Sei sulla pagina 1di 11

CFD APPLIED TO TURBULENT FLOWS IN CONCENTRIC

AND ECCENTRIC ANNULI WITH INNER SHAFT ROTATION


J. L. Vieira Neto,
1
A. L. Martins,
2
A. Silveira Neto,
3
C. H. Atade
1
and M. A. S. Barrozo
1
*
1. Federal University of Uberl andia, School of Chemical Engineering, Building 1K, Campus Santa M onica, POB 593, 38400-902
Uberl andia, MG, Brazil
2. PETROBRAS R&D Center, Well Technology Sector, 21949-900 Rio de Janeiro, RJ, Brazil
3. Federal University of Uberl andia, School of Mechanical Engineering, Building 1M, Campus Santa M onica, 38400-902
Uberl andia, MG, Brazil
Turbulent ows in concentric and eccentric annuli with and without rotating inner cylinder were investigated by numerical simulations. Similar
ows occur in drilling operations of oil wells. Several turbulence models with Reynolds Average NavierStokes approach were used in the simulations
and the respective models predictions were compared with experimental data from the literature. The simulated results of axial and tangential
velocities show a good agreement with the experimental data. As compared with another turbulence models, the simulations with the standard
Reynolds Stress model presented a slightly better prediction for most of the responses studied.
On a analys e des turbulences en espaces annulaires concentriques et excentriques avec ou sans cylindre interne tournant ` a laide de simulations
num eriques. Des ecoulements semblables se produisent lors dactivit es de forage de puits de p etrole. Plusieurs mod` eles de turbulence avec
lapproche de NavierStokes en moyenne de Reynolds ont et e utilis es dans les simulations et les pr edictions respectives des mod` eles ont et e
compar ees aux donn ees exp erimentales de la litt erature. Les r esultats simul es des v elocit es axiales et tangentielles cadrent bien avec les donn ees
exp erimentales. Comparativement ` a dautres mod` eles de turbulence, les simulations avec le mod` ele de tension de Reynolds standard ont pr esent e
une pr ediction l eg` erement sup erieure pour la plupart des r eponses etudi ees.
Keywords: computational uid dynamics, turbulence, rotating annular ows
INTRODUCTION
T
he main interest of annular ows study is concerned with
the petroleumindustry, especially for application of the per-
forations mud used in drilling of oil wells. Drilling uids are
usually either water- or oil-based colloidal suspensions. Additives
are usually mixed with the uid to modify its drilling properties,
such as barite to increase the mud density, lignosulphonates to
reduce occulation, polymers to modify ltration rates or to act
as viscosiers, each of which may have a signicant effect on the
uid rheology (Escudier et al., 1995).
In drilling operations, the annular ow cannot be driven with
very slow velocities (in laminar regime) to avoid the gravel sed-
imentation to the bottom of the well, while extremely turbulent
ows can harmthe stability of the well. Therefore, it is very impor-
tant to get a better understanding of the rotating ow regimes in
annular sections.
Experimental investigations of the ow velocity characteristics
of rotating and non-rotating Newtonian ows in concentric and
eccentric annuli have been discussed by Nouri et al. (1993); and
Nouri and Whitelaw (1994, 1997). The Laser Doppler Velocimeter
(LDV) technique was used in these works to obtain the axial,
radial, and tangential mean velocity proles and their respective
uctuations and crossed correlations.
The experimental study of Kaye and Elgar (1958) also com-
bined axial and rotational ow in an annulus section with a
rotating inner wall. These authors identied the following basic
ow regimes in the annular gap: laminar ow, laminar ow
with vortices, turbulent ow, and turbulent ow with vortices.
Another experimental work of Yamada (1962) showed that the

Author to whom correspondence may be addressed.


