This article was downloaded by: [Aston University]

On: 23 January 2014, At: 02:57

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

A MODIFICATION TO THE SIMPLE METHOD FOR

BUOYANCY-DRIVEN FLOWS
a a b c
Y. Sheng , M. Shoukri , G. Sheng & P. Wood
a
Department of Mechanical Engineering , McMaster University , Hamilton, Ontario, L8S 4K1,
Canada
b
Faculty of Environmental Studies, York University , North York, Ontario, M3J1P3, Canada
c
Department of Chemical Engineering , McMaster University , Hamilton, Ontario, L8S 4K1,
Canada
Published online: 25 Apr 2007.

To cite this article: Y. Sheng , M. Shoukri , G. Sheng & P. Wood (1998) A MODIFICATION TO THE SIMPLE METHOD FOR
BUOYANCY-DRIVEN FLOWS, Numerical Heat Transfer, Part B: Fundamentals: An International Journal of Computation and
Methodology, 33:1, 65-78, DOI: 10.1080/10407799808915023

To link to this article: http://dx.doi.org/10.1080/10407799808915023

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
 A MODIFICATION TO THE SIMPLE METHOD FOR
BUOYANCY-DRIVEN FLOWS

Y. Sheng and M. Shoukri

Department ofMechanical Engineering, McMaster University,
Hamilton, Ontario, Canada L8~ 4KI

G. Sheng
Faculty ofEnvironmental Studies, York University,
North York, Ontario, Canada M3J IP3
Downloaded by [Aston University] at 02:57 23 January 2014

P. Wood
Department of Chemical Engineering, McMaster University,
Hamilton, Ontario, Canada L8S 4KI

TIu! essence of the SIMPLE method lies in its coupling between the momentum and
continuity equations. Almost all the algorithms of the SIMPLE family are based on one
precondition, thai is, the corrected velocity is obtained from the corrected pressure only.
However, in buoyancy-drivenfluws, there are two majorfactors driving the fluid movement:
the temperature grrulients and the pressure (including kinetic pressure) grrulients. In this
article, the effect of the temperature correction on the velocity correction is considered
during the derivation of the pressure-linked equation. A modification /Q the SIMPLE
algorithm, SIMPLEr, is proposed and two cases are tested. It is shown thai this
modification provides a better convergence rate and more robust results.

INTRODUCTION
The finite-difference method with primitive variables is one of the most
widely used computation methods for solving convective heat transfer problems.
This method solves the discretized conservation equations for continuity, momen-
tum, energy, and others (if any). In 1972, Patanker and Spalding [1] successfully
developed the SIMPLE method, which stands for Semi-Implicit Method for Pres-
sure-Linked Equations, with a staggered grid system. The essence of the method
lies in its treatment of the coupling between the momentum and continuity
equations. Since then, several variants of SIMPLE have been proposed to improve
its convergence rate. In 1980, Patankar [2] introduced the SIMPLER method, in

Received 26 July 1996; accepted 31 July 1997.

The authors would like to express thanks to Prof. G. Round and P. Bewer for their help and
support. The computational work was carried out on the C-90 supercomputer at San Diego Supercom-
puter Center, and their support also is appreciated, as is the financial support of NSERC.
Address correspondence 10 Professor Philip Wood, Department of Chemical Engineering,
McMaster University, Hamilton, Ontario, Canada 1.8S 4KI. E-maii: woodpe@mcmaster.ca

Numerical Heat Transfer, Part B, 33:65-78, 1998

Copyright © 1998 Taylor & Francis
1040-7790/98 $12.00 + .00 6S
 66 Y. SHENG ET AL.

NOMENCLATURE
a coefficient in discretized x Cartesian coordinate
equation y Cartesian coordinate
b constant in discretized equation p volumetric thermal expansion
B volumetricexpansioncoefficient coefficient
d pressure coefficient in velocity- r diffusion coefficient
correctionexpression s temperature coefficient in
g gravitational acceleration velocity-correction expression
Gr Grashof number /!IT temperature difference
[= gP/!lT/ 3/ ( IL/ p )' j /!lV volume of control volume
/ length IL molecular viscosity
p pressure p density
Pr Prandtl number <P general dependent variable
Ra Rayleigh number (= Gr Prj
Re Reynolds number(= Vlp/IL)
Downloaded by [Aston University] at 02:57 23 January 2014

