Вы находитесь на странице: 1из 126

Master of Science Thesis

The influence of laminar-turbulent


transition on the performance of a
propeller
A numerical and experimental investigation
R.F. Janssen
April 24, 2015

Ad

The influence of laminar-turbulent


transition on the performance of a
propeller
A numerical and experimental investigation
Master of Science Thesis

For obtaining the degree of Master of Science in Aerospace Engineering


at Delft University of Technology

R.F. Janssen
April 24, 2015

Faculty of Aerospace Engineering

Delft University of Technology

Delft University of Technology

Copyright Aerospace Engineering, Delft University of Technology


All rights reserved.

DELFT UNIVERSITY OF TECHNOLOGY


DEPARTMENT OF AERODYNAMICS

The undersigned hereby certify that they have read and recommend to the Faculty of
Aerospace Engineering for acceptance the thesis entitled The influence of laminarturbulent transition on the performance of a propeller by R.F. Janssen in fulfillment of the requirements for the degree of Master of Science.

Dated: April 24, 2015

Supervisors:
prof. dr. ir. L.L.M. Veldhuis

prof. Dr.-Ing. G. Eitelberg

Dipl.-Ing. C. Lenfers

dr. ir. B.W. van Oudheusden

Preface
This report is written as the final thesis as part of the Aerodynamics Master track for obtaining
the Masters Degree of Aerospace Engineer at Delft University of Technology.
This way I would like to thank my supervisors. I would like to thank Leo and Georg for their
supervision and healthy discussions. They also taught me that the flow direction is always
from left to right and I should never forget that. I would also like to thank Carsten for his
supervision during my time at DLR and his continued support via email when I was back at
the university.
I would also like to thank the technical staff of the High Speel Lab, Nico van Beek, Henk-Jan
Siemer, Peter Duyndam and Frits Donker-Duyvis, without their support I could not have
performed my wind tunnel tests.
Thanks to the rest of the people at the Transport Aircraft department of the DLR Insitute of
Aerodynamics and Flow Technology. If I had questions I could alway ask them and everyone
would offer their help.
I would also like to thank all the people I met during my time at Delft University of Technology. Especially the people from the aerodynamics basement, with whom I had some healthy
discussions, and who were always willing to lend a helping hand.
Finally I would like to thank my whole family for their support and understanding. In
particular I would like to thank my parents, if it was not for their support I would have never
been able to finish this study. And I would also like to thank my sister, whom Ill miss now
she moved to Australia.
Delft,
April 24, 2015
Ruud Janssen

MSc Thesis

R.F. Janssen

vi

R.F. Janssen

Preface

MSc Thesis

Summary
The influence of laminar-turbulent transition on a propeller blades performance is discussed in
this report. Numerical as well as experimental work has been performed. The computational
fluid dynamics (CFD) is performed using the German Aerospace Center (DLR) developed
TAU code, as well as an existing propeller lifting-line code. The laminar-turbulent transition
is simulated using the Ret correlation based transition model, which is compared to the
results of the Spalart-Allmaras one-equation turbulence model. To validate the CFD data,
experiments are performed in the Open Jet Facility of Delft University of Technology where
the laminar-turbulent transition is measured using an infrared camera.
The Ret correlation based transition model showed strange behaviour for advance ratios
below 1. After investigation it was found that vortical structures were forming in the flow
near the propeller blade. The results showed that at an advance ratio of 1.2, a separation
bubble was formed on the aft part of the blade. At and advance ratio of 1.0, the separation
region moves toward the leading edge, and at the same time the vortical structures started to
form on the aft part of the blade. Decreasing the advance ratio to 0.8, the separation region
moves to the leading edge and it can be seen that the strenght of the vortical structures
become stronger. On the lower half of the blade the separation region is near to the leading
edge, while near the blade tip still laminar flow is present up to around half chord. For an
advance ratio of 0.65 a strong separation can be seen, and again the vortical structures are
formed.
From the numerical results it was found that the difference in thrust coefficient between the
turbulence model and transition model changed from around 1% at an advance ratio of 0.65 to
about 27% at an advance ratio of 1.2. The differences in power coefficient changed from 1% at
an advance ratio of 0.65, to about 23% at an advance ratio of 1.2. The propeller efficiency did
not have these large changes, at an advance ratio of 1.2 the difference in propeller efficiency
change approximately 6%, while at an advance ratio of 0.8 the change was about 0.06%.
The differences between the blade element momentum theory and two RANS simulations
was larger, 57% lower values of the thrust coefficient for the RANS simulations. Due to
the two-dimensional nature of the blade element moment theory it is concluded that the lift
coefficients are high compared to RANS.
Qualitative infrared measurements have been performed on the propeller in a wind tunnel,
this was done because the location of laminar-turbulent transition was of interest. From this

MSc Thesis

R.F. Janssen

viii

Summary

it was found that for decreasing advance ratio, the separation line moves toward the leading
edge. While at an advance ratio of 0.8, leading edge transition could be seen on the middle
part of the blade. It was also checked what the influence of the Reynolds number was on
the propeller blade. If the Reynolds number is changed, the state of the boundary layer also
changes.
It can be concluded that the influence of laminar-turbulent transition on a propeller blades
performance can be determined to some degree. Comparing the RANS simulations with the
qualitative infrared measurements, some agreement between the results can be seen. The main
separation regions are predicted quite accurate using the correlation based model. Due to
the formation of vortical structures on the blade there is still some doubt about the accuracy
results of the simulations.

R.F. Janssen

MSc Thesis

Contents

Preface

Summary

vii

Nomenclature

xi

1 Introduction

1.1

Transition and Propellers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.1.1

Laminar-Turbulent Transition . . . . . . . . . . . . . . . . . . . . . . . .

1.1.2

Laminar-Turbulent Transition Research on Propellers . . . . . . . . . . .

1.2

Research Aim and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

1.3

Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

2 Numerical Setup

13

2.1

Propeller Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

13

2.2

Blade Element Momentum Theory . . . . . . . . . . . . . . . . . . . . . . . . .

15

2.3

Reynolds-Averaged Navier-Stokes equations . . . . . . . . . . . . . . . . . . . .

17

3 Experimental Setup

25

3.1

Wind Tunnel Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

3.2

Propeller Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

MSc Thesis

R.F. Janssen

Contents
3.3

Infrared Thermography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

3.4

Infrared Thermography Post-Processing Techniques . . . . . . . . . . . . . . . .

30

4 Numerical Results

33

4.1

Results for the Blade Element Momentum Theory . . . . . . . . . . . . . . . . .

33

4.2

Results for the Reynolds-Averaged Navier-Stokes Equations . . . . . . . . . . . .

34

4.2.1

Turbulent Model Results . . . . . . . . . . . . . . . . . . . . . . . . . .

35

4.2.2

Transition Model Results . . . . . . . . . . . . . . . . . . . . . . . . . .

46

4.2.3

Comparison Results Turbulent model and Transition model . . . . . . . .

57

Comparison BEMT and both RANS models . . . . . . . . . . . . . . . . . . . .

66

4.3

5 Experimental Results

73

5.1

Integration Time Influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

73

5.2

Advance Ratio Influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77

5.3

Reynolds Number Influence . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

78

6 Comparison Numerical and Experimental results

81

7 Conclusions and Recommendations

85

7.1

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85

7.2

Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87

Bibliography

89

A Additional Numerical Results

93

A.1 Cp values on propeller blade surface . . . . . . . . . . . . . . . . . . . . . . . .

94

A.2 Cp values slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

100

A.3 Blade loading characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . .

106

R.F. Janssen

MSc Thesis

Nomenclature

Latin Symbols
clstall
c
cd
Cf

Lift coefficient increment to stall


Chord
Drag coefficient
2
Skin friction coefficient, Cf = U
2

Ch

Stanton number, Ch =

q
u (h hw )

cl
cl
cm
CP

Lift gradient
Lift coefficient
Pitching moment coefficient
Power coefficient, CP = nP2 D4

Cp
CT

Pressure coefficient
Thrust coefficient, CT =

D
DL
Eb
fv1
G
H

Propeller diameter
Digital Level
Radiated power from black body
Spalart-Allmaras function
Gain
Shape Factor

Enthalpy

IT
J
x
M
n
O
P
p
Pr

Integration time

Advance Ratio, J = UnD


Distance
Mach number
Propeller rotational speed
Offset
Power
Pressure
Prandtl number

MSc Thesis

[]
[m]
[]
[]

T
n2 D4

[]

1
rad

[]
[]
[]
[]
[]
[m]
[J]
W 
m2

[]
[]
[]
h i
J
kg

[s]
[]
[m]
[]
1
s

[]
[kW ]
[P a]
[]
R.F. Janssen

xii
R
r
Re
Ret
Re
Rev
RP M
S
Sij
T
T
U
u
v
y+

Nomenclature
Propeller radius
Radial coordinate
Reynolds number
Transition onset momentum-thickness Reynolds number
Momentum-thickness Reynolds number
Strain-rate Reynolds number
Propeller RPM
Absolute value of strain rate
Strain-rate tensor
Temperature
Thrust
Velocity
Velocity in X-direction
Velocity in Y -direction
Non-dimensional wall distance

[m]
[m]
[]
[]
[]
[]
 rev 
min
1
 1s 
s

[K]
[N ]
m
s
m
s
m
s

[]

Greek Symbols

99


Angle of attack
Zero-lift angle of attack
Blade pitch angle
Intermittency
Relative difference
Displacement thickness
99% Boundary layer thickness
Emissivity coefficient
CT
Efficiency, = J C
P
Momentum thickness
Wavelength

Dynamics viscosity

Eddy viscosity

Eddy viscosity
Spalart-Allmaras variable

Density

Stefan-Boltzmann constant, 5.670 108


Shear stress
Vorticity vector

[o ]
[o ]
[o ]
[]
[%]
[m]
[m]
[]
[]
[m]
[m]
h i
kg
ms

kg
ms

[P a s]
[P a s]
h i
kg
m3



/f racW m2 T 4
[P a]
 rad 
s

Subscripts
0
0.7R
R.F. Janssen

Minimum
70% blade span
MSc Thesis

xiii

c
cr
crit
e
LIT
max
min
s
SIT
T rans
T urb
uc
w
x
y
z

Freestream
Corrected
Critical
Critical
External
Long Integration Time
Maximum
Minimum
Surface
Short Integration Time
Transition RANS simulation
Turbulent RANS simulation
Uncorrected
Wall
X-direction
Y -direction
Z-direction

Abbreviations
BEMT
BNF
BPR
CFD
DLR
IR
NETD
NLR
NUC
OJF
QESTOL
RANS
RotGISF
RPM
TDI
TS

MSc Thesis

Blade Element Momentum Theory


B
urgernahes Flugzeug
Bad Pixel Replacement
Computational Fluid Dynamics
German Aerospace Center
Infrared
Noise equivalent Temperature Difference
National Aerospace Laboratory of the Netherlands
Non Uniformity Correction
Open Jet Facility
Quiet-Efficient-Short-Take-Off-and-Landing Aircraft
Reynolds-Averaged Navier-Stokes Equations
Rotational-Flow Global Interferometry Skin Friction
Rotations Per Minute
Tech Development Inc.
Tollmien-Schlichting

R.F. Janssen

xiv

R.F. Janssen

Nomenclature

MSc Thesis

Chapter 1
Introduction
Air traffic is growing steadily over the last couple of decades, doubling every 15 years. It is
expected this trend will continue for the next years [1], this can be seen in figure 1.1a. With
the increase in air traffic the demand for new aircraft will increase. It is expected that most
of these new aircraft will be of the category of single-aisle aircraft, this can be seen in figure
1.1b.
This increase in aircraft can lead to serious airport congestion at the major airports around the
world. This could be solved by expanding the current airports, but this is not always possible
due to environmental or residential constraints. Because most of the aircraft deliveries fall in
the single-aisle category it should be possible to shift air traffic to smaller, already existing,
airports.
These small airports often have shorter runways which can not handle the current generation
of single-aisle aircraft. The runway length could be increased, but as with the larger airports
this is not always possible. Another way to solve this is by designing a short take-off and
landing aircraft. In the German research project B
urgernahes Flugzeug (BNF) the key technologies for a new quiet-efficient-short-take-off-and-landing aircraft (QESTOL) are developed
[2]. In the BNF project a QESTOL configuration is investigated using computational fluid
dynamics (CFD) and wind tunnel experiments. The QESTOL aircraft is defined as a twinengine single-aisle aircraft with a passenger capacity of 150, utilizing a gapless blown trailing
edge flap high-lift system. To get the best performance of the high-lift system an advanced
turboprop propulsion system is chosen.
Because of the importance of the interaction between the high-lift system and the turboprop
propulsion system, it is important to get a detailed and correct analysis of the turboprop
propulsion system. To get a detailed analysis of the turboprop propulsion system a CFD
analysis, using the Reynolds-averaged Navier-Stokes (RANS) equations, is done to simulate
the propeller. Using RANS to simulate the propeller usually assumes the whole flowfield
to be turbulent, but it is expected that the influence of laminar flow and laminar-turbulent
transition has a big influence on the performance of the turboprop propulsion system.
MSc Thesis

R.F. Janssen

Introduction

(a) Worldwide annual air traffic

(b) New aircraft demand for the period 20132032

Figure 1.1: Annual air traffic and new aircraft demand as predicted by Airbus [1]

1.1

Transition and Propellers

A fluid is not frictionless and therefore at walls and other interfaces a no-slip condition is
present. This means that the velocity of the fluid at the wall is equal to the velocity of the
wall itself. At a stationary flat plate, this means that u and v are both zero at the wall when
looking at a two-dimensional flow. This no-slip condition leads to the existence of a thin
boundary layer [3], a region in the flow field where the velocity of the flow changes from the
no-slip condition to the velocity of the flow outside the boundary layer. The main controlling
parameter in viscous flows is the dimensionless Reynolds number, given in equation (1.1).
Re =

U L

(1.1)

Boundary layers are subdivided in two types, laminar boundary layers and turbulent boundary
layers. Laminar boundary layers have a smooth profile, whereas turbulent boundary layers
are fluctuating [4]. In figure 1.2 the friction coefficient can be seen for increasing Reynolds
number for a flat plate [5]. As can be seen there are two regions, first a laminar flow and next
a turbulent flow after a Reynolds number of around 5 105 . It can be seen that the value of
the friction coefficient increases when the flow changes from laminar to turbulent flow. From
this it can be seen that a laminar boundary layer is preferred over a turbulent one to minimize
friction losses.
The flat plate boundary layer has been studied extensively and much is known about this type
of boundary layer. Using the analysis by Karman one can derive the Karman type integral
relations. In figure 1.3 the flow past a flat plate is shown. The outer streamline is taken as
the position where the velocity has reached 99% of the external flow velocity. When the inlet,
outlet, plate and outer streamline are taken as the control volume for the integral analysis,
one arrives at the compressible integral relations given in equations (1.2), (1.3) and (1.4) [4].
These equations can be very useful in analysing the properties of boundary layers. The value
of shows what the displacement effect of the boundary layer is, as shown in figure 1.3.
Here it can be seen that the boundary layer in effect adds thickness to the body contour.
The momentum thickness, , is related to the drag. H is called the shape factor and this is
a value which is often used in boundary layer analysis [4].

Z 99 
u

=
1
dy
(1.2)
Ue
0
R.F. Janssen

MSc Thesis

1.1 Transition and Propellers

Figure 1.2: Friction coefficient for flat plate flow [5]

99

=
0

H=

u
Ue



u
1
dy
Ue

(1.3)

(1.4)

Figure 1.3: Displacement effect of a boundary layer [6]

Pure laminar boundary layers are well understood nowadays. As well as pure turbulent
boundary layers, even though the finest scales of turbulence are still part of thorough investigation. The problem is in the connection of laminar and turbulent boundary layers, the
laminar-turbulent transition.
MSc Thesis

R.F. Janssen

Introduction

1.1.1

Laminar-Turbulent Transition

Laminar-turbulent transition in flows normally arises from small instabilities in the flow [7].
The way these instabilities grow depends on the initial conditions provided by the freestream
conditions [8].

Two-Dimensional flows
Transition in two-dimensional flows can be classified in three cases [9]. These three cases are
Natural transition. Natural transition is the laminar-turbulent transition occuring due
to small instabilities in the flow which grow and form two- dimensional TollmienSchlichting (TS) waves. These TS waves grow and form three-dimensional loop vortices
with high fluctuations. After this the high fluctuations in the flow develop into turbulent spots which eventually develop into a fully turbulent flow. This process is shown
in figure 1.4.
Bypass transition. When the freestream disturbances are very strong it is possible that
the flow transitions to turbulent very quickly because of turbulent spots or subcritical
instabilities. This kind of transition is then called bypass transition [10]. Main causes
for this kind of transition are roughness and high freestream turbulence levels [11].
Separated-flow transition. Transition due to laminar separation is called separation
induced transition. When a laminar boundary layer separates under the influence of a
pressure gradient it is possible a transition zone develops within the separation bubble
[12]. This type of transition is shown in figure 1.5.

Figure 1.5: Sketch of separated-flow


transition [13]

Figure 1.4:
Idealized sketch of
laminar-turbulent transition on a flat
plate [13]
R.F. Janssen

MSc Thesis

1.1 Transition and Propellers

Three-Dimensional flows
In three-dimensional flows the behaviour is completely different compared to the corresponding two-dimensional flow. Most studies on three-dimensional boundary layers are directed
toward swept wing flows. In these swept wing flows, four instability types were found that
lead to transition: leading edge instability, streamwise instability, centrifugal instability and
crossflow instability. In figure 1.6 a sketch is shown with the instabilities working on a swept
wing.