E-mail address: masbarrozo@ufu.br
Can. J. Chem. Eng. 89:636646, 2011

2011 Canadian Society for Chemical Engineering


DOI 10.1002/cjce.20522
Published online 30 March 2011 in Wiley Online Library
(wileyonlinelibrary.com).
| 636 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|
pressure drop increased sharply when the inner shaft was rotat-
ing and decreased with increasing axial bulk ow (Chung and
Sung, 2005).
On the turbulence modelling domain, the works of Chung and
Sung (2005); Chung et al. (2002); and Ninokata et al. (2006) can
be mentioned. These authors used the experimental data of Nouri
et al. (1993); and Nouri and Whitelaw (1994, 1997) to validate
their numerical results. Chung and Sung (2005) studied the tur-
bulent concentric ow with rotation of an inner cylinder using
the Large-Eddy Simulation (LES) methodology and their numer-
ical results showed that the alteration of the near-wall turbulent
structures can be attributed to the destabilising effect of rotation
of the inner wall.
Chung et al. (2002) used Direct Numeric Simulation (DNS)
technique for the study of turbulent ows in a concentric annu-
lus channel at low Reynolds numbers. Their numerical results
showed that the turbulent structures near the outer wall are more
activated than those near the inner wall, which may be attributed
to the different vortex regeneration processes between the inner
and outer walls.
Ninokata et al. (2006) applied the DNS approach to a fully
developed single-phase turbulent ow analysis for an eccentric
annulus channel conguration. This study was focused on the
heterogeneity and local laminarisation phenomena. First, the
authors performed a comparison between the DNS calculation
and the numerical benchmark results with the spectral method
to demonstrate the accuracy of the DNS for ows in concen-
tric annuli. Then, the computations were extended to ows in
eccentric annuli, rst to get qualitative insights into the turbulent
structure in reference to the experimental data available in the lit-
erature, and then, to see the local laminarisations near the narrow
gap region.
Due to the high computational cost required by the LES and DNS
methodologies, it is important to investigate if turbulence models
with Reynolds Average NavierStokes (RANS) approach can lead
velocity proles with good agreement with the experimental data
and with less computational effort.
Torii and Yang (1994) employed several k models to study
ows through concentric annuli with rotating inner walls and
they found that increasing the Taylor number (Ta) amplied the
turbulence kinetic energy, resulting in a substantial enhancement
of the heat transfer. Hanjalic (1974) studied turbulence features
in annular ducts based on the Reynolds-averaged NavierStokes
(RANS) equations with a differential transport model. Azouz and
Shirazi (1998) used RANS approach to predict the turbulent ow
in annular channels. Kang et al. (2001) carried out numerical
simulation with RANS method on the isothermal and heated tur-
bulent upow in a vertical concentric annular channel, with its
inner wall heated.
The purpose of the present work are: to simulate rotating and
non-rotating turbulent ows of Newtonian uids in concentric
and eccentric annular sections using CFD with different turbu-
lence models based on RANS approach and, thus, to compare
the results obtained with the experimental data reported in the
literature (Nouri et al., 1993; Nouri and Whitelaw, 1994, 1997).
MATERIALS AND METHODS
Experimental Conditions
The experimental apparatus for concentric and eccentric (with
0.5 of eccentricity) annular gaps were extracted from Nouri and
Whitelaw (1994, 1997) studies with these dimensions: 40.3 mm
of outer diameter (D
o
) and 20.0 mmof inner diameter (D
in
), hence
the hydraulic diameter (D
H
) was 20.3 mm. The Newtonian uid
was a 31.8%mixture of tetraline in turpentine at 25

Cwith density
(,) of 896 kg/m
3
and kinematic viscosity (u) of 1.63 10
6
m
2
/s.
The axial bulk velocity (U
b
) for fully developed turbulent owwas
2.14 m/s that gives a bulk Reynolds number (Re =U
b
D
H
/u) of
26 600. The inner cylinder tip velocity (V
t
) was 0.315 m/s that
corresponds to 300 rpm.
Modelling
Assuming an isothermal, incompressible and turbulent ow of
a Newtonian uid, the ow modelling should be described by
Reynolds averaging approach, in which the ow variables in the
instantaneous (exact) NavierStokes equations are decomposed
into a mean (ensemble-average or time-average) component and
a uctuating component. For the velocity components:
u
i
= u
i
+u

i
(1)
where u
i
and u

i
are the mean and uctuating velocity components
(i =1, 2, 3).
Likewise, for pressure and other scalar quantities:
=

+

(2)
where denotes a scalar, such as pressure, energy, or concentra-
tion of species.
Substituting expressions of this form for the ow variables into
the instantaneous continuity and momentum equations and tak-
ing ensemble (or time) average (and dropping the overbar on
the mean velocity, u) yields the ensemble-averaged momentum
equations. The momentum equations can be written in Cartesian
coordinates form as:

t
(,u
i
) +

x
j
_
,u
i
u
j
_
=
p
x
i
+

x
j
_

_
u
i
x
j
+
u
j
x
i

2
3

ij
u
l
x
l
__
+

x
j

_
,u

i
u

j
_
(3)
These equations, called RANS equations, have the same gen-
eral form as the instantaneous NavierStokes equations, with
the velocities and other solution variables now representing
ensemble-averaged (or time-averaged) values. Additional terms
appear to represent the turbulence effects (Fluent, 2008).
In this work, the RANS turbulence models used in the CFD
simulations were the k Standard, k Renormalisation Group
(RNG), k Standard, and k Shear-Stress Transport (SST)
with two additional transport equations, and the RSM Linear
model with seven additional transport equations. The k Stan-
dard model proposed by Launder and Spalding (1972) is a
semi-empirical model based on model transport equations for
turbulence kinetic energy (k) and its dissipation rate ():

