FLUID DYNAMICS

RESEARCH
ELSEVIER Fluid Dynamics Research 20 (1997) 127-142

Modeling the turbulent heat and momentum transfer in flows under different thermal conditions
Y. Nagano a,*, H. Hattoria, K. A b e b
aDepartment of Mechanical Engineerin9, Nagoya Institute of Technology, Gokiso-cho, Showa-ku, Nagoya 466, Japan b Toyota Central Research and Development Laboratories, Inc., Nagakute-cho, Aichi-gun, Aichi-ken 480-11, Japan

Abstract
Two-equation turbulence models for velocity and temperature (scalar) fields are developed to calculate wall shear flows under various flow conditions and related turbulent heat transfer under various wall thermal conditions. In the present models, we make the modified dissipation rates of both turbulent energy and temperature variance zero at a wall, though the wall limiting behavior of velocity and temperature fluctuations is reproduced exactly. Thus, the models assure computational expediency and convergence. Also, the present k-e model is constructed using a new type of expression for the Reynolds stress uiuj proposed by Abe et al. [Trans. JSME B 61 (1995) 1714-1721], whose essential feature lies in introducing the explicit algebraic stress model concept into the nonlinear k-E formulation, and the present twoequation heat transfer model is constructed to properly take into account the effects of wall thermal conditions on the eddy diffusivity for heat. The models are tested with five typical velocity fields and four typical thermal fields. Agreement with experiment and direct simulation data is quite satisfactory.

I. Introduction
Two-equation turbulence models are n o w powerful tools for solving flow and heat transfer problems encountered in engineering applications. Since the 1980-1981 Stanford Conference (Kline et al., 1981), a great deal o f effort has been directed to the d e v e l o p m e n t and i m p r o v e m e n t o f the model in this category. In particular, the near-wall limiting behavior o f turbulence was discussed in detail, and several models representing the correct behavior were proposed, a m o n g which the earliest were those developed b y M y o n g and Kasagi (1990) and b y N a g a n o and T a g a w a (1990) (hereafter referred to as the N T model). Recently, based on these two models, various extended models o f k-~ type have been constructed. For example, Abe et al. (1994) proposed an i m p r o v e d version o f the N T

* Corresponding author. 0169-5983/97/$17.00 @ 1997 The Japan Society of Fluid Mechanics Incorporated and Elsevier Science B.V. All rights reserved.

PIIS0169-5983(96)00049-4

128

Y. Nagano et al. / Fluid Dynamics Research 20 (1997) 127-142

model (Nagano and Tagawa, 1990). The most essential improvement made in this model is that the Kolmogorov velocity scale us ---- (re) 1/4, where v is the kinematic viscosity, is introduced instead of the friction velocity u~ to account for the near-wall and low-Reynolds-number (LRN) effects. On the other hand, Hattori and Nagano (1995) suggested a developed version of the Nagano and Hishida model (NH) (Nagano and Hishida, 1987), which reproduced the wall limiting behavior exactly. As for the heat-transfer calculations, Nagano and Kim (NK) (1988), Nagano et al. (NTT) (1991) have evolved new approaches at the two-equation level of turbulence modeling. Generally, it is known that numerical calculations with a two-equation model which satisfies the wall-limiting behavior of turbulence do not necessarily provide a good convergence. This problem mainly stems from the imposed boundary conditions for the dissipation rates of turbulent energy and temperature variance at the wall, i.e., non-zero values. On the other hand, the NH and NK models make the (modified) dissipation rates zero at the wall, so that these models are very stable in computation. The NH and NK models, however, do not reproduce exactly the wall limiting behavior; thus the models are less accurate in predicting the heat transfer under arbitrary wall thermal conditions. In the present study, we propose modified two-equation models in which the apparent dissipation rates of both turbulent energy and temperature variance are set to zero at the wall though the wall limiting behavior is reproduced exactly. The present k-e model is a nonlinear k-e model of Abe et al. type (Abe et al., 1995b) which incorporates some essential characteristics of Reynolds stress and algebraic stress models. The proposed k-e turbulence model was tested by application to a channel flow and boundary layers with/without pressure gradients. The proposed heat-transfer model was tested in a channel flow and in turbulent boundary layers with four different wall thermal conditions; i.e., a uniform wall temperature, a uniform wall heat flux, a constant wall temperature followed by an adiabatic wall, and intermittent stepwise heat-flux input at the wall.
2. Governing equations
2.1. Velocity field

