On The Performance and Wake Aerodynamics of The Savonius Wind Turbine

ON T H E P E R F O R M A N C E AND WAKE AERODYNAMICS
OF T H E SAVONIUS WIND TURBINE
by
M A H A M A R A K K A L A G E SAMAN UDAYA K U M A R FERNANDO
B . S c . E n g . ( H o n s ) , University of Peradeniya, Sri L a n k a , 1980
A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F
THE REQUIREMENTS FOR T H E DEGREE OF
DOCTOR OF PHILOSOPHY
in
T H E F A C U L T Y O F G R A D U A T E STUDIES
(Department of M e c h a n i c a l Engineering)
We accept this thesis as conforming
to the required standard
T H E UNIVERSITY OF BRITISH C O L U M B I A
July 1987
© Mahamarakkalage Saman U d a y a K u m a r Fernando, 1987

In presenting this thesis in partial fulfillment of the requirements for an advanced
degree at T h e University of B r i t i s h C o l u m b i a . I agree that the library shall make it
freely available for reference and study. I further agree that permission for extensive
copying of this thesis for scholarly purposes may be granted by the Head of my
Department or by his or her representatives. It is understood that copying or
publication of this thesis for financial gain shall not be allowed without my written
permission.
Department of Mechanical Engineering

T h e University of B r i t i s h C o l u m b i a
2324 M a i n M a l l
Vancouver, B r i t i s h C o l u m b i a
Canada
V 6 T 1W5
July 1987
c
iii
ABSTRACT
T h e objective of the thesis is to establish methodology for development of a wind

turbine, simple in design and easy to m a i n t a i n , for possible application in devel-
oping countries. T o that end the Savonius configuration is analyzed in detail both
experimentally and analytically to lay a sound foundation for its performance eval-
uation.
Following a brief review of relevant significant contributions in the field (Chapter

I), an extensive w i n d tunnel test-program using scale models is described which
assesses the relative influence of system parameters such as blade geometry, gap-size,
overlap, aspect ratio, Reynolds number, blockage, etc., on the rotor output. The
parametric study leads to an o p t i m u m configuration w i t h an increase in efficiency
by around 100% compared to the reported efficiency of « 12 — 15%. O f particular
interest is the blockage correction procedure which is v i t a l for application of the wind
tunnel results to a prototype design, and facilitates comparison of data obtained by
investigators using different models and test facilities.
W i t h the design a n d performance results i n h a n d , Chapters III — V I focus

attention on a n a l y t i c a l approaches to complement the test procedure. Using the
concept of a central vortex, substantiated by a flow visualization study, Chapter III
develops a semi-empirical approach to predict the rotor performance using measured
stationary blade pressure data. T h e objective here is to provide a simple yet reliable
design tool which can replace dynamical testing w i t h a significant saving in time,
effort, and cost. T h e simple approach promises to be quite effective in predicting
the rotor performance, even in the presence of blockage, and should prove useful at
least in the preliminary design stages.
Chapter I V describes in detail a relatively more sophisticated and rigorous

B o u n d a r y Element A p p r o a c h using the Discrete Vortex M o d e l . T h e method at-
tempts to represent the complex unsteady flow field w i t h separating shear layers
in a realistic fashion consistent w i t h the available computational tools. Important
steps i n the numerical analysis of this challenging problem are discussed at some
length in Chapter V and a performance evaluation algorithm established. O f con-
siderable importance is the effect of computational parameters such as number of
elements representing the rotor blade, time-step size, location of the nascent vor-
tices, etc., on the accuracy of results and the associated cost.
iv
Results obtained using the Discrete Vortex M o d e l are presented and discussed
in C h a p t e r V I , for b o t h stationary as well as rotating Savonius configurations. A
detailed parametric study provides fundamental information concerning the starting
and dynamic torque time histories, power coefficient, evolution of the wake, Strouhal
number, etc. A comparison w i t h the flow visualization and w i n d tunnel test data
(Chapter II) shows remarkable correlation suggesting considerable promise for the
approach.
T h e thesis ends w i t h concluding remarks and a few suggestions concerning pos-

sible future research in the area.
V
T A B L E OF C O N T E N T S
Abstract iii
L i s t of Tables viii
List of Figures ix
Nomenclature xiv
Acknowledgement xvii
Chapter I INTRODUCTION 1
1.1 Development of W i n d Turbines 4

1.1.1 Betz m a x i m u m efficiency 4
1.1.2 Types of wind turbines 5
1.1.3 Vertical axis w i n d turbines 6
1.2 T h e Savonius Turbine 7
1.2.1 Experimental work 7
1.2.2 Theoretical investigations 14
1.3 Scope of the Present Investigation 16
C h a p t e r II O P T I M U M CONFIGURATION STUDIES 19
2.1 Definition of the B l a d e Shape 19

2.2 Experimental Methodology 21
2.3 E x p e r i m e n t a l Setup 24
2.3.1 W i n d tunnel 24
2 3.2 M o d e l fabrication and mounting arrangement 24
2.3.3 Measuring devices 28
2.3.4 Bearing friction calibration 30
2.3.5 Pressure distribution measurements 33
2.3.6 Starting torque 35
2.4 E x p e r i m e n t a l Results 36
2.4.1 Gap-size 36
2.4.2 Blade overlap 40
vi
2.4.3 Blade aspect ratio 40

2.4.4 B l a d e shape factor pjq 43
2.4.5 B l a d e arc angle 47
2.4.6 Reynolds number 47
2.4.7 Blockage ratio 52
2.5 Pressure Distribution 55
2.6 Starting Torque 63
2.7 Error Estimation 65
C h a p t e r III A SEMI-EMPIRICAL APPROACH FOR

PERFORMANCE PREDICTION 67
3.1 Inadequacy of the Quasi-Steady A p p r o a c h 67

3.2 Modified Quasi-Steady A p p r o a c h 69
3.2.1 Flow visualization study 71
3.2.2 Central vortex approach 73
3.2.3 E m p e r i c a l relations 73
3.3 Results and General Remarks 79
Chapter I V A MATHEMATICAL MODEL FOR

ROTOR ANALYSIS 84
4.1 Boundary Element M e t h o d ( B E M ) 85

4.2 Mathematical Formulation 87
4.2.1 Basic assumptions 87
4.2.2 Governing equations 87
4.2.3 Uniqueness of the solution 90
4.2.4 Indirect B o u n d a r y Element formulation 91
4.3 Discrete Vortex M e t h o d ( D V M ) 94
4.3.1 M o d e l geometry 94
4.3.2 Flow model 97
4.3.3 Vortex shedding model 99
4.3.4 Pressure, torque and power 102
4.4 M o d e l l i n g of the W i n d Tunnel Blockage 103
VII
Chapter V NUMERICAL INVESTIGATION 106
5.1 F o r m a t i o n of the S y s t e m M a t r i c e s 106

5.1.1 Unconfined flow model 106
5.1.2 W i n d tunnel blockage model 108
5.2 C a l c u l a t i o n of L o a d i n g on the Blade 109
5.2.1 Velocity distribution along the blade 109
5.2.2 C a l c u l a t i o n of | f term 115
5.2.3 E v a l u a t i o n of torque and power coefficient 118
5.3 C a l c u l a t i o n Procedure 119
5.3.1 C o m p u t a t i o n a l algorithm 119
5.3.2 C o m p u t a t i o n a l parameters 121
5.4 Effect of C o m p u t a t i o n a l Parameters 123
5.4.1 Number of elements m 123
5.4.2 Vortex core radius (limiting distance o) 127
5.4.3 Position of the nascent vortex t s 130
5.4.4 B a l a n c i n g parameter n 130
5.4.5 Time-step size At 130
Chapter V I RESULTS A N D DISCUSSION 138
6.1 Stationary Blade 139

6.2 Rotating Blade 170
6.2.1 Effect of tip-speed ratio A 171
6.2.2 Effect of p/q variation 193
6.3 Effect of W i n d Tunnel Blockage 193
6.3.1 Stationary blade 200
6.3.2 R o t a t i n g blade 204
Chapter VII CONCLUDING REMARKS 211
BIBLIOGRAPHY 216
viii
LIST O F T A B L E S
TABLE Page
5.1 Effect of number of elements on the computational cost 127

5.2 Effect of time-step size on the computational cost 137
6.1 Effect of k on the computational cost 155
6.2 Strouhal number at different blade angles 168
6.3 Effect of K on the computational cost 171
ix
LIST OF F I G U R E S
FIGURE Page
1- 1 A n illustrative schematic diagram of a w i n d energy

operated irrigation system 3
l-2a T h e Savonius configuration 8
1- 2b Geometry of the B a c h rotor 8
2- 1 Blade geometry and the associated parameters 20
2-2 B o u n d a r y layer w i n d tunnel used in the test program 25
2-3 M o d e l mounting frame and associated subassemblies 27
2-4 H y d r a u l i c dynamometer 29
2-5 Modified hydraulic dynamometer 29
2-6 Test arrangement for bearing friction calibration 31
2-7 Bearing friction characteristic 32
2-8 Pressure tap locations and numbering scheme 34
2-9 A schematic diagram showing model of the single-stage Savo-
nius rotor used in the gap-size and blade overlap studies 37
2- 10 Plots showing the effect of gap-size on
power coefficient^/** = 11.5%, i E n = 4 x 10 ) 5
38
2-11 Effect of gap-size on peak power coefficient
{b/d = 11.5%, R n = 4 x 10 )
5
39
2-12 Variation of the m a x i m u m power coefficient
w i t h percentage overlap (a/d= 0,R n = 4 x 10 )5
41
2-13 Effect of the blade aspect ratio on the peak
power coefficient (a/d = 0, b/d = 0,R =4x n 10 ) 5
42
2-14 A typical model of the two-stage Savonius rotor used
to study the effect of blade geometry parameter p/q 44
2-15 P o w e r - R P M characteristic of the two-stage model (p/q = 1.0),
showing the effect of bearing friction loss 45
2-16 Variation of the uncorrected power coefficient with
wind speed for the two-stage rotor (p/q — 1.6) 46
X
FIGURE Page
2-17 Bearing loss corrected characteristic curves at three

different w i n d speeds for the two-stage (p/q — 1.6) model 48
2-18 Effect of the blade geometry parameter p/q on
the power coefficient at a constant blockage 49
2-19 Variation of the peak power coefficient w i t h the blade geometry
parameter p/q at a constant blockage of 16.4% 50
2-20 Effect of the blade circular arc angle 6 on the power
coefficient for the single-stage model (B = 10%,p/q = 0.2) 51
2-21 Effect of blockage on power output
of the Savonius rotor 53
2-22a Effect of blockage on peak power coefficient
for two different blade geometries 54
2-2 2b Effect of blockage on the tip-speed ratio at
peak power for two different geometries 54
2-23a-f Surface pressure distribution over the Savonius
rotor as affected by the blade orientation: /? = 0° — 150° 56
2- 24 Variation of the starting torque w i t h angular position
of the Savonius rotor (p/q = 0.2, B = 10%) 64
3- 1 A typical velocity triangle at the blade element i combining
the effect of free stream velocity and blade rotation 68
3-2 Results of the quasi-steady analysis 70
3-3 A schematic diagram showing the towing
tank facility and associated equipment used in the
flow visualization study 72
3-4 A typical set of flow visualization pictures
showing the presence of the central vortex filament 74
3-5 Velocity distribution due to the potential vortex 75
3-6 Identification of the modified quasi-steady approach
parameter S2 as a function of p/q when F2
and blockage are kept constant 77
xi
FIGURE Page
3-7 Identification of the modified quasi-steady approach

parameter F2 as a function of blockage ratio when
5 2 and p/q are kept constant 78
3-8 E m p i r i c a l relationship between S2 and
the geometric parameter p/q 80
3- 9 E m p i r i c a l relationship between F2 and the blockage ratio B 81
3-10 Predictions of the modified quasi-steady approach
compared to the experimental results 82
4- 1 General description of the problem 89
4-2 Simpified geometry of the Savonius rotor
under investigation 95
4-3 Description of the problem parameters — .96
4- 4 Geometry for blockage modelling 105
5- 1 Uniformly distributed vortex 110
5-2 Velocity distribution for a uniformly distributed
vortex sheet compared to that for a point vortex 112
5-3a Velocity at the centre of the i th
element 114
5-3b Points at a radial distance t from the blade 114
5-4 T h e block diagram of the computation process 120
5-5 T i m e history of the torque coefficient as
affected by the number of elements 125
5-6 Flow patterns after 60 time-steps as affected
by the number of elements 126
5-7 Effect of the l i m i t i n g distance parameter o
on the time history of CT 128
5-8 Flow patterns after 60 time-steps as affected by a 129
5-9 T i m e history of the torque coefficient for
different nascent vortex positions 131
5-10 F l o w patterns after 60 time-steps for different t s 132
5-11 V a r i a t i o n of the torque coefficient as affected by 77 133
5-12 Flow patterns after 60 time-steps for different 77 134
XII
FIGURE Page
5-13 T i m e history of CT as affected by the time-step size 135

5- 14 Flow patterns at t = 0.792 s for different time-step sizes 136
6- 1 T i m e history of the torque coefficient at 0 — 0° 141
6-2 Flow patterns at 0 = 0 ° : k - 30,60,90 142
6-3 T i m e history of C T at 0 - 30° 143
6-4 F l o w patterns at 0 = 3 0 ° : k = 30 - 150 145
6-5 T i m e history of C T at• 0 - 60° 147
6-6 Flow patterns at 0 = 6 0 ° : ifc = 10 — 90 148
6-7 T i m e history of C T at 0 = 90° 151
6-8 Flow patterns at 0 = 9 0 ° : k = 10 - 150 152
6-10 Flow patterns at 0 = 120°: k = 40 - 200 158
6-12 Flow geometry at the blade orientation of 0 = 150° 163
6-13 Flow patterns at 0 = 150°: k = 10 - 200 164
6-14 V a r i a t i o n of the average starting torque as affected
by the blade orientation 169
6-15 T i m e history of C? at A = 0.4
as affected by the resolution parameter K 172
6-16 T i m e history of C P at A = 0.4 as affected by K 173
6-17 Flow patterns after k = 70,90,110, time-steps
at A = 0.4 w i t h K = 1 174
6-18 Flow patterns after k = 35,45,55, time-steps
at A = 0.4 w i t h K = 2 175
6-19 Evolution of the flow pattern for A = 0.4 and its compari-
sion w i t h the flow visualization study: k = 10 — 150 177
6-20 F l o w patterns after k = 20 - 100 time-steps: A '= 0.4,0.8,1.2;
p/q = 0.2; K = 2 184
6-21 T i m e history of C T at A = 0.8 189
6-22 T i m e history of C T at A = 1.2 190
FIGURE Page
6-23 Theoretically predicted variation of the average

torque coefficient as affected by the tip-speed
ratio A for p/q = 0.2 191
6-24 Theoretically predicted variation of the average
power coefficient as affected by the tip-speed
ratio A for p/q = 0.2 192
6-25 E v o l u t i o n of the flow pattern as affected by
p/q for A = 0.6 (Jk = 20 - 100) 194
6-26 T i m e history of C T as affected by p/q: A = 0.6; p/q = 0,1.0 . . . . 197
6-27 Theoretically predicted Cp vs. A curves for p/q = 0,0.2,1.0 199
6-28 T i m e history of the torque coefficient for a stationary rotor,
w i t h p/q - 0.2 and f3 = 120°, as affected by the blockage 201
6-29 Flow pattern after 100 time-steps for a stationary rotor,
w i t h p/q = 0.2 and 0 = 120°, as affected by the blockage 202
6-30 Variation of the average starting torque coefficient w i t h
blade angle (3 for a Savonius rotor: p/q = 0.2; B = 10% 203
6-31 Theoretically predicted characteristic curves
for a rotor w i t h p/q = 1.0 as affected by the blockage 205
6-32 Theoretically predicted characteristic curves
for a rotor w i t h p/q — 0.2 as affected by the blockage 206
6-33 Theoretically predicted variation of Cp max
w i t h blockage for different p/q 207

6-34 Flow pattern after 100 time-steps for a rotor w i t h p/q — 0.2
as affected by the blockage(A = 0.8,1.2) 209
XIV
NOMENCLATURE
a blade gap-size
A aspect ratio, h/d
A square matrix of order (m + 2 ) x (m + 2 )
b blade overlap
b column m a t r i x of order (m + 2 )
B blockage ratio, (h.d)/(H.W)
Cp pressure coefficient, (p — Poo)/\pU 2
base pressure coefficient

c P power coefficient, Pj\pV hd
z
CT torque coefficient, Tj\pXJ hdR2
d rotor diameter
ddisc end plate diameter
dA elemental area in 0
dS elemental length of the boundary S
DVM discrete vortex method
FuF 2 empirical parameters (Equation 3 . 2 )
F normal derivative of the Green's function. 4 ^
' a n
G Green's function
h blade height
H height of the w i n d tunnel
H space derivative of the Green's function, 4^-
k number of time-steps
K resolution parameter
m number of elements representing the blade contour
M w number of elements representing the wind tunnel w a l l
n normal vector
parameters defining blade geometry (Figure 2 - 1 )
Vii Poo pressure at the i th
tap and infinity, respectively
p power
Pf frictional power loss

XV
(jy*y) polar coordinates of the blade contour w i t h

respect to the main radius
R rotor radius, d/2
R n Reynolds number
As length of the blade element
Si, S2 empirical parameters (Figure 3-5)
S ,S<p
n boundary where | ^ and $ known, respectively (Figure 4-1)
t time
t* time elapsed since the generation of the vortex
t" period over which the power is averaged (Equation 4.29)
At time-step for convection calculation
AT time interval at which the boundary conditions are satisfied
T torque
u,v velocities in x and y directions, respectively
U free stream velocity
v n velocity normal to the surface
V velocity vector
w(z) complex velocity
W(z) complex potential
W w i d t h of the tunnel
x,£ s B points on the problem boundary
z , Zb
a end points of the uniformly distributed vortex
a angle between the element and the x-axis
0. angle of attack of the blade (Figure 2-8)
0] initial blade angle
7(f ) s vorticity density at £ s
r, , T s w strengths of the bound vortices

T(f) known vorticity density at point r
e s distance between the nascent vortex and the edge
of the blade (Figure 4-3)
r field point in fi
77 balancing parameter (Equation 4.26)
A tip-speed ratio, du/2U
/i viscosity
xv i
V kinematic viscosity
p density of air
o core radius of the free vortices
4> angle between the x-axis and the normal to the blade contour
$ potential function
w angle between the normal and radius vector at the blade contour
stream function
angular velocity of the rotor
n problem domain (Figure 4-1)
§c integration along an arbitrary path around the rotor

excluding singular points
h integration along the blade contour
Subscripts
indices
n nodal points
s singular points
wake
max peak value
xvii
ACKNOWLEDGEMENT
The author wishes to thank D r . V . J . M o d i for his generous assistance during the
course of this research work. His help and sharing of experience during the conduct
of the experiments and the preparation of this thesis are very much appreciated.
The author is grateful to the departments of Mechanical Engineering and Physics

for the use of various facilities. T h a n k s are also due to the technicians of the machine
shop for assistance during construction of the models.
The project was supported by the N a t u r a l Sciences and Engineering Research

council of Canada's grant (A - 2181) to D r . M o d i . I also want to recognize the
K i l l a m Foundation's generosity in awarding the Predoctoral Fellowship.
F i n a l l y , the author would like to express his gratitude to his wife, D h a m m i k a

and children for their support and encouragement in completing this thesis.
xviii
To my parents.
CHAPTER I
INTRODUCTION
Since the beginning of time, man has worshipped the elements of nature, often
out of fear, and has also strived to utilize them to advantage. In particular, light,
water and w i n d represent the elements frequently exploited by the Indus-Valley
C i v i l i z a t i o n of Ariyans and the E g y p t i a n Dynasty of the Pharaohs. W i n d m i l l s were
in existence in Persia as early as in 134 B . C . ( A h a m e d i , 1978). In western Europe,
windmills have been used for grinding grains and p u m p i n g water since the middle
ages. In the United States, more than six million w i n d m i l l s have been built since
1880, particularly in the midwest and southwest, and around 100,000 of them are
still in use (Savino and E l d r i d g e , 1975).
T h e attention towards utilization of w i n d energy (besides other alternative en-
ergy sources) may be attributed to increased consciousness about the depleting
character of the fossil fuel supply. For the technologically advanced countries, a
favourable balance of payments has always depended heavily on the exported prod-
ucts of their technologies. Unfortunately, the heavy manufacturing industry, which
supplies these products, uses an inordinate quantity of energy. Heavy consumption
of fossil fuels with the accompanying degradation of the quality of air, excessive
d a m m i n g of the rivers for hydro-electric power production, thermal p o l l u t i o n , and
possible risk of nuclear-electric plants; all have come under attack by environmental-
ists and other concerned groups. Consequently, the " clean " natural energy sources
1
Chapter I: Introduction 2
such as solar . geothermal, tidal and wind have gained in importance (Blackwell et
al., 1975).
It has been demonstrated that the per capita energy consumption is directly
related to the standard of living. Conversely, by improving the energy supply to
a certain community the standard of living can be upgraded. A large percentage
of the population in the developing countries live in rural areas. Being away from
major cities and scattered over a vast region, economical distribution of the grid
electric power is practically impossible for such communities. A l t h o u g h the rural
community subsists on farming, it does not produce enough for the village's au-
tonomy whereas,in the U n i t e d States, 1% of the population feed the nation and
export the surplus. T h i s disparity is largely due to old farming techniques and the
lack of energy. P u m p i n g water for irrigation is an important use of energy in rural
communities. As pointed out earlier, needed power cannot be supplied from con-
ventional energy sources due to remoteness of these areas and a small scale of the
energy use. In this situation, windmills appear particularly promising. A schematic
diagram of a wind energy operated irrigation system suitable for rural communities
is shown in Figure 1-1. Obviously, the w i n d turbine for rural application should
be simple in design, use locally available materials, and be easy to fabricate and
maintain. Furthermore, preferably, it should be self-starting and have a reasonably
high efficiency.
A l t h o u g h the wind has been used as an energy source for many centuries, the
potential of wind energy for today's world as an alternative energy source needs
to be emphasized. Consider the energy available in a wind stream. T h e kinetic
energy per unit mass of a moving air parcel is | c 7 , and the mass flow rate through
2
a stream tube of cross-sectional area A is pAU. T h e product of these two is the

F I G U R E 1-1 A n illustrative schematic diagram of a w i n d energy operated
irrigation system
power available,
P = [\U ){ AU)
2
P
= - U A.
P
3
Density of air (p), w i n d velocity (U), and stream tube area (A) are the variables
that influence the power availability in the w i n d . Seasonal density variation may
be about 20%. T h e stream tube area, which is directly related to the w i n d turbine
frontal area, increases w i t h the size of the turbine. W i n d velocity is the primary
variable that affects the available power because of the cubic relationship. For
example, a l O m / s wind contains 8 times more power than a 5m/s w i n d . Since size
of the turbine is a design parameter, it is convenient to use power per unit area
(power flux) for the meteorological purposes. T h e power flux being proportional
to the cube of the w i n d speed, mean wind speed data alone are not adequate for
estimating power availability. Instead, the mean of U , which is always greater than
3
the cube of mean w i n d speed, should be used to estimate the power availability at
a given site.
1.1 D e v e l o p m e n t o f W i n d T u r b i n e s
1.1.1 Betz maximum efficiency
Design of ancient windmills was based on empiricism and engineering skill. A d -
vent of the fluid mechanics as a science, and more specifically aerodynamics of wind
turbines, is more recent. Perhaps the earliest significant development i n this area
came about after the first world war and may be attributed to B e t z (1926). Betz
power coefficient, Cp Betz

t = 16/27, is considered the m a x i m u m possible efficiency
for w i n d turbines. In particular, since the derivation of the Betz efficiency appears
to be firmly based on the fundamental laws of conservation, it has been accorded
the same status as the Carnot efficiency for heat engines. However, such an analogy
Chapter 1: Introduction 5
is rather misleading because Betz derived his theory under restrictive assumptions
-the velocity field of the wind must be stationary, non-viscous and incompressible.
In addition, it must allow for a control volume which envelopes the rotor area and
is bounded by streamlines. Glauert (1935) improved Betz efficiency prediction by
taking wake rotation into account. A s a result Glauert efficiency is dependent on
the tip-speed ratio. A c c o r d i n g to Glauert, the efficiency is zero at zero tip-speed
ratio and reaches the B e t z efficiency as the tip-speed ratio tends to infinity. There
exists in the literature remarks to the effect that the efficiency of the real windmills
can exceed the Betz-Glauert limiting prediction (Hutter 1977, Inglis 1979). The
derivation of the B e t z efficiency is also critisized by Greet (1980), and by R a u h and
Seelert (1984). They claim that the Betz o p t i m u m efficiency cannot be rigorously
justified within Betz model itself and conclude that there exists no theoretical opti-
m u m for the power coefficient of windmills which is firmly based on the fundamental
laws.
1.1.2 Types of wind turbines
M a n y different types of w i n d turbines have been invented. A distinction can be
made between turbines driven mainly by drag forces versus those powered mainly
due to the lift. Further classification may be based on the orientation of the axis
of rotation: turbines with the axis parallel to the ground (Horrizontal A x i s W i n d
Turbines - H A W T ) ; and those w i t h the axis perpendicular to the ground (Vertical
A x i s W i n d Turbines - V A W T ) . T h e efficiency of the w i n d turbines driven primarily
by drag forces is usually low (De Vries, 1979) compared to the lift force driven
turbines. T h i s may be attributed to a large viscous dissipation of energy due to
flow separation in the drag devices. One the other hand, lift based devices tend to
be relatively more efficient. M o s t vertical axis wind turbines have the advantage of
their performance being independent of the w i n d direction, whereas a horrizontal
axis w i n d turbine has to be yawed into the w i n d direction to optimize performance.

