Engineering Applications of Computational Fluid

Mechanics

ISSN: 1994-2060 (Print) 1997-003X (Online) Journal homepage: http://www.tandfonline.com/loi/tcfm20

Hydromechanics Characterization of The

Turbulent Flow Generated by Anchor Impellers

S. Karray, Z. Driss, H. Kchaou & M. S. Abid

To cite this article: S. Karray, Z. Driss, H. Kchaou & M. S. Abid (2011) Hydromechanics
Characterization of The Turbulent Flow Generated by Anchor Impellers,
Engineering Applications of Computational Fluid Mechanics, 5:3, 315-328, DOI:
10.1080/19942060.2011.11015374

To link to this article: http://dx.doi.org/10.1080/19942060.2011.11015374

Copyright 2011 Taylor and Francis Group

LLC

Published online: 19 Nov 2014.

Submit your article to this journal

Article views: 87

View related articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tcfm20

Download by: [109.161.202.118] Date: 24 February 2016, At: 17:55

 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3, pp. 315–328 (2011)

HYDROMECHANICS CHARACTERIZATION OF THE TURBULENT

FLOW GENERATED BY ANCHOR IMPELLERS
S. Karray*, Z. Driss, H. Kchaou and M. S. Abid

Laboratory of Eletro-Mechanic Systems (LASEM), National School of Engineers of Sfax (ENIS),

University of Sfax (US), B.P 1173, Km 3.5 Soukra, 3038 Sfax, TUNISIA
* E-Mail: [email protected] (Corresponding Author)

ABSTRACT: Hydromechanics of a stirred tank generated by a classical anchor and an anchor blade are
numerically simulated. The numerical results gives good prediction of the hydrodynamics such as the velocity field,
the turbulent kinetic energy and its dissipation rate. These results are obtained by the computational fluid dynamics
(CFD) code. The mechanical deformation of these anchors is calculated by the computational structural dynamics
(CSD) code. The deformed shape of the anchor blade has been found in turbulent flow. However, the anchor blade
has not been affected by this flow. Predictions of the numerical results have been compared with literature data and a
satisfactory agreement has been found.
Keywords: classical anchor, anchor blade, stirred tank, computer simulation, finite volume, CFD, CSD
Downloaded by [109.161.202.118] at 17:55 24 February 2016

1. INTRODUCTION increase of rotational speed and by the use of

impellers with a higher blade height and with a
A mixing operation can be defined as an artificial blade support. Akiti et al. (2005) have studied
creation of a fluid flow to decrease its also the behavior of the fluid in a big tank with an
heterogeneity, to accelerate its transfers and to anchor impeller. They observed that the flow is
achieve a certain degree of homogeneity. These weak at the bottom of the anchor impeller.
factors are related to the flow type and However, this agitator can be used in industrial
geometrical shape of the agitator. For this applications with non-viscous fluids. Certainly,
purpose, we have to take into account especially this situation is not that common, but it does exist.
the agitator geometry. For example, the anchors Baccar and Abid (1999) used two agitators for the
impellers have two arms to clean up the vessel thermal treatment of non-viscous media in which
wall by bringing back into the bulk the clay case heat transfer occurs in a turbulent regime.
agglomerates that tend to accumulate on the wall. Other researchers were interested in the study of
So, it has a minimal clearance between the the thermal performances of an anchor impeller.
outside of the blade and the tank wall. This type For this purpose, they determined the global
of agitators is commonly used for mixing highly exchange coefficient in the stirred tanks (Triveni
viscous fluids or heating sensitive materials et al., 2008; Baccar et al., 2001). However, all of
currently encountered in the food industry. these works were not concerned with the effect of
Despite the vast amount of publications on the the anchor deformation. This phenomenon
stirred tank, there is a paucity of published work appears essentially in the case of thin arms and at
that dealt with the anchor impeller. Abid et al. the highest speed. In the present work,
(1992), Pedrosa and Nunhez (2000) and Savreux hydrodynamics of a stirred vessel by a classical
et al. (2007) studied already this type of agitator. anchor and an anchor blade are numerically
These works are concerned with anchor impellers simulated in order to analyze the in-depth
in laminar flow. Indeed, Savreux et al. (2007) turbulence characteristics, the hydrodynamics and
have studied numerically the laminar 2D mixing the effect of the anchor arms deformation. For
of viscoplastic fluids with an anchor impeller in a this purpose, a numerical method is developed to
stirred tank. They examined in detail the flow characterize the structure of the turbulent flow
structures and particularly the static and moving generated by anchors. The rest of this paper is
rigid zones induced by the plasticity such as the organized as follows: Section 2 describes the
presence of vortices. Pedrosa and Nunhez (2000) geometry of the system. Section 3 provides the
have used a CFD code to study a model of a governing equations of the fluid and the structure.
stirred vessel using the anchor impeller. They Section 4 describes the numerical code. Section 5
showed that mixing can be improved by the

