders array, turbulent flow, Large Eddy Simulation (LES), Laser Doppler Anemometry (LDA), Digital Particle
Image Velocimetry (DPIV)

1. Introduction  found. For example, Sayers[3,4] conducted experiments


The wake interference around the cylinder arrays on four cylinders in a square configuration with two
haV been widely investigated in the past because of its end plates in the spacing ratio range of 1.5 to 5 and the
inherent importance and practical significance in aspect ratio 11.5 at the Reynolds number of 3×104.
engineering applications. Norberg[1] and Sumner et Lam et al. [5-8] carried out extensively experimental
al.[2] summarized recent research work on different investigations using Laser Induced Fluorescent (LIF)
kinds of flow patterns for two cylinders in visualization and Particle Image Velocimetry (PIV)
side-by-side arrangement and in tandem arrangement. for the cross-flow around four cylinders in different
The more complex wake flow around the spacing ratios and orientations at different Reynolds
four-cylinder arrays configuration has not been numbers. However, these studies were mostly focus
studied as extensively as the single and two cylinders on the force and pressure coefficient distributions as
configurations. However, a number of studies of the well as the transformation of the flow patterns around
fluid dynamics for the four-cylinder arrays can still be the four cylinders. The quantitative data of the mean
velocity distributions, the instantaneous velocity and
the vorticity fields for the turbulent flow around four

* Project supported by the Council of the Hong Kong
cylinders in an in-line square configuration at
Special Administrative Region, China (Grant No. PolyU subcritical Reynolds numbers are not available yet.
5299/03E). Due to the difficulties in data acquisition in
Biography: ZOU Lin (1970-), Female, Ph. D. certain positions and configurations in the
experimental study, numerical simulation results for
404

such complex turbulent flows have been sought. wui


Numerical simulations of the flow around one or two =0 (1)
cylinders have been reported in the past for turbulent
wxi
flows. However, there are still relatively little
numerical investigations on the four-cylinder arrays. wui wui u j 1 wp w 2ui wW
Farrant et al. [9] captured numerically two-dimensional + = +Q  ij ,
wt wx j U wxi wx j wx j wx j
laminar flow characteristics and interactive forces
associated with flows around four equi-spaced
cylinders at Re = 200 using a cell boundary element (i = 1, 2,3) (2)
method. Lam et al. [10] attempted to carry out a 3-D
numerical simulation for laminar cross-flow around where ui are the filtered velocity components along
the four in-line cylinders using the finite volume the Cartesian coordinates xi, p is the pressure, U
method at Re=200 and predicted accurately some of
wake behaviour around the four cylinders. Since the is the fluid density and Q is the kinematic viscosity
simulations for the flows around the four cylinders are of the fluid. The influence of the small scales on the
for laminar flows only, and more engineering large (resolved) scales takes place through the SGS
applications are at high Reynolds numbers. In order to stress defined by
fill the aforementioned gap, capture the detailed 3-D
vortex structures and obtain the full field W ij = ui u j  ui u j (3)
instantaneous and mean physical information such as
the mean and fluctuating pressure, forces, the vortex resulting from the filtering operation, which are
shedding characteristics, etc., the 3-D numerical unknown and must be modeled with a SGS model.
simulations of the turbulent flows around the The majority of SGS models are based on the eddy
four-cylinder array in an in-line square configuration viscosity models of the following form:
with the spacing ratio at 1.5 and 3.5 are carried out
using the Large Eddy Simulation (LES) at
1
Re = 1.5 u 104 (refer to the similar LES methods W ij  W kk G ij = 2Q t Sij (4)
adopted by the previously investigators[11,12]). At the 3
same time, advanced experimental techniques such as
the Laser Doppler Anemometer (LDA) and Digital where the trace of the SGS stresses W kk is
Particle Image Velocimetry (DPIV) techniques will be incorporated in the pressure resulting in a modified
employed to acquire the quantitative data[13,14]. The pressure term, Q t is the SGS kinematic viscosity,
purposes of comparison for numerical and and Sij is the strain rate tensor for the resolved
experimental results could primarily lead to a better
understanding of the effects of wake interference of scale defined by
the turbulent flows around the four-cylinder array in
an in-line square configuration. Moreover, the authors 1 wui wu j
will examine the suitability of the LES to predict Sij = ( + ) (5)
2 wx j wxi
accurately the complex turbulent flows around the
cylinder array thus minimising the costly and
complicated experimental work for the investigation The most basic SGS models was proposed by
of complex turbulent wake flows. Smagorinsky[15] and further developed by Lilly[16]. In
the Smagorinsky-Lilly model, the sub-grid kinematic
viscosity Q t is modeled by
2. Computational details
2.1 Governing equations and the sub-grid model Q t = ls2 Sij (6)
By using the 3-D LES turbulence model, the
large scale eddies are studied directly by solving the
filtered Navier-Stokes equations, and the small eddies where ls is the mixing length for the SGS, and
are modeled using a Sub-Grid Scale (SGS) model. Sij { 2 Sij Sij . ls can be computed using
The incompressible Navier-Stokes equations for the
evolution of the large-scale motions are obtained. The
governing equations employed for the LES are ls = min( ky , C s '1/ 3 ) (7)
405

