i An update to this article is included at the end

International Journal of Heat and Mass Transfer 142 (2019) 118447

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Enhancing the flow and heat transfer in a convective cavity using

symmetrical and adiabatic twin fins
Yang Liu a,⇑, Shuaikun Zhang a, Han Huang d, Qinghui Suo b, Yongning Bian a,⇑, Yongling Zhao c,⇑
a
School of Ocean Science and Technology, Dalian University of Technology, Panjin 124221, China
b
School of Engineering and Technology, Southwest University, Chongqing 400715, China
c
School of Aerospace, Mechanic and Mechatronic Engineering, University of Sydney, NSW 2006, Australia
d
Eindhoven University of Technology, Netherlands

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

Article history: The enhancement of natural convection flow and the associated heat transfer in a differentially heated
Received 3 April 2019 cavity with two horizontal adiabatic fins attached to each sidewall are numerically investigated at differ-
Received in revised form 14 July 2019 ent Rayleigh numbers and with various fin positions in this study. The numerical approach is validated by
Accepted 17 July 2019
both representative scales and experimental shadowgraph results. A typical double-plume flow regime is
Available online 25 July 2019
identified at both the early and fully-developed stages. The interaction of the two plumes is investigated
in detail in this paper. With the increasing of the Rayleigh number, the separation frequency of the plume
Keywords:
above both fins increases. It is also revealed that the separation frequency of the upstream plume at the
Natural convection
Heat transfer
quasi-steady stage does not only depend on the characteristics of the upstream flow, but also greatly
Differentially heated cavity depends on the position of the downstream fin. The frequency varies non-monotonically as the down-
Plume interaction stream fin moving toward the cavity ceiling, suggesting complex nonlinear characteristics. Compared
Two fins to the scenario without any fin, the flow rate across the cavity is significantly boosted, which is increased
by up to 136.2% at the early stage and 124.8% at the fully-developed stage. The heat transfer rate across
the cavity having symmetrical fins on sidewalls is enhanced by up to 12.7% at a Rayleigh number of
3.68
109.
Ó 2019 Elsevier Ltd. All rights reserved.

1. Introduction with analytical, experimental and numerical approaches since

1950s (see for example [25–29]).
Natural convection is an important branch of convective flow Previous studies [25,27] suggested that the Rayleigh number
apart from forced convection and the heat transfer by this process (Ra), Prandtl number (Pr) and the aspect ratio of the cavity (A)
is completely passive without the need for extra energy and/or are the governing parameters of the natural convection flow in a
noisy mechanical components, which keeps attracting fundamen- differentially heated cavity. The flow in the cavity is steady at
tal and applied research in recent years [1–7]. It occurs in various low Rayleigh numbers, e.g. Ra < 103, where the flow is mainly dom-
household applications and industrial systems, such as in roof solar inated by heat conduction. One of the earliest studies of steady nat-
collector [8] and heat exchangers of refrigeration appliances [9]. In ural convection in the differentially heated cavity was presented in
addition, the natural convection in various heat exchangers filled [30]. The authors pointed out that a weakly stable unicellular lam-
with nanofluids has been extensively investigated in [10–16]. inar circulation is generated for Ra < 103. For 103 < Ra < 105, the
The features of the flow and heat transfer of nanofluids in the cav- flow is similar to that adjacent to a heated semi-infinite vertical
ity have also been analyzed in [17–24]. A differentially heated rect- plate except that the vertical growth of the boundary layers is
angle cavity is a representative model from a variety of civilian and inhibited in the central region of the cavity. A steady laminar sec-
industrial applications and it is widely used for the studies of nat- ondary flow starts to form at approximately Ra = 105. As the fur-
ural convection flows. The coupled flow and heat transfer charac- ther increase of Rayleigh number to around Ra = 106, the
teristics have been extensively investigated and documented secondary cell grows larger and a further steady cellular motion
is generated in the weak shear regions between each cell.
The effects of single and multiple conductive fins [31–34] on a
⇑ Corresponding authors. steady cavity flow were also investigated within this Ra range. It
E-mail addresses: [email protected] (Y. Liu), [email protected] (Y. Bian), was demonstrated that heat transfer is enhanced due to the
[email protected] (Y. Zhao).

https://doi.org/10.1016/j.ijheatmasstransfer.2019.118447
0017-9310/Ó 2019 Elsevier Ltd. All rights reserved.
2 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447

Nomenclature

A aspect ratio Dt time step

f1 dominant frequency at P1 T equivalent dimensional temperature (k)
f2 dominant frequency at P2 T0 equivalent dimensional value of the initial temperature
f3 dominant frequency at P3 in the cavity (k)
g gravitational acceleration (m/s2) DT equivalent dimensional value of the temperature differ-
H, L equivalent dimensional values of the height and length ence between the heated and cooled sidewalls (k)
of the cavity (m) DTlayer equivalent dimensional value of the temperature differ-
h, l height and length of the cavity ence in the unstable layer above the fin (k)
h1, h2 height of the upstream and downstream fins u, v velocities in x and y directions
Dh gap between the two fins ui velocity of the downstream intrusion
Nu Nusselt number vp velocity of the downstream plume
Nufin, Nunofin time-averaged Nusselt number at the heated side- x, y horizontal and vertical coordinates
wall with two fins and without any fin h temperature
p pressure b coefficient of thermal expansion (1/K)
Pr Prandtl number dlayer thickness of the unstable layer above the fin
Q volumetric flow rate dts thickness of the steady thermal boundary layer
Ra Rayleigh number j thermal diffusivity (m2/s)
RaDh local Rayleigh number based on Dh m kinematic viscosity (m2/s)
Ral local Rayleigh number above the fin q density (kg/m3)
Rel local Reynolds number above the fin e enhancement factor of heat transfer
t time

