Accepted Manuscript

Aircraft icing: an ongoing threat to aviation safety

Yihua Cao, Wenyuan Tan, Zhenlong Wu

PII: S1270-9638(17)31760-1
DOI: https://doi.org/10.1016/j.ast.2017.12.028
Reference: AESCTE 4358

To appear in: Aerospace Science and Technology

Received date: 26 September 2017

Revised date: 13 December 2017
Accepted date: 14 December 2017

Please cite this article in press as: Y. Cao et al., Aircraft icing: an ongoing threat to aviation safety, Aerosp. Sci. Technol. (2018),
https://doi.org/10.1016/j.ast.2017.12.028

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing
this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is
published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 Aircraft icing: an ongoing threat to aviation safety
Yihua Cao1, Wenyuan Tan1, Zhenlong Wu1, 2,*
1
School of Aeronautic Science and Engineering, Beihang University, Beijing 100191, China
2
Alexander von Humboldt research fellow, University of Stuttgart, Stuttgart 70569, Germany
*Corresponding author: jackilongwu@gmail.com
___________________________________________________________________________________________________

ABSTRACT
Keywords: Flight safety is undoubtedly the most important requirement for
modern aircraft design and operation. However, in reality, aircraft
Aircraft flight performance and safety are inevitably affected by adverse
meteorological conditions, one such weather is icing. Aircraft icing
can cause severe aerodynamic and flight mechanical effects, thus
Icing threatens aircraft flight safety. In this paper, a comprehensive review
of the past research on aircraft icing is presented. Special attentions
are paid in the following aspects. First, the causes, types, severity
Flight safety and natural parameters of aircraft icing are introduced. Then, the
various effects of ice accretion on aircraft aerodynamic performance,
stability, controllability as well as the existing estimation methods
SLD icing are summarized and analyzed. Following is a simple introduction to
the recently rising issue on supercooled large droplet (SLD) icing.
Finally, a series of aircraft flight accidents caused by icing in recent
Aviation accidents years are analyzed. The previous lessons should be accepted and
disseminated by later generations to avoid accidents by aircraft icing.

____________________________________________________________________________________________________

Contents

1. Introduction .........................................................................................................................................................................2
2. Causes of icing on aircraft ...................................................................................................................................................4
3. Types and severity of aircraft icing ......................................................................................................................................5
3.1 Aircraft icing types .......................................................................................................................................................5
3.2 Aircraft icing severity level ..........................................................................................................................................6
4. Icing parameters ..................................................................................................................................................................9
4.1 Environmental parameters............................................................................................................................................9
4.1.1 Water droplet diameter ........................................................................................................................................9
4.1.2 Liquid water content ..........................................................................................................................................11
4.1.3 Ambient temperature .........................................................................................................................................12
4.1.4 Cloud characteristics .........................................................................................................................................12
4.1.5 Other environmental parameters........................................................................................................................15
4.2 Aircraft-related parameters.........................................................................................................................................16
4.2.1 Altitude and speed .............................................................................................................................................16
4.2.2 Angle of attack...................................................................................................................................................16
4.2.3 Airfoil ................................................................................................................................................................16

1
 4.2.4 Other aircraft-related parameters .......................................................................................................................17
5. Icing effects on aircraft ......................................................................................................................................................18
5.1 Aerodynamic performance .........................................................................................................................................18
5.1.1 Drag increase .....................................................................................................................................................18
5.1.2 Lift and stall angle of attack reduction ..............................................................................................................18
5.1.3 Flight speed reduction .......................................................................................................................................20
5.2 Stability and controllability ........................................................................................................................................20
5.2.1 Longitudinal stability and controllability degradation .......................................................................................20
5.2.2 Rolling and overturn ..........................................................................................................................................22
5.2.3 Yawing stability and controllability degradation ...............................................................................................23
5.3 In different flight phases ............................................................................................................................................23
5.3.1 Take-off phase ...................................................................................................................................................23
5.3.2 Cruise phase ......................................................................................................................................................24
5.3.3 Approach or waiting phase ................................................................................................................................24
5.3.4 Landing phase....................................................................................................................................................24
5.4 Estimation methods of iced aerodynamic parameters ................................................................................................25
5.4.1 Lift estimation ...................................................................................................................................................25
5.4.2 Drag estimation .................................................................................................................................................25
5.4.3 Arbitrary aerodynamic parameters estimation ...................................................................................................26
5.5 Parameter identification .............................................................................................................................................27
5.6 Control techniques of aircraft icing ............................................................................................................................28
5.7 Other icing effects ......................................................................................................................................................30
6. SLD icing ...........................................................................................................................................................................30
6.1 Characteristics of SLD icing process .........................................................................................................................30
6.2 Review of SLD icing effects on aircraft .....................................................................................................................31
7. Analysis of typical aircraft icing accidents ........................................................................................................................32
7.1 Statistics of aircraft icing accidents ............................................................................................................................33
7.2 Analysis of parameters in aircraft icing accidents ......................................................................................................34
8. Summary and conclusions .................................................................................................................................................37
Fundings ................................................................................................................................................................................37
References .............................................................................................................................................................................37