t
(,k) +

x
i
(,ku
i
) =

x
j
__
+

t
o
k
_
k
x
j
_
+G
k
+G
b
,Y
M
+S
k
(4)

t
(,) +

x
i
(,u
i
) =

x
j
__
+

t
o

_

x
j
_
+C
1

k
(G
k
+C
3
G
b
)C
2
,

2
k
+S

(5)
|
VOLUME 89, AUGUST2011
| |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| 637 |
The k RNG turbulence model of Choudhury (1993) is derived
from the instantaneous NavierStokes equations, using a mathe-
matical technique called RNG method. The analytical derivation
results in a model with constants different from k Standard
model, and additional terms and functions in the transport equa-
tions:

t
(,k) +

x
i
(,ku
i
) =

x
j
_

k

eff
k
x
j
_
+G
k
+G
b
,Y
M
+S
k
(6)

t
(,) +

x
i
(,u
i
) =

x
j
_


eff

x
j
_
+C
1

k
(G
k
+C
3
G
b
)
C
2
,

2
k
R

+S

(7)
The k Standard model proposed by Wilcox (1998) is an
empirical model based on the transport equations for the tur-
bulence kinetic energy (k) and the specic dissipation rate (),
which can also be thought of as the ratio of to k. The k
Standard model has the following transport equations:

t
(,k) +

x
i
(,ku
i
) =

x
j
_
I
k
k
x
j
_
+G
k
Y
k
+S
k
(8)

t
(,) +

x
i
(,u
i
) =

x
j
_
I

x
j
_
+G

+S

(9)
The k SST model is a variation of the k Standard model.
In k SST, the denition of the turbulent viscosity is modied to
account for the transport of the principal turbulent shear stress.
This feature gives to the k SST model an advantage in terms
of performance over both the k Standard model and the k
Standard model. Other modications include the addition of a
cross-diffusion term in the () equation and a blending function
to ensure that the model equations behave appropriately in both
the near-wall and far-eld zones:

t
(,k) +

x
i
(,ku
i
) =

x
j
_
I
k
k
x
j
_
+

G
k
Y
k
+S
k
(10)

t
(,) +

x
i
(,u
i
) =

x
j
_
I

x
j
_
+G

+D

+S

(11)
The Reynolds Stress model (RSM; Launder et al., 1975; Gib-
son and Launder, 1978; Launder, 1989a) involves calculation of
the individual Reynolds stresses, u

i
u

j
, using differential trans-
port equations. The individual Reynolds stresses are then used
to obtain closure of the Reynolds-averaged momentum equation
(Equation 3). The exact form of the Reynolds stress transport
equations may be derived by taking moments of the exact momen-
tum equation. This is a process wherein the exact momentum
equations are multiplied by a uctuating property, the product
then being Reynolds averaged. Unfortunately, several of the terms
in the exact equation are unknown and modelling assumptions
are required in order to close the equations. The exact transport
equations for the transport of the Reynolds stresses, ,u

i
u

j
, may
be written as follows:

t
_
,u

i
u

j
_
. .
Local Time Derivative
+

x
k
_
,u
k
u

i
u

j
_
. .
C
ij
Convection
=

x
k
_
,u

i
u

j
u

k
+p
_

kj
u

i
+
ik
u

j
_
_
. .
D
T.ij
Turbulent Diffusion
+

x
k
_


x
k
_
u

i
u

j
_
_
. .
D
L.ij
Molecular Diffusion
,
_
u

i
u

k
u
j
x
k
+u

j
u

k
u
i
x
k
_
. .
P
ij
Stress Production
,
_
g
i
u

j
+g
j
u

_
. .
G
ij
Buoyancy Production
+p
_
u

i
x
j
+
u

j
x
i
_
. .

ij
Pressure Strain
2
u

i
x
k
u

j
x
k
. .

