Anda di halaman 1dari 19

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1, pp.

3957 (2010)

FINITE VOLUME ALGORITHM TO STUDY THE EFFECT OF TUBE


SEPARATION IN FLOW PAST CHANNEL CONFINED TUBE BANKS
S. Jayavel and Shaligram Tiwari*
Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
* E-Mail: shaligt@iitm.ac.in (Corresponding Author)
ABSTRACT: Numerical investigations are carried out to study the effect of tube separation on flow and heat
transfer characteristics for flow past circular tube bundles confined within a rectangular channel. The tube separation
has been varied for cross flow of air past inline and staggered arrangements of tubes in such a way that in staggered
arrangement, center of an upstream tube and centers of two respective downstream tubes from an equilateral
triangle. A three-dimensional computational code based on SIMPLE method using finite volume technique has been
developed to solve the governing equations. A body-fitted structured grid has been generated for discretization of the
incompressible form of conservation equations of mass, momentum and energy. The flow and heat transfer
characteristics for different values of tube separation have been compared using streamline plots, temperature
contours, span-averaged pressure drop along length of the channel and span-averaged Nusselt number variation near
heated/cooled surfaces.
Keywords:
method

tube separation, tube banks, channel confinement, flow and heat transfer characteristics, finite volume

may be divided into two categories. The first type


is comprised of those flows in which the
recirculation region is large, typically of the same
order of magnitude as the characteristic body
dimension. The second type consists of those
flows in which the recirculation region
downstream of separation is small. It was first
shown by Catherall and Mangler (1966) that
smooth solutions could be computed beyond the
point of separation. They did that by replacing the
standard boundary layer problem by one in which
the displacement thickness is specified and the
pressure-gradient term is unknown. Further, a
brief and clear discussion of singular versus
regular flow at separation is given in Brown and
Stewartson (1969). Son and Hanratty (1969) and
Dennis and Chang (1970) numerically
investigated the nature of the large recirculation
region downstream of separation for low and
moderate values of Re. However, they did not
address the problem of separation directly. On the
other hand, Briley (1971) carried out numerical
solutions for a separation bubble that occurs after
separation in a linearly retarded flow. The flow
contains a small separation bubble of limited
stream-wise extent which is embedded within the
boundary layer. The flow far above the wall is
assumed to be irrotational with a constant xcomponent of velocity while the usual no-slip and
impermeable wall conditions are imposed at the
wall. The boundary conditions are thus prescribed

1. INTRODUCTION
Cylindrical geometries often appear in
engineering
applications.
Although
such
cylindrical bodies are of regular and simple shape,
the behavior of fluid flow and heat transfer
around them is quite complicated. The circular
cylinder is a common bluff body and forms a
large separated stagnant wake. The characteristics
of separated wakes cannot be predicted
analytically and hence must be analyzed either
numerically or experimentally. Unlike square or
rectangular cylinders, where the flow separates
from the leading edges, the flow separation may
occur from any location of a circular cylinder,
further complicating the analysis. The difficulties
in predicting flow and heat transfer around
circular cylinders get multiplied when two or
more of these cylinders are placed in proximity to
each other. The large separated wakes behind
each of the cylinders interact with each other to
give rise to a flow that is characteristically much
more different than the flow past a single cylinder.
The exact form of the interaction is highly
dependent on the cylinder spacing and the
Reynolds number (Re). There exist a large
number of ways in which cylinders may be placed
in proximity of one another.
Numerical solutions to the Navier-Stokes
equations obtained for flows based on the size of
the recirculation region downstream of separation

Received: 12 Mar. 2009; Revised: 13 Jul. 2009; Accepted: 31 Jul. 2009


39

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

study with the assumption of laminar flow, the


significant role of tubes arrangements in heat
transfer and pressure drop characteristics of fintube heat exchangers is studied. The heat transfer
and pressure drop characteristics for such tube
bundles in cross flow are of practical interest.
Therefore, the arrangement of tubes in cross-flow
involving multiple rows having inline or
staggered
layout
requires
comprehensive
investigations of flow and heat transfer. ElShaboury and Ormiston (2005) have conducted
two-dimensional numerical investigations on heat
transfer in laminar, steady cross flow in banks of
plain tubes with 10 longitudinal rows having inline arrangement. Wang, Penner and Ormiston
(2000) conducted similar study for staggered
arrangement of tubes, where they reported the
local and overall heat transfer and fluid flow
results at nominal pitch-to-diameter ratios of 1.25,
1.5, and 2.0 for equilateral triangle and rotated
square tube arrangements for Reynolds number
values of 100 and 300.
Griffiths and Awbery (1933) presented
measurements on heat transfer from constant
diameter iron tubes with their axes placed
transverse to the flowing air stream. They
observed that the second layer of pipes met by the
air loses more heat than the first layer with the
third layer losing the same as the second. They
explained that this is due to the eddy motion set
up in air as it passes the first layer, no further
increase in the eddy motion occurring afterwards.
They also reported the case with tubes maintained
at a temperature lower than that of the air stream.
Faghri and Rao (1987) carried out numerical
computations for laminar flow over a wide range
of Re and Pr. They observed a decrease in the
heat transfer rate, and small change in the
pressure drop, as a result of introducing fins. They
claimed that a decrease in heat transfer due to fins
placed in the stagnation regions occurred at the
front and rear of the tubes. Jang, Wu and Chang
(1996)
considered
incompressible,
threedimensional and laminar flow over a multi-row
(16 rows) plate-fin and tube heat exchangers.
The effects of different geometrical parameters
such as tube arrangement, number of tube rows
and fin pitch were investigated in detail for Re
ranging from 60 to 900. They reported that the
average heat transfer coefficient of staggered
arrangement is 15%27% higher than that of
inline arrangement, while the pressure drop of
staggered arrangement is 20%25% higher than
that of inline arrangement. They also reported that
the number of tube rows influences the heat
transfer such that the average Nusselt number