Figure 1.6: Instabilities found on


swept wing flow [14]
Figure 1.7: Boundary layer profile
with crossflow present [7]

Leading edge instability is connected to the basic instability of the attachment-line flow
or to turbulent disturbances that propagate along the wing leading edge [15].
Streamwise instability is similar to the two-dimensional instability, where TS waves are
developed.
Centrifugal instability plays a role in concave surface regions on the swept wing. Here
Gortler vortices appear in the flow leading to transition [16].
Crossflow instability is linked to the in-plane curvature of the streamlines [14], which
cause centrifugal forces. Outside the boundary layer these centrifugal forces are balanced
with the pressure forces. But inside the boundary layer these centrifugal forces decrease
when nearing the wall, leading to the crossflow component. This is shown in figure
1.7. Because the crossflow component of the flow must tend to zero when nearing the
boundary layer edge, an inflection point must be present. Inflection points in boundary
layers are a necessary and sufficient condition for instability [17], making crossflow an
inflectional instability.

Most of the times in three-dimensional flow all of these instabilities interact with one other
and make it difficult to tell which instability is observed [14].
MSc Thesis

R.F. Janssen

1.1.2

Introduction

Laminar-Turbulent Transition Research on Propellers

A propeller blade is a three-dimensional system which is rotating at high speed. This rotating
motion increases the difficulty of the analysis of the laminar-turbulent transition. The first
well known investigation of the effect of rotation on boundary layers was carried out by H.
Himmelskamp [18]. Himmelskamp performed point-wise pressure measurements at different
radial sections on a rotating propeller blade. Figure 1.8 shows local lift coefficients at different
radial sections for different angles of attack. The results obtained by Himmelskamp show that

Figure 1.8: Local lift coefficients on a rotating propeller at various radial sections [18]

near the hub the local lift coefficient is increased, Himmelskamp concluded that this was due
to a delay of separation. The separation is delayed because of the additional accelaration
forces which are present due to the Coriolis forces caused by the rotation of the blade. These
additional accelaration forces have the same effect as a favorable pressure gradient, thus
delaying the separation. Another effect could be due to the centrifugal forces felt by the
particles in the boundary layer, making the boundary layer thinner compared to the stationary
two-dimensional case. Thin boundary layers are less sensitive to adverse pressure gradients,
this means that due to the centrifugal forces transition to turbulence is delayed on a rotating
propeller.
More recent is the research performed by Sch
ulein et al [19]. Here a new measurement
technique was proposed called rotational-flow global interferometry skin friction (RotGISF).
The RotGISF technique was tested on a generic propeller
model rotating at speeds up to 240
m
[Hz] at freestream flow speeds between 30 and 70 s . RotGISF is an optical interferometry
which is based on the relation between the thinning of an oil film, applied on the blade surface,
and the local shear stress. A schematic of the method is given in figure 1.9.
After processing, the interference images are obtained, two examples are shown in figure 1.10.
Figure 1.10a shows the propeller operating at a high advance ratio, while figure 1.10b shows
the propeller operating at a low advance ratio. It can be seen that the flow pattern changes
when the advance ratio is changed. The change in flow topology is sketched in figure 1.11 for
R.F. Janssen

MSc Thesis

1.1 Transition and Propellers

Figure 1.9: Schematic RotGISF technique used by Sch


ulein et al [19]

(a)

(b)
Figure 1.10: Interference images after image processing, (a) J = 1.15 and (b) J = 0.81 [19]

MSc Thesis

R.F. Janssen

Introduction

advance ratios between 0.50 and 1.20.


Figure 1.11a shows the flow topology for J = 0.50. It can be seen that near the hub the flow
is laminar and a separation bubble is occuring near the trailing edge of the blade. Beyond the
critical radius, rcr , a leading edge separation bubble is starting to form. This leading edge
separation bubble leads to a sudden transition to turbulent flow. At the outer part of the
blade a large three- dimensional conical vortex is observed. It is thought this conical vortex
is closely related to the delta wing vortices occuring at high angles of attack, because of the
high leading edge sweep angle.
At an advance ratio of J = 0.80 it can be seen that the flow topology has changed, see
figure 1.11b. The critical radius is moving to the outer part of the propeller blade, while no
conical separation vortex can be seen. Below the critical radius the flow is laminar and near
the trailing edge a separation bubble is occuring. At J = 1.00 a large part of the blade is
laminar, with a separation bubble at the trailing edge, this can be seen in figure 1.11c. A
leading edge separation bubble can only be seen at the tip of the blade, rRcr 0.8. Finally
it can be seen in figure 1.11d that at high advance ratios the flow is fully laminar, where a
separation bubble forms over the trailing edge of the blade.
The results from Sch
ulein et al show some agreement with the work of Kuiper, who did an
optical investigation on a ship propeller [20]. Figure 1.12 shows an overview of the boundary
layer regimes found on the ship propeller blade. It can be seen that at the tip of the propeller,
line AB, a short laminar separation bubble was observed near the leading edge. This short
separation bubble makes the boundary layer turbulent over the rest of the blade chord. Line
BC defines the critical radius, and Kuiper found it was strongly dependant on the propeller
loading. Below the critical radius it was found that a transition region was formed, given
by line CD. Very close to the hub it was found that the flow was laminar and a laminar
separation bubble was formed, this is due to a low sectional Reynolds number combined with
a thick propeller section.

R.F. Janssen

MSc Thesis

1.1 Transition and Propellers

(a) J = 0.50

(b) J = 0.80

(c) J = 1.00

(d) J = 1.20

Figure 1.11: Sketch of the flow topology over the suction side of the propeller blade at different
advance ratios [19]

MSc Thesis

R.F. Janssen

10

Introduction

Figure 1.12: Sketch of the flow topology on a ship propeller blade [20]

R.F. Janssen

MSc Thesis

1.2 Research Aim and Objectives

1.2

11

Research Aim and Objectives

This section describes the proposed thesis question and objectives. The goal of the research
can be reached by answering the main question and its sub-questions.
First the main research question:
Can the influence of laminar-turbulent transition on a propeller blades performance be determined through the use of CFD and wind tunnel experiments?
To be able to answer the main question a couple of sub-questions should be answered first.
These sub-questions are formulated as follows:
Is it possible to use current CFD models to calculate accurately the position of the
laminar-turbulent transition?
Is it possible to measure the laminar-turbulent transition using infrared thermography?
Do the results from the CFD simulations agree with the results from the wind tunnel
experiment?
Does laminar-turbulent transition influence the performance of a propeller?
From the main question and sub-questions the main objective of the thesis can be defined as:
Determine the influence of the laminar-turbulent transition on the performance of
a propeller, by means of CFD simulations and wind tunnel experiments.

1.3

Thesis Outline

This report consists of seven chapters. This first chapter is an introduction to work to
be presented. Chapter 2 will explain the numerical setup, here the propeller model to be
investigated will be described as well as the computational techniques which will be used
for the analysis. To be able to validate the results obtained by computations, chapter 3
will give a description of the experimental setup as well as the post-processing techniques
which will be used. Having finished describing the experimental setup, chapter 4 will present
the numerical results. Following the numerical results, chapter 5 will present the obtained
experimental results. Having both numerical and experimental results, chapter 6 will compare
the numerical and experimental results. Finally, chapter 7 will present the conclusions drawn
from this work together with recommendations for future research.

MSc Thesis

R.F. Janssen

12

R.F. Janssen

Introduction

MSc Thesis

Chapter 2
Numerical Setup
This chapter describes the numerical methods which are used to analyse the isolated propeller. First the XPROP propeller will be shown in section 2.1. Section 2.2 will explain
the Blade Element Moment Theory, with a short description about MSES and XROTOR.
Finally, section 2.3 the Reynolds-Averaged Navier-Stokes equations will be elaborated upon,
together with the computational grid and the used turbulence models.

2.1

Propeller Model

The propeller which is used to perform the simulations and experiments is the Delft University of Technology XPROP propeller. Figure 2.1 shows the wind tunnel model of the XPROP
propeller and figure 2.2 shows the XPROP model used in the CFD simulations. When comparing the two models, it can be seen that the connection between the hub and the propeller
blade is not modelled correctly. This is due to missing information about the hub-propeller
connection in the XPROP documentation.
In the CFD model, a right handed coordinate system is used. As can be seen in figure 2.2, the
Y -axis is defined in positive up direction, the Z-axis is defined to the right as positive, which
means the X-axis is pointing away from the reader. The origin of the coordinate system is
defined at r = 0 [m] at the location where the spinner starts.
The XPROP propeller is a six-bladed propeller which has a diameter of 0.4064 [m]. The
hub of the propeller is 0.0884 [m] in diameter. The blade pitch angle of the propeller can be
adjusted, and it was chosen to set the blade pitch angle at 30[o ] at 70% of the blade span.
An overview of the XPROP propeller characteristics is given in Table 2.1.

MSc Thesis

R.F. Janssen

14

Numerical Setup

Figure 2.1: Front view wind tunnel


model of the XPROP propeller

Figure 2.2: Front view CFD model of


the XPROP propeller

Table 2.1: XPROP propeller characteristics

Number of blades
Propeller diameter
Hub diameter
Blade pitch angle at Rr = 0.7

R.F. Janssen

6
0.4064 [m]
0.0884 [m]
30 [o ]

MSc Thesis

2.2 Blade Element Momentum Theory

2.2

15

Blade Element Momentum Theory

A classical approach in the design and analysis of propeller blades is the Blade Element
Momentum Theory (BEMT). BEMT is a combination of the blade element theory and the
momentum theory for propellers. The BEMT used in this work uses a combination of MSES
and XROTOR. MSES is used to obtain the blade section characteristics, while XROTOR
calculates the propeller performance.

MSES
To obtain the XPROP airfoil data the program MSES, written by M. Drela, is used. MSES
is based on the steady Euler equations which are discretised using a finite volume method.
To account for boundary layers, a two-equation integral formulation with lagged-dissipation
closure model is used. The displacement thickness is used to couple the viscous and inviscid
regions of the flow [21].
The inviscid flow is resolved using the steady Euler equations [22]. These Euler equations are
solved by a full Newton method on an conservatively discretised intrinsic streamline grid.
The boundary layer is calculated using a two-equation integral formulation based on dissipation closure which is valid for both laminar and turbulent flows. To account for transition the
eN method developed by van Ingen [23] and Smith and Gamberoni [24] is utilised. The eN
method assumes transition when the most unstable Tollmien-Schlichting wave in the boundary layer has grown by a factor of eN , where often N is chosen as 9.
Coupling the discrete inviscid Euler equations together with the discrete boundary layer
equations, a system of nonlinear equations is obtained. To solve this system of nonlinear
equations a Newton solution procedure is used [22]. Because the resulting Newton matrix is
very sparse and structured it can be solved very fast using a direct Gaussian block-elimination
method. Figure 2.3 shows a simplified diagram of the solution procedure MSES uses.

Blade geometry
Freestream conditions

Panel solution

Initial surface
grid

Grid
smoothing

Initial solution

Circulation
converged?

cl
cd
cm

Figure 2.3: Simplified diagram of solution procedure MSES

The input of a simulation consists of the blade section geometry, freestream conditions and
simulation parameters [21]. Using these inputs, first a panel solution is obtained. This panel
MSc Thesis

R.F. Janssen

16

Numerical Setup

solution divides up the domain in blocks, giving the skeleton to generate the grid. After the
panel solution is obtained an initial surface grid is constructed, which distributes the surface
nodes over the blade geometry. The obtained surface grid is smoothed using an elliptic grid
smoother to remove kinks and overlaps, and makes the grid streamlines correspond to the
streamlines of the panel solution. Using the obtained grid an initial solution is calculated.
After this, the Euler equations and boundary layer equations are solved using a Newton
iteration method. After the solution is converged the lift, drag and moment coefficients are
given as outputs which are used as input for the XROTOR program.

XROTOR
The XROTOR program is written by the same person as the MSES program, M. Drela.
XROTOR is released under the GNU General Public License in 2011. XROTOR utilizes
the classical propeller lifting-line theory, which allows for the specification of lift and drag
properties, without giving geometrical data. To take compressibility into account for higher
Mach number flows, Prandtl-Glauert correction is applied.
XROTOR uses an iterative solution procedure, which can be seen in figure 2.4. The initial

Compute
initial
circulation

Set circulation

Compute
induced
velocities

Compute
updated
circulation

Compute
section
values

Compute
section cl and
cd values

Compute
circulation

Circulation
converged?

Compute
CT, CP and

Figure 2.4: Simplified diagram of solution procedure XROTOR. Reproduced from [25]

circulation is calculated assuming no induced effects are present. Knowing the initial circulation a Newton iteration method is used to obtain a converged solution for the circulation.
When the circulation has converged the values for the thrust coefficient, torque coefficient
and efficiency are calculated.
For each blade section the following parameters have to be provided:

Maximum lift coefficient, clmax


Minimum lift coefficient, clmin
Zero-lift angle, 0
l
Lift gradient, c

h
l
Lift gradient after stall, c
stall
Lift increment to stall, clstall
R.F. Janssen

MSc Thesis

2.3 Reynolds-Averaged Navier-Stokes equations

Minimum drag, cd0


cd =cd0
Lift coefficient at minimum drag coefficient, cl
Gradient of drag coefficient with respect to lift coefficient squared,
Critical Mach number, Mcrit

17

cd
c2l

These parameters are obtained from the results of the airfoil calculations performed by MSES.
The obtained lift and drag curves for each blade section are fitted using multi-variable optimizations using a Matlab routine provided by T. Sinnige which uses a Nelder-Mead optimization
method [25].

2.3

Reynolds-Averaged Navier-Stokes equations

For the Reynolds-Averaged Navier-Stokes (RANS) simulations of the XPROP propeller the
German Aerospace Center (DLR) developed TAU code is used. The development of the TAU
code was started in the German CFD project MEGAFLOW [26], which was a collaboration
between DLR, German universities and the German aircraft industry.
The TAU code is an unstructured CFD solver which can solve the unsteady RANS equations.
It is well suited for mixed-element meshes consisting of tetrahedron, prisms, hexahedra and
pyramids [27].
The XPROP propeller will be evaluated using a fully turbulent model and a model including laminar-turbulent transition. Table 2.2 gives the operating points which are simulated
using both the fully turbulent and transition model, where n is the rotation frequency of the
propeller. For the spatial discretisation a central finite volume approach is chosen. The time
discretisation is done using a Runge-Kutta scheme. In the past these settings have proven to
give good results for propeller simulations using the DLR TAU code [28] [29].
Table 2.2: Propeller operating points for the CFD simulations, 0.7R = 30 [o ]

J [-]
0.3
0.5
0.65
0.8
1.0
1.2

 
v m
s
26
26
26
26
26
26

 
n 1s
213.25
127.95
98.45
79.97
63.98
53.31

Computational Grid
The computational grids for the XPROP propeller are generated using the commercial software Centaur, developed by CentaurSoft [30]. As stated before the mesh can consist of
MSc Thesis

R.F. Janssen

18

Numerical Setup

mixed-elements. Near the wall prism elements will be used, while the remaining domain will
consist of tetrahedral elements. The grid is constructed using best practice guidelines and
personal experience obtained working on the BNF project [28].
The mesh size can be reduced by simulating only one propeller blade and utilising periodic
boundary conditions, this is shown in figure 2.5 where the simulated part is highlighted in
black and red. The use of periodic boundary conditions puts a few restictions on flows which
can be investigated, namely only flows where the flow direction is parallel to the rotation axis.
The periodic boundary conditions would also give rise to an unphysical rotating boundary
layer over the spinner and nacelle surfaces, so the spinner and nacelle surfaces are set as
inviscid walls, highlighted in black in figure 2.5. From a previous study [28] it was also found
that a small error in surface mesh discretisation led to a region of separated flow on the aft
part of the nacelle. To overcome this problem it was opted to extend the nacelle to the end of
the domain. To further simplify the modeling, the whole domain is rotating with the propeller
blade. Because of this simplification no sliding mesh or Chimera approach is needed, further
reducing the computational costs.
The surface of the propeller, the spinner and the nacelle are discretised using triangular
elements. The surface grid of the propeller blade is shown in figure 2.6. As the blade is thin,
additional refinement is done at the leading edge, this is shown in figure 2.7.
To ensure the boundaries have no influence on the final solution, the domain extends from
10 [m] upstream of the propeller blade, to 10 [m] downstream of the blade and the domain
has a height of 10 [m]. The 10 [m] was chosen because it was around 50 times the propeller
radius. To resolve the flow more accurate near the blade the volume mesh is fine near
the blade surface, while it is coarse far away form the blade. This is allowed, as far away
from the blade the flow is almost uniform and no large variable changes are present. A fine
refinement region of the volume mesh, having a radius of 1.2 times the propeller radius, is
chosen to start at an X position where the spinner starts, and this region extends to about
6 propeller diameters downstream. To avoid sudden changes in tetrahedral size, regions of
gradual refinement are defined around the fine refinement area. The complete volume mesh
is constructed of approximately 41 million elements, 13 million prism elements and 28 million
tetrahedral elements. The volume mesh for the complete domain can be seen in figure 2.8,
where a slice at Z = 0 is shown. The regions of coarse and finer elements can be seen. To
better show the volume mesh near the propeller blade, figure 2.9 shows the volume mesh
zoomed in more around the blade.
Finally in figure 2.10 the blade tip is shown where also the distinction can be made between
the prism elements and the tetrahedral elements. The height of the first cell over the blade
surface is constructed to be y + = 1. This is needed for resolving the boundary layer in the
simulations.