An incompressible turbulent velocity field is described with the following governing equations (Nagano and Tagawa, 1990) (the equation of continuity, the ensemble-averaged Navier-Stokes equation, and the transport equations of the turbulent energy and its dissipation rate):
Ui,i 7_ O,

(1)
P'g + (vUg, j - u ~ ) , j , p

D U~ _ Dr

(2)
(3)

Dk -- =D~+Tk+Pk+Hk-e, Dr

e -e2 De De + T~ + H~ - C ~ U i j - C ~ 2 f ~ + E, (4) Dr where D/Dz implies the substantial derivative and the Einstein summation convention applies to repeated indices. The terms on the right-hand sides in Eqs. (3) and (4) are Dk = vkjj, D~ =

Y. Na#ano et al. / Fluid Dynamics Research 20 (1997) 127-142

129

w,jj (molecular diffusion); Tk = - ( k ' u j ) d, T~ = - ( d u j ) o (turbulent diffusion); Pk = -uiujU~,j (production); Hk = --uj(p/p) o, fI~ = - ( 2 v / p ) ~ j (pressure diffusion); e = vu~,sui,s (dissipation); and E is the extra production term, respectively, where k' = uiuj2, e' = V(U~,ku,,k) and k = k'. In Eq. (4), C~l and Ca are the model constants, and f~ is the model function to satisfy the physical requirement in the near-wall region.

2.2. Thermal field

Using the concept of eddy diffusivity for heat at, the governing equations of the two-equation heat transfer model can be written (Nagano et al., 1991) as

D~
Dz = (~T,j - h~jt),j, -u j--7 = ~, T,j,
Dkt

(5) (6) (7)

gt----

Dz = D~, + Tk, + Pk, -- et,

Dat
Dz

D~, + L, - C p l f e l ~ u j t T j - Cp2fe2~uiujUi,j
" '

gt----

Cmfm~t

2
-

~ F.tg
+
t

c'o2jD2 T

Et,

(8)

where Dk, = ~ktjj, D~, = ~at,jj (molecular diffusion); Tk, = -(kt'uj)o, T~, = -(etuj) 4 (turbulent diffusion); Pk, = - ~ j t T 4 (production); at = ~t, jt o (dissipation); and k~ = t2/2, eI = ~(t, kt~) and kt = k~. In Eq. (8), Cm, Cp2, Cm and CD2 are model constants, and fro, fP2, fD1 and f m are model functions, respectively.

3. Modeling
3.1. Nonlinear k-e model
In this study, we adopt a nonlinear formulation for the k-e model constructed by Abe et all (1995b) (hereafter referred to as the AKN model), and modify it so that we can make e zero at the wall. This formulation includes a new type of expression for the Reynolds stress uiuj by introducing the explicit algebraic stress model concept (Pope, 1975; Taulbee, 1992; Gatski and Speziale, 1993) to the nonlinear k-~ representation. The present nonlinear k-e model is thus described with the following equation (Abe et al., 1995b):

uiuj = ~kbij +
x

1 1 + ( c ~ c ) 2 { ~ 2 + 2 (02 _ $2) fB}

- 4c~ (&a~j - ~i~&) + 4 c ~ t ( s i ~ & k - xSm.Sma~j) (9)

-2v,&

where CD is the model constant and Sij[= (Ui,j + Us, i)~2] and f2is[= (Ui, j - Uj,/)/2] are the strain-rate and vorticity tensors, respectively, and S 2 = SijS~j and f22 = ~2oY20-. According to Abe et al. (1995b),

130

Y.

Nagano et al./Fluid Dynamics Research 20 (1997) 127-142

in Eq.(9), the parameter (CD'cc)2(f2 2 - S 2) is one of the most important measures in turbulence since it indicates how the flow field deviates from the condition of pure shear flow. In complex flows, sometimes the normal strain rate becomes much larger than the shear strain rate, i.e., $2>>f2 2. Therefore, to guarantee non-negative turbulent intensities, the model function f~ is introduced as follows: fB : 1 + Cn (K2 z - $2). The eddy viscosity vt in Eq. (9) is expressed as
k2

(10)

vt = C~f~--. e
Note that the following time scale is employed as the characteristic time scale of turbulence:
_ vt - C k

(11)