was first developed in the 1930s [8]. In 1948, Preston and

1. Introduction Blackman [9] performed the first successful iced flight
experiment in which the drag coefficient increased by 81%
Aircraft icing refers to ice accretion on surfaces of an and the pilot obviously felt that the aircraft was almost
aircraft flying in icing conditions. According to analyses of beyond control.
the causes for the aircraft accidents in recent years, aircraft
icing is a major external cause. Until 1988, the number of
flight accidents and events caused by aircraft icing was as
high as 542. Most of the accidents were fatal, originating
from the general aviation. After doing a statistic work on
aircraft icing incidents and accidents in the recent 15 years,
the US National Transportation Safety Board (NTSB)
acquired the above conclusion (Fig. 1). According to the
International Civil Aviation Organization’s (ICAO) statistics
in the years from 1986 to 1996, the passenger mortality rate
reached 39% among 42 aircraft icing accidents [1]. These
aircraft icing accidents have caused wide public concern
[2-7]. In view of the serious harms caused by aircraft icing, Fig. 1 Statistics on aircraft icing accidents in the recent 15 years [10]
the research about aircraft icing can be traced back to the Based on the data accumulated by the Beech's King Air
beginning of the last century, and a simple de-icing system aircraft for nearly 20 years, Ashenden et al. [11] studied the
2
effects of aircraft icing, especially under freezing drizzle. low cost. LEWICE, ONERA, and FENSAP-ICE are current
They also found that the degradation rate of flight well-known icing simulation software.
performance best reflected the severity of icing. Based on Aircraft icing influences the flight safety in many
limited flight data, Leckman [12] estimated the performance ways:
parameters of the Cessna Centurion and Air Overlord 1) Under icing conditions, the aircraft’s maximum lift
aircraft in the icing conditions and obtained satisfactory coefficient and the slope of lift curve decrease, the drag and
results. the critical stall speed increases. At this point if the pilot
To date, the most complete flight data are NASA's does not pay attention to the changes in the airspeed and
experiments on the DHC-6 twin otter aircraft (Fig. 2). The climb rate, the aircraft may be near the stall boundary
performance of the aircraft under different icing conditions [22-27].
or with different ice shapes was measured in detail and the 2) Wing icing can lead to a decrease in the airfoil stall
aerodynamic derivatives were identified by the modified angle of attack, while tailplane icing may cause a tailplane
maximum likelihood estimate method and modified stall, especially in flap downwash flows. Both situations
stepwise regression [13]. Simultaneously, the effects of may cause pitch instability and further lead to a crash.
aircraft thrust, landing flaps, and angle of attack were also 3) Additional rolling moment can be caused by the
studied. NASA also conducted a test flight in conjunction following situations: asymmetric icing, one side anti-icing
with the FAA on the Tailplane Icing Program [14-15] (TIP), system breakdown, and formation of the ice ridge by water
which provided information on the behavior of the stability droplets beyond the ice protection system. The first two
and control characteristics [16]. It was found that the effect may only limit the lateral control, but the third is likely to
of aircraft icing increased with the increasing angle of cause rolling and overturn. One typical accident was the
attack. ATR-72 crash [101] in 1994. At that time, the airplane was
at a severe level of icing condition. The co-effect of the
electric heating de-icing system at the wing leading edge
and the natural conditions formed an ice ridge on the second
half of the wing, resulting in a negative pressure zone on the
one side’s aileron. In a routine turn, the negative pressure
zone led to an automatic deflection of the aileron, therefore
the aircraft immediately lost control, rolled, capsized and
finally crashed.
4) The control efficiency reduces [28]. Flight tests
show that, icing on tailplane may produce increased stick
force that may reach hundreds of pounds in weight and
increase difficulty in the manipulation. Icing on rudder
movable structure may cause the rudder surfaces stuck and
the aircraft uncontrollable. Icing on flap leading edge will
cause the air separation in advance on the flap surfaces and
Fig. 2 NASA Glenn Research Center Icing Research Aircraft [15] slump in the flap efficiency.
Recently, research on the experimental technology of This paper systematically reviews the various issues of
icing wind tunnel has been obtaining continuous progress. aircraft icing. The cause of aircraft icing is presented in the
IRT wind tunnel at NASA Glenn Center and IWT wind second chapter. The types of aircraft icing and definition of
tunnel in Italy are famous icing wind tunnels in the world. the severity are presented in the third chapter, including
Addy et al. used the icing ratio theory and its scale-size three main types of aircraft icing and four standards of
equation for large-scale or full-scale experiments [ 17 ]. determining aircraft icing severity level. The main
Papadakis et al. conducted an experiment on the impinging parameters of aircraft icing are discussed in the fourth
characteristics of supercooled water droplets and chapter, including the environmental and aircraft-related
investigated the effects of back sweep on these parameters. The effects of ice accretion on aircraft are
characteristics [18]. Bragg et al. simulated the growth of ice presented in the fifth chapter, which are classified into three
on iced surface and the aerodynamic characteristics of the parts, i.e., effects on aerodynamic performance, effects on
iced wing [19-20]. Perkins studied the ice accretion under stability and controllability and effects on different flight
anti-icing system and designed an anti-icing system by its phases. Estimation methods of iced aerodynamic parameters
results. Bragg and his team also designed a smart anti-icing are also reviewed in this chapter. Then, the characteristics of
system [21] using technologies like neural networks. At the Supercooled Large Droplet (SLD) icing, which has been
same time, numerical simulation has become a popular focused on by some airworthiness regulations in recent
method of research due to its short experimental cycle and years, are presented in the sixth chapter where the
3
differences in the icing process between conventional and
large supercooled water droplet ice accretions are
emphasized. Followed is a review of the SLD effects on
aircraft icing. Finally, some analyses of typical aircraft icing
accidents are performed in the seventh chapter.

2. Causes of icing on aircraft

The common causes of aircraft icing are as follows:

1) Aircraft Encounters clouds with supercooled water
droplets in flight. Supercooled water droplets refer to
droplets that remain liquid under icing temperature [21]. Fig. 4 Fusion as the main phenomenon of the formation of supercooled
Supercooled water droplets are often present in clouds at droplets [99]
temperature between 0 and -20°C, sometimes even -40°C Supercooled water droplets are very unstable and will
and can be formed in a variety of ways. Supercooled water freeze once disturbed. During flight, due to its small mass
droplets can be formed when snow falls from the sky and and inertia, the air molecules can completely bypass the
melts through warm air. This happens in a greater chance aircraft surface and flow downstream. But supercooled
when cold and warm air masses encounter and form an water droplets are larger in mass and inertia, so a fraction of
inversion layer. In Fig. 3, the horizontal axis is the them will track through the airflow and impact the airplane
temperature distribution and the vertical axis is the altitude surface, resulting in freezing and icing of supercooled water
distribution. Cold air mass is at a higher altitude. When droplets, as shown in Fig. 5 [29].
snow is formed in a cold air mass, it inevitably falls through
a warm air mass at a lower altitude, then the ice crystals
melt and form supercooled water droplets.

Fig. 5 Typical air and water droplet flow around a typical airfoil [29]
In the case of aircraft icing, the most common situation
is that a large number of supercooled water droplets impact
onto the aircraft surface. Water droplets with low energy are
quickly frozen on the aircraft surface due to the effect of
low temperature. Other droplets with higher energy are not
immediately frozen. They flow along the surface until the
Fig. 3 Inversion layer mechanism [99]
energy is depleted and are then frozen in the downstream
Another way of forming supercooled water droplets is
area. This is why there exist different shapes of ice.
the impacting small water droplets in the clouds (Fig. 4).
2) Aircraft surface has already been contaminated
Since the water droplets are different in size and falling
before take-off.
speed in the clouds, they tend to merge with each other to
3) Aircraft encounters a high concentration of ice
become larger droplets. Other disturbances in the clouds
crystals in flight. The phenomenon is not common because
also aggravate this situation. Droplets with diameter larger
ice crystals are not viscous and do not easily freeze on
than 50 ȝm are referred to SLD. Freezing drizzle and
aircraft surfaces, but it can adversely affect the engine
freezing rain are two ways in which SLD exist.
performance.

