Professional Documents
Culture Documents
On The Performance and Wake Aerodynamics of The Savonius Wind Turbine
On The Performance and Wake Aerodynamics of The Savonius Wind Turbine
On The Performance and Wake Aerodynamics of The Savonius Wind Turbine
by
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
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)
T H E UNIVERSITY OF BRITISH C O L U M B I A
July 1987
c
iii
ABSTRACT
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 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
C h a p t e r II O P T I M U M CONFIGURATION STUDIES 19
BIBLIOGRAPHY 216
viii
LIST O F T A B L E S
TABLE Page
LIST OF F I G U R E S
FIGURE Page
FIGURE Page
FIGURE Page
FIGURE Page
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
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
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)
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.
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,
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
character of the fossil fuel supply. For the technologically advanced countries, a
favourable balance of payments has always depended heavily on the exported prod-
of fossil fuels with the accompanying degradation of the quality of air, excessive
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
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
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
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
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
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
energy per unit mass of a moving air parcel is | c 7 , and the mass flow rate through
2
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
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
for w i n d turbines. In particular, since the derivation of the Betz efficiency appears
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
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
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.
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
by drag forces is usually low (De Vries, 1979) compared to the lift force driven
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
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
logically developed countries, for generation of electric power on a large scale. This
prime mover is required to start the machine adding more sophistication to the
system.
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
both aerodynamically and structurally, already exists. However, due to its unde-
sirable starting characteristics and complexity in design, the Darrieus rotor is not
Chapter I: Introduction 7
1.2 T h e S a v o n i u s T u r b i n e
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:
- acceptance of w i n d from any direction thus eliminating the need for re-
orientation;
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
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
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
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
rotor. He altered the basic shape of the blade by connecting two circular arc portions
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-
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
geometry were carried out by several researchers in late 1970's and early 1980's.
Chapter I: Introduction 10
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-
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
Sivasegaram (1977, 1978a, 1978b, 1982) has done a remarkable amount of work
the results he concludes that the two-bladed configuration has more potential for
of the Savonius rotor. His tests were carried out in a closed-circuit w i n d tunnel.
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
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
the nature of the flow around the rotor at a tip-speed ratio of 0.2.
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
using smoke to better understand the flow around the Savonius rotor were also
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
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
housing. The design proved to be effective independent of the wind direction and
Sivasegaram (1979) has tested several straight walled ducted concentrator con-
design and fabrication was the m a i n objective. He was able to increase the power
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
-•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
rotor.
power results were obtained over a range of wind speeds. Design criteria based on
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
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
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
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
T h e flow visualization results and the comments made on the flow patterns
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
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
cluded that the lift force produces the torque over a large range of angular positions.
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
Chapter I: Introduction 14
m a x i m u m efficiency.
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
(b) .Almost all the tests were carried out using models in fairly small wind
(d) In all the test results, bearing friction losses were neglected. Unfortu-
nately, due to small models, they may affect the performance substan-
tially.
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
Chapter I: Introduction 15
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
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
Vortex and panel methods are also widely used in the analysis of Darrieus type
a better model analogous to the theory for a w i n g w i t h a finite span can now be
(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
based on local circulation method to calculate the air loading on the Darrieus tur-
et a l . , 1984).
Due to the completely different nature of the flow around the Savonius rotor the
For the Savonius rotor an analytical model was developed by Wilson et al.
Chapter I: Introduction 16
(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
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
H a t a y a m a et al. (1984) analysed the problem using the discrete vortex method
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
also directed towards the aerodynamic problem of modelling the unsteady separated
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
tunnel test program. Effect of the blade geometry, w i n d tunnel blockage as well
distribution and starting torques have been measured for the stationary blade at
different angles.
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
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
T h e flow fields corresponding to different angular positions of the blade are also
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
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
study.
CHAPTER II
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
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
circular arc by a straight line (arc of infinite radius, Figure 2-1). He conducted
provement that can be realized proved to be inconclusive. In the present study, the
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 ;
h blade height ;
W w i d t h of the w i n d tunnel ;
To help establish the two dimensionality of the flow, end plates are used.
