A GOOD METHODOLOGY (FYP2)

EXPERIMENT & SIMULATION BASED RESEARCH

The wind tunnel tests were conducted at the Aerodynamics Laboratory at Sheffield University.
The University low speed wind tunnel used for this study has a square test section of
dimensions
1.2 m_1.2 m. The tunnel consists of a large bell mouth screened inlet with a contraction
ratio of (2.5:1) before the flow encounters another screen and an array of honeycomb flow
straighteners. The tunnel has a working section of length 3.0 m. Any obstruction placed
within a wind tunnel will alter the characteristics of the flow to some degree. If this
obstruction is too large then the area available to the flow is significantly reduced, and so
the speed of the flow around the model will increase. If the blockage ratio is high enough
then the effects of the tunnel walls may begin to interfere with the flow over the model. The
literature suggests that blockage ratios below about 6–7.5% have a negligible effect on the
flow. However, the difficultly in this case comes in
defining the frontal area of the turbine as it spins. The area of the wind turbine when it is
stationary is small, but the area it sweeps out as it spins is significant. For the most
conservative value of blockage ratio, the frontal swept area of the turbine could be used
and using the wind tunnel dimensions, this would result a 20 cm_43 cm turbine (i.e. area
0.086 m2). Clearly this would be a tiny model, and the difficulties in testing it would be
numerous including the problem that the Reynolds number would be too much smaller
than that of a realistic VAWT. It was therefore more practical and reasonable to assume
the apparent frontal area of the turbine would lie somewhere between the small area of the
components and the large area of the swept frontal area. It was decided that blockage
ratios based on frontal swept area could be tolerated beyond 7.5%, so a notable amount of
the frontal swept area would not actually be occupied by the turbine. Preliminary
experiments with the turbine in the wind tunnel showed that there was no correlation
between the rotational speed of the wind turbine and the approach velocity within the
tunnel (measured five rotor blade chords upstream). Furthermore, during the testing of the
Turby VAWT [7], a turbine with blockage ratio of 14% based on frontal swept area was
used with a view to it having a ‘‘negligible blockage effect’’.

2.1. Aerofoil selection

Traditionally NACA 4-digit series have been employed for Darrieus- type VAWTs. For
example, a NACA0015 profile has been used in the Sandia investigations. In the late 1970s,
Healy investigated the effect of thickness on VAWT performance [8]. More recently, a
systematic numerical study of various aerofoils including NACA 4- digit series based on the
unsteady RANS solutions of the VAWT flows by the authors indicated that thicker sections
performed
much better under the flowconditions of interest. As a result of our simulations, the
NACA0022 profile was chosen as the primary profile candidate for this research
programme. In order to create a geometrically accurate profile the rotor blades were
manufactured using CNC milling machine from high-density foam. In order to give the
blades sufficient strength to withstand the centrifugal bending forces cause by the high
rotational speeds of the turbine, the foam had to be of a minimum thickness. The
NACA0022 profile was therefore created with a thickness of 22 mm and a resulting chord
of 100 mm. This, combined with a height of 400 mm (limited by the CNC machine), gave
the blades an aspect ratio of 4 and gave the turbine a solidity of 1.0 for the three bladed
turbine and a value of 0.67 for the two bladed turbine. These values are rather high
given other research clearly shows a lower solidity will result in a higher performance
coefficient.

However, the aim of this research turbine was to provide validation data for computational
methods and an understanding of the aerodynamic inefficiency (such as those generated by
the tip flows and resulting vorticies) so maximum performance in itself was unimportant.
As will be shown later, the low aspect ratio did indeed give rise to large tip vorticies.
Testing was carried out using 300 mm rotor arms resulting in a blockage ratio (based on
frontal swept area) 16.7%.
During initial testing at some conditions, the turbine would suddenly reduce its rotation
speed and eventually stop despite no changes to the applied torque. It transpired that the
turbine blades
were slowly rotating about an axis centred through the bolts that fix them to the support
struts. This was caused by the centre of rotor lift not being aligned properly with the fixing
bolts. It is clear that the centre of lift for VAWT rotor blades is constantly changing
throughout every rotation so it was not a simple process to determine the best location of
the fixing holes on the rotors. The maximum change in the rotor blade fixing angle was
measured to be less than 5_ and this small angle change caused a complete loss of lift and
hence the eventual complete loss of power from the turbine.

