Topic A 2

nd
International Conference on Energy Conversion and Conservation

Laminar flow of a shear-thinning fluid in 90pipe bends

1
Abdallah Iman,
2*
Kahine Khalil,
3, 4
Accary Gilbert,

1
Lebanese University, Ecole doctoral de science et technologie, physics department
Hadath, lebanon
2
Lebanese University, University Institute of Technology
Sada, Lebanon
3
Holy-Spirit University of Kaslik, Faculty of Science and Computer Engineering
Kaslik, Lebanon
4
Lebanese University, Faculty of Engineering
Tripoli, Lebanon
* E-mail: kkahine@ul.edu.lb


Keywords: Shear-thinning, Rheological models, Pipe bend flow, Numerical modeling, Pressure loss.

Abstract
There are many complex and important industrial fluids such as polymer solutions, suspensions, blood, paints, oils
and greases exhibiting complex rheological behavior. Several empirical rheological models have been proposed
for representing the viscosity function of these non-Newtonian fluids. Among them, the Power law (Ostwald-de
Waele) model is the most widely used and it was found to be very simple and versatile in describing shear-thinning
(suspensions, emulsions, polymeric fluids) as well as shear-thickening (highly loaded suspensions of very fine
particles; e.g., starch, plaster. . .) behavior. In this paper, the non-Newtonian fluid flow in a 90 curved pipe is
studied numerically to obtain the additional pressure loss coefficient resulting from the bend for different values of
Reynolds.

I. Introduction
The implementation of fluids in numerous industrial
activities often requires their transportation in
networks of conducts including the oddness, the
pumps and the intersections of heat. A good
dimensionality of these networks, pass by a precise
knowledge of the governing laws for energy losses
(regular and singular).
The fluids presenting a non linear behavior obeys the
power law as well as the evolution of their parameters
in the flow which put an interest in many industrial
sectors; they are nevertheless distinctly studied
compared to the Newtonians fluid. The consequences
of the shear-thinning have to be identified for a better
understanding for flow structure. In this paper, one is
interested in the laminar flow of the non Newtonian
shear thinning fluids, in a 90 degree pipe bend (Fig.1).
The mathematical model is represented by the
equations of continuity and conservation of movement
quantity. The Rheological behavior is given by the
power law of (Ostwald of Waele). We intend to
determine by numeric way the structures of the
dynamic fields of this flow. Also, we are going to study
the influence of Reynolds number and the shear
thinning characters on the dynamic field and on the
evolution of the pressure loss. The development or
the use of software is indispensable to the resolution
of this kind of problems. The discretization of the
system of equations is done numerically by finite
volume method that is implemented to numerical
modeling software Fluent.
In pressure drop calculations it is an accepted
practice to consider that the flow is fully developed in
straight pipes, while all other effects (flow distortions,
redevelopment) in sudden symmetric
expansions/contractions, pipe bends and other
obstacles are introduced using local loss coefficients.
The mathematical model consists of the
incompressible steady-state equations of continuity
and momentum and the rheological behavior of the
fluid is given by the power law of Ostwald de Waele
as follows:
= K |y |
n-1
y (1)
Where is the shear stress, y is the shear rate, K and
n are respectively the consistency and the structure
index of the model. In this work we consider only the
case of a shear-thinning fluid, hence n 1. A fully
developed velocity profile is considered at the inlet,
non-slip condition is considered at the pipe wall and an
outflow condition is used at the outlet of the pipe (see
Fig. 1).
In Section 2, a theoretical approach is presented in
order to obtain the equation for the pipe bend
pressure loss.Sections 3 numerical modeling.section
4 talkes about results and obtained correlations.










-1-
Topic A 2
nd
International Conference on Energy Conversion and Conservation

-2-
Nom nclature e
: : Velocity (m/sec)
p : Pressure (N.m
-2
)
R
e


d

D
r
: Reynolds number.

: Pipe diameter.

: Pipe bend curvature diameter.
: Radius of the pipe cylinder.

Greek letters

a
: Apparent Viscosity (Pa.s)
Subscripts
equiv : Equivelent
elbow : Elbow bend
K
L
: Coefficient of pressure loss

II. Theoretical approach
The flow problem, which is the subject of study in this
work, is perfectly described by the fundamental
equations of fluid dynamics governing flow laws ;the
continuity equation, equations of motion in both
directions (axial and radial), which must be added the
equation of conservation of flow and the Law
rheological behavior of the fluid considered. It is
therefore proposed in this section to represent
mathematical equations governing flow.


