3-D Simulation for small vertical axis turbine
Numerical Methodology
Geometry Generation
Using CAD software both the channel and the turbine were designed
(Ansys Design Modeler & Solidworks)
Grid Discretization Strategy
A structured mesh is generated for both the rotor and stator domains including non-
conformal interfaces with Skewness lower than 0.5 (Sk = 0.46 )
Mesh Sensitivity Study
Three different mesh resolutions: coarse, medium, and fin were simulated in the worst scenario
at (λ = 1.5) where high instantaneous angles of attack leads to boundary layer separation.
Turbulence Model
U-RANS Solution
Two-equation turbulence models were applied to solve the Reynolds-averaged Navier Stokes
equations: K-ϵ including (Standard, RNG and Realizable) and k-ω SST model.
Time Dependency Study
Three temporal resolutions were carried out to determine how the time step affects the
accuracy to obtain the minimum sufficient value.
Solution and Results
Computational flow field and Boundary conditions
The computational domain as shown in Fig. 1 was modelled with the real dimensions of the
experimental, having the cross-section of 0.3 m width × 0.5 m water height as a maximum
height (0.35 m in the current study) and 1.5 m in length. The rotor having the diameter of 0.15
m and height of 0.15 m was simulated in the channel. Three NACA-0015 aerofoils are used for
Darrieus rotor coupled with two supporting caps (upper and lower).
The sliding mesh technique is used to simulate the rotational movement of the turbine rotor. In
order to solve the Unsteady Reynolds Averaged Navier Stokes (URANS) equations, k-ω SST
turbulence model is utilized (after intensive comparison between other turbulence models as
discussed later) to simulate the flow over the rotating flow field across the turbine and to model
the circulated flow vortices and eddies. Furthermore, the Coupled pressure-based solver is
selected according to a comparison between the Coupled and Simple pressure-velocity
coupling. As shown in Fig. 2, the Coupled algorithm can reach the periodical solution by
reaching the third cycle and the power coefficient remain constant after more rotational cycles.
Walls (No slip)
Interface (Rotating Zone)
Outflow
Free surface (Specified shear)
Inflow
(a)
0.75 m Interface Region Channel walls
Y
X
0.3 m
Velocity Inlet Pressure Outlet
1.5 m
(b)
Fig 1. Different views and geometrical parameters of the computational domain: (a) isometric view, (b)
top view.
Conversely, the Simple algorithm need more revolutions to reach the converged solution which
affect the computational cost negatively. The Coupled pressure-based solver with second order
implicit transient formulation were used for better results. The instantaneous torque coefficient
vs azimuthal angle for a complete rotation of the turbine blade for both Simple and Coupled
algorithms is shown in Fig. 3. All the solution variables were calculated with second order
Upwind discretization scheme. The 3D computational domain conditions provided to the
Fluent solver are listed in Table. 1.
0.5
Simple
0.41
Coupled
0.4 0.37
0.32 0.31
0.3 0.3 0.29
0.29 0.29
0.3
Cp
0.2
0.1
0
1st 2nd 3rd 4th 5th
Cycle
Fig 2. A comparison between Simple and Coupled pressure–velocity coupling effect on reaching the
periodic solution at v = 0.62 m/s and λ = 1.5.
0.6
0.5 Coupled
Simple
Torque Cofficient, Cm
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
0 45 90 135 180 225 270 315 360
Azimuth Angle (deg)
Fig 3. A comparison between Simple and Coupled pressure–velocity coupling for a complete rotation
of the turbine blade at v = 0.62 m/s and λ = 1.5.
Table 1. Boundary conditions for the present numerical simulation.
Location Boundary Condition Boundary type
Inlet Velocity inlet Flow Velocity (0.62 m/s)
Outlet Pressure outlet Constant pressure
Channel walls (sides and bottom) Wall No-slip wall
Free surface Wall Specified shear
Contact region Interface Interface
Turbine including shaft Wall No-Slip wall
Grid sensitivity study
Both of the quality and density have an effective influence on the CFD results. Thence, to
ensure grid independent results and to avoid high computational cost, simulations were carried
out using three different aggressive structured mesh resolutions: coarse (L), medium (XL), and
fin (XXL). Statistics and power coefficients corresponding to each mesh are listed in Table 2.
Solutions were ascertained grid independent when negligible difference was achieved in the
average power coefficients of at least two consecutive meshes. Comparisons among the
forgoing three meshes were made for the NACA 0015 fixed-pitch mechanism at a water flow
velocity of 0.62 m/s (free stream Reynolds number calculated based on the chord length of
blade is estimated to be about 2.5×104) and TSR of 1.5. This case was selected to put the tested
meshes under a critical working conditions as lower tip speed ratios involve massive flow
separation due to higher range values of angle of attack and thus require more iterations per
time step to converge.
To keep a constant value of y+ on all blades, the number of inflation layers (50 layers) and the
height of the first grid node (y) were kept constant in both of (XL) and (XXL) high density
meshes at y = 1.2×10-5 m, resulting in values of y+ of 0.5 on all three blades. In fact, the
spanwise discretization was the key to decrease the mesh size without affecting results as the
measured properties don’t change noticeably in this direction so the central part of the blade is
uniform with acceptable number of nodes but near the two plates is refined to catch the
boundary layers near blade tips and caps. The instantaneous torque coefficient vs azimuthal
angle for a complete rotation post the transient starting instabilities for the different three
meshes are shown in Fig. 4&5. The medium (XL) mesh with skewness lower than 0.5
(SK=0.46) as shown in Fig. 6&7 was selected for the current simulation that achieving the high
accuracy and moderate computational time as it has total number of cells near 1 million.
