Home Search Collections Journals About Contact us My IOPscience

LBM study on heat transfer and fluid flow past a rotating isothermal cylinder

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2012 IOP Conf. Ser.: Earth Environ. Sci. 15 062009

(http://iopscience.iop.org/1755-1315/15/6/062009)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 160.75.33.93
This content was downloaded on 18/08/2014 at 12:24

Please note that terms and conditions apply.

26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

LBM study on heat transfer and fluid flow past a rotating

isothermal cylinder

F Yang, X M Shi and X Y Guo

School of Energy and Power Engineering, University of Shanghai for Science and
Technology, Shanghai, 200093, P. R. China

E-mail: usstyf@126.com

Abstract. In this paper, a viscous fluid flowing past a rotating isothermal cylinder with heat
transfer is studied by lattice Boltzmann method (LBM) based on double-population approach.
A numerical strategy presents for dealing with curved boundaries of second order accuracy for
both velocity and temperature fields. The system, with constant and temperature-dependent
viscosity, of fluid flowing and heat transfer are simulated numerically. The effects of the
peripheral-to-translating-speed ratio on flow and heat transfer are discussed.

1. Introduction
Fluid flow around a rotating isothermal cylinder is a common occurrence in a variety of industrial
processes. Although the configurations are normally rather simple, the flow past the rotating cylinder
and the heat transfer performance between the fluid and the cylinder are very complex [1]. Following
factors involved in the flow, such as the features of viscous wakes including dissipation, diffusion and
cancellation of the vortices, the local heat transfer performance around the cylindrical surface, and the
evolution of surrounding temperature field with time, have greatly increased the difficulties to build up
mathematical formulations and to simulate properly of such complex transport phenomena.
The lattice Boltzmann method, a derivative of lattice gas automata (LGA) method, has been
successfully demonstrated to be an alternative numerical scheme to traditional numerical methods for
solving partial differential equations and modeling physical systems, particularly for simulating fluid
flows with the Navier-Stokes equations. In traditional numerical methods, the Navier-Stokes equations
are solved by some specific numerical discretization. In contrast, the fundamental principle of lattice
Boltzmann method is to construct a simplified molecular dynamic that incorporates the essential
characteristics of the physical microscopic processes so that the macroscopic averaged properties obey
the desired macroscopic equations. This microscopic approach in the lattice Boltzmann method
incorporates several advantages of kinetic theory. It includes clear physical pictures, easy
implementation of boundaries and fully parallel algorithms [2].
In this paper, a two-dimensional nine-velocity (D2Q9) LBM model with double distribution
functions is employed to simulate incompressible viscous thermal flows. A distribution function of
density is used to calculate density and velocity fields. And the temperature field is obtained by
solving the distribution function of temperature. The system, with constant and temperature-dependent
viscosity, of fluid flowing and heat transfer are simulated numerically. And the effects of the
peripheral-to-translating-speed ratio on flow and heat transfer are discussed.

Published under licence by IOP Publishing Ltd 1

26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

2. The LBM model

2.1. Lattice Boltzmann equation of flow field

The lattice Boltzmann equation of velocity field can be discretised into the following form:
1
f ( xi  e  t , t   t )  f ( xi , t )   [ f ( xi , t )  f( eq ) ( xi , t )] (1)

where:  represents the lattice direction and  denotes a relaxation factor. At each time t , we
consider the particle distribution function f ( xi , t ) located in lattice node xi and moving with a
lattice velocity e . In the D2Q9 model (Fig. 1), e denotes the discrete velocity set, namely,

0 1 0 1 0 1 1 1 1 
e   c
0 0 1 0 1 1 1 1 1
where: c   x /  t ,  x and  t are the lattice constant and the time step size, respectively.
The equilibrium distribution for D2Q9 model (Fig. 1) is of the form [3]
3 9 3
f( eq )   [1  2
(e  u)  4 (e  u)2  2 u  u] (2)
c 2c 2c
where:  is the weighting factor given by

4 1 1 1 1 1 1 1 1
    (3)
9 9 9 9 9 36 36 36 36 

The speed of sound in this model is cs  c / 3 and in the discretized velocity space, the density
 and momentum fluxes  u are defined as particle velocity moments of the distribution function,
8 8
   f ,  u   e f (4)
 0  0

