Energy Conversion and Management 144 (2017) 275–285

Contents lists available at ScienceDirect

Energy Conversion and Management

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

CFD investigation to quantify the effect of layered multiple miniature

blades on the performance of Savonius rotor
Sonu Sharma ⇑, Rajesh Kumar Sharma
Department of Mechanical Engineering, National Institute of Technology Hamirpur, HP 177005, India

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

Article history: Present study is focussed on improvement of the coefficient of performance (COP) of a Savonius rotor
Received 9 February 2017 using numerical simulation software. Quantification of the improvement is based on the comparison
Received in revised form 10 April 2017 of the coefficient of performance (COP) of a basic configuration constituting a conventional Savonius rotor
Accepted 17 April 2017
to that of a modified configuration developed by adding concentric multiple miniature blades inside the
Available online 28 April 2017
rotor blades of the basic configuration. Validation and grid convergence studies are carried out using k
e
and Shear Stress Transport (SST) turbulence models. Validation study suggested Shear Stress Transport
Keywords:
(SST) model as more accurate and better option in present study. Shear Stress Transport (SST) turbulence
VAWT
CFD
model is used in the numerical simulations of the modified configuration. Optimum level of grid refine-
Wind turbine ment is achieved through grid convergence study. Boundary layer mesh is created on the rotor blades, by
Savonius rotor estimating distance of first mesh node from the wall using desired values of yþ for both k
e and Shear
Turbulence Stress Transport (SST) turbulence models. An improvement in COP spanning between 8.1% and 11.34% is
achieved with the modified configuration.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction low aerodynamic efficiency as compare to Darrieus type wind tur-

bines [7,14]; but still preferable for many applications due to its
In recent decades, global concern over pollution, global warm- good starting characteristics [4,15–17]. Also, as it is a VAWT, it
ing and possible depletion of non-renewable sources of energy, can accept wind from any direction because of its omnidirectional
such as oil has led to exploration of renewable and sustainable characteristics and hence needs no yaw mechanism [7,8,11,17–19].
energy resources. Conventionally, wind energy is harnessed on Many theoretical and experimental studies have been performed
large scale using Horizontal Axis Wind Turbines (HAWTs) spread by various researchers in the past to improve the aerodynamic per-
over large area. But the application of Horizontal Axis Wind Tur- formance of the Savonius rotor [8]. Better performance can be
bines (HAWTs) is limited due to requirement of large wind veloc- achieved by using high aspect ratios [20]; low overlap ratios
ities, high initial cost and large area required for their installation [21]; end plates [22]; two bladed Savonius rotor rather than three
[1,2]. The use of unconventional methods of energy generation bladed [23]; deflector plate [12,24]; multiple quarter blades [25].
such as Savonius wind rotors in micro generation is growing, but Improvement in the COP is desirable, without compromising the
it is still less widespread. The problem is that it has a low Coeffi- advantages of Savonius rotor and increasing the complexity of
cient of Performance (COP) [3–7]; theoretically COP ffi 0:18 [6]. design [6].
Despite of low power generation capacity, Savonius rotors are A cost effective study can be performed using Computational
gaining popularity among the decentralized power generation Fluid Dynamics (CFD), saving material and manufacturing costs.
methods due to their simple assembly and high starting torque The flow through and around the rotor is turbulent. It is very diffi-
at small rotational speeds [3,8]; low operating costs [9]; easy cult to achieve accuracy in such highly turbulent and transient
installation, manufacturing and maintenance [4,9–12]. Savonius conditions, so the adopted methodology must be validated to
rotor is a brilliant design proposed by Finnish inventor S.J. Savonius ensure accuracy [6]. Also, turbulence being an important factor
[5,8,11,13]. Conventional Savonius rotors are drag type hydroki- in analysis, the prediction of input values of turbulence parameters
netic Vertical Axis Wind Turbines (VAWTs). Savonius rotor has to be used in the numerical procedure must be done carefully.
Commercial CFD software, Ansys CFX 13 is used to carry out the
numerical simulations. This choice was motivated by the availabil-
⇑ Corresponding author. ity of licenses for the use of software and the existing practice of its
E-mail address: [email protected] (S. Sharma). use in the research environment.

http://dx.doi.org/10.1016/j.enconman.2017.04.059
0196-8904/Ó 2017 Elsevier Ltd. All rights reserved.
276 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285

Nomenclature

CAD Computer Aided Design L characteristic length

CFD Computational Fluid Dynamics q degrees by which domain is rotated per timestep
COP Coefficient of Performance N rpm of the rotor
DNS Direct Numerical Simulation Pin input power
HAWTs Horizontal Axis Wind Turbines Pout output power
HPC High Performance Computing Re Reynolds number
LES Large Eddy Simulation R radius of the rotor
RANS Reynolds Averaged Navier Stoke’s Equations t length of each time step in seconds
SST Shear Stress Transport T torque output
TSR Tip Speed Ratio u0 root mean square of the velocity fluctuations
VAWTs Vertical Axis Wind Turbines uav g average velocity of air flow
VSM Viscous Sublayer Model ut friction velocity
V inlet velocity
Symbols y distance of first node from wall
l dynamic viscosity of fluid yþ non dimensional parameter
A swept area of rotor b turbulent viscosity ratio
C 0f local coefficient of friction e turbulent dissipation rate
degr total number of degrees rotated per second q density of fluid
I turbulence intensity sx wall shear stress
k turbulent kinetic energy x specific dissipation rate
l turbulent length scale x0 angular velocity

