Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

This article was downloaded by: [Tel Aviv University]

On: 09 March 2014, At: 11:51


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK
Numerical Heat Transfer, Part B: Fundamentals:
An International Journal of Computation and
Methodology
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/unhb20
OPEN BOUNDARY CONDITIONS FOR THE
STREAMFUNCTION -VORTICITY FORMULATION OF
UNSTEADY LAMINAR CONVECTION
G. Comini
a
, M. Manzan
a
& G. Cortella
a
a
Departmento di Energetica e Macchine , Universit degli Studi di Udine , Via delle
Scienze 208, Udine, 33100, Italy
Published online: 17 Apr 2007.
To cite this article: G. Comini , M. Manzan & G. Cortella (1997) OPEN BOUNDARY CONDITIONS FOR THE STREAMFUNCTION
-VORTICITY FORMULATION OF UNSTEADY LAMINAR CONVECTION, Numerical Heat Transfer, Part B: Fundamentals: An
International Journal of Computation and Methodology, 31:2, 217-234, DOI: 10.1080/10407799708915106
To link to this article: http://dx.doi.org/10.1080/10407799708915106
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
OPEN BOUNDARY CONDITIONS FOR THE
STREAMFUNCTION -VORTICITY FORMULATION
OF UNSTEADY LAMINAR CONVECTION
G. Comini, M. Manzan, and G. CorteUa
Unioersita degli Studi di Udine, Departmento di Energetica e Macchine,
Vw delle Scienze 208, 33100 Udine, Italy
The streamj'wu:tion-vortiJ:ily formulation is used to tmIllyze unsteady laminar-conoeaion
problems in two-dimensional incompressible flows. The Bubnov-Gakrlcin jinile..lemetu
method and a sequential procedure are employed to discretize and sollJe the governing
differential eqUllliDns. Very accurate resulss are obtained by employing "advedive
tkrivaJive condmons" at the ouJjlowfor 00 thevariables invohJed. The boundJuy conditions
for the streamfunction at internal woHs are imposed during the assembly pTOC1!SS, and the
vortiJ:ilyat inflow and woOboundaries is eva1UlJled in theframework ofthe streamfunction
equation. The accuracy of the approach is tkmonstrated by the solution of two weU-known
benchmark problems concerningforced convedion over a circular <yUnlkr in cross flow and
mixed convection in a plane channel heatedfrom below.
INTRODUCTION
Streamfunction-vorticity (1/1, w) formulations of two-dimensional incompress-
ible flows allow the elimination of pressure from the problem and automatically
satisfy the continuity constraint. On the other hand, the specification of boundary
conditions for the (1/1, w) equations is rather critical and, if not done properly, may
lead to serious difficulties in obtaining a converged solution [1-9]. A key question
is the treatment of open boundaries which, in the framework of velocity-pressure
formulations, has been discussed extensively at the Open Boundary Condition
Minisymposium [10], and has also been the subject of a letter to the editor of this
journal [11].
In a recent study, we have investigated the influence of different specifica-
tions of the boundary conditions at inflow and outflow, in the framework of the
streamfunction-vorticity formulation for stationary flows [12]. In this article, we
extend the analysis to transient problems, such as flows over a circular cylinder [13]
and plane channel flows heated from below [14]. Since we have already presented
some preliminary results concerning these flows at two recent conferences [15, 16],
here we concentrate mainly on the complete unified formulation of unsteady
laminar-convection problems. In particular, we advocate the use of the so-called
advective derivative conditions at outflow, and the imposition of proper boundary
conditions for the streamfunction at internal walls in multiply connected domains.
Received 17 May 1996; accepted 26 July 1996.
This research was supported by Ministero dell' Universite della Ricerca Scientifica e Technologica.
Address correspondence to Professor Gianni Comini, Dipartmento di Energetica e Macchine,
Universitadegli Studi di Udine, Via delle Scienze 208,33100 Udine, Italy. E-mail: gianni.comini@uniud.it
Numerical Heat Transfer, Part B, 31:217-234,1997
Copyright 1997 Taylor & Francis
1048-7790/97 $12.00 + .00 217
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

