You are on page 1of 13

Delft University of Technology

Rotor / Wake Aerodynamics

Assignment 2: Lifting Line model

Patrick Duffy (4756444)


Bose Sumantraa (4774302)
Simone Tamaro (4714229)

May, 2018
Contents

Introduction 2

Assumptions 2

Limitations 3

Code flow chart 5

Lifting line theory results 6


Comparison with Blade Element Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Induction factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Inflow angle and angle of attack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Normalized forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Power Coefficient, Thrust coefficient and cosine distribution . . . . . . . . . . . . . . . . . . . . . . 9

Sensitivity analysis 10
Time length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Azhimutal discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Conclusions 11

1
Introduction
This report presents the performance results of a wind turbine rotor obtained using an implementation of
the Lifting Line Theory (LLT). The wind turbine blade analyzed is a version of the blade given in Assign-
ment 1 which was modified to improve its performance based on an implementation of the Blade Element
Momentem (BEM) Theory. This optimized blade is considered for this assignment.

The essence of the Lifting Line Theory for analyzing wings and rotors relies on applying linear superposition
of potential flow solutions to Laplace’s equations for incompressible flow. This is done by analyzing the
velocity components uω induced at specific points from different vortex filaments, or sections of vorticity
concentrated along lines of constant circulation, with the Biot-Savart law, given below. An image depicting
this is presented in Figure 1.
b
Z
Γ dl × r
uω =
4π |r3 |
a
The Helmholtz and Kelvin-Stokes theorems ensure the conservation of circulation in time and space. With
the ability to solve for the velocities at control points representing simplified airfoil elements, it is possible
to obtain lift and drag forces at each element. An analysis of the full wing/rotor performance is possible
when the force contributions from the local velocities and angles of attack at individual blade elements are
computed and numerically integrated. In the case of a wind turbine the wake vortices and bound vortices
impact the induced velocities. MATLAB was used to implement the LLT using a blade element approach with
a frozen wake vortex model. In the following sections, the key assumptions behind the implementation of
a solution are discussed and a flowchart of the computational solution process is presented along with the
results of the rotor performance analysis.

Figure 1: Illustration of application of Biot-Savart along a constant circulation line

Assumptions
In the process of implementing the LLT to assess wind turbine rotor performance with the blade element
method involved the following assumptions:

• Steady, uniform, incompressible, 2-Dimensional flow normal to the rotor plane. The 2-D
assumption enables the rotor to be divided airfoil into independent blade elements and aerodynamic
properties of each airfoil section to be computed based on the local angle of attack and measured
airfoil polars. Any blade section is considered to work under two-dimensional flow conditions when
the complete influence of the induced, rotational and axial velocities of the flow field is considered. This
assumption means that the induced radial velocity is neglected. The steady and normal assumptions
mean one does not have do consider any wind shear or yawed flows, this simplifies the computations
because of symmetry. Incompressiblilty means that no shocks or shock waves.
2
• Frozen wake. The fundamental shape of the wake is considered to be constant as it is convected
downstream. This frozen wake is made up of the shed vortices of the spinning rotor blades. The
implications of the frozen wake assumption involve analyzing the velocities induced at the rotor control
points and does not account for changes to the shape, magnitude, and direction of the wake in time.
No wake self-induction is considered and turbulent mixing is also ignored. The frozen wake assumption
means that the shed sections of the vortex rings may be considered to cancel out, which results in the
’horseshoe’ vortex shape. The horseshoe vortex consists of three segments: a bound vortex spanning
along the wing (centered at c/4 distance from the leading edge), connected to two trailing vortices at
the endpoints of the bound vortex filament.
An image depicting the wake horseshoe vortices is presented in Figure 2.

Figure 2: Visualization of the vortex ring structure with a horseshoe vortex highlighted in red. Control points are
indicated in green along the quarter chord line.

• Simplified airfoil geometry. It is assumed that flow field around an airfoil is represented with
a lifting line passing through the quarter chord point of each cross-section and all the flow field in
chord-wise direction is concentrated at that point. For the sake of reducing computational time, the
physical description of the flow field is simplified to a lifting line by assuming that the blade may be
represented as a line positioned at the quarter chord position. The induced velocities are computed
at control points along this quarter chord line and represent different blade elements at which local
forces are computed based on local airfoil properties.

