You are on page 1of 9

International Journal of Heat and Mass Transfer 54 (2011) 1563–1571

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/ijhmt

Unsteady natural convection heat transfer from a heated horizontal circular


cylinder to its air-filled coaxial triangular enclosure
Zi-Tao Yu a,⇑, Xu Xu b, Ya-Cai Hu a, Li-Wu Fan c,⇑, Ke-Fa Cen d
a
Institute of Thermal Science and Power Systems, Department of Energy Engineering, Zhejiang University, Hangzhou 310027, PR China
b
College of Metrological and Measurement Engineering, China Jiliang University, Hangzhou 310018, PR China
c
Department of Mechanical Engineering, Auburn University, Auburn, AL 36849-5341, USA
d
State Key Laboratory of Clean Energy Utilization, Zhejiang University, Hangzhou 310027, PR China

a r t i c l e i n f o a b s t r a c t

Article history: Unsteady natural convection heat transfer in a horizontal annular region bounded by a heated inner cir-
Received 4 August 2010 cular cylinder and a coaxial outer triangular cylinder is numerically studied for a wide range of Grashof
Received in revised form 28 October 2010 numbers, aspect ratios, and inclination angles of the triangular enclosure. Different phases are identified
Accepted 28 October 2010
during the course of flow development through the evolutions of the average Nusselt number over the
Available online 20 December 2010
inner circular wall. Snapshots of streamlines and isotherms for two typical cases are presented to exhibit
identification among these phases. The flow development time and time-averaged Nusselt number over
Keywords:
the inner circular wall are predicted and scaled with Grashof number. Additionally, the onset and evolu-
Numerical simulation
Unsteady natural convection
tion of pitchfork bifurcation are quantitatively investigated.
Heat transfer Ó 2010 Elsevier Ltd. All rights reserved.
Triangular cylinder
Circular cylinder

1. Introduction et al. [6] reported their numerical results on transient natural


convection in horizontal isosceles triangular enclosures for
Natural convection has been the subject of growing interest in 103 6 Gr 6 108. Thereafter, unsteady natural convection in a trian-
the literature with numerous applications to thermal-fluid engi- gular enclosure induced by absorption of radiation was investi-
neering, e.g. heat transfer in tanks for solar energy storage [1,2] gated by Lei and Patterson [7] and was recently revisited by Mao
and for storage of liquid natural gas [3] and in attics of buildings et al. [8] using improved scaling analysis. Lei and Patterson [9] also
[4]. The Boussinesq approximation for buoyancy has successfully studied unsteady natural convection in a triangular enclosure in-
been used in analytical and numerical studies of natural convec- duced by surface cooling, which was then extended by Mao et al.
tion. By applying this approximation, two dimensionless variables [10] to explore natural convection driven by surface cooling in a
are present in the non-dimensionalized governing equations, i.e. near-shore region. In addition, Lei et al. [11] investigated experi-
the Grashof (or Rayleigh) and Prandtl numbers. Parametric studies mentally and numerically unsteady natural convection in a
on Grashof and Prandtl-number dependence of steady-state and water-filled isosceles triangular enclosure that is heated on the
unsteady features of natural convection have been presented over bottom and cooled symmetrically from the inclined side walls. Fur-
a wide range of both numbers. Numerical and experimental results thermore, a reservoir model is also able to be represented by a hor-
of natural convection in enclosures have also been extensively doc- izontal triangular enclosure. Experimental and numerical results
umented for various geometries and boundary conditions. A brief on unsteady natural convection in such model were presented by
literature survey of unsteady natural convection heat transfer in Bednarz et al. in a series of papers for a variety of boundary condi-
enclosures was given by Yu et al. [5]. tions, i.e. cooling from the water surface with a constant tempera-
In practice, unsteady natural convection in horizontal enclo- ture [12] and with a constant heat flux [13], periodic heating and
sures that have a triangular cross-section has attracted consider- cooling [14], and diurnal temperature variations [15].
able attention. To the best of the authors’ knowledge, pioneering Another cross-sectional geometry that is of practical signifi-
work of this subject goes back to late 1980s when Karyakin cance is concentric or eccentric annulus bounded by circular or
non-circular cylinders. Investigations of unsteady natural convec-
⇑ Corresponding authors. Tel./fax: +86 571 87952378 (Z.-T. Yu); tel.: +1 334 844 tion between horizontal concentric cylinders were conducted
8386 (L.-W. Fan). [16–20] for different aspect ratios and various types of boundary
E-mail addresses: yuzitao@zju.edu.cn (Z.-T. Yu), liwufan@gmail.com (L.-W. Fan). conditions. Usually, by assuming that the flow behavior does not

0017-9310/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2010.11.032
1564 Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571