1.1.3 Vertical axis wind turbines
T h e first modern vertical axis wind rotor was developed by S. J . Savonius (Savo-
nius, 1928), who published the results of its performance in 1931 (Savonius, 1931).
Another vertical axis configuration was patented about the same time by G . J . M .
Darrieus (Darrieus, 1931). A l t h o u g h both are vertical axis turbines, they work on
completely different principles. The Savonius turbine is, in part at least, a drag
driven machine whereas the Darrieus turbine is a lift force device. Over the years,
attention has centred particularly on the Darrieus rotor because of its relatively
high aerodynamic efficiency. Furthermore, its high rotational speed is favourable in
the generation of electric power. It has received attention, particularly in techno-
logically developed countries, for generation of electric power on a large scale. This
type of turbine is also amenable to theoretical modelling due to its well-investigated
aerodynamic shape. Consequently, a considerable body of theoretical and exper-
imental data exists on the Darrieus turbine performance. A l t h o u g h it has higher
efficiency, it suffers from the disadvantage of a low starting torque. A n external
prime mover is required to start the machine adding more sophistication to the
system.
Research on w i n d energy received a considerable boost due to the "energy crisis"
of 1973-74. A e r o d y n a m i c s of Darrieus rotors was investigated in depth by Lissaman
et a l . ( l 9 7 4 , 1976), H o l m e (1976), Base and Russel (1976), T e m p l i n (1974), Strick-
land et a l . ( l 9 7 5 , 1976) and many others. Corresponding experimental results on the
performance of the Darrieus rotor have also been presented by Lissaman et al.(l974,
1976), R a n g i et al. (1972, 1974), Feltz and Blackwell (1975), etc., to quote a few
important contributions. Blade element theory, vortex theory and panel methods
are the most commonly used tools in the analysis of the Darrieus turbine. Thus
a considerable amount of literature aimed at development of the Darrieus turbine,
both aerodynamically and structurally, already exists. However, due to its unde-
sirable starting characteristics and complexity in design, the Darrieus rotor is not
suitable for small scale applications, especially in the rural environment.
1.2 T h e S a v o n i u s T u r b i n e
1.2.1 Experimental work
T h e Savonius rotor concept never became popular, until recently, probably be-
cause of its low efficiency. However, it has the following advantages over the other
conventional w i n d turbines:
- simple and cheap construction;
- acceptance of w i n d from any direction thus eliminating the need for re-
orientation;
- high starting torque;
- relatively low operating speed (rpm).
T h e above advantages may outweigh its low efficiency and make it an ideal
economical source to meet small scale power requirements, especially in the rural
parts of developing countries. It has also been proposed as an auxiliary starting
device for the Darrieus turbine (Templin and South, 1976) and as a t i d a l power
generator (Manser and Jones, 1975). T h e concept of the Savonius rotor was based
on the principle developed by Flettner. Savonius used a rotor which was formed
by c u t t i n g the Flettner cylinder into two halves along the central plane and then
moving the two semi-cylindrical surfaces sideways along the cutting plane so that
the cross-section resembled the letter 'S'(Figure l-2a). A n ' o p t i m u m ' geometry was
obtained by systematically testing more than 30 different models in a w i n d tunnel,
and Savonius reported encouraging results. He conducted further tests in natural
w i n d and observed that the rotor ran at a higher speed than that i n the wind
tunnel for the same wind velocity. A c c o r d i n g to Savonius the best of his rotors
h a d a m a x i m u m efficiency of 31% while the m a x i m u m efficiency of the prototype

F I G U R E l-2a T h e Savonius configuration
F I G U R E l-2b Geometry of the B a c h rotor

was 37% (Savonius, 1931). However, other researchers who have conducted similar
experiments w i t h the Savonius rotor have not agreed w i t h the claimed efficiencies
(Shankar 1976, Sivasegaram 1978).
Following Savonius, B a c h (1931) made some investigations on the Savonius
rotor. He altered the basic shape of the blade by connecting two circular arc portions
w i t h different radii (Figure l-2b) which resulted in a considerable increase in the
power coefficient. B a c h reported a m a x i m u m efficiency of 24% for his rotor without
correcting for the blockage effect (B « 10%).
For nearly 40 years the field remained essentially dormant except for some iso-
lated contributions. Simonds and Bodek (1964) reported the possibility of using
empty oil drums to fabricate Savonius turbines for water p u m p i n g in rural areas.
They used a prototype machine to assess feasibility of the rotor in real life. The
Savonius design was found to be relatively insensitive to the quality of its aerody-
namic surfaces compared to the Darrieus turbine.
Newman (1974) conducted experiments to assess the effect of gap-size on the
performance. He used a two-bladed Savonius configuration in a closed-circuit w i n d
tunnel at different Reynolds numbers and at a blockage ratio of about 20%. New-
man's results indicate that while the o p t i m u m gap is of the order of 10% of the
diameter, a rotor without a gap has almost the same peak efficiency. He also found
the efficiency of the rotor to increase w i t h the increasing Reynolds number in the
operating range of his experiment. Newman applied corrections of the order of 50%
to his results for C p . O n the basis of his model tests it was concluded that the full
scale performance cannot be predicted with certainty from the model test results
because of the w i n d tunnel interferrence, however, model tests can be used to assess
relative merits of different rotor designs.
Further experimental investigations on the effect of Reynolds number and blade
geometry were carried out by several researchers in late 1970's and early 1980's.
Shankar (1976) studied the effect of the number of blades, blade gap and blade
curvature on the rotor performance. His experiments were carried out in an open-
circuit w i n d tunnel and w i t h a considerably low blockage ratio ( « 3%) at Reynolds
numbers between 1 x 1 0 and 2 x 1 0 . Shankar found the efficiency of the rotor

5 5
to increase w i t h the Reynolds number in this range although the increment was
not very significant. He concluded that except for a slightly higher starting torque,
the three-bladed rotors have up to 35% lower peak efficiency t h a n the two-bladed
rotors. T h e o p t i m u m blade geometry suggested by Shankar had a m a x i m u m power
coefficient of about 20%.
Sivasegaram (1977, 1978a, 1978b, 1982) has done a remarkable amount of work
on the development of the Savonius configuration. His experiments were conducted
in an open test-section w i n d tunnel using fairly small models. In summarizing
the results he concludes that the two-bladed configuration has more potential for
improvement compared to the multibladed design; and an increase in the Reynolds
number has a favourable effect on the m a x i m u m power coefficient. He indicates a
possibility of further improvements in performance although it might be relatively
small. Use of augmentors to concentrate the w i n d power is also recommended.
K h a n (1978) has done a fair amount of model tests to improve performance
of the Savonius rotor. His tests were carried out in a closed-circuit w i n d tunnel.
He used several different configurations of the two-bladed design and achieved a
considerable improvement in performance. O n the basis of the results obtained
from the model tests, a prototype unit w i t h the o p t i m u m configuration was built
and tested in an open site. K h a n concluded that depending on the basic geometry,
the blade over-lap and the blade-gap may have different o p t i m u m values. The
prototype test results confirmed what Savonius referred to as " the S-rotor instantly
utilizes any increase in the w i n d speed without having to lose time in getting into
the right position."

Chapter I: Introduction 1.1
A l d e r (1978) also tested the Savonius rotor in a closed-circuit w i n d tunnel. The
mean and periodic components of torque, drag and side force were measured. He did
not observe any significant effect of the Reynolds number on the power coefficient
in the range 2 x 1 0 — 6 x 1 0 . However, Alder d i d notice a significant change in

5 5
the nature of the flow around the rotor at a tip-speed ratio of 0.2.
G o v i n d a Raju and N a r a s h i m a (1979) investigated the use of sails in the fabri-
cation of Savonius rotors. They built a prototype turbine for pumping water which
showed encouraging results. Sundaram and G o v i n d a Raju (1980) tested sail type
Savonius w i n d turbines in an open-circuit wind tunnel. T h e o p t i m u m configura-
tion gave a m a x i m u m power coefficient of 0.18. Some flow visualization studies
using smoke to better understand the flow around the Savonius rotor were also
conducted. Unfortunately, the flow visualization results are insufficient to provide
a clear understanding of the flow.
T h e special aspects such as the effect of augmentors, shear flow (earth boundary
layer), w i n d tunnel blockage, etc., have also received some attention. Sivasegaram
(1979), Sabzevari (1978), and Morcos and Khalafallah (1981) investigated the effect
of various types of augmentors. Sabzevari tested several ductings, concentrators and
diffusers on a s p l i t ' S ' Savonius rotor in a w i n d tunnel. T h e results of these tests have
led to the design of the circularly ducted Savonius rotor equipped w i t h a number
of identical w i n d concentrators and diffusers along the periphery of its cylindrical
housing. The design proved to be effective independent of the wind direction and
showed about 125% increase in the m a x i m u m power coefficient compared to the
base value of 0.20 (without blockage correction).
Sivasegaram (1979) has tested several straight walled ducted concentrator con-
figurations in the w i n d tunnel and obtained ' o p t i m u m ' geometric parameters. A l -
though this type of augmentors are not directionally independent, a change in w i n d
direction by ss 10° d i d not show a significant change in performance. Simplicity in

design and fabrication was the m a i n objective. He was able to increase the power
by a factor of about 1.5 through moderate size concentrators.
M o r c o s et a l . ( l 9 8 l ) introduced a simple flat plate shield as a power augmentor.
A systematic series of model tests in a low speed w i n d tunnel gave an o p t i m u m
configuration which increased the peak power coefficient from 0.22 to 0.34. However,
this design is also not directionally independent. Sivasegaram et al.(l983) have also
shown the possibility of an 80% increase in power coefficient by using two moderate
size-stationary vanes in their o p t i m u m position.

•(•V
-•Aldos (1984) developed a mechanism allowing the rotor blade to swing back
and align with the wind when it is on the upwind stroke. T h i s reduces the drag
on the upwind blade and hence improves the performance. A c c o r d i n g to Aldos the
system is simple to construct and independent of the w i n d direction.
Ogawa et al.(1986a) tested in a wind tunnel a Savonius rotor w i t h straight
guide vanes installed axisymmetrically along the periphery. T h r o u g h a systematic
variation in geometric parameters they were able to improve the performance by a
considerable amount. Use of guide vanes improved starting characteristics of the
rotor.
' M a j o l a (1986) has examined, under field conditions, performance characteristics
of Savonius rotors w i t h seven different overlap ratios. R o t a t i o n a l speed, torque and
power results were obtained over a range of wind speeds. Design criteria based on
the avilable data were recommended.
Ogawa et al.(1986b) have also tested the effect of end plates on the performance
of the Savonius rotor. The results suggested an o p t i m u m size for the end plates.
Furthermore, contribution of the lift force to the rotor power was found to be
significant. Use of a deflecting plate in front of the rotor increased the power by
about 24%. It was also suggested that the deflecting plate can serve as a speed
control and safety device for Savonius turbines.

Bowen and McAleese (1984) carried out detailed measurements of the pulsating
w i n d flow around a Savonius rotor using: i) tuffts and a stroboscope; ii) a hot-wire
anemometer; and iii) a turbulence meter. D a t a suggested that "active coupling"
between the rotors might be useful in "redirecting" the wind more efficiently. In
particular, it was shown that if two counter-rotating rotors are placed side by side
in a w i n d tunnel, a natural phase locking occurs.
M a j o l a and Onsanya (1981) investigated the effect of shear flows on performance
of the Savonius rotor. T h e y used the conventional Savonius rotor in an open test-
section w i n d tunnel simulating a shear flow. It was concluded that the major effect
of w i n d shear is to reduce the power coefficient below the inviscid flow level. The
strong correlation between the power coefficient and the tip-speed ratio was noticed
even in the shear flow. T h e field testing of the Savonius rotor has indicated that
unsteadiness of the w i n d appears to be a far more important parameter governing
the power-loss than the w i n d shear.
T h e flow visualization results and the comments made on the flow patterns
by Jones et al.(l979) deserve some attention. It was observed that, i n general, a
Savonius rotor with zero gap describes the characteristic flow behaviour quite well,
and the performance can be improved by a careful blade design to delay separation
on the convex surfaces.
For the purpose of clarifying the mechanism of rotation of the Savonius geom-
etry, loading on the rotating and stationary blades was measured in a water filled
towing tank (Sawada et al, 1985). The flow around the rotor was visualized using
a l u m i n u m powder on the water surface. After a series of experiments it was con-
cluded that the lift force produces the torque over a large range of angular positions.
Alexander (1978), using a variation of Maskell's theory (Maskell, 1965) for
bluff bodies, predicted the w i n d tunnel correction for Savonius rotors. Several
w i n d tunnel tests were carried out to substantiate the approach. Considering the
magnitude of corrections, the agreement was good, particularly at the point of
m a x i m u m efficiency.
Suzuki and O k i t s u (1982) investigated the characteristics of a Savonius rotor
connected to a synchronous generator. The effect of changes in w i n d speed and load
resistance on the performance was discussed.
Based on a careful evaluation of the experimental procedures, measuring tech-
niques and analyses of the available data, following remarks can be made:
' (a) A s can be expected, earlier tests did not have the advantage of reliable
and sophisticated instrumentation.
(b) .Almost all the tests were carried out using models in fairly small wind
tunnels. A s a result the corrections due to blockage were substantial and
hence the accuracy of the measurements suffered.
(c) In a number of cases blockage corrections were completely ignored. W h e n
applied, the procedure was mostly empirical and unreliable.
(d) In all the test results, bearing friction losses were neglected. Unfortu-
nately, due to small models, they may affect the performance substan-
tially.
A systematic o p t i m i z a t i o n program for the two-bladed Savonius configuration
w i t h a desirable blade geometry (easy to construct) representing a circular arc
followed by a straight line remains unrecorded in the available literature.
1.2.2 Theoretical investigations
Available theoretical work on the analysis of the Savonius turbines is indeed very
scarce, perhaps due to the complex time-dependant separated nature of the high
Reynolds number flow. Several researchers have commented that a reliable analysis
of the flow around a Savonius turbine is extremely difficult if not impossible (Wilson
et a l . , 1976; Sivasegaram, 1977).
O n the other hand, there are several well developed theories to analyse the
Darrieus and propeller type w i n d turbines where lift is the main driving force. The
blade element theory (Glauert, 1935; W i l s o n and L i s s a m a n , 1974; Holme, 1981;
and others), which is one of the simplest, can predict the performance of a Darrieus
turbine reasonably well. It assumes that the different spanwise blade elements are
independent of each other and that the forces on the elements can be determined
from the local flow conditions. The flow field is estimated by either a momentum
or a vortex consideration or both.
Vortex and panel methods are also widely used in the analysis of Darrieus type
turbines. Instead of estimating induced velocities from the m o m e n t u m equation,
a better model analogous to the theory for a w i n g w i t h a finite span can now be
formulated. These methods are based on the "lifting line" approach of P r a n d t l
(1918), the "lifting surface" theory of Weissinger (1947), the "panel" methods of
Hess (1972) and others. The flow around the wing is assumed to be irrotational
everywhere (except in the t h i n layer of trailing vortices), and the velocity is given
in terms of a perturbation potential.
A z u m a and K i m u r a (1983) have developed a computationally sound approach
based on local circulation method to calculate the air loading on the Darrieus tur-
bine. T h e method when checked experimentally showed good agreement ( K i m u r a
et a l . , 1984).
Due to the completely different nature of the flow around the Savonius rotor the
above mentioned methods cannot be applied directly in its analysis. A l t h o u g h the
classical m o m e n t u m theory was used by Betz (1926), it fails to distinguish between
different types of w i n d turbines, not to mention the effect of various parameters of
the Savonius geometry.
For the Savonius rotor an analytical model was developed by Wilson et al.
(1976). Another model using a vortex sheet was presented by V a n Dusen and
Kirchhoff (1978). These are the very first theories found in the literature for the
prediction of the performance and flow field of a Savonius rotor. However, these
models assume attached flow throughout, which is unrealistic. Available literature
describing the modelling of separated flows reveals the considerable potential of the
discrete vortex method. A theoretical analysis using the discrete vortex method and
assuming flow separation at the tips of the blades has been developed by Ogawa
(1984). A l t h o u g h the performance predicted is not quantitatively accurate, the
model provides fairly accurate qualitative description of the flow.
H a t a y a m a et al. (1984) analysed the problem using the discrete vortex method
together with conformal mapping. A conventional Savonius rotor w i t h zero gap
between the blades was studied. T h e rotor was transformed into a unit circle using
a rather complicated mapping function and the analysis carried out in the mapped
plane. T h e unsteady lift and drag forces as well as the moment were evaluated
for three different geometries and at two different tip-speed ratios. A s a result of
the conformal mapping, the prediction curves were found to be relatively smooth
compared to the results of the discrete vortex method used in the physical plane
(Ogawa, 1984).
1.3 S c o p e o f t h e P r e s e n t Investigation
T h i s thesis describes the development of a Savonius rotor configuration, simple
in design, fabrication and maintenance, yet having a reasonably high efficiency. It
is suitable for rural small scale applications in developing countries. Attention is
also directed towards the aerodynamic problem of modelling the unsteady separated
flow around a Savonius rotor in order to predict its performance.
In Chapter II the experimental methodology and the corresponding results of
the systematic o p t i m i z a t i o n process undertaken, to develop an efficient Savonius

Chapter I: Introduction .17
rotor configuration, are presented. Special care has been taken to overcome the
problems which led to unreliable results found in the literature. Focus is on the
two-bladed configuration w i t h geometric parameters varied i n an extensive wind
tunnel test program. Effect of the blade geometry, w i n d tunnel blockage as well
as frictional losses in the model mounting bearings are assessed. T h e pressure
distribution and starting torques have been measured for the stationary blade at
different angles.
C h a p t e r III presents a semi-empirical approach to predict performance of the
Savonius rotor. T h e analysis is based on a quasi-steady method, similar to the
blade element theory, w i t h a central vortex at the core, and the measured pressure
data as an input. T h e model is based on the flow visualization study carried out in
a tow-tank. The experimental results of Chapter II are used to evaluate empirical
parameters.
In Chapter I V , the boundary element method and its application to the analysis
of a Savonius rotor are discussed in detail. The flow model and the corresponding
governing equations are developed using first principles and assumptions made in
the analysis are discussed.
Discretization of the governing equations and their numerical solution procedure
is described at length in Chapter V . The effect of flow-model parameters as well
as computational problems are discussed. In Chapter V I results are obtained for
starting torque at different angular positions and compared w i t h experimental data.
T h e flow fields corresponding to different angular positions of the blade are also
obtained. T h e vortex shedding behaviour of the stationary blade at several angles
is also discussed w i t h reference to the Strouhal number. For the rotating blade,
flow patterns at different tip-speed ratios for several geometries are obtained. T h e
wake characteristics are compared w i t h the corresponding flow visualization results.
T h e time history of the torque coefficient, and hence the power coefficient, has been
Chapter J: Introduction 18
calculated over a range of tip-speed ratio for different geometries of the rotor and
compared w i t h the experimental results of Chapter II. The mathematical model is
also used to account for the wind tunnel blockage.
T h e thesis ends w i t h concluding remarks and recommendations for further
study.
CHAPTER II
OPTIMUM CONFIGURATION STUDIES
T h i s chapter discusses a comprehensive w i n d tunnel test program aimed at opti-
mization of a simple blade geometry to give the m a x i m u m possible power coefficient.
T h e simplest of blade geometries would be the original Savonius configuration (Fig-
ure l-2a), but several researchers ( K h a n , 1978, Sivasegaram, 1978) have concluded
that the possible improvement in efficiency through modifications of the blade would
be little. Bach(1931) suggested a rather different geometry (Figure l-2b) involving
portions of two circular arcs. Performance of the Bach rotor is indeed better com-
pared to the original Savonius design, however, the configuration is relatively more
complicated. In 1978 Sivasegaram altered this configuration replacing the second
circular arc by a straight line (arc of infinite radius, Figure 2-1). He conducted
some experiments aimed at improving the performance, however, the extent of i m -
provement that can be realized proved to be inconclusive. In the present study, the
circular arc-straightline blade geometry is systematically studied in a parametric
fashion to arrive at an o p t i m u m configuration.
2.1 D e f i n i t i o n o f t h e B l a d e S h a p e
T h e basic blade shape shown in Figure 2-1 can be denned by the following param-
eters:
19
Chapter II: Optimum Configuration Studies 21
a blade gap-size ;
b blade overlap ;
D shaft diameter ;
d rotor diameter ;
ddi sc end plate diameter ;
H height of the w i n d tunnel ;
h blade height ;
p -+ b length of the straight line portion ;
q radius of the circular arc portion ;
W w i d t h of the w i n d tunnel ;
6 blade circular arc angle .
To help establish the two dimensionality of the flow, end plates are used.
In general, all length parameters (except p,q) are non-dimensionalized w i t h
respect to the rotor diameter d. T h e non-dimensionalized blade shape parameter
p/q 'defines the basic geometry.
2.2 E x p e r i m e n t a l M e t h o d o l o g y
T h e power P developed by a rotor, operating under unconfined conditions, would be
a function of the blade parameters, density p, wind speed U, rotor angular velocity
u and viscosity p,. W i t h the w i n d tunnel wall restriction the additional blockage
parameter B has to be introduced,
^ g m a x i m u m projected area of the model (h d) ^

cross-sectional area of the tunnel ( H W ) '
i.e.,
P = P{a, b, rf, h, p, g, D, d ,0, di3C w, p, fi, U, B) (2.2)
U s i n g the B u c k i n g h a m IT-theorem, this expression can be rewritten i n terms
of nondimensional parameters (assuming a geometric constraint given by E q u a t i o n
2.10) as,
Cp = Cp
{l^^l^^W ^) B
' (2
- 3)
where
is referred to as the average power coefficient. T h e parameter
du>
is k n o w n as the tip-speed ratio and
A =
1 ( 2
- 6 )
is called the blade aspect ratio. R n is the Reynolds number based on free stream
velocity and rotor diameter.
Similarly, the average torque coefficient CT can be expressed in non-dimensional
terms as
Cj — CT{-,, A,-, — , d t s c
,6, A,B,R )n , (2.7)
a a q a a
where
C t = T ^ H l . . (2.8)
Note, A is the only independent parameter that varies w i t h the loading. There-
fore, Cp vs A curve represents the fundamental characteristic of each rotor.

Chapter 11; Optimum Configuration Studies 23
As pointed out in Chapter I, the flow past the turbine blade is so complex that
no exact theoretical method exists to predict it. T h e flow is unsteady relative to
the blades and, at any given instance, it is partly separated. Even if a theoretical
description of the flow past the rotor may appear reasonable, predictions based on
such conjecture cannot be accepted w i t h certainty. Under these circumstances the
logical approach is model testing.
Experimental approach to performance prediction requires that the model sat-
isfy certain fluid dynamical similarity w i t h the prototype machine. Geometric sim-
ilarity (similarity in ratio between the principal dimensions) is relatively easy to
achieve in model fabrication. K i n e m a t i c similarity (similarity in ratios between
the velocities at various corresponding locations) is achieved by maintaining similar
geometry and operating conditions. In the present study, this is ensured by main-
taining the same tip-speed ratio for models and the prototype. D y n a m i c similarity
(similarity between the forces acting on the turbine) is basically achieved when the
operating Reynolds n u m b e r s ( J 2 „ ) for the model and the full scale turbine are equal.
In general, the Reynolds number has relatively small influence on the model per-
formance. In laboratory tests, a model frequently has a smaller Reynolds number
compared to the prototype. This leads the performance predicted by the model to
be marginally smaller than that for a full scale machine. However, it is advisable to
perform the model tests at Reynolds numbers comparable to those of the prototype
turbines under field conditions.
In the present test program, relatively larger models and higher w i n d speeds
are used to achieve comparable Reynolds numbers. T h i s has an added advantage
of better accuracy i n the measurement of power and torque as well as the w i n d
velocity.
However, the use of larger models presents the problem of w i n d tunnel block-
age, and the associated correction could become significant. A closed w i n d tunnel
requires negative corrections whereas a w i n d tunnel w i t h open test-section needs
positive corrections. Obviously, the prediction would not be reliable unless a proper
method for blockage correction is available. In this study, a set of experiments is
carried out to establish the effect of blockage, thus providing an approach to obtain
reliable data.
T h e test program involves a systematic variation of variables affecting the per-
formance curve C p vs. A, thus leading to an ' o p t i m u m ' combination of parameters.
2.3 E x p e r i m e n t a l S e t - u p
2.3.1 Wind tunnel
M o s t of the model tests were carried out in the boundary layer w i n d tun-
nel(Figure 2-2). T h e partially return type tunnel has a 2.44 m wide and 24.4 m
long test-section consisting of eight 3.05 m long bays. T h e height of the test-section
can be varied using rachets operating on the top panels. For this study the first
bay w i t h smooth uniform flow was used using a height setting of 1.6 m at the test-
section. T h e estimated m a x i m u m turbulence level in this section is less than 0.4 %.
W i n d is driven by a variable pitch blower running at a constant speed of 700 rpm.
Power consumption of the blower is rated at 80 kW. T h e tunnel can provide a
stable wind speed in the range 2.5 m/s — 25 m/s and is fitted w i t h emergency stop
switches located every 6 m along the test-section.
2 3.2 Model fabrication and mounting arrangement
Dimensions of a model are determined basically by the blockage ratio (B),

Axivane Series 2000 Rotor.
2.44 m dia..
16 Cast aluminum blades.
125 h.p. electric motor,

1 honeycomb and
175,000 cfm at 700 rpm. 4 screens in 4 x 4 i
settling section
Fisher 4 8 0 - 6 0 pneumatic
variable pitch control
Test-Section,
2.44 x 1.6 m
F I G U R E 2-2 Boundary layer wind tunnel used in the test program.

Chapter It: Optimum Configuration Studies 26
aspect ratio (A), shape factor (p/q), blade gap (a), blade overlap (b) and blade
arc angle (0). T h e blockage ratio fixes the value of the projected area whereas the
aspect ratio limits the diameter to height ratio. These two parameters determine
the blade diameter (d), and height (h). For the convenience of using the same
mounting structure, the diameter of the main shaft (D) was kept constant at 3 8 m m .
Depending on the p/q, a and 0 value, the dimension q can be evaluated using:
S V - 2S Cq 2 3
- (d - C ) { — + 1 )q
2 2 2
+^ ~ ^ Cq + (fjl^l) *= 0 ; (2.10)
2 | 2
(a + 0 > n)
or
(J2 _ (-,2)
q {S
2 2
+ 2 S s i n 0 + 2 ( l - c o s t ? ) } -qC(l-cos0)- '- - 0 ; (2.11)
(a + 0 < 7 r )
where:
5 = ?;
C = D + a; (2.12)
- i t S
a = tan
1 -
2q
Note, for C < 2q, a ^ t a n " S. 1
T h e models were fabricated using 16 gauge a l u m i n u m sheets. Blades were rolled
to the desired shape using a sheet metal rolling machine, and pop-rivets were used
for fastening. B o t h single and two stage models were built depending on the re-
quirement.
A frame made of angle iron (Figure 2-3) supported the models in the wind tun-
nel. T h e model, mounted on a vertical steel shaft, was supported by two self aligning
Chapter II: Optimum Configuration Studies
ball bearings. The lower end of the shaft was flexibly coupled to a dynamometer.
T h e frame was tied to the w i n d tunnel walls to avoid undesirable vibrations.
2.3.3 Measuring devices
T h e wind velocity was measured using a pitot-static tube connected to an in-
clined alcohol manometer. Since the model was mounted in the first bay, the wind
velocity measurement was not possible upstream of the model. Therefore, the pitot-
static tube was placed in the last bay (w 25 diameters downstream), at the centre
line of the flow parallel to the tunnel, where the wake can be assumed to have
reached steady state. T h e roof height was adjusted to compensate for the growing
boundary layer in the tunnel. Velocity measurements at the first and the last bay
showed no difference in the absence of the model. Even i n the presence of a model
w i t h 10% blockage, a scan of the velocity field, at the centre line of the last bay
showed the deviation to be less than 1%. T h e rotor rpm was measured using a
strobotac.
T h e emphasis was on measurement of the torque. Since the torque varied over
a large range, different types of dynamometers were required in different ranges.
However, the basic approach to the measurement was similar in each case. Es-
sentially the force was measured using a cantilever beam with four strain gauges
attached near its root, two on either side. T h e output signal from the strain gauges,
forming a part of the Wheatstone bridge, is amplified using a Bridge Amplifier M e -
ter ( B A M ) and measured by a digital voltmeter. T h e sensitivity of the system was
calculated as 0.5 x 1 0 ~ Nm.

3
A t the beginning of each trial the force measuring
system was calibrated using known loads.
A higher accuracy is required when measuring smaller torques. T h e hydraulic
dynamometer was designed and fabricated for this purpose(Figure 2-4). Load was
applied by filling the outer cylinder w i t h high viscosity oil ( S A E lOWfJO). The
shaft of the turbine is connected to the central rotating vanes through a flexible
A A
Dynamomete Flexible couplings Dynamometer if Flexible couplings
support frame rn support frame _l*fl
Bearings
Bearings Bearings
Strain gage Strain gage

torquemeter torquemeter
Stationary cylinder Stationary cylinder
Vanes Vane
Additional
Vanes
Oil drain Oil drain
F I G U R E 2-4 H y d r a u l i c dynamometer. F I G U R E 2-5 M o d i f i e d hydraulic dynamometer.

Chapter fj: Optimum Configuration Studies 30
coupling. T h e motion of the outer cylinder is restricted by a string connecting the
cylinder to the strain-gauged cantilever. T h e m a x i m u m torque that can be measured
using this type of dynamometer depends on the rotational speed of the shaft. The
present device is capable of measuring torques up to 2.2 Nm at a rotational speed
of 180 rpm. The m a x i m u m measurable torque decreases rapidly w i t h a decrease in
shaft rotational speed.
To measure a higher torque the dynamometer was modified by inserting verti-
cal vanes on the inside wall of the outer cylinder (Figure 2-5). T h o u g h the concept
seems promising, improvement in results was not as expected. In fact, performance
of the modified dynamometer deteriorated at higher rotational speeds, due to up-
ward creeping motion of the oil between the stationary vanes of the outer cylinder,
leaving a clear gap between stationary and rotating vanes. T h e concept is indeed
sound, however, design modifications are needed to make the device effective. To
save time it was decided to leave this development as a part of a separate project.
A prony brake type dynamometer was routinely used for larger torque measure-
ments. One end of the rope-brake is connected to a known weight through a pulley
while the other end is attached to the strain-gauged cantilever. T h e system was
loaded through dead weights. T h e arrangement yields reasonably accurate measure-
ments i n the higher torque range. Since the weight imposes transverse loading on
the m a i n shaft, the bearing friction is slightly affected by this type of dynamometer.
2.3.4 Bearing friction calibration
A t higher rotational speeds the frictional loss due to bearings cannot be ne-
glected. Therefore a set of experiments was carried out to establish the frictional
power loss characteristics of the bearings. T h e arrangement shown in F i g u r e 2-6 was
used for this purpose. T h e bearings support a shaft driven by a variable speed D . C .
Belt drive
Variable speed
drive
Bearings
Frame
*"' T» In
Bearings
Strain gage Mr
under test T
transducer
F I G U R E 2-6 Test arrangement for bearing friction calibration.

