Vous êtes sur la page 1sur 7

Proceedings of IMECE02

Proceedings
2002 ASME International Mechanical Engineering of IMECE2002
Congress & Exposition
ASME International Mechanical Engineering Congress & Exposition
New Orleans, Louisiana, November 17-22, 2002
November 1722, 2002, New Orleans, Louisiana

IMECE2002-33925

IMECE2002-33925

TURBULENT HEAT TRANSFER IN A BACKWARD-FACING STEP FLOW USING A


NON-LINEAR k- MODEL

Marcelo Assato
Marcelo J.S. de Lemos *
Departamento de Energia - IEME
Instituto Tecnolgico de Aeronutica - ITA
12228-900 - So Jos dos Campos - SP Brazil
* ASME Member. Corresponding author. E-mail: delemos@mec.ita.br

ABSTRACT established in the literature that the standard k- model do not,


This work presents numerical results for heat transfer in on the whole, cope well with strong streamline curvature
turbulent flow past a backward-facing step. It is shown that non- whether it arises from flow over curved surfaces or imparted
linear k- models perform better than their linear counterparts swirling. And yet, turbulence-driven secondary motion and
when simulations are compared with experimental values. Wall directional effects due to buoyancy cannot, due to absence of
functions are used for simplicity of the simulations. The finite- information on individual stresses, be fully simulated with the
volume technique is employed for discretizing the transport standard k- model. In spite of that, it is frequently used for
equation set on a non-orthogonal grid system. The SIMPLE engineering computations due to the numerical robustness
method is used for correcting the pressure field. Results for the obtained via its linear stress-strain rate relationship (Jones &
reattachment length using the non-linear model are closer to the Launder (1972)).
experimental values when compared with similar calculations Non-linear models represent an extension of the standard
using the standard linear closure. two-equation closure and have shown good performance in
flows with recirculation and where normal stresses play an
1. INTRODUCTION important role (Assato & de Lemos (1998), (1999), (2000),
Flows presenting separation and reattachment have been 2001). Basically, they follow the procedures used in obtaining
the subject of various research efforts during the last decades. constitutive equations for laminar flow of non-Newtonian fluids
Such flows occur in gas turbine, nuclear reactors, and heat (Rivlin (1957)). Example is the work of Speziale (1987).
transfer devices, for example. The recirculating zone close to Essentially, the observed relationship between laminar flow of
heated wall causes large variations in the local heat transfer viscoelastic fluids and turbulent flow of Newtonian substances
coefficient, ultimately leading to overall heat transfer has motivated developments of such Non-Linear Models
augmentation around the reattachment point (Vogel & Eaton (Lumley (1970)). Other works on this subject are of Nisizima &
(1985)). Thus, accurate prediction of heat transfer in such areas Yoshizawa (1987), Rubinstein & Barton (1990) and Shih et al.
is important for reliable thermal design of engineering systems. (1993). In those papers, quadratic products were introduced
Accordingly, proper evaluation of turbulent heat transfer involving the strain and vorticity tensors with different
coefficients requires the use of turbulence models which can derivations and calibrations for each model. The basic
predict both velocity and temperature fields within acceptable advantage of these non-linear techniques over more complex
tolerance. closures, such as the Algebraic Stress Models (e.g. de Lemos &
In the literature, the standard k- model, in combination Sesonske (1985), de Lemos (1988)) lies on the achieved
with the assumption of a constant turbulent Prandtl number, has computational savings (roughly 25-50% less computing time).
been often used to predict heat transfer in separating and In the present study, velocity and temperature fields are
reattaching flows. However, a literature survey reveals that the obtained by means of linear and non-linear models for
standard k- model under-predicts the reattachment point convective heat transfer past a backward-facing step. Wall
location by an amount on the order of 20-25%. In fact, it is well functions for velocity and temperature are used for handling

1 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


wall proximity. It was noted that numerical results obtained by w Wall surface
the non-linear model of Shih et al. (1993) are closer to
experimental data of Vogel & Eaton (1985) when compared to 2. TRANSPORT AND CONSTITUTIVE EQUATIONS
simulations using isotropic modes. The governing equations to be solved are the continuity,
momentum and scalar equations which take the general form
r
div ( v ) = 0 , div (v U i ti ) = s u i , div (v q ) = s
r r r v
NOMENCLATURE (1)
Cf
2
Mean skin friction coefficient, w / U 0 / 2( ) r
where v is the time-mean velocity vector, ti contains the
r
C Coefficient of velocity-field turbulent model stresses acting in the i-direction and s ui represents all source
C1 , C2 Constants in velocity-field turbulent model terms including the pressure gradient and gravitational forces.
c1NL , c2 NL , c3 NL Coefficients of non-linear terms
The symbol represents the scalar quantity, q is its diffusive
cp Specific heat at constant pressure
flux and s represents the sources or sink of . The
k Turbulent kinetic energy, uiui / 2

