Numerical simulation of ice accretion on a

propeller blade
Thomas Laoh (s1012932)
Preface

Part of the master curriculum Mechanical Engineering is to do an internship in either an industrial or

research environment. The internship is a great opportunity for students to gain insight in the working
life after their study and to put their knowledge into practice. This is an excellent chance to not only work
at a company for a while, but also to go abroad and live in a different culture. To experience working
abroad eventually led to my choice of an internship at CIRA (International Aerospace Research Centre) in
Capua, Italy. During the months July to October I have been researching the ice accretion on a propeller
of an aircraft numerically. This report describes the process of my research to study ice accretion and an
anti-icing system as well.

I would like to thank my supervisor from the University of Twente, Harry Hoeijmakers of the Engineering
in Fluid Dynamics group and Giuseppe Mingione from the CIRA Fluid Mechanics department for all their
efforts to realize my internship at CIRA. Furthermore, I would like to thank Francesco Petrosino for his
close guidance during my research and always being able to answer questions when possible. My time
at CIRA was very pleasant and the ambiance and experiences very useful for my future job(s) and studying.

Thomas Laoh

November 2015, Utrecht

2
Contents

1 Introduction 4

2 Literature study 5

3 First aerodynamic simulations 7

4 Propeller 2D method 11

5 Full propeller analysis 18

6 Three dimensional check 28

7 Conclusion 31

3
1. Introduction

Icing on aircrafts is known to be a problem for a long time and many studies have been conducted to
determine the effects of this phenomenon. These studies focused on ice accretion and the effects on the
plane of the accumulated ice. Over the years many methods have been designed to de-ice or prevent ice
accretion on a plane. These methods differ for the different parts of the plane. So far, most research
has been done on the wing. Ice accretion on a propeller will reduce the aerodynamic performance of the
propeller meaning the thrust will reduce significantly. This could eventually mean that the plane is not
able to maintain the velocity needed to stay airborne.

The goal of this study will be: To perform an assessment of methods for propeller ice protection and to
perform a case study for propeller ice accretion and ice protection design.

The set up of this research is to first get acquainted and understand the numerical approaches used
to predict icing on aircraft geometries. After which the best method is searched to make a simplified
simulation of the propeller by attempting to make the simulation two dimensional in Chapter 4. The best
method found is then used to fully asses the propeller in Chapter 5. These simulations are then verified
with simulations from a three dimensional flow field in Chapter 6.

Nomenclature
SLD Supercooled Large Droplet
LWC Liquid Water Content
MVD Median Volumetric Diameter
rpm rounds per minute
W Watt
c chord
Ω Rotational velocity (rad/s)
T Temperature (K)
ρ Density (kg/m3 )
U Velocity (m/s)
r radius (m)
a Speed of Sound (m/s)
U∞ free stream velocity

4
2. Literature study

In-flight icing could occur when an aircraft flies in a temperature range between −40°C and 0°C. In these
conditions Supercooled Large Droplets (SLD) can form from small droplets, below these temperatures
SLD can not exist. These droplets are unstable and when contact is made with an airfoil or propeller,
they immediately freeze to the surface. The first ice accretion is usually on the propeller before it is visible
on the wing. While freezing the droplets increase in temperature until 0°C is reached. When the droplet is
not completely frozen after the temperature rise, the remaining fluid will start to run back on the airfoil or
on already formed ice. The shape of the ice is dependent on the liquid water content (LWC), the droplet
diameter, the airspeed and ambient temperature. For low temperatures (−40 to −10°C), low airspeed
and low LWC, the ice immediately freezes and forms ‘Rime ice’. It alters the shape of the airfoil changing
the aerodynamic characteristics and adding weight, possibly bringing it airfoil out of balance. At higher
temperatures (−18 to 0°C), higher airspeed and high LWC, the forming of ‘Glaze ice’ is more likely. Glaze
ice has a completely different profile than the airfoil, forming for example a double horn. These glaze ice
shapes form a danger as they significantly change the aerodynamic performance. Examples of both ice
forms can be found in figure 2.1.[1, 2, 3]

(a) Rime ice (b) Glaze ice

Figure 2.1: Two different types of ice on a propeller blade

Over time systems have been designed to remove the ice (de-ice), or to prevent ice accretion (anti-ice) to
airfoils. Not all of these are applicable on propellers as they are designed for larger (stationary) surfaces.
For propeller blades, only two methods are found to be applicable. Firstly, there are a fluid based ‘weeping
wing’ systems, where alcohol or another freezing point depressant is distributed along the propeller by
a discharge nozzle or slinger ring. The nozzle distributes the fluid at the centre on the leading edge.
The fluid is then distributed along the leading edge by the centripetal force through grooves in the boot.
Advantages of this system are that the airfoil contour can be retained. No runback ice or any residual is
left after the system is actuated. The system has a low power demand. Some drawbacks of this system
are that it can only carry a limited supply, so it will need to be refilled. The liquid system is least effective
when the adhesive bond is the greatest, with low LWC, low temperatures and low MVDs. Lastly, if the
turboprop aircraft uses engine bleed air, the liquid cooling is not possible as it can cause harmful vapours
inside the cabin.

A second option is to heat the surface with electrical heating pads. Power generated by the engine is

5
used to transfer power along wires to a heat a conductive layer of insulation on the airfoil. The electrical
energy should then be applied either continuously to keep the profile clean of ice or intermittently to loosen
and shed the accumulated ice. To maintain an ice free airfoil, the temperature of the surface should be
continuously above the freezing point, it should be checked however that no runback water can freeze
further on the propeller. The main benefit of this system is the simplicity. Also, the response time of the
system is very short compared to the other systems. A disadvantage of this system is that it is power
consuming, especially when evaporative anti-icing is desired. A quick search in available systems learns
that the power used for these kind of anti-icing systems is around 15 kW .