• Starting vortex considered to be infinitely far away. This assumes that enough time has
elapsed since the start of the flow such that the induced velocity contribution from the starting vortex
is therefore ignored in the calculations. This is due to the fact that large distances between a vortex
filament and a control point when using the Biot-Savart law make for negligible contributions to
induced velocities from that filament.

• The tip vortices are counter-rotating. This means certain effects cancel out due to symmetry.
• The circulation is assumed to be constant along an entire horseshoe. This means that the
Biot-Savart law can be used for vortex filaments.
• The wake is assumed to be convected at a constant speed.

Limitations
As given by Helmholtzs theorem, the circulation (gamma) is constant along the entire vortex line. But, this
is not well suited for accurate predictions of overall lift induced drag. The main deficiency is that its local
lift/span is constant across span, which is not very realistic for a real wing/ blade, as lift falls gradually to
zero at blade tip.
3
The geometric simplification decreases the accuracy of the flow description in the space close to the airfoil.
Three dimensional flow effects are ignored. This may be important at the tips where some spill over may
occur. This is the region where vortices play the most signifant role in shaping the flow.

4
Code flow chart

Define Blade
Geometry: Radius,
Blades, Pitch, TSR,
Airfoil data

Compute Induced Velocities 


due to trailing vortices
& Sum the total contribution in 
X,Y & Z direction indepedently

Discretize Blade &


define control points
along the span

Compute Induced Velocities due to Bound


Vortex @ control points along the blade
(c/4 distance)

Define Horseshoe
Wake Geometry for 3 blades:
 X, Y & Z component of wake 
find perceived velocity of blade=
Freestream vel + Induced velocities

Coordinates of Wake
Blade1:                                                Blade2: 
X= t * Uwake                              X= t * Uwake
Y= R,i * Sin(wt)           Y= R,i * Sin(wt+2pi()/3) Find,
Z= R,i * Cos(wt)        Z= R,i * Cos(wt)+2pi()/3)
* Resultant Velocity
                         Blade3: * Inflow angle
                         X= t* Uwake * Angle of Attack
                         Y= R,i * Sin(wt-2pi()/3) * Cl & Cd at found AoA
                         Z= R,i*  Cos(wt-2pi()/3) * Gamma

Start Iteration:
Initialize Circulation Strength
(gamma=1)

If (Gamma,new- Gamma) Yes


Evaluate loads, Cp
< Tolerance

No

Goto Iteration step again

5
Lifting line theory results
Using the LLT implemented in MATLAB, the induced velocities at the control points of the blade are computed.
A very important part of the LLT coding process lies in setting up the induced velocity matrices. This was
done by taking into account the influence of all vortex filaments from all blades on each control point on
the blade. Examples of these induced velocity matricies depict the induced velocity contribution from each
horseshoe on each control point. Note that the strongest contributions are along the diagonals which makes
sense because contribution from the Biot-Savart law is inversely proportional to the distance. The results
are presented in Figure 3.

Induced velocity u Induced velocity v


20 0.05 20

18 18
0.08
16 16
0
0.06
14 14
Spanwise element

Spanwise element
12 12
0.04

10 -0.05 10
0.02
8 8

6 6
-0.1 0

4 4
-0.02
2 2
-0.15
0 0
0 5 10 15 20 0 5 10 15 20
Horseshoe Horseshoe

(a) u (b) v
Induced velocity w
20 6

18
4
16

14 2
Spanwise element

12
0
10

8 -2

6
-4
4

2 -6

0
0 5 10 15 20 10-3
Horseshoe

(c) w

Figure 3: Matrices of induced velocities computed for λ = 7 ∆t = 0.1s, total time of 200 s.

Comparison with Blade Element Momentum


The results obtained with the LLT code are compared with previously obtained results from BEM in the
following sections. The input parameters used in the computations are summarized in Table 1.

Chord distribution Twist distribution Pitch angle U∞ ρ Time (LL) N λ aw


−3.48 log(r) + 6.88 0.005r2 − 0.55r + 15 -2.3◦ 10 m/s 1.225 kg/m3 160 s (∆ = 0.1 s) 30 7 0.23

Table 1: Parameters used in the computations. Parameters obtained from BEM optimization [1].