2. Objective Design (CAD) models of basic configuration and modified configu-

ration consist of two domains separated by contact interfaces, in
The purpose of present study is to quantify the effect of layered which one large stationary domain representing the surroundings,
multiple miniature blades on the performance of Savonius rotor. A completely contains the smaller rotating domain as shown in
modified configuration, thus obtained by adding layered multiple Fig. 3.
miniature blades to the basic configuration; neither alters the max-
imum space requirements nor the complexity of the basic 4. Governing equations
configuration.
The equations governing the fluid flow representing conserva-
tion laws can be summarised as below [26].
3. Computer Aided Design (CAD) models
Continuity Equation (Conservation of mass):

A virtual model of the Savonius rotor with same dimensions as @

two bladed single stage Savonius rotor of Saha et al. [23] is referred ðui þ u0i Þ ¼ 0 ð1Þ
@xi
to as the basic configuration (Fig. 1) and is used for validation
study. A modified configuration (Fig. 2) is prepared by addition where
of two layers, concentric with the existing blades in basic configu- ui = average velocity of air flow
ration. Each layer consists of two co-radial miniature blades sepa- u0i = velocity fluctuation due to turbulence effects
rated circumferentially by a small distance. Computer Aided x = direction of flow

Fig. 1. Basic configuration [23].

 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285 277

Fig. 2. Modified configuration with layered multiple miniature blades.

balance between accuracy and performance. RANS are formulated

by splitting the variables into fluctuating and mean components in
the Navier Stoke’s equations. More number of variables introduced
in the RANS due to this splitting results in requirement of more
equations to close the system of equations which are provided
by turbulence closure models [28]. Standard k
e, k
x and Shear
Stress Transport (SST) are two equation turbulence closure models
with their own advantages and disadvantages. Standard k
e is a
robust turbulence model, fairly accurate with low resource
requirement [29,30]; but inefficient in accurately modelling flows
with turbulent boundary layer separation [28]. Standard k
e
model provides equations for turbulent kinetic energy ðkÞ and tur-
bulent dissipation rate ðeÞ as presented below respectively [31,32]:

 
@ðqkÞ @ðqkui Þ @ lt @k
þ ¼ lþ þ Gk þ Gb
qe
Y m þ Sk
@t @xi @xj rk @xj
ð3Þ

 
@ðqeÞ @ðqeui Þ @ lt @ e e
þ ¼ lþ þ C 1e ðGk þ C 3e Gb Þ
@t @xi @xj re @xj k
e2

C 2e q þ Se ð4Þ
k

Fig. 3. Detailed CAD model.

where

Gk ¼ generation of turbulent kinetic energy

ð5Þ
due to the mean velocity gradients ¼ lt S2
Momentum Equations (Conservation of momentum):
S ¼ modulus of the mean rate of strain tensor
@ui @ui  l @ 2 ui
1 @p @ 0 0 Gb = generation of turbulent kinetic energy due to buoyancy
þ uj ¼
þ
uu ð2Þ
@t @xj q @xi q @xi @xj @xj i j
Y m ¼ contribution of the fluctuating dilatation incompressible
where
k
t = time turbulence to the overall dissipation rate ¼ 2qe ð6Þ
cRT
 = average pressure
p
q = density of air rk and re ¼ turbulent Prandtl numbers for k and e
l = dynamic viscosity of air Sk and Se ¼ user defined source terms

The equations can be solved using different approaches. Direct Values of Constants:
Numerical Simulation (DNS) and Large Eddy Simulation (LES) pro- C l ¼ 0:09
vide accurate results, but are very resource intensive and thus, C 1e ¼ 1:44
popular only in the high end research labs with high performance C 2e ¼ 1:44
computing (HPC) resources. Reynolds Averaged Navier Stoke’s rk ¼ 1:0
Equations (RANS) with the closure models [27], struck a perfect re ¼ 1:3
278 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285

It can be used with coarser meshes (with modest computational In present study, validation is performed using Standard k
e
resources) as near wall flow is modelled using empirical relations model and SST model available in Ansys CFX 13 and results are
based on the law of wall known as wall functions [2,28,31,33,34]. compared to identify the turbulence closure model that provides
Higher values of yþ can be desired as requirement of fine mesh is more accurate results.
not a necessity here. A superior variant of Standard k
e model
known as Realizable k
e model provides better performance in 5. Input parameters
flows involving flow separation, rotation and recirculation, without
an appreciable increase in resource requirement [29]. It is better Boundary details are shown in Fig. 3. A boundary condition of
than Standard k
e in resolving flow phenomena near the surfaces No-slip wall is imposed on rotor blades. The inputs include a veloc-
of interest [35]. The equations for Realizable k
e model are given ity boundary condition at the inlet, angular velocity for the rotating
by [32,36]: domain, interfaces at the contact faces of both domains, opening
with atmospheric pressure at all other faces of the stationary

 
@ðqkÞ @ðqkuj Þ @ lt @k domain and turbulence parameters. The input parameters can be
þ ¼ lþ þ Gk þ Gb
qe
Y m þ Sk
@t @xj @xj rk @xj estimated as discussed further.
Accuracy of the solution is considerably affected by the correct-
ð7Þ
ness of the estimated values of input turbulence parameters and
boundary layer parameters. Estimation of values of turbulence

 
@ðqeÞ @ðqeuj Þ @ lt @ e e2 parameters at the inlet of the stationary domain is carried out
þ ¼ lþ þ qC 1 Se
qC 2 pffiffiffiffiffi
@t @xj @xj re @xj k þ me using the empirical relations listed below [32]:
e u0
þ C 1 e C 3 e G b þ Se ð8Þ ¼ 0:16ðReÞð 8 Þ