Other methods which work fine on a wing or stationary part are not applicable for propellers. One of
these methods is hot gas de-icing. It is found to be an effective way to transfer heat to the blade by using
bleed air of the engines transferred to the surface through small holes. However, this way of transferring
heat remains limited for wing surfaces and is thus not applicable on the desired propeller. Also, the use
of pneumatic boots (with bleed air) is not applicable. [2, 3, 4, 5, 6, 7]

Some experiments
Some tests have already been done with icing on a propeller, for example by Dumont et al. who conducted
an experiment based on a MU-2B propeller. They state that mainly the performance loss of the propellers
has been documented, but the accretion characteristics are not. In their experiment they measured the
engine rpm, torque, blade angle, boot voltage(s), thrust, wind speed and temperature. They were not able
to measure the exact LWC and MVD during the tests, but calibrated the set up before the experiment.
The icing conditions were correspondent to the icing regulations as presented in Appendix C of the federal
aviation regulations. These are the conditions to which the aviations systems have to be able to withstand
icing to get a safety certification. The velocity profile of the airflow decreased after a span of 75% was
reached, this they assumed acceptable. Despite the fact that exact profiles have not been shown in the
work of Dumont et.al. they have presented some results about the thickness of the ice on various span
widths. They measured the thickness on the leading edge at span widths 28% (just after the icing boot),
50% and 75%. Due to a lack of time they only performed tests with the de-icing device turned on. The
resulting ice accretion on the propeller was then measured in thickness at the three mentioned points.
The results state that in case of icing, most of the times it was found at the span width of 50% (nearly
half of the cases) and a runner up at 28% (one third of the documented data). [8, 9, 10]

Also, research has been done discussing propeller performance using the vortex theory by Busch et al.
First they developed numerical code to be able to simulate the propeller performance. After this code was
developed they used experimental data from Corson and Maynard, who performed a full scale icing test
in 1946, to validate the code. With these experimental results the effect of icing is included. Combining
their code to the results of the thrust, power coefficient and the propeller efficiency found by Corson
and Maynard gives good icing effect results. After this validation they also try to use the experimental
results from (Dumont et al.) to validate the effect of icing on the propeller. They conclude that the
properties found numerically for icing are reasonable but can be improved by adding more experimental
results between the 25% − 50% range instead of interpolating. Also, it is stated that differences between
experimental and simulated effects might also have occurred due to shedding of ice after the measurements
of for example the thrust. Recommending the addition of an ice shedding model could be the solution in
predicting an accurate solution of the performance degradation of a propeller. [11, 12, 13]

6
3. First aerodynamic simulations

For every numerical simulation it must be made sure that the numerical simulations provide realistic
results. Therefore, before any propellers are researched, a simulation has been made of airfoils which
are shaped quite similar to propellers, for which experimental data is available. These simulations are
both run to check the validity of the results for these wings and to learn the process of running these
simulations. This chapter will contain information about these trial simulations, the generation of results
and the results compared to the experimental results. This is all documented extensively such that the
obtained results can be reproduced. [14]

To perform any kind of numerical simulation, a mesh must be made which will be the domain for the flow.
To do so, the geometry of the airfoil is loaded into Ansys ICEM CFD. These are (x, y)-coordinates from
which the airfoil geometry is generated by manually connecting the points to describe the profile. Also, a
far field boundary is built, the result is shown in figure 3.1(a).

To set up the mesh of the flow field, a 2D block can be created. The edges of the block are associated with
the far field profile (figure 3.1(b)). To be able to make a mesh refinement near important flow features like
the leading edge of the profile, the block is then divided into five domains. On the front a domain for the
leading edge which has to be fairly fine. Near the upper and lower edge of the airfoil the mesh should also
be relatively fine such that a boundary layer can be simulated. The two domains for the wake behind the
airfoil do not have any significant influence on the calculations and can thus be relatively coarse further
away from the airfoil (figure 3.1(c)).
The sections on the leading edge and the upper and lower side of the airfoil are provided with a hyperbolic
spacing of the cells along the profile. Over the whole mesh the spacing of cells is set to become exponentially
smaller as it approaches the airfoil. The far field cells will not have changes in velocities, but closer to the
airfoil, and especially inside the boundary layer large fluctuations in velocities will occur. Subsequently,
the upper and lower sections behind the airfoil are set to increase their spacing exponentially moving away
from the profile. This all results in a fine grid near the airfoil, as desired and far away at the far field
regions the grid is coarse as shown in figure 3.1(d).

Having created the mesh, the flow field still has to be determined. The program used is Ansys Fluent. To
do so a couple of extra steps have to be carried out. Firstly, the pre-mesh generated needs to be converted
into an unstructured mesh, which is the type of mesh used in Fluent. Secondly, the output variables need
to be qualified. The solver used is Fluent V6 and the common structural solver is NASTRAN. Thirdly,
the type of boundary conditions need to be provided. The far field is set to be a pressure-farfield, the
airfoil is set as wall and the internal mesh is set as fluid. Lastly, the output needs to be written as a 2D
simulation without any scaling.

To finally begin with the simulation of the flow field, the mesh is opened in Fluent. To perform this sim-
ulation the program is opened in a 2D - double precision mode. Firstly, the mesh quality is checked and
the domain reordered to reduce bandwidth. The solution to be made is a steady, planar, pressure-based
solution. Making use of the 1-equation Spalart-Allmaras viscous model which is set to be strain/vorticity
based. Subsequently, the properties of the air as fluid are set to be an ideal gas, where the Sutherland Law
is used as viscosity model. Next up the boundary conditions are set as presented in the experimental data,
such that a comparison can be made between the computational model and the experimental results. The
data for the practice geometries is presented in table 3.1.[14, 15]

7
 (a) The airfoil geometry and far field (b) The block associated with the edges

(c) Splitting the block in five different blocks (d) The mesh of the flow field

Figure 3.1: Creating a mesh using ANSYS ICEM CFD

Property GLC-305 NFL-414

Static temperature T 263.6 K 267.3 K
Velocity U∞ 128.6 m/s 66.9 m/s
Mach number M∞ 0.396 0.204
Chord c 0.9144 m 0.9144 m
Angle of Attack 1.5° 2°
LWC 0.5 g/m3 0.54 g/m3
MVD 15 µm 20 µm
Spray time 472 s 1350 s

Table 3.1: Experimental flow properties of both airfoils.

Now the flow field corresponding to the experimental set up can be determined. The Solution method
used is a coupled scheme and apart from the Fluent defaults, the Modified Turbulent Viscosity is set to
Second Order Upwind. Convergence of the solution is not searched in the residuals of the continuity, the
velocities or the energy equation as these converge too fast. Therefore, the lift and drag coefficient are also
plotted and written and when these converge the solution is assumed converged. Depending on, among
other things, the Mach number and angle of attack convergence of the solution can already be reached
within 100 iterations.

After these calculations, the flow field with corresponding velocities and pressures are known and can be
plotted. An example is shown in figure 3.2. To be able to simulate the icing on the airfoils, the computer
code of Multi-Ice is used. This program uses the airfoil geometry and the (dimensionless) velocities in the

8
flow field. At first, to obtain the data needed, the results of the Fluent simulation are written into case
and data files. These are then opened using tecplot 360 to write the data in an ASCII -file. Doing this,
the velocities are made dimensionless by dividing them with the far field velocity. After this, the icing
simulation can finally begin.

(a) Velocity magnitude (b) Static pressure

Figure 3.2: Plots of the NFL-414 flow field characteristics.