Ri Richardson number Superscripts

(= Gr/Re')
Res Residual variablecorrection
S source term • variable before correction
Sc constant part of linearized
source term Subscripts
Sp coefficient of <pp in linearized
source term E, W,P,N,S grid points
u velocity component in nb neighbor grid points
x direction e, W, n, S control-volume faces
v velocity component in r reference
y direction h hot
V velocity c cold

which an extra equation was solved for the evaluation of pressure. In 1984, Van
Doormaal and Raithby [3] proposed the SIMPLEC method to improve the consis-
tency. In 1985, PISO, a similar method to SIMPLER, was proposed by Issa [4]. In
the same year, Latimer and Polard [5] developed a method called FIMOSE to
introduce a new, fully implicit solution algorithm. In 1991, based on the minimiza-
tion of the global residual norm, Chatvani and Turan [6] proposed a
pressure-velocity coupling algorithm to determine the underrelaxation factor in
the pressure equation. In 1992, Lee and Tzong [7] introduced an artificial source
term into the pressure-linked equation. In 1993, Yen and Liu [8] introduced an
additional explicit correction step to decrease the number of iterations.
It can be seen that all these modifications are based on one precondition:
The velocity correction is evaluated with respect to the pressure correction only.
However, in buoyancy-driven flows, two major factors drive the fluid movement:
the temperature gradients and the pressure (including kinetic pressure) gradients.
In order to reflect this physical phenomenon, the pressure-linked equation should
be extended. In the present work, a modification to the SIMPLE method, SIM-
PLET, is developed specifically for buoyancy-driven (i.e., natural and mixed con-
vective) flows. It considers the effect of both the pressure correction and the
temperature correction on the velocity correction during the derivation of the
 THE SIMPLET METHOD FOR BUOYANCY·DRNEN FLOWS 67

pressure-linked equation and therefore makes the pressure correction and velocity
correction more reasonable. Furthermore, the energy equation is solved based on a
consistent velocity and pressure field. Therefore, the convergence rate is also
improved.

FORMULATION

Pressure-Linked Equation and SIMPLE Method

The differential equations expressing conservation of mass, momentum, and
energy for two-dimensional steady flows in Cartesian coordinates can be written as

a(pu,p) + a(pV<!» = !......(ra<!» + !......(ra<!» + s

Downloaded by [Aston University] at 02:57 23 January 2014

(1)
ax ay ax ax ay ay

Using a control-volume method with a staggered grid system as shown in

Figure 1 to discretize Eq. (1) and noting that the source term over the control
volume ~V is

we have

(3)

~
...l- J. I
I ; I
I
-- --+--1 -f- i -N - - r - - I

I
--~-~ f-_.
.J.. -'- .J..
r ~ ,n I
-~ f--- ..
.l..-w w-~
I
-"1--'1
::% p~e- E.L
--
I

-- --~-~ -fa;;c:~
~---t
I
.-'--j ..

--
I s I
.l.. I I
I I S I
I
I
--I --r---j -~--!
__ .1 _ _

,
I -~-4 --
:..
I
I
, I
.i,
,
I I I I I
Lx
-.. ::: u; t ::: u; 0 = other variables.
Figure I. Staggered grids.
 68 Y. SHENG ET AI.

where the summation is over the appropriate neighbor points, and the coefficient
a p is

The term b is the summation of all the other terms that are treated as constants.
Now, to develop the algorithm, and without loss of generality, let cf> be the
velocity u and write Eq. (3) for the control volume centered at e. After extracting
the pressure term from b, we have

(4)

Based on a guessed p*, this u-momentum equation is solved to obtain a u* that

satisfies
Downloaded by [Aston University] at 02:57 23 January 2014

(5)

Subtracting Eq. (5) from Eq. (4), we have an equation for the velocity correction
U 'C""

(6)

Assuming that Eanbu'nb = zero (reference to SIMPLE method, Patanker [2]) gives

where
p = p* + p'
u = u* + u'
(8)
Ac
dc = -
ac

(9)

As discussed in [1], the omission of Ea nb U~b is acceptable, for it will not

change the ultimate solution.
Similarly, for u w ' un' and Us we have

Us = ui + d,( Ps - p'p)

Substituting these into the following discretized continuity equation,

 THE SIMPLET METHOD FOR BUOYANCY·DRlVEN FLOWS 69

gives the pressure-linked equation

(10)

where

(11)

as = (pAd), (12)

[ - b also represents a mass source] (13)

where
Downloaded by [Aston University] at 02:57 23 January 2014

Cw = (pu*A)w C
s = (pv*A), (14)

With a guessed pressure field p*, the momentum equations are solved to
obtain u* and v*. The pressure-linked equation is then solved to obtain p'. Finally,
we make the correction to the velocity field and pressure field and then solve the
energy equation to find the solution for velocity u, v, pressure p, and temperature
T. Since the equation set is coupled and nonlinear, the coefficients in the
discretized equations are related to the dependent variables and need to be
updated. Therefore, sufficient iterations should be executed until a specified
convergence criterion is satisfied.