1

k I ¼ Turbulence intensity ¼ ð14Þ

uav g
where L
  l ¼ Turbulent length scale ¼ 0:07 3 ð15Þ
g C 4l
C 1 ¼ max 0:43; ð9Þ
gþ5 k ¼ Turbulent kinetic energy ¼ 1:5ðuav g IÞ2 ð16Þ
1
k 2

g¼S
k
ð10Þ x ¼ Specific dissipation rate ¼ ð17Þ
e Cl l
3
Values of constants: k2
e ¼ Turbulent dissipation rate ¼ ð18Þ
C 1e ¼ 1:44 l
C 2 ¼ 1:9 qC l k2
rk ¼ 1:0 b ¼ Turbulent v iscosity ratio ¼ ð19Þ
le
re ¼ 1:2
where
Another turbulence model in the k
e family is u0 ¼ root mean square of the v elocity fluctuations
Renormalization-group (RNG) k
e model which provides accurate uav g ¼ mean flow v elocity
results in flows involving swirling effect behind the rotating blade L ¼ characteristic length
[37]; large flow separations [38]; and is more responsive to the C l ¼ 0:09
effects of rapid strain and streamline curvature than the Standard
k
e model [32,36]. The Renormalization-group (RNG) k
e model The values of turbulence parameters are calculated with the
equations are summarised below [32,36,39]: input values from Saha et al. [23]. The variation of turbulence

 parameters with the Reynolds number are tabulated in Table 1.
@ðqkÞ @ðqkui Þ @ @k
þ ¼ ak leff þ Gk þ Gb
qe
Y m þ Sk ð11Þ In present study, COP is used as a measure of the improvement
@t @xi @xj @xj in performance by comparing its value for the basic configuration

 and the modified configuration. The value of COP is directly
@ðqeÞ @ðqeui Þ @ @e e affected by the value of torque measured on the blades of the rotor.
þ ¼ ae leff þ C 1e ðGk þ C 3e Gb Þ
@t @xi @xj @xj k So, accuracy can be enhanced if quality of mesh near the rotor sur-
e2 faces is good enough to capture the smallest of the changes in flow

C 2e q
R e þ Se ð12Þ variables. This near wall mesh is known as boundary layer mesh.
k
Boundary layer mesh is generated automatically by the software,
where obeying the input mesh parameter values provided to it. The
ak and ae ¼ inverse
 effective Prandtl numbers for k and e boundary layer formed due to the no-slip condition applied on
C l qn3 1
gg e
2 the rotor surfaces, can be accurately handled by a quality boundary
Re ¼ ð13Þ
0

ð1 þ bg3 Þ k layer mesh. The values of mesh parameter are dependent on the
turbulence model chosen for the simulation. The values of yþ  1
Values of Constants: are desirable when using k
x based models such as SST turbu-
C 1e ¼ 1:42 lence model, which ensures the presence of first mesh node from
C 2e ¼ 1:68 the wall completely inside the viscous sublayer [28]. The genera-
tion of fine mesh near the wall lowers the yþ value, which is good
SST model is a combination of Standard k
x model and trans- for accuracy with Viscous Sublayer Model (VSM) but increases the
formed k
e model formulated by multiplying both models by a number of nodes and hence the computational expenses. Wall
blending function and then adding the equations. The blending functions are used with Standard k
e model for the near wall
function is one in the boundary layer region (near to the wall) acti- treatment. Larger values of yþ are recommended when using wall
vating the Standard k
x model and zero for the far field activat- functions for boundary layer region. But the use of larger values of
ing the transformed k
e model [32,40]. yþ compromises the grid independence [28].
 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285 279

Table 1
Variation of input turbulence parameters with Reynolds number.

Inlet velocity (V) (m/s) Re I b e k x

6 1.474E05 0.036 250.312 0.117 0.070 18.472
6.5 1.597E05 0.035 268.471 0.144 0.081 19.813
7.27 1.787E05 0.035 296.102 0.194 0.098 21.852
8.23 2.023E05 0.034 330.045 0.268 0.122 24.357
9.21 2.263E05 0.034 364.188 0.361 0.149 26.876
9.88 2.428E05 0.034 387.268 0.434 0.168 28.580
10.17 2.499E05 0.033 397.196 0.468 0.177 29.312

In this research, the distance of first node from the rotor surface is 2pN
x0 ¼ angular v elocity ðrad=sÞ ¼ ð26Þ
calculated with desired yþ ¼ 1 for SST model and yþ values of 30, 60
100, 200, 300 for Standard k
e model using following relations
[41]: N ¼ rpm of the rotor
qut y
yþ ¼ ð20Þ 1
l Pin ¼ input power ðWÞ ¼ qAV 3 ð27Þ
2
where
ut ¼ friction v elocity A ¼ swept area of rotor ðm2 Þ
yþ ¼ non dimensional parameter
y ¼ distance of first node from wall V ¼ inlet v elocity ðm=sÞ
q ¼ density of fluid Tip Speed Ratio (TSR) is calculated using Eq. (28) [2,39,43]
l ¼ v iscosity of fluid
x0 R

12 TSR ¼ ð28Þ
sx V
ut ¼ ð21Þ
q where
R ¼ radius of the rotor
where
Input timestep value for transient simulation is calculated using
sx ¼ wall shear stress
Eqs. (29)–(31).
sx can be calculated as follows [42]:
N  360
1 Total number of degrees rotated per second ðdegrÞ ¼
sx ¼ C 0f qV 2 ð22Þ 60
2
ð29Þ
where
C 0f ¼ Local coefficient of friction Let q be the degrees by which domain is rotated per timestep (in
present study 2).
V ¼ Inlet v elocity
q
Timestep in seconds ðtÞ ¼ ð30Þ
C 0f is calculated using following empirical formula [42]: degr
ðNumber of rotations  360  tÞ
C 0f ¼ ½2logðReÞ
0:65

