American Institute of Aeronautics and Astronautics

1

An Experimental Study on Icing Physics for Wind Turbine

Icing Mitigation

Linyue Gao1, Yang Liu

2, Hui Hu

3()

Department of Aerospace Engineering, Iowa State University, Ames, Iowa, 50011

An experimental study was conducted to investigate the dynamic ice accretion process over

the surface of a wind turbine blade with DU96-W-180 airfoil profile in order to elucidate the

underlying physics for wind turbine icing mitigation. The experimental study was conducted

in the Icing Research Tunnel of Iowa State University (ISU-IRT). Six test cases with three

typical liquid water content (LWC) levels (i.e., LWC =0.3 g/m3, 1.1 g/m

3, and 3.0 g/m

3) and

two typical oncoming airflow temperatures (i.e., T∞ = -5 oC, and -10

oC) were selected to

represent rime icing, mixed icing and glaze icing conditions that wind turbines may

experience in winter. In addition to using a high-speed imaging system to quantify the

transient behavior of surface water runback and dynamic ice accretion processes over the

airfoil surface, an infrared (IR) thermal imaging system was also utilized to simultaneously

measure the airfoil surface temperature distributions in the course of the ice accreting

process. The measurement results reveal clearly that, both the length of the surface water/ice

film formed near the airfoil leading edge and the subsequent water/ice rivulet length would

increase as the liquid water content level and the temperature of the oncoming airflow

increase. The thickness of the ice accreted at the airfoil leading edge was found to grow

rapidly with the increasing liquid water content level and decreasing oncoming airflow

temperature for both glaze and mixed icing test cases. The temperature distribution over the

airfoil surface was also found to vary significantly, depending on the type of ice accreted

over the airfoil surface. A stream-wise ‘plateau’ region was observed in the measured

surface temperature distributions for the glaze and mixed icing cases, due to the formation

of surface water runback over the ice accreting airfoil surface.

Nomenclature AoA = angle of attack

C = chord length

IEA = International Energy Agency

IR = infrared

ISU-IRT = Icing Research Tunnel at Iowa State University

LE = leading edge

L, Lf, Lr, LLE = length, stream-wise lengths of ice film, ice rivulet, and ice accretion on LE

LWC = liquid water content

MVD = median volumetric diameter

Q = heat flux

RTV = relative temperature variation

T, T∞ = temperature, ambient temperature

t = time

TE = trailing edge

V∞ = oncoming flow velocity

X,Y = span-wise and stream-wise coordinates

2D, 3D = two-dimensional, three-dimensional

1 Graduate Student, Department of Aerospace Engineering.

2 Postdoctoral Associate, Department of Aerospace Engineering.

3 Martin C. Jischke Professor, Dept. of Aerospace Engineering, AIAA Associate Fellow, Email: huhui@iastate.edu.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

35th Wind Energy Symposium

9 - 13 January 2017, Grapevine, Texas

AIAA 2017-0918

Copyright © 2017 by Linyue Gao, Yang Liu, Hui Hu. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

AIAA SciTech Forum

American Institute of Aeronautics and Astronautics

2

I. Introduction

Wind turbines in cold climate are becoming one of the most popular trends for wind energy developments. Cold

climate areas retain wind resources superiority ascribed to the increased air density and wind speeds, providing a

vast market potential, particularly in offshore and hilly projects. But, meanwhile, cold climate exerts profound

negative effects on the aerodynamic performance and lifespan of wind turbines due to the occurrence of icing

events. According to IEA Icing Climate site classification1, five classes are recommended on behalf of the severity

of the icing events whose annual production losses are ranging from less than 0.5% to over 20%. One reason for the

loss of energy production would be the ice accretion on wind turbine blades. The ice accumulation enlarges the

blades’ surface roughness and, more seriously, changes the geometric shapes, which perturbs the aerodynamic

performance of wind turbines2,3

. The downtimes would be the second reason4. Wind turbines without the icing

mitigation equipment have to shut down during the icing events to prevent mechanical damages. The downtimes

usually last days, even weeks, associate with favorably high wind speeds, resulting in the most acute energy loss. In

addition to the power loss, the lifespan of wind turbines subjected to the icing events would be shortened to different

degrees due to the additional loads, unbalance, and edgewise vibration5 or resonance

6. Furthermore, the secondary

induced effects of wind turbine icing, including increased noise and projected ice chunks7, would impair human’s

health and cause damages to their personal and property security.

In order to reduce these dampening effects caused by icing events, multifarious methods for wind turbine icing

mitigation have been proposed and some of them have been commercially utilized. These de-/anti-icing protection

techniques comprise passive and active modes. Hydrophobic/ice phobic coatings8–13

, black paint6 and operational

stops14,15

are most commonly-used passive ice protections, while active ice protections can be subdivided into

chemical, mechanical, thermal, and active pitch control methods. In IEA Ice Class 1 and 2, black paint and active

pitch control are comparably effective and efficient due to their low costs and energy consumption6,16

. For more

intensive icing events, such as those in IEA Ice Class 3, 4, and 5, inside/outside resistive heaters with/without

hydrophobic coatings are most promising in both new and existing wind farms16

. The resistive heaters have both

anti-/de-icing capabilities, very effective but expensive. They are ordinarily installed covering 1/3 leading edge

(LE), 2/3 LE or all blades16

. Recommendations of their mounting positions are imperatively needed, which is greatly

based on comprehensive perceptions with respect to icing physics.