218 G. COMINl ET AL
NOMENCLATURE
a thermal diffusivity (= klpCp) T dimensionless temperature
A area projected perpendicular to the I vector of nodal values of temperature
free-stream velocity u,w velocity components in the
A advection matrix [Eq. (22) (x, z) directions
b vector of buoyancy contributions U,W velocity components in the (X, Z)
[Eq. (23)] directions, dimensionless
B body force per unit volume V volume
c
p
specific heat at constant pressure x,z Cartesian coordinates
C
x
drag coefficient, dimensionless X,Z Cartesian coordinates, dimensionless
C
z
lift coefficient, dimensionless
/3
coefficient of thermal expansion
D diameter it time
p
force 6 time, dimensionless (= it uI L)
r vector of tangential velocity contributions 8 period, dimensionless
[Eq. (24)] A wavelength, dimensionless
Fr Froude number
I"
dynamic viscosity
[= Re
2/Gr
= u
2/g
/3(l
h
- le)L)
P
density
s
modulus of the gravity vector (T normal component of the stress vector,
g gravity vector dimensionless
Gr Grashof number[ = p2g /3(l
h
- le)L) I 1"'] r tangential component of the stress vector,
H height of a channel dimensionless
k thermal conductivity
'"
streamfunction
K Laplacian matrix [Eq. (21)]
'"
vector of nodal values of the
K
o
Laplacian matrix referred to the outflow streamfunction
boundary [Eq. (27) 'I' streamfunction, dimensionless (= "'luL)
L reference length w vorticity
M mass matrix [Eq. (20) ... vector of nodal values of vorticity
M
o
mass matrix referred to the outflow n vorticity, dimensionless (= wLjU)
boundary [Eq. (25)
n number of nodes Subscripts

shape and weighting functions related to
the node i c cold
Nu Nusselt number, equal to the h hot
dimensionless temperature gradient at I inflow
the surface i, j indices or node numbers
Nu space-averaged Nusselt number n normal component, positive when
(Nul time-averaged Nusselt number outward oriented
(Nul space- and time-averaged Nusselt number 0 outflow
p pressure p prescribed
p
pressure, dimensionless [ = pI( pu
2)]
s tangential component, or at the local
Po
outflow boundary vector surface coordinate s
[Eqs. (26) and (28)] S synunetry
Pe Peclet number (= Re Pr = uLla) X,Z in the (X, Z) directions
Pr Prandtl number (= c
p
I"lk) 0 reference value
Re Reynolds number (= puLI 1")
s linear coordinate in the direction of s Superscripts
s unit vector tangent to a boundary surface
S surface
I
m at the time step m
St Strouhal number C1/8)
-
average value
.
expended matrix or vector, I temperature
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARY CONDmONS FOR UNSTEADY C0NVECI10N 219
After having obtained a temperature solution, we compute local and average
Nusselt numbers from the resulting nodal rates of heat transfer. Similarly, from the
streamfunction and vorticity solutions we evaluate the normal and tangential
components of the stress vector at wall boundaries, in order to compute such
derived quantities as drag, lift, and pressure distributions.
In the finite-element discretization we employ the Bubnov-Galerkin method;
i.e., we use the same functions for interpolating the variables and for weighting the
residuals. In this way we do not rely on the effects of numerical diffusion
introduced by upwinding techniques related to Petrov-Galerkin methods. Finally,
we demonstrate the accuracy of the approach by the comparisons with well-estab-
lished literature results.
STATEMENT OF THE PROBLEM
For a constant-property Boussinesq fluid, when B, = - pg is the only contri-
bution to the body force per unit volume, in the absence of volumetric heating and
neglecting the effects of viscous dissipation, the streamfunction, vorticity, and
energy equations can be written in dimensionless form as [1]
(1)
(2)
and
In the above equations, the space coordinates, the velocities, and the time are
normalized with respect to a reference length L, to the average flow velocity ii,
and to a characteristic time 11
0
= L Iii, respectively, while the dimensionless
variable T = (r - t)I(t
h
- t
c
) is referred to suitably defined "hot" and "cold"
temperatures.
In the problems considered here, we can have inflowboundaries, internal and
external wall boundaries, symmetry boundaries, and outflow boundaries. At inflow
boundaries, external walI boundaries, and symmetry boundaries, we assume that
the streamfunction is known and its distribution can be prescribed as a boundary
condition of the first kind:
(4)
At internal wall boundaries belonging to multiply connected domains, the stream-
function does not change with. the space coordinate, but has an unknown value
which can be time-dependent and must be determined as part of the solution
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