R.F. Janssen

MSc Thesis

2.3 Reynolds-Averaged Navier-Stokes equations

19

Figure 2.5: Simulated part of the XPROP propeller, highlighted in black is the spinner and
nacelle, and in red the propeller blade

Figure 2.6:
blade

MSc Thesis

Surface grid propeller

Figure 2.7: Zoom of top corner surface grid propeller blade

R.F. Janssen

20

Numerical Setup

Figure 2.8: Volume mesh of the complete domain, slice at Z = 0

Figure 2.9: Volume mesh, slice at Z = 0

R.F. Janssen

MSc Thesis

2.3 Reynolds-Averaged Navier-Stokes equations

21

Figure 2.10: Zoom of volume grid at blade tip, slice at Z = 0

Turbulent model
The Reynolds averaging of the Navier-Stokes equations leads to the RANS equations. Next to
the unknown pressure and three velocity components, Reynolds averaging also introduces six
unknowns, the Reynolds stress components u0i u0j [31]. This means there are ten unknowns,
while there are only four equations: mass conservation and the three momentum equations.
This means that the system of equations is not closed, and to close the system additional
equations are needed which model the Reynolds stress components. These Reynolds stress
models can be divided in multiple categories:

Algebraic (zero-equation) models;


One-equation models;
Two-equation models;
Second-order closure models (Reynolds stress equation models).

Here the zero-, one- and two-equation models use the Boussinesq approximation to model
the Reynolds stresses. The Reynolds stress equation (RSE) models use the exact differential
equation for the Reynolds stress tensor, which makes them computationally more intensive.
For this thesis it was chosen to use the Spalart-Allmaras one-equation model [32] as it has
shown good results compared to experiments, see St
urmer et al [29].
The Spalart-Allmaras model is written in terms of the eddy viscosity t , equation (2.1), with
the Reynolds stresses given by ui uj = 2t Sij [32].
t = fv1
MSc Thesis

(2.1)
R.F. Janssen

22

Numerical Setup

Here is the working variable, which is obtained solving a transport equation, and fv1 is a
function. If the reader is interested in the full description of the Spalart-Allmaras model, the
reader is referred to the paper written by P. Spalart and S. Allmaras [32].

Transition Prediction model


The prediction of laminar-turbulent transition in CFD applications can be accomplished using
different methods. First there are the weak coupling methods, which calculate the mean
flow field, followed by a boundary layer calculation [33]. The second method is using socalled stability techniques such as the eN -method developed by van Ingen [23] and Smith
and Gamberoni [24]. Finally there are the correlation based techniques, one example is the
method developed by Langtry and Menter [34].
Here it is chosen to use the Ret correlation based transition model of Langtry and
Menter as it has shown promising results in the simulation of turbomachinery [35] and wind
turbines [36]. The weak coupling methods and stability techniques also have difficulties with
complex three dimensional flows, as is the case with the flow around a propeller blade. The
big advantage of the correlation based transition model is that it uses only local variables
which makes it ideal for implementation in a parallel CFD solver. In the DLR TAU code the
Ret correlation based transition model is coupled with the SST k turbulence model
as was suggested by Langtry and Menter [34].
The model uses the strain-rate Reynolds number to link the transition onset Reynolds number
from an empirical correlation and local boundary layer quantities. The equation for the strainrate Reynolds number is given in equation (2.2).
y 2
Rev =


u y 2
=
S
y

(2.2)

From equation (2.2) it can be seen that Rev is a local property as it only depends on the
density , the wall distance y, viscosity and the shear strain rate S. The strain-rate Reynolds
number is scaled to have a maximum value of one inside the boundary layer. From this it
follows that the maximum of the strain-rate Reynolds number profile is proportional to the
momentum-thickness Reynolds number,
given

 in equation (2.3). The value of 2.193 is chosen
2.193Re
such that for a Blasius profile max
= 1.
Rev
Re =

max (Rev )
2.193

(2.3)

The link between the strain-rate Reynolds number and the empirical tranition correlations
is accomplished using two transport equations. One for the intermittency , which is used
to trigger the tranisition locally, given in equation (2.4). And one transport equation for the
t , which is required to capture the
transition onset momentum thickness Reynolds number, Re
nonlocal influence of the turbulence intensity, given in equation (2.7). This second transport
equation connects the empirical transition correlations to the transition onset criteria in the
R.F. Janssen

MSc Thesis

2.3 Reynolds-Averaged Navier-Stokes equations

23

intermittency transport equation.

() (Uj )
= P E +
+
t
xj
xj




t
+
f xj

(2.4)

The value f in equation (2.4) is a constant. The transition sources in the intermittency
transport equation are given by P1 , see equation (2.5). Flength is an empirical correlation
which controls the transition region length. S is the strain-rate magnitude. Fonset is a
correlation which controls the transition onset location. It should be noted that both Flength
and Fonset are dimensionless functions, while ca1 and ce1 are constants.
P1 = Flength ca1 S [Fonset ]0.5 (1 ce1 )

(2.5)

The E is given as the destruction/relaminarization source and is given in equation (2.6).


Again ca2 and ce2 are constants, while is the vorticity magnitude. Fturb is a function which
used to disable the destruction/relaminarization source outside of a laminar boundary layer
or in the viscous sublayer.
E = ca2 Fturb (ce2 1)

(2.6)

To transform a non-local correlation for Ret into a local form, the tranport equation for the
transition momentum thickness is introduced, see equation (2.7).




"
#
t
t
Re
Uj Re
t

Re
+
= Pt +
t ( + t )
(2.7)
t
xj
xj
xj
t to match the local values
Here Pt is the source term which forces the transported scalar Re
of the empirical correlations for Ret . The equation for Pt is given in equation (2.8). t is
a model constant which controls the diffusion coefficient.


t (1.0 Ft )
Pt = ct
Ret Re
(2.8)
t
In equation (2.8) ct is a constant which controls the magnitude of the source term. The time
scale t is there for dimensional reasons and was determined on basis it had to scale with the
convective and diffusive terms in the transport equation. Ft is a blending function which
t in from the freestream.
turns off the source term in the boundary layer and diffuses Re
For a detailed description of the model and how it was conceptualised the reader is referred
to the dissertation of Langtry [37] and the paper by Langtry and Menter in which the full
model correlations have been published [34]. For the implementation in the DLR TAU code
the reader is referred to the paper by Grabe and Krumbein [38].

MSc Thesis

R.F. Janssen

24

R.F. Janssen

Numerical Setup

MSc Thesis

Chapter 3
Experimental Setup
This chapter explains the experimental setup used during the testing of the XPROP propeller
in the wind tunnel to determine the transition location using infrared thermography. Section
3.1 shows an overview of the wind tunnel. After this, section 3.2 gives an overview of the
propeller test setup. Subsequently, infrared thermography is elaborated upon in section 3.3.
And finally in section 3.4 the post-processing techniques are explained.

3.1

Wind Tunnel Facility

The experiments were performed in the Open Jet Facility (OJF) of the Delft University of
Technology.
  The OJF is a closed circuit open jet facility which has a maximum wind velocity
of 30 m
s , a schematic of the OJF is given in figure 3.1. The OJF features a test section
which is 6.0 [m] wide, 6.5 [m] high and 13.5 [m] long. The settling chamber is equipped
with a honeycomb flow rectifier and five fine-mesh screes to reduce the turbulence level of the
flow and remove spatial velocity deviations. The velocity deviations are smaller than 0.5% in
the vertical plane at two meters from the outlet, and the OJF has a longitudinal turbulence
intensity level lower than 0.24%. Noise levels are reduced using perforated plates installed on
mineral wood and sound absorbing foam which cover the entire tunnel. There are no special
measures taken to attanuate the fan noise [39].

MSc Thesis

R.F. Janssen

26

Experimental Setup

Figure 3.1: Schematic of the Open Jet Facility with an wind turbine in the test section. Reproduced from [39]

3.2

Propeller Test Setup

The experiments were conducted on the Delft University of Technology XPROP propeller.
The chosen configuration was an isolated propeller to investigate the transition of laminar to
turbulent flow on the blades. The test setup consists of several parts: the propeller itself, the
air turbine motor, a heat source in the form of a 1 [kW ] power lamp, and finally the data
measurement hardware.
The XPROP propeller is driven using a Tech Development Inc. (TDI) 1999A air turbine
motor, which is supplied with air from the central air supply system of the Delft University
of Technology High Speed Wind Tunnel Laboratory. The air turbine motor is designed to
deliver 73 [kW ] at 22500 [RP M ] with a drive air total pressure of 34.47 [Bar] and a drive total
air temperature of 288.15 [K] [40]. A detailed description of the operation and installation of
the air turbine is described in [41].
The temperature on the propeller blade surface was measured using an infrared camera, as
will be explained in more detail in section 3.3. To measure the temperature on the surface
of a rotating propeller blade, phase locking is needed. This is done by supplying the infrared
camera with a trigger signal coming from the air turbine motor. This signal from the air
turbine motor is generated by two magnetic pickups, which produce one pulse per revolution.
The signal coming from the air turbine motor cannot be directly connected to the infrared
camera, a Stanford Research Systems DG535 Digital Delay/ Pulse Generator was used, as
shown in Figure 3.2. The DG535 can measure external signals and give out precisely controlled
pulses. These pulses can be used to trigger the infrared camera.
A schematic of the test set-up is shown in figure 3.3. The propeller test setup, consisting of
R.F. Janssen

MSc Thesis

3.2 Propeller Test Setup

27

Figure 3.2: Stanford Research Systems DG535 Digital Delay/ Pulse Generator. Reproduced
from [42]

the XPROP propeller, TDI 1999A air turbine motor, infrared camera and a 1 [kW ] lamp,
can be seen in Figures 3.4 and 3.5. Figure 3.4 shows a side view of the test setup, and Figure
3.5 shows a view of the propeller, lamp and infrared camera.

Figure 3.3: Schematic of the experimental test set-up

MSc Thesis

R.F. Janssen

28

Experimental Setup

Figure 3.4: Side view of the wind tunnel setup

3.3

Figure 3.5: View of the experimental


setup

Infrared Thermography

In aerodynamics it can be useful to know the temperature distribution over a body. Especially
the heat flux at the wall is of interest, as this can be related to the skin friction through the
Reynolds analogy, given in equation (3.1). Here, Cf is the friction coefficient, Ch the Stanton
number and P r the Prandtl number [4].
2
Cf
= 2P r 3
Ch

(3.1)

Using the characteristic that laminar boundary layers have lower skin friction than turbulent
boundary layers, using infrared thermography one can determine the state of the boundary
layer by looking at the temperature. Regions of high temperature on the propeller blade
indicate low heat flux and thus low skin friction, which would suggest the flow is laminar
in those regions. Regions of low temperature on the other hand indicate a high heat flux,
meaning high skin friction and thus a region of turbulent flow. During the measurements the
propeller blade will be heated using a heat source in the form of a high power lamp.
When an object has a temperature above absolute zero, it radiates electromagnetic radiation,
which is linked to the temperature of the objects surface. For a black body, the wavelength
at which maximum radiation is emitted is given by Wiens displacement law, equation (3.2)
R.F. Janssen

MSc Thesis

3.3 Infrared Thermography

29

[43].
max Ts = 2.898 103

[m K]

(3.2)

For a body at 300 [K] this means max is around 10 [m]. The total power radiated from a
black body can be given by the Stefan-Boltzmann equation, which is the maximum radiated
power possible at a certain temperature. For real bodies the Stefan-Boltzmann equation is
multiplied by , the spectral emissivity coefficient, which is 1 for a black body.
Eb = T 4

(3.3)

Here is the Stefan-Boltzmann constant, 5.670 108


ature.

W
m2 K 4

, and T is the surface temper-

The camera used during the experiments was the CEDIP Titanium 530L, as is shown in
Figure 3.6. The camera uses a Focal Plane Array (FPA) sensor, consisting of 320 x 256
detectors. These are Mercury Cadmium Telluride quantum detectors, which have a spectral
response in the range of 7.7 [m] to 9.3 [m]. At full resolution, the maximum frame rate is
250 [Hz]. The full specifications of the camera are given in Table 3.1. Here the abbrevation
NETD stands for Noise Equivalent Temperature Difference, which is the smallest variation
in detectable temperature.

Figure 3.6: CEDIP Titanium 530L camera. Reproduced from [44]

Table 3.1: CEDIP Titanium 530L specifications. Reproduced from [44].

Detector Material
Spectral Response
Frame Rate
Pitch
NETD
Aperture
MSc Thesis

Mercury Cadmium Telluride


7.7 - 9.3 [m]
250 [Hz]
30 x 30 [m]
< 25 [mK] @ 25 [o C] & 350 [s] integration time without filter
F/2
R.F. Janssen

30

Experimental Setup

The camera is connected to a computer, which controls the camera and is used to store the
obtained images. The software which is used to control the camera is Altair [45]. The main
functions of Altair are:
Set integration time
Non Uniformity Correction (see section 3.4)
Camera trigger
Recording of the images
The integration time is the time a detector is exposed to the electromagnetic radiation, and
as such determines how much radiation can be measured. The selection of integration time
determines which temperature range can be measured.
The propeller is a rotating system, which makes it difficult to measure the temperature
when no phase locking is used. Using phase-locking, the camera makes a measurement when
the blade is at the same position each rotation cycle. To accomplish this phase-locking the
camera is equiped with a trigger function.
The obtained infrared images are saved in the Cedip-PTW file format. To be able to view
the images outside of the Altair software, a Matlab routine is available to read in the data as
a matrix. After that operation is completed the data can be post-processed.

3.4

Infrared Thermography Post-Processing Techniques

The CEDIP Titanium 530L infrared camera utilizes an FPA sensor. This FPA sensor is made
up of multiple detectors, in the case of the CEDIP Titanium 530L 320 x 256 detectors, where
each detector has its own characteristics. To correct for these differences in characteristics a
Non Uniformity Correction (NUC) is applied. This NUC will introduce, for each detector, a
gain value and an offset value. In this way all detector response curves will be brought in line
with the average curve for the entire sensor.
The NUC can be performed by the software provided by CEDIP itself. However, to obtain
more control over the resulting correction, a NUC is programmed in Matlab . The NUC which
is used is the two point time integration method. This method is chosen because it does not
need a black body at two different temperatures, but only one temperature is needed. Using
the two integration times one can get a detector response curve. To get a good detector
response curve, multiple measurements need to be performed at each integration time.
A linear NUC is used, as from preliminary investigations it was found that a quadratic NUC
did not improve the results. The formula for the linear NUC is given in equation (3.4). Here
DL stands for Digital Level, the output signal of the detectors in [J]. The subscript c stands
for the corrected signal and uc is the uncorrected detector signal. The value G (i, j) is the
gain of the detector and O (i, j) is the offset.
DLc (i, j) = G (i, j) DLuc (i, j) + O (i, j)
R.F. Janssen

(3.4)
MSc Thesis

3.4 Infrared Thermography Post-Processing Techniques

31

The values of G and O need to be determined for each detector on the sensor. Using two
different integration times, long and short, the values for G and O can be determined. Introducting the term DLLIT for the detector signal for a long integration time, and DLSIT
for the short integration time detector signal. As was said before, multiple measurements at
each integration time are needed to perform a correct NUC. The equations for the gain and
offset are given in (3.5) and (3.6). Here the values hDLLIT i and hDLSIT i are the mean DL
of the complete sensor at a long integration time and short integration time respectively. The
values DLLIT (i, j) and DLSIT (i, j) are the mean values of the single detectors for the long
and short integration time, respectively.
G (i, j) =

O (i, j) =

hDLLIT i hDLSIT i

(3.5)

DLLIT (i, j) DLSIT (i, j)


DLLIT (i, j) hDLLIT i DLSIT (i, j) hDLSIT i
DLLIT (i, j) DLSIT (i, j)

(3.6)

In figure 3.7 an example of obtained gain and offset values for a short integration time of 15
[s] and long integration time of 25 [s] are shown. From the figures it can be seen that at
some places on the sensor the gain and offset have values which are completely out of range
in comparison with the rest of the sensor values. During the measurements these pixels will
also show up, and degrade the obtained image. A bad pixel is defined as a pixel where the
gain value is 25% compared to the mean gain value of the complete sensor [46]. In figure
3.8 it is shown where these pixels are located on the sensor. To correct for this a Bad Pixel
Replacement (BPR) is used.
The BPR uses weighted averaging of the surrounding pixel DLc values to obtain an interpolated value for the bad pixel. The weighting formula which was used is given in (3.7). The
influence of BPR can be seen in figure 3.9. In figure 3.9a the resulting infrared image can be
seen without BPR, while in figure 3.9b the same infrared image is shown with BPR applied.
It can be seen that the use of BPR greatly reduces the defects in the infrared image.
DLc,BP R =
+

MSc Thesis

DLc,i+1,j + DLc,i1,j + DLc,i,j+1 + DLc,i,j1


+
1 + 1 + 1 + 1 + 12 + 12 + 12 + 12

1 DLc,i+1,j+1
2

1 DLc,i+1,j1
2

1+1+1+1+

1 DLc,i1,j+1 + 1 DLc,i1,j1
2
2
1 + 1 + 1 + 1
2
2
2
2

(3.7)

R.F. Janssen

32

Experimental Setup

(a) Gain values

(b) Offset values

Figure 3.7: Infrared sensor gain and offset values obtained from a short integration time of 15
[s] and a long integration time of 25 [s]

Figure 3.8: Bad pixel locations on


the infrared camera sensor, indicated
by white dots

(a) No BPR

(b) With BPR

Figure 3.9: Effect of BPR for an infrared image obtained at an integration time of 20 [s], using
the gain and offset values from figure 3.7

R.F. Janssen

MSc Thesis

Chapter 4
Numerical Results
This chapter will show the results obtained using the Blade Element Momentum Theory and
the Reynolds-Averaged Navier-Stokes simulations. First the results from the BEMT will be
discussed in section 4.1. After the BEMT results, the results from the RANS simulations will
be shown in section 4.2. Finally the results from the BEMT and RANS will be compared in
section 4.3.