Recent years, researchers pay more attentions to wind turbine icing. Yirtici17

estimated energy production losses

of wind turbines by using Blade Element Momentum theory with a 2D-based ice accretion prediction tool. Hudecz18

numerically analyzed the 2D iced profiles gained from wind tunnel tests by using ANSYS Fluent and then their iced

profiles were further compared with those generated in TURBICE, a numerical ice accretion software from VTT.

Li19

investigated icing area and icing rate considering different angles of attack (AoA) and wind speeds through a

wind tunnel experimental study, and they found that the maximum icing area ratio of icing area against airfoil area is

21.8% at AoA = -80 °. Kraj8 experimentally simulate four typical phases of icing on wind turbine blades, including

the pre-icing, operational-icing, stopped-icing and post-icing phases and further compared the effects of coatings,

heat treatments and some combination thereof. These studies provide variously practicable contributions to the wind

turbine icing issue. However, literatures of more fundamental icing physics are mostly related to aircraft icing20–22

.

Few of them are in the wind turbine field. The significant differences between wind turbines and aircrafts cannot be

neglected. Wind turbines usually operate beneath the Ekman layer, where climate, underlying surface conditions,

and the rainfall intensity, etc. are distinct from those aircrafts may experience. Besides, wind turbine dedicated

airfoils, different from aircraft airfoils, e.g. NACA0012, are thicker and asymmetric, possessing particular

aerodynamic, structural and thermal properties, which would affect icing formation and accretion process. So that an

investigation in terms of the underlying physics during the process of supercooled water droplets impingement,

water runback and ice accumulation, as well as the related thermal phenomena, over a wind turbine blade is valuable

and highly desired for enhancing the fundamental comprehension and providing references on the icing mitigation.

The widely-used icing models, i.e. LEWISE and TURBICE, are proved successful to generate 2D ice shapes

under certain icing conditions, followed with CFD analysis for aerodynamic performance23

. Based upon the given

initial and boundary information, they are efficient to some extent to yield acceptable predictions to meet the

industry requirements, especially in rime ice conditions. But, for more complicated and commonplace icing

conditions, such as glaze mixed icing, experimental techniques have undoubted advantages in providing more

accurate and extensive data to dig into the meticulous physical phenomena.

Therefore, we have conducted a series of experimental tests in the Icing Research Tunnel at Iowa State

University (ISU-IRT). In this paper, we concretely focus on the full life-cycle icing process over a DU96-W-180

airfoil model, as well as the thermal characteristics, to give a better understanding of the underlying physics for wind

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

3

turbine icing. The high speeding imaging technique was used to investigate the transient water runback and dynamic

ice accretion process. Simultaneously, an infrared (IR) imaging thermometry technique was developed and utilized

to quantitatively measure the variations of temperature distributions on the pressure-side surface of the airfoil model.

A thorough experimental setup and case design are illustrated in the Section II. Six cases with two typical ambient

temperatures and three LWC levels were conducted, compared and analyzed in terms of the dynamic behaviors of

windward ice accumulation on the LE, water/ice film and rivulet, and surface temperature variations in Section III.

Section IV concludes the main findings.

II. Experimental Setup and Case Study A. Experimental Setup

Figure 1 shows the configuration of experimental setup for measuring the dynamic ice accreting process and

surface water runback over the DU96-W-180 model in ISU-IRT. The ISU-IRT, a research-grade multifunctional

icing research tunnel, has a capability of generating a maximum wind speed of 60 m/s and a minimum air

temperature of -25 °C. The test section is of 400 mm × 400 mm × 2,000 mm in height, width and length,

respectively, with an approximately 3.0% turbulence level of oncoming flow at its entrance. All the walls of the test

section are optically transparent and interchangeable for various test requirements. The spray system, installed at the

entrance of the contraction section, consists arrays of pneumatic atomizing spray nozzles (Spraying Systems Co.,

1/8NPT-SU11). This system can inject micro-sized water droplets (10 ~ 100 µm in size) into the test section. The

desired LWC levels in the wind tunnel is achieved by regulating the water flow rate while the needed medium

volumetric diameter (MVD) of the oncoming droplets is provided by adjusting the air/water pressure thorough the

spray nozzles. Various icing conditions have been simulated/duplicated in ISU-IRT and a series of

experiments21,22,24

have been successfully completed, which provides an excellent icing environment for wind

turbine icing physics studies.

Figure 1. Schematic diagram of experimental setup.

The high speed camera (PCO Tech, Dimax), aiming at capturing the whole icing dynamic process over the

model, was mounted right above the model with a 24 mm lens (Nikon, 24 mm Nikkor 2.8D). It provides a top view

of 528 × 888 pixels with a resolution of 5.08 pixels/mm. The frame rate of high speed camera is set as 30 frames per

second, the same with the IR camera, triggered by a digital delay/pulse generator (BNC, Model 577) to

simultaneously record the icing process. IR camera (FLIR, A615) and IR inspection window (FLIR, IR Window-

IRW-4C) constitute the IR system for temperature measurements. The IR camera can detect waves within a range of

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

4

7.5-14 µm. It was fixed above the model, providing a field of 640 × 480 pixels with a resolution of 4.77 pixels/mm.