Fig. 1: The geometry of the 90 bend pipe under
consideration with boundary conditions. d and D are
respectively the pipe and the bend curvature diameters.

Contin

o(ru)
oz
uity equation :
+
o(r:)
or
= u. ( 2)

Momentum equation in z-direction:
pu
u
z
+p:
u

= -
p
z
+
1

[rp
u
u

+

z
[p
u
u
z
. (3)
Mo ntum quati in d n:

pu
u
z
me e on r irectio
+p:
u

= -
p

()

+

z
[p
u

z
. (4)

Taking into acco t th
almost zero, we will ha
oP
or
un at the radial pressure gradient is
ve:
= u onJ
oP
oz
=
JP
Jz


Mo ntum quati in irec n:

pu
u
z
me e on z-d tio
+p:
u

= -
dp
dz
+
1

[rp
u
u

+

z
[p
u
u
z
. (5)

Mo ntum quation in r direction com

pu

z
me e be es:
+p:

()

+

z
[p
u

z
. (6)

The movement of isothermal and steady laminar flow
of an incompressible non-Newtonian fluid is governed
by mass and momentum conservation equations.


.


III.Numerical modelling:
III.1 geometry and boundary condition:
The geometry of a pipe bend was shown
schematically in (Fig.1) together with the coordinate
system adopted. It is composed of a 90 bend of
curvature radius R , upstream and downstream
tangents of length L
1
and L
2
.For solving the governing
equations, at the inlet, an axial velocity is assigned a
fully-developed laminar velocity profile. The fully
developed velocity profile for the inlet conditions was
prepared for each Reynolds number. To obtain
significant results, simulating as much as possible the
actual fluid flow, passes first through a good
discretization of the flow field. Recall that the golden
rule of a good mesh is more finely meshed possible
areas subject to strong variations of flow parameters
and especially the area of entry conduct, where the
maximum disturbance is due to the singularity of
startup. In our case, the mesh is made of "Gambit",
this software allows the design of the geometry of the
flow, the discretization of its edges, the mesh flow
areas and to define the type of each side (boundary
conditions) as shown in figure 2. It was agreed to
appoint a mesh with a pair of numbers corresponding
respectively the numbers of discretization points of
the sides of the flow field, respectively in the axial and
radial direction.
25 d


Fig. 2: the 3-dimensional mesh draw using Gambit.


d
5d
D
Inlet
Outlet
Topic A 2
nd
International Conference on Energy Conversion and Conservation

-3-
III.2 Numerical procedure and simulation
The components of speed in the equations of
momentum are bound to the pressure, which creates
a main difficulty for their mathematical resolution.
Indeed, if the field of pressure is taken in account in
the treatment of these equations, the gotten speeds
verify the equation of continuity, but the pressure is a
fully-fledged variable, although it doesn't have an
equation. We use a method of speed
prediction-correction then according to the pressure;
the equation of conservation of the mass is
transformed to get pressure equation, so that the
solution of the streamlined problem gives a field of
speed and pressure verifying the equations of
momentum and continuity equation.
We introduce the flow parameters in the software
where the law of (Ostwald de waele) is defined
rheologically for the non-Newtonian fluid of index n
and consistency K; Where K is obtained from
d n u the following equation for
a m:
Reynolds an sing
cylindric l coordinate syste
Rc
n
=Rc [
3n+1
s
4n

-n
8
1-n
(7)

Where Rc
s
=
pu
m
2-n

n
K

The transport equations are solved by a fully-implicit
finite volume method in a segregated formulation on
an unstructured staggered mesh. Space discretization
is based on the power law scheme for the convection
terms while diffusion terms are approached by central
difference approximation and the velocity pressure
coupling is treated using SIMPLER algorithm. A
steady-state solution is supposed to be obtained
when the residuals of all transport equations reach
1u
-8
in non-dimensional form.

Fig. 3: Evolution of pressure for different values of residuals
at the centerline of the pipe.
The residual effect shows that for every point we have
a number of iterations take place in order to converge
to the solution the process repeats itself continuously
to have a convergence limit. The final convergence is
determined by measuring residual source of each
variable. The residual sources are defined for each
variable at each point and for each iteration.
In our paper we did three convergence criterions in
order to distinguish between good and close results,
the values, 1u
-04
, 1u
-06
onJ 1u
-08
for one selected
sh and the results are shown in figure 3. me
As we can see from the above graph(fig 3) that values
1u
-06
and 1u
-08
coincide on each other, the value of
1u
-08
was enough as convergence criterion. It starts
the simulation and it makes the number of iterations
needed to reach convergence.