P Mean pressure expressions for t i , q and su i are readily obtained as,


Pk Production of turbulence energy k P
Pr, Prt Molecular and turbulent Prandtl numbers ti = ij .i j , q = grad = j i j , su i = , (2)
y y i
q w
Wall heat flux
where, P = p + gh + 23 k . For easy of computation, the total
Re Reynolds number based on step height, U 0 H /
s Non-dimensional strain pressure P involves also a term containing the turbulent kinetic
s Source term energy k = uiu i / 2 where ui is the fluctuating part of the
Sij Strain rate tensor instantaneous velocity in the i-direction. The transport
coefficient , and the source terms, s , for k, and
St Stanton number, q w / (c pU 0 (Tw T0 ))
temperature T are written in Table 1.
T Mean temperature
T+ Non-dimensional mean temperature
U ,V Mean velocity components Table 1. Transport coefficient and source terms
XR Flow reattachment length
k T
X
Non-dimensional coordinate, (x xR ) / x R t t t
x Coordinate in streamwise direction +
k Pr Prt
y Coordinate normal to streamwise direction
2
C1 Pk C 2
Greek symbols s Pk k k 0
H Height of the step
Dissipation rate of k, (u i x j )(u i x j ) k2 U i
t = C , Pk = ij
Generalized coordinate x j
Von Krmns universal constant
, t Dynamic viscosity and eddy dynamic viscosity
Density In the table above, C = 0.09 , k = 1.0 , = 1.33 , C1 = 1.44
Transport coefficient and C2 = 1.92 . The non-dimensional groups ( Pr and Prt )
k , Model constants for turbulent diffusion of k and appearing in the definition of coefficient for the energy
ij Reynolds stress tensor, u iu j equation are the molecular and turbulent Prandtl numbers,
respectively. Also, t and Pk are the eddy viscosity and
w Wall shear stress
turbulent kinetic energy production rate, respectively.
Non-dimensional vorticity
A general non-linear expression for the Reynolds stress-
ij Vorticity tensor
kept to second order, can be written as

Subscripts
R Flow reattachment point

2 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


NL1 1 1

( )
L k
[
i j = t Si j + c1NL t Si kS k j 13 Sk lSk l i j

] + +
y =
C 4 k P 2 n
. (10)
P

NL 2
k
[
c2 NL t i k Sk j + j k Sk i

] + (3) Although it is recognized the limitation of the wall law for
simulating recirculating flows, the use of the above equations is
NL3 considered here for simplicity. Also, equation (9) is conditioned
k
[
c3NL t l k

jk

13 l k l k i j ] , to y P+ higher than 11.6.

Thermal Field: The wall function for temperature was
where i j is the delta Kronecker, the superscripts indicate developed in analogy to the wall function for velocity. A heat
flux per unit area crossing the control volume surface adjacent
Linear and Non Linear contributions. Also, Sij and i j are the
to the wall can be given by:
strain-rate and vorticity tensors, respectively, given as
T
q w = T c p (11)
u uj u uj n
Si j = i + , i j = i . (4)
x x x
j xi j i where c p is the specific heat at constant pressure of the fluid,
T is the difference between the temperature at the near wall
For the non-linear model of Shih et al. (1993) the values of node P and the wall and T is a coefficient for the thermal field
c1 NL , c2 NL and c3 NL are calculated by the following defined as,
expressions:

0.75 C 3.8 C 4.8 C T = for laminar flow, (12)
c1 NL = , c2 NL = , c3 NL = , (5) Pr
1000 + s 3 1000 + s 3 1000 + s 3
1 1
C k P
4 2

where, and T = n for turbulent flow, (13)


T+
23 k k
C = , s= 1
2
Sij Sij , = 1
2
ij ij (6) In (13), T + represents the dimensionless temperature at node
1.25 + s + 0.9
P given by:

3. WALL FUNCTIONS Prt


T+ = ln y + + (14)