Received: 7 Jul. 2010; Revised: 7 Feb. 2011; Accepted: 10 Feb. 2011

315
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

presents the numerical results. Section 6 contains 3. GOVERNING EQUATIONS

concluding remarks.
The fluid flow generated in the coaxial mixer is
2. GEOMETRICAL SYSTEM governed by the momentum and mass
conservation principles. If we adopt the classical
The geometrical system was made up of a flat Eulerian description, this flow can be described
bottom cylindrical vessel of diameter D=0.3 m by the Navier-Stokes equations. The mass
and the liquid height in the vessel was H=D. The conservation is written as follows:
classical anchor of thickness e=1 mm is located
D 
in the cylindrical tank with height h1 =0.75   divV  0 (1)
D=0.225 m. It has two arms of diameter d=0.9 Dt
D=0.27 m, its width b=0.1 D=0.03 m and the
If the fluid flow is stationary and incompressible,
off-bottom clearance C=0.1 D=0.03 m. The
this equation can be written as follows:
geometry of this system resembles the one 
already studied by Nagata (1975). The anchor div V  0 (2)
blade is obtained by adding two blades to the
classical anchor with height h2 =0.5 D=0.15 m. In The momentum conservation equation for the
these conditions, the coordinate system origin is fluid flow is written as follows:

Downloaded by [109.161.202.118] at 17:55 24 February 2016

located on the bottom of the tank (Fig. 1). In

   div
 V
 V  V  pI 
turbulent flow, the fluid used in the CFD  
simulation is water with the density ρ and the t (3)
 div    R      r  2    V  
viscosity μ equal respectively to 1000 kg/m3 and   
2
1 mPa.s. The flow is fully turbulent, where the
Reynolds number Re and the Froude number Fr
are respectively 104 and 0.02. The rotation speed Three components of the velocity vector U, V and
N is equal to 60 rpm. The anchor material used in W are obtained by the projection following the
the CSD simulation is steel with the Young’s radial, tangential and axial directions respectively.
module E, the yield stress σ and the Poisson’s These dimensionless equations are written as
ratio υ equal respectively to 210 MPa, 215 MPa follows:
and 0.3.

⎡ ⎡ U 1 ∂V ⎤ 1 ∂ ⎡ ∂U ⎤ ⎤
⎢ -2 ⎢ 2 + r 2 ∂ ⎥ + r ∂r ⎢ re ∂r ⎥ ⎥
⎡ 2 ⎛ d ⎞ 1  ⎤
2 2
∂U ∂p 2 ⎛ d ⎞ 1 ⎢ ⎣ r ⎦ ⎣ ⎦ ⎥ V2
+ div ⎢VU − ⎜ ⎟ e gradU ⎥ = − + ⎜ ⎟ + + r + 2V (4)
∂t ⎢⎣  ⎝ D ⎠ Re ⎥⎦ ∂r  ⎝ D ⎠ Re ⎢ ∂ ⎡ ∂ ⎛ V ⎞ ⎤ ∂ ⎡ ∂W ⎤ ⎥ r
⎢ + 
⎢ e r ⎜ ⎟⎥ + 
⎢ e ⎥⎥
⎣ r ∂ ⎣ ∂r ⎝ r ⎠ ⎦ ∂z ⎣ ∂r ⎦ ⎦

⎡ ⎡ 1 ∂U ∂ ⎛ V ⎞ ⎤ 1 ∂ ⎡ ⎛ ∂U ⎞⎤ ⎤
⎢e ⎢ 2 + ⎜ ⎟⎥ + ⎢e ⎜ − V ⎟⎥ ⎥
⎡ 
 ⎤ ∂p 2 ⎛ d ⎞ 1 ⎢ ⎣ r ∂ ∂r ⎝ r ⎠ ⎦ r ∂r ⎣ ⎝ ∂
2 2
∂V  2⎛ d ⎞ 1 ⎠ ⎦ ⎥ UV
+ div ⎢VV − ⎜ ⎟ e gradV ⎥ = − + ⎜ ⎟ + − 2U (5)
∂t ⎢⎣  D
⎝ ⎠ Re ⎥⎦ r ∂  D
⎝ ⎠ Re ⎢ ∂ ⎡ ⎛ ∂V 2U ⎞ ⎤ ∂ ⎡ ∂W ⎤ ⎥ r
⎢ + e ⎜ + ⎟ + e ⎥
⎢⎣ r ∂ ⎢⎣ ⎝ r ∂ r ⎠ ⎥⎦ ∂z ⎢⎣ ∂ ⎥⎦ ⎥⎦

⎡1 ∂ ⎡ ∂U ⎤ ∂ ⎡ ∂V ⎤ ⎤
⎡  2

 ⎤ 2 ⎢ r ∂r ⎢ re ∂z ⎥ + r ∂ ⎢e ∂z ⎥ ⎥
∂W 2⎛ d ⎞ 1 ∂p 2 ⎛ d ⎞ 1 ⎢ ⎣ ⎦ ⎣ ⎦⎥ 1
+ div ⎢VW − ⎜ ⎟ e gradW ⎥ = − + ⎜ ⎟ + (6)
∂t ⎢⎣  ⎝ D ⎠ Re ⎥⎦ ∂z  ⎝ D ⎠ Re ⎢ ∂ ⎡ ∂W ⎤ ⎥ Fr
⎢ + ⎢e ⎥
⎣ ∂z ⎣ ∂z ⎥⎦ ⎦

