Professional Documents
Culture Documents
Final Project Report415
Final Project Report415
Final Project Report415
Team Poisson
ENME331-0201
Fluid Mechanics
Keya Gemechu
Cooper Gilbert
Sahil Kulgod
Kyle McDaniel
Joy Stephens
Page 2
Table of Contents
Introduction
Physical Phenomena
Dimensional Analysis
Buckingham Pi Theorem
Generation of Pi Terms
Equation of Motion
Scaling Analysis
10
12
Results
15
Conclusion
19
Sources
20
Page 3
Introduction
Commercial aviation is a ubiquitous feature of modern life. Hundreds
of millions of people around the world each year travel by air on millions of
individual flights, using what is considered one of the safest methods of
mass transportation after decades of operation with a relatively few number
of accidents. This reputation for safety is well deserved, and largely due to
the diligence of those investigators who examine the accidents that do
occur, to ascertain their causes and prevent the circumstances that
precipitated those accidents from happening again. These investigations are
not limited only to mechanical failures but also include conditions that
might be experienced by an otherwise normally functioning aircraft and
which can lead to catastrophe. It is with this in mind that we consider what
happens to an aircraft under icing conditions. Aircraft often encounter
conditions where ice may accumulate on the skin of the fuselage or
elsewhere on the body, both during flight and on the ground. However,
there are certain circumstances during flight when such accumulation,
along the wing surfaces specifically, can cause the aircraft to rapidly lose
aerodynamic character and stall. Given the aforementioned prevalence of
commercial aviation, if there is a specific set of circumstances where a
catastrophic accumulation of ice can occur, it must be identified so that it
can be avoided, and possible accidents averted.
In the following study, we will assess the viability of using scaledmodel wind tunnel testing to simulate actual conditions at altitude and
identify the critical conditions at which ice will collect on the wing surfaces
and to what degree under those conditions. To comprehensively preform
this analysis, we will mathematically describe the forces acting on a
representative particle of supercooled water, from which the ice is formed,
in the air flow field, use known significant quantities to create a model
scale, and finally use a complicated computer simulation in MatLab, of the
Page 4
force equations to show the system in a Lagrangian and Eulerian
specification. This simulation will both graphically display and
computationally solve the efficiency of the wing surfaces at collecting ice
under the specified conditions. From this computation, we will determine if
the scaled-model is an appropriate representation of the actual phenomenon
of wing icing, and thus if this method can be used in the future to help
prevent accidents and continue the reputation of air travel as being one of
the safest methods of travel in the world today.
Page 5
Physical Phenomena
If we are to generate a model that is capable of simulating wing icing
conditions experienced in flight, we must first understand the factors
involved in determining how that ice accumulates: first qualitative, then
mathematical. In this section, we will determine the qualitative aspects.
For wing icing to occur, an aircraft must fly through a cloud in which
the droplets of water are both sufficiently small and supercooled that they
freeze instantly upon contact with a solid surface. These droplets travel
relative to the free stream velocity until the aircraft is close enough to the
droplet that the airfoil changes the airflow in which the particle is
suspended. The particles that will accumulate to form ice are those for
which the drag forces acting on the droplet are such that the streamlines of
the particle intersect with the airfoil, causing the particle to impact the
wing and freeze. With this understanding of the process of icing, we can
immediately identify several factors that impact the efficiency of ice
collection on the wing.
The first of these significant factors is the size of the droplet. By using
the density of water, we can use the size of the droplet to determine its
weight, which acts as a downward force in the equations of motion
describing the droplets path in the flow field. Another factor that also
comes from the size of the droplet, and is more significant in its impact, is
the cross-sectional area, a component of the drag force. The larger the
cross-sectional area of the droplet, the greater the drag force acting upon
the particle, driving the particle into the wing and increasing the collection
of ice.
Another component of the drag force that we must include in our
analysis is the velocity of the particle. In real world conditions, the particle
would be relatively stationary and the velocity of the aircraft is the variable
with which we are concerned, however for the model analysis, the airfoil is
Page 6
stationary and the particle is given a relative velocity by the air flow field.
