You are on page 1of 39

Natural Convection Heat Transfer of Viscoelastic Fluids

in a Horizontal Annulus

A Thesis Submitted
in Partial Fulfillment of the Requirements for the Degree of

Master of Technology
in
Chemical Engineering

by

Pyari Mohan Sahu


(2019CHM1008)

to the
Department of Chemical Engineering
Indian Institute of Technology Ropar
Punjab, India-140001
July 2021

1
Certificate

It is certified that the work contained in this thesis entitled “Natural Convection Heat Transfer
of Viscoelastic Fluids in a Horizontal Annulus” by Pyari Mohan Sahu, has been carried under
my supervision and that this work has not been submitted elsewhere for a degree.

Supe
rvisor’s signature Date: 12.07.2021

Name: Dr. Chandi Sasmal


Po
sition: Assistant Professor
D
epartment: Chemical Engineering

2
Acknowledgements

First and foremost, I would like to convey my sincere thanks to my thesis supervisor, Dr.
Chandi Sasmal for his continuous invaluable support, encouragement, suggestions and
guidance for completion of this work. Moreover, thanks again to him for giving me the
opportunity to get the insight into the areas of “Computational Fluid Dynamics, Heat Transfer
and Complex Fluids”. I heartily thank to all my course instructors who exposed me to many
new frontiers of knowledge.
I would like to thank to all my lab mates for giving me such a welcoming and energetic
environment in the lab. I would also like to thank the Department of Chemical Engineering,
IIT Ropar for giving me the opportunity to be a part of this family. My heartfelt thanks to all
of my friends in IIT Ropar for giving me the emotional support during the worst time of my
life, and also giving me the many enjoyable and hilarious moments which I will cherish for
the rest of my life.

Date: 12.07.2021 Pyari Mohan Sahu

IIT Ropar

3
Contents
Certificate
Acknowledgements
List of figures
List of tables
Nomenclature
1. Introduction
2. Literature review
3. Problem statement
4. Governing equations and boundary conditions
5. Numerical solution procedure and grid independence study
6. Results and discussion
6.1. Validation
6.2. Streamlines and velocity magnitude plots
6.3. Isotherm contours
6.4. Variation of the local Nusselt number
6.5. Variation of the average Nusselt number
7. Conclusions and future perspectives

References

4
List of Tables

Table 1: Details of grid independence study

List of Figures

Figure 1: Natural convection heat transfer from a heated vertical flat plate [1].
Figure 2: Schematic of the present problem

Figure 3: Schematic of the grid structure at different grid densities (a) G1 (b) G2 and (c) G3.

Figure 4: (a) Natural convection in a square cavity (b) Comparison between the present
results on the temperature distribution with the results of Dixit and Babu [29]

Figure 5: Comparison between the present results with that of Kuehn and Goldstein [16] on
the distribution of the temperature distribution. Here the symbols and broken lines are the
experimental and numerical results of Kuehn and Goldstein [16], whereas the solid lines are
the present results.

Figure 6: Comparison between the present results with that of Kuehn and Goldstein [16] on
the variation of the equivalent thermal conductivity.

Figure 7: Streamlines and velocity magnitude plots at

(a) Ra  10 ,   0.9, Wi  1, L  10 (b) Ra  10 ,   0.9, Wi  1, L  10


3 2 6 2

(c) Ra  10 ,   0.5, Wi  1, L  500 (d) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

Figure 7: Streamlines and velocity magnitude plots at

(e) Ra  10 ,   0.9, Wi  100, L  500 (f) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

(g) Ra  10 ,   0.5, Wi  100, L  10 (h) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

Figure 8: Surface distribution of isotherm contours

(a) Ra  10 ,   0.9, Wi  1, L  10 (b) Ra  10 ,   0.9, Wi  1, L  10


3 2 6 2

(c) Ra  10 ,   0.5, Wi  1, L  500 (d) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

Figure 8: Surface distribution of isotherm contours

5
(e) Ra  10 ,   0.9, Wi  100, L  500 (f) Ra  10 ,   0.5, Wi  100, L  500
6 2 6 2

(g) Ra  10 ,   0.5, Wi  100, L  10 (h) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

Figure 9: Distribution of local Nusselt number along the surface of the inner cylinder.

Figure 10: Variation of the average Nusselt number with the Rayleigh number both for

Newtonian and FENE-P viscoelastic fluids with   0.5, Wi  10, L  10 .


2

Figure 11: Variation of the average Nusselt number with the Weissenberg number and
polymer extensibility parameter at Ra = 103 (a) and 106 (b).

Figure 12: Variation of the average Nusselt number with the polymer viscosity ratio at two
values of the Rayleigh number, namely, 103 and 106.

6
1. Introduction

Natural or free convection is a mode of heat transfer which happens without the presence of
any external agency like pump, blower, etc. In this mode of heat transfer, the driving force is
the buoyancy force which is originated due to the density difference in the fluid, resulting
from the presence of a temperature difference in the system. For example, in the following
figure, a hot plate whose surface is maintained at a temperature Ts is placed in a cold fluid

environment of temperature T such that Ts  T . Due to this existing temperature difference,


the fluid which is nearer to the hot plate gets hotter and rises up due to the decrease in the
density; whereas, the fluid which is far away from the surface and whose density is more,
moves towards the hot surface. This results a circulation in the fluid near hot plate, which in
turn, causes the heat transfer which is known as the natural convection heat transfer [1].

Figure 1: Natural convection heat transfer from a heated vertical flat plate [1].

This mode of heat transfer has many practical applications. In particular, natural convection
heat transfer in a differentially heated confined space like horizontal annulus or cavity is

7
considered as one of the benchmark problems in the helm of convective heat transfer
phenomena for many decades. This problem has not only of theoretical importance, but is
also encountered in many practical applications like sterilization of foods, concentric heat
exchangers, electrical and nuclear industries, solar energy, solar desalination systems, melting
and solidification processes, material processing, etc [2-6]. Due to these extensive
applications over a wide range of industrial settings, over the years, an enormous amount of
studies for this benchmark geometry, consisting of both numerical and experimental, have
been carried out when the confined space is filled with simple Newtonian fluids like air or
water.