presence of the fins. In addition, the percentage of enhancement aration frequency increases with Rayleigh number and decreases
augments with fin length in general whilst the effect of fin thick- with fin height. Moreover, it was found that heat transfer across
ness on heat transfer is minor. Fin position was also found to dra- the cavity is enhanced by the thin fin [45]. The flow and heat trans-
matically affect the flow features, which subsequently alters heat fer in the cavity with two horizontal fins of different lengths were
transfer in the cavity [32]. Moreover, it was demonstrated that experimentally and numerically investigated in [46]. The authors
by attaching two fins to the sidewall, the heat exchange across suggested that the heat transfer is improved by up to 17.1% over
the cavity is significantly strengthened and the optimal fin position the Rayleigh numbers examined compared to the case without
in terms of heat transfer enhancement was discussed and identi- any fin. Transient natural convection flow in the finned cavity
fied in [34]. The impacts of multiple pin fins and sinusoidal ele- was also studied using a scaling analysis and the corresponding
ment on the natural convective flow and heat transfer were scaling relations quantifying different flow regimes were obtained
investigated in [35,36]. in [47,48]. These studies revealed that Rayleigh number and
Convection becomes more important than conduction with the Prandtl number play a vital role in the development of transient
increase of Rayleigh number. The flow and heat transfer in the cav- flow, which was confirmed and discussed in recent years, e.g.
ity then enter convection dominance gradually. The steady natural [49,50].
convection flow transits to unsteady natural convection flow as the The early-stage transient flows in a differentially heated cavity
Rayleigh number increases beyond a critical value, which depends with an adiabatic horizontal fin attached to the sidewalls were
on the Prandtl number of the fluid as well as the aspect ratio of the reported by Xu [41,42] and Ma [49,50]. The corresponding scaling
cavity [27]. This transition corresponds to the onset of the so-called analyses outlining the force and energy balances at different flow
convective instability [28,37]. Moreover, it was demonstrated that stages were also performed [47]. The development of the thermal
if the Rayleigh number is further increased, the flow may transit to boundary layer, the intrusion underneath the fin and the plume
a turbulent state [38–40]. Apart from the flow in a cavity without bypassing the fin have been elucidated for the early stage. How-
any fin, the transition to an unsteady flow in a differentially heated ever, the flow features in a differentially heated cavity with two
finned cavity was also investigated in recent years [41,42]. It was adiabatic horizontal fins attached to the sidewalls have not been
found that, besides the Rayleigh number, Prandtl number and the well understood.
aspect ratio, the transition from a steady flow to an unsteady flow, To better understand the flow and heat transfer improvement
or the so called quasi-steady flow, also depends on fin length and by using twin fins, especially how the two plumes interact with
fin height. The corresponding stability map was obtained in [43]. each other, the cavity with two horizontal adiabatic thin fins
The development of the unsteady flow in the finned cavity con- attached at each sidewall is investigated in the present study. It
sists of three distinct stages: an early stage, a transitional stage and is expected that the convective flow in the cavity with two hori-
a fully-developed stage [41,44]. The early stage is featured by the zontal adiabatic fins attached to each sidewall may be significantly
development of a thermal boundary layer flow along the sidewall, different from the traditional scenario with only one fin, which is
horizontal intrusions underneath the fin and the ceiling, the forma- the motivation of the present study. Two plumes will develop,
tion of a vertical plume caused by the intrusion bypassing the fin, evolve and interact owing to the presence of the two fins, which
and the reattachment of the plume to the downstream thermal may lead to a significant unsteady flow, ultimately enhancing the
boundary layer. A double-layer structure of the thermal boundary cavity-wide heat transfer significantly. This study provides useful
layer forms gradually in the lengthy transitional stage [44]. The reference for the design of high-performance heat exchanger.
temperature ultimately stratifies in the cavity at the fully- Besides, the present work is also significant for the design of com-
developed stage and it is found that plumes periodically separate pact heat exchangers utilized in limited space in various industrial
from the upper surface of the fin, which in turn causes significant applications, such as electrical heat sinks, cooling system of under-
oscillations of parameters of the downstream flow. The plume sep- water ROVs and AUVs. In this paper, a finite volume method (FVM)
 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447 3

based numerical method is employed to account for the flow and tional non-dimensional parameters govern the transient flows in
heat transfer properties in the finned cavity (with two fins at each the cavity, i.e. the Rayleigh number (Ra), Prandtl number (Pr) and
sidewall), in which three Rayleigh numbers and five spatial set- the aspect ratio of the cavity (A), which are defined as,
tings of the fins are studied. The numerical method is validated
against representative analytical scales and experimental shadow- gbDTH3 m H
Ra ¼ ; Pr ¼ ; A¼ ð5Þ
graph results of the flow. As a reminder of this paper, numerical mj j L
approach is given in Section 2. The flow features in the early stage
The Prandtl number and aspect ratio considered are 7.02 and
and the fully-developed stage as well as the associate heat transfer
0.24 respectively, while three different Rayleigh numbers, i.e.
properties are analyzed and discussed in Section 3 followed by
Ra = 9.2
108, 1.84
109 and 3.68
109, are examined in the pre-
conclusions of this study given in Section 4.
sent study. It is noted that the specific aspect ratio, Prandtl number
and Rayleigh number are the same as those adopted in [45,46].
2. Numerical approach The governing Eqs. (1)–(4) are solved iteratively by the finite
volume method (FVM) approach where the SIMPLE algorithm is
A two-dimensional model can properly quantify the flow in a used to handle the pressure-velocity coupling [51] and the QUICK
differentially heated cavity with fins attached to the sidewalls scheme is applied for the discretization of advection terms [52].
and at Ra < 7.7
109, which has been extensively demonstrated The pressure term is discretized by the PRESTO scheme [53]. The
in e.g. [41,46,47]. Therefore, a two-dimensional model, as schemat- discretization of all second-derivative terms and linear first-
ically depicted in Fig. 1, is adopted in the present study. Water is derivative terms adopts the second-order center-differenced
chosen as the working fluid and it is initially at rest with a uniform scheme and the discretization of transient integration uses the
temperature of h0 . The top and bottom walls of the cavity are adi- second-order implicit differenced scheme. The solution is con-
abatic. The flow initiates when the right and left sidewall is simul- verged when the scaled residual of the computational cell falls
taneously heated and cooled to h0 þ Dh=2 and h0  Dh=2. It is noted below 106 for the temperature equation and 103 for the continu-
that the adiabatic thin fins are symmetrically attached at each ity and momentum equations.
sidewall to keep the thermal equilibrium across the cavity. Fur- A non-uniform mesh with finer grids towards the cavity side-
thermore, all wall surfaces and fins are considered to be no-slip walls is constructed to capture the details of the flow features.
and rigid. The schematic of the computational mesh in the vicinity of the fins
The flow and heat transfer in the cavity are mathematically is depicted in Fig. 2(a). For the purpose of ensuring the accuracy of
described by the following dimensionless continuity, Navier- numerical computations, mesh dependency test is performed with
Stokes and energy equations incorporated with the Boussinesq two mesh systems, i.e. 360
520 and 450
600, for the case with
approximation [51], Ra = 3.68
109 and h2 = 7/24. Meanwhile, the Nusselt number at
@u @ v the heated sidewall is monitored and recorded to evaluate the
þ ¼0 ð1Þ impact of different meshes and time steps on numerical results.
@x @y
In the present study, the Nusselt number is defined as,
!
@u @u @u @p Pr @2u @2u Z
þv @h
1
þu ¼  þ 1=2 þ ð2Þ
@t @x @y @x Ra @x2 @y2 Nu ¼ dy ð6Þ
0 @x
! Time series of the Nusselt number at the heated sidewall are
@v @v @v @p Pr @2v @2v
þu þv ¼  þ 1=2 þ þ Prh ð3Þ presented in Fig. 2(b), and the time-averaged Nu and velocity com-
@t @x @y @y Ra @x2 @y2 ponents in x and y direction, i.e. u and v, at the monitoring point P3
! at the fully-developed flow stage, are summarized in Table 1. It is
@h @h @h 1 @2h @2h found that the relative difference of Nu originated from grid system
þu þv ¼ 1=2 þ ð4Þ is found to be less than 1%, and the relative difference with regard
@t @x @y Ra @x2 @y2
to u and v is approximately 4%. Therefore, either mesh system can
where the length, time, velocity and pressure are nondimensional- be utilized. A time step dependency test is also carried out with
ized by H, H2j1Ra1/2, H1jRa1/2 and qH2j2Ra respectively. The two dimensionless time steps of 0.01 and 0.03 using the
temperature is normalized according to h = (T – T0)/DT. 360
520 mesh. The corresponding time series of Nusselt number
It is worth noting that the height of the upstream fin h1 is fixed are also given in Fig. 2(b). It is clearly seen that the time series of
at 1/4 for all the scenarios under consideration whilst the height of Nusselt number obtained with different time steps converge which
the downstream fin varies and appears at five locations, respec- indicates that both time steps are appropriate. Considering the
tively, i.e. h2 = 7/24, 1/3, 5/12, 1/2 and 7/12. In the meantime, the computational cost, the mesh system with 360
520 computa-
fin length is fixed at 1/6, which is equivalent to that adopted in tional cells together with the time step of 0.03 are employed for
[45]. Apart from the height of the downstream fin h2, three addi- the following simulations.