where k is the von Kármán constant ( k = 0.42 ), y the estimated value of O z / D is 0.2. The
the distance to the nearest wall. C s the computational domain in the z direction is set at 3D.
Smagorinsky constant, and ' the volume of the
computational cell. Lilly[16] derived a value of 0.23 for
C s for homogeneous isotropic turbulence in the
inertial subrange. However, this kind of large C s
value was found to cause excessive damping of the
large-scale fluctuations in the presence of mean shear
or in transitional flows, and small values ( Cs  0.1 )
may cause convergence problems. For many
investigators, the Smagorinsky constant
( Cs = 0.1  0.14 ) has been found to yield good results
Fig.1 Schematic of the computational domain
for a wide range of flows [11,12]. For example, Zhao[12]
simulated flows around a circular pier at Re = 7040
The computational domain is divided into a
successfully by using the similar LES methods with
number of unstructured hexahedral grids (see Fig.2).
Cs = 0.1 . Considering the results above, all The total grid number is 2.8×106 for L/D = 1.5 and
computations in the present work were carried out 2.98×106 for L/D = 3.5 , respectively. The grid is
with a Smagorinsky constant of Cs = 0.1 , which is non-uniform in the x-y plane but uniform along the z
found to be a suitable value for the applications of the direction. The distance from the cylinder surface to
Smagorinsky model to turbulent wake simulation. the nearest grid points are fixed at y  close to 1.
2.2 Numerical method At the inlet boundary, a uniform velocity profile
In the present simulation, the Finite-Volume ( u = 1 , v = w = 0 ) is imposed, while the vanishing
Method (FVM) applied on unstructured grids is
gradient of velocity is used at the outlet boundary. A
employed to calculate the 3-D unsteady
periodic boundary condition is employed at the
incompressible Navier-Stokers equation. A second-
boundaries along the spanwise direction and a no-slip
order central differencing scheme is used for
boundary condition is prescribed at the surface of the
momentum discretization while a second-order
cylinders. The lateral surfaces are treated as slip
implicit scheme is employed to advance the equations
surfaces using symmetry conditions. The maximum
in time. The well-known Pressure Implicit method
CFL number is close to 2 to ensure sufficiently small
with Splitting of Operators (PISO) algorithm is used
CFL numbers, less than 1 for most part of the
to deal with the pressure-velocity coupling between
computational domain.
the momentum and continuity equations.
2.3 Computational domain and boundary conditions
Figure 1 shows a schematic diagram of the
computational domain in the present study. The origin
of the coordinate system is located at the center of the
four cylinder arrangement with (x, y, z) denoting the
coordinates along the streamwise x-direction, the
transverse y-direction and the cylinder spanwise
z-direction, respectively. The computational domain in
the x and y directions are set at 32D and 20D,
respectively. The upstream boundary is set at 8D away
from the coordinate origin and the downstream
boundary is 24D away from the origin. Williamson et
al. [17] showed that the wavelength of the streamwise
vortex structures in the near wake of a circular
cylinder scales is given by Oz / D | 25 Re-1/2 . Further
downstream, the large scale structures for streamwise
vortices with the wavelengths O z / D | 1 [18]. Here, Fig.2 Grid around the four-cylinder array
O z is the spanwise wavelength of the vortices. In the
present simulations, the Reynolds number is fixed at
3. Datails of experiment
Re = U f D / v = 1.5 u 104 for all the cases studied. So The experimental investigation on the turbulent
flows around the four-cylinder array in an in-line
406