10.0-1
.A"
8.0-
P f =0. 0308 f RPM-2 .1538

7.0
A ^
00
^ 6.0 .A'
Â
8 !
o P = 0 .0250 RPM- I.O

f
1
^
CL
3.0
A"
2.0
1.0
0.0
80 100 120 140 160 180 200 220 240 260 280 300 320 340
RPM
F I G U R E 2-7 Bearing friction characteristic. GO

1x3
motor. T h e frictional torque of the bearing was measured using the strain-gauged
cantilever over a range of speeds. It is evident from Figure 2-7 that, i n general, the
bearing loss characteristic is piece-wise linear and can be expressed as:
f 0.025RPM - 1.0, 0 < RPM < 200; . .

f
~ \ 0.0Z0SRPM - 2.1538, RPM > 200: ^ ' '
where Pf is the frictional power loss.
2.S.5 Pressure distribution measurements
To better understand behaviour of the flow around the Savonius rotor, it was
considered desirable to measure the pressure distribution along the blade. Mea-
surements on the rotating blade are extremely difficult and need very sophisticated
instrumentation. In this study the pressure measurement is limited to the station-
ary blade at different angles. This provided rather important information such as
the-positions of the separation points, base pressure coefficient, etc.,which proved
to be useful in theoretical modelling.
A single-stage model w i t h 46 pressure taps was used for this purpose. The
pressure taps were mounted w i t h the openings on the convex surface of one blade
and the concave surface of the other blade (Figure 2-8), and were connected to a
scanivalve (48J9-1331) v i a 'tygon' tubes (<£§")• Pressure at a tap was monitored
using a transducer ( B A R O C E L 511J-10) attached to the scanivalve. Care was
taken to minimize the interference due to the tubes by keeping t h e m in the wake
region. T h e wind speed was set at 6.76 m/s for all pressure measurements, which
corresponds to & ,R n of 3.2 x 1 0 (based on free stream velocity and rotor diameter).
5
T h e model had a blockage ratio of 10%.
T h e Savonius turbine model was mounted on two clamps w i t h the facility to

Front Side
F I G U R E 2-8 Pressure tap locations and numbering scheme.

Chapter If: Optimum Configuration Studies 35
lock it in a desired angular position. T h e angle was measured using a protracter
scale attached to the lower clamp w i t h 1° accuracy. The pressure was measured
at 10° intervals from 0° to 360°. Measurements over one cycle were appropriately
combined to give the pressure distribution on both sides of each blade.
T h e pressure transducer was calibrated before each run and all the taps were
checked for leakage.
T h e pressure coefficient at the i th

tap is defined as,
where pi is the pressure measured at the i th

tap; Ptx>, the static pressure at infinity;
p. the density of air; and U, the free stream velocity.
2.3.6 Starting torque
Starting torque for the rotor w i t h p/q = 0.2, 10% blockage ratio, single-stage
model was measured over a range of angular positions. The model was mounted
vertically on two bearings as explained before. A d r u m of 127 mm diameter,attached
to the lower end of the shaft, had a string tangential to its surface connected to
the strain gauge arrangement mentioned earlier to measure the static torque. The
angular position of the turbine was recorded using a protractor scale mounted at the
lower bearing. B y varying the length of the string, different angular positions were
obtained. The arrangement permitted measurement of negative starting torque
when rotated through 1 8 0 ° . T h e strain-gauged cantilever system was calibrated
using known loads w i t h sufficient time allowed for w a r m up.
T h e torque measurements were carried out at two different wind speeds (R n =

Chapter II; Optimum Configuration Studies 36
2.1 x 1 0 , 2.7
5
x 10 ).
5
2.4 E x p e r i m e n t a l R e s u l t s
Basically two models were used in the test program. A smaller two blade model
(Figure 2-9) with a m a x i m u m blade diameter of 3 3 0 m m and a height of 3 0 1 m m ,
resulting in the aspect ratio of 0.91 and projected area of 0 . 1 m 2

was primarily
designed to study the effect of gap-size and overlap. The blades, rolled into the de-
sired shape were supported by two plexiglass end plates, 6.35 mm thick and 381 m m
in diameter. Rigidity of the plates was augmented by reinforcing t h e m w i t h alu-
m i n u m discs, 203 m m in diameter and 6.35 m m in thickness. A 25.4 m m diameter
high prescision straight shaft supported the blade assembly in a pair of self-aligning
bearings.
For blade-gap and overlap studies a model w i t h p/q = 1.9, 0 — 112° was
used. The model was tested in a low speed low turbulence return type w i n d tunnel
w i t h a test-section of 0.91 x 0.68 (blockage 17% ) at a w i n d speed of 17.9 /s. The
estimated turbulence level of the tunnel is less than 0.1%.
2.4-1 Gap-size
T y p i c a l results for the variation of power coefficient w i t h tip-speed ratio and
blade-gap are shown in Figure 2-10. M o r e informative would be the effect of percent-
age gap-size on the m a x i m u m power coefficient as presented in Figure 2-11. This
clearly shows that as separation a is increased, the m a x i m u m power diminishes.
T h e peak power coefficient of 0.158 occurred at zero gap-size.

Chapter II: Optimum Configuration Studies
381 mm
End Plate
Rotar Blades 301 mm
203 mm-
330 mm
F I G U R E 2-9 A schematic diagram showing model of the single-stage

Savonius rotor.
0.16-1
0.14'
0.12-
0.10-
0.08-
0.06-
A a/d = 0.000
0.04 X a/d = 0.018
• a/d = 0.039
= 0.058
0.02 a/d
S a/d = 0.106
0.00 0.5 I
0.6
I
0.7
I
0.8
I
0.9. 1.0
I
1.1 1.2 1.3 1.4 1.5
A
F I G U R E 2-10 Plots showing the effect of gap-size on power coefficient

[b/d = 11.5%, i E n = 4 x, 1 0 ) .
5
0.20
0.18 H
0.10 | i 1 j 1 ' 1 1 1 1 1 ' 1

0.0 3.0 6.0 9.0 12.
a/d %
F I G U R E 2-11 Effect of gap-size on peak power coefficient (b/d - 11.5%,

R = 4 x 10 ).
n
5
2.4.2 Blade overlap
To investigate the effect of blade overlap, models similar to that described earlier
but w i t h zero blade gap and specified overlap ratios were used. B y combining power
coefficient vs. tip-speed ratio curves for different blade overlaps, a more informative
curve of m a x i m u m power coefficient vs. percentage overlap was established (Figure
2-12). It is apparent that an o p t i m u m value for blade overlap is around 10%. The
plot also shows that the m a x i m u m power coefficient remains essentially unaffected
for overlaps less than 10%.
These results are uncorrected for blockage and bearing power loss. A s the
blockage ratio for models in this set of experiments is the same, it is not likely to be a
significant parameter affecting the o p t i m u m . O n the other hand, the bearing friction
correction may have a significant, effect on the final prediction since it is proportional
to the rpm. Fortunately, since the measured power outputs are reasonably high,
the error due to bearing friction is not likely to affect the o p t i m u m blade setting by
a significant amount.
2.4-8 Blade aspect ratio
O f considerable interest is the effect of the aspect ratio (A) on the wind turbine
performance. To that end, a set of models similar to the one discussed earlier but
w i t h different aspect ratios (keeping projected area constant) and fixed a/d, b/d
(•- 0) were fabricated and tested in the wind tunnel in a similar manner. The
results showing the variation of m a x i m u m power coefficient vs. aspect ratio (A) are
presented in Figure 2-13. T h e results are corrected for bearing loss. T h i s shows an
o p t i m u m aspect ratio of 0.77 giving a m a x i m u m power coefficient of 23.5%.
Besides providing the useful information concerning the o p t i m u m blade config-

F I G U R E 2-12 V a r i a t i o n of the m a x i m u m power coefficient w i t h percentage

overlap showing the peak around 10% (a/d = 0,R n = 4 X 10 ).
5
0.25
0.24 H
F I G U R E 2-13 Effect of blade aspect ratio on the peak power coefficient

(a/d = 0,6/rf = 0 , , R = 4 x 1 0 ) .
n
5
uration, the single-stage model study emphasized, as expected, the presence of dead
spots when the blades are aligned with the w i n d and the rotor fails to start on its
own. T h u s for self-starting of a two blade rotor, it is necessary to have at least
a two-stage system w i t h blades in the individual stage oriented orthogonal to one
another. Furthermore, results suggest that to generate even 100 W of power at a
w i n d speed of 25 km/h it would require a projected area of 3.5 m . In addition, the

2
ease of fabrication being a guiding criterian, particularly in a rural environment,
suggested a multi-stage construction. It was, therefore, decided to conduct tests
w i t h models of a two-stage rotor to assess interference effects due to staging.
To establish the effect of blade shape, two-stage models w i t h a projected area
« 0.6 m 2
were used.
2.4-4 Blade shape factor p/q
It is reasonable to assume that the basic blade shape has a significant effect
on its performance. A s shown in Figure 2-1 the parameter p/q governs the basic
shape of the blade. A set of two-stage models, diameter 635 m m , stage height
489 m m (A = 0.77), and end-plate diameter 8 4 7 m m was constructed with gauge 16
a l u m i n u m sheet to study the effect of shape factor p/q (Figure 2-14).
For simplicity in fabrication 6 was taken to be 135° for all models. A typical
power vs. r p m plot for p/q — 1, without friction correction, is presented in Figure
2-15 . It shows the m a x i m u m power of 39 W at 215 rpm. Note, the corrected curve
indicates a peak power of 44 W at 222 rpm. Thus the bearing losses, if unaccounted
for, would lead to an error of approximately 11% in peak power and 3% i n the
prediction of the corresponding r p m .
Uncorrected Cp vs. A curves at three different w i n d speeds for p/q = 1.6 are
F I G U R E 2-14 A typical model of the two-stage Savonius rotor used to study the
effect of blade geometry parameter p/'g (projected areas: 0 . 6 m ) .
2
45.0
<x x x
- X> X
X x
X
X x
x x
x
40.0- X x
i A
* A
X A * l
A X
X
A
A
35.0- A ... .A.
" A.. X
A
A
A >
30.0
•
A
X
% 25.0
•
L\ X
£ 20.0
O A
Q_ •
15.0-
10.0-
U = 6.76 m/s
5.0- A Uncorrected
•
x Corrected
0.0
150 170 190 210 230 250 270 290 310 330 350
RPM
F I G U R E 2-15 P o w e r - R P M characteristic of the two-stage model (p/q = 1.0),

showing the effect of bearing friction loss.
0.35^
0.30 _ rn • n .
•
1
u c
• •
< X X
• X >
x >< c
X
X
X •
0.25H ><
A *^ A A
A A X
X
L
A A
L
0.20H
0.15 H
0.10 H
A u
0.05H
X u
• u
0.00 1 1 1 1 1 ,
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3
A
F I G U R E 2-16 V a r i a t i o n of power coefficient w i t h w i n d speed for the two-stage

rotor (p/q = 1.6) emphasizing the effect of bearing power loss.
Chapter H: Optimum Configuration Studies 47
shown in Figure 2-16. In general, one would expect variation of the power coefficient
w i t h tip-speed ratio to be essentially independent of the w i n d speed. However,
Figure 2-16 suggests a marked dependence due to the uncorrected character of the
data. A c c o u n t i n g for the bearing dissipation led to near collapse of the results on
a single curve as shown in Figure 2-17. T h e slight discrepancy between corrected
curves may be attributed to the Reynolds number effect, which in this case is
relatively insignificant.
Corrected C p vs. A curves for different p/q values are shown in Figure 2-18. In
this set of experiments the blockage ratio (B) was kept constant at 16.4%. Note, the
geometric parameter p/q has a significant effect on the power coefficient. T h e more
informative graph in Figure 2-19 shows the variation of m a x i m u m power coefficient
w i t h p/q. Under the given conditions, the m a x i m u m power coefficient was found to
be as high as 0.5 (uncorrected for blockage) for p/q — 0.2.
2-4-5 Blade arc angle
To investigate the effect of blade arc angle a set of single-stage rotors w i t h
p/q = 0.2 and 10% blockage were constructed. E a c h had a diameter of 704 m m
and a height of 542 m m giving the aspect ratio of 0.77, w i t h circular end plates
measuring 939 m m diameter. T h e projected area was « 0.37 m . The associated p

2
and q were found using equation (2.10) or (2.11).
T h e corrected Cp vs. A curves for 6 = 112°, 135° and 1 5 0 ° , presented in Figure
2-20, suggest the o p t i m u m value for 6 to be around 1 3 5 ° .
2.4-6 Reynolds number
T h e operating Reynolds number of the prototype turbine would be around

0.40
0.35H 1 fi-53 • n
s «i
\
L
' S
8 A
0.30 H •
R
0.25 H ..A
C p 0.20 H
0.15H
0.10
Corrected
A u
0.05 A
X u
• u
0.00 I I I I I I I
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3
F I G U R E 2-17 B e a r i n g loss corrected characteristic curves at three different

w i n d speeds for the two-stage (p/q = 1.6) model.
0.55-1
•
0.50 <• ^
• >> •>
0.45
>
N
-
j
c: c
\
0.40
0.35
•
[
a"*
x.--...
\
» %
3
0.30
r
•
/
0.25 . . 1I
0.20-
B = 16.4%
0.15- A p/q = 0.0
x p/q - 0.2
0.10- • p/cj = 0.4
B p/q = 1.0
0.05- H p/q = 1.6
0.00
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
A
F I G U R E 2-18 Effect of the blade geometry parameter p/q on the power

coefficient' at a constant blockage.
0.55-1
p/q
F I G U R E 2-19 V a r i a t i o n of the peak power coefficient w i t h the blade geometry

parameter p/q at a constant blockage of 16.4%.
f-
>
>
L\ L^ N <
^ .[: [ : c
D-
cy'
X3 N
[ \
-
p/q = 0.2
A e -112°
x 6 = 135°
• e -150°
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
A
F I G U R E 2-20 Effect of the blade circular arc angle 6 on the power coefficient
for single-stage model (p/q = 0.2, B = 10%).
Chapter II: Optimum. Configuration Studies 52
3 x 1 0 -- 9 x 1 0 . O n the other hand, most of the wind tunnel tests were carried
5 5
out in the Reynolds number range 1.7 x 1 0 — 4 x 1 0 . Fortunately, C

5 s
P vs. A plots
in this range suggest that the effect of the Reynolds number in the operating range
is likely to be insignificant.
2.4-7 Blockage ratio
W i n d tunnel results presented by different investigators often do not correlate
because of different test conditions. One of the major parameters affecting the test
data is the blockage.
To have some appreciation of the wall confinement effects, four single-stage and
one two-stage rotor models with an identical geometric shape but with different
blockage ratios were tested in the boundary layer w i n d tunnel. T h e results presented
in F i g u r e 2-21 clearly show a dramatic increase in the m a x i m u m power coefficient,
p r i m a r i l y due to an increase in the local velocity, with blockage. Note, an increase
in wall confinement from 5 to 20% can raise the C p m a i by around 70% thus leading
to a highly optimistic performance if the blockage effect is not corrected.
Figure 2-22a shows the variation of Cp max w i t h blockage (B) for p/q = 1. It
should be noted that when operating in the unconfined environment the power co-
efficient reduces to 0.2. To obtain similar information at other p/q values would
involve an extensive test program. However, of particular interest here is the cor-
rected power coefficient corresponding to the o p t i m u m p/q of 0.2. To this end, two
models w i t h blockage ratios of 10% and 16.4% were constructed. Recognizing the
fact that the associated wake aerodynamics remains essentially the same, the vari-
ation of Cp rnax w i t h blockage is expected to have the same trend. T h i s infers that
the Savonius rotor w i t h an o p t i m u m combination of parameters has an efficiency of

0.45
0.40-
0.35
-• .. K
XI"
0.30
0.25
•1— ~x
0.20-
0.15
p/q = 1.0
A B = 2.0%
0.10- x B = 5.0%
• B - 10.0%
ia B = 16.4%
0.05-
E B = 20.0%
0.00 1 1 1 1 1 1 1 I 1
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
F I G U R E 2-21 Effect of blockage on power output of the Savonius rotor.

0.6
0.5-
F I G U R E 2-22a Effect of blockage on peak power coefficient for two different

blade geometries.
F I G U R E 2-22b Effect of blockage on the tip-speed ratio at peak power coeffi-

cient for two different blade geometries.
about 32% in unconfined conditions.
T h e plot of tip-speed ratio (A) at the peak power coefficient vs. B% (Figure
2-22b) shows a linear variation. T h i s yields the zero blockage values of A as 0.71
and 0.79 for p/q = 1.0 and 0.2, respectively. T h e corresponding results of Maskell
(1965) for the flat plate blockage correction also showed a similar linear variation for
small B. However, the free air power coefficient predicted using Maskell's procedure
gives an error of about 25%. The considerably large side-force and the asymmetric
wake due to blade rotation, which had been neglected in the Maskell derivation may
account for this discrepancy.
T h e systematic optimization process has indeed proved to be quite rewarding.
B y conducting a series of carefully planned experiments it has been possible to
improve the rotor performance by a factor of two.
T h e o p t i m u m blade configuration (for the end plate parameter d/ddi

sc = 0.75)
can be summarized as follows:
a/d non-dimensional blade gap-size = 0 ;
b/d non-dimensional blade overlap = 0 ;
A blade aspect ratio = 0.77 ;
p/q blade shape parameter = 0.2 ;
6 blade arc angle = 135°.
2.5 P r e s s u r e D i s t r i b u t i o n
A few typical results showing pressure distributions over a single-stage Savonius
rotor w i t h o p t i m u m geometric parameters and 10% blockage are presented in Figure
2-23. T h e tap numbering scheme was indicated in Figure 2-8. T h e side of the blade
1.5
1.0 X
'45
0.5-
0.0-A A ' A A' A ^

AAA' A
A A A A A A A A A A A A A A
AAA
A A A A A A
-0.5-
x x
x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x
t
X .A
-1.0-1
: A
A.. .A
-1.5- A
-2.0-
/? = 0
A Front S i d e
-2.5-
x Back Side
-3.0 T
5 10 15 20 25 30 35 40 45
Pressure Tap No.
F I G U R E 2-23a Surface pressure distribution over the Savonius rotor as affected
by the blade orientation: 0 = 0°. Note, two separation points
(tap 0, 39) on the front (lower) surface and one on the back
(upper) surface (tap 11).
1.5
1-0 " A A A ' A A' A A A A A' A A ' A' A ' A " A A A A

' A
A A A A
0.5
0.0 x
A A A A A A A A
-0.5 x X X X X X X X X X X X X X X X X X X 5^ X X X X V V' A X X X X X X X X
A A
-1.0
-1.5
X X
-2.0
(3 = 30
A Front Side
-2.5-
x Back Side
•3.0 -r "T
0 5 10 15 20 25 30 35 40 45
Pressure Tap No.
FIGURE 2-23b Surface pressure distribution over the Savonius rotor as affected
by the blade orientation: 3 = 3 0 ° . Note, separation points at
tap 11 and 38 on the back and front surfaces, respectively.
1.0 A A A A A A A A A A A A A' A A A A A A A A A A A A
0.5
0.0
-0.5 x x x x x x x x x x x x
X X
X. XX. X X XX X X A. XX. X X X X X X X X X x
X
x
x X X X X
-1.0 A A A
A A A
-1.5
-2.0
p= 60
A Front Side
-2.5
x Back Side
-3.0 T " —r— -1

5 10 15 20 25 30 35 40 45
Pressure Tap No.
F I G U R E 2-23c Surface pressure distribution over the Savonius rotor as affected
by the blade orientation: 0 = 6 0 ° . Note, wake flow on the back
surface and the separation at tap 41 on the front surface.
1.5-
1.0 A A A A ' A A- A A- A A- A - A A A A A A- A A A A A' A A A A A
0.5 45
o.o-
-0.5-1 x: X x x x x"x:"*"x X X x x x x x x x x x x x x x x Y x x x x x x
X X X X x X x x x xx
A A
-1.0
A A
-1.5-
-2.0
p= 90
A Front S i d e
-2.5
x Back Side
-3.0 T
5 10 15 20 25 30 35 40 45
Pressure Tap No.
F I G U R E 2-23d Surface pressure distribution over the Savonius rotor as affected

by the blade orientation: 0 = 9 0 . Note near constant pressure
c
along the back surface and smooth pressure variation on the

front side.
1.5
1.0- A A A A
A A
A A
A A
•
A c
0.5- ^
0.0-
A
x* :
-0.5-' A
>
X
A X X X X >( X X X X >< X X X X >< X X X X )< X X X X >
' X x
x X x* X
•1.0
(
xx x x >
-1.5-
-2.0 _ ...
(3 = 120°
—
A Front Side
-2.5
o I x Back Side
-3.0
10 15 20 25
Pressure Tap No.
F I G U R E 2-23e Surface pressure distribution over the Savonius rotor as affected

by the blade orientation: 3 = 120". Note the absence of surface
flow separation.
1.5
1.0- A- A - 45
0.5-
0.0- A A A A A A A A A
A A
A A
A A A A A A A A A ^ A A A A
x x x x
-0.5 -x X X X X x X X X X X X X X X X X X X X X X X X X x X X X X x" x x x x x x x x
x
-1.0-
-1.5-
-2.0
(3 = 150
A Front Side
-2.5
x Back Side
-3.0
10 15 20 25 30 35 40 45
Pressure Tap No.
F I G U R E 2-23f Surface pressure distribution over the Savonius rotor as affected
by the blade orientation: /3 = 1 5 0 ° . Note near constant back
pressure and the separation at tap 15 on the front side.
facing the free stream is referred to as the front side while the opposite is called
the back side. Similarly, the upstream blade is called the leading blade and the
downstream blade is referred to as the trailing blade(Figure 2-8).
W h e n the blades are aligned w i t h the flow direction (/? = 0 ° , Figure 2-23a),
stagnation point occurs at tap 0 on the front side of the leading blade. Pressure
drops along the front surface reaching a m i n i m u m at tap 9 followed by a brief rise i n
pressure. Unable to negotiate the adverse pressure gradient, the flow separates at
around the 11 th
t a p . T h e flow remains separated and the pressure continues to be
essentially uniform at the separation value. O n the back side, the flow separates at
the t i p of the leading blade and reattaches i n the vicinity of the 2 3 r d

tap resulting
in a region of rise in pressure. Now the circular cylinder type pressure profile begins
on the trailing blade reaching a m i n i m u m i n the vicinity of the 3 3 r d

tap on the
backward curvature, showing separation near the 39 th

tap, and maintaining a near
constant value thereafter. Thus the flow separates twice on the back side and once
on the front side. T h i s pattern remains unchanged until about 0 = 3 0 ° .
Obviously the local flow is rather complex w i t h cause and effects often eluding
detection. A t /? = 30° (Figure 2-23b), pressure on the front side remains at the
stagnation value for the leading blade followed by a sharp drop. T h e reverse curva-
ture causes an adverse pressure gradient and the flow separates around the 38 th
tap.
T h e pressure on the back side decreases when the flow accelerates along the convex
surface of the leading blade, goes through the familiar variation and separates at
the \ \ t h
tap. Thus there are two separation points, one on the back side of the
leading blade and the other on the front side of the trailing blade.
A t (3 — 6 0 ° , the front side pressure remains at the stagnation value for the
leading blade followed by a drop and a p a r t i a l recovery on the trailing blade (Figure
Chapter 11: Optimum Configuration Studies 63
2-23c). T h e back side pressure remains essentially uniform suggesting a wake flow.
T h e same characteristics were observed for the range 40° < 0 < 8 0 ° . In this
configuration there are only two separation points, one at the tip of the leading
blade and the other at the front side of the trailing blade.
Separation points move to the tips at 0 — 90° (Figure 2-23d) as characterized
by the constant back pressure. The smooth pressure drop along the front side of
the trailing blade suggests accelerating flow and is followed by a partial pressure
recovery before separation at the trailing edge. Similar pressure variations were
observed for 0 up to 140°(e.g., refer to the pressure distribution at 0 = 120°. Note
the stagnation point near the 10 th

tap on the front side). T h e configurations for
3 greater than 90° are similar to those w i t h negative 0 values (e.g., 0 = 120°
corresponds to 0 = —60°). To keep the direction of tap numbering the same as the
flow direction, for 0 > 9 0 ° the equivalent negative angle has been considered.
A t 0 — 1 5 0 ° ( = - 3 0 ° ) , stagnation occurs in the vicinity of the 4 th

tap on the
front side and separates at the leading edge of the blade. Proceeding in the direction
of the flow, the pressure drops from the stagnation value, briefly rises, and becomes
nearly uniform suggesting the wake condition. T h e back side pressure remains
essentially constant indicating that this part of the blade is also in the wake (Figure
2-23f).
2.6 Starting Torque
T h e starting torque coefficient for the 10% blockage model w i t h o p t i m u m blade
geometry was measured at two w i n d speeds and is shown in Figure 2-24. Note,
the effect of Reynolds number is m i n i m a l substantiating the earlier observations.
T h e line represents the torque calculated through pressure integration assuming two
/'AXAxX
k* \ x . . . . . . ^ x\ y V
A A
tX
1
/ V
t
1
1 X
1
>
1
A
; /
legend
A R = 2.1x10
n
5
x R = 2.7x10
n
s
Calculated
n 1
1 • 1 1
1 1
1 ' 1 1
1 1
1 , —
i 1
1
0 20 40 60 80 100 120 140 160 180

(3 (Degrees)
F I G U R E 2-24 Variation of the starting torque w i t h angular position of the

Savonius rotor (p/q = 0.2, B = 10%). Calculated results based
on pressure data are also indicated.
Chapter ././» Optimum Configuration Studies 65
dimensionality. The deviation between measured and calculated values is within an
acceptable error margin (maximum error « 5%). T h i s suggests that the assumption
of two dimensional flow is reasonable for this class of models.
Figure 2-24 shows a positive torque over the range of /? « 0° — 130° followed by
the torque reversal. T h e peak positive torque occurs at around 30° suggesting that
this rotor is not exactly a drag device.
2.7 Error Estimation
T h e important parameters evaluated in this study are :
CT Torque Coefficient ;
Cp Power Coefficient ;
C p Pressure Coefficient ;
A Tip-Speed R a t i o .
T h e quantities actually measured and the possible ranges of errors were:
N ± 0.5 R P M of the rotor ;
/ ± 0.05 mm reading of the inclined alchol manometer ;
V ± 0.005 v voltmeter reading of the load measuring system ;
R ± 1mm rotor radius ;
h ± 1mm blade height ;
p ± 0.08 N/m 2
reading of the pressure transducer.
T h i s gives the m a x i m u m possible error margins as
\, 0.025 , .0046"
x
6X = ± (0.0032 + — — )A + — - 7
Chapter IF: Optimum Configuration Studies 66
which amounts to
6X = ± 0.0038 + 0.0135A for w i n d speed = 4.43 m/s ;
= ± 0.0025 + 0.0039A for w i n d speed = 6.76 m/s.
Similarly
bC r = ± 0.0016+ 0 . 0 7 5 1 C T
at U = 4.43 m/s ;
± 0.0007 + 0 . 0 3 7 0 C T at U = 6.76 m/s.
T h e m a x i m u m error margin for Cp is given by:
0.0016 0.0038
SCP = ± + 0.1052 C p , at U = 4.43m/s;
A
0.0007 0.0025
= ± + 0.0481 C p , at U = 6 . 7 6 m / s .
CT A
F i n a l l y , the m a x i m u m error limits for pressure coefficient C P can be written as
6C p = ± 0.0029 + 0.0286C , 7 at U = 6.76 m/s.
Thus the possible m a x i m u m errors in the measurements of tip-speed ratio,
torque coefficient, power coefficient and pressure coefficient decrease w i t h an in-
crease in w i n d velocity.
CHAPTER III
A SEMI-EMPIRICAL APPROACH FOR

PERFORMANCE PREDICTION
T h i s chapter discusses application of a semi-empirical approach to the Savonius
rotor performance prediction using the stationary blade pressure distribution data
presented in Chapter II. Such an approach, if successful, can replace dynamic testing
of models w i t h significant, reduction in cost, time and effort.
3.1 I n a d e q u a c y o f t h e Q u a s i - S t e a d y A p p r o a c h
Consider the classical quasi-steady approach which has been successfully applied in
the analysis of the Darrieus rotor as well as various other fluid dynamics problems.
Here the blade is divided into a finite number of elements assuming two dimen-
sional flow. For a given angular position of the blade, contributions of the free
stream velocity and the.bla.de rotation at each element is determined using velocity
triangles (Figure 3-1). A s shown in Figure 3-1, ( r ; , ^ ) are the polar coordinates at
the centre of the i th

element w i t h respect to the diameter of the rotor, ifci is the
angle between the normal vector and the radius vector. ,/? is the blade angle with
t
respect to the relative velocity V - at the i

t
th
element, and OJ is the angular velocity of
the turbine. T h e mean differential pressure coefficient obtained experimentally, at
the i th
element for the blade angle f3 = /?, was used in conjunction w i t h the relative
velocity Vj to evaluate the force acting on this element. In the present study, pres-
67
Chapter III: Semi-Empirical Approach 68
F I G U R E 3-1 A typical velocity triangle at the blade element i combining the

effect of free stream velocity and blade rotation.
Chapter If]: Serni-Empirical Approach (59
sure distribution was measured at 10° intervals (section 2.5). Linear interpolation
was used to estimate the pressure coefficients at intermediate locations. Using an
appropriate moment a r m for the blade element, integration over the blade gave the
torque for a specified position /? of the rotor. Finally, by evaluating the work done
over a cycle, the mean power and power coefficient were established. Repeating the
procedure for different angular velocities the effect of tip-speed ratio on the power
coefficient was estimated.
T h e results obtained using this procedure are compared w i t h the corresponding
experimental results (for p/q = 0.2, B — 10%, 6 = 135° model) in Figure 3-2.
T h e large descrepancy between results suggests that the classical quasi-steady ap-
proach is not successful in predicting the Savonius rotor performance. This may
be attributed to disparity in the flow features associated w i t h the stationary and
rotational modes of the blade. T h e method, however, proved effective with the
Darrieus rotor, where the flow character does show a degree of similarity. In the
present case, since the relative direction of the flow varies significantly along the
blade, interactions between elements of the rotor cannot be neglected. Unfortu-
nately, the classical quasi-steady approach does not account for these interactions.
Thus this approach cannot, be used effectively to analyse the Savonius rotor.
3.2 M o d i f i e d Q u a s i - S t e a d y A p p r o a c h
Failure of the classical quasi-steady procedure emphasized the need for an alterna-
tive. It was apparent that clear understanding of the flow character was a prerequi-
site to approach the problem. To better appreciate the basic character of the flow
around the Savonius rotor a flow visualization study was undertaken.

