You are on page 1of 8

Computers & Fluids 99 (2014) 116123

Contents lists available at ScienceDirect

Computers & Fluids


j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

CFD-based estimation and experiments on the loss coefcient


for Bingham and power-law uids through diffusers and elbows
}s
Pter Csizmadia , Csaba Ho
Budapest University of Technology and Economics (BME), Faculty of Mechanical Engineering, Department of Hydrodynamic Systems (HDS), Hungary

a r t i c l e

i n f o

Article history:
Received 19 July 2013
Received in revised form 17 February 2014
Accepted 2 April 2014
Available online 24 April 2014
Keywords:
CFD
Diffuser
Elbow
Loss coefcient
Non-Newtonian uid

a b s t r a c t
This paper presents results on the CFD-based estimation of loss coefcients for typical pipeline elements
in the case of Bingham and power law uids. The actual elements are diffusers with angles between 7.5
and 40 and elbows of R=D ratios between 1 and 10. Three uids were studied; water (for validation purposes), a power-law uid (0.13 m/m% Carbopol 971 solution) and Bingham plastic uids with Hedstrm
number varying from 10 to 106 . Experiments were conducted with water and power-law uid to validate
the CFD computations. Loss coefcients are given for the Reynolds number regime that is usual in engineering applications. The behavior of different turbulence models (eddy-viscosity and Reynolds stress
models) are also reported.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
The importance of non-Newtonian uids in engineering applications is continuously increasing as these materials are often
encountered in e.g. energy, petroleum or food industry. Due to
their complex rheological behavior standard uid mechanical techniques are often not applicable or have to modied: on one hand,
the denition of such basic quantities as e.g. the Reynolds number
is highly nontrivial for a power-law uid or, on the other hand, for
a Bingham plastic uid, new dimensionless properties appear such
as the Hedstrm number. However, when designing pipeline systems for such uids, one has to estimate the pressure losses to
be able choose the appropriate pump and pipe diameters, which
is often cumbersome due to the lack of experimental and theoretical guidelines. Computational uid dynamics (CFD) codes seem to
be suitable for such computations, however, there is little experience on the behavior of such codes (notably the performance of
the turbulence models developed for Newtonian uids) in the case
of non-Newtonian uids. This paper reports on the experiences of
the as-is application of a commercial CFD code for predicting loss
coefcients in the case of some non-Newtonian uids, i.e. we did
not modify or ne-tune the standard settings of the code ANSYS
CFX 13.0.

Corresponding author. Tel.: +36 1 463 2553.


E-mail addresses: csizmadia@hds.bme.hu (P. Csizmadia), csaba.hos@hds.bme.hu
} s).
(C. Ho
http://dx.doi.org/10.1016/j.compuid.2014.04.004
0045-7930/ 2014 Elsevier Ltd. All rights reserved.

Several experimental and numerical studies can be found in the


literature on the pressure drop in pipelines and ttings due to friction losses and shape effects for non-Newtonian uids. Cabral et al.
[1] reports friction losses in valves and ttings, including seat, ball,
no-return, and buttery valves, bends, tees, and unions. The uids
in this study were food products, which behave like power-law uids. Monterio and Bansal [2] examines the pressure drop in ice
slurry ow and compared different rheological models, as the Bingham, Casson, power-law and HerschelBulkley models. This paper
contains the denition of a generalized Reynolds number RePL . The
behavior of the same slurry in horizontal pipes is also examined in
[3]. Experimental study is performed in [4,5] to predict non-Newtonian head losses through diaphragm valves at different opening
position from laminar to turbulent ow regimes. Neofytou and Drikakis [6] and Ternik et al. [7] studied the sudden and planar symmetric expansion ows and their stability for non-Newtonian
uids numerically. In [8], the non-Newtonian uid ow in a 90
curved pipe is studied numerically to obtain the pressure loss coefcient and the authors split the pressure drop in the pipe bend into
two parts: wall friction and effects of curved ow. Chowdhury and
Fester [9] studied the pressure losses for Newtonian and non-Newtonian (e.g.: power-law, Bingham plastic) ow in long square
edged orices in the wide Reynolds number range
1 6 Re 6 105 . Gradual contractions, sudden contractions, and
90 horizontal bends were investigated in [10] in which the local
resistance characteristic of coalwater slurry ow was examined.
Both [11,12] and the present study deals with friction losses in
valves and ttings for power-law uids and their rheological

