Вы находитесь на странице: 1из 8

American Institute of Aeronautics and Astronautics

1
Icing Numerical Simulation for Single and Multi-Element
Airfoils
Zhihong Zhou
1
, Fengwei Li
2
, Guangning Li
3
,Weimin Sang
4

College of Aeronautics, Northwestern Polytechnical University
We developed an ice accretion prediction code applies a time-stepping procedure to
accrete ice on single and multi-element airfoils, each time step consists of a flowfield
calculation, a particle trajectory and impingement computation, and an ice accumulation
prediction. The computational grids for the multi-element airfoils were generated by using a
multi-block grid technique. We use a four-stage Runge-Kutta method to solve the droplet
trajectory equation and a recursive box searching method is proposed to compute the
position of the droplet, compared with the traditional direct method, it obtains the same
result but reduces computational time by one order of magnitude. Ice shape is determinated
with assumption that all droplets freeze at their points of impact and the ice grows in the
direction normal to the surface. The Reynolds-averaged Navier-Stokes equations are
conducted to calculate the air flow field and a time-accurate, fully implicit LU-SGS method
is used to solve the viscous flow problems. It uses the flux-vector splitting (FVS) method of
van Leer, and Spalart-Allmaras turbulence model was used in the numerical procedure.
With the above methods and techniques, the computational results we got are in good
agreement with the available experiments data, and show preliminarily that our method of
icing predicted model and flow field numerical simulation is feasible.
Nomenclature
= Air mass density
P = The pressure per unit mass
H = Total enthalpy per unit mass
E = Total energy per unit mass
d
= Warter droplet mass density
u , v = Two components of the velocity vector
d
v = The volume of the drople
d
A = The area normal to the free stream direction
d
C = The drag coefficient of the droplet.
= Dynamic viscosit
C
p
= pressure coefficient
Re

= The Reynolds number
r
Re

= The relative Reynolds number
LWC

= Liquid water content, kg/m3
MVD

= Mean Volumetric Diameter,
Dp

= Droplet diameter, m



1
Graduate Student, College of Aeronautics, Northwestern Polytechnical University.
2
Professor, College of Aeronautics, Northwestern Polytechnical University
3
Graduate Student, College of Aeronautics, Northwestern Polytechnical University.
4
Associate professor, College of Aeronautics, Northwestern Polytechnical University.
28th AIAA Applied Aerodynamics Conference
28 June - 1 July 2010, Chicago, Illinois
AIAA 2010-4232
Copyright 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

American Institute of Aeronautics and Astronautics


2
I. Introduction
t is well known and appreciated that the accretion of ice on lifting surfaces can significantly degrade the
performance and handling characteristics of affected aircraft. Ice accretion and its subsequent build-up is one of
the potential hazards in airplane flight. To improve flight safety, a better understanding of the effect of ice accretion
on the aerodynamic performance of airplane is required. Prediction of ice accretion is thus an important part of
airplane design and the need for ice accretion prediction capabilities has been growing over the past several years.
As we known, the ice on wing is the most important hazardous factor to the aerodynamics performance of
airplane on icing problems. Computational studies of aircraft icing have primarily concentrated on the more
common leading-edge ice shapes. This work presents an improved lagrangian approach to calculate the droplet
collection efficiency on arbitrary two dimensional bodies such as single and multi-element airfoils, a recursive box
searching method is proposed. Based on the above methods, we developed a 2-D ice accretion prediction code
which can simulate the ice accretion on single and multi-element airfoils.
Just like the recognized general methodology used for the simulation of ice accretion, our code is based on the
successive computation of air flow, water droplet trajectories, collection efficiency, heat transfer balance, and
accreted ice. For this paper, a four-stage Runge-Kutta method is used to solve the droplet trajectory equation, a
recursive box searching method is proposed to compute the position of the droplet. To determine the capabilities of
our software codes, it is necessary to complete quantitative experiments for assessing the codes. This paper is
divided into four sections, plus a conclusions section. Section II describes the governing equations of the air flow
field, Section III provides a description of the grid generation method. Section IV describes the calculation of water
collection and computation of new foils. Section V presents some results of the test cases.
II. Governing Equations and Numerical Method
The first and most important step in icing analysis is to investigate the effectiveness of a body in collecting the
water droplets in the atmosphere. The collection efficiency calculation is a two-step process in our code. First, the
external flow is solved over the body. The subsequent droplet impingement calculation is performed using the
Lagrangian approach.
The air flowfield surrounding the airfoil is simulated by solving the N-S equations. Our numerical approach to
solve the N-S equations is based on a finite volume form of the integral equations. The equations are the expression
of the conservation principle for mass, momentum and energy. In a domain of volume with boundary S, the
equations may be written in the following form:

