Aircraft Icing

2-D Ice accretion prediction

Prof. Dr. Serkan ÖZGEN

Dept. Aerospace Engineering, METU
November 2020
 Outline
• Flow field solution
• Droplet trajectories and collection efficiencies
– Governing equations
– Droplet impact
– Parameters effecting collection efficiency
• Boundary-layer calculations
• Ice growth model; Extended Messinger Model
• Modified geometry
• Runback water
• Single and multi-step approach
• Validation results

Serkan ÖZGEN 2
 Flow Field Solution
The flow field solution is needed to determine:
• The velocity and pressure distribution on the surface of
the geometry for subsequent boundary-layer calculations
and determination of aerodynamic performance.
• Off-body velocities to be used for droplet trajectory
calculations.

Serkan ÖZGEN 3
 Flow Field Solution
Alternatives:
• Panel method coupled with a boundary-layer solution:
less accurate but fast and cheap.
• Navier-Stokes: more accurate but slow and expensive.
No significant improvement has been observed by using a
Navier-Stokes solver instead of a panel code  use a Hess-
Smith panel code.

Serkan ÖZGEN 4
 Flow Field Solution
• In the panel method, the geometry is divided into quadrilateral panels
each associated with a source singularity element together with a
vortex singularity that is constant for all panels.
• The strengths of the singularities are constants and are unknowns of the
problem. The developed computer program uses N quadrilateral panels
to solve N+1 singularity strengths using the flow tangency boundary
condition at the collocation points of the panels and an additional
equation is introduced for the Kutta condition. The collocation points are
the centroids of each panel.
• Once the singularity strengths are calculated, one can construct a
velocity potential and hence calculate the air flow velocity components
at any location in the flow field including the boundaries of the
geometry. The velocity components at a given point are the x-, y-
derivatives of the velocity potential constructed at that point.

Serkan ÖZGEN 5
Flow field solution
NACA 23012, α=-0.27o

Serkan ÖZGEN 6
Flow field solution
NACA 23012, α=-1.05o

Serkan ÖZGEN 7
 Droplet trajectories and collection
efficiencies
For droplet trajectories, Lagrangian approach is used with the
following assumptions:
• Droplets are assumed to be spherical.
• The flow field is not affected by the presence of the droplets.
• Gravity and aerodynamic drag are the only forces acting on the
droplets.
These assumptions are valid for droplet sizes with 𝑑𝑝 ≤ 500 𝜇𝑚.

Serkan ÖZGEN 8
 Droplet trajectories and collection
efficiencies
The governing equations for droplet trajectories are:

𝑚𝑥ሷ 𝑝 = −𝐷 cos 𝛾,
𝑚𝑦ሷ𝑝 = −𝐷 sin 𝛾 + 𝑚𝑔,
−1 𝑦ሶ 𝑝 −𝑉𝑦
𝛾= tan ,
𝑥ሶ 𝑝 −𝑉𝑥
12
𝐷 = 𝜌𝑉𝑟𝑒𝑙 𝐶𝐷 𝐴𝑝 ,
2

𝑉𝑟𝑒𝑙 = (𝑥ሶ 𝑝 − 𝑉𝑥 )2 +(𝑦ሶ𝑝 − 𝑉𝑦 )2

Serkan ÖZGEN 9
 Droplet trajectories and collection
efficiencies
Definitions of parameters:
𝐶𝐷 : droplet drag coefficient,
𝑉𝑥 , 𝑉𝑦 : components of the flow field velocity at the droplet
location,
𝑥ሶ 𝑝 , 𝑦𝑝ሶ : components of the droplet velocity,
𝑥ሷ 𝑝 , 𝑦𝑝ሷ : components of the droplet accelaration,
𝜌: atmospheric density,
𝜋 2
𝐴𝑝 = 𝑑 :
4 𝑝
droplet cross-sectional area.