4
3. Types and severity of aircraft icing continues to freeze and form a single or double-horn shape.
The results from five icing research flights by Mikkelson
3.1 Aircraft icing types [33] demonstrated that glaze ice affected the performance of
aircraft far more seriously than rime or mixed ice. Kind et al.
Usually, aircraft icing can be classified into three types: [34] reviewed experimental and computational methods for
rime ice, glaze ice and mixed ice [30]. A comparison of the simulating in-flight ice accretion and indicated that glaze
rime ice and glaze ice conditions is presented in Table 1, icing is much more complex than rime icing and thus much
where “LWC” refers to liquid water content. more difficult to simulate computationally. Cao et al. [35-37]
Table 1 Comparison of rime and glaze icing conditions [31] developed some simple and reasonable mathematical
Conditions Rime ice Glaze ice models to simulate ice accretions on three-dimensional
Cold: less than bodies directly. The predicted glaze ice on the GLC-305
Temperature Warm: 0 to -10°C
-10°C
LWC Low High wing shows an obvious double-horn shape. Unlike rime ice,
Density Low High this irregular shape may severely damage the flowfield
Airspeed Low High characteristics of the wing, and cause lift loss, stall angle of
Color Milky/Opaque Glossy/Clear attack decrease and drag increase [38-39].
Texture Rough Smooth
Runback No Yes
Fragility Fragile Hard
Water Droplet
Small Large
Size
Airfoil Ice Single or Double
Streamlined/Spearheaded
Shape Horn
1) Rime ice
Rime ice is formed through the contact of tiny
supercooled water droplets with surfaces below freezing
temperatures. It usually appears in an environment of low
airspeed, low-temperature, low liquid water content and Fig. 7 Typical glaze ice shape [32]
small droplet diameter. As soon as the supercooled water 3) Mixed ice
droplets reach the wing surface, they immediately freeze In reality, aircraft icing often has the characteristic of a
and remain a hemispherical shape, thus forming a spear-like mixture of glaze and rime ice with no fixed ice shape,
ice shape on the leading edge (Fig. 6). because liquid water content and water droplet diameter in
the atmosphere vary widely. Mixed ice is formed in the
clouds with a temperature from -20°C to -10°C. In this
range of temperature, usually a variety of sizes of
supercooled water droplets exist in the clouds, so the mixed
ice characterizes both glaze ice with double-horn shape and
milky white rime ice (Fig. 8).

Fig. 6 Typical rime ice shape [32]

2) Glaze ice
Glaze ice is formed in relatively warm conditions. An
environment of large water droplets, high liquid water
content, and high airspeed makes it difficult for water
droplets to immediately freeze on the wing surface.
Therefore, the droplets flow along the surface until they are
blown away by the aerodynamic forces, otherwise they will
freeze somewhere. As the water droplets are closely
combined, glaze ice generally has a greater density and a
smoother appearance. Glaze ice is usually tightly attached
to the aircraft surface, thus is more difficult to remove. It is Fig. 8 Typical mixed ice shape [40]
not easily observed, especially at night. Mostly, its shape is Different types of ice may have different influences on
a single or double horns (Fig. 7), because the water the flight performance, and sometimes even the same kind
overflows along the direction of airflow. In this process, it of ice may show different effects under different icing
5
conditions. Therefore, the aircraft icing severity is needed to n is the freezing fraction, first proposed by Messinger,
determine the level of influence that aircraft icing has on the and was defined as the mass of the frozen supercooled water
flight performance. droplets divided by the mass of the supercooled water
droplets impacting on the surface.
3.2 Aircraft icing severity level ݉௙௥௘௘௭௘
݊ൌ                                  ሺ͵Ǥʹሻ
Standard by Aeronautical Information Manual: ݉௜௠௣௔௖௧
The following standards come from the Aeronautical E is the collection efficiency. It is defined as the mass
Information Manual [41] which was originated in the 1960s, of the impacting supercooled water droplets divided by the
and are the most widely accepted criteria for aircraft icing mass of the supercooled water droplets coming from the
severity evaluation. forward projection across the wing section.
݉௜௠௣௔௖௧
1) Trace icing ‫ܧ‬ൌ                               ሺ͵Ǥ͵ሻ
Trace refers to the slightest ice formation and acts as a ݉௣௥௢௝௘௖௧௜௢௡
“white line” only on the wings’ leading edge. At this point, LWC is the liquid water content, V is the flight speed,
the ice formation rate is only slightly higher than the and ߩ௜௖௘ is the density of ice.
sublimation rate. Unless the icing state lasts more than one The meaning of this formula is the thickness of ice per
hour, the icing protection system (IPS) is generally not minute, where the numerator represents the mass of ice per
required. unit area in an hour. The criterion for judging the icing
2) Light icing severity level by ‫ܬ‬଴ is shown as follows (Table 3).
If the airplane is in a state of icing for more than an The NCCAM icing standard (Table 4) is similar to this
hour, it may affect the normal flight. At this point, the above standard [44].
"white line" on the wings’ leading edge begins to expand, Standard by aircraft icing thickness:
and the flight performance is affected. Thus, the ice In reality, it is the maximum thickness and the position,
protection system is required. rather than the rate of accretion, that affect the flight
3) Moderate icing performance. If the airplane does not stay for long in an
In this case, even a very short period of icing may pose icing state, it will be little affected even if the icing
a threat to the flight safety, because the consequent condition is intensive. On the contrary, if the airplane is in
performance degradation is obvious. Even with the help of icing state for a long time, it is possible to form a more
icing protection systems, the icing state should last no salient ice angle or ice ridge, and cause severe damage to
longer than 45 minutes, as required by airworthiness the flight performance. So there also exists an aircraft icing
regulations. Meanwhile, the ices should be cleared away severity level evaluation standard based on icing thickness
from the dangerous airspace immediately if there is no icing (Table 5).
protection system. Due to the complexity of aircraft icing, the ice shapes
4) Severe icing are often complex and varied, solely ice thicknesses cannot
At this point, even if the icing protection system works accurately describe the severity of aircraft icing. Therefore,
well, it still can not prevent the disastrous ice accretions, so on the one hand, if there is an abundant data accumulation,
all kinds of aircraft should return to the safe airspace as the existing ice can be compared with the most severe ice to
soon as possible. determine a dimensionless scale for expressing the
The standard is set to give the average pilot a reference proportion of the icing coefficient. On the other hand,
for reporting the state of icing to the control tower, so the simulation may be carried out directly to observe the
classification is qualitative and vague. Due to this feature, performance degradation of the iced aircraft relative to the
the standard in recent years is widely criticized. In addition clean one as the basis for assessment.
to this qualitative classification, the US Air Force also Standard by ice angle or ice ridge:
defines the severity of aircraft icing with liquid water Ice angle or ice ridge refers to the angular protrusion
content and ice collection rate (Table 2) [42]. In the past, on a glaze ice surface. It will seriously damage the local
this definition was rarely used. However, with the flowfield and result in deterioration of the airfoil
development of satellite detection technology, the purely aerodynamic performance. Therefore, the characteristics of
meteorological parameters, such as liquid water content and ice angle or ice ridge can also be used to assess the severity
temperature, have become important criteria for predicting level of aircraft icing.
the risk of aircraft icing in recent years. Based on the NACA 0012 airfoil, Jacobs used wind
Standard by aircraft icing intensity: tunnel tests to study the effect of the height and position of
Aircraft icing intensity ‫ܬ‬଴ is often used in the actual the artificial ice ridge on the maximum lift coefficient (Fig.
quantitative analysis to represent the icing conditions [43]. 9 and Fig. 10) [45]. The results show that changes in the ice
݊‫ܸܧ‬ሺ‫ܥܹܮ‬ሻ ridge position and height have a significant effect on the
‫ܬ‬଴ ൌ                              ሺ͵Ǥͳሻ maximum lift coefficient. If the ice ridge is not at the
͸Ͳߩ௜௖௘
6
leading edge point, the more front the ice ridge position and the University of Illinois (Fig. 11). The research by Kim [47]
the higher the ice ridge height are, the greater the maximum also pointed that ice shape height and location had a
lift coefficient loss is. If the ice ridge is at the leading edge significant effect on the measured reduction of ‫ܥ‬௟ǡ௠௔௫ and
point, the maximum lift coefficient no longer declines after other measures of aerodynamic performance.
the height reaches 0.005c [ 46 ]. Similar results were
obtained on the NACA 23012 airfoil in the wind tunnel at
Table 2 1956 US Air Force standards of aircraft icing severity [42]
Ice Collection Rates on Small Probes
Descriptive Liquid Water Content
Aircraft Performance Criteria Inches per 10
Terminology (g/m3) Miles per 1/2 inch
miles
Barely perceptible formations on unheated
Trace 0 to 0.125 0 to 0.09 56 or more
aircraft components
Evasive action unnecessary (No perceptible
Light 0.125 to 0.25 0.09 to 0.18 28 to 56
effects on performance)
Evasive action desirable (Noticeable effects
Moderate 0.25 to 0.60 0.18 to 0.36 14 to 28
on performance)
Eventual, evasive action necessary (Aircraft
Heavy is unable to cope with icing situation and 0.60 to 1.0 0.36 to 0.72 7 to 14
extended operation is not possible)
Immediate evasive action is required (Aircraft
uses climb power to hold altitude, and
Severe 1.0 or more 0.72 or more 0 to 7
continued operation is limited to a few
minutes)
Table 3 Icing severity level [43]
Icing severity level Light Moderate Heavy Severe
Icing intensity < 0.6 0.6-1.0 1.1-2.0 > 2.0
Table 4 1964 NCCAM Icing Standards [44]
Accretion Rate on a
Definition Effects on Aircraft Pilot Response
Small Probe
The presence of ice on the airframe is
perceptible, but the rate of accretion is nearly
The use of deicing equipment is
Trace 1/2-inch in 80 miles balanced by the rate of sublimation. Therefore,
unnecessary.
thus it is not a hazard unless for an extended
period of time.
The rate of accretion is sufficient to create a
hazard if the flight is prolonged in these Occasional use of deicing
Light 1/2-inch in 40 miles
conditions, but is insufficient to make equipment may be necessary.
diversionary action necessary.
On the airframe, the rate of accretion is Immediate diversion is necessary,
Moderate 1/2-inch in 20 miles excessive, making it hazardous even in short or use of deicing equipment is
encounters. mandatory.
Under these conditions, the deicing equipment Immediate exit from the icing
Heavy 1/2-inch in 10 miles
fails to handle the hazards. condition is mandatory.
Table 5 Icing severity level based on icing thickness [43]
Icing severity level Light Moderate Heavy Severe
Maximum thickness (mm) 0.1-5.0 5.1-15 15.1-30 > 30