4.1

Results for the Blade Element Momentum Theory

In this section the results from the BEMT will be shown. BEMT is a relative simple technique,
only the propeller performance and the forces acting on the propeller blades are given.
The performance of the XPROP propeller as computed with the BEMT is given in table 4.1,
and is visualised in figure 4.1. From the table and the figure it can be seen that for decreasing
Table 4.1: Results obtained using BEMT

J [-]
1.40
1.30
1.20
1.10
1.00
0.90
0.80
0.70
0.65
0.60
MSc Thesis

CT [-]
0.0264
0.0782
0.1277
0.1747
0.2188
0.2599
0.2976
0.3316
0.3466
0.3606

CP [-]
0.0985
0.1483
0.1949
0.2380
0.2771
0.3118
0.3412
0.3644
0.3729
0.3794

[-]
0.3752
0.6855
0.7862
0.8074
0.7896
0.7502
0.6978
0.6370
0.6042
0.5703
R.F. Janssen

Numerical Results

0.85

0.45

0.8

0.4

0.75

0.35

0.7

CT [-], CP [-]

0.5

0.3

0.65

0.25

0.6

0.2

0.55

0.15

0.5

0.1

0.45

0.05

0.4

0
0.6

0.8

1.2

[-]

34

0.35
1.4

J [-]

Figure 4.1: Performance XPROP propeller, BEMT results. Black line indicates CT , red line
indicates CP the vertical axis for both of these line is shown on the left. The blue line indicates
and its axis is shown on the right of the graph

J the values of CT and CP increase as would be expected. The maximum efficiency of the
propeller, at a blade pitch angle 0.7R = 30 [o ], is around an advance ratio of 1.1, with a value
of approximately 0.81. At the lowest simulated advance ratio, J = 0.60, the value of CP is
nearing its maximum, while for the CT value this is not yet the case.
The radial distributions of the lift and drag are shown in figures 4.42 to A.43 in section 4.3
where it is compared to the results from the RANS analysis.

4.2

Results for the Reynolds-Averaged Navier-Stokes Equations

Following the computations of the propeller performance using BEMT, this section will show
the results obtained using RANS simulations. RANS computes the values of aerodynamic
properties in volume cells surrounding the propeller blade. Because of this, more information
is known about the flow surrounding the blade as compared to the BEMT analysis. The
RANS analysis was executed using two different models. The first model, in which the flow
is assumed to be fully turbulent, is discussed in section 4.2.1. The simulation in which the
laminar-turbulent transition is modeled using a correlation based model is treated in section
4.2.2.
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

4.2.1

35

Turbulent Model Results

The fully turbulent simulations are obtained using the Spalart-Allmaras one-equation turbulence model, explained in section 2.3. As this model assumes the flow to be turbulent it is
expected there is a modeling error because of the exclusion of laminar flow.
To see if the solution of the simulation has converged enough, figure 4.2 shows the convergence
history of the six turbulent simulations.

10

Density Residual

10

J = 1.20
J = 1.00
J = 0.80
J = 0.65
J = 0.50
J = 0.30

-1

10-2

10-3

10-4

10-5

10-6
0

5000

10000

15000

20000

25000

Iteration

Figure 4.2: Simulation history of the density residual for the turbulent simulation

The convergence of the solution is checked by looking at the residual of the density, which
should go to zero. After 25000 iterations it can be seen that all residuals are in the range of
1e5 , which is the range in which the solution is accepted as being converged. The simulations
still show some oscillatory behaviour, but are accepted as they stay within the range of 1e5 .
Figure 4.3 shows the Cp distribution on the suction side of the propeller for the investigated
advance ratios. On the rotating propeller the Cp is defined as:
Cp =

p p
1
2

2 + (2nr)2
v

(4.1)

As the advance ratio is decreased it can be seen that the pressure coefficient distribution
changes. At an advance ratio of J = 1.20 it can be seen that the pressure coefficient stays
near zero and is almost constant, as every change in color is a Cp difference of 1. Decreasing
the advance ratio one can see that the leading edge suction peak increases, both in strength
and size. As expected the suction peak is the largest and strongest near the tip of the blade,
where the velocities are highest. To clarify the results a slice is investigated at Rr = 0.7, shown
MSc Thesis

R.F. Janssen

36

Numerical Results

(a) J = 0.30

(b) J = 0.50

(c) J = 0.65

(d) J = 0.80

(e) J = 1.00

(f) J = 1.20

Figure 4.3: Cp distributions suction side propeller blade for the fully turbulent simulations, solid
black lines indicating Rr = 0.7
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations


in figure 4.4. In appendix A.1 additional results are shown for the slices at
0.6, 0.8 and 0.9.

37
r
R

of 0.3, 0.4, 0.5,

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

Figure 4.4: Cp distributions at

r
R

= 0.7 for the turbulent simulations

To investigate the boundary layer in more detail, figures 4.5 to 4.10 show the skin friction
coefficient distributions over the suction side of the blade, where the skin friction coefficient
is given as [4]:
Cf =

2
U 2

(4.2)

Figure 4.5 shows the distributions for J = 1.20. Figure 4.5a shows the distribution of the
absolute value of the skin friction, |Cf |. It can be seen that the highest skin friction is in the
stagnation point region. If looking at the components of the skin friction, given in figures
4.5b to 4.5d, it can be seen that the value of the skin friction coefficient is dominated by the
Z-component. This means that near the surface of the blade most of the flow is directed in
the positive Z-direction, as Cf,z is everywhere positive.
The advance ratio of 1.00 is given in figure 4.6 and this shows the same trend as for J = 1.20.
The skin friction coefficient is dominated by the z-component. In comparison to J = 1.20 it
can be seen that at a lower advance ratio the value of |Cf | increases on the outward part of
the blade. This is due to the increasing flow velocity at increasing radial position along the
blade. While in the hub region of the blade the skin friction coefficient is comparable. This
same trend can be seen for decreasing advance ratio down to J = 0.50. Comparing figures
4.7, 4.8 and 4.9 it can be seen that the absolute skin friction coefficient is almost completely
described by the Z-component.
MSc Thesis

R.F. Janssen

38

Numerical Results

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.5: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 1.20

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.6: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 1.00

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a) |Cf |

(b) Cf,x

(c) Cf,y

39

(d) Cf,z

Figure 4.7: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.80

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.8: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.65

MSc Thesis

R.F. Janssen

40

Numerical Results

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.9: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.50

At J = 0.30, figure 4.10, it can be seen that the distribution of |Cf | shows a different pattern
compared to the higher advance ratio values. Looking at the individual components it can
be seen that near the tip is a big region of reversed flow when looking at the X- and Zcomponents, see figures 4.10b and 4.10d. It is thought this is caused by the tip vortex created
by the preceding blade. To visualise the tip vortex, an isosurface plot of the absolute value of
the vorticity, ||, is created, see figure 4.11. Figure 4.11a shows the tip vortex at J = 0.65.
Here it can be seen that the vortex is not interfering with the blade. At J = 0.30 it can be
seen that the tip vortex is very close to the blade surface, see figure 4.11b. Because the tip
vortex is very close to the blade surface it can be expected that regions of reversed flow are
forming on the blade surface as the ones which can be seen in figure 4.10.
Again a slice is made at Rr = 0.7, and the skin friction coefficient on the suction side of the
blade is compared, shown in figure 4.12. In figure 4.12a all investigated advance ratios can be
seen. J = 0.30 shows a large value of Cf at the leading edge, after which it drops quickly to
zero at an Xc of around 0.11. From here on a region of negative Cf is present. From Xc = 0.375
to about Xc = 0.64 a region of positive skin friction can be seen. After Xc = 0.64 a trailing
edge separation bubble can be seen. The advance ratio of J = 0.50 shows a short trailing edge
separation bubble in the last 3% of the chord. The rest of the investigated advance ratios
do not show the regions of negative Cf as could be seen from the figures 4.5 to 4.9. From
figure 4.12b it can be seen that a region of almost constant Cf can be seen after the first
steep decline, between Xc = 0.2 and 0.3, after which the Cf continues a slow decline toward
the trailing edge.
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a) |Cf |

(b) Cf,x

(c) Cf,y

41

(d) Cf,z

Figure 4.10: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.30

The boundary layer over the blade surface is further investigated by looking at the shape
factor, H, given in equation (1.4). To simplify this analysis only a 2-dimensional slice is
investigated, as it is difficult to investigate this in 3 dimensions. To calculate H the flow field
variables perpendicular to the blade surface are extracted. The boundary layer height, 99 , is
estimated as the position where the parallel velocity is at its maximum. There is a maximum
velocity because there is a positive pressure gradient present in the direction perpendicular
to the blade surface because of the blade curvature. This means the velocity decreases with
increasing distance to the surface. After the boundary layer height is determined the values
for the displacement thickness, , and the momentum thickness, , are calculated using
equations (1.2) and (1.3). Finally the shape factor can be calculated using the calculated
and . The results for Rr = 0.7 for the turbulent simulations can be seen in figure 4.13.
The results are only shown for Xc > 0.1 as in the beginning there is a big influence from the
stagnation point on the calculated results. In figure 4.13a all advance ratios can be seen. For
J = 0.30 it can be seen there is immediately a peak in the shape factor, indicating separation.
After the initial peak, the value of H drops to about 2.1. Following this the value of H
steadily increases again to another peak near the trailing edge of the blade, again indicating
a region of separated flow.
Figure 4.13b shows the shape factor for the other advance ratios. These values all stay all
stay below 2, except for J = 0.50 at the trailing edge showing an increase in H to about 2.15,
which could explain the separation at the trailing edge.
Every iteration the DLR TAU code calculates the forces and moments in all directions. From
MSc Thesis

R.F. Janssen

42

Numerical Results

(a) J = 0.65

(b) J = 0.30
Figure 4.11: Vorticity isosurface plot, || = 1000, of the turbulent simulations at two different
advance ratios with CP overlay

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

43

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

0.5

Cf [-]

-0.5

-1

-1.5

-2

0.2

0.4

0.6

0.8

X/c [-]

(a)
J = 0.65
J = 0.80
J = 1.00
J = 1.20

0.2
0.15
0.1

Cf [-]

0.05
0

-0.05
-0.1
-0.15
-0.2

0.2

0.4

0.6

0.8

X/c [-]

(b)
Figure 4.12: Cf distributions of the blade suction side at Rr = 0.7 for the turbulent simulations,
(a) showing the full range of simulated advance ratios and (b) range without J = 0.30 and 0.50

MSc Thesis

R.F. Janssen

44

Numerical Results

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

5
4.5
4

H [-]

3.5
3
2.5
2
1.5
0

0.2

0.4

0.6

0.8

X/c [-]

(a)
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

2.4

2.2

H [-]

1.8

1.6

1.4

0.2

0.4

0.6

0.8

X/c [-]

(b)
Figure 4.13: Shape factor H for the blade suction side at Rr = 0.7 for the turbulent simulation,
(a) full range of investigated advance ratios and (b) range without J = 0.30

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

45

Table 4.2: Results obtained using the turbulent RANS model

J [-]
1.20
1.00
0.80
0.65
0.50
0.30

CT [-]
0.1115
0.1893
0.2597
0.3084
0.3553
0.3962

CP [-]
0.1713
0.2482
0.3018
0.3294
0.3508
0.4021

[-]
0.7813
0.7627
0.6885
0.6084
0.5063
0.2956

these forces and moments the thrust, T , and torque, Q, can be calculated, which can be used
to determine the performance of the XPROP propeller. The performance results calculated
from the fully turbulent simulations are given in table 4.2 and are shown in figure 4.14.

0.6

0.85
0.8

0.5

0.75
0.7
0.65
0.6

0.3

0.55

[-]

CT [-], CP [-]

0.4

0.5
0.2

0.45
0.4

0.1

0.35
0.3

0
0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.1

0.25
1.2

J [-]

Figure 4.14: Performance XPROP propeller, turbulent RANS results. Black line indicates CT ,
red line indicates CP the vertical axis for both of these line is shown on the left. The blue line
indicates and its vertical axis is shown on the right of the graph

From the results of the fully turbulent performance values, given in table 4.2 and figure 4.14,
it can be seen that for decreasing J the values of CT and CP are increasing. It is expected
that at low advance ratios, the values of CT and CP would start decreasing. However, in
the case of CP the opposite can be seen, at low advance ratios the value CP increases. It is
thought this is due to the tip vortex of the preceding blade passing close to the blade surface.
Because of this regions of reversed flow are forming on the blade surface, which increases the
drag of the propeller and thus increasing the power needed. The efficiency of the propeller
increases for increasing J, where it can be seen that at J = 1.20, is almost at its maximum.
MSc Thesis

R.F. Janssen

46

4.2.2

Numerical Results

Transition Model Results

Using the transition model developed by Langtry and Menter [34], the flow field around the
XPROP propeller is calculated. The model is explained in more detail in section 2.3. Due to
the inclusion of laminar-turbulent transition it is hoped that this model is able to predict the
performance of the XPROP propeller more accurately compared to the simulations without
laminar-turbulent transition.
The history of the density residual is checked to see if all simulations are converged, these
can be seen in figure 4.15. As is the case with the turbulent simulations 25000 iterations are
taken to see if the solution is converged. Two simulations show density residuals which are in
the range of 1e3 . The J = 0.30 simulation does not reach density residual values far below
1e3 whereas the simulation for J = 0.50 appears to converge for the first 16000 iterations
after which it quickly jumps to a density residual of approximately 2e3 . From this it can
be concluded that the results from the J = 0.30 and J = 0.50 cases are not to be trusted.
Because no intermediate solutions are available, it is not known what caused the solutions to
not converge.

10

J = 1.20
J = 1.00
J = 0.80
J = 0.65
J = 0.50
J = 0.30

Density Residual

10-1

10-2

10

-3

10-4

10

-5

5000

10000

15000

20000

25000

Iteration

Figure 4.15: Simulation history of the density residual for the transition simulation

In figures 4.16a to 4.16f the Cp distributions on the suction side of the propeller blade are
shown for all simulated advance ratios, including the two cases which were not fully converged.
The simulation for the case of J = 1.20, given in figure 4.16f, shows only small differences
in pressure coefficient compared to the other simulations. From an advance ratio of J =
1.00 strange structures start occurring on the blade surface which increase in strength with
decreasing advance ratio, until no clear structures can be distinguished at an advance ratio of
J = 0.50. Figure 4.17 shows the Cp values on the blade surface at Rr = 0.7. For J = 0.3 three
bumps can be seen in the Cp distribution on the suction side. One can see that for J = 0.65
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a) J = 0.30

(b) J = 0.50

(c) J = 0.65

(d) J = 0.80

(e) J = 1.00

(f) J = 1.20

47

Figure 4.16: Cp distributions suction side propeller blade for the transition simulations

MSc Thesis

R.F. Janssen

48

Numerical Results

two peaks can be seen on the suction side, at Xc = 0.3 and Xc = 0.475. At higher advance
ratios these peaks are no longer clearly visible. In appendix A.1 additional results are shown
for the slices at Rr of 0.3, 0.4, 0.5, 0.6, 0.8 and 0.9.
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

Figure 4.17: Cp distributions at

r
R

= 0.7 for the transition simulations

It is thought that the peaks occuring in the Cp distribution on the blade surface are caused
by the strange structures starting to form at J = 1.00. To see what causes these strange
structures, first the Cf distribution on the propeller surface is investigated. Figure 4.18a
shows the value of Cf on the suction side of the propeller, where a solid line at Rr = 0.5 is
shown as a slice is extracted from the domain at this location which is used to investigate the
flowfield surrounding the propeller blade.
Figure 4.18b focusses more at the region surrounding Rr = 0.5, here the strange structures
can be seen more clearly. In figure 4.18b the skin friction direction is visualised, this is done
to see the direction of the flow just above the surface. At the leading edge of the blade it can
be seen that the flow has only a small component in the Y -direction. After a short distance
it can be seen that the flow is directed more and more in positive Y -direction, this means
that both the X- and Z-component of the skin friction are becoming smaller. At around 40%
chord it can be seen that the flow is directed completely in the positive Y -direction, meaning
that both the X- and Z-component approach zero, indicating the start of flow separation.
After a while the strange structures are forming on the blade surface. Looking at the flow
topology of one of the strange structures it looks as though some sort is vortex is formed. To
investigate this in more detail, a slice is extracted at Rr = 0.5.
Figure 4.19 shows the Cp values in the neighbourhood of the propeller blade together with
some flow streamlines, only the X- and Z-components of the velocity are used. At the aft part
of the blade it can be seen that a second low pressure area is forming, see figures 4.19a 4.19b.
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a)