Flow field: The classical wall function for the velocity
where is a function of the molecular Prandtl number
field is used in conjunction with the k- model to describe the
flow near the walls. The wall shear force can be expressed in expressed as (Kader & Yaglom (1972) ),
terms of the control volume located adjacent to wall, such that: 2

r V p = 12.5 Pr + 2.12 ln Pr 5.3


3
if Pr 0.5 , (15)
w = (7)
n 2
= 12.5 Pr + 2.12 ln Pr 1.5
3
if Pr < 0.5 . (16)
where subscript P refers to the node P next to the wall; V p is
the difference between the velocity parallel to wall and the wall
velocity; n is the distance between the node P and the wall; 4. NUMERICAL METHOD
and is a coefficient given as:
The numerical method employed for discretizing the
= for laminar flow (8) governing equations is the control-volume approach applied to
a generalized coordinate system. The SIMPLE algorithm was
and
used to relax the pressure field. Figure 1 shows a typical
1
C k P n
4
1
2 control-volume with detailed notation, distances and indexing
= for turbulent flow, (9) used in transforming the original equations into the
(
ln E y P+ ) coordinate system.
In (13) E is a constant which can be varied to simulate the
surface roughness (Launder & Spalding (1974)), = 0.41 is
the Krmn constant and y P+ represents the dimensionless
distance of node P to the wall, defined as:

3 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


The entire numerical treatment and discretization process U (y n ys )(U e U w ) (y e y w )(U n U s ) yP . U P y P . U P P
= = = a
of convection and diffusion terms in the momentum equation, as x P (y n y s )(x e x w ) (y e y w )(x n xs ) y P
. x P
y P
. x P
P

well as the handling of equations for k-, are presented in detail U (U n U s )(x e x w ) (U e U w )(x n xs ) U P . xP U P . xP
= =
P
= b
in a series of papers for linear (Pedras & de Lemos (2001) ) and y P (y n ys )(x e xw ) (ye y w )(x n xs ) y . x y . x P
P P P P

non-linear models (Assato & de Lemos (1998, 1999, 2000, V (y n ys )(Ve Vw ) (ye y w )(Vn Vs ) yP . VP yP . VP P
= = = c
x P (y n ys )(x e xw ) (ye y w )(x n xs ) y . x y . x P
P P P P

y
x
n V (Vn Vs )(x e x w ) (Ve Vw )(x n xs ) VP . xP VP . xP P
= = = d
y P (y n ys )(x e xw ) (ye y w )(x n xs ) y . x y . x
line N

P P P P
P
n An
nw y
n
y
n
(19)
w
- line x
n
x
e
ne
Thus, the discretized expressions for s and are
P
respectively
sw
e

e
y y
e
s=
k
P
2
( ) ( 2 2
2 aP + bP + cP + 2 dP , ) ( )
Ae
s
k
( )
e
x E
i = j se = bP cP (20)
P
2
- line
line
i = i
1

Similarly, algebraic equations for the mean flow and


x
temperature are obtained and resolved with the SIMPLE
method, as mentioned before.
Figure 1- Control-volume and notation
5. RESULTS AND DISCUSSION
2001). For the sake of clarity, only a few details are here
repeated.
The problem to be considered in this study consists of
With the help of Figure 1, the following geometric simulation of turbulent heat transfer in a backward-facing step
distances can be identified: flow (illustrated in Figure 2). This geometry is often employed
x e = (x ne x se ) , x e = (x E x P ) , y e = (y ne y se ) , for benchmarking the performance of turbulence models. The
results presented in this section were obtained using linear and
y e = (y E y P ) , x n = (x ne x nw ) , x n = (x N x P ) , non-linear versions of the k- model.
Here, H = H 2 H 1 represents the height of the step and
y n = (y ne y nw ) , y n = (y N x P ) , x P = (x e x w ) , L is the length of the separation bubble. The results obtained
by the turbulence models were compared with the experimental
x P = (x n x s ) , y P = (y e y w ) , y P = (y n x s ) .
data of Vogel & Eaton (1985). The inlet Reynolds number of
(17) Re = 28000 is based on the height of the step, H . The
channel expansion ratio is H 2 / H 1 = 1.25 and the working fluid
From equation (6), s and can be, respectively,
is air (Pr=0.71) with a uniform inlet temperature of T = 25 oC .
expressed as:
2 2 2
The boundary conditions for the thermal field are constant heat
k U U V V flux ( q w = 270W ) on the bottom wall, not including the step,
s= 2 + + + 2 ,
x y x y
k U V
= . (18)
y x
In order to obtain the discrete forms of the transformed
governing and constitutive equations (1)-(2)-(3), the velocity
derivatives in curvilinear coordinates are given as,

Figure 2- Geometry of back-step.

and qw = 0 on the other surfaces. The computational domain


was within the range 3.8H x 50H , calculated with a
grid of size 144x50 . For accuracy, the mesh was refined in the
recirculating region. The constant value for the turbulent
Prandtl number was Prt = 0.9 . The inlet conditions in Figure 2