Nomenclature

AR aspect ratio, AR = Rout/Rin Greek symbols


g gravitational acceleration (m/s2) a thermal diffusivity (m2/s)
Gr Grashof number b thermal expansion coefficient (1/K)
k thermal conductivity (W/mK) h inclination angle
l length (m) m kinematic viscosity (m2/s)
L characteristic length, L = Rout  Rin (m)
Nu Nusselt number, defined by Eq. (1) Subscripts
Nu  average Nusselt number, defined by Eq. (2) dev development
Nu time-averaged Nusselt number, defined by Eq. (3) diff difference
Pr Prandtl number in inner circular cylinder
q local heat flux (W/m2) left left side wall
R radius of circle (m) out outer triangular cylinder
R2 coefficient of determination right right side wall
t time (s) total total length/duration
T temperature (K)
x, y Cartesian coordinates

vary in the axial direction, a two-dimensional cross-section was tion. Initially, the bottom of the triangle is horizontally positioned
considered to simplify the problem. However, in order to take so that the angle formed by the segment connected between the
the end effects into consideration, Vafai and Ettefagh [21] carried apex and the center of the triangle and the positive y-axis is equal
out a three-dimensional study on transient natural convection heat to zero, i.e. h = 0°. When the triangle is turned about its center in
transfer in a closed horizontal annulus. Borjini et al. [22] paid spe- clockwise direction, this inclination angle will increase, as shown
cial attention to unsteady natural convection in a horizontal annu- in Fig. 1. The entire physical domain is initially at a uniform tem-
lus due to radiation. When the annulus is filled with a porous perature Tout = 300 K. At t = 0 s, a temperature Tin = 320 K that is
medium, the resulting modification on transient features of natural higher than Tout is applied suddenly on the inner wall, and this
convection has also been of interest in the literature [23–25]. Mod- temperature is maintained for t > 0 s. Hydrodynamically, no-slip
ification of unsteady natural convection due to the presence of an boundary conditions are adopted on both walls.
adiabatic baffle and an inner body within an enclosure was numer- The governing equations and definitions of dimensionless num-
ically studied by Fu et al. [26] and Ha et al. [27,28], respectively. bers for the present model are the same as those for the physical
Additionally, periodic thermal boundary conditions, which are ex- model considered by Yu et al. [5]. The properties of air are assumed
pected to have significant influence on unsteady natural convec- to be constant and evaluated at 310 K. As a result, the Prandtl num-
tion, were considered by many researchers, e.g. Lakhal et al. [29] ber (Pr = m/a) is a constant in this study (Pr = 0.74) and the Grashof
and Abourida et al. [30]. number (Gr = gb(Tin  Tout)L3/m2) is only able to be varied by chang-
As a combination of triangular and circular cylinders, transient ing the characteristic length. The variables in the definitions of
natural convection heat transfer from a heated triangular cylinder these two dimensionless numbers are defined in the Nomenclature.
to its coaxial cylindrical enclosure, filled with either air or a low-
Prandtl-number fluid, was numerically studied by Yu et al. [5,31].
On the other hand, the shapes of the inner and outer cylinders
may be exchanged. Steady-state natural convective heat transfer
around a horizontal cylinder inside a concentric triangular enclo-
sure for Pr = 0.71 and 0.01 < Pr < 10,000 was studied by Xu et al.
[32] and Yu et al. [33], respectively. Such geometric configuration
might have potential applications in nuclear reactors, thermal stor-
age systems, and thermal management/cooling of electronic de-
vices. In the literature, however, there is a lack of effort devoted
to studying unsteady features of natural convection heat transfer
in such geometry. A numerical study is therefore conducted to
investigate unsteady laminar natural convection heat transfer in
an air-filled horizontal triangular enclosure that contains a coaxial
inner circular cylinder.

2. Problem formulation

As schematically shown in Fig. 1, the two-dimensional cross-


section, i.e. an annular region bounded by an outer equilateral tri-
angle and an inner circle, is considered. The centers of the circum-
scribed circle of the triangle and the inner circle coincide with each
other. The aspect ratio and characteristic length of the model are
defined as the ratio and gap between the outer and inner radii,
i.e. AR = Rout/Rin and L = Rout  Rin, respectively. A Cartesian coordi-
nate system is adopted in such a way that the origin coincides with
the center of the circles and gravity acts along the negative y-direc- Fig. 1. Schematic diagram of the physical model.
Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571 1565