Fig. 9 NACA 0012 airfoil and spanwise protuberance geometry [45]

7
 the lift and pitching moment curves (Fig. 12 and Fig. 13). It
is obvious that an increase in the ice ridge height
significantly increases the risk of wing stall in advance and
may lead to a sudden change in the pitching moment, which
adversely affects the aircraft stability. And the effect is
particularly obvious for the ice ridge positioning behind.

Fig. 10 Variation of maximum lift with spanwise protuberance height

on NACA 0012 [46]

(a) Lift

Fig. 11 Variation of maximum lift with simulated ridge ice ridge height
on NACA 23012m [46]

(b) Pitching moment

Fig. 13 Effect of simulated ridge ice of various heights at x/c = 0.10
location on the modified NACA 23012m [46]
The University of Illinois also studied the effect of
glaze ice geometry on the maximum lift coefficient (Fig. 14
to Fig. 17). When the ice ridge height remains unchanged,
‫ܥ‬௟ǡ௠௔௫ increases as the ice ridge shape becomes rounder
(namely, the leading edge radius r increases) and as the ice
ridge is closer to the front edge. Similarly, ‫ܥ‬௟ǡ௠௔௫ increases
(a) Lift as the ice ridge height decreases and r increases. If the front
edge of the ice ridge is sharp, the aerodynamic performance
damage will be greater.

(b) Pitching moment

Fig. 12 Effect of simulated ridge ice of various heights at the leading
edge location on the modified NACA 23012m [46]
Further, they obtained the effect of ice ridge height on Fig. 14 Glaze-ice horn radius effects on NLF 0414, k/c=0.0433 [46]

8
 height are key parameters, but the geometry of the ridge is
also proved to be important.
According to the results above, if the detailed
aerodynamic parameters of an airfoil under different
artificial ice ridge heights, positions and leading edge
radiuses are known, the aerodynamic performance
degradations under natural icing conditions could be
deduced as the evaluation standard of the aircraft icing
severity level.

Fig. 15 Glaze-ice horn radius effects on NLF 0414, s/c=0.017 [46]

(a) Simulation results

Fig. 16 Ridge ice shape tested on NACA 23012m [46]

(b) Training data

Fig. 18 The effects of ice horn locations on ࡯࢒ [48]