X
*x
X
legend
Q-S Method
x Experimental
0.0 0.3 0.6 0.9 1.2 1.5

A
FIGURE 3-2 Results of the quasi-steady analysis. Note, the approach is

woefully inadequate in predicting the actual performance.
Chapter III: Serni-Empirical Approach 71
8.2.1 Flow visualization study
T h e flow visualization study was carried out using the tow-tank facility located
in the plasma physics laboratory of Professor B . A h l b o r n . T h e tank is 5 m long and
has a cross-section of 82 c m X 62 c m (Figure 3-3). A Savonius rotor model 20 c m in
diameter and 30 c m high was used in the test. A 35 m m still camera and a video
camera moving w i t h the model captured the flow patterns. T h e still camera had a
motor drive so that a series of pictures could be taken at a rate of two to four frames
per second. T h e model was towed at a desired uniform speed using a step pulley
combination driven by a half horse power constant speed motor. Towing speed
could be varied from 2 to 200 cm/s. T h i s gave a Reynolds number of the order 1 0 . 3
Towing speed was estimated by electronically measuring the elapsed time between
two points 10 c m apart. Safety switches were placed at either end of the cartway to
guard against overshoot.
T h e Savonius rotor model was mounted on an a l u m i n u m frame using two ball
bearings. A pulley assembly and a flexible shaft connected to a variable speed
electric motor was used to t u r n the model at a desired constant angular velocity.
T h e rotor speed was measured using a " S H I M P O " noncontact type tachometer.
For the surface flow investigations, a l u m i n u m filing tracers of « 0.5 m m size
were applied by scraping an aluminum block w i t h a file. T h e tracers were illumi-
nated w i t h four flood lamps placed at a low angle of incidence w i t h respect to the
fluid surface in order to avoid undesirable reflection of the surface waves. Special
attention was paid to the fluid surface condition. Periodic s k i m m i n g of the sur-
face was needed to remove static electrical charges and contaminants, which altered
properties of the surface significantly.
A t higher tip-speed ratios, presence of large surface waves disturbed the flow sig-
Still/Video
Camera
Camera
Carriage
Flood
Lights
T — R q i |
—r
Water Surface
Model Seeded with
Aluminum Filings
Model
Carriage
• Speed
o Control
a
F I G U R E 3-3 A schematic diagram showing the towing tank facility and

associated equipment used in the flow visualization study.
Chapter III; Semi-Empirical Approach 73
nificantly thus affecting the simulation. In order to obtain better flow-visualization
pictures the motor drive was removed and self-driven blades were used.
A typical set of pictures obtained are shown in Figure 3-4. Perhaps the most
significant feature is the central vortex filament in the time averaged flow. The
existence of this vortex has also been reported by Jones, L i t t e r and Manser (1979).
Several other relevant features are discussed later in Chapter V I .
8.2.2 Central vortex approach
T h e flow visualization study provided a rational modification of the classical
quasi-steady approach to the problem by introduction of a central vortex filament.
T h e flow is simulated by a potential vortex w i t h a core radius S\R, a core angular
velocity S OJ
2 and a uniform free stream velocity U. Here R and ui represent radius
and angular speed of the rotor, respectively and S\, S are two empirical parameters.
2
T h e velocity distribution of the vortex filament is shown in Figure 3-5.
T h e relative velocity at each element is evaluated using the modified flow field
and the angular speed of the rotor. Force at the individual element is evaluated
using the differential pressure coefficient and the velocity of the resultant flow as
discussed in section 3.1. The procedure leads to evaluation of power coefficient Cp
in terms of tip-speed ratio A, Si and S 2 .
8.2.3 Empirical relations
Si and 5 2 were systematically varied and the corresponding Cp estimated over
a range of tip-speed ratio A = 0.6 — 1.5. T h e results obtained, when compared to
the experimental results presented in section 2.4, suggested the functional relations:
F I G U R E 3-4 A typical set of flow visualization pictures showing the presence
of the central vortex filament.
Chapter III: Semi-Empirical Approach
F I G U R E 3-5 Velocity distribution due to the potential vortex.

Chapter 111: Semi-Empirical Approach
Cp -;- C (Si, P S2) (3.1)
5, • FAX) +F {B,R )
2 n
(3.2)
S 2 •• S {p/q,R )
2 n (3.3)
as logical. T h e effect of tip-speed ratio could be absorbed in Si while 52 showed a
direct dependence on the variation of the geometric parameter p/q.The remainder of
Si , when varied, appeared to model the blockage effect. Influence of the Reynolds
number (R ) n is included here in both S\ and however, as seen before it is
insignificant in the operating range. Using the experimental results for p/q — 0,
B — 16.4% (Figure 2-18), F\(X) can be expressed as,
Fi(X) = 0.27778A - 1.08333A 2

(3.4)
w i t h corresponding values for F and 52 as 1.7 and 1.25, respectively. According to

2
the suggested functional forms, F = 1.7 combined w i t h different 52 should produce

2
a set of results corresponding to performance of constant blockage (l6.4%)models
w i t h different p/q values. A s shown in Figure 3-6, the calculated results for 52 =
1.25, 1.3, '1.37. agree w i t h experimental curves of p/q — 0, 1.0, 0.4 at B = 16.4%,
respectively when F 2 is kept constant at 1.7. Similarly, when 52 is kept constant
at 1.3 (corresponding to p;q -- l ) and F 2 is varied, this should simulate results
for p/q = 1 models with different blockage ratios. T h e results shown in Figure
3-7 verifies this prediction. Now it is evident that the above mentioned functional
forms can be used to predict, the performance of Savonius rotors, if the relationship
between F and blockage ratio, and 52 and p/q can be established for the operating
2
R n range. A s shown in Figures 3-6 and 3-7, there is a slight descrepancy between
the experimental and calculated results at lower tip-speed ratios. T h i s may be due
i
to the fact that the pressure data for p/q — 0.2 were used to predict the performance
0.5
0.4-
F = 1.7, B = 16.4%
2
MQSM S = 1.30
x
MQSM . S a = 1-37
Exper. .PA=.°.
Exper. _PA=A
Exper. p/o = 0.4
t
"1
1.5
X
F I G U R E 3-6 Identification of the modified quasi-steady approach parameter 5 2
as a function of p/q when i<2 and blockage are kept constant.

0.5-1
0.4
0.3-
0.2-
S 2 = 1.3, p/q = 1.0

. 2 - i v v
MQSM F a = 1.65
0.1-
MQSM r = i.7o2
Exper. 6 =5%
Exper. B = 10%
Exper. B = 16.4%
0.0
0.5 0.7 . 0.9 1.1 1.3 1.5
X
F I G U R E 3-7 Identification of the modified quasi-steady approach parameter F 2
as a function of blockage ratio when 5 and p/q are kept constant.

2
Chapter ill: Semi-Empirical Approach 79
of other geometries. However, at moderate tip-speed ratios the effect of the central
vortex is more dominant and the results may not be affected significantly by a slight
change in pressure data.
T h e agreement between calculated and experimental results shown in Figures
3-6 and 3-7 is used to establish 5 2 and F 2 in terms of p/q and B, respectively.
T h e empirical curves established for 52 and F for the operating Reynolds number
2
range are shown in Figure 3-8 and Figure 3-9. T h e variation of S 2 w i t h p/q shown
in Figure 3-8 resembles the Cp >max vs. p/q curve (Figure 2-19). Similarly, F 2 vs.
B curve (Figure 3-9) has the same basic shape as the C p , m a i vs. B curve shown
in Figure 2-22. T h i s indicates that 52 and F 2 are directly related to the m a x i m u m
power coefficient Cp .
imax
3.3 R e s u l t s a n d G e n e r a l R e m a r k s
Cp vs X curve for p/q = 0.2 and p/q = 1.6 at B = 16.4% were obtained experimen-
tally in section 2.4.4 and were not used to establish the empirical relations in the
modified quasi-steady approach. Similarly, the experimental results for p/q = 0.2,
B = 10% were not involved i n the development of the method. Therefore, these
three cases can be used to verify the validity of the above approach. T h e com-
parison of the predictions given by the modified quasi-steady approach w i t h the
experimental results for these three cases are shown in Figure 3-10.
From engineering design considerations, the agreement may be considered quite
acceptable although the results are under-predicted at lower tip-speed ratios.
Superposition of the central vortex filament assumes a potential flow field. B u t
the time-dependent shear layers, separated flow and high turbulence levels suggest
that the.flow field is far from being potential. Furthermore, the wake is not con-
F I G U R E 3-9 E m p i r i c a l relationship between F

2 and the blockage ratio B.
0.6
0.5-
0.4-
0.3-
legend
0.2
MQSM F a = 1.70, S a = 1.45
MQSM F 2 = 1.65, = JL45_
MQSM F a = 1.70, S a = 1.26
Exper. B = 16.4% p/q_ = 0.2
0.1
Exper. B = 10% p / q := 0.2
Exper. B = 16.4% p/q_ = 1.6
0.0 T
0.5 0.7 0.9 1.1 1.3 1.5
A
F I G U R E 3-10 Predictions of the modified quasi-steady approach compared to

the experimental results.
sidered at all in this analysis. However, w i t h i n its limitations, the method permits
prediction of model performance in the wind tunnel as well as that of the prototype
in unconfined condition with a reasonable accuracy.
A more sophisticated mathematical model considering the unsteadiness of the
flow, shear layers, vortex shedding and the wake is discussed later i n Chapters I V ,
V and V I .
C H A P T E R IV
A MATHEMATICAL MODEL FOR

ROTOR ANALYSIS
W i t h some appreciation as to the character of the complex flow associated with
a Savonius rotor and-significant parameters affecting it, a more sophisticated ap-
proach to its analysis is attempted here. T h e problem is indeed formidable, being
characterized by viscous, turbulent, separated and time-dependent behaviour of the
flow, and will demand application of the complete Navier-Stokes equations. Even
if a suitable a l g o r i t h m is developed to account for the above mentioned complexi-
ties, appropriate boundary conditions and mathematical consistency, the numerical
solution w i l l be enormously time consuming w i t h the available computer ( A m d a h l
5850) and equally costly. One is, therefore, forced to turn to an approximate ap-
proach which makes the problem amenable to the numerical tools at hand without
sacrificing its essential features.
A n approximate method of solution of the Navier-Stokes equations in the oper-
ational range of the Reynolds number (3 x 10 — 9 x 10 ) would involve use of an

5 5
appropriate turbulent flow model and an integration scheme. Since the turbulent
scales are relatively small around this Reynolds number the grid required in finite
element and difference integration schemes becomes extremely small thus adding to
the computational cost.
Several such approaches, applicable to specific geometries have been reported in
84
Chapter IV: Mathematical Analysis 85
literature. They involve time stepping, finite element or difference solutions of the
Navier-Stokes equations through E u l e r i a n (Jordan and F r o m m , 1972), Lagrangian
( C h o r i n , 1973; Stansby and D i x o n , 1983) and hybrid (Eulerian approach in the
boundary layer region with Lagrangian procedure in the potential flow regime,
Spalart et al., 1983) schemes for simple geometries like circular cylinders. Since
the boundary layer is simulated, the flow separates on its own and the point of
separation need not be specified. Unfortunately, these methods are not applicable
to bluff bodies w i t h geometrical irregularities and sharp edges. Furthermore, they
often suffer from computational instability or require finite element or difference
grid of impracticably small size to obtain reliable results at high Reynolds numbers.
T h i s chapter briefly describes the basis of an approach which, through judicious
simplifications, arrives at a mathematical model that can predict the performance
of the Savonius rotor w i t h an acceptable accuracy.. Since the rotor possesses sharp
edges and hence fixed initial separation c o n d i t i o n , the difficulty of having a very fine
grid for stable results is overcome through the use of a Boundary Element M e t h o d
( B E M ) as against the F i n i t e Element M e t h o d or F i n i t e Difference M e t h o d ( F E M
or F D M ) .
4.1 B o u n d a r y E l e m e n t M e t h o d ( B E M )
T h e essence of this technique is the transformation of the governing differential
equations into an equivalent set of integral equations as a first step to.their solution.
T h i s would result in a set of equations which would involve only values of the
variables at the extremes of the range of integration (i.e., boundaries of the region).
Thus any discretization scheme needed subsequently would only involve subdivisions
of the boundary surfaces of the problem (Banerjee, 1981). In contrast to the F E M ,

Chapter IV; Mathematical Analysis 86
which involves subdivisions of the whole flow field, the B E M proves to be less
expensive in terms of computer time.
A l l boundary integral methods utilize the principle of superposition and there-
fore are applicable to linear systems or those which can be approximated as incre-
mentally linear (Hess et al., 1962-76; Zienkiewicz, 1978; Banerjee et al., 1979; M a r -
j a r i a et a l . , 1980). In problems involving elliptic differential equations the solutions
are direct, whereas for parabolic and hyperbolic systems of equations, marching
processes in time have to be introduced (Banerjee, 1981).
In general, there are only a few problems solvable by a F E M which cannot be
solved at least as efficiently by a B E M . These are the problems where:
i) the material properties of almost every individual element are different;
ii) those in which the general geometry of the problem is such that one or
more dimensions are smaller compared to others but not sufficiently so
as to genuinely reduce its effective dimensionality (e.g., moderately thick
plates and shells, narrow t h i n strips, etc.).
A B E M reduces the dimensionality of the basic problem by one, i.e., for two
dimensional problems the analysis generates a one dimensional boundary integral
equation. Similarly, for three dimensional problems only two dimensional surface
integral equations arise. The method proves to be more advantageous compared
to the F E M for problems w i t h some boundaries at infinity. Since the B E M solu-
tion procedure automatically satisfies admissible boundary conditions at infinity,
no subdivisions of these boundaries arise. O n the other hand w i t h the F E M infinite
boundaries have to be approximated by an appreciable number of distant elements.
T h e indirect formulation of the B E M (Banerjee, 1981) can be easily applied to

Chapter IV: MatharmUtdl Analysis 87
the Savonius rotor analysis. In this formulation, the integral equations are expressed
in terms of a unit singular solution of the original differential equation distributed
at a specific density over the boundaries of the region of interest.
4.2 M a t h e m a t i c a l F o r m u l a t i o n
4.2.1 Basic assumptions
T h e physical problem involves a time-dependent separated flow at high Reynolds
number. B u t for the purpose of analysis a simplified model is considered which
partly accounts for the effect of viscosity. It is assumed that the influence of viscos-
ity is confined to an infinitely thin layer of fluid adjacent to the solid surface of the
rotor (i.e., boundary layer of zero thickness which does not affect the external invis-
cid flow), and vorticity is contained w i t h i n "wake sheets" issuing from sharp edges.
These "wake sheets" correspond to physically identifiable regions of rotational fluid
which transport the vorticity generated by viscous effect w i t h i n the boundary layer
on the blade. T h e vorticity generated is carried away downstream w i t h i n a sheet
of fluid whose shape is such that the pressure difference across it, and hence the
difference in velocity magnitude, is zero at all points on its surface. T h e velocity
is discontinuous across the shear layer. These shear layers roll up to form strong
vortices which shed downstream w i t h the flow.
In practise, the effect of viscosity and diffusion lead to finite cores of rotational
fluid.
4.2.2 Governing equations
T h e problem in hand is to estimate performance of the Savonius rotor under
various conditions. W i t h the above mentioned assumptions, torque a n d power gen-

erated can be determined uniquely from the velocity field which can be considered
as the only unknown of the problem.
T h e governing equations for the idealized incompressible, inviscid, irrotational
two-dimensional flow outside the wake surface can be written as:
£ H<.' .'>=/<«)
+ + ; <«>
where <& is the velocity potential; p, the pressure; p, the density; / ( £ ) , an arbitrary
function of time t; and u, v are the velocity components in the x and y directions,
respectively. The velocity components (u,v) can be obtained from the potential
V = V$,
or
ay
It is necessarj^ to specify conditions on the boundary 5 of the domain of interest
fl (Figure 4-1) in order to solve for <fr, which is the only unknown. T h i s falls into
the general class of boundary value problems.
The boundary condition at the blade surface specifies that the flow normal to
the surface be zero,
^ - - " „ = 0, • (4.4)
on
where n is the vector normal to the blade surface and v n is the normal velocity of
the blade.
F I G U R E 4-1 General description of the problem.

Theoretically, the wake extends indefinitely downstream of the body. However,
due to viscous dissipation, it effectively ceases to exist beyond some characteristic
distance downstream. In the present analysis, the wake is artificially terminated at
a large but finite distance downstream such that this truncation produces negligible
effect on the body and the near wake.
B y virtue of the definition of S, the potential in the idealized physical problem
is single valued and continuous throughout fi and S (although it is possible to
include point singularities outside the domain O ) .
4-2.3 Uniqueness of the solution
T h e uniqueness theorem for a function $ satisfying Equation (4.1) throughout
fi can be expressed as:
If at each point of the boundary of a given region either $ or its normal
derivative d$jdn is known (Dirichlet, Neumann problem) and if § is
single valued, then there exists at most one function $ satisfying these
conditions, unless the value of $ is nowhere known, in which case $
is indeterminate to the extent of an additive constant (Hunt, 1980).
Suppose a mathematical domain fi m w i t h boundary S m geometrically identical
to that of the physical problem (Figure 4-1) is constructed, then the potential $ m
w i l l be identical to $ at all corresponding points of fi m and fi (at least to w i t h i n
an additive constant) provided the following condition is satisfied:
the potential 5> m and $ are identical at all corresponding points of
selected regions S$> (where <& is known) of the surface S m and S, and
(d<& jdn) m and d<&/dn are identical at all corresponding points of the
Chapter TV: Mathematical Analysis 91
remainder Sn (where d$jdn is known ) of these boundaries (Hunt,
1980).
It is important to note that in principle any mathematical model which satis-
fies the appropriate conditions may be used, even though the model may bear no
physical resemblance to the flow.
In the so called panel methods, the exact mathematical expressions derived in
the following section are approximated. Note that it is not necessary to consider
singularity distribution on the boundary surface in order to simulate the external
flow field. For example, a point doublet of strength 27r, set w i t h its axis parallel to
a uniform flow of unit magnitude, produces a velocity field identical to the potential
flow about a sphere of unit radius, in the region outside of the sphere.
It follows that use of the geometric surface for the purpose of singularity dis-
t r i b u t i o n does not act as a source of error in the mathematical formulation of the
problem(Hunt, 1980).
4-2.4 Indirect Boundary Element formulation
T h e fundamental solution of E q u a t i o n (4.1) relates the potential $(x) generated
at a field point x to a singularity of strength 7 ( f ) applied at a load point £. Note,
here x and £ are coordinate vectors. A l t h o u g h the origin of coordinates x and £ are
identical, it is absolutely necessary to reserve each set for a specific purpose. T h e
classical singular solution can be written as,
<D(x) = G ( x , £ h ( £ ) , (4-5)
where G(x, £) is the 'two-point' function (generally known as the Green's function)
involving the coordinates (x, £) of the two points.

By differentiating $ with respect to x the velocity vector V ( x ) is obtained,
Let dG/dx be H ( x , f ) , then
V(x) = H(x,eh(0. (4-6)
T h e Green's function G ( x , £) and the principle of superposition can be used to
construct the solution to the physical problem i n terms of an integral equation.
T h e field points on S are designated by x s while £ locate load points on S. Now

s
singularities of unknown intensity 7 ( £ ) per unit length of S are introduced. T o

8
complete the problem definition any specified singularity distribution T(c) per unit
area in Q is also included. T h e net result of the 7 ( £ ) and T(c) distributions at any
s
field point x can be determined by integrating the fundamental solution,
*(x)= / G ( x , £ h ( £ ) r f S ( £ ) + / G(x,f)r(f)df7(c) + C .
s s s (4.7)
Js Jn
For existence of the solution it should be insured that the net flux across the infinite
boundary is zero. T h i s is achieved when
/ i(L)dS(e )+ s [ r ( ) r f n ( c ) = o.
f (4.8)
Js Jn
Similar operations using H ( x , £ ) would yield the velocity at any field point x ,
s
v(x) = / H ( x , £ b ( £ H S ( £ ) + / H(x,f)r(<r)dn(f).
s s s (4.9)
Js Jn
Chapter TV: Mathematical Analysis 93
In principle, the only remaining formal step to arrive at a solution to the problem is
to bring the field point x onto the boundary S. W i t h this, Equation (4.7) becomes
(x )=
B / G ( x , £ b ( £ ) d S ( £ ) + / G{x ,t)T{t)dn[$)
B B B 8 B + C. (4.10)
Js Jn
Similarly, from E q u a t i o n (4.9), velocity normal to the boundary S is given by
v M=
n I F(x ,£ b(£ )<2S(£ )+
Js
8 8 s s / ^(x ,f)r(f)dfi(f),
Jn
s (4.11)
where
Fix £ ) =
'(x,^ j s ^ ,
and f s represents an improper integral due to the singularity in F as £ —• x . s s
In a well-posed problem, either $ ( x ) or u(x ) s s is specified at every point of the
boundary and Equations (4.10) and (4.11) are simultaneous integral equations which
can be solved for the only unknown 7 ( £ ) . Once 7 ( £ ) is determined, Equations (4.7)
s s
and'(4.9) yield 3>(x) and V ( x ) , respectively, at any field point x of interest. It is
important to notice that Equations (4.10) and (4.11) are scalar integral equations
since the kernel functions (G,F), the singularities ( I \ 7 ) , and therefore are
scalar quantities.
If it is possible to integrate Equations (4.10) and (4.11) in closed forms and
solve for 7 then the solution w o u l d be exact; however this is practically impossible
and approximations have to be introduced. In the B E M inaccuracies arise from, and
only from, numerical discretization and integration procedures. Thus by refining
the approximations,, theoretically, any degree of precision is achievable. In practise,
there is a trade off between computing time, effort and solution accuracy. Several
different algorithms can be used to obtain worthwhile practical results.

Chapter IV: Matherriatical Analysis 94
4.3 D i s c r e t e V o r t e x M e t h o d ( D V M )
In essence, this method is a version of the B E M where a point vortex is used as the
distributed fundamental solution of E q u a t i o n (4.1). T h e approach is physically more
meaningful as a vortex shedding model is incorporated in the analysis. Equations
(4.7) to (4.11) are discretized so that a numerical computational scheme can be
utilized.
In general, in the discrete vortex method, the physical plane is mapped con-
formally into a circle and analysed in the transformed plane. However, a complex
physical configuration makes determination of the mapping function difficult. Fur-
thermore, even if a complex mapping function is obtained, the rotating character
of the system would require derivatives of the function thus reducing the inherent
advantages of the mapping technique.
4-3.1 Model geometry
In the present study, a Savonius rotor with t h i n blades is considered. A s shown
in Figure 4-2 the basic geometry is defined by the parameters: p/q - shape param-
eter; 0 - size of the arc portion; R - radius of the rotor; and A - the blade aspect
ratio. The blade contour is divided into m equal straight line segments (elements,
Figure 4-3). A l l dividing points including the blade edges are considered as control
points ( m - f 1 nodal points). U n k n o w n discrete bound vortices T i(i s — l,...,m),
are located at the mid-points of each segment (singular points). The origin of the
coordinate system coincides with the centre of rotation.
T h e blade angular position is given by angle f3, whereas r , 7 { (i = 1,...

s t S , m)
and r i,
n 7 - (i = 1,..., ra+ l ) represent the polar coordinates of singular and nodal
m
points, respectively, when /? is zero (Figure 4-3). These coordinates are obtained in
FIGURE 4-2 Simpified geometry of the Savonius rotor under investigation.

