Ice Accretion

AIAA 2015-3715
AIAA Propulsion and Energy Forum

July 27-29, 2015, Orlando, FL
13th International Energy Conversion Engineering Conference
In-Cloud Ice Accretion Modeling on Wind Turbine Blades

Using Extended Messinger Model
Ali, M. Anttho1 and Lakshmi, N. Sankar 2

School of Aerospace Engineering, Georgia Institute of Technology, Atlanta GA 30332-0150
Wind turbines operating under cold weather conditions may accumulate ice on its blades.
Icing causes the blade sections to stall prematurely reducing the power production at a given
wind speed. The unsteady aerodynamic loads associated with icing can accelerate blade
structural fatigue and creates safety concerns. In this work, the combined blade element-
Downloaded by UNIVERSITY OF TORONTO on November 18, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2015-3715
momentum theory is used to compute the air loads on the baseline rotor blades, prior to icing.
At each blade section, a Lagrangian particle trajectory model is used to model the water
droplet trajectories and their impact on the blade surface. The extended Messinger model is
next used to solve the conservation of mass, momentum, and energy equations in the boundary
layer over the surface, and to determine ice accretion rate. The 2-D ice shapes obtained are
compared against experimental data at several representative atmospheric conditions with
acceptable agreement. The performance of a generic experimental wind turbine rotor exposed
to icing climate is simulated to obtain the power loss and identify the critical locations on the
blade. The results suggest the outboard of the blade is more prone to ice accumulation causing
considerable loss of lift at these sections. Iced formed on the leading edge strongly influences
the flow resulting in an early separation over the upper surface. Also, for a tapered and twisted
blade, the blades operating at a higher pitch are expected to accumulate more ice. The loss in
power ranges from 10% to 50% of the rated power for different pitch settings under the same
operating conditions.
Nomenclature
Cpa = specific heat capacity of air
Cpi = specific heat capacity of ice
Cpw = specific heat capacity of water
ds = length of panel section
ki,w = thermal conductivity of ice, water
LWC = liquid water content
LF = latent heat of fusion
MVD = median volumetric diameter
r = adiabatic recovery factor
Ta = air temperature
Tf = freezing temperature of water
T∞ = ambient temperature
χe = evaporation coefficient
ε = emissivity of ice
ρg = density of glaze ice
ρr = density of rime ice
ρw = density of water
σr = Stefan-Boltzmann constant
φ = blade pitch angle
1
Graduate Student, School of Aerospace Engineering, Georgia Institute of Technology, 270 Fesrt Dr, Atlanta GA
30332-0150, Student Member AIAA.
2
Regents Professor and Associate Chair for Undergraduate Programs, School of Aerospace Engineering, Georgia
Institute of Technology, 270 Fesrt Dr, Atlanta GA 30332-0150, AIAA Fellow.
1
American Institute of Aeronautics and Astronautics
Copyright © 2015 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
I. Introduction
W IND energy is one of the most promising forms of renewable energy. Wind farms continue to expand around
the world including cold regions and high altitudes, which offer some of the best locations for wind farms.
These locations, however, are innately susceptible to atmospheric icing events during late fall and winter seasons.
Cold climate favors accumulation of ice on the wind turbine blade which adversely influences the airfoil aerodynamic
characteristics, and decreases the power production. It also causes accelerated structural fatigue of the blades and the
gear system, and increases associated maintenance costs. Ice shedding from the blades is also a serious safety concern
for nearby communities, roads, power lines etc. Heavy icing prevents continued operation of wind turbines and the
disruption may be prolonged in severe conditions. The Swedish statistical incident database reported 161,523 hours
of total downtime between 1998-2003, 7% of which were related to cold climate and resulted in 25% or more of
production loss [1].
Regions with cold climates are preferred locations for wind farms as these areas have high potential of energy
harvesting. The International Energy Agency (IEA) estimates around 60 GW of wind power installed in such regions
across the globe, one third of which is located in regions favorable to icing [2]. As the IEA notes, “the large-scale
exploitation of cold climate sites has been limited by our lack of knowledge about their special challenges and the lack
of proven and economical technological solutions [2].”
“The expansion of wind power in cold regions requires an improved understanding of icing on wind turbines and
the resulting production losses and safety risks. On-site measurements are time consuming, labor intensive, and require
long term collection of data for a wide range of ambient conditions [3].” There is a need of mathematical and numerical
models that complement these field studies and improve understanding and prediction of icing events. Such
simulations are helpful in designing turbine blades and deicing systems for better performance and improved life. An
a priori knowledge of ice accumulation and loss of power would also enable power companies to prepare in advance
for a steady level of power supply with appropriate backups.
The current models for ice accretion on wind turbines are based on two popular concepts. A commonly used
approach for forecasting icing events in wind farms is the model proposed by Makkonen [4], which describes in-cloud
icing on a vertically placed, freely rotating cylinder. The ice accretion rate calculated by the model mainly depends
on droplet collision efficiency, adhesion efficiency, and accretion efficiency. Icing characteristics on a vertical, freely
rotating cylinder are most likely different from icing on a wind turbine blade in circular motion in a vertical plane as
the shape, dimensions and air flow are very different. As a result, the Makkonen model can only yield a qualitative
estimate of the amount of ice accreted on a given blade surface. A second approach, similar to the approach for
aircraft/rotorcraft icing, is based on Computational Fluid Dynamics where aerodynamic parameters are evaluated
using commercial CFD solvers followed by ice growth calculations employing numerical codes such as LEWICE,
TURBICE or FENSAP [5] [6]. These ice growth codes are based on Messinger model [7]. Some of these models have
three dimensional capabilities. Since these codes are developed strictly for aviation needs, these analyses include
redundant features not applicable to wind turbines. These analyses also ignore some of the important aspects of wind
turbines operational environment such as prolonged exposure to icing conditions and temporal variations in external
meteorology. For application on wind turbines, the icing tools must be coupled with numerical weather models and
include more of the physics.
This research work is an effort to better understand the fundamentals of ice accretion on wind turbines and to more
reliably predict operational production losses due to icing. The extended Messinger model is implemented in
MATLAB and validated using available 2-D data. Finally, the applicability to wind turbines is demonstrated by
modeling the performance of wind turbines under representative operating conditions, and extracting the distribution
of ice shape and mass over the blade surface.
II. Problem Definition