2:3
ð23Þ Total time in seconds ¼ ð31Þ
q
The calculated values of mesh parameters are shown in Table 2.
Also, the values of first node distance from the wall for different
6. Results and comparison
desired yþ values are listed in Table 3. A section of the generated
mesh on the rotating domain of the basic configuration is shown
A validation study is performed by comparing the experimental
in Fig. 4.
results of Saha et al. [23] for two bladed single stage Savonius rotor
Performance calculations are carried out using Eq. (24).
and a numerical simulation of basic configuration with same
Pout dimensions, initial conditions and boundary conditions setup in
COP ¼ ð24Þ
Pin Ansys CFX 13 in present study. Separate validation studies are per-
formed employing Standard k
e model and SST model. A 3D sim-
where,
ulation approach yields better results as compared to 2D [44]. So, a
Pout ¼ output power ðWÞ ¼ T x0 ð25Þ 3D transient simulation approach is used to carry out all the sim-
ulations discussed in this research. A 2° rotation of the rotating
T ¼ torque output ðNmÞ domain is desired at every timestep of the transient simulation
to achieve accuracy. A smaller rotation angle at every timestep

Table 2
Calculated values of mesh parameters for q ¼ 1:1871 kg=m3 and l ¼ 0:00001811 Ns=m2 .
Inlet velocity (V) (m/s) Re C 0f sx (kg/ms2) ut (m/s)

6 1.474E05 5.4E
03 0.115 0.311

6.5 1.597E05 5.3E
03 0.133 0.334
7.27 1.787E05 5.2E
03 0.162 0.370
8.23 2.023E05 5.1E
03 0.203 0.413
9.21 2.263E05 4.9E
03 0.248 0.457
9.88 2.428E05 4.9E
03 0.282 0.487
10.17 2.499E05 4.8E
03 0.297 0.500
280 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285

Table 3
Calculated values for distance of first mesh node from wall ðyÞ in metres for different yþ values.

yþ ¼ 1 yþ ¼ 30 yþ ¼ 100 yþ ¼ 200 yþ ¼ 300

4.897E
05 1.469E
03 4.897E
03 9.794E
03 1.469E
02
4.557E
05 1.367E
03 4.557E
03 9.115E
03 1.367E
02
4.121E
05 1.236E
03 4.121E
03 8.243E
03 1.236E
02
3.686E
05 1.106E
03 3.686E
03 7.373E
03 1.106E
02
3.331E
05 0.999E
03 3.331E
03 6.663E
03 0.999E
02
3.127E
05 0.938E
03 3.127E
03 6.254E
03 0.938E
02
3.046E
05 0.914E
03 3.046E
03 6.093E
03 0.914E
02

Fig. 4. Boundary layer mesh for basic configuration rotating domain.

ensures small timestep duration, capturing the changes in flow 6.2. Grid convergence study
behavior more frequently leading to accuracy. Input values of sim-
ulation total time, duration of a timestep are calculated using Eqs. Optimal use of the computational resources is needed when
(29)–(31). A yþ study is also carried out to choose the best desired working with low resource infrastructure. The refinement level of
yþ value for the Standard k
e model validation study. Air at 25  C the grid determines the resources required for a given study. More
is the working fluid. number of element and nodes require more number of calcula-
tions. A grid converged solution is achieved, when no improvement
in the output of the simulation is observed with further increase in
6.1. yþ study the grid refinement. Thus, an optimal value of mesh refinement is
found for maximum achievable accuracy in a particular case. This
yþ value for a simulation study determines the fineness of the checks the over refinement of grid and hence ensures the optimal
mesh required to achieve the desired yþ value to obtain accurate use of the resources. In present study, a very fine meshing is used
results. yþ ¼ 1 is desired value for the SST model. But to find out for the boundary layer mesh to capture the near wall phenomena
most appropriate yþ value for the Standard k
e model, a compar- accurately. This resulted in large number of nodes and elements
ative study is performed. A separate mesh is generated to achieve on which the governing equations are to be solved. As a large num-
required refinement level for a set of four different yþ values of ber of 3D transient simulations were to be performed, it was a
30,100, 200 and 300 at V ¼ 6 m=s. The most desired value is short- necessity to ensure the optimal use of the available resources. As
listed by comparing the output torque values summarised in needed, a grid convergence study was carried out for two different
Table 4. The value of torque has been observed to be close to the grid sizes at V ¼ 6 m=s for both Standard k
e model and SST
experimental results of Saha et al. [23] was corresponding to model as presented in Table 5. In the case of Standard k
e model,
yþ ¼ 300. the coarse mesh with 517,942 nodes and 1,605,542 elements was
found to provide a slightly better solution than a fine mesh with
859,150 nodes and 2,624,829 elements, while in the case of SST
Table 4 model, coarse mesh with 1,226,120 nodes and 3,491,868 elements
Output torque for different yþ values with Standard k
e and fine mesh with 1,962,162 nodes and 5,332,718 elements gen-
turbulence model for V ¼ 6 m=s.
erated a similar output with a very small difference.
yþ Torque (Nm)

30 1.420E
02 6.3. Validation study

100 1.350E
02
200 1.280E
02
300 1.270E
02
Validation study was performed on the basic configuration for
velocities ranging between 6 m/s and 10.17 m/s. A comparison
 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285 281