SIMPLET: A Modification to SIMPLE for Buoyancy-Driven Flows

One common feature of the SIMPLE family of methods (such as SIMPLER,
SIMPLEC, etc.) is that the velocity change is caused by the pressure change only.
Since the changed velocity should satisfy mass continuity, a pressure-linked equa-
tion can be derived to replace the continuity equation. Mathematically, only a
pressure-related term is extracted from the constant term, b, in the momentum
equation, and the remaining terms are considered constant during an iteration. For
most incompressible flows, this mathematical treatment reflects the physical situa-
tion adequately. However, for natural-convection and mixed-convection problems,
the buoyancy-driving force cannot be neglected and sometimes it is the major force
driving the flow. Noting that the velocity change is caused not only by pressure
change but also by temperature change, the buoyancy-driving force should also be
extracted from the constant term, b, in addition to the pressure term. Therefore,
Eqs. (4)-(14) should be rewritten as follows:
 70 Y. SHENG ET AL.

where
T = T* + T'
gxB~V
f3e = 2 (8')

se = f3e
ae

(9')

Similarly,

U~ + dw(p'w - p'p) + ow(T~ + T~)

Downloaded by [Aston University] at 02:57 23 January 2014

Uw =

un = U: + dn(p'p - P'N) + 0n(T~ + T~)

u, = U: + d,(pS - p'p) + o,(T;' + T~)

The final pressure-linked equation keeps the same form,

(10')

where ap ' Ea nb , a E , a w ' aN' as are the same as in Eqs. (11) and (12)

where c E ' c w, CN' Cs are the same as in Eq. (14).

as = (pAo), (15)

It can be seen that the extra bracketed term in Eq. (13') is the only difference
between the original pressure-linked equation and the new one.

SOLUTION PROCEDURE
Based on the previous discussion, the solution procedure can be summarized
as follows.

1. Guess a pressure field p* and a temperature field T*.

2. Solve the momentum equation to obtain u* and u*.
3. Solve the energy equation to obtain T and evaluate T' = T - T*.
4. Solve the pressure-linked equation to obtain p'.
5. Correct the velocity field and the pressure field to obtain u, u, and p.
6. Solve the energy equation to obtain the temperature field T.
7. Solve other </> equations (if any, such as k, E for turbulent flows).
8. Update properties to prepare for the next iteration.
 THE SIMPLET METHOD FOR BUOYANCY-DRIVEN FLOWS 7l

Obviously, this procedure is very similar to the original SIMPLE method

procedure. The major difference is that there is an extra step 3 to evaluate T. It
should also be noted that an approximation has been introduced in the evaluation
of T. We solve the energy equation at step 3 using u* and v* instead of u and v
so that the u' and v' are omitted. When the solution approaches convergence, u'
and v' approach zero. This kind of omission will not change the ultimate solution,
as mentioned in the development of the original SIMPLE method.
The updated pressure-linked equation gives a more reasonable pressure
correction since it considers the effect of the temperature change on the velocity
change. The corrected velocity field and pressure field are consistent, so the
convergence rate is improved. This modification provides a faster convergence for
buoyancy-driven flows as illustrated below in the two case studies examined.
Downloaded by [Aston University] at 02:57 23 January 2014

CASE STUDIES
The updated SIMPLE method for buoyancy-driven flows is tested by using
two typical examples. The first involves a natural-convection problem in a square
cavity; while the second considers a mixed-convection problem behind a vertical
backward-facing step.

Case Study I
The heated square cavity problem is solved for air with P, = 0.71 using an
array of 32 x 32 irregularly spaced grids. The nonuniformity of the grids is
generated in an exponential fashion with the refined grids along the boundaries. By
introducing the Boussinesq approximation, the governing equations can be written
as follows:
au av
-+-=0 (16)
ax ay
a(uu) + a(vu) = -1 ap + I-L[!-(au) + !-(au)] (17)
ax ay p ax p ax ax ay ay
a(uv)
- a(vv)
-+ - - = --1 -ap + -I-L[a(av)
- - + - a(av)]
- + g /3(T - T) (18)
ax ay p ay p ax ax ay ay r

a(uT) + a(vT) =a[!-(aT) +!-(aT)] (19)

ax ay ax ax ay sy
The boundary conditions are given by
x = 0; u = 0, v = 0, T= Th
x = I; u = 0, v = 0, T= T;
aT
y = 0; u = 0, v = 0, -=0
ay
aT
y =h; u = 0, v = 0, -=0
ay
 72 Y. SHENG ET AL.