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

117

Nomenclature
A1
A2
Bi
d
D
Din
Dout
He
k
L
Lin
Ll
Lout
nPL
p
Q
R
Re
ReB
RePL
v
v in
v out
c_

cross section of the inlet (m2)


cross section of the outlet (m2)
Bingham number
smallest diameter in the Venturi pipe (m)
internal diameter (m)
internal diameter in the inlet pipe section (m)
internal diameter in the outlet pipe section (m)
Hedstrm number
roughness (m)
length of the straight pipe (m)
length of the inlet pipe section (m)
length of the local ttings (m)
length of the outlet pipe section (m)
ow behavior index, power-law uid
pressure (Pa)
ow rate (m3/s)
radius of curvature (m)
Reynolds number
Reynolds number in the case of Bingham plastic uids
generalized Reynolds number snPL
average velocity (m/s)
average velocity in the inlet pipe section (m/s)
average velocity in the outlet pipe section (m/s)
shear rate (s1)

properties. In [13] experimental research was performed to evaluate the frictional pressure drop across different piping components
such as orices, gate and globe valves, elbows, and bends for nonNewtonian pseudo-plastic liquids in laminar ow. Rosa and Pinho
[14] deals with pressure drop via antisymmetric diffusers for Newtonian ow in laminar case. In [15,16] non-Newtonian slurries are
studied experimentally and the friction losses for laminar, transitional and turbulent ow through straight pipe and in bends, ttings, valves, and Venturi meters are presented. The study
includes a wide range of geometries and material properties. Gzel
et al. [17] Deals with Carbopol solutions and their owing and rheological properties.
As it can be seen, there is a large body of experimental (and analytical) studies on non-Newtonian loss coefcients, however the
penetration of CFD codes is rather weak. The results of the present
study encourages the use of these codes for such purposes.
This paper is organized as follows. After dening the loss coefcient in Sections 2 and 3 presents the rheological properties of the
uids. Sections 4 and 5 contain the experimental and numerical
set-up. Section 6 includes the validation of the CFD computations
and the main results, while we conclude the study in Section 7.

d
Dp
Dpall
Dpdown

Dpin
Dpl
Dpout
Dpup
f
k
kanalytical
kin
kout

lB
lPL
q
s
syield

bevel-angle ()
total pressure drop caused by the ttings (Pa)
total pressure drop via the complete model (Pa)
total pressure drop caused by the downstream disturbance (Pa)
total pressure drop due to wall friction in the inlet pipe
section (Pa)
total pressure drop within the disturbance local ttings
(Pa)
total pressure drop due to wall friction in the outlet pipe
section (Pa)
total pressure drop caused by the upstream disturbance
(Pa)
loss coefcient
friction factor
friction factor, analytical value
friction factor in the inlet pipe section
friction factor in the outlet pipe section
consistency index, Bingham plastic uid (Pa s)
consistency index, power-law uid Pa snPL
density (kg/m3)
shear stress (Pa)
yield stress (Pa)

sides (see Dpup and Dpdown in Fig. 1). For clarity we dene now the
evaluation of (1), similar to [8,10,16].
The measured (overall) pressure drop consists of three parts as
shown in Fig. 1: Dpin and Dpout being the frictional pressure losses
of the upstream and downstream straight pipes of length Lin and
Lout and friction coefcient kin and kout , plus the pressure drop of
the element, denoted by Dp:

Dp Dpall  Dpin  Dpout Dpall  kin

Lin q 2
L q
v  kout out v 2out :
Din 2 in
Dout 2
2

The pressure drop of the element Dp is made up of three contributions: pressure disturbances at the inlet (Dpup ) and outlet
(Dpdown ) of the element (which also changes the linear pressure distribution of the connecting pipes in the vicinity of the element) and
the local pressure drop of the element itself Dpl , see Fig. 1. We have

Dp Dpup Dpl Dpdown :

However, from the practical point of view, it is unnecessary to


determine these values separately as our aim is not to differentiate

2. Denition of the loss coefcient


The loss coefcient is dened as the non-dimensional difference
in total pressure between the two ends of a pipe or tting (see
[10,18]):

q 1
f Dp v 2in
;
2

in which Dp is the total pressure drop across the element, q is the


uid density and v in is the average velocity at the upstream side.
In the case of laboratory experiments or CFD computations one
usually does not measure the pressure directly at the upstream and
downstream side of the actual element but straight pipe segments
are attached to both sides. Moreover, the element which is being
measured (or simulated) changes the pressure distribution on both

Fig. 1. Pressure distribution along centreline of pipeline elements.