The IR window was embedded in the top panel of the test section. Three materials relevant to the tests are White

Protective Enamel covering the airfoil model surface, water and ice. Their emissivity coefficients are 0.945, 0.980

and 0.960, respectively, which are considered in the temperature calibration of the IR system. A surface

thermocouple (Omega), taped onto the model surface and connected to a J-type thermocouple (Omega), was also

used to record the one-point temperature on the model surface during the test to guarantee that the surface

temperature equals to that of the ambient airflow before each test. The results of this thermocouple were validated

by comparing with the data gathered by IR camera. They are in good agreement and linearly fitted with an

uncertainty less than ± 0.1 °C.

The test model was designed based on a wind turbine dedicated airfoil, DU96-W-180, with dimensions of 152.4

mm and 400 mm in stream-wise and span-wise, respectively, as shown in Fig. 2. DU96-W-180 is an asymmetric,

cambered, low-speed airfoil profile with a thickness-to-length ratio of 0.18, which is widely applied to make the

wind turbine blades as outboard airfoil due to its high lift-to-drag ratio, insensitivity to roughness and low noise25,26

.

It is representative and thus was selected as the objective in this study. The test model, supported by three stainless

steel rods, was mounted at its quarter-chord across the middle of the test section with a desired AoA by using a

digital inclinometer. The model was produced by using the 3D printing technique with a hard plastic material, Vero

White. Its surfaces were coated with Primer and White Protective Enamel, and then polished with up to 2,000 grit

sandpapers to assure the sufficient smoothness.

Figure 2. DU96-W-180 airfoil model. (a) Geometry of DU96-W-180 Profile; (b) Lateral view; (c) Top view.

B. Case Study

Table 1 shows the case design for icing physics study based on DU96-W-300 model. As an outboard airfoil,

DU96-W-180 possesses a relatively high lift-to-drag ratio when its lift-to-AoA curve situates at the linear growth

period where AoA is from 0 ° to 10 °, and one of them AoA = 5 ° is selected for the tests. For most multi-megawatt

wind turbines, their cut-in, rated and cut-out wind speeds, as well as rotor rated rotational speed, are around 3 m/s,

10 ~ 12 m/s, 22 ~ 25 m/s, and 12 ~ 16 rpm, respectively. The relative velocity of a blade element is the resultant

velocity of absolutely oncoming flow velocity and blade rotational linear velocity. The rotational linear velocity is

the product of rotor rotational speed and the local radius. GW87/1500 wind turbine, for example, has a tip speed

ranging from 40 m/s to 80 m/s. DU96-W-180 is designed to be mounted at 60% ~ 99% span-wise position, where

the relative velocity ranges approximately from 24 m/s to 80 m/s, so that the oncoming flow velocity in this study is

set as V∞ = 40 m/s covering the majority situation. MVD is controlled around 10 ~ 100 µm for six cases, which

satisfies the size of impinging water droplets, ranging from 10 µm to 150 µm, onto the wind turbine blades in the

real field environment17,27

. The ice accretion process over the pressure-side surface of the airfoil is highlighted

because the impinging super-cooled water droplets exert more significant influence on it in comparison with the

suction-side surface.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

5

Besides the parameters mentioned above, ambient temperature and LWC level are the most critical and decisive

factors for the icing issue. Three typical LWC levels, 0.3 g/m3, 1.1 g/m

3 and 3.0 g/m

3, are selected to show the low,

medium and high LWC situations. Two temperatures, -5 °C and -10 °C, are chosen. There are mainly two types of

ice, rime ice and glaze ice. Rime ice is formed when supercooled water droplets hit a surface and freeze

immediately, while glaze ice is constructed when the water droplets that hit the surface do not freeze completely and

the non-frozen water flow over the object1, leading rivulet formation. Compared to glaze icing condition, rime icing

usually occurs at a relative lower temperature with a smaller LWC level. There is also a transition ice type between

rime and glaze ice, named mixed ice. Given the oncoming flow velocity, ambient temperature and LWC level, as

well as the MVD, three typical types of icing conditions, rime, mixed and glaze icing, are expected to be observed

among the following six cases, which will be further validated with the obtained data.

Table 1. Case design for icing physics study on DU96-W-180 airfoil.

Case No. AoA [°] V∞ [m/s] T∞ [°C] (± 1.5 °C) LWC [g/m3] Expected Ice type

1 5 40 -5 0.3 Glaze

2 5 40 -10 0.3 Rime

3 5 40 -5 1.1 Glaze

4 5 40 -10 1.1 Mixed

5 5 40 -5 3.0 Glaze

6 5 40 -10 3.0 Glaze

III. Results and Discussions

To give a concrete to integral concept of the icing process, the dynamic icing process under various ice types is

elaborately illustrated, followed by the time evolutions of surface temperature distributions, dynamic ice accretion

upon the LE, and quantification of the icing impacted regions. And then icing process mechanism is summarized

and discussed.

A. Dynamic Icing Process

Figure 3 shows time evolutions of the dynamic ice accretion process among the three cases with an ambient

temperature of -5 °C. From Fig. 3 (a) to Fig. 3 (c), LWC level increases, and seven typical moments are selected.

Both LE and trailing edge (TE) are portrayed in the figure to help identify the special scale. When t = 0 s, the water

droplets are located upstream near the airfoil model without impinging onto it. The snapshot previous to the first

with water droplets impingement is identified as the beginning when t = 0 s.

It can be seen clearly that, supercooled water droplets impinge onto the airfoil surface and form a thin water film

that runs back over the airfoil surface. The film expands downstream and develops into isolated water rivulets

further downstream, indicating all of these three cases are in glaze icing conditions. Meanwhile, the water film and

rivulets freeze into ice film and ice rivulets. The building up ice icicles block the subsequent water runback and lead

to ice accretion concentrated near the LE. Beyond the basic process, more intuitive ice accreting process are shown.