However, it is readily acknowledged that many fluids comprising high molecular weight
polymers like polymer melts and solutions, or of multi-phase nature like slurries, emulsions,
etc. exhibit complex non-Newtonian characteristics like shear-dependent viscosity, yield
stress, viscoelasticity, etc. [7]. These fluids are frequently encountered in many industrial
settings like petroleum and polymer processing, personal and healthcare product
manufacturing, food processing, etc. In spite of their frequent occurrence in a broad spectrum
of industrial settings, very little is known about the natural convection heat transfer
phenomena in these fluids, especially with regard to the influence of fluid viscoelasticity. It
has been found both experimentally and numerically that the addition of small amount of
polymers into a Newtonian solvent like water dramatically changes the flow behaviour of the
resulting viscoelastic polymer solution as compared to that seen in water alone. For instance,
the existence of elastic instability and elastic turbulence at low Reynolds numbers regime [8-
10] and the turbulent drag reduction phenomena at high Reynolds numbers regime [11-13]
are typical examples of distinct flow characteristics of a viscoelastic polymer solution which
are not evident in a Newtonian fluid like water. Therefore, one would expect a substantial
change in the corresponding heat and mass transfer phenomena in these viscoelastic fluids in
comparison to that seen in a Newtonian fluid. For instance, in recent studies, it has been
observed that the heat transfer rate in a viscoelastic fluid was substantially increased in a
curved or serpentine microchannel due to the existence of elastic turbulence [14, 15].

Despite the presence of non-Newtonian fluids, particularly the viscoelastic fluids, in scores of
industrial settings and practical applications, very little is known about the natural convection
heat transfer in these fluids. Although some studies have been carried out, but those are again
mostly pertained to generalized Newtonian fluids (GNF) or inelastic non-Newtonian fluids.

8
There is almost no study available on how the fluid viscoelasticity could influence this mode
of heat transfer. Therefore, this study aims to fill this gap of knowledge in the literature. In
particular, this study considers the problem of natural convection heat transfer in a
viscoelastic fluid when it is confined between two concentric cylinders and/or in a horizontal
annulus.
2. Literature review

As mentioned in the previous section, the study on natural convection heat transfer in a
horizontal annulus is mostly carried out for simple Newtonian fluids like water or air, and
there are some limited studies available for non-Newtonian fluids which again mostly include
generalized non-Newtonian fluids or GNF fluids. Therefore, in the first half of the literature
review, some important studies which have been carried out for Newtonian fluids like water
or air, are reviewed and then those for non-Newtonian fluids are summarized in the second
half of the literature review.

Kuehn and Goldstein [16] were probably the first who carried out a detailed experimental as
well as numerical study on natural convection heat transfer of both water and air in a
horizontal annulus. Their experiment was based on the Mach-Zehnder interferometer method,
whereas simulation was based the finite difference method. The study was carried out for a
fixed diameter ratio of 0.8 and for a range of Rayleigh number between 10 3 and 105. They
presented the results in terms of the variation of the local temperature distribution, velocity
field and local heat transfer coefficients, and found a good match between their experimental
and numerical results. Later, Date [17] also carried out a numerical study based on the finite
volume method and found a good agreement with the results of Kuehn and Goldstein [16].
Kumar [18] also performed a numerical study based on the finite difference method for a
wide range of the Rayleigh number and diameter ratios. They found the formation of a
crescent shaped eddy in the annuli for the small diameter ratio whereas a kidney shaped flow
pattern for large diameter ratio. They observed an increase in the heat transfer rate with the
Rayleigh number; however, this increment was larger for the constant wall temperature
boundary condition than that for the constant heat flux boundary condition. Mizushima et al.
[19] found the existence of a dual stable steady solutions once the Rayleigh number exceeded
a critical value, and they explained the origin of this dual solution based on the bifurcation
analysis. The effect of the Prandtl number on the existence of this dual solutions was studied
by Yoo [20] who determined the bifurcation points as a function of the Prandtl number. On
the other hand, Hu et al. [21] found the existence of multiple steady solutions when the

9
Rayleigh number exceeded a critical value and their simulations were based on the immersed
boundary-lattice Boltzmann method. On the other hand, the turbulent natural convection of a
Newtonian fluid in a horizontal annulus was studied by Fukuda et al. [22] both numerically
and experimentally. The numerical simulations were based on the explicit leap-frog scheme,
whereas the experiments were conducted using the hot-wire anemometer. In their analysis,
they obtained the critical values of the Rayleigh number at which the transition from the
laminar to turbulent occurs. Desai and Vafai [23] conducted numerical simulations of
turbulent natural convection with the standard    turbulent model and using the Galerkin
method of weighted residuals. The results were obtained over a wide range of Rayleigh
number (106 < Ra < 109), Prandtl number (0.01 < Pr < 5000) and radius ratio (1.5 < R < 11).
They found a good agreement between their results and prior results available in the
literature.

Therefore, from the aforementioned literature cited herein, it is clear that a significant number
of studies have been carried out for the present problem with simple Newtonian fluids like
water or air. Some studies were also carried for generalized Newtonian fluids like power-law
or Bingham plastic fluids. For instance, Matin and Khan [24] performed a numerical study
based on the finite volume method and SIMPLE algorithm for non-Newtonian power-law
fluids over a range of Rayleigh number between 103 and 105 and Prandtl number between 10
and 103. They found that the average Nusselt number decreases as the value of the power-law
index increases. Therefore, they suggested that pseudo-plastic fluids (with power-law index <
1) can be used as efficient coolants, whereas dilatant fluids (with power-law index > 1) can
be used as efficient insulating material. Moreover, they observed that the tendency of either
increasing or decreasing heat transfer rate with the power-law index becomes more
pronounced as the Rayleigh number increases. Recently, Laidoudi and Ameur [25] performed
numerical simulations for the horizontal annulus geometry, and once again found that the
pseudoplastic fluids promote the heat transfer rate in comparison to a Newtonian fluid under
otherwise identical conditions. On the other hand, Masoumi et al. [26] carried out a numerical
study based on the Galerkin’s weighted residual method to investigate the natural convection
of Bingham plastic fluids in a horizontal annulus over a wide range of the Rayleigh number,
Yield number, radius ratio and for a fixed value of the Prandtl number. They observed a
positive dependence of the heat transfer rate on the Rayleigh number likewise the Newtonian
fluid; however, the Yield number has a negative influence on the heat transfer rate. In fact,

10
they showed that after a critical value of the Yield number, the heat transfer is totally
governed by the conduction mode and there is no convection present.