Chapter IV: Mathematical Analysis
F I G U R E 4-3 Description of the problem parameters.

terms of the number of elements and geometric parameters. In addition, the angle
between the radius vector and the outward normal to the blade contour is calculated
at all nodal and singular points, i.e., [i — 1,. • • ,m + l) andip i (i = 1,. . . , m ) .

s
For simplicity, the model reference frame is treated as complex Z - p l a n e . The blade
rotates in the positive j3 direction (counterclockwise) and the uniform flow is in the
negative x-direction.
4.3.2 Flow model
As discussed earlier, the flow past a bluff body separates at the sharp edges
and the shear layers are shed downstream from the separation points. To represent
starting shear layers an unknown free vortex is placed close to each separation
point. These points are known as nascent points (Figure 4-3). Introduction of the
first free vortices can be achieved in several ways. One approach would be to fix
the positions of the first nascent vortices and make the strengths of the vortices
unknown (Kuwahara, 1973; K i y a and A r i e , 1977). Alternatively, one may specify
the strengths of the vortices and to make their positions unknown (Sarpkaya, 1975).
In this study the former approach is used.
In the discrete vortex method, it is easier to carry out the analysis in the
complex plane. A fundamental solution to E q u a t i o n (4.1) can be given as a bound
vortex
Wi{z) = ^-T ln{z-

ei z ),
st (4.12)
where
Wi{z) = (4.13)
and clockwise T is considered'positive. W^(z) is the complex potential at z due
to a bound vortex of strength T i at Zg . 5>i and ^ , are the potential and stream
s %
functions, respectively. Thus the Green's function in E q u a t i o n (4.5) has the form
G{z,z ) sl = J-ln(*-*„•). (4.14)
Now E q u a t i o n (4.7) can be written for the flow after k time-steps in the discretized
form as follows,
m 2k
W{z) = ]Tr ln(z - z) s i si + ^~Y1 T

WJ \n{z - z ) wj + C, (4.15)
1=1 y=i
where C is a constant at a given time instant. Note, the $ in E q u a t i o n (4.7)
corresponds to the real part of W(.z). Similarly, E q u a t i o n (4.9) can be written as,
~y -^—, + — y
m r . 2fc _ Twi
w{z) = T l (4.i6)
V
^ 27T {Z - 2 s i ) 27T ^ ( 2 - 2 u ; y) V
'
Here w(z) is the wellknown complex velocity in the potential flow theory,
w(z) = u — iv, (4-17)
where u and v are the velocity components in x and y directions, respectively.
B y looking at Equation (4.16) it is evident that w{z) —» 0 as z —» oo. B u t
the model requires a uniform velocity — U in ^-direction at infinity. T o satisfy this
condition, E q u a t i o n (4.16) and hence E q u a t i o n (4.15) were modified i n the following
way:
m „ . 2k -p
w{z) = — y- s i
, +—y -1/ ; (4.18)
.V ;
2TT^(Z-Z ) SI 2TT^(Z-Z ) WJ
V ;
m . 2fc
^(^ = R
« ( - *«) + T~ Yl ^ ( -
LN 2 T ln 2
- +- Uz c
(- ) 4 19
. t'=l • . j \
Now the only remaining step is to apply E q u a t i o n (4.18) at the (m + 1) nodal points
of the blade contour. A t the begining of the k th

time-step this equation contains
m - f 2 unknowns r s i (i --- 1.. . . , ro) and T j

w (j — 2k — 1,2A;). B y satisfying the
boundary condition that the velocity normal to the blade at the nodal points is
zero:
(4.20)
<$>ni — tpni + Ini + 0 • angle between the x-axis
and the normal vector at the i th

node;
u! — angular velocity of the rotor;
(m -+ 1) simultaneous linear algebraic equations are obtained. E q u a t i o n (4.8) yields
the remaining relation
(4.21)
which is also known as K e l v i n ' s theorem: the sum of strengths of all bound and free
vortices reduces to zero.
B y solving Equations (4.20) and (4.21), the strengths of m bound vortices and
two nascent vortices are determined, w h i c h , in t u r n , through Equations (4.18) and
(4.19) establish the flow field.
4-3.3 Vortex shedding model
A t any instant t. the strengths of m b o u n d vortices and two nascent vortices are
found by solving Equations (4.20) a n d (4.21). D u r i n g the incremental time interval
At, the shed free a n d the nascent vortices, move w i t h the velocity induced by other
vortices and the uniform flow at their respective locations, and the rotor advances
by an angle to At. A t the subsequent instant another set of unknown vortices are
placed at the singular and the nascent points, and their strengths determined by
solving the ( m + 2) linear simultaneous equations.
T h e position of the I th
vortex shed from the separation point is advanced in a
small time interval At by the first order scheme:
z {t + At) = z (t)
wl wi + w{z i)At-w (4.22)
• m p . 2k „
27T ^ ( 2 u ; / - Z) sl 27T ^ (z w / - Z )
wj
where w(z i)
w is the complex conjugate of w(z i), w the complex velocity at the I th
free vortex due to all vortices except the I th

vortex itself and the uniform flow. A n y
interception of a vortex by the blade is avoided as discussed later.
T h e unit inviscid vortex expressed in E q u a t i o n (4.14) induces a velocity whose
magnitude is inversely proportional to the distance from its centre. Therefore, the
vortices which approach each other too closely aquire extremely large velocities.
To avoid this problem a rotational core of radius o is imparted to the free vortices
while calculating velocities. In the core region the velocity is given by:
w(z) = { 2
tr
^ f - i ^ (4.24)
27TC7
W h e n a shedding vortex gets too close to the blade surface, to satisfy the zero
normal flow condition, an equal and opposite image vortex will be generated on
the boundary surface, thus cancelling the effect of its presence. Physically this
cancellation is due to the viscous dissipation (Fage and Johansen, 1927; Ogawa,
1984). In the present study the l i m i t i n g spacing between a vortex a n d the blade
surface for this cancellation is considered as a.

In real flow the core of a vortex grows w i t h time. Giesing (1969) and Nagano
et a l . (1981), i n their studies w i t h flat plates and cylinders, suggested that o be of
the form
a = 2.24v î ,
/ 7
(4.25)
where v is the kinematic viscosity and V is the time elapsed since the vortex
was shed from the separation point.
To reduce the computational cost, several investigators in their studies w i t h flat
plates and cylinders have suggested to replace the free vortices i n a given cluster
by a single equivalent vortex when located more than two diameters downstream
(Sarpkaya, 1975). However, i n this singular point vorticity model, it is important
to preserve the mechanism of vorticity dissipation which i n the above procedure
is affected. Thus the cancellation effect of vorticity when two vortices of opposite
circulation come close to each other is minimized. Hence, in the present study,
representation of a cluster by a single vortex is avoided.
In real flows only a fraction (=s 60%) of the circulation fed into the shear layers
is found i n the vortex clusters or in the concentrated vortices of the K a r m a n vortex
street ( M a i r and M a u l l , 1971). T h e loss of vorticity as a result of coming too
close to the blade surface, and a small amount of cancellation between elementary
vortices of opposite sign which come close to each other (Clements, 1973) are the
only two mechanisms that could bring about some loss in vorticity, in this approach.
Evidently these two mechanisms are not sufficient to account for an approximately
40% loss of circulation. Therefore Clements (1973), through his study w i t h flat
plates, concluded that the mechanism by which a large amount of vorticity is lost
must be due to viscosity.

Chapter IV: .Mathematical Ana-lysis 102
To account for this Stansby (1985) suggested a vorticity dissipation model
T{t + At) = T{t)e- > , r At

(4.26)
where rj is a balancing parameter. Increasing rj increases the rate of shedding of
vorticity into the wake by reducing the vorticity in the wake. In the present study
the above model is used and the effect of different values of rj evaluated (Stansby
suggested n = 0.03).
4.3.4 Pressure, torque and power
Once the flow field is k n o w n , the corresponding pressure distribution and hence
the torque and power produced by the rotor can be evaluated quite readily. Pressure
distribution along the blade can be obtained by applying the unsteady flow Bernoulli
E q u a t i o n (4.2). T h e pressure coefficient defined in E q u a t i o n (2.13) can be written
as
n P ~ Poo
w\\l \ o2 ^5 $ ( - )
2 4 2 ?
1
' U J U 2
dt
where |tu| is the magnitude of complex velocity.
M u l t i p l y i n g the pressure by elemental area the force exerted on each element
and hence the elemental torque, can be found. Integrating over the blade the total
torque at a given instant can be calculated.
Alternatively, by applying the generalized unsteady flow Blasius equation (Imai,
1974),
Chapter fV: Mathematical Analysis 103
where: C represents an arbitrary path around the rotor excluding singular points;
B is a closed loop along the blade contour; and W is as given by E q u a t i o n (4.19).
Average power generated over a period V* is now given as
(4.29)
Torque and power coefficients are calculated as defined by Equations (2.8) and
(2.4), respectively.
4.4 Modelling of the Wind Tunnel Blockage
T h e mathematical model discussed above can be extended quite readily to account
for the blockage imposed by the w i n d tunnel walls as shown in Figure 4-4. The two
walls are placed symmetrically, at a distance H from the centre of the rotor. The
blockage ratio in a two-dimensional case can be defined as
d
B =
2E"
where d is the diameter of the rotor. E a c h wall was divided into a finite number
of elements (Mw) and unknown concentrated singularities placed at the centre of
each element. T h i s adds 2Mw additional unknowns to the problem. To satisfy
the K u t t a condition at the downstream end of the wall, an unknown fixed nascent
vortex is placed at a distance e from the end of each wall. Upstream ends of the
£
walls being nodal points, automatically satisfy the tangential flow condition. The
formulation results in 2(Mw + l ) additional unknowns. T h e boundary condition
of no flow through the w a l l , applied at 2(Myy + l ) nodal points, yields 2(Mw + 1)
additional equations (similar to E q u a t i o n 4.20). Thus there are 2(Mw + l) + m + 2

linear algebraic simultaneous equations and the same number of unknowns. The
solution to this set of equations establishes the flow field and hence the required
loading.
Details of the numerical representation and computer application of the above
discussed mathematical model are presented in Chapter V .

Singularities
Tunnel
Wall Mw 1^1
i « 1 • i I 4 1-
Rotor
U
2H
—t •
i • • i
2 3 Mw Nascent
Vortex
Nodes
F I G U R E 4-4 Geometry for blockage modelling.

CHAPTER V
NUMERICAL INVESTIGATION
5.1 F o r m a t i o n o f t h e S y s t e m M a t r i c e s
5.1.1 Unconfined flow model
T h e set of linear simultaneous algebraic Equations (4.20) and (4.21) derived in
Chapter I V can be represented in the m a t r i x form
A x = b, (5.1)
where A is a ( m + 2) x (m + 2) square m a t r i x and x , b are order ( m + 2) x 1
column matrices, x represents the unknown strength of the distributed and nascent
vortices, and b is a known m a t r i x . T h e matrix A can be partitioned as
(5.2)
where A j is known as the influence coefficient m a t r i x and A 2 is a unit row m a t r i x .
Similarly, matrix b can be partitioned as
(5.3)
where vector b i defines normal v e l o c i t y at the nodal points due to known inputs
106
Chapter V.: .Numerical Investigation .107
(free stream, rotation, wake vortices, etc.,) and
2k
b
2 = - Y , T
3=1
after k time-steps. T j w is the strength of the j t h

shedded vortex. Note, the last
row of m a t r i x equation (5.1) represents E q u a t i o n (4.21).
Let one of the nascent vortices be specified as (Ti,zy) where T j is the strength
and 2 l 5 the position. Similarly, let the distributed vortices be designated as (Tj,Zj)
(j' — 2 , . . . , ro -f- .1.) and the nascent vortex at the other end as ( T m + 2: - The
unknown m a t r i x x can then be written as
r 2
x= (5.4)
Vr m + 2 J
T h e element a,j of the influence coefficient m a t r i x represents the velocity normal
to the blade surface at the i th

node due to a unit vortex at the j t h
singular point.
Using Equations (4.16) and (4.20), a,j can be written as:
Real
f_L_^!_l f,-=l,...,m+l; ,
\2-K{z ni - )\Z j \j — 1 , . . . , m + 2; v
" }
where i and j are two positions fixed w i t h respect to the blade. N o w (z — Zj) l
can be expressed as
Zni Zj — TiyC
where (r{j,&ij) are constants for given i, j and blade geometry. 0 is the blade
angular position. Recalling that <j) i — yj i + Ini + 0 (Equation 4.20), it is evident

n n
that a y is a constant irrespective of the blade rotation, i.e., the m a t r i x A j need be

t
calculated only once for a given geometry.

Chapter V: Numerical Investigation 108
T h e element bn of the m a t r i x b i is the velocity at the i th

node due to free
stream, free vortices and blade rotation. Using Equations (4.16) and (4.20), after k
time-steps, bu can be written as
2k _
»• }_wl
bit = T iU! sin ip i — Real e -' \ - Ucos<j> ,
l4>
ni (5.6)
27T ^
n n
{z
ni - Z l) _
w
where T ^ / (/ = 1 , . . . , 2k) are the free vortices shedded in A; time-steps.
T h e m a t r i x equation (5.1) is solved using the Gaussian E l i m i n a t i o n Procedure
w i t h partial pivoting, forward and backward substitution, followed by iterative i m -
provement.
Once the unknown strengths of the vortices are found, the nascent vortices are
renamed as shedding free vortices, i.e.,
z
w(2k-\) ~ l iz
Tw2k = + (5.7)
z
w2k = z
m + 2i .
T 8i = Ti +i i = 1,... , m .
5.1.2 Wind tunnel blockage model
In the modelling of the w i n d tunnel wall confinement, matrix A in E q u a t i o n
(5.1) is of order ( m -f Myy -f 4) x (m -r Mw + 4) and is a constant for the station-
ary blade. However, for the rotating blade certain parts of the m a t r i x A become
functions of time, i.e., the influence coefficients relating the blade t o the tunnel
boundaries vary w i t h orientation of the rotating blade. In contrast, the unconfined
model had a constant A matrix for both rotating and stationary blade. T h i s implies
an increase in the computational effort not only due to an increase in the number of
boundary elements but also as a result of the variable influence coefficient matrix.
T i m e marching process and the rest of the computational procedure remain
essentially the same.
5.2 Calculation of Loading on the Blade
T h i s is achieved by applying E q u a t i o n (4.27) along the blade. Several problems were
encountered during application of Equations (4.18) and (4.19) along the blade, due
to distributed concentrated vorticies. T h e improper integral which has a singularity
in its kernel function in the domain of integration (Equation 4.11) was the reason
for most of these problems. It is advised to use the Cauchy P r i n c i p a l Value Integral
w i t h an added 'free t e r m ' from the singularity (Banerjee, 1981) to overcome this
difficulty in the B E M . Definition of the 'free t e r m ' is rather complicated when
concentrated, instead of distributed, vortices are used.
5.2.1 Velocity distribution along the blade
For the purpose of calculating the velocity on either side of the i th

element, the
concentrated vortex in that element is treated as uniformly distributed. Thus the
density of this uniformly distributed vortex is T /AS.SI where AS is the length of
the element. T h e complex potential for a vortex sheet, with uniformly distributed
vorticity of density 7 , which makes an angle 0 . with the positive real axis (Figure
5-1) can be written as
(5.8)
where z a and z\, are the coordinates of the end points. O n evaluating the integral
F I G U R E 5-1 Uniformly distributed vortex.

Chapter V : Numerical Investigation 111
it takes the form,
t"ye ~ ia
\ ( z —z a \
F(z) - — z In -r zi l n ( 2 - 2 )
b - z \r\(z - z ) + ( z -
a a a 25) • (5-9)
27T L V 2
- 2
b /
T h e associated complex velocity is given by
dF(z) iê
where l n ( f ^ ) is the key t e r m . Let 2 - z a = r e " and 2 - zi, — ^ê ^ • (Figure

a
i e ^ f
5-1). Now
\z-z J b n
T h e velocity normal to the vortex sheet has a magnitude
v=
n — In — ,
27T r\
whereas the velocity tangential to the sheet is
Vt = ^-{6a ~ Ob)-
Depending on the side of the sheet considered, (6 a - 61) would be ZL7I. Figure
5-2 shows the velocity distribution due to a vortex sheet, compared to that of
a concentrated vortex. The normal velocity tends to infinity at both ends of the
element and is zero at the middle of the element. T h e tangential velocity is constant
at -i- ^ and - ^ on either side of the element and is zero along the line of the element
everywhere else. In the case of a concentrated vortex at the middle of the element,
the normal velocity tends to infinity at the centre when approaching from either
direction and is finite everywhere else. The tangential velocity is zero along the
F I G U R E 5-2 Velocity distribution for a uniformly distributed vortex sheet com-

pared to that for a point vortex.
line of the element and becomes infinite at the centre. B o t h distributions are fairly
similar away from the element.
In the analysis, boundary conditions were satisfied only at the nodes. L o o k i n g
at the velocity distribution for a concentrated vortex it is obvious that the 'no flow
through the blade' condition is not satisfied along the element. This also suggests
that calculation of the velocity at the centre of the i th

element may cause some
error. To avoid this situation calculations were made at the nodes i and i - r 1
(where the boundary conditions are satisfied) using E q u a t i o n (4.18), and a linear
variation in magnitude of the velocity along the element assumed. Now the velocity
at the centre of the i th

element can be given as
\w{z )\ ni + |u>(2NI+1)|
w(z )
si exp(-i(j) i)
s ± —~- exp(-i<p ), sl (5.11)
where w{z i)n and w{z i+i)

n are the complex velocities at z ni and respectively,
and <t>si is the inclination of the i th

element to the positive real axis. -I- and - signs
correspond to top and b o t t o m sides of the blade, respectively.
Several other methods to determine the velocity distribution along the blade
were also tested. In one approach calculations were made at the nodes and the
velocity at the centre of the i th

element obtained by averaging the velocities at i th
and (i + l) th
nodes, and adding the free term ± ( r t / 2 A s ) exp(-id )
s sl to it, i.e.,
w{z ) sl = ± êxp(-?.0 S ! ).
T h i s yields the velocity at the centre, different from that in the tangential direction,
on either side of the element (Figure 5-3a).
A n o t h e r method was to evaluate velocities at the nodes, and add averaged "free
F I G U R E 5-3b Points at a radial distance t from the blade.

t e r m " to obtain the total velocity at the node, i.e.,
w{z )ni = w(z ) ni ± ( ^'

r
'J 3
' - 1
) exp(-^ n i ),
4As
where w(z i)
n is the velocity found by using E q u a t i o n (4.18) at the i th
node. T h e
procedure is rational and gives logical results.
Evaluation of E q u a t i o n (4.18) was also attempted along points which were at
a distance t away from the blade . T h i s was tried at the nodes as well as at the
singularities given by:
ZA,Bni ~ Zm ± t exp(i>„i);
ZA,Bsi = z i ± t exp(i<f> i):

s s
where — and — signs denote the two sides of the blade (Figure 5-3b). A s expected,
the results were very sensitive to t and the method failed due to the singular char-
acter of the solution at the blade.
Alternatively, the velocity can be calculated at the i th

singularity, using E q u a -
tion (4.18) neglecting the i th

singularity, and adding the "free term" ±(T s i/2As)e~ '
t<t
T h i s method also yields a non-tangential velocity on either side of the blade. How-
ever, the results showed considerable improvement compared to those of the above
method.
None of these methods were as successful as the method used in the present
analysis (Equation 5.11).
5.2.2 Calculation of 4jy term
To find the pressure distribution using E q u a t i o n (4.27), it is necessary to eval-

uate the term ^ where $ is the real part of W(z) in E q u a t i o n (4.19),
W(z) = * + t*,
<9$ „ f dW
— = R e a <^ — (5.12)
dt | df
Since the vorticity distribution is assumed to be uniform along the element i n the
calculation of velocity, it is necessary to use the same assumption in the calculation
of A l t h o u g h E q u a t i o n (5.9) appears to have singularities, contributed by the
first t e r m , it can be rewritten as
i~fe (z - z ) a
z
(z - z ) h
zi
,
F(z) M
* >+ («. - * )
2TI
Thus in the l i m i t i n g case as z —> z or z —» Zj, the associated terms cancel leavinga
the relation free of singularity. Therefore, the complex potential at the centre of
the i th
element after k time-steps can be written using E q u a t i o n (5.9) together with
contributions from the free stream, bound and free vortices as
le z \n(z - z ) + (z - z )
F(z )sl = ——T ei zi, ln(z - - zt,) sl a sl a a b
2wAs
2k
-r — E
II = 1
T s n l n
( 2
s * ~ sn)
z
- —
2TT
^2 T w l ln
( *'
s Zwl) + Uz ai + C[t),
(5.13)
where z a and z^, are the end points of the i th

element. B y considering the time
dependency of each term and simplifying it using the chain rule the following ex-
pression for ^ is obtained
dF(z )si i dT f
2k
i v-- r,
+ {^si z
wl)
ie~ " l(p
+ 2TTAS
i uT i(z i S n — z i i)
n +
t -»>..-
e dT si
+ 2TTAS ~dT
Z i\n(z i
n s — Z {) — 2ru+l
n l n
( 2
s t ~~
z
ni+\) + +l —
ni)
z
e- *'' dT
1
ei
± 2-nAs dt
z Bi + i uUz £i + C(t),
(5.14)
where OJ is the angular velocity of the rotor and C is a constant at a given instant
of t i m e .
E v a l u a t i o n of the general term.
2.71
in E q u a t i o n (5.13) using a computer leads to some error in the estimation of
Recognizing that z — z,- = r,-c ', tfl
A,- = — In r, + i $i
2^
6, i
In r , .
27T ' 27T
Here, - ^ corresponds to $ where the d a t u m for 0, is somewhat ambiguous. A s s u m -
ing 0, in the range 0 < 0, < 27r, and with a series of z,-'s such that the corresponding
0 's are in the 4

t
th
and 1 st
quadrants, the real part of A w i l l have a discontinuity T
of 2TT w-hen crossing the positive real axis. It could be avoided by m a k i n g the range
— Ti < 8i < 7T. T h i s would then result in a similar discontinuity when crossing the
negative real axis. T h u s a discontinuity can be avoided by appropriately selecting
the d a t u m corresponding to the situation. In the computer algorithm, this problem
was carefully studied and treated accordingly so that no discontinuity in 6 would t
occur along the blade.
T h e results were found to be quite sensitive to the value of dT i/dt.

s Therefore,
it was evaluated w i t h a second order accuracy by using the Gaussian Three Point
Differential Formula at the middle point.
T h e experimental data available in this study are not sufficient to evaluate
the time-dependent c o n s t a n t C , which contributes to the value of ^ y , as unsteady
pressure measurements were not carried out. To make the matter worse, variations
in the range of 0, to obtain a smooth $ along the blade adds yet another time-
dependent constant to Qjj term. It is indeed demanding, from the computational
point of view, to keep track of these constants.
Fortunately, the final results obtained for the average pressure distributions
showed proper trends and none of these unknown constants affect prediction of
the rotor performance (torque, power) since it depends only on the difference in
pressure on the two sides of the blade.
5.2.3 Evaluation of torque and power coefficient.
The torque coefficient can be found either by integrating the elemental torques
due to the differential pressure over the blade or by applying E q u a t i o n (4.28). B o t h
methods were tested in the present study and led to similar answers.
E q u a t i o n (4.28) was simplified by using the 'residue theorem' and the chain
rule for differentiation. T h e discretized, simplified form of E q u a t i o n (4.28) can be

w r i t t e n as
l_yy
. m 2k
2
i -»-{: ££ 71
i si - Zwl)
z
+ 2U
Sr"2"+1 *™)- (5'15)
where r SJ is the magnitude of the radius vector at the i th
singularity. Now the
torque coefficient is estimated using the equation
c
- = WvW • < 5
' 1 6 )
T h e power produced is calculated by discretizing the E q u a t i o n (4.29)
^ * = —y^w (5.i7)
p k A i
h^
where is the average power over k time-steps, T\ is the torque generated at the I th
time-step and UJ is the angular velocity of the rotor. T h e average power coefficient
after k time-steps is given by
5.3 C a l c u l a t i o n P r o c e d u r e
5.3.1 Computational algorithm
A block diagram for the computational process is shown in Figure 5-4. At
t = 0 the unknown distributed and nascent vortices T ; (j = l , . . . , m -j-2) are
calculated by satisfying the boundary conditions and K e l v i n ' s theorem (Equation
5.1). N e x t , the nascent vortices r^.j and r„,2 are shed downstream and all the
vortices in the wake are convected according to Equations (4.22) and (4.23), in
time-steps of interval At up to an instant t î n = t n -f AT [AT = KAt, K =

Establish
Problem
Geometry
t=0
k= 1
Set Initial
Conditions
I E
Calculate
Solve
Ax = fe
Find the
Flow Field G
No
k = k+1
F I G U R E 5-4 T h e block diagram of the computation process.

an integer; Inamuro et al., 1983). D u r i n g the interval At vortices are convected
without satisfying the boundary conditions at the blade. A t the end of the K ih
interval, i.e., at t = t +i-, strength and position of the distributed and nascent
n
vortices are again evaluated by satisfying the boundary conditions and the same
process is repeated. Here, the time interval At is chosen smaller than AT inorder
to reduce the computational cost and to improve the accuracy of the convection
calculation.
To evaluate using the Gauss Three Point Formula at the middle point,
the computational process is advanced by three time-steps and then returned to
the second step to calculate the loading. Thus the earliest loading information is
available only at the second step.
5.3.2 Computational parameters
There are several important parameters which must be established before the
calculation process begins. They are:
ts distance between the nascent vortex and the edge of the blade;
m number of elements representing the blade;
At time-step size;
o core radius of the vortices;
n balancing parameter (Equation 4.26);
K number of small time-steps At per large
time-step AT (Resolution Parameter).
Obviously, an increase in the number of elements would lead to a better accu-
racy, however, it would also escalate the computational cost. Thus a compromise
between these two competing parameters is necessary. T h e effect of the number of

elements on the final results was systematically tested inorder to arrive at a cost
effective value.
The size of the viscous core (o) is an important parameter which indirectly
determines the amount of vorticity dissipation in the wake. K i y a and Arie (1977)
suggest o to be 0.05 x R where R is the characteristic radius. W i t h this as a
guideline, three different values for o were tested.
Considering the fact that e should be greater than o, Ogawa (1984) selected e
s s
as 1.1 x o. It can be presumed that there is an o p t i m u m value for t s and location
of the nascent point, which may depend on the position and the tip-speed ratio of
the blade. In the present study, however, the nascent vortex position is fixed on the
extension line of the blade contour.
Stansby (1985) has suggested a value for the balancing parameter rj to be 0.03.
In the present study, several different values of n were used to assess their effect.
Based on accuracy and computational cost considerations, the resolution pa-
rameter K was taken to be 2. A change of K from 1 to 2 reduces the computational
cost by at least a factor of 4, without substantially affecting the final results.
The most important parameter in the computational process is the time-step
size At. T h i s should be small enough for the convecting vortices to follow stream
lines, and large enough to reduce the computational cost. A smaller value of At
was necessary so that the free vortex may not intercept the blade after one time-
step. This is particularly necessary at a large tip-speed ratios; the blade rotation
angle uAt, in one time-step may become large raising the possibility of vortex
interception. Based on the above consideration together w i t h values suggested by
Clements (1973), K i y a and A r i e (1977), and K a t z (1981), At was selected in the
range 0.03 - 0.1 x (R/U), depending on the tip-speed ratio A.

T a k i n g a constant angular displacement of the rotor per time-step of 3° the
following relation can be obtained,
0.04363372
At = (5.19)
XU
A t larger values of A, At gets quite small resulting in a slow rate of convection
downstream. To avoid this the following model was used at higher tip-speed ratios,
(5.20)
where 6 is determined according to the range of A of interest.
5.4 Effect of Computational Parameters
In order to determine the computational parameters which may lead to an i m -
proved calculation efficiency, several test runs were carried out and the effect of
each parameter on the final result established.
T h e calculations were performed, in double precision, using the A m d a h l 5850
main frame computer. A l t h o u g h the increasing 0 direction was taken as counter-
clockwise in the modelling procedure, and the free stream flow along the negative
x-direction, the results are presented w i t h the flow from left to right w i t h clock-
wise rotation of the blade. T h e computer C P U time required for the calculations
increases exponentially with the number of time-steps. Therefore, a restart facil-
ity was incorporated in the program so that the computations can be stopped and
restarted as desired.
5.4-1 Number of elements m
A word concerning the choice of number of elements would be appropriate. T h e

m i n i m u m length of the element is limited by the size of the vortex core o to satisfy
the boundary condition at the blade. T h i s gives the m a x i m u m value for m as 33.
W i t h this as a background three sets of results were obtained w i t h the number of
elements being 20, 25, and 30 w i t h all the other parameters held constant:
p/q Geometry parameter = 0.2;
Q C i r c u l a r arc radius = 0.191 m ;
A Tip-speed ratio = 0;
0 Blade angle = 120°;
At Time-step size = 0.004 s;
K Resolution parameter = 2;
U W i n d speed = — 7 m/s;
V Balancing parameter = 0;
k Number of time-steps = 100;
6 Blade arc angle = 135°.
T h e results obtained for torque coefficient ( C V ) as affected by the number of
elements is shown in Figure 5-5. Note, the plot extends over the time interval of
0.8 5 during which the first vortex cluster moves approximately two rotor diameters
downstream. T h e corresponding flow patterns in the wake after 60 time-steps are
also presented in Figure 5-6.
It is apparent that the results for m = 25 and 30 are essentially the same w i t h
the corresponding flow patterns remaining virtually unaffected. T h e C P U time
involved in the calculation process (a measure of cost) is compared in Table 5.1.

