Habchi 1986
Habchi 1986
Habchi 1986
1711~172’2,1986
Printed in Great Britain
Abstract-A numerical investigation is made of laminar mixed convection of air in a vertical channel
containing a partial rectangular blockage on one channel wall. The wall containing the blockage is assumed
to be heated while the other wall is assumed to be either adiabatic (asymmetric heating) or heated
(symmetric heating). Results indicate that at low values of Gr/Re’ the maximum velocity occurs near the
adiabatic wall in the asymmetrically heated channel. As Gr/Re2 increases, this peak shifts towards the hot
wall. Reverse flow is predicted beyond the blockage. In this reverse flow region temperature variations are
small. The average Nusselt number in the blockage and the pre-blockage regions increases with decreasing
Gr/Re2 values. Beyond the blockage, the average Nusselt number decreases with Gr/Re’ at high Rayleigh
numbers. For the symmetrically heated channel, the velocity profiles are depressed at the center. For both
thermal conditions (symmetric and asymmetric), the Nusselt numbers are smaller than the corresponding
smooth duct Nusselt numbers.
1711
r b charmel width
N~MENC~TUR~
temp~ratnre
1
5 specific heat at constant pressure temperature of hot surfaces
P a~~ierat~on due to gravity temperature of inlet air
N height of the blockage dimensio~iess velocities
k thermal conductivity of fluid velocity components
L, Lf , L2 lengtk of the blockage, distance dimensional inlet velocity
before the blockage, distance after the dimensionless coordinate, s/t
blockage dimensionless coordinate, y/L
Nu,, NM, Nusselt number in tbe x and y coordinates.
directions
Nu average Nusseh number Greek symbols
p, P dimensional and d~rne~si~~Iess pressme 8 coefficient of the~al expansion
Pr Prandtl comber 0 ~mensionl~s temperature
PC? P&let number v kinematic viscosity
& Rayleigh number P density
Re Reynolds number Pn density of inlet air.
upwind profile approximations, the assumption of the overall conservation of momentum and energy is
exponential profile is associated with lower numerical satisfied to within 1% with the 56 x 46 grid.
diffusion and, is therefore, more accurate. To avoid
checkerboard fields, the velocities are stored at
RESULTS AND DISCUSSION
staggered locations. The pressure-velocity linkage is
resolved by a predictor-corrector technique. Results presented include streamline and isotherm
To determine the appropriate grid size with which patterns, dimensionless temperature and velocity pro-
grid independent solutions can be obtained, the cal- files, and average and local Nusselt number dis-
culations were done on increasingly finer grid size tributions. In the discussion that follows, results are
distributions. A 56 x46 non-uniform grid with a presented first for the asymmetrically heated channel
denser clustering near the walls was considered to followed by those of the symmetrically heated
give grid-independent results. To corroborate this, the channel.
56 x 46 grid results were compared with the solution on
an 80 x 80 non-uniform grid. The two results compare Asymmetrically heated channel (adiabatic left plate)
very well with each other with a maximum local Streamline and isotherm patterns. Figures 2 and 3
difference of 4% in the two solutions. In addition, show the streamline and the isotherm patterns for
ADIABATIC PLATE
ADIABATIC
/
PLATE I
ADIABATIC F’LATE
STREAMLINES ADIABATIC PLATE
/
HOT PCATE -0.003
(a) (b)
FIG. 2. Isotherms and streamlines for Re = 1195, H/L = 0.1 : (a) Gr/Re2 = 0.1; (b) Gr/Re’ = 1.
ISOTHERMS
STREAMLINES
ADIABATIC PLATE ADlAs/ATlC PLATE
/
(0) Lb)
FIG. 3. Isotherms and streamlines for Gr/Re’ = 1.0, H/L = 0.1 : (a) Re = 1195; (b) Re = 37.8.
1714
different parameter values~ As the Aow approaches decreasing Rey~oi~s number. In fact, at Re = 37.8,
the blockage, the streamlines are defiected towards the no boundary-layer b&vior is observed.
adiabatic wall and therefore, in the partially biocked When the height of the blockage is decreased fram
region, the streamlines are more densely packed. 0.2 to 0.1, the size ofthe recirculating eddy decreases.
