Materials Today: Proceedings xxx (xxxx) xxx

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

Computational fluid dynamics simulation of a stirred tank reactor

Gaurav Mittal a,⇑, Rafael Issao Kikugawa b
a
Department of Mechanical Engineering, Graphic Era Deemed to be University, Dehradun, Uttarakhand, India
b
Escola Politécnica da Universidade de São Paulo, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Stirred tank reactors are used routinely in chemical processing industries. Mixing and dispersion in such
Received 4 January 2021 reactors depends on the geometry of the reactor and impeller. In addition, CFD simulations are an impor-
Received in revised form 30 January 2021 tant tool for characterizing flow and turbulence. Towards this end, the flow field inside a stirred tank
Accepted 4 February 2021
reactor is simulated by means of ANSYS Fluent CFD package. Computational geometry and mesh are gen-
Available online xxxx
erated in ANSYS. Simulations are conducted using the Multiple Rotating Frame (MRF) technique in con-
junction with the k

turbulence model. MRF is a computationally efficient technique, however the
Keywords:
solution may depend on the size of the inner domain near the impeller. Therefore, it is important to val-
Stirred tank
Multiple rotating frame
idate the technique by comparison with the experimental data. In this work, the model is implemented in
CFD ANSYS Fluent and the simulated velocity profiles are assessed against the experimental profiles available
in the literature. The simulations do describe the overall features of the flow well, however some discrep-
ancies are noted. Suggestions for extending this work for proper evaluation of the MRF technique and var-
ious turbulence models are given. Once properly validated, the technique can be used for realistic
evaluation of the effect of reactor and impeller geometry on mixing and turbulence.
Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Confer-
ence on Technological Advancements in Materials Science and Manufacturing.

1. Introduction dynamic characteristics of mixing and dispersion, testing of new

reactor concepts and getting quantitative information for scale-
In chemical industries, agitating of fluids using impellers is up or scale-down of reactors [1–4].
often used to accelerate the mixing process by converting the There are several models for turbulence as well as impeller
energy of rotating impeller into hydrodynamic motion. The flow modeling approaches with relative merits and demerits in terms
developed inside the reactor vessel is strongly governed by the of computational expense and physically closer approximation to
geometrical parameters of the reactor and rotor. The design of stir- the actual problem. Impeller-baffle interaction has been modeled
red reactors requires detailed understanding of the velocity and by various steady state techniques including IBC (impeller bound-
turbulence field inside the reactor. The mean flow causing convec- ary condition) technique, Source-sink technique, Inner-outer tech-
tive transport of heat, mass and momentum is caused by the nique, MRF (Multiple rotating frame) technique, and Snapshot
impeller, whereas turbulence characteristics determine eddy diffu- technique [1,2]. The time dependent interaction of fluid and impel-
sion. Thus, the performance of a stirred reactor is strongly depen- ler has also been modeled through unsteady approaches such as
dent on the flow field and the turbulence caused by the rotor [1]. Sliding Mesh [3] and Moving-deforming grid approach [2]. Among
Since the flow induced by the impeller in a stirred reactor with these approaches, the MRF technique and the Sliding Mesh tech-
baffles is intricate, 3D and recirculating turbulent flow, the general nique have attracted greater attention due to their simplicity and
approach is either to develop empirical correlations for flow char- are also incorporated in commercial CFD packages.
acteristics or resort to computational fluid dynamic (CFD) simula- Use of MRF approach has been reported in a number of papers
tions when validated computational models are available. CFD [e.g. 5–10]. Aubin et al. [5] asserted that the stirred tank can be
modeling of stirred reactors is very beneficial for obtaining fluid assumed to be in a quasi-steady state due to the periodic nature
of interaction of impeller with baffles. Abdul Rasool et al. [6]
numerically studied the effect of impeller speed in a cylindrical
⇑ Corresponding author. tank using 3 blade radial impeller. In their work, MRF model was
E-mail address: gauravmittal@geu.ac.in (G. Mittal).

https://doi.org/10.1016/j.matpr.2021.02.102
2214-7853/Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Conference on Technological Advancements in Materials Science and
Manufacturing.