Prediction of the power losses of a wind turbine under inclement weather conditions requires an understanding of
the physics of the problem. Depending upon the ambient conditions, ice may form in a variety of ways. These may be
roughly classified as in-cloud icing, precipitation icing, and frost. In cloud icing, subject matter of the present study
and the most frequently encountered, is an ice accretion phenomenon caused by water droplets impinging on a cold
surface. The resulting ice shape can be further categorized as Rime and Glaze Ice. Glaze ice is the result of liquid
precipitation striking surfaces at temperature below the freezing point. Glaze ice is rather transparent, hard and
attached firmly to the surfaces. Rime ice occurs when surfaces below the freezing point are exposed to clouds or fog
2
composed of super-cooled water droplets. Its white and opaque appearance is caused by the presence of air bubbles
trapped inside the ice. Rime ice is of primary importance in high elevation locations such as hill or mountain tops.
Precipitation icing is the freezing of snow or rain after striking a cold surface. Frost is the sublimation of water vapors
directly on a cold surface and is common at lower temperatures. However, frost is of low density and strength and
therefore does not have significant effects.
The rate of accretion and the amount of ice accumulated on an unheated surface depends on the shape and size of
the body, surface roughness, wind speed, ambient temperature and pressure, liquid water content in the air, and the
size of the water droplets in the cloud. The process starts with the interception of water droplets by the body. The rate
at which the droplets are intercepted is a function of the ability of the body to collect the water droplet, known as the
collection efficiency, the amount of water present in the cloud or liquid water content, and the wind speed or free
stream velocity. The collection efficiency depends on the size and shape of the body, effective angle of attack, and the
diameter of the droplets. Once the droplets are intercepted, heat transfer occurs from the droplets to the cold surface.
Modeling the heat transfer rate requires modeling the kinetic heating, cooling by convection, evaporation or
sublimation of the water droplets, and other factors. For freezing to occur, the droplets must lose energy equivalent to
the heat of fusion. The heat losses are caused by convection and evaporation. A balance equation of all the heat
exchanges can be used to determine the formation of ice.
III. Problem Formulation