6
Circulation
The resulting plots of circulation vs. radial position along the blade are presented in Figure 4 for LLT with
uniform and cosine blade discretizations and BEM. Note that the results presented have been normalized
by U∞2
/Nblades πω.

Circulation
0.7

0.6

0.5

0.4
[/]

0.3 Lifting line theory


BEM
LLT cosine distribution
0.2

0.1

0
10 15 20 25 30 35 40 45 50
r [m]

Figure 4

It may be seen that the results from LLT and BEM show good agreement.

Induction factors
The computed axial induction factors vs. radial location are presented in Figure 5 for LLT with even
discretization and BEM. A good agreement is observed in the middle sections of the blade. In the inner and
outer regions however, higher differences are observed. These may be due to the corrections applied in BEM.
It is considered that the lifting line theory performs better than BEM in these regions of the blade, where
tip and trailing vortices highly influence the performance of the blade. The visually high offsets observed
for the tangential induction factor are not considered to be relevant as they are of the order of 10−2 .

Axial induction factor a Tangential induction factor


0.5 0.18
Lifting line theory
0.16 BEM
0.45
0.14
0.4
0.12
Lifting line theory
0.35 BEM 0.1
a1 [/]
a [/]

0.3 0.08

0.06
0.25
0.04
0.2
0.02

0.15 0
10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50
r [m] r [m]
(a) Axial (b) Tangential

Figure 5: Induction factors

Inflow angle and angle of attack


The velocity triangle is computed once the circulation and induced velocities are known. The inflow angle
and angle of attack are very high near the hub as expected. This feature led to some bugs in the LLT code
when the discretization was too fine. This was due to very high angle of attack for which lift and drag
7
coefficients could not be interpolated.
The difference between BEM and LLT are more pronounced in the Angle of attack distribution near the
hub. The BEM theory with correction factors fails to capture the 3D flow physics near the hub region
whereas LLT seems to more accurately predict the flow physics.

Inflow angle Angle of attack


25 18
Lifting line theory Lifting line theory
BEM BEM
16
20

14
15

[°]
[°]

12
10
10

5
8

0 6
10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50
r [m] r [m]
(a) Inflow angle Φ. (b) Angle of attack α.

Figure 6: Angles

Normalized forces
Once the lift and drag forces are computed at each element, the normalized axial and azimuthal forces may
be calculated. Good agreement is noted between BEM and LLT. The normalization used is 0.5ρU∞ 2
R.

Normalized axial force Normalized azhimutal force


1.2 0.15

0.8 0.1
F azhimutal [/]
F axial [/]

0.6
Lifting line theory
BEM
0.4 0.05
Lifting line theory
BEM
0.2

0 0
10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50
r [m] r [m]
(a) (b)

Figure 7: Normalized forces

8
Power Coefficient, Thrust coefficient and cosine distribution

Power coefficient Thrust coefficient


0.6 0.9

0.55 0.8

0.5
0.7
0.45

C T [/]
C p [/]

0.6
0.4
0.5
0.35 LL linear distribution LL linear distribution
BEM BEM
0.3 LL cosine distribution 0.4 LL cosine distribution
LL linear - corrected a w LL linear - corrected a w

0.25 0.3
6 6.5 7 7.5 8 8.5 9 9.5 10 6 6.5 7 7.5 8 8.5 9 9.5 10
[/] [/]

(a) (b)

Figure 8

A good agreement can be seen between BEM and lifting line theory for power and thrust coeficient. It
seems like lifting line accuracy can be enhanced by applying a cosine distribution discretization of the blade
instead of a linear one. The two different discretizations are displayed in Figure 9.

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50

(a) Linear distribution (b) Cosine distibution

Figure 9: Blade discretization with 15 elements.