on the entire boundary of the domain in which the


bubble is embedded. Horton (1974) and Carter
and Wornom (1975) have also reported
computations of laminar boundary layers
associated with region of flow reversal. Flow
visualizations of Tanida, Okajima and Watanabe
(1973) showed that the flow for Re = 100 is
laminar for all values of cylinder spacing. For
Re = 300, they report the formation of a turbulent
vortex street when the cylinder spacing is greater
than a critical value. They reported about the
presence of a critical spacing for low Re flows.
In the present work, a systematic study of inline
and staggered tube bundles for tubes in different
locations and built-in with a confining channel
has been conducted. Computations are carried out
for a given laminar flow corresponding to a
fixed Reynolds number (ReD = 600). The
hydrodynamics of this problem has already been
studied by Bossel and Honnold (1976) without
considering the effect of tube separation. It is
known from literature that when the distance
between the channel walls is very small, the flow
is of Hele-Shaw type. As the channel separation is
increased, a horseshoe vortex is formed just
upstream of the tube, followed by a separated
region behind the tube. This region becomes
larger and eventually communicates with the fluid
downstream of the tube. Moreover, a peak in the
variation of local Nusselt number near channel
walls occurs at the location of the horseshoe
vortex. In the wakes behind tubes, the Nusselt
number is very small but increases when there is
exchange between the wake and downstream fluid.
Numerical investigations were carried out by
Tiwari et al. (2003) and Tiwari et al. (2005) to
study the behavior of flow and heat transfer for a
circular tube in cross-flow in the absence and
presence of an integral wake splitter attached at
the rear of the tube, respectively. Moreover,
Tiwari et al. (2006) studied unsteady flow past
confined circular tube and observed that the onset
of vortex shedding gets delayed due to the effect
of channel confinement. Pratish and Tiwari (2009)
have carried out two-dimensional numerical
investigations to study the behavior of unsteady
wake for flow past an inline arrangement of
square cylinders confined in a channel. They
studied the influence of the relative size and
arrangement of the two cylinders on vortex
shedding characteristics in their wakes. In
engineering applications complex turbulent flows
commonly occur. Paul, Ormiston and Tachie
(2008) presented the results of measurements and
numerical predictions of turbulent cross-flow in a
staggered tube bundle. However, in the present
40

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

flow is steady at the entrance of the tube banks,


but becomes oscillatory downstream beyond a
location of onset of vortex shedding. They found
that the location of onset of vortex shedding
moves upstream with increasing Re, and the
upstream development of flow transition are
much faster for the staggered array of tubes than
for the in-line array.
Finite-volume method is a powerful numerical
tool used to solve the governing flow and energy
equations. El-Shaboury and Ormiston (2005) and
Wang, Penner and Ormiston (2000) also have
used finite-volume method to solve the NavierStokes equations and energy equation using
boundary-fitted grid with co-located variables for
their study of flow and heat transfer past in-line
and staggered arrangement of tube. A pressurebased method on a collocated grid arrangement
for steady and unsteady flows was developed by
Darbandi and Vakilipour (2008). They estimated
the advection terms on the cell faces using an
inclusive pressure-weighted up-winding scheme
extended to unstructured grids. However, to avoid
a non-physical spurious pressure field pattern,
two mass fluxes per volume expressions were
employed at the cell interfaces. They compared
their results based on finite-volume and finiteelement methods. From the finite-volume
perspective, the approximation of fluxes at the
cell faces can be enormously improved by
incorporating the correct physics of flow in the
cell-face expressions. For example, a basic
upwind estimation for the convection terms
produces false diffusion (Raithby, 1976a).
However, the false diffusion can be decreased
drastically if up-winding is skewed in the flow
direction (Raithby, 1976b). Unfortunately, the upwinding does not contain all the physics behind
approximating the cell-face velocity. A more
inclusive suggestion is to incorporate the weight
of other key parameters in the approximation.
Schneider and Raw (1986) presented a massweighted, skewed, positive influence coefficient
upwind scheme on quadrilateral mesh. Schneider
and Raw (1987) and Darbandi and Schneider
(1998) showed that the approximation could be
effectively improved if the pressure was also
included in the advection term approximations.
Derksen
(1990)
developed
a
simple
computational scheme to compute laminar flows
inside axisymmetric ducts. Interactive vortex
shedding from a pair of circular cylinders in a
transverse arrangement at Re = 100 was reported
by Chang and Song (1990). Several reports on
solution algorithm and analysis of finite volume
method are available from studies of Ju and Du

decreases as the number of tube rows is increased


from 1 to 6. Wilson and Bassiouny (2000)
predicted the pressure drop and heat transfer
characteristics of laminar and turbulent flow of air
across tube bundles (single and two rows), where
the tube surfaces are kept at constant temperature.
Buyruk
(2002)
carried
out
numerical
investigations
to
predict
heat
transfer
characteristics for flow past inline and staggered
arrangement of cylinders using finite element
method. The author observed that for inline
arrangement of tubes, laminar boundary layer heat
transfer of the first cylinder is not affected by
decreasing the gap between the cylinders, but
decreases downstream of the first cylinder by
decreasing the gap, due to flow blockage by the
second cylinder. They also reported that when
cylinder spacing lsd = 6D, first and second cylinder
surface heat transfer coefficient distributions have
similar trend. For staggered arrangement of tubes,
local heat transfer coefficient in the front region
of the second cylinder is higher than that in the
front of the first cylinder due to the effect of flow
acceleration.
Baker (1991) studied the oscillatory behavior of
vortices formed at a single tube-plate junction in
the transitional regime, between the steady
laminar horseshoe vortices formed at low Re and
the fully turbulent horseshoe vortices that occur at
higher Re values. He reported that vortex
oscillations begin at ReD = 1000 and break down
into full turbulence at ReD = 1600. Roychowdhury,
Sarit
and
Sundararajan
(2002)
studied
numerically the effect of Re and tube spacing on
flow and heat transfer over staggered tube banks.
They observed that both the Re and tube spacing
influence the vortex formation and growth in the
region between the tubes. For sufficiently small
spacing, eddy formation gets completely
suppressed even at higher Reynolds number.
Tatsutani, Devarakonda and Humphrey (1993)
studied the effect of one tube on another towards
changes in the characteristics of flow and heat
transfer for different values of separation distance
between centers of the tubes. Nishimura,
Hisayoshi and Hisashi (1991) studied numerically
and experimentally the effect of tube layout on
the surface shear stress distribution for steady
laminar flow. They found significant difference in
the nature of shear stress distribution on the
upstream tubes for the staggered and in-line
arrays of tubes. Nishimura, Hisayoshi and Hisashi
(1993) studied flow characteristics past tube
banks in staggered as well as inline arrangement
in the transitional flow regime at intermediate
Reynolds numbers (50 ReD 1000), where the
41