4. Icing parameters
Fig. 17 Effect of simulated ridge ice geometry on lift (NACA 23012m),
4.1 Environmental parameters
k/c=0.0139 simulation height [46]
Cao et al. [48] developed a methodology for predicting According to FAR-25 Appendix C [50], the aircraft
the effects of glaze ice geometry on airfoil aerodynamic icing severity level depends on three variables: the liquid
coefficients by using neural network (NN) prediction. Fig. 18 water content (LWC), the mean effective diameter (MED),
shows that the ice horn position plays a critical role in the and the ambient air temperature.
aerodynamic performance degradation due to icing. It is
clear that larger s/c causes a larger decrease in ‫ܥ‬௟ǡ௠௔௫ and 4.1.1 Water droplet diameter
lower stall angle as compared with the clean case.
The water droplet diameter is usually characterized in
Bragg’s review [49] indicates that spanwise-ridge ice
two ways: the cloud mean effective diameter (MED) and
usually forms farther back on the airfoil surface than horn
median volumetric diameter (MVD). The difference
ice and, while there are similarities to horn ice, has a
between MED and MVD is that the former calculation is
different flowfield. Spanwise-ridge ice is a flow obstacle,
based on the hypothetical water droplet distribution, while
since the airfoil boundary layer develops along the airfoil
the latter is based on the measurement of the actual droplet
surface before encountering the ridge. Ridge location and
size. In general, the two are considered equivalent. There
9
are also some literatures [51] suggesting that the radar generate and can easily be converted to precipitation loss.
estimated size (RES) is better suited to reflect the icing Those small droplets (less than 15 ȝm) do not contribute
threat because RES can better represent the amount of much to icing thus is not involved in the FAR.
supercooled water droplets in the cloud. However, flight records in recent years indicate that
As can be seen from the results [52] from the CANICE aircraft may also encounter SLD in the freezing drizzle
icing prediction software in Fig. 19, the droplet diameter can (MVD 50 to 500 ȝm) or freezing rain (MVD 500 ȝm and
affect the height of the ice angle and the thickness of the ice larger). The size of these droplets is greater than the droplets
layer. The larger the water droplets within a certain range, in the definition of the operating specification under any
the thicker the ice, the more obvious impact on the aircraft previous icing conditions and is prone to cause accidents
(Airfoil NACA 0012, angle of attack 0°, flow speed 129 m/s, because ice accretions by the large size droplets in such
ambient temperature -12.6 °C, LWC = 1.0 kg/m3, icing time freezing drizzles and freezing rains often occur behind the
120 s, airfoil chord length 0.3 m). The droplet diameter can anti-icing system protection area.
affect the distribution of icing on the airfoil surface (Fig. 20). Therefore, in recent years, the FAA has also added the
The larger the diameter of water droplets, the greater the conditions of SLD in the airworthiness regulations. The
inertia is. It is more difficult that water droplets are deviated associated parameters for the definition of icing conditions
from the original trajectory. are height, vertical extent, horizontal extent, temperature,
liquid water content and water mass distribution. Appendix
O specifies the icing condition with maximum water droplet
diameter greater than 100 ȝm as the SLD icing condition, in
which maximum water droplet diameter in the range of 100
ȝm to 500 ȝm is freezing drizzle icing condition, and
maximum water droplet diameter greater than 500 ȝm is
freezing rain icing condition. And then by treating an
average droplet diameter of 40ȝm as a boundary, the SLD
icing conditions are further divided into four categories, as
shown in Table 6. The MVD critical value chosen to be 40
ȝm is consistent on the one hand with the maximum MVD
limit for continuous icing conditions in FAR-25 Appendix C
and on the other hand with the SLD data in the existing
researches. The ‫ܦ‬௠௔௫ critical value of 500 ȝm is based on
Fig. 19 The relationship between water droplet size and ice shape [52] the consistent definition of freezing drizzle and freezing rain
in meteorology. The ‫ܦ‬௠௔௫ lower limit of 100 ȝm is
consistent with the maximum water droplet diameter
discussed in FAR-25 Appendix C. Therefore, FAR-25
Appendix C, together with Appendix O, covers almost all
supercooled water droplet icing conditions.
Table 6 Categories of icing conditions for SLD [53]
Definition ‫ܦ‬௠௔௫ (ȝm) MVD (ȝm)
100-500 <40
Freezing drizzle
environment
100-500 >40

>500 <40
Freezing rain
environment
>500 >40
Fig. 20 The relationship between water droplet size and local collection
efficiency [52] The traditional formation mechanism of supercooled
In general, the supercooled water droplet diameter is water droplets in the atmosphere can be divided into two
approximately in the range of 15 to 40 ȝm, thus FAR-25 kinds. One is the traditional mechanism of melting and
Appendix C establishes the airworthiness standard on this re-cooling, which means the high-altitude snow encounters
basis. In this range, with the diameter of water droplet the rise warm air group and then melts and falls until the
increases, the wing front edge is more likely to be iced, formation of supercooled water droplets cooled by the cold
causing greater aerodynamic damages. Exceptionally large air layer. The other one is related to water droplet
droplets beyond this range require special conditions to condensation, collision and fusion, known as the

10
non-traditional mechanism [52]. The results show that 88%
of the freezing drizzle icing conditions are formed by the
non-traditional mechanism while 92% of the freezing rain
icing conditions are formed by the traditional mechanism.
The difference between the two conditions lies not only in
the size of the water droplet diameter, but also in the
different formation mechanisms in the atmosphere. Fig. 21
shows the droplet mass distributions for the four kinds of
supercooled water droplets, which can be used to determine
the distribution of water droplet diameter in numerical
simulations and wind tunnel tests.

Fig. 22 Altitude-temperature envelopes of ZLE and ZRE [55]

It is worth noting that, for the classification of SLD,
Marwitz, etc. [56] had proposed another classification. They
specified the water droplets diameter less than 40 ȝm for
general cloud icing conditions, water droplets diameter
between 40 to 400 ȝm for freezing drizzle icing conditions
and water droplets diameter greater than 400 ȝm for
freezing rain icing conditions, as shown in Table 8.

4.1.2 Liquid water content

Fig. 21 Cumulative mass fraction distributions for SLD environments

LWC refers to the liquid water per unit volume. It only
that include (a) freezing drizzle drops > 100 ȝm, and with MVD < 40
counts the liquid state of supercooled water droplet, but
ȝm and MVD > 40 ȝm and (b) freezing rain drops > 500 ȝm, and with
does not contain other states (such as ice crystals) of water.
MVD < 40 ȝm and MVD > 40 ȝm [53]
Liquid water content is closely related to the types of cloud.
Table 7 shows the comparison of the pressure altitude Water content varies widely for different types of cloud.
range, the maximum vertical extent, the horizontal extent, LWC is one of the most important parameters affecting
and the ambient temperature range for the icing conditions aircraft icing. It reflects the amount of supercooled water
in FAR-25 Appendix C and Appendix O. It can be seen that droplets that an aircraft can hit in a given air mass. The
the altitude-temperature envelope of the continuous icing higher the LWC, the greater the risk of icing is. Since the
conditions is the largest one, followed by that of freezing amount of liquid water contained in the air mass (or cloud)
drizzle environment. Then that of freezing rain environment depends on the available energy and the temperature, the
is smallest. With ο, CMI, ZLE and ZRE representing the most serious risk of icing tends to occur at ambient
area of the altitude-temperature envelope, continuous temperatures above -15°C (stratiform cloud) or -20°C
maximum icing, freezing drizzle environment and freezing (cumuliform cloud). Cumuliform cloud is generally
rain environment respectively, Fig. 22 suggests that within unstable relative to the stratiform clouds. The internal
the ο௓ோா there is possibility that all the three icing airflow for cumuliform cloud is relatively intense, which
conditions exist; within the scope of (ο௓௅ா െ ο௓ோா ), the makes it more energetic inside, and also accommodates
icing conditions of ZLE and CMI possibly appear; within more supercooled water droplets, resulting in higher liquid
(ο஼ெூ െ ο௓௅ா ), only the CMI conditions will appear [54]. water contents. Similarly, the thickness of ice increases with
the increasing liquid water content, and the more severely
the flight safety is threatened, as shown in Fig. 23.
Table 7 Extend or ranges of different icing conditions [54]
Pressure altitude Maximum vertical Horizontal extent Ambient temperature
Types of icing conditions
range (ft, MSL) extent (ft) (nautical miles) range (°C)
Appendix O, Freezing Drizzle
0~22000 12000 17.4 0 ~ -25
(FZDZ)
Appendix O, Freezing Rain
0~12000 7000 17.4 0 ~ -13
(FZRA)

11
 Appendix C, Continuous
0~22000 6500 17.4 0 ~ -30
maximum icing (CMI)
Appendix C, Intermittent
4000~22000 -- 2.6 0 ~ -40
maximum icing (IMI)
Table 8 Definition of cloud, drizzle, and rain drops [56]
Drops Diameter (ȝm) Fall speeds
Cloud <40 <5cm/s
Drizzle 40 ~ 400 5-160cm/s
Rain >400 >1.6m/s