Fig. 1. Schematic of the computational domain and boundary conditions. Data are monitored at the points P1 (4.083, 0.271), P2 (4.083, h2 + 0.0625) and P3 (4.158, 0.917)
during the calculations.
4 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447

Fig. 2. (a) Local grid near the two horizontal adiabatic fins. (b) Time series of
Nusselt number at the heated sidewall with Ra = 3.68
109 and h2 = 7/24.

3. Results and discussion

Fig. 3. Development of the thermal boundary layer. (a) Shadowgraph experiment.
3.1. Flow features in the early stage (b) Numerical results in the present study. (c) The comparison of thermal boundary
layer between the experimental and numerical results.
Previous studies [49,50,54] demonstrated that, immediately
after the initiation of the flow, the thermal boundary layer is dom-
inated by conduction and it experiences a one-dimensional growth identifies the outer edge of the thermal boundary layers from the
in the first place. After the leading-edge effect (LEE) is completely shadowgraph images.
convected away, the boundary layer transits to a two-dimensional The numerical results are further validated against representa-
steady stage which is dominated by convection, if the Rayleigh tive scales of the flow. Fig. 4 plots the thickness of the present
number exceeds O(106) [27]. The thermal boundary layer was visu- steady-stage thermal boundary layer dts against the scale
1=4
alized and investigated by the shadowgraph experiment in [26]. RaDh (1 + Pr1/2)1/2 [47]. A clear linear fit is obtained, indicating
Fig. 3(a) presents snapshots obtained at different time instants, that this scale can appropriately describe the thickness of the ther-
demonstrating the growth of the corresponding thermal boundary mal boundary layer under consideration. Note that the RaDh is the
layer (observing from right to left). Fig. 3(b) shows the develop- Rayleigh number calculated using the length scale from the
ment of thermal boundary layer in the present numerical study streamwise locations of the two fins (refer to Dh in Fig. 1).
and they compare reasonably with the shadowgraph images. It is Due to the presence of the fins and the ceiling, the fluid in the
clearly seen that the convective boundary layer becomes thicker vertical thermal boundary layer is discharged into the interior of
with the streamwise position after the passage of the one- the cavity, resulting in intrusion flows underneath the fins and
dimensional growth stage. The thickness of the present steady- the ceiling respectively. The isotherms near the heated sidewall
stage thermal boundary layer developed between the downstream in the early stage are presented in Fig. 5, where the front positions
fin and the ceiling is compared with the experimental results with of the upstream intrusion are marked by black dash lines. It is
the same equivalent Rayleigh number in Fig. 3(c). A reasonably clearly demonstrated in Fig. 5(a)–(c) that, the intrusion underneath
good agreement is obtained with the average relative error being the upstream fin propagates faster than that underneath the down-
approximately 3%, suggesting that the numerical approach stream fin. It is interesting to note that when the height of the
employed in this work can properly describe the characteristics downstream fin is 1/2 the development of the two intrusions is
of the boundary layer flow in the cavity. It is also worth clarifying identical, as is shown in Fig. 5(d). As the height of the downstream
that the thicknesses obtained from experimental results are fin exceeds 1/2, i.e. h2 = 7/12, the downstream intrusion propagates
slightly greater than those determined from present numerical faster that the upstream intrusion, as suggested in Fig. 5(e). All the
ones, as shown in Fig. 3(c). This is associated with the way one above-mentioned phenomena are associated with the various evo-

Table 1
Grid and time step dependence test.

Test run Mesh (h
l) Time step Time-averaged Nu Time-averaged u Time-averaged v

1 450
600 0.03 40.80 0.00561 0.0923
2 360
520 0.03 40.60 0.00540 0.0888
3 360
520 0.01 40.33 0.00536 0.0872
 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447 5

Fig. 4. The scale of thickness of the thermal boundary layer Ra1=4

Dh (1 + Pr
1/2 1/2
) Fig. 6. The normalized downstream intrusion velocity uiRa1=20
Dh (1 + Pr
1/2 1/5
)
versus the numerically calculated dts. against t7/10.

