Академический Документы
Профессиональный Документы
Культура Документы
Introduction unity or greater, provides local heat transfer coefficients in the at-
tached flow regions. Empirical information from cavities and steps
The purpose of this paper is to present an analytical model for [12-16] supplies coefficients in the separated flow areas.
the flow over a rough surface. This model complements the work of Attention is restricted to the steady, incompressible, turbulent
Kays [I], 1 Kacker [2], Lavalee and Popovitch [3], and Perry, et al. flow of a constant property, single-phase, Newtonian fluid in a
[4]. It fills the gap between the wholly empirical treatment of channel of constant cross section with rough walls. Any curvature
rough surfaces [5, 6] and the extensive investigations on single ele- of the channel is small and a constant mean heat-flux is applied at
ments [7]. The analysis provides a quick and simple means of eval- the surfaces. Extensions of the simple model to other roughness
uating or optimizing rough surfaces [8], it should help to avoid un- shapes and to other flow situations, and a fuller development of
necessary and costly experiments [9], and it will provide a frame- the equations presented here, may be found elsewhere [17].
work for more sophisticated methods.
A simple surface—equally spaced, rectangular ribs—is consid- T h e Physical Nature of the Flow Over a Rough
ered and the flow is represented by a series of attached and sepa- Surface and an Approximate Flow Model
rated flow regions. For widely spaced roughness the flow reattach- We consider a fully developed channel flow and ribs with h « D.
es between the elements, but for closely spaced roughness the flow The flow may be divided into two layers; one of thickness 5, which
does not reattach in the gaps and each element shields the one is of order h and which envelopes the roughness elements, and the
downstream. Wall shear stresses for the attached flows are deter- other the core between the roughness elements and the channel
mined from an assumed velocity profile, a logarithmic law-of-the- center line. We are concerned solely with the former layer. T h e
wall [10] for a smooth surface, which also provides the dynamic basic pattern of the flow over the ribs is illustrated by Pig. 1(a).
pressure for the form drag [11] of each element. This dynamic The channel Reynolds number is assumed to be high enough and
pressure is reduced by the shielding effect of closely spaced ele- the ribs widely spaced such that the separated flow regions shown
ments. A modified Reynolds analogy [10], which restricts the heat- are formed. Separation and reattachment imply that the shear
transfer analysis to fluids with Pr (see Nomenclature) of order stress, wall temperature, and velocity profiles are cyclic functions
of z in the vicinity of the roughness.
To simplify the problem the physical flow in Fig. 1(a) is re-
placed by the model of Fig. 1(b), where four regions 1, 2, 3, and 4
1
are indicated. The separation "bubble" on top of each rib in region
Numbers in brackets designate References at end of paper. 1 is now completely ignored. Separated flow regions are specified
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer by characteristic separation lengths C3 and e 4 . The flow is assumed
Division May 29,1974. Paper No. 75-HT-JJ. to remain attached to the top of the rib giving a mean wall stress
TI, over b. Similarly T2 is defined for region 2. Both TI and r<i are
smaller than the total average wall shear stress rw which contains
contributions from the form drag of each element. ; u(6)
i ^
As the pitch of the ribs is reduced the flow will no longer reat- 8
!© ©
tach between the ribs and a cavity will be formed. This occurs ! +y
when p — 6 5 C3 + C4. The dynamic pressure acting on the front-
face of each element is reduced, as well as the area over which this
pressure acts, by the shielding from each upstream element. Flow
777Y. bmjraa^
Basic Analysis, p - b > c 3 + c 4 . A force balance over a length small compared with Df. Contributions to TW from the shear stress-
L of channel of unit width containing a large number of ribs gives es in the separated flow regions are assumed to be incorporated in
Ajlu," = SArf!" + S A ^ z " + SAtf3" + W / 4 " uSA* = Alny* + BSA = ufiyi) = u2*(y2) = u3%y3) (11)
+ T,AFqs" + ^AFqR" (5)
with the origin for u 3 taken from y 3 = k.