Table 5 was also performed based on the results of validation study con-
Grid convergence study.
ducted with Standard k
e model and SST model. Initially for the
Inlet velocity (V) Mesh Torque (Nm) Standard Torque (Nm) lower velocities of 6 m/s and 6.5 m/s, the Standard k
e model
(m/s) type k
e SST generated more accurate output torque and COP values as com-
6 Coarse 1.130E
02 1.940E
02 pared to the SST model, but for higher velocities ranging between
6 Fine 1.270E
02 1.960E
02 7.27 and 10.17, the accuracy of the results diverged very steeply. A
better and stable trend of output results was generated with SST
model. The high degree of flow separation involved in the present
Table 6
Comparison of torque for basic configuration simulation results with experimental case rendered the Standard k
e model inaccurate. The formula-
results of Saha et al. [23]. tion of SST model proved to be robust in handling transient high
turbulence problems with highly complex flows involved in Turbo
Inlet velocity (V) Torque(Nm) Torque(Nm) Standard Torque(Nm)
(m/s) [23] k
e SST machinery simulations. Based on the comparison, SST model was
an obvious choice for simulations of the modified configuration,
6 0.093E
02 1.270E
02 0.019
6.5 0.142E
02 1.407E
02 0.022
which due to more number of blade surfaces increases the complex
7.27 0.219E
02 1.689E
02 0.028 flow phenomena. The results of the validation study for torque and
8.23 0.363E
02 2.160E
02 0.037 COP are documented in Tables 6and 7 respectively.
9.21 0.443E
02 3.115E
02 0.047
9.88 0.474E
02 3.831E
02 0.056
10.17 0.397E
02 4.151E
02 0.060
6.4. Effect of layered multiple miniature blades

The effect of the layered multiple miniature blades is quantified

Table 7
by comparing the output of the numerical simulation of modified
Comparison of COP for basic configuration simulation results with experimental
results of Saha et al. [23]. configuration and the basic configuration with identical initial
and boundary conditions, simulation methodology and turbulence
Inlet velocity (V) (m/s) COP [23] COP Standard k
e COP SST
models. The net positive torque generated on the blades of rotor is
6 0.078 0.120 0.183 the driving force that makes the rotor to rotate. This net positive
6.5 0.108 0.119 0.193
torque output is used to calculate the COP of the configuration
7.27 0.137 0.119 0.203
8.23 0.180 0.120 0.208 under the given conditions. The values of torque generated on
9.21 0.159 0.125 0.192 the rotor, COP and the percentage improvement are tabulated in
9.88 0.139 0.126 0.186 Table 8. The mesh generated for rotating domain of basic and mod-
10.17 0.108 0.126 0.184 ified configuration is shown in Fig. 5. The boundary layer mesh was
optimally generated to capture the flow characteristics in the

Table 8
Comparison of Torque and COP (SST) between basic configuration and modified configuration.

Inlet velocity Torque (Nm) basic Torque (Nm) modified COP basic COP modified % increase
(V) (m/s) configuration configuration configuration configuration in COP
6 0.019 0.020 0.183 0.198 8.141
6.5 0.022 0.024 0.193 0.211 9.362
7.27 0.028 0.031 0.203 0.220 8.443
8.23 0.037 0.040 0.208 0.226 8.921
9.21 0.047 0.053 0.192 0.213 11.345
9.88 0.056 0.062 0.186 0.204 9.918
10.17 0.060 0.065 0.184 0.201 8.915

Fig. 5. Mesh for basic and modified configuration rotating domain.

282 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285

Fig. 6. Velocity distribution in the rotating domain of basic configuration (above) and modified configuration (below) for V ¼ 8:23 m=s.

12
Percentage improvement in

0.07
Torque(Saha et. al 10
0.06
2008)
0.05 8
Torque(Nm)

Torque(Basic
COP

0.04
Configuration with k-ε 6
0.03 model)
0.02 Torque(Basic 4
Configuration with SST
0.01 model)
2
0 Torque(Modified
6 6.5 7.27 8.23 9.21 9.88 10.17 Configuration) 0
Inlet Velocity(m/s) 6 6.5 7.27 8.23 9.21 9.88 10.17
Inlet Velocity(m/s)
Fig. 7. Variation of torque with inlet velocity for Saha et al. [23], Basic Configuration
and Modified Configuration. Fig. 9. Percentage improvement in COP with inlet velocity.

5.00E-02
Basic Configuration Modified Configuration
0.25
4.00E-02
COP(Saha et. al 2008)
0.2
3.00E-02
0.15
Torque (Nm)

COP(Basic
COP

Configuration with k-ε

model) 2.00E-02
0.1
COP(Basic
Configuration with SST
0.05 model) 1.00E-02
COP(Modified
0 Configuration) 0.00E+00
6 6.5 7.27 8.23 9.21 9.88 10.17 0 120 240 360 480 600 720 840
Inlet Velocity(m/s)
-1.00E-02 Angle of Rotation (deg)
Fig. 8. Variation of COP with inlet velocity for Saha et al. [23], Basic Configuration
and Modified Configuration. Fig. 10. Variation of torque on the rotor with angle of rotation for V ¼ 6 m=s.
 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285 283

5.00E-02 Basic Configuration Modified Configuration 1.20E-01

Basic Configuration Modified Configuration

4.00E-02 1.00E-01

3.00E-02 8.00E-02

Torque (Nm)
Torque (Nm)

2.00E-02 6.00E-02

1.00E-02 4.00E-02

0.00E+00 2.00E-02
0 120 240 360 480 600 720 840
-1.00E-02 0.00E+00
0 120 240 360 480 600 720 840
-2.00E-02 Angle of Rotation (deg) -2.00E-02 Angle of Rotation (deg)

Fig. 11. Variation of torque on the rotor with angle of rotation for V ¼ 6:5 m=s. Fig. 14. Variation of torque on the rotor with angle of rotation for V ¼ 9:21 m=s.

