You are on page 1of 11

Sample Cover

AIAA 2003-1076
Prediction of 2D Airfoil Ice Accretion
by Bisection Method And by Rivulets
and Beads Modeling
G. Fortin and J.-L. Laforte
Universite du Quebec a Chicoutimi
Quebec, Canada
A. Ilinca
Université du Quebec a Rimouski
Quebec, Canada
V. Brandi
Centro Italiano Ricerche Aerospaziali
Capua, Italy

39th Aerospace Sciences Meeting & Exhibit


12–15 January 2002
Reno, Nevada

For permission to copy or to republish, contact the copyright owner named on the first page.
For AIAA-held copyright, write to AIAA Permissions Department,
1801 Alexander Bell Drive, Suite 500, Reston, VA, 20191-4344.
41st Aerospace Sciences Meeting and Exhibit AIAA 2003-1076
6-9 January 2003, Reno, Nevada

PREDICTION OF 2D AIRFOIL ICE ACCRETION BY BISECTION METHOD


AND BY RIVULETS AND BEADS MODELING
Guy Fortin*, Adrian Ilinca†, Jean-Louis Laforte‡, Vincenzo Brandi§

ABSTRACT
The paper presents recent developments in wet and dry ice accretion simulation at AMIL (Anti-Icing Materials
International Laboratory) in a joint project with CIRA (Italian Aerospace Research Center). The thermodynamic model of
ice accretion is similar to existing ones developed by LEWICE in USA, DRA in British, ONERA in France and Ecole
Polytechnique de Montreal in Canada. However, this paper introduces an analytical model to calculate the surface
roughness in the wet regime based on the residual, runback and shedding liquid water mass on an airfoil surface. Also, a
new geometric model to build the ice surface, based on panel angle bisections, is presented. In the wet regime, the
empirical LEWICE correlation used to determine the equivalent sand-grain roughness is replaced by two analytical
formulations to calculate the local roughness height. The first one considers the maximal height that the bead can reach
before moving, while the second computes the wave height on the water film. The maximum bead height before moving is
determined from the equilibrium between aerodynamic, gravitational and surface tension forces. The bead behavior in dry
and wet regimes was described analytically Based on the work of Al-Khalil and Hansman, which led to the determination
of a water surface state (film, rivulets or beads). A mass balance is used to determine the residual and runback mass of
water when the water state and the maximal bead height before moving are known. The water shedding mass is equal to
the runback water mass for the lower surface and is zero for the upper surface of the airfoil respectively. A geometrical
model based on panel bisection allows the ice growth in normal direction to the surface. This method simulates the ice
surface accretion continually without gaps between panels. The accretion model is validated with icing profiles obtained
experimentally in wind tunnel by Shin and Bond for a NACA0012 wing profile with a 0.5334 m chord, a 20 µm median
volume droplet diameter, a 1 g/m³ liquid water content and a 65 m/s airspeed. These results cover both ice accretion
regimes in the -4.4° to -28.3°C temperature interval. The roughness calculated analytically is in the same order of
magnitude as LEWICE correlation. The use of analytical models for roughness generated the complex icing shapes (horn)
as the ones observed experimentally. However, in most cases, the accreted ice was slightly bigger than the measured.

NOMENCLATURE LWC : Liquid water content, g/m3.


Ap : Control volume surface, m². mb : Slope.
Ai : Ice section, m² mi : Ice mass, kg.
Ccal : Calibration parameter. mcap : Captured water mass, kg.
Cf : Friction coefficient. mevap : Evaporative water mass, kg.
CG : Flow coefficient. mrbin : Runback incoming water mass, kg.
Cpa : Air specific heat, J/kg/K. mrbout : Runback outgoing water mass, kg.
dd : Median volume droplet diameter, m. mresw : Residual mass, kg.
eb : Bead height before moving, m. mrmw : Resident mass, kg.
ef : Film height, m. mshw : Shedding mass, kg.
efmin : Minimal film height, m. mw : Liquid water mass, kg.
er : Rivulet height, m. mwadm : Admissible liquid water mass, kg.
f : Solid fraction. Pr : Air Prandtl number.
fbw : Liquid bead fraction. Pr : Air turbulent Prandtl number.
fθ Ue : Air speed at the edge boundary, m/s.
: Shape factor.
fw : Wet fraction. U∞ : Air speed, m/s.
g : Gravitational acceleration, m/s². Uτ : Friction speed, m/s.
hcv : Convection heat transfer coefficient, W/m²/K. St : Stanton number.
Tf : Water solidification temperature, K.
*
Graduate student, Applied Sciences Department, Universite du Quebec a Chicoutimi, 555 boulevard de l’Universite,
Chicoutimi, Quebec, Canada, G7H 2B1, guy_fortin@uqac.ca, Student member AIAA.