Fig. 23 Effect of LWC on ice shapes [52]

Fig. 24 Effect of temperature on ice shapes due to large droplets [52]
4.1.3 Ambient temperature
4.1.4 Cloud characteristics
The higher the temperature, the greater the heat
contained in the atmosphere, thus the atmosphere can Cloud type:
absorb more water vapor. As the temperature drops, the Among the parameters that affect icing, the external
water vapor from the air precipitation congeals into water temperature is often dependent on the height, while the
droplets [57]. Supercooled water droplets are generally cloud type determines the liquid water content, water
generated at 0 to -40°C. The temperature range in which droplet diameter and icing time, as shown in Table 9.
aircraft icing is most likely to occur is -2 to -10°C. Table 9 Cloud types, height and distribution of liquid water
Moderate icing occurs in the range of -2 to -12°C and LWC
Height Cloud type Distribution
serious icing occurs in the range of -8 to -10°C. Generally, water type
water below -40°C is in the form of ice crystals, which has Stratiform
High LWC Dispersion
almost no threat to normal flights. The exception is when Low clouds clouds
the body surface is warmed by an electrothermal deicing (below 6500 Nimbostratus High LWC Dispersion
system (or when the airspeed exceeds about 250 knots, ft)
resulting in dynamic warming near the leading edge), the Stratocumulus High LWC Concentration
ice particles will melt after impacting the body and freeze Solid
again on the back of the body surface due to a relative lower Medium Altostratus Dispersion
/liquid
temperature. It is noteworthy that when the temperature clouds
approaches 0°C, the fluidity of the droplets is enhanced, (6500-20000
Altocumulus Solid/liquid Concentration
resulting in a decrease in the ability of predicting aircraft ft)
icing. This is because even a very small disturbance can
change the local energy distribution, resulting in Cirrus Ice crystal
fluctuations of the icing probability fluctuations. This High clouds
requires a deliberate analysis of the reliability of icing (above Cirrostratus Ice crystal
prediction when the temperature is near the freezing point. 20000 feet)
Cirrocumulus Ice crystal
1) Stratiform cloud
In general, stratiform cloud has smaller droplet
12
diameter and lower liquid water content than cumuliform
cloud, but the horizontal distribution area is larger. It has the
following relationship with the occurrence of precipitation
(Table 10).
Table 10 Relationship between stratiform clouds and precipitation
Stratiform clouds type Precipitation type
Stratiform cloud Drizzle
Stratocumulus Drizzle or light rain
Nimbostratus Light rain or moderate rain
Altostratus Rain or snowfall
FAR-25 Appendix C defines the occurrence of icing
under these conditions as continuous maximum icing (CMI).
The relationship among MED, LWC, ambient temperature
and pressure altitude in this case is given in Appendix C, as
shown in Fig. 25 and Fig. 26.

Fig. 26 FAR-25 Appendix C CMI conditions: ambient temperature vs

pressure altitude [50]
Stratiform cloud icing usually occurs below 10000 ft,
located between low-level clouds and hollow clouds. At
high altitudes above 20000 ft, water is usually in the form of
Fig. 25 FAR-25 Appendix C CMI conditions: LWC vs MED [50]
ice crystals, which is difficult to cause severe icing. Fig. 27
shows the risk range for continuous maximum icing.

Fig. 27 Risk range for CMI [30]

2) Cumuliform cloud
Cumuliform cloud appears as a mass of unstable air
13
masses distributed in the vertical direction and has larger
liquid water content. The distribution of cumuliform cloud
is vertical, which also promotes precipitation, as shown in
Table 11. The water droplet diameter in the cumuliform cloud
is larger than that in the stratiform cloud. Therefore, the
FAR-25 defines icing condition occurring in the cumuliform
cloud as intermittent maximum icing (IMI) condition. The
relationship among MED, LWC, ambient temperature and
pressure altitude in this case is also given in Appendix C, as
shown in Fig. 28, Fig. 29 and Fig. 30.

Fig. 30 FAR-25 Appendix C IMI conditions variation LWC parameter

with cloud horizontal extent [50]
Table 11 Relationship between cumuliform cloud and precipitation

Cumuliform clouds type Precipitation type

Altocumulus Virga

Cumuliform cloud Showers or small hail

Fig. 28 AR-25 Appendix C IMI conditions LWC vs MED [50] Cumulonimbus Heavy rain, thunderstorms or hail

IMI generally occurs at altitude between 4000 and

22000 ft, but is also likely to occur at altitudes up to 30000
ft. The MED is typically between 15 and 50 ȝm and the
temperature usually ranges from 0 to -30 °C and sometimes
can approach -40 °C. The risk range for the occurrence of
intermittent maximum icing is given in Fig. 31.

Fig. 31 Risk range for IMI [30]

It is noteworthy that the data in the FAR standard were
derived from the NASA’s measurement reports many years
ago and the relevant data appeared in the 1960s’ technical
Fig. 29 FAR-25 Appendix C IMI conditions ambient temperature vs manuals. In view of the limitations of the measurement
pressure altitude [50] methods at that time, these data do not accurately reflect the
14
reality of the meteorological parameters. In fact, in recent Geographical parameters also affect the formation of
years some of the data and the standard do not fit very well. clouds and the liquid water content, thereby affecting the
Pressure altitude: icing risk [30]. For example, as an air mass climbs along the
According to the International Standard Atmosphere mountain, its temperature will drop. If the humidity at this
(ISA), below 36089 feet, the temperature decreases with time is right, it will form a cloud. If the air mass is stable, it
altitude, while above this height, the temperature is at a will form a stratiform cloud, otherwise, it will form a
constant value of -56.5°C. According to the FAA Aircraft cumuliform cloud. Or when the air mass flows through a
Icing Handbook [58 ], aircraft icing above 22000 ft is large water body, the lower air mass will absorb the
virtually non-existent because the atmospheric energy moisture and heat, thereby enhancing the overall liquid
thereby is not sufficient to contain enough liquid water, water content, which also makes the air mass less stable.
which corresponds to an ISA temperature of -28.6°C. Seasonal characteristics:
Therefore, in general, researches on aircraft icing should The warm season is prone to strong updrafts, and
focus on the altitude below 22000 ft. contains abundant water vapor. The resulting cumuliform
Horizontal distribution of cloud: cloud tends to have very high liquid water content, thus the
The NACA TN 2738 report [59] shows that the more icing risk is also significant, though its influence extent is
horizontally distributed is the stratiform or cumuliform small. The air mass in cold season is often cold and stable,
cloud, the lower the liquid water content is. The standard and its liquid water content is not high. However, due to the
stratiform and cumuliform cloud ranges are 17.4 and 2.6 low temperature and the large range of stratiform cloud, it
nautical miles. The time required for aircraft to travel at will also cause a serious icing risk.
different cruising speeds (true airspeeds) over the clouds is Weather process:
given in Table 12 for the standard cloud range and the As supercooled water droplets appear in the cloud
maximum cloud range, as icing severity is closely related to above the isotherm of 0°C, as long as the temperature is
the flight time in the clouds. In addition, the severity and appropriate, it is prone to form the aircraft icing
extent of aircraft icing are also positively correlated with the phenomenon. Such systems include fronts, upper-level
icing time, as shown in Fig. 32. On the other hand, the icing troughs, low vortices, low-pressure vortexes, shear lines, etc.
time is related to the flight speed and the horizontal [57]. Weak icing mostly appears inside the air mass,
distribution of the clouds. whereas moderate and strong icing has an opportunity of
appearing in the front area that is several times higher than
in the internal area. The main reason is that in the frontal
cloud system, the cloud and precipitation horizontal range is
relatively large, thus the flight time is longer. In addition,
the icing probability in a warm front is larger than that in a
cold front. Near the upper-level trough line and shear line,
the water vapor is generally more abundant. At this moment,
the supercooled water droplets diameters are relatively large
and icing is prone to occur. Low-pressure vortexes often
form a wide range of clouds, and low-pressure vortex
weather often maintains for a long time, even for several
days. The stronger are the low-pressure vortexes, the greater
the liquid water content and droplet diameter in the cloud
Fig. 32 Comparison of glaze and rime icing for cruise conditions [60]
and the more severe the aircraft icing will be.
4.1.5 Other environmental parameters
Geographical parameters:
Table 12 Comparison of continuous and intermittent maximum horizontal extents [30]
Time in CM clouds Time in IM clouds
True airspeed (Knots) LWC factor CM LWC factor IM
-17.4 nm (minutes) -2.6 nm (minutes)