Serkan ÖZGEN 10
 Droplet trajectories and collection
efficiencies
The drag coefficients of the droplets are calculated from:
𝑅𝑒
𝐶𝐷 24 = 1 + 0.197𝑅𝑒 0.63 + 2.6 ∗ 10−4 𝑅𝑒 1.38 , 𝑅𝑒 ≤ 3500,
𝑅𝑒
𝐶𝐷 24 = 1.699 ∗ 10−5 𝑅𝑒 1.52 , 𝑅𝑒 > 3500.
𝑅𝑒 = 𝜌𝑉𝑑𝑝 Τ𝜇

Serkan ÖZGEN 11
 Droplet trajectories and collection
efficiencies
Droplet impact:
𝑥𝑖+1 , 𝑦𝑖+1

𝑥𝑑𝑝 , 𝑦𝑑𝑝 𝑥𝑑 , 𝑦𝑑

𝑥𝑖 , 𝑦𝑖

Serkan ÖZGEN 12
 Droplet trajectories and collection
efficiencies
𝑙1 = 𝑥𝑑 − 𝑥𝑑𝑝
𝑛1 = 𝑦𝑑 − 𝑦𝑑𝑝 ൠ vector along particle trajectory
𝑙2 = 𝑥𝑖+1 − 𝑥𝑖
ൠ vector along panel
𝑛2 = 𝑦𝑖+1 − 𝑦𝑖

𝑙1 𝑡1 + 𝑙2 𝑡2 = 𝑥𝑖 − 𝑥𝑑𝑝
ൠ system of two equations
𝑛1 𝑡1 + 𝑛2 𝑡2 = 𝑦𝑖 − 𝑦𝑑𝑝
with two unknowns
𝑙1 𝑙2 𝑡1 𝑥𝑖 − 𝑥𝑑𝑝
= 𝑦 −𝑦
𝑛1 𝑛2 𝑡2 𝑖 𝑑𝑝

Serkan ÖZGEN 13
 Droplet trajectories and collection
efficiencies
𝑛2 𝑥𝑖 −𝑥𝑑𝑝 −𝑙2 (𝑦𝑖 −𝑦𝑑𝑝 )
𝑡1 = ; along the trajectory,
(𝑙1 𝑛2 −𝑙2 𝑛1 )
𝑙1 𝑦𝑖 −𝑦𝑑𝑝 −𝑛1 (𝑥𝑖 −𝑥𝑑𝑝 )
𝑡2 = ; along the panel.
(𝑙1 𝑛2 −𝑙2 𝑛1 )
Additional constraints to prevent fake impact locations:
0 ≤ 𝑡1 ≤ 1,
0 ≤ 𝑡2 ≤ 1.
Exact impact location:
𝑥𝑖𝑚𝑝 = 𝑥𝑖 + 𝑙2 𝑡2 ,
𝑦𝑖𝑚𝑝 = 𝑦𝑖 + 𝑛2 𝑡2 ,

Serkan ÖZGEN 14
 Droplet trajectories and collection
efficiencies
Collection efficiency:

Serkan ÖZGEN 15
Particle trajectories

Serkan ÖZGEN 16
Droplet collection efficiency
(effect of airframe size)

Serkan ÖZGEN 17
Droplet collection efficiency
(effect of droplet size)

Serkan ÖZGEN 18
 Droplet collection efficiency
(effect of freestream velocity)

Serkan ÖZGEN 19
Droplet collection efficiency
(effect of angle of attack)

Serkan ÖZGEN 20
 Droplet trajectories and collection
efficiencies
Factors increasing collection efficiency:
• Smaller airframe size; geometry creates a smaller obstacle for
the incoming droplets and the deviation of the particles away
from the body is not sufficient for them to avoid it.
• Greater droplet size; increases droplet inertia and droplets
follow ballistic trajectories  collection efficiencies increase,
impingement zone gets wider.
• Higher airspeed; increases droplet inertia.

Serkan ÖZGEN 21
 Boundary-layer calculations

• Integral boundary layer method is used for the calculation of

the convective heat transfer coefficients.
• With this method, the details of the laminar and turbulent
boundary layers are calculated fairly accurately.