Beyond the blockage (x]C > 2.0), the flow separates The comparison of the reattachment fength for the
and reattaches further downstream on to the hot @ate. two blockage heights is shown in Table 1. AIso, the
The isot~~ d~st~but~on ex~t~y shows a mono- therm& gradients near the hot waft are larger for
tonic decrease of tern~mtu~ from the hot wall to the H/L = 0. ?t_ A~~~~g~ other ~uan~t~t~ve digerenees
adiabatic surface. exist, the ~ua~ita~ve behavior of the streamfines and
The influence of buoyancy (or Grashof number) the isotherms in both cases (H/L = 0.1, H/L = 0.2)
can be observed in Fig. 2 which shows the streamline are similar.
and isotherm patterns for NIL = 0.1, & = 1195, and Temperature distrJ&ion. The temperature dis-
GrjW = 0. I and f .(1,respectively. For GrjXe2 = 0.1, tribution across the channel is presented in Fig. 4
a small eddy is observed near the bztding edge of the for different Gr/ReZ values. At x/L = 0.77, the tem-
blocknge, As GrjRe’ increases to 1, this smaHleading perature distribution exhibits a boundary-layer be=
edge eddy disappears while the recirculation region havior near the hot surface. As Gr/Re’ is increased,
behind the blockage becomes smaller. This can also the influence of natural convection becomes increas-
be seen in TabIe 1 where the reattachment lengths ingly important. Since natural convection boundary
for the various eases considers are presented. This layers are thicker compared tea forced convection
behavior is expected since, forlarge GP/.R$~the strong bonds Iayers, the ~ern~~at~e gradients at the wall
b~oy~t upBow along the vertical surface inhibits the are expec&edfy smaller and the ~rn~eratu~ profile is
si~o~theseparatededdy.In~ewofthei~c~~sing~dy more uniform at higher Gr/Re” values. Beyond the
sizes 8s Gr/Re* values become smaller, the streamlines blockage (x/L = 2.17), the presence of the recir-
beyond the blockage are shifted towards the adiabatic culating eddy distorts the temperature profile (Fig. 4).
wall as GrjRe’is decreased. This distortion is explplainedby noting that as the flow
The itiuerze of the Reynolds number {at a con- moves backwards along the hot surface and towards
stant GrjRe”) can be stxn by ~ornpa~~~ the dif&rent the blockage, it transports energy towards the interior
plots in Fig. 3. In view of the stronger motion at of the eddy and thu$ arrests the temperature decrease.
Re w 1195, the recirculating eddy is stronger and Therefore, a decrea.se in the slope of the temperature
larger in Fig. 3(a) as compared to that in Fig. 3(b) prafile is observed in the recirculating region. It may
(R$ = 37.8). Furth~~ore, and as expected, the be noted that in the recircuIating eddy, the dependence
thermal boundary layer increases in th~~k~~s with ofthe temperature profile on C&-j&?is opposite to that
Laminar mixed convection 1715
0.8
0.7
0.6
8 0.5
0.4 L
0.3
0.2
0.1
0.0 L
0.0 0. I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Y/L
FIG. 4. Temperature distribution across the asymmetrically heated channel : Re = IO’,H/L = 0.2.
observed outside the eddy, i.e. temperatures decrease adiabatic plate while at Gr/Re’ = 5.0, the maximum
with increasing Gr/Re*. This trend reverses around velocity occurs near the hot plate. At low values of
y/L = 0.25. Gr/Re’, the forced convection effects are dominant
Velocity distribution. The U-velocity profile across and the presence of the blockage causes the flow to
the channel for different Gr/Re’ values is plotted in be diverted towards the adiabatic wall resulting in a
Fig. 5 for Ra = 10’ and H/L = 0.2. At x/L = 0.77, velocity maximum near the adiabatic plate. As Gr/Re’
the maximum velocity at Gr/Re’ = 0.1 occurs near the is increased, the natural convection effect becomes
2.0
Gr/Fd=5
1.8
1.6
0.8
0.6
FIG. 5. Velocity distribution across the asymmetrically heated channel for Ra = IO’, H/L = 0.2 (main
figure) and H/L = 0.1 (inset).