The convergence criteria are taken from the default ones set in the FLUENT
package [9]:
Res no rm = 1.0e-3 for u, U,p
Res norm = 1.0e-6 for T
The problem is first solved with the FLUENT package choosing the options
of SIMPLE and SIMPLEC. As we know from the FLUENT User's Guide, there is
no Boussinesq approximation in the solution procedure. The residual histories for
the two options, SIMPLE and SIMPLEC, are shown in Figure 2a. It can be seen
that the application of these two methods result in almost the same convergence
rates.
Using the same grid distribution and convergence criteria but introducing the
Boussinesq approximation, the same problem is solved using the SIMPLE and
Downloaded by [Aston University] at 02:57 23 January 2014

SIMPLET codes written by us. The residual histories are plotted in Figure 2b.
Further tests for different Rayleigh numbers are executed and the results are
shown in Figure 3. The required iterations for different Rayleigh numbers are
summarized in Table 1.
In each case, the underrelaxation factors for the two methods, SIMPLE and
SIMPLET, are the same since the convergence rates are related to the underrelax-
ation factors being used. By comparing Figures 2a and 2b, it can be seen that for
the same SIMPLE method, the introduction of Boussinesq approximation improves
the convergence rate, especially at the initial stages of the iteration (say, the first
300 iterations in the particular example). From Figures 2b and 3, it is obvious that
in all the cases, the SIMPLET method provides significantly faster convergence
rate compared to the conventional SIMPLE method.

Case Study II
The mixed-convection problem behind a vertical backward-facing step shown
in Figure 4 is solved also for air using a 72 X 29 grid array with the refined grids
around the walls.
The governing equations are the same as those in Eqs. (16)-(19). The
boundary conditions are given by
au auor
x=O and o < y < [step; -ax = 0, -
0, -
ax
ax
= = 0
aT
x = h - h step and o <y < [step; U = 0, U = 0, -ax = 0
x=h and [step < Y < I; u = 0, U = 0, T = Th

y=o and 0< X < h - h step ; U = U in ' U = 0, T = Tin

aT
and h - h step < X < h; u = 0, u = 0, -
=0
ay
au au aT
y = [ and O<x<h'-=O - = 0 - = 0
'ay 'ay 'ay
 THE SIMPLET METHOD FOR BUOYANCY·DRIVEN FLOWS 73

Convergence Rates

(Ra = l.e6)

SIMPLEC (Solid)

SIMPLE (Dashed)
Downloaded by [Aston University] at 02:57 23 January 2014

100 200 300 400

Iteration Number

(a)

Convergence Rates

(Ra = 1.e6)

SIMPLET (Solid)
SIMPLE (Dashed)

--:::~: ~::::
---== ., ., ----III_
100 200 300
Iteration Number
(b)

Figure 2. Residual histories for test case I (Ra ~ t.eo): (a) solved with SIMPLE
and SIMPLIC (using FLUENT); (b) solved with SIMPLE and SIMPLET.
 74 Y. SHENG ET AI..

~10' Convergence Rates

"'iiI (Ra=l.e4)
0:::'0'1
SIMPLET (Solid)
10·
SIMPLE (Dashed)

10·

10·

10'
Downloaded by [Aston University] at 02:57 23 January 2014

10' " ... _... -- .......

10.7

100 200 300. 400

IterationNumber

Convergence Rates
~10' (Ra = 1.e5)

~ 10. 1 SIMPLET (Solid)

SIMPLE (Dashed)

10" \

10·

10'
'-
10'
--- -',

100 200 300

IterationNumber
Figure 3. Residual histories for test case I.

In the test cases,

[= 173 mm [step = 5.6 mm

h = IOmm h step = 10 mm
 THE SIMPLET METHOD FOR BUOYANCY·DRIVEN FLOWS 7S

Convergence Rates
(Raa 1.e7)

SIMPLET (Solid)

10' SIMPLE(Dashed)

10'

..... _--
Downloaded by [Aston University] at 02:57 23 January 2014

10~ ".

50 100 150 200 250 300

IterationNumber
Figure 3. Residual histories for test case I (Continued).