Professor, Engineering Department, Universite du Quebec a Rimouski, 300, allee des Ursulines, C.P. 3300, Rimouski,
Quebec, Canada, G5L 3A1, adrian_ilinca@uqar.ca.

Professor, Applied Sciences Department and Director of the Anti-Icing Materials International Laboratory, Universite du
Quebec a Chicoutimi, 555 boulevard de l’Universite, Chicoutimi, Quebec, Canada, G7H 2B1, jean-
louis_laforte@uqac.ca.
§
Research Engineer, Applied Aerodynamic Laboratory, Centro Italiano Ricerche Aerospaziali, 81043 Capua (CE),
Maiorise, Italy, v.brandi@cira.it.

1
American Institute of Aeronautics and Astronautics
Copyright © 2003 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
To : Static temperature, K. where the friction coefficient is defined as a function of the
Ts : Surface temperature, K. equivalent sand-grain roughness and the boundary layer
Rg : Gravitational flow ratio. displacement thickness:
Rw : Aerodynamic flow ratio.
0.1681
∆b : Control volume width, m. 1
2 ⋅Cf = 2
(3)
∆t : Time step, s.   θ 
∆θc : Hysteresis contact angle, °. ln 864 ⋅ t + 2.568 
κ : Local roughness, m.   κs 
κo : Initial object roughness, m. In this paper, the computation of the Stanton number
κs : Equivalent sand-grain roughness, m. and of friction coefficient was improved by adding
µw : Water dynamic viscosity, Pa s. analytical models to determine the local height roughness
νa : Air kinematics viscosity, m/s. based on the residual, runback and shedding water mass
θc : Contact angle, °. balance. A detailed description of this method is presented
θt : Boundary layer displacement thickness, m. in the following section.
ρa : Air density, kg/m3. To predict the ice shape we present a new technique
ρb : Bead density, kg/m3. that uses the bisection of angles between two adjacent
ρi : Ice density, kg/m3. panels. The ice growth is done along this bisection by
ρw : Water density, kg/m3. solving a two equation system for the coordinates of the
σw : Water surface tension, N/m. new node location that conserves the accreted volume
τw : Wall shear stress, Pa. without empty gaps like other methods1-5.
INTRODUCTION The predicted ice shape is compared with experimental
In flight icing of aircraft represented an intensive results for dry and wet regimes, but no computation are
research topic in recent years1, 2, 3, 4 and 5. A recognized performed for liquid regime (ice melting). Computed heat
general methodology used for the simulation of ice convection coefficient distribution are also presented as
accretion on airfoils is based on the successive computation well as comparisons of the average roughness with the
of air flow, water droplet trajectories, collection efficiency, equivalent sand grain roughness obtained by Shin and
heat transfer balance and accreted ice. A panel method with Bond7.
boundary layer correction is used for the aerodynamic ANALYTICAL MODELS
prediction of air flow around an airfoil. The balance of Liquid water mass on a surface can be in a film, rivulet
forces acting on supercooled water droplets determines or bead state. According to these alternatives, three
their trajectories and the collection efficiency. The analytical models were developed to calculate the residual
aerodynamic model based on a potential flow, the droplet water mass. For each model a roughness height is
trajectories and collection coefficients used in this paper introduced.
were developed by CIRA5.
Film
The heat balance is made on each panel using energy The film height is calculated by performing a mass
equation: summation of the latent heat of solidification, balance over the control volume and considering that the
evaporation and sublimation and the sensible, convection, flow is laminar8.
and conduction heats with the adiabatic and kinetic heating.
This thermodynamic model is similar to those available in 2 µ w ⋅ mw
the literatture1-5. The sensitive part of the solution is the ef = ⋅ ⋅C (4)
τw ρ w ⋅ ∆b ⋅ ∆t cal
convection heat transfer coefficient, a key parameter for
improved results. This coefficient is very sensitive to the The surface water forms a film when its height is
surface roughness and it is responsible for most of the greater than the minimal film height. The minimal film
inaccurate results during simulations under a wet regime height corresponds to the bead height before moving,
ice accretion. The convection heat transfer coefficient is multiplied by a shape factor which depends on the bead
based on Chilton-Colburn analogy defined in terms of the form and the contact angle.
Stanton number, air speed at the edge of the boundary layer θ c − sin(θ c )⋅ cos(θ c )
e f min = fθ ⋅ eb where fθ = (5)
2 ⋅ sin(θ c )
and the air properties:
hcv = ρ a ⋅ Cpa ⋅ U e ⋅ St (1) The residual water mass of the film is calculated by
6
The Spalding analogy is used to define the Stanton performing a mass balance over the control volume (for a
number for a roughness flat plate as a function of the dry regime, the residual water mass is zero):
equivalent sand-grain roughness, friction speed, friction
coefficient and air properties:
mrmw = ρ w ⋅ Ap ⋅ e f min (6)
The roughness height is considered to be equal to the
1 ⋅Cf
St =
2 (2) wave height over the film9.
0.45
U ⋅κ 
Prt + 1
2 ⋅ C f ⋅ 0.52 ⋅  τ s  ⋅ Pr 0.8
 νa 