that the intrusion velocity decreases rapidly with time for the case
with small h2, i:e: h2  1=3, whilst it decreases mildly when
h2  5=12.
A starting plume forms gradually as the intrusion bypasses the
fin, which subsequently triggers unsteadiness of the flow adjacent
to the sidewall. Fig. 7 presents the isotherms near the heated side-
wall for different fin positions at Ra = 1.84
109 to demonstrate
the effect of the fin position on the development of the plume. It
is clearly seen in Fig. 7(a) and (b) that the downstream intrusion
is so weak that only one plume presents when the two fins are
close to each other, i.e. h2  1/3. However, a double-plume flow
regime can be identified for the cases with h2  5/12, as shown
in Fig. 7(c) to (e).
In addition, the position of the downstream fin dramatically
affects the process of plume interaction. The upstream plume
ascends vertically until it hits the downstream fin and it is subse-
quently divided into two halves, one being the plume flow
between the two fins and the other continuing to ascend. If the
Fig. 5. Intrusion flows in the early stage (t = 1.71) for different fin positions with gap between the two fins is not very large, e.g. h2 ¼ 5=12, part of
Ra = 1.84
109 (isotherms from 0 to 0.5 with an interval of 0.021). (a) h2 = 7/24, (b) the upstream plume directly merges into the downstream plume,
h2 = 1/3, (c) h2 = 5/12, (d) h2 = 1/2, (e) h2 = 7/12. as given in Fig. 7(c). Nevertheless, with the downstream fin moved
to 1/2, the downstream plume reattaches to the downstream ther-
lutionary length of the thermal boundary layer between the two mal boundary layer on the arrival of the upstream plume, as shown
fins. That is, the development of each intrusion depends on its in Fig. 7(d). With further lifting the downstream fin, e.g. h2 ¼ 7=12,
respective upstream thermal boundary layer. the downstream plume hits the intrusion underneath the ceiling of
It is also worth noting that the local Rayleigh number RaDh of the cavity, as the upstream plume reaches the downstream fin. In
the boundary layer flow between the two fins is less than the meanwhile, it is worth noting that with the lifting of the down-
Pr10(1 + Pr1/2)10 for all the cases studied, implying that the down- stream fin, the entrainment of the thermal boundary layer between
stream intrusion is dominated by a buoyancy-viscous balance [47]. the two fins also becomes stronger which results in a more tilted
Meanwhile, the flow rate to the intrusion is unsteady because the plume as shown in Fig. 7.
present fin length is less than the critical value for the unsteady- As discussed above, the velocity of the downstream intrusion
steady transition. Therefore, the dimensionless velocity scale of increases with time according to O(t7/10), which corresponds to
the intrusion under the downstream fin can be expressed as:

t 7=10
ui   1=5 ð7Þ
1 þ Pr1=2
1=20
RaDh

Fig. 6 plots the normalized velocity uiRaDh (1 + Pr1/2)1/5 of

1=20

the downstream intrusion against the time scale of t7/10. A clear

linear correlation can be discerned in Fig. 6 for all the cases inves-
tigated, supporting the scaling equation (7). In the meanwhile,
slight deviation from the linear correlation is noticed for the cases
with relatively small downstream fin height, i.e. h2  1/3. This is
because the evolutionary lengths of the thermal boundary layer
between the two fins are very limited in these circumstances,
which leads to a very low RaDh at an order of O(104–106). There-
Fig. 7. Starting plumes in the early stage for different fin positions with
fore, the effect of conduction becomes significant in comparison Ra = 1.84
109 (isotherms from 0 to 0.5 with an interval of 0.021). (a) h2 = 7/24,
with convection, in which cases the prescribed dynamics of equa- t = 4.17. (b) h2 = 1/3, t = 4.17. (c) h2 = 5/12, t = 5.66. (d) h2 = 1/2, t = 6.41. (e) h2 = 7/
tion (7) does not hold strictly. Additionally, it is also found in Fig. 6 12, t = 5.88.
6 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447

Fig. 10 presents the isotherms in the proximity of the heated

sidewall for different fin positions at Ra = 1.84
109 in the fully-
developed stage. Similar to the flow in the early stage, when the
gap between the two fins is small, i.e. Dh = 1/24, the equivalent
local Rayleigh number of the corresponding thermal boundary
layer flow is at an order of O(104). The flow is hence dominated
by heat conduction and there is only one plume, as shown in
Fig. 10(a). With the increase of the distance between the two fins,
i.e. Dh  1/12, two plumes form as demonstrated in Fig. 10(b)–(e).
Nevertheless, the plume could not ascend vertically in this stage
due to the entrainment of the boundary layer and it travels hori-
zontally on top of the thin fin, forming an unsteady thermal layer.
 1=3 In fact, the origination of this unstable plume is the reverse tem-
Fig. 8. The normalized velocity vP 1 þ Pr 1=2 of the downstream plume perature gradient above the thin fin which is associated with the
against t.
Rayleigh-Bénard instability. The related discussion is not carried
out herein since it has been particularly elaborated in [42,46].
an unsteady flow rate of O(t3/2) for the downstream plume flow. Further, it can be found that the position of the downstream fin
Therefore, the balance of the plume is between the buoyancy and greatly affects development of the plume flow. As demonstrated in
viscous [47], then a vertical dimensionless velocity scale may be Fig. 10(b), the upstream plume directly hits the intrusion under-
obtained for the plume as, neath the downstream fin, merging into it, and eventually becomes
a part of the downstream plume. With the downstream fin moved
t to a higher position, for example h2  5/12, the upstream plume
vp   4=3 ð8Þ
firstly reattaches to the thermal boundary layer between the two
1 þ Pr1=2
fins due to the entrainment effect. Then it travels upward and
The normalized velocity vp(1 + Pr1/2)1/3 of the downstream merges into the downstream intrusion. Furthermore, it is worth
plume is plotted against the time scale t in Fig. 8. Considering noting that the downstream plume is more chaotic than the
the facts that the plume velocity should be measured prior to the upstream one due to the strong mixing of the two streams of fluid
arrival of the upstream plume and the downstream plume must which is clearly seen in Fig. 10(c)–(e).
have enough space to grow, two fin positions, i.e. h2 = 5/12 and The above analyses demonstrate that the flow behavior
h2 = 1/2, are hereby chosen. It is found in Fig. 8 that the normalized between the two fins is dramatically affected by the position of
downstream plume velocity increases linearly with time, which is the downstream fin. To quantitatively evaluate its impact, Fig. 11
consistent with the scaling relation (8). However, it is also clear (a)–(j) present the temperature time series and the corresponding
that the linear fit lines are separate for different fin positions and power spectra at the monitoring point P1 in the fully-developed
Rayleigh numbers. This can be explained by the fact that the time stage for different fin positions at Ra = 1.84
109. The dominant
for the downstream intrusion to reach the fin tip is case dependent. frequency discerned from the spectral analysis is recognized to
be the separation frequency of the upstream plume f1. As shown
in Fig. 11(a), for the case with h2 = 7/24, no temperature fluctuation
3.2. Flow features in the fully-developed stage
is seen, suggesting the flow above the upstream fin is stable, which
is consistent with the isotherm results given in Fig. 10(a). Hence,
After a lengthy transitional stage, the flow in the cavity
this corresponds to the absence of a dominant frequency, i.e. we
ultimately enters a fully-developed stage, in which a cavity-wide
can take f1 = 0 in this scenario. Strong temperature oscillations
temperature stratification is established. Fig. 9 presents the tem-
are then seen for the cases with h2  1/3 where distinct dominant
perature profile at the x = 2.083 plane and an approximate linear
frequency can be obtained from the corresponding spectral analy-
temperature stratification can be discerned. It is interesting to note
sis in Fig. 11(f)–(j). It is further noticed that with the downstream
that the temperature profiles subject to different fin configurations
almost overlap with each other suggesting the effect of fin position
on the temperature profile in the cavity interior is minor for the
current problem.