At first the airfoil geometry and the flow field are read into Multi-Ice. Subsequently, the corresponding
conditions of the flow field are set, again. The reason for this is that the data files created in Fluent and
altered with tecplot do not contain data of the temperatures and droplets. The simulation ran is then a
Field method - Predictor run. After running the Multi-Ice simulation and with the use of MATLAB the
data of the iced airfoils have then been layered over the presented experimental data to be able to give a
comparison of both simulations. The results of the icing on over the clean profile are presented in figure
3.3. [16]

Now that the simulation has been layered over the experimental data, the results can be compared. Firstly,
the GLC-305 simulation (figure 3.3(a)) is discussed. It can be noticed is that the simulation has about a
similar ice thickness as the experimental results have, only the horn is located lower on the leading edge.
Especially the lower horn almost exactly corresponds to the results of the LEWICE simulations, whereas
the upper horn is clearly better simulated by the other simulation type. However, again the thickness of
the ice profile is about the same in all results. Finally, one thing stands out in this simulation. This is the
ice glitching through the surface of the airfoil, which of course should not be possible. When the data is
extracted from Fluent to Multi-Ice, the geometry file for the airfoil is not documented with an increasing
s-coordinate and has to be manually sorted. There was a minor error which resulted in the glitch. This
is visible in figure 3.4, here it shows that the glitch is indeed on that same location.

Secondly, the results of the NFL-414 simulation, which are presented in figure 3.3(b), show a general
agreement in the shape of the ice. However, the thickness is much less than found in the experimental
results. This thickness, amount of ice and especially runback ice is less in the simulation than one would
expect. As the temperature is in the range of 0 − 10°C, a lot of runback ice is expected according to the
literature study. In the experimental results the glaze ice is much thicker than one would expect as well
but there is also clearly runback ice present. The simulations of LEWICE seem to correspond better to
this simulation, apart from one peak which is more appropriately simulated in Multi-Ice. Probable cause
for the reduced ice thickness is the temperature range, as the code works better with colder temperatures
to compute the icing. When water does not completely freeze, the Multi-Ice code does not let it flow to
other regions, but build up on the same location. This is an explanation for the much smaller amounts of
ice and the absence of runback ice in this simulation.

9
 (a) GLC-305 (b) NFL-414

Figure 3.3: The icing simulation with the use of Multi-Ice (coloured), compared with the experimental data and
NASA’s LEWICE simulation. Using the data from table 3.1

−3
x 10

−2

−4

−6

−8

−10

−0.02 −0.015 −0.01 −0.005 0 0.005 0.01 0.015 0.02 0.025

Figure 3.4: The GLC-305 airfoil zoomed in on the leading edge

10
4. Propeller 2D method

Knowing the methods of analysing a flow around an airfoil, an analysis can be made of a flow around the
propeller of a turboprop aircraft. The propeller analysed does contain four propeller blades and rotates
at 420 rpm. This system is placed in flow of 102.89 m/s along the x-axis, a 0° angle of attack on the set
up. A graphical representation of the system - containing the four propeller blades (red), the domain in
which the flow field is calculated (yellow cylinder) and the wakes (blue) which are not further used during
this analysis - is shown in figure 4.1.

Figure 4.1: The numerical domain for the flow field (yellow), the propellers(red), propeller wakes (blue)

The solution provided is given in nine different time steps, as the flow around the propeller blade could
vary during the rotation. Furthermore, a 2D solution is sought for an obvious 3D problem. Where all
previous analyses were made of airfoils, the propeller blades do not only move in the free stream direction,
the blades are also rotating. This gives an extra velocity component in radial direction. Moreover, near
the tip of the propeller blade downwash will be present, creating vortices in the flow field. Near the nacelle
of the propeller (the centre of the blades) the flow will also be disturbed and can not be assumed 2D.
Lastly, the propeller blade has a curl. To be able to make a 2D solution of this problem, the propeller
is cut into multiple slices. This approximates the different curls at different speeds. The geometry of a
propeller blade is shown in figure 4.2.

The unsteadiness of the solution due to the rotating is dealt with by taking the a slice the same blade
section of all time steps. This has been realised by rotating the complete grid to the same position. Such
that the first propeller blade with corresponding flow field is rotated back with α degrees using rotation-
around the x-axis. This way the same same propeller blade will be directed along the z-axis making it
able to obtain the exact same slice each time. The equations used are given in equation 4.1 (for a 45°
rotation). The same rotation matrix can be applied to the velocity components. This way the flow field
of different time steps can be properly evaluated. To deal with the 3D effects near the nacelle and of the
downwash at the propeller tip, they have been cut off the geometry of the propeller blade.. To be certain
the measured flow is not influenced by this three dimensional effects, the slices have not been taken too

11
close to the edge. The locations of the slices are shown in figure 4.2 as the red lines (only z1 to z4 are
used in the current analysis).

x0 =x
45π 45π
y 0 =y cos ( ) − z sin ( ) (4.1)
180 180
45π 45π
z 0 =y sin ( ) − z cos ( )
180 180

Figure 4.2: The geometry of the propeller blade, with the shown slices (black lines) and corresponding numbers

To prove that it does not matter what angle blade has to analyse the flow around on a relative coordinate
system, a visual check has been made and afterwards the flow properties were compared in magnitudes.
In this flow case the aircraft flies horizontally, this means that the angle of attack for the propeller blades
is the same in each location of the cycle. This means the expectation is that the flow will be equal at
all locations. Four positions of the blade section at radius z = 1.95m have been presented in figure 4.3.
Apart from very little differences the flow is indeed the same. An explanation of the minor differences is
found in the rotation of the grids, due to the rotations some grids in the xy-plane become more refined
than others, therefore presenting a better interpolated result in the desired direction. The same results
are obtained for the other flow properties and sections of the blade, but these are omitted from the report
due to the lengthiness.

When the flow is analysed more closely something odd is visible. Figure 4.3 also has the streamlines
plotted for the comparison. These streamlines clearly show that the flow does not deflect as much as one
would expect. The flow does not seem to comply with the no slip condition as the flow goes straight
through the blade. Moreover, the flow due to the rotation of the propeller system does seem to be absent
here as the flow is almost entirely horizontal, except for a small disturbance by the blade.

12
 (a) Pitch angle 0° (b) Pitch angle 45°

(c) Pitch angle 90° (d) Pitch angle 135°

Figure 4.3: The 2D flow field (contour of velocity v in Y direction) around the propeller at a radius of z = 1mvfor
different time steps.

Figure 4.4: The clean airfoil with ice plotted over it (T = 244.7K, U∞ = 102.89m/s), and the limit trajectories.

Despite the realisation this flow field is not realistic, an analysis has been made in Multi-Ice. This is

13
done to be able to see the accretion on the profile with the current flow. In figure 4.4 is visible that the
accretion profile is not at all what is expected. The plot has been made with various accretion times, but
in the given flow field no ice is attached to the profile. Also the droplet trajectories confirm that the flow
field is not good compared to the expected results, as the droplets (which have a standard MVD of 20µm)
should be deflecting around the profile further away from the stagnation point, which is obviously never
going to happen as the streamlines of figure 4.3 show.