The turbulent kinetic energy k and the turbulent dissipation rate ε are calculated using the k-ε turbulence
model (Hammemi et al., 2007; Bridgeman et al., 2009). The k-ε model equations are given in the following
form:

∂k ⎡  2 ⎛ d ⎞ 2 1 ⎛ t ⎞  ⎤ 2 ⎛ d ⎞ 2 1
+ div ⎢Vk − ⎜ ⎟ ⎜1 + ⎟ grad k ⎥ = ⎜ ⎟ G − (7)
∂t ⎢⎣  ⎝ D ⎠ Re ⎝  k ⎠ ⎥⎦  ⎝ D ⎠ Re

316
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

⎡ ⎞  ⎤  ⎡ ⎤
2 2
∂ 2 ⎛ d ⎞ 1 ⎛ t 2⎛ d ⎞ 1
+ div ⎢V  − ⎜ ⎟ ⎜1 + ⎟ grad ⎥ = ⎢Cc1 ⎜ ⎟ G − Cc 2 ⎥ (8)
∂t ⎢⎣  ⎝ D ⎠ Re ⎝   ⎠ ⎥⎦ k ⎢⎣  ⎝ D ⎠ Re ⎥⎦

The turbulent kinetic energy production is given in the following form:

⎡ ⎡⎛ ∂U ⎞ 2 ⎛ ∂V U ⎞ 2 ⎛ ∂W ⎞ 2 ⎤ ⎡ ∂V V ∂U ⎤ 2 ⎡ ∂W ∂V ⎤ 2 ⎡ ∂U ∂W ⎤ 2 ⎤
G = t ⎢ 2 ⎢ ⎜ ⎟ +⎜ + ⎟ +⎜ ⎟ ⎥+⎢ − + ⎥ +⎢ + ⎥ +⎢ + ⎥ (9)
⎢⎣ ⎢⎣⎝ ∂r ⎠ ⎝ r ∂ r ⎠ ⎝ ∂z ⎠ ⎥⎦ ⎣ ∂r r r ∂ ⎦ ⎣ r ∂ ∂z ⎦ ⎣ ∂z ∂r ⎥⎦ ⎦⎥

The viscosity equations are given in the following 4.1 Pre-processor

form:
The pre-processor tool allows the definition of the
 D
2
k 2 geometry, the mesh generation and specification
t  C    Re (10) of the appropriate boundary conditions. It defines
2d   also the mesh size for both CFD and CSD codes.
 e  t  1 (11) The mesh generation is performed by the
computational grid for the CFD code. This grid
must fit the contours of the tank and its internals,
4. COMPUTATIONAL SIMULATION
even if the components are geometrically
Downloaded by [109.161.202.118] at 17:55 24 February 2016

complex. For this reason, the computer aided

To predict numerically the hydromechanics of the
design (CAD) is used first to build the anchor
turbulent flow, we developed a numerical code
shape (Fig. 1). After that, the mesh of this anchor
simulating the behavior of an incompressible fluid
is created by the CSD code and consists of 1409
in a three-dimensional domain (Driss et al., 2010).
nodes and 1311 finite elements. Finally, the mesh
This code has three modules: the pre-processor,
of the fluid domain is automatically generated for
the processor and the post-processor. Data are
the three-dimensional simulations. It contained
interchanged between these three modules. In the
54000 nodes (Fig. 2). This choice of the mesh
processor, the coupling interface is used to allow
will be verified in Section 5.3. The average
the transfer of the forces and displacements
dimensional element size is equal to 0.005 m in
between the computational fluid dynamics (CFD)
the near-wall regions and we used a wall function.
code and the computational structural dynamics
This approach is that of Launder and Spalding
(CSD) code.

b
s
b
H
h1
b h2
C

a. Classical anchor b. Anchor blade

Fig. 1 Geometrical system.

317
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

(Kchaou et al., 2008). The spatial discretization is

obtained by the hybrid scheme (Baccar et al.,
2001; Kchaou et al., 2008; Patil and Tiwari,
2009). For the temporal discretization, the
implicit formulation is adopted. The integration in
the time is undertaken by using the directions
scheme presented by Douglas and Gunn (1964).
Fig. 3 presents the CFD code algorithm in the
turbulent flow.

Start

Parameters initialization
Fig. 2 Meshing.
Downloaded by [109.161.202.118] at 17:55 24 February 2016

(1974). This is due to the fact that the transport Wall law
equation for the eddy dissipation has a singularity
at the wall, where k is zero. Thus the equation for
ε must be treated in a different manner. The
k and ε resolution
number of element in the tip clearance is equal to
4.