Two features should be highlighted. The first one is the distinction of water and ice film and rivulets. Water film and

rivulets can be driven by the boundary layer flow due to their liquidity. For three cases, water film splits into rivulets

before 15 s and then continues expanding over the airfoil surface. Approximately at t = 80 s, water film and rivulets

almost completely convert into ice film and rivulets, and ice starts to primarily accumulate upon the region near LE.

Water film/rivulets are adaptable to investigate the transient behavior while ice film/rivulets are more suitable for

identifying of the icing impacted area. Another one is the ice configuration. As LWC level increases, an increasing

number of supercooled water droplets are impinging onto the airfoil, which leads to an enhancement in liquidity and

further results in more complex and coarser surface topography due to the increased uncertainty of ice accumulation.

Meanwhile, more crystal-clear ice deposits with tiny gaps in cases with relatively larger LWC levels are observed. In

Case No.5 where LWC = 3.0 g/m3, icicles can be seen standing closely side by side towards the upstream direction.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

6

Compared to other cases, it displays several complete water channels from LE to TE. These channels are much

thicker than the rivulets with dead ends shown in Case No.1 and Case No.3 due to the increased LWC level.

Figure 3. Snapshots of dynamic ice accretion process on DU96-W-180 airfoil where AoA = 5°, V∞ = 40 m/s.

Compared to the cases shown in Fig.3, the cases in Fig. 4 have a lower ambient temperature. Due to temperature

change, the appearance of ice accretion on the airfoil surface changes to a great extent. Only in Case No.6, rivulets

are formed and developed, but they are much thinner than those in Case No.5. Lower temperature would lead to a

fast freezing and thus block water runback, as well as the rivulet formation and extension. In Case No.2 and Case

No.4, there is no obvious rivulets. Case No.2 has a uniform ice layer with tiny structures, prone to be rime ice, the

same as expected. Compared to it, Case No.4 has a severe ice accretion, and the configuration of ice aggregated on

airfoil surface is quite different from that in Case No.2. Due to the unique characteristics, Case No.4 pertains to

mixed icing condition where the icicles stand on the surface towards the flow direction with a compact structure.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

7

Therefore, due to the combined action of ambient temperature and LWC level, the type of icing changes.

Additionally, ice accretion on LE can be observed clearly when t >80 s. But within the first 15 s, the snapshots have

a limitation to show more information due to the tiny changes, especially in Case No.2 and Case No.4.

Figure 4. Snapshots of dynamic ice accretion process on DU96-W-180 airfoil where AoA = 5°, V∞ = 40 m/s.

B. Temperature Variation Analysis

The icing process is along with complicated heat transfers. When the supercooled water droplets impinge onto

the airfoil, they entirely or partially solidify into ice, releasing the latent heat. The heat is immediately absorbed by

the upper unfrozen water layer, leading to a temperature raise in the water/air or ice/air interfaces, and then transfers

to the ambient flows mainly by convection. Due to the surface temperature alteration during the icing process, the

temperature contours over the surface are applied to provide a thorough view of the icing development.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

8

Figure 5 compares three typical moments of six cases. The increments of LWC level accompanies an

enlargement of icing impacted areas. At t = 8 s, this phenomenon can be observed more clearly. The rivulets that

start to form, stretch and advance downstream beyond 0.5C are shown in Fig. 5 (a), (c) and (e). A higher

temperature occurs near the LE due to a locally larger water collection and freezing rate. The impact of ambient

temperature on the surface temperature distributions can be seen directly, such as the absence of rivulet formation,

but due to the different temperature scales, they cannot be quantitatively compared, which would be further

discussed in the following section. Two interesting phenomena can be observed. Firstly, IR images show a nonlinear

temperature gradient along the chord, particularly in the cases with rivulet formation, which will be further

explained in the next section. Secondly, in the cases without rivulet formation, Case No.2 and Case No.4, their time

evolutions of surface temperature along the chord are also different. A clear expansion of surface temperature

variation in stream-wise is shown in Case No.4 due to the water runback above the ice layer, which help validate

that Case No.4 is in the mixed icing condition while Case No.2 is in the rime icing condition.

Figure 5. Time evolutions of surface temperature distribution on DU96-W-180 airfoil model where AoA = 5°

and V∞ = 40 m/s.

To further analyze the dynamic process of surface temperature variation, time series of temperature distributions

are extracted and shown in Fig. 6. A ‘peak’ can be seen in each curve after a precipitous rise, representing the

stream-wise largest temperature. At downstream locations of ‘peak’, surface temperature drops gradually. The

‘peak’ occurs near the LE and slightly moves downstream as time goes on, indicating the location of largest water

collection (stagnation points) in case without water runback24

. As LWC level increases from 0.3 g/m3 to 3.0 g/m

3,

maximum ‘peak’ value of temperature increases correspondingly due to incremental amount of latent heat released

during the water solidification. The time used for reaching the maximum ‘peak’ temperature also reduces, and all of

the cases can reach stable ‘peak’ values within 10 s.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

9

Figure 6. Span-wise normalized surface temperature variations along the chord in six cases.