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

118

between local losses but to provide an overall loss coefcient for the
element. We shall use (3) later to evaluate the loss coefcient. We
shall mention here that in the case of the CFD computations, Lin is
not the actual total length of the straight pipe segment attached
on the upstream. As the fully developed velocity prole approaching the element (e.g. bend) is usually not known a priory, only an
approximate prole is prescribed at the inlet boundary and Lin
stands for only that segment of the upstream pipe, in which the
velocity prole fully develops.

yield stress. The former values were D 0:1 m; q 1250 kg=m3


and lB 0:07 Pa s, the yield stress, the Hedstrm numbers and
the Bingham numbers are given in Table 1.
In the case of the Bingham uids, the Reynolds number is
dened as (see [3,6,20])

ReB

Dqv

lB

4. Measurements
3. Rheology
3.1. Carbopol solution
In this part we introduce the material properties of the Carbopol
solution, which contains 0.13 m/m% Carbopol 971, 0.05 m/m%
NaOH and 99.82 m/m% water. The NaOH was added to set the
pH-value to approximately 7. The measurement of the rheology
was performed on a Rheotest 2 RV2 rotational viscometer. The
results of the rheological measurements and the curve t of the
shear stressshear rate relationship is described by

s lPL c_ nPL ;

with lPL being the consistency index and nPL standing for the ow
behavior index, see e.g. [2] for details. The actual values were found
to be lPL 0:1334 Pa snPL and nPL 0:7266. For the power-law uid,
we use the generalized Reynolds number (see [2]) given by



1
1 3nPL
RePL v 2nPL DnPL q lPL 8nPL 1
;
4nPL

 and D stand for average velocity and pipe diameter,


where v
respectively.
3.2. Bingham plastic uids

Fig. 2 presents a sketch of the experimental set-up with the


main elements. The centrifugal pump (1) conveys the 25 C uid
through the system from tank (5) via the pipe of 20 mm inner
diameter. The ow rate is set by the control valve Eq. (2), after
which more than 10D (310 mm) straight pipe was mounted to provide a fully developed velocity prole (see [21,22]). The pressure
drop on the straight segment of length L was measured with the
help of two pressure taps p1 and p2 . The computed from thus pressure difference friction factors will be used in Section 5.4 for validation purposes. Only water and Carbopol solution were used
during the measurements.
The curvature of the elbow (3) was 20 mm (which is equal to
the inlet diameter of the pipe), i.e. we have R=D 1 for the measured elbow. The pressure drop of the elbow was measured with
the help of the p2 and p3 taps and evaluated as described in Section (2). The length of the outlet section beyond the elbow was
more than 15 diameters long (348 mm). With the help of (1) and
(2), the loss coefcient can be calculated from the measured pressure drop p2  p3 .
The system also contains a Venturi pipe (4) to measure the ow
rate based on its calibrated DpQ curve from the pressure drop
between p4 and p5 . The vena contract of the Venturi pipe was
11 mm. All pressure drops were measured with the help of a
multimanometer.

In the case of the Bingham plastic uids, the rheological behavior is described by

5. CFD set-up

s syield lB c_ ;

5.1. Governing equations

with lB viscosity consistency and syield yield stress, see [2]. In the
case of Bingham uids, the dimensionless number highlighting
the importance of yield stress is the Hedstrm number, which is
dened as

He

D2 qsyield

l2B

with D standing for the pipe diameter and q for density, see [3].
Another important dimensionless number is the Bingham number
(not including the density), see [19] which is dened as

Bi

syield D
:
lB v

The CFD computations were performed with the commercial


code ANSYS CFX 13.0 and ICEM CFD was employed as a meshing
tool. The software solves the Reynolds-averaged NavierStokes
equations, the continuity equation (see e.g. [2326]), and the
transport equations associated with the actual turbulence model.
Both eddy-viscosity models (k  , SST) and Reynolds stress model
(BSL Reynolds stress) were tested. The actual equations can be
found in [26], it should be emphasized that the model constants
were not modied and the default values were used. Near-wall
treatment greatly inuences the results as skin friction represents

Note that the Bingham uid appears only in the CFD computations (i.e. no experiments were performed with thus uids), were
the pipe diameter, the density and the viscosity consistency were
xed and the Hedstrm number was controlled by varying the

Table 1
Properties of the Bingham plastic uids.

syield (Pa)

3:92  103

3:92  102

He

10

10

Bi/5.6

103

102

3:92  101
10

101

3:92
4

10
1

39:2

392

10

106

101

102

Fig. 2. The sketch of the experimental set-up.