1.40E-01
7.00E-02 Basic Configuration Modified Configuration
Basic Configuration Modified Configuration
1.20E-01
6.00E-02
1.00E-01
5.00E-02
8.00E-02
Torque (Nm)
4.00E-02
Torque (Nm)

3.00E-02 6.00E-02

2.00E-02 4.00E-02

1.00E-02 2.00E-02
0.00E+00
0 120 240 360 480 600 720 840 0.00E+00
-1.00E-02 0 120 240 360 480 600 720 840
-2.00E-02 Angle of Rotation (deg)
-2.00E-02 Angle of Rotation (deg)
Fig. 15. Variation of torque on the rotor with angle of rotation for V ¼ 9:88 m=s.
Fig. 12. Variation of torque on the rotor with angle of rotation for V ¼ 7:27 m=s.

modified configuration suggesting smooth mechanical operations

1.00E-01 and energy transfer to the generator. Velocity contours shown in
Basic Configuration Modified Configuration
Fig. 6 present a view of the velocity distribution in the rotating
8.00E-02 domain. As Savonius rotor is drag based turbine, the miniature
blades increase the surface area of the rotor, which utilizes the
6.00E-02
kinetic energy possessed by the fluid in more efficient manner.
The miniature blades being very small and simple in shape, as
Torque (Nm)

4.00E-02
compared to the complete size of the rotor, don’t add much to
the weight and complexity of the rotor, thus the improvement in
2.00E-02
the performance is worth the effort. The averaged torque value
0.00E+00
for the second complete rotation is deployed for comparison.
0 120 240 360 480 600 720 840 Fig. 7 shows a comparative plot of the variation of torque with
-2.00E-02

-4.00E-02 Angle of Rotation (deg) 1.40E-01

Basic Configuration Modified Configuration
Fig. 13. Variation of torque on the rotor with angle of rotation for V ¼ 8:23 m=s. 1.20E-01

1.00E-01
region near the blades, as the parameter of interest is the torque on
the surface of the blades. Improvement ranging between 8.1% and 8.00E-02
Torque (Nm)

11.34% is observed with the use of layered multiple miniature

6.00E-02
blades. The total time and timestep of the transient simulation
are calculated in order to achieve two complete rotations of the 4.00E-02
rotating domain. Each timestep corresponds to 2 degrees rotation,
which accounts for 180 timesteps for one complete revolution. The 2.00E-02
output generated in the second rotation was considered due to less
0.00E+00
variation in the torque values. The variation of torque with angle of 0 120 240 360 480 600 720 840
rotation made by the rotor blades with the axis parallel to the -2.00E-02 Angle of Rotation (deg)
direction of flow is plotted in Figs. 10–16 for different inlet veloc-
ities. Figs. 10–16 show a comparatively smooth variation for the Fig. 16. Variation of torque on the rotor with angle of rotation for V ¼ 10:17 m=s.
284 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285

0.25 to study the effect of layered multiple miniature blades on the per-
formance of Savonius rotor. A remarkable improvement is con-
Coefficient of Performance (COP)