At the early time, the temperature decreases linearly and then a ‘plateau’ region appears in Case No.1, Case

No.3, Case No.4, Case No.5 and Case No.6. This phenomenon is greatly related to the rivulet formation and water

runback. When coming into the second stage, more oncoming water droplets supplement the water layer and the

water film expanding downwards breaks into several isolated rivulets by overwhelming the gravity and water

surface tension. To obtain sufficient surface stress driven by the boundary layer flow, large amount of water

accumulates at the certain position resulting in a thicker water layer and more latent heat release, which in turn

causes a rise of local temperature. From the eight moments selected in the Fig. 6 (a), (c) and (e) with the same

temperature, we can roughly pick up the time of occurrence of the ‘plateau’, t = 8 s, t = 3 s, and t = 2 s, respectively.

This time has a similar trend with the time arriving at the ‘peak’ value, depending on the LWC. In Fig. 6 (d), there is

no obvious bulge on the chord-wise temperature distribution, and it owns the largest ‘peak’ value among the cases

under the same temperature, even larger than the case with a higher LWC ascribed to the intensive ice building

around the LE. Similar findings18,20,28,29

about the more severe effects of mixed ice than both glaze and rime ice have

been presented. In Fig. 6 (b), generally, the curves at different moments are roughly parallel with each other due to

the almost evenly ice building up along the pressure-side airfoil surface, forming a smooth and streamline-shaped

ice layer and indicating a rime icing process. The slight difference between the chord-wise temperatures of two

moments decreases with the increased value of Y/C, opposite with other cases. In Case No.2, the water droplets

impinging onto the airfoil surface freeze immediately. The LE area has higher temperature due to larger water

droplets collection and as distance away from the LE increases, the capability of capturing water droplets smoothly

decreases with respect to the geometry and posture of airfoil model. For other cases, the enlargement of the

difference is caused by water runback. Therefore, according to time series of chord-wise surface temperature

variation, types of icing can be identified more directly than only using the configuration characteristics of them.

In order to achieve a clear comparison of temperature alterations among six cases, a parameter, Relative

temperature variation (RTV), is defined in Eq. (1). It cannot provide the actual value, but it is superior to be applied

to illustrate the dynamic icing process. For the icing conditions without rivulet runback phenomenon, the ice mass

change linearly corresponds to the temperature variation24

, and thus Eq. (1) can be seen as the water collection

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

10

efficiency ratio. But, for the icing conditions with rivulet formation, there is no linear relationship between the

variations of ice mass and temperature so that RTV is not equal to the water collection efficiency ratio. Due to the

icing conditions set up in this study, RTV is appropriate for the further comparison between various types of icing

conditions.

( )

( )

( )

max

t i

jt i

j t i

TRTV

T

(1)

where ( )

j

t i

T

is the difference between span-wise-averaged temperatures ( )

j

t i

T

and ( 0)

j

t

T

at chord-wise location

Y/C = j; ( )

max

t i

T

is the maximum value of ( )

j

t i

T

.

( ) ( ) ( 0)t i t i t

j j jT T T

(2)

( ) ( ) ( 0)

max

t i t i t

j peak j peakT T T

(3)

Figure 7. Comparison of time series of RTV among six cases at different stream-wise locations away from LE.

(a) Y/C = 0, LE; (b) Y/C = 0.1C; (c) Y/C = 0.2C; (d) Y/C = 0.3C; (e) Y/C = 0.4; (f) Y/C = 0.5.

Figure 7 shows the RTV variations at six typical locations on the airfoil model. According to the Eq. (1), (2) and

(3), ( 1) ( 1) ( 1) ( 1) ( 1) ( 1), and constantt t t t t t

j a j b j a j b j a j bRTV RTV RTV RTV RTV RTV

when a <b correspond to the growing,

decreasing and stable trends of curves in Fig. (7). Since the maxT occurs near the LE, RTVs are almost equal to

‘1.0’ at position Y/C = 0 in six cases. As the distance apart from LE increases, RTV values decrease and approach

‘0.0’ at Y/C = 0.5 where the water/ice have not completely developed. Within the time duration of 14 s, cases under

lower ambient temperature generally have larger RTV values than those cases with higher ambient temperature

because the low temperature aids to accelerate the latent heat release and convection between the ice/air or water/air

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

11

interfaces. The RTV values in most cases are fixed after t = 3 s at Y/C = 0.1, as shown in Fig. 7 (b), except for Case

No.1 whose RTV becomes stable when t >12 s. That time equals to the time consumed for reaching the maximum

surface temperature at a certain location. As Y/C increases, duration time for reaching the stable state increases

ascribed to the evolutions of water/ice film and rivulets.

C. Quantification of the Icing Impacted Region

Based on the underlying physics of dynamic icing process and surface temperature distributions, there are

mainly three icing impacted regions, ice accretion upwards the LE, ice film and runback rivulets. Quantification of

them can provide a practical reference for wind turbine icing mitigation, especially the optimization of resistant

heaters.

Figure 8 illustrates the icing impacted region and the parameters used to quantify them. LLE is the length of ice

accretion in the windward direction ahead the LE. Lf and Lr are the lengths of ice film and ice rivulet, respectively.

Notably, the Lr is determined by Eq. (4). As mentioned in Section III A, water rivulets run back over the airfoil

surface, which means the lengths of water rivulet would increase due to the wind-driven boundary layer flow. But

for the ice rivulets, their lengths are constant. Hence, the dimensions of ice film/rivulet, instead of those of water

film/rivulet, are used for the further comparison. Three special shapes of ice rivulets and their length measurement

methods are portrayed in Fig. 8 (b). If there is a large gap within one rivulet and the downstream part is small, its

length is determined as that of upstream part. If the rivulet develops and split into more than one branches, the

lengths of all branches are counted and measured from the dividing line of film/rivulet. If only part of the rivulet

appears in the image, it is not included in the calculation.