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

(2009), Piller and Stalio (2008), Lacor, Smirnov


and Baelmans (2004) and Pereira, Kobayashi and
Pereira (2001).

methods of grid generation and grid smoothening


have been reported by Thompson, Warsi and
Mastin (1982), Thompson (1984), Connell and
Braaten (1995), Smith (1982) and Warsi and
Ziebarth (1982). Sensitivity of parameters, such
as the local Nusselt number and friction
coefficient distributions to the computational grid,
was reported by Wang, Penner and Ormiston
(2000). In the present study, structured, bodyfitting type orthogonal grid has been generated in
the solution domain and appropriate boundary
conditions are applied. Based on the grid, control
volumes are defined in order to evaluate the
integrals of the convective and viscous fluxes as
well as of the source term.
The grid is basically generated in two-dimension
(X-Y), which has a rectangular boundary with
circular regions embedded within it. This plane is
then extruded along the Z-direction to provide the
necessary three-dimensional grid, where the
location and diameter of the circles in the 2-D
plane decides the location and diameter of the
tube in 3-D. In the grid generation procedure, the
boundaries of the rectangular channel and the
circles within it are specified and the interior grid
nodes are joined by straight edges. The advantage
of this type of meshing is that the boundary points
and the forthcoming elemental cells can be easily
addressed by the indices (i, j, k). The grid nodes
within the boundary of rectangular channel
(excluding the circular regions) have been
generated using transfinite interpolation method
of Gordon and Thiel (1982). The grid nodes are
smoothened using PDE method.

2. COMPUTATIONAL DOMAIN
The computational domain for flow past circular
tubes, cross-confined and built-in with a
rectangular channel is shown in Fig. 1. The
channel is designed to mimic a passage formed by
any two neighboring fins in a fin-tube heat
exchanger. In the computational domain, the
location of the upstream tube is fixed whereas the
location of the downstream tube is suitably
altered to study the effect of tube separation and
also to represent inline and staggered
arrangements. All the length scales have been
non-dimensionalized with respect to tube
diameter, D. The computational domain is of
dimension 120.5 along X and Z-directions
respectively. The dimension along Y-direction is
dependent on the tube separation in such a way
that the Y-dimension equals the span-wise tube
separation for staggered arrangement of tubes.
For staggered tube arrangement, the centers of an
upstream tube and the two respective downstream
tubes lie at the vertices of an equilateral triangle.
3. GRID GENERATION AND GEOMETRY
OF CELL ELEMENTS
The physical space of the computational domain
is sub-divided into a number of grid cells
quadrilaterals in two-dimensional, hexahedra in
three-dimensional. Several descriptions and

Fig. 1

Computational domain.

42

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

(a)

Fig. 2

(b)

Schematic of two-dimensional projection of grid-mesh in X-Y plane: (a) inline and (b) staggered arrangement
of tubes.

Vinokur (1983). Defining rij rj ri where ri

and rj are the position vectors of points i and j,


respectively, the volume of the cell can be the
computed using position vectors of vertices,
assuming that the cell corners are joined by linear
segments to form the six cell faces.
1



r71 r31 r21 r21 r61 r41 r31
6



r81 r41 r51 r81 r61 r51

3.1

Fig. 3

(1)

Grid independence study

Grid-independence has been confirmed for three


different mesh-sizes of 1735919, 1836321
and 1936723. Variation of span-averaged
Nusselt number along the length of the channel
for both inline and staggered arrangement of
tubes has been compared in Figs. 4(a) and 4(b) for
the three grid sizes. Since the maximum
difference is well within 3% of the mean Nusselt
number, the grid used for all computations that
follow is chosen to be 1836321.

Representation of a typical 3-D computational


cell.

The primary two-dimensional grid in X-Y plane


is shown in Figs. 2(a) and 2(b) for inline and
staggered tube arrangements. It has skewed and
rectangular shaped elements with four-nodal
corner points near the region of circular tubes and
almost regular rectangular-shape type element in
the remaining region. This 2-D grid is uniformly
stacked along Z-direction to obtain the volume
grid for the entire 3-D flow domain. The 3-D grid
consists of deformed hexahedral type of elements
near the tube surfaces with eight-nodal corner
points. In the 3-D grid, the areas of faces of
control volumes (Fig. 3) having face surface
vector S are calculated. The computational nodes
at which the flow variable values are to be
calculated are placed at the geometric centre of
the control volume. The coordinates of cell and
face centers, the surface area vectors and cell
volume required for numerical evaluation of
discrete integrals in the discretized governing
equations are computed and stored prior to flow
field solution.
The finite volume vertices are numbered 1 to 8 as
shown in Fig. 3. The outward surface normal and
volume are found as suggested by Kordulla and

4. NUMERICAL TECHNIQUE: FINITE


VOLUME METHOD
The general conservation equation in convectiondiffusion source integral form given in Eq. (2) is
discretized and solved for each computational cell
using finite volume technique.

V
n dS
d
S
t
grad n dS q d
S

(2)

where is the general transport variable for all


the governing equations. By setting equal to 1, u,
v, w and T and selecting appropriate values for the
diffusion coefficient and source terms, the
equations for mass, momentum and energy
conservation are obtained. In Eq. (2) the first term
43

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

d V n dS

t P
S

on the left hand side signifies the rate of change


of the total accumulation of transport property

in the control volume. The product V n


expresses the flux component of the property
due to the fluid flow along the outward face
normal vector n . Therefore, the net rate of
change of fluid property for the fluid element
due to convection is indicated by the second term
on the left hand side of Eq. (2). The first term on
the right hand side represents the net rate of
increase of fluid property of the fluid element
due to diffusion and the last term gives the rate of
increase of property due to sources present
inside the fluid element. The solution domain is
subdivided into a finite number of small control
volumes (CVs) as described in the previous
section on grid generation. Collocated grid
scheme has been employed, where all the
transport variables (conservative variables: , u,
v, w and dependent variable: T) are assigned
values at cell center. The convective fluxes at cell
faces of all CVs are suitably interpolated.
4.1

grad n dS

q d

The surface integrals are approximated using


quadrature formulae. The net flux through the CV
surface S is the sum of integrals over the six CV
faces as given below.