d d d
1
V
S S
t Re

+ =


U F S F S

(1)
For two-dimensional case, the vector of the conserved variables and both the inviscid and viscous flux vector
function are given by

u
v
E





=





U
2
2
u v
u P uv
uv v P
uH vH







+

=

+




F
0 0
xx xy V
yx yy
x y






=






F (2)
where , P , H ,andE are the density, pressure, total enthalpy and total energy per unit mass respectively, and
u , v are the Cartesian velocity components, V is the velocity vector, the variables in
V
F can be found in Reference
1
1
.As used for the present computations, the cell-centered finite volume methods are employed to solve the Navier-
Stockes equations. The flux-vector splitting (FVS) method of van Leer is implemented
2
. A time-accurate, fully
implicit method based on LU-SGS has been used to solve the viscous flow problems in this paper, the details of the
method can be found in Reference 3
3
. At the wall surface boundary the immovability boundary condition is applied,
the wall velocity is zero. At the far-field boundary some special boundary conditions must be imposed. The viscosity
coefficient computed as the sum of laminar and turbulent viscosity coefficients, which are evaluated by the
Sutherlands law. The Spalart-Allmaras model is implemented to handle turbulence flows.

III. Grid Generation
A. Structured Grid Generation Method
I

American Institute of Aeronautics and Astronautics


3
The computational grids about icing research are obtained by using the elliptic grid generation method with
forcing-function control. The elliptic grid generation equations of Thompson are solved implicitly
4

5
. Control of the
grid behavior is achieved through the introduction of forcing function terms in the manner of Hilgenstock. It can
adjust the angle and spacing at the boundary. The main idea of this technique is to use the difference between
desired and actual value (angle or spacing) during the iteration to estimate the corrections to be added to the forcing-
function terms. The detailed process is given in Reference 4. With the control technique, the grid lines can be
matched along the common boundaries in the multi-
block grid generation. The numerical procedure to
generate the structured grid consists of the following
steps:
1) Generate an initial grid by an algebraic method
based on transfinite interpolation.
2) Calculate initial values of the forcing-function
terms at the boundaries.
3) Interpolate these forcing-function terms into the
whole domain.
4) Solve the elliptic equations by SSOR method.
5) Change the forcing-function terms at the
boundaries with the desired values.
6) Repeated The step 3 to 5 until the grid we need
is generated.
Figure 1 is the flow chart of generate a
structured grid.
B. Multi-block Grid Technique
For complex configuration, such as multi-
element airfoil, it is difficult to generate a single-
block, high quality and body-fitted orthogonal grid
6
. To overcome such difficulty, a multi-block grid technique was
used through the combination of the transfinite interpolation and the elliptic method with boundary control. The
basic idea of the multi-block grid technique is to decompose the domain into different appropriate blocks, and then
subgrids are generated in each block separately and patched smoothly along common boundaries to form the global
grid for the whole field.
IV. Icing simulation
The first and most important step in icing analysis is to investigate the effectiveness of a body in collecting the
water droplets in the atmosphere, often referred to as the collection efficiency of the body. Collection efficiency is
predominantly a function of body shape and size, droplet size and the flowfield. Collection efficiency is a local
quantity defined at a given location on the body surface and is defined as the ratio of the mass flux of the droplets
impinging on the surface at that location to the mass flux at the free stream. The collection efficiency calculation is a
two-step process. First, the external flow is solved over the body. The subsequent droplet impingement calculation is
performed using the Lagrangian approach.
A. Droplet Trajectory Equation
Supercooled water droplets follow trajectories that may or may not result in an impact on the surface of the
airfoil. The first and most important step in icing analysis is to get the locations of the impacts on the airfoil surface
range and compute the effectiveness of a body in collecting the water droplets. Collection efficiency decides the
shape and size of ice on body. The impacted locations are usually found using a Lagrangian approach, where
differential equations describing Newton's Second Law for a single droplet are solved iteratively. As opposed to the
Eulerian approach, the Lagrangian methodology follows a control volume (the liquid droplet) as it passes through
the region of interest. For the reason of the constringency is very slowly on multi-block grid with Eulerian method,
we used a improved lagrangian trajectory approach to obtain droplet impingement limit and collection efficiency.
Calculation of water droplet trajectory is the most time-consuming step in the ice accretion process. This is due
to the reason that hundreds of droplets must be tracked individually over their entire path
7