220 G. COMINI ET AI..
process. Thus, for the time being, we will express this condition as
'If = const
while, in a later section, we will show how to enforce Eq. (5) during the assembly
process.
At inflow, wall, and symmetry boundaries, the normal derivative of the
streamfunction is also known. At inflow and wall boundaries we have
a'If
u = -- =0
s an
while at symmetry boundaries we have
a'If
u = -- * 0
s an
(6)
In a sequential procedure, we cannot use the additional conditions (6) and (7) for
the solution of the streamfunction equation because, in this way, we would
overspecify the problem. However, the information on the normal derivative of the
streamfunction can be incorporated in the boundary condition for the vorticity [1,
9]. Thus at inflow, wall, and symmetry boundaries, we can obtain the values of the
vorticity from a solution of Eq. (1) where conditions (6) and (7) are taken into
account. At this point, we must remark that, by using Eq. (6) also at inflow
boundaries (instead of specifying the inlet vorticity directly, as we did in [1] and [9]),
we have followed the mathematically elegant approach suggested by Gresho [17].
Numerical experiments have shown that, in this way, we obtain slightly more
accurate results at the expense of a decrease in the stability limits for the time
integration algorithms [15, 16].
To complete the specification of inflow, wall, and symmetry boundary condi-
tions, we must deal with the temperature variable. However, this presents no
difficulty, since at inflow and wall boundaries we have
while at symmetry boundaries we can confidently assume that
aT
-=0
an
(8)
(9)
In transient problems, we suggest using the outflow boundary conditions
identified as "advective derivative conditions." By using the average normal veloc-
ity lJ" as an estimate of the constant average phase speed [18], these conditions can
be written directly as
an _ an
-+u-=o
ao n an
(10)
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARYCONDmONS FOR UNSTEADY CONVECTION 221
and
aT _ aT
-+u-=o
a(J n an
(11)
for the vorticity and the temperature, respectively. The advective derivative condi-
tion for the streamfunction can be obtained from the corresponding advective
condition for the tangential component of the velocity,
au. - au.
-+u-=o
a(J n an
(12)
Taking into account the definition U. = -(a'ltjan) and the auxiliary expression
(13)
derived from Eq. (1), we can first write Eq. (12) as a second-order differential
equation,
a (a'lt) _(a
2'1t
)
- - =U --+0
a(J an n as
2
(14)
defined on the outflow boundary and subjected to the conditions of 'It = const at
the two endpoints. Then we can discretize Eq. (14) with respect to the time
variable, arriving at the desired final expression of the outflow boundary condition
for the streamfunction,
(
a'lt )m+! _ (a2'1t )m
- :;; U 1i(J -- + 0 +
an n as
2
(15)
In stationary problems, Eqs. (10) and (11) yield again the usual zero normal
derivative conditions, while Eq. (14) leads to the expression
(16)
which represents the "least constraining" condition for the streamfunction [19].
At this point it must be remarked that, in the context of integrated solution
procedures, a set of alternative downstream boundary conditions has been derived
in [20, 21] as the streamfunction-vorticity equivalent of the traction-free conditions
in primitive variables [21, 22]. However, these alternative conditions are not easily
implemented and cannot be employed in completely sequential procedures of
solution.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

2ZZ G. COMINI ET AL
FINITE-ELEMENT FORMULATION
Governing differential equations (1)-(3), with boundary conditions (4)-(11)
and (15), are first discretized in space using a Bubnov-Galerkin procedure. Then, a
Crank-Nicolson time stepping scheme is applied to the partially discretized ver-
sions of Eqs. (1)-(3), subjected to the boundary conditions (10) and (11). In matrix
form, the final algebraic equations are [I, 9, 12, 15, 16]
for the streamfunction,
Kt/I = Mw - Po - r
(17)

(M +.2-M
o
) = (_1 K+A)]wm+ 1
se u; Re 2 Re
= +.2-M )- A)]W
m
+ (18)
so u; Re 0 2 Re Fr
for the vorticity, and