2
3 τw e3 RW (eb ) − RW2 (eb ) − 4 ⋅ Rg ⋅ ∆θ c
κ= ⋅ ⋅ f (7) eb = (12)
4 µw g 2 ⋅ Rg
Rivulets • When the aerodynamic and gravitational forces act in
The water on the surface will form rivulets when the opposite directions and the gravitational force
film height is less than the minimal film height and the dominates, referred to as gravitational flow hysteresis:
control volume is not exposed to precipitation. The surface
is wavy and rough. The rivulets are treated as small films RW (eb ) + RW2 (eb ) + 4 ⋅ Rg ⋅ ∆θ c
of cylindrical form that flow parallel to the wind. The eb = (13)
rivulet height is equal to the bead height before moving: 2 ⋅ Rg
The gravitational flow ratio, Rg, is the projection of the
er = eb (8)
gravitational force parallel to the surface, divided by the
The residual water mass for the rivulet is calculated by rigidity force:
performing a mass balance over the control volume (for a
2 ρ  2 + cos(θ c )
 ⋅ sin (ϕ )
dry regime, the residual water mass is zero):
Rg = ⋅g⋅ w ⋅ (14)
mrmw = ρ w ⋅ Ap ⋅ e f ⋅ (1 − f ) (9) 3 σw  sin 2
(θ c ) 
The roughness height is equal to the rivulet height: The component parallel to the surface of the
gravitational force is, for a non-deformed bead:
κ = er (10)
π 2 + cos(θ c ) 3
Bead Fg = ⋅ g ⋅ ρw ⋅ ⋅ eb ⋅ sin (ϕ ) (15)
The water on the surface will form beads when the film 3 1 − cos(θ c )
height is less than the minimal film height and the control
The aerodynamic flow ratio, Rw, is the projection of the
volume is in the collection zone. The bead height is aerodynamic force parallel to the surface divided by the
calculated considering the bead growth in turbulent flow. rigidity force:
When the bead reaches a specific height, it starts moving
CG 2 ⋅ τ w θ c − sin (θ c ) ⋅ cos(θ c )
and a new bead develops. The maximum height that the
RW (eb ) = ⋅ ⋅ (16)
bead can reach before moving depends on the longitudinal π σ w sin 2 (θ c )⋅ [1 − cos(θ c )]
surface tension (rigidity force) and the combined effect of
The flow coefficient is the non-linear part of the
the aerodynamic and gravitational forces. Submitted to
aerodynamic flow ratio:
these forces, the bead is deformed as in Figure 1. This
2
deformation, called hysteresis, is treated as a rigidity that  τ 