A CAD model was used to carry out the design before manufacture, see Fig. 1. Two deep-
grove ball race bearings were used to support the turbine and allow free rotation of the
rotor shaft, where the lower bearing supporting the weight of the turbine by acting as
thrust bearings. Two support arm/strut brackets were designed to fit on the main drive
shaft so that they could be
mounted anywhere along the height of the shaft accommodating 2, 3 or 4 blades to
accommodate future blades of different lengths (aspect ratios). Rotor blade radial
arms/spokes were machined
from aluminium bars and aerodynamically profiled with an elliptical leading edge (major
to minor axis ratio of 2:1) and sharp trailing edge. All components were attached such that
they could be assembled or adjusted with ease. The finished assembly is pictured in Fig. 1.
To measure the power output from the turbine a simple torque brake was employed. The
torque applied to the turbine rotor drive shaft was increased and decreased by changing
the separation
distance between two spring balances. The torque applied was calculated using the
difference in the forces applied by the spring balances and the drive shaft radius. To
calculate the power output, this torque was combined with the rotational speed of the
turbine, itself picked up from a once per revolution optical tachometer. A feedback control
system was not used for the control of the turbine so it was not possible to measure the
characteristics of the turbine past its maximum torque (and therefore minimum stable
rotational speed). This is because beyond this limit any small increase in applied load
(torque) causes a drop in the rotation speed of the turbine and a drop in the lift (and
driving torque) it generates. If there is no control system to reduce the applied torque to
match the new aerodynamics condition, the turbine rotor will continue to
drop in speed so the applied torque becomes more unmatched to the conditions of the
slower rotating turbine.

SIMULATION BASED RESEARCH

CFD domain characteristics

The numerical investigation is used for the aerodynamic behavior of a three-bladed Darreius vertical axis
wind turbine characterized. This turbine consists of S1046 airfoil (as started airfoil) with solidity ( σ = Nc/2R)
equals 0.1. Wind speed at computational domain entrance has been kept constant and equals to 5  m/s. The
angular velocity of the turbine (rotating zone) is varied to evaluate the turbine performance. Table 1shows
the geometrical and kinematic features of the tested rotor, while in Table 2, the main flow field characteristics
are illustrated.
Table 1.
Main geometrical features of the analyzed rotor.
Denomination Value
Rotor diameter (D) [mm] 2000
Rotating zone diameter 1.25D
Rotor height (H) [m] 1
Blade section S104
6
Blade chord (c) [mm] 70
Number of blades (n) [-] 3
Turbine solidity (σ) [-] 0.1

Table options
Table 2.
Main Characteristics of the simulated flow field.
Denomination Value
Pressure [Pa] 101325
Density (ρ) [kg/m3] 1.225
Inlet velocity (V∞) [m/s] 5
 Table options

3.1. Description of the numerical flow field

Darrieus turbine operates in open field conditions, so it is necessary to consider a large domain, to avoid solid
blockage. Fig. 1 displays the main geometrical features of the computational domain. The appropriate size of
the computational domain has been investigated to eliminate the effect of the domain size on the results. A
computational domain is a square domain of size 2L × 2L, it has been considered in the CFD computation.
Mohamed et al. [12] indicated that the smaller domains are associated with a large variation of the rotor
coefficient. This demonstrates that the computational domain should extend at least over 20 times the rotor
radius in each direction. In a smaller domain, the boundary conditions influence the results in an
inappropriate manner. Therefore, the domain (25 D) as shown in Fig. 1 has been retained for all further
computations in this paper.