This difference should not significantly affect our modeling results. As the
velocity of the particle increases in the wind tunnel, the drag force acting
on the particle increases proportional to the square of the relative velocity
in both the x and y directions. The resultant increase in drag is, as it was
with the increase due to cross-sectional area, leading to an increase in the
collection of ice. Velocity is also a component in the Reynolds number,
another factor in the collection of ice on the airfoil, however here we are
somewhat limited by the model. Because we are modeling the aircraft on a
smaller scale, the length of the airfoil is smaller as well. The Reynolds
number for the airfoil is a measure of the characteristics of the air flowing
over the airfoil, but because the chord length is less than the actual aircraft,
the air does not have the same distance to develop the airflow
characteristics that it would for an actual flight. As such, the Reynolds
number will not be accurately modeled by our experiment. Our study may
be able to show later that this is acceptable as long as another factor, called
the Stokes number, is roughly equivalent between the model and the full
scale. For the purposes of qualitative explanation, the Stokes number is a
measure of the response time of the particle to the flow field. If the particle
has a small response time, then the resulting Stokes number is also small,
and if that is less than one, than the droplet should follow the flow field
closely. As long as both the model and the full scale have Stokes numbers
below one, we can ignore difference in the Reynolds numbers and accept
our model conditions as showing similitude, scaled equivalence of the
values between the model and the full scale.
A final factor that must be considered as significant in the trajectory
of the droplet is the geometry of the wing. There are two primary geometric
variables: chord length, which is the straight line length from the leading
edge to the trailing edge of the wing, and the angle of attack, which is the
pitch of the wing in the flow. While it may not be obvious, the chord length
is not significant in the trajectory of the droplet. The reason for this is that
Page 7
the chord length does not affect the streamlines of the airflow. The angle of
attack, however, is significant to the accumulation of ice, as it dramatically
changes the flow characteristics. As the angle of attack increases, the
disturbance of the airflow field also increases, and since the particle is
acted upon by drag, which separates the particle from the streamlines, the
droplet has a greater opportunity to impact the wing surface. These
descriptions are strictly qualitative, however with this qualitative
understanding, we can now select the variables that we will need to
describe the system mathematically.
Page 8
Dimensional Analysis
The following variables were chosen for analysis based on significant
factors that affect the amount of ice accumulation in the system.
Symb
ol
Units
Description
D/I
m
m
kg
m s
radians
m
s2
kg
m3
kg
m3
lc
va
m
s
I
I
Page 9
Buckingham Pi Theorem:
This type of analysis uses the concept of repeating variables to develop
dimensionless products. This concept simplifies the calculations to
accurately predict the behavior of a similar system.
First we choose three of our selected variables to become the repeating
variables.
a
M L3
lc
L
L
va
T
LT 1
Table 2: Repeating Variables
M
The selected variables (k) contain M (mass), L (length), and T (time), thus
the repeating variables (n) must contain all three components of the MLT
system.
Number of Pi Terms:
k n
103
Pi Terms Needed
Calculating Pi Terms:
Sample calculation of a non-trivial pi term:
x
2=( a ) ( v a ) ( l c ) ( d ) =
L y ( )z M
L
T
[ ][ ] [ ] ( ) ( )
kg
m3
m
[ m ] kg = M3
s
m s
L
1
0 0
3 1 1
0 1 0
][][ ]
x 1
y = 1
z
1
x=1
y=1
z=1
2=( a )1 ( v a )1 ( l c )1 ( d )1
d
2=
a v a l c
( )
Page 10
This pi term dimensionally corresponds to the inverse of Reynolds
number.
Page 11
Generation of Pi Groups
1
0
0
0
1
( a ) ( v a ) ( l c ) ( )
( a ) ( v a ) ( l c ) ( d )
d
a v a l c
0
0
1
1
( a ) ( v a ) ( l c ) ( D )
D
lc
0
0
0
1
( a ) ( v a ) ( l c ) ( )
0
2
1
1
( a ) ( v a ) ( l c ) ( g )
0
0
1
1
( a ) ( v a ) ( l c ) ( t )
( a ) ( v a ) ( l c ) ( a )
g lc
v a2
t
lc
w
a
Table 3: Pi Groups
1= ( 2 , 3 , 4 , 5 , 6 , 7 )
a v a l c D
gl c t w
, , , 2 , ,
d
lc
v a l c a
Page 12
Page 13
Equation of Motion
Equations of Motion:
F x =m a x
d2 x
=F dx m g x
dt2
Equation 2
d x F dx
=
g sin
2
m
dt
2
F y =m ax
Equation 1
d2 y
m 2 =F dy m g y
dt
d y F dy
=
g cos
2
m
dt
Equations 1 and 2 are the dimensional equations of motion for a single
droplet exposed to a variable fluid velocity.