8
. To calculate the
trajectory of a droplet, the following assumptions are sensible
9
.
. The droplets are spherical without any deformation or breaking
Figure 1 The process of generate a structured grid

American Institute of Aeronautics and Astronautics


4
. No droplets collision, coalescence or splashing;
. No heat and mass exchange between the droplets and the surrounding air
. Turbulence effects on the droplets can be neglected;
. The only forces acting on the droplets are due to air drag, gravity and buoyancy.
The basic equation relating droplet trajectories is derived from Newton's law. In this equation, we take into
account the buoyancy, gravity and drag force. The main objective of the droplet trajectory calculations is the
simulation of the motion of droplets which impinge on the foil. The droplet trajectory is calculated by integrating the
droplet equation of motion over many time steps and an efficient scan procedure of the frontal surface of the wing
permits the calculation of the flux of water impinging on each panel forming the surface. The equation of motion of
the droplet is given by
10

( ) ( )
1
2
d
d d d a d a d d d d
du
v v g AC u u u u
dt
= +


3
where
a
and
d
are the densities of air and water,
d
v is the volume of the droplet,
d
A is the area normal to the free
stream direction, u

and
d
u

are the velocity vectors of the air and the droplet, and
d
C represents the drag coefficient
of the droplet. The drag coefficient for a spherical particle is given by the following empirical relation:
0.63 4 1.38
24
(1.0 0.197 2.6 10 )
d r r
r
C Re Re
Re

= + + (4)
where
r
Re is the relative Reynolds number which is based on the relative speed between the gas flow and droplet
flow and is defined as:
r eq d
a
Re d u u =

(5)
B. Impact point calculation
The droplet equation of motion can write as the form as a one order Differential equation:
( )
( )
0
,
d
d
d
du
f u
d
u u


(6)
It can be solved using a fourth-order Runge-Kutta numerical integration procedure. The initial droplet position is
determined and its new position is recalculated after a time interval t , while a test on impact points gives whether
the droplet hits the body or is outside.
( )
1 1 2 3 4
1
2 2
6
n n
u u t K K K K
+
= + + + + (7)
In the equation,
1
K ,
2
K ,
3
K ,
4
K is :

( )
( )
1
2 1
3 2
4 3
,
1 1
,
2 2
1 1
,
2 2
,
n n
n n
n n
n n
K f t u
K f t t u tK
K f t t u tK
K f t t u tK

= + +


= + +



= + +

(8)
The position of the droplet at + can be computed as follow:
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
d d d
x
d d d
y
x x u
y y u

+ = +

+ = +

(9)
We must calculate the grid cell which the droplets belong to and compute the velocity, density and temperature in
that location every time step. Its the most time-consuming step of the calculation of droplet trajectory. Traditionally,
we calculate the cell by they grid serial number directly, this method must search the cell in the entire gird block one
grid cell after another, its very slowly. Based on dichotomy, we propose a recursive box method to calculate the
locations of particles. It is efficient, general and robust. Compared with the traditional direct method, it obtains the