reacts to the combined effect of the aerodynamic and CG = 1.725 ⋅ ln  w ⋅ eb  + 1.4  (17)
gravitational forces and is induced by the surface tension.   µ a  
The aerodynamic force is determined by the bead drag,
acceleration and object’s orientation. The force is generated
by the air flow acting in the same direction as the wind.
The force is calculated by considering that the flow is
a) non-deformed b) deformed turbulent and the bead drag coefficient is 0.44. It is
Figure 1: Forces acting on a bead. calculated for a non-deformed bead as:
The maximum height that a bead can reach before θ c − sin (θ c )⋅ cos(θ c ) 2
FW = τ w ⋅ CG ⋅ ⋅ eb (18)
moving is determined by calculating the bead height that
corresponds to the maximum hysteresis along the [1− cos(θ c )]2
movement axis. When the maximum hysteresis is reached, The wall sheer stress is divided by a calibration
the equilibrium between the aerodynamic, gravitational and parameter, Ccal:
rigidity force is broken and the bead begins to move. Three
hysteresis equations are possible when force intensities and
1
2 ⋅ ρ a ⋅ C f ⋅ U e2
τw = (19)
orientations are considered. The maximum bead height is, Ccal
for these three situations, respectively :
The rigidity force is induced by the surface tension
• When the aerodynamic and gravitational forces act in parallel to the wall. The force is calculated by integrating
the same direction, referred to as mixed flow the surface tension force component acting in the same
hysteresis: direction as the moving axis when the radius is expressed
− RW (eb ) + RW2 (eb ) + 4 ⋅ Rg ⋅ ∆θ c in terms of the non-deformed bead height and by
eb = (11) considering that the contact hysterisis is low. The deformed
2 ⋅ Rg bead geometry is complex. A simplifying hypothesis is
used: that the hysteresis distribution over the bead
• When the aerodynamic and gravitational forces act in perimeter in contact with the wall follows a cosinusoidal
opposite directions and the aerodynamic force distribution as shown in Figure 2.
dominates, referred to as wind flow hysteresis:

3
volume. The residual water mass is equal to the resident
water mass at the preceding time step.
Shedding Mass
Different liquid water behavior are considered on the
Figure 2: Hysteresis contact angle distribution. lower and upper surfaces. The runback water mass on the
lower surface will be shed by the combined action of
Following integration, aerodynamic and gravitational forces instead of going to
π the following control volume. Accordingly, a runback
Fσ = ⋅ σ w ⋅ [1 + cos(θ c )]⋅ eb ⋅ ∆θ c (20) model is used on the upper surface and a residual model on
2 the lower surface. Therefore, the shedding water mass is
The resident water mass for the bead is determined by zero on the upper surface and the runback water mass is
the admissible liquid water mass when the bead is fully zero on the lower surface.
grown. This corresponds to the quantity of water left after
partial solidification. This mass is the same for each of the m on the lower surface
beads uniformly spread over the surface. mrmw =  rbout (29)
 0 on the upper surface
mwadm = ρ w ⋅ (1− f bw )⋅ Ap ⋅ eb (21)
BISECTION METHOD
Here, the liquid fraction, fbw, is determined when the The geometric shape of accreted ice is difficult to
bead is fully grown. It is the ratio between the liquid and model because the growth direction is unknown. The use of
total height of the bead plus the ratio between the solid the panels’ bisection method allows for ice growth normal
fraction and the volume fraction of a fully grown bead. to the object’s surface which is the most natural growing
ρb 1− f ρ f 2 + cos(θ c ) (22) direction.
f bw = ⋅ ⋅ [2 + cos(θ c )] + f ⋅ b ⋅ w ⋅
ρw 3 ρ i 3 1 + cos(θ c ) The use of the panels’ bisection method is illustrated on
Figure 3. Each bisection is characterized by a slope mb,
The bead density is expressed in terms of the liquid and and the x and y coordinates of the origin node.
solid bead parts.
y − yi = mbi ⋅ (x − xi )
ρ b = ρ i ⋅ f + ρ w ⋅ (1− f )
(30)
(23)
The new panels are calculated using previous nodes
Ice density was determined by Laforte10 for a rotating coordinates, the bisection equation and accreted ice
cylinder. This equation is independent of the cylinder surface. The accreted ice on the panel is equal to the
diameter and it is valid when the surface temperature is addition of two triangular surfaces delimited by the
lower than the solidification temperature. bisection crossing the nodes 1 and 2 (Figure 3).
2
  If the positions of nodes 1, 2 and n1 are known then the
dd ⋅U ∞
ρ i = 917 ⋅   (24)
 [
 d d ⋅ U ∞ + 2.6 × 10 −6 ⋅ T f − Ts ] 