The ice accretion may be empirically modeled by an approach proposed by Messinger [7], and further developed
by Myers et al [8]. The Messinger model treats the growth of ice as a Stephan phase change problem, which can be
described mathematically as,
𝜕𝑇 𝑘𝑖 𝜕 2 𝑇
=
𝜕𝑡 𝜌𝑖 𝐶𝑝𝑖 𝜕𝑦 2 (1)
𝜕𝜃 𝑘𝑤 𝜕 2 𝜃
=
𝜕𝑡 𝜌𝑤 𝐶𝑝𝑤 𝜕𝑦 2 (2)
𝜕𝐵 𝜕ℎ
𝜌𝑖 + 𝜌𝑤 = 𝜌𝑖 𝛽𝑉∞ + 𝑚̇𝑖𝑛 − 𝑚̇𝑒,𝑠 (3)
𝜕𝑡 𝜕𝑡
𝜕𝐵 𝜕𝑇 𝜕𝜃
𝜌𝑖 𝐿𝐹 = 𝑘𝑖 − 𝑘𝑤
𝜕𝑡 𝜕𝑦 𝜕𝑦 (4)
Here, 𝑇 and 𝜃 are the temperatures in the ice and water layers, respectively; y is the normal distance measured
from the wall; 𝑘, 𝜌 and 𝐶𝑝 are the thermal conductivity, density and specific heat capacity, respectively; and, the
subscripts 𝑖 and 𝑤 refer to ice and water respectively. The quantities ρiβV∞, 𝑚̇𝑖𝑛 and 𝑚̇𝑒,𝑠 are the impinging, runback
and evaporating (or sublimating) water mass flow rates for a control volume respectively. Finally 𝐿𝐹 is the latent heat
of solidification of water. Equations (1) and (2) give the temperature distributaions in the ice and water layers
respectively. Equation (3) is the mass balance in the control volume whereas Eq (4), also called the phase change law,
is the heat balance at the ice-water interface.
The turbine blade is didivided in several radial sections, where flow velocities and angles are computed, followed
by the ice growth calculations at each section. For each section, the basic Stephan problem is formulated in terms of
the ambient conditions and geometric parameters to give the shape and thickness of ice formed. These parameters
include ambient temperature and pressure, liquid water content, wind speed and blade geometry. The solution includes
heat and mass balances for each control volume. All possible heat transfers are estimated based on the governing
parameters. A Lagrangian based droplet trajectory calculation is employed to determine 𝑚̇𝑖𝑛 , the amount of water
deposited on the blade surface. Hess-Smith panel method coupled with integral boundary layer approximations gives
the necessary flow field parameters. The modified shapes are analysed to determine their aerodynamic characteristics,
which is then used to determine the power production.
3
Wind Speed, RPM, Blade Pitch,
LWC, Ambient Conditions
BEMT
(Angle of Attack, Lift, Drag
and Power Coefficients)
CFD or XFOIL
(Iced Airfoil aerodyanmic Panel Method
analysis to generate lift, drag (Pressure and Velocity Distribution)
coefficients vs alpha)
Messinger Model Droplet Solver

(Ice Thickness and Shape) (Collection Efficiency)
Figure 1: Schematic of the solution procedure.
IV. Implementation
A. Blade Element Momentum Theory
The flow velocities and angles, illustrated in the Figure 2 below, at each blade section from a wind turbine
aerodynamic analysis, based on the blade element-momentum theory [9]. The flow consists of the free stream velocity
V∞ and the rotational velocity Ωr accompanied by induced veloicties in both axial and tangential directions. The
induced velocity factors are determined iteratively with an initial estimate of the tangential factor a’=0 and axial factor
a given by
1
𝑎= [2 + 𝜋𝜆𝜎 − √4 − 4𝜋𝜆𝜎 + 𝜋𝜆2 𝜎(8𝛽 + 𝜋𝜎)] (5)
4
where λ= Ωr/V∞ and σ is the rotor solidity. From here, the inflow angle ψ is given by,
1−𝑎
tan 𝜓 = (6)
𝜆(1 − 𝑎′ )
4
Figure 2: Flow Velocities and Angles at a Blade Section
The induction factors are then given by
−1
4𝐹𝑠𝑖𝑛2 𝜑
𝑎 = [1 + ] (7)
𝜎(𝐶𝑙 𝑐𝑜𝑠𝜑 + 𝐶𝑑 𝑠𝑖𝑛𝜑)
−1
4𝐹 sin 𝜑 cos 𝜑
𝑎′ = [−1 + ] (8)
𝜎(𝐶𝑙 sin 𝜑 − 𝐶𝑑 cos 𝜑)
The effective ange of attack, from figure 3, is
𝛼 =𝜓−𝜑 (9)
Once the effective angle of attack is computed, the associated lift and drag coefficients for the clean or iced airfoil
section may be computed from a table look up. The sectional lift and drag forces may be integrated to compute the
sectional thrust and torque as shown below. Empirical corrections may be made for tip and hub loss of lift in terms of
F.
2 (1
𝑑𝑇 = 4𝜋𝜌𝑟𝑈∞ − 𝑎)𝑎𝐹𝑑𝑟 (10)
1
𝑑𝑄 = 𝜌𝑟𝑉𝑇2 (𝐶𝑙 𝑠𝑖𝑛𝜑 − 𝐶𝑑 𝑐𝑜𝑠𝜑) 𝑐 𝐹𝑑𝑟 (11)
2
This approach has been programmed in Matlab, and validated for the NREL Phase VI rotor for which experimental
data is available [10]. Sample results are shown in Fig. 3 below.
5
Power Curves
12
10
Power (kW)
6
BEMT
4 NASA
2
0
4 5 6 7 8 9 10
Wind Speed (m/s)
Figure 3: Validation of the Blade Element Model for the NREL Phase VI Rotor
B. Flow Field Calculations
Hess-Smith panel method [11] [12] is employed for the calculation of pressure and velocity distributions. This
method is based on a vortex and source distribution around an airfoil. The vortex strength is assumed to be constant
over the whole airfoil. The Kutta condition is invoked to fix its value. The source strength is allowed to vary from
panel to panel. This source distribution, together with the constant vortex distribution and the free-stream velocity
should satisfy the flow tangency boundary condition on the blade surface. A comparison of the pressure distributaion
over NACA 0012 aifoil by this method is compared with that obtained from XFOIL is shown in the fiure below.
Pressure Distribution AoA = 4° Pressure Distribution AoA = 8°