1716 S. HAWXI and S. AC~.ARYA
FIG. 6. Velocity distribution across the ~y~metricaliy heated channel for different GriXe* values and
Rayleigh numbers.
increasingly important and therefore, the peak vel- is the heat transfer coefficient based on the wail to the
oeity shifts closer to the hot surface. In the blockage inlet temperature difference.
region (x/E = 1.45) the velocity levels are generally Locut NW& number. Plots for the local Nusselt
higher due to the smaller flow area. At x/L = 2.17 numbers (Nu,) along the hot plate and the blockage
(beyond the blockage), negative velocities are surface parallel to it are presented in Fig. 7. In general,
observed due to the recirculating eddy (Fig. 6). The the Nusselt number values decrease with Rayleigh
magnitude of the negative velocity increases as Gr,/Re’ number. The Nusselt number attains a maximum
increases. However, as noted eariier, the recirculation value near the channel inlet and decreases as x/L
region is smaller at higher Gr!W values. increases towards the blockage. This is expected
If the blockage height is decreased from 0.2 to 0.1, because the surface near the entrance is washed by the
the qualitative behavior of the velocity profiles cold fluid (zero dimensionless temperature). As the
remains the same. However, there are quantitative fluid flows along the hot surface, its temperature
differences. The inset of Fig. 5 presents the U-velocity increases and the heat transfer coefficient corre-
profiles at xjL = 0.71 for H/L = 0.1‘ As H/L decreases, scoldingly decreases. Along the vertical surface of
the bounce-Iayer behavior near tbe hot surface the blockage (parallel to the x-axis) the same behavior
becomes more pronounced. is observed. Beyond the blockage, the Nusselt number
The effect of changing the Rayleigh number and behavior is noticeably different due to the recir-
Reynolds number, and keeping GrlReZ constant is culating eddy in that region. The Nusselt number
shown in the inset of Fig. 6. As the Rayleigh number increases up to the point of reattachment and then
is increased, velocity boundary layers develop near decreases gradually towards the exit of the channel.
the hot side of the channel. At Ra = 103, the natural At the point of reatta~hment~ the heat flux and there-
convection effects are small and the velocity profiie fore, the Nusselt number values peak. On either side
is almost symmetrical. At Ra = 106, the effects of of this point of reattachment, the fiow is directed
buoyancy and the presence of the blockage result in outwards and, therefore, the Nusselt number (Nu,)
local peaks near the two walls with a depression in- values expectedly decrease. For Ra =L:103, the recir-
between. culation eddy is very small and weak. Therefore,
~~~e~~ ~~~~~. To present the heat transfer data, unlike the other Rayleigh numbers, a monotonic
the Nusselt number is calculated along the heated increase towards the channel exit is observed.
surfaces using the following definition. The influence of Gr/Re’ should also be noted. For
the higher Rayleigh rmmbers, as Gr/Re2increases, the
MAL ==hL/k (13) strength of the recirculating eddy increases leading to
where
higher Nu, values (Fig. 7). Since the size of the eddy
is reduced at higher Gr/Re”, Nu, peaks at a smaller x/f,
Laminar mixed convection 1717
60
40
t
N”x
0
30 00 0.4 06 1.2 16 20 2.4 26 3.2 36 40
X/L
0.0 0.4 0.6 1.2 1.6 2.0 2.4 2.6 3.2 3.6 4.0
X/L
FIG. 7 Nusselt number distribution along the hot wall for the unsymmetrically heated channel, H/L = 0.2.