coordinates of node n2 are given by:
xn2 = x2 −
(y 2 − yn1 ) ⋅ (x2 − x1 ) − (x2 − xn1 ) ⋅ ( y 2 − y1 ) + 2 ⋅ Ai (31)
The maximum value of the resident water mass is less ( y2 − yn1 ) − mb2 ⋅ (x2 − xn1 )
than, or equal to, the admissible liquid water mass (for a and
dry regime, the resident water mass is zero):
yn2 = y 2 − mb2 ⋅
(y 2 − yn1 ) ⋅ (x 2 − x1 ) − (x2 − xn1 ) ⋅ ( y 2 − y1 ) + 2 ⋅ Ai (32)
mwadm if mwadm < mw ( y 2 − yn1 ) − mb2 ⋅ (x2 − xn1 )
mrmw =  (25)
 mw if mwadm ≥ mw
The liquid water mass is the part of the incoming water
mass on the panel that remains liquid. It is the summation
of the captured water mass, the runback incoming water
mass, and the residual mass, minus the evaporative mass:
( )
mw = mcap + mrbin + mresw − mevap ⋅ (1 − f ) (26)
The roughness height is equal to the height that the
bead will reach before moving.
Figure 3: New panel building.
κ = eb (27)
Runback Water Mass The new panels are build starting from the stagnation
The outgoing runback water mass is equal to the panel, where the runback water mass is zero, going toward
difference between the liquid water mass and the resident the lower and upper surface (Figure 4).
water mass:
mrbout = mw − mrmw (28)
The runback incoming water mass is equal to the
runback outgoing water mass of the preceding control

4
and residual water mass. On the lower surface, the input
liquid water mass is composed of collection and residual
water, since the runback water is shed by the combined
effect of the aerodynamic and gravitational forces.

Figure 4: Stagnation panel.


The method, as described, is used at the stagnation
panel with an additional equation necessary to close the
system that considers that the new panel is parallel to the
old one:
y2 − y1 yn2 − yn1 Figure 5: Liquid water mass at -28.3 °C.
= (33)
x2 − x1 xn2 − xn1
RESULTS
Study cases
The CIRALIMA code was evaluated by performing
simulations on a NACA0012 wing profile. The geometric,
aerodynamic and meteorological conditions are shown in
Table 1. The computation was performed at seven different
temperatures: -28.3°C, -19.4°C, -13.3°C, -10.0°C, -7.8°C,
-6.1°C and -4.4°C.
Accretion time 360 s
Angle of attack 4°
Chord 0.5334 m
Figure 6: Liquid water mass at -19.4 °C.
Airspeed 67.05 m/s
Atmospheric pressure 101 300 Pa
Liquid water content 1 g/m³
Median water droplet diameter 20 µm
Table 1: Meteorological conditions.
The predicted ice shapes are compared to the
experimental results of Shin and Bond11. The roughness
obtained by simulation is compared to the empirical
correlation developed by Shin et Bond7:
κ o ⋅ (0.6839) ⋅ (0.047 ⋅ To − 11.27) (34)
κs =
(
⋅ 0.5714 + 0.2457 ⋅ LWC + 1.2571⋅ LWC2 )
The initial object roughness is 0.628 mm.
This section shows the liquid water mass, the Figure 7: Liquid water mass at -13.3 °C.
convection heat transfer coefficient and the roughness
thickness distribution as a function of the curvilinear
abscissa for the final time step and the predicted and
measured icing shape after six minutes of accretion. The
regular time step used is about 32s and the last time step is
10s.
Liquid Water Mass
Figures 5 to 11 show the input liquid water mass
distribution for the final 10 s time step. In a dry regime (at -
28.3 °C and -19.4°C), the input liquid water mass is only
the collection mass, except at the stagnation panel for the -
19.4°C temperature, where the input mass contains some
residual water. In the wet regime on the upper surface, the
Figure 8: Liquid water mass at -10.0°C.
input liquid water mass is composed of collection, runback

5
increases to 706 W/m2/K and decreases to 238 W/m2/K on
the lower surface. For the -7.8°C case (Figure 14), the
coefficient increases to 1891 W/m2/K and decreases to
261 W/m2/K on the upper surface and it increases to 546
W/m2/K and decreases to 232 W/m2/K on the lower
surface. For the -6.1°C case (Figure 14), the coefficient
increases to 1385 W/m2/K and decreases to 182 W/m2/K on
the upper surface and it increases to 450 W/m2/K and
decreases to 220 W/m2/K on the lower surface.

Figure 9: Liquid water mass at -7.8°C.

Figure 12: Convection coefficient in dry regime.


For the -4.4°C case (Figure 14), the coefficient
increases to 1842 W/m2/K and decreases to 300 W/m2/K on
the upper surface and it increases to 449 W/m2/K and
decreases to 215 W/m2/K on the lower surface. The
Figure 10: Liquid water mass at -6.1°C.
maximum coefficient is found on the upper surface and it
increases with increasing temperature (from 615 W/m²/K at
-28.3°C to 1842 W/m²/K at -4.4°C). The average
coefficient on the upper surface is about the same for all
temperatures and, on the lower surface it decreases with
decreasing temperature (from 360 W/m²/K at –28.3°C to
210W/m²/K at -4.4°C).