1.2-1
Time (Sec)
F I G U R E 5-5 T i m e history of the torque coefficient as affected by the number

of elements.
1.5-1
1.0- x x
X
X x x
0.5-
x x x
0.0
-0.5H
M = 30
-1.0
-1.5 T
xx X
1.0
x* x X
x
x X
* * *
0.5
^ x x x *
Y 0.0
-0.5
M = 25
-1.0
-1.5
1.0
xxx<
x x x
0.5 H
0.0
-0.5
M = 20
-1.0
-1.5 H ^ i i i 1 1 l 1 i 1 .
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
X (meters)
F I G U R E 5-6 Flow patterns after 60 time-steps as affected by the number of
elements!
N u m b e r of C P U Time
Elements (m) (s)
20 173
25 217
30 260
T A B L E 5.1 Effect of number of elements on the computational cost
Based on this information m — 25 was selected in the subsequent analysis.
5.4-2 Vortex core radius (limiting distance o)
A n o t h e r important parameter in the flow model is a , the core radius of the
vortex or the l i m i t i n g distance between a vortex and a solid surface. A s mentioned
before, it indirectly determines the dissipation of vorticity i n the flow. To establish
the effect of o on the final results, several calculations were carried out w i t h a —
0.017?, 0.05i? and 0.1 R, keeping all the other parameters fixed. T h e results for
time variation of the torque coefficient are shown in Figure 5-7. T h e plots for
o = 0.05/? and O.li? are similar in trend except for a phase shift suggesting the
process of vorticity dissipation to be dependent on o. However, the character of
torque variation w i t h o = O.Oli? is fundamentally different. T h i s disparity is also
reflected in the corresponding flow fields (Figure 5-8) where the flow pattern for
o = O.Oli? at the upstream blade is rather unrealistic. T h e vortex core radius o is
directly related to the position of the nascent vortex (e — 1.1a). W h e n o is small,

s
the nascent vortex being too close to the singularity at the tip of the blade causes
this unrealistic behaviour. In the present study, the value of o was fixed at 0.05/?.
Time (Sec)
F I G U R E 5-7 Effect of the l i m i t i n g distance parameter o on the time history of

CT-
Chapter V: Numerical Investigation
1.5
1
x x x x
x v X
x X
x*
xx x x
X
X XX
X x
0.5H
0.0
a = 0.10 R
-0.5-J
-1.0-1
-1.5
T 1 1 r T r
XX X
1.0
x * x x
x x ^ x
X
0.5 x X
xxX><
Y 0.0
-0.5
ff = 0.05 R
-1.0
-1.5- I I T r
i i r
1.0-
x x X X
x^x x
a = 0.01 R
-1.0
-1.5-i \ 1 1 1 T 1 1 1 1 p
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
X (meters)
F I G U R E 5-8 Flow patterns after 60 time-steps as affected by o.

5.4-8 Position of the nascent vortex e s
Three different values, e = l.lo, 1.5a and 2.0a, of the distance between the
s
nascent vortex and the tip of the blade were tested to evaluate its effect on the final
result. O f course, e should be greater than o to satisfy K u t t a condition at the tip.

s
T h e results for the torque coefficient variation for different e are presented in
3
Figure 5-9. T h e curves show basically similar time histories w i t h a slight difference
in the phase and the average magnitude. The corresponding flow patterns obtained
at 60 time-steps are shown in Figure 5-10. They are also quite similar suggesting
the effect of e on the results to be relatively insignificant. However, as t = 1.1a

e s
predicts the torque coefficient closest to the experimental value shown in Figure
2-30 it was selected for the model.
5-4-4 Balancing parameter n
A similar procedure was adopted to evaluate the effect of the balancing param-
eter (n) on the final predictions. T h e results are shown in Figure 5-11. Except for a
small change in amplitude, the rotor response is essentially the same. T h e similarity
of the wake is also apparent (Figure 5-12). However, as the computing cost involved
is slightly higher for none-zero r/ it was set at zero in the present model.
5-4-5 Time-step size At
In a time marching computational process, the size of the time-step is an impor-
tant parameter. Conceptually, the smaller the time-step the better the prediction.
However, this may not be true for the flow field. Calculation of the blade loading
using smaller time-steps may give misleading results due to a large number of small
vortices close to the blade causing an unrealistically large velocity. T h i s effect is

1.2 7—
//^ — ^
\
DVM - Solu. is X \
0.9H
—11/, V.
\^
C-S i. i v
e = 1.5 o
s
c s = 2.0 a
0.6H
/ /
0.3 H \ V
0.0 H
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Time (Sec)
F I G U R E 5-9 T i m e history of the torque coefficient for different nascent vortex

positions.
1.5
X
x
1.0- x x x
X x
x X
v
x x x x
0.5 XX X v
x ^ x
0.0
-0.5H rn • • n • •
c = 2.0 a
s
-1.0
-1.5
Xx
1.0 x
x x x
v
0.5 x x X
x
x
Y 0.0
® -
-0.5
c , = 1.5 a
-1.0
-1.5 i i r
XX X
1.0- x ^ x x
x x ^ xX
0.5- x X
xxX><
0.0-
-0.5-
c = 1.1 a
s
-1.0-
-1.5 1 """""T™* i i i i i ^ ""T™ "•""T"™"" T~""

1 - - -
i I
- 0 . 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
X (meters)
F I G U R E 5-10 Flow patterns after 60 time-steps for different e . s

T i m e (Sec)
F I G U R E 5-11 V a r i a t i o n of the torque coefficient as affected by n.

1.5
>x x
1.0-
X ^ x x x
x
x
X x x
0.5- XX X
x x x
0.0-
-0.5-
•n = 0.06
-1.0-
-1.5
X
1.0 xx
x X
x ^ x
0.5 x X
xxX><
0.0
-0.5
fid* •n = 0.03
-1.0
-1.5 i r
1
XX X
/ x x
x x
x £ >
x X
xxxX
0.5-1
0.0
to* r) - 0.00
-0.5
-1.0 f
-1- 5
i l l 1 1 I 1 1 1 |
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
X (meters)
F I G U R E 5-12 Flow patterns after 60 time-steps for different n.
F I G U R E 5-13 T i m e history of Cj as affected by the time-step size.

1.5-
1.0
0.5H xxx
0.0 n X^
X
O n
or • • •
-0.5- •
At = 0.008
-1.0-
-1.5
XX *
1.0
x x x
X X
0.5 * xxx*
x
0.0
-0.5 At = 0.004
-1.0
"I P
At = 0.002
-1.5 1 1 I I 1 1 I I 1 I '
- 0 . 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
X (meters)
F I G U R E 5-14 F l o w patterns at t = 0.792 s for different time-step sizes

evident in the predicted torque coefficient shown in Figure 5-13. A larger value
of At leads to a relatively smaller amplitude for CT as the vortices are convected
downstream to a longer distance. T h i s is v i v i d l y shown by the flow patterns in
Figure 5-14 (at t = 0.792 s). Several interesting features are apparent. Although
the wake vortices are similar in the basic shape, it is evident that there is an in-
crease in the radius of curvature of the separating shear layers with an increase in
At. Note, a larger At causes the vortices to come closer to the blade and hence
be destroyed whereas at a smaller At they continue to persist. Since dissipation of
vorticity in the wake plays an important role in the analysis a smaller At may not
always give better results. In fact, in the present case, At = 0.008 s predicted the
torque coefficient closest to the experimental results.
There is yet another aspect to the selection of At: the computational cost
involved in the execution of the program. The C P U time for the results shown in
Figures 5-13 and 5-14 are given in Table 5.2.
Time-Step C P U Time
Size (s) (s)
0.002 1275
0.004 217
0.008 75
T A B L E 5.2 Effect of time-step size on the computational cost
A n enormous increase in cost with a decreasing At is apparent. Based on these
considerations, the stationary rotor analysis was carried out using At — 0.004 s.
CHAPTER VI
RESULTS A N D DISCUSSION
F r o m the foregoing discussion it is apparent that the mathematical model and its
numerical evaluation, as developed in the last two chapters, are indeed quite chal-
lenging b o t h in terms of fluid dynamical phenomena and computational complexity.
T h e difficulty is further accentuated by a large number of variables, some rather
ill defined, e.g., the l i m i t i n g distance parameter a and A r , which indirectly govern
vorticity dissipation; the balancing parameter 77, which directly controls the vis-
cous effects; and the nascent vortex position c . Obviously by a systematic change
s
of wind speed, rotor geometric variables, wall confinement, initial conditions, and
computational parameters an enormous body of data can be generated. However,
the objective here is to develop methodology for approaching such a complex situa-
tion and assess its effectiveness in the light of experimental data presented earlier.
Hence, for conciseness, only a typical set of results useful in establishing trends are
recorded here.
To begin w i t h , the static Savonius rotor is analyzed w i t h different orientation
of the blade; and the starting torque, wake geometry and frequency characteristics
discussed. T h i s is followed by the results for the rotating case accounting for the
effect of tip-speed ratio (A), and blade geometry parameter [p/q). F i n a l l y , influence
of the w i n d tunnel wall confinement for b o t h stationary and rotating blades is
assessed.
138
Chapter VI: Results and Discussion 139
6.1 Stationary Blade
T h e flow around a stationary blade at different angular positions and the cor-
responding average strating torque characteristics for the Savonius rotor were es-
tablished using the mathematical model discussed in Chapters I V and V with the
computational parameters:
m = 25 ;
o =0.05# ;
i
s = 1.1a ;
n = 0 ;
At = 0.004 s.
To minimize the computational cost the resolution parameter K (section 5.3.1)
was fixed at 2. q (Figure 2-1) was taken as 0.191 m and p/q fixed at 0.2 to facilitate
comparison w i t h the experimental results presented in Chapter II (section 2.6). The
wind speed was maintained at 7 m/s simulating the experimental conditions.
A s discussed in section 2.5 the pressure distribution over the blade varies with
the blade angle of attack. Depending on the angle the flow may have different
separation points. However, the mathematical model developed assumes the flow
to separate at the tips of the blade, which is valid only for angles (/?) between 90° and
120°. A l t h o u g h the model agrees with the separation points in this range it cannot
be expected to give a constant base pressure on the backside of the blade as observed
experimentally due to its potential character ( D V M ) . A t other angles, the flow does
not separate at the tips. However, due to high Reynolds number the separated
shear layer remains close to the tip of the blade making the flow geometry not much
different from the separation at the tips. B u t in the calculation of loading it makes
a significant difference due to a considerable change in the pressure distribution.

T w o distinct procedures were used to account for these effects.
T h e results (section 2.5) show that separation occurs due to an adverse pressure
gradient and the pressure remains at the separation value in the wake. T o minimize
the effect of this discrepancy, the experimentally established separation points were
used in the analysis and the pressure calculated at the separation point was held
constant in the wake. T h e results obtained using these assumptions are named as
" D V M with surface separation".
In the second approach, the shear layers separate at the blade tips as implicit in
the D V M model, however, the base pressure is assumed constant. T h i s is referred
to as the " D V M w i t h C b = constant".

p It is of interest to point out that actual
value of Cpb is not required as it does not contribute to the torque.
T h e time variations of the torque coefficient predicted at 0 = 0° using the above
three approaches are compared in Figure 6-1. Constant base pressure cases give
almost the same result whereas the potential flow solution yields relatively higher
values for Cj. H i g h frequency irregularities present in all the three solutions may
be attributed to the free vortices shed from the front tip passing close to the blade.
There is no evidence of low frequency vortex shedding ( K a r m a n vortex shedding)
at this angle. T h e flow patterns at k =30, 60 and 90 time-steps are presented in
Figure 6-2 substantiate the above observation.
Figure 6-3 shows variation of the torque coefficient with time at 0 = 3 0 ° . Per-
haps the most striking feature in these curves is their smoothness. Plots representing
the constant base pressure assumption are fairly close to each other. However, the
solution using the potential flow assumption gives a torque coefficient well above
the other two as before. A g a i n there is no evidence of the K a r m a n vortex shedding
at this angle, i.e., the rotor continues to behave as a fiat plate.

1.2
1.0-
0.8-
0.6-
0.4-
0.2-
[3 = 0
DVM
Seporatton C p b
tip recovery
tip consfT
surface const.
0.0-
0.0 0.1 0.2 0.3 0.4 0.5
Time (Sec.)
F I G U R E 6-1 T i m e history of the torque coefficient at 0 = 0°.
1.5-r
1.0-
0.5-
-1.0-
-1.5-j 1 1 1 1 1 1 1 : ! 1 1 1
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-2 F l o w patterns at (3 = 0°: k = 30,60,90.
(3 = 30°
PVM
Separation C p b
tip recovery
tip const.
surface const.
n
1
1
1 ri 1 1 1 f
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Time (Sec.)
F I G U R E 6-3 T i m e history of C
T at 0 = 3 0 ° .
T h e corresponding flow patterns obtained at time-steps 30, 60, 90, 110, 130, and
150 are shown in Figure 6-4 . These patterns are quite similar to that for a flow
past an aerofoil. A l l the vortices shed from the front edge of the blade are destroyed
and the flow separates at the rear tip. Thus validity of the results predicted by the
constant base pressure assumption are in doubt. The flow patterns suggest attached
flow over the blade simulating an aerofoil at a high angle of attack without stalling.
However, this is rather unrealistic. Therefore the results obtained at this angle
using the potential flow solution would be highly optimistic and unrealistic.
W h e n the blade is oriented at 6 0 ° to the free stream, a trace of a low frequency
vortex shedding becomes apparent (Figure 6-5) which is also evident in the flow
patterns (Figure 6-6). As can be expected, the time needed to reach a steady state
depends on the initial conditions and the blade orientation.
Note, the solution corresponding to tip separation w i t h a constant base pressure
shows a periodic character. O n the other hand, the case of the surface separation
leads to a periodic solution with a low frequency mean associated w i t h a high
frequency modulation (Figure 6-5). T h e potential flow solution was observed to
reach a steady state at around 0.35 seconds, with several peaks, corresponding to
formation of vortex clusters, having a periodicity of K 0.41 s.
The periodic nature of the flow is evident in the flow patterns shown in Figure
6-6. E v o l u t i o n of the vortex clusters and their drift downstream is apparent. The
first vortex cluster from the bottom edge separates at around 30 time-steps fol-
lowed by a similar separation from the top at around k =60. A r o u n d 50 time-steps,
the first vortex cluster separated from the blade passes two diameters downstream
where the solution can be approximated as having reached a steady state. Immedi-
ately afterwards, at around 80 time-steps, the second cluster from the b o t t o m edge
1.5
1.0H
0.5
0.0-
nuiiEiiirrBiiiMirwrm-T
-0.5- k = 30
-1.0-
-1.5 i r T r
1.0-
0.5-
Y 0.0 X
^\ _^jnrrjm:n L- I • li I- I !• !• I*P H !• H •
s> 11. H t. <. ii .t i , |. n
M - - r --r-t—I -
HHHf7Trn n pTTr TTrF
1
-0.5 k = 60
-1.0-1
-1.5 T r T r
1.0
0.5
0.0
jffiiiiniiirni.1 •i.iH'Mi, i .|||||
l M T i r m :
TOW w i n ; n-i i t.ri.|..m.»:.itirC33-
-0.5-
k = 90
-1.0-
-1.5 i 1 1 1 i i 1 1 r
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-4 F l o w patterns at f3 = 3 0 ° :fc= 30,60,90.
2.0-
1.5-
1.0-
0.5-
0.0- y n||lll—III
'"»"•"'"" "• "•iiiniiiiiiiiiii"'»n»"iMiiiiiiiiiiimrmT
-0.5-
k = 110
-1.0-
-1.5
-2.0 T—i r T P T P
1.5-
1.0-
0.5-
Y 0.0
miHimiimm • millliTTTTTTTTT
-0.5
k = 130
-1.0-
-1.5-
-2.0 T P i r T 1 1 1 1 P T 1 P
1.5
1.0
0.5
0.0
muHUBHiHHnimiHmminimi mm ••••.•• | , „„„„
Ill ll lm n „ m m t T ^
-0.5
k = 150
-1.0
-1.5
-2.0 H 1 1 1 1—I—I—i—i—i—i—i—i—i—i—i—• '

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
X (meters)
F I G U R E 6-4(cont.) F l o w patterns at 0 = 30°: k = 110,130,150.
1.8
1.54
124
13 =
flp
tip"
DVM
surface
60
Seporatfon 9pb
recovery
const.
const.
A
0.9H
Time (Sec.)
F I G U R E 6-5 T i m e history of C T at 0 = 6 0 ° .
1.5
1.0H
0.5
x * x
0.0-
-0.5
-i.oH k = 10 k = 40
-1.5
1.0
0.5-
Y o.o-
rftb
-0.5-
-1.0- k = 20 k = 50
-1.5
1.0
0.5-1
X X
x X x*
o.o-l
-0.5
-1.0 k = 30 k = 60
-1.5
-0.5 0.0 0.5 • 1.0 1.5 2.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
X (meters)
F I G U R E 6-6 Flow patterns at 0 = 60°: k = 10,20,30,40,50,60.
1.5-
1.0-
0.5- x x
0.0-
-0.5-
So
k = 70
-1.0-
-1.5
1.0 H
x
xx
x x X
x x x
x
0.5 X x x
xxxxxx
X v
Y o.o XxX
-0.5
k = 80
-1.0-1
-1.5
1.0-
x x^
0,5- xxxxxx
X X x >$<
x x x x
0.0 x x x
-0.5-j
-1.0-I
-1.5-*
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-6(cont.) Flow patterns at /? = 6 0 ° : k = 70,80,90.

separates which clearly shows mixing of vortices with opposite rotation (k =90)
resulting in a partial cancellation effect among the vortices. Note, evolution of the
flow closer to the blade follows the same pattern and the process continues in a
similar way shedding vortex clusters of opposite signs intermittently. The effect
of the vortices far downstream becomes progressively negligible on the blade. T h e
periodic variation of CT (Figure 6-5) is p r i m a r i l y governed by the periodic nature
of the flow described above. A t 90 time-steps (Figure 6-6) the distance between
the two vortex clusters shed from the b o t t o m edge was found to be approximately
2.5 m . W i t h the free stream velocity of 7 m / s , it gives a vortex shedding frequency
of fa 2.8 Hz resulting in a period of fa 0.38 s which is comparable to the period of
the CT fluctuation as shown in Figure 6-5.
T h e nature of the flow at 0 = 90° is quite similar to that at 6 0 ° . The flow
separates at the tips as in the real case. Therefore only two sets of results are
presented for the variation of the torque coefficient (Figure 6-7). B o t h numerical
solutions show periodic nature once the steady state is established. Compared to
the case of 0 = 6 0 ° the time taken to reach the steady state is significantly larger
(ss 0.65 S for 0 = 9 0 ° compared to fa 0.35 s for 0 = 6 0 ° ) . Obviously, this is due to
a higher bluffness at 0 = 9 0 ° , which also results in a lower Strouhal frequency as
reflected in the time history of CT-
T h e corresponding flow patterns are shown in Figure 6-8. Unlike the earlier
case, here vortex clusters from both top and b o t t o m tips grow almost at the same
rate initially. A s a result shedding of the first vortex cluster is delayed. A s the
time increases, the top cluster advances further and ultimately sheds around 0.7s
(k fa 90). T h e cluster from the bottom edge also separates soon after(k = 110).
Consequently attainment of the steady state is delayed. A t 150 time-steps, the two
1.2
1
_1
[\
i »
1
i •
i »
1
.11
1
0.8-
1
i
• >
•
1 • 1
1 •
\ fi \ / i
A. !
1 t • i• < i
\
/ i •
V. (
\
\ • • '• -\ ]
• \>
V Ws r t»
J
«
li
*
0.4-
/ A+i
f3 = 90°
0.0- PVM
Separation C p b
tip recoveryt
tip const.
-0.2
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Time (Sec.)
F I G U R E 6-7 T i m e history of C T at 0 = 9 0 ° .
1.5-
1.0 H
0.5 ^^^^^^
^ x
x x * X
* x *
0.0
-0.5-
k = 10 k = 40
-1.0-
-1.5 i i l I i i i i i * i -
0.5
6
3<
x xx X x
\ X
) x
x x x
-0.5H
-1.0 k = 20 k = 50
-1.5 i i i i
.0.5
CxXXxxx x*
-1.0 k = 30 k = 60
-1.5 i r
-0.5 0.0 0.5 1.0 1.5 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
X (meters)
F I G U R E 6-8 Flow patterns at (3 = 90°: k= 10,20,30,40,50,60.
1.5-,
1.0H
-1.0-
-5 -1
_1
1 1 1 1 1 ; ! : 1 : 1 1
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-8(cont.) Flow patterns at f3 = 9 0 ° : k = 70,80,90.
1.5
1.0
XX
KXXXXX,
0.5 X* X
0.0
-0.5-1
r o FID D
k = 110
-1.0H
-1.5H
x>x
*>S<>K
1.0
x*
0.5-
Y o.o-
-0.5
k = 130
-1.0H
-1.5
x
*>* xxxx
XT** # „
1.0-1 *xx x*x><
x
JL>& x
xXXxXxE xxx £ x x
0.5, xxx ^
\ X
x
x X
x
0.0 xxx
<*£xx
X
•0.5- k = 150
• r
-1.0-
-1.5 i . I I I I T i i i . i
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-8(cont<) Flow patterns at 0 = 90°: k = 110,130,150.

vortex clusters shedded from the top edge were observed to be approximately 3
meters apart corresponding to a vortex shedding frequency of % 2.3 Hz or a period
of 0.43 s, which is comparable to the period of CT variation (Figure 6-7).
It was necessary to carry out the computational process long enough to establish
periodic nature of the flow. In general, it took around 0.35 — 0.7 s for the flow
to attain a steady state (k 40 — 60). T h e true periodicity of the flow can be
assessed only after this point. For k = 100, usually it was possible to observe
only one cycle of variation. To help establish the periodic character of the flow
more conclusively, the computational process was extended to 200 time-steps for
0 — 120° and 150°. A s can be expected, the number of time-steps has an enormous
effect on the computational cost. Since every time-step adds two vortices, the
number of free vortices and hence the amount of calculations needed to evaluate
the flow field, increases with the number of time-steps. Table 6.1 shows variation
of the computational cost ( C P U time) with number of time-steps.
N u m b e r of C P U Time
Time-Steps (k) 00
50 75
100 217
150 550
200 1278
250 2811
T A B L E 6.1 Effect of k on the computational cost
Thus the cost increases exponentially w i t h the number of time-steps. Hence,
having established periodic behaviour of the flow, the computation was limited to
the first cycle.
T h e prediction of CT variation w i t h time for 0 = 120° is presented in Figure
6-9. T h i s clearly shows a regular periodic behaviour specially with the constant
base pressure case. The potential flow solution also shows a periodic pattern with
essentially the same period.
T h e corresponding flow patterns are shown in Figure 6-10 up to A" = 200 (note
the scales on different pages are not the same). Shedding of the first vortex cluster
from the top edge is apparent at k — 40. Formation of the vortex cluster at the lower
edge and its shedding follows. T h e first cluster travels two diameters downstream
around k = 50. A t k = 100 shedding of the second vortex cluster from the top
edge is noticeable. D u r i n g the period k = 100 — 200 (0.8 — 1.6 s), it is obvious that
the distance between the vortex clusters shed from the top edge remains constant.
At 200 time-steps shedding of the t h i r d cluster from the top edge can be seen.
T h e distance between the second and the third cluster is approximately equal to
that between the first and the second cluster (=: 3.5 m ) . T h i s represents a vortex
shedding frequency of about 2 Hz which corresponds to a period of 0.5 s, the same
as the CT variation (Figure 6-9). However, the vortices shed from the bottom
edge do not seem to have as regular a behaviour as the vortices shed from the top
edge. It is of interest to note that distance between the two clusters shed from
the b o t t o m edge gets shorter with time. A t k — 200 m i x i n g of the second clusters
from the top and b o t t o m edges, leading to a cancellation effect, is apparent. T h i s
results in a spacing between vortices in the top row to be different from that in
the b o t t o m row in the near wake. However, for a distance greater than around
ten rotor diameters downstream an evenly spaced street of vortex clusters w i t h
alternate signs is established.

1.4 f
P = 120
1.1- DVM
Seporatfon 9pb
tip _ recovery
tip consfT
0.84
C T 0.5
0.2 H
-o.H
-0.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Time (Sec.)
F I G U R E 6-9 T i m e history of C T at .0 = 120

c
1.5
1.0
* x
x \
0.5
0.0
-0.5
k = 40
-1.0
-1.5 T r
X
1.0 xx
0.5 x X
x x X x
Y o.o
-0.5
k = 60
-1.0
-1.5
T 1
1.0 xVx
X XX
x
x x
Y
0.5- V* x
0.0-
fe
-0.5
•5U 5Bfe%^ k = 80
-1.0
i i i i i i l I | i 1 | |
- 0 . 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-10 Flow patterns at /? = 1 2 0 ° : k = 40,60,80.

2.5-1
xVx^
k = 160
-2.5 T— —i—•—r
1
"i— —r
1
1.5
0.5 •
-0.5
k = 180
-1.5
-2.5 ' I ' I— —I—

1 1
I 1
T — 1
— r
1.5
4*
0.5
-0.5
k = 200
-1.5-
-2.5 -> 1 — . — | ,—p—.—p—,—p i 1

i • i—•—r
-0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5
X (meters)
F I G U R E 6-10(cont.) Flow patterns at 0 = 120°: k = 160,180,200.

Note, the cancellation of vortices is identified through following development of
the wake over a relatively long distance (around ten rotor diameters) thus adding
substantially to the computing cost. In general, the distance required for the wake
to attain steady state would also depend on the blade inclination. In the earlier
results for /5 = 6 0 , 9 0 ° , as the numerical integration of the equation extended only

c
over 4 — 5 diameters to minimize cost, the vorticity cancellation process won't be
complete, and hence the calculated frequency must be corrected to account for this
effect.
As in the case of /3 = 1 2 0 ° , the integration process was extended to k = 200 to
firmly establish periodic character of the flow. The computed transient character-
istics of CT for f3 — 150° is shown in Figure 6-11. A l l the three cases show fairly
regular periodic behaviour, however, as can be expected, the local details differ.
For the potential flow solution, time as well as spatial variation of the base pres-
sure leads to high frequency modulation of the torque coefficient and the results
cannot be expected to represent the real situation. T i p separation w i t h constant
base pressure dramatically improves the prediction and the frequency of the torque
coefficient reflects that of the fluctuating pressure on the upstream face of the blade.
T h e surface separation case models the real situation still better. T h e attached flow
is only over a small portion of the blade (Figure 6-12) w i t h the vortex separating
from the upper tip having little influence on the pressure in that region. O n l y the
vortex shedded from the lower t i p , being closer, affects the pressure thus reducing
the frequency of the transient torque by around a factor of two.
Figure 6-13 shows the corresponding flow patterns. Basic features governing
evolution, separation, clustering, downstream convection and periodic cancellation
of vortices as discussed before are all quite apparent here. For example, for k «
p= 150°
DVM
I ' 1
0.2 0.4
F I G U R E 6-11 = 150°.
F I G U R E 6-12 Flow geometry at the blade orientation of 0 = 150°. Note,

pressure in the attached flow region is p r i m a r i l y affected by the
vortex cluster formed near the lower tip due to its proximity.
1.5
** * * x
0.5
x x
* x x v
-0.5
k = 10 k = 40
' I I I I 1 1 1 1 1 1
* x
x x ^ x
x
Y 0.0
-0.5
-1.0 k = 20 k = 50
' I I I I I I I 1 1 1
1.0
* * * x
X
x
x x x
k = 30 k = 60
-1.5
-0.5 0.0 0.5 1.0 1.5 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
X (meters)
F I G U R E 6-13 Flow patterns at 0 = 150°: k = 10,20,30,40,50,60.
1.5-1
x#
x x x
X X
X
v x x X
k = 70
-1.5-
1.0-
y x * xxx
0.5- x x x x
B x x
X
X
Y 0.0-
X x
-0.5-
k = 80
-1.0-
-1.5
T 1 1 r i r
1.0
X y
x \
,0.5-
c
x x x x<
x^
0.0- x x xx A
-0.5-
k = 90
• •
-1.0-
-1.5 . . , ! ! ! ! , ! ^ ! 1
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-13(cont.) Flow patterns at (3 = 150°: k = 70,80,90.
2.0-1
X (meters)
F I G U R E 6-13(cont.) Flow patterns at (3 = 150°: k = 100,120,140.

2.5 n
X (meters)
F I G U R E 6-13(cont.) Flow patterns at 0 = 150°: k = 160,180,200.