American Institute of Aeronautics and Astronautics


5
same result but reduces computational time by one order of magnitude.
Moreover , the denser the grids are , the higher efficiency it is.
Obviously, the four boundaries of a grid block could compose a close
curve, if the position of the water droplet is inside of the close curve, we
can consider the water droplet is in the grid block. In this paper, we use a
radial method to estimate the position relation between a point and a
closed curve. For example, we want to judge if the point A is inside or
outside of the close curve 123n. We could make a radial through the
point A to any direction, if the radial has odd number intersections with the
close curve, the point is inside of the close curve. Otherwise if the radial
has even number intersections with the close curve, the point is
outside of the close curve. In the code used in this paper, the
curves are composed by several line segments, if the
intersections between the radial and the curves contain one or
more vertex of the line segments, we should make a new radial
to calculate the relation. The process of finding the grid cell
which was the droplet belongs to is as follows.
1) Estimate the position relation between the water droplet
and the close curve which is composed by grid boundaries.
If the water droplet is outside of the close curve, go out of
the circulation.
2) Divide the grid block into 2 grid blocks along the edge on
which have the most grid points
3) Estimate the relation of water droplet position and the
close curve composed by the boundaries. if the water
droplet is not inside of the first grid block, it must in the
second grid block.
4) Do the step 2 and 3 until the grid points on each edges of
the block is no more than 2. Then the water droplet is in
this grid cell.
Take a simple grid block for example. As shown in figure 4,
to search the grid unit in the grid block ABCD which the point
O belongs to. The first step in recursive box searching method,
according to the radial method, we could calculate the Point O
was inside or outside of the close curve ABCD which is composed by the 4 grid boundaries. If the point O is outside
of the closed curve, end the circulation. Obviously, in this example, the point is inside of the closed curve. Then
continue to the second step, because the points on the direction of I is
more than the direction of J, gird block ABCD is divided into two grid
blocks of ABFE and CDEF along the I direction. We can see the point O
is in the closed curve EFCD by the radial method. Continue with the
cycle, we could found that point O belonging to the grid cell KLIG at
last. At this moment, the number of points on each edge is only 2, we
marked the grid number and end the circulation.
The method used in this paper need to divide the grid block
continuously, so this method is called recursive box method. This method is general,
it need few additional information, it only need to know the beginning and the
ending grid index of the grid block.
The velocity, density and temperature in the Point O can be calculated with the
rule as follows. If point O is belong to the quadrangle 1234 as figure 5, and the
velocity, density and temperature in the four verticals 1,2,3,4 are
1 1 1
( , , ) u T ,
2 2 2
( , , ) u T ,
3 3 3
( , , ) u T ,
4 4 4
( , , ) u T .
14
d ,
23
d ,
12
d ,
34
d are the distance
from point O to the line 14,23,12,34.We could get the value in point O with the
following formula.

Figure 2 The radial method
Figure 5 a calculate cell
Figure 3 The flow chart of recursive box method
Figure 4 process of searching a grid unit