square configuration was conducted using a


closed-circuit water tunnel with a working square
cross-section of 0.3m u 0.6m and a length of 2.4 m.
The four circular cylinders of diameter D = 20mm
were placed horizontally with two end-plates at the
mid of the water tunnel. The spacing ratio ( L/D )
could be set at 1.5 and 3.5 and the aspect ratio ( H/D )
at 20, here, where H is the length of the cylinders
along the spanwise direction. This leads to a blockage
ratio (per cylinder) of ~ 5%. The incoming flow
velocity U f was adjusted to 0.75 m/s, corres-
ponding to the Reynolds number 1.5×104. The LDA
and DPIV systems used in the present work were
described in detail by Lam et al. [7].

4. Results and discussions


To obtain all the instantaneous and mean
physical information as well as examine the suitability
of the LES for complex turbulent flow simulation, 3-D
numerical simulations of the turbulent flow around the
four-cylinder array in an in-line square configuration
with the spacing ratio 1.5 and 3.5 have been carried
out using LES model at Re = 1.5 u 104 . The
experimental results provided solid foundation for the
validation of numerical models. Initially, attention has
been focused on the accuracy of the LES calculations.
In the following section, the results of the simulations
will be compared with the experimental data for the
vortex structure, velocity and pressure coefficients
provided by the present DPIV and LDA
measurements as well as by Lam and Lo[5].
4.1 Flow patterns
Figure 3 presents the instantaneous flow patterns
in the x-y sections for the four-cylinder array with the
spacing ratios 1.5 and 3.5 at Re = 2100 . The
computations captured the complicated instantaneous
vortex structure behind the downstream cylinders of
the different biased flows for the case of L/D = 1.5
(see Figs.3(a) and 3(b)). The free shear layers from the
upstream cylinders shield the corresponding
downstream cylinders completely and show a
bi-stable flow pattern behind the downstream
cylinders. The wake structure is similar to that
observed for L/D = 1.54 and Re = 2100 by Lam
and Lo [5] and the intermittent change of the wide and
narrow wakes behind the downstream cylinders was
also observed. However, this bi-stable nature of the
wake flow appears to occur randomly. The wake
dynamics associated with the amalgamation and the
Fig.3 Instantaneous vortex structures for the four-cylinder
squeezing effect of the vortices behind the
downstream cylinders at this configuration are clearly array at Re = 1.5 u 104 using LES method
demonstrated. Furthermore, at L/D = 3.5 , the near
wake vortex structures between the four cylinders anti-phase vortex shedding (see Fig.3(d)). The free
show the in-phase vortex shedding (see Fig.3(c)) and shear layers from the upstream cylinders roll up into
mature vortices between the upstream and
407

shedding patterns (in-phase and anti-phase) occur


randomly for the case of L/D = 3.5 . All the captured
instantaneous flow patterns above are found to be in
good qualitative agreement with the observed flow
visualization results derived from DPIV at the same
Reynolds number (see Fig.4).
4.2 Force and Strouhal numbers
The total drag coefficient and lift coefficient are
defined by

2 FD
CD =
UU f2 DH

and

2 FL
CL =
UU f2 DH

respectively. And the Strouhal number is defined by


St = f s D / U f , where FD , FL , f s are the total drag
force, lift and the vortex shedding frequency obtained
from the frequency of lift fluctuating, respectively. In
the present simulations, the mean drag coefficients
CD1 , CD 2 , CD3 and CD 4 where the subscripts
“1” , “2” , “3” and “4” denote the upstream cylinders
and downstream cylinders, respectively. (refer to
Fig.1), the root-mean-square fluctuating lift
coefficients C Lc 1 , C Lc 2 , C Lc 3 and C Lc 4 and the
corresponding Strouhal number compared with the
experimental results of Sayers[4], Lam et al.[6,8] are
shown in Table 1. There is only a slight disparity
between the present results and those obtained from
experiments. It may be due to the effects related to the
differences in Re, the spacing ratio ( L/D ), the
ambient turbulence, etc.. It should be noted that CD 4
and C Lc 4 differ from CD3 and C Lc 3 at the case of
L/D = 1.5 . This could be attributed to the different
formation lengths and wake regions behind the
downstream cylinder 3 and 4 (see Figs.3(a), 3(b) and
4(a), 4(b). Those effects contrast with the difference in
the transverse direction and thus lead to corresponding
different mean drag coefficients and associated
fluctuating lifts.
4.3 Mean velocity distributions
Figure 5 shows the distributions of normalized
mean streamwise velocity ( U / U f ) and mean
transverse velocity ( V / U f ) in the near wake of the
Fig.4 Instantaneous vortex structures for the four cylinders at four cylinders with the spacing ratios L/D = 1.5 and
Re = 1.5 u 104 using DPIV measurements 3.5 and the Reynolds number at 1.5×104. The present
LES results of both the mean streamwise and
downstream cylinders and impinge on the downstream transverse velocity distributions show excellent
cylinder surfaces at Re = 1.5 u 104 . Also the vortex agreement with those by LDA measurements. From
408