ij
Dissipation
2, O
k
_
u

j
u

ikm
+u

i
u

jkm
_
. .
F
ij
Production by System Rotation
+ S
user
..
User-Dened Source Term
(12)
The turbulent diffusion, D
T.ij
, can be modelled by the gen-
eralised gradient-diffusion model of Daly and Harlow (1970).
Table 1. Geometric dimensions and boundary conditions adopted in the numerical simulations with the Newtonian uid (31.8% mixture of
tetraline in turpentine)
Without rotation With inner pipe rotation
(=0rpm) (=300rpm)
Geometric dimensions
Outer diameter (D
o
, mm) 40.3 40.3
Inner diameter (D
in
, mm) 20.0 20.0
Hydraulic diameter (D
H
, mm) 20.3 20.3
Computational length (L
z
=5D
H
, mm) 101 101
Boundary conditions
Volume ow rate (m
3
/s) 2.0610
3
2.0610
3
Periodic mass ow rate (kg/s) 1.837 1.837
Inner cylinder surface velocity (V
t
, m/s) 0 0.315
Bulk velocity (U
b
, m/s) 2.14 2.14
Bulk Reynolds number (Re =U
b
D
H
/) 26600 26600
Mixture temperature (

C) 25 25
Density of the mixture (, kg/m
3
) 896 896
Kinematic viscosity of the mixture (, m
2
/s) 1.6310
6
1.6310
6
| 638 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|
Figure 1. Concentric computational grid: (a) Face with plans 14 and (b) periodic section.
Figure 2. Eccentric computational grid: (a) Face with plans 14 and (b) periodic section.
However, this equation can result in numerical instabilities, so
it has been simplied to use a scalar turbulent diffusivity (Lien
and Leschziner, 1994). The pressurestrain term,
ij
, is modelled
according to the proposals of Gibson and Launder (1978); Fu
et al. (1987); and Launder (1989b). In general, when the tur-
bulence kinetic energy is needed for modelling a specic term,
it is obtained by taking the trace of Reynolds stress tensor: k =
(1,2)u

i
u

i
. However, there is another option to model the turbu-
lence kinetic energy, using the following equation:

t
(,k) +

x
i
(,ku
i
) =

x
j
__
+

t
o
k
_
k
x
j
_
+
1
2
(P
ii
+G
ii
),(1 +2M
2
t
) +S
k
(13)
Simulation Conditions
The experimental parameters for concentric and eccentric (with
0.5 of eccentricity) annular gaps were extracted from Nouri and
Whitelaw (1994, 1997) studies. The dimensions of the experi-
mental apparatus, as well as the boundary conditions adopted
in the simulations are described in Table 1. A periodic bound-
ary condition was imposed in the axial direction by a mass
ow rate, and a non-slip boundary condition was imposed at
the solid walls. In all the cases, the computational length in
the axial direction was L
z
=5 D
H
. The average velocity pro-
les obtained with this periodic condition were invariant along
the axial direction. The standard deviations between the veloc-
ities proles obtained along the axial direction were less then
1 10
04
.
Figure 3. Simulated and experimental results for axial velocity proles normalised with bulk velocity (U
b
) in a concentric gap: (a) Experimental data of
Nouri et al. (1993) without inner shaft rotation and (b) experimental data of Nouri and Whitelaw (1994) with inner shaft rotation speed of 300rpm.
|
VOLUME 89, AUGUST2011
| |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| 639 |
Figure 4. Simulated and experimental (Nouri and Whitelaw, 1994) results for tangential velocity proles in a concentric gap with an inner shaft
rotation speed of 300rpm: (a) normalised with bulk velocity (U
b
) and (b) normalised with tip velocity (V
t
).
The numerical simulations were carried out using the software
FLUENT

6.3.26 with segregated solver strategy and convergence


criteria of 1 10
04
(Pereira et al., 2007). The simulation time
was 2 s with a time step size of 1 10
05
s. The SIMPLE algo-
rithm was used for the pressurevelocity coupling, the PRESTO
(pressure staggered option) scheme was adopted for the pressure
discretisation, and the First Order Upwind algorithm was used for
the momentum equations discretisation and for the turbulence
models equations. The RANS turbulence models used in the
CFD simulations were the k Standard, k RNG, k Standard,
and k SST with two additional transport equations, and the
RSM Linear model with seven additional transport equations. The
default values (Fluent, 2008) were adopted for models constants
of each turbulence model.
Figure 5. Simulated and experimental (Nouri et al., 1993) results for axial velocity proles normalised by bulk velocity (U
b
) in an eccentric annulus
without inner shaft rotation in the plans of the eccentric arrangement: (a) plan 1, (b) plan 2, (c) plan 3, and (d) plan 4.
| 640 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|
A grid independence test was carried out with three different
grid sizes. In the grid independence study, it was veried that
additional grid lines did not change the average simulated velocity
proles. It was observed that the narrowest gap is more sensitive
to grid size, when compared to the other plans. In most high-
Reynolds number ows, the wall function approach substantially
saves computational resources, because the viscosity-affected near
wall region, in which the solution variables change most rapidly,
does not need to be resolved. Thus, the wall function approach
becomes popular because it is economical, robust, and reason-
ably accurate. For the standard wall functions adopted in the CFD
simulations, the wall-adjacent cells centroid should be located
within the log-law layer (30 -y
+
-300). In the present work the
y
+
value calculated was close to the lower bound (y
+
30).
Figures 1 and 2 show the concentric and eccentric tri-
dimensional computational grids (with the same dimensions of
the experimental apparatus), built through the mesh generation
software GAMBIT