value. Between x/L = 0 and x/L = 2, Nu, decreases (Fig. 8), due to the increase in the recirculating eddy
as Gr/Re* increases. This is because the temperature strength with Gr/Re’. At the lower Rayleigh number
profiles are more uniform at higher Gr/Re2 values (Ra = 103), the recirculating eddy beyond the block-
resulting in lower heat transfer coefficients. age is very small and weak and the values of Nu,
The Nusselt number along the horizontal surfaces decrease as Gr/Re2 increases.
of the blockage (Nu,) is presented in Fig. 8. It should Average Nusselt number. The average Nusselt
be noted that Nu, at x/L = 1 (inset of Fig. 8), is of number along a length AS is defined as
the same order of magnitude as Nu, while Nu, at
x/L = 2 is much smaller than Nu, at x/L = 1 due, in s+As
part, to the fluid being warmer at x/L = 2 and there-
fore less amenable to heat transfer.
N”=rs
- ‘S s
Nu(s) ds (15)
At a higher Rayleigh number, Nu,, along the upper where s represents either x or y. The average Nusselt
surface of the blockage increases as Gr/Re’ increases number along the vertical surfaces and along the
IO 45
40
9
35
6 30
25
7 N"Y
20
6 15
IO
N”Y
5
5
0
4 006 012 0.16
Y/L
3
H/L=0.2
x/L=2.0
2
LI
FIG. 8. Local Nusselt number along the lower (x/L = 1.0) and upper (x/L = 2.0) surface of the blockage.
1718 S. J&mm and S. ACRARVA
Ra
to3 104 IO5 106
Rc. 9. Average NnsseIt number distributions along the wtical surfaces (s&d lines) and along the
horizontal surfaces of the blockage (dotted lines), H/L = 0.2.
upper and lower s~r%xs of the blockage are pre- This result is s~nle~h~~ u~~~~t~ and is due, in part,
sented in Fig. 9 for ff/L ‘= 0.2. With increasing Ray tr, the near-isothermal recircularion region behind the
Ieigh number, the strength of the &XVfiefd and, there- blockage which, as noted earlieer, is characterized by
fore, the average Nusselt number, atong the vertical low heat transfer rates.
surfaces increases. Between X/L = 0 and x/t = 2, the
average Nussclt nur&r increases a5 @/Be’ decreases.
However, above the blockage ⅈ > 2) and at high in describing the results for the syrnrnet~~a~~~
Rayleigh numbers, the average Nusseft number heated channel, the discussion is limited primarily
increases with increasing Gr{Re2, 70 explain this, it to the observed similarities and diReren@% witi the
should be noted that at the high Rayleigh nttmhers, results of the asymmet~~ally heated channel.
as &r/8&’ decreases the size oE the recirculation Te~pern~slre distribution.The texnperature pr~Eiles
beyond the blockage increases (Fig. 2) and, since the across the channel for Ra = IO’, E$/L = 0.2, and the
recirculation region is associated with low heat trans- various Gr/Re’ values are shown in Fig. 10. At x/t “-
fer rates, the average heat transfer or Nussett number 0.77, it is observed t&t the t~~perat~~ exhibits
from the: $ate is therefore expected to deerease with a botmdary-layer hehatior near both plates. How-
decreasing Gr/&‘. At iow Ra and Re, the recirculation ever, the temperature values are generally higher near
zone is very small and its effects are not significant, y/L = 0, due to the higher heat transfer area. Beyond
utemx, the HK @ends fcdfow &me aiong 0 G xJL G 2, the bto&kage (at x/L = 2, I?), the ~rn~ra~~ pro&e
i.6 decreasing Nrc with increasing C&+/I@. is distorted, as for the asymmet&a& heated charmel,
The average Nusselt number along the upper and due to the recirculating eddy.
lower surfaces of the blockage is shown by dotted Y&ci?y &sirthution. The U-wlocity profifes across
tines in Fig. 9. At the lower stIrface (X/L = 1j, the the channel are pre=ted in Fig. f 1 far Rra= IF,
average Nusselt number increases with increasing HjL = 02, and the various GqRe” values. At
Rayleigh number and decreasing Er/Re’. On the other x/L = 0.77, it is observed that the velocity peaks Itear
hand, at the upper surface (x/L = 2), the dependence l3ofh plates with the maximum occurring near @mei&t
of the average Nusselt number on GrjRe’ reverses at
higher Rayleigh numbers due to the effect of recir-
culating eddy beyond the blockage.