4.2 Processor Test No

In the processor, we used the CFD code for the Yes

fluid flow and the CSD code to study the anchor
deformation. The implementation of the Viscosity actualization
computational simulation requires many
hypotheses such as the geometrical concept, the
choices of the meshing type, the physical models, Velocities filed resolution
the boundary conditions and the discretization
method.
Pressure-velocity coupling
4.2.1 CFD code
The CFD code allows the execution of the
program written in Fortran (Driss et al., 2010). No
Test
This code was used to model the fluid flow in a
stirred tank. The standard k-ε turbulent model was Yes
used with the hybrid discretization scheme for the
convective terms. The SIMPLE algorithm of Velocities filed correction
Patankar (1980) was adopted to couple the
pressure and the velocity. The governing
equations for the mean radial (Eq. (4)), tangential Km=Km+1
(Eq. (5)) and axial (Eq. (6)) velocities; the
transport equations for the turbulence kinetic Yes
energy (Eq. (7)) and its dissipation rate (Eq. (8)) No
are discretised using the finite-volume method Test
(Jagadeesh and Murali, 2010). These equations
are written in cylindrical coordinates (Montante
et al., 2001; Javed et al., 2006). This method
ensures the continuity between the nodes. It End
consists of dividing the fluid domain in
elementary volumes surrounding every node Fig. 3 Flow chart of CFD code.

318
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

4.2.2 CSD code 4.2.3 Coupling interface

The CSD code allows us to achieve three tasks: The coupling interface allows the exchange
reading the input file, calculating and visualizing between the CFD code and the CSD code. This
the results. After reading the input file already coupling is used in the partitioned method
created, we can proceed with the editing of the (Bobovnika et al., 2005; Sternel et al., 2008)
implementation plan. At this stage, we introduce when the fluid and structural parts are resolved
new parameters of our application. More separately. This method has been introduced by
particularly, we are interested in redefining the Park and Felippa (1983), and further investigated
analysis type, the mesh shape, the boundary by Wood (1990). Various experimental and
conditions and the result field of the problem. numerical simulations study the fluid behaviour,
After the program execution, we can visualize the when the movement of the structure is prescribed
various results like the Von Mises stress, the analytically (Ralph and Pedley, 1989; Natarajan
strain and the nodes displacement (Fig. 4). and Mokhtarzadeh-Dehghan, 2000). Other
researchers concentrated on the fluid part, while a
simple structural model for a rigid body was used
(Cossu and Morino, 2000). The coupling interface
Start allows the exchange of the forces at interface,
when the meshes in the fluid and structure fields
Downloaded by [109.161.202.118] at 17:55 24 February 2016

are different and hence difficult to place within

Construction of the model the same code, thus the utility of the partitioned
algorithm methodology. The coupling interface
organizes the transfer of information and makes
Definition of the properties an effective coupling between the fluid and the
structure fields. At the fluid-structure interface,
the forces from CFD code are applied as a load to
the CSD code. The displacements are calculated
Assembly of the geometry using the CSD code to characterize the anchor
arms deformation.

Specification of the results 4.3 Post-processor

The post-processor defines the third module of
our numerical code. This module presents the
Boundary conditions results obtained after the execution of our
numerical code. Several results can be obtained,
namely the velocity and pressure fields, the
Meshing turbulent characteristics and the structural
deformations.
5. SIMULATION RESULTS
Execution of the program
5.1 Flow patterns in r-θ planes
Figs. 5 and 6 show the velocity vector plot (U,V )
Visualization of the results in the r-θ planes located respectively at the
bottom and the top of anchors. They are
respectively defined by the axial positions z=0.16
No and z=1.5. In the first plane, the region swept by
Test the classical anchor, the velocity field is very
important (Fig. 5a). In the case of the anchor
Yes blade, the flow in the tank is slightly different
from that in a tank with a classical anchor
(Fig. 5b). This result is already expected because
End of the effect of the two horizontal blades. At the
top of the classical anchor, the values of the
Fig. 4 Flow chart of CSD code. velocity are relatively important near the wall.
Near the axis, the flow becomes centripetal and
the values of the velocity remain small (Fig. 6a).

319
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

0.7

a. Classical anchor b. Anchor blade

Fig. 5 Flow patterns induced in the r-θ plane defined by z =0.16.

Downloaded by [109.161.202.118] at 17:55 24 February 2016

0.7

a. Classical anchor b. Anchor blade

Fig. 6 Flow patterns induced in the r-θ plane defined by z =1.5.