180 — 200, the third vortex cluster is clearly visible. A s in the case of 0 = 1 2 0 ° ,
the distance between the first and second vortex clusters ( « 4 ro) is substantially
larger than that between the second and third vortex clusters ( « 2 . 7 5 m ) . Similar
behaviour can be observed for the vortices shed from the lower edge. T h i s gives an
approximate steady state vortex shedding frequency of 2.54 Hz leading to a period
of s= 0.4 s, which compares with the steady state period of CT (Figure 6-11). The
Strouhal number data for various blade orientations are summerized in Table 6.2.
Blade Strouhal
Angle (0°) Number
60 0.125
90 0.12
120 0.09
150 0.13
T A B L E 6.2 Strouhal number at different blade angles
Numerically obtained variation of the average torque coefficient as affected by
the blade orientation is compared w i t h the test data presented earlier (Figure 2-
29, uncorrected for blockage) in Figure 6-14. T h i s clearly brings to light strengths
and limitations of the different approaches developed in the thesis. In particular,
it provides designers w i t h a tool to predict static performance of the Savonius
rotors using several different methods, depending upon the available data, and the
associated measure of accuracy.
A s can be expected, although the potential flow model qualitatively predicts
correct trends, the actual magnitudes are rather off. O f course, this is primarily due
to the method's inability to predict constant pressure in the wake. On the positive
1.8
•
-
Experimental
-
PVM
1.4
Sepgrgtion C p b
-
• tip recovery
1.2 x tip const.
- • surface const.
1.0
- 11
CjO.8-
X
0.6 p •
><
• •
0.4
c: —> •
X / 11 >
0.2 • / • C
/ X
0.0-
E
^ /
-0.2 1
1 •
1
i 1 '
0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0
F I G U R E 6-14
P
Variation of the average starting torque as affected by the blade
CO
orientation. Note, the experimental results are uncorrected for
the blockage of 1 0 % .
side, the method is self contained, i.e., it does not require any input information.
Hence, it may be used to advantage during the preliminary design requiring trends
in torque variation.
O n the other hand, the approach involving tip separation w i t h constant base
pressure appears quite attractive. Note, it assumes constant base pressure but does
not require its explicit value. Thus, as before, no input information is required,
however, the correlation between experimental and numerical results is vastly im-
proved.
O f course, in practice, the flow does not always separate from the tip. If the
surface separation information, as obtained from the pressure data or by simple
flow visualization study, is used in conjunction with the constant base pressure,
the correlation is, as expected, better. A s discussed later, blockage corrections will
further improve the numerical predictions.
6.2 R o t a t i n g B l a d e
T h e main goal of the mathematical model developed earlier is to predict perfor-
mance of the rotating blade. In this case further complications arise in modelling
of the flow. Since the shear layer separation positions are not known for rotating
blades, the assumption of tip separation has to be contended w i t h . There is no ev-
idence to assume a constant base pressure for a rotating blade. Therefore, at least
in the present case, one is forced to resort to the potential flow approach discussed
earlier. T h i s may prove to be favourable in the case of pressure recovery through
reattachment of the flow, as the mathematical model is able to partially account
for it.
To confirm that satisfying the boundary conditions at every other time-step

(A' = 2), as in the stationary blade case, continues to yield satisfactory results, trial
runs were conducted using K =1 and 2, w i t h :
p/q = 0.2 ;
X = 0.4 ;
Pi (initial blade angle) = 30° ;
At = 0.0057 s
The torque and power coefficients as well as flow fields are compared in F i g -
ures 6-15 to 6-18. Surprisingly the correlation is excellent even in the transient
state (Figures 6-15, 6-16). Furthermore, the vortex patterns in the wake compare
favourably (Figures 6-17, 6-18). Note, a saving in the computational cost w i t h
K — 2 is significant as indicated in the following table.
K C P U (s)
1 350
2 95
T A B L E 6.3 Effect of K on the computational cost.
In this analysis, the tip-speed ratio X was taken to be 0.4. As the angle of
rotation of the blade per time-step is taken to be constant irrespective of A, the
solution w i l l improve w i t h an increase in tip-speed ratio due to an increase in the
number of free vortices in the field.
6.2.1 Effect of tip-speed ratio X
Effect of the tip-speed ratio for a rotor with p/q - 0.2 (optimum value given by
the experimental data, Figure 2-19) was analysed using the numerical model rather
thoroughly through a systematic variation of A in the range 0.4 — 1.8. O n l y three

0.8
\
\\
\ \
0.6
0.5
\\
\»
0.4
\ \
\ \
\ \
0.3 \ \
\
- - ^
0.2
DVM
K=1
K=2
U . U H 1 1 1 1 1 1 1 1 1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time (Sec.)
F I G U R E 6-16 T i m e history of C P at A = 0.4 as affected by K.

1.5-
K =1
1.0-
0.5-
0.0-
k = 70
-0.5- x T, +ve
v T, -ve
-1.0- o T +ve 2
+ T -ve 2
-1.5-
1 1 1 1
1.0-
0.5-
Y o.o-
k = 90
-0.5- x r, +ve
*~ ^ #
+
+ ++
v
o
F, -ve
T +ve
-1.0- 2
+ F -ve2
-1.5-
1 1 1 1
1.0-
0.5-
0.0-
k = 110
-0.5- x T, +ve
v r, -ve
-1.0- o T +ve2
+ F -ve
2
-1.5- I 1 1 1 1 1 1 1 1 1 1 1 1
- 0 . 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-17 Flow patterns after k = 70,90,110, time-steps at A = 0.4 w i t h
K = 1.
1.5-
1.0- K =2
X r, +ve
0.5H v r, -ve
V x x x
x
0 o +v« r2
x x ^
0.0 + T -ve
x
x
++ 2
+++ ^ + ++
-0.5- +
+ +
k = 35
-1.0-
-1.5H
1.0
0.5H x 3* x„ v
+ ^xx X^xxv
Y o.o +
.++
-0.5 ++
k = 45
-1.0H
-1.5
1.0 H
X x X X
y
0.5
^ 4 X
X X X
V
+
+ v X
0.0 + t +
+ +
A
+
-0.5 +
+ + + + ++ +
k = 55
-1.0
-1.5 1 1 1 1 1 1 1 1 1 1 1 1 T
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-18 F l o w patterns after k = 3 5 , 4 5 , 5 5 , time-steps at A = 0.4 w i t h

K = 2.
typical sets of results are presented here to help establish trends. A sample of flow
visualization photographs are also included for the case of A = 0.4.
It should be emphasized that there are three fundamental aspects of interest
here:
a) effect of blade rotation on evolution of the wake compared to the station-
ary case studied earlier;
b) flow visualization study of the wake to assess validity of the numerical
model; and
c) effect of the tip-speed ratio on the wake and the rotor performance.
Figure 6-19 shows evolution of the wake for rotating blades w i t h A = 0.4.
A l t h o u g h , the basic process of generation of the wake is the same as in the stationary
case there are some interesting differences.
T h e rotation is initiated w i t h a disturbance /?, = 30° at t = 0, and evolution of
the wake, which is now sensitive to the instantaneous position of the blade, remains
essentially the same until k » 40. For the stationary case, vortices associated
w i t h the two tips were of opposite sign as expected. However, w i t h the rotating
blades, the situation can be different as seen at A- = 50 where there is shedding of
counter-clockwise (—ve) vortices from both the tips. The process continues until
k =s 70, beyond which the wake pattern bears no resemblance to the stationary
case. Note interaction between vortices shed from the two tips which was not
present in the stationary case. A t k > 100, the wake attains a steady state and
the important features of the flow field, also substantiated by the flow visualization,
can be summarized as follows:
a) a clockwise (+ve) vortex is captured near the centre of the rotor;

Chapter VI: Results and Discussion 17'
1.5-
x F,+ve
1.0 v r, -ve
o r 2 +ve
0.5
0.0
-0.5H :
-1.0 k = 10 k = 40
-1.5 i i i i i
0.5
Y o.o c — +t+
-0.5
-1.0 k = 20 k = 50
-1.5 i t i i i
0.5
-1.0 k = 30 k = 60
-1.5
-0.5 0.0 0.5 1.0 1.5 2.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
X (meters)
F I G U R E 6-19 E v o l u t i o n of the flow pattern for A = 0.4 and its comparison w i t h
the flow visualization study: k = 10,20,30,40,50,60.
1.5-
x r,+ve
1.0 H v P, -ve
o F +ve 2
0.5
•» T -ve 2
k = 70
-1.5 i i i i i i i i
0.5
++ +
-1.0
k = 80
-1.5 i i i i i i i i
i i i i
0.5
0.0
k = 90
-1.5 i r
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-19(cont.) Evolution of the flow pattern for A =- 0.4 and its compar-
ison with the flow visualization study: k — 70,80,90.
1.5-
1
* r,4ve
v P, - v e
o T 2 4 v e
•* T 2 -ve
0.5-1 4 4
4++
4+ +
+44
0.0 +
k = 100
-0.5-1
-1.0-1
-1.5-
1.0-
0.5
Y o.o
+4-4
-0.5
4 4 4 k = 110
4+ 4+
-1.0
-1.5 H I I
1.0
X ' ,
+ + 4 ^ X X
^ X
X „
4 4
+
4 v £ : + , ^
4 A k = 120
- 1 . 5 Hi i : 1 : 1 1 1 1 : 1 1 : — .
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-19(cont.) E v o l u t i o n of the flow pattern for A = 0.4 and its compar-

ison w i t h the flow visualization study: k = 100,110,120.
F I G U R E 6-19(cont.) Evolution of the flow pattern for A = 0.4 and its compar-
ison w i t h the flow visualization study (k s= 100,110.120,
flow visualization pictures).
1.5
x r,+ve
1.0 ^ T, -ve
o f +ve
0.5 XX X o
2
4 f 2 -ve
o.oH
v X
. #1
-0.5
-1.0H
k = 130
-1.5
1.0
0.5-1
Y o.o-l ^ L +
+ V
x
~
x „
x
-0.5 + ++ + +
" +
+ -H-
• -i.oH
k = 140
-1.5
1.0-
0.5-
o. xxxx
X X
0.0- v
+ ++ +
-0.5-
V + + +* *x V
+ +
-1.0-
k = 150
-1.5
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-19(cont.) Evolution of the flow pattern for A = 0.4 and its compar-
ison with the flow visualization study (k = 130,140,150).
F I G U R E 6-19(cont.) E v o l u t i o n of the flow pattern for A = 0.4 and its compar-

ison w i t h the flow visualization study (k = 130.140.150.
flow visualization pictures).
b) a shear layer w i t h negative vorticity, separating from the upstream blade,
has a small part of it captured at the centre. T h e captured vortex slowly
advances towards the tip of the upper blade and is periodically shed;
c) vortices shed from the tips form shear layers close to the blade;
d) separating shear layers produce large vortex structures which are even-
tually shed downstream.
A video of the flow visualization clearly showed the existence of the central vortex,
shear layers eminating from the tips and large vortex structures moving down-
stream. The typical flow visualization photographs corresponding to k = 100 — 150
reveal the same features although they are a bit obscure. It is of interest to point
out that an independent flow visualization study, using a smaller model of the ro-
tor, by Yokomizo et a l . (1987) also revealed the same basic flow pattern. Note, the
assumption of the central vortex made in the semi-empirical approach (Chapter III)
is supported by b o t h the theoretical analysis and the flow visualization thus giving
it a firm foundation.
In practice, one would like to establish the o p t i m u m geometry of the rotor
leading to a peak power coefficient which, in t u r n , depends on the tip-speed ratio
A. Figure 6-20 attempts to assess the effect of A on evolution of the wake for several
values of k (k = 2 0 , 4 0 , 6 0 , 8 0 , 1 0 0 ) .
For k in the range of 20 — 40, variation of the tip-speed ratio has no significant
effect on the character of the wake. However, for k = 60 the central vortex, which
is distinctly present at A = 0.4and0.8, disappears at A = 1.2. T h i s suggests that
entrainment of vorticity at the centre is reduced or even eliminated as the rotor
tip-speed ratio increases. The same trend persists at k — 80,100. A s mentioned
earlier, now the central vortex is shed and a new one formed from the vorticity
1.5
x T,+ve
k = 20 k = 40
1.0-
o T +ve
2
+ T -ve
0.5-
2
0.0-
-0.5- +
-1.0- X = 0.4 X = 0.4
-1.5-
1.0
0.5-
Y 0.0
-0.5
-1.0 X = 0.8 X = 0.8
-1.5- T 1 I P
i i r
1.0-
0.5
o.oH
-1.0-1 X = 1.2 X = 1.2
-1.5 T 1 P T I P
-0.5 0.0 0.5 1.0 1.5 2.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
X (meters)
F I G U R E 6-20 F l o w patterns after k = 20 and 40 time-steps: A = 0.4,0.8,1.2;

p/q = 0.2; K = 2.
1.5
"i I I 1 1 1 1 1 1 1 1 1 1 1
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-20(cont.) Flow patterns after k = 60, time-steps: A = 0.4,0.8,1.2;

p/q = 0.2; K = 2.
1.5-
X k = 80
1.0-
V V x
v r, -ve
0.5- xx ^
x
o T 2 +ve
9 +
4-
+ n
y yx S
X
+ + T -ve
0.0- o V OO oo 2
+ ++v +
+ +++
+ +
-0.5- 7
?w w
V
+ +
+
-1.0- A = 0.4
-1.5
1.0H
0.5
Y o.o
-0.5-1
-1.0-1 X - 0.8
-1.5 -I T"
A = 1.2
1-5 j | i | | | | i i i | | i
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-20(cont.) F l o w patterns after k = 80, time-steps: A = 0.4,0.8,1.2;

p/q = 0.2; K = 2.
2.0
k = 100
1.0H x r, +ve
v T, - v e
po o T 2 +ve
0.0H .++++ + T 2 -ve
++
V V
-i.oH
X = 0.4
-2.0H
I ' I ' I ' I ' I ' I
1
I 1
I 1
I 1
I
1.0 H
Y o.o
-1.0
I it'
x = 0.8
-2.0 i r— 1
r - 1
i — 1
i 1
i — r
i I 1
I
1.0
o
o.o H
-1.0H
X = 1.2
-2.0 "i
1
r
-0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5
X (meters)
F I G U R E 6-20(cont.) Flow patterns after k = 100, time-steps: A = 0.4,0.8,1.2;

p/q = 0.2; K = 2.
shed from the upper blade. Jones et al. (1979) observed similar phenomenon of the
central vortex formation and shedding through a flow visualization study.
Corresponding time variation of the torque coefficient for A = 0.8 and 1.2 is
presented in Figures 6-21 and 6-22, respectively. In general, the torque coefficient
approaches a periodic profile, around an average value, asymptotically. The results
for average torque coefficient as a function of the tip-speed ratio are presented in
Figure 6-23. The plot shows the m a x i m u m value of the average torque coefficient
as 0.42 occuring at A = 0.4. T h e experimental data presented earlier showed the
same trend w i t h Cj^max (corrected for blockage) as 0.40 for A ss 0.45.
T h e average power coefficient obtained by integrating the steady state CT VS.
time curves, at different tip-speed ratios, is given in Figure 6-24. T h i s shows a max-
i m u m power coefficient of about 0.38 at a tip-speed ratio 1.2. T h e corresponding
experimentally obtained Cp vs. A curve was presented earlier (Figure 2-17). It
shows a peak power coefficient of about 0.50 at a tip-speed ratio of about 1.2. It
is important to notice that the theoretical prediction is for zero blockage whereas
the experimental results were obtained at a 16.4% blockage. The experimentally
obtained m a x i m u m power coefficient w i t h blockage correction for the p/q = 0.2
model is around 0.32 compared to the theoretical prediction of 0.38. Considering
the highly complex, transient and separated character of the flow, the numerical pre-
dictions should be considered rather good and of sufficient accuracy for engineering
applications.
T h e mathematical mode) forecasts the peak power coefficient at a tip-speed
ratio greater than one, suggesting that the Savonius configuration is not exactly a
drag type w i n d turbine. The experimental as well as the theoretical starting torque
characteristics also suggest the same. Ogawa et a l . (1986 b) and Sawada et a l .

Time (Sec.)
F I G U R E 6-22 Time history of C T at A = 1.2.

0.45-1
A
F I G U R E 6-23 Theoretically predicted variation of the average torque coefficient
as affected by the tip-speed ratio A for p/q — 0.2.
0.40-1
n 1
i 1
i 1
i 1
i 1
1 1
1 1
1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
FIGURE 6-24 Theoretically predicted variation of the average power coefficient

as affected by the tip-speed ratio A for p/q = 0.2.
(1985) have also come to the same conclusion.
6.2.2 Effect of p/q variation
To assess the model's ability to predict the effects of geometric changes on
performance of the Savonius rotor, several test runs were executed w i t h different
P/Q.
A set of typical results obtained at a tip-speed ratio of 0.6 for p/q — 1.0, 0.2 and
0 is presented in Figure 6-25. T h e flow patterns are essentially similar in character to
the ones seen earlier (Figure 6-19). T h e presence of the central vortex is evident at
time-steps 60, 80, and 100 irrespective of the value of p/q. T h e corresponding typical
variation of the torque coefficient is presented in Figure 6-26 for two representative
values of p/q.
T h e average power coefficient variation w i t h the tip-speed ratio (Figure 6-27)
shows the same trends as the experimental results presented earlier (Figure 2-27).
Note, the theory predicts the peak Cp for p/q = 0.2, which is the same as that given
by the test results. A n error of around 16% in prediction of Cp jTnax for p/q — 0.2 is
indeed remarkably good considering the challenging character of the problem and
the quasi-potential nature of the model.
6.3 E f f e c t o f W i n d T u n n e l B l o c k a g e
T h e numerical computational scheme has the capability of changing the number
of tunnel elements as well as the length of the tunnel. In the present analysis the
tunnel wall was divided into 32 boundary elements. T h e finite tunnel walls extended
three diameters upstream and five diameters downstream. T h e size of the rotor was
kept constant while the w i d t h of the tunnel was altered with a consequent change
Chapter VI: Results and Discussion
x T, +ve
k = 20 k = 40 k = 60 * r, -ve
o T 2 +ve
+ T 2 -ve
p/q =o . P/q =0 p/q =0
P/q = 0.2 . P/q = 0.2 P/q = 0.2
. | . | i | i | i | i i I i I i I i I i I i
p/q = 1-0 . P/q = 1-0 p/q = 1.0
I 1
I 1
I 1
I 1
I 1
I ' i • I 1
-0.5 0.5 -0.5 0.5 1.5 -0.5 0.5 1.5 2.5

X (meters)
F I G U R E 6-25 E v o l u t i o n of the flow pattern as affected by p/q for A =

(k = 20,40,60).
k = 80
x r, +ve
v T, -ve
o r +ve
v
+ c D
fe 0 2
+ T -ve
2
*x* +++ +
p/q = 0
OV
p
, xxx xy
p/q = 0.2
£ 2 *X
X
p/q = 1-0
•1.5-1 1 1 1 1 1 1 1 1 1 1 r
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-25(cont.) E v o l u t i o n of the flow pattern as affected by p/q for A =

0.6 (A; = 80).
1.5-1
k = 100
1.0
x T, 4ve
+ CPQ
0.5 & 4 v T, - v e
5'0 o 4
O
O T 2 4ve
4
0.0 X4 + T 2 -ve
XX
-0.5 ^44+ 4+
x^ 4
XX 4+ +
-1.0 x x
p/q = 0
x
4 +++
x v
f
+ 4 +
-1.5 —I H
1.0
0.5 +
4 + eô
Y 0.0 *8o°
X
V V XX
-0.5 \ V X
4X 44
X
4+
-1.0 4 4
p/q = 0.2
w ++ 1
-1.5
1.0
0.5
0.0
-0.5
-1.0
P/q = 1-0
-1.5 1 1 1 1 1 1 1 1 1 1 1 1
0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-25(cont.) E v o l u t i o n of the flow pattern as affected by p/q for A =

0.6 (k = 100).
1.8-1
i I i i i i i | i
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (Sec.)
F I G U R E 6-26 T i m e history of C T as affected by p/q: A = 0.6; p/q = 1.0.

1.8-1 : : : : • : !
1.5H
-0.3-f j i i i 1 i 1 1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (Sec.)
F I G U R E 6-26(cont.) T i m e history of CT as affected by p/q: A = 0.6; p/q = 0.

0.40-1
n , , , ! , ! , ( , ! , ! , ,
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
A
F I G U R E 6-27 Theoretically predicted C P vs. A curves for p/q = 0 , 0 . 2 , 1 . 0 .
in blockage.
6.8.1 Stationary blade
Theoretically obtained transient torque coefficient for a stationary rotor, w i t h
p/q = 0.2 and (3 = 1 2 0 ° , as affected by the blockage is presented in Figure 6-
28. T h e procedure assumed was that of the surface separation with a constant
base pressure. Corresponding results for unconfined condition are also included to
facilitate comparison. Three interesting features become apparent :
a) T i m e to attain steady state is governed by the wall confinement and
increases w i t h the blockage.
b) Period of the torque fluctuations increases w i t h the blockage. T h i s is
related to the vortex formation and shedding. A s shown in Figure 6-29,
distance between the successive vortices increases as the blockage ratio
increases suggesting a delay in the vortex formation.
c) A s can be expected due to an increase in the local velocity w i t h blockage,
amplitude of the torque coefficient as well as the velocity with which the
vortices are convected increases.
Similar results were obtained for different angular positions of the stationary
blade with identical trends. Theoretically obtained average starting torque coeffi-
cient for a rotor with 10% blockage is compared with the corresponding experimental
data over a range of blade orientation (/?) in Figure 6-30. T h e correlation between
the results is indeed good. A s expected, the effect of blockage is not significant for
angles in the range 0 < j3 < 30° and 150° < p < 180°.
0.5
0.4-
*
DVM - C = Const.
p b
*
0.3 B= 0 %
7o */
B= 10 *
B = 20 %
0.2 /
t
t
/
0.1-
t
*
0.0-
/
*
y .
3^
-0.1-
-0.2-
-0.3
-0.4-
-0.5
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (Sec.)
F I G U R E 6-28 Time history of the torque coefficient for a stationary rotor, with
p/q = 0.2 at /? = 120°, as affected by the blockage.
1.5
x x x X
x
1.0
X ^ X * * XX
XX
0.5 x xx?x
X
xx:X*
x
^xxXx
0.0 "^XsTxx
k = 100
-0.5
-1.0
4-HX B= 0 %
-1.5
X ^
1.0 xxxx y
X X x x x
0.5 :^<xxx5< xxx x
X
XX x
+
xx&xx^xx
Y 0.0
4 +
4 4
-0.5
N
xx
+±.
4 4+4^ "
+
+
-1.0
+4 4 B = 10 %
-1.5
x "x x x x
1.0 X
xx x
X
x
0.5-
4 + 4+44 *****
0.0- x X
-0.5-
-1.0-
B = 20 %
-1.5-
I I 1 1 1 1 1 1 1 1 1 1 ,
0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-29 Flow pattern after 100 time-steps for a stationary rotor, with
p/q — 0.2 and /? = 120°, as affected by the blockage.
6.S.2 Rotating blade
Effectiveness of the analytical model for predicting wall confinement effects for
a m o v i n g rotor was also assessed through a comparison w i t h the experimental data
presented earlier. Four blockage ratios of 5%, 10%, 15% and 20% were used w i t h
the rotor configurations defined by p/q — 0.2 and 1. Variation of the average power
coefficient w i t h the tip-speed ratio is shown in Figures 6-31 and 6-32 for p/q = 1
and 0.2, respectively. A t the outset it is apparent that the theory is able to predict
the trends quite accurately. Even the magnitudes correlated rather well as shown
by the representative experimental results (dots). In general, the effect of blockage
to increase the peak power coefficient which now occurs at a higher tip-speed ratio.
Note, the theory consistently predicts the peak power coefficient, at a higher tip-
speed ratio. T h i s may be attributed to the incapability of the potential flow theory
to model the surface separation and the associated constant base pressure in the
wake. T h e numerical predictions can be improved further but at a computational
cost. In the present study, walls were modelled using fairly large boundary elements,
and the boundary conditions were satisfied only at the nodal points. Thus in essence,
the model treats the rigid wall as permeable. Even a small increase in the number
of elements (from 32 to 40) showed an improvement in the correlation between the
theoretical and experimental results. However, the computation cost increased by
about 40%.
Figure 6-33 summarizes the variation of Cp max with blockage for two values of
p/q. Note, the effect of the blade geometry parameter is to shift the plot without
altering its shape. Thus the functional variation remains essentially the same to a
constant (this concept was used earlier in Figure 2-22 to obtain the C P r n a x variation
for p/q = 0.2 using the experimental data for p/q = 1). E x p e r i m e n t a l data for the
0.50
p/q = 1.0
0.45- B= 0 %
B= 5 %
B = 10 %
0.40 B = 15 %
B = 20 %
Exper. B = 20 %
0.35-
0.30-
0.25
0.20-
0.15 —I 1
1—
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
X
F I G U R E 6-31 Theoretically predicted characteristic curves for a rotor with

p/q = 1.0 as affected by the blockage.
0.50
p/q = 0.2
0.45- B = 0 %
B = 5 %
B = 10 %
B = 15 %
0.40 \
B = 201 %
E x p e r . B = 10 %
\
Y
X
0.35-
0.30-
\
0.25-
0.20
0.15 -1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
A
F I G U R E 6-32 Theoretically predicted characteristic curves for a rotor with
p/q — 0.2 as affected by the blockage.
0.55
Theoretical
0.50- A p/q = 0.2
x p/q - 1.0
— A
Exper. p/q = 0.2
0.45-
F I G U R E 6-33 Theoretically predicted variation of Cp maxw i t h blockage for dif-

ferent p/q. Note, the agreement w i t h the experimental data is
quite good.
o p t i m u m configuration of p/q = 0.2 gave the m a x i m u m error of less than « 15%.
T h e theoretical model was also used to predict the evolution of the wake for a
rotating Savonius turbine in the presence of blockage as would be the case during
w i n d tunnel tests. Only a small sample of data for the o p t i m u m blade configuration
(p/q = 0.2) are presented in Figure 6-34 for A =• 0.8 and 1.2. T h e basic flow features
are the same as those in the unconfined case. For example, existence of the central
vortex is noticeable, even in the presence of blockage, at A = 0.8, however, it is
absent at A = 1.2 as in the unconfined case. A s discussed for the stationary case,
position of the vortex cluster continues to be farther downstream w i t h blockage
during rotation. Thus the vortex convective speed increases w i t h blockage even
during the blade rotation.