4 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


were obtained by simulating first the flow in a channel of height agreement than the non-linear model next to the step. In other
H 1 . The calculated exit values were then used as inlet stations, the results of both models practically coincide. The
condition for the problem shown in Figure 2. distribution of skin friction coefficient C f is shown in Figure
Comparison of results with experimental values is shown 7(a). The non-linear model presents better agreement mainly
in Figure 3 for the case of a circular tube. Predictions for planar within the recirculation region. After the reattachment point
channels are of similar quality indicating good agreement with X > 0 , both models do not present good predictions in
experimental values. relation to the work of Vogel & Eaton (1985). From Figure 7(b)
1.4 4.0 it can be observed better results for the Stanton number using
1.2 3.5
the non-linear model, except in the region next to the
k reattachment point where both models do not perform well.
U 1.0 3.0

Um ut2
0.8 2.5

2.0
X*=-0.55 -0.44 -0.22 -0.11 0.0 X*= 0.33
0.6

0.4 1.5
3 3 3 3 3 3

0.2 1.0

0.0 0.5

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
2 2 2 2 2 2
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
r/R r/R
Y/h
a) b)
45.0
1 1 1 1 1 1
40.0
R
ut3
35.0
Laufer (1954)
0 0 0


30.0 0 0 0

25.0 Linear 0 1 0 1 0 1
__
0 1 0 1 0 1

20.0
---- Non-linear U / U0
15.0
Experimental ---- linear Non-linear
10.0

5.0

0.0
Figure 4 Dimensionless mean velocity U / U o .
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
r/R

c)

Figure 3 - Validation of computed values for circular tubes 6. CONCLUSION


a) Mean velocity U in ,
In this work convection heat transfer past a backward-
b) Turbulent kinetic energy, kin
facing step was simulated using linear and non-linear turbulence
c) Dissipation rate of k, in . models. It was observed that the non-linear theory presented
Experimental data of Laufer, (1954). results closer to experimental values when compared with the
standard closure. More specifically, better predictions were
calculated for the mean velocity, turbulent intensity and friction
The predicted flow reattachment length X R = L / H by
coefficient when using the non-linear theory. For the
the linear k- model was 5.0, while the non-linear k- model of temperature field it was observed that the linear model
Shih et al. (1993) gave a separation length of 6.0. Experiments presented better results than the non-linear counterpart in the
of Vogel & Eaton (1985) indicate X R 6.67 . Figure 4 shows station next to the step. Along other axial stations, both models
__ presented similar results. On the overall, the non-linear model
the dimensionless mean velocity field U / U 0 in several stations presented inaccurate results in the region next to the step, as can
along the channel. The experimental value X R is taken as be observed in the first stations in Figures 4, 5 and 6. However,
reference and the velocity U 0 is the inlet mean velocity. Except the use of equation (3) predicts reasonable values for the wall
shear stress w , and wall temperature T w . These quantities are
in the first station X = 0.55 , where X = (x L ) L , better
responsible for the good quality of the calculations of C f and
agreement with experimental data is obtained with the non-
linear model. The same behavior is observed for the St along the wall, particularly within the recirculating zone.
dimensionless turbulent intensity distribution u / U 0 shown in
Figure 5. The mean temperature profiles are shown in Figure 6.
It can be seen that both models predict the profiles very
similarly. The linear turbulence model presents a better