This feature is valid for low values of the tip speed ratio. For λ > 8, the lifting line theory over-predicts
power and thrust coefficients. Furthermore, the cosine distribution does not perform well anymore. There
may be two reasons for this. The first one is that at high tip speed ratios the thrust coefficient gets high
and hence in BEM the Glauert correction is widely applied. This may cause disagreement with the lifting
line model.
The second one is that the axial convection speed aw used for the wake discretization was initially set
to 0.23 for all simulations. This was the first approach taken as the mean induction factor from λ = 7.
This however does not reflect the physics at all operating contitions as there is likely greater axial thrust
at higher λ. It was found that adjusting this aw according to the value obtained from BEM for the corre-
sponding λ led to results which better resembled BEM. These are depicted above as black crosses in Figure 8.

The results clearly indicated that this parameter is essential to obtain good agreement with BEM. In fact,
the wake obtained are extremely different, as showed in Figure 10. This is especially the case in heavily
loaded rotor conditions, where the axial induction factor increases.

9
10 10

5 5

0 0
z

z
-5 -5

-10 -10
10 10
800 800
0 600 0 600
400 400
y 200 y 200
-10 0 x -10 0 x

(a) λ = 6 (b) λ = 10

Figure 10: Wake convection and expansion at the blade root

This provided great accuracy to the linear distribution, while the cosine distribution was not improved by
the correction.

Sensitivity analysis
In this section the sensitivity of the results to different input parameters and discretization is investigated.

Time length
The influence of the wake length was inspected. The results are presented in Figure 11. It is reasonable
to expect that the more wake is considered, the more precise the computations will get. This effect gets
less pronounced with the wake length, because the induced effects of the vortex filaments far downstream
become negligible at a certain point. It is also important to underline that the assumption of frozen wake
ignores turbulence and the interaction of the wake. Hence, the results will not be totally accurate, but they
do paint a picture of the wind turbine performance which is fairly close to reality. Obviously, the longer
wake considered, the more computational time increases if the ∆t remains constant.
It is observed that convergence is obtained with a time length of 30 seconds, which corresponds to approx-
imately 3 turbine rotations (ω = 13.37 RPM at λ = 7). The power coefficient is seen to decrease because
the more wake is taken into account, the more wake losses are introduced due to the induced velocities.

0.65 0.398

0.396
0.6
0.394

0.55 0.392
C p [/]
[/]

0.39
0.5 BEM 0.388
t=2 s
t=5 s
0.386
t=10 s
0.45
t=30 s
t=80 s 0.384

0.4 0.382
0 5 10 15 0 1 2 3 4 5 6
Radial position Turbine rotations [/]
(a) Circulation Γ. (b) Convergence of Cp .

Figure 11: Time length influence, with a timestep ∆t = 0.1s, λ = 7.

10
Azhimutal discretization
The number of wake segments per rotation was inspected as well. As expected, it is obtained that a finer
discretization of the wake provides the best results. Figure 12 shows that a time step of 1 second may be
already enough accurate to obtain precise results.

0.65 0.46

0.44
0.6

0.42
0.55

C p [/]
[/]

0.4
0.5
0.38
BEM
0.45 t=0.1 s
t=0.5 s 0.36
t=2 s
t=4 s
0.4 0.34
0 2 4 6 8 10 12 14 0 0.5 1 1.5 2 2.5 3 3.5 4
Radial position t [s]

(a) Circulation Λ. (b) Convergence Cp .

Figure 12: Wake discretization influence, total time = 80 s, λ = 7.

Oscillations are visible in Figure 12b. This may be explained by the fact that the time step considered gets
too high to obtain a realistic and valid discretization of the wake.

Conclusions
LLT was applied to assess the performance of an improved rotor blade design from Assignment 1. These
results were compared with those obtained with the BEM theory and despite being based on two different
physical theories, it was observed that they resembled each other quite well. This gives more confidence in
the results.

BEM calculates the velocity triangle and circulation based on conservation of momentum, whereas LLT
does this based on conservation of vorticity.
The assumptions made in BEM ignore the wake effects and thus tip and root corrections as well as correc-
tions for heavily loaded rotors are added. In LLT, this is not necessary since the flow at the control points
is computed from vortex filaments.

It can be concluded that the LLT describes very well what occurs at the root and tip of the blade, since
these locations are large sources of trailing vortices which are directly used in computations.

11
References
[1] AE4135 - Rotor/Wake Aerodynamics, Assignment 1: BEM, Delft University of Technology, April 2018

12

You might also like