The previous results are clearly not as what was hoped to achieve. To have at least one proper simulation
of this set-up a simulation in Fluent has been made. This was done in a similarly to the meshing as
performed in the preparatory paragraph (First aerodynamic simulations). The profile used was of the
propeller blade section 2.5m away from the centre of the propeller. Furthermore, the remaining prop-
erties, previously unused in the other flow field, to be able to simulate a result which might be possible
to compare. These are given in table 4.1. Furthermore, in this simulation use have been made of the
apparent speed as shown in figure 4.5. This flow angle is computed by adding the vectors of the speed in
horizontal direction and the one causing the speed experienced due to the rotation.

Property Value
RPM 426
Ω 44.61 rad/s
Ω×r 111.53 m/s
U∞x 102.89 m/s
a 339.12 m/s
T 286.2 K
ρ 1.190 kg/m3
Uapparent 151.73 m/s

Table 4.1: Properties of the air used to simulate flow field of figure 4.6

In figure 4.6 it is visible how the streamlines now are tangent to the profile instead of the odd horizontal
streamlines as plotted in figure 4.3. This simulation gives a better result regarding the flow near the
profile. Now that the flow field has been obtained, it can be used for a computation in Multi-Ice. For
this simulation, the standard settings regarding temperature, MVD and accretion time are used, as the
flow field is computed with the same settings as the previous flow field. Which was generated with a flow
above freezing temperature. As the flow will be altered insignificantly due to the temperature difference,
this is assumed as a valid simplification.
The shape of the ice accretion is presented in figure 4.7. The ice is accumulating underneath the leading
edge of the section blade. There is as seen earlier, a small glitch in the airfoil geometry where two points
are not connected properly, resulting in a spike of ice out of (and into) the blade. This means the airfoil
geometry should be computed with more care to make sure the geometry is ordered correctly. Below
the leading edge is the place where most ice accretion would be expected. On the other hand, this flow
simulation assumes the apparent flow as presented in figure 4.5 to be also in this matter in the far field.
It needs to be checked, however, if this is a reasonable assumption.

An ideal solution for computing this obvious three dimensional problem in a two dimensional simulation
would be when the flow field could be combined, imposing the rotational velocity onto the propeller blade
section. To do so, the same mesh can be used as used for the earlier simulation. However, the difference is
that from the flow properties of table 4.1 the apparent velocity Uapparent is not used any more. The free
stream velocity imposed on the pressure-farfield zone, is set to U∞x with a 0° angle of attack. And the
section of the propeller blade is no longer set to be a stationary wall. But it is now to translate with the
velocity of Ω × r in the positive direction along the Y-axis. This way, the physics which the droplets in the
free stream experience should correspond closer to the reality as they are only diverted as they approach
the propeller.
In figure 4.8(a) this flow field is shown, where the droplets are carried in the horizontal flow, diverted
around the profile. Behind the profile, two vortices appear as the flow fully separates from the profile.
Although some effects are visible from the translating of the propeller blade, it must be confirmed using a
three dimensional flow field. The level of separation of the flow is however not promising for the validity of

14
Figure 4.5: The apparent velocity (green) to the airfoil, combining the Free stream velocity U∞ (black) and the
rotational velocity Ω × r (red)

8.5 Y Velocity
20
10
0
-10
8 -20
-30
-40
-50
Y

-60
7.5 -70
-80
-90
-100
-110
7 -120
-130
-140
-150
-160
6.5

-2.5 -2 -1.5 -1 -0.5

X

Figure 4.6: The flow field around the propeller blade with the conditions of table 4.1

the simulation, as this would mean there is a huge amount of drag inflicted on the propeller. The results
regarding the ice accretion with the standard Multi-Ice settings (Table 4.2) are shown in figure 4.8(b).
As to be expected, the ice build up is now located much lower than the leading edge. The absence of
ice on the leading edge is worrisome, because despite that ice would be expected on the lower side of the
propeller, the leading edge should also be fully covered.

15
Figure 4.7: Multi-Ice simulation with the flow field of figure 4.6 and outside temperature T = 244.7K and
accretion time t = 360s

8.5 Y Velocity
180
160
140
120
8 100
80
60
40
Y

20
7.5 0
-20
-40
-60

6.5
-2.5 -2 -1.5 -1 -0.5
X

(a) Flow field with induced translation of the propeller blade (b) Ice accretion profile at corresponding flow as
in Table 4.2

Figure 4.8: The flow field with an induced velocity on the propeller blade and flow properties of 4.1.

Property Value
LWC 1 g/m3
Accretion time 360 s
Air Temperature 244.7K
MVD 20 µm

Table 4.2: Standard Multi-Ice settings regarding flow properties

Now that these initial simulations have been made, the choice has been made to further examine the
solution using an apparent velocity from the far field. The strongest reason is the doubt of the validity
and lack of knowledge present about the solution with the translating section. The droplet size of 20 µm

16
is evaluated, which are tiny droplets, which will easily follow the flow. Therefore, the assumption is made
that the plotted streamlines tangential to the propeller sections are plausible for the droplets to be their
trajectory as well. This can hopefully later be confirmed with the use of a (slice of a) three dimensional
flow field.

17
5. Full propeller analysis

To obtain a proper analysis, all simulations are made using the same conditions for creating a flow field.
The conditions used are still the same as the original (faulty) three dimensional flow field, as a new three
dimensional flow field was promised. The assumption is that the same conditions for generating the flow
are used. These conditions are listed in table 5.1. To obtain a view over the whole span of the propeller
blade, several sections of the geometry have been used. The sections used are shown in figure 4.2 including
the corresponding section numbers. Please not that the bottom of the shown geometry is not the centre of
the propeller set. This centre is located close to one meter lower. The distances from the centre of the set
of propellers (r) and the corresponding apparent velocities are presented in table 5.2. In the presentation
of the results only slices z1 to z4 are shown. The remaining three slices have been added to better view a
trend and to give a better insight in the three dimensional picture of the ice accretion on the geometry.

Property Value
a 339.12 m/s
T 286.2 K
ρ 1.190 kg/m3
U∞ 102.89 m/s

Table 5.1: Flow properties for creating the 2D flow field of the propeller sections.

Slice z1 z2 z3 z4 z5 z6 z7
r[m] 1.35 1.95 2.55 3.15 1.65 2.25 2.85
Uapparent [m/s] 119.24 134.77 153.43 174.21 126.54 143.78 163.60

Table 5.2: The slices in the z-plane and their distance from the center of the propeller blades as well as the
apparent free stream velocity

A flow field has been created for each of the slices. As the geometry of the blade is fairly curled, all
sections need their own mesh. In figure 5.1 the necessity for this is clearly visible as the geometry curls
along with the apparent flow for the relative sections on the propeller blade. This is a slight indication
that this approach could be right. The geometry also gets thinner on the outer parts of the blade, causing
less distortion of the flow. But as also visible in table 5.2, the apparent velocity increases significantly.