Fig. 10. Flow adjacent to the heated sidewall at t = 1282.05 for different fin
Fig. 9. Temperature profiles at the center plane of the cavity in the fully-developed position at Ra = 1.84
109 (isotherms from 0.5 to 0.5 with an interval of 0.021).
stage at Ra = 1.84
109. (a) h2 = 7/24, (b) h2 = 1/3, (c) h2 = 5/12, (d) h2 = 1/2, (e) h2 = 7/12.
 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447 7

Fig. 11. Temperature time series and the corresponding power spectra at the monitoring point P1 at the fully-developed stage at Ra = 1.84
109. (a) and (f) h2 = 7/24, (b) and
(g) h2 = 1/3, (c) and (h) h2 = 5/12, (d) and (i) h2 = 1/2, (e) and (j) h2 = 7/12, (k) f1 versus h2.

fin moving towards the ceiling, the temperature varies differently ing horizontally towards the fin tip. It eventually becomes a part of
with time at the monitoring point P1 which is mainly caused by the subsequent downstream plume after bypassing the down-
the presence of the downstream fin. Temperature oscillates irregu- stream fin. Further, it is seen that the downstream plume is dis-
larly for the case with h2 = 5/12. Meanwhile, for the case with torted in comparison to the upstream one, which indicates that
h2 = 7/12, bifurcation of temperature can be seen in Fig. 11(e). the flow above the downstream fin is generally more unstable than
These findings suggest that the downstream fin has a dramatic the one above the upstream fin due to the strong interaction
effect on the upstream unstable thermal layer flow which is less between the two plumes.
recognized in the literature.
Temperature time series of the cases at the other two Rayleigh
numbers are analyzed simultaneously by the spectral analysis and
the corresponding dominant frequencies are summarized in Fig. 11
(k). It can be found that f1 is zero for all the Rayleigh numbers
investigated in this study for h2 = 7/24. It is also clear that f1
increases with Rayleigh number for all the other fin positions. In
addition, f1 first grows at the increasing of h2, which then decreases
beyond a critical value at h2 = 5/12. It is also worth noting that the
maximum of f1 is less than the numerical results reported in [45],
in which the case was with only one thin fin attached to the side-
wall of the cavity. This indicates that the presence of the down-
stream fin tends to suppress the thermal flow above the
upstream fin.
To understand the interaction process of the two plumes, Fig. 12
presents series of isotherms in the vicinity of the heated sidewall of
a typical case with h2 = 5/12 and Ra = 1.84
109. The front posi-
tions of the upstream plume are marked by the black arrows. It
Fig. 12. Thermal flow adjacent to the heated sidewall with h2 = 5/12 and
is seen in Fig. 12 that the upstream plume first reattaches to the Ra = 1.84
109. (isotherms ranging from 0.5 to 0.5 with an interval of 0.02). (a)
sidewall and it ascends vertically in the thermal boundary layer. t = 1283.12, (b) t = 1283.55, (c) t = 1284.40, (d) t = 1285.15, (e) t = 1285.68, (f)
Then it gradually merges into the downstream intrusion and mov- t = 1286.22.
8 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447

As discussed above, when the evolutionary length between the

two fins is long enough, the upstream plume is well developed and
it separates from the upstream fin periodically with a frequency of
f1. This leads to a homogenous downstream plume flow with an
oscillation frequency of f2, and this oscillatory frequency subse-
quently propagates adjacent to heated sidewall resulting in one
peak frequency in the downstream thermal boundary layer, i.e.
f3. In order to further understand the impact of plume interaction
on the flow, spectral analyses are carried out.
Fig. 13 presents the spectral power of temperature time series
at the monitoring points P1, P2 and P3 for different fin positions
at Ra = 1.84
109. Fig. 13(a) shows that f2 equals f3 for the case
with h2 = 7/24, which can be explained by the fact that only one
plume presents in this scenario and this is in fact similar to the cav-
ity flow with only one fin at the sidewall (see e.g. [50,45]). In com- Fig. 14. Dominant frequency at P3 in the thermal boundary layer.
parison, it is found in Fig. 13(b) and (c) that f3 is equal to the sum of
f1 and f2 when h2 = 1/3 and 5/12, respectively. Nevertheless, it is
interesting to note that the behavior of f1 = f2 = f3 is observed for Local Reynolds number (Rel) and Rayleigh number (Ral) are
the cases with h2 = 1/2 and h2 = 7/12, as shown in Fig. 13(d) and adopted to quantify and analyze the flow above the upstream
(e). In addition, it needs to be clarified that for all the fin configu- and downstream fins. They are defined as:
rations examined, f1, f2 and their harmonic frequencies also present
at the downstream thermal boundary layer (refer to the results of umax dlayer
Rel ¼ ð9Þ
f3) even though they might be much weaker than the dominant fre- m
quencies. This complicated and compelling process reflects the
strong non-linear characteristics of the plume interaction process. gbDT layer d3layer
Fig. 14 presents the results of f3 at different fin positions and Ral ¼ ð10Þ
mk
Rayleigh numbers. It is clear that f3 increase with the Rayleigh
number. Meanwhile, f3 depends on the downstream fin position. where the umax, dlayer and DT layer are the maximum horizontal veloc-
It first increases with h2, then it decreases beyond h2 = 5/12 for ity, thickness and temperature difference of the unstable fluid layer
all the Rayleigh numbers studied, which indicates that the thermal which are measured at the fin tip position as schematically marked
boundary layer is more unstable for the cases with h2 = 5/12. by the black dash lines in Fig. 10.

Fig. 13. Power spectra of the temperature time histories obtained at P1, P2 and P3 with Ra = 1.84
109. (a) h2 = 7/24, (b) h2 = 1/3, (c) h2 = 5/12, (d) h2 = 1/2, (e) h2 = 7/12.
 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447 9

Fig. 15 plots the Rel and Ral of the fluid layers at the upstream
and downstream fins respectively with Ra = 1.84
109. It can be
found in Fig. 15(a) that both Rel and Ral at the upstream fin tip
minimize at h2 = 7/24 and maximize at h2 = 7/12. It suggests that
the presence of the downstream fin generally suppresses the flow
above the upstream fin. On the other hand, it is seen in Fig. 15(b)
that, for the thermal flow above the downstream fin, the Rel peaks
at h2 = 5/12, while Ral reaches maximum at h2 = 7/24 as a result of
greater DT layer in the unstable layer.