Table 1 Experimental and LES results on mean drag coefficients C D , r.m.s. lift coefficients CLc and Strouhal numbers
St for four cylinders in an in-line square configuration

Sayers [4] Exp. Lam [6] Exp. Lam [8] Exp. Present 3-D LES
Results H/D = 11.6 results
H/D = 28.4 H/D = 17

Re 3.0×104 1.28×104 4.1×104 1.5×104

L/D 1.5 4.0 1.56 3.58 1.69 3.4 1.5 3.5

C D1 1.21 1.27 1.19 1.35 1.40 1.19 1.36 1.28

CD2 í í 1.24 1.36 í í 1.35 1.27

CD 3 í í  0.29 0.59 í í  0.23 0.70

CD 4  0.13 0.38  0.37 0.59  0.20 0.25  0.19 0.69

CLc 1 í í í í 0.11 0.06 0.10 0.66

CLc 2 í í í í í í 0.06 0.62

CLc 3 í í í í í í 0.28 1.22

CLc 4 í í í í 0.19 0.38 0.34 1.25

St1 0.124 0.186 í í í í í 0.196

St2 í í í í í í í 0.192

St3 í í í í í 0.20 0.125 0.192

St4 0.103 0.177 í í 0.22 0.16 0.169 0.196

these quantitative and qualitative comparisons, the ( x/D = 2.25 , y/D = 0.75 ) and the vortex formation
validation provides a confidence in employing the length is different for the two downstream cylinders. It
LES model to replicate and predict the complex suggests that the flow structure behind the
turbulent flow characteristics of the four in-line downstream cylinders is biased to on side and it
cylinders configuration. Therefore, the three- exhibits a bi-stable state of a wide and narrow wake.
dimensional numerical investigation can be used to From x/D = 4.25 to x/D = 11.25 , the normalized
reveal more quantitative information which may not mean sreamwise velocity distribution indicates that
be very easy to obtain by experimental investigation. the wake structure from the downstream Cylinders 3
At L/D = 1.5 and x/D = 2.25 , the value of and 4 forms an amalgamated structures with the wide
mean streamwise velocity ( U / U f ) shows obvious wake dominating the narrow wake. This phenomenon
difference at the positions of y/D = 0.75 and of wake characteristics is strongly supported by the
y/D = 0.75 . It is positive in the plane y/D = 0.75 flow pattern results obtained from LES and DPIV (see
Figs.3(a), 3(b) and 4(a), 4(b)). The same as the flow
while is negative in the plane y/D = 0.75 (see characteristics showed with the mean transverse
Fig.5(a)). This implies the flow is still in the reverse velocity ( V / U f ) at L/D = 1.5 (see Fig.5(b)).
flow region behind the downstream Cylinder 3
409

cylinders occurs for Re = 1.5 u 104 . In the plane


x/D = 2.75 where the distance from the centre of the
downstream cylinders is about 1D (see also Fig.5(d)),
the mean transverse velocity ( V / U f ) distributions
show the same flow characteristics behind upstream
Cylinders 1 and 2, and downstream Cylinders 3 and 4.
This kind of flow pattern may be clearly seen in Figs.
3 and 4. Furthermore, at Re = 1.5 u 104 , the flow
behaviour such as in-phase vortex shedding,
anti-phase vortex shedding and synchronized vortex
shedding from the upstream cylinders were also
observed using the LES and the DPIV technique. The
instantaneous streamlines around the four cylinders
derived from the 3-D LES results at L/D = 1.5 and
L/D = 3.5 are shown in Fig.6. The clearly bi-stable
flow characteristics at L/D = 1.5 could be observed.
It shows a deflected gap flow with a narrow and wide
wake patterns. The vortex streets become asymmetric
with different vortices formation lengths behind the
downstream cylinders at the same spacing ratio (Refer
to Figs.3 and 4). Moreover, the symmetrical vortex
shedding characteristics at L/D = 3.5 are observed.