Although some studies on generalized non-Newtonian fluids are present on the natural
convection heat transfer in a concentric annulus; however, there is no such corresponding
study available for viscoelastic fluids. For this class of fluids, the studies are mainly
constrained to relatively simple geometries like vertical or horizontal flat plates or cone. For
instance, Zhao et al. [27] performed a numerical study based on the finite difference method
for the natural convection of the fractional Maxwell viscoelastic fluids in horizontal annulus.
They found that the fractional order of the viscoelastic fluid model increases the thickness of
the velocity and thermal boundary layers, and hence decreases the heat transfer rate. Mustafa
et al. [28] did an analytical study (based on the similarity solution method) on the natural
convection heat transfer from a vertical flat plate of a viscoelastic second-grade fluid. They
observed that with the increase in the Deborah number and/or the fluid viscoelasticity, both
the local skin-friction and Nusselt number decrease. Apart from the studies on the flat plate,
there are studies also available on the natural convection of viscoelastic fluids in a square
cavity. The literature on this has been reviewed in our recent study [29]. However, as
mentioned earlier, there is no such study available for a horizontal annulus despite having
many practical applications of this particular system in scores of industrial settings and
process designs. The aim of this present study is to fill this gap of knowledge in the literature.
In particular, we aim to investigate how the fluid elasticity, shear-thinning properties,
polymer concentration, polymer size, etc. would tend to influence the heat transfer rate in a
horizontal annulus.

11
3. Problem statement

The problem considered in this study is to investigate the natural convection heat transfer of
viscoelastic fluids confined between two concentric horizontal cylinders or in a horizontal
annulus as schematically shown in Fig.2. The inner cylinder of the annulus is maintained at a
higher temperature TH whereas the outer cylinder is kept at a relatively lower temperature TC,
such that TH > TC. Both the cylinders are chosen to be sufficiently long in the z-direction so

 ()
0
that the gradients of any variable in that direction is neglected, i.e., Z and also, we have

assumed that there is no flow in the z-direction, i.e., U z  0 . Due to the existing temperature

difference ( T  TH  TC ) between the inner and outer cylinders, a buoyancy induced flow
current will be generated within the annulus, which in turn, will cause the heat transfer by the
natural convection. The thermo-physical properties of the viscoelastic fluid, e.g., specific heat

capacity (Cp ), thermal conductivity ( k ), zero-shear rate viscosity ( 0 ), polymer relaxation


time (  ), etc., are assumed to be independent of temperature except the density in the body
force term appearing in the Y-component of the momentum equation, which will be
discussed in the next section. This assumption is known as the Boussinesq approximation
which is valid for a relatively small temperature difference. According to this approximation,

   ref 1  T  T  Tref  


the variation of density with temperature is given by where  ref is the

reference fluid density at the reference temperature Tref and T is the coefficient of thermal

1 
T  
expansion at constant pressure defined as  T P . In the present study, this reference

temperature is that of the cold wall temperature of the outer cylinder, i.e., Tref  TC .

12
4. Governing equations and viscoelastic constitutive equation

Under the abovementioned simplifying assumptions, the present flow and heat transfer
phenomena will be governed by the following equations, written in their dimensionless
forms:

Equation of continuity:

ui
0
x j
(1)

Momentum equation:

ui u p    ui  1  Cij


 uj i         i 2
t x j xi Ra Pr x j  x j  Wi Ra Pr x j (2)

Energy equation:

  1    
uj   
t x j RaPr x j  x j  (3)

In the above equations, variables like length


 x  , velocity  u  , time  t  , pressure  p  ,
*
i
*
i
* *

temperature (T) and polymeric conformation tensor


C 
*
ij
have been non-dimensionalized as
follows:

xi* ui* t *uc p* T  TC Cij*


xi  , ui  , t  , p ,  , Cij  2
R uc R  ref uc2 TH  TC L0

uc  Rg T T
Where is the characteristic velocity scale, L0 is the equilibrium length of a

 ...
*

polymer molecule and is a dimensional variable. Here the polymeric conformation

Cij  Ri R j 
tensor is defined by where Ri is the dimensionless end-to-end vector of a
polymer molecule. In the present study, we have used the FENE-P (finitely extensible non-

13
linear elastic spring with the Perterlin approximation) viscoelastic constitutive equation to
realize the rheological behaviour of the polymer solution for the following reasons:

first, this model can effectively capture the finite extensibility of polymer molecules [30]. It
can thus provide bounded polymer stress and hence bounded solutions, particularly for the
problems with large strain rates and large Weissenberg numbers. Second, this model can
efficiently predict the shear-thinning property of a fluid which is shown by most of the
polymer solutions. Linear viscoelastic spring models for polymers such as Oldroyd-B cannot
represent either this finite extensibility of a polymer molecule or predict the shear-thinning
property of a polymer solution. Lastly, the FENE-P model has been widely used in many
prior studies, particularly for high Reynolds number turbulent flow problems, and captured
many corresponding experimental observations [31, 32]. This model can capture the
rheological response of many polymers like polyacrylamide (PAA), polyethylene oxide
(PEO), poly-isobutylene (PIB), etc.

The addition of polymer molecules into a Newtonian solvent does not directly influence the
convection-diffusion equation for the temperature distribution and/or the energy equation. It
directly influences the velocity field (through the momentum equation), which in turn,
influences the temperature field as the energy equation is associated with the velocity field.
Additionally, in the case of natural convection heat transfer, one can see that the momentum
equation contains a temperature dependent term and hence the momentum and energy
equations are coupled with each other in this mode of heat transfer. Therefore, one has to
solve these two equations simultaneously.

The transport equation for the evaluation of the conformation tensor for the FENE-P
viscoelastic model is given as follows [30]

Cij Cij u j ui f ( R )Cij   ij


 uk  Cik  Ckj 
t xk xk xk Wi

 ij
In the above equation, is the Kroneker delta and f ( R) is the Peterlin’s approximation of

L2  3
f ( R) 
the finite extensibility of the FENE-P model defined as L2  R where R  tr (Cij )
and L are the extension length and maximum possible extension length of a polymer

14
molecule, respectively. The relation between the conformation tensor and viscoelastic stress

 ijp  f ( R)Cij   ij
tensor is given by .