(M +.2- M
o
) + K+A)]t
m
+
1
/i.O u; Pe 2 Pe
= [_1 (M +.2-M
o
) - +A)]t
m
(19)
s u; Pe 2 Pe
for the temperature. In the above equations t/I, w, t, r, b, and Po are vectors of
length n (the number of node points), while M, K, A, and M
o
are (n x n) matrices.
The mass matrices M, the Laplacian matrix K, the advection matrix A, the
body force vector b, and the load vector r are evaluated from the usual definitions:
Mj j = 1
v
1
(
aN; aN;
K .. = --+-- dV
'I v ax ax az az
(20)
(21)
1
(
aN aN) 1 [(a'l') aN (a'l') aN] A .. = N U _I + V _I dV = N _ _I - - -r-_I dV (22)
II v I ax aY v I az ax ax oz
b. = ( N( aT) dV (23)
I lv I ax
and
(24)
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARY CONDmONS FOR UNSTEADY CONVEcrlON 223
where N ; , ~ are shape/weighting functions, with i, j = 1, n. The derivatives
a'I'/ ax, a'I'/ az, and aT/aX are computed at the integration points using the
most recent values of '" and t. The expanded matrix M
o
and the expanded vector
Po have nonzero entries only on the outflow boundary. The nonzero entries in M
o
correspond to the entries in the matrix M
o
' evaluated from the definition
(25)
with i, j = 1, no' The nonzero entries in Po correspond to the entries in the vector
Po' and Po is computed, at each new time step m + 1, from the discretized version
of Eq. (15),
(26)
using the null vector p ~ = 0 as a starting value at tm = 1). The additional
definitions
and
f.
aN aN
(K ) .. = -' _J dS
o IJ So as as
a'I'
(P)i = - f. N;-dS
So an
(27)
(28)
hold good for the outflow boundary calculations, again with i,j = 1, no' It must be
pointed out, however, that we do not form Eq. (28) explicitly, since the vector Po is
evaluated directly from Eq. (26).
At this point, we must still impose the condition (5) in Eq. (17) on the nodes
that belong to internal wall boundaries in multiply connected domains. This can be
done by following an assembly procedure which can be better explained by
referring to a practical situation, such as the one illustrated in Figure 1. Let us
assume, for example, that we look for a solution of Eq. (17) where the streamfunc-
tion values at nodes 3, 5, and 7 must coincide with the streamfunction value at .
4
Figure 1. Internal boundary 1357 sur-
rounded by elements (1), (2), (3), and (4).
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

224 G. COMINI ET AI..
Table I. Connectivity lists for the internal boundary represented in Figure 1
Element Node Numbers Connections
1 1 2 4 3 Real
1 1 2 4 1 Auxiliary
2 3 4 6 5 Real
2 1 4 6 1 Auxiliary
3 5 6 8 7 Real
3 1 6 8 1 Auxiliary
4 7 8 2 1 Real
4 1 8 2 1 Auxiliary
node 1. Then, in addition to the "real" connectivity lists for the element nodes, we
can define "auxiliary" lists, where the "master" node 1 substitutes the "slave"
nodes 3, 5, and 7. These lists are reported in Table 1 for the elements (1)-(4) that
surround the internal wall boundary 1357. Obviously, we use the real lists for
computing the element contributions to the global matrices, while we refer to the
auxiliary lists in the assembly process. In this way, the algebraic equations corre-
sponding to the slave nodes are effectively eliminated from the global system of
algebraic equations (17). The only penalty is the bandwidth of the system increases,
but this can be a minor difficulty if one uses, as we do, an iterative solver.
Alternative methods for dealing with multiply connected domains, in the context of
streamfunction-vorticity formulations, are illustrated in [4], but we believe that the
procedure suggested here has the advantage of simplicity.
Equations (17)-(19) and (26) can be decoupled and solved in sequence,
provided that special care is taken to represent properly the vorticity. boundary
conditions at inflow and wall boundaries. The calculation procedure, to advance
from the step m ;;. 0 to the step m + 1, can be described as follows.
Once we have determined the contributions of the outflow boundary condi-
tion from Eq. (26), we can impose the boundary conditions (4) and (5) at the
appropriate nodes. Therefore, we can eliminate from the system (17) the nodal
equations that correspond to "slave" nodes or to points where the streamfunction
is known. Afterward, the new values of the streamfunction at the remaining nodes
can be computed from the reduced system (17) as
(29)
where w
m
is the vorticity solution at the previous step, while pm+ 1 has been
obtained from the solution of Eq. (26). Once Eq. (29) has been solved, we can go
back to the original equation (17) and, using the most recent values of the
streamfunction '"m+ 1, we can calculate the vorticity at inflow boundaries, wall
boundaries, and symmetry boundaries as
(30)
Actually, in Eq. (30), we do not have to consider a complete new solution, since the
computations involve only a strip of elements adjacent to the boundaries consid-
ered. Finally, having obtained the boundary values of vorticity from Eq. (30), we
can solve in sequence Eqs. (18) and Eq. (19), advancing to the next time step.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARYCONDmONS FOR UNSTEADY CONVECTION 225
CALCULATION OF STRESS COMPONENTS
In the solution of fluid flowproblems, a very important issue is the evaluation
or pressure distributions along internal or external wall boundaries. To this
purpose, in the streamfunction-vorticity formulation, we can utilize the projection
of the momentum equation in the direction s tangential to the wall boundary [221,
au. 1 a(u.
2
+ U}) ap 1 an 1 g
- + - - un + - + - - + - T- s = 0 (31)
ao 2 as n as Re an Fr g
At no-slip walls we assume U; = U. = 0 and thus, by integrating Eq. (31) from a
reference point So to the point of interest s, we obtain
,( 1 an 1 g )
P - Po = - f - - + - T- . s tis
'0 Re an Fr s
(32)
where the unit gravity vector is defined as gig = (0, -1). After having computed
the pressure distribution from Eq. (32) we can calculate, if required, the normal
and tangential components of the stress vector, using the dimensionless expressions
u= -P
and
1
r > --n
Re
which are normalized with respect to pu
2