Figure 11: Liquid water mass at -4.4°C.


Convection Heat Transfer Coefficient
Figures 12 to 14 show the convection heat transfer
coefficient distribution for the final time step. The
coefficient is at a minimum at, or near, the stagnation
panel, then it increases to a maximum value and decreases
towards the separation panels. For the -28.3°C case
(Figure 12), the coefficient increases to 615 W/m2/K and Figure 13: Convection coefficient in mixed regime.
decreases to 399 W/m2/K on the upper surface and
increases to 478 W/m2/K and decreases to 333 W/m2/K on
the lower surface. For the -19.4°C case (Figure 12), the
coefficient increases to 569 W/m2/K and decreases to 387
W/m2/K on the upper surface and increases to 455 W/m2/K
and decreases to 308 W/m2/K on the lower surface. For the
-13.3°C case (Figure 13), the coefficient increases to
1623 W/m2/K and decreases to 400 W/m2/K on the upper
surface and it increases to 515 W/m2/K and decreases to
280 W/m2/K on the lower surface. For the -10.0°C case
(Figure 13), the coefficient increases to 1107 W/m2/K and
decreases to 350 W/m2/K on the upper surface and it

6
Figure 14: Convection coefficient in wet regime.
Roughness Thickness
Figures 15 to 17 show the roughness thickness
distribution for the final time step. Under a dry regime, the
roughness thickness is at a maximum at the stagnation
panel and decreases towards the separation panels. The
roughness is created by supercooled water droplets which
freeze on impact and form beads on the surface. The frozen
beads pile up and reach a height equivalent to the
maximum height before moving. At this moment the new
droplets fall between existing beads and the process
restarts. For the -28.3°C case (Figure 15), the average
thickness is 0.48 mm on the lower surface and 0.23 mm on
the upper surface. For the -19.4°C case (Figure 15), the Figure 15: Roughness in dry regime.
average thickness is 0.44 mm on the lower surface and On the lower surface the thickness decreases and
0.21 mm on the upper surface. Under the mixed and wet reaches a minimum, after which it increases (1.95 mm at -
regimes, the liquid water on the upper surface flows as a 7.8°C to 1.19 mm at -6.1°C and 1.27 mm at -4.4°C). The
film to the next panel. The roughness is formed by the average thickness is independent of the temperature and is
wave height. For this flow state, the thickness is at a very small (0.05 mm). On the lower surface the average
minimum at the stagnation panel, and increases gradually thickness decreases with increasing temperature (0.40 mm
to reach a maximum value, after which it decreases towards at -13.3°C to 0.13 mm at -4.4°C)
the separation panels. For the -13.3°C case, (Figure 16),
the thickness increases to 3.34 mm and decreases to
0.09 mm. For the -10.0°C case, (Figure 16), the thickness
increases to 3.70 mm and decreases to 0.05 mm. For the
-7.8°C case, (Figure 17), the thickness increases to 5.62
mm and decreases to 0.03 mm. For the -6.1°C case,
(Figure 17), the thickness increases to 4.95 mm and
decreases to 0.04 mm. For the -4.4°C case, (Figure 17), the
thickness increases to 4.65 mm and decreases to 0.02 mm.
For the panel on the lower surface near the stagnation, the
liquid water, before shedding, is under film or bead form.
When the liquid water forms beads, the beads will grow to
reach maximum height before moving. The maximum
height reached by the beads is 1.20 mm (at -13.3°C) and Figure 16: Roughness in mixed regime.
decreases to 0.40 mm towards the separation panel. When
the liquid water forms a film, the roughness is formed by
the wave height. For this flow state, the thickness is at a
minimum at the stagnation panel, and increases gradually
to reach a maximum value, after which it decreases towards
the separation panels. For the -10.0°C case, (Figure 16),
the thickness increases to 1.95 mm and decreases to 0.24
mm. For the -7.8°C case, (Figure 17), the thickness
increases to 1.41 mm and decreases to 0.19 mm. For the -
6.1°C case, (Figure 17), the thickness increases to 1.19
mm and decreases to 0.14 mm. For the -4.4°C case,
(Figure 17), the thickness increases to 1.27 mm and
decreases to 0.13 mm. For the other panels on the lower
surface, the roughness is created by the supercooled water Figure 17: Roughness in wet regime.
droplets which freeze on impact and form beads on the Figure 18 shows the time dependence of the roughness
surface in similar manner as on the upper surface. Under a distribution expressed in terms of equivalent sand-grain
dry regime, the maximum thickness is greater on the lower roughness. Under a dry regime, the thickness oscillates
surface (0.46 mm) than on the upper surface (0.22 mm) and lightly around an average value. In a wet regime, the
is independent of temperature. Under a wet regime, the roughness increases after which it oscillates greatly around
maximum thickness is greater on the upper than the lower an average value.
surface (0.22 mm). On the upper surface the thickness
increases and reaches a maximum, after which it decreases
(3.34 mm at -13.3°C to 5.62 mm at -7.8°C and 4.65 mm at
-4.4°C ).