49

(b)

Figure 4.18: Skin friction coefficient, Cf , distribution on the suction side of the propeller blade
at J = 1.00, (a) full propeller blade, solid line indicating Rr = 0.5 and (b) focus around Rr = 0.5
where both Cf and the skin friction direction is visualised

MSc Thesis

R.F. Janssen

50

Numerical Results

(a)

(b)
Figure 4.19: Cp distribution, J = 1.00, in the flowfield surrounding the propeller blade including
streamtraces at Rr = 0.5, (a) full blade section and (b) aft part of the blade section

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

51

Looking at the streamlines at this location it can be seen a rather large separation bubble is
forming. The flow above the separation bubble is accelarated and thus creating the second
low pressure area. In figure 4.19b it can be seen that a smaller separation bubble already was
present upstream of the large separation bubble, which is originating from another vortex
structure, see figure 4.19b.
To see where the vortices are originating from, the value of the intermittence, , is investigated,
see figure 4.20. The value of varies between 0 and 1, where 0 means laminar flow and 1 is
turbulent flow. Looking at figure 4.20a it can be seen that in the farfield the value of is 1
and near the propeller blade it decreases to 0. When zooming in on the trailing edge of the
propeller blade, see figure 4.20b, it can be seen that the value of decreases to 0 at the wall,
even at the location where turbulent flow is expected. From the description of the transition
model [34] it is known that the boundary condition for at a wall is a zero normal flux, so
a Neumann boundary condition. But when looking at figure 4.20b it looks like a Dirichlet
boundary condition is implemented in the TAU code. It is thought that because of this the
vortex like structures are forming over the propeller blade surface at lower advance ratios.
The skin friction coefficient is investigated on the suction side of the blade, as shown in figures
4.21 to 4.25. Figure 4.21a shows |Cf | for J = 1.20. Here a region of low skin friction can be
seen at around the halfway chord. Looking at the single components of Cf , it can be seen that
after the low skin friction region, a region of negative Cf,x and Cf,z is shown. Also it can be
seen that Cf,y has a big influence on the absolute skin friction value, indication flow directed
toward the blade tip. Combining this leads to the conclusion that at J = 1.20 trailing edge
separation occurs.
Decreasing the advance ratio to 1.00, see figure 4.22, one can also see the vortical structures
forming as could be seen in the Cp distributions. Looking at the three components, one can
see a separation line, indicated by the purple regions at 31 to 12 chord in Cf,x and Cf,z . The
structures consist of both positive and negative values of Cf,x and Cf,z , indicating a vortex
like structure, while having a mostly positive value of Cf,y , which means that the flow again
is directed toward the blade tip.
At J = 0.80, the strength of the vortex like structures is increasing as can be seen by the
increase in the value of the skin friction. Looking at the Cf,x and Cf,z components, it looks
like the vortex like structures start to form a strong separation line, right behind a separation
line which is weaker, indicated by the blue/purple colour. This is an indication that the flow
separates, has a short reattachment and then separates again, creating a bigger separation
bubble. The Cf,y component shows a strong flow direction toward the tip of the blade after
the flow starts to separate, see figures 4.23c and 4.23d. Another thing of interest is the shape
of the separation line. Where at J = 1.20 and J = 1.00 the separation line is relatively
straight, starting from J = 0.80 it can be seen that at the outward portion of the blade the
separation is postponed until approximately half chord, while in the middle part of the blade
separation starts to occur near the leading edge.
Figure 4.24 shows the breakdown of the skin friction coefficient in its three components for
J = 0.65. Comparing J = 0.80 to J = 0.65, one can see the same structure occurring. But
the differences being a stronger separation, the region where leading edge separation occurs
MSc Thesis

R.F. Janssen

52

Numerical Results

(a)

(b)
Figure 4.20: Distribution of the intermittency, , in the surrounding flowfield, at J = 1.00, of
the propeller blade at Rr = 0.5, (a) full blade section and (b) focus on the trailing edge part of
the blade section

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a) |Cf |

(b) Cf,x

(c) Cf,y

53

(d) Cf,z

Figure 4.21: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 1.20

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.22: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 1.00

MSc Thesis

R.F. Janssen

54

Numerical Results

(a) |Cf |

(b) Cf,x

(c) Cf,y

(d) Cf,z

Figure 4.23: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.80

moves upward to the tip and the flow structure at the tip is changing.
Looking at the Cf distributions for J = 0.50 and J = 0.30, shown in figures 4.25a and 4.25b,
it is difficult to make out what happens. For J = 0.50 it looks as if the separation line has
bursted, giving rise to the failure in convergence.
At J = 0.30 it looks as if the same problem is occurring as with the turbulent RANS simulation, the interaction of the tip vortex of a preceding blade with the flow over the surface,
this can be seen in more detail in figure 4.26. Figure 4.26a shows an isosurface plot of the
absolute vorticity value at J = 0.65, while figure 4.26b shows the isosurface plot at J = 0.30.
Comparing the two figures it can be seen that, at the low advance ratio, the tip vortex is very
near to the surface of the blade and it is possible this has an effect on the flow, as was the
case with the simulations using the Spalart-Allmaras model from section 4.2.1.
When a slice is analysed at Rr = 0.7, see figure 4.27, one can see that at J = 0.30 and J = 0.50
Cf values are out of range compared to the other simulations. In figure 4.27b J = 0.30 and
J = 0.50 are removed. In this figure it can be seen that at Rr = 0.7 the flow for J = 0.65
separates early, at Xc = 0.18, but it reattaches again at Xc = 0.55, after which the flow stays
attached up to Xc = 0.96. The flow at J = 0.80 separates at Xc = 0.46 and reattaches again
at Xc = 0.92. Where the flows for J = 0.65 and J = 0.80 reattach, at J = 1.00 and J = 1.20
no reattachment of the flow occurs. J = 1.00 flow separates at Xc = 0.575, and for J = 1.20
flow separates at Xc = 0.66. It can be seen from this, decreasing the advance ratio, decreases
the separation position. But early separation leads to reattachment of the flow, as is the case
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

(a) |Cf |

(b) Cf,x

(c) Cf,y

55

(d) Cf,z

Figure 4.24: Absolute value of the skin friction coefficient and a breakdown of the skin friction
coefficient in the three axial directions at J = 0.65

(a) J = 0.50

(b) J = 0.30

Figure 4.25: Absolute value of the skin friction coefficient |Cf |

MSc Thesis

R.F. Janssen

56

Numerical Results

(a) J = 0.65

(b) J = 0.30
Figure 4.26: Vorticity isosurface plot, || = 1000, for the transition simulations with CP overlay

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

57

for J = 0.65 and J = 0.80.


As was done in the turbulent case, a 2-dimensional boundary layer analysis is done to investigate the boundary layer. In figure 4.28 the shape factor is shown for a slice at Rr = 0.7. In
figure 4.28a the shape factor for all simulations, except J = 0.50, can be seen. H for J = 0.30
shows a couple of peaks in the graph corresponding to the bumps in the Cf graph, figure
4.27a, and the final peak shows that trailing edge separation occurs. At J = 0.65 two peaks
can be seen on the first half of the chord. After these two peaks, the value of H decreases,
which coincides with the reattachment of the flow. After reattachment the shape factor stays
in the range of 3 to 4, this would not explain the sudden separation at the trailing edge
and because of that it is thought that at the trailing edge a 3-dimensional phenomenon is
responsible for the separation of the flow. For the cases of J = 0.80, J = 1.00 and J = 1.20 an
increase of H can be seen for increasing chord coordinate leading to flow separation. Where
for J = 0.80 the value of the shape factor decreases by a significant amount near the trailing
edge, which shows that the flow is attached at the trailing edge. This is not the case for
J = 1.00 and J = 1.20. Figure 4.28b shows H for the J = 0.50, here it can be seen that the
values for the shape factor are not physical, reaching highly negative values.
The performance of the XPROP propeller can be calculated using the forces calculated by
the DLR TAU code. These calculated performance values for the transition simulations are
given in table 4.3 and shown in figure 4.29.
Table 4.3: Results obtained using the transition RANS

J [-]
1.20
1.00
0.80
0.65
0.50
0.30

CT [-]
0.0812
0.1747
0.2564
0.3055
0.3412
0.3667

CP [-]
0.1323
0.2292
0.2978
0.3255
0.3471
0.3845

[-]
0.7370
0.7623
0.6889
0.6100
0.4915
0.2861

The performance results of the RANS simulation including transition show that for decreasing
advance ratio, the values of both CT and CP increase, while it is expected that at low advance
ratios these values would decrease. As is the case in the turbulent RANS simulation, the value
of CP starts to increase at very low advance ratios. Again, it is thought this is due to the
increase in drag which is due to the tip vortex of the preceding blade. has a maximum at
J = 1.04, which means the efficiency of the propeller starts to decrease at too high advance
ratios.

4.2.3

Comparison Results Turbulent model and Transition model

As is expected, the results between the models with and without laminar-turbulent transition
show differences. By comparing the two models these differences are illustrated.
First the results of Cp are compared, this is shown in figure 4.30. It should be noticed that
MSc Thesis

R.F. Janssen

58

Numerical Results

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

0.5

Cf [-]

-0.5

-1

-1.5

-2

0.2

0.4

0.6

0.8

X/c [-]

(a)
J = 0.65
J = 0.80
J = 1.00
J = 1.20

0.2
0.15
0.1

Cf [-]

0.05
0

-0.05
-0.1
-0.15
-0.2

0.2

0.4

0.6

0.8

X/c [-]

(b)
Figure 4.27: Cf distributions of the blade suction side at Rr = 0.7 for the transition simulations,
(a) showing the full range of advance ratios and (b) range without J = 0.30 and 0.50

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

59

J = 0.30
J = 0.65
J = 0.80
J = 1.00
J = 1.20

18
16
14

H [-]

12
10
8
6
4
2
0

0.2

0.4

0.6

0.8

X/c [-]

(a)
J = 0.50
1500

1000

H [-]

500

-500

-1000

-1500

0.2

0.4

0.6

0.8

X/c [-]

(b)
Figure 4.28: Shape factor H for the blade suction side at Rr = 0.7 for the turbulent simulation,
(a) investigated advance ratio range without J = 0.50 and (b) advance ratio J = 0.50

MSc Thesis

R.F. Janssen

60

Numerical Results

0.6

0.85
0.8

0.5

0.75
0.7
0.65
0.6

0.3

0.55

[-]

CT [-], CP [-]

0.4

0.5
0.2

0.45
0.4

0.1

0.35
0.3

0
0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.1

0.25
1.2

J [-]

Figure 4.29: Performance XPROP propeller, transition RANS results. Black line indicates CT ,
red line indicates CP the vertical axis for both of these line is shown on the left. The blue line
indicates and its vertical axis is shown on the right of the graph

the Cp distribution on the pressure side of the blade is in good agreement for almost all
investigated advance ratios, even J = 0.30 and J = 0.50. For J = 1.20 a separation bubble
can be seen on the pressure side of the blade, see figure 4.30. Even though the simulation
of J = 0.30 in the case of transition was not converged, it can be seen that the pressure
coefficient for both cases are in the same order of magnitude. For J = 0.50 a difference can
be seen at the leading edge, where in the case of transition the pressure coefficient is lower
compared to the case of the turbulent simulation. Near the leading edge the opposite occurs,
the value of the transition simulation is higher compared to the case of fully turbulent flow.
Looking at figure 4.30 one sees that for a J = 0.65 the values between the turbulent and transition simulation are in the same order of magnitude, where the transition case is fluctuating
around the value of the turbulent case. The advance ratio of 0.80 shows good agreement on
the first 60% of the chord, where in the case of transition a plateau in the pressure distribution
is occurring, a sign of separation. J = 1.00 shows almost the same as J = 0.80, except the
plateau occurs later at Xc = 0.75. When J = 1.20 it can be seen a small plateau is occuring
at the trailing edge of the blade in the case of transition, whereas this is not the case when
the flow is simulated as fully turbulent.
In figures 4.31 to 4.36 the Cp values near the blade are shown at Rr = 0.7. From this it can
be seen that in some cases the Cp values are similar, J = 0.80, J = 0.65 and J = 0.30.
When J = 1.20, see figure 4.31, a second negative pressure gradient region can be seen on
the pressure side of the propeller blade. As is the case at lower advance ratios it seems that
a separation bubble is forming, but now on the pressure side of the blade. This effect can
also be seen in the Cp distribution on the blade surface shown in figure 4.27. On the suction
R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

61

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-3
-2.5
-2

Cp [-]

-1.5
-1
-0.5
0
0.5
1

0.2

0.4

0.6

0.8

X/c [-]

Figure 4.30: Cp distributions at Rr = 0.7 for both the turbulent and transition simulations, solid
turbulent and dashed transition data

side it can also be seen that the region of low pressure has significantly decreased in the case
of transition, this is due to the separation at the trailing edge. At J = 1.00, see figure 4.32,
it can be seen that at the trailing edge small differences are present. On the pressure side of
the blade the value of Cp is lower compared to the turbulent simulation, it is thought that
this is due to the flow separation occuring on the suction side. J = 0.50 shows artifacts
at the trailing edge of the suction side of the blade. This was to be expected, because the
solution has not converged and it is even possible that the failure to converge is caused by
these artifacts.
In the pressure coefficient distributions there are no big differences. When looking at Cf it
can be seen that these differences are increased. This is expected as laminar and turbulent
flows have different skin friction characteristics. Figure 4.37 shows the difference in Cf for
the turbulent and transition case. In all cases it can be seen that the transition simulations
have a lower value of Cf . This can be explained by the fact that at the leading edge of the
blade the flow is laminar and laminar flow generates less skin friction. For J = 1.20 it can be
seen that the skin friction up to Xc = 0.66 is always smaller for the transition case compared
with the turbulent case. After this the skin friction of the transition case becomes negative
and this can influence the characteristics of the blade in a negative way. J = 1.00 shows
the same behaviour as in the case of J = 1.20, only the separation in the case of transition
occurs earlier, at Xc = 0.575. When the advance ratio is 0.80, the Cf at the beginning of the
blade is almost half of the Cf in case of the fully turbulent flow. At Xc = 0.46 a separation
bubble is occurring, which reattaches at the trailing edge of the blade. In the transition case,
when J = 0.65, a separation bubble is present between Xc = 0.34 and Xc = 0.55, which is not
MSc Thesis

R.F. Janssen

62

Numerical Results

(a) Turbulent

(b) Transition

Figure 4.31: Cp comparison at


J = 1.20

(a) Turbulent

r
R

= 0.7,

(b) Transition

Figure 4.35: Cp comparison at


J = 0.50

R.F. Janssen

= 0.7,

(b) Transition

Figure 4.33: Cp comparison at


J = 0.80

(a) Turbulent

r
R

r
R

= 0.7,

(a) Turbulent

(b) Transition

Figure 4.32: Cp comparison at


J = 1.00

(a) Turbulent

= 0.7,

(b) Transition

Figure 4.34: Cp comparison at


J = 0.65

(a) Turbulent

r
R

r
R

= 0.7,

(b) Transition

Figure 4.36: Cp comparison at


J = 0.30

r
R

= 0.7,

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

63

present in the turbulent simulation. After the reattachment, the value of Cf for the case of
transition stays constant, up to a point near the trailing edge where trailing edge separation
occurs. For J = 0.50 there is a big difference between the turbulent and transition case, where
the transition case show large regions of separated flow. Both the transition and turbulent
simulation at J = 0.30 show separated flow behaviour, they show very similar behaviour,
which could be explained by the tip vortex of the preceding blade influencing the flow on the
blade surface.
Comparing the shape factor of the turbulent and transition cases should give different results,
as almost no flow has separation in the fully turbulent flow cases. In figure 4.38 the comparison
between the shape factors at different advance ratios is shown. The value of J = 0.50 is left
out, as the calculated shape factor in the transition case was unphysical. For J = 1.20 and
J = 1.00 it can be seen that the turbulent flow has an H between 1.5 and 2. While in the
case of transition, it can be seen that the value of H is increasing, leading to a peak where
separation occurs beforehand. In both cases separation occurs when the value of H passes
3.5. When J = 0.80, the turbulent case shape factor is again between 1.5 and 2, while in the
transition case separation occurs around H = 3.5. J = 0.65 shows H = 1.5 for almost all the
chord length for the turbulent case, only increasing to 2 in the last 20% of the chord. In the
transition case, separation now occurs at a value around H = 3.25. In the case of J = 0.30
the turbulent flow is attached when H < 0.225. For the flow with transition no single value
of H gives a clear separated flow.
The difference in the calculated performance between the turbulent RANS simulation and the
RANS simulation including transition is shown in tables 4.4 to 4.6. Here the values for both
the turbulent and transition RANS simulations is shown, as well as the difference between the
performance values. The values of CT , CP and are shown in figures 4.39 to 4.41 in section
4.3.
Table 4.4 shows the values of CT for the turbulent and transition RANS simulations. It can
be seen that at high advance ratios, the difference between the two models is significant. At
J = 1.20 the obtained CT is 27% lower when the transition model is used. This is probable
due to the laminar separation bubble forming on the aft part of the blade. CT decreases
with decreasing advance ratio, up to J = 0.65, which is not suprising as more of the flow over
the propeller blade becomes turbulent. After J = 0.65 CT increases again for J = 0.50 and
J = 0.30. The results for these low advance ratios is not reliable, as the simulations did not
converge at these values.
CP is shown in table 4.5 for the turbulent and transition RANS simulations. At J = 1.20,
again a big difference can be seen between the two RANS models, a difference of almost 23%.
For decreasing advance ratio, CP is also decreasing, until J = 0.50. The decrease in CP
for lower advance ratio can be explained by the fact that at lower advance ratios more flow
is turbulent.
For the propeller efficiency a similar trend can be seen as with CT and CP , but is
much smaller compared to the other two. At J = 1.20 the biggest difference in propeller
efficiency can be seen, where is almost 6% lower for the propeller simulated using the
transition model compared to the turbulent RANS model. It can be seen that increases
MSc Thesis