In order to perform a parametric study, the Grashof number is time-step intervals is only 0.01% when the total number of cells
varied from 105 to 107, ensuring that the flow will be laminar and is 10,000. For the time-step interval of 0.1 s, increasing the number
dominated by convection. In addition, three values are considered of cells from 6000 to 10,000 decreases the time-averaged Nusselt
for the aspect ratio (AR = 3, 4, and 5) and inclination angle (h = 0°, number by about 2%. In addition, as the time-step interval is low-
30°, and 60°) of the triangular enclosure. ered to 0.05 s, increasing the number of cells from 10,000 to 16,000
decreases the measure by less than 1.5%. This indicates that for the
3. Numerical solution examined case, even the coarsest temporal and spatial resolutions
are able to yield reasonable accuracy as compared to the finest res-
3.1. Computational details and code validation olutions. When the total number of cells is greater than 6000 and
the time-step size is smaller than 0.2 s, further improvement on
The computations are performed using the commercial CFD temporal and spatial resolutions would not result in desirable gain
code ANSYS Fluent 6.3, which is based on the finite volume ap- in accuracy and the numerical results become nearly independent
proach. The computational domain is created and then divided into of the grid and time-step intervals. On the other hand, if the
control volumes and the boundary conditions are specified using temporal and spatial resolutions are greatly improved, then the
the preprocessing code Gambit 2.3. In order to capture precisely computational expenses will be dramatically increased as well.
the flow and heat transfer characteristics next to the walls, thin Therefore, trade-off between numerical accuracy and computa-
boundary layers that consist of fine structured quadrilateral cells tional cost suggests using intermediate temporal and spatial reso-
are generated. The rest of the computational domain is then lutions for all the computations in this study, i.e. a total number of
meshed using unstructured triangular cells. around 10,000 cells and a time-step size of 0.1 s. Since unstruc-
The second-order implicit scheme is utilized for discretizing the tured cells are used, the total number of cells for each case is not
temporal terms in the governing equations and spatially, the sec- exactly 10,000 and slightly varies.
ond-order upwind scheme is adopted. Pressure–velocity coupling
is implemented using the SIMPLE algorithm and the convergence 4. Results and discussion
criterion is set to be 106 for each primitive variable. For all the
cases studied, the computations are carried out up to a physical 4.1. Evolutions of average Nusselt number
time of 200 s. As pointed out by previous studies [5,31], this phys-
ical duration is long enough to capture the overall transient char- Based on the characteristic length L, the local Nusselt number
acteristics of the evolution of natural convection. along a wall is defined as
The commercial code ANSYS Fluent has been widely used by
many researchers for solving unsteady natural convection prob- qL
Nu ¼ ; ð1Þ
lems and was recently validated by Xu et al. [32] for investigating kðT in  T out Þ
steady-state natural convection heat transfer in the same geometry
where k is the thermal conductivity and q the local heat flux along
under consideration. Therefore, a validation test of the code is not
the wall. The average Nusselt number over the wall is then evalu-
required for the present study.
ated by
3.2. Independence test for grid and time-step intervals Z ltotal
1
Nu ¼ Nudl; ð2Þ
ltotal 0
For computations of an unsteady process, the numerical accu-
racy strongly depends on both grid and time-step intervals. In or- with ltotal being the length or perimeter of the wall of interest. Addi-
der to determine proper temporal and spatial resolutions, an tionally, the time-averaged Nusselt number is predicted over a per-
independence test is carried out for a duration of 100 s for the case iod, given by
of AR = 3, h = 0°, and Gr = 107, which is the highest Grashof number Z
  t total
under consideration. The grid interval is represented by the total 1
Nu ¼ Nudt; ð3Þ
number of cells generated in the computational domain. A total t total 0
number of 6000, 10,000, and 16,000 cells and a time-step size of
with ttotal being the duration of the process of interest.
0.2, 0.1, and 0.05 s, giving rise to five combinations as listed in
Table 1, are considered. The time-averaged Nusselt number, which
will be defined in the following section, is used as a measure for
comparing numerical accuracy among the five cases. The relative
deviation of the time-averaged Nusselt number for each case with
respect to the case of the smallest grid and time-step intervals, i.e.
total number of 16,000 cells and 0.05 s, is calculated as well.
It is found that the maximum deviation between the two ex-
treme cases is less than 4%. At constant number of cells, the
time-averaged Nusselt number is nearly independent of the
time-step interval. The relative deviation between two different

Table 1
Independence test of grid and time-step intervals for AR = 3, h = 0°, and Gr = 107.
 
Number of cells Time-step size (s) Nuin Deviation (%)

6000 0.20 18.0487 3.436


6000 0.10 18.0675 3.544
10,000 0.10 17.7102 1.496
10,000 0.05 17.7086 1.487
16,000 0.05 17.4491 0.000 Fig. 2. Time histories of the average Nusselt number over the inner circular wall for
different aspect ratios and inclination angles for Gr = 105.
1566 Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571