Table 2. Grid sensitivity analysis for NACA-0015 at λ = 1.5 and V = 0.68 m/s
Grid Name Number of control volumes Total Number Power coefficient
of cells (Cp)
Rotor Channel
L Coarse 294320 213015 507335 0.235
XL Medium 741320 213015 954335 0.259
XXL Fine 1596780 284384 1881164 0.261
Mesh Sensitivity
0.60
0.55 XXL
0.50 XL
0.45 L
Torque Cofficient, C m
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0 30 60 90 120 150 180 210 240 270 300 330 360
Azimuth Angle (deg)
Fig 4. Mesh sensitivity for three different grids for one complete cycle of the turbine at V = 0.62
m/s and λ = 1.5.
0.6
0.5
XXL
XL
Torque Cofficient, Cm
0.4
L
0.3
0.2
0.1
0.0
-0.1
-0.2
0 45 90 135 180 225 270 315 360
Azimuth Angle (deg)
Fig 5. Mesh sensitivity for three different grids for one complete cycle of the blade at V = 0.62 m/s
and λ = 1.5.
Turbulence model
In the current study, RANS two-equation turbulence models were tested to select the most
proper one, K-ε Turbulence model including (RNG, Realizable and standard) and k-ω SST
(Shear Stress Transport). Basically, k-ω SST model is a combination of: k-ω model that is used
near walls and k-ε that solve the far field. While k-ε uses a wall function to resolve the boundary
layer, the k-ω SST solve RANS equations up to the wall giving superior results for flow with
strong adverse pressure gradient such that in our turbine. A comparison between the different
Interface
Rotating zone
Fig. 6. Mesh visualization for the interface region and rotating zone (Skewness 0.46)
Fig. 7. Mesh visualization near wall treatment around the blade
turbulence models are shown in Fig. 8&9. Although, both RNG and Realisable k-ε models give
higher torque generation than the standard k-ε model, the k-ω SST give the highest average
power coefficient value as listed in Table 3. Furthermore, the RNG and Realisable k-ε models
have a limitation that they could sometimes produce non-physical turbulent viscosities when
the computational domain contains both rotating and stationary fluid zones as in our case.
0.6
k-ω SST
0.5 RNG
K-ε Realizable
0.4
k-e Standard
Torque Cofficient, Cm
0.3
0.2
0.1
0.0
-0.1
0 45 90 135 180 225 270 315 360
Azimuth Angle (deg)
Fig. 8. Comparison of different turbulence models for one complete cycle of the turbine at V = 0.62
m/s and λ =2
0.6
k-ω SST
0.5
Realizable
Torque Cofficient, C m
0.4 RNG
0.3
k-e Standard
0.2
0.1
0.0
-0.1
0 45 90 135 180 225 270 315 360
Azimuth Angle (deg)
Fig. 9. Comparison of different turbulence models for one complete cycle of the blade at V = 0.62
m/s and λ =2
Table 3. The average power coefficient values corresponding to each turbulence model for one
complete cycle of the turbine at v = 0.62 m/s and λ = 2
Turbulence model k-ω SST Realizable RNG Standard
Average Cp 0.3119 0.3275 0.3237 0.2454
Convergence Criteria
A convergence criterion of 10−5 was set for all parameters at each time step. At the beginning
of the simulation it was observed that the periodic flow field has instabilities in the solution
that is called starting instability as shown in Fig.10. In order to achieve the periodical converged
solution, the rotor was allowed to rotate 5 revolutions with 1800 time steps. A minimum
number of rotations is necessary to ensure that a repeatable power extraction cycle is achieved
but this number is case-dependant and must be estimated for each simulation by comparing the
CP for the last cycle to the previous ones. Generally, it was observed that the converged solution
was obtained after 3 revolutions, but in our case the turbine was allowed to rotate 4 revolutions
with 1440 time steps to ensure that it reaches the periodical converged solution.
1.00
0.80
Torque Cofficient, Cm
Starting Instabilitiy
0.60
Periodical Solution
0.40
0.20
0.00
-0.20
0 360 720 1080 1440 1800
Time step
Fig. 10. Periodical torque results for one blade over 5 cycles at v = 0.62 m/s and λ = 2.
0.60
Blade 1 Blade 2 Blade 3 Total
0.50
Torque Cofficient, Cm
0.40
0.30
0.20
0.10
0.00
-0.10
-0.20
0 45 90 135 180 225 270 315 360
Azimuth Angle (deg)
Fig 11. Torque coefficient for three individual rotor blades and the total torque developed by the
turbine after 3 revolutions at λ = 1.75.
0.5
0.45 3-Blades
0.4
Power Cofficient, Cp
Turbine
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9
Tip speed ratio, λ
Fig. 11. Average power coefficient for both 3- blades and the turbine including (two caps and shaft)
at V = 0.62 m/s.