From the above non-dimensionalization, it can be seen that the present coupled flow and heat
transfer phenomena is governed by the following dimensionless parameters, namely,

 Rayleigh number (Ra)


T g TG 3
Ra 


    k 
  0   
  ref    ref Cp 
Where   and   are the kinematic viscosity and thermal diffusivity of
the viscoelastic fluid, respectively.

 Prandtl number (Pr)


Pr 

 Weissenberg number (Wi)


 uc
Wi 
G

 Viscosity ratio (  )
s

0

   s   p 
where  s is the zero-shear rate viscosity of the solvent and 0 is that of the total

p
polymer solution with as that of the contribution from the polymer. The value of  is
bounded between 0 and 1. The viscoelastic polymer solution approaches a polymer melt

when   0 and a Newtonian solvent when   1 . In addition to these dimensionless


numbers, the maximum allowable extension of a polymer molecule L2 is another
dimensionless parameter which is involved in the present study.

15
Finally, the following boundary conditions are used to complete the problem formulation.

 At all solid walls: the standard no-slip and no-penetration


 ui  0  boundary conditions
are used for the velocity.
 For the temperature,   1 and   0 are imposed at the inner and outer cylinders,
respectively.
 At all solid walls, the polymeric extra stresses are linearly extrapolated to 0.

5. Numerical solution procedure and grid independence study

All the governing equations, namely, mass, momentum and energy equations (Eqs.1-3) have
been solved using the open-source finite volume method based CFD code OpenFOAM
(version 7) [33]. Additionally, a recently developed rheoHeatFoam solver available in
rheotool [34], has been used to solve the FENE-P viscoelastic constitutive equation, Eq.4. All
the advective terms in the momentum, energy, and constitutive equations were discretized
using the high-resolution convergent and universally bounded interpolation scheme for
treatment of advection (CUBISTA) scheme for its improved iterative convergence properties
[35]. The diffusion terms in both the momentum and energy equations were discretized using
the second-order accurate Gauss linear orthogonal interpolation scheme. While all the
gradient terms in the governing equations were discretized using the Gauss linear
interpolation scheme, the time derivative terms were discretized using the Euler scheme. The
linear systems of the pressure and velocity fields were solved using the preconditioned
conjugate (PCG) solver with diagonal-based incomplete Cholesky (DIC) preconditioner and
the stress and temperature fields were solved using the preconditioned biconjugate gradient
solver (PBiCG) solver with diagonal-based incomplete LU (DILU) preconditioner [36, 37].
The pressure-velocity coupling was accomplished using the SIMPLE method, and the log-
conformation tensor approach was used to stabilize the numerical solution [38, 39].
16
Furthermore, the relative tolerance level for the pressure, velocity, temperature, and stress
fields was set as 10-10.

Next, we turn our attention to choose an optimal grid density which denotes a trade-off
between the accuracy of the numerical results on one hand and the computational time on the
other. This was achieved by performing a systematic grid independence study. For doing so,
three different grids, namely, G1 (), G2 () and G3 () with different number of hexahedral cells
on the cylinder wall as well as in the whole computational domain, were created where n and 1

n are the number of grid points in the 1 and 2 directions, respectively. After creating different
2

grid densities, simulations were run at the highest values of the Raleigh number, Weissenberg
number, polymer extensibility parameter and lowest value of the polymer viscosity ratio, i.e.,

at Ra = 106; Wi = 100; L2 = 800 and  = 0.5. After inspecting the results obtained at different
grid densities, the grid G2 with a total number of 40000 cells was found to be adequate to
carry out the present study. This was confirmed by comparing the results (obtained with
different grid densities) not only in terms of the variation of the average Nusselt number
(Table 1), but also in terms of the variation of the temperature, velocity and polymeric stress
profiles at different positions of the system. The latter is considered as the more stringent test
for choosing an optimal grid density, particularly for the simulation of viscoelastic fluids.
Furthermore, one can see that grids are more concentrated near the solid walls in order to
capture the steep gradients of velocity, temperature and polymeric stress components in these
regions, see Fig.3.

17
Figure 2: Schematic of the present problem

Figure 3: Schematic of the grid structure at different grid densities


(a) G1 (b) G2 and (c) G3.

Table 1: Details of grid independence study

Grid Elements Parameters Nu


Ns Nr Nt Ra Wi L2 Pr
G1 256 78 19968 106 100 800 7 11.72
G2 400 100 40000 106 100 800 7 11.87
G3 500 120 60000 106 100 800 7 11.88

6. Results and discussion

In this study, an extensive numerical investigation has been carried in order to elucidate the
influence of the fluid viscoelasticity on the natural convection heat transfer in a horizontal
annulus over a wide range of pertinent dimensionless numbers, namely, Rayleigh number (

103  Ra  106 ) , Weissenberg number  1  Wi  100  , polymer extensibility parameter

 10  L 2
 500 
, viscosity ratio
 0.5    0.9  and for a fixed value of the Prandtl number of
Pr = 7. However, prior to undertaking a detailed discussion, it is customary to validate the

18
present numerical set up to establish the accuracy of the present new results. This is achieved
by comparing the present results with the prior results available in the literature, as presented
in the ensuing subsection.

6.1. Validation

Before going into the detailed discussion of the present results, it is needed to validate the
present code that we have used in this study. It was achieved by presenting some validation
studies that we have carried out with the prior results available in the literature. One of the
benchmark problems, which is used to validate any new numerical code, is the natural
convection in a square cavity, see sub-Fig.4(a). This problem is being extensively used as a
benchmark problem to validate any numerical code. Therefore, at the onset of our validation,
we have also carried a validation study for this benchmark geometry. Figure 4(b) shows the
comparison between the present results on the temperature distribution with that of Dixit and
Babu [29], and excellent agreement between the two results can be seen. Next, we have done
some validation studies for the geometry that is considered in the present study, i.e., for the
horizontal annulus filled with Newtonian fluids like water or air. Figures 5 and 6 show the
comparison among the experimental and numerical results of Kuehn and Goldstein [16] and
the present numerical results on the distribution of the non-dimensional temperature and
equivalent thermal conductivity when the cavity was filled with air. In both the cases, an
excellent match can be seen. All these validation studies provide us the confidence to present
and discuss the new results.