( )

1

ni

r

i

r

L

Ln

(4)

where (Lr)i is the length of ith

rivulet and n represents the number of rivulets shown in the image.

Figure 8. Dimensional parameters of icing impacted regions. (a) Main parameters; (b) Three special types of

ice rivulets.

Figure 9 shows ice accreting thickness variation on airfoil LE within 250 s in six cases. The boundary line of ice

is detected by using the method proposed in Ref. [21]. The data are averaged in span-wise and both the mean value

and standard deviation are presented. Generally, the time series of LLE/C are linearly increasing with time, especially

when t is larger than about 80 s. After that time, ice accretion concentrates upon the LE. The nonlinear section is

caused by the water runback over airfoil surface. The final LLE/C when t = 250 s are 2.67%, 1.43%, 3.91%, 5.43%,

4.81%, 7.17% for six cases, respectively. As LWC level increases, the ice thicknesses upon LE correspondingly

increase for both cases in T = −5 °C and T = −10 °C. Since the process of ice accumulation is approximatively

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

12

linear, the linear fitted curves are plotted. The normalized slope by using the slope in Case No.1 are 1.00, 0.54, 1.47,

1.89, 2.40, 3.01 for cases from Case No.1 to Case No.6, respectively. The slopes of the linear curves represent the

growth rate of the ice accretion on LE and inherit the similar trends with values of LLE/C, indicating that LWC aids

both increment of ice thickness and its growth rate. Besides LWC level, ambient temperature is another significant

factor. As shown in Fig. 9 (b), (c) and (d), the cases with lower ambient temperature are compared with the cases

with higher ones. Both Case No.5 and Case No.6 are in typical glaze icing conditions. The LLE/C is larger when

ambient temperature is lower mainly due to two reasons. Firstly, lower temperature leads to an inhibiting effect on

both water rivulet formation and evolution, resulting in the water/ice massing near LE. Secondly, lower temperature

accelerates heat transfers between the water layer and ambient air, which shortens the duration of water turning to

ice, so that water droplets are prone to building up at LE area. Similar phenomenon is observed in Case No.3 and

Case No.4, but differences between of these two cases are smaller. An opposite circumstance occurs in Fig. 9 (b)

where case with lower temperature owns a smaller ice increments on LE. In these two cases, Case No.1 is under

glaze icing condition while Case No.2 in rime icing condition. Due to the different types of icing, ice in Case No.2

with less bubble cells is denser than that in Case No.1 and it almost doesn’t have obvious gaps between the icicles

observed in glaze icing cases, so that the ice has smaller volume and thickness upon LE.

Figure 9. Comparison of ice accretion process on the LE among six cases. (a) Time series of span-wise-

averaged LLE/C; Span-wise-averaged LLE/C and linear fitted curves for (b) Case No.1 and Case No.2; (c) Case

No.3 and Case No.4; (d) Case No.5 and Case No.6. The error bands in light colors are ±1 standard deviation

around the mean.

Figure 10 shows the normalized lengths of ice films and ice rivulets in six cases when t = 300 s. For glaze icing

cases under the same temperature (Case No.1, Case No.3 and Case No.5), ice film length increases correspondingly

with the increment of LWC level. But no obvious pattern occurs in terms of ice film length in Case No.2, Case No.4

and Case No.6 and they are almost the same around 41% due to ice type change. Beyond the environmental factors,

the position of dividing line of ice film/rivulet might be influenced greatly by geometric shape of airfoil due to this

phenomenon.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

13

Four cases are compared in terms of ice rivulet dimension because Lr is zero in both Case No. 2 and Case No.4.

The number of rivulets are 10, 12, 9 and 15 for Case No.1, Case No.3, Case.5 and Case.6, respectively. When T∞ =

−5 °C, compared with Case No. 1, Lr increases 13.52% and 23.07% in Case No.3 and Case No. 5, respectively,

indicating that Lr increases as LWC level increases. The cumulative percentage of normalized length of ice film and

ice rivulet in Case No.5 is almost reaching 100% due to several completed channels from LE to TE observed in

Fig.3. When LWC levels are the same, such as in Case No.5 and Case No.6, the rivulet length is larger when

temperature increases ascribed to the longer duration of freezing. The lengths of ice films and ice rivulets are

relatively stable when the icing process completely moves into the third stage, and their lengths could be used as the

reference of heater design for wind turbine ice mitigation.

Figure 10. Comparison of both ice film length and ice rivulet length among six cases at t = 300 s.