7
lower surface, but different on upper surface because the
horn is not developed.

Figure 18: Equivalent sand-grain roughness.


Figure 19 shows a comparison between the average
thickness calculated over the time period and the roughness Figure 20: Ice shape at -28.3 °C.
calculated with the empirical Shin and Bound correlation.
The roughness values are of the same order of magnitude
and behave similarly.

Figure 21: Ice shape at -19.4 °C.

Figure 19: Average equivalent sand-grain roughness.


For the empirical Shin and Bound correlation, the
roughness increases linearly with temperature and the
predicted roughness increases according to a power law
inversely proportional to the temperature in absolute values
and in degrees Celsius.

κ s = 2.82 ⋅ T( ) −0.6
(35)
Ice accretion shapes
Figures 26 to 32 show the ice shapes obtained after six
minutes of accretion. For the -28.3°C (Figure 20) and -
Figure 22: Ice shape at -13.3 °C.
19.4°C (Figure 21) cases, only rime ice occurs and the
shape is characteristic of the dry regime. The lower surface
is more channeled than the upper surface. Predicted and
measured icing shapes are essentially the same, except that
the predicted volume is slightly greater than the measured.
For the other cases (-13.3°C, Figure 22; -10.0°C, Figure
23; -7.8°C, Figure 24; -6.1°C, Figure 25 and -4.4°C,
Figure 26), the predicted ice shape is characteristic of a
wet regime. A horn forms on the upper surface and a small
hollow forms at the stagnation panel. The ice surfaces on
the upper and lower surfaces near the attached icing points
are channeled. The predicted measured shapes are similar,
but the predicted volume is slightly greater then the
measured volume. In some cases, the attached icing points Figure 23: Ice shape at -10.0 °C.
are not located at the same position. For the -4.4°C case,
the predicted and measured icing shapes are similar on the

8
Figure 24: Ice shape at -7.8 °C. Figure 27: Time step effect at -28.3 °C.
Panel length effect
Figure 28 shows the influence of panel length on the
final icing shape for the -4.4°C temperature case. The final
icing shape is not affected by increasing the percentage
chord panel length by a factor of 2 or 3.

Figure 25: Ice shape at -6.1°C.

Figure 28: Panel length effect at -4.4 °C.


ANALYSIS
Thermodynamic Model
The proposed model can simulate ice accretion on an
airplane wing with a precision comparable to and, in some
cases, better than the existing models. The precision
increases as temperature decreases. The predicted and
measured ice shapes are similar and the asperity and
channeling observed experimentally are present on the
Figure 26: Ice shape at -4.4°C. predicted ice surface. The maximum value of the
For the -7.8°C case, the predicted and measured ice convection heat transfer coefficient found on upper surface
shapes are very similar because this case was chosen to facilitate the horn formation, as observed experimentally.
calibrate the roughness model. The numerical value of the The length of the wet zone near the stagnation panel
calibration parameter, Ccal, used in equation 4 and 19. increases with temperature.
Time step effect Roughness Models
Figure 27 shows the influence of the time step on the The roughness models used in dry and wet regimes
final icing shape for the -28.3°C temperature case. The produce an average roughness in the same order of
final icing shape is not affected by reducing or increasing magnitude as Shin and Bond’s empirical correlation.
the regular time step by a factor of 2. However, the local roughness obtained by simulation are
very different. This difference can favor horn formation
and asperity since the local convection heat transfer
coefficient depends on the roughness. The interdependence
between the roughness and friction coefficient does not
lead to numerical instability, but, rather, to oscillations
around an average value. It has been shown that the models
used to calculate the roughness are independent of the
panel length and the time step.