The evolution of the average Nusselt number over the inner the presence of multi-cellular pattern, the steady-state Nusselt
circular wall is used as the indicator for distinguishing different number is increased. This might be desired in engineering applica-
phases during the course of convective flow development. For tions for heat transfer enhancement.
Gr = 105, the time histories of this indicator are presented in Evolutions of the average Nusselt number for Gr = 106 are
Fig. 2 for different combinations of aspect ratio and inclination an- shown in Fig. 4. For the two highest aspect ratios studied (AR = 4
gle. Recall that the entire domain is isothermally kept at 300 K and and 5), a steady state is able to be attained for each combination
the temperature on the inner wall is suddenly raised to 320 K at of aspect ratio and inclination angle after about 30 s, which is
t = 0. Therefore, a great temperature gradient is established imme- greatly longer than the several seconds for Gr = 105. The variation
diately between the inner wall and the air layer in the vicinity to it, for AR = 4 is similar to those observed for AR = 3 and 5 when
giving rise to an infinitely large average Nusselt number at t = 0. Gr = 105, as shown in Fig. 2. As natural convection is enhanced,
These initial impulses are truncated at a reasonable value in the unusual undershoot that is present for AR = 4 and h = 60° in
Fig. 2 in order to present the relatively small values at a late stage. the previous case has disappeared. For AR = 5, the two curves for
It is shown that the impulses at t = 0 decay very fast and attain h = 0° and 60° nearly assemble to a single one and the curve corre-
their corresponding steady-state values within several seconds. A sponding to h = 30° becomes the lowest. The local peaks on the
steady state is signified by a flat curve, showing that the average curves during the developing phase become more pronounced
Nusselt number does not vary with time. Although it is hard to and most importantly, the curves for AR = 3 exhibit apparent unu-
identify in Fig. 2, there is a short peak on each curve during the sual characteristics when the triangular enclosure is turned. When
dropping period, which might be caused by transition of convec- the top side wall is horizontal (h = 60°), after developing about 30 s
tive flow pattern or short-term instability. However, because the a pseudo-steady state is attained at a value that is greater than that
spans and magnitudes of these peaks are relatively small, their for h = 0°, followed by a transition of nearly 20 s. The presence of
effects on the overall characteristics of flow development are the pseudo-steady state is due to a neutrally stable flow pattern
expected to be insignificant and hence are not of interest in this during that period, which will be presented in detail in the follow-
study. In general, two phases are identified during the course of ing section. The flow finally becomes oscillatory, corresponding to
flow development for Gr = 105, i.e. developing and steady state. unstable flow patterns, with a magnitude of nearly unity and a per-
As shown in Fig. 2, the steady-state values of the average Nus- iod of 10 s. The peak and trough values of the oscillations are
selt number are increased as the aspect ratio is raised and exhibit slightly lower than the pseudo-steady-state value in this case
insignificant differences as the inclination angle is changed. For and the steady-state value for h = 0°, respectively. Note that if the
AR = 3, the three curves are very close to one another with the computation is terminated during the pseudo-steady-state period
one corresponding to h = 0° being the highest, indicating that turn- (30 s < t < 60 s), one may incorrectly conclude that a steady state
ing the triangular enclosure from its original orientation will has already attained for this case. When the triangular enclosure
slightly decrease the overall heat transfer. A similar trend is found is turned by 30°, i.e. the right side wall is vertical, the flow finally
for AR = 5 with the two lower curves almost coinciding with each becomes stably oscillatory as well, with much smaller magnitude
other. In this case, the attenuation of the average Nusselt number and period than those for h = 60°. However, unlike the previous
due to the turning of the triangular enclosure becomes more pro- case, the pseudo-steady state is absent in this case and a secondary
nounced as compared to that for AR = 3. When AR = 4, the behav- peak is found on the curve around t = 15 s, followed by a drop to a
iors of the curves for h = 0° and 30° are still similar to what have trough value that is lower than the local value for the original case
been observed for the other two aspect ratios. However, an excep- (h = 0°). The average Nusselt number is then gradually increased
tion is found for h = 60° in such a way that the dropping period is towards the late oscillatory status. Note that for AR = 3, since turn-
followed by a growth towards the steady state, thus leading to that ing of the triangular enclosure leads to oscillatory flows, the aver-
the steady-state value in this case becomes the largest one among age Nusselt number at a late stage becomes greater than that for
the three angles. This unusual behavior is attributed to a unique the original orientation, as opposed to the trend for the other
flow pattern with multiple recirculating cells. Note that when two aspect ratios.
h = 60°, i.e. the top side wall of the inner triangle is horizontal, When natural convection is further enhanced to Gr = 107, as
the upper middle part of the enclosure is analogous to the classical presented in Fig. 5, the flow exhibits high-frequency oscillations
bottom-heated parallel plates for investigating Rayleigh–Bénard for most of the cases and no stably oscillatory state is able to be at-
convection. Therefore, there is a chance to form multiple recircu- tained any more. However, under certain circumstances, a steady
lating cells in that region. As shown in Fig. 3, for Gr = 105, only state or quasi-steady state is still available, e.g. for the three cases
the case corresponding to AR = 4 will be able to lead to multi- of h = 60°. Although in each of these three cases, the flow is oscilla-
cellular flow pattern at steady state, which disappears while either tory with small magnitude during the developing phase, a nearly
increasing or decreasing the aspect ratio. As seen in Fig. 2, due to flat curve is finally achieved after t = 80 s. This means that in case

(a) AR = 3 (b) AR = 4 (c) AR = 5


Fig. 3. Comparison of the steady-state flow patterns by means of streamlines among cases with different aspect ratios: (a) AR = 3, (b) AR = 4, and (c) AR = 5, for h = 60° and
Gr = 105.
Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571 1567