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

119

Table 2
Number of elements of the CFD model for the elbows.
R/D

10

Number of nodes along the elbow


Number of elements

60
448k

90
486k

120
524k

150
675k

200
877k

Table 3
Number of elements of the CFD model for the diffusers.

Number of elements
Number of elements

Fig. 4. The sketch of the diffuser geometry (left). Mesh control edges (right, see text
for details).

d ()

7.5

15

20

40

A2 =A1 2:25
A2 =A1 4

1146k
762k

685k
685k

637k
657k

589k
618k

the majority of the losses. CFX 13.0 uses automatic wall treatment,
which allows a gradual switch between empirical wall functions (if
the computational mesh is not ne enough to resolve the viscous
sublayer) and low-Reynolds number models. Finally, the rheological constitutive equation (see Section 3) relates the stress tensor
(s) to the rate-of-strain tensor (c_ ).
5.2. Geometry and meshing
Fully structured 3D mesh was used, which was optimized via a
grid-independence study: three different meshes were tested in
the case of a the R=D 1 elbow containing 449k, 1.19 M, and
2.07 M elements. Special care was devoted to the correct resolution
of the boundary layer in the vicinity of the walls. The largest deviation between the coarsest and nest meshes was found to be less
than 3%, hence the coarsest mesh was used for the rest of computations. The mesh details are given in Figs. 3 and 4, the numbers of
the nodes along different directions are highlighted, see also Tables
2 and 3. Note that this mesh is slightly denser than the one used in
[8] (containing 382k elements). Additional straight pipes of 10D
length were added to the upstream and downstream sides in both
cases (elbow and diffuser) to allow proper boundary conditions
(see [21]).
5.3. Numerics
Uniform velocity distribution was prescribed at the inlet and
average static pressure at the outlet. The rest of the surfaces were
set to no-slip walls. The built-in material models were employed
with the material properties prescribed in Section 3.
Regarding to the turbulence modeling, three built-in models
were tested: k  e, SST (2-equation eddy-viscosity models) and
BSL Reynolds stress model (6 equations). As it will be shown later
(see Figs. 5, 6, 8, 9 and 11), the SST model turned out to be a reasonable compromise: the k  e model gave results far from the
measurements whereas the BSL Reynolds stress model did not

improve the results signicantly (compared to the ones obtained


with SST). It is worth mentioning that the usefulness of advanced
two-equation eddy viscosity models was also conrmed by
[23,27], where Fluents RNG k  e model was used. Unless indicated explicitly, the results correspond to the SST turbulence
model.
Steady state computations were performed, similar to the ones
in [21]. High-resolution spatial scheme was used for all equation
classes, except for the turbulence model, which was disretized by
rst-order scheme. One full computation needed typically approx.
10025,000 iteration steps and about 125 h on a standard desktop PC (3.4 GHz CPU, 4 GB RAM), depending on the number of
the elements, the Reynolds number, and the material properties.
For the Carbopol solution, the computations were relatively quick,
approx. 1 h was needed if the auto timescale option of CFX (that
detects automatically the suitable time step) was switched on.
However, in the case of Bingham plastic uids, the computation
time escalated with increasing Hedstrm number, for the higher
values (He  104  106 ), approx. 1525 h was needed and only
the use of local timescale option resulted in stable computations
(for details, see [26]). Convergence was judged by monitoring the
total pressure drop between the inlet and outlet boundary.
5.4. Validation
The measurement accuracy and the CFD results are tested rst
with the help of the straight-pipe friction loss case, as it is fully
described for Newtonian uids and well-documented for nonNewtonian cases, see [2,20]. The friction factor k is dened as

Dp
:
v 2

L q
D 2

There are numerous formulae to relate the friction factor to the


pipe geometry and ow conditions. For hydraulically smooth
pipes, the Blasius equation [2830]

0:316
Re0:25

11

is generally used. One of the most wide-spread way of describing


the effect of pipe roughness is the ColebrookWhite equation [31]



1
k=D 2:51
p 2 log
p ;
3:7 Re k
k

Fig. 3. The sketch of the elbow geometry (left). Mesh control edges (right, see text
for details).

10

12

in which k=D stands for relative roughness.