Serkan ÖZGEN 22
 Boundary-layer calculations
• Transition prediction is based on the roughness Reynolds number:
𝜌𝑈𝑘 𝑘𝑠
𝑅𝑒𝑘 =
𝜇

𝑘𝑠 : roughness height,
𝑈𝑘 : local flow velocity at the roughness height:

𝑈𝑘 𝑘𝑠 𝑘𝑠 3 𝑘𝑠 4 1 𝛿 2 𝑑𝑈𝑒 𝑘𝑠 𝑘𝑠 3
= 2 −2 + + 1 −
𝑈𝑒 𝛿 𝛿 𝛿 6 𝜈 𝑑𝑠 𝛿 𝛿

𝑈𝑒 : flow velocity outside the boundary-layer at the roughness

location.
𝑠: streamwise distance along the airfoil surface starting at the
stagnation point.
Serkan ÖZGEN 23
 Boundary-layer calculations
• Roughness height (NASA model):
𝑘𝑠
= 0.00117𝑘𝑉∞ 𝑘𝐿𝑊𝐶 𝑘 𝑇∞ 𝑘𝑑𝑝
𝛿
𝑘𝑉∞ = 0.4286 + 0.0044139𝑉∞ ,
𝑘𝐿𝑊𝐶 = 0.5714 + 0.2457 ∗ 1000𝐿𝑊𝐶 + 1.2571 ∗ (1000𝐿𝑊𝐶)2
𝑇∞
𝑘 𝑇∞ = 46.8384 − 11.2037
1000
𝑘𝑑𝑝 = 1 𝑖𝑓 𝑑𝑝 ≤ 20 𝜇𝑚,
𝑘𝑑𝑝 = 0.033 𝑖𝑓 20𝜇 ≤ 𝑑𝑝 ≤ 50𝜇𝑚,
5 1
𝑘𝑑𝑝 = − 𝑑𝑝 ∗ 106 𝑖𝑓 𝑑𝑝 > 50𝜇𝑚.
3 30

Serkan ÖZGEN 24
 Boundary-layer calculations
• Laminar boundary-layer thickness:
315
𝛿= 𝜃𝑙
37
• Laminar momentum thickness (Thwaites formulation):
𝜃𝑙2 0.45 𝑠 5
= 6 න 𝑈𝑒 𝑠 𝑑𝑠
𝜈 𝑈𝑒 (𝑠) 0
• For laminar flow, 𝑅𝑒𝑘 ≤ 600, method of Smith & Spalding for
the calculation of the convective heat transfer coefficient:
0.296𝑘𝑈𝑒1.435
ℎ𝑐 =
𝑠
𝜈 ‫׬‬0 𝑈𝑒1.87 𝑠 𝑑𝑠

𝑘: thermal conductivity of air.

Serkan ÖZGEN 25
 Boundary-layer calculations
• For turbulent flow, 𝑅𝑒𝑘 ≥ 600, method of Kays & Crawford
for the calculation of the convective heat transfer coefficient:
ℎ𝑐 = Stρ𝑈𝑒 𝐶𝑝
• Stanton number:
𝐶𝑓 /2
𝑆𝑡 =
𝑃𝑟𝑡 + (𝐶𝑓 Τ2) /𝑆𝑡𝑘
𝑃𝑟𝑡 = 0.9: turbulent Prandtl number.
• Roughness Stanton number:
𝑆𝑡𝑘 = 1.92𝑅𝑒𝑘−0.45 𝑃𝑟 −0.8
𝑃𝑟 = 0.72: laminar Prandtl number.

Serkan ÖZGEN 26
 Boundary-layer calculations
• Skin friction coefficient is calculated from Makkonen relation:
𝐶𝑓 0.1681
= 2
2 ln 864 𝜃𝑡 Τ𝑘𝑠 + 2.568
• Turbulent momentum thickness:

𝑠 0.8
0.036𝜈 0.2 3.86
𝜃𝑡 = 3.29 න 𝑈𝑒 𝑠 𝑑𝑠 + 𝜃𝑡𝑟
𝑈𝑒
0

Serkan ÖZGEN 27
 Boundary-layer calculations
• Boundary-layer calculations start at the leading edge and
proceed downstream using the marching technique for the
upper and lower surfaces of the airfoil.
• Transition is fixed at the streamwise location, where
𝑅𝑒𝑘 = 600 according to Von Doenhoff criterion.

Serkan ÖZGEN 28
Boundary-layer calculations

Serkan ÖZGEN 29
Boundary-layer calculations

Serkan ÖZGEN 30
 Extended Messinger Model
See lecture notes.