In the case of the anchor blade, the flow in the

tank remains similar to that described in the case
of the classical anchor, but with a slightly higher
intensity (Fig. 6b). 0.4

5.2 Flow patterns in r-z planes

The influence of the anchor geometry was
analyzed in various r-z planes including the
revolution axis of the tank. Figs. 7, 8 and 9 show
the secondary flow (U,W ) in different r-z planes
defined respectively by angular coordinates
θ=35°, θ=45° and θ=55°. These planes are
located respectively at the upstream, the level and
the downstream of anchors. In the first plane, we
distinguish two circulation zones (Fig. 7). The
first is situated below the horizontal blade, a. Classical anchor b. Anchor blade
whereas the second is located at the top of the
Fig. 7 Flow patterns induced in the r-z plane defined
anchor. Also, two vertical jets are created; the
by θ =35°.
first is ascending near the tank wall and

320
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

0.4 0.4

a. Classical anchor b. Anchor blade a. Classical anchor b. Anchor blade

Fig. 8 Flow patterns induced in the r-z plane defined Fig. 9 Flow patterns induced in the r-z plane defined
by θ =45°. by θ =55°.
Downloaded by [109.161.202.118] at 17:55 24 February 2016

the other is descending near the axis. These jets dimensionless radial velocity component V(z) for
end on the interface of the horizontal blade different grid resolutions. This figure shows that
(Fig. 7a). In the case of the anchor blade, the flow the accuracy of the dimensionless radial velocity
in the tank remains roughly similar to that
described in the case of the classical anchor.
However, the downward jet ends at the meeting
point of the second horizontal blade (Fig. 7b).
Fig. 8a shows the correct non-sliding condition
which assumes that the velocity field is null. This
zone is exactly localized on the anchor. Also, we
observed a strong penetration of the fluid towards
the tank interior. This penetration ends when
meeting the horizontal blade. The flow between
the horizontal blades seems centripetal going to
the classical anchor axis. In the case of an anchor
blade, this penetration ends at the meeting point
of the second horizontal blade. In these
conditions, the flow between the horizontal blades
seems centripetal going to the anchor blade axis
(Fig. 8b). Downstream of the anchors, the
velocity field converges towards the axis of the
classical anchor (Fig. 9). The phenomenon of
brewing was detected along the arm. It was
developed into two opposite directions. The first
moves towards the tank wall while the second
moves towards the horizontal blade (Fig. 9a). In
the case of the anchor blade, we observe that this
penetration ends on the meeting of the second
horizontal blade. In these conditions, the flow
between the two horizontal blades becomes
centripetal moving towards the anchor blade axis
(Fig. 9b).

5.3 Grid resolution effects: mesh dependence

analysis Fig. 10 Axial profiles of the radial velocity for
different grid resolutions.
Fig. 10 shows the predicted axial profiles of the

321
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

predictions is not strongly influenced by the grid 5.4 Axial profiles of radial velocity
resolution. Thus, all predicted results are component
independent of the grid size for meshes with
Fig. 11 illustrates the predicted axial profiles of
54000 nodes and above. Therefore, a mesh with
the dimensionless radial velocity component U(z).
54000 nodes was selected for the simulations.
The dimensionless radial coordinates are r=0.5,
This result is verified by numerical results found
r=0.8 and r=0.9. The superimposed profiles
in the literature. For example, Murthy Shekhar
correspond to two anchor shapes. These profiles
and Jayanti (2003) used a 1145712 grid in the show a parabolic shape function which develops
axial, radial and circumferential directions to with the level of the horizontal blade of the
discretize the flow domain. In order to verify the classical anchor. The extrema of these functions
fineness of the grid, he added the calculations are respectively defined by the axial positions
with a 1145724 grid as well as with a z=0.33, z=0.37 and z=0.35. In the case of the
22811424 grid. Since the computational time anchor blade, two parabolic shape functions are
memory requirements for these time-dependent developed at the level of the two horizontal blades
calculations were very high in these simulations, of the anchor blade. The extrema of these
especially in the last case, the 1145712 grid functions are respectively defined by the axial
was used in all subsequent calculations. Deglon positions equal to z=0.36, z=0.39 and z=0.34 at
and Meyer (2006) used four grids to investigate the level of the inferior horizontal blade. But, at
Downloaded by [109.161.202.118] at 17:55 24 February 2016

the effects of grid resolution and discretization the level of the superior horizontal blade the
scheme on the CFD simulation of fluid flow in a extrema are respectively defined by the axial
stirred tank using the MRF impeller rotation positions z=0.99, z=1.08 and z=0.92. Roughly, it
model and the standard k- turbulence model. is noted that the radial effect of the anchor blade
is more important than that of the classical
anchor. This phenomenon is especially localized
at the level of the second blade.

Classical anchor Anchor blade

a. Radial coordinate r =0.5 b. Radial coordinate r =0.8 c. Radial coordinate r =0.9

Fig. 11 Axial profiles of the radial velocity.

322
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

0.07

0.06

0.05

0.04

0.03

0.02

0.01

a. Classical anchor b. Anchor blade

Fig. 12 Turbulent kinetic energy k in the r-θ plane defined by z =0.16.

0.16
Downloaded by [109.161.202.118] at 17:55 24 February 2016

0.14

0.12

0.1

0.08

0.06

0.04

0.02

a. Classical anchor b. Anchor blade

Fig. 13 Turbulent kinetic energy k in the r-θ plane defined by z =1.5.