(12)

where f~ is the model function to account for the near-wall and LRN effects originating from the rigorous physical requirements. We employ the following representation of ~ in the eddy viscosity vt and the turbulent Reynolds number Rt = k2/(ve):

e=~2v(x/-k,y) 2

(x/-k,y ~>0)

(13)

with "g - 0 at the wall (y = 0). Since the above representation is used, we should solve the transport equation for the pseudo-dissipation ~. Thus, we replace e with ~ in Eq. (4). In Eq. (11 ), because the limiting behavior of e is e ~ y0, the model function f~ has to satisfy fu e< y - l , where y is the distance from wall. Therefore, we adopt the representation similar to Abe et al. (1995b) as follows: f~= 1 + Rt---t3-~ exp [1 - fw(28)], (14)

where fw(~) is the wall reflection function (Nagano and Shimada, 1995) given in the present model by fw(~) = exp , (15)

where y* = u~y/v is the dimensionless distance from the wall based on the Kolmogorov velocity scale us = (re) 1/4. It is appropriate to discuss the argument regarding the use of the wall distance in the model functions. In this study, the wall distance at a point, y, is definitely determined by the definition as "the distance between that point and the nearest point on all the wall surfaces in a flow field" (Abe et al., 1995b, 1996). By following this definition, we can uniquely determine the wall distance y used in the model functions in any geometrical configuration and in any coordinate system. Even near the comer, it can be determined by considering the limit of a wall with a very small curvature. The distance between two points is not essentially harmful to the tensorial invariance, as is readily recognized by some expressions of the two-point-correlation equations

Y. Nagano et al./ Fluid Dynamics Research 20 (1997) 12~142

131

and general solution of the pressure-strain correlation. Furthermore, Abe et al. (1995b; 1996) have confirmed that the calculations with the AKN model, in which the wall distance is uniquely determined by the above-mentioned definition, give the identical solution when the coordinate system is rotated some degrees, e.g., 45 , in backward-facing step flows. Thus, the use of the wall distance may not violate the tensorial invariance, so long as the wall distance is determined uniquely in the problem, e.g., the nearest distance from all the wall surface as employed in the present study. 1 An important feature of the present model functions is the introduction of the Kolmogorov velocity scale, u~. = (re) U4, instead of the friction velocity u~. In adverse pressure gradient (APG) flows, as will be discussed in the later section, the friction velocity changes more rapidly in the streamwise direction than turbulent quantities such as Reynolds stresses. Therefore, in our previous study (Hattori and Nagano, 1995) the dimensionless pressure-gradient parameter P + [ = v(dP/dx)/pu3~] was adopted in order to restrain the variation of the friction velocity, leading to reasonable predictions of the APG flows. In addition, it is elucidated in this study that model functions with y*, i.e. us, work more appropriately than those with y+ for APG flow predictions even though we have excluded P+ (see Section 4.3). Abe et al. (1994) have also confirmed that in backward-facing step flows, the model functions with y* improve the prediction accuracy in the region downstream of the reattachment point, where the flow is subjected to a strong adverse pressure gradient. The above discussions, however, do not directly indicate that the Kolmogorov velocity scale is more suitable for scaling the near-wall turbulence than the friction velocity. The law of scaling the near-wall turbulence in APG flows is now investigated on the basis of the experiment of Nagano et al. (1992), the details of which will be reported in an another paper. Turbulent diffusion terms T~ and T~ in Eqs. (3) and (4) are modeled using the generalized gradient diffusion hypothesis by Daly and Harlow (1970) as follows:
Tk = C ftlruu-k t,

(16) (17)

T~ + I1~ = Gf,2~.ujul Kl,

where Cs = 1.4 and C~ = 1.4 are the model constants, and ftt = 1 +6.0fw(5) and ft2 = 1 ~-5.0/w(5 ) are the model functions working in the vicinity of the wall. Other model constants are assigned the same value as those in the AKN model, i.e., C u = 0.12, CD = 0.8, C, = 5.0, C,1 = 1.45 and C~2 = 1.9, respectively. For f,, we adopt the conventional model function: f~ = 1 - 0 . 3 exp[-(Rt/6.5)2]; and the extra production term E in Eq.(4) is E = O.Olfw(28)v(k['g)u--~UijlU~jk.

3.2. Two-equation heat transfer model

Main issue in improving a two-equation heat transfer model is to obtain a simplified form satisfying the wall limiting behavior. Under the nonuniform wall temperature conditions, kt is generally not null at the wall, and the pseudo-dissipation rate gtt = e t - 2c~(x/-~t,y) z in the NK model cannot 1 From a practical viewpoint, a further effort to develop a new model without the wall distance is now being made (Shimada and Nagano, 1996).