Fig.5 Comparison of the mean streamwise velocity U / U f


and the mean transverse velocity V / U f distributions
derived from LDA and LES at different x-positions for
L / D = 1.5 and 3.5, Re = 1.5 u 104 Fig.6 Instantaneous streamlines at z = H/2 and
4
Re = 1.5 u 10
For the case of L/D = 3.5 (see Fig. 5(c)), in the
plane the x/D = 0.625 , the velocity values are
Figure 7 shows the instantaneous vorticity
negative (reaching about  0.2 and  0.1 for Re
contours in the x-z plane with the streamwise vorticity
=1.5×104) behind the upstream cylinders and in the
( Z x ), crosswise vorticity ( Z y ) and spanwise vorticity
planes x/D = 0 and x/D =0. 625, the mean streamwise
velocity ( U / U f ) values are positive at the ( Z z ), respectively. Considering the discussions above,
y/D = r1.75 . It indicates that the flow behind the the vortex shedding disappears between the Cylinders
upstream cylinders is in the high reverse flow 1 and 4 with L/D = 1.5 at such Reynolds number, so
condition and the vortex shedding of the upstream the strength of the vorticity between the Cylinders 1
410

and 4 with L/D = 1.5 are less than that with L/D = 1.5 (see Fig.8(a)). The maximum of TKE
L/D = 3.5 . However, the similar wake structures behind the downstream cylinders positions appears at
behind the Cylinder 4 are observed for both the the two points of x/D = 2.1 , y/D = 0.4 and
cylinder arrays with L/D = 1.5 and 3.5. The evident x/D = 2.2 , y/D = 0.6 . It shows asymmetry along
3-D effects of the vortex structures are also observed the streamwise direction with the axis line at y/D = 0 .
for such cylinder configurations.
This characteristic also reflects the bistable flow
characteristics which we discussed above. While the
TKE distributions in the x-y plane for the case
L/D = 3.5 shows a symmetrical characteristic (see
Fig. 8(b)). The maximum of TKE is found behind the
upstream cylinders. That means the evident vortex
shedding appears from the upstream Cylinders 1 and 2.
It also confirmed the phenomenon which we observed
using both LES method and DPIV measurements
(refer to Figs.3(c), 3(d) and Figs.4(c), 4(d)). Figure 9
shows that, along the spanwise direction (z direction),
the mean TKE distributions in the x-z plane show a
little difference for the four cylinder arrays of
L/D = 1.5 and the same as that of L/D = 3.5 .

Fig.7 Instantaneous vorticity contours in the x-z plane


( Z x D/U f = Z y D / U f = r10 , contour levels equal to
10)

4.4 Turbulent Kinetic Energy (TKE) Fig.8 Mean turbulent kinetic energy distributions in the x-y
Figure 8 shows the mean Turbulent Kinetic plane and z = H/2
Energy (TKE) distributions in the x-y plane for the
four cylinders cases of L/D = 1.5 and 3.5 at 4.5 Mean and fluctuating pressure coefficient
Re = 1.5 u 10 4 . The distributions of TKE show the distributions
difference in the near wake positions of downstream The pressure coefficient on the cylinder surface is
Cylinders 3 and 4 for the cylinder arrays with
411

defined as C p = ( p  pf )/(1/2UU f2 ) , where pf is coefficient C cp distributions of the upstream