According to the limit of incompressible flow, Ma  u / cs 1 and the Chapman-Enskog

expansion, the mass and momentum equations can be derived from the D2Q9 model as [4]
u  0 (5)

u 1
 (u )u   p  2 u (6)
t 
where: the pressure p satisfies the equation of state as p   cs2 ; the kinematical viscosity  is
determined by  (  0.5) cs  t .
2

2.2. Lattice Boltzmann equation of temperature field

For incompressible flow the lattice Boltzmann equation of temperature field can be given by [5]:
1
g ( xi  e  t , t   t )  g ( xi , t )   [ g ( xi , t )  g( eq ) ( xi , t )] (7)


2
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

where: g is temperature distribution function in the  direction and   is the relaxation time.
g( eq ) is the corresponding equilibrium distribution function and can be expressed as:
3
g( eq )   T [1  (e  u)] (8)
c2
where: T is the fluid temperature and can be evaluated from:
8
T   g (9)
 0

In addition, similar to mass and momentum equations, equation of temperature can be obtained as:
T
   (uT )   2T (10)
t
where  is the diffusivity coefficient which is represented as   (   0.5) cs  t
2

In general, both Eqs. (1) and (7) can be solved in two steps:
collision step:

~f a xi , t  t   f  xi , t   1

 f x , t   f   x , t 
a i 
εq
i (11a)

g~ a xi, , t  t   g  xi , t  
1

g x ,t   g   x , t 
a i 
εq
i (11b)

streaming step:

f ( xi  e  t , t   t )  f ( xi , t   t ) (12a)

g α ( xi  eαt, t  t )  g~α ( xi , t  t ) (12b)

where: ~ ~ represent the post-collision state.

f and g

3. Solid boundary treatment of LBM

In Fig. 2, a curved wall separates the solid region from the fluid region. The lattice node on the
fluid side of the boundary is denoted as x f and that on the solid side is denoted as xb . The filled small
circles on the boundary, x w , denote the intersections of the wall with various lattice links. The
boundary velocity at x w is uw . The fraction of an intersected link in the fluid region is  , that is,

x f  xw
 , 0   1 (13)
x f  xb

The horizontal or vertical distance between xb and x w is  x on the square lattice. The particle
momentum moving from x f to xb is e and the reversed one from xb to x f is e  e .

3.1. No-slip velocity condition in curved boundary

3
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

To construct f ( xb , t ) based upon known information in the surrounding fluid nodes, Bouzidi et al.
[6] presented a simpler boundary condition based on the bounce-back for the wall located at arbitrary
position. In their work, both linear scheme and the quadratic schemes were given to obtain
f ( x f , t   t ) which is equivalent to f ( xb , t ) . The linear version was as follow:

1 2  1
f ( x f , t   t )  f ( x f , t )  f ( x f , t ) for   0.5 (14a)
2 2

f ( x f , t   t )  2 f ( x f , t )  (1  2) f ( x ff , t ) for   0.5 (14b)

Figure 1. 2-Dimensional 9-velocity Figure 2. Layout of the regularly spaced lattices

lattice (D2Q9) model. and curved wall boundary.

3.2. Temperature condition in curved boundary

For temperature field in curved boundary this study use the method is based on the method reported in
[7]. Distribution function for temperature divided two parts, equilibrium and non equilibrium
g ( xb , t )  g( eq ) ( xb , t )  g( neq ) ( xb , t ) (15)

By substituting Eq. (15) into Eq. (12b) we have

1 ( nεε )
g~α ( xb , t  t )  g (q ) ( xb , t )  (1  ) g  ( xb , t ) (16)
τγ
Equilibrium and non equilibrium parts of Eq. (15) are define as
3
g( eq ) ( xb , t )   Tb* (1  e  ub ) (17)
c2 
where: Tb* is determined by linear extrapolation using either:

Tb*  Tb1 , if   0.75 (18a)

Tb*  Tb1  (1  )Tb 2 , if   0.75 (18b)

and

4
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

1
Tb1  [Tw  (  1)T f ] (19b)


1
Tb 2  [2Tw  (  1)T ff ] (19b)
1 
here: T f and T ff denote the fluid temperature in node x f and x ff , respectively.
Similarly, g( neq ) ( xb , t ) can be evaluated as:

g( neq ) ( xb , t )   g( neq ) ( x f , t )  (1    g( neq ) ( x ff , t ) (20)

4. Numerical simulation
Fig. 3 shows the physical model of flow and heat transfer around a rotating isothermal circular
cylinder of radius R . At the entrance, fluid with the constant temperature Tc is injected into the
domain with constant uniform velocity U in x-direction. Meanwhile, a free outflow boundary with
zero velocity and temperature gradients in x-direction is set at the right hand side boundary of the
domain. The upper and lower boundaries are set as free-slip velocity and heat insulated boundaries. In
the present simulations, the boundaries at upstream and downstream are set, respectively, as 6.6 and
12.7 times of the radius away from the centre of the cylinder; the upper and lower walls are both set as
8.07 times of the radius away from the cylinder centre.