2
Page 14
m= w V
m= w
4 D
3
2
1
Fd = a V 2 A C D
2
1
2
Fd = a V rel|V rel| D C D
2
4
2
a V rel|V rel| D C D
Fd =
8
( )
w D 3
m=
6
Therefore:
d2 x
=
2
dt
a V relx|V rel| D C D
8
w D
6
g sin
d 2 x 3 a V relx |V rel|C D
=
g sin
w D
d t2 4
d2 y
=
2
dt
w D
6
g cos
d 2 y 3 a V rel y |V rel|C D
=
g cos
w D
dt2 4
If certain conditions are met (shown in next section), dimensionless
variables can be expressed as:
t |V rel|
x
y
x =
y =
t=
lc
lc
lc
Therefore:
Page 15
2
d ( x lc )
t lc
d
|V rel|
=
2
( )
3 a V relx|V rel|C D
g sin
4
w D
Equation 4
g lc
d 2 x 3 a V relx l c
=
C D
sin
2
2
4 w |V rel| D
dt
|V rel|
Equation 3
d ( y l c ) 3 a V rel y|V rel|C D
=
g cos
2
4
w D
t lc
d
|V rel|
2
( )
2
g lc
d y 3 a V rel y l c
=
C
cos
2
4 w |V rel| D D |V rel|2
dt
D
v 2a
3=
7= w
5=
lc
a
gl c
are the most important pi terms (Table 3) are needed to maintain during
scaling analysis.
Page 16
Scaling Analysis
Variab
le
Angle of
Attack
Unit
s
[rad ]
[ ]
[ ]
[ ]
Prototy
pe
Value
Model
Analysis
0.1396263
kg
3
m
1.225
kg
3
m
1000
m
s2
9.81
Chord
Length
[m]
1.5
l c =1.5
l cm=0.1
Wing
Thickne
ss
[m]
0.15
t m=
l cm
t
lc
Droplet
Diamete
r
[m]
d dm=
l cm
d
lc d
Speed
[ ]
Density
of Air
Density
of Water
Gravity
m
s
Dynamic
Viscosity
[ ]
kg
m s
8.3(105 )
217.5
V m=
l cm
V
lc o
Model
Value
Approach/Rationale
( am v am l cm )
d
a v a l c
Table 4: Scaling Analysis
5
1.79(10 ) dm=
Page 17
for angle of attack, density of water, density of air, and the value of gravity
for the model and prototype value. For the rest of the model, we used a
ratio of 1/15 scaling value.
We used different sources to come up with model scale values. For the
aircraft, we used the McDonnell Douglas MD-11 airplane, which represents
a wide-body jet airliner. The cruising speed, angle of attack, and different
body dimensions were also determined from this aircraft.
As stated earlier, certain conditions must be met in order to make the
assumption of dimensionless variables. Equal values of Reynolds number,
, for each system.
Calculate
=
a v a lc
d
PROTOTYPE
a=1.225
kg
m3
l c =1.5 m
v a =217.5
5
m
s
d =1.79(10 )
kg
m s
MODEL
a=1.225
kg
m3
8
d =7.956 ( 10 )
=1131.24
=1131.18
Equation 5
m
s
l c =0.1 m
kg
ms
kg
m
14.5
(0.1
m)
s
m3
kg
7.956(108 )
ms
kg
m
1.225 3 217.5
(1.5 m)
s
m
=
kg
1.79(105 )
ms
v a =14.5
1.225
Equation 6
Although these two values are equivalent, in the actual simulation, the
Reynolds number is a function of relative velocity, which changes
throughout the experiment. Further MatLab analysis will demonstrate the
effect of a changing Reynolds number.
Page 18
We anticipate that the Reynolds number will not accurately model under
these limitations, therefore, Stokes number,
long as both the model and the full scale have Stokes numbers less than
one, the Reynolds number can be ignored.
Stk=
Vo
dc
where:
= relaxation time of particle
V o= fluid velocity away from particle
d c =
PROTOTYPE
w =1000
kg
m3
w =1000
D=0.076 ( 103 ) m
kg
m3
Dm=5.0667 ( 106 ) m
6
d =1.79 ( 105 )
d =1.1933 ( 10 )
m
kg
m s
2
kg
1000 3 ( 0.076 ( 103 ) m )
m
=
kg
18 1.79 ( 105 )
m s
=0.017927 s
Equation 7
MODEL
106
5.0667()m
2
kg
m3
1000
m=
kg
m s
m =0.001195 s
Equation 8
Page 19
Calculate
Stk
Stk=
Vo
t
PROTOTYPE
t=0. 15 m
V 0=217.5
0. 017927 s 217.5
Stk=
MODEL
m
s
m
s
t m=0. 0 1m
0.15 m
Stk=2 5. 9938
Equation 9
V 0 =14 .5
m
0.001195 s 14.5
St k m =
m
s
m
s
0.01 m
St k m =1.73299
Equation 10
Page 20
Results
Water Droplet Trajectory Model w/ Streamlines
airfoil
droplet misses airfoil
droplet hits airfoil
0.1
y (m)
0.05
-0.05
-0.1
-0.2
-0.15
-0.1
x (m)
-0.05
0.05
Page 21
the prototype exactly; this is likely the cause for the discrepancy between
the two models.