. The grids have 49 152 cells with 64 16 48


subdivisions in the circumferential, radial, and axial directions,
respectively.
RESULTS AND DISCUSSION
Simulated Velocity Proles and Experimental Data
Figures 37 present simulated and experimental results for a
fully developed turbulent ow (Re =26 600) in concentric and
eccentric arrangements. In the eccentric arrangement the simu-
lations were carried out with 0.5 of eccentricity (e) dened as:
e =L/(R
o
R
in
), with L the distance between the centres of the
inner and outer pipes and R
in
and R
o
the inner and outer radii,
respectively.
Figure 3a presents a comparison between simulation results
and the experimental data obtained by Nouri et al. (1993) for
a non-rotation of the inner cylinder. Figure 3b shows a similar
comparison for an inner shaft rotation speed of 300 rpm, using
the experimental data of Nouri and Whitelaw (1994). It can be
observed in both comparisons (Figure 3a and b) a good agreement
between the simulated and experimental proles of axial velocity,
especially for the case of rotating the inner cylinder. Moreover,
simulations results obtained by the ve turbulence models are
very close to each other, and the relative errors computed were
near 5.5% for non-rotating ows and less than 5% for rotating
ows.
Figure 4 shows the numerical simulations results and the
experimental data (Nouri and Whitelaw, 1994) for the radial dis-
tribution of tangential velocity proles normalised with the bulk
velocity, U
b
(Figure 4a) and normalised with the inner shaft tip
velocity, V
t
(Figure 4b) in the concentric arrangement. It also can
be seen in Figure 4 a good agreement between the numerically
simulated results and the experimental data. Among the used tur-
bulence models, the RSM Standard model gives slightly better
predictions, with relative errors around 11%, against nearly 17%
for the two-equations models.
Figure 6. Simulated and experimental (Nouri and Whitelaw, 1997) results for axial velocity proles normalised with bulk velocity (U
b
) in an eccentric
gap with inner shaft rotation of 300rpm in the eccentric gap plans: (a) plan 1, (b) plan 2, (c) plan 3, and (d) plan 4.
|
VOLUME 89, AUGUST2011
| |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| 641 |
Figure 5 presents experimental (Nouri et al., 1993) and
simulated results for the radial distribution of axial velocities nor-
malised with bulk velocity (U
b
) without the rotation of the inner
cylinder (0 rpm), in the four studied plans of the eccentric arrange-
ment (see details of these plans in Figure 2). Figure 5c shows that,
in the larger annular space (plan 3) the axial normalised veloci-
ties are nearly 1.5 U
b
that indicates a canalisation of the axial
ow for this region. The simulation results have a good agree-
ment with the experimental data, for all the plans, except for the
smallest gap (plan 1), in which the simulation under predicts the
experimental results. The relative errors computed for the small-
est region were about 20% while the relative errors for the other
regions of the eccentric annuli were less than 1%.
Figure 6 shows experimental (Nouri and Whitelaw, 1997) and
numerical results for the radial distribution of axial velocities nor-
malised with bulk velocity (U
b
), with an inner shaft rotation
speed of 300 rpm, in the four plans of the eccentric arrange-
ment. In this condition, it can also be observed a good agreement
between numerical results and the experimental data of Nouri
and Whitelaw (1997). The RSM Linear Standard model appears
to yield results that have better agreement with the experimental
data with relative errors varying within a range of 1.65%, except
for the largest annular space (plan 3) in which the k SST and
k Standard models have gotten better results with relative errors
around 3%. The RSM turbulence model deals with the effects of
streamline curvature, swirl, and rapid changes in the strain rate
in a more rigorous manner than the classical two transport equa-
tions models. Therefore, this model (RSM) has a greater potential
to accurately predict complex ows, as in the case of the rotat-
ing ows in annular sections. However, as the k models have
been modied over the years, production terms have been added
to both the k and equations, which have improved the accuracy
of these models for predicting free shear ows, which could be
a possible explanation to the better results with the k models
in the largest annular space. A more detailed discussion of the
turbulence models is given in Fluent (2008).
Figure 7 presents the radial distribution of tangential velocity
normalised with bulk velocity (U
b
) with an inner shaft rotation
speed of 300 rpm, in the plans of the eccentric arrangement. It
can be observed a good agreement between the simulated and
experimental data of Nouri and Whitelaw (1997) for the larger
annulus space (plan 3) with relative errors around 6% for the
RSM and k SST models, while for the smallest gap (plan 1)
the simulated results under predict the experimental data with
relative errors of about 20%.
Contours of Velocities in the Eccentric Annulus
Figure 8 shows the contours of axial velocity in the eccentric
gap without inner tip rotation (0 rpm) for each turbulence model.
A preferential axial ow can be observed in the largest annular
region (plan 3, details see Figure 2) with maximum velocity of
Figure 7. Simulated and experimental (Nouri and Whitelaw, 1997) results for tangential velocity proles normalised with bulk velocity (U
b