Figure 3. Flow field set-up.

In the present study, the flow and heat transfer are simulated at Pr  0.5 and Re  200 , where the
Reynolds number is defined as Re  2UR  . Parameter k is introduced to define the rate of the
peripheral velocity V   R to the constant uniform inflow velocity U , i.e. k  V / U . t *  Ut / R is
dimensionless time.
The flow and heat transfer, with temperature-dependent viscosity, are also simulated. The viscosity
variation with temperature is characterized by F ( )   (T )  0 (T0 )  exp[C (1  1  )] , where the
viscosity  0 is evaluated at some reference temperature T0 . The constant C is taken to be equal to 2
in this study, and the dimensionless temperature is defined as   T T0 .
The heat transfer convected from the cylindrical surface is estimated in terms of Nusselt number.
Once the temperature field is determined, the following Nusselt numbers are defined as
2R T
Nu   ( ) wall (21)
(Th  Tc ) n

5
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

where n is the outer-normal vector of cylindrical wall; angle  equals zero at the rearmost point of
the cylinder and increases in anticlockwise.

5. Results and discussion

(a) t *  2.0 (b) t *  4.0 (c) t *  6.0 (d) t *  8.0 (e) t *  10.0

(f) t *  2.0 (g) t *  4.0 (h) t *  6.0 (i) t *  8.0 (j) t *  10.0

Figure 4. Evolution of velocity streamlines and temperature contours for Re  200 , k  0 , Pr  0.5
(a)-(e): results of constant viscosity; (f)-(j): results of temperature-dependent viscosity.

(a) t *  2.0 (b) t *  4.0 (c) t *  6.0 (d) t *  8.0 (e) t *  10.0

(f) t *  2.0 (g) t *  4.0 (h) t *  6.0 (i) t *  8.0 (j) t *  10.0

Figure 5. Evolution of velocity streamlines and temperature contours for Re  200 , k  0.1 , Pr  0.5
(a)-(e): results of constant viscosity; (f)-(j): results of temperature-dependent viscosity.

(a) t *  2.0 (b) t *  4.0 (c) t *  6.0 (d) t *  8.0 (e) t *  10.0

(f) t *  2.0 (g) t *  4.0 (h) t *  6.0 (i) t *  8.0 (j) t *  10.0

Figure 6. Evolution of velocity streamlines and temperature contours for Re  200 , k  0.5 , Pr  0.5

6
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

(a)-(e): results of constant viscosity; (f)-(j): results of temperature-dependent viscosity.

The effects of parameter k on system, with constant and temperature-dependent viscosity, of flow and
heat transfer are predicted numerically. Figs. 4, 5, and 6 show the early stage of temperature contour
evolution and corresponding velocity streamlines at k  0.0 , 0.1, and 0.5, Re  200 , Pr  0.5 ,
respectively.
In Fig. 4, the cylinder does not rotate as k  0 ; two opposed symmetrical vortices are formed
simultaneously because of the symmetry of the velocity gradient. The upper vortex rotates clockwise,
and the lower vortex rotates anticlockwise. These two vortices grow in width and length with time,
remain as symmetrical for a certain period of time and are stably attached to the cylinder. During this
period, heat transfer also remains symmetry between the upper and lower part of the cylinder. After
this, the vortices become asymmetrical and are shed alternately downstream, and meanwhile, the
symmetry of heat transfer is destroyed.
When k  0 (see Figs. 5 and 6), the velocity and temperature fields are quite similar to those for
k  0 at the start of cylinder rotation, but as t * increases slightly, differences in the flow become
obvious. The wall shear gradient becomes asymmetrical because the upper side of the cylinder moves
against the free stream whereas the lower one moves in the direction of the free stream. As a result, the
symmetry of the wall vorticity is destroyed and asymmetrical vortices appear. Moreover, the upper
vortex appears earlier than that on the lower side.