resulting temperature distributions inside the enclosure are to be


presented by means of streamlines and isotherms, respectively.
According to prior discussion, the two oscillatory cases at AR = 3
and Gr = 106 that exhibit complicated variations of the average
Nusselt number are of great interest. Typically, the flow develop-
ment of these two cases can be identified as three phases: develop-
ing, transitioning, and oscillating. Representative streamlines and
isotherms at different phases for the case of AR = 3, h = 60°, and
Gr = 106 are presented in Fig. 6, in which the unit of the color column
for temperature is Kelvin. As shown in Fig. 6(a), the flow pattern and
temperature distribution are symmetric about the vertical central
line of the geometry at an early stage. However, the two Rayleigh–
Bénard-like secondary cells on the top middle part of the enclosure
are neutrally stable and become slightly asymmetric during the
transitioning phase, as shown in Fig. 6(b), whereas the two primary
cells seem to be nearly unaffected by the deformation of the second-
ary cells. As mentioned before, the flow patterns shown in Fig. 6
clearly elucidate the reason for the presence of the pseudo-steady
Fig. 4. Time histories of the average Nusselt number over the inner circular wall for state for this case. At a later stage, the flow starts oscillating
different aspect ratios and inclination angles for Gr = 106. and the snapshots at a typical instant are only shown. One of the
secondary cells becomes much larger than the other and occupies
the middle region. The development in flow patterns is verified by
the movement of locations of the pair of thermal plumes in iso-
therms. During the oscillating phase, the relative sizes and positions
of the two secondary cells change periodically to some extent and
hence the average Nusselt number varies periodically as well.
The snapshots of streamlines and isotherms for AR = 3, h = 30°, and
Gr = 106 are presented in Fig. 7, where an apparent difference is that
the streamlines and isotherms are no longer symmetric because the
geometric symmetric line does not coincide with gravity. It is shown
that the convective flow and heat transfer inside the left part and the
lower right part of the enclosure are nearly unchanged as flow devel-
ops. Therefore, attention should be paid to the upper right corner. At
the beginning, there is only one small secondary cell and hence one
distinguishable thermal plume in the upper middle region. Multiple
cells exist at later phases and two thermal plumes are identified. As
shown in Fig. 7(b), among the three cells in the upper corner, the mid-
dle one gradually grows and becomes more intensive as the stream-
lines become more concentrated. However, the flow patterns at the
oscillating phase do not essentially differ from those at the transition-
Fig. 5. Time histories of the average Nusselt number over the inner circular wall for
ing phase. As a result, it is difficult to distinguish these two phases
different aspect ratios and inclination angles for Gr = 107.
through the isotherms shown in Fig. 7(b) and (c).

of intensive natural convection, the configuration with the top side 4.3. Flow development times and correlations
wall being horizontally positioned leads to most stable flow pat-
terns among all the three inclination angles studied. Recall that As shown in Figs. 2–5, the total duration of the developing and
when Gr = 106, this is the most unstable configuration due to the transitioning phases is significantly increased as the Grashof num-
presence of unstable secondary cells, which will be presented in ber is raised. The flow development time is of interest for transient
the following section. The sizes and positions of the cores of the natural convection because it indicates how long does the flow take
secondary cells mutually change, thus leading to oscillatory flow to develop towards the steady/quasi-steady state. In practice, a short
behavior. However, as natural convection is enhanced, the second- development time suggests that the steady-state Nusselt number,
ary cells are more confined with nearly fixed sizes and positions. which is usually well-documented in the literature, can be used in-
The flow therefore becomes stable for Gr = 107 in this case. Also, stead of the transient Nusselt number. The flow development time is
the overall heat transfer rate at a late stage is somewhat enhanced measured as the instant at which the average Nusselt number has at-
by turning the triangular enclosure by 60° for each aspect ratio. tained 99.9% of its steady-state value or time-averaged value over a
The enhancement is rather remarked for AR = 3 such that the aver- late period, if the flow is finally oscillatory. The flow development
age Nusselt number for h = 60° at a late phase (t > 80 s) is increased times are thus predicted and listed in Table 2. Due to the limitation
by about 12% as compared to those for the other inclination angles of the time-step interval in this study, the flow development time is
at the same instant. only obtained accurately to a resolution of 0.1 s.
As shown in Fig. 8(a), the flow development time (unit: s) for
4.2. Development of flow patterns and temperature distributions AR = 3, h = 0°, and Gr = 106 is scaled with Grashof number as a
power function, given by
Having shown the flow development by presenting the evolu-
tdev ¼ 0:0221 Gr0:4858 ; ð4Þ
tions of the average Nusselt number over the inner circular wall,
convective flow patterns seem to have significant influence on vari- with a nearly perfect coefficient of determination of 0.9999. The
ations of overall heat transfer. In this section, flow patterns and the scaling as a power function is not available for the other inclination
1568 Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571

320
318
316
314
312
310
308
306
304
302
300

(a) Developing (b) Transitioning (c) Oscillating


Fig. 6. Typical snapshots of streamlines (top row) and isotherms (bottom row, unit: K) at three different phases: (a) developing, (b) transitioning, and (c) oscillating, during
the course of flow development for AR = 3, h = 60°, and Gr = 106.

320
318
316
314
312
310
308
306
304
302
300

(a) Developing (b) Transitioning (c) Oscillating


Fig. 7. Typical snapshots of streamlines (top row) and isotherms (bottom row, unit: K) at three different phases: (a) developing, (b) transitioning, and (c) oscillating, during
the course of flow development for AR = 3, h = 30°, and Gr = 106.

angles for AR = 3 because the flow becomes oscillatory at higher there is only one inclination angle, i.e. h = 60° that leads to an expo-
Grashof numbers and hence the flow development time is not expo- nential growth of flow development time. This is indicated by the
nentially grown. Similarly, for each of the other two aspect ratios, fact that only for h = 60° the flow is able to attain a steady or
Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571 1569

Table 2
Predicted flow development times (unit: s) for various aspect ratios and inclination angles.