3.3. Heat transfer enhancement

As discussed above, the flows adjacent to the heated sidewall

are appreciably affected by the two fins. Accordingly, the heat
transfer across the cavity is expected to be different from the cavity
without any fin. For the purpose of precisely evaluating the
enhancement or depression of the heat transfer through the side-
wall, an enhancement factor, e, is defined as below, which is simi-
lar to that used in [42],
e ¼ ðNufin  Nunofin Þ=Nunofin ð11Þ
Fig. 16(a) presents the time histories of e at the heated sidewall.
It is clear that e is positive, implying that the heat transfer through-
out the heated sidewall is improved for all the fin configurations
examined in the present study compared to the case without any
fin. It is also found that e is higher in the early stage, reaching a
maximum value of 0.356. In addition, it is seen from Fig. 16(b) that
e increases monotonically with h2 for the cases with Ra = 9.2
108.
It is also demonstrated that the e first increases and then decreases,
Fig. 16. (a) The time series of heat transfer enhancement factor e for the case with
reaching maximum at h2 = 1/2 for the cases with Ra = 1.84
109 Ra = 3.68
109 and h2 = 5/12. (b) Time-averaged e at different Rayleigh numbers
and it peaks at h2 = 5/12 for the cases with Ra = 3.68
109. The and fin positions in the fully-developed stage.
best heat transfer improvement achieved is 12.7% at the Rayleigh
number of 3.68
109 and h2 = 5/12. The flow rate and heat transfer obvious that the flow rate and heat transfer rate are distinctly
enhancement factor in the one-fin cavity at Ra = 1.84
109, as enhanced. In addition, it is also seen in Fig. 16(b) that, the optimal
reported in [42], are approximately 0.0046 and 0.07 respectively, h2 shifts to a lower streamwise location when the Rayleigh number
while the corresponding values in the present study are 0.0065 increases.
and 0.127 for the case with h2 = 1/2 and Ra = 1.84
109. It is

Fig. 15. Rel and Ral against fin position with Ra = 1.84
109. (a) At the upstream fin Fig. 17. Horizontal flow rate Q at the centerline of the cavity at Ra = 1.84
109. (a)
tip. (b) At the downstream fin tip. Time series of Q, (b) time-averaged Q in the fully-developed stage.
10 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447

The heat transfer enhancement across the cavity is not only can be further augmented if both fins are properly configured. This
associated with the flow near the heated and cooled sidewalls, is also worth a separate study.
but also the horizontal flow across the whole cavity. The horizontal
flow rate at the centerline of the cavity is defined as: Declaration of Competing Interest

Z 1
1 The authors declared that there is no conflict of interest.
Q¼ jujdy ð12Þ
2 0
Acknowledgements
Fig. 17 presents the time series of Q and its time-averaged val-
ues for Ra = 1.84
109 with different fin positions. For the case The financial support from the Fundamental Research Funds for
with Ra = 1.84
109 and h2 = 7/12, the flow rate is increased by the Central Universities of China (DUT19LK23) is acknowledged.
up to 136.2% in the early stage, whilst it is improved by up to The financial support by the National Science Foundation of China
124.8% in the fully-developed stage compared to the case without (Project 11602215) and the Natural Science Foundation of Jiangsu
any fin. In addition, it is found that the flow rate Q keeps oscillating Province (BK20160453) is also acknowledged.
in the cases with h2 = 5/12 and h2 = 1/2, whilst it remains constant
for the other cases as demonstrated in Fig. 17(a). In the meanwhile,
References
it can be seen in Fig. 17(b) that the time-averaged Q in the fully-
developed stage for all the two-fin cases is higher than the non- [1] J.F. Torres, F. Ghanadi, I. Nock, M. Arjomandi, J. Pye, Mixed convection around a
finned results, which suggests that the horizontal flow in the cavity tilted cuboid with an isothermal sidewall at moderate Reynolds numbers, Int.
is greatly enhanced due to the presence of the two thin fins. J. Heat Mass Transf. 119 (2018) 418–432.
[2] W.X. Lin, S.W. Armfield, Scalings for unsteady natural convection boundary
layers on a vertical plate at time-dependent temperature, Int. J. Therm. Sci. 111
(2017) 78–99.
[3] Y.L. Zhao, C.W. Lei, J.C. Patterson, A PIV measurement of the natural transition
4. Conclusions of a natural convection boundary layer, Exp. Fluids 56 (9) (2015) 1–10.
[4] Y.L. Zhao, C.W. Lei, J.C. Patterson, The K-type and H-type transitions of natural
In this paper, the transient natural convection flow and heat convection boundary layers, J. Fluid Mech. 824 (2017) 352–387.
[5] Y.L. Zhao, C.W. Lei, J.C. Patterson, PIV measurements of the K-type transition in
transfer in a differentially heated cavity with two horizontal adia- natural convection boundary layers, Exp. Therm. Fluid Sci. 101 (2019) 62–75.
batic fins on each sidewall are numerically studied at a series of [6] Y.S. Chen, J. Tian, S.D. Sun, Q. Sun, Y. Fu, Z.F. Tang, H.H. Zhu, N.X. Wang,
Rayleigh numbers and fin positions. The flow characteristics at Characteristics of the laminar convective heat transfer of molten salt in
concentric tube, Appl. Therm. Eng. 125 (2017) 995–1001.
the early and fully-developed stages are described and analyzed
[7] M. Salari, A. Kasaeipoor, E.H. Malekshah, Influence of static bubbles at the
in detail, whilst the heat transfer enhancement at the fully- surface of electrodes on the natural convection flow for application in high
developed stage is also discussed. performance lead-acid battery, Therm. Sci. Eng. Prog. 5 (2018) 204–212.
At the early stage, the downstream intrusion is relatively weak [8] J. Khedari, P. Yimsamerjit, J. Hirunlabh, Experimental investigation of free
convection in roof solar collector, Build. Environ. 37 (5) (2002) 455–459.
and only one plume develops when the two fins are close to each [9] L. Tagliafico, G. Tanda, Radiation and natural convection heat transfer from
other, i.e. h2  1/3 (h2 referring to the distance form the leading wire-and-tube heat exchangers in refrigeration appliances, lnt. J. Refrig. 20 (7)
edge of the sidewall to the position of the downstream fin). In com- (1997) 461–469.
[10] A. Rahimi, A. Surendar, A. Kasaeipoor, P. Hooshmand, E.H. Malekshah, Lattice
parison, a double-plume flow regime is identified for the cases Boltzmann simulation of nanofluid flow and heat transfer in a hollow multi-
with h2  5/12. Likewise, at the fully-developed stage, the pipe heat exchanger considering nanoparticles’ shapes, Powder Technol. 339
double-plume flow regime is observed except for the case with (2018) 974–984.
[11] A. Rahimi, A.D. Saee, A. Baghban, A. Kasaeipoor, H. Ashrafi, E.H. Malekshah,
h2 = 7/24. It is further found that the position of the downstream Double-MRT lattice Boltzmann simulation of natural convection in a C-shaped
fin dramatically affects the process of plume interaction both at heat exchanger, Powder Technol. 336 (2018) 465–480.
the early and fully-developed stages. [12] A. Rahimi, A. Amiri, A. Kasaeipoor, E.H. Malekshah, Heat transfer enhancement
using Al2O3-EG/W(60/40vol%) in multiple-pipe heat exchanger, J. Mol. Liq.
The power spectral analyses of the flow characteristics obtained 261 (2018) 319–336.
at the fully-developed stage demonstrate that f1 presents a non- [13] A. Rahimi, A. Kasaeipoor, E.H. Malekshah, A.S. Far, M. Sepehr, Heat transfer
monotonic dependence of h2, which peaks at h2 = 5/12. It is also intensification using CuO-water nanofluid in a finned capsule-shaped heat
exchanger using lattice Boltzmann method, Chem. Eng. Proces. – Process
interesting to note that the maximum frequency identified with
Intensification 127 (2018) 17–27.
h2 = 5/12 is less than the one obtained in the one-fin cavity, which [14] A. Rahimi, A. Kasaeipoor, E.H. Malekshah, M. Palizian, L. Kolsi, Lattice
indicates that the downstream fin suppresses the flow above the Boltzmann numerical method for natural convection and entropy generation
upstream fin. In addition, f3 first increases with h2, then it in cavity with refrigerant rigid body filled with DWCNTs-water nanofluid-
experimental thermo-physical properties, Therm. Sci. Eng. Prog. 5 (2018) 372–
decreases beyond h2 = 5/12. It is also found that both Rel and Ral 387.
at the upstream fin tip minimize at h2 = 7/24 and maximize at [15] A. Rahimi, M. Sepehr, M.J. Lariche, A. Kasaeipoor, E.H. Malekshah, L. Kolsi,
h2 = 7/12. Nevertheless, for the thermal flow above the down- Entropy generation analysis and heatline visualization of free convection in
nanofluid (KKL model-based)-filled cavity including internal active fins using
stream fin, the Rel peaks at h2 = 5/12, while Ral reaches its maxi- lattice Boltzmann method, Comput. Math. Appl. 75 (5) (2018) 1814–1830.
mum at h2 = 7/24. [16] A. Rahimi, M. Rahjoo, S.S. Hashemi, M.R. Sarlak, M.H. Malekshah, E.H.
Furthermore, heat transfer through the cavity is enhanced sig- Malekshah, Combination of Dual-MRT lattice Boltzmann method with
experimental observations during free convection in enclosure filled with
nificantly in comparison with the case without any fin. It is MWCNT-MgO/Water hybrid nanofluid, Therm. Sci. Eng. Prog. 5 (2018) 422–
improved by up to 12.7% with the Rayleigh number of 3.68
109 436.
and h2 = 5/12. Additionally, for the case with Ra = 1.84
109 and [17] A.S. Dogonchi, A.J. Chamkha, D.D. Ganji, A numerical investigation of magneto-
hydrodynamic natural convection of Cu–water nanofluid in a wavy cavity
h2 = 7/12, the flow rate is increased by 136.2% at the early stage, using CVFEM, J. Therm. Anal. Calorim. 135 (4) (2019) 2599–2611.
whilst it is improved by 124.8% at the fully-developed stage com- [18] A.S. Dogonchi, M.A. Ismael, A.J. Chamkha, D.D. Ganji, Numerical analysis of
pared to the case without any fin. The time-averaged flow rates at natural convection of Cu–water nanofluid filling triangular cavity with
semicircular bottom wall, J. Therm. Anal. Calorim. 135 (6) (2019) 3485–3497.
the fully-developed stage for all the present two-fin cases are much
[19] S.M. Seyyedi, A.S. Dogonchi, D.D. Ganji, M. Hashemi-Tilehnoee, Entropy
greater than the non-finned result, which suggest that the horizon- generation in a nanofluid-filled semi-annulus cavity by considering the
tal flow in the cavity is greatly intensified due to the presence of shape of nanoparticles, J. Therm. Anal. Calorim. (2019).
the two adiabatic thin fins. Nevertheless, the position of the [20] S.M. Seyyedi, N. Sahebi, A.S. Dogonchi, M. Hashemi-Tilehnoee, Numerical and
experimental analysis of a rectangular single-phase natural circulation loop
upstream fin may also have great impact on the flow and heat with asymmetric heater position, Int. J. Heat Mass Transf. 130 (2019) 1343–
transfer. It is expected that the heat transfer across the cavity 1357.
 Y. Liu et al. / International Journal of Heat and Mass Transfer 142 (2019) 118447 11