Serkan ÖZGEN 31
 Modified geometry
• After the ice growth rates given by equation (10) for rime ice
and equation (52) for glaze ice are solved, an ice thickness
results for a given control volume (panel).
• In order to obtain the modified geometry with ice formation,
it is assumed that the ice grows perpendicularly to the
surface.

Serkan ÖZGEN 32
 Modified geometry

Ψ𝑖−1 𝑥𝑖+1 , 𝑦𝑖+1

Ψ𝑖

𝑥𝑖 , 𝑦𝑖 Ψ𝑖 : panel inclination angle

𝐵𝑖 : ice thickness calculated at node 𝑖,
Ψ𝑖−1 +Ψ𝑖
𝑥ҧ𝑖 = 𝑥𝑖 + 𝐵𝑖 sin 2
Ψ𝑖−1 +Ψ𝑖
𝑦ത𝑖 = 𝑦𝑖 + 𝐵𝑖 cos 2

𝑥𝑖−1 , 𝑦𝑖−1

Serkan ÖZGEN 33
 Runback water
• Under glaze ice conditions, freezing fraction is less than one
(𝐹𝐹 < 1), meaning that only a fraction of the impinging
water freezes.
• Remaining water may either flow downstream as runback
water or may be shed due to high shear and gravity.
• According to Fortin et al:
– Upper surface: All the unfrozen water passes to the neighboring
downstream panel as runback water,
– Lower surface: All unfrozen water is shed (due to gravity).
• Including the runback water introduces some modifications to
the Messinger Model formulation (see lecture notes).

Serkan ÖZGEN 34
 Single and multi-step approach
• Computing the ice shapes in a single cycle or computational
step for the entire duration of the exposure may yield
unrealistic and incorrect ice shapes.
• This is particularly true for:
– Glaze ice conditions where runback water effect is prominent,
– Icing exposures where the icing conditions are varying, i.e. climbing
flight, descending flight.
– Long exposures.

Serkan ÖZGEN 35
 Single and multi-step approach
• In the current solution method, an unsteady problem is solved by a
quasi-steady approach.
• Therefore a continuous flow of water, impingement, ice and water
accumulation has to be analyzed in as small as possible intervals to
represent these faithfully, but this increases computational time.
• With a multi-step approach, the effect of ice formation on the flow
field, droplet trajectories, ice formation and runback water are
taken into account to some degree.
• At each computational step, flow field solution, droplet trajectories,
collection efficiencies, energy terms and the ice formation routines
are repeated.
• Experience shows that Δ𝑡 = 60 − 120 seconds is a good
compromise between computational time and accuracy.

Serkan ÖZGEN 36
 Validation results
Variables Value
Angle of attack 4o
Chord length 0.53 m
Freestream velocity 58.1 m/s
Ambient pressure 95610 Pa
Liquid water content 1.3 g/m3
Exposure time 480 s
Droplet median diameter 20 μm

Serkan ÖZGEN 37
Validation results (Ta=-27.8o C)

Serkan ÖZGEN 38
Validation results (Ta=-19.8o C)

Serkan ÖZGEN 39
Validation results (Ta=-13.9o C)

Serkan ÖZGEN 40
Validation results (Ta=-6.7o C)

Serkan ÖZGEN 41
Multi-step approach (Ta=-27.8o C)

Serkan ÖZGEN 42
Multi-step approach (Ta=-19.8o C)

Serkan ÖZGEN 43
Multi-step approach (Ta=-13.9o C)

Serkan ÖZGEN 44
Multi-step approach (Ta=-6.7o C)

Serkan ÖZGEN 45
 Airfoil with flap
Variables Value
Angle of attack 4o
Chord length 1. element: 0.53 m
2. Element: 0.1 m
Freestream velocity 60 m/s
Ambient pressure 101300 Pa
Liquid water content 1.0 g/m3
Exposure time 360 s
Droplet median diameter 20 μm

Serkan ÖZGEN 46
Particle trajectories
(airfoil with flap)

Serkan ÖZGEN 47
Ice shapes in two element airfoil
(δf=10o)

Serkan ÖZGEN 48
Ice shapes in two element airfoil
(main airfoil, δf=10o)

Serkan ÖZGEN 49
Ice shapes in two element airfoil
(flap, δf=10o)

Serkan ÖZGEN 50