static pressure of the oncoming flow. The mean cylinders displays a similar distribution to the
pressure coefficient C p and fluctuating pressure downstream cylinders. It shows symmetry
distributions base on the x-axis. The lowest values of
coefficient C cp distributions on the four cylinders are
C cp for four cylinders are all around 0o . The higher
shown in Figs.10 and 11. The experiment data of Lam
values of C cp of the upstream cylinders are around
and Fang [6] at approximately similar spacing ratios
( L/D = 1.56 and 3.58 ) and Reynolds number 80o and 280o , while those for both downstream
Re = 1.28 u 10 4 are also included in the figures. The cylinders occur at about 50o and 310o . It suggests that
present 3-D LES results predict the mean pressure the periodic vortex shedding from the upstream
coefficient distributions on the four cylinders within cylinders occurs and impinges on the surfaces of
the experimental accuracy. For the case L/D = 1.5 downstream cylinders. The present DPIV flow
visualization results also support this flow pattern
(see Fig.10(a)), the mean pressure coefficient C p characteristic for the two spacing ratios (see Fig.4).
distributions exhibits asymmetry about the point
T = 180o for the upstream Cylinders 1 and 2. This is
due to the interactions between the inner shear layers
from the upstream Cylinders 1 and 2. It suggests that a
bi-stable flow pattern has been created. The mean
pressure coefficient distributions display only one
maximum point at 300o for the downstream Cylinder
3 and 60o for the downstream Cylinder 4. This implies
a fully shielding flow pattern occurring around the
four cylinders at this spacing ratio. These calculated
data can clearly be explained by the present DPIV
visualization study (see Fig.4). For L/D = 3.5 (see
Fig.11(a)), the mean pressure coefficient C p of the
downstream Cylinders 3 and 4 at the front point
increases sharply up to the value of  0.1, which
suggests that the flow pattern of the cylinders is the
vortex shedding flow pattern. (see Figs.4(c), 4(d)).
The present LES data are also in excellent agreement
with the data determined experimentally by Lam and
Fang[6]. Furthermore, the calculated fluctuating
pressure coefficient distributions C cp for L/D = 1.5
and 3.5 are also shown in Figs.10(b) and 11(b). The
fluctuating pressure coefficient C cp distributions of
the upstream cylinders is in general varying within a
small range from 0.08 to 0.12, but vary up and down
rapidly with one local maximum in the range of
0o d T d 180o and two local maxima in the range of
180o d T d 360o for L/D = 1.5 . The magnitudes of Fig.9 Mean turbulent kinetic energy distributions in the x-z
C cp of the downstream cylinders are much larger plane
compared with that for corresponding upstream
cylinders. The first maximum peak of C cp occurs at
5. Conclusions
42o and the second maximum peak of C cp occurs at The turbulent flow around the four cylinder
295o for the down Cylinder 4. It indicates that the arrays with spacing ratios L/D = 1.5 and 3.5 at a
outer shear layer of the downstream Cylinder 4 could subcritical Reynolds number of 1.5×104 have been
be actually developed from the upstream Cylinder 1 investigated using LES method and experimental
due the vortex-shielding flow pattern and the inner techniques (LDA and DPIV). The major results are
shear layer of the downstream Cylinder 4 is affected summarized as follows:
by the shear layer of the downstream Cylinder 3. For Depending on the spacing ratio, at the subcritical
the case L/D = 3.5 , the fluctuating pressure Reynolds number, the distinct flow patterns have been
412

observed. The bi-stable flow nature has been observed


at L/D = 1.5 both in experimental studies and in
numerical simulations. The distinct vortex shedding
(in-phase and anti-phase) of the upstream cylinders
occurs randomly at L/D = 3.5 . The vortex formation
length for the upstream and downstream cylinders is
affected by the spacing ratio ( L/D ). Moreover, the
flow patterns of the turbulent flow around four
cylinder arrays have an essential influence on the
mean and fluctuating pressure and forces coefficient
distributions. 3-D numerical simulation using the LES
method well replicates the experimental features of
the complex turbulent flow around the four cylinder
arrays. It can be used to predict the mean velocity
fields, the mean pressure and forces coefficient, etc..
Hence it is a reliable method to reveal the full field
quantitative information for any instance which may
be extremely difficult to obtain by experimental
investigation.