III.3. Velocity profile
In the entrance zone before the pipe bend we have an
established velocity profile which is in an increasing
way until it takes the parabolic form as shown in the
following figure.
The velocity profile in the entrance zone is expressed
chapter 1. in equation (1.14) shown in
0()
0
m
= [
3n+1
n+1
_1 -[

n+1
n
_

Where r is the radius of input flow where the radius of
the circle is R=0.5
and 0 < r < R . u is calculated and the obtained
results for n= 0.3 , 0.4 , 0.6 , 0.8 and 1are listed in the
following graph (3.4):


Fig. 4: Inlet Velocity profile


We observe in fig. 4 that as n decreases the
plasticity zone increases and the shear rate
deformation occurs more on the edges of the surface,
and when the index structures reaches 1 which
corresponds to Newtonian behavior ;the fluid is not
affected by shear deformation and it moves smoothly.
Once the mesh is achieved; we can start the
numerical calculation. The simulations are performed
on "fluent". This software has a user interface that
allows
easy handling, a significant display and easy access
to Digital results, which makes software responsible
to run a simulation.
The components of speed in the equations of
momentum are bound to the pressure, which creates
a main difficulty for their mathematical resolution.
Indeed, if the field of pressure is taken in account in
the treatment of these equations, the gotten speeds
verify the equation of continuity, but the pressure is a
fully-fledged variable, although it doesn't have an
equation.
Topic A 2
nd
International Conference on Energy Conversion and Conservation

-4-
We use a method of speed prediction-correction then
according to the pressure; the equation of
conservation of the mass is transformed to get
pressure equation, so that the solution of the
streamlined problem gives a field of speed and
pressure verifying the equations of momentum and
continuity equation.
we introduce flow parameters software and especially
the law is defined theologically for the fluid used by the
index structure n and by the consistency K as shown
in the following figure(fig.5).


Fig. 5: Input parameter

We note that the equation used is Ostwald de waele,
To circumvent this problem, and although our study
fluid is that of Ostwald de Waele, we chose to adopt
as law behavior the equation corresponding
Hurschel-Bulkley fluids in non-Newtonian critical
stress. Recall that the model of Herschel-Bulkley is
given by:

=
:
0
+K_y
n
-[

0
y

n
_
y


Having
0
the yield stress threshold equal to zero the
preceding equation besomes the Ostwald de waele
equatio : n
=K |y |
n-1
y

IV. Pressure drop and correlation
IV.1. Pressure evolutions
Numerical results are obtained for different values of
Reynolds number ranging from 50 up to 1000, for a
structure index varying between 0.3 and 1, and for
different values of the geometric ratio D/d. we export
the results to a new program Tecplot which plot the
pressure as a function of the center line of pipe bend.
We plot simultaneously for every Reynolds number
the corresponding power law indexes n ranges
between 0.3 and 1.For example figure 6 shows the
evolution of the pressure (made dimensionless using
the square root of the flow mean velocity) at the
centerline of the pipe for Re = 250, D/d = 6 and for
different values of n. Figure 7 shows the evolution of
the pressure drop coefficient as a function of the
Reynolds number for D/d = 4 and for different values
of n.



Fig.6: Evolution of the pressure at the centerline of the pipe
for Re = 250, D/d = 6 and for different values of n.


Fig. 7: Evolution of the pressure drop coefficient as a
function of the Reynolds number for D/d = 4 and for different
values of n.

The above figures are two of many graphs that we
have achieved by making different simulations varying
(D/d, Re, n) .in the next section we are interested to
get a correlation for the pressure drop coefficient
covering the range of the control parameters (Re, n,
and D/d).

IV.2. Correlation and results
In this study, correlations for the pressure drop
coefficient were obtained covering mentioned the
range of the control parameters (Re, n, and D/d) due
the pre nce of the elbow in the cylindrical pipe
n followin eq :
to se
applyi g the g uation
I
cqu
I
cIbow
=
K
L
S2
R
c
,

4
n[

J
,
(8)

Where K
L
is the pressure drop coefficient in the
pipe, I
cqu
is the equivalent length of the pipe
assuming it is straight and I
cIbow
is the length of the
Topic A 2
nd
International Conference on Energy Conversion and Conservation

-5-
pipe bend.

Knowing K
L
we can set up
L
cq
L
clbcw
-1 as a function
of Reynolds and trace it in the following graph.
For example figure (5.1) represent the evolution
L
cq
L
clbcw

for D/d =2 as a function of Reynolds for different values
of n ranged from n=0.3 to 1.