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
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
i.e.,
2.10) as,
Cp = Cp
{l^^l^^W ^) B
' (2
- 3)
where
du>
A =
1 ( 2
- 6 )
is called the blade aspect ratio. R n is the Reynolds number based on free stream
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-
As pointed out in Chapter I, the flow past the turbine blade is so complex that
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
isfy certain fluid dynamical similarity w i t h the prototype machine. Geometric sim-
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-
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
In the present test program, relatively larger models and higher w i n d speeds
velocity.
Chapter II: Optimum Configuration Studies 24
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
positive corrections. Obviously, the prediction would not be reliable unless a proper
carried out to establish the effect of blockage, thus providing an approach to obtain
reliable data.
2.3 E x p e r i m e n t a l S e t - u p
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-
stable wind speed in the range 2.5 m/s — 25 m/s and is fitted w i t h emergency stop
2.44 m dia..
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
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
Chapter II: Optimum Configuration Studies 28
ball bearings. The lower end of the shaft was flexibly coupled to a dynamometer.
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
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,
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
Vanes Vane
Additional
Vanes
using this type of dynamometer depends on the rotational speed of the shaft. The
cal vanes on the inside wall of the outer cylinder (Figure 2-5). T h o u g h the concept
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
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.
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
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
.A"
8.0-
^ 6.0 .A'
^A
8 !
2.0
1.0
0.0
80 100 120 140 160 180 200 220 240 260 280 300 320 340
RPM
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
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
ary blade at different angles. This provided rather important information such as
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
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
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
T h e pressure transducer was calibrated before each run and all the taps were
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
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
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
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
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
2.4-1 Gap-size
blade-gap are shown in Figure 2-10. M o r e informative would be the effect of percent-
381 mm
End Plate
203 mm-
330 mm
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
0.20
0.18 H
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
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
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.
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
0.25
0.24 H
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
« 0.6 m 2
were used.
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
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
Uncorrected Cp vs. A curves at three different w i n d speeds for p/q = 1.6 are
Chapter II: Optimum Configuration Studies 44
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
Chapter II: Optimum Configuration Studies 45
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
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
shown in Figure 2-16. In general, one would expect variation of the power coefficient
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
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
w i t h p/q. Under the given conditions, the m a x i m u m power coefficient was found to
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
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
0.55-1
•
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
0.55-1
p/q
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
in this range suggest that the effect of the Reynolds number in the operating range
is likely to be insignificant.
because of different test conditions. One of the major parameters affecting the test
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 wall confinement from 5 to 20% can raise the C p m a i by around 70% thus leading
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
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
0.6
0.5-
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
2.5 P r e s s u r e D i s t r i b u t i o n
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-
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
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
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.
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.
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.
Chapter II: Optimum Configuration Studies 62
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
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
in a region of rise in pressure. Now the circular cylinder type pressure profile begins
constant value thereafter. Thus the flow separates twice on the back side and once
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.
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
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.
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).
geometry was measured at two w i n d speeds and is shown in Figure 2-24. Note,
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
acceptable error margin (maximum error « 5%). T h i s suggests that the assumption
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
CT Torque Coefficient ;
Cp Power Coefficient ;
C p Pressure Coefficient ;
A Tip-Speed R a t i o .
p ± 0.08 N/m 2
reading of the pressure transducer.
\, 0.025 , .0046"
x
6X = ± (0.0032 + — — )A + — - 7
Chapter IF: Optimum Configuration Studies 66
which amounts to
Similarly
bC r = ± 0.0016+ 0 . 0 7 5 1 C T
at U = 4.43 m/s ;
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
crease in w i n d velocity.
CHAPTER III
rotor performance prediction using the stationary blade pressure distribution data
presented in Chapter II. Such an approach, if successful, can replace dynamic testing
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
angle between the normal vector and the radius vector. ,/? is the blade angle with
t
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
sure distribution was measured at 10° intervals (section 2.5). Linear interpolation
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
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
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
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
X
*x
X
legend
Q-S Method
x Experimental
T h e flow visualization study was carried out using the tow-tank facility located
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
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.
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
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
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).