Fig. 1. 
Geometrical features and main dimensions of the computational domain.
Figure options
Inlet has been set as a velocity inlet, with a constant wind velocity profile of 5 m/s, while, outlet has been set as
a pressure outlet with atmospheric pressure value. Two symmetry boundary conditions have been used for
the two side boundaries. To ensure the continuity of the flow field, the circumference around the turbine
rotational axis was set as an interface. The inner zone of the flow field is named as turbine zone and also is set
as interface to ensure mesh linkage between domain and turbine. Turbine zone is characterized by a moving
mesh and it is revolving at the same angular velocity of the rotor. Fig. 2 shows the main dimensions and the
boundary conditions of the rotating zone (turbine). All the blade profiles inside the rotating zone rotate in
circle of 2000 mm diameter while the rotating zone is 1.25 times the diameter.
 Fig. 2. 
Scheme of rotating zone area.
Figure options

3.2. Domain meshing

A paved (Quad Dominant) mesh has been chosen for the whole computational domain in both ANSYS (Fig. 3)
and GAMBIT (Fig. 4) meshing tools. This type of mesh is selected in order to reduce the computational time
and increasing the accuracy in spite of increasing the preparation time of the CFD simulations. The mesh on
both sides of the interface (domain and turbine) has approximately the same characteristics cell size to obtain
faster convergence.

Fig. 3. 
 ANSYS mesh for a three-bladed VAWT.
Figure options

Fig. 4. 
GAMBIT mesh for a three-bladed VAWT.
Figure options
Being the area close to the blade profiles, great attention where the computational grids have been
constructed from lower topologies to higher ones and adopting appropriate size. Clustering grid points near
the leading edge and the trailing edge of the blade profile is to capture the physics and improves the CFD
code capability. Fig. 5 and Fig. 6 represent the mesh near wall using ANSYS and GAMBIT, respectively.

Fig. 5. 
Grid at airfoil section using ANSYS.
Figure options
 Fig. 6. 
Grid at airfoil section using GAMBIT.
Figure options

4. Numerical solution

A finite volume CFD solver ANSYS FLUENT is used in this work, that implements Reynolds averaged Navier–
Stokes equations. The fluid has been assumed to be incompressible since the inlet velocity equals 5 m/s. The
hypothesis of incompressible flow is consistent taking into account the relative speed to the blade. It does not exceed
45 m/s and it is acceptable (Ma < 0.15). By studying the effect of time step size on torque coefficient using 0.001,
0.002, 0.004 and 0.005 s time step size, it is observed that is the change in torque coefficient doesn't exceed 5%. In
addition, same study performed for the maximum number of iteration as the the calculation performed using 50, 70,
80 and 90 maximum number of iteration; the change in torque coefficient doesn't exceed 1%.

The time step size is set to be 0.005 by performing the computations on 7 revolutions with 70 maximum number of
iteration. As a global convergence criterion, each simulation has been run until instantaneous torque coefficient is
less than 1% compared with the relative values of the previous period. Residuals convergence criterion for each
physical time step has been set to 10−5. Furthermore, Courant number can be calculated for 0.005 time step size to
ensure that is within the acceptable range. As the minimum cell size is 7.07 × 10−3 ms, Courant number will be about
3.5 which is acceptable for such cases. The selected solver is pressure based, well suited to compute an
incompressible flow field. Two turbulence models have been checked in this step of work, Realizable k–ε and the
SST k–ω. These two models are recommended by the CFD worker for the rotating zones. The power coefficient
shown in the figures is an average over an entire revolution.

The Realizable k–ε model has been adopted for viscous computations. This model differs from the Standard k–ε in
two points; 1) it contains a new formulation for the turbulent viscosity and 2) a new transport equation for the
dissipation rate ε has been derived from an exact equation for the transport of the mean-square vorticity fluctuations.
One of the benefits of the Realizable k–ε model is that it provides superior performances for flows involving
rotation, boundary layers under strong adverse pressure gradients, separation and recirculation.

SST k–ω turbulence model is a two-equation eddy-viscosity model which has become very popular. Usage of a k–
ω formulation in the inner parts of the boundary layer makes the model directly usable all the way down to the wall
through the viscous sub-layer. Hence, the SST k–ω model can be used as a low-Re turbulence model without any
extra damping functions. SST formulation also switches to a k–ε behavior in the free-stream to avoid the common k–
ω problem. SST model is too sensitive to the inlet free-stream turbulence properties. So, firstly the computations are
taken place using Realizable k–ε for both models meshed by ANSYS and GAMBIT and compare it with SST k–
ω results. The computations use SST k–ω for ANSYS model by keeping the a fine near-wall mesh capable of
resolving the viscous sub-layer (y+ < 5) and compare it with the GAMBIT model performed by Realizable k–ε.