With this flow fields, the results have been examined using Multi-Ice. Firstly shown are the trajectory
plots for the different sections. In figure 5.2 the trajectories corresponding to the flow field are shown.
These show the trajectories of the droplets which will collide with the airfoil section. The two outer
trajectories are also displayed, showing the first droplets that will not hit the propeller blade. In figure
5.2(d) there is a strange sudden angle in this outer trajectory. But as the droplet on this trajectory line
do not influence any of the icing results it is left in.

Now that the trajectories of the droplets are known, it is also desired to know how much water is actually
impinging the propeller blade. To do so, the β-coefficient is checked. This shows is an impinging factor of
the droplets on the different sections. The plot in figure 5.3(a) shows the impinging factor for the slices
z1 to z4 . What can be seen from this plot, is that further outwards on the propeller, more droplets are
impinging the profile. This could be explained by the significantly higher velocity the section experiences
on these outer regions. The reason why slice z4 has a much thinner collection span than the other three is

18
mainly because the geometry itself is also much thinner. Therefore, the leading edge is as well. Visible is
however, that the impinging factor is much higher on this smaller surface. What can even been seen in slice
z3 and z4 is that in some regions, the factor is above 1. This is odd, as it means more water is impinging
than present in the trajectory line. A possible explanation is that water from different trajectories are
impinging on the same locations, resulting in more water than would be present in one trajectory of air.
Putting this all together, it can be seen that for an increasing velocity the impinging factor increases as
well. Even resulting in more water than to be expected initially.

A bit more chaotic is the plot of the ice thickness of the different sections in figure 5.3(b). As can be
expected from the β-coefficient plot, the ice thickness is largest on the outer slices. Due to the amount of
water impinging the most ice will also be created here. The shape of the plots all describe a horn on the
leading edge. With locally less ice on the on both sides of the stagnation point. The thickness corresponds
to the shown shapes in figure 5.4. This shows that the horns are becoming not only larger in size, but also
more rounded towards the outer edge. While looking closer to the shape of the ice on each section, it is
very well possible that the ice would gradually build up in the given shapes for the sections. Apart from
the horn on the leading edge, there is no ice present on the propeller sections. When looking at figure
5.5, the values from the ice thickness on the sections for all seven slices have been numerically integrated,
calculating the volume of ice per span. The figure shows a nearly linear increase on the middle sections
(r = 1.95 to r = 2.85), in these sections the geometry is quite similar. Meaning the ice accretion is indeed
stronger for higher velocities. The highest section(s) show a decrease in ice volume, which is a result of a
smaller geometry to impinge on.
For the ultimate comparison, the three dimensional geometry of the propeller blade, as used, has been
plotted in figure 5.6. On this three dimensional mesh, the ice horns are added in their corresponding
height. This means the geometry of ice seen on the blade is in scale with the propeller.

A point of discussion for the analysis of the ice accretion is that the programmed code in Multi-Ice does
not cope well with high velocities and high temperatures. In these simulations, the temperature is set
sufficiently low at 244K. But the Mach number of the apparent velocity of the most outer simulated
section is 0.48. This is very high and could result into ice breaking off, which is not embedded into
the numerical solution. Also, in literature Tan and Papadakis , found that droplets impinging at high
velocities have a greater rebound of the (iced) airfoil. Meaning it is possible that the impinged droplets do
not all freeze to the surface as Multi-Ice thinks it does. However, to be able to confirm this, experimental
data is needed as current icing software cannot handle these physical properties of the droplets. Also,
little is known about the exact influence of the Mach number to these physics. [17] [18]

19
 1
1.5
Y Velocity Y Velocity
80 60
70 50
60 1 40
50 30
0.5 40 20
30 10
20 0
10 0.5 -10
Y

Y
0 -20
-10 -30
-20 -40
-30 -50
-40 0 -60
0 -50 -70
-60 -80
-70 -90
-80 -100
-90 -0.5 -110
-100 -120

-2.5 -2 -1.5 -1 -3 -2.5 -2 -1.5 -1 -0.5

X X

(a) z1 (b) z2

1 Y Velocity
40
20
0
-20
0.5 -40
-60
-80
-100
Y

-120
-140
0 -160

-0.5

-3 -2.5 -2 -1.5 -1
X

(c) z3 (d) z4

Figure 5.1: Steamlines of the flow field and the Y-velocity, with conditions used as presented in tables 5.1 and
5.2.

20
 The droplet limit trajectories on the propeller slide z1 The droplet limit trajectories on the propeller slide z2
0.2 0.2

0 0
Position y [m]

Position y [m]
−0.2 −0.2

−0.4 −0.4
−2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55
Position x [m] Position x [m]

(a) z1 (b) z2

The droplet limit trajectories on the propeller slide z3 The droplet limit trajectories on the propeller slide z4
0.2 0.2

0 0
Position y [m]

Position y [m]

−0.2 −0.2

−0.4 −0.4
−2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55
Position x [m] Position x [m]

(c) z3 (d) z4

Figure 5.2: The droplet trajectories for the different sections. (Flow conditions tables 5.1 and 5.2.)

Beta coefficient vs airfoil coordinate s (stagnation point on s=0, slices z1−z4 Ice thickness vs airfoil coordinate s (stagnation point on s=0, slices z1−z4
Slice z1
Slice z1
Slice z2
Slice z2
Slice z3
Slice z3
Slice z4
1.2 Slice z4

1
0.1
Ice thickness [m]
Beta Coefficient

0.8

0.6

0.05
0.4

0.2

0 0
−0.08 −0.06 −0.04 −0.02 0 0.02 0.04 0.06 0.08 −0.08 −0.06 −0.04 −0.02 0 0.02 0.04 0.06 0.08
Coordinate on wing s [m] coordinate on wing s [m]

(a) Beta coefficient (b) Ice thickness

Figure 5.3: The beta coefficient and Ice thickness on the coordinate of the airfoil with s=0 is the stagnation
point.(Flow conditions tables 5.1 and 5.2.)

21
 Geometry of ice on propeller slice z1 Geometry of ice on propeller slice z2
Clean airfoil Clean airfoil
Ice geometry Ice geometry
0.1 0.1

0.05 0.05

0 0

−0.05 −0.05
y [m]

y [m]
−0.1 −0.1

−0.15 −0.15

−0.2 −0.2

−0.25 −0.25

−2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −1.5 −2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −1.5
x [m] x [m]

(a) z1 (b) z2

Geometry of ice on propeller slice z3 Geometry of ice on propeller slice z4

Clean airfoil Clean airfoil
Ice geometry Ice geometry
0.1 0.1

0.05 0.05

0 0

−0.05 −0.05
y [m]

y [m]

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2

−0.25 −0.25

−2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −1.5 −2 −1.95 −1.9 −1.85 −1.8 −1.75 −1.7 −1.65 −1.6 −1.55 −1.5
x [m] x [m]

(c) z3 (d) z4

Figure 5.4: The shape of the ice accretion on various slices among the propeller blade (Flow conditions tables 5.1
and 5.2.)