R.F. Janssen

64

Numerical Results

Turbulent
Transition
0.1

Turbulent
Transition
0.15

0.1
0.05

Cf [-]

Cf [-]

0.05

-0.05
-0.05
-0.1

-0.1

0.2

0.4

0.6

0.8

-0.15

0.2

0.4

0.6

X/c [-]

X/c [-]

(a) J = 1.20

(b) J = 1.00

0.8

Turbulent
Transition

Turbulent
Transition

0.15

0.2
0.15

0.1
0.1
0.05

Cf [-]

Cf [-]

0.05
0

-0.05
-0.05
-0.1
-0.1
-0.15
-0.15

0.2

0.4

0.6

0.8

-0.2

0.2

0.4

X/c [-]

0.6

0.8

X/c [-]

(c) J = 0.80

(d) J = 0.65
Turbulent
Transition

Turbulent
Transition
1
0.8

0.2

0.6
0.4

Cf [-]

Cf [-]

0.2
0

0
-0.2
-0.4
-0.6

-0.2

-0.8
0

0.2

0.4

0.6

0.8

-1

0.2

0.4

0.6

X/c [-]

X/c [-]

(e) J = 0.50

(f) J = 0.30

0.8

Figure 4.37: Comparison between the turbulent and transition simulations of the skin friction
coefficients on the suction side of the blade at Rr = 0.7 for the different advance ratios

R.F. Janssen

MSc Thesis

4.2 Results for the Reynolds-Averaged Navier-Stokes Equations

65

Turbulent
Transition

Turbulent
Transition
10

14
12

10

H [-]

H [-]

6
8
6

4
2
2
0

0.2

0.4

0.6

0.8

0.2

0.4

0.6

X/c [-]

X/c [-]

(a) J = 1.20

(b) J = 1.00

0.8

Turbulent
Transition

Turbulent
Transition

10

12

10

H [-]

H [-]

6
6

4
4
2

0.2

0.4

0.6

0.8

0.2

0.4

X/c [-]

0.6

0.8

X/c [-]

(c) J = 0.80

(d) J = 0.65
Turbulent
Transition

14

12

10

H [-]

0.2

0.4

0.6

0.8

X/c [-]

(e) J = 0.30
Figure 4.38: Comparison between the turbulent and transition simulations of the shape factor
on the suction side of the blade at Rr = 0.7 for the different advance ratios (J = 0.50 is excluded)

MSc Thesis

R.F. Janssen

66

Numerical Results
Table 4.4:
Comparison of CT
between the turbulent and transition
RANS models

CTT urb [-]


0.1115
0.1893
0.2597
0.3084
0.3553
0.3962

J [-]
1.20
1.00
0.80
0.65
0.50
0.30

CTT rans [-]


0.0812
0.1747
0.2564
0.3055
0.3412
0.3667

CT [%]
-27.17
-7.723
-1.271
-0.940
-3.968
-7.446

Table 4.5:
Comparison of CP
between the turbulent and transition
RANS models

J [-]
1.20
1.00
0.80
0.65
0.50
0.30

CPT urb [-]


0.1713
0.2482
0.3018
0.3294
0.3508
0.4021

CPT rans [-]


0.1323
0.2292
0.2978
0.3255
0.3471
0.3845

CP [%]
-22.77
-7.655
-1.325
-1.184
-1.055
-4.377

Table 4.6: Comparison of between


the turbulent and transition RANS
models

J [-]
1.20
1.00
0.80
0.65
0.50
0.30

T urb [-]
0.7813
0.7627
0.6885
0.6084
0.5063
0.2956

T rans [-]
0.7370
0.7623
0.6889
0.6100
0.4915
0.2861

[%]
-5.670
-0.052
0.058
0.263
-2.923
-3.214

for decreasing advance ratio. At J = 0.65 and J = 0.80 it can be seen that is higher for the
transition model compared to the turbulent model, meaning more thrust is generated for the
same amount of power delivered. At J = 0.50 and J = 0.30 it can be seen that decreases
rapidly.

4.3

Comparison BEMT and both RANS models

In this section the obtained results of the BEMT and RANS simulations will be compared. As
BEMT only shows performance results and blade loading characteristics, these are the only
ones that will be showed in this section. First the obtained performance results are shown,
after this the results of the blade loading are shown.
Figures 4.39, 4.40 and 4.41 show the results obtained from tables 4.1, 4.2, 4.3 and the results
from the National Aerospace Laboratory of the Netherlands (NLR) shoptest are added [47].
The NLR shoptest is a test campaign to characterise the performance of the XPROP propeller.

Figure 4.39 shows a comparison of the CT obtained from the three different computational
methods and the NLR shoptest results. CT between BEMT and the turbulent RANS
simulation at J = 1.20 is 14.5%, this value decreases to 12.4% at J = 0.65. The difference
between BEMT and the RANS simulations with transition is larger, 57.3% difference at
R.F. Janssen

MSc Thesis

4.3 Comparison BEMT and both RANS models

BEMT
RANS, Fully Turbulent
RANS, Transition
NLR shoptest

0.4

0.35
0.3

0.25

0.25

CP [-]

0.3

CT [-]

BEMT
RANS, Fully Turbulent
RANS, Transition
NLR Shoptest

0.4

0.35

0.2

0.2

0.15

0.15

0.1

0.1

0.05

0.05

0
0.3

67

0.4

0.5

0.6

0.7

0.8

0.9

1.1

1.2

1.3

0
0.3

1.4

0.4

0.5

0.6

0.7

0.8

0.9

1.1

1.2

1.3

1.4

J [-]

J [-]

Figure 4.39: Comparison of CT for


BEMT, RANS turbulent, RANS transition and NLR shoptest

Figure 4.40: Comparison of CP for


BEMT, RANS turbulent, RANS transition and NLR shoptest
BEMT
RANS, Fully Turbulent
RANS, Transition
NLR Shoptest

1
0.9
0.8

[-]

0.7
0.6
0.5
0.4
0.3
0.2
0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.1

1.2

1.3

1.4

J [-]

Figure 4.41: Comparison of for


BEMT, RANS turbulent, RANS transition and NLR shoptest

MSc Thesis

R.F. Janssen

68

Numerical Results

J = 1.20 and 13.5% at J = 0.65. This larger difference for the RANS with transition could
be because of the separation that occurs on the aft part of the blade. Comparing the three
computational methods to the shoptest, it can be seen that at high advance ratios BEMT is
closest to the wind tunnel results. This could mean that the effect of trailing edge separation
at high advance ratios is not as significant on CT as the RANS simulations show. Below
J = 0.75 it can be seen that the results from the RANS simulations come close to the
values of the wind tunnel data. The big difference between the computations and the wind
tunnel data is the gradient of the curves, where the wind tunnel data has a lower gradient in
comparison to the computations in the investigated range of advance ratios.
In figure 4.40 the value of CP is shown for the three computational methods and the wind
tunnel data from NLR. When comparing BEMT to the fully turbulent simulation, it can be
seen that CP ranges from 13.8% at J = 1.20 to 13.2% at J = 0.65. CP between BEMT
and the RANS simulations including transition range from 47.3% at J = 1.20, to 14.6% at
J = 0.65. From this it can be seen that the relative difference between BEMT and turbulent
RANS stays constant, while the difference between BEMT and RANS including transition
shows a decrease in the relative difference between the CP values. If the graphs are compared,
it can be seen that the absolute difference between BEMT and the RANS including transition
have an almost constant offset, with the RANS results having a lower value of CP . Comparing
the simulations with the wind tunnel data, it can be seen that BEMT has the best agreement
of the computational techniques.
The differences between the propeller efficiency is shown in figure 4.41 for the three computational methods and the wind tunnel data. At high advance ratio it can be seen that there is a
difference between BEMT and the turbulent RANS, as BEMT has its maximum efficiency at
J = 1.10, while the turbulent RANS is still increasing a little at J = 1.20. At J = 0.65 there
can be seen a good agreement between the turbulent RANS and BEMT, = 0.70%. The
RANS model including the transition show a maximum efficiency around J = 1.05, which
is at a slightly lower advance ratio compared to BEMT. At high advance ratios is large
between BEMT and the transition RANS model, = 6.7% at J = 1.20. At J = 1.20 it
can be seen that for the wind tunnel data is higher in comparison to the computational
models. At low advance ratios, J < 1.00, it can be seen that is smaller compared to the
computational models.
From the above analysis it can be seen there are big differences between the performance
data obtained through BEMT, turbulent RANS, RANS model including transition and the
wind tunnel data of the NLR shoptest. The results from BEMT and RANS can be directly
compared, as the freestream parameters were the same. The problem with the wind tunnel
data is that it is not known what the freestream conditions are. It is possible that the tests
were performed at a different Reynolds number, which could lead to different results as it could
change the boundary layer properties on the blade, such as the location of laminar-turbulent
transition.
The load on the blade, given as cl and cd distributions, can be compared between BEMT
and RANS. These comparisons are given in figures 4.42 to A.43 for BEMT, fully turbulent
RANS and RANS using the transition model. While figures A.44 and A.45 only show the
comparison between the two RANS methods as the BEMT model could not compute these
R.F. Janssen

MSc Thesis

4.3 Comparison BEMT and both RANS models

69

low advance ratios.


0.5
0.4
0.3

cl [-]

0.2
0.1
0
0.1
RANS Transition
RANS Fully Turbulent
BEMT

0.2
0.3
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.07
RANS Transition
RANS Fully Turbulent
BEMT

0.06

cd [-]

0.05

0.04

0.03

0.02

0.01
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure 4.42: Lift and drag comparison BEMT, RANS turbulent and RANS transition for J =
1.20, (a) cl and (b) cd

In figure 4.42 the differences between the BEMT and the two RANS models is shown for
J = 1.20. In figure 4.42a it can be seen that the local lift coefficient is higher for BEMT
in comparison to the turbulent RANS and RANS with transition modeling. One possible
explanation for this is that BEMT does not take into account three-dimensional effects. It
is thought that due to the three-dimensional effect, the addition of flow in Y -direction, the
direction of the force could change, thus lowering the lift coefficient. When the transition
is taken into account in the RANS simulation, it can be seen that the local lift coefficients
decrease further. This can be caused by the separation bubble occurring on the aft part of
the propeller blade.
The local drag coefficient, cd , is shown in figure 4.42b. The drag predicted by BEMT is
MSc Thesis

R.F. Janssen

70

Numerical Results

significantly lower compared to the RANS data. As BEMT does not take three-dimensional
effects into account, it is possible that a lot of drag is added due to a possible rotation of the
force vector due to the flow in Y -direction in the RANS simulations. It can also be seen that
the drag coefficient is lower for the RANS simulations using the transition model compared
to the turbulent RANS simulations. This can be explained by the fact that a large portion
of the flow over the blade is laminar.
0.8
0.7
0.6

cl [-]

0.5
0.4
0.3
0.2
RANS Transition
RANS Fully Turbulent
BEMT

0.1
0
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.09
RANS Transition
RANS Fully Turbulent
BEMT

0.08
0.07

cd [-]

0.06
0.05
0.04
0.03
0.02
0.01
0
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure 4.43: Lift and drag comparison BEMT, RANS turbulent and RANS transition for J =
1.00, (a) cl and (b) cd

Figure 4.43 shows the blade loading at J = 1.00. Looking at the lift distribution, figure 4.43a,
it can be seen that the values for cl are the largest for BEMT. Again this can be explained by
the effect of BEMT using only two-dimensional data. Comparing the RANS models with one
another, it can be seen that for Rr < 0.6 the values of cl are very similar. The cl -distribution
of the RANS simulations using transition shows a more wavy structure, this is due to the
formation of vortices on the aft part of the blade. It is thought this is because a larger part
R.F. Janssen

MSc Thesis

4.3 Comparison BEMT and both RANS models

71

of the flow over this part of the blade is turbulent, see figure 4.22. When Rr > 0.6 it can be
seen that still a large part of the blade chord is laminar and separates on the aft part of the
blade which could lead to the loss of lift on the outer part of the propeller blade.
Comparing cd , see figure 4.43b, it can be seen that BEMT has a significantly lower cd compared
to the RANS simulations. Again it is thought this is due to the simplification of a twodimensional flow in BEMT. The two RANS models show similar behaviour, except the values
of cd on the RANS model incorporating laminar-turbulent transition are smaller. It is thought
this is due to the flow being laminar for a large part of the blade chord.
1.2
1.1
1

cl [-]

0.9
0.8
0.7
0.6
0.5
0.4

RANS Transition
RANS Fully Turbulent
BEMT

0.3
0.2
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.14
RANS Transition
RANS Fully Turbulent
BEMT

0.12

cd [-]

0.1
0.08
0.06
0.04
0.02
0
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure 4.44: Lift and drag comparison BEMT, RANS turbulent and RANS transition for J =
0.80, (a) cl and (b) cd

At J = 0.80, see figure 4.44, it can be seen that BEMT exhibits the highest lift characteristics,
as in the earlier cases. It is more interesting to look at the RANS simulations, as from
J = 0.80 it can be seen that the distribution of lift is equal for the turbulent and transition
MSc Thesis

R.F. Janssen

72

Numerical Results

RANS model. This can be explained by the fact that now a large part of the flow over the
blade is turbulent, see figure 4.23.
The cd distribution, figure 4.44b, shows a similar trend compared to J = 1.00. The cd
distribution obtained using BEMT is again lower, but the differences between BEMT and the
RANS models have become smaller on a part of the blade. It can be seen that in the inner
part of the blade, the cd values are higher for the case of J = 0.80, while around Rr = 0.7 it
can be seen that the values of cd are lower. It is thought this is due to the centrifugal forces,
thinning the boundary layer compared to the two-dimensional case, making the boundary
layer more resistant to separation. From figure 4.23 it can be seen that on the outer part of
the blade the flow is still for a large part of the chord is still laminar. The final thing to note
is that leading to the tip it can be seen that cd increases to high values. It is thought this is
happening because of the tip vortex present in the RANS simulations.
The figures for J = 0.65 are not shown here, because the results are very similar to J = 0.80.
For the interested reader, these graphs can be found in the appendix in section A.3. The
results for J = 0.50 and J = 0.30 are not shown here, because BEMT could not be resolved
at these advance ratios and the RANS model which included laminar-turbulent transition did
not converge at these operating points. For completeness they are supplied in the appendix
in section A.3.

R.F. Janssen

MSc Thesis

Chapter 5
Experimental Results
This chapter will give the results of the qualitative infrared measurements of the XPROP
propeller. First a short investigation will be done to the influence of the integration time on
the displayed results in section 5.1. After that, the infrared results will be shown in section
5.2 for experiments at different advance ratios. And finally the influence of Reynolds number
will be investigated in section 5.3.