cd

dS

f e , w,n , s ,t ,b S f

cd

dS

(4)

where Fcd is the convective flux vector ( V )


or diffusive flux vector (grad ) at the CV face,
f. While computing the surface integrals in the
Eq. (4), only the nodal values of are known, but
integrand Fcd is required everywhere on the
surface, S f . An approximation of mid-point rule
is used in which the mean value over the surface
is used as the cell face center value.

d
S

S
cd , f
f
cd

Discretization technique

(5)

Sf

The integral conservation equation in Eq. (2)


applies to each of the CV cells as given in Eq. (3).
Since the surface integrals over inner CV faces
cancel out, the sum of the equations for all CVs
provides the global conservation equation. Eq. (2)
is converted into set of algebraic equations for
values at centroids of the CVs. The centroid value
p and a number of neighboring nodal values ( E ,
W , etc.) are involved in the algebraic equation of
each CV. For a typical CV, Eq. (2) may be
expressed as

The transient and the source terms in the transport


equations require integration over the volume of
the CV, for which the second-order
approximation is used to replace the volume
integral by the product of the mean value of the
integrand and the volume, where the mean values
are approximated by the values at the CV center.

Q d Q

QP P

Here Q p is the value of Q at CV center.

Fig. 4

(3)

Grid independence for variation of span-averaged Nusselt number: (a) inline arrangement of tubes and
(b) staggered arrangement of tubes.
44

(6)

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

4.2

The equations of motion and energy equation


require appropriate boundary conditions for their
solution. The boundary conditions for the given
computational domain are summarized in Table 1.
Boundary conditions of isothermal wall type are
used for top and bottom channel walls and tube
surfaces. Symmetry boundary conditions are
applied for side walls. Fully developed parabolic
velocity profile at the channel inlet and Orlanski
(1976) type of convective boundary condition at
the exit are used. This approach at the exit
boundary condition may be eliminated by
computing the flow in the whole bundle which
automatically includes the developing flow in the
first few rows of the bundle as reported by Wang,
Penner and Ormiston (2000). At the initial instant
of time (t = to), the values of all the dependent
variables are specified.

Solution method

Applying the above approximations of surface


and volume integrals, the discretized form of the
momentum equation has been solved along with
the discretized form of the continuity equation for
each finite volume cell. QUICK (Quadratic
Upwind Interpolation for Convective Kinematics)
scheme of Leonard (1979) is used to find the face
value f , which involves one downstream and
two upstream neighboring cells. Even though
QUICK offers lower order accuracy for nonorthogonal grid arrangements, the present grid is
made nearly orthogonal except the four diagonal
cells around the tube surface. SIMPLE (Semi
Implicit Method for Pressure Linked Equations)
method of Patankar and Spalding (1972) has been
employed in the flow solver. The diffusion flux
consists of two distinct parts, viz normal
derivative diffusion and cross-derivative diffusion.
The normal derivative diffusion flux has been
treated implicitly and is coupled with the implicit
part of the convective flux to calculate the main
coefficients of the discretized equations while the
cross-derivative diffusion flux is treated explicitly
to avoid any possibility of producing negative
coefficients in an implicit treatment. This term
together with explicit part of convective flux is
added to the source term. The diffusion flux of
variable f through the cell faces are evaluated by
considering contributions from normal-derivative
diffusion alone. Since the grid employed has been
maintained close to orthogonal, the contributions
appearing from cross-derivative diffusion fluxes
are expected to be quite small and have been
ignored.

5. RESULTS AND DISCUSSION


Numerical computations are carried out for flow
past circular tubes having inline and staggered
arrangements representing a unit of a tube-fin heat
exchanger. Tube separation has been varied for
both inline and staggered arrangement of tubes
and the results are presented in the form of
streamline plots, temperature contours, spanaveraged Nusselt number and span-and-heightaveraged pressure variation. Instead of the flow
fields over the domain, flow pattern in the
stagnant wake zone has been presented for
different cases of tube separation. The tube
separation is varied simultaneously with spanwise
blockage for inline arrangement of tubes, whereas
for staggered arrangement of tubes, the centers of
tubes are located in such a way that lines joining
the centers always form an equilateral triangle.

Table 1 Summary of boundary conditions.


p
0 and T Tw
z
p
T
0, and
0
y
y

Top and bottom walls; u v w 0 ,

Side walls; u w 0,

Inlet to the channel;

v 0,

p
2
u 3 2 U 1 2 z H , v w 0,
0 and T T

x
Channel exit U av 0 , (Orlanski, 1976)
t
x
Tube surfaces;

u v w 0,

p
0 and T Tw ;
n

where n denotes normal to the tube surface

45

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 5

Instantaneous streamlines in the stagnant wake of the tubes for different values of tube separation:
(a) upstream tube, (b) downstream tube of inline arrangement and (c) downstream half-tubes of staggered
arrangement.

arrangement is observed to become smaller in


size. This is considered to be a favorable flow
behavior, because smaller size of the stagnant
wake betters the heat transfer. In the case of
staggered arrangement of tubes, when the tubes
are located far apart, it is observed that the
variation in tube separation has less effect on the
characteristic nature of flow as seen from
Fig. 5(c). On the other hand, when the tubes are
closely placed, even a small variation in tube
separation
influences
the
flow
pattern
significantly.
Figure 6(a) shows the instantaneous temperature
distribution in the wake of the upstream tube.
Figures 6(b) and 6(c) show the respective
temperature contours in the wake of the
downstream tube for inline and staggered
arrangement of tubes. It is observed from Fig. 6(a)
that with a decrease in tube spacing the

The actual flow behavior can be visualized with


the help of streamlines plots. Figure 5(a) presents
streamline plots corresponding to instantaneous
velocity field in the stagnant wake of the
upstream tube for different value of tube
separation. It is observed that the variation in the
wake characteristics of the upstream tube occurs
mainly due to the variation in transverse blockage
(which is varied simultaneously with tube
separation) rather than due to variation in tube
separation. Due to the variation of longitudinal
separations, the streamlines near the wake of the
downstream tube get affected for both inline and
staggered arrangement of tubes. Streamline plots
in the near wake of the downstream tubes are
presented in Figs. 5(b) and 5(c) for inline and
staggered arrangement of tubes respectively. As
the tube spacing decreases, the wake region
behind the downstream tube of the inline

46

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 6

Instantaneous temperature distribution in the stagnant wake of the tubes for different tube separations:
(a) upstream tube, (b) downstream tube of inline arrangement and (c) downstream half-tubes of staggered
arrangement.