−3
x 10
Ice found on the propeller after icing analysis
2.6
Volume of ice per unit span [m3]

2.4

2.2

1.8

1.6

1.4
1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
Radius r [m]

Figure 5.5: The integrated values of the ice thickness along the s-coordinate (figure 5.3(b)).

22
 Figure 5.6: Three dimensional plot of the ice build up along the leading edge using slice 1 to 7

Anti-Ice analysis
The ice accretion on the propeller blade from this study is dangerously high and will most definitely
decrease its performance significantly. This means the propeller needs to be protected from flying in the
conditions analysed. In the literature study it is found that the most common and easy anti-icing system
on a propeller blade is by heating it with an electrical path. Since there is a different amount of ice
accretion for each section of the blade as previously studied, for each section an anti-icing analysis has
been made. This will give an insight in how much energy is needed to clear the whole propeller of ice.

To show how such an analysis has been performed, the in-between data to get to the final design of
the anti-icing on the first slice (z1 ) has been presented. To obtain the data, the simulations for the ice
accretion have been re-used. Then on the surface where ice is present, an (electrical) heat is added of a
certain amount of W/m2 . Multi-ice does not simulate a flow of water on the profile, the water that does
not freeze to the propeller will stay on the same location, resulting in a relatively thick layer of runback
water. When no heat is added to the regions with water, the water will then freeze to the next point
as the temperature of the propeller is still below the freezing point. Therefore, it has been decided to
only clear the leading edge plus 15% of the chord length. This is assumed sufficient, but it can only be
validated with the use of experiments. To define the leading edge, the original section containing ice has
been used. So 0.15 m away (along the airfoil coordinate s) from the point where ice was first found will
be the point where ice may build up in this numerical simulation.

In the graph in figure 5.7 the different tests to obtain the ice free leading edge using as little energy as
possible are shown. Firstly, a constant value of 100, 000 W/m2 (test1 ) has been applied everywhere until
0.15 m away from where the ice originally formed. The results of the ice thickness and the runback water
are shown in figure 5.8. Here it is clearly visible that the ice is practically gone. The thickness is below
a measurable level (the black line is not even visible). Also, looking ahead, in figure 5.8(b) can be seen
that some water even evaporates with this heat as much less water is present in test 1 than the other
simulations. If more ice is melted, water should increase, as for evaporating the water too much energy
is added. So for the second run, still a constant power was applied, but now 80, 000 W/m2 (test2 ). This
power clears most of the ice, except for one peak near the stagnation point. This means more heat must
be applied there, not as much as test1 but just enough. The rest of the section might be able to still be
ice-free using less power. For the third test shown, a peak of power is presented near the point where the
remaining ice was found during test2. Moving away from this peak the applied power drastically drops

23
and eventually will go to zero at the point where the ice is allowed again. In the third to fifth (and final)
presented series, the power curves have been slightly adjusted, such that the ice will form behind the
marked 15% zones. As visible, each simulation the ice moves further away from the leading edge, whilst
also attempting to use minimal power.

Now for the six other slices (z2 to z7 ), the same kind of analysis has been made. The results of slice z2
to z4 are presented in the figures 5.9, 5.10 and 5.11. What stands out from these figures is that on the
upper side of the leading edge on all geometry sections, the largest power is applied in the form of a peak.
This power then rapidly declines and remains constant with a slight decrease for the remaining part of the
desired ice free zone. It was shown in the previous section that the amount of ice build up is increasing
towards the outer part of the propeller blade. However, the simulation shows that the required power
to prevent icing from occurring does not have a direct relation to the amount of ice. This is due to the
fact that not only the amount of ice built up determines how much power is needed to keep it clean. The
geometry and corresponding velocity will also influence these results. In table 5.3 the amount of power
added per section is checked by integrating the graphs of the power density over the s-coordinate. The
amount of power per slice does not follow any particular trend, but is fairly constant around 15000 W/m.

Slice z1 z2 z3 z4 z5 z6 z7
Power added [W/m] 15546 15286 15503 15700 14383 15561 14468

Table 5.3: The amount of power to be added in one slice

4
x 10 Power added to propeller surface
12
test 1
test 2
test 3
test 4
10 Result

8
Power Q W/m2

0
−0.25 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0.25
s [m]

Figure 5.7: Applied power to the propeller on slice z1 per square meter

24
 x 10
−3 Runback water after anti−icing
Ice thickness after heating the leading edge 2
0.03 test 1
test 1 test 2
test 2 1.8 test 3
test 3 test 4
test 4 Result
0.025 Result 1.6
15% marker

1.4
0.02

Water layer [m]

Ice Thickness [m]

1.2

1
0.015

0.8

0.01
0.6

0.4
0.005

0.2

0 0
−0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 −0.3 −0.2 −0.1 0 0.1 0.2 0.3
s [m] s [m]

(a) Resulting ice (b) Runback water

Figure 5.8: The ice thickness for several simulations of anti-ice set-ups for slice z1 .

4
x 10 Power added to propeller surface, slice2
12 Ice thickness after heating the leading edge, slice z2 Runback water with anti−icing, slice z2
z2 0.03 0.004
z2 z2
15% marker

10 0.0035
0.025

0.003

Runback water height [m]

8
0.02
Ice Thickness [m]
Power Q W/m2

0.0025

6
0.015 0.002

0.0015
4 0.01

0.001

2 0.005
0.0005

0 0 0
−0.25 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0.25 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 −0.3 −0.2 −0.1 0 0.1 0.2 0.3
s [m] s [m] s [m]

(a) Power density (b) Ice thickness (c) Runback water

Figure 5.9: Anti-icing results for slice z2

4
x 10 Power added to propeller surface, slice3
12 Ice thickness after heating the leading edge, slice z3 Runback water with anti−icing, slice z3
z3 0.03 0.004
z3 z3
15% marker

10 0.0035
0.025

0.003
Runback water height [m]

8
0.02
Ice Thickness [m]
Power Q W/m2

0.0025

6
0.015 0.002

0.0015
4 0.01

0.001

2 0.005
0.0005

0 0 0
−0.25 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0.25 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 −0.3 −0.2 −0.1 0 0.1 0.2 0.3
s [m] s [m] s [m]

(a) Power density (b) Ice thickness (c) Runback water

Figure 5.10: Anti-icing results for slice z3

25
 4
x 10 Power added to propeller surface, slice4
12 Ice thickness after heating the leading edge, slice z4 Runback water with anti−icing, slice z4
z4 0.03 0.004
z4 z4
15% marker