Table 2 shows the overall average Nusselt rmmber
values for a blocked vertical channel (H/L = 0.2) and
a smooth. vertical channel (H!L if Oj at J&2= 103.
Results for the smooth channel were obtained using
a parabolic finite-difference procedure of Patankar
and Spafding 1121.It is noted that the average Nusseit
number values are higher for the smooth channel. -..-.l.” __-_elllil
Laminar mixed convection 1719
I.0
0.9
0.8
0.7
plate (y/L = 1.O). The presence of the blockage causes observed in the ~y~et~~al~y heated channel. Nega-
the flow to be deflected towards the left plate causing tive velocities are once again observed (Fig. 12) beyond
the velocity to attain its maximum in that region. With the blockage due to the recirculating eddy.
increasing x/L the velocity near both walls increase The effect of the Rayleigh number can be seen in
and since the channel flow rate is constant, this results Fig. 12. As expected, at a high Rayleigh number
inaconcavityintheprofileinthecenter.Themagnitude (Ra = 106), natural convection effects near the wall
of this concavity increases as the flow moves down- are dominant and a large depression in the velocity
stream as seen in the inset of Fig. 12 (x/L = 2.17).This profile is observed around the center line. On the
significant depression in the velocity profile is not other hand, at a low Rayleigh number (Ra = 103), the
1.0
U/U, 0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0. I
no
_._ I I
Fro. 12. Velocity distribution across the symmetrically heated channel at x/L = 2.17.
v&&y profile is more uniform and no concavity is the entrance of the channel and decreases as the Aow
observed. approaches the exit. As Cr/Re2 increases, the Nusselt
NW& tzupnbers. Tbe general behavior, as well as numbers decrease. The inset of Fig. 13 shows the
the effect of the governing parameters on the local local Nusselt number aiong the kft plate for both the
Nusselt number distribution along the right side of blocked channel (H/L = 0.2) and the smooth channel
the channel are the same as observed in the asym- (H/L = 0). As for the asymmetrically heated channel,
metrically heated charm&i, but the magnitudes of Nu, the values of Nu, are higher for the smooth channel.
are lower by about 10%. The local Nusselt number Figure 14 shows that the average Nusselt number
profiles atong the left plate of the heated channel are along the right plate is slightly tower for the sym-
presented in Fig. 13 far H/L = 0.2. Expectedly, the metrically heated channel case. This is expected
local Nusselt number attains a maximum value near because the thermal boundary layer along the left
FIG. IS. Nusselt number distribution afonp the smooth plate of the s~~et~ca~ly heated channel.
Laminar mixed convection 1721
16
- Asymmetric Healinq
15
- - - Symmetric Haatinq
14
13
I2
FIG. 14. Overall average Nusselt number along the hot wall containing the blockage (H/L = 0.2)
R&sum&-On etudie numeriquement la convection mixte laminaire d’air dans un canal vertical contenant
un blocage partiel rectangulaire sur une paroi d’un canal; cette paroi est supposie chauffee tandis que
l’autre est soit adiabatique (chauffage asymetrique) ou chauffee (chauffage symetrique). Les resultats
indiquent qu’aux faibles valeurs de Gr/Re*, le maximum de vitesse se produit pres de la paroi adiabatique
dans le cas de chauffage asymetrique. Lorsque Gr/Re’ augmente ce pit se d&place vers la paroi chaude. Un
Ccoulement de retour est derriere le blocage. Dans cette region les variations de temperature sont petites.
Les nombres de Nusselt moyen dans le blocage et avant le blocage augmentent quand diminue Gr/Re*.
Au-dessus du blocage, le nombre de Nusselt moyen diminue avec Gr/Re’ aux grands nombres de Rayleigh.