-2.0 -4.5
XFOIL XFOIL
-1.5 -3.5 Current
Current
-1.0 -2.5
Cp
-0.5
-1.5
Cp
0.0
-0.5
0.5
0.5
1.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c
Figure 4: Pressure distribution over NACA 0012
6
C. Heat Transfer Coefficeinet
Surface heat transfer rate is calculated for both laminar and turbulent regions separately. Transition is based on
roughness height given by
𝜌𝑈𝑘 𝑘𝑠
𝑅𝑒𝑘 = (12)
𝜇
where the rouhgness height ks [13] is estamited as

1
4𝜎𝑤 𝜇𝑤 3 (13)
𝑘𝑠 = ( )
𝜌𝑤 𝐹𝜏
and the local velocity at the roughness height is given by
𝑈𝑘 𝑘𝑠 𝑘𝑠 3 𝑘𝑠 4 1 𝛿 2 𝑑𝑈𝑒 𝑘𝑠 𝑘𝑠 3
= 2 − 2( ) +( ) + (1 − ) (14)
𝑈𝑒 𝛿 𝛿 𝛿 6 𝜈 𝑑𝑠 𝛿 𝛿
Here, δ is the boundary layer thickness computed by Thwaite’s method and ds is the distance along the airfoil
surface. Rek = 600 is the transition location, as of von Doenhoff [14]. Other necessary parameters are evaluated by
well-known empirical relations. Viscosity of air is computed from Sutherland’s Law. The heat transfer rate coefficient
in the laminar region is based on Smith and Spalding [14] correlation.
0.296𝑘𝑈𝑒1.87
ℎ𝑐 =
𝑠 (15)
√𝜈 ∫0 𝑈𝑒1.87 𝑑𝑠
In turbulent flows, the method of Kays and Crawford [14] is employed for computing the heat transfer rate.
ℎ𝑐 = 𝑆𝑡𝜌𝑈𝑒 𝐶𝑝 (16)
while the following realtions are used for the intermediate calculations,
𝐶𝑓
𝑆𝑡 = 2
(17)
𝐶𝑓 1
𝑃𝑟𝑡 + √
2 𝑆𝑡𝑘
𝑆𝑡𝑘 = 1.92𝑅𝑒𝑘−0.45 𝑃𝑟 −0.8 (18)

𝐶𝑓 0.1681
= 2
2 864𝜃𝑡 (19)
[ln ( + 2.568)]
𝑘𝑠
0.8
0.036𝜈 0.2 𝑠
𝜃𝑡 = (∫ 𝑈𝑒3.86 𝑑𝑠 ) + 𝜃𝑡𝑟 (20)
𝑈𝑒3.29 𝑠𝑡𝑟
7
D. Collection Efficiency
Assuming spherical droplets, acted upon only by gravity and aerodynamic drag, in a flow field, unaffected by their
presence, collection efficiency is determined based on Lagrangian approach. The only force in the x-direction that we
are considering is the component of drag in x-direction while in y-direction, the drag component is accompanied by
the gravitational force. Thus, the equations of motion becomes, which are solved for the entire flow field to get the
droplet trajectories.
𝑚𝑥̈ 𝑝 = −𝐷 𝑐𝑜𝑠 𝛾 (21)
𝑚𝑦̈𝑝 = −𝐷 𝑠𝑖𝑛 𝛾 + 𝑚𝑔 (22)

where
𝑦̇𝑝 − 𝑉𝑦
𝛾 = 𝑡𝑎𝑛−1 (23)
𝑥̇ 𝑝 − 𝑉𝑥
1
𝐷 = 𝜌𝑉 2 𝐶𝐷 𝐴𝑝 (24)
2
2 2
𝑉 = √(𝑥̇ 𝑝 − 𝑉𝑥 ) + (𝑦̇𝑝 − 𝑉𝑦 ) (25)
The drag coefficient for the droplets are determined by the following drag law.
1 + 0.197𝑅𝑒 0.63 + 2.6 × 10−4 𝑅𝑒 1.38 , 𝑅𝑒 ≤ 3500