where q{' are mean wall heat transfer rates per unit area in regions To determine BSA we recall that BSM is defined in terms of a
1, 2, 3, 4, and over the front S and rear R faces. Defining the tem- laminar sublayer thickness ISM^ i~ 11.63)
perature T+(6) = [T,„ - T(S)]pCurJqw", where Tw - T(S) is the
temperature difference a t y i = <5 a n d y 2 = <5, and introducing T+(<5) Bstt - ^sti Alnlsil ~ 5,5 (12)
into equation (5) gives
and, by analogy,
pCuT /T + (5) = afi/p + ot2(p - 6 - c 3 - cA)/p + a,c 3 //> B
SA = lSA*-AlnlSA* (13)
+ afjp + ciji/p + aRh/p (6)
+
where ISA is representative of the laminar sublayer thickness in
where m is an average heat-transfer coefficient for each region, de- regions 1 and 2. We assume that any increase in local turbulence
fined by at = qi"l[Tw - T(<5)], and the convective transport intensity, generated by the separated flows, is reflected by a de-
through the ends of the control volume is neglected. crease in ISA+ and that this decrease is given by
T+(<5) is a parameter suitable for characterizing the average
heat-transfer properties of the whole rough surface and may readi- hA* = hn\SA/Urw = W«W«s/(5) (14)
ly be related to the roughness function g(h+, Pr) [18]. The a; are
determined from empirical relations for cavities and steps. T h e H e a t T r a n s f e r C o e f f i c i e n t s «,•
Employing the modified Reynolds analogy [10] to the layer be-
M o d i f i c a t i o n s f o r p - b < c3 + c 4 tween ISA+ and 8 in each region 1 and 2, we obtain
We extend the foregoing equations to this case by simply assum-
ing that each element shields its downstream neighbor, reducing
a , = a 2 =pCuT u*(5)/{Pvt[us/{5)f
its effective frontal area. Also, because the flow outside the sepa- IV
rated flow region 3 is a shear layer, the dynamic pressure acting on + Zs/Ms/(6)[PrM~Prt]} (15)
the effective area is reduced. For the flow field of Fig. 1(c), the ef-
fective height he of each element is he = (h - k), and u now takes For regions 3, 4, and for R and S, we use Seban's [12-14] experi-
the form: mental information for rectangular cavities, which is summarized
2
in Fig. 3. The variation of etc in the cavity is given by
(/, - k)u = / ' " Vrfj-3 (7)
where y 3 is measured from the base of the cavity and uz is the pro- «c s Q«"/{T« - Ta) = iu„x/vnm/X)iPr/o.nr (ie)
file in region 3. Strictly, a separated flow profile should be em-
where, following Presser [16], the Prandtl number is introduced to
ployed for U3, but we retain simplicity by assuming us = u% with
generalize Seban's results in air. [/», X, and Ta define a flat plate
U3 = 0 for 0 < yz < k. Equations (4) and (6) now take the form:
heat-transfer coefficient upstream of the cavity itself, and the vari-
[H*(5)]-2 = [«, + (6)]-V/> + CDp\uVu(6)]Vij2p (8) ation of ac is determined from the given variation of W. From Se-
ban's information it is difficult to estimate the effect on ac of the
and
cavity length d = (p — b). We infer this effect from the results of
pCuryT*(5) = afi/p + a3c3he/hp + otiCihe/hp Charwat, et al. [15], where a linear approximation gives ac a (0.5 +
0.5d/c) until the flow reattaches in the cavity, that is, until p — b>
+ ashjp + aRh/p (9) C3 + C4. We have anticipated the result that ac is a maximum when
with he = (h - k) = hip - fe)/(c3 + c 4 ). The contribution to the d = c. This, in fact, contradicts Charwat's measurements for d > c
heat transfer of the frontal area between 0 < y3 ^ k is assumed and is a weakness of the present model in which region 2 is sud-
negligible. denly introduced. Combining Charwat's and Seban's results gives
h/b =1 p/h = 6
a3= Ui = 0.025(fe/o)(M(6)6/i;]"(Pr/0.71)" , (0.5 + 0 . 5 d / e )
(18) CD= 1.2
12
" X\
c/h= 8.5 -
^ *xtun,
0,03
1
AVERAGE W
^ $ ^ r\
CD and c/h are expected to be functions of some local Reynolds
number and, possibly, of the rib width [10], but we assume here
that they are constants. The constants A, BSM, and ISM+ are cho-
sen 2.5, 5.5, and 11.63, respectively.
0,02
Predicted values of R are compared with some representative
measurements [21] from an annulus test rig in Fig. 4. This compar-
-
0,01
-A 1jA 1
CAVITY FLOOR D' D
ison is typical of the discrepancy between theory and experiment
with good agreement at high values of h+, but too high values of R
at lower values of h+. Correct trends are shown: R increasing with
decreasing h+ for closely spaced elements but decreasing with h+
for widely spaced elements. The reason for this behavior is not
clear. To give some idea of the discrepancies over a very wide range
of roughness shapes a comparison is made in Fig. 5 between pre-
dicted values of R and a correlation of experimental information
[22]. Considering the approximations in the theory and the wide
scatter (not indicated on the correlation) in the experiments the
Fig. 3 Heat transfer distribution in a shallow rectangular cavity 2 < d/h <
5, after Seban, ei al. [12-14] agreement is very reasonable. The theory indicates R ^ R(b) for p
^ -
6
cC\Nx\\ „
120
b v \ \ \ \ \ y^X
100
4
80
i
**""~~- *-**'
<;>
60
40 a = 11.63 P r 2 / 3 _ 2 . 5 Prt[17
mk^m
CD=1.2
20 c/h = 8.5
REYNOLDS N0=sl0 5
30 p/h h/b THEORY C D c/h 4£L_ THEORY
i#SSS 10 1-4 1.2 8.5 r
SM H^j EXPERIMENT '////////,[ 261
7" y////,. 20 2 1.2 8.5 h/b = 1
llll 40 2 1.2 8.5
Pr = 0.71 p/h
20
DATA FROM [251
m h h/b
l«4in
,r \
10
i_L
AA
i I
A A'
ii
D" D A
: FLOW v y FLOW
g= 0 6 3 P r 2 / 3 - 2.5 Pr t [171 A it. D A
— h+
-J 1 I
V77777777777777777777T/. 3 O T XMr #7/7777//' W77
10 100 A' D'
Fig. 8 Comparison with experiment—local heat-transfer coefficient ratios
Fig. 6 Comparison with experiment, rectangular rib over rectangular ribs