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 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
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
5, • FAX) +F {B,R )
2 n
(3.2)
S 2 •• S {p/q,R )
2 n (3.3)
Si , when varied, appeared to model the blockage effect. Influence of the Reynolds
insignificant in the operating range. Using the experimental results for p/q — 0,
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%,
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
Chapter III: Semi-Empirical Approach 77
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
0.5-1
0.4
0.3-
0.2-
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
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
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
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-
experimental results for these three cases are shown in Figure 3-10.
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-
Chapter III: Semi-Empirical Approach 81
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
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
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
flow, and will demand application of the complete Navier-Stokes equations. Even
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
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
84
Chapter IV: Mathematical Analysis 85
literature. They involve time stepping, finite element or difference solutions of 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
grid of impracticably small size to obtain reliable results at high Reynolds numbers.
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
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 )
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
which involves subdivisions of the whole flow field, the B E M proves to be less
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 -
are direct, whereas for parabolic and hyperbolic systems of equations, marching
ii) those in which the general geometry of the problem is such that one or
A B E M reduces the dimensionality of the basic problem by one, i.e., for two
equation. Similarly, for three dimensional problems only two dimensional surface
the Savonius rotor analysis. In this formulation, the integral equations are expressed
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
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.
which transport the vorticity generated by viscous effect w i t h i n the boundary layer
of fluid whose shape is such that the pressure difference across it, and hence the
is discontinuous across the shear layer. These shear layers roll up to form strong
In practise, the effect of viscosity and diffusion lead to finite cores of rotational
fluid.
erated can be determined uniquely from the velocity field which can be considered
£ 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
fl (Figure 4-1) in order to solve for <fr, which is the only unknown. T h i s falls into
The boundary condition at the blade surface specifies that the flow normal to
^ - - " „ = 0, • (4.4)
on
where n is the vector normal to the blade surface and v n is the normal velocity of
the blade.
Chapter IV: Mathematical Analysis 89
a large but finite distance downstream such that this truncation produces negligible
single valued, then there exists at most one function $ satisfying these
to that of the physical problem (Figure 4-1) is constructed, then the potential $ m
selected regions S$> (where <& is known) of the surface S m and S, and
(d<& jdn) m and d<&/dn are identical at all corresponding points of the
Chapter TV: Mathematical Analysis 91
1980).
fies the appropriate conditions may be used, even though the model may bear no
the following section are approximated. Note that it is not necessary to consider
flow field. For example, a point doublet of strength 27r, set w i t h its axis parallel to
a uniform flow of unit magnitude, produces a velocity field identical to the potential
flow about a sphere of unit radius, in the region outside of the sphere.
It follows that use of the geometric surface for the purpose of singularity dis-
problem(Hunt, 1980).
here x and £ are coordinate vectors. A l t h o u g h the origin of coordinates x and £ are
<D(x) = G ( x , £ h ( £ ) , (4-5)
where G(x, £) is the 'two-point' function (generally known as the Green's function)
complete the problem definition any specified singularity distribution T(c) per unit
area in Q is also included. T h e net result of the 7 ( £ ) and T(c) distributions at any
s
*(x)= / G ( x , £ h ( £ ) r f S ( £ ) + / G(x,f)r(f)df7(c) + C .
s s s (4.7)
Js Jn
For existence of the solution it should be insured that the net flux across the infinite
/ i(L)dS(e )+ s [ r ( ) r f n ( c ) = o.
f (4.8)
Js Jn
Similar operations using H ( x , £ ) would yield the velocity at any field point x ,
s
v(x) = / H ( x , £ b ( £ H S ( £ ) + / H(x,f)r(<r)dn(f).
s s s (4.9)
Js Jn
Chapter TV: Mathematical Analysis 93
In principle, the only remaining formal step to arrive at a solution to the problem is
to bring the field point x onto the boundary S. W i t h this, Equation (4.7) becomes
(x )=
B / G ( x , £ b ( £ ) d S ( £ ) + / G{x ,t)T{t)dn[$)
B B B 8 B + C. (4.10)
Js Jn
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 ^ ,
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
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.
solve for 7 then the solution w o u l d be exact; however this is practically impossible
there is a trade off between computing time, effort and solution accuracy. Several
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
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
of the system would require derivatives of the function thus reducing the inherent
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
are located at the mid-points of each segment (singular points). The origin of the
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
Chapter IV: Mathematical Analysis 95
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
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.