132

Y. Nayano et al. / Fluid Dynamics Research 20 (1997) 127-142

be applied as it is. Thus, we propose a new expression for st, with reference to Youssef et al. (1992):

where Akt = kt(x,y,z)- kt(x,O,z) = k t - ktw. With a Taylor series expansion of t near the wall, i.e., t = tw + bly + b z y 2 + O(y3), we obtain that the wall limiting behavior of st is eb 2 + O ( y ) for tw = 0 and 2~b2tw + O ( y ) for tw # 0 (see Youssef et al., 1992). It can be readily seen that Eq. (18) satisfies this requirement since gt = 0 at the wall. The turbulent diffusivity for heat can be written as

~xt = C;f ;kzm,

(19)

where C;~ = 0.1 is the model constant, rm is the hybrid (mixed) time scale proposed by Nagano and Kim (1988) and f;. is the wall reflection function. In the present model, using e in Eq.(13) and e, in Eq.(18), we write Zm with reference to some recent studies (Abe et al., 1995a; Shikazono and Kasagi, 1993) as follows: Zm=-

k{2R
e

- - + 0.5+R

(23v/2R/pr3/4 ~

R~/4

[ (Rt']2]} exp J \150J /

(20)

where R = (kt/et)/(k/e) is the time-scale ratio with e and et defined by Eqs. (13) and (18). The following wall reflection function f~ is adopted on the basis of the discussion by Cebeci (1973): f = [1 - fw(A)] 1/2 [1 - fw(B)] 1/2, (21)

where the constants A and B are set to A = 26 and B = A/Pr ~/3. Turbulent diffusion terms Tk, and T~, should be modeled. We adopt gradient-type diffusion modeling, and write Tk, and T~, as rk, = 3 k,j
,J

(22)

T~t =

~'t,j .a ,j

where ~h = 1.4 and a = 1.0 are the model constants for diffusion, ft3 = 1 + 9.0fw(5) and ft4 = 1 + 1.0fw(5) are the model functions working in the neighborhood of the wall. Other model constants and functions in the present kt-et model are assigned the standard values: Ce~ = 0.85, Cp2 = 0.64, CD1 = 1.0, CD2 0.9, fD1 fP1 : fP2 1.0 and f m = (I/CD2)(C~2f~ - 1). The extra production term E, has a form similar to the NK model: E, = ~tfw(26)7~jk.
= = :

4. Discussion on predictions with the proposed models

To assess the performance of the present k-e and kt-et models, several representative test cases are calculated. The numerical technique used are a finite-volume method in the wall-bounded flows and a Runge-Kutta method in the homogeneous shear flow. Full details of the present numerical

Y. Nagano et al./ Fluid Dynamics Research 20 (1997) 127-142

133

method of solution are given in Hattori and Nagano (1995) and Abe et al. (1995b; 1996). The computations were performed on an SGI Indigo 2 computer.

4.1. Rotating homogeneous shear flow

In this study, we propose a new model formulation including the invariants of not only the strain rate but also the vorticity tensor. It is expected that this new formulation gives more reasonable predictions for a flow field where there exist considerable differences between the strain rate and the vorticity tensor. Thus, we first applied the proposed model to the rotating homogeneous shear flow, which is a typical test case to evaluate the model performance in the above-mentioned flow field. In the rotating homogeneous shear flow, the most important phenomenon which a turbulence model should reproduce is the "bifurcation" as follows: Unsteady (developing with time) solutions are obtained under the condition of 0 ~<(2/S <<. 0.5. Steady (decaying with time) solutions are obtained under the condition of f2/S << 0 and f2/S >> 0.5. Note that S is the shear rate and f2 is the frame-rotating speed. The present model is applicable to the rotating homogeneous shear flow just by defining the non-dimensional vorticity tensor f2~ as follows:

*= (2~j

2CDz ~ j - 1.2e~jkQk ,

f2ij = f2~j -- e,jkQk,

(24)

where Ok is the angular-velocity vector of the frame rotation, i.e., ~2 = If f2k = 0, the above expression completely coincides with that in the inertial frame. This model extension follows the discussion by Gatski and Speziale (1993). The "bifurcation diagram" for the rotating homogeneous shear flow, which is the variation of the equilibrium values of c/kS versus (2/S, is shown in Fig. 1. The computational results obtained by the standard k-~ model and the GS model (Gatski and Speziale, 1993) are also included for comparison. From Fig. 1, we can readily see that the present and the GS model sufficiently predict the above-mentioned bifurcation phenomenon, whereas the standard k-e model evidently fails to predict it. This is a definite advantage.