9
Mass Models
The proposed model to calculate the resident, runback 3 Guffond, D., Hedde, T. and Henry, R., Overview of
and shedding water masses are a good representation of the Icing Research at ONERA, AGARD/FDP Joint
reality because the final icing shapes are validated by the International Conference on Aircraft Flight Safety -
measured icing shapes. The free water at the surface that Actual Problems of Aircraft Development, Zhukovsky,
takes the shape of a film, bead or rivulet, flows on the Russia, August 31 - September 5, 1993.
upper surface and is shed on the lower surface. A small 4 Paraschivoiu I., Overview of J.-A. Bombardier
water mass will reside on the surface depending on the Aeronautical Chair Projects at École Polytechnique de
intensity and direction of aerodynamic and gravitational Montréal, 1st Bombardier International Workshop
forces. The runback water mass on the upper surface and Aircraft Incing/Boundary-Layer Stability and
the shedding water mass on the lower surface increases Transition, Montréal, Québec, Canada, September 20-
with temperature as well as the resident water mass. The 21, 1994.
icing mass and volume decrease as temperature increases. 5 Mingione, G. and Brandi, V., A Code for the
Ice density increases with temperature. Evaluation of Ice accretion on Multi-Element Airfoils,
Centro Italiano Recerche Aerospaziali, CIRA Technical
Bisection Method
Manual 96-089, Juliet, 26, 1996.
Based on the results, the proposed geometric accretion
6 White, F. M., Viscous Fluid Flow, Mechanical
model, based on panels’ bisection method, is a good
Engineering, Second Edition, McGraw-Hill, Inc., 1991.
representation of reality. The method can reproduce horn
7 Shin, J. and Bond, T., Experimental and Computational
and spikes in wet regime as well as the asperity and
Ice Shapes and Resulting Drag Increase for a NACA
channeling observed experimentally. The method is
0012 Airfoil, NASA Technical Manual 105743, 1992.
independent of the panel length and time step. The angle
8 Al-Khalil, K.M., Keith, T. G., De Witt, Jr., K. J.,
between panels are limited to a minimum value of 140° and
Nathman, J. K. and Dietrich, D. A., Thermal Analysis
a maximum value of 240° since the flow code cannot
of Engine Inlet Anti-Icing Systems, American Institute
support a complex surface. The panel lengths are limited to
of Aeronautics and Astronautics Journal, AIAA 89-
a minimum value of 0.25% of the chord and a maximal
0759, 1989.
value of 1% of the chord because of the droplets
9 Gerhart, P. M., Gross, R. J. and Hochstein, J. I.,
trajectories model. The geometry model is robust enough to
Fundamentals of Fluid Mechanics, Second Edition,
work with small and large ice masses on the panels without
Addison-Wesley Publishing Compagny, 1992.
altering the final shape.
10 Laforte, J.-L. et Allaire, M. A., Évaluation du
CONCLUSIONS givromètre d’Hydro-Québec à différentes intensités de
Based on the final icing shapes, the analytical model givrage sec et humide, Rapport HQ-92-02, 1992.
additions (local roughness thickness, resident runback and 11 Shin, J. and Bond, T., Results of an Icing Test on a
shedding water masses and the bisection method) to the NACA 0012 Airfoil in the NASA Lewis Icing Research
thermodynamic model currently used for ice accretion on Tunnel, American Institute of Aeronautics and
airplane wings, generated the complex icing shapes Astronautics Journal, AIAA-92-0647, 1992.
observed experimentally (horn and channeling) with the
same precision as the other models, or better, in some
cases. However, in the majority of cases, the predicted
icing volume is slightly larger than the measured. The
simulated results show that the runback water is the main
mass process on the upper surface while shedding water is
the main mass process on the lower surface. The ice shape
is strongly dependent on the local roughness which
controls the friction coefficient and, indirectly, the
convection heat transfer coefficient.
ACKNOWLEDGMENTS
This work was supported financially by AMIL. The
authors would like to thank the CIRA who provided the
flow, trajectories and collection codes. We also thank
Doctor Patrick Louchez who started the project.
REFERENCES

1 Wright, W.B., Update to the NASA Lewis Ice


Accretion Code LEWICE, NASA report #195387, pp.
48, 1994.
2 Gent, R. W., TRAJICE2, A Combined Water Droplet
and Ice Accretion Prediction Program for Aerofoil,
DRA Technical Report Number TR90054, 1990.

10

You might also like