Pour le canal chat& symetriquement, les profils de vitesse sont deprimes au centre. Pour les deux conditions
thermiques (symttrique et asymitrique), les nombres de Nusselt sont plus petits que ceux relatifs au canal
lisse correspondant.
Zusammenfassung-Es wurde eine numerische Untersuchung der laminaren gemischten Konvektion von
Luft in einem vertikalen Kanal, der durch ein rechteckiges Hindernis auf einer Kanalwand teilweise
versperrt ist, durchgefiihrt. Es wurde angenommen, daBdie Wand mit dem Hindernis beheizt wird, wahrend
die andere Wand entweder adiabat (asymmetrische Beheizung) oder beheizt (symmetrische Beheizung) sein
~011. Die Ergebnisse zeigen, da13 fur kleine Werte von Gr/Re2 die maximale Geschwindigkeit im asym-
metrisch beheizten Kanal nahe der adiabaten Wand auftritt. Fur zunehmende Gr/Re*-Werte verschiebt
sich das Maximum in Richtung auf die beige Wand. Jenseits des Hindernisses ergibt sich eine Umkehrung
der Striimung. In dem Gebiet der Rdckwartsstromung sind die Temperaturunterschiede sehr gering. Die
mittlere Nusselt-Zahl in den Gebieten vor und innerhalb des Hindernisses steigt mit abnehmenden Gr/Re’-
Werten. Jenseits des Hindernisses nimmt die mittlere Nusselt-Zahl mit Gr/Re* fur groDe Rayleigh-Zahlen
ab. Beim symmetrisch beheizten Kanal sind die Geschwindigkeitsprofle in der Mitte abgeflacht. Fiir beide
thermischen Bedingungen (symmetrisch und asymmetrisch) sind die Nusselt-Zahlen kleiner als die fiir
einen entsprechenden glatten Kanal.
A~OTa~-~pO~eAeHOq~~AeHHoe~ccneAoaaH~e JIaMuHapHoiicMe~aHHo~KoH~xu~~BO3AyXaB~ep-
T5iKanbHoMKaHaAe,K 0~~0SicTeH~e~0~0p0f0np~K~nAe~binp~~oyronb~b~eneperopoAK~.~peA~oAa-
raeTcr, '~TO cTeHKa c neperoponrtawf isarpesaeTcn, B TO Bpehfir KaK Apyrar C'fBTaeTcK ~1~460
aA~a6aT~qeCKO~ (ac~MMeTp~YH~~ Harpen), na60 narperoii (cuMMerpesnbtii narpee). Pesynbrarbi noza-
3btBa~T,qTO~p~MaA~X3HaqeH~KX Gr/Pe'MaKCEfMyMCKOpOCfB BMeeTMeCTO B611B3H aA5ia6aTwszcKo2i
cTeHKU ac~MMeTp~~H0 xarpeeaebforo KaHaAa. C ysena~emfek4 Gr/Pe’ 3~oT ,waKceMyM cMewaeTcK 3
Ha~paB~eH~~ ropnqel crerimf. Ilpencxa3ano cy.mecraonanwe o6pareoro re~erinx 38 neperoponxoir. B
3oHe o6parnoro Te'teHWKB3MeHeHHK TeMnepaTypbi hianbI.CpeAHee wicno HyCC%?nbTa Ha neperopoAKe
B B 30Hax nepen Hei? pacTeT c yMeHbLUeHHeM 3HaveHaii Gr/Pe’. 3a neperopoAKoii CpeAHee SWCJIO
HyCCeJIbTa yMeHbmaercn c Gr/Pe2 npn 6onbnmx rncnax P3nea. Ana cnyran ceMhfeTpa%roro narpe-
BaeMoro Kanana npo@inn CKO~~CT~ CxoiMarorca B uearpe. Ana 06oax rennoabrx ~~~HMOB (CNhiMe-
TpE,YHO$-0H H~~MMeTp~qHOrO) 9HCAa HyCCeAbTa MeHbLUe COOTBeTCTBy~~~X 98CeA AAIi rAaAKOf0
rcaiiana.