10 0.0035
0.025

0.003

Runback water height [m]

8
0.02

Ice Thickness [m]

Power Q W/m2

0.0025

6
0.015 0.002

0.0015
4 0.01

0.001

2 0.005
0.0005

0 0 0
−0.25 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0.25 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 −0.3 −0.2 −0.1 0 0.1 0.2 0.3
s [m] s [m] s [m]

(a) Power density (b) Ice thickness (c) Runback water

Figure 5.11: Anti-icing results for slice z4

To check if indeed all the ice was cleared of the propeller sections, and the water is not vaporised by the
added heat, the plots of the total runback water on the four main sections have also been presented, next
to the remaining ice residue sufficiently far from the leading edge.
Resulting from the anti-icing study is that the droplets impinging now form water instead of ice. To
see how much water stays onto the propeller, these results per section are also added. Moreover, the
remaining ice which supposedly forms after the heated section, is included as well. To see if the leading
edge is indeed ice-free. These lines have also been integrated numerically and the resulting water and ice
per section are plotted in figure 5.12. What can be concluded from this graph, is that with the heat in
each section, the amount of runback water follows the exact same trend as the amount of ice (as presented
in figure 5.5). To check if this result could be indeed valid, also the remaining ice has been plotted into
the figure and it shows that this is a constant volume of ice present in all simulations. So it looks like the
applied power to each section is appropriate.

−3
x 10
Resulting Runback water and remaining ice
1.8
amount of water and ice left [m3/m]

1.6

1.4

1.2

0.8

0.6 Runback water

Remaining ice

0.4
1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
Propeller radius r [m]

Figure 5.12: The resulting runback water per section and remaining ice far from leading edge.

Lastly, the added power of each section has been interpolated onto the three dimensional blade for a
visualisation. Here it indeed visible that in the middle of the propeller the highest power is applied. The
leading edge is heated about the same width everywhere. The interpolation is however not perfect in
radial direction, as the mesh on which is interpolated is very coarse here. But the general picture is clear.

26
 Z Z

X
X Y
Y
Q/m2 Q/m2
110000 110000
100000 100000
90000 90000
80000 80000
70000 70000
60000 60000
50000 50000
40000 40000
30000 30000
20000 20000
10000 10000

(a) 3D plot (b) 3D plot inverse angle

Figure 5.13: Interpolated values of the power density of the 7 slices visualised on a three dimensional profile

27
6. Three dimensional check

It has been said before that the simplifications made must be validated to ensure that the simulations
are a representation of the actual flow around the propeller. As there is no experimental data available
to validate the flow field and icing analyses made in the previous chapters, a numerical three dimensional
flow field is used to see if the local two dimensional analyses are valid.

The three dimensional flow field obtained has a slightly different geometry than that of the earlier simula-
tions. Also, different slices have been used, but this should not influence the result much, as the previous
analysis showed that the geometry of the ice is quite similar but for the volumes of ice. In the two dimen-
sional analyses it was assumed that the far field velocity would consist of two velocity components. These
components are the free stream velocity of the plane and the velocity induced by the rotation. The latter
is measured by multiplying the radius times the rotational speed. The graphical representation of the
apparent velocity was shown in figure 4.5. The velocity component is the radial direction was neglected
as this was assumed to be minimal, especially in the mid-sections of the propeller. This is known to be
the case close to the propeller. However, it might not be the case for the droplets moving in from further
away. Therefore, it needs to be checked if the droplet trajectory / flow streamline will be following this
assumed apparent velocity as they approach the propeller.

This flow field has been generated in slightly different flow conditions than the earlier simulations. The
properties of the flow are presented in table 6.1 For the analysis of this three dimensional flow field, four
slices have been used. The three extra slices in the two dimensional analysis are a good comparison to
find a correlation between the radial differences. Unfortunately, there is not enough time to analyse them
as well. The properties of the slices analysed are presented in table 6.2.

Property Value
M 0.4
T 249.19 K
ρ 0.6601 kg/m3
U∞ 126.57 m/s
Ω 44.67 rad/s

Table 6.1: Flow properties for the creation of the 3D flow field (figure 6.1) of the propeller sections.

Slice z1 z2 z3 z4
r [m] 1.2 1.8 2.4 3.0
Uapparent [m/s] 138.4 150.9 167.4 186.5

Table 6.2: The different flow properties of the four slices.

In figure 6.1 the streamlines constructed from the absolute velocity (x, y, z) directions is viewed in an
Y -Z-plane. Also, the contour plot on the background shows the local values of the velocity in radial direc-
tion. Close to the propeller, the radial component of this velocity indeed approaches zero. Moreover, the
streamlines show an full agreement with the assumption made in the earlier analysis. The flow approaches
the propeller in a tangent angle, sufficiently far away from the airfoil to make the droplets travel in this
direction as well.

28
 Z Z

X Y X Y

Vx_Earth_ms Vx_Earth_ms
50 100
40 80
30 60
20 40
10 20
0 0
-10 -20
-20 -40
-30 -60
-40 -80
-50 -100

(a) Slice z1 (b) Slice z2

Z Z

X Y X Y

Vx_Earth_ms Vx_Earth_ms
100 100
80 80
60 60
40 40
20 20
0 0
-20 -20
-40 -40
-60 -60
-80 -80
-100 -100

(c) Slice z3 (d) Slice z4

Figure 6.1: The streamlines along the slices of the new three dimensional flow field.

To be able to load the data into Multi-Ice, a flow from left to right is desired. Furthermore, the flows
need to be normalised with their free stream apparent velocity (from table 6.2). Moreover, the program
cannot handle three-dimensional input. Therefore, as the x-velocity component does not have influence
on the streamlines for this section, it has been set to 0. Also, to be able to compare the flow slightly more
to the earlier solution. In Multi-Ice, different settings regarding the temperatures have been set. These
are still the same as the earlier computations, presented in table 4.2.

The simulations in Multi-Ice show that the ice has a different shape compared to the earlier computations.
It is less horn shaped as the earlier simulations. Please note that as the flow field was given in dimen-
sionless coordinates, in this case also the multi-ice simulations have been made in these dimensionless
coordinates. This should however, not influence the results of the simulation. What is also visible in
these plots is that for the first slice, in figure 6.2(a), the slice was chosen to be too close to the centre of
the propeller set up. As this geometry catches significantly more ice then the others. This is also visible
when plotting the accreted ice, similar to the earlier results (figure 6.3). This is the point where more
simulations are necessary to be able to view a trend in the ice accretion. Assumed is that the trend is
similar to the 2D analysis, only the geometry of the radius at 1.2 meter is too small, maybe here the three
dimensional effect also plays a part. As visible in figure 6.1(a) the velocity in x-direction is between 10
and 20 m/s close to the leading edge.