Gr AR = 3 AR = 4 AR = 5
h = 0° h = 60° h = 30° h = 0° h = 60° h = 30° h = 0° h = 60° h = 30°
105 5.8 8.0 7.1 6.1 10.8 7.8 6.5 8.5 8.1
106 18.3 70.2 45.3 39.0 26.4 28.9 25.9 26.7 25.6
107 55.7 81.6 49.5 48.9 76.8 31.6 41.4 73.3 22.9

Fig. 8. Variations of the flow development time for (a) AR = 3, h = 0°, (b) AR = 4,
h = 60°, and (c) AR = 5, h = 60°.

quasi-steady state at Gr = 107. The correlation for AR = 4 and h = 60°


appears as Fig. 9. Variations of the time-averaged Nusselt number over the inner circular wall
for (a) AR = 3, (b) AR = 4, and (c) AR = 5.
0:4498
t dev ¼ 0:0544 Gr ; ð5Þ
with a coefficient of determination of 0.9974, and the correlation for different inclination angles. Therefore, generic correlations that
AR = 5 and h = 60° is found to be are valid for all three inclination angles are proposed for both
t dev ¼ 0:0532 Gr0:4486 ; ð6Þ aspect ratios. The correlation for AR = 4 is given by
 
with a coefficient of determination of 0.9987. Nuin ¼ 0:3461 Gr 0:2527 ; ð10Þ

with a coefficient of determination of 0.9989, and for AR = 5 the cor-


4.4. Time-averaged Nusselt numbers and correlations
relation appears as
 
As defined in Eq. (3), the overall heat transfer over the duration Nuin ¼ 0:4201 Gr0:2478 ; ð11Þ
(ttotal = 200 s) is presented as the time-averaged Nusselt number,
which is expected to exponentially grown with respect to Grashof with a coefficient of determination of 0.9997.
number. The predicted values of the time-averaged Nusselt num- The indices of all the cases are around 0.25 except for the case of
ber for various aspect ratios and inclination angles are plotted in AR = 3 and h = 60° (Eq. (8)). In that case, the overall heat transfer is
Fig. 9. For AR = 3, an individual correlation is obtained for each of significantly enhanced and the index is greater than 0.25. It is
the three inclination angles, given by worthwhile emphasizing that all the correlations are valid within
  the range of 105 6 Gr 6 107, which is in convection-dominated
Nuin ¼ 0:2939 Gr0:2528 ; for h ¼ 0 ; ð7Þ laminar regime.
 
Nuin ¼ 0:2259 Gr0:276 ; for h ¼ 60 ; ð8Þ
  4.5. Observation of pitchfork bifurcation
Nuin ¼ 0:3447 Gr0:2527 ; for h ¼ 30 ; ð9Þ

with the coefficients of determination being 0.9996, 1, and 0.9981, Having presented in Fig. 6, the transitioning phase might be sig-
respectively. nified by the appearance of asymmetric flow patterns and temper-
Note that the correlations for h = 0° and h = 30° may be inter- ature distributions. The asymmetry is considered as an indicator
changeable with reasonable accuracy. Having shown in Fig. 5, it for the presence of pitchfork bifurcation. For h = 30°, since gravity
is confirmed by Fig. 9(a) that for AR = 3 the overall heat transfer does not act along the symmetric line, the flow and heat transfer
is significantly enhanced as the triangular enclosure is positioned are inherently asymmetric. In order to examine the onset and evo-
in such a way that the top side wall is horizontal (h = 60°). How- lution of pitchfork bifurcation, an indicator is introduced as the dif-
ever, this enhancement effect becomes negligible when the aspect ference between the average Nusselt numbers along the left and
ratio is increased. As shown in Fig. 9(b) and Fig. 9(c), for AR = 4 and right side walls for h = 0° and 60°, given by Nudiff = Nuleft  Nuright.
5 the time-averaged Nusselt numbers are nearly identical among This difference should be zero when the flow and heat transfer
1570 Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571

contrary, for Gr = 107 the magnitude of the indicator is oscillatory


to a great extent and the sign alternates irregularly.
The time histories of the Nusselt number difference for h = 60°
are shown in Fig. 11. The magnitude of the indicator in this case
is much smaller than that for the previous case, meaning that the
flow is more stable when the triangular enclosure is positioned
with the top side wall being horizontal. Similar to the previous
case, there is no pitchfork bifurcation for Gr = 105. However, when
Gr = 106, the flow and heat transfer are still nearly symmetric for
AR = 4 and 5. For AR = 3, the indicator is grown and finally becomes
stably oscillatory after t = 70 s, which agrees with the variation of
the average Nusselt number in this case, as presented in Fig. 4.
When the Grashof number is further increased to 107, pitchfork
bifurcation is present for each of the three aspect ratios such that
the indicator significantly departs from zero shortly after begin-
ning, followed by a damping oscillation. The magnitude of the
oscillations are strongly attenuated and particularly, for AR = 3
the oscillation finally converges to a constant value.