(a) k  0.0 (b) k  0.1 (c) k  0.5

(d) k  0.0 (e) k  0.1 (f) k  0.5

Figure 7. Distributions of local Nusselt numbers for Re  200 , Pr  0.5 ,  [  2,  2]

(a)-(c): results of constant viscosity; (d)-(f): results of temperature-dependent viscosity.

7
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

The other effect of parameter k on flow is to accelerate the detachment of vortices. For k  0 (see
Fig. 4) the vortices remain symmetrical and are stably attached to the cylinder when t *  0  10 ,
whereas for k  0.1 , as shown in Fig. 5, the first vortex is completely shed downstream at t *  10.0 ;
while, for k  0.5 , this happens more rapidly at t *  6.0 . Indeed, this effect is very evident at the
early stage of the flow and becomes weaker with the development of time.
The variable viscosity effects on the flow and heat transfer are also simulated. It is observed from
Figs. 4, 5, and 6 that the overall flow patterns for a variable viscosity are similar to those for a constant
viscosity, however, the thermal boundary layer near the cylinder is thinner. It is because the local
viscosity near the hot cylinder is smaller for high temperature, which has influences on flow pattern
and heat transfer.
The effect of k on local heat transfer coefficient is also predicted. The distributions of local
Nusselt numbers at Re  200 , Pr  0.5 ,  [  2,  2] for k  0.0 , 0.1, and 0.5 are shown in
Fig. 7. It can be seen from the figure (a) that, for k  0 , the local Nu distribution is strictly
symmetrical with respect to   0 , and there are two local minimum values, and a local maximum
value.
This phenomenon may be connected with the fact that, for k  0 , the two vortices formed at the
rear surface of the cylinder (see Fig. 4) rotate in opposite directions, roll up more and more heat from
the upper and lower surfaces of the cylinder symmetrically while amplify the thickness of the thermal
layers, and consequently diminish much more temperature gradient. However, as shown in Fig. 7b and
c, the symmetry of local Nusselt number distribution is destroyed by the unsymmetrical motion of
vortices when k  0 . For example, for k  0.5 , the region with revised flow does not distribute along
the x-axis but rocks up and down along with the migration and the alternative shedding of the vortices.
The variable viscosity effects on distributions of local Nusselt numbers are also similar to those for
a constant viscosity (see Fig. 7d, 7e and 7f), but the values of Nusselt number are not shift with
dimensionless time.

6. Conclusions
A multi-distribution function LBM model for simulating the system, with constant and temperature-
dependent viscosity, of viscous fluid flow past a rotating isothermal cylinder with heat transfer is
presented. The model can deal with curved moving flow and thermal boundaries of second order
accuracy. The velocity and temperature distributions are presented. The effects of peripheral-to-
translating-speed ratio k on flow and heat transfer are evaluated. With the increase of k, the formation
of lower vortices is inhibited, vortices shedding are greatly accelerated at the early stage of flow, and
the asymmetry distributions of velocity and temperature are enhanced. The results showed that the
variation in viscosity has significant influences on both flow and heat transfer behaviors.

Acknowledgments
Project supported by the National Natural Science Foundation of China (Nos. 10902070 and
51176127), Open Research Fund Program of State Key Laboratory of Hydroscience and Engineering
(No. sklhse-2011-E-02), Open Research Fund Program of State Key Laboratory of Water Resources
and Hydropower Engineering Science (No. 2010B076), Special Science Foundation for Selection and
Cultivation of Excellent Young Scholars in Shanghai (No. slg09002), and Leading Academic
Discipline Project of Shanghai Municipal Education Commission (No. J50501).

References
[1] Yan Y and Zu Y 2008 Int. J. Heat Mass Transfer 51 p 2519
[2] Qian Y, d’Humières D and Lallemand P 1992 Europhys. Lett. 17 p 479
[3] Frisch U, Hasslacher B and Pomeau Y 1986 Phys. Rev. lett. 56 p 1505
[4] Yu D, Mei R, Luo L and Shyy W 2003 Prog. Aerosp. Sci. 39 p 329

8
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 062009 doi:10.1088/1755-1315/15/6/062009

[5] Guo Z, Shi B and Zheng C 2002 Int. J. Numer. Methods Fluids 39 p 325
[6] Bouzidi M, Firdaouss M and Lallemand P 2001 Phys. Fluids 13 p 3452
[7] Fattahi E, Farhadi M and Sedighi K 2010 Int. J. Therm. Sci. 49 p 2353