Finally, the results of the full two dimensional analyses and the slices of the three dimensional field are
compared. The flow fields consist of almost the same characteristics. The rotational speed of the propellers
is almost equal. Unfortunately, the free stream velocity of the three dimensional field is much higher than
the two dimensional free stream analysis. This could inflict some differences in the ice accretion as we
already know that for high velocities Multi-Ice might not work perfectly. The stream lines in the three
dimensional flow field and the two dimensional field show the exact same flow behaviour. As assumed the
droplets appear to also follow the apparent velocity which was constructed from the rotational and free

29
 Ice geometry on slice z1 Ice geometry on slice z2
ice ice
0.2 clean airfoil 0.2 clean airfoil

0.1 0.1

y coordinate [−]

y coordinate [−]
0 0

−0.1 −0.1

−0.2 −0.2

−0.3 −0.3

−0.4 −0.4
−0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2
x coordinate [−] x coordinate [−]

(a) Slice z1 (b) Slice z1

Ice geometry on slice z3 Ice geometry on slice z4

ice ice
0.2 clean airfoil 0.2 clean airfoil

0.1 0.1
y coordinate [−]

y coordinate [−]
0 0

−0.1 −0.1

−0.2 −0.2

−0.3 −0.3

−0.4 −0.4
−0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2
x coordinate [−] x coordinate [−]

(c) Slice z1 (d) Slice z1

Figure 6.2: The ice geometry on the various slices.

Total ice build up for the different radii

0.004

0.0035

0.003

0.0025
Ice [m2]

0.002

0.0015

0.001

0.0005

0
1 1.5 2 2.5 3
Radius r [m]

Figure 6.3: The total ice build up for different radii

stream velocity. The ice geometry of the three dimensional analyses are a bit odd and should be further
researched.

30
7. Conclusion

The goal of this study was:

To perform an assessment of methods for propeller ice protection and to perform a case study for propeller
ice accretion and ice protection design.

From the literature study, two solutions of propeller ice protection systems have been found. These are
the weeping wing system, which distributes a freezing point depressant along the propeller. Using the
centrifugal force to distribute it along the leading edge. With a large disadvantage that the it can only
be used for systems where no bleed air is sent into the cabin. Because dangerous fumes may spread when
these depressants are heated.
The second system, the system analysed, is the addition of electrical heating pads on the propellers leading
edge. These heating pads can either heat the profile intermittently, to de-ice the propeller. Or continuously
on a lower power, to keep the propeller clean of ice. The latter is further analysed using computational
fluid dynamics.

After a few attempts for icing simulations on wings, an icing simulation was made for the propeller system
as provided. Great difficulty was found in the obtaining of a proper flow field to describe a realistic flow.
Therefore, an assumption has been made to use only the propeller geometry and simulate a flow using
only the free stream properties as constants. An apparent velocity was constructed by combining the free
stream and rotational velocity. This was done for seven different slices of the propeller geometry. To be
able to gain an insight in the different levels of ice accretion. It showed that an increasing radius led to
more ice. However, after the geometry started to get more slim near the tip, less droplets were caught,
resulting into less ice accretion.

A three dimensional flow field was delivered which satisfies the Navier-Stokes equations. This is also
analysed to try and compare if the assumptions in the two dimensional analyses were valid. It could be
concluded that the flow field has identical properties. Also, for the mid-sections of the propeller geometry,
the radial velocity component is nearly zero close by the propeller. For slices near to the nacelle and the
propeller tip this is not the case. The assumption for the two dimensional flow is therefore only valid for
a limited section of the propeller geometry.

On the ice found in the two dimensional analysis, an anti-icing analysis has been conducted. It became
clear that to keep the propeller clean of ice, a power peak is necessary near the stagnation point of the
geometry. Also, independent of the section and amount of ice built up on the different sections. All slices
needed about the same amount of total energy, which was 15 kW/m. This is a value similar as seen while
looking at suppliers of these anti-icing systems.

Unfortunately, this research has some flaws which can be better evaluated in a further research. The flow
conditions of the two dimensional analysis are not equal to those of the three dimensional analysis. They
were based on an earlier given three dimensional flow field, which appeared to be wrong. Therefore, it
is strongly advised to simulate the two dimensional flow again with the same conditions as this will take
less computing time. Furthermore, the icing results should be validated with the use of experiments, to
ensure that Multi-Ice predicts the ice accretion correctly.

31
Bibliography

[1] I. Paraschivoiu and F. Saeed. Aircraft icing. Wiley-Interscience publication, 2004.

[2] Air owners and pilot association. Aircraft icing. http://flighttraining.aopa.org/pdfs/SA11_
Aircraft_Icing.pdf.

[3] Scott K. Thomas, Robert P. Casoni, and Charles D. MacArthur. Aircraft anti-icing and de-icing
techniques and modeling, 1996.
[4] James P. Lewis. De-icing effectiveness of external electric heaters for propeller blades. Flight Propul-
sion Research Laboratory, Ohio, 1948.

[5] Anti-ice and deice systems. http://www.aboutflight.com/

handbook-of-aeronautical-knowledge/ch-6-aircraft-systems/anti-ice-and-deice-systems,
2015.
[6] R.B. Rutherford. De-ice and anti-ice system and method for aircraft surfaces, February 27 2001. US
Patent 6,194,685.

[7] UTC Aerospace Systems. Icing conditions appendix c.

[8] C. Dumont, P. Pellicano, T. Smith, and J. Riley. Results from a full-scale propeller icing test. In
46th AIAA Aerospace Sciences Meeting and Exhibit, 2008.
[9] FAR (Federal Aviation Regulations). Icing conditions appendix c, 2015.

[10] C. Dumont, P. Pellicano, T. Smith, and J. Riley. Propeller icing tunnel test on a full-scale turboprop
engine. 2010.
[11] B.W. Corson, Jr. and J.D. Maynard. The effect of simulated icing on propeller performance. 1946.
NACA TN-1084.

[12] G. Busch, M. Bragg, and A. Broeren. Prediction of propeller performance in icing conditions using
vortex theory. In 1st AIAA Atmospheric and Space Environments Conference, 2009.
[13] K.D. Korkan, L. Dadonet, and R.J. Shaw. Performance degradation of propeller system due to rime
ice accretion, 1984.
[14] William B. Wright and Adam Rutkowski. Validation results for lewice 2.0, 1999.

[15] Inc ANSYS. Ansys fluent tutorial guide, release 14.0, 2011.
[16] Giuseppe Mingione and Francesco Petrosino. Cira multi-ice code version 3.3.3.
[17] S.C. Tan and M. Papadakis. Droplet breakup, splashing and re-impingement on an iced airfoil, 2005.

[18] Alan J. Bilanin. Proposed modifications to ice accretion/icing scaling theory, 1991.

32