5. Conclusion

A numerical study on unsteady natural convection heat transfer


around a heated horizontal circular cylinder to its air-filled coaxial
Fig. 10. Time histories of the difference between average Nusselt numbers along
the left and right side walls for h = 0° and for (a) AR = 3, (b) AR = 4, and (c) AR = 5. triangular enclosure is conducted. The effects of Grashof number, as-
pect ratio, and inclination angle of the triangular enclosure on devel-
opment of flow and heat transfer are parametrically investigated.
Evolutions of the convective flow and heat transfer are pre-
sented by means of the time histories of the average Nusselt num-
ber over the inner circular wall and different phases are identified
during the course of flow development. Snapshots of streamlines
and isotherms for two typical cases are presented to exhibit iden-
tification of these phases. The flow development times are pre-
dicted and for three of the cases, are scaled with Grashof number
by curve fitting. The correlations are given by Eqs. (4)–(6). The
time-averaged Nusselt number over the inner circular wall is also
scaled with Grashof number, given by Eqs. (7)–(11). In addition,
based on the time histories of an indicator that is defined as the
difference between the average Nusselt numbers along the left
and right side walls of the triangular enclosure, the onset and evo-
lution of pitchfork bifurcation are explored quantitatively for the
cases of h = 0° and 60°.

Acknowledgement

This work was supported by the National Natural Science Foun-


dation of China (NSFC) through Grant Nos. 50706044 and
50876094.

References
Fig. 11. Time histories of the difference between average Nusselt numbers along
[1] D.T. Reindl, W.A. Beckman, J.W. Mitchell, Transient natural convection in
the left and right side walls for h = 60° and for (a) AR = 3, (b) AR = 4, and (c) AR = 5.
enclosures with application to solar thermal storage tanks, J. Sol. Energ. Eng.
114 (1992) 175–181.
[2] I. Hmouda, I. Rodriguez, C. Bouden, A. Oliva, Unsteady natural convection
cooling of a water storage tank with an internal gas flue, Int. J. Therm. Sci. 49
are symmetric, indicating that a non-zero value signifies the pres- (2010) 36–47.
ence of pitchfork bifurcation. The transient variations of this indi- [3] M.S. Khelifi-Touhami, A. Benbrik, D. Lemonnier, D. Blay, Laminar natural
cator for h = 0° are presented in Fig. 10. It is shown that the flow convection flow in a cylindrical cavity application to the storage of LNG, J.
Petrol. Sci. Eng. 71 (2010) 126–132.
and heat transfer are always symmetric for different aspect ratios [4] S.C. Saha, J.C. Patterson, C. Lei, Natural convection in attics subject to
when Gr = 105. As the Grashof number is increased, pitchfork bifur- instantaneous and ramp cooling boundary conditions, Energ. Buildings 42
cation appears during the early developing phase. The instant at (2010) 1192–1204.
[5] Z.-T. Yu, X. Xu, L.-W. Fan, Y.-C. Hu, K.-F. Cen, Transient natural convection heat
which the indicator starts deviating from zero is considered as
transfer from a heated triangular cylinder to its air-filled coaxial cylindrical
the onset of pitchfork bifurcation. Obviously, the onset time is infi- enclosure, Int. J. Heat Mass Transfer 53 (2010) 4296–4303.
nitely large for the symmetric cases and is significantly decreased [6] Yu.E. Karyakin, Yu.A. Sokovishin, O.G. Martynenko, Transient natural
as the Grashof number is raised. For Gr = 106, the asymmetry is convection in triangular enclosures, Int. J. Heat Mass Transfer 31 (1988)
1759–1766.
grown to a small magnitude and retains this value, which indicates [7] C. Lei, J.C. Patterson, Unsteady natural convection in a triangular enclosure
that the flow and heat transfer are stably asymmetric. On the induced by absorption of radiation, J. Fluid Mech. 460 (2002) 181–209.
Z.-T. Yu et al. / International Journal of Heat and Mass Transfer 54 (2011) 1563–1571 1571