0.4

I Present Model S t a n d a r d k E Model

0.3 ........ GS Model

0 -0.5

mm

--m.'"

* . - -

u/s

0.5

Fig. 1. Bifurcation diagram for rotating homogeneousshear flow.

134

Y.

Nagano et al./Fluid Dynamics Research 20 (1997) 12~142

4.2. Channel flow

We have calculated a fully developed channel flow. This test case is the most fundamental in actual flows, and often occurs in problems relevant to engineering. Figs. 2-5 show the present results of the channel flow calculated under both the DNS conditions of Kim et al. (1990) (Re~ = u~g/v = 395) and of Kasagi et al. (1992) (Re, = 150). From Figs. 2-4, it can be seen that the mean velocity, Reynolds shear stress and turbulent energy are quite successfully predicted for both cases. Fig. 5 shows all the terms of the turbulent energy budget. The predicted dissipation and molecular diffusion terms give the profiles slightly different from the DNS very near the wall (y+ < 3). It seems, however, that this difference does not affect the overall performance of model predictions. Next, we perform the model assessment in a fully developed channel flow with heat transfer, for which a trustworthy DNS database (Kasagi et al., 1992) (Re~ = 150 and P r = 0 . 7 1 ) is available. Comparisons of the predicted mean temperature and temperature variance with DNS are shown in Fig. 6. The model predictions are in almost perfect agreement with the DNS data. Fig. 7 shows the
. . . . . . . . ! . . . . . . . . I . . . . . .

DNS
20 0 Kasagi et al.(Re.=150)

f .t~

Predictions
- ---Present model AKN model 102

0o

10 i

y+

Fig. 2. Profiles of mean velocity in a channel.

o 10

o5

Y/~

1]

10~

10-]

10o

10 t

y+

102

0 10:]

Fig. 3. Profiles o f Reynolds shear stress in a channel.

Y. Nagano et at/Fluid Dynamics Research 20 (1997) 12~142

. . . . . . . . I . . . . . . . . I . . . . . . . .

135

DNS
0

[]

Kasagi et al.(Re.=150) Kim et al. (Re~=395)

Predictions - Present model - - - - - - AKN model

0100

101

y+

102

"

.,

103

Fig. 4. Profiles o f turbulent energy in a channel.

0.2

~
" "

DNS Prediction
- - -

t
J

< O

-0.2

~.-"
Re~=395
I
I I

T~+H~
-e
I

a 40

......
I

20

y+

60

Fig. 5. Budget o f turbulent energy in channel flow.

20

o.5
I I I I I I

u/~
I I I

O [] []

DNS(Kasagi et al.) Predictions

10

01 100

! 101

. y+

! 10 2

Fig. 6. Profiles o f mean temperature and temperature variance in a channel.

136

Y.
0.2

Nagano et at/Fluid Dynamics Research 20 (1997) 12~142

!

I Z rJ

DNS Prediction Dk, <> ~

1 l

"''-.o."

P~,

-- ---

-1

~.._Lt...~....,,

,....

,,, . . . .

,'-

~ t , zxzx h ZX ZX iX iX

~..A-"

Tkt

[]

Re~=150 -0.~ 0
I

-et
I
I

a
I

......
I

20

40

y+

60

Fig. 7. Budget of temperature variance in channel flow.

30

........

........

........

.... 20
+

Prediction (Ro=1620) DNS (Spalart, Ro=1410) O Experi _ --..

10

U + = 2.441ny + +5.0

010o

101

102 y+

10a

Fig. 8. Mean velocity profiles in boundary-layer flow (dfi/dx = 0).

predicted budget of temperature variance, compared with the DNS data. Obviously, agreement of each term with DNS is also very good.
4.3. Boundary layer flows with and without pressure gradients