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
the strengths of the vortices and to make their positions unknown (Sarpkaya, 1975).
In the discrete vortex method, it is easier to carry out the analysis in the
vortex
where
Wi{z) = (4.13)
to a bound vortex of strength T i at Zg . 5>i and ^ , are the potential and stream
s %
Chapter IV: Mathematical Analysis 98
functions, respectively. Thus the Green's function in E q u a t i o n (4.5) has the form
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
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,
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 \
Chapter IV: Mathematical Analysis 99
Now the only remaining step is to apply E q u a t i o n (4.18) at the (m + 1) nodal points
boundary condition that the velocity normal to the blade at the nodal points is
zero:
(4.20)
(4.21)
which is also known as K e l v i n ' s theorem: the sum of strengths of all bound and free
B y solving Equations (4.20) and (4.21), the strengths of m bound vortices and
A t any instant t. the strengths of m b o u n d vortices and two nascent vortices are
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
Chapter IV: Mathematical Analysis 100
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
T h e position of the I th
vortex shed from the separation point is advanced in a
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
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
In real flow the core of a vortex grows w i t h time. Giesing (1969) and Nagano
the form
a = 2.24v ^i ,
/ 7
(4.25)
where v is the kinematic viscosity and V is the time elapsed since the vortex
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
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,
In real flows only a fraction (=s 60%) of the circulation fed into the shear layers
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
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).
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
as
n P ~ Poo
w\\l \ o2 ^5 $ ( - )
2 4 2 ?
1
' U J U 2
dt
and hence the elemental torque, can be found. Integrating over the blade the total
1974),
Chapter fV: Mathematical Analysis 103
where: C represents an arbitrary path around the rotor excluding singular points;
(4.29)
Torque and power coefficients are calculated as defined by Equations (2.8) and
(2.4), respectively.
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
d
B =
2E"
where d is the diameter of the rotor. E a c h wall was divided into a finite number
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
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.
Rotor
U
2H
—t •
i • • i
2 3 Mw Nascent
Vortex
Nodes
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
A x = b, (5.1)
column matrices, x represents the unknown strength of the distributed and nascent
(5.2)
(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
2k
b
2 = - Y , T
3=1
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
r 2
x= (5.4)
Vr m + 2 J
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
stream, free vortices and blade rotation. Using Equations (4.16) and (4.20), after k
2k _
»• }_wl
bit = T iU! sin ip i — Real e -' \ - Ucos<j> ,
l4>
ni (5.6)
27T ^
n n
{z
ni - Z l) _
w
provement.
Once the unknown strengths of the vortices are found, the nascent vortices are
z
w(2k-\) ~ l iz
Tw2k = + (5.7)
z
w2k = z
m + 2i .
T 8i = Ti +i i = 1,... , m .
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
model had a constant A matrix for both rotating and stationary blade. T h i s implies
Chapter V: Numerical Investigation 109
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.
encountered during application of Equations (4.18) and (4.19) along the blade, due
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
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
(5.8)
where z a and z\, are the coordinates of the end points. O n evaluating the integral
Chapter V: Numerical Investigation 110
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 /
dF(z) i^e
5-1). Now
\z-z J b n
v=
n — In — ,
27T r\
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
Chapter V: Numerical Investigation 112
line of the element and becomes infinite at the centre. B o t h distributions are fairly
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
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
\w{z )\ ni + |u>(2NI+1)|
w(z )
si exp(-i(j) i)
s ± —~- exp(-i<p ), sl (5.11)
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
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 = ± ^exp(-?.0 S ! ).