Please cite this article as: G. Mittal and R. Issao Kikugawa, Computational fluid dynamics simulation of a stirred tank reactor, Materials Today: Proceedings,
https://doi.org/10.1016/j.matpr.2021.02.102
G. Mittal and R. Issao Kikugawa Materials Today: Proceedings xxx (xxxx) xxx

used without adequate validation. The results however showed

zones of poor mixing in the reactor as well as increase in velocity
fluctuations with rotational speed. In an optimization work, the
optimum position of impellers was attained by using Evolutionary
Programming along with MRF based CFD approach employing MRF
approach [7]. Ameur et al. [8] computationally studied the effect of
blade diameter, curvature and numbers on power requirement in a
mixing tank. They highlighted that certain types of straight and
large blades lead to dead zones near the walls of the tank. In con-
trast, curved blades yielded better mixing and reduced power con-
sumption. Sharan et al. [9] computationally explored a Rushton
impeller for enhanced mixing and desired turbulence. MRF tech-
nique was used for modelling. Study was conducted employing
single and multiple impellers.
The ultimate objective of this work is to adequately assess the
MRF approach. Various researchers have used the MRF model
without considering the importance of the size of the inner zone
near the impeller. In principle, the size of the inner zone can affect
the simulated results because of non-validity of the quasi-steady
assumption at the interface of the inner and outer zones. This work
is an important step towards the ultimate objective of thorough
assessment of MRF approach. In this work, high fidelity experimen-
tal data from literature is selected and computational model of a
reactor with an axial flow impeller is developed in ANSYS Fluent
using the MRF approach. Furthermore, the simulated results for a
fixed inner zone size are compared with experimental data. In
future, the work will be extended for thorough assessment of the
MRF approach, including the sensitivity of results to the size of
the inner zone as well as baffle orientation. This work will eventu- Fig. 2. Model geometry [a] Tank [b] Baffle [c] Impeller shaft [d] Impeller [e]
ally provide guidelines for reliable CFD simulations. In the follow- Rotating zone.
ing, the computational details and comparison of the CFD results
with experimental data is presented sequentially.

2. Computational modelling

2.1. Geometry

The computational modelling is done by using ANSYS Fluent

CFD package. The geometry of the baffled stirred tank is taken from
Ranade et al. [4] and is shown in Figs. 1 and 2.
The tank is a cylinder of diameter, T = 30 cm, and height,
H = 30 cm. It has four baffles of T/10 width arranged at 900 interval
along the tank. The impeller is a 45o pitched turbine blade
‘pumping-down’ impeller with four blades. It is located at an axial
location of C = T/3 from the bottom of the reactor and has diameter
d = T/3. A surface mesh of the reactor and grid distribution in
selected vertical and horizontal planes are shown in Figs. 3 and
4. ANSYS mesh tool was used to create the volume mesh of the
model, composed by tetrahedrons. In the following, the models

Fig. 3. Mixing tank surface mesh.

and details of the discretization and solution procedure are

presented.

2.2. Computational specifications

RANS equations are solved in conjunction with the standard k-e

turbulence model. The approach of enhanced wall treatment is
Fig. 1. Details of the tank and impeller. used, which mitigates the problem of restriction on near-wall grid
2
G. Mittal and R. Issao Kikugawa Materials Today: Proceedings xxx (xxxx) xxx

Fig. 4. Mesh display in selected plans; [A] Vertical plane; [B] Horizontal impeller plane.