Fig.8: Evolution of
L
cq
L
clbcw
as a function of Reynolds number
for different values of n for D/d=2.

As we can see from the above graph that each line
passes through the origin which insure the fact that As
Re - u we have
L
cq
L
clbcw
- 1 and D/d - we
have
L
cq
L
clbcw
= 1

After doing similar graphs for D/d=3, 4, 5, 6 and 8.
The function
L
cq
L
clbcw
= (Rc
n
) is similar relation to:

y = A (n) Re

Where A (n) is the slop of the curve above.

In order to find the correct correlation between (n, Re
and D/d) we draw the evolution of the slope A(n) as a
function of n=0.3,0.4,0.6.0.8 and 1 for different values
of D/d as shown in figure 9.


Fig. 9: Evolution of the slopes for different index structure (n)
for D/d=6.
The other graphs are similar to this one but instead of
showing them we show their respective equation in the
following table.



D/d A(n)
D/d=2 1.560 n E
-02
D/d=3 8.104 n E
-03
D/d=4 5.228 n E
-03
D/d=5 3.894 n E
-03
D/d=6 3.201 n E
-03
D/d=8 2.375E
-03
n
Tab. 1: slope A (n) as a function of D/d and n

Now, the same procedure we are interested in the
evolution of the slope A(n) as a function of D/d.
The evolution is presented in fig.10.




Fig. 10: Evolution of the slopes as a function of ( D/d).

We have A (n) = a /d) (D n
and a (D/d) = 3.74E
-02
(J)
-1.368
( 9)

then,

A (n) = 3.74E
-02
(J)
-1.368
n (10)
Finall w e: y e hav

L
cq
L
clbcw

= {S.74E
-02
(J)
-1.368
n] Rc +1 (11)

N w pressure dr
K =
S2
c
o op coefficient correlation:
L
R

I
cqu

:
K
L
=8 j{S.74E
-02
(J)
-1.368
n] +
1
Rc
[ (12)










Topic A 2
nd
International Conference on Energy Conversion and Conservation

-6-

V. Conclusions
Steady and laminar flow of a non-Newtonian fluid in a
90 pipe bend was studied numerically taking the
advantage of fluent, gambit, Tec plot program
numerical codes.
First a theoretical approach was used in order to
obtain the equation for pressure loss that occurs due
to a presence of a pipe bend. To do so,
one-dimensional analysis for the fully developed
pressure drop in a straight pipe was incorporated in
the overall pressure drop that occurs in a pipe bend
as well as in an upstream and downstream tangent.
Meshes, was performed to study the influence of grid
refinement on numerical predictions. In addition, a
technique was used to quantify the numerical results
for the total pressure drop (ptot) and the pressure
drop between the entrance and exit of a bend
(pbend).
After we found the pressure drop coefficient we had
made a correlation to find the equivalent pipe bend
length; such that:
For equivalent length


I
cq
I
cIbow
= {S.74E
-02
(J)
-1.368
n] Rc +1

For p ssure drop coefficient:
K
L
= 8 j{S.74E
-02
(J)
-1.368
n] +
1
Rc
re
[.

Acknowledgements
I would like to express my thanks to Dr. Khalil Kahine
and Dr. Gilbert Accary for devoting their expensive time
in helping me to achieve this work; I will never forget
their kindness helping me to finish my project.

References
A. B. METZNER, Heat Transfer in Non-Newtonian fluids.
Adv. Heat Transfer, vol 2, Pp 357 (1965).
N.MIDOUX, Mcanique et Rhologie des fluides en gnie
chimique (1985).
D. C. H. CHENG, A design procedure for pipeline flow of
non-newtonian dispersed systems.Proc. Hydrotransport 1,
1st-10th September 1970, Page J5-77 J5-95.
D. QUEMEDA, Relation Comportement-Structure dans les
dispersions Concentres.Proceeding du sminaire :
Thermique des Liquides Complexes. Pp 3-40, LA BAULE 26
Septembre (1984).
K. Kahine, Simulation Numrique de lEcoulement de
Fluides non-Newtoniens Lors du Franchissement dun
divergent. Lois de transfert de chaleur en conduite
Thse Universit Henry POINCARE, NANC I. Y
M. LEBOUCHE, M. LUCIUS, Grandeurs caractristiques en
Rhologie et Mcanique.Proceeding du 2
nd
congrs sur
les Techniques Avances en Hmorhologie.Pp 75-131
NANCY (1983)

Y. I. CHO, J. P. HARTNETT, Non-Newtonian fluids in circular
pipe flow. Adv. Heat. Transfer, vol 15, Pp 67-69 (1982).