D. Icing Process Mechanism

To have a more integral concept of the icing process, a full life-cycle icing process from both high-speed

imaging and IR thermal imaging results is extracted and portrayed in Fig. 11. The six cases can be grouped into two

types by their distinct features, i.e. with and without rivulets formation. As shown in the flow diagram, three stages

generally constitute the full life cycle of the glaze icing process. These three stages are not separated with clear

timelines, and they always overlap with their adjacent one or two stages. In the first stage, supercooled water

droplets carried by the oncoming airflow impinge onto the airfoil surface, particularly in the region near the airfoil

LE. The water droplets soon form a thin film expanding downstream along the surface. Notably, the water runback

happens in the upper/water layer of the film while the ice layer is located underneath it. As the film develops, it will

split into several isolated water rivulets stretching further downstream, which is the second stage. The lengths of the

rivulets greatly depend on the temperature, LWC level, and oncoming flow velocity, which is quantitatively

discussed in Section III C. If the lengths of the rivulets are large enough to reach the airfoil TE, water channels will

form and transport the water mass from the LE to the TE. Otherwise, they will freeze before the water mass can

shedding from the TE. After the rivulets formation, a retreating phenomenon is observed because the accumulated

small icicles on the surface impede the subsequent water droplets. The ice is building up around the LE in the

windward direction. For the rime and mixed icing process, there are only two stages without the second stage,

rivulet formation. To further distinguish the rime ice and mixed ice, we should focus on the water runback of the

water layer above the ice layer. For rime ice, there is no water runback while mixed ice has it. One method is to

observe the ice configuration, as mentioned in Section III A. Another method is based on the surface temperature

distribution, particularly the ‘plateau’ region, which is discussed in Section III B.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

14

Figure 11. Flow diagram for the full life cycle of the icing process over DU96-W-180 airfoil model with and

without rivulets formation.

IV. Conclusions

Icing process on a wind turbine dedicated airfoil model, DU96-W-180, have been experimentally investigated by

simultaneously using high speed imaging and IR imaging thermometry techniques. The full life cycle of icing

process can be roughly divided into three stages and they have overlaps across the timeline. The first stage is

oncoming water droplets impinging onto the airfoil surface followed by water film expansion. The second stage is

the rivulet formation, stretching and freezing. The third one highlights the ice accretion upon LE. Six cases, with

two ambient temperatures (-10 °C, -5 °C) and three LWC levels (0.3 g/m3, 1.1 g/m

3, 3.0 g/m

3), have been conducted

and compared to present the impact of ambient temperature, LWC level and icing conditions on the icing

performance on DU96-W-180 airfoil model. Among the six cases, three typical types of icing conditions are

included, one case for rime icing, one for mixed icing and the rest for glaze icing. For rime icing (Case No.2), there

are no rivulet formation and the ice layer poses a streamline configuration. For mixed icing (Case No.4), no rivulet

is observed under the conditions offered in this study, while icicles building up on the airfoil surface are transparent.

For glaze ice (Case No.1, Case No.3, Case No.5 and Case No.6), four cases with rivulet formation are analyzed and

compared.

The snapshots for dynamic icing accreting process are presented, followed by the quantitative analysis on the

impacted regions. In order to give a reference for wind turbine icing mitigation, the regions are further subdivided

into three parts with respect to the time evolution of dynamic icing process. They are ice film, ice rivulet and ice

accretion upon the LE in the windward direction. Their dimensions are calculated and provided. The 5-min ice film

length achieves about 41% for most cases, and larger LWC level and ambient temperature are inclined to enhance

the ice film length. The length of span-wise-averaged ice rivulets in glaze icing cases increases as LWC increases

and decreases as temperature decreases. Normalized ice accretion thickness on LE towards flow direction within

250 s increases linearly, especially after the first 80 s. As LWC level increases, the ice accretion thickness on LE

increases correspondingly regardless of the ice types. As ambient temperature decreases, the ice accretion thickness

upon LE increases due to the rapid solidification from water to ice for both glaze and mixed icing cases. Rime icing

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

15

case possesses the smallest value of ice accretion thickness upon LE due to its denser structure. The growth rate of

ice accretion thickness has the same trends. Besides, surface temperature distributions are also provided, as well as

the time series of temperature variations. The time series of temperature variations in stream-wise own different

features due to ice types, which can be used for ice type identification. The ‘plateau’ region observed only in glaze

and mixed icing cases is mainly caused by the water runback phenomenon.

In the future study, the thickness of water/ice film and rivulets will be detected to provide a thorough picture of

the icing effects on wind turbine blades.

Acknowledgments

The authors want to thank Dr. Rye Waldman and Mr. Linkai Li of Iowa State University for their help in

conducting for the present study. The support of National Science Foundation (NSF) under award numbers of

CBET- 1064196 and CBET-1435590 is gratefully acknowledged.

References 1 Lehtomäki, V., “Wind Energy in Cold Climates Available Technologies - report,” 2016. 2 Lamraoui, F., Fortin, G., Perron, J., and Benoit, R., “Canadian icing envelopes near the surface and its impact

on wind energy assessment,” Cold Regions Science and Technology, vol. 120, 2015, pp. 76–88. 3 Tammelin, B., Cavaliere, M., Holttinen, H., Morgan, C., Seifert, H., and Säntti, K., “Wind energy production in

cold climate (WECO),” ETSU contractor’s report W/11/00452/REP, UK DTI, 1999, pp. 1–38. 4 Lamraoui, F., Fortin, G., Benoit, R., Perron, J., and Masson, C., “Atmospheric icing impact on wind turbine

production,” Cold Regions Science and Technology, vol. 100, 2014, pp. 36–49. 5 Yin, C., Zhang, Z., Wang, Z., and Guo, H., “Numerical simulation and experimental validation of ultrasonic de-

icing system for wind turbine blade,” Applied Acoustics, vol. 114, 2016, pp. 19–26. 6 Seifert, H., “Technical Requirements for Rotor Blades Operating in Cold Climate,” Proceedings of Boreas VI,

2003, pp. 50–55. 7 Seifert, H., Westerhellweg, A., and Kröning, J., “Risk analysis of ice throw from wind turbines,” Paper