SOLUTION OF BENCHMARK PROBLEMS


(34)
The examples presented here concern transient solutions of two well-known
laminar benchmark problems: forced convection over a circular cylinder in cross
flow and mixed convection in a plane channel heated from below. In the computa-
tions, we have used eight-node isoparametric elements. The space-discretization
errors and the sensitivity of the computed fields to the location of the outflow
boundary have been discussed in [IS, 161. At each time step, the discretized
equations have been iterated in sequence until convergence has been reached,
employing underrelaxation both for the vorticity and the energy equation. The
calculations were continued in time until the spatially averaged Nusselt numbers
differed less than 0.1% at two successive extremum points.
The space-averaged and the time-averaged Nusselt numbers have been
defined as
- If
Nu = S s Nus dS
1 f92
(Nus) = 8 Nus ae
O
2
- I 9,
(35)
(36)
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

226
and
G. COMINI ET AL.
- 1 f62 -
(Nu) = Nud8
8
2
- 8
1
6,
(37)
In the above equations, the time interval (8
2
- 8
1
) is large compared to the period
of oscillations, and the local Nusselt numbers Nu, are computed directly from the
dimensionless temperature gradients, that is, from the "reactions" (1).
Forced Convection over a Circular Cylinder in Cross Flow
In the present test problem we have followed [13] by taking Re = 100 and an
average normal velocity equal to the total inlet velocity. We have also assumed a
value of the Prandtl number Pr = 0.71, and we have disregarded the natural
convection contributions by taking llFr = O. The boundary conditions and the
finite-element mesh utilized are shown in Figure 2, where the flow domain has
been divided into 952 eight-node, isoparametric elements, yielding a total of 3,002
nodes.
,,-0
1-0
d"
--I .-0
(a)
dl
--0
iI.
(b)
1.(illf).
iIIJ iI.
(8.8) 12$.8)
(e)
(25.-8)
Figure 2. Forced convection over a circu-
lar cylinder in cross flow: (a) boundary
conditions for the flow problem; (b)
boundary conditions for the thermal
problem; (e) finite-element mesh.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARY CONDmONS FOR UNSTEADY CONVECfION 227
The basic characters of the flow are illustrated in Figures 3 and 4, where we
show the periodic evolution of streamlines and temperature contours starting from
a reference time (J = 0, which corresponds to the instant when a minimum is
reached in the temperature at point (X, Z) = (4,0). The qualitative agreement
between our streamline plots and those of [13] is very good. Moreover, these
figures demonstrate that we have very little reflection and distortion at the outflow,
thanks to the advective derivative conditions. We can also note that, by allowing
the value of the streamfunction at the cylinder wall to be unknown, we have arrived
at a solution that changes periodically with time in an interval of amplitude 0.0256
around 'It = 8. Since, as pointed out in [4], solutions obtained by assuming 'It = 8
compare well with the benchmark solution, we can infer that the agreement
between out results and such a resonable assumption implicitly validates our
procedure.
By taking into account the projections of the stress vector components in the
direction of the main flow, we have evaluated the drag coefficient,
and the lift coefficient,
F
z
C = -----,,--
z A pu
2
j 2
(38)
(39)
Ell
Ell
I!!I
EiII
Figure 3. Forced convection over a circu-
lar cylinder in cross flow at Re = l()():
periodic evolution of streamlines at time
intervals of tEl.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