T h i s yields the velocity at the centre, different from that in the tangential direction,
A n o t h e r method was to evaluate velocities at the nodes, and add averaged "free
Chapter V: Numerical Investigation 114
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
a distance t away from the blade . T h i s was tried at the nodes as well as at the
ZA,Bni ~ Zm ± t exp(i>„i);
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-
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
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
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
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)
dependency of each term and simplifying it using the chain rule the following ex-
Chapter V: Numerical Investigation 117
dF(z )si i dT f
2k
i v-- r,
+ {^si z
wl)
+ 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 .
2.71
A,- = — In r, + i $i
2^
6, i
In r , .
27T ' 27T
ing 0, in the range 0 < 0, < 27r, and with a series of z,-'s such that the corresponding
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
Chapter V: Numerical Investigation 118
it was evaluated w i t h a second order accuracy by using the Gaussian Three Point
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-
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
The torque coefficient can be found either by integrating the elemental torques
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
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
c
- = WvW • < 5
' 1 6 )
^ * = —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
5.3 C a l c u l a t i o n P r o c e d u r e
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
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
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 second step to calculate the loading. Thus the earliest loading information is
There are several important parameters which must be established before the
ts distance between the nascent vortex and the edge of the blade;
At time-step size;
racy, however, it would also escalate the computational cost. Thus a compromise
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)
Considering the fact that e should be greater than o, Ogawa (1984) selected e
s s
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
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.
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
0.04363372
At = (5.19)
XU
downstream. To avoid this the following model was used at higher tip-speed ratios,
(5.20)
proved calculation efficiency, several test runs were carried out and the effect of
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
ity was incorporated in the program so that the computations can be stopped and
restarted as desired.
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.
elements being 20, 25, and 30 w i t h all the other parameters held constant:
A Tip-speed ratio = 0;
K Resolution parameter = 2;
U W i n d speed = — 7 m/s;
V Balancing parameter = 0;
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
It is apparent that the results for m = 25 and 30 are essentially the same w i t h
1.2-1
Time (Sec)
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!
Chapter V: Numerical Investigation 127
N u m b e r of C P U Time
Elements (m) (s)
20 173
25 217
30 260
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
reflected in the corresponding flow fields (Figure 5-8) where the flow pattern for
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/?.
Chapter V: Numerical Investigation 128
Time (Sec)
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)
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
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
predicts the torque coefficient closest to the experimental value shown in Figure
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.
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
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
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-
T i m e (Sec)
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.
Chapter V: Numerical Investigation 135
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)
evident in the predicted torque coefficient shown in Figure 5-13. A larger value
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
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
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
Time-Step C P U Time
Size (s) (s)
0.002 1275
0.004 217
0.008 75
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-
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
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.
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
assessed.
138
Chapter VI: Results and Discussion 139
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.
was fixed at 2. q (Figure 2-1) was taken as 0.191 m and p/q fixed at 0.2 to facilitate
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
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
In the second approach, the shear layers separate at the blade tips as implicit in
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.
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
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°.
Chapter VI: Results and Discussion 142
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.
Chapter VI: Results and Discussion 143
(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 ° .
Chapter VI: Results and Discussion 144
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.
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
shows a periodic character. O n the other hand, the case of the surface separation
reach a steady state at around 0.35 seconds, with several peaks, corresponding to
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-
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
Chapter VI: Results and Discussion 145
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 :
-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.
Chapter VI: Results and Discussion 146
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
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 ° .
Chapter VI: Results and Discussion 148
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.
Chapter VI: Results and Discussion 149
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)
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 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
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
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
Chapter VI: Results and Discussion 151
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 ° .
Chapter VI: Results and Discussion 152
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.
Chapter VI: Results and Discussion 153
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.
Chapter VI: Results and Discussion 154
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)
vortex clusters shedded from the top edge were observed to be approximately 3
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
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
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
having established periodic behaviour of the flow, the computation was limited to
Chapter VI: Results and Discussion 156
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
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
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
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
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.)
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)
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
1.5
4*
0.5
-0.5
k = 200
-1.5-
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
complete, and hence the calculated frequency must be corrected to account for this
effect.
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
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
Figure 6-13 shows the corresponding flow patterns. Basic features governing
of vortices as discussed before are all quite apparent here. For example, for k «
Chapter VI: Results and Discussion 162
p= 150°
DVM
I ' 1
0.2 0.4
F I G U R E 6-11 = 150°.