𝐶𝐷 = { (26)
(1.699 × 10−5 )𝑅𝑒 1.92 , 𝑅𝑒 > 3500
where Re = ρVdp /μ is the Reynolds Number based on the droplet diameter dp, the relative velocity 𝑉 and air viscosity
μ.
Finally, the collections efficiency is given by the ratio of distance between two water droplets at the release plane
and at the impact location.
∆𝑦0
𝛽= (27)
∆𝑠
Figure 5: Definition of collection efficiency.
The data generated is compared with the data from Kim et al [15], shown in which the collection efficiency
is plotted against the curvilinear distance over the surface of the airfoil, staring from the stagnation point. The negative
x-axis corresponds to the lower surface whereas the positive x-axis corresponds to the upper surface. Acceptable
agreement is found between the collection efficiency distributions.
8
0.8 0.8
0.7 Current 0.7 Current

Kim Kim
0.6 0.6
0.5 0.5
0.4
β
0.4
β
0.3 0.3
0.2 0.2
0.1 0.1
0 0
-0.3 -0.2 -0.1 0 0.1 -0.3 -0.2 -0.1 0 0.1
s/c s/c
(a) (b)
Figure 6: Collections efficiency validation.
(a) MVD = 20 μm, c = 0.53 m, α = 4.0°, V ∞ = 58.10 m/s
(b) MVD = 20 μm, c = 0.53 m, α = 3.5°, V∞ = 102.8 m/s
E. Heat and Mass balance
The thermodynamic model, based on the heat and mass transfer equations derived by Messinger, and extended by
Myers et al, is used to determine the quantity of ice accreted on each panel of the airfoil. A number of heat transfer
modes are accounted for, which are expressed in terms of the ambient conditions. Readers are refered to [14] for the
detailed derivations and the underlying assumptions.
The type of ice formed, rime or glaze, needs to be identified, which is achieved quantitatively by computing the
parameters Bg and tg, , given by,
𝑘𝑖 (𝑇𝑓 − 𝑇𝑠 )
𝐵𝑔 =
𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 − 𝑚̇𝑠𝑢𝑏 (28)
𝜌𝑔 𝐿𝐹 ( ) + (𝑄0 + 𝑄1 𝑇𝑓 )
𝜌𝑟
𝜌𝑟
𝑡𝑔 = ( )𝐵
𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 − 𝑚̇𝑠𝑢𝑏 𝑔 (29)
Bg is the thickness and tg is the time at which glaze ice first appears. Q1 and Q0 are functions of ambient conditions
and physical properties of air, water and ice and are given by [14],
𝑉∞2 𝑉∞2
𝑄0 = 𝜌𝑎 𝛽𝑉∞ + 𝑟ℎ𝑐 + 𝜌𝑎 𝛽 𝑉∞ 𝐶𝑝𝑤 𝑇𝑎 + ℎ𝑐 𝑇𝑎 + 4𝜖𝜎𝑟 𝑇𝑎4 + 𝜒𝑒 𝑒0 𝑇𝑎 + 𝑚̇𝑖𝑛 𝐶𝑝𝑤 𝑇𝑓
2 2 𝐶𝑝𝑎 (30)
𝑄1 = 𝜌𝑎 𝛽𝑉∞ 𝐶𝑝𝑤 + ℎ𝑐 + 4𝜖𝜎𝑟 𝑇∞3 + 𝜒𝑒 𝑒0 + 𝑚̇𝑖𝑛 𝐶𝑝𝑤

(31)
9
Equations (18) and (19) give a quantitative measure for the type of ice formed. Only rime ice formed when 𝐵𝑔 ≤ 0
and 𝑡𝑔 > 𝑡𝑒𝑥𝑝 . In this case, the thickness of ice B and the freezing fraction FF are given by
𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 − 𝑚̇𝑠𝑢𝑏
𝐵=( )𝑡
𝜌𝑟 (32)
𝜌𝑟 𝐵
𝐹𝐹 =
(𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 )𝑡 (33)
Glaze ice forms when 𝐵𝑔 > 0 and𝑡𝑔 ≤ 𝑡𝑒𝑥𝑝 . Glaze ice is calculated by solving the following differential equation
using Runge-Kutta-Fulhberg method. In this case, we get freezing fraction by Eq. 12.
𝜕𝐵 𝑇𝑓 − 𝑇𝑎 𝑄0 − 𝑄1 𝑇𝑓
𝜌𝑔 𝐿𝐹 = 𝑘𝑖 − 𝑘𝑤
𝜕𝑡 𝐵 𝑘𝑤 − 𝑄1 ℎ (34)
𝜌𝑟 𝐵𝑔 + 𝜌𝑔 (𝐵 − 𝐵𝑔 )
𝐹𝐹 = (35)
(𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 )𝑡
Runback water mass flow rate is given by