228 G. COMINI ET AI..
Figure 4. Forced convection over a circu-
lar cylinder in cross flow at Re = 100:
periodic evolution of temperature con-
tours at time intervals of teo
where A is the cylinder frontal area. The time histories of these coefficients,
represented in Figure 5, start from the previously defined reference time and are
found again to be in good agreement with the benchmark results [13]. From the
periodic evolution of the lift coefficient, we estimate the Strouhal number, defined
as the inverse of the dimensionless period, to be St = 1.70. This value is in close
agreement with the best estimate (St = 1.73) obtained in (13) using a 14,000-node
mesh.
The variation of the local, time-averaged Nusselt number with the angular
coordinate t/> is shown in Figure 6. From the periodic evolution of the space-aver-
aged Nusselt number (not r ~ r t here), we estimate the time- and space-aver-
aged Nusselt number to be (Nu) = 5.14. The empirical Hilpert correlation
(Nu) = 0.683 ReO.
466
Pr
1
/
3
(40)
suggested in (23), yields (Nu) = 5.21 for Re = 100 and Pr = 0.71. This kind of
agreement must be considered good, since experimental results are undoubtedly
influenced by the finite length of the cylinder.
Mixed Convection in a Plane Channel Heated from Below
In the present test problem we have followed (5) by assuming that Re = 10,
Pe = 20/3, and Fr = 1/150. The boundary conditions and the finite-element mesh
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARY CONDmONS FOR UNSTEADY C0NVECI10N 229
, f\
,
1\
' ,
,-
I-
t- -
f-
V V V V V
, ,
I I I I I
1.342
1.334
o 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
8
(a)
1.338
1.350
0.3 r--r-.--,--,---,-,---,,-,
Cz
02
0.1
0.0
-0.1
-0.2
02 0.4 0.6 0.8 1.0 12 1.4 1.6
8
(b)
Figure 5. Forced convection over a circu-
lar cylinder in cross flow at Re = 100:
periodic evolution of drag (a) and lift
(b) coefficients.
Figure 6. Forced convection over a circu-
lar cylinder in cross flow at Re = 100:
circumferential variation of the local
time-averaged Nusselt number.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

230 G. COMINI ET AI..
utilized are shown in Figure 7, where the flow domain has been divided into 8 X 20
eight-node isoparametric elements, yielding a total of 537 nodes.
The basic characters of the flow after are illustrated in Figures 8 and 9, where
we represent the periodic behavior of streamlines and temperature contours
starting from a reference time 8 = 0, which corresponds to the instant when a
minimum is reached in the temperature at point (X, Z) = (5,0.5). As can be seen,
the advective derivative conditions lead to very little reflection and distortion at the
outflow, even with a very short mesh.
More quantitative comparisons, taken from [16], concern the period e, the
time- and space-averaged Nusselt number (Nu), the wavelength A, and the
thermal wave speed A/e. These results, reported in Table 2, demonstrate good
accuracy, even if we always utilized relatively coarse meshes.
From Eq. (33), we have evaluated the pressure at the Z = 0 wall boundary,
finding a pressure distribution that is not spatially periodic. However, as demon-
strated in Figure 10, where we plot the auxiliary variable P + 12/Re X, we find
that a periodic component is superimposed on the overall linear pressure distribu-
tion,
12
P= --X
Re
(41)
which corresponds to the Poiseuille flow. These results are, once again, in good
agreement with those of [14].
(a)
(b)
roo
L"{al_'
(c)
(0.1) (S.I)
-
(0.0)
(d)
(S.O)
Figure 7. Mixed convection in a plane
channel at Re = 10, Pe = 20/3, and
Fr = 1/150: (a) boundary conditions for
velocitycomponents; (b) boundary condi-
tions for streamfunction and vorticity;
(c) boundary conditions for temperature;
(d) finite-element mesh.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARYCONDmONS FOR UNSTEADY CONVECTION 231
Figure 8. Mixed convection in a plane
channel at Re = 10, Pe = 20/3, and
Fr = 1/150: periodic evolution of
streamlines at time intervals of teo
Figure 9. Mixed convection in a plane
channel at Re = 10, Pe = 20/3, and
Fr = 1/150: periodic evolution of tem-
perature contours at time intervals of
teo
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