5.5 Turbulent kinetic energy in r-θ planes 5.6 Turbulent kinetic energy in r-z plane
Figs. 12 and 13 present respectively the evolution Fig. 14 presents the distribution of the turbulent
of the distribution of the turbulent kinetic energy kinetic energy k in the r-z plane located upstream
k in the horizontal planes located in the bottom of the anchors. The angular position is θ=30°.
and in the top of the anchors. These planes are This figure shows an important production of the
respectively defined by the axial positions z=0.16 turbulent kinetic energy due to the agitation of the
and z=1.5. The first plane shows that the area fluid behind the blade. In these conditions, the
defined by the maximum values is located near maximum values region of the turbulent kinetic
the tank wall. Away from this area, the turbulent energy is located near the classical anchor arm.
kinetic energy becomes very weak (Fig. 12a). In Away from this domain, it decreases quickly.
the case of the anchor blade, the maximum values This region has a length z=1.1 (Fig. 14a). For the
of the turbulent kinetic energy are located near the anchor blade case, the distribution of the turbulent
tank wall. However, the turbulent kinetic energy kinetic energy remains roughly similar to that
quickly decreases (Fig. 12b). In the second plane, described in the case of the classical anchor.
it is noted that the region deformed by the However, it appears that the length of the
maximal values of the turbulent kinetic energy is maximum values of the turbulent kinetic energy
located near the walls (Fig. 13). Also, it is becomes more important, it is equal to z=1.3
observed that the maximum values of the (Fig. 14b).
turbulent kinetic energy are reached in the case of
the anchor blade (Fig. 13b).

323
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

5.7 Dissipation rate of turbulent kinetic

energy in r-θ planes
0.2 Figs. 15 and 16 show the distribution of the
dissipation rate of the turbulent kinetic energy ε in
various horizontal planes. These planes
0.15 correspond to various axial positions defined
respectively by z=0.16 and z=1.5. According to
these results, the maximum values are located in
0.1 the wake developed at the end of the anchor arm.
In Fig. 15a, this wake is defined by an angular
sector α=70°. For the anchor blade case, this
0.05
sector increases and becomes equal to α=80°
(Fig. 15b). In the second plane, a progression of
the formation of this wake is observed (Fig. 16).
a. Classical anchor b. Anchor blade In these conditions, the angular sector is α=80°
with a classical anchor (Fig. 16a). With the
Fig. 14 Turbulent kinetic energy k in the r-z plane anchor blade case, this sector increases and
defined by θ =30°. becomes α=90° (Fig. 16b).
Downloaded by [109.161.202.118] at 17:55 24 February 2016

0.07

0.06

0.05

0.04

0.03

0.02

0.01

a. Classical anchor b. Anchor blade

Fig. 15 Dissipation rate of the turbulent kinetic energy ε in the r-θ plane defined by z =0.16.

0.14

0.12

0.1

0.08

0.06

0.04

0.02

a. Classical anchor b. Anchor blade

Fig. 16 Dissipation rate of the turbulent kinetic energy ε in the r-θ plane defined by z =0.16.

324
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

values are close to experimental values found in

0.35 the literature. In fact, Pollard and Kantyka (1969)
as well as Nagata (1975) found values between
0.3 0.5 and 0.6 for the classical anchor. In the case of
the anchor blades, Nagata (1975) gave a value of
0.25 the power number NP =0.7. The comparison of the
computed power number with the experimental
0.2
results shows good agreement. Fig. 18 presents
0.15
the radial profiles of the dimensionless angular
velocity component V(r). The angular coordinate
0.1 is θ=90°. In this figure, the numerical result
found by Murthy Shekhar and Jayanti (2003) is
0.05 superposed with an average difference of 8%. The
good agreement between Murthy Shekhar and
Jayanti (2003) and the numerical results confirms
a. Classical anchor b. Anchor blade
the validity of the analysis method, although for a
Fig. 17 Dissipation rate of the turbulent kinetic energy final validation an experimental analysis would be
ε in the r-z plane defined by θ = 30°. required, for example using particle image
Downloaded by [109.161.202.118] at 17:55 24 February 2016

velocimetry (PIV).