𝑚̇𝑜𝑢𝑡 = (1 − 𝐹𝐹)(𝜌𝑎 𝛽𝑉∞ + 𝑚̇𝑖𝑛 ) − 𝑚̇𝑒
(36)
V. Validation
In flight icing on both aircrafts

and rotorcrafts have been extensively Table 1. Conditions for test cases
studied as compared to icing on wind
turbines. Hence, there is sufficient Parameters Units CASE 27 CASE 28 CASE 37 CASE 38
aircraft icing data available for
c m 0.53 0.53 0.152 0.152
comparison and validation. These
data include data from wind tunnel α ° 4.0 4.0 0.0 8.5
experiments as well as numerical V∞ m/s 58.1 58.1 130.5 130.5
codes like LEWICE, TURBICE,
T∞ K 245.2 266.3 260.7 260.7
FENSAP, etc. The cases picked for
the validation of the current code are P∞ kPa 95.61 95.61 90.50 90.50
given in table 2. The results are LWC g/m 3
1.3 1.3 0.5 0.5
shown in figures Figure 7 to Figure
10. The results are compare with MVD µm 20 20 17.5 17.5
experimental data and data from Texp s 480 480 120 120
LEWICE [13]. The results are
acceptable within numerical
inaccuracies. The ice shapes are not perfectly in agreement with the experimental results because of the inherent
shortcomings of the model as well as numerical inaccuracies. Nevertheless, encouraging agreement with test data is
found.
10
0.05 0.05
0 0
-0.05 -0.05
-0.05 0 0.05 0.1 0.15 -0.05 0 0.05 0.1 0.15
Current LEWICE Experimental Current LEWICE Experimental

Figure 7: Ice shape Case 27. Figure 8: Ice shape Case 28.
0.015
0.015
0 0
-0.015
-0.015
-0.01 0 0.01 0.02 0.03
-0.01 0 0.01 0.02 0.03
Current Experimental LEWICE Current Experimental LEWICE
Figure 9: Ice shape for Case 37. Figure 10: Ice shape for Case 38.
VI. Case Studies
As a test case, the NREL Phase

VI rotor was studied for a set of Table 2: Conditions for cases studies
ambient conditions, given in Table
Parameters φ RPM V∞ T∞ P∞ LWC MVD Texp
2. The blade uses an S809 airfoil
with a nonlinear twist and linear Units ° RPM m/s K kPa g/m3 µm s
taper. The 2-bladed wind turbine
rotated at a constant speed of 72 Case I 3.0 72 10 270.4 101325 0.22 20 1800
RPM with a 3 degree pitch and a Case II 2.0 72 10 270.4 101325 0.22 20 1800
stall regulated motor. The
turbine’s radius from the center of Case III 6.0 72 10 270.4 101325 0.22 20 1800
rotation was 5.03 m which
includes both the blade and the hub.
A. Radial Distribution of the Collection Efficiency
The collection efficiency is a function of the chord length, angle of attack, free stream velocity and the droplet
MVD [16]. Except for the MVD, all the other three parameters vary along the radius of the blade. Thus each section
will have a different collection efficiency depending on their radial location, because of the taper, the chord decreases,
as do the angle of attack. The distribution of collection efficiency for Case III is shown in figure below. In the figure
11
below, the x-axis is the radial locations and y-axis is the length along the chord non-dimensionalized by the chord
starting from the leading edge with the lower surface as negative. The z-axis gives the collection efficiency. There is
no catchment of water droplets on the blade except beyond 75% radius.
Figure 11: Distribution of collection along the radius.

B. Ice Shapes
For case I and II, only two sections, at 92.5% and 97.5% radii, accumulate ice mainly on the bottom surface, and
shall not be discussed further. Case III is the worst case with five outboard iced sections starting at 77.5% radius. For
section at 77.5 % radius, there is a little droplet catchment and hence a concentrated mass of ice on the bottom surface.
The other four sections, at 82.5%, 87.5%, 92.5% and 97.5% radii, gets significant amount of ice, shown in Figure 12.
The ice layers start at the leading edge and extend on the lower surface.
82.5% 87.5%
0.15 0.15
0.05 0.05
-0.05
y/c
y/c
-0.05
-0.15
-0.05 0.15 0.35 0.55 -0.15
-0.05 0.15 0.35 0.55
x/c x/c
12
92.5% 97.5%
0.15 0.15
0.05 0.05
-0.05
y/c
y/c
-0.05
-0.15 -0.15
-0.05 0.15 0.35 0.55 -0.06 0.05 0.15 0.25 0.35 0.45
x/c x/c
Figure 12: Ice shapes for sections at 82.5%, 87.5%, 92.5% and 97.5% radii.
C. Power Losses
Dramatic loss in lift have been observed for the four sections because of the sharp ice layer at the leading edge.
The performance of the four out board sections is reduced to power consumption rather than producing power. The
net effect is around 60% of power loss at the rated operating conditions. The loss of lift may be attributed to the early
separation because of the leading edge ice. The increase in drag is a result of both separation on the upper surface and
increased skin friction on the lower surface. The roughness of the ice layer causes the skin friction to rise.
Sectional Power Power Curve