U inis set to different values to provide different Richardson numbers as shown in

Table 2.
Based on the same default convergence criteria, the required iterations for
different Richardson numbers are summarized in Table 2. It can be seen that when
the Richardson number is large, indicating that buoyancy is the dominant factor
driving the flow, the SIMPLET method provides significantly faster convergence.
The residual histories for Ri = 00 and 100 are shown in Figure 5.
When the Richardson number is small, the number of iterations required for
SIMPLET is only slightly less than that for SIMPLE and therefore, the faster
convergence effect of SIMPLET becomes more moderate. As with the SIMPLER
method, an extra equation (step 3 of the solution procedure) must be solved in
each iteration for SIMPLET. This extra work partly offsets the benefit from the
decrease of number of iterations required.

SUMMARY AND CONCLUSIONS

For buoyancy-driven flows, there is no direct link between velocity and
temperature in the original momentum equations. The temperature change influ-

Table 1. Required iterations for test case I

Ra

1.e4 1.e5 1.e6 1.e7

SIMPLE 420 375 380 340

SIMPLET 215 170 200 245
 76 Y. SHENG ET AL.
Downloaded by [Aston University] at 02:57 23 January 2014

u· u,.
Figure 4. Computation domain for test
v·o case II (the mixed-convection problem
T_T. behind a vertical backward-facing step).

ences the velocity through the density. By introducing the Boussinesq approxima-
tion, a direct link between the velocity and temperature is established, and the
temperature, T, is treated as an explicit variable in the discretized momentum
equations. The introduction of the Boussinesq approximation makes the solution
easier to reach convergence. However, in the SIMPLE family of methods, the
numerical solution procedure did not incorporate this approximation and thus
there is no direct link between velocity change and temperature change. The
derivation of the pressure-linked equation considers only the velocity change
caused by pressure change in the usual manner. Fortunately, lack of this update
only slowed the convergence rate and did not change the final result.

Table 2. Required iterations for test case II

Ri

100 10 0.1 0.01

SIMPLE 1640 1210 545 770 615 650

SIMPLET 1140 595 455 670 515 500
 THE SIMPLET METHOD FOR BUOYANCY·DRNEN FLOWS 77

Convergence Rates
(Natural Convection)

SIMPLET (Solid)

SIMPLE (Dashed)

--, '"

'-,
Downloaded by [Aston University] at 02:57 23 January 2014

"

500 1000 1500

Iteration Number

Convergence Rates
(Ri=l00)

SIMPLET (Solid)

SIMPLE (Dashed)

'"
".

200 400 600 BOO 1000 1200

Iteration Number
Figure 5. Residual histories for test case II.

The idea of SIMPLET is to take full advantage of the Boussinesq approxima-

tion. It makes the direct link between velocity and temperature not only in the
discretized momentum equations but also in the solution procedure. The derivation
of the pressure-linked equation considers the velocity change caused by both
pressure change and temperature change. Therefore, the solution procedure is
more concise and the convergence rate is improved.
 78 Y. SHENG lIT AL.

REFERENCES
1. S. V. Patankar and D. B. Spalding, A Calculation Procedure for Heat Mass and
Momentum Transfer in Three-Dimensional Parabolic Flows, Int. J. Heat Mass Transfer,
vol. 15, pp. 1787, 1972.
2. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington, DC,
1980.
3. J. P. Van Doormaal and G. D. Raithby, Enhancements or the Simple Method for
Predicting Incompressible Fluid Flows, Numer. Heat Transfer, vol. 7, pp. 147-163, 1984.
4. R. I. Issa, Solution of the Implicitly Discretized Fluid Flow Equations by Operator-Split-
ting, J. Comput. Phys., vol. 62, pp. 40-65, 1985.
5. B. R Latimer and A. Pollard, Comparison of Pressure-Velocity Coupling Solution
Algorithms, Numer. Heat Transfer, vol. 8, pp. 635-652, 1985.
6. A. U. Chatwani and A. Turan, Improved Pressure-Velocity Coupling Algorithm Based
on Minimization of Global Residual Norm, Numer. Heat Transfer, Pari B, vol. 20, pp.
Downloaded by [Aston University] at 02:57 23 January 2014

115-123, 1991.
7. S. L. Lee and R Y. Tzong, Artificial Pressure for Pressure-Linked Equation, Int. J. Heat
Mass Transfer, vol. 35, pp. 2705-2716, 1992.
8. R-H. Yen and C.-H. Liu, Enhancement of the SIMPLE Algorithm by an Additional
Explicit Corrector Step, Numer. Heat Transfer, Pari B, vol. 24, pp. 127-141, 1993.
9. FLUENT User's Guide, Vol. 3, Version 4.3, chap. 16, FLUENT, Inc., Lebanon, NH,
January 1995.