19
(b)

Figure 4: (a) Natural convection in a square cavity (b) Comparison between the present
results on the temperature distribution with the results of Dixit and Babu [29].

Figure 5: Comparison between the present results with that of Kuehn and Goldstein [16] on
the distribution of the temperature distribution. Here the symbols and broken lines are the
experimental and numerical results of Kuehn and Goldstein [16], whereas the solid lines are
the present results.

20
Figure 6: Comparison between the present results with that of Kuehn and Goldstein [16] on
the variation of the equivalent thermal conductivity.

6.2. Streamlines and velocity magnitude plots

To show explicitly the velocity field inside the annulus, we have plotted the streamlines along
with the velocity magnitude in Fig. 7. To show the effect of the Rayleigh number on the
streamlines and velocity magnitude plots, results are presented for two values of Ra, namely,
103 (sub-Fig.7(a)) and 106 (sub-Fig.7(b)) for fixed values of the Weissenberg number,
polymer extensibility parameter and viscosity ratio. At low Ra numbers, for instance at 103
(sub-Fig.7(a)), there is a perfect symmetry present along both the horizontal and vertical
midplane passing through the origin of the annulus both in the streamlines and velocity
magnitude plots. This is mainly due to the fact that at this low Ra numbers, the strength of the
buoyancy-induced convection current is very weak and hence, the heat transfer mainly occurs
by the conduction mode. Furthermore, at this Ra number, a kidney shaped vortex is formed
inside the annulus and the velocity magnitude is always seen to be large near the vertical
sides of both the inner and outer cylinders of the annulus. However, as the Rayleigh number
gradually increases, the buoyancy-induced convection current inside the annulus also
increases, which can be evident from the ten-fold increase in the velocity magnitude
presented in sub-Fig.7(b) at Ra = 106. Due to this increase in the convection current, the
symmetry along the horizontal midplane passing through the origin is totally lost; however,
the vertical symmetry still exists. Furthermore, the kidney shaped vortex formed at low Ra
numbers is now transformed into a dumbbell one at this high Rayleigh numbers, whose
position is also now shifted upwards. Once again, a high velocity magnitude zone is seen to
form in the vicinity of vertical sides of both the cylinders; however, now the maximum
velocity is seen to appear at the top middle position of the two cylinders. A very little
convection is seen to present at the bottom space between the two cylinders.

On the other hand, the effect of the Weissenberg number on the streamlines and velocity
magnitude is shown in sub-Figs.7(c) and (d) for Wi = 1 and 100, respectively for fixed values
of the polymer extensibility parameter, viscosity ratio and Rayleigh number. At Wi = 1 (sub-
Fig.4(c)), once again, a kidney shaped recirculation region is formed. As the Weissenberg
21
number increases to 100, this recirculation region is still there; however, its shape becomes
little bit distorted, see sub-Fig.7(d). Furthermore, the velocity magnitude also increases inside
the annulus with the increases in the Weissenberg number. The reason for this can be
explained as follows: as the Weissenberg number increases, the deformation rate in the
annulus increases. This results in the increase of the shear-thinning tendency of the polymer
solution. Due to the increase of this tendency of the polymer solution, the apparent viscosity
of the polymer solution will also tend to decrease with the Weissenberg number, which
thereby resulting in the increase of the velocity magnitude as well as of the recirculation
region formed in the annulus.

22
Figure 7: Streamlines and velocity magnitude plots at

(a) Ra  10 ,   0.9, Wi  1, L  10 (b) Ra  10 ,   0.9, Wi  1, L  10


3 2 6 2

(c) Ra  10 ,   0.5, Wi  1, L  500 (d) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

On the other hand, the effect of the polymer viscosity ratio and/or the polymer concentration
on the streamlines and velocity magnitude is shown in sub-Figs.7(e) and (f) for two values of
the polymer viscosity ratio, namely, 0.9 and 0.5, respectively. It can be seen that as the
polymer concentration increases, the buoyancy-induced convection current inside the annulus
also increases, which can be evident from the increase in the velocity magnitude. Moreover,
the size of the recirculation region also increases, whose shape becomes more distorted at
lower values of the polymer viscosity ratio. This is mainly due to the increase in the shear-
thinning behaviour of the polymer solution. The effect of the polymer extensibility parameter
in shown in sub-Figs.7(g) and (h) for L2 = 10 and 500, respectively. One can see that as the
value of the polymer extensibility parameter decreases, the size of the recirculation region
inside the annulus increases. Not only that, but also a small recirculation region is formed at
the bottom space for L2 = 10, see sub-Fig.7(g). However, for L2 = 500, no such recirculation
region is present. This suggests that the flow becomes more chaotic inside the annulus as the
polymer extensibility parameter decreases. This is, once again, because of the increase in the
shear-thinning behaviour of the polymer solutions with the decreasing value of L2. Therefore,
one can expect that the heat transfer will also increase with the decreasing value of L2, which
will be discussed in the next sub-section.

23
Figure 7: Streamlines and velocity magnitude plots at

(e) Ra  10 ,   0.9, Wi  100, L  500 (f) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

(g) Ra  10 ,   0.5, Wi  100, L  10 (h) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

24
6.3. Isotherm contours

The corresponding temperature distribution inside the annulus is shown in Fig.5 in terms of
the isotherm contours. The effect of the Rayleigh number on the isotherm contours is shown
in sub-Figs.8(a) and (b). At low Rayleigh numbers, for instance at Ra = 103, the isotherm
contours look like concentric circles without any distortion in it, sub-Fig.8(a). This is due to
the fact that at this low Ra numbers, the heat transfer happens mainly due to the conduction
mode, and there is almost convection present. As a result, it causes the isotherm contours as
concentric circles. However, as the Rayleigh number progressively increases, the buoyancy-
induced convection becomes pronounced in the annulus, for instance see the results in sub-
Fig.8(b) presented at Ra = 106. At this value of Ra, the isotherms become much more
distorted in their distribution as compared to that seen at Ra = 103. Moreover, the formation
of a thermal plume can also be seen at the top space. At high Rayleigh numbers, the thermal
boundary layers are now become more concentrated in the vicinity of the cylinder surface. It
is more evident at the lower surface of the inner cylinder and at the upper surface of the outer
cylinder. Therefore, one can expect an increase in the heat transfer rate with the Rayleigh
number. On the other hand, the effect of the Weissenberg number on the isotherm contours is
depicted in sub-Figs.8(c) and (d) for the values of Wi = 1 and 100, respectively. It can be
clearly seen that as the value of the Weissenberg number increases, the thickness of the
thermal boundary layer decreases, which can be evident from the fact that the isotherms are
more crowded in the vicinity of the cylinder surface for Wi = 100 (sub-Fig.8(d)) than that
seen for Wi = 1 (sub-Fig.8(d)). This is because of the increase in the shear-thinning property
of the polymer solution with the increase in the Weissenberg number.