100 10.4 1.00 1.6 1.00

200 5.2 1.00 0.8 1.00

250 4.2 1.00 0.6 1.00

15
 Time in CM clouds Time in IM clouds
True airspeed (Knots) LWC factor CM LWC factor IM
-310 nm (minutes) -5.21 nm (minutes)

100 186.0 0.20 3.1 0.85

200 93.0 0.20 1.6 0.85

250 74.4 0.20 1.3 0.85

region of the droplets moves to the lower surface of the

4.2 Aircraft-related parameters wing and the ice accretion is mostly found on the lower
4.2.1 Altitude and speed surface accordingly. Furthermore, in their latest research
[62], it was pointed out that though the angle of attack still
If the aircraft cruising altitude is beyond the icing risk affects the ice accretion region, the influence has been
range, then only the climb and descent phases are needed weakened due to the three-dimensional flow features. It is
where the icing risk is taken into account. However, the worth noting that severe ice accretion may occur under
majority of general airplanes can not climb to sufficiently sideslip conditions, which should be simulated in a direct
high altitudes, so the icing possibility exists in their entire manner due to its intrinsic nature.
flight envelopes [30].
The faster is the flight, the greater the mass flow rate of 4.2.3 Airfoil
the supercooled water droplets is. Although the It was reported by statistics that the NACA 230XX
aerodynamic heat increases with the increasing flight speed, series of airfoil make up the majority of the airfoils used on
the increase in the mass flow rate does not dominate the the crashed aircraft, of which the proportion of the NACA
icing process until the true airspeed reaches 430 knots. 23012 airfoil is close to 25%, as shown in Fig. 35. Some
Meanwhile, the impact of aerodynamic heat on the ice experiments have also confirmed that this airfoil can lose
shape can not be ignored. Due to the air compression and more lift than other tested airfoils under icing conditions, as
frictional heat, the temperature near the stagnation point of shown in Fig. 36. Gadebusch [63] remeshed the predicted ice
the wing leading edge is significantly higher than that of the shapes and presented 2-D NACA 23012 airfoil
other positions, leading to the supercooled water droplets computations for the clean and iced wing by using CFD
not freeze immediately but flow to other locations and form analysis as inputs to Lewice3D. The results show a severe
double-horn or trough ices. At true airspeed above 430 50% reduction of lift coefficient on NACA 23012 airfoil at
knots (sea level) or 475 knots (20000 ft), the flowing water AOA of 8ι. Yet, considering the holding of aircrafts using
droplets on the surface no longer freeze. Moreover, at true this airfoil is large, the incidence of accident should also be
airspeed above than 530 knots (sea level) or 580 knots higher, thus the proportion does not necessarily objectively
(20000 ft), de-icing devices can be completely eliminated in reflect the real icing disaster risk. However, this urges
accordance with the US military standard (MIL-A-9482), designers to study the airfoil icing aerodynamic
since the aerodynamic heat can completely melt the ice [30]. characteristics in the early stages of the aircraft design.
Therefore, to control icing risks, the airspeed must be kept More effective de-icing measures should be adopted for
appropriate. This appropriate airspeed is neither as low as those airfoils with poor icing tolerance.
stall speed nor too high to cause a high mass flow rate of
supercooled water droplets.

4.2.2 Angle of attack

Changes in the aircraft angle of attack can cause
changes in the flowfield, the local collection efficiency, and
the ice shape and thickness, etc., as shown in Fig. 33 and Fig.
34. Therefore, it is necessary to pay attention to the iced
aerodynamic performance at different angles of attack,
especially at relatively large angles of attack. Cao et al. [61]
developed an approach based on the Eulerian two-phase
flow theory to numerically simulate ice accretions on an
aircraft wing. The distribution of local collection efficiency
and ice shape at different angles of attack were evaluated. Fig. 33 Effect of angle of attack on ice shapes [52]
As the angle of attack is equal to 6ι, the main impingement
16
 horizontal tail icing tends to be more severe than wing icing.
Horizontal tail airfoils often have relatively small thickness
and sharp front edges, thus it is easier to be iced than the
main wings with blunt front edges. In addition, increase in
the airfoil relative thickness reduces the collection
efficiency, but at the same time causes increase in the total
mass of the impinging water. Therefore, the resultant effect
of the relative thickness change on the ice shape depends on
the above two aspects.