[8] Y. Mao, C. Lei, J.C. Patterson, Unsteady natural convection in a triangular [22] M.N. Borjini, C. Mbow, M. Daguenet, Numerical analysis of combined radiation
enclosure induced by absorption of radiation – a revisit by improved scaling and unsteady natural convection within a horizontal annular space, Int. J.
analysis, J. Fluid Mech. 622 (2009) 75–102. Numer. Methods Heat Fluid Flow 9 (1999) 742–764.
[9] C. Lei, J.C. Patterson, Unsteady natural convection in a triangular enclosure [23] G.N. Facas, B. Farouk, Transient and steady-state natural convection in a
induced by surface cooling, Int. J. Heat Fluid Flow 26 (2005) 307–321. porous medium between two concentric cylinders, J. Heat Transfer 105 (1983)
[10] Y. Mao, C. Lei, J.C. Patterson, Unsteady near-shore natural convection induced 660–663.
by surface cooling, J. Fluid Mech. 642 (2010) 213–233. [24] C.-K. Chen, S.-W. Hsiao, P. Cheng, Transient natural convection in an eccentric
[11] C. Lei, S.W. Armfield, J.C. Patterson, Unsteady natural convection in a water- porous annulus between horizontal cylinders, Numer. Heat Transfer A. Appl.
filled isosceles triangular enclosure heated from below, Int. J. Heat Mass 17 (1990) 431–448.
Transfer 51 (2008) 2637–2650. [25] M. Kumari, G. Nath, Unsteady natural convection from a horizontal
[12] T.P. Bednarz, C. Lei, J.C. Patterson, An experimental study of unsteady natural annulus filled with a porous medium, Int. J. Heat Mass Transfer 51 (2008)
convection in a reservoir model cooled from the water surface, Exp. Therm. 5001–5007.
Fluid Sci. 32 (2008) 844–856. [26] W.-S. Fu, J.-C. Perng, W.-J. Shieh, Transient laminar natural convection in an
[13] T.P. Bednarz, C. Lei, J.C. Patterson, A numerical study of unsteady natural enclosure partitioned by an adiabatic baffle, Numer. Heat Transfer A. Appl. 16
convection induced by iso-flux surface cooling in a reservoir model, Int. J. Heat (1989) 325–350.
Mass Transfer 52 (2009) 56–66. [27] M.Y. Ha, M.J. Jung, Y.S. Kim, Numerical study on transient heat transfer and
[14] T.P. Bednarz, C. Lei, J.C. Patterson, An experimental study of unsteady natural fluid flow of natural convection in an enclosure with a heat-generating
convection in a reservoir model subject to periodic thermal forcing using conducting body, Numer. Heat Transfer A. Appl. 35 (1999) 415–433.
combined PIV and PIT techniques, Exp. Fluids 47 (2009) 107–117. [28] M.Y. Ha, I.-K. Kim, H.S. Yoon, K.S. Yoon, J.R. Lee, S. Balachandar, H.H. Chun,
[15] T.P. Bednarz, C. Lei, J.C. Patterson, Unsteady natural convection induced by Two-dimensional and unsteady natural convection in a horizontal enclosure
diurnal temperature changes in a reservoir with slowly varying bottom with a square body, Numer. Heat Transfer A. Appl. 41 (2002) 183–210.
topography, Int. J. Therm. Sci. 48 (2009) 1932–1942. [29] E.K. Lakhal, M. Hasnaoui, P. Vasseur, E. Bilgen, Natural convection in a square
[16] E. Van De Sande, B.J.G. Hamer, Steady and transient natural convection in enclosure heated periodically from part of the bottom wall, Numer. Heat
enclosures between horizontal circular cylinders (constant heat flux), Int. J. Transfer A. Appl. 27 (1995) 319–333.
Heat Mass Transfer 22 (1979) 361–370. [30] B. Abourida, M. Hasnaoui, S. Douamna, Transient natural convection in a
[17] T.H. Nguyen, P. Vasseur, L. Robillard, Natural convection between horizontal square enclosure with horizontal walls submitted to periodic temperatures,
concentric cylinders with density inversion of water for low Rayleigh Numer. Heat Transfer A. Appl. 36 (1999) 737–750.
numbers, Int. J. Heat Mass Transfer 25 (1982) 1559–1568. [31] Z.-T. Yu, X. Xu, L.-W. Fan, Y.-C. Hu, K.-F. Cen, Transient natural convective heat
[18] Y.T. Tsui, B. Tremblay, On transient natural convection heat transfer in the transfer of a low-Prandtl-number fluid inside a horizontal circular cylinder
annulus between concentric, horizontal cylinders with isothermal surfaces, with an inner coaxial triangular cylinder, Int. J. Heat Mass Transfer 53 (2010)
Int. J. Heat Mass Transfer 27 (1984) 103–111. 5102–5110.
[19] T. Sano, Transient natural convection between horizontal concentric cylinders, [32] X. Xu, Z.-T. Yu, L.-W. Fan, Y.-C. Hu, K.-F. Cen, A numerical study of laminar
Fluid Dyn. Res. 1 (1986) 33–47. natural convective heat transfer around a horizontal cylinder inside a
[20] A. Castrejon, D.B. Spalding, An experimental and theoretical study of transient concentric air-filled triangular enclosure, Int. J. Heat Mass Transfer 53 (2010)
free-convection flow between horizontal concentric cylinders, Int. J. Heat Mass 345–355.
Transfer 31 (1988) 273–284. [33] Z.-T. Yu, Y.-C. Hu, L.-W. Fan, K.-F. Cen, A parametric study of Prandtl number
[21] K. Vafai, J. Ettefagh, An investigation of transient three-dimensional buoyancy- effects on laminar natural convection heat transfer from a horizontal circular
driven flow and heat transfer in a closed horizontal annulus, Int. J. Heat Mass cylinder to its coaxial triangular enclosure, Numer. Heat Transfer A. Appl. 58
Transfer 34 (1991) 2555–2570. (2010) 564–580.

You might also like