5.8 Dissipation rate of turbulent kinetic

energy in r-z plane
Fig. 17 shows the distribution of the dissipation
rate of the turbulent kinetic energy ε in the r-z
plane defined by the angular position θ=30°. This
plane is located upstream of the anchors. With the
classical anchor, the maximum values region of
the dissipation rate of the turbulent kinetic energy
is situated near the walls. It decreases quickly in
the tank interior. This region has a length z=0.87
(Fig. 17a). In the case of the anchor blade, the
dissipation rate of the turbulent kinetic energy
remains roughly similar to that described in the
classical anchor case. However, it is observed that
the length of the maximal values region of the
turbulent kinetic energy is equal to z=1.1 Fig. 18 Radial profiles of the angular velocity.
(Fig. 17b).
5.10 Deformed shape of impeller
5.9 Comparison with experimental results
Fig. 19 presents the results of the CSD code. In
To compare between several mechanical this study, we are particularly interested in the
agitators, it is important to calculate the static behavior of the anchor and the evolution of
consumption power P to choose the installation the displacement field of the arms. We would like
power range. In turbulent flow, we must calculate to observe the nodal displacement of the anchor
P via the integration of the dissipation rate of the arm to allow us to mesh the fluid domain. Under
turbulent kinetic energy (Driss et al., 2010). Also, these conditions, it is supposed that the
we can characterize the consumption power, deformation of the anchor axis due to torsion is
through the calculation of the power number NP negligible. These results show that the biggest
defined as follows: variation in displacement is at the top of the
P classical anchor arm. It is fluctuating around
NP  (12) 0.0254 m. On the other hand, the variation in
 N3 d 5 displacement at the top of the anchor blade is
equal to 0.00169 m. Hence, the deformation of
According to our computational results, it is noted the anchor blade is less important than that of the
that the power number NP =0.51 for the classical classical anchor. This is due to the addition of
anchor and NP =0.64 for the anchor blades. These

325
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)
Downloaded by [109.161.202.118] at 17:55 24 February 2016

a. Classical anchor b. Anchor blade

Fig. 19 Displacement field.

the horizontal blade located in the top of the arm. To solve this problem, we suggest using an
anchor. This blade might eliminate the vibrations anchor blade. Therefore, we can avoid the
that have fatal effects during the agitation. For deformation and resulting vibrations in the
this reason, we demonstrated that it is very mechanical agitation system. The comparison of
important to study also the deformed shape of the the power number and the velocity profile
impeller to avoid all structure problems. between experimental and numerical results
shows good agreement. This proves the validity
6. CONCLUSIONS of the numerical method. Experimentally, we
suggest characterizing the hydrodynamic structure
To investigate the performance of the anchor, a of this agitator by the particle image velocimetry
numerical code is implemented for the (PIV).
hydromechanical characterization of the turbulent
flow in a vessel equipped with a classical anchor
ACKNOWLEDGEMENTS
and an anchor blade. This approach appeals to a
computational fluid dynamics (CFD) code, a
The authors would like to thank Mr. A. Dammak
computational structure dynamics (CSD) code
from the Engineering School of Sfax for his help
and a coupling interface. The numerical results
in English writing.
obtained allow us to visualize the velocity field,
the turbulent kinetic energy, the dissipation rate of
the turbulent kinetic energy and the structural NOMENCLATURE
deformation. These various results show that the
use of the classical anchor in turbulent flow b width of the two blades
engenders an important deformation of the anchor D internal diameter of the vessel tank, m

326
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

d impeller diameter, m ε dissipation rate of the turbulent kinetic

Fr Froude number, dimenionless, energy, dimensionless,
 2 N 
2
d *
Fr  
g 2 N 3 R 2
G turbulent kinetic energy production, θ angular coordinate, rad
dimensionless υt turbulent viscosity, dimensionless,
g gravity acceleration, m2.s-1 t
h turbine position, m t 

H vessel tank height, m
k turbulent kinetic energy, dimensionless, σk constant in the standard k-ε model
 stress tensor of the fluid
k*
k r*, z*, dimensional variables
 2 N R 
2
t*, U*,
N velocity of the anchor impeller, rad.s-1 V*, W*,
NP power number, dimensionless, k*, ε*,
P p*
NP 
 N3 d 5
Downloaded by [109.161.202.118] at 17:55 24 February 2016

P power, W REFERENCES
p pressure, dimensionless,
p* 1. Abid MS, Xuereb C, Bertrand J (1992).
p Hydrodynamics in vessels stirred with
 2 N R 2 anchors and gate agitators: necessity of a 3-D
Re Reynolds number, dimensionless, modeling. T. Instn. Chem. Eng. 70:377–384.
Nd2 2. Akiti O, Yeboah A, Bai G, Armenante P
Re  (2005). Hydrodynamic effects on mixing and
 competitive reactions in laboratory reactors.
R internal ray of the vessel tank, m Chem Eng Sci. 60:2341–2354.
r radial coordinate, dimensionless, 3. Baccar M, Abid MS (1999). Caractérisation
r* de l'écoulement turbulent et du transfert
r thermique générés par des mobiles ancre et
R
barrière dans une cuve agitée. Int. J. Therm.
s shaft diameter, m
Sci. 38:892–903.
t time, s, t  2 N t * 4. Baccar M, Mseddi M, Abid MS (2001).
U radial velocity components, Contribution numérique à l’étude
U* hydrodynamique et thermique des
dimensionless, U  écoulements turbulents induits par une turbine
2 N R
radiale en cuve agitée. Int. J. Therm. Sci.
V angular velocity components,
40:753–772.
V* 5. Bobovnika G, Mole N, Kutina J, Stokb B,
dimensionless, V 
2 N R Bajsic I (2005). Coupled finite-volume/finite-
W axial velocity components, element modelling of the straight-tube
W* Coriolis flowmeter. Journal of Fluids and
dimensionless, W  Structures 20:785–800.
2 N R 6. Bridgeman J, Jefferson B, Parsons A (2009).
z axial coordinate, dimensionless, Computational fluid dynamics modeling of
z* floccutation in water treatement: a review.
z Engineering Applications of Computational
 R
u displacement vector Fluid Mechanics 3(2):220–241.
 7. Cossu C, Morino L (2000). On the instability
V velocity vector of the fluid
of a spring-mounted circular cylinder in a
I identity tensor viscous flow at low Reynolds numbers.
μ fluid viscosity, Pa.s Journal of Fluids and Structures 14:183–196.
ρ fluid density, kg.m3 8. Deglon DA, Meyer CJ (2006). CFD
modelling of stirred tanks: Numerical