We have calculated boundary layers with/without pressure gradients, i.e., with a zero pressure gradient (ZPG; dfi/cLY = 0), with an adverse pressure gradient (APG; d f i / d x > 0) and with a favorable pressure gradient (FPG; dP/dx < 0). In Fig. 8, the prediction of the mean velocity profile in the ZPG flow is presented in comparison with the experimental data of Nagano et al. (1992) (Ro = 1620) and the DNS data of Spalart (1988) ( R o = 1410). Again, the present predictions are in almost perfect agreement with the experiment, the DNS and the standard log-law profile: /7 + = 2.44 In y+ + 5.0. Predictions of flows under different pressure-gradient conditions are shown in Figs. 9-11, in comparison with the experimental data of Nagano et al. (1992) (APG, ZPG) and with the DNS data of Spalart (1986) (FPG). Note that including the additional production of e due to irrotational strains

Y. Nagano et al. / Fluid Dynamics Research 20 (1997) 12~142

40

137

20

0
+

0 100

101

102

y+

10 a

Fig. 9. Mean velocity profiles in flows with various pressure gradients.

I 4~-

I ["

Experints(Nagan~ al) Predictions 0 ZPG(Ro = 1 6 2 0 ) - - ~ - ZPG O APG(Ro=1290) - APG A P G (Ro = 1880) ..... FPG [] A P G (Ro = 2660) A P G (Re = 3350) ~/

10 0

101

10 2 !I+

[0 a

Fig. 10. Profiles of Reynolds shear stresses in flows with various pressure gradients.

is significant in calculating APG flows (Hanjali6 and Launder, 1980; Hanjalid, 1994; Nagano and Tagawa, 1990; Hattori and Nagano, 1995) as: D~
=

Dz

De + T~ + 17~ - -

CglT.~U

-- v -

"~

--( 2I A C~I-7

7.

~'2
--

v2)Ua

--

C~2f,:-:- + E .

(25)

138

Y.

Nagano et al./Fluid Dynamics Research 20 (1997) 127-142

. . . . . . . . u . . . . . . . . i . . . . . . . .

15

Experints(Nagano et al.) O ZPG (Ro = 1620) O APG (Ro = 1290) APG (Ro = 1880) rn APG (Ro = 2660)

Predictions ZPG APG nm

10

.5

100

101

102

y+

10a

Fig. 11. P r o f i l e s o f t u r b u l e n t e n e r g y in f l o w s w i t h v a r i o u s p r e s s u r e g r a d i e n t s .

Here, the velocity fluctuation components u 2 and v2 are calculated by the present nonlinear k-e model, and the model constant of C~1 = 2.5C,1 is given as in the NT model. Fig. 9 shows the mean velocity profiles. Obviously, the present model works very satisfactorily in flows with a variety of pressure gradients. Fig. 10 compares the predicted results with the experimental and DNS data of Reynolds shear stresses in APG, FPG and ZPG flows. The proposed model reproduces well the phenomena of increasing -~--~+ in the outer region along the flow on APG and decreasing one on FPG. These cannot be observed in the ZPG flow. The predictions of turbulent energy are shown in Fig. 11. Again, it can be seen that good accordance with the experimental data (Nagano et al., 1992) is achieved.

4.4. Heat transfer from uniform temperature or uniform heat-flux wall

In the following, we assess the present two-equation heat transfer model in boundary layer flows. The most basic situations encountered in engineering applications are the heat transfer from a uniform temperature or uniform heat-flux wall. The results of thermal-field calculations under the constant wall temperature or the constant wall heat-flux condition along a flat plate, are shown in Fig. 12. The present predictions indicate good agreement with the experimental data (Gibson et al., 1982; Antonia et al., 1977).

4.5. Constant wall temperature followed by adiabatic wall

The next test case for which calculations have been performed is concerned with a more complex thermal field in a boundary layer along a uniformly heated wall followed by an adiabatic wall. This case is of interest in the de-icing of aircraft wings. The same situations also occur frequently in compact heat exchangers. Fig. 13 shows a comparison of the predicted results with the experimental data (Reynolds et al., 1958) of temperature differences between the wall and the freestream A~. The proposed model gives generally good predictions for the rapidly changing thermal field.

Y. Nagano et at/Fluid Dynamics Research 20 (1997) 127-142

139

30

20

Expriments 0 T~=const. (Gibson et al.) q,j,=const. (Antonia et al.) Predictions 3~ =const.

10

100

10~

102

y+

10a

Fig. 12. Mean temperature profiles in boundary layer flows. 20 '

-o---o---o---o-~

:xperilent ( ~ynolds' et al.) t

<1 10

0,4

0.8

[m]

, 1.2

Fig. 13. Comparison o f the predicted variation of wall temperature with the measurement.