5 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Assato, M., de Lemos, M.J.S., 2000, Secondary Flow in Ducts
of Non-Circular Cross Section Using a Non-linear k-
X*=-0.55 -0.44 -0.22 -0.11 0.0 X*= 0.33 Model, Proc. of ENCIT2000 - 8th. Braz. Congr. of Thermal
Eng. and Sciences (on CD-ROM), Porto Alegre, RS, Brazil,
3 3 3 3 3 3
October 3-6 (In Portuguese).
Y/h Assato, M., de Lemos, M.J.S., 2002, A Novel Implicit
2 2 2 2 2 2 Numerical Treatment For Non-Linear Turbulence Models
Using High And Low Reynolds Formulations, Int. Journal
1 1 1 1 1 1
of Computational Fluid Dynamics (submitted).
de Lemos, M.J.S., Sesonske, A.,1985, Turbulence Modeling In
Combined Convection In Liquid-Metal Pipe Flow, Int. J.
0
0.0 0.1
0
0.2 0.0 0.1
0
0.2 0.0 0.1
0
0.2 0.0 0.1
0
0.2 0.0 0.1
0
0.2 0.0 0.1 0.2
Heat & Mass Transfer, vol. 28, #6, pp. 1067-1088.
de Lemos, M.J.S.,1988, Anisotropic Turbulent Transport
u / U 0 Modeling For Rod-Bundle, Int. J. of Heat & Technology,
Experimental ---- linear Non-linear vol. 6, # 1-2, pp. 27-37.
Jones, W.P. and Launder, B.E., 1972, The Prediction of
Figure 5 Dimensionless turbulent intensity u /U 0 . Laminarization with Two-Equation Model of Turbulence,
Int. J. Heat & Mass Transfer, Vol. 15, pp. 301 - 314.
Kader, B.A., Yaglom, A.M., 1972, Heat and Mass Transfer
Laws for Fully Turbulent Wall Flows, Int. J. Heat Mass
Transfer, vol. 15, pp. 2329-2351.
ACKNOWLEDGMENTS Launder, B.E. and Spalding, D.B., 1974, The Numerical
Computation of Turbulent Flows, Comp. Meth. Appl.
The authors are thankful to FAPESP and CNPq, Brazil, for Mech. Eng., vol. 3, pp. 269-289.
their financial support during the preparation of this work. Laufer, J., 1954, The Structure of Turbulence in Fully
Developed Pipe Flow, NACA-TR-1174.
REFERENCES Lumley, J.L., 1970, Toward a Turbulent Constitutive

Assato, M., de Lemos, M.J.S., 1998, Development of a Non-


3
Linear Turbulence Model For Recirculating Flows Using (a)
Generalized Coordinates, Proc. of ENCIT98- 7th Braz. 2
Experimental
Cong. Eng. Th. Sci., vol. 2, pp. 1386-1391, Rio de Janeiro, ---- Linear
RJ, Brazil, November 3-6. 1 Non-linear
Cf x 1000

Assato, M., de Lemos, M.J.S., 1999, Implicit Numerical


Treatment for Implementing a Non-linear Turbulence 0
w
Models in Generalized Coordinates, Proc. of COBEM99 -
( )
Cf =
2
15th. Braz. Congr. Mech. Eng. (on CD-ROM), ISBN: 85- -1 U 0 / 2

85769-03-3, guas de Lindia, So Paulo, Brazil,


November 22-26 (In Portuguese). -2
-1 0 1 2 3

4 X
X*=-0.75 -0.35 0.05 X*= 0.45 (b)
2.0 2.0 2.0 2.0
3
St x 1000

1.5 1.5 1.5 1.5


2

qw

( ))
1.0 1.0 1.0 1.0 St =
Y/h 1
(
C p U 0 T w T0

0.5 0.5 0.5 0.5


0
-1 0 1 2 3

0.0 0.0 0.0


0 2 4 6 8
0.0
0 2 4 6 8
X
0 2 4 6 8 10 12 14 -1 0 1 2 3 4 5 6 7

T T0 ( C) o Figure 7- (a) Friction coefficient;


(b) Stanton number.
Experimental ---- linear Non-linear
Figure 6 Mean temperature field. Relation, J. Fluid Mech., 41, 413.

6 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Nisizima, S. and Yoshizawa, A., 1987, Turbulent Channel and
Couette Flows Using an Anisotropic k- Model, AIAA J.,
Vol. 25, N0 3, p. 414.
Pedras, M.H.J., de Lemos, M.J.S, 2001, Simulation of
Turbulent Flow In Porous Media Using A Spatially Periodic
Array and a Low Re Two-Equation Closure, Numerical
Heat Transfer - Part A Applications, vol. 39, n.1, pp. 35-59.
Rivlin, R.S., 1957, The Relation Between the Flow of Non-
Newtonian Fluids and Turbulent Newtonian Fluids, Q.
Appl. Maths, 15, 212.
Rubinstein, R. & Barton, J.M., 1990, Renormalization Group
Analysis of the Stress Transport Equation, Phys Fluids A
2, pp. 1472.
Shih, T.H., Zhu, J., Lumley, J.L., 1993, A Realisable Reynolds
Stress Algebraic Equation Model, NASA TM-105993.
Speziale, C.G., 1987, On Nonlinear k-l and k- Models of
Turbulence, J. Fluid Mech., vol. 176, pp. 459-475.
Vogel, J.C., Eaton, J.K., 1985, Combined Heat Transfer and
Fluid Dynamic Measurements Downstream of a Backward-
Facing-Step, J. Heat Transfer, vol. 107, pp. 922-929.

7 Copyright 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Vous aimerez peut-être aussi