) in an
eccentric gap with inner shaft rotation of 300rpm in the plans of eccentric annulus: (a) plan 1, (b) plan 2, and (c) plan 3.
| 642 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|
Figure 8. Simulated contours of axial velocity (m/s) in an eccentric annulus (e =0.5) without inner shaft rotation for each turbulence model: (a) k
Standard, (b) k RNG, (c) k Standard, (d) k SST, and (e) RSM Linear Standard.
approximately 2.94 m/s, while, in the narrow section (plan 1) the
axial velocity is reduced to approximately 1.2 m/s, because the
resistence to ow is increased as the gap between the two pipes
decreases. For directional drilling operation, the low velocity in
the narrow part causes particle settling leading to a cuttings bed
formation. The same tendency was veried in the Mustafa (1989);
and Hussain (1999) numerical studies using bipolar coordinates
to obtain velocity and viscosity proles of non-Newtonian ows
in varying eccentric annuli.
Figures 9 and 10 show, respectively, the contours of axial and
tangential velocities in an eccentric annulus with an inner shaft
rotation speed of 300 rpm, for each turbulence model. Figure 9
|
VOLUME 89, AUGUST2011
| |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| 643 |
shows a displacement of the preferential axial ow from the wider
gap (plan 3) to the narrowest areas corresponding to the plans 2
and 1, due to the rotation of the inner shaft. The axial velocity in
the larger gap is reduced to approximately 2.78 m/s, while, in the
strait region the velocity is increased to 2 m/s. The increment of
the axial velocity in the strait region is relevant in cleaning oper-
ations and cuttings transportation in horizontal and directional
drilling systems. The axial velocity contours also shows a similar
Figure 9. Simulated contours of axial velocity (m/s) in an eccentric gap (e =0.5) with inner shaft rotation (300rpm) for each turbulence model: (a) k
Standard, (b) k RNG, (c) k Standard; (d) k SST, and (e) RSM Linear Standard.
| 644 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|
Figure 10. Simulated contours of tangential velocity (m/s) in an
eccentric gap (e =0.5) with inner shaft rotation (300rpm) for each
turbulence model: (a) k Standard, (b) k RNG, (c) k Standard,
(d) k SST, (e) RSM Linear Standard.
core ow displacement for all the turbulence models, except for
the RSM model that presents higher values in the narrowest areas
(plans 1 and 4) of the eccentric annulus. The same tendency have
been observed in Figure 6, in which RSM model presents better
agreement with experimental data, with relative errors varying
within a range of 1.65%. In Figure 10 the tangential velocity
contours shows a similar velocity distribution for all the turbu-
lence models, except for the RSM model in which the contour
lines present some undulations, and higher tangential velocity in
the narrow gap (plan 1).
CONCLUSIONS
In the concentric arrangement there was a good agreement in sim-
ulated scenarios of axial and tangential velocities compared with
the experimental data (Nouri et al., 1993; Nouri and Whitelaw,
1994), especially with the rotation of the inner cylinder (300 rpm).
The simulations results obtained by the ve turbulence models
were very close to each other, although, the RSM Standard model
got slightly better predictions of tangential velocity.
The predicted values from CFD simulation of axial velocity in
the eccentric (e =0.5) arrangement for a fully developed turbulent
ow (Re =26 000) without rotation of the inner cylinder (0 rpm)
agree very well with the experimental data of Nouri et al. (1993)
for all the plans, except in the smallest gap (plan 1), where the
simulation under predict the experimental data.
It was also observed that the axial velocity in the narrow part of
the eccentric annulus is signicantly lower than the axial velocity
in the wider part in the non-rotation condition. This makes the
narrow lower region of the annulus vulnerable to cuttings bed
deposition under directional drilling situation. This, however, is
not benecial in cuttings removal because under practical drilling
applications the larger eccentricity causes cuttings bed to develop
that partially blocks the passages. Otherwise, the rotating inner
shaft effect increases the axial velocity in this narrow part, and
can thus reduce the cuttings bed deposition.
It can be also veried in the eccentric arrangement that when
the rotation to the inner shaft was incorporated (300 rpm), the
simulated axial velocity results also showed a good agreement
with the experimental data of Nouri and Whitelaw (1997) and
the simulated proles obtained by the ve turbulence models were
very close to each other. The turbulence model that had slightly
better agreement with the experimental data was the RSM Linear
Standard model, except for the largest annular space (plan 3),
in which the k SST and k Standard models got little better
results.
The numerical simulations for tangential velocities in an eccen-
tric channel with rotation of the inner shaft (300 rpm) also showed
a good agreement with the experimental data for the largest region
(plan 3) and underestimated the experimental data for the narrow
gap (plan 1).
NOMENCLATURE
C
1
, C
2
, C
3
the constants of the Standard k model
C
ij
the convection for the RSM Linear model
D
H
hydraulic diameter
D
in
inner cylinder diameter
D
L.ij
the molecular diffusion for the RSM Linear model
D
o
outer pipe diameter
D
T.ij
the turbulent diffusion for the RSM Linear model
D