Panel (a) in Fig. 5 presents the friction factor in the case of
water as a function of the Reynolds number for hydraulically
smooth and rough pipes. The black dashed line presents the Blasius law (11), the (black) crosses refer to the CFD results of a
100D long pipe with smooth wall while the (green) plus signs
with error bars stand for the measurement. Clearly in the measurements the pipes were not hydraulically smooth but it would
have been cumbersome to measure the internal roughness
directly. Thus, (12) (blue dashdot line) was used to nd the
roughness value providing the closest results to the measure-

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

120

a: water

b: Carbopol solution

0.03

0.3

0.025

0.25

Friction factor

Friction factor

0.035

0.02
0.015
0.01

Measurement

0.2
0.15
0.1

CFD, SW
CFD, RW, k/D=0.15%
Analytical, smooth wall (B)
Analytical, rough wall (CW)
Measurement

0.005

CFD SST, SW
CFD SST, RW
CFD BSL, SW
CFD k, SW
64/ReGeneralized

0.05

200

Reynolds number

x 10

400

600

Generalized Reynolds number

Fig. 5. Friction factor in the case of water (panel a) and Carbopol solution (panel b). (B) and (CW) stand for the Blasius Eq. (11) and ColebrookWhite Eq. (12), respectively
(SW: smooth wall, RW: rough wall).

100

b: Carbopol solution

a: water

2
CFD
Miller [18]
Measurement

1.8
1.6
1.4

elbow

elbow

1.2
1
0.8
0.6
CFD SST
CFD BSL
CFD k
Measurement

0.4
0.2
10

104

105

106

Reynolds number

200

400

600

800

Generalized Reynolds number

Fig. 6. Loss coefcients in the case of an elbow of R=D 1. Panel (a): water, panel (b): Carbopol solution (power-law uid).

ments in the least-squares sense, which turned out to be


k=D 0:15% (D 20 mm, k 0:03mm). The CFD simulations
were then repeated with this roughness (blue asterisks). We
observe a reasonable agreement between the measurements
and the computed results validating the CFD settings.
Panel (b) in Fig. 5 depicts the friction factor corresponding to
the Carbopol solution as a function of the generalized Reynolds
number RePL (see 3.1 for details). As it can be seen this generalization of the Reynolds number (see e.g. [2]) allows the use of
the classic 64=Re formula in the laminar regime: see the black
dashed line. The red crosses, the magenta circles and the cyan
asterisks stand for the CFD results with the SST, BSL and k  e
turbulence models, respectively, for hydraulically smooth pipe.
The blue plus signs represent the CFD simulations with rough
wall and SST turbulence model. It is striking to what extend
the standard k  e model fails to predict the friction factor. On
the other hand, the SST model and the BSL model give practically
the same values, while the BSL model requires far more computational effort. Finally, the green markers with the error bars represent the measurements, which agree reasonably well with the
theory and the CFD computations.
Now we turn to Bingham plastic uids, for which the BuckinghamReiner equation (see [20]) allows the analytical computation
of the friction factor as

"
!#
64
He
64 He4
k
1

;
ReB
6ReB
3 k3 Re7B

13

which was simplied to iteration-free form in [20] to

kanalytical

 
64 10:67 0:1414He=ReB 1:143 He
:

1:16
ReB 1 0:0149He=ReB ReB ReB

14

Table 4 compares the friction factors at ReB 178:6


 0:1 m=s) and ReB 1786 (v
 1 m=s), computed by means
(v
of (14) and by CFD for a wide range of the Hedstrm number. In
the CFD simulations syield was varied, the rest of the material properties were kept constant. For low Reynolds numbers, we observe
an acceptable agreement (mostly below 10%). For the higher Reynolds number which is close to the laminarturbulent transition
region (see Fig. 1 in [20], Fig. 3.5 in [19], or Fig. 6 in [32]) the discrepancy is larger for low Hedstrm numbers whereas for higher
Hedstrm numbers (beyond 104 ) that represent purely laminar
ow again, the agreement is acceptable again.
6. Loss coefcients of elbows and diffusers
In this section, we present the main results of the current study:
loss coefcients for elbows and diffusers, for power-law and Bingham uids.

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

121

Table 4
Friction factor values for Bingham plastic uids at ReB 178:6 (v 0:1 m=s) and ReB 1786 (v 1 m=s). kanalytical refers to (14).
He

101

102

103

Bi/5.6

103

102

101

104
1

101

105

106
102

Re = 178.6

kanalytical
kCFD;SST
kCFD;laminar

0.3619
4.12%
+0.65%

0.3918
3.86%
+0.67%

0.6883
1.96%
+1.16%

3.3556
+2.06%
+4.08%

27.356
+5.69%
+6.55%

257.5
+10.08%
+10.29%

Re = 1786

kanalytical
kCFD;SST
kCFD;laminar

0.0359
+54.63%
+42.72%

0.0362
+53.97%
+42.17%

0.0392
+46.88%
+37.12%

0.0688
+12.68%
+12.82%

0.336
+0.75%
+3.17%

2.74
+5.26%
+4.89%

6.1. Elbow

5.5
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
0

R/D = 1
R/D = 2
R/D = 4
R/D = 6
R/D = 10

6.2. Diffuser

100

200

300

400

500

600

700

800

900

1000

Generalized Reynolds number


Fig. 7. Elbow loss coefcients in the case of Carbopol solution.