American Institute of Aeronautics and Astronautics


6

0 1 4
2 3
(1 )(1 ) (1 )
(1 )
W W W
W W


= +
+ +
10
where and are genes relative to the distances between
the point O and four borders.
13 13 24
/ ( ) d d d = +
12 12 34
/ ( ) d d d = +
W in (10) represent u and T .
In the grid of multi-element foil, every element could
compose a close curve. We should get the control domains of every closed curve as the Figure 6 shows. When the
trajectory and the control domain have overlapping domains, compute the number of intersection by the trajectory
and the closed curve. If the number is not zero, end the calculation of that trajectory. When the number is more than
one, the impact point is the intersection near the start point of the trajectory. The collection efficiency can be
computed at all the wall boundaries. Thermodynamic module for heat transfer characterictics computation by means
of a boundary layer integral method. Ice accretion on single or multi-element airfoils based on the classical
Messinger model. With this model, we could get the quality of the ice in every control volume, and the volume of
the ice can be calculated. After that, we could compute the boundary displacement due to ice accretion. The
geometry description and grid need to change in response to the ice accretion on the airfoil. Then do the new icing
cycle as follows until reaching the target icing time, generate new multi-block grid, simulate the air flow field, and
calculate the water collection efficiency, then computer the new foil.
V. Results
The numerical results are compared to the experimental data to
assess the CFD capabilities of the developed methods including grid
generation and numerical solver.
We calculate 2000 droplet trajectories in a grid block which have
the points in I and J direction are 197 and 40 need 145s with the direct
method, and it need 112s with the recursive box method. It can save
22.8% time in this grid block of 7880 cells. We calculate 2000 droplet
trajectories in a grid block which have the points in I and J direction
are 245*50 need 185s with the direct method, and it need 125s with
the recursive box method. It can save 32.4% time in this grid block of
12250 cells. We calculate 2000 droplet trajectories in a three-block
grid which have 176*35,176*30,176*35 points need 317S with the
direct method, and it only need 121s with the recursive box method. It
can save 61.8% time in this grid block of 17600 cells. As show in
figure 7, obviously, the recursive box method can reduce
computational time, the denser the grids are , the higher efficiency of
recursive box is.
A. Cylinder
We computed for a circular cylinder with diameter of 0.1016m. The free-
stream gas velocity is 80 m/s, and the ambient pressure is 89867 Pa. The
incoming droplet LWC is 1g/m
3
, and the mvd is 16m. Figure 8 shows the
comparison of collection efficiency between the simulation results and
experimental data. We can see that only the calculation results in the stagnation
point is bigger than the experimental value, but the values in the other points is
good agree with experiment.
B. Ice Accretion for NACA 0012 Airfoil
TABLE 1: METEOROLOGICAL AND DYNAMIC CONDITIONS
Case 1 Case 2 Case 3 Case 4
v(m.s
-1
)

67.05 67.05 67.05 67.05
T(
o
C) -4.4 -6.1 -11.1 -11.1
Figure 6 control domains of foil
The grid cell number
T
h
e
s
a
v
e
d
t
i
m
e
(
%
)
10000 15000 20000 25000
0
10
20
30
40
50
60
70

Figure 7 Comparisons of collection efficiency
S/C
c
o
lle
c
t
io
n
e
f
f
ic
ie
n
c
y
-0.6 -0.4 -0.2 0 0.2 0.4
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Cal
Exp
Figure 8 Comparisons of collection
efficiency

American Institute of Aeronautics and Astronautics


7
lwc(g.m
-3
)

1 1 1.05 1.05
mvd(m) 20. 20. 20. 20.
(deg) 4 4 4 0
Chord(m) 0.5334 0.5334 0.5334 0.5334
P(Pa) 101300 101300 101300 101300
t(s) 360 360 240 180

The ice accretion on NACA0012 airfoil is simulated by the above methods. The numerical simulations of ice
accretion are performed under the geometric, aerodynamic, and meteorological conditions shown in Table 1. Figure
9 and 10 give the computed ice shape with the experimental results. We can see in figure9 and 10 that the ice shape
we calculated is agreement with experimental data on the whole.
Figure 11 shows the local collection efficiency for different MVD against the arc length s of the airfoil. We can
see that the larger the droplet size, the larger are the impingement limits and the peak collection efficiency values.
This is because the larger inertia of the larger droplets makes them less affected by the gas flows and have more
impact domain.
C. Ice Accretion for Multi-Element Airfoil
Based on the above method, we simulate the ice accretion of a two-element airfoil (GA-W(1)) and three-element
airfoil. Figure 13 and 15 show the droplet track around the foils, the area where iced can see clearly from the figures.
Figure 12,14 are the grids generated by the method we have introduce above. Figure 17 and figure 18 shows the lift
coefficient and drag coefficient in different time in the process of icing on the three-element airfoil. We can see in
-0.05 0 0.05 0.1 0.15
-0.05
0
0.05
Cal
Exp
Foil