the cross-diffusion term for the SST k model


e eccentricity
F
ij
production by system rotation for the RSM Linear
model
G
b
the generation of turbulence kinetic energy due to
buoyancy
G
ij
the production terms due to buoyancy for the RSM
Linear model
G
k
the generation of turbulence kinetic energy due to
the mean velocity gradients

G
k
generation of turbulence kinetic energy due to mean
velocity gradients for the k SST model
G

the generation of
k turbulent kinetic energy
L distance between the centres of the inner and outer
pipes
M
t
turbulent Mach number
p pressure
P
ij
the stress production for the RSM Linear model
r1/S radial position normalised with the gap between the
outer and inner cylinders
Re bulk Reynolds number
R
in
inner cylinder radius
R
o
outer pipe radius
R

additional term in the equation of the RNG k


model
S the gap between the outer and inner cylinders
S
k
user-dened source term for k
S

user-dened source term for


S

user-dened source terms for


S
user
user-dened source term of the RSM model
u velocity component in z direction
u, v, w velocity magnitude, also written with directional
subscripts (e.g., v
x
, v
y
, v
z
, v
r
)
U
b
bulk velocity
u
i
mean velocity components (i =1, 2, 3)
u

i
uctuating velocity components (i =1, 2, 3)
u

i
u

j
individual Reynolds stresses
V
t
inner cylinder tip velocity
w velocity component in direction
|
VOLUME 89, AUGUST2011
| |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| 645 |
Y
k
and Y

the dissipation of k and due to turbulence


Y
M
contribution of uctuating dilatation in compress-
ible turbulence to the overall dissipation rate
Greek Letters

k
and

the inverse effective Prandtl numbers for k and ,


respectively

ij
delta function
dissipation rate of k

ij
the dissipation tensor for the RSM Linear model
denotes a scalar, such as, pressure, energy, or con-
centration of species