Temperature distributions on tube surfaces are


studied for tubes in inline and staggered
arrangements and presented in Fig. 7, in the form
of temperature contours. The angular positions
shown in the figures are with respect to the line
joining center of respective tube and forward
stagnation point and it is measured in clockwise
sense. Figure 7(a) shows the temperature
distribution on the upstream tube surface for
different values of tube separations. In Fig. 7(a),
the comparison of temperature contours for tube
separations of 5D and 4D shows that with
decreasing separation, a small improvement in
heat transfer takes place near the rear stagnation
surface of the upstream tube. As the tube
separation
is
further
decreased,
more
improvement of heat transfer over the entire tube
surface is observed from Fig. 7(a). A similar heat
transfer enhancement over downstream tube for
inline arrangement is observed from Fig. 7(b). For
all values of tube spacing considered, the halftubes in staggered arrangement have surface

non-dimensional temperature in the wake also


decreases showing an enhancement in heat
transfer. The effect of tube spacing on the
upstream tube is found to be almost similar for
both inline and staggered arrangements of tubes.
A comparison of Figs. 6(a) and 6(b) indicates that
the heat removal from the wake of downstream
tube is less when the tube spacing is more. This is
due to the large gap between tubes where the fluid
carrying heat from the upstream tube is further
heated by the channel walls before reaching the
location of downstream tube. Moreover, the
disturbed flow due to the upstream tube gets
streamlined by the channel walls before it reaches
the downstream tube. On the other hand, for
closely placed tubes (with small tube separation),
the disturbances generated by the upstream tube
get further amplified by the downstream tube and
this flow in the presence of horseshoe vortices
enhances heat transfer near the wake of the
downstream tube.

47

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 7

Instantaneous temperature distribution on tube surfaces for different tube separations: (a) upstream tube,
(b) downstream tube of inline arrangement and (c) downstream half-tubes of staggered arrangement.

temperatures lower (more heat transfer) as


compared to inline downstream tubes. This higher
heat transfer is due to the accelerated flow near
the half-tubes. Moreover, with decrease in tube
separation, half-tubes show a better heat transfer
as observed from Fig. 7(c).
Friction factor is defined on the basis of an
equivalent shear force in the flow direction per
unit heat transfer (or friction) area. For the
channel walls and tube surfaces, this equivalent
shear force is a combination of viscous shear
(skin fiction) and pressure force (form drag).
Using this definition of the friction factor, a
common treatment for all surfaces is adopted
(Eq. 8) which is also followed by Roychowdhury,
Sarit and Sundararajan (2002).

dp 1

f D / V 2
dx 2

The friction factor increases along the tube


surface from the forward stagnation line and
attains a maximum at an angular location of /2
and become small on the rear surface of the tube
it is found to become small. For both the
arrangements of tubes, it is observed that the
friction factor on the tube surface increases with
decreasing tube separation as shown in
Figs. 8(a)(c). In the downstream location of
inline and staggered arrangements of tubes, a
comparison of friction factor values shows a
higher friction factor on the surface of half-tubes
in the staggered arrangement. This is due to the
accelerated flow in the region.
The local Nusselt number near channel walls is
defined as

(8)

Nu ( x , y )

48

1 b

z z 0, H

(9)

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

( x, y, z ) u( x, y, z ) dA
b

b ( x )

Ac

u( x, y, z ) dA

(11)

Ac

where Ac is the flow area in the cross-stream


plane.
The span-averaged Nusselt number near the
bottom wall at a particular cross-stream location
becomes
B

Nu sp ( x )

Fig. 8

In a similar way, the local Nusselt number on the


tube surfaces can be written as
1

1 b

r r R

(12)

where B is the width of the channel.


Figures 9(a) and (b) show the span-averaged
Nusselt number variation near the bottom channel
wall for different values of tube separation of
inline and staggered arrangement of tubes
respectively. For both the tube arrangements,
higher Nusselt number is observed near the tube
locations. The higher Nusselt number in the
location of tube is due to the formation of horseshoe vortices. This higher heat transfer regions
are observed to shift closer to each other along the
tube location as the tube separation is decreased
and it is observed to be highest for a tube
separation of 2D. An overall increase in Nusselt
number due to a decrease in tube separation is
clearly observed.
Figures 10(a) and 10(b) present the variation of
span-averaged pressure along length of the
channel for both the arrangements of tubes. It is
apparent from the results that for small tube
separation, the pressure drop increases due to
increased skin friction drag for both types of tube
arrangements.
Consequently,
the
overall
performance of a particular tube arrangement has
to be judged by considering the ratio of heat
transfer enhancement to increase in pressure drop.
For the present computational domain, the top and
bottom walls of the channel provide no-slip
boundary conditions whereas, the side walls are
symmetric with zero velocity gradients across
them. Therefore, not all fluid particles travel at
the same velocity within channel. For better
understanding and to visualize the physical
significance of the boundary conditions, velocity
profiles in cross-stream plane are presented in
Figs. 11(a) and 11(b) respectively, for inline and
staggered arrangement of tubes. The velocity of
the fluid in contact with the channel walls is
essentially zero and increases further away from
the wall. Therefore, the velocity profile for the
flow in the channel is studied at mid-horizontal
plane parallel to both the channel walls (at

Height-averaged friction factor along tube


surfaces for different tube separations:
(a) upstream tube, (b) downstream tube of
inline arrangement and (c) downstream halftubes of staggered arrangement.

Nu ( x , y )

1
Nu( x, y ) dy
B 0

(10)

Here r is the radial coordinate and is positive


along outward normal of the tube surface and R is
the radius of the tube. b is the bulk-mean
temperature of the fluid at a particular crossstream location and is defined as
49

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 9

Span-averaged Nusselt number variation along the flow direction: (a) inline arrangement of tubes and (b)
staggered arrangement of tubes.