1.5-1
X = 0.8
1.0
x P. 4ve
0.5 v T, - v e
-to o r 4ve
2
0.0 + T -ve
2
+
+
-0.5
-1.0H B=0 %
-1.5
1.0-
0.5-
Y 0.0-
-0.5-
-1.0
B = 10 %
-1.5
1.0
0.5
0.0
+4
-0.5-
+ + 4 +_+
4
-1.0- #x 4
r
B = 20 %
-1.5 1 1 1 1 1 1 1 1 I 1 1 1
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-34 Flow pattern after 100 time-steps for a rotor w i t h p/q - 0.2 as
affected by the blockage(A = 0.8).
1.5
-1-5 T 1
I 1 1 1 1 1 l 1 1 1 1 1
- 0 . 5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
X (meters)
F I G U R E 6-34 Flow pattern after 100 time-steps for a rotor w i t h p/q = 0.2 as
affected by the blockage(A = 1.2).
C H A P T E R VII
CONCLUDING REMARKS
T h e thesis has attempted to lay a sound foundation for the design, development
and performance evaluation of a wind turbine, economical to fabricate and easy to
maintain. Using the classical Savonius configuration as reference, it has tried to
approach the objective through an organized experimental program, complemented
by theoretical approaches and flow visualization, w i t h innovative contributions at
all the stages.
T h e extensive experimental program, through a carefully planned parametric
study, has led to an o p t i m u m configuration for the Savonius rotor w i t h a twofold
increase in efficiency. Equally important are the results of blockage study which are
helpful in the design of a prototype. They also provide a basis for comparison of
results obtained by researchers using different sizes of models and test facilities.
O n the analytical side attention was focussed on two aspects of practical signif-
icance:
i) To develop a simple and quick approach of engineering accuracy that can
be used w i t h confidence, at least during the preliminary design stage.
T h e semi-empirical model, using the concept of a central vortex in con-
j u n c t i o n w i t h the measured pressure data on stationary blades, can ad-
mirably play this role. It can account for the effect of blade geometry
211
Chapter VII: Concluding Remarks 212
as well as blockage, and can effectively replace d y n a m i c a l testing w i t h a
considerable saving in cost.
ii) To develop a sophisticated mathematical model that can effectively sim-
ulate the transient, separated viscous flow associated w i t h the rotor, yet
maintaining simplicity of the potential flow. Furthermore, it was neces-
sary to achieve a desired accuracy w i t h an acceptable cost using available
computational tools. The numerical, time-marching approach based on
the B o u n d a r y Element M e t h o d is developed in the thesis which meets
all of the criteria mentioned above. T h i s elegant tool is able to predict
performance of the rotor and the wake dynamics w i t h an accuracy that
is limited, to an extent, by the computational cost. Its effectiveness can
be further improved by judiciously incorporating information that can be
measured quite readily. Furthermore, the approach is rather general and
can be applied quite readily to a large class of transient aerodynamics
problems involving a variety of slender a n d / o r bluff geometries.
Based on the w i n d tunnel test program, flow visualization study and mathe-
m a t i c a l modelling of the Savonius rotor, following general conclusions can be made:
a) T h e o p t i m u m configuration of the Savonius rotor corresponds to the blade
geometry as follows:
gap-size (a/d) 0 ;
blade overlap (b/d) 0 ;
aspect ratio (^4) 0.77 ;
blade shape parameter [p/q) = 0.2 ;
blade arc angle (0) = 135°.
It is inferred that this has a peak power coefficient of around 0.32 at a tip-
Chapter VIJ: Concluding Remarks 213
speed ratio of 0.79 (with a possible error of ±5 — 10%) in the unconfined
condition representing a twofold improvement in performance compared
to the reported values in literature ( « 0.12 — 0.15). T h i s brings it close
to more sophisticated and costly configurations like the Darrieus design
in terms of performance, yet retaining its simplicity and self starting
character.
b) Characteristic plots of Cp vs. A suggest that the Savonius rotor is not a
pure drag device. T h i s is logical as at small 0 the rotor behaves like a
slender body w i t h lift contributing to the power.
c) T h e effect of blockage is to increase the peak power coefficient and the
corresponding tip-speed ratio. A variation in the blade shape parameter
(p/q) shifts the C p > m a x vs. blockage plot along the ordinate without
affecting its general shape.
d) T h e semi-empirical approach using experimentally established parame-
ters can predict performance of the rotor with a reasonable accuracy.
e) Presence of a vortex filament at the centre of the rotor and its shedding
which is predicted by the analytical model is also confirmed through the
flow visualization study.
f) A mathematical model for viscous separated flow w i t h an unsteady wake,
developed in the thesis, represents a versatile tool applicable to a large
class of transient fluid dynamics problems. W h e n applied to a Savonius
rotor it is able to predict performance and evolution of the wake geometry
w i t h a reasonable accuracy.
g) T h e wake of the stationary blade shows periodic behaviour over the range
/? « 6 0 ° —150°. T h e vortex shedding frequency corresponds to a Strouhal
number of « 0.12. T h e theoretically predicted peak starting torque coef-
ficient occurs at (3 ss 30° which is confirmed by the experimental results.
Starting torque becomes negative over the range of f3 « 130° — 180°.
h) For the rotating blade, existence of the central vortex and its shedding is
evident for tip-speed ratio less than one. However, for A > 1 theoretical
results do not show presence of the central vortex. Periodicity of the
wake geometry is reflected in the time history of CT for b o t h stationary
and rotating cases.
i) For a given blockage ratio, the effect of p/q is to shift the C p vs. A
characteristic curve along the ordinate. In general, p/q does not have a
significant effect on the wake geometry.
There are several avenues for future studies which are likely to be profitable
and satisfying :
a) Measurement of torque and pressure distribution d u r i n g blade rotation
should provide valuable information for comparison w i t h and further re-
finement of the mathematical model.
b) W i t h an o p t i m u m blade configuration in hand, the next logical step would
be to build a prototype model for field tests to investigate its operational
performance, structural and mechanical integrity, and maintenance re-
quirements. W i n d tunnel tests of the o p t i m u m configuration simulating
the earth boundary layer and the turbulent character as encountered in
field operation should provide valuable information. Its performance cor-
relation w i t h the field data has not been reported in the literature.
c) T h e use of augmentors to further improve the performance presents an
exciting possibility.
d) Improvement in the semi-empirical model through the use of more sophis-
ticated functional relations is likely to make the method more attractive.
e) There are several possible avenues to further improve the Boundary E l -
ement M o d e l . Perhaps the most obvious one is to incorporate a suitable
boundary layer model to predict separation and base pressure w i t h added
accuracy.
f) A p p l i c a t i o n of the B o u n d a r y Element M e t h o d in the solution of transient
fluid dynamics problems of aircraft and industrial aerodynamics as well
as ocean engineering presents a challenging possibility.

216
BIBLIOGRAPHY
A h m e d i , G . , 1978,"Some P r i l i m i n a r y Results on the Performance of a Small Vertical

A x i s C y l i n d r i c a l W i n d Turbine", Wind Engineering, V o l . 2, N o . 2, pp. 65-
74.
A l d e r , G . M . , 1978, "The A e r o d y n a m i c Performance of the Savonius R o t o r " , Inter-

national Symposium of Wind Enregy Systems, BHRA, A m s t e r d a m , Nether-
land, pp. E 1 0 119-126.
Aldos, T . K . , 1984, "Savonius R o t o r Using Swinging Blades as an Augmentation

System", Wind Engineering, V o l . 8, N o . 4, p p . 214-220.
Alexander, A . J . , 1978. " W i n d Tunnel Corrections for Savonius R o t o r s " , Interna-

tional Symposium of Wind Enregy Systems, BHRA, A m s t e r d a m , Netherland,
V o l . 1, pp. E 6 69-80.
A k i r a , A . , Shigeo, K . , 1983, " A M e t h o d of Calculation on the A i r Loadind of

Vertical A x i s W i n d Turbines", Proceedings of the \% Intersociety Energy
th
Conversion Engineering Conference, Orlando, F l o r i d a , U S A .
B a c h , V o n G . , 1931, "Vutersuchungen Uber Savonius Rotoren und Verwandte Stro-

mungsmaschinen", Forsh. auf dem Gebiete des Ingenierwesens, V o l . 2, pp.
218-231 (in G e r m a n ) .
Banerjee, P . K . , Butterfield, R . , 1981 Boundary Element Methods in Engineering

Science, Publisher M c G r a w - H i l l Book C o , U . K . .
Banerjee, P . K . , Davis, T . G . , 1979, "Analysis of Some Case Histories of Later-

ally Loaded Pile G r o u p s " , Proceedings of the International Conference on
Numerical Methods in Offshore Piling, Institute of C i v i l Engineers, London.
Base, T . E . , Russel, B . , 1976, "Computer A i d e d Aerogenerator Analysis and Perfor-

mance", International Symposium of Wind Enregy Systems, BHRA, Cam-
bridge, p p . A 4 53-74.
B e t z , A . , 1926, " W i n d Energy and its E x p l o i t a t i o n by W i n d m i l l s " , Gottingen: Van-

denhoeck und Ruprecht,p. 64 (in G e r m a n ) .
217
B l a c k w e l l , B . F . , Feltz, L . V . , 1975, " W i n d Energy a Revitalized P u r s u i t " , Report

Sandia Laboratories, Albuquerque, New M e x i c o , SAND-75-01666.
B o w d e n , G . J . , M c Aleese, S. A . , 1984, "Properties of Isolated and C o u p l e d Savonius

R o t o r s " , Wind Engineering, V o l . 8, N o . 4, pp. 271-288.
C h o r i n , A . J . , 1973, "Numerical Study of Slightly Viscous F l o w " , Journal of Fluid

Mechanics, V o l . 57, pp. 785-796.
Clements. R . R . , 1973, " A lnviscid M o d e l of T w o Dimensional Vortex Shedding",

Journal of Fluid Mechanics, V o l . 57, part 2, pp. 321-336.
Darrieus, G . J . M . , 1931, "Turbine H a v i n g its R o t a t i n g Shaft Transverse to the

Flow of the C u r r e n t " , U n i t e d States Patent N o . 1835, 018.
De Vries, O . , 1979, " F l u i d D y n a m i c Aspects of W i n d Energy Conversion", AGAR

Do graph, N o . 243, A G A R D , Neuilly-suv-Seine, France, p. 150.
Fage. A . , Johansen, F . C , 1927, " O n F l o w of A i r B e h i n d an Inclined F l a t Plate of

Infinite S p a n " , Proceedings of Royal Society, A-116, pp. 170-197.
Feftz. L . V . . Blackwell, B . F . , 1975, " A n Investigation of R o t a t i o n Induced Stresses

of Straight and C u r v e d Vertical A x i s W i n d Turbine Blades", Report Sandia
Laboratories, Albuquerque, New M e x i c o , SAND-74-0379.
Fernando, M . S. U . K . , M o d i , V . J . , O h t a , G . , Yokomizo, T . , 1987, ' A Numerical

Analysis of the Unsteady Flow Past a Savonius W i n d T u r b i n e " , Proceedings
of the II Canadian Congress of Applied Mechanics, V o l . 2, pp. E 82-83.
th
Gieseng. J. P., 1969, "Vorticity and K u t t a C o n d i t i o n for Unsteady Multi-energy

Flows", Transactions of ASME. Journel of Applied Mechanics. V o l . 36, pp.
608-613.
G l a u e r t , H . , 1935a, " W i n d m i l l s and Fans", Aerodynamic Theory, E d i t o r : F . W .

D u r a n d , V o l . 4, D i v . L , pp. 324-341.
218
G l a u e r t , H . , 1935b, " A i r Screw T h e o r y " , Aerodynamic Theory. E d i t o r : F . W . D u -

rand, V o l . 4, D i v . L , pp. 170-293.
G o v i n d a R a j u , S. P., Narashima, R . , 1979, " A Low Cost Water P u m p i n g W i n d m i l l

Using a Sail T y p e Savonius R o t o r " , Dept. of Aeronautical Engineering,
Indian Institute of Science, Report 79 F M 2 , Bangalore, India.
Greet, R . J . , 1980, " M a x i m u m W i n d m i l l Efficiency", Journel of Applied Physics,

V o l . 51, N o . 9, pp. 4680-4681.
Hatayama, E . , N a k a t a n i , H . , M i y a i , Y . , 1984, " A n Analysis of the Flow Past an

S-Shaped R o t o r by Discrete Vortex M e t h o d " , Transactions of the Japan
Society for Aeronautical and Space Sciences, V o l . 27, N o . 77, p. 169.
Hess, J . L . , 1972, "Calculation of Potential Flow A b o u t A r b i t r a r y Three Dimen-

sional Lifting Bodies", M D D Report N o . J 5679-01, M c Donnell-Douglas
C o r p . , St. Louis, M o . , U S A .
Hess. J . L . , S m i t h , A . M . O., 1962, "Calculation of N o n - L i f t i n g Potential Flow

A b o u t A r b i t r a r y Three Dimensional Bodies", Douglas Aircraft Report N o .
E S 40622, L o n g Beach, California.
Hess. J . L . , 1974, "The P r o b l e m of Three Dimensional Lifting Potential Flow and

its Solution by Means of Surface Singularity D i s t r i b u t i o n " , Computational
.Methods in Applied Mechanics and Engineering, V o l . 4, pp. 283-319.
Holme, O . A . M . , 1976, " A Contribution to the Aerodynamic Theory of the Vertical

A x i s W i n d Turbine", International Symposium of Wind Enregy Systems.
BUR A, Cambridge. U . K . , pp. C4 55-72.
Holme. O . A . M . , 1981, "Detailed A e r o d y n a m i c Analysis of H o r r i z o n t a l Axis W i n d

Turbines", V K I Lecture Series 9, Rhode, St. Genese, B e l g i u m , V o n - K a r m a n
Institute of F l u i d D y n a m i c s , p. 39.
H u n t , B . , Semple, W . G . , 1976. "Economic Improvements to the M a t h e m a t i c a l

M o d e l in a Plane-Constant Strength Panel M e t h o d " , Paper Presented at
Euromech Colloq., N o . 75, Rhode, West Germany.
219
H u n t , B . , 1980, "The M a t h e m a t i c a l Basis and Numerical Principles of the B o u n d -

ary Integral M e t h o d for Incompressible and Potential Flow Over 3-D Aero-
dynamic Configurations", Numerical Methods in Applied Fluid Dynamics,
A c a d e m i c Press, L o n d o n , pp. 49-135.
H u t t e r , U . , 1977, " O p t i m u m W i n d Energy Conversion Systems", Annual Review of

Fluid Mechanics, V o l . 9, pp. 399-419.
Imai, I., 1974, Fluid Dynamics, V o l . 1, Shokabo, Tokyo, pp. 408-415 (in Japanese).
Inamuro, T . , et a l . , 1983, " A N u m e r i c a l Analysis of Unsteady Separated Flow by-

Vortex Shedding M o d e l " , Bulletin of JSME, V o l . 26, N o . 222, pp. 2106-
2112.
Inglis. D . R . , 1979, " A W i n d m i l l s Theoretical M a x i m u m E x t r a c t i o n of Power From

the W i n d " , American Journal of Physics, V o l . 47, N o . 5, pp. 416-420.
Jones. C . N . , L i t t e r . R . D . , Manser, B . L . , 1979, "The Savonius Rotor-Performance

and F l o w " , Proceedings of the First BWEA Wind Energy Workshop, pp.
102-108.
J o r d a n . S. K . , F r o m m , J . E . , 1972, "Oscillatory D r a g , Lift and Torque on a Circular

C y l i n d e r in a Uniform F l o w " , The Physics of Fluids. V o l 15, pp. 371-376.
K a t z , J . , 1981, " A Discrete Vortex M e t h o d for the Non-Steady Separated Flow-

Over an A i r f o i l " , Journal of Fluid Mechanics, V o l . 102, pp. 315-328.
K h a n , M . H . , 1978, "Mode! and Prototype Performance Characteristics of the Savo-

nius R o t o r W i n d m i l l " , Wind Engineering, V o l . 2, N o . 2, pp. 75-85.
K i m u r a . S., et a l . . 1984, " W i n d Tunnel Test for the Verification of a New Approach
to the Analysis of V A W T " , European Wind Energy Conference, Hamburg.
F.R.G..
K i y a , M . , A r i e , M . , 1977, " A C o n t r i b u t i o n to an Inviscid Vortex Shedding Model

for an Inclined Flat Plate in Uniform F l o w " , Journal of Fluid Mechanics,
V o l . 82, pp. 223-240.
220
K u w a h a r a , K . , 1973, "Numerical Study of Flow Past an Inclined F l a t Plate by an

Inviscid M o d e l " , Journal of Physical Society of Japan. V o l . 35, pp. 1545-
1551.
Lissaman, P. B . S., 1976, "General Performance Theory for Cross W i n d A x i s Tur-

bines", International Symposium of Wind Enregy Systems, BHRA, Cam-
bridge, pp. C2 21-38.
M a i r , W . A . , M a u l l , D . J . , 1971, "Bluff Bodies and Vortex Shedding", Journal of

Fluid Mechanics, V o l . 45, p. 209.
M a j o l a , O . O . , 1986, "The Aerodynamic Design of the Savonius R o t o r " , Journal of

Wind Engineering and Industrial Aerodynamics, V o l . 21, N o . 2, pp. 223-231.
M a j o l a , O . O . , Onasanya, O . E . , 1981, "Performance Testing of a Savonius W i n d m i l l

Rotor in Shear F l o w " , Proceedings of the 16 Intersociety Energy
th
Conversion
Engineering Conference, A t l a n t a , Georgia, pp. 2041-2046.
M a r j a r i a , M . , Mukherjee, S., 1980, "Improved Boundary Integral Equations for

Three Dimensional Elasto-Plastic F l o w " , International Journal of Solids and
Structures, V o l . 17, pp. 144-151.
Manser, B . L . , Jones, C . N . , 1975, "Power from W i n d and Sea-the Forgotten

panemon", Paper presented al Thermo-Fluids Conference: Energy-Transport
-ation, Storage, and Conversion, Brisbane, A u s t r a l i a .
M a s k e l l , E . C , 1965, " A Theory of the Blockage Effects on Bluff Bodies and Stalled
Wings in a Closed W i n d T u n n e l " , Aero. Research Council. Reports and
M e m o r a n d a 3400.
Morcos, S. M . , Khalafallah, M . G . , Heikel, H . A . , 1981, "The Effect of Shieldind on

the A e r o d y n a m i c Performance of the Savonius W i n d Turbines". Proceedings
of the 16 Intersociety Energy Conversion Engineering Conference, A t l a n t a ,
th
Georgia. V o l . 2, pp. 2037-2040.
Nagano, S., et a l . , 1981, "Discrete Vortex M e t h o d Analysis of the Flow Past a

Rectangular P r i s m " , Transactions of JSME, V o l . 47, N o . 413, p. 32 (in
Japanese).
221
N e w m a n , B . G . , 1974, " Measurements on Savonius R o t o r w i t h Variable G a p " ,

Proceedings of the Sherbrooke University Symposium on Wind Energy, Sher-
brooke, C a n a d a , p. 116.
Ogawa, T . , 1984, "Theoretical Study on the Flow A b o u t Savonius R o t o r " , ASME

Journal of Fluids Engineering, V o l . 106, pp. 85-90.
Ogawa, T . , et al., 1986a, " W i n d Tunnel Performance D a t a on the Savonius Rotor

W i t h C i r c u l a r G u i d e Vanes", Bulletine of JSME, V o l . 29, N o . 253, pp.
2109-2114.
Ogawa, T . , Yoshida, H . , 1986b, "Effects of a Deflecting Plate and R o t o r E n d Plates

on Performance of Savonius T y p e W i n d Turbine", Bulletine of JSME, V o l .
29, N o . 253, pp. 2115-2121.
P r a n d t l , L . , 1918, "Theory of the Lifting W i n g " , I. Nachr. Ges. Wiss. Gottingen,

M a t h . P h y s . K l . , pp. 451-477 (in German).
R a n g i , R . S., South, P., T e m p l i n , R . J . , 1974, " W i n d Power and the Vertical Axis
W i n d Turbines Developed at the National Research C o u n c i l " , DME/NAE
Quarterly Bulletine, N o . 2.
R a u h , A . , Seelert, W . , 1984, "The Betz O p t i m u m Efficiency for W i n d m i l l s " , Applied

Energy, V o l . 16, pp. 1-9.
Sabzevari, A . , 1978, "Power A u g m e n t a t i o n in a D u c t e d Savonius R o t o r " , Interna-

tional Symposium of Wind Enregy Systems, BHRA, A m s t e r d a m , Netherland,
V o l . 1, pp. F 3 25-34.
Sarpkaya, T . . 1975, " A n Inviscid M o d e l of T w o Dimensional Vortex Shedding

for Transient and A s y m p t o t i c a l l y Steady Separated Flow Over an Inclined
P l a t e " , Journal of Fluid Mechanics, V o l . 68, Part 1, pp. 109-128.
Savino, J . M . , E l d r i d g e , F . R . , 1975, " W i n d Power", Astronautics and Aeronautics,

Vol.13, N o . 11, pp. 53-57.
Savonius, S. J . , 1928, "The W ing R o t o r in Theory and P r a c t i c e " , Savonius C o . ,

7
Finland.
222
Savonius, S. J . , 1931, "The S-Rotor and its A p p l i c a t i o n " , Mechanical Engineering,

V o l . 53, N o . 5, pp. 333-338.
Sawada, T . , et al., 1985, "Blade Force Measurement and Flow Visualization of

Savonius Rotors", Nippon Kikai Gakkai Ronbunshu B Hen, V o l . 51, N o .
471, p p . 3743-3747.
Shankar, P . N . , 1976, "The Effects of Geometry and Reynolds Number on Savonius

T y p e Rotors'", Technical M e m o r a n d u m , A E - T M - 3 - 7 6 , National Aeronautical
Laboratory. Bangalore.
Simonds, M . H . , Bodek. A . , 1964, "Performance Test of a Savonius R o t o r " , Tech-

nical R e p o r t , No. T 10. Published by Brace Research Institute, M c D o n a l d
College of M c G i l l University, Ste. A n n e de Bellevue 800, Quebec, Canada.
Sivasegaram, S., 1977, "Design Parameters Affecting the Performance of the Resis-
tant T y p e Vertical A x i s W i n d R o t o r s " , Wind Engineering, V o l . 1, N o . 3,
pp. 207-217.
Sivasegaram, S., 1978a, " A n E x p e r i m e n t a l Investigation of a Class of Resistance

type Direction Independant W i n d Turbines", Energy, V o l . 3, pp. 23-30,
Pergamon Press, U . K . .
Sivasegaram, S., 1978b, "Secondary Parameters Affecting the Performance of Re-

sistance T y p e Vertical A x i s W i n d R o t o r s " , Wind Engineering, V o l . 2, N o .
1, pp. 49-58.
Sivasegaram. S., 1979, "Concentration Augmentation of Power in a Savonius Type

W i n d R o t o r " , Wind Engineering, V o l . 3, N o . 1, pp. 52-61.
Sivasegaram. S., 1982, "Improvement of the Performance of Slow-Running Vertical

A x i s W i n d Rotors", Renewable Energy Review Journal, V o l . 4, p. 36.
Sivasegaram, S., Sivapalan, S., 1983, "Augmentation of Power in Slow Running

Vertical A x i s W i n d Rotors U s i n g M u l t i p l e Vanes", Wind Engineering, V o l .
7, N o . 1, pp. 12-19.
South, P., R.angi, R . S., 1972, " A W i n d Tunnel Investigation of a 14ft. Diameter
Vertical A x i s W i n d m i l l " , N a t i o n a l Research C o u n c i l . Canada, Report L T R -
LA-105.
223
Spalart, P . R . , et a l . , 1983, "Numerical Simulation of Separated F l o w s " , Report

N A S A TM-84328.
Stansby, P . K . , D i x o n , A . J . , 1983, "Simulation of Flows A r o u n d Cylinders by

Lagrangian Vortex Scheme", Applied Ocean Research, V o l . 5, p p . 167-178.
Stansby, P . K . , 1985, " A Generalized Discrete Vortex M e t h o d for Sharp Edge C y l i n -

ders", AIAA Journal, V o l . 23, N o . 6, p p . 856-861.
S t r i c k l a n d , J . H . , F o r d , S. R . , Reddy, G . B . , 1975, " T h e Darrieus Turbine: A

Summery R e p o r t " , Proceedings of the Second US National Conference on
Wind Engineering Research, Colarado State University, p. V 2 1.
S t r i c k l a n d , J . H . , 1975, " T h e Darrieus Turbine: A Performance P r e d i c t i o n M o d e l

U s i n g M u l t i p l e Streamtubes", Sandia Laboratory Report, S A N D 75-0431.
S t r i c k l a n d , J . H . , 1976, " A Performance Prediction M o d e l for the Darrieus Turbine",

International Symposium of Wind Enregy Systems, BHRA, Cambridge, pp.
C 3 39-54.
S u n d a r a m , P . , G o v i n d a R a j u , S. P . , 1980, "Performance of Savonius R o t o r s " , F l u i d

Mechanics Reports, Dept. of Aeronautical Engineering, Indian Institute of
Science. Report 8 0 F M 7 . Bangalore, India.
S u z u k i , T . , O k i t s u , H . , 1982, "Characteristics of a Savonius W i n d m i l l Power System

with a Synchronous Generator", Wind Engineering, V o l . 6, N o . 3, pp. 131-
139.
T e m p l i n , R . J . , 1974, "Aerodynamic Performance Theory for the N R C Vertical A x i s

W i n d Turbine", National Research Council Laboratory Technical Report,
LTR-LA-160.
V a n Dusen, E . S., Kirchhoff, R . H . , 1978, " A T w o Dimensional Vortex Sheet

M o d e l of a Savonius R o t o r " , Fluids Engineering in Advanced Energy Sys-
tems, A S M E , pp. 15-31.
Weissinger, J . , 1947, " T h e Lift D i s t r i b u t i o n on Swept Back W i n g s " , N A C A T M

1120, N A C A , Washington D . C , p. 48.
224
W i l s o n , R . E . , Lissaman, P. B . S., 1974, " A p p l i e d Aerodynamics of W i n d Power

Machines", N S F / R A / N - 7 4 1 1 3 , Oregon State University, Corvallis, p. 118.
W i l s o n , R . E . , Lissaman, P . B . S., Walker, S. N . , 1976, "Aerodynamic Performance

of W i n d Turbines", E R D A / N S F / 0 4 0 1 4 - 7 6 1 1 , pp. 111-164.
Zienkiewicz. O . C , 1978, The Finite Element Method, 3 r d

E d i tion. M c - G r a w - H i l l .
London.

On The Performance and Wake Aerodynamics of The Savonius Wind Turbine

Uploaded by

Copyright:

Available Formats

You might also like

On The Performance and Wake Aerodynamics of The Savonius Wind Turbine

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

On The Performance and Wake Aerodynamics of The Savonius Wind Turbine

Uploaded by

Copyright:

Available Formats

ON T H E P E R F O R M A N C E AND WAKE AERODYNAMICS

OF T H E SAVONIUS WIND TURBINE

M A H A M A R A K K A L A G E SAMAN UDAYA K U M A R FERNANDO

B . S c . E n g . ( H o n s ) , University of Peradeniya, Sri L a n k a , 1980

THE REQUIREMENTS FOR T H E DEGREE OF

We accept this thesis as conforming

to the required standard

© Mahamarakkalage Saman U d a y a K u m a r Fernando, 1987

Department of Mechanical Engineering

T h e objective of the thesis is to establish methodology for development of a wind

Following a brief review of relevant significant contributions in the field (Chapter

W i t h the design a n d performance results i n h a n d , Chapters III — V I focus

Chapter I V describes in detail a relatively more sophisticated and rigorous

T h e thesis ends w i t h concluding remarks and a few suggestions concerning pos-

1.1 Development of W i n d Turbines 4

2.1 Definition of the B l a d e Shape 19

2.4.3 Blade aspect ratio 40

C h a p t e r III A SEMI-EMPIRICAL APPROACH FOR

3.1 Inadequacy of the Quasi-Steady A p p r o a c h 67

Chapter I V A MATHEMATICAL MODEL FOR

4.1 Boundary Element M e t h o d ( B E M ) 85

Chapter V NUMERICAL INVESTIGATION 106

5.1 F o r m a t i o n of the S y s t e m M a t r i c e s 106

Chapter V I RESULTS A N D DISCUSSION 138

6.1 Stationary Blade 139

6.3.2 R o t a t i n g blade 204

Chapter VII CONCLUDING REMARKS 211

5.1 Effect of number of elements on the computational cost 127

1- 1 A n illustrative schematic diagram of a w i n d energy

2-17 Bearing loss corrected characteristic curves at three

3-7 Identification of the modified quasi-steady approach

5-13 T i m e history of CT as affected by the time-step size 135

6-23 Theoretically predicted variation of the average

w i t h blockage for different p/q 207

base pressure coefficient

CT torque coefficient, Tj\pXJ hdR2

Pf frictional power loss

(jy*y) polar coordinates of the blade contour w i t h

r, , T s w strengths of the bound vortices

§c integration along an arbitrary path around the rotor

h integration along the blade contour

The author is grateful to the departments of Mechanical Engineering and Physics

The project was supported by the N a t u r a l Sciences and Engineering Research

F i n a l l y , the author would like to express his gratitude to his wife, D h a m m i k a

water and w i n d represent the elements frequently exploited by the Indus-Valley

C i v i l i z a t i o n of Ariyans and the E g y p t i a n Dynasty of the Pharaohs. W i n d m i l l s were

in existence in Persia as early as in 134 B . C . ( A h a m e d i , 1978). In western Europe,

still in use (Savino and E l d r i d g e , 1975).

T h e attention towards utilization of w i n d energy (besides other alternative en-

ergy sources) may be attributed to increased consciousness about the depleting

ucts of their technologies. Unfortunately, the heavy manufacturing industry, which

supplies these products, uses an inordinate quantity of energy. Heavy consumption

d a m m i n g of the rivers for hydro-electric power production, thermal p o l l u t i o n , and

related to the standard of living. Conversely, by improving the energy supply to

a certain community the standard of living can be upgraded. A large percentage

electric power is practically impossible for such communities. A l t h o u g h the rural

lack of energy. P u m p i n g water for irrigation is an important use of energy in rural

energy use. In this situation, windmills appear particularly promising. A schematic