Fig. 9 shows the loss coefcients of the diffusers in the case


of Carbopol solution. In panel (a) (A2 =A1 2:25) of Fig. 9 we
observe an increase in the loss coefcient, apart from the results
with d 7:5 which is nearly constant. Panel b (A2 =A1 4) of the
same Figure shows similar tendencies, but now d 15 seems to
be the border between progressive and degressive characteristics. Unlike in the case of elbows, the main source of the losses

a: ReB =1786

102

100

b: ReB =17860

R/D=1
R/D=2
R/D=4
R/D=6
R/D=10

elbow

101

elbow

elbow

Fig. 6 compares the numerical and experimental results in the


case of an elbow of relative radius R=D 1, for water (panel a)
and the Carbopol solution (panel b). In the case of water, we experience that the CFD computation (red crosses) underpredicts the
loss coefcient compared to either the measurements (green plus
signs) or the literature value of [18] (blue asterisk). Similar to the
straight-pipe case, we speculate that this is due to the fact that
the CFD computation assumes hydraulically smooth pipe. Panel b
of the same Figure depicts the same results for the Carbopol solution. The tendencies are the same as in Fig. 5: the k  e model fails
to predict the loss coefcients while the SST and BSL models give
practically the same results.
Fig. 7 presents the loss coefcients of elbows with different R=D
ratios in the case Carbopol solution. The results show that loss
coefcients are decreasing with increasing Reynolds number,
which means that out of the shape and wall friction contributions

the latter dominates. This fact can be veried with another perspective: the larger the radius of curvature is, the longer the curved
bend becomes and the larger the loss coefcient becomes for a
given generalized Reynolds number [10]. Reports similar
values for higher Reynolds numbers, for example at
RePL 800; felbow  1 assuming similar R=D ratios. However, for
smaller RePL values the loss coefcients in [10] are higher.
Fig. 8 depicts the loss coefcients for Bingham plastic uids at
 1 m=s, panel a) and ReB 17860 (v
 10 m=s,
ReB 1786 (v
panel b) for various R=D ratios. It can be seen that the loss coefcients are nearly constant for both ReB numbers up to He  104 ,
beyond which we experience a sharp increase in the loss coefcient. By investigating the details of the ow eld (obtained by
means of CFD) directly, it was found that the sudden increase
in the loss coefcient is not due to the appearance of a separation
as such a bubble was continuously present between
He 103 . . . 106 . We speculate that the sudden increase is related
more to the skin friction as in the case of straight pipes, there is
also a sudden loss increase around He 104 . . . 105 . The smallest
loss coefcient can be observed in the case of R=D 2, which
seems to be an optimum between shape losses (dominating sharp,
short elbows) and friction losses (dominating long, straight
pipe-like elbows).

100

101
100

102

104

Hedstrm number

106

101
100

R/D=1
R/D=2
R/D=4
R/D=6
R/D=10
R/D=1,BSL
R/D=1,k

102

104

106

Hedstrm number

Fig. 8. Elbow loss coefcients in the case of Bingham plastic uids. Panel a: ReB 1786, panel b: ReB 17; 860.

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

122

a: A2 /A1 =2.25

0.25

b: A2 /A1 =4

0.5

0.2

diffuser

0.3

diffuser

0.4
0.15

0.1

=7.5
=15
=20
=40
=40,BSL
=40,k

0.2
=7.5
=15
=20
=40

0.05

0.1

500

1000

1500

Generalized Reynolds number

500

1000

1500

Generalized Reynolds number

Fig. 9. Diffuser loss coefcients in the case of Carbopol solution, panel a: A2 =A1 2:25, panel b: A2 =A1 4.

102

a: ReB =1786

100

=7.5
=15
=20
=40

b: ReB =17860
=7.5
=15
=20
=40

diffuser

diffuser

101

100

101
100

102

104

101
100

106

Hedstrm number

102

104

106

Hedstrm number

Fig. 10. Diffuser loss coefcients for Bingham plastic uids, A2 =A1 2:25, panel a: Re 1786, panel b: Re 17860.