Chapter VI: Results and Discussion 163
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.
Chapter VI: Results and Discussion 165
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.
Chapter VI: Results and Discussion 166
2.0-1
X (meters)
2.5 n
X (meters)
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
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
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
rotors using several different methods, depending upon the available data, and the
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 % .
Chapter VI: Results and Discussion 170
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
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
6.2 R o t a t i n g B l a d e
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
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
for it.
(A' = 2), as in the stationary blade case, continues to yield satisfactory results, trial
p/q = 0.2 ;
X = 0.4 ;
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
K C P U (s)
1 350
2 95
In this analysis, the tip-speed ratio X was taken to be 0.4. As the angle of
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
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.)
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.
Chapter VI: Results and Discussion 175
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)
typical sets of results are presented here to help establish trends. A sample of flow
here:
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
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
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,
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.
Chapter VI: Results and Discussion 178
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.
Chapter VI: Results and Discussion 179
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.) 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).
Chapter VI: Results and Discussion 181
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).
Chapter VI: Results and Discussion 182
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-
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-
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.
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
earlier, now the central vortex is shed and a new one formed from the vorticity
Chapter VI: Results and Discussion 184
1.5
x T,+ve
k = 20 k = 40
1.0-
o T +ve
2
+ T -ve
0.5-
2
0.0-
-0.5- +
-1.5-
1.0
0.5-
Y 0.0
-0.5
-1.5- T 1 I P
i i r
1.0-
0.5
o.oH
-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)
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)
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)
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)
shed from the upper blade. Jones et al. (1979) observed similar phenomenon of the
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
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
time curves, at different tip-speed ratios, is given in Figure 6-24. T h i s shows a max-
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 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.
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
Time (Sec.)
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.
Chapter VI: Results and Discussion 192
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
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
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.
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
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
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
. | . | i | i | i | i i I i I i I i I i I i
I 1
I 1
I 1
I 1
I 1
I ' i • I 1
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)
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 + e^o
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)
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.)
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.)
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 .
Chapter VI: Results and Discussion 200
in blockage.
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
amplitude of the torque coefficient as well as the velocity with which the
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°.
Chapter VI: Results and Discussion 201
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.
Chapter VI: Results and Discussion 202
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.
Chapter VI: Results and Discussion 203
Chapter VI: Results and Discussion 204
Effectiveness of the analytical model for predicting wall confinement effects for
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
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
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
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
Chapter VI: Results and Discussion 205
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
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.
Chapter VI: Results and Discussion 207
0.55
Theoretical
0.50- A p/q = 0.2
x p/q - 1.0
— A
Exper. p/q = 0.2
0.45-
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
absent at A = 1.2 as in the unconfined case. A s discussed for the stationary case,
during rotation. Thus the vortex convective speed increases w i t h blockage even
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).
Chapter VI: Results and Discussion 210
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
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:
mirably play this role. It can account for the effect of blade geometry
211
Chapter VII: Concluding Remarks 212
ulate the transient, separated viscous flow associated w i t h the rotor, yet
Based on the w i n d tunnel test program, flow visualization study and mathe-
geometry as follows:
gap-size (a/d) 0 ;
It is inferred that this has a peak power coefficient of around 0.32 at a tip-
Chapter VIJ: Concluding Remarks 213
character.
(p/q) shifts the C p > m a x vs. blockage plot along the ordinate without
e) Presence of a vortex filament at the centre of the rotor and its shedding
w i t h a reasonable accuracy.
g) T h e wake of the stationary blade shows periodic behaviour over the range
Chapter VII: Concluding Remarks 214
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
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
There are several avenues for future studies which are likely to be profitable
and satisfying :
relation w i t h the field data has not been reported in the literature.
Chapter VII: Concluding Remarks 215
exciting possibility.
accuracy.
BIBLIOGRAPHY
Imai, I., 1974, Fluid Dynamics, V o l . 1, Shokabo, Tokyo, pp. 408-415 (in Japanese).
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..
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.
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.
Finland.
222
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.
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