1500 12
Clean
Case III 10
1000
8
Power (kW)
Power (W)
500 Clean
6 Case III
0
4
-500
2
-1000 0
0.25 0.50 0.75 1.00 5 6 7 8 9 10 11 12
r/R Wind Speed (m/s)
Figure 13: Sectional power loss at 10 m/s wind speed Figure 14: Power curve at 3° pitch angle.
and 3° pitch angle.
VII. Discussion
The development of a stand-alone and dedicated tool to predict the ice accumulation on wind turbine blades will
help in ensuring the continued operation in cold climates. This will enable the wind energy industry to take a priori
measures in a cost efficient manner. The inclusion of weather data in the simulation will make the tool high fidelity.
The current study lays out the basic structure of such a numerical tool. So far, the results obtain are reasonable enough
within numerical inaccuracies which can be improved by using more robust and efficient algorithms. Moreover, the
following observations can be stated based on the study.
13
A. Computational Cost
As for every computational setup, it is necessary to keep track of the computational time, the computation time for
the problem at hand has also been investigated. Table 3 presents the breakdown of the computation time for each
module for a typical simulation with 4 layers and 15 radial sections of the blade. The breakdown shows the
computation of collection efficiency to be the most expensive one. It is because of the time stepping nature of the
problem formulation as the calculations involves marching in time throughout the flow domain.
Table 3: Computational cost breakdown.
Module Computation Time (s) Percentage of Total Time

Blade Element Momentum Theory 5.8 0.05%
Panel Method 166 1.40%
Convective Heat Transfer Coefficient 4.3 0.04%

Collection Efficiency 11566 97.65%
Extended Messinger Model 2.28 0.02%
Miscellaneous 100.2 0.85%
Total 11844.7 100%
B. Critical Locations
As have discussed earlier, the water droplet collection efficiency is dependent on the chord length, angle of attack
and velocity at a given location. These parameters varying along the span of the wind turbine blade. The combination
of these parameters for the inboard sections are not favorable for the catchment of the droplets. On the contrary, they
form a favorable combination on the outboard sections. Thus, the outboard sections of the blade are more sensitive to
ice accretion than the inner sections. Also, as the contribution of outboard sections in power produciotn is prominent,
the loss of power generation at these sections is felt as a heavy penalty.
Iced formed on the leading edge is high detrimental for the power production as it completely changes the flow
over the upper surface. So, if somehow the leading edge is protected from ice accumulation, the performance
degradation may be reduced.
C. Sensitivity
For the current study, only variation in pitch angle was considered. The collection efficiency is heavily dependent
on the pitch angle and hence the overall ice accumulation. The study shows a power loss of 10% to 60% for different
pitch settings. So, it is apparent from this study that controlling the pitch may be very effective way of avoiding the
ice accumulation on the turbine blades.
VIII. Recommendations
Based on the observations in this study, it is recommended that;
1) The code may further be developed to enable its used for unsteady ambient conditions which will make it
possible to predict the performance evaluation of wind turbines operating in cold climates using the weather
forecast.
2) The calculation of collection efficiency consumes most of the time, so it is strongly recommended to use
alternate methods, if the simulations are run for longer period of time and for higher number of steps.
3) Operating the wind turbines at low pitch angles during the icing climate may reduce the chances of ice
accumulation.
4) If anti-icing or de-icing is sought, they should be place outboard and towards the leading edge of the airfoil.
5) The data generated should be used to determine the effects on the structural dynamics of the blades which in
turn will predict the fatigue life of the blades.
14
Acknowledgments
I would like to thank US Department of State and the Fulbright Commission for granting financial assistance for
my study program which also extends to my program officers and the staff members at both the Institute of
International Education (IIE) and United States Education Foundation in Pakistan (USEFP) who played a vital role in
making my dream of graduate studies a reality.
References
1. Dalili, N., A. Edrisy, and R. Carriveau, A review of surface engineering issues critical to wind turbine
performance. Renewable and Sustainable Energy Reviews, 2009. 13(2): p. 428-438.
2. Ronsten, G., et al., State-of-the-art of Wind Energy in Cold Climates. IEA Wind Task XIX, VTT, Finland,
2012.
3. Pedersen, M.C. and C. Yin, Preliminary modelling study of ice accretion on wind turbines. Energy Procedia,
2014. 61: p. 258-261.