232 G. COMINI ET AL.
Table 2. Mixed convection in a plane channel at Re = 10, Pe = 20/3, and Fr = 1/150
Period (8) (Nul Wavelength (Al Aje Reference
1.3319 2.5583 1.4465 1.09 Evans and Paolucci [I4]
8,000 nodes
1.273 2.574 1.45 1.14 Our results [161
2,097 nodes
Finally, in Figure 11, we show the time evolution of the spatially averaged
Nusselt number, starting from a Poiseuille velocity distribution and a linear
temperature field. As we can see, at first heat transfer takes place only by
conduction (Nu = 1), while advection is the dominating effect after the formation
of the vortices. When the flow becomes unstable, the space-averaged Nusselt
number Nu suddenly increases and starts oscillating around the final space- and
time-averaged value (Nu). It must be pointed out, however, that the value
(Nu) = 2.34, obtained in Figure 11, is referred to the short domain of Figure 7d,
and thus cannot be compared directly with the results of Table 2, which are
referred to a domain four times longer.
CONCLUSIONS
A finite-element method has been presented to solve transient laminar-con-
vection problems in two-dimensional incompressible flows, using the streamfunc-
tion-vorticity formulation. Advective derivative conditions have been used at the
outflow for all the variables involved. In multiply connected domains, the boundary
conditions for the streamfunction at internal walls have been imposed during the
assembly process. The vorticity at inflow and wall boundaries has always been
evaluated in the framework of the streamfunction equation. The accuracy of the
approach has been demonstrated by the solution of two well-known benchmark
problems, concerning forced convection over a circular cylinder in cross flow and
mixed convection in a plane channel heated from below.
x
Figure 10. Mixed convection in a plane
channel at Re = 10, Pe = 20/3, and
Fr = 1/150: pressure distribution at
Z = 0 with the Poiseuille pressure field
superimposed.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

OPEN BOUNDARY CONDmONS FOR UNSTEADY C0NVECI10N 233
01234.5678910
9
Figu", 11. Mixed convection in a plane
channel at Re = 10, Pe = 20/3, and
Fr = 1/150: time history of the spatially
averaged Nusselt number.
REFERENCES
1. G. Comini, G. Cortella, and M. Manzan, A Streamfunction-Vorticity-Based Finite-Ele-
ment Formulation for Laminar-Convection Problems, Numer. Heat. Transfer, Part B
vol. 28, pp. 1-22, 1995.
2. A. Campion-Renson and M. J. Crochet, On the Stream Function-Vorticity Finite
Element Solutions of Navier-Stokes Equations, Int. J. Numer. Meth. Eng., vol. 12, pp.
1809-1818, 1978.
3. M. F. Peeters, W. G. Habashi, and E. G. Dueck, Finite Element Stream Function-
Vorticity Solutions of the Incompressible Navier-Stokes Equations, Int. J. Numer. Meth.
Fluids, vol. 7, pp. 17-27, 1987.
4. T. E. Tezduyar, R. Glowinski, and J. Liou, Petrov-Galerkin Methods on Multiply
Connected Domains for the Vorticity-Stream Function Formulation of the Incompress-
ible Navier-Stokes Equations, Int. J. Numer. Meth. Fluids, vol. 8, pp. 1269-1290, 1988.
5. C. F. Kettleborough, S. R. Husain, and C. Prakash, Solution of Fluid Flow Problems
with the Vorticity-Streamfunction Formulation and the Control-Volume-Based Finite-
Element Method, Numer. Heat Transfer, Part B, vol. 16, pp. 31-58,1989.
6. T. E. Tezduyar, J. Liou, D. K. Ganjoo, and M. Behr, Solution Techniques for the
Vorticity-Streamfunction Formulation of Two-Dimensional Unsteady Incompressible
Flows, Int. J. Numer. Meth. Fluids, vol. 11, pp. 515-539, 1990.
7. R. J. Mackinnon, G. F. Carey, and P. Murray, A Procedure or Calculating Vorticity
Boundary Conditions in the Stream-Function-Vorticity Method, Commun. Appl. Numer.
Meth., vol. 6, pp. 47-48, 1990.
8. E. Barragy and G. F. Carey, Stream Function Vorticity Solution Using High-p Element-
by-Element Techniques, Commun. Appl. Numer. Meth. Eng., vol. 9, pp. 387-385, 1993.
9. G. Comini, M. Manzan, and C. Nonino, Finite Element Solution of the Stream
Function-Vorticity Equations for Incompressible Two-Dimensional Flows, Int. J. Nu-
mer. Meth. Fluids, vol. 19, pp. 513-525, 1994.
10. P. M. Gresho and R. L. Sani, Introducing Four Benchmark SOlutions, Int. J. Numer.
Meth. Fluids, vol. 11, pp. 951-952,1990.
11. P. M. Gresho, Letter to the Editor, Numer. Heat Transfer, Part A, vol. 20, p. 123, 1991.
12. M. Manzan and G. Comini, Inflow and Outflow Boundary Conditions in the Finite
Element Solution of the Stream Function-Vorticity Equations, Commun. Numer. Meth.
Eng., vol. 11, pp. 33-40, 1995.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