firmed with the use of layered multiple miniature blades on the

0.2
performance of the basic rotor configuration of Savonius rotor.
The energy efficiency of a rotor is the measure of its ability to con-
0.15 COP (Saha et. al vert the available wind energy into mechanical energy and is rep-
2008) resented as COP. Maximum COP improvement was observed to be
COP (Basic 11.34% for V ¼ 9:21 m=s. Layered multiple miniature blades
0.1 Configuration) enhance the performance of Savonius rotor without appreciable
COP (Modified increase in the complexity of the design, weight and maximum
Configuration)
0.05 space requirements for the installation. Future studies can involve
studying the effect of more number of layers of miniature blades
on the performance of basic rotor configuration.
0
0.60 0.65 0.70 0.75
Tip Speed Ratio (TSR)
References
Fig. 17. Variation of COP with Tip Speed Ratio (TSR) for Saha et al. [23], Basic
Configuration and Modified Configuration. [1] Tabatabaeikia S, Ghazali NNBN, Chong WT, Shahizare B, Izadyar N,
Esmaeilzadeh A, et al. Computational and experimental optimization of the
exhaust air energy recovery wind turbine generator. Energy Convers Manage
the inlet velocity of the fluid for experimental data from Saha et al. 2016;126:862–74. http://dx.doi.org/10.1016/j.enconman.2016.08.039.
[2] Sarma NK, Biswas A, Misra RD. Experimental and computational evaluation of
[23], basic configuration validation study with SST model and Savonius hydrokinetic turbine for low velocity condition with comparison to
modified configuration with SST model. The plot shows an over Savonius wind turbine at the same input power. Energy Convers Manage
prediction of torque values in the basic configuration validation 2014;83:88–98. http://dx.doi.org/10.1016/j.enconman.2014.03.070.
[3] Reupke P, Probert SD. Slatted-blade Savonius wind-rotors. Appl Energy
study as compared to the experimental values of Saha et al. [23], 1991;40:65–75. http://dx.doi.org/10.1016/0306-2619(91)90051-X.
but the similar trends of both the curves indicate the successful [4] Menet JL. A double-step Savonius rotor for local production of electricity: a
prediction of the flow behaviour and characteristics. The plotted design study. Renew Energy 2004;29:1843–62. http://dx.doi.org/10.1016/j.
renene.2004.02.011.
curve for the modified configuration shows an improvement in
[5] Bhuyan S, Biswas A. Investigations on self-starting and performance
the output torque for each inlet velocity case against the basic con- characteristics of simple H and hybrid H-Savonius vertical axis wind
figuration, suggesting a positive effect of layered multiple minia- rotors. Energy Convers Manage 2014;87:859–67. http://dx.doi.org/10.1016/j.
enconman.2014.07.056.
ture blades on the performance of Savonius rotor. As the same
[6] Mohamed MH, Janiga G, Pap E, Thévenin D. Optimal blade shape of a modified
numerical methodology is adopted for simulation of the basic con- Savonius turbine using an obstacle shielding the returning blade. Energy
figuration validation study and the modified configuration study, Convers Manage 2011;52:236–42. http://dx.doi.org/10.1016/j.enconman.
both studies have the same advantages and disadvantages in terms 2010.06.070.
[7] Kamoji MA, Kedare SB, Prabhu SV. Experimental investigations on single stage
of numerical procedures. Similar comparative plot for the COP is modified Savonius rotor. Appl Energy 2009;86:1064–73. http://dx.doi.org/
shown in Fig. 8. The COP plot shows a considerable improvement 10.1016/j.apenergy.2008.09.019.
in the performance of the modified configuration with a maximum [8] Altan BD, Atilgan M. The use of a curtain design to increase the performance
level of a Savonius wind rotors. Renew Energy 2010;35:821–9. http://dx.doi.
COP of 0.226 with an improvement of 8.92% achieved for org/10.1016/j.renene.2009.08.025.
V ¼ 8:23 m=s, as compared to the basic configuration. Fig. 9 shows [9] Yassi Y, Hashemloo S. Improvement of the efficiency of the Agnew micro hydro
a dedicated plot of percentage improvement in COP with the inlet turbine at part loads due to installing guide vanes mechanism. Energy Convers
Manage 2010;51:1970–5. http://dx.doi.org/10.1016/j.enconman.2010.02.029.
velocity showing the maximum improvement of 11.34% for [10] Roy S, Ducoin A. Unsteady analysis on the instantaneous forces and moment
V ¼ 9:21 m=s. The improvement in COP with respect to the tip arms acting on a novel Savonius-style wind turbine. Energy Convers Manage
speed ratio for the modified configuration is shown in Fig. 17. 2016;121:281–96. http://dx.doi.org/10.1016/j.enconman.2016.05.044.
[11] Goodarzi M, Keimanesh R. Numerical analysis on overall performance of
Savonius turbines adjacent to a natural draft cooling tower. Energy Convers
7. Conclusion Manage 2015;99:41–9. http://dx.doi.org/10.1016/j.enconman.2015.04.027.
[12] Golecha K, Eldho TI, Prabhu SV. Influence of the deflector plate on the
performance of modified Savonius water turbine. Appl Energy
A 3D transient CFD simulation based study is performed using 2011;88:3207–17. http://dx.doi.org/10.1016/j.apenergy.2011.03.025.
Ansys CFX 13. A numerical validation study is performed by com- [13] Gupta R, Biswas A, Sharma KK. Comparative study of a three-bucket
Savonius rotor with a combined three-bucket Savonius-three-bladed
paring the simulation results with the experimental results of Saha Darrieus rotor. Renew Energy 2008;33:1974–81. http://dx.doi.org/10.1016/j.
et al. [23]. Standard k
e turbulence model and SST turbulence renene. 2007.12.008.
model are compared by carrying out validation studies for both [14] Saha UK, Rajkumar MJ. On the performance analysis of Savonius rotor with
twisted blades. Renew Energy 2006;31:1776–88. http://dx.doi.org/10.1016/j.
the models. Based on these validation studies, SST model being renene.2005.08.030.
more accurate in present case was chosen for the simulation of [15] Zhou T, Rempfer D. Numerical study of detailed flow field and performance of
modified configuration. Input parameters for simulations were cal- Savonius wind turbines. Renew Energy 2013;51:373–81. http://dx.doi.org/
10.1016/j.renene.2012.09.046.
culated with data from the Saha et al. [23] for all the simulations. [16] Kamoji MA, Kedare SB, Prabhu SV. Performance tests on helical Savonius
All initial conditions, boundary conditions and numerical method- rotors. Renew Energy 2009;34:521–9. http://dx.doi.org/10.1016/j.renene.
ology were kept same for each particular velocity case in all the 2008.06.002.
[17] Zheng M, Li Y, Tian Y, Hu J, Zhao Y, Yu L. Effect of blade installation angle on
studies. A validated methodology is an inexpensive way of explor-
power efficiency of resistance type VAWT by CFD study. Int J Energy Environ
ing the performance characteristics of new designs, within the lim- Eng 2015;6. http://dx.doi.org/10.1007/s40095-014-0142-0.
itations of the numerical techniques. A grid convergence study was [18] Alaimo A, Lo Iacono F, Navarra G, Pipitone G. Numerical and experimental
comparison between two different blade configurations of a wind generator.
also carried out to find the optimal mesh quality to reach an
Compos Struct 2016;136:526–37. http://dx.doi.org/10.1016/j.compstruct.
achievable accuracy, thus suggesting the optimized usage of the 2015.10.04.
available resources. The accuracy of Standard k
e turbulence [19] Nahas MN, Mohamad AS, Akyurt M, EL-Kalay AK. Wind energy: an engineering
model was also compared for different yþ values suggesting the survey. Energy Sources 1987;9:137–48. http://dx.doi.org/10.1080/
00908318708908691.
use of yþ ¼ 300 in present validation study for the model. A [20] Alexander AJ, Holownia BP. Wind tunnel tests on a Savonius rotor. J Ind
numerical study on the modified configuration was also performed Aerodyn 1978;3:343–51. http://dx.doi.org/10.1016/0167-6105(78)90037-5.
 S. Sharma, R.K. Sharma / Energy Conversion and Management 144 (2017) 275–285 285