5.1

Integration Time Influence

First the integration time will be examined, as it is important that the information captured
by the infrared sensor shows no distorted image. For this the results from three different
tests will be compared at 3 integration time settings, where the integration time of 20 [s]
is corrected using two different NUCs (see section 3.4 for the explanation of the NUC), the
integration time settings are given in table 5.1. The integration time of 73 [s] is corrected
using the internal NUC of the IR camera. These 4 settings will be applied to measurements
at an advance ratio of J = 0.80, J = 1.00 and J = 1.20.
Looking at the first results, J = 0.80 see figure 5.1, it can be seen that the smoothest picture
is the one corrected using the internal NUC, but it also has the highest integration time,
shown in figure 5.1a. Looking at the three remaining results obtained using the own NUC
Table 5.1: Integration time settings

(a)
(b)
(c)
(d)
MSc Thesis

IT [s]
73
15
20
20

SIT [s]
14
19
15

LIT [s]
16
21
25
R.F. Janssen

74

Experimental Results

2290

9660

x 10

x 10

9650

1.066

1.066

1.064

1.064

1.062

1.062

1.06

1.06

1.058

1.058

1.056

1.056

1.054
DLc,BPR

1.054
DLc,BPR

2280
9640
2270

9630

9620
2260
9610
2250

9600

9590

2240

9580
2230
9570

DL

(a)
IT = 73 [s]

2220

9560
DLc,BPR

(b)
IT = 15 [s]
SIT = 14 [s]
LIT = 16 [s]

(c)
IT = 20 [s]
SIT = 19 [s]
LIT = 21 [s]

(d)
IT = 20 [s]
SIT = 15 [s]
LIT = 25 [s]

Figure 5.1: Infrared comparison of integration time for an advance ratio of J = 0.80

method, it can be seen there is not much difference between the results. The measurement
obtained using the 15 [s] IT shows more striping compared to the images obtained at 20
[s]. Comparing the two results using the 20 [s] IT, it can be seen that measurement (d)
shows some measurement errors at the trailing edge of the blade in the middle.
The results for J = 1.00 are shown in figure 5.2. This is the measurement with the highest
rotational speed of the propeller, 3835 [RP M ], this is due to limits on the maximum RPM
of the propeller imposed by the technical staff of the high speed lab. The smoothest result is
given again by the correction done by the IR camera itself. The results obtained with the own
correction are pretty much alike. Again the striping on the measurements obtained at an IT
of 15 [s] make it not the best option. It looks like there are again small pixel inconsistencies
in measurement (d).
As J = 1.00 is the most critical case an extra measurement result set is given in figure 5.3,
where the results are shown in grey tones and the image is left in its original state. In figure
5.3a it can be seen that at the IT of 73 [s] blurring of the image occurs, which is expected at
high rotation speeds and long integration times. This result shows why the long integration
time of 73 [s] is not suited for measurements on the propeller. Comparing the other three
cases, figures 5.3b, 5.3c and 5.3d, it can be seen that the other three measurements show
comparable results.
Finally the advance ratio of J = 1.20 is investigated, see figure 5.4. If the previous results
R.F. Janssen

MSc Thesis

5.1 Integration Time Influence

75

2370

2360

9840

9840

9840

9820

9820

9820

9800

9800

9780

9780

9760

9760

9740

9740

9720

9720

9700

9700

9800
2350
9780
2340
9760
2330
9740
2320
9720

2310

DL

(a)
IT = 73 [s]

2300

9700

9680
DLc,BPR

(b)
IT = 15 [s]
SIT = 14 [s]
LIT = 16 [s]

DLc,BPR

(c)
IT = 20 [s]
SIT = 19 [s]
LIT = 21 [s]

DLc,BPR

(d)
IT = 20 [s]
SIT = 15 [s]
LIT = 25 [s]

Figure 5.2: Infrared comparison of integration time for an advance ratio of J = 1.00

are taken into account: IT 73 [s] too long, previous striping results at IT 15 [s] and more
bad pixels at measurement (d), it can be seen that the setup of (c) shows the best results. In
the rest of the report the results obtained using setup (c) will be used.

MSc Thesis

R.F. Janssen

76

Experimental Results

9800

2350

9750

2300
9700

9650
2250

9600

DL

9550
DLc,BPR

2200

(a)
IT = 73 [s]

(b)
IT = 15 [s]
SIT = 14 [s]
LIT = 16 [s]

DL

9800

9800

9750

9750

9700

9700

9650

9650

9600

9600

9550

DL

c,BPR

(c)
IT = 20 [s]
SIT = 19 [s]
LIT = 21 [s]

9550

c,BPR

(d)
IT = 20 [s]
SIT = 15 [s]
LIT = 25 [s]

Figure 5.3: Comparison of the complete infrared image at J = 1.00

R.F. Janssen

MSc Thesis

5.2 Advance Ratio Influence

77
4

x 10

9780

1.084

1.084

9760

1.082

1.082

1.08

1.08

1.078

1.078

1.076

1.076

1.074

1.074

1.072

1.072

1.07

1.07

2340

9740

2330

x 10

2350

9720
2320
9700
2310
9680

2300
9660

DL

(a)
IT = 73 [s]

2290

9640
DLc,BPR

(b)
IT = 15 [s]
SIT = 14 [s]
LIT = 16 [s]

1.068
DLc,BPR

(c)
IT = 20 [s]
SIT = 19 [s]
LIT = 21 [s]

1.068
DLc,BPR

(d)
IT = 20 [s]
SIT = 15 [s]
LIT = 25 [s]

Figure 5.4: Infrared comparison of integration time for an advance ratio of J = 1.20

5.2

Advance Ratio Influence

To see what the influence of the advance ratio is on the flow surrounding the propeller, three
different measurements were done at different
rotation speeds at the same wind tunnel speed.
 
The wind tunnel speed is set to 19.4 m
,
results
shown in figure 5.5.
s
When comparing the measurement results in figure 5.5, it can be seen that the lowest advance
ratio shows a large hot area at the trailing edge of the propeller blade, see figure 5.5a, this
region is interpreted as a region of separated flow. At a third of the blade a hot spot can be
seen.
Looking at J = 1.00, it can be seen that the region of high temperature is moving a little bit
toward the leading edge, becoming a smaller region. This probably means that the separated
flow attaches again at the trailing edge. Seeing how a cool spot occurs behind the separation
line it is expected that at this location the flow has become turbulent.
Decreasing the advance ratio to 0.80, it can be seen that in the middle of the blade a short
transition region can be distinguished, after which immediatly a cold region is forming, suggesting turbulent flow. At the hub region not much is changed, only a slight movement of
the separation region toward the leading edge. The interesting flow phenomena occur at the
blade tip. It can be seen that in the tip region a large hot spot is forming. If investigated
MSc Thesis

R.F. Janssen

78

Experimental Results
4

x 10

1.066

x 10

x 10

1.066

1.066

1.064
1.064

1.064

1.062
1.062
1.062

1.06

1.058

1.06
1.06

1.056
1.058
1.058
1.054
1.056
1.052

1.056
1.054

1.05
DLc,BPR

(a) J = 1.20

DLc,BPR

(b) J = 1.00

1.054
DLc,BPR

(c) J = 0.80

Figure
 5.5: Comparison of different advance ratios at a constant wind tunnel speed v =
19.4 m
s , (a) 2400 [RP M ], (b) 2865 [RP M ] and (c) 3575 [RP M ]

more thoroughly one could distinguish two regions, one continuing from the separation line
formed over the blade. And another region at the trailing edge of the blade tip.

5.3

Reynolds Number Influence

The Reynolds number can be very influential on the state of the boundary layer. As was stated
in section 1.1, the Reynolds number is one of the main controlling parameter of viscous fluid
flows. Because of this it is of great interest to see what the influence is of a change in Reynolds
number on the state of the boundary layer.
To see the influence of the Reynolds number on the boundary layer, four measurements were
done, two at J = 1.20 and
at a
 two
 at J = 1.00. The tests at J = 1.20 were performed
m
wind tunnel speed of 19.4 m
and
2400
[RP
M
]
and
a
wind
tunnel
speed
of
25
and
3075
s
 ms 
[RP M ]. The advance ratio of 1.00 was tested
at
a
wind
tunnel
speed
of
19.4
s and 2865
 
[RP M ] and at a wind tunnel speed of 26 m
and
3835
[RP
M
].
s
First the results of the tests at an advance ratio J = 1.20 will be shown in figure 5.6. In figure
5.6a it can be seen that a separation line can be seen on the trailing edge region of the blade,
with a hot spot at about a third of the blade. Comparing this to the test at a higher Reynolds
number, figure 5.6b, it can be seen that there is still a separation line. At a third of the blade
still a larger region of backflow can be seen, but the hot spot on the blade is now moved to
R.F. Janssen

MSc Thesis

5.3 Reynolds Number Influence

79
4

x 10

x 10

1.066

1.084

1.064

1.082

1.062

1.08

1.06

1.078

1.058
1.076
1.056
1.074
1.054
1.072
1.052
1.07
1.05
1.068
DLc,BPR

DLc,BPR

(a)

(b)

Figure 5.6: Infrared comparison


  of Reynolds number for J = 1.20, (a) v = 19.4
RP M = 2400, (b) v = 25 m
s and RP M = 3075

m
s

and

the trailing edge of the blade tip, and the separation region at the blade tip is expanded. It
is expected that due to the rotating motion of the propeller, the flow in the separated region
of at the trailing edge is transported in the direction of the tip of the propeller blade.
The measurements done at J = 1.00 are shown in figure 5.7. At the low Reynolds number
flow, figure 5.7a a thick separation region can be seen over the whole length of the blade,
ending in a hot spot at the trailing edge of the tip. Looking at the high Reynolds number
flow, shown in figure 5.7b it can be seen that the separation region ends in a bigger hot spot
at the trailing edge of the blade tip. Again it is thought that because of the increased rotation
speed, more mass is transported to the blade tip, leading to a larger separated region, leading
to a larger hot spot.

MSc Thesis

R.F. Janssen

80

Experimental Results

x 10

9840

1.066

9820

1.064

9800

1.062

9780

1.06

9760

1.058

9740

1.056

9720

1.054

9700
DLc,BPR

DLc,BPR

(a)

(b)

Figure 5.7: Infrared comparison


 m  of Reynolds number for J = 1.00, (a) v = 19.4
RP M = 2865, (b) v = 26 s and RP M = 3835

R.F. Janssen

m
s

and

MSc Thesis

Chapter 6
Comparison Numerical and Experimental
results
This chapter compares some of the results obtained from the RANS simulations with the
experiments performed in the OJF of the Delft University of Technology. First the results at
an advance ratio of 1.20 are compared, after this the results of J = 1.00 and this chapter will
finish with the results of the advance ratio J = 0.80.
First the results of J = 1.20 are shown, see figure 6.1. To compare the RANS |Cf | values
with the infrared results, the colour map of the IR camera results is inverted, meaning blue
is hot and red is cold. Using the Reynolds analogy, see equation (3.1), the skin friction can
be related to the heat flux at the wall. Low values of skin friction would mean low values of
heat flux, thus low heat exchange between the propeller blade surface and the air flow over
the blade surface. This means that regions of lower skin friction have higher temperatures.
Comparing figures 6.1a and 6.1b one can see big differences. In the simulation using RANS
it can be seen that on the trailing edge of the propeller blade a separation region is forming.
In the RANS simulation assuming fully turbulent flow this can not be seen. Comparing the
simulation data to the experimental data, it can be seen that during the wind tunnel test a
region of flow separation is occurring at the trailing edge of the propeller blade. The difference
between the RANS simulation including transition and the experimental data is the apparent
larger region of separation at the blade tip and the formation of the hot spot at one third of the
blade length. In the numerical results there is some bulge forming at the region comparable
to the place of the hot spot on the real propeller blade. There is definitely a difference at the
hub of the propeller, as the connection of the blade to the hub is not modelled correctly, this
way the simulation of the hub region is wrong.
Looking at the results obtained at an advance ratio J of 1.00, shown in figure 6.2, more
differences can be noticed. Looking at the fully turbulent RANS simulation, it can be seen
that the values of |Cf | suggest a completely different temperature distribution compared to
MSc Thesis

R.F. Janssen

82

Comparison Numerical and Experimental results

x 10

1.084

1.082

1.08

1.078

1.076

1.074

1.072

1.07

1.068
DLc,BPR

(a) Turbulent

(b) Transition

(c) Experimental

Figure 6.1: Comparison numerical and experimental result, J = 1.20

9840

9820

9800

9780

9760

9740

9720

9700
DLc,BPR

(a) Turbulent

(b) Transition

(c) Experimental

Figure 6.2: Comparison numerical and experimental result, J = 1.00

R.F. Janssen

MSc Thesis

83
4

x 10

1.066

1.064

1.062

1.06

1.058

1.056

1.054
DLc,BPR

(a) Turbulent

(b) Transition

(c) Experimental

Figure 6.3: Comparison numerical and experimental result, J = 0.80

the experimental data. Comparing the transitional RANS data with the experimental data,
some similarities can be seen. First off, there is the shape of the separation region, it follows
from both the experiment and the RANS simulation including transition that the separation
line bends towards the leading edge after which it will turn back to the trailing edge a little
and it ends in a separation region in the blade tip region which covers almost half the chord
length. There are also differences, for example the vortex like structures behind the separation
line from the simulation are not found in the experiment. This may be due to simulation
faults, or due to an inadequate resolution of the IR camera which makes it impossible to
capture small scale effects. It also seems as if the flow separates a little bit too early in the
middle region of the simulated blade.
Finally J = 0.80 is investigated, see figure 6.3. It should be noted that there is a difference
between the RANS simulations and the
m wind tunnel tests, as the simulations were executed

assuming a wind tunnel
m speed of 26 s , while the experiments were conducted at a wind


tunnel speed of 19.4 s . This means that the Reynolds numbers of the flows do not match
and this can have some influence on the obtained results, see section 5.3.
Again it can be seen that the value of |Cf | of the fully turbulent simulation does not match
the result from the experiment. If the transition RANS simulations are compared to the
experimental data similarities in the flow can be seen. The bottom part of the blade shows
an early separation of the flow near the leading edge. Almost halfway up the blade, it can
be seen that the separation line moves to the trailing edge of the blade ending in a large
separation region at the trailing edge of the blade tip. Again there are some dissimilarities
between the simulation and the experiment. Mainly the occurrence of vortex structures at
MSc Thesis

R.F. Janssen

84

Comparison Numerical and Experimental results

the trailing edge of the blade. It also seems as if the separation region curves more toward
the trailing edge in the experiment compared to the simulation.

R.F. Janssen

MSc Thesis

Chapter 7
Conclusions and Recommendations
This chapter will present the conclusions and recommendations for future research based on
the results presented in this report. First section 7.1 will discuss the conclusions and in section
7.2 the recommendations are presented.

7.1

Conclusions

In this report the influence of laminar-turbulent transition on an isolated propeller was investigated, both numerically and experimentally. The aim of this work was to investigate if it
was possible to determine the influence of laminar-turbulent transition on a propeller blades
performance through the use of computational fluid dynamics (CFD) and wind tunnel experiments.
It was found that the results from the Langtry-Menter model showed strange behaviour at
J 1.0. At these advance ratios, vortical structures started forming in the flow near the
propeller blade after the flow had separated. After investigation it was found there could be
a possible implementation error in the Langtry-Menter model in the used DLR TAU code.
Comparing the turbulent RANS simulations with the RANS simulations incorporating
laminar-turbulent transition, large differences between the propeller performance were found.
The differences in thrust coefficient, CT , between the two methods varied from approximately
1% at J = 0.65 up to 27% at J = 1.2. The same could be said about the differences in the
power coefficient, CP , varying from approximately 1% at J = 0.65 up to 23% at J = 1.2. It
is concluded that the big differences at J = 1.2 are probably due to the laminar separation
bubble on the aft part of the propeller blade. The small difference at J = 0.65 can be explained by the fact that most of the flow over the propeller blade is turbulent. Even though
the values of CT and CP show a large difference between the two RANS models, the value
of the propeller performance varies only slightly over the investigated range. At J = 1.2 the
MSc Thesis

R.F. Janssen

86

Conclusions and Recommendations

difference between the computed s is around 6%, while at J = 0.8 only a 0.06% difference
was calculated.
The differences between BEMT results and the RANS models are more pronounced. For
J = 1.2 the difference between CT obtained using BEMT and the turbulent RANS simulation
is 14.5%, while at J = 0.65 the difference decreased to about 12.4%. CT between BEMT
and the transition RANS simulations varied from 57.3% at J = 1.2 to 13.5% at J = 0.65.
The differences between CP obtained through BEMT and turbulent RANS simulations range
between 13.8% when J = 1.2 and 13.2% at J = 0.65. CP for BEMT and RANS simulations
using transition vary between 47.3% when J = 1.2 and 14.6% at J = 0.65. The values of
the propeller efficiency start to deviate at advance ratios Jgeq0.8. From J = 0.8 the values
of for the two RANS models increase slower compared to BEMT. At J = 1.05 the RANS
model including transition reaches its maximum, while the turbulent RANS model has not
reached its maximum yet up to J = 1.2. BEMT evaluated the maximum propeller efficiency
at J = 1.1.
From the blade loading characteristics it can be concluded that BEMT performs best as it is
only a two-dimensional technique, where the RANS simulations are three-dimensional. Due
to a force in Y -direction, it is expected that the force vector is rotated, leading to a lower lift
and a higher drag in case of the RANS simulation, thus showing an decrease in performance.
From the wind tunnel experiments it could be concluded that the best setup for the integration
time is 20 [s]. For the non uniformity correction (NUC) the best settings to use were a short
integration time of 19 [s] and a long integration time of 21 [s]. At too long an integration
time it could be concluded that the images were not focussed, too short an integration time
made the images degrade. When the difference between the short and long integration time
was increased, it was concluded that the resulting images also degraded.
When the advance ratio was increased during the experiments, it could be concluded that at
first the separation region is moved to the leading edge, especially in the inner part of the
propeller blade. While at J = 0.8 it can be concluded that on the middle part of the blade
leading edge transition is present, as a large region of colder surface can be observed, meaning
high skin friction, thus turbulent flow. On the outer part of the blade a lot of laminar flow
is present, as from the infrared image it can concluded that the hot region indicates low skin
friction and thus laminar flow. If the Reynolds number of the flow is changed, it can be
concluded that the state of the boundary layer also changes. This could be seen from the
change in the temperature profile over the blade surface. It was observed that at a higher
Reynolds number, same advance ratio, the separation region size at the tip of the blade
increased.
After comparing the RANS simulations with the experimental data it could be concluded that
the agreement between the RANS model using transition and the experimental data showed
some similarities, despite the vortical structures being present in the RANS data.
From the numerical and experimental evaluations it is concluded that the influence of laminarturbulent transition on a propeller blades performance can be determined to some degree
through CFD and wind tunnel experiments. The CFD results showed some similarity with
R.F. Janssen

MSc Thesis

7.2 Recommendations

87

the experimental data, but there are still some uncertainties which need to be solved such as
the vortical structures forming on the blade surface. If the problems with the transition model
could be solved, a good analysis tool would be available with which the more complicated
flow features of propellers can be investigated.