On the other hand, the effect of the polymer viscosity ratio and polymer extensibility
parameter on the isotherm contours is presented in sub-Figs.8(e)-(h). As the polymer
viscosity ratio decreases and/or polymer concentration increases in the solution, the isotherms
become more distorted and crowded in the vicinity of the cylinder surface, for instance see

the results in sub-Figs.8(e) and (f) for the polymer viscosity ratio   0.9 and 0.5,
respectively. Similarly, a same trend is seen when the polymer extensibility parameter
decreases, for instance see the results presented in sub-Figs.5(g) and (h) for the values of

L2  10 and 500, respectively. Therefore, one can expect an increase in the heat transfer rate

with both the decreasing values of the polymer viscosity ratio and polymer extensibility

25
parameter. This is due an increase in the shear-thinning behaviour of the polymer solution in
both the cases.

Figure 8: Surface distribution of isotherm contours

(a) Ra  10 ,   0.9, Wi  1, L  10 (b) Ra  10 ,   0.9, Wi  1, L  10


3 2 6 2

(c) Ra  10 ,   0.5, Wi  1, L  500 (d) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

26
Figure 8: Surface distribution of isotherm contours

(e) Ra  10 ,   0.9, Wi  100, L  500 (f) Ra  10 ,   0.5, Wi  100, L  500


6 2 6 2

27
(g) Ra  10 ,   0.5, Wi  100, L  10 (h) Ra  10 ,   0.5, Wi  100, L  500
6 2 6 2

6.4. Distribution of local Nusselt number

The local Nusselt number can indicate the local rate of heat transfer in the annulus, defined
by

h Ri 
Nu  
k n s innercylinder
where h is the local heat transfer coefficient and n s is the unit
outward normal vector drawn on the surface. The variation of the local Nusselt number along
the surface of the inner cylinder in shown in Fig. 9 for two values of the Rayleigh number,
namely, 103 (a) and 106 (b). It can be seen that the local Nusselt number has the maximum

value in the vicinity of   0 , and then it gradually increases as one moves towards the top

point of the cylinder or   180 . This trend is same for both Newtonian and FENE-P
viscoelastic fluids. This is due to the fact that the thermal boundary layer thickness gradually

increases as one travels along the cylinder surface from   0 towards   180 , thereby
resulting a decrease in the local Nusselt number. It was also evident in the isotherm plots
presented in preceding subsection. At low Ra and Wi numbers, the results for the Newtonian
and FENE-P viscoelastic fluids are almost same. However, as the Weissenberg number

increases, say to 100, the local Nusselt number is seen to increase at   0 and decrease at
  180 . This is mainly due to the variation of the thermal boundary layer thickness at these
two positions wherein it decreases at first position and increases at the second position. On
the other hand, at Ra = 106, a similar patten is observed; however, the local values of the
Nusselt number increases simply due to decrease in the thermal boundary thickness. In this
case, the local Nusselt number increases with the Weissenberg number. This is due to the
increase in the shear-thinning property of the polymer solution with Wi. In both the cases,
there is a perfect symmetry present in the distribution of the local Nusselt number along the
vertical mid-plane passing through the origin. This is due to the fact that the flow remains in
the steady and laminar flow state within the ranges of conditions encompassed in the present
study. This causes a perfect symmetry in the local Nusselt number distribution.

28
Figure 9: Distribution of local Nusselt number along the surface of the inner cylinder.

6.5. Variation of average Nusselt number

29
The variation of the average Nusselt number with the Rayleigh number is shown in Fig.10

both for Newtonian and FENE-P viscoelastic fluids with   0.5, Wi  10, L  10 . It can be
2

seen that irrespective of the fluid type, i.e., whether it is Newtonian or viscoelastic fluid, the
average Nusselt number always increases with the Rayleigh number. This is due to the fact
that with the increase in the Rayleigh number, the buoyancy-induced convection inside the
annulus increases, which in turn, increases the heat transfer rate and/or the Rayleigh number.
Another thing, one can see that at low Ra numbers, for instance at Ra = 103, the value of the
average Nusselt number is almost indistinguishable for Newtonian and viscoelastic fluids.
The difference between then becomes prominent as the Rayleigh number gradually increases.
This is due to the fact that as low Ra numbers, the heat transfer happens mainly due to the
conduction mode, and hence it does not matter what fluid it is (only the thermal conductivity
of the fluid matters), thereby leading to almost same heat transfer rate for both the fluids.
However, as the Ra number increases, the convection becomes dominant in the annulus, and
the viscous and elastic properties also become crucial to determine the heat transfer rate.

Figure 10: Variation of the average Nusselt number with the Rayleigh number both for

Newtonian and FENE-P viscoelastic fluids with   0.5, Wi  10, L  10 .


2

30
The corresponding variation of the average Nusselt number with the Weissenberg number
and polymer extensibility parameter is shown in Fig.11 for two values of the Rayleigh
number, namely, 103 and 106. At low Ra numbers, as pointed out above, the heat transfer
mainly happens due to the conduction mode, and as a result, there is a little variation in the
heat transfer rate with the Weissenberg number and polymer extensibility parameter, see sub-
Fig.11(a). However, as the Rayleigh number increases, say to 10 6, a clear dependence of the
average Nusselt number on these parameters can be seen. One can see that as the
Weissenberg number gradually increases, the heat transfer rate also progressively increases.
However, after a critical value of the Weissenberg number, it remains almost constant and/or
a plateau in the value of the average Number can be seen. This can be explained as follows:
as the Weissenberg number increases, the shear-thinning behaviour of the polymer solution
also increases which tends to increase the heat transfer rate. Whereas, the elastic properties of
the polymer solution also increase with the Weissenberg number, which have a tendency to
decrease the heat transfer rate by increasing the thermal boundary layer thickness. Therefore,
there is a competition present of these two oppositely acting effects on the heat transfer rate,
which ultimately leads to an almost constant heat transfer rate. On the other hand, the effect
of the polymer extensibility parameter on the heat transfer rate is more prominent at low
Weissenberg numbers. The heat transfer rate increases with the decreasing value of L2. This
is due to the dominance of the shear-thinning behaviour of the polymer solution at low values
of L2. However, at high values of the Weissenberg numbers, the dependence of the heat
transfer rate on the value of L2 is seen to be minimal.