Fig.11 Pressure coefficient distributions around the four


cylinders at L / D = 3.5 and Re = 1.5 u 104 . The
solid lines denote the data obtained from present LES
results, while the dashed lines denote the data obtained
from experimental measurements for L / D = 3.58 and
Re = 1.28 u 104 by Lam and Fang[6]

Fig.10 Pressure coefficient distributions around the


References
four-cylinder array at L / D = 1.5 and Re = 1.5 u 104 .
The solid lines denote the data obtained from present
LES results, while the dashed lines denotes the data [1] NORBERG C. LDV-measurements in the near wake of a
obtained from experimental measurements for circular cylinder[C]. Proceedings of the 1998 ASME
Fluids Engineering Division Summer Meeting.
L / D = 1.56 and Re = 1.28 u 104 by Lam and Fang[6] Washington, USA, 1998, FED-Vol. 245: 1-12.
[2] SUMNER D., PRICE S. J. and PAIDOUSSIS M. P.
Flow-pattern identification for two staggered circular
cylinders in cross-flow [J]. Journal of Fluid Mechanics,
Acknowledgements
2000, 411(1): 263-303.
This work was supported by the Research [3] SAYERS A. T. Flow interference between four
Program of the Wuhan University of Technology, equispaced cylinders when subjected to a cross flow[J].
China(Grant No. 471-38650324). Journal of Wind Engineering and Industrial
Aerodynamics, 1988, 31(1): 9-28.
413

[4] SAYERS A. T. Vortex shedding from groups of three London, UK: Springer. 2008, 7: 409-413.
and four equispaced cylinders situated in cross-flow[J]. [11] XU Chang-yue, CHEN Li-wei and LU Xi-yun. Large-
Journal of Wind Engineering and Industrial eddy and detached-eddy simulations of the separated
Aerodynamics, 1990, 34(2): 213-221. flow around a circular cylinder[J]. Journal of
[5] LAM K., LO S. C. A visualization study of cross-flow Hydrodynamics, Ser. B, 2007, 19(5): 559-563.
around four cylinders in a square configuration[J]. [12] ZHAO Wei, Huhe Aode. Large-eddy simulation of
Journal of Fluids and Structures, 1992, 6(1): 109-131. three-dimensional turbulent flow around a circular
[6] LAM K., FANG X. The effect of interference of four pier[J]. Journal of Hydrodynamics, Ser. B, 2006,
equispaced cylinders in cross flow on pressure and force 18(6): 765- 772.
coefficients[J]. Journal of Fluids and Structures, 1995, [13] QI E-rong, LI Guo-ya and LI Wei et al. Study of vortex
9(2): 195-214. characteristics of the flow around a horizontal circular
[7] LAM K., LI J. Y. and CHAN K. T. et al. Flow pattern cylinder at various gap-rations in the cross-flow[J].
and velocity field distribution of cross-flow around four Journal of Hydrodynamics, Ser. B, 2006, 18(3):
cylinders in a square configuration at low Reynolds 334-340.
number[J]. Journal of Fluids and Structures, 2003, [14] Daichin, KANG Wen and ZHAO Li-li. PIV mea-
17(5): 665-579. surements of the near-wake flow of an airfoil above a
[8] LAM K., LI J. Y. and SO R. M. C. Force coefficient and free surface[J]. Journal of Hydrodynamics, Ser. B,
Strouhal numbers of four cylinders in cross flow[J]. 2007, 19(4): 482- 487.
Journal of Fluids and Structures, 2003, 18(3): [15] SMAGORINSKY J. General circulation experiments
305-324. with the primitive equations, I. The basic experiment[J].
[9] FARRANT T., TAN M. and PRICE W. G. A cell Monthly Weather Review, 1963, 91(3): 99-164.
boundary element method applied to laminar [16] LILLY D. K. A proposed modification of the Germano
vortex-shedding from arrays of cylinders in various subgrid-scale closure method[J]. Physics of Fluids,
arrangements[J]. Journal of Fluids and Structures, 1992, 4(3): 633-635.
2000, 14(3): 375-402. [17] WILLIAMSON C. H. K., WU J. and SHERIDAN J.
[10] LAM K., ZOU L. The effects of aspect ratio and end Scaling of streamwise vortices in wakes [J]. Physics of
condition on the control of free shear layers Fluids, 1995, 7(10): 2307-2309.
development and force coefficients for flow past four [18] WILLIAMSON C. H. K. Vortex dynamics in the
cylinders in the in-line square configuration [C]. cylinder wake [J]. Annual Review of Fluid Mechanics,
IUTAM Symposium on Flow Control and MEMS. 1996, 28: 477- 539.

Das könnte Ihnen auch gefallen