Fig. 10 Span-averaged-normalized pressure variation along the flow direction for (a) inline arrangement of tubes
and (b) staggered arrangement of tubes.

depend upon whether the flow is laminar or


turbulent.
A normalized estimate of overall heat transfer
from the channel walls and tube surfaces can be
presented in terms of Colburn factor (j) defined as

Z = l3 / 2). The velocity profiles in Y direction for


different X locations show zero velocity gradients
at the symmetric side walls for inline arrangement
of tubes. Moreover, in the wake zone, the velocity
magnitude is found to be small and recovers at
distances away from the circular tube. Similar
locations are identified for staggered arrangement
of tubes and the velocity profiles are studied as
shown in Fig. 11(b). It is interesting to observe
the steep gradient in the wake of the downstream
half-tubes, even with symmetric boundary
conditions being imposed. Velocity profiles also

j Nu o / Re Pr 0.33

(13)

where the overall mean Nusselt number, Nu o for


all the heated surfaces, i.e. the top channel wall,
the bottom wall, and the tube surface is calculated
using Eq. (14) given below.

Nuo
Nu( x, y) dx dy Nu( x, y) dx dy Nu( , z) r d dz Nu( , z) r d dz

Atotal top
downstream
bottom
upstream

tube surface
wall
tube surface

wall
(14)
where A total = 2(A bottom wall + A tube surface).
50

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 11 Velocity profile at different X location for varying tubes spacing: (a) inline arrangement of tubes and
(b) staggered arrangement of tubes.

surface in the absence of confinement against the


numerical results of Chun and Boehm (1989) and
the experimental results of Eckert and Soehngen
(1952) for a Reynolds number of 200. Results
compare favorably well with numerical
computations of Chun and Boehm (1989).
However, there is small mismatch of results at
inlet and exit sections from the experimental
results of Eckert and Soehngen (1952) which may
be mainly due to no turbulence model being
employed in the present computations. The
considered Reynolds number of ReD = 200
appears to be sufficiently large and small scale
structures may exist in the flow field. The effect
of these structures can be captured provided that a
suitable turbulence model is employed in the
present computations. In addition, the values of
coefficient of friction (friction factor) on half-tube
surface obtained from the present computations
(Re = 600) are compared against those reported by

Figure 12 compares the performance of both


inline and staggered arrangement of tubes in
terms of j and f factors. From Fig. 12(a) it is
observed that the tube separation has less effect
on heat transfer for both inline arrangements of
tubes whereas, the friction factor increases at a
higher rate with decreasing tube separation as
shown in Fig. 12(b). The j factor which gives
estimate of heat transfer shows a negligible
increase in heat transfer with decreasing tube
separation. Therefore, tube separation less than
4.5D may be used, where heat transfer is of more
importance irrespective of pressure drop. An
optimum tube separation is found to be 4.5D as
observed from Fig. 12(c), which compares j/f
factors for different tube separation for both inline
and staggered arrangement of tubes.
Figure 13 presents the validation of the computed
results by comparing the variation of heightaveraged local Nusselt number on half tube
51

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 12 Performance variation with tube separation


for both inline and staggered arrangement of
tubes (a) j-factor, (b) f-factor and (c) j / f
factor.

Fig. 13 Local Nusselt number


distribution on the
cylinder half-surface
from forward stagnation
point towards rear
stagnation point
(Re =200).

Wang, Penner and Ormiston (2000) for staggered


arrangement of tubes and have been presented in
Fig. 14. However, the flow geometry of the twodimensional computational domain considered by
Wang, Penner and Ormiston (2000) is different
from the one used in the present computations.
They have considered equal values of transverse
and longitudinal pitch for the staggered
arrangement of tubes while in the present
computations, the tube centers lie on the vertices
of an equilateral triangle. Moreover, they have
reported their results for ReD = 54. Hence, the

comparison of the present results against their


results is made to confirm only the qualitative
nature of variation of the friction factor on the
tube surface and not with an expectation of close
matching of the numerical values. Variation of
friction factor on the tube surface reported by
Wang, Penner and Ormiston (2000) corresponds
to transverse and longitudinal pitch of 2D while
the results from the present computations consider
different values of tube separations. Thus, the
over-estimated value of friction factor from the
present study for sd = 2D may be attributed to the
52

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

Fig. 14 Coefficient of friction along


tube surface for downstream
half-tube of staggered
arrangement of tubes for various
separation distances.

Fig. 15 Comparing height-averaged Nusselt number variation along tube surface for inline arrangement of tubes:
(a) upstream tube and (b) downstream tube.

in the presence of heated channel walls is higher


for the three-dimensional study as compared with
the values reported by El-Shaboury and Ormiston
(2005) from their two-dimensional study.
However, the effect of increasing channel
confinement (decreasing height) is not directly
related to an increase in Nusselt number. In fact,
on the downstream tube surface, the Nusselt
number is much higher for H = 0.8D and further
increase in confinement causes reduction in local
height-averaged Nusselt number. Moreover, the
increase in Nusselt number near forward and rear
stagnation points is more than that in the
accelerated flow region. This may be attributed to
the change in flow structure due to the presence
of channel walls. Even though the effect of heated
channel walls on increase in height-averaged
Nusselt number seems apparently due to more
contributions in heat transfer due to the vicinity of
heated channel walls, the convective heat transfer
from the tube surfaces should also get influenced

presence of confining channels walls and


difference in the value of Reynolds number
considered.
The present three-dimensional computational
result have also been validated against the results
reported from two-dimensional numerical
investigations of El-Shaboury and Ormiston
(2005) on heat transfer in laminar, steady cross
flow. For the purpose of validation and to bring
out the effect of channel confinement, additional
computations have been carried out for three
different channel heights (l3 = 0.8, 0.6 and 0.4) for
Reynolds number, ReD = 120, even though all
other reported results correspond to a channel
height of l3 = 0.5 or H = 0.5D and ReD = 600. The
Nusselt number around the tube surfaces from the
present study and those reported by El-Shaboury
and Ormiston (2005) are presented in Figs. 15(a)
and 15(b). A comparison of the results shows that
the height-averaged Nusselt number on the
surfaces of both upstream and downstream tubes
53