for complex flows and allows improved solution for coarse as well
as fine near-wall meshes. Impeller is modelled using the MRF
approach. MRF is a steady state approach which renders a situation
which is transient with respect to the fixed frame of reference, into
steady with respect to a reference frame that moves with the rotor.
The computational domain is broken into several fluid/solid
domains. The zones which include the rotors are solved by using
the equations for moving reference frame. In contrast, the equa-
tions for stationary frame are used in the non-moving zones. Effec-
tively, the grid remains static for computational purpose. In this
model the impeller stays at the same position, this means that
the position of the baffles relative to the blades does not change.
Thus, this method is not recommended when the interaction of
the impeller and baffles is significant. This is similar to freezing
the movement of the rotor at a particular spot and perceiving the
instantaneous flow field in that particular position.
As shown in Fig. 2, the computational domain is allocated into a
zone around the impeller and an outside zone. The inner zone
involving the impeller has 4.2 cm height and 7 cm as radius. A Fig. 5. Profiles of axial velocities for different levels of grid refinement, at z/R = 0.2.
rotating frame of reference is used for the inside zone around the
rotor, and a fixed frame for the flow outside the rotor zone.
The governing equations for momentum and turbulence are 3.2. Simulated flow field
discretized using the second-order upwind scheme and solved
using the SIMPLE method. Simulations are done for an rpm of The general features of the simulated flow field are presented in
212, which corresponds to a Reynolds number in the turbulent Fig. 6, which shows velocity vectors and contours of turbulence
region based on Re ¼ ND2 q=l. A no slip condition is given to all intensity.
surfaces, including the rotor, shaft, baffles and cylinder. The con- The intake flow to the rotor originates from the top and from
vergence is confirmed by monitoring residues. the vertical periphery. The rotor develops high speed flow under
the impeller in the downward direction. After impinging at the
bottom of the rank, the fluid moves along the bottom surface and
finally moves upward along the cylindrical surface. As pointed
out by Ranade et al. [4], there is a distinct cone like zone under
3. Results
the rotor where speed is small and flow occurs in the upward
direction. Intense turbulence is noted particularly in the region
3.1. Mesh independence
below the impeller.
In order to confirm mesh independence, simulations were con-
ducted using grids in the range of approximately 200,000 to 3.3. CFD simulations vs experiment
600,000. The differences between the simulations from the coars-
est and finest grids used were not enormous. As an example, pro- CFD simulations were assessed using the data from Ranade
files of axial velocity in radial direction for a selected plane for the et al. [4]. Ranade et al. [4] measured mean velocities using LDA
coarsest and finest grid using the MRF approach is shown in Fig. 5. for the pitched blade impeller simulated here and provided exten-
The horizontal planes in the reactor are defined based on Z/R ratio, sive experimental data. Simulated and experimental axial velocity
where Z axis is directed axially downward in the reactor with ori- curves for various Z/R ratios are shown in Fig. 7. A positive value of
gin at the centre of the rotor and R is the radius of the tank. Z/R Z/R corresponds to downstream side of impeller (i.e. towards the
ratio spans from
1.33 at the reactor top to 0.66 at the bottom. bottom) and a negative value corresponds to upstream.
Furthermore, all subsequent simulations were done with a grid For Z/R =
0.366, the flow is in the downward direction for r/
of approximately 600,000. R < 0.7 and in upward direction for r/R > 0.7. This is due to the cir-
3
G. Mittal and R. Issao Kikugawa Materials Today: Proceedings xxx (xxxx) xxx

Fig. 6. (a) Velocity vectors (b) turbulence intensity on a plane at 45° to baffles.

Fig. 7. Simulated and experimental axial velocity profile along radius, at z/R =
0.366, z/R = 0.2, z/R = 0.366, z/R = 0.53.

culation vortex set up by the downward pumping impeller. The cir- 1. As Z/R increases, the peak axial velocity in the downward direc-
culation can be clearly noted in Fig. 6. For Z/R =
0.366, simula- tion decreases monotonically for the experimental data,
tions match the trend of experimental data very well, however whereas it first increases and then decreases in simulations.
velocities are slightly underpredicted. In Fig. 7, the plots for Z/ Significant discrepancies in simulations are noted at Z/R of
R = 0.2, 0.366 and 0.533 represent velocity profiles progressively 0.2, and the agreement becomes very good as Z/R increases.
downstream of the impeller. Certain features can be noted in the 2. As Z/R increases, the radial location of the peak axial velocity (in
experimental and simulated data. These are as follows: the downward direction) moves outward. This trend is noted in
both experimental and simulated data.