4.6. Double-pulse heat input

To further verify the effectiveness of the present model for calculating various kinds of turbulent thermal fields, comparison of the predicted variation of wall temperature ( A T w = Tw - Te; Tw and Te being the wall and free-stream temperatures) with the measurement (Reynolds et al., 1958) for double-pulse heat input case is shown in Fig. 14. Fig. 15 shows how a wall turbulent thermal layer changes when heat input is intermittent, in which AT = T - Te is normalized by the temperature difference between the wall and flee-stream AT0 = Tw0 - Te just before the first heat-input/cut-off point. It can be seen that very abrupt decrease and increase in mean fluid-temperature occur in the wall region, which is the consequence of the no heat-input condition followed by heat input, i.e., T,ylw = 0 --+ T y]w = constant. Within a short distance from the discontinuity point, the mean temperature profile becomes uniform over most of the thermal layer. Also, variations of the temperature variance are shown in Fig. 16. It can been seen that the temperature fluctuations are growing along the thermal boundary layer, then sudden heat cut-off at the wall makes the temperature variance decrease close to the wall. This phenomenon agrees with the experimental evidence obtained by Subramanian and Antonia ( 1981 ).

140

Y. Nagano et aL / Fluid Dynamics Research 20 (1997) 127-142

Prediction Heat input 15

f
f

-I I

I
>

~,
00' -~"

I 0 Exi)eriment

al.)
o oOl

~i"

0.4

0.8 z[m]

1.2

Fig. 14. Comparison o f the predicted variation o f wall temperature with the measurement (double-pulse heat input).

~. " ~ " ",~ ~

'
' ~

Predictions - . . . .

x[nl] 0,278
0.oi,

--~\\~
~

"~. ~

"\ N

..... ........

-'3

-""<
0.5

"4\\\

--

--

0.93

0.579 0.760 0.876

~ -i~-i-~~ i - .... I
101

......

$
102

....
y+

; .,.,~,~,~x,
103

100

Fig. 15. Mean temperature profiles for double-pulse heat input.

0.1 ~\ [~'~ $ ~'\ |'t~\ ~ , I Predicti:ns - . . . .

..... ........
-

II~ k " ,

x[m] I 0.278 0.517 0.579 0.760 0.876

0.937

- -

0.4

y/5

0.8

1.2

Fig. 16. Variation o f temperature variance for double-pulse heat input.

Y. Nagano et al./ Fluid Dynamics Research 20 (1997) 12~142

141

5. Conclusions
Advanced turbulence models are developed to calculate flows in the presence of pressure gradients and complex turbulent heat transfer. In the present models, we can make the apparent dissipation rates of both turbulent energy and temperature variance zero at walls, though the wall limiting behavior of turbulence in the velocity and thermal fields is reproduced exactly. Thus, the models assure computational expediency and stability. The model predictions indicate that agreement with the experiments and the DNS data is generally very good. It is also shown that the present model works very well for calculating the heat transfer under different wall thermal conditions.