V1;X
V
2
;Y
-0.05 0 0.05 0.1 0.15
-0.05
0
0.05 Cal
Exp
Foil

s/c
c
o
lle
c
t
io
n
e
f
f
ic
ie
n
c
y
-0.2 -0.1 0 0.1 0.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
MVD=20
MVD=30
MVD=40
MVD=50

Figure 9 iced foil (case 1) Figure 10 iced foil (case 2) Figure 11 Comparisons of collection
efficiency with different mvd




Figure 12 Grid of a two-element foil Figure 13 The water trace(case 3) Figure 14 Grid of a three-element foil



' t'
'C
l'
50 100 150 20
1.9
1.91
1.92
1.93
1.94
1.95
1.96

't'
C
d
0 50 100 150 200
0.0775
0.078
0.0785
0.079
0.0795
0.08
0.0805
0.081
0.0815
0.082
0.0825
0.083
0.0835
0.084
0.0845

Figure 15 The water trace (case 4) Figure 16 Ice shape of Figure 17 Cl in different time Figure 18 Cd in different time
leading-edge flap (case 4)

American Institute of Aeronautics and Astronautics


8
the figure than the lift coefficient is more and more smaller and the drag coefficient is more and more bigger with
icing accretion.

VI. Conclusion
In order to study icing accretion on airfoil and multi-element airfoil, one software package is being developed
including grid generation, air flow simulation, water droplet trace calculation, ice shape computation. This study is
part of a large research program to develop smart icing system technology. The presented results show preliminarily
that the above methods are feasible. This study is primary and the further researches are necessary. Three-
dimensional problems will be simulated next step. This work is now underway.
References

1
Dome, N.D., Karman, S.L., SPLITFLOW: Progress in 3D CFD with Cartesian Omni-tree Grids for Complex
Geometries,AIAA 2000-1006.
2
Van Leer, B, Flux Vector Splitting for Euler Equations, Lecture Notes in Physics,Vol. 170, pp. 501-512, 1982.
3
Edward K. Buratynski and David A. Caugheyt, An Implicit LU Scheme for the Euler Equations Applied to
Arbitrary Cascades, AIAA JOURNAL ,VOL. 24, NO. 1, JANUARY 1986
4
Thomas, P. D., Middlecoff, J. F., Direct Control of Grid Point Distribution in Meshes Generated by Elliptic
Equations,AIAA Journal, Vol.18, No.6, 1979.
5
Hilgenstock, A., A Fast Method for the Elliptic Generation of Three Dimensional Grids with Full Boundary
Control,Numerical Grid Generation In Computational Fluid Mechanics88, Pineridge Press Limited, pp.137-146,
1988
6
Weimin Sang Shengju Jiang and Fengwei Li, Icing Research for Airfoil and Multi-element Airfoil,AIAA 2006-3647
7
Jaiwon, S., Thomas, H. B., Repeatability of Ice Shapes in the NASA Lewis Icing Research, Journal of Aircraft,
Vol.31,No.5, September-October, 1994
8
Jaiwon, S., Brian, B., Hsun, H. C., and Tuncer, C., Prediction of Ice Shapes and Their Effect on Airfoil Drag,
Journal ofAircraft, Vol.31, No.2, March-April, 199
9
Yves Bourgault,Wagdi G Habashi,Julien Dompierre and Guido S.Baruzzi A Finite Element Method Study of
Eulerian Droplets Impingement Models , Internernational Journal for Numerical Methods in Fluids Int. J. Numer.
Meth. Fluids 29: 429449 (1999)
10
S. Nilamdeen and W. G. Habashi , FENSAP-ICE: Modeling of Water Droplets and Ice Crystals, ,AIAA 2009-4128
bout to be published

Вам также может понравиться