102

a: ReB =1786

b: ReB =17860

=7.5
=15
=20
=40

100

diffuser

diffuser

101

=7.5
=15
=20
=40
=40,BSL
=40,k

100

101
100

102

104

106

Hedstrm number

101
100

102

104

106

Hedstrm number

Fig. 11. Diffuser loss coefcients for Bingham plastic uids, A2 =A1 4, panel a:Re 1786, panel b: Re 17; 860.

is the shape effect for diffusers, which is strongly related to the


onset of a separation.

The two panels of Figs. 10 and 11 depict the loss coefcients of


diffusers in the case of Bingham plastic uids, for A2 =A1 2:25 and
A2 =A1 4, respectively. It can be seen that for the lower Reynolds

}s / Computers & Fluids 99 (2014) 116123


P. Csizmadia, C. Ho

number, the loss coefcients are nearly constant up to He  104


and rapidly increase beyond that value. The same tendency is
found at He  105 for ReB 17860. These critical Hedstrm numbers are approximately the same for both area ratios. In the examined range of the Reynolds number, the smallest loss coefcient is
observed with d 7:5 .
7. Conclusion
This study presented a CFD-based study on the prediction of
loss coefcients of elbows and diffusers for power-law and Bingham plastic uids. When compared against measurements, it was
found that the CFD technique generally underpredicts the loss
coefcients, which is most likely due to the assumption of hydraulically smooth internal walls. Our computations conrmed that the
modied Reynolds number dened by (5) can conveniently be
used to characterize the ow of power law uids. It was also found
that for elbows with R=D 1 . . . 10, the wall friction dominates
over the shape factor. In the case of Bingham uid ow in elbows
and diffusers, we found that there exists a critical Hedstrm number (approx. 104 for ReB 1786 and 105 for ReB 17; 860), below
which the loss coefcient is independent of the Reynolds number.
We also found that in the case of the diffusers, the critical angle is
close to 7.5 for A2 =A1 2:25 and 15 for A2 =A1 4. By having a
closer look on the velocity distributions it becomes clear that
as expected the change in the behavior of the loss curve and
the critical angle are related to the onset of a separation bubble
in the diffuser. Finally, our simulations suggested that the standard
k  e model fails to predict the ow eld accurately whereas the
eddy-viscosity SST turbulence model and the BSL Reynolds-stress
model give very close results, suggesting the use of the SST model.
Future studies are planned to determine the loss coefcients for
uids with different rheology and other pipeline elements (e.g. Tjunctions, Y-junctions, etc). Such results would help the design of
such pipeline systems in a great manner.
It was interesting for the authors to realize how little is known
about turbulence in non-Newtonian uids, especially in terms of
the behavior of turbulence models implemented in CFD codes,
which are developed and (more importantly) validated for Newtonian uids.
Acknowledgments
The work reported in the paper has been developed in the
framework of the project, Talent care and cultivation in the scientic workshops of BME project. This project is supported by the
Grant TMOP-4.2.2.B-10/1-2010-0009.
References
[1] Cabral RAF, Telis VRN, Park KJ, Telis-Romero J. Friction losses in valves and
ttings for liquid food products. Food Bioprod Process 2011;89(4):37582.
[2] Monteiro ACS, Bansal PK. Pressure drop characteristics and rheological
modeling of ice slurry ow in pipes. Int J Refrig-Rev Int Froid 2010;33(8,
SI):152332.
[3] Grozdek M, Khodabandeh R, Lundqvist P. Experimental investigation of ice
slurry ow pressure drop in horizontal tubes. Exp Therm Fluid Sci
2009;33(2):35770.