collection efficiency vs angle of attack
120
100
80
model eta
prototype eta - model eta
60
40
20
7
8
9
angle of attack (degrees)
10
11
12
Plot 2: Collection efficiency at various angles of attack of scaledmodel and prototype values.
Our original prediction was that the collection efficiency would be
under 100 percent because in the real world phenomenon of wing icing, the
supercooled large droplets (SLDs) never fully covered along a length that
was greater than the entire projected frontal height of the wing. The ice
only collected on a certain portion of the wing, changing the airfoil profile
enough to affect the flow drastically and cause an accident. According to
Plot 2, both the model and prototype efficiency predict a greater-than 100%
collection efficiency. One possible explanation has to do with the assumption
that the SLDs will always freeze when they contact the airframe. While this
is likely, there may be other factors at play, such as ambient heat from
Page 22
engine exhaust. However, it is more likely that the scaling needs to be
changed in order for the model to work. Unfortunately, due to physical
limitations of wind tunnel testing, this cannot be accomplished, and may
explain why in the real world, this case was tested at full scale, on the
actual airplane.
5
x 10
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.5
-0.4
-0.3
-0.2
-0.1
x position (m)
0.1
0.2
Page 23
number is more realistic can also help guide us to a solution that is more
realistic and better representative of the actual phenomenon.
Page 24
Conclusion
This study analyzes the feasibility of using scaled-model wind tunnel
testing to simulate actual conditions at which ice will collect on the wing
surfaces of a full-scale aircraft. From our research, it can be concluded that
using scaled-model data is not adequate in predicting the accumulation of
ice on a full-scale aircraft wing.
In order to have complete similarity, and thus accurate predictions,
between the model and prototype, three types of similarity must be
maintained. Using model design conditions based on actual parameters of
the McDonnell Douglas MD-11 airplane achieved geometric similarity.
However, we found that this scale could not be used to equate all pi terms,
and attempted use Reynolds number similarity. Due to physical limitations
of model chord length and the resulting airflow characteristics, we found
that Reynolds number would not be ideal for modeling this experiment. We
postulated that if Stokes number could be shown to be less than one for
both the model and full scale, then we could ignore the discrepancies in the
Reynolds number. However, this was not proven to be the case, as indicated
in the Scaling Analysis section. The failure to equate Stokes number results
in a lack of dynamic similarity between model and prototype.
In the Results section, MatLab analysis illustrates general correlation
between the collection efficiency of the prototype and model data, but the
actual values of the model scale are shown as greater than 100%. This is
most likely explained by inaccuracies in the scaling analysis, and cannot be
corrected due to physical constraints of the wind tunnel. Further MatLab
calculations demonstrate the increased Reynolds number of a droplet that
misses the wing. If we were able to adjust the variables to obtain a smaller
Reynolds number, then our predictions of prototype behavior would become
more accurate due to more dynamic similarity of the system.
Page 25
In conclusion, the possibility of catastrophic accumulation of ice
during flight is a very serious and important concern to all of us. The
complex interaction between the many variables shown in this experiment
requires a level of diligence and accuracy that simply cannot be modeled in
a small-scale laboratory wind tunnel. The critical conditions that cause
aerodynamic failure need to be determined at a larger facility to ensure
comprehensive and accurate data to maintain the safety and reputation of
commercial aviation.
Page 26
Sources
Anderson, John D. Elements of Airplane Performance, Introduction to
Flight. 7th Edition, 2012.
Ciornei, Simona. "Aircraft Structural Design." Department of Automotive
and Aeronautical Engineering. 31 May 2005. Web. 20 Oct. 2014.
<http://courses.washington.edu/ie337/Boeing Tour Facts.pdf>.
Munson, Bruce et. al. Dimensional Analysis, Similitude, and Modeling.
Fundamentals of Fluid Mechanics. 7th Edition, 2013.
Phases of Flight. Phases of Flight. n.d. Web. 13 Nov. 2014.
<http://www.fp7-restarts.eu/index.php/home/root/state-of-theart/objectives/2012-02-15-11-58-37/71-book-video/parti-principles-offlight/126-4-phases-of-a-flight>.