327
 Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 3 (2011)

considerations. Minerals Engineering International Journal for Numerical Methods

19:1059–1068. in Fluids 32:711–724.
9. Douglas J, Gunn JE (1964). A general 21. Park KC, Felippa C (1983). A Partitioned
formulation of alternating direction implicit analysis of coupled systems. Computational
methods. Num. Math. 6:428–453. Methods for Transient Analysis in
10. Driss Z, Bouzgarrou G, Choutorou W, Computational Methods in Mechanics 1:157–
Kchaou H, Abid MS (2010). Computational 218.
studies of the pitched blade turbine design 22. Patankar SV (1980). Numerical Heat
effect on the stirred tank flow characteristics. Transfer and Fluid Flow (Series in
European Journal of Mechanics—B/Fluids computational methods in mechanics and
29:236–245. thermal sciences). McGraw Hill, New York.
11. Hammemi M, Mseddi M, Baccar M (2007). 23. Patil PP, Tiwari S (2009). Computation of
Numerical study of coupled heat and mass flow past complex geometries using MAC
transfer in a trapezoidal cavity. Engineering algorithm on body-fitted coordinates.
Applications of Computational Fluid Engineering Applications of Computational
Mechanics 3:216–226. Fluid Mechanics 1:15–27.
12. Jadadeesh P, Murali K (2010). RANS 24. Pedrosa SMCP, Nunhez JR (2000). The
predictions of free surface effects on behavior of stirred vessels with anchor type
Downloaded by [109.161.202.118] at 17:55 24 February 2016

axisymetric underwater Body. Engineering impellers. Computers & Chemical

Applications of Computational Fluid Engineering 24:1745–1751.
Mechanics 4(2):301–313. 25. Pollard J, Kantyka TA (1969). Heat transfer
13. Javed KH, Mahmud T, Zhu JM (2006). to agitated non-Newtonian fluids. T. Instn.
Numerical simulation of turbulent batch Chem. Eng. 47:21–27.
mixing in a vessel agitated by a Rushton 26. Ralph ME, Pedley TJ (1989). Viscous and
turbine. Chemical Engineering and inviscid flows in a channel with a moving
Processing 45:99–112. indentation. Journal of Fluid Mechanics
14. Kchaou H, Driss Z, Bouzgarrou G, 209:543–566.
Chtourou W, Abid MS (2008). Numerical 27. Savreux F, Jay P, Magnin A (2007).
investigation of internal turbulent flow Viscoplastic fluid mixing in a rotating tank.
generated by a flat-blade turbine and a Chemical Engineering Science 62:2290–
pitched-blade turbine in a vessel tank. 2301.
International Review of Mechanical 28. Sternel DC, Schäfer M, Heck M, Yigit S
Engineering 3:427–434. (2008). Efficiency and accuracy of fluid-
15. Kchaou H, Kaffel M, Chehadi F, Baccar M structure interaction simulations using an
(2008). Modélisation numérique du implicit partitioned approach. Comput Mech.
comportement thermique d'un disque de frein. 43:103–113.
Revue européenne de mécanique numérique 29. Triveni B, Vishwanadham B, Venkateshwar S
17:1057–1075. (2008). Studies on heat transfer to Newtonian
16. Launder BE, Spalding DB (1974). The and non-Newtonian fluids in agitated vessel.
Numerical Computation of Turbulent Flows. Heat Mass Transfer 44:1281–1288.
Comput. Methods Appl. Mech. Engrg. 3:269– 30. Wood WM (1990). Practical Time-Stepping
289. Schemes. Oxford University Press, New
17. Montante G, Lee KC, Brucato A, York.
Yianneskis M (2001). Numerical simulations
of the dependency of flow pattern on impeller
clearance in stirred vessels. Chem. Eng. Sci.
56:751–3770.
18. Murthy Shekhar S, Jayanti S (2003). Mixing
of power-law fluids using anchors: Metzner-
Otto concept revisited. AIChE Journal
49(1):30–40.
19. Nagata S (1975). Mixing Principles and
Applications. Halstead Press, New York.
20. Natarajan S, Mokhtarzadeh-Dehghan MR
(2000). Numerical prediction of a (potential)
soft acting peristaltic blood pump.

328