[21] A.S. Dogonchi, M. Waqas, S.M. Seyyedi, M. Hashemi-Tilehnoee, D.D. Ganji, [38] S. Paolucci, D.R. Chenoweth, Transition to chaos in a differentially heated
CVFEM analysis for Fe3O4–H2O nanofluid in an annulus subject to thermal vertical cavity, J. Fluid Mech. 201 (1989) 379–410.
radiation, Int. J. Heat Mass Transf. 132 (2019) 473–483. [39] Z.L. Gao, A. Sergent, B. Podvin, S.H. Xin, P. Le Quere, L.S. Tuckerman, Transition
[22] A.S. Dogonchi, Hashim, Heat transfer by natural convection of Fe3O4-water to chaos of natural convection between two infinite differentially heated
nanofluid in an annulus between a wavy circular cylinder and a rhombus, Int. vertical plates, Phys. Rev. E 88 (2) (2013) 1–18.
J. Heat Mass Transf. 130 (2019) 320–332. [40] I.V. Miroshnichenko, M.A. Sheremet, Turbulent natural convection heat
[23] A.H. Pordanjani, S.M. Vahedi, F. Rikhtegar, S. Wongwises, Optimization and transfer in rectangular enclosures using experimental and numerical
sensitivity analysis of magneto-hydrodynamic natural convection nanofluid approaches: a review, Renew. Sust. Energ. Rev. 82 (2018) 40–59.
flow inside a square enclosure using response surface methodology, J. Therm. [41] F. Xu, J.C. Patterson, C.W. Lei, An experimental study of the unsteady thermal
Anal. Calorim. 135 (2) (2019) 1031–1045. flow around a thin fin on a sidewall of a differentially heated cavity, Int. J. Heat
[24] S.M. Vahedi, A.H. Pordanjani, S. Wongwises, M. Afrand, On the role of Fluid Flow 29 (4) (2008) 1139–1153.
enclosure side walls thickness and heater geometry in heat transfer [42] F. Xu, J.C. Patterson, C.W. Lei, Transition to a periodic flow induced by a thin fin
enhancement of water–Al2O3 nanofluid in presence of a magnetic field, J. on the sidewall of a differentially heated cavity, Int. J. Heat Mass Transf. 52 (3–
Therm. Anal. Calorim. (2019). 4) (2009) 620–628.
[25] G.K. Batchelor, Heat transfer by free convection across a closed cavity between [43] F. Xu, J.C. Patterson, C.W. Lei, Effect of the fin length on natural convection flow
vertical boundaries at different temperatures, Q. Appl. Math. 12 (3) (1954) transition in a cavity, Int. J. Therm. Sci. 70 (2013) 92–101.
209–233. [44] F. Xu, J.C. Patterson, C.W. Lei, Experimental observations of the thermal flow
[26] F. Xu, J.C. Patterson, C.W. Lei, Shadowgraph observations of the transition of around a square obstruction on a vertical wall in a differentially heated cavity,
the thermal boundary layer in a side-heated cavity, Exp. Fluids 38 (6) (2005) Exp. Fluids 40 (3) (2006) 364–371.
770–779. [45] Y. Liu, C.W. Lei, J.C. Patterson, Plume separation from an adiabatic horizontal
[27] J.C. Patterson, J. Imberger, Unsteady natural-convection in a rectangular cavity, thin fin placed at different heights on the sidewall of a differentially heated
J. Fluid Mech. 100 (1980) 65–86. cavity, Int. Commun. Heat Mass Transfer 61 (2015) 162–169.
[28] S.W. Armfield, R. Janssen, A direct boundary-layer stability analysis of steady- [46] Y. Liu, C.W. Lei, J.C. Patterson, Natural convection in a differentially heated
state cavity convection flow, Int. J. Heat Fluid Flow 17 (6) (1996) 539–546. cavity with two horizontal adiabatic fins on the sidewalls, Int. J. Heat Mass
[29] Y. Liu, Y. Bian, Y. Zhao, S. Zhang, Q. Suo, Scaling laws for the transient Transf. 72 (2014) 23–36.
convective flow in a differentially and linearly heated rectangular cavity at Pr > [47] F. Xu, J.C. Patterson, C.W. Lei, Transient natural convection flows around a thin
1, Phys. Fluids 31 (4) (2019) 043601. fin on the sidewall of a differentially heated cavity, J. Fluid Mech. 639 (2009)
[30] J.W. Elder, Laminar free convection in a vertical slot, J. Fluid Mech. 23 (1965) 261–290.
77–98. [48] J. Ma, F. Xu, Dynamic and heat transfer of a Pr < 1 thermal flow around a fin on
[31] A. Elatar, M.A. Teamah, M.A. Hassab, Numerical study of laminar natural the thermal wall, Int. J. Heat Mass Transf. 111 (2017) 696–708.
convection inside square enclosure with single horizontal fin, Int. J. Therm. Sci. [49] J. Ma, F. Xu, S.C. Saha, Flows and heat transfer of the transition to an unsteady
99 (2016) 41–51. state in a finned cavity for different Prandtl numbers, Int. Commun. Heat Mass
[32] H.T. Chen, M.C. Lin, J.R. Chang, Numerical and experimental studies of natural Transfer 88 (2017) 220–227.
convection in a heated cavity with a horizontal fin on a hot sidewall, Int. J. Heat [50] J. Ma, F. Xu, Unsteady natural convection and heat transfer in a differentially
Mass Transf. 124 (2018) 1217–1229. heated cavity with a fin for high Rayleigh numbers, Appl. Therm. Eng. 99
[33] X.D. Shi, J.M. Khodadadi, Laminar natural convection heat transfer in a (2016) 625–634.
differentially heated square cavity due to a thin fin on the hot wall, J. Heat [51] S.V. Patankar, D.B. Spalding, Calculation procedure for heat, mass and
Transfer 125 (4) (2003) 624–634. momentum-transfer in 3-dimensional parabolic flows, Int. J. Heat Mass
[34] W. Al-Kouz, A. Alshare, S. Kiwan, A. Al-Muhtady, A. Alkhalidi, H. Saadeh, Two- Transf. 15 (10) (1972) 1787–1806.
dimensional analysis of low-pressure flows in an inclined square cavity with [52] B.P. Leonard, Stable and accurate convective modeling procedure based on
two fins attached to the hot wall, Int. J. Therm. Sci. 126 (2018) 181–193. quadratic upstream interpolation, Comput. Meth. Appl. Mech. Eng. 19 (1)
[35] H.-J. Hsu, Y.-H. Huang, Y.-H. Liu, Natural convection in an oscillating (1979) 59–98.
cylindrical enclosure with pin fins, Int. J. Heat Mass Transf. 93 (2016) 720–728. [53] S.V. Patankar, Numerical Heat Transfer and Fluid Flow. Hemisphere,
[36] M. Yousaf, S. Usman, Natural convection heat transfer in a square cavity with Washington, D.C., 1980.
sinusoidal roughness elements, Int. J. Heat Mass Transf. 90 (2015) 180–190. [54] W.X. Lin, S.W. Armfield, J.C. Patterson, C.W. Lei, Prandtl number scaling of
[37] J. Tao, P. Le Quere, S. Xin, Absolute and convective instabilities of natural unsteady natural convection boundary layers for Pr > 1 fluids under isothermal
convection flow in boundary-layer regime, Phys. Rev. E 70 (6) (2004) 7. heating, Phys. Rev. E 79 (6) (2009) 1–8.
 Update
International Journal of Heat and Mass Transfer
Volume 170, Issue , May 2021, Page