Presented at Boreas VI, 2003, pp. 1–9. 8 Kraj, A. G., and Bibeau, E. L., “Phases of icing on wind turbine blades characterized by ice accumulation,”

Renewable Energy, vol. 35, 2010, pp. 966–972. 9 Wang, Z. J., Kwon, D. J., Lawrence DeVries, K., and Park, J. M., “Frost formation and anti-icing performance

of a hydrophobic coating on aluminum,” Experimental Thermal and Fluid Science, vol. 60, 2015, pp. 132–137. 10

Peng, C., Xing, S., Yuan, Z., Xiao, J., Wang, C., and Zeng, J., “Preparation and anti-icing of superhydrophobic

PVDF coating on a wind turbine blade,” Applied Surface Science, vol. 259, 2012, pp. 764–768. 11

Liu, Q., Yang, Y., Huang, M., Zhou, Y., Liu, Y., and Liang, X., “Durability of a lubricant-infused Electrospray

Silicon Rubber surface as an anti-icing coating,” Applied Surface Science, vol. 346, 2015, pp. 68–76. 12

Lopera-Valle, A., and McDonald, A., “Flame-sprayed coatings as de-icing elements for fiber-reinforced

polymer composite structures: Modeling and experimentation,” International Journal of Heat and Mass

Transfer, vol. 97, 2016, pp. 56–65. 13

Arianpour, F., Farzaneh, M., and Kulinich, S. A., “Hydrophobic and ice-retarding properties of doped silicone

rubber coatings,” Applied Surface Science, vol. 265, 2013, pp. 546–552. 14

Pinar Pérez, J. M., García Márquez, F. P., and Ruiz Hernández, D., “Economic viability analysis for icing

blades detection in wind turbines,” Journal of Cleaner Production, vol. 135, 2016, pp. 1150–1160. 15

Baring-Gould, I., Tallhaug, L., Ronsten, G., Horbaty, R., Cattin, R., Laakso, T., Durstewitz, M., Lacroix, A.,

Peltola, E., and Wallenius, T., Expert Group Study on Recommendations for wind energy projects in cold

climates, 2009. 16

Fakorede, O., Feger, Z., Ibrahim, H., Ilinca, A., Perron, J., and Masson, C., “Ice protection systems for wind

turbines in cold climate: characteristics, comparisons and analysis,” Renewable and Sustainable Energy

Reviews, vol. 65, 2016, pp. 662–675.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918

American Institute of Aeronautics and Astronautics

16

17 Yirtici, O., Tuncer, I. H., and Ozgen, S., “Ice Accretion Prediction on Wind Turbines and Consequent Power

Losses,” Journal of Physics: Conference Series, vol. 753, 2016, p. 22022. 18

Hudecz, A., Koss, H., Hansen, M. O. L., “Ice Accretion on Wind Turbine Blades,” 2014, p. 115. 19

Li, Y., Tagawa, K., Feng, F., Li, Q., and He, Q., “A wind tunnel experimental study of icing on wind turbine

blade airfoil,” Energy Conversion and Management, vol. 85, 2014, pp. 591–595. 20

Gregorio, F., Ragni, A., Airoldi, M., and Romano, G., “PIV Investigation on Airfoil with Ice Accretions and

Resulting Performance Degradation,” Ieee, 2001, pp. 94–105. 21

Waldman, R., and Hu, H., “High-Speed Imaging to Quantify the Transient Ice Accretion Process on a NACA

0012 Airfoil,” 53rd AIAA Aerospace Sciences Meeting, 2015, p. AIAA 2015-0033. 22

Zhang, K., Wei, T., and Hu, H., “An experimental investigation on the surface water transport process over an

airfoil by using a digital image projection technique,” Experiments in Fluids, vol. 56, Aug. 2015, p. 173. 23

Pedersen, M. C., and Sørensen, H., “Towards a CFD Model for Prediction of Wind Turbine Power Losses due

to Icing in Cold Climate,” 2016, pp. 1–6. 24

Liu, Y., and Hu, H., “An Experimental Investigation on the Convective Heat Transfer Process over an Ice

Roughened Airfoil,” 54th AIAA Aerospace Sciences Meeting, vol. 2, 2016, pp. 1–31. 25

Mandell, J. F., Samborsky, D. D., and Miller, D. ., “Effects of resin and reinforcement variations on fatigue

resistance of wind turbine blades,” Advances in wind turbine blade design and materials, 2013, pp. 210–250. 26

Timmer, W. A., and van Rooij, R. P. J. O. M., “Summary of the Delft University Wind Turbine Dedicated

Airfoils,” Journal of Solar Energy Engineering, vol. 125, 2003, p. 488. 27

Agent, O., “Task 19 Work plan 2016-2018,” 2016. 28

Özgen, S., and Canıbek, M., “Ice accretion simulation on multi-element airfoils using extended Messinger

model,” Heat and Mass Transfer, vol. 45, 2009, pp. 305–322. 29

Han, Y., Palacios, J., and Schmitz, S., “Scaled ice accretion experiments on a rotating wind turbine blade,”

Journal of Wind Engineering and Industrial Aerodynamics, vol. 109, 2012, pp. 55–67.

Dow

nloa

ded

by I

OW

A S

TA

TE

UN

IVE

RSI

TY

on

Febr

uary

13,

201

7 | h

ttp://

arc.

aiaa

.org

| D

OI:

10.

2514

/6.2

017-

0918