123

[4] Kabwe AM, Fester VG, Slatter PT. Prediction of non-Newtonian head losses
through diaphragm valves at different opening positions. Chem Eng Res Des
2010;88(8):95970.
[5] Fester VG, Kazadi DM, Mbiya BM, Slatter PT. Loss coefcients for ow of
Newtonian and non-Newtonian uids through diaphragm valves. Chem Eng
Res Des 2007;85(9):131424.
[6] Neofytou P, Drikakis D. Non-Newtonian ow instability in a channel with a
sudden expansion. J Non-Newtonian Fluid Mech 2003;111(23):12750.
[7] Ternik P, Marn J, Zunic Z. Non-Newtonian uid ow through a planar
symmetric expansion: shear-thickening uids. J Non-Newtonian Fluid Mech
2006;135(23):13648.
[8] Marn J, Ternik P. Laminar ow of a shear-thickening uid in a pipe bend. Fluid
Dyn Res 2006;38(5):295312.
[9] Chowdhury MR, Fester VG. Modeling pressure losses for Newtonian and nonNewtonian laminar and turbulent ow in long square edged orices. Chem Eng
Res Des 2011;90:8639.
[10] Liu M, Duan YF. Resistance properties of coalwater slurry owing through
local piping ttings. Exp Therm Fluid Sci 2009;33(5):82837.
[11] Etemad SG. Turbulent ow friction loss coefcients of ttings for purely
viscous non-Newtonian uids. Int Commun Heat Mass Transfer
2004;31(5):76371.
[12] Polizelli MA, Menegalli FC, Telis VRN, Telis-Romero J. Friction losses in valves
and ttings for power-law uids. Braz J Chem Eng 2003;20(4):45563.
[13] Bandyopadhyay TK, Das SK. Non-Newtonian pseudoplastic liquid ow through
small diameter piping components. J Pet Sci Eng 2007;55(12):15666.
[14] Rosa S, Pinho FT. Pressure drop coefcient of laminar Newtonian ow in
axisymmetric diffusers. Int J Heat Fluid Flow 2006;27(2):31928.
[15] Turian RM, Ma TW, Hsu FLG, Sung DJ. Flow of concentrated non-Newtonian
slurries: 1. Friction losses in laminar, turbulent and transition ow through
straight pipe. Int J Multiphase Flow 1998;24(2):22542.
[16] Turian RM, Ma TW, Hsu FLG, Sung MDJ, Plackmann GW. Flow of concentrated
non-Newtonian slurries: 2. Friction losses in bends, ttings, valves and Venturi
meters. Int J Multiphase Flow 1998;24(2):24369.
[17] Gzel B, Frigaard I, Martinez DM. Predicting laminarturbulent transition in
Poiseuille pipe ow for non-Newtonian uids. Chem Eng Sci
2009;64(2):25464.
[18] Miller DS. Internal ow systems. BHRA uid engineering; 1978 [ISBN 0 900983
78 7].
[19] Richardson JF, Chhabra RP. Non-Newtonian ow and applied
rheology. Elsevier; 2008.
[20] Swamee PK, Aggarwal N. Explicit equations for laminar ow of Bingham
plastic uids. J Pet Sci Eng 2011;76(34):17884.
[21] Kfuri SLD, Silva JQ, Soares EJ, Thompson RL. Friction losses for power-law and
viscoplastic materials in an entrance of a tube and an abrupt contraction. J Pet
Sci Eng 2011;76(34):22435.
[22] Pal Gy, Ugron , Szikora I, Bojtr I. Flow in simplied and real models of
intracranial aneurysms. Int J Heat Fluid Flow 2007;28(4):65364 [Including
Special Issue of Conference on Modelling Fluid Flow (CMFF06), Budapest 13th
event of the international conference series in uid ow technologies:
Conference on Modelling Fluid Flow].
[23] Singh RK, Singh SN, Seshadri V. CFD prediction of the effects of the upstream
elbow ttings on the performance of cone owmeters. Flow Meas Instrum
2010;21(2):8897.
[24] Filali A, Khezzar L, Mitsoulis E. Some experiences with the numerical
simulation of Newtonian and Bingham uids in dip coating. Comput Fluids
2013;82(0):11021.
[25] Sherman FS. Viscous ow. McGraw-Hill; 1990.
[26] ANSYS, Inc. ANSYS CFX-Solver Theory Guide, Release 13.0.
[27] Erdal A, Andersson HI. Numerical aspects of ow computation through
orices. Flow Meas Instrum 1997;8(1):2737.
[28] Fox RW, McDonald AT. Introduction to uid mechanics. John Wiley & Sons
Inc.; 1994.
[29] Zierep J, Bhler K. Grundzge der Strmungslehre. Springer; 2008.
[30] Lorenzini M, Morini GL, Salvigni S. Laminar, transitional and turbulent friction
factors for gas ows in smooth and rough microtubes. Int J Therm Sci
2010;49(2):24855.
[31] Mays LW. Water distribution systems handbook; 2008 [ISBN 0 07 134213 3].
}s C. Predicting the friction factor in straight pipes in the case of
[32] Csizmadia P, Ho
Bingham plastic and the power-law uids by means of measurements and CFD
simulation. Period Polytech-Chem Eng 2013;57(12):7983.

You might also like