234 G. COMINI ET AL.
13. M. S. Engelman and M. A. Jamnia, Transient Flow past a Circular Cylinder: A
Benchmark Solution, Int. J. Numer. Meth. Fluids, vol. 11, pp. 985-1000,1990.
14. G. Evans and S. Paolucci, The Thermoconvective Instability of Plane Poiseuille Flow
Heated from Below: A Benchmark Solution for Open Boundary Flows, Int. J. Numer.
Meth. Fluids, vol. 11, pp. 1001-1013, 1990.
15. M. Manzan, G. Cortella, and G. Comini, Finite Element Solution of the
Streamfunction-Vorticity Equations for Flow past a Circular Cylinder, Proc. Ninth Int.
Conf. on Numerical Methods in Laminar and Turbulent Flow, C. Taylor and P. Durbetaki
(eds.), vol. 9, part II, pp. 1515-1526, Pineridge Press, Swansea, UK, 1995.
16. G. Cortella, G. Comini, and M. Manzan, A Streamfunction-Vorticity Based Finite
Element Solution for the Poiseuille-Benard Channel Flow, Proc. Ninth Int. Conf. on
Numerical Methods in Thermal Problems, R. W. Lewis and P. Durbetaki (eds.), vol. 9, part
I, pp. 5-16, Pineridge Press, Swansea, UK, 1995.
17. P. M. Gresho, Some Interesting Issues in Incompressible Fluid Dynamics, both in the
Continuum and in Numerical Simulations, Adv. Appl. Mech., vol. 28, pp. 45-140,1992.
18. R. L. Sani and P. M. Gresho, Resume and Remarks on the Open Boundary Condition
Minisymposium, Int. J. Numer. Meth. Fluids, vol. 18, pp. 983-1008,1994.
19. A. J. Baker, Finite Element Computational Fluid Mechanics, chap. 5, Hemisphere,
Washington, DC, 1983.
20. M. Behr, J. Liou, R. Shih, and T. E. Tezduyar, Vorticity-Streamfunction Formulation of
Unsteady Incompressible Flow past a Cylinder: Sensitivity of the Computed Flow Field
to the Location of the Outflow Boundary, Int. J. Numer. Meth. Fluids, vol. 12, pp.
323-342, 1991.
21. T. E. Tezduyar and J. Liou, On the Downstream Boundary Condition for the Vorticity-
Streamfunction Formulation of Two-Dimensional Incompressible Flows, Comput. Meth.
Appl. Mech. Eng., vol. 85, pp. 207-217, 1991.
22. T. E. Tezduyar, J. Liou, and D. K. Ganjoo, Incompressible Flow Computations Based on
the Vorticity-Streamfunction and Velocity-Pressure Formulations, Computers & Struc-
tures, vol. 35, pp. 445-472, 1990.
23. F. P. Incropera and D. P. De Witt, Fundamentals of Heat and Mass Transfer, p. 411, John
Wiley, New York, 1990.
D
o
w
n
l
o
a
d
e
d

b
y

[
T
e
l

A
v
i
v

U
n
i
v
e
r
s
i
t
y
]

a
t

1
1
:
5
1

0
9

M
a
r
c
h

2
0
1
4

You might also like