ij
the pressurestrain term for the RSM Linear model
I
k
and I

the effective diffusivity of k and , respectively


dynamic viscosity

eff
effective viscosity for the RNG k model

t
turbulent viscosity
, uid density
o
k
and o

the turbulent Prandtl numbers for k and , respec-


tively
v uid kinematics viscosity
specic dissipation rate
ACKNOWLEDGEMENTS
The authors wish to thank the nancial support from Petrobras.
REFERENCES
Azouz, I. and S. A. Shirazi, Evaluation of Several Turbulence
Models for Turbulent Flow in Concentric and Eccentric
Annuli, ASME J. Energy Resour. Technol. 120, 268275
(1998).
Choudhury, D., Introduction to the Renormalisation Group
Method and Turbulence Modelling, in Fluent Inc. Technical
Memorandum, TM-107. Fluent Inc., USA (1993).
Chung, S. Y. and H. J. Sung, Large-Eddy Simulation of
Turbulent Flow in a Concentric Annulus With Rotation of an
Inner Cylinder, Int. J. Heat Fluid Flow 26, 191203 (2005).
Chung, S. Y., G. H. Rhee and H. J. Sung, Direct Numerical
Simulation of Turbulent Concentric Annular Pipe Flow. Part
1: Flow Field, Int. J. Heat Fluid Flow 23, 426440 (2002).
Daly, B. J. and F. H. Harlow, Transport Equations in
Turbulence, Phys. Fluids 13, 2634 (1970).
Escudier, M. P., I. W. Gouldson and D. M. Jones, Flow of
Shear-Thinning Fluids in a Concentric Annulus, Exp. Fluids
18, 225238 (1995).
Fluent 6.3.26. Users Guide, ANSYS, Inc., USA (2008).
Fu, S., B. E. Launder and M. A. Leschziner, Modelling Strongly
Swirling Recirculating Jet Flow With Reynolds-Stress
Transport Closures, in Proc. Sixth Symp. Turbulent Shear
Flows, Toulouse, France (1987).
Gibson, M. M. and B. E. Launder, Ground Effects on Pressure
Fluctuations in the Atmospheric Boundary Layer, J. Fluid
Mech. 86, 491511 (1978).
Hanjalic, K., Prediction of Turbulent Flow in Annular Ducts
With Differential Transport Model of Turbulence, Warme
Stoffubertragung 7, 7278 (1974).
Hussain, Q. E., Numerical Investigation of Viscoplastic Fluid
Flow in Irregular Annuli, PhD Thesis, Department of
Aerospace Engineering and Mechanics, University of
Alabama, Alabama, USA (1999).
Kang, S., B. Patil, J. A. Zarate and R. P. Roy, Isothermal and
Heated Turbulent Upow in a Vertical Annular ChannelPart
I. Experimental Measurements, Int. J. Heat Mass Transfer
44, 11711184 (2001).
Kaye, J. and E. C. Elgar, Modes of Adiabatic and Diabatic Fluid
Flow in an Annulus With an Inner Rotating Cylinder, Trans.
ASME J. Heat Transfer 80, 753765 (1958).
Launder, B. E., Second-Moment Closure: Present and Future?
Int. J. Heat Fluid Flow 10, 282300 (1989a).
Launder, B. E., Second-Moment Closure and Its Use in
Modelling Turbulent Industrial Flows, Int. J. Numer.
Methods Fluids 9, 963985 (1989b).
Launder, B. E. and D. B. Spalding, Lectures in Mathematical
Models of Turbulence, Academic Press, London, England
(1972).
Launder, B. E., G. J. Reece and W. Rodi, Progress in the
Development of a Reynolds-Stress Turbulence Closure, J.
Fluid Mech. 68, 537566 (1975).
Lien, F. S. and M. A. Leschziner, Assessment of Turbulent
Transport Models Including Non-Linear RNG Eddy-Viscosity
Formulation and Second-Moment Closure, Comp. Fluids 23,
9831004 (1994).
Mustafa, H., Non-Newtonian Fluid Flow in Eccentric Annuli
and Its Application to Petroleum Engineering Problems, PhD
Thesis, Department of Petroleum Engineering, Louisiana
State University and Agricultural and Mechanical College,
Louisiana, USA (1989).
Ninokata, H., T. Okumura, E. Merzari and T. Kano, Direct
Numerical Simulation of Turbulent Flows in an Eccentric
Annulus Channel, in Annual Report of the Earth Simulator
Center, Earth Simulator Center, Yokohama, (2006), 293297.
Nouri, J. M. and J. H. Whitelaw, Flow of Newtonian and
Non-Newtonian Fluids in a Concentric Annulus With
Rotation of the Inner Cylinder, J. Fluids Eng. 116, 821827
(1994).
Nouri, J. M. and J. H. Whitelaw, Flow of Newtonian and
Non-Newtonian Fluids in an Eccentric Annulus With the
Rotation of the Inner Cylinder, Int. J. Heat Fluid Flow 18,
236246 (1997).
Nouri, J. M., H. Umur and J. H. Whitelaw, Flow of Newtonian
and Non-Newtonian Fluids in Concentric and Eccentric
Annuli, J. Fluid. Mech. 253, 617641 (1993).
Pereira, F. A. R., M. A. S. Barrozo and C. H. Atade, CFD
Predictions of Drilling Fluid Velocity and Pressure Proles in
Laminar Helical Flow, Braz. J. Chem. Eng. 24, 587595
(2007).
Torii, S. and W. J. Yang, Numerical Study on Turbulent Flow
and Heat Transfer in Circular Couette Flows, Numer. Heat
Transfer, Part A: Appl. 26, 321336 (1994).
Wilcox, D. C., Turbulence Modelling for CFD, DCW Industries,
Inc., La Canada, California (1998).
Yamada, Y., Resistence of a Flow Through an Annulus With an
Inner Rotating Cylinder, Bull. JSME 5, 302310 (1962).
Manuscript received January 27, 2010; revised manuscript
received December 23, 2010; accepted for publication January 6,
2011.
| 646 |
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
| |
VOLUME 89, AUGUST 2011
|

Potrebbero piacerti anche