4. Makkonen, L., Models for the Growth of Rime, Glaze, Icicles and Wet Snow on Structures. 2000, The Royal
Society. p. 2913.
5. Switchenko, D., et al. FENSAP-ICE Simulation of Complex Wind Turbine Icing Events, and Comparison to
Observed Performance Data. in 32nd ASME Wind Energy Symposium. Place of Publication: Reston, VA,
USA; National Harbor, MD, USA. Country of Publication: USA.: American Institute of Aeronautics and
Astronautics.
6. Ping, F. and M. Farzaneh, A CFD approach for modeling the rime-ice accretion process on a horizontal-axis
wind turbine. Journal of Wind Engineering and Industrial Aerodynamics, 2010. 98(4): p. 181-188.
7. Messinger, B.L., Equilibrium temperature of an unheated icing surface as a function of air speed. Journal of
the Aeronautical Sciences., 1953. 20: p. 29-42.
8. Myers, T.G., Extension to the Messinger Model for Aircraft Icing. AIAA Journal, 2001. 39(2): p. 211.
9. Moriarty, P.J. and A.C. Hansen, AeroDyn theory manual. 2005.
10. Giguere, P. and M. Selig, Design of a tapered and twisted blade for the NREL combined experiment rotor.
NREL/SR, 1999. 500: p. 26173.
11. Moran, J., An introduction to theoretical and computational aerodynamics / Jack Moran. 1984: New York :
Wiley, c1984.
12. Houghton, E.L., Aerodynamics for engineering students. 5th ed. ed, ed. P.W. Carpenter. 2003, Butterworth-
Heinemann: Oxford ;.
13. Wright, W.B., R.W. Gent, and D. Guffond, DRA/NASA/ONERA collaboration on icing research. Part II,
Prediction of airfoil ice accretion [microform] / William B. Wright, R.W. Gent, Didier Guffond. NASA
contractor report: 202349. 1997: [Washington, DC : National Aeronautics and Space Administration ;
Springfield, Va. : National Technical Information Service, distributor, 1997].
14. Ozgan, S., Cambek, M., Ice accretion simulation on multi-element airfoils using extended Messinger model.
Heat & Mass Transfer, 2009. 45(3): p. 305-322.
15. Kim, J., et al. Ice Accretion Modeling Using an Eulerian Approach for Droplet Impingement. 2013. Red
Hook NY, Curran Associates.
16. Gent, R.W., N.P. Dart, and J.T. Cansdale, Aircraft Icing. 2000, The Royal Society. p. 2873.
15

Ice Accretion

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ice Accretion

Uploaded by

Copyright:

Available Formats

AIAA 2015-3715

AIAA Propulsion and Energy Forum

In-Cloud Ice Accretion Modeling on Wind Turbine Blades

Ali, M. Anttho1 and Lakshmi, N. Sankar 2

II. Problem Definition

III. Problem Formulation

Messinger Model Droplet Solver

Figure 1: Schematic of the solution procedure.

Figure 2: Flow Velocities and Angles at a Blade Section

The induction factors are then given by

Wind Speed (m/s)

B. Flow Field Calculations

Pressure Distribution AoA = 4° Pressure Distribution AoA = 8°

Figure 4: Pressure distribution over NACA 0012

where the rouhgness height ks [13] is estamited as

𝑆𝑡𝑘 = 1.92𝑅𝑒𝑘−0.45 𝑃𝑟 −0.8 (18)

𝑚𝑥̈ 𝑝 = −𝐷 𝑐𝑜𝑠 𝛾 (21)

𝑚𝑦̈𝑝 = −𝐷 𝑠𝑖𝑛 𝛾 + 𝑚𝑔 (22)

1 + 0.197𝑅𝑒 0.63 + 2.6 × 10−4 𝑅𝑒 1.38 , 𝑅𝑒 ≤ 3500

Figure 5: Definition of collection efficiency.

0.7 Current 0.7 Current

(b) MVD = 20 μm, c = 0.53 m, α = 3.5°, V∞ = 102.8 m/s

E. Heat and Mass balance

𝑄1 = 𝜌𝑎 𝛽𝑉∞ 𝐶𝑝𝑤 + ℎ𝑐 + 4𝜖𝜎𝑟 𝑇∞3 + 𝜒𝑒 𝑒0 + 𝑚̇𝑖𝑛 𝐶𝑝𝑤

Runback water mass flow rate is given by

In flight icing on both aircrafts

Current LEWICE Experimental Current LEWICE Experimental

VI. Case Studies

As a test case, the NREL Phase

A. Radial Distribution of the Collection Efficiency

Figure 11: Distribution of collection along the radius.

Sectional Power Power Curve

Table 3: Computational cost breakdown.

Module Computation Time (s) Percentage of Total Time

Convective Heat Transfer Coefficient 4.3 0.04%

2014. 61: p. 258-261.

You might also like