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

by the momentum transport of the fluid. Hence,


there exists an optimum channel separation for
which the heat transfer from tube surfaces reaches
its maximum value. Any further decrease in
channel height reduces the height-averaged
Nusselt number due to viscous effect of closely
approaching wall boundary layers. For the
upstream tube, the trend of variation of Nusselt
number with decreasing channel height does not
seem to be similar to that for the downstream tube
(or may be for other downstream tubes, if they
had been present). This could be mainly because
the upstream tube is present in the thermally
developing region while for the other downstream
tubes the flow gets thermally more developed.

n
Pr
p

q
ReD

Reynolds number based on tube


diameter (= VD / )

position vector
r
surface vector
S
temperature
T
time
t
free stream velocity
V
axial velocity component
u
volume

span-wise velocity component


v
velocity component normal to channel
w
wall
X, Y, Z dimensionless rectangular coordinates
x, y, z
dimensional rectangular coordinates

6. CONCLUSIONS
A 3-D computational code to solve
incompressible form of flow and energy equations
using finite volume method has been presented.
Air flow past inline and staggered tube bundles
confined between channel walls has been
investigated as the practical application of a tubefin heat exchanger. Effect of tube separation for
both the arrangements has been investigated on
heat transfer and pressure drop characteristics.
The results show that overall heat transfer and in
particular, heat transfer from tube surfaces are
enhanced with decreasing tube separation for both
inline as well as staggered arrangement of tubes.
Moreover, with decreasing separation, the
staggered arrangement of tubes is relatively more
effective in heat transfer enhancement compared
to inline arrangement of tubes. The results also
show that the decrease in tube separation
increases the heat transfer, but at the cost of
increase in pressure drop. For the investigated
range of tube separation, 2D lsd 5D, lsd =4.5D is
associated with highest goodness factor (j/f) for
both the arrangements of tubes.

Greek symbols

difference/increment
diffusion coefficient
dynamic viscosity
kinematic viscosity
transport property (u, v, w or T )
non-dimensional temperature
[= (T T)/(Tw T)]
density
angle measured clockwise from forward
stagnation point with respect to tube
center

Subscripts

1
2
3
b
c1

NOMENCLATURE
A
CV
D

F
f
H
j
k
l
Nu
Nu

normal to surface
Prandtl number
pressure
span-averaged and height-averaged
pressure
volumetric source term

area
control volume
diameter of circular tube
flux vector
friction factor
channel height (= l3)
Colburn factor
thermal conductivity
length
Nusselt number
average Nusselt number

e
m
max
n
norm
o
s
sd
sp
w

54

span-wise direction
transverse direction
direction normal to channel wall
bulk
centre of upstream tube from the
channel inlet
east
mean
maximum
north
normalized
overall
south
separation distance
span-averaged
west

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

REFERENCES
1. Baker CJ (1991). Oscillations of horseshoe
vortex systems. ASME Journal of Fluids
Engineering 113(3):489495.
2. Bossel U, Honnold FV (1976). On the
formation of horse-shoe vortices in plate fin
heat exchangers. Archives of Mechanics
28:773780.
3. Briley WR (1971). A numerical study of
laminar separation bubbles using the NavierStokes equations. Journal of Fluid Mechanics
47:713736.
4. Brown SN, Stewartson K (1969). Laminar
separation. Annual Review of Fluid
Mechanics 1:4572.
5. Buyruk E (2002). Numerical study of heat
transfer characteristics on tandem cylinders,
inline and staggered tube banks in cross-flow
of air. International Communications in Heat
and Mass Transfer 29(3):355366.
6. Carter JE, Wornom SF (1975). A forward
marching procedure for separated boundarylayer flows. AIAA Journal 13:11011103.
7. Catherall D, Mangler KW (1966). The
integration of the two dimensional laminar
boundary-layer equations past the point of
vanishing skin friction. Journal of Fluid
Mechanics 26:163182.
8. Chang Keun-Shik, Song Chang-Joon (1990).
Interactive vortex shedding from a pair of
circular cylinders in a transverse arrangement.
International Journal for Numerical Methods
in Fluids 11(3):317329.
9. Chun W, Boehm RF (1989). Calculation of
forced flow and heat transfer around a
cylinder in cross flow. Numerical Heat
Transfer Part-A 15:101122.
10. Connell
SD,
Braaten
ME
(1995).
Semistructured mesh generation for threedimensional Navier-Stokes calculations.
AIAA Journal 33:10171024.
11. Darbandi M, Schneider GE (1998). Analogybased method for solving compressible and
incompressible
flows.
Journal
of
Thermophysics and Heat Transfer 12:239
247.
12. Darbandi M, Vakilipour S (2008).
Developing
implicit
pressure-weighted
upwinding scheme to calculate steady and
unsteady flows on unstructured grids.
International Journal for Numerical Methods
in Fluids 56:115141.
13. Dennis SCR, Chang G (1970). Numerical
solutions for steady flow past a circular

14.

15.

16.

17.

18.

19.

20.
21.

22.

23.
24.

25.

26.

55

cylinder at Reynolds number up to 100.