DOI: https://doi.org/10.1016/j.ijheatmasstransfer.2019.118736
 International Journal of Heat and Mass Transfer 170 (2021) 118736

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Erratum

Erratum to “Enhancing the flow and heat transfer in a convective

cavity using symmetrical and adiabatic twin fins” [International
Journal of Heat and Mass Transfer 142 (2019) 118447]
Yang Liu a,∗, Shuaikun Zhang a, Han Huang d, Qinghui Suo b, Yongning Bian a,∗,
Yongling Zhao c,∗
a
School of Ocean Science and Technology, Dalian University of Technology, Panjin 124221, China
b
School of Engineering and Technology, Southwest University, Chongqing 400715, China
c
School of Aerospace, Mechanic and Mechatronic Engineering, University of Sydney, NSW 2006, Australia
d
Eindhoven University of Technology, Netherlands

The publisher regrets that the Acknowledgments section of this paper was published with the incorrect information. The correct Ac-
knowledgments statement is given here:

Acknowledgements
The financial support from the Fundamental Research Funds for the Central Universities of China (DUT19LK23) is acknowledged. The
financial support by the National Science Foundation of China (Project 11602215) and the Natural Science Foundation of Jiangsu Province
(BK20160453) is also acknowledged.
The publisher would like to apologise for any inconvenience caused.

DOI of original article: 10.1016/j.ijheatmasstransfer.2019.118447

∗
Corresponding authors.
E-mail addresses: [email protected] (Y. Liu), [email protected] (Y. Bian), [email protected] (Y. Zhao).

https://doi.org/10.1016/j.ijheatmasstransfer.2019.118736
0017-9310/© 2019 Elsevier Ltd. All rights reserved.