[21] Fujisawa N. On the torque mechanism of Savonius rotors. J Wind Eng Ind [33] Larin P, Paraschivoiu M, Aygun C. CFD based synergistic analysis of wind
Aerodyn 1992;40:277–92. http://dx.doi.org/10.1016/0167-6105(92)90380-S. turbines for roof mounted integration. J Wind Eng Ind Aerodyn
[22] Fleming PD, Probert SD. A proposed, three-sail, Savonius-type wind-rotor. Appl 2016;156:1–13. http://dx.doi.org/10.1016/j.jweia.2016.06.007.
Energy 1982;12:327–31. http://dx.doi.org/10.1016/0306-2619(82)90068-X. [34] Hellström JGI, Marjavaara BD, Lundström TS. Parallel CFD simulations of an
[23] Saha UK, Thotla S, Maity D. Optimum design configuration of Savonius rotor original and redesigned hydraulic turbine draft tube. Adv Eng Softw
through wind tunnel experiments. J Wind Eng Ind Aerodyn 2008;96:1359–75. 2007;38:338–44. http://dx.doi.org/10.1016/j.advengsoft.2006.08.013.
http://dx.doi.org/10.1016/j.jweia.2008.03.005. [35] Joo S, Choi H, Lee J. Aerodynamic characteristics of two-bladed H-Darrieus at
[24] Ogawa T, Yoshida H, Yokota Y. Development of rotational speed control various solidities and rotating speeds. Energy 2015;90:439–51. http://dx.doi.
systems for a Savonius-type wind turbine. J Fluids Eng 1989;111:53. http://dx. org/10.1016/j.energy.2015.07.051.
doi.org/10.1115/1.3243598. [36] Roy S, Saha UK. Computational study to assess the influence of overlap ratio on
[25] Sharma S, Sharma RK. Performance improvement of Savonius rotor using static torque characteristics of a vertical axis wind turbine. Procedia Eng
multiple quarter blades – a CFD investigation. Energy Convers Manage 2013;51:694–702. http://dx.doi.org/10.1016/j.proeng.2013.01.099.
2016;127. http://dx.doi.org/10.1016/j.enconman.2016.08.087. [37] Lee JH, Lee YT, Lim HC. Effect of twist angle on the performance of Savonius
[26] Akwa JV, Alves Da Silva Júnior G, Petry AP. Discussion on the verification of the wind turbine. Renew Energy 2016;89:231–44. http://dx.doi.org/10.1016/j.
overlap ratio influence on performance coefficients of a Savonius wind rotor renene.2015.12.012.
using computational fluid dynamics. Renew Energy 2012;38:141–9. http://dx. [38] Howell R, Qin N, Edwards J, Durrani N. Wind tunnel and numerical study of a
doi.org/10.1016/j.renene.2011.07.013. small vertical axis wind turbine. Renew Energy 2010;35:412–22. http://dx.doi.
[27] ANSYS Inc. ANSYS CFX Reference Guide; 2013. org/10.1016/j.renene.2009.07.025.
[28] Coughtrie AR, Borman DJ, Sleigh PA. Effects of turbulence modelling on [39] Lee YT, Lim HC. Numerical study of the aerodynamic performance of a 500W
prediction of flow characteristics in a bench-scale anaerobic gas-lift digester. Darrieus-type vertical-axis wind turbine. Renew Energy 2015;83:407–15.
Bioresour Technol 2013;138:297–306. http://dx.doi.org/10.1016/j.renene.2015.04.043.
[29] Dragomirescu A. Performance assessment of a small wind turbine with [40] Menter FR. 2-Equation eddy-visocity turbulence models for engineering
crossflow runner by numerical simulations. Renew Energy 2011;36:957–65. applications. Aiaa J 1994;32:1598–605. http://dx.doi.org/10.2514/3.12149.
http://dx.doi.org/10.1016/j.renene.2010.07.028. [41] ANSYS Inc. ANSYS CFX-Solver Theory Guide; 2013.
[30] Pope K, Dincer I, Naterer GF. Energy and exergy efficiency comparison of [42] Schlichting Herrmann, Gersten K. Boundary layer theory, vol. 82. Springer;
horizontal and vertical axis wind turbines. Renew Energy 2010;35:2102–13. 1979. p. 114–5. http://dx.doi.org/10.1016/S0997-7546(00)01101-8.
http://dx.doi.org/10.1016/j.renene.2010.02.013. [43] Sunny KA, Kumar NM. Vertical axis wind turbine: aerodynamic modelling and
[31] Gupta R, Biswas A. CFD analysis of flow physics and aerodynamic performance its testing in wind tunnel. Procedia Comput Sci 2016;93:1017–23. http://dx.
of a combined three-bucket Savonius and three-bladed darrieus turbine. Int J doi.org/10.1016/j.procs.2016.07.305.
Green Energy 2011;8:209–33. http://dx.doi.org/10.1080/15435075.2010. [44] Kacprzak K, Sobczak K. Numerical analysis of the flow around the Bach-type
548541. Savonius wind turbine. J Phys: Conf Ser 2014;530:12063. http://dx.doi.org/
[32] Ansys Inc., Ansys Fluent User Guide; 2013. 10.1088/1742-6596/530/1/012063.