4
G. Mittal and R. Issao Kikugawa Materials Today: Proceedings xxx (xxxx) xxx

3. The disagreement close to impeller (Z/R = 0.2) could be due to CRediT authorship contribution statement
the size of the inner computational zone, which was set to a
height of 4.2 cm and radius of 7 cm. The MRF model assumes Gaurav Mittal: Conceptualization, Methodology, Writing -
that the unsteady effects are negligible at the interface of the reviewing & editing. Rafael Issao Kikugawa: Investigation.
inner and outer zones. A different choice of the inner zone could
lead to better agreement with the experimental data and needs Declaration of Competing Interest
to be explored.
The authors declare that they have no known competing finan-
4. Concluding remarks
cial interests or personal relationships that could have appeared
to influence the work reported in this paper.
CFD simulations of stirred tank reactors are valuable for design
of reactor and impeller geometries that can give desired mixing
References
and dispersion and avoid dead zones. There are many techniques
for modelling of rotating impeller. In particular, the technique of [1] J.B. Joshi et al., CFD simulation of stirred tanks: Comparison of turbulence
Multiple Rotating Frames (MRF) offers the advantage of computa- models. Part I: Radial flow impellers, Can. J. Chem. Eng. 89 (1) (2011) 23–82.
tional efficacy. An important point to consider, however, is that the [2] J.B. Joshi et al., CFD simulation of stirred tanks: Comparison of turbulence
models (Part II: Axial flow impellers, multiple impellers and multiphase
size of the domain near the impeller can affect the solution. In dispersions), Can. J. Chem. Eng. 89 (4) (2011) 754–816.
order to understand and evaluate this aspect, as a first step, the [3] J. Aubin, D.F. Fletcher, C. Xuereb, Modeling turbulent flow in stirred tanks with
model of a reactor was made using ANSYS Fluent and CFD simula- CFD: the influence of the modeling approach, turbulence model and numerical
tions were done using k

turbulence model. The MRF or ‘frozen
scheme, Exp. Therm. fluid Sci. 28 (5) (2004) 431–445.
[4] V.V. Ranade, J.B. Joshi, Flow generated by pitched blade turbines I:
rotor’ approach was used. The reactor had a height as well as diam- measurements using laser Doppler anemometer, Chem. Eng. Commun. 81 (1)
eter of 30 cm, and employed 45o pitched turbine blades. (1989) 197–224.
[5] J. Aubin, S.M. Kresta, J. Bertrand, C. Xuereb, D.F. Fletcher, Alternate operating
The model was assessed using the experimental data. The methods for improving the performance of continuous stirred tank reactors,
results mimic the overall trend in the velocity profiles well. How- Chem. Eng. Res. Des. 84 (7) (2006) 569–582.
ever, some discrepancies are also noted. Most significant is the dis- [6] A.A.A.R. Rasool, S.S. Ahmad, F. Hamad, Effect of impeller rotational speed on
flow behavior in fully baffled mixing tank, Int. J. Adv. Res. (2017).
crepancy in the axial velocity profile near the impeller in the [7] J.A. Alfaro-Ayala, V. Ayala-Ramirez, A. Gallegos-Munoz, A.R. Uribe-Ramirez,
downstream direction. It is conjectured that the discrepancy could Optimal location of axial impellers in a stirred tank applying evolutionary
partially be due to the size of the inner zone near the impeller. programing and CFD, Chem. Eng. Res. Des. 100 (2015) 203–211.
[8] H. Ameur, Y. Kamla, D. Sahel, Optimization of the operating and design
There are no definite guidelines for the choice of the size of the
conditions to reduce the power consumption in a vessel stirred by a paddle
zone. The general understanding is that the zone interface should impeller, Period. Polytech. Mech. Eng. 62 (4) (2018) 312–319.
not be close to impeller or baffles. A proximity of the interface to [9] V. Sharan, K. Rohit, M. Ravishankar, D. Bhuvaneshwar, R. Harish, CFD
baffles or impeller can, in principle, increase unsteady effects simulation of turbulent flow behaviour in a mixing reactor with Rushton
impeller, J. Phys. Conf. Ser. 1716 (1) (2020) 12025.
which are not accounted for in the MRF model. [10] M. Jaszczur, A. Młynarczykowska, L. Demurtas, Effect of impeller design on
In order to properly comprehend the cause of discrepancy, fur- power characteristics and Newtonian fluids mixing efficiency in a
ther simulations should be conducted by varying the size of the mechanically agitated vessel at low Reynolds numbers, Energies 13 (3)
(2020) 640.
inner domain as well as the relative angle between the baffles
and rotor blades. Then only the adequacy of MRF approach and
various turbulence models for the problem at hand can be
addressed.