On the other hand, the effect of the polymer viscosity ratio and/or the polymer concentration
on the variation of the average Nusselt number is shown in Fig.8. Once again, at low Ra
numbers, there is almost no effect of the polymer viscosity ratio present on the heat transfer
rate, see sub-Fig.12(a). This is because of the dominance of the heat transfer rate by the
conduction mode. As the Rayleigh number increases, there is a clear dependence can be seen,
see sub-Fig.12(b). In particular, it can be seen that the heat transfer rate increases with the

decreasing value of the polymer viscosity ratio  and/or increasing the polymer

concentration. This is mainly due to that fact that as  decreases, both the shear-thinning and
elastic properties increase due to the increase in the polymer amount. However, within the

31
present conditions encompassed here, the shear-thinning tendency will dominate over and
above the viscoelastic properties. This results in the increase of the heat transfer rate.

Figure 11: Variation of the average Nusselt number with the Weissenberg number and
polymer extensibility parameter at Ra = 103 (a) and 106 (b).

32
Figure 12: Variation of the average Nusselt number with the polymer viscosity ratio at two
values of the Rayleigh number, namely, 103 and 106.

33
7. Conclusions and future scope

In this study, we have carried out an extensive numerical investigation on the natural
convection heat transfer of viscoelastic fluids confined between two concentric cylinders
and/or in a horizontal annulus. The simulations have been carried based on the FENE-P
(Finitely Extensible Non-Linear Elastic Spring with the Peterlin’s approximation)
viscoelastic fluid model, and encompassed a wide range of the pertinent dimensionless

numbers, namely, Rayleigh number ( 10  Ra  10 ), Weissenberg number ( 1  Wi  100 ),


3 6

polymer extensibility parameter ( 10  L  500 ), polymer viscosity ratio ( 0.5    0.9 ) and
2

a fixed value of the Prandtl number of Pr = 7. As expected, and seen in earlier studies, the
buoyancy-induced convection current inside the annulus increases as the Rayleigh number
increases for Newtonian fluids. A same trend is also seen for FENE-P viscoelastic fluids.
However, the flow pattern is seen to be more chaotic and distorted as compared to that seen
for Newtonian fluids, particularly at high Rayleigh and Weissenberg numbers. This is due to
the dominance of the shear-thinning properties of the viscoelastic polymer solution. The
isotherms are seen to be more concentrated or crowded in the vicinity of both the cylinders
for viscoelastic polymer solutions than that seen for Newtonian under otherwise identical
conditions. Therefore, the heat transfer rate is always high in the viscoelastic polymer
solutions than that seen in the Newtonian fluid, at least within the ranges of conditions
covered in this study. The heat transfer rate and/or the average Nusselt number increases with
the Rayleigh number regardless of the fluid type, i.e., Newtonian or viscoelastic. It also
increases with the Weissenberg number; however, beyond a critical value, the average
Nusselt number remains almost constant and it does not vary with the Weissenberg number.
This is due to the competitive influence of shear-thinning and elastic properties of the
viscoelastic polymer solutions. Furthermore, the heat transfer rate increases with the
decreasing value of the polymer viscosity ratio and/or increasing the polymer concentration
in the solution. This is again due to the increase in the shear-thinning tendency of the polymer
solution.

34
This study has been carried for a fixed for a fixed value of the annulus gap. Therefore, in
future, it can be extended for different gap ratios, as it will surely influence the heat transfer
rate. Furthermore, the present study was carried out with the FENE-P viscoelastic model
which shows both shear-thinning and elastic properties in its rheological behaviour.
However, it would be good if the study is further carried out with a viscoelastic model which
does not shear-thinning behaviour, for instance, FENE-CR model. In this way, one can show
the explicit effect of the fluid elasticity on the heat transfer rate. Furthermore, a universal
correlation for the present results on average Nusselt number can also be provided, which
could be useful in predicting the average Nusselt number for intermediate values of the
governing parameters. Finally, the study can be extended to other geometries confined and
unconfined geometries in order to investigate the influence of fluid viscoelasticity on the heat
transfer rate.

35
References

1. Y. A. Cengel, A. J. Ghajar, Heat and Mass Transfer: Fundamentals and Applications,


Fifth edition, McGraw Hill, 2015.
2. A. G. A., M. M. Farid, X. D. Chen, P. Richards, Numerical simulation of natural
convection heating of canned food by computational fluid dynamics, J. Food. Eng., 41
(1999) 55-64.
3. A. K. Datta, A. A. Teixeira, Numerically predicted transient temperature and velocity
profiles during natural convection heating of canned liquid foods, J. Food Sci., 53
(1988) 191-195.
4. F. Espinosa, R. Avila, J. G. Cervantes, F. J. Solorio, Numerical simulation of
simultaneous freezing melting problems with natural convection, Nuclear Eng.
Design 232 (2004) 145-155.
5. K. S. Reddy, N. S. Kumar, An improved model for natural convection heat loss from
modified cavity receiver of solar dish concentrator, Solar Energy 83 (2009) 1884-
1892.
6. S. N. Akour, M. A. Jarrah, Experimental and numerical analysis of natural convection
for al-5.5% cu alloy, J. Materials Pro. Tech. 164 (2005) 1479-1486.
7. R. P. Chhabra, J. F. Richardson, Non-Newtonian Flow and Applied Rheology:
Engineering Applications, Butterworth-Heinemann, 2011.
8. R. G. Larson, E. S. G. Shaqfeh, S. J. Muller, A purely elastic instability in Taylor-
Couette flow, J. Fluid Mech. 218 (1990) 573-600.
9. A. Groisman, V. Steinberg, Elastic turbulence in a polymer solution ow, Nature 405
(2000) 53-55.
10. B. Qin, P. F. Salipante, S. D. Hudson, P. E. Arratia, Upstream vortex and elastic wave
in the viscoelastic ow around a confined cylinder, J. Fluid Mech. 864 (2019) 1-11.