Acknowledgements The authors acknowledge the financial support through a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science and Culture of Japan (No. 05240103). References
Abe, K., T. Kondoh and Y. Nagano (1994) A new turbulence model for predicting fluid flow and heat transfer in separating and reattaching flows - I. Flow field calculations, Int. J. Heat Mass Transfer 37, 139-151. Abe, K., T. Kondoh and Y. Nagano (1995a) A new turbulence model for predicting fluid flow and heat transfer in separating and reattaching flows - II. Thermal field calculations, Int. J. Heat Mass Transfer 38, 1467-1481. Abe, K., T. Kondoh and Y. Nagano (1995b) A low-Reynolds-number k-e model reflecting characteristics of a Reynolds stress model, Trans. JSME B 61, 1714 1721 (in Japanese). Abe, K., T. Kondoh and Y. Nagano (1997) On Reynolds-stress expressions and near-wall scaling parameters for predicting wall and homogeneous turbulent shear flows, Int. Heat and Fluid Flow, to appear. Antonia, R.A., H.Q. Danh and A. Prabhu (1977) Response of a turbulent boundary layer to a step change in surface heat flux, J. Fluid Mech. 80, 153 177. Cebeci, T. (1973) A model for eddy conductivity and turbulent Prandtl number, Trans. ASME, J. Heat Transfer 95, 227 234. Daly, B.J. and F.H. Harlow (1970) Transport equations in turbulence, Phys. Fluids 13, 2634-2649. Gatski, T.B. and C.G. Speziale (1993) On explicit algebraic stress models for complex turbulent flows, J. Fluid Mech. 254, 59 78. Gibson, M.M., C.A. Verriopoulos and Y. Nagano (1982) Measurements in the heated turbulent boundary layer on mildly curved surface, in: Turbulent Shear Flows 3, eds. L.J.S. Bradbury et al. (Springer, Berlin) pp. 80-89. Hanjalid, K. and B.E. Launder (1980) Sensitizing the dissipation equation to irrotational strains, Trans. ASME, J. Fluids Eng. 102, 34-40. Hanjalid, K. (1994) Advanced turbulence closure models: a view of current status and future prospects, Int. J. Heat and Fluid Flow 15, 178~03. Hattori, H. and Y. Nagano (1995) Calculation of turbulent flows with pressure gradients using a k e model, JSME Int. J. II 38, 518-524. Kasagi, N., Y. Tomita and A. Kuroda (1992) Direct numerical simulation of passive scalar field in a turbulent channel flow, Trans. ASME, J. Heat Transfer 114, 598-606. Kim, J. et al. (1990) Collaborative Testing of Turbulence Models, Data Disk No. 4. Kline, S.J., B.J. Cantwell and G.M. Lilley (1981) in: The 1980-81 AFOSRHTTM-Stanford Conf. on Complex Turbulent Flows, Comparison of Computational and Experiment I, II and III, Stanford University. Myong, H.K., and N. Kasagi (1990) A new approach to the improvement of k-e turbulence model for wall-bounded shear flows, JSME Int. J. II 33, 63 72.

142

E Nagano et al./Fluid Dynamics Research 20 (1997) 127-142

Nagano, Y. and M. Hishida (1987) Improved form of the k-e model for wall turbulent shear flows, Trans. ASME, J. Fluids Eng. 109, 156 160. Nagano, Y., and C. Kim (1988) A two-equation model for heat transport in wall turbulent shear flows, Trans. ASME, J. Heat Transfer 110, 583-589. Nagano, Y. and M. Shimada (1995) Rigorous modeling of dissipation-rate equation using direct simulations, JSME Int. J. II 38, 51-59. Nagano, Y. and M. Tagawa (1990) An improved k-e model for boundary layer flows, Trans. ASME, J. Fluids Eng. 112, 33-39. Nagano, Y.M. Tagawa and T. Tsuji (1991) An improved two-equation heat transfer model for wall turbulent shear flows, in: Proc. ASME-JSME Thermal Eng. Joint Conf. 3, eds. J. Lloyd and Y. Kurosaki, pp. 233-240. Nagano, Y., M. Tagawa and T. Tsuji (1992) Effect of adverse pressure gradients on mean flows and turbulence statistics in a boundary layer, in: Turbulent Shear Flows 8, eds. F. Durst et al. (Springer, Berlin) 7 ~ 1 . Pope, S.B. (1975) A more general effective-viscosity hypothesis, J. Fluid Mech. 72, 331-340. Reynolds, W.C., W.M. Kays and S.J. Kline (1958) NASA MEMO, 12-3-58W. Shikazono, N., and N. Kasagi (1993) Modeling Prandtl number influence on scalar transport in isotropic and sheared turbulence, in: Proc. 9th Symp. Turbulent Shear Flows, 18.3.1-18.3.6. Shimada M., and Y. Nagano (1996) Advanced two-equation turbulence model for complex flows in engineering, in: Proc. 3rd Int. Symp. Engineering Turbulence Modelling and Measurements, 111 120. Spalart, P.R. (1988) Direct simulation of a turbulent boundary layer up to R0=1410, J. Fluid Mech. 187, 61-98. Spalart, P.R. (1986) Numerical study of sink-flow boundary layers, J. Fluid Mech. 172, 307-328. Subramanian C.S., and R.A. Antonia (1981) Response of a turbulent boundary layer to a sudden decrease in wall heat flux, Int. J. Heat Mass Transfer 24, 1641-1647. Taulbee, D.B. (1992) An improved algebraic Reynolds stress model and corresponding nonlinear stress model, Phys. Fluids A 4, 2555~561. Youssef, M.S., Y. Nagano and M. Tagawa (1992) A two-equation heat transfer model for predicting turbulent thermal field under arbitrary wall thermal conditions, Int. J. Heat Mass Transfer 35, 3095-3104.