Journal of Fluid Mechanics 42:471489.
Derksen
RW
(1990).
Computing
axisymmetric, laminar internal flows.
International Journal for Numerical Methods
in Fluids 11(4):361377.
Eckert ERG, Soehngen E (1952). Distribution
of heat transfer coefficients around circular
cylinders in cross flow at Reynolds numbers
from 20500. Trans. ASME 74:343347.
El-Shaboury AMF, Ormiston SJ (2005).
Analysis of laminar forced convection of air
crossflow in in-line tube banks with
nonsquare arrangements. Numerical Heat
Transfer Part-A 48(2):99126.
Faghri M, Rao N (1987). Numerical
computation of flow and heat transfer in
finned and unfinned tube banks. International
Journal of Heat Mass Transfer 30(2):363
372.
Gordon William J, Thiel Linda C (1982).
Transfinite mappings and their application to
grid generation. Applied Mathematics and
Computation 1011:171233.
Griffiths E, Awbery J (1933). Heat transfer
between metal pipes and stream of air.
Proceedings of Institute of Mechanical
Engineering 125:319382.
Horton HP (1974). Separating laminar
boundary layers with prescribed wall shear.
AIAA Journal 12:17721774.
Jang Jiin-Yuh, Wu Mu-Cheng, Chang WenJeng (1996). Numerical and experimental
studies of three-dimensional plate-fin and
tube heat exchangers. International Journal
of Heat Mass Transfer 39(14):30573066.
Ju Lili, Du Qiang (2009). A finite volume
method on general surfaces and its error
estimates. Journal of Mathematical Analysis
and Applications 352:645668.
Kordulla W, Vinokur M (1983). Efficient
computation of volume in flow predictions.
AIAA Technical Notes 21(6):917918.
Lacor C, Smirnov S, Baelmans M (2004). A
finite volume formulation of compact central
schemes on arbitrary structured grids. Journal
of Computational Physics 198:535566.
Leonard BP (1979). A stable and accurate
convective modeling procedure based
on
quadratic
upstream
interpolation.
Computational
Methods
in
Applied
Mechanics Engineering 19:5998.
Nishimura T, Hisayoshi I, Hisashi M (1991).
Experimental validation of numerical analysis
of flow across tube banks for laminar flow.

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

27.

28.
29.

30.

31.

32.

33.

34.

35.

36.

37.

38. Schneider GE, Raw MJ (1987). Control


volume finite-element method for heat
transfer and fluid flow using collocated
variables. Numerical Heat Transfer Part-B
11(4):363390.
39. Smith RE (1982). Algebraic grid generation.
Applied Mathematics and Computation 10
11:137170.
40. Son JS, Hanratty TJ (1969). Numerical
solution for flow around a cylinder at
Reynolds numbers of 40, 200 and 500.
Journal of Fluid Mechanics 35:369386.
41. Tanida Y, Okajima A, Watanabe Y (1973).
Stability of circular cylinder oscillating in
uniform flow or in a wake. Journal of Fluid
Mechanics 61:769784.
42. Tatsutani K, Devarakonda R, Humphrey JAC
(1993). Unsteady flow and heat transfer for
cylinder pairs in a channel. International
Journal of Heat and Mass Transfer
13(36):33113328.
43. Thompson JF, Warsi ZUA, Mastin WC
(1982). Boundary-fitted coordinate systems
for numerical solutions partial differential
equations A review. Journal of
computational Physics 47:1108.
44. Thompson JF (1984). Grid generation
techniques in computational fluid dynamics.
AIAA Journal 22:15051523.
45. Tiwari S, Biswas G, Prasad PLN, Sudipta
Basu (2003). Numerical prediction of flow
and heat transfer in a rectangular channel with
a built-in circular tube. Journal of Heat
Transfer (ASME) 125:413421.
46. Tiwari S, Chakraborty D, Biswas G,
Panigrahi PK (2005). Numerical prediction of
flow and heat transfer in a channel in the
presence of a built-in circular tube with and
without
an
integral
wake
splitter.
International Journal of Heat and Mass
Transfer 48:439453.
47. Tiwari Shaligram, Pratish P Patil, Jayavel S,
Biswas G (2006). Flow field characteristics
near first transition for flow past a circular
tube confined in a narrow channel.
Proceedings of 33rd National & 3rd
International Conference on Fluid Mechanics
and Fluid Power. 79 December 2006, IIT
Bombay, India, Paper No. NCFMFP20061132.
48. Wang YQ, Penner LA, Ormiston SJ (2000).
Analysis of laminar forced convection of air
for crossflow in banks of staggered tubes.
Numerical Heat Transfer Part-A 38(8):819
845.

Journal of Chemical Engineering of Japan


24(5):666669.
Nishimura T, Hisayoshi I, Hisashi M (1993).
The influence of tube layout on flow and
mass transfer characteristics in tube banks in
the transitional flow regime. International
Journal of Heat and Mass Transfer
36(3):553563.
Orlanski I (1976). A simple boundary
condition for unbounded flows. Journal of
Computational Physics 21:251269.
Patankar SV, Spalding DBA (1972).
Calculation procedure for heat mass and
momentum transfer in three dimensional
parabolic flows. International Journal Heat
and Mass Transfer 1:17871805.
Paul SS, Ormiston SJ, Tachie MF (2008).
Experimental and numerical investigation of
turbulent cross-flow in a staggered tube
bundle. International Journal of Heat and
Fluid Flow 29:387414.
Pereira JMC, Kobayashi MH, Pereira JCF
(2001). A fourth-order-accurate finite volume
compact method for the incompressible
Navier-Stokes
solutions.
Journal
of
Computational Physics 167:217243.
Piller M, Stalio E (2008). Compact finite
volume schemes on boundary-fitted grids.
Journal of Computational Physics 227:4736
4762.
Pratish P Patil, Shaligram Tiwari (2009).
Numerical investigation of laminar unsteady
wakes behind two inline square cylinders
confined in a channel. Engineering
Application
of
Computational
Fluid
Mechanics 3:369386.
Raithby GD (1976a). A critical evaluation of
upstream differencing applied to problems
involving fluid flow. Computer Methods in
Applied Mechanics and Engineering 9:75
103.
Raithby GD (1976b). Skew upstream
differencing schemes for problems involving
fluid flow. Computer Methods in Applied
Mechanics and Engineering 9:153164.
Roychowdhury Ghosh D, Sarit Kumar Das,
Sundararajan T (2002). Numerical simulation
of laminar flow and heat transfer over banks
of staggered cylinders. International Journal
of Numerical Methods in Fluids 39:2340.
Schneider GE, Raw MJ (1986). A skewed,
positive influence coefficient upwinding
procedure for control-volume-based finiteelement convection-diffusion computation.
Numerical Heat Transfer Part-B 9(1):126.

56

Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 1 (2010)

49. Warsi ZUA, Ziebarth JP (1982). Numerical


generation of three-dimensional coordinates
between bodies of arbitrary shapes. Applied
Mathematics and Computation 1011:717
728.
50. Wilson Safwat A, Khalil Bassiouny M (2000).
Modeling of heat transfer for flow across tube
banks. Chemical Engineering and Processing
39:114.

57

Anda mungkin juga menyukai