36
11. C. M. White, M. G. Mungal, Mechanics and prediction of turbulent drag reduction
with polymer additives, Annu. Rev. Fluid Mech. 40 (2008) 235-256.
12. R. J. Gordon, Mechanism for turbulent drag reduction in dilute polymer solutions,
Nature 227 (5258) (1970) 599-600.
13. V. N. Kalashnikov, Dynamical similarity and dimensionless relations for turbulent
drag reduction by polymer additives, J. Non-Newt. Fluid Mech. 75 (2-3) (1998) 209-
230.
14. D.-Y. Li, X.-B. Li, H.-N. Zhang, F.-C. Li, S. Qian, S. W. Joo, Efficient heat transfer
enhancement by elastic turbulence with polymer solution in a curved microchannel,
Micro. Nano. 21 (1) (2017) 10.
15. W. M. Abed, R. D. Whalley, D. J. C. Dennis, R. J. Poole, Experimental investigation
of the impact of elastic turbulence on heat transfer in a serpentine channel, J. Non-
Newt. Fluid Mech. 231 (2016) 68-78.
16. T. H. Kuehn, R. J. Goldstein, An experimental and theoretical study of natural
convection in the annulus between horizontal concentric cylinders, J. Fluid Mech., 74
(695-719) 1976.
17. A. W. Date, Numerical prediction of natural convection heat transfer in horizontal
annulus, Int. J. Heat Mass Transfer, 29 (1457-1464) 1986.
18. R. Kumar, Study of natural convection in horizontal annuli, Int. J. Heat Mass
Transfer, 31 (1137-1148) 1988.
19. J. Mizushima, S. Hayashi, T. Adachi, Transitions of natural convection in a horizontal
annulus, Int. J. Heat Mass Transfer, 44 (1249-1257) 2001.
20. J-S. Yoo, Prandtl number effect on bifurcation and dual solutions in natural
convection in a horizontal annulus, Int. J. Heat Mass Transfer, 42 (3279-3290) 1999.
21. Y. Hu, D. Li, S. Shu, X. Niu, Study of multiple steady solutions for the 2D natural
convection in a concentric horizontal annulus with a constant heat flux wall using
immersed boundary-lattice Boltzmann method, Int. J. Heat Mass Transfer, 81 (591-
601) 2015.
22. K. Fukuda, Y. Miki, S. Hasegawa, Analytical and experimental study on turbulent
natural convection in a horizontal annulus, Int. J. Heat Mass Transfer, 33 (629-639)
1990.
23. C. P. Desai, K. Vafai, An investigation and comparative analysis of two-and three-
dimensional turbulent natural convection in a horizontal annulus, Int. J. Heat Mass
Transfer, 37 (2475-2504) 1994.

37
24. M. H. Matin, W. A. Khan, Laminar natural convection of non-Newtonian power-law
fluids between concentric circular cylinders, Int. Comm. Heat Mass Transfer, 43
(112-121) 2013.
25. H. Laidoudi, H. Ameur, Investigation of the mixed convection of power-law fluids
between two horizontal concentric cylinders: Effect of various operating conditions,
Ther. Sci. Eng. Progress, 20 (100731) 2020.
26. H. Masoumi, M. S. Aghighi, A. Ammar, A. Nourbakhsh, Laminar natural convection
of yield stress fluids in annular spaces between concentric cylinders, Int. J. Heat Mass
Transfer, 138 (1188-1198) 2019.
27. J. Zhao, L. Zheng, X. Zhang, F. Liu, Unsteady natural convection boundary layer heat
transfer of fractional Maxwell viscoelastic fluids over a vertical plate, Int. J. Heat
Mass Transfer, 97 (760-766) 2016.
28. N. Mustafa, S. Asghar, M. A. Hossain, Natural convection flow of second-grade fluid
along a vertical heated surface with variable heat flux, Int. J. Heat Mass Transfer, 53
(5856-5862) 2010.
29. A. Chauhan, P. M. Sahu, C. Sasmal, Effect of polymer additives and viscous
dissipation on natural convection in a square cavity with differentially heated side
walls, Int. J. Heat Mass Transfer, 175 (121342) 2021.
30. R. B. Bird, C. F. Curtiss, R. C. Armstrong, O. Hassager, Dynamics of Polymeric
Liquids, Volume 2: Kinetic Theory, Wiley-Interscience, 1987.
31. R. Sureshkumar, A. N. Beris, R. A. Handler, Direct numerical simulation of the
turbulent channel ow of a polymer solution, Phys. Fluids 9 (3) (1997) 743-755.
32. C. D. Dimitropoulos, Y. Dubief, E. S. G. Shaqfeh, P. Moin, S. K. Lele, Direct
numerical simulation of polymer-induced drag reduction in turbulent boundary layer
ow, Phys. Fluids 17 (1) (2005) 011705.
33. https://www.openfoam.org.
34. F. Pimenta, M. Alves, rheotool, https://github.com/fppimenta/rheoTool (2016).
35. M. A. Alves, P. J. Oliveira, F. T. Pinho, A convergent and universally bounded
interpolation scheme for the treatment of advection, Int. J Num. Method Fluids 41
(2003) 47-75.
36. J. Lee, J. Zhang, C.-C. Lu, Incomplete LU preconditioning for large scale dense
complex linear systems from electromagnetic wave scattering problems, J. Comp.
Physics 185 (1) (2003) 158-175.

38
37. M. A. Ajiz, A. Jennings, A robust incomplete Choleski-conjugate gradient algorithm,
Int. J. Num. Methods Eng. 20 (5) (1984) 949-966.
38. R. Fattal, R. Kupferman, Constitutive laws for the matrix-logarithm of the
conformation tensor, J. Non-Newt. Fluid Mech. 123 (2-3) (2004) 281-285.
39. F. Pimenta, M. A. Alves, Stabilization of an open-source _nite-volume solver for
viscoelastic fluid flows, J. Non-Newt. Fluid Mech. 239 (2017) 85-104.

39

You might also like