7.2

Recommendations

The author of this work gained insight into the influence of laminar-turbulent transition on
a propeller blade, but there are still some areas where the results could be improved. To
improve the results obtained during this thesis some recommendations are listed here.
During the RANS simulations using the Langtry-Menter transition model, vortical structures
were present in the flow near the propeller blade. After investigating these vortical structures
it was found that it could be possible these vortical structures started forming because of
an implementation error of the Langtry-Menter transition model in the DLR TAU code.
It is thought the TAU code implements a Dirichlet boundary condition, while a Neumann
boundary condition should have been specified.
It is also possible that these vortical structures start forming in the flow near the blade, but
could not be detected by the infrared camera due to limited resolution. Using an infrared
camera using a higher resolution could be a possible, but expensive, solution to see if the
vortical structures are truly present in the flow near the blade. Another solution could be to
investigate the propeller using Particle Image Velocimetry (PIV). But it is very difficult to
investigate boundary layer on the propeller using PIV, as the boundary layer is very thin.
During the wind tunnel tests, there was no Rotating Shaft Balance (RSB) available for the
XPROP propeller. This made it very difficult to measure the performance of the propeller in
the wind tunnel and it was opted to use the performance data in the XPROP documentation.
As no information about the freestream parameters is known of these performance data,
it possible these tests were performed at a different Reynolds number, leading to different
boundary layer behaviour, which in turn could influence performance of the propeller.
In this work, only qualitative measurements were performed using the infrared camera. In a
next study quantitative measurements could provide useful information, as the skin friction
could be calculated and compared to the results from CFD.

MSc Thesis

R.F. Janssen

88

R.F. Janssen

Conclusions and Recommendations

MSc Thesis

Bibliography
[1] Airbus S.A.S. Global Market Forecast 2013-2032. , Airbus, Blagnac, France, 2013.
[2] C. Lenfers. Propeller Design for a future QESTOL Aircraft in the BNF Project. 30th
AIAA Applied Aerodynamics Conference, June 2012.

[3] L. Prandtl. Uber


Fl
ussigkeitsbewegung bei sehr kleiner Reibung. Verhandl. III. Intern.
Math. Kongr. Heidelberg, feb 1904.
[4] F.M. White. Viscous fluid flow. McGraw-Hill, third edition, 2006.
[5] H. Schlichting and K. Gersten. Grenzschicht-Theory. Springer Berlin Heidelberg, tenth
edition, 2006.
[6] F.M. White. Fluid Mechanics. McGraw-Hill, fourth edition, 1998.
[7] W.S. Saric, H.L. Reed, and E.B. White. Stability and Transition of Three -Dimensional
Boundary Layers. Annual Review of Fluid Mechanics, 35:413440, 2003.
[8] W.S. Saric, H.L. Reed, and E.J. Kerschen. Boundary-layer receptivity to freestream
disturbances. Annual Review of Fluid Mechanics, 34:291319, 2002.
[9] R.E. Mayle. The 1991 IGTI Scholar Lecture: The Role of Laminar-Turbulent Transition
in Gas Turbine Engines. Journal of Turbomachinery, 113(4):509536, October 1991.
[10] M.V. Morkovin. Bypass-Transition Research: Issues and Philosophy. In D.E. Ashpis,
T.B. Gatski, and R. Hirsh, editors, Instabilities and Turbulence in Engineering Flows,
volume 16 of Fluid Mechanics and Its Applications, pages 330. Springer Netherlands,
1993.
[11] E Reshotko. Boundary-Layer Stability and Transition. Annual Review of Fluid Mechanics, 8(1):311349, 1976.
[12] E. Malkiel and R.E. Mayle. Transition in a separation bubble. Journal of turbomachinery,
118(October), 1996.
[13] D.E. Halstead, D.C. Wisler, T.H. Okiishi, G.J. Walker, H.P. Hodson, and H.-W. Shin.
Boundary layer development in axial compressors and turbines: Part 1 of 4composite
picture. Journal of Turbomachinery, 119(January), 1997.
MSc Thesis

R.F. Janssen

90

Bibliography

[14] H. Bippes. Basic experiments on transition in three-dimensional boundary layers dominated by crossflow instability. Progress in Aerospace Sciences, 35(4):363412, may 1999.
[15] H.L. Reed and W.S. Saric. Stability of three-dimensional boundary layers. Annual Review
of Fluid Mechanics, 21:235284, 1989.
[16] R.J. Volino and T.W. Simon. Measurements in a transitional boundary layer with gortler
vortices. Journal of fluids engineering, 119(September), 1997.
[17] W Tollmien. General Instability Criterion of Laminar Velocity Distributions. Technical
Report NACA-TM-792, National Advisory Commision for Aeronautics, 1936.
[18] H. Himmelskamp. Profile Investigations on a Rotating Airscrew. AVA 45 A 20, Aerodynamische Versuchsanstalt Goettingen, 1947.
[19] E. Sch
ulein, H. Rosemann, and S. Schaber. Transition Detection and Skin Friction Measurements on Rotating Propeller Blades. 28th Aerodynamic Measurement Technology,
Ground Testing, and Flight Testing Conference, June 2012.
[20] G. Kuiper. Cavitation inception on ship propeller models. Technische Hogeschool Delft,
1981.
[21] M. Drela. A Users Guide to MSES 3.04, December 2006.
[22] M. Drela. Newton Solution of Coupled Viscous/Inviscid Multielement Airfoil Flows. 21st
Fluid Dynamics, Plasma Dynamics and Lasers Conference, June 1990.
[23] J.L. van Ingen. A suggested semi-empirical method for the calculation of the boundary
layer transition region. Technical Report V.T.H.-74, Technische Hogeschool Vliegtuigbouwkunde, 1956.
[24] A.M.O. Smith and N. Gamberoni. Transition, pressure gradient and stability theory.
Technical Report ES 20388, Douglas Aircraft Company, 1956.
[25] T. Sinnige. The effects of pylon blowing on pusher propeller performance and noise
emissions. Masters thesis, Delft University of Technology, 2013.
[26] N. Kroll and J.K. Fassbender. MEGAFLOW-Numerical flow simulation for aircraft
design. Springer-Verlag Berlin Heidelberg, 2006.
[27] T. Gerhold. Overview of the hybrid rans code tau. In N. Kroll and J.K. Fassbender,
editors, MEGAFLOW - Numerical Flow Simulation for Aircraft Design, volume 89 of
Notes on Numerical Fluid Mechanics and Multidisciplinary Design (NNFM), pages 81
92. Springer Berlin Heidelberg, 2005.
[28] C. Lenfers, R.F. Janssen, N. Beck, N. Friedrichs, and A. Rezaeian. Experimental Investigation of the Propeller Design for future QESTOL Aircraft in the BNF Project. 52nd
Aerospace Sciences Meeting, January 2014.
[29] A. St
urmer, C.O. Marquez Gutierrez, E.W.M. Roosenboom, A. Schroder, R. Geisler,
D. Pallek, J. Agoc, and K. Neitzke. Experimental and Numerical Investigation of a
Contra Rotating Open-Rotor Flowfield. Journal of Aircraft, 49(6):18681877, nov 2012.
R.F. Janssen

MSc Thesis

Bibliography

91

[30] CentaurSoft. Centaur. http://www.centaursoft.com [Accessed 23 January 2015], 2015.


[31] D.C. Wilcox. Turbulence modeling for CFD. DCW Industries, 1998.
[32] P. Spalart and S. Allmaras. A one-equation turbulence model for aerodynamic flows.
30th Aerospace Sciences Meeting and Exhibit, January 1992.
[33] F. Moens, J. Perraud, A. Krumbein, T. Toulorge, P. Ianelli, P. Eliasson, and A. Hanifi.
Transition Prediction and Impact on a Three-Dimensional High-Lift-Wing Configuration.
Journal of Aircraft, 45(5):17511766, sep 2008.
[34] R.B. Langtry and F.R. Menter. Correlation-Based Transition Modeling for Unstructured
Parallelized Computational Fluid Dynamics Codes. AIAA Journal, 47(12):28942906,
December 2009.
[35] R.B. Langtry, F.R. Menter, S.R. Likki, Y.B. Suzen, P.G. Huang, and S. Volker. A
Correlation-Based Transition Model Using Local VariablesPart II: Test Cases and Industrial Applications. Journal of Turbomachinery, 128(3):423434, July 2006.
[36] N.N. Srensen. CFD modelling of laminar-turbulent transition for airfoils and rotors
model. Wind Energy, 12(8):715733, 2009.
using the -Re
[37] R.B. Langtry. A correlation-based transition model using local variables for unstructured
parallelized CFD codes. Universitat Stuttgart, 2006.
[38] C. Grabe and A. Krumbein. Correlation-Based Transition Transport Modeling for ThreeDimensional Aerodynamic Configurations. Journal of Aircraft, 50(5):15331539, September 2013.
[39] Delft
University
of
Technology.
Open
jet
facility.
http://www.lr.tudelft.nl/organisatie/afdelingen/
aerodynamics-wind-energy-flight-performance-and-propulsion/facilities/
wind-tunnel-lab/open-jet-facility-hsl/ [Accessed 12 January 2015], 2015.
[40] Tech Development Inc. Installation and Operation Manual Model 1999A Air Turbine
Motor, 1987.
[41] Sinnige, T. Propeller Test Rig Setup and Operating Manual, 2014.
[42] Stanford Research Systems. Model DG535 Digital Delay/ Pulse Generator. http://
www.thinksrs.com/products/DG535.htm [Accessed 12 January 2015], 2015.
[43] F.F.J. Schrijer. Experimental Investigation of Re-entry Aerodynamic Phenomena. Delft
University of Technology, 2010.
[44] Cedip Infrared Systems. Titanium User Manual.
[45] Cedip Infrared Systems. Altair User Manual.
[46] O. Riou, S. Berrebi, and P. Bremond. Nonuniformity correction and thermal drift
compensation of thermal infrared camera. Proceendings of SPIE, Thermosense XXVI,
5405:294302, April 2004.
[47] NLR. N250 propeller shoptest. Technical report.

MSc Thesis

R.F. Janssen

92

R.F. Janssen

Bibliography

MSc Thesis

Appendix A
Additional Numerical Results
This appendix will show additional results obtained from the numererical simulations. Section
A.1 shows the values of Cp on the blade surface for the different advance ratios and radial
positions. In section A.2 slices at different radial positions are shown for all investigated
advance ratios. Finally section A.3 shows the cl and cd distributions over the blade span for
J = 0.65, J = 0.50 and J = 0.30.

MSc Thesis

R.F. Janssen

94

A.1

Additional Numerical Results

Cp values on propeller blade surface


J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-3

Cp [-]

-2

-1

1
0

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.1: CP distribution at

R.F. Janssen

r
R

= 0.3 for turbulent model and transition model

MSc Thesis

A.1 Cp values on propeller blade surface

95

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.2: CP distribution at

MSc Thesis

r
R

= 0.4 for turbulent model and transition model

R.F. Janssen

96

Additional Numerical Results

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.3: CP distribution at

R.F. Janssen

r
R

= 0.5 for turbulent model and transition model

MSc Thesis

A.1 Cp values on propeller blade surface

97

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-4

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.4: CP distribution at

MSc Thesis

r
R

= 0.6 for turbulent model and transition model

R.F. Janssen

98

Additional Numerical Results

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-2

Cp [-]

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.5: CP distribution at

R.F. Janssen

r
R

= 0.8 for turbulent model and transition model

MSc Thesis

A.1 Cp values on propeller blade surface

99

J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-3

Cp [-]

-2

-1

0.2

0.4

0.6

0.8

X/c [-]

(a) Turbulent
J = 0.30
J = 0.50
J = 0.65
J = 0.80
J = 1.00
J = 1.20

-2

Cp [-]

-1

0.2

0.4

0.6

0.8

X/c [-]

(b) Transition
Figure A.6: CP distribution at

MSc Thesis

r
R

= 0.9 for turbulent model and transition model

R.F. Janssen

100

Additional Numerical Results

A.2

Cp values slices

J = 1.20

(a) Turbulent

(b) Transition

Figure A.7: Cp comparison at


J = 1.20

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.11: Cp comparison at


J = 1.20

R.F. Janssen

= 0.3,

(b) Transition

Figure A.9: Cp comparison at


J = 1.20

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.8: Cp comparison at


J = 1.20

(a) Turbulent

= 0.4,

(b) Transition

Figure A.10: Cp comparison at


J = 1.20

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.12: Cp comparison at


J = 1.20

r
R

= 0.9,

MSc Thesis

A.2 Cp values slices

101

J = 1.00

(a) Turbulent

(b) Transition

Figure A.13: Cp comparison at


J = 1.00

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.17: Cp comparison at


J = 1.00

MSc Thesis

= 0.3,

(b) Transition

Figure A.15: Cp comparison at


J = 1.00

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.14: Cp comparison at


J = 1.00

(a) Turbulent

= 0.4,

(b) Transition

Figure A.16: Cp comparison at


J = 1.00

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.18: Cp comparison at


J = 1.00

r
R

= 0.9,

R.F. Janssen

102

Additional Numerical Results

J = 0.80

(a) Turbulent

(b) Transition

Figure A.19: Cp comparison at


J = 0.80

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.23: Cp comparison at


J = 0.80

R.F. Janssen

= 0.3,

(b) Transition

Figure A.21: Cp comparison at


J = 0.80

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.20: Cp comparison at


J = 0.80

(a) Turbulent

= 0.4,

(b) Transition

Figure A.22: Cp comparison at


J = 0.80

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.24: Cp comparison at


J = 0.80

r
R

= 0.9,

MSc Thesis

A.2 Cp values slices

103

J = 0.65

(a) Turbulent

(b) Transition

Figure A.25: Cp comparison at


J = 0.65

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.29: Cp comparison at


J = 0.65

MSc Thesis

= 0.3,

(b) Transition

Figure A.27: Cp comparison at


J = 0.65

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.26: Cp comparison at


J = 0.65

(a) Turbulent

= 0.4,

(b) Transition

Figure A.28: Cp comparison at


J = 0.65

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.30: Cp comparison at


J = 0.65

r
R

= 0.9,

R.F. Janssen

104

Additional Numerical Results

J = 0.50

(a) Turbulent

(b) Transition

Figure A.31: Cp comparison at


J = 0.50

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.35: Cp comparison at


J = 0.50

R.F. Janssen

= 0.3,

(b) Transition

Figure A.33: Cp comparison at


J = 0.50

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.32: Cp comparison at


J = 0.50

(a) Turbulent

= 0.4,

(b) Transition

Figure A.34: Cp comparison at


J = 0.50

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.36: Cp comparison at


J = 0.50

r
R

= 0.9,

MSc Thesis

A.2 Cp values slices

105

J = 0.30

(a) Turbulent

(b) Transition

Figure A.37: Cp comparison at


J = 0.30

(a) Turbulent

r
R

= 0.5,

(b) Transition

Figure A.41: Cp comparison at


J = 0.30

MSc Thesis

= 0.3,

(b) Transition

Figure A.39: Cp comparison at


J = 0.30

(a) Turbulent

r
R

r
R

= 0.8,

(a) Turbulent

(b) Transition

Figure A.38: Cp comparison at


J = 0.30

(a) Turbulent

= 0.4,

(b) Transition

Figure A.40: Cp comparison at


J = 0.30

(a) Turbulent

r
R

r
R

= 0.6,

(b) Transition

Figure A.42: Cp comparison at


J = 0.30

r
R

= 0.9,

R.F. Janssen

106

Additional Numerical Results

A.3

Blade loading characteristics

This section shows the blade loading characteristics for J = 0.65, J = 0.50 and J = 0.30.

J = 0.65
1.6
1.4

cl [-]

1.2
1
0.8
0.6
RANS Transition
RANS Fully Turbulent
BEMT

0.4
0.2
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.2
RANS Transition
RANS Fully Turbulent
BEMT

0.18
0.16

cd [-]

0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure A.43: Lift and drag comparison BEMT, RANS turbulent and RANS transition for J =
0.65, (a) cl and (b) cd

R.F. Janssen

MSc Thesis

A.3 Blade loading characteristics

107

J = 0.50
1.1
1
0.9

cl [-]

0.8
0.7
0.6
0.5
0.4
0.3
RANS Transition
RANS Fully Turbulent

0.2
0.1
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.35
RANS Transition
RANS Fully Turbulent

0.3
0.25

cd [-]

0.2
0.15
0.1
0.05
0
0.05
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure A.44: Lift and drag comparison RANS turbulent and RANS transition for J = 0.50, (a)
cl and (b) cd

MSc Thesis

R.F. Janssen

108

Additional Numerical Results

J = 0.30
1.2
1

cl [-]

0.8
0.6
0.4
0.2
0
0.2
0.2

RANS Transition
RANS Fully Turbulent
0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(a)
0.8
RANS Transition
RANS Fully Turbulent

0.7
0.6

cd [-]

0.5
0.4
0.3
0.2
0.1
0
0.1
0.2

0.3

0.4

0.5

r
R

0.6

0.7

0.8

0.9

[-]

(b)
Figure A.45: Lift and drag comparison RANS turbulent and RANS transition for J = 0.30, (a)
cl and (b) cd

R.F. Janssen

MSc Thesis

Вам также может понравиться