4.2.4 Other aircraft-related parameters

Skin temperature:
A prerequisite for aircraft icing is that the skin
Fig. 34 Effect of angle of attack on local collection efficiency [52]
temperature is lower than the freezing point. The skin
temperature also affects the icing shape. If the skin
temperature is too low, the water droplets will immediately
freeze to form rime ice. On the contrary, the water droplets
will still maintain liquidity for a period of time, resulting in
glaze ice.
Surface roughness:
Obviously, a relatively rough profile, both physically
and aerodynamically, provides very good conditions for
icing, which has the same principle as the sharp leading
edge [65].
Sweep angle:
Fig. 35 Percentage of accidents by airfoil type [64] Sweep angle changes the wing flow and further the ice
shape. Fig. 37 shows the ice shape on the NACA 0012 wing
in a sweep condition. It can be seen that the double-horn ice,
which is common on two-dimensional wings, turns into a
scallop ice.
Shielding effect:
Because the supercooled water droplet has a certain
inertia, it is not completely fit the airflow. If the upstream
body produces a flowfield that deflects the droplet trajectory,
the body downstream may not freeze because the water
droplets have not yet had time to hit the aircraft surface.
Similarly, the downwash caused by the lifting surfaces will
affect the collection efficiency of the rear of the aircraft.
Fig. 36 Effect of different airfoils on ࡯࢒ǡ࢓ࢇ࢞ [46] Flap:
As the radius of airfoil leading edge increases, the Opening the flaps allows the aircraft to fly more slowly
maximum local collection efficiency decreases. The and simultaneously the aircraft is more susceptible to icing,
maximum local collection efficiency ߚ is defined as the resulting in stall. If the flap leading edge is exposed to the
limit of the ratio of the distance ‫ݏ‬ଵ between the trajectories far flow, there is also a risk of ice accretion. It is more
of the two adjacent water droplets in the unperturbed serious if the downwash caused by flap causes changes in
airflow to the peripheral distance ‫ݏ‬ଶ between the two water the horizontal tail local angle of attack changes, which may
droplets and the airfoil at infinity, which characterizes the cause horizontal tail to stall.
local water droplet aggregation speed.
‫ݏ‬ଵ
ߚ ൌ Ž‹                                ሺͶ െ ͳሻ
௦మ ՜଴ ‫ݏ‬ଶ
The reason is that the bow shock wave of wing leading
edge generates the droplets track deflection, and only large
enough droplets have enough inertia to avoid the effects of
the bow shock wave. Conversely, if the leading edge is
sharper, its icing efficiency is higher. This explains why
17
 method to simulate flowfield and aerodynamic performance
of complex iced airfoil. The simulation results show a sharp
increase in drag coefficient when ice accretes at the leading
edge of commercial transport jet main airfoil (B757/767).
The drag coefficient increases about 575% at 10ι of AOA
(Fig. 38). Increased drag will require greater engine thrust
to maintain level flight, resulting in increased fuel
consumption and shortened the voyage, flight time and
activity radius.

Fig. 37 Fully developed scallops [66]

5. Icing effects on aircraft

Similar to the effects of rainfall on aircraft

aerodynamics [ 67 ], aircraft icing can change the
aerodynamic configuration of the aircraft, affect the airflow,
and reduce the aircraft aerodynamic performance, resulting
in degradation of controllability and stability and
Fig. 38 Drag coefficient of B757/767 airfoil with 145-m ice [68]
unpredictable disasters. A great deal of research has been
carried out on the icing effects on aircraft. However, 5.1.2 Lift and stall angle of attack reduction
different weather conditions and aircraft configurations can
lead to different icing effects, so there are many specific Likewise, icing damage to the wing profile results in
directions waiting for research. reduced lift and stall characteristics, which is particularly
pronounced in the case of glaze ice. When the air passes the
5.1 Aerodynamic performance ice angle on the leading edge, the speed increases rapidly,
the upper wing pressure drop at the leading edge is even
5.1.1 Drag increase
lower. Aft of the ice angle, the airflow separates from the
Ice can increase the surface roughness and further body and forms leading-edge bubbles [69]. The airfoil flow
increase the drag in flight. The degree depends on the field is changed with ice accretions compared with the clean
aircraft model and icing conditions. The drag of the wing case, resulting in deteriorated aerodynamic performances.
makes up only a small fraction of the total drag increase. A At the same time, there is a chance of air separation at a
typical rime ice may increase the total drag by as much as small angle of attack, as well as a reduced stall angle of
100%, while the glaze ice is likely to increase the drag by attack. Valarezo [70] pointed out that the typical minimum
200% to 300%. Even with anti-icing systems, the drag will initial leading-edge ice could lead to a loss in maximum-lift
increase by 20% to 90%. capability by 40%. For a multi-element airfoil, the research
Data in this regard can be referred to in the study of by Khodadoust [71] indicated that ice accretion on the main
Ashenden et al. [11]. Based on the Beech's King Air model, element and flap leading edges did not have a significant
they studied the flight data of 20 years in icing condition. impact on the maximum lift performance and
The results show that the drag is averagely increased by angle-of-attack-margin to stall. Only the parasite drag
about 50% in the icing environment and the maximum increased. This result further illustrated that the ice
value can reach 200%. Similar results can also be obtained accretion on the leading edge of airfoil is the most
from Leckman’s experiments [12]. In the absence of ice dangerous situation. Papadakis [72] studied aerodynamic
protection system, the increase of zero-lift drag coefficient characteristics of a symmetric NACA section with
ο‫ܥ‬஽଴ ൌ ͲǤͲͷͷ . The zero-lift drag coefficient ‫ܥ‬஽଴ is simulated ice shapes. The results showed that the upper and
increased by 275%. While, with the ice protection system, lower spoiler-ice shapes had a dramatic impact on
ο‫ܥ‬஽଴ ൌ ͲǤͲͳ͹ͻ and ‫ܥ‬஽଴ is increased by 90%. Cao et al. aerodynamic performance. This was due to the long-bubble
[ 68 ] developed a point-to-point hybrid grid generation flow separation observed over the upper and lower airfoil
18
surfaces throughout the angle of attack range. Gurbacki
[ 73 - 74 ] pointed out that the iced airfoil exhibited a
thin-airfoil stall in which the separation-bubble length
increased with the increasing angle of attack to form
complete airfoil separation. Mirzaei [ 75 ] studied
characteristics of separated bubbles and unsteady features of
the flow field around a glaze-iced airfoil (NLF-0414). It was
found that the primary separation bubble grew with the
increasing angle-of-attack. The shear layer formed between
the primary separation bubble and the high velocity inviscid
flow above the airfoil and consisted of two highly turbulent
regions, The first region was at the tip of the ice accretion
where the flow separated and the second was upstream of
the reattachment location.
Cebeci's [76] show that even a short period of icing can
seriously deteriorate the lift characteristics, including the
slope of lift curve, the maximum lift coefficient and the stall
angle of attack. Based on the results of wind tunnel testing,
Fig. 40 Lift loss after icing on NACA 65A413 airfoil [60]
a process of deterioration of a lift coefficient and declining
Ohio State University conducted further wind tunnel
of critical angle of attack is shown in Fig. 39. The
tests on lift changes caused by different ice shapes, as
configurations are the wing profiles after icing for 1, 5, 15,
shown in Table 13, Fig. 41 and Fig. 42. Failed Boot ice shape
30, 60 seconds. In the calculation, the forward flow
refers to the ice accretion over a period of time after the
temperature 260.44 K, the pressure 907548 Pa, the LWC 0.5
system failure. Both ice shapes are from the NASA IRT
g/m3, the wind speed 129 m/s and the droplet diameter 20
wind tunnel test results. S&C ice shape is recorded in the
ȝm. As can be seen from Fig. 40, as for the NACA 65A413
real-time camera during the DHC-6 Twin Otter flight
airfoil, the maximum lift coefficient is reduced by nearly 50%
experiment, and is often named Stability & Controllability
in icing, the stall angle of attack is reduced by 6° and the
because it is often used to study the aircraft stability and
slope of the lift curve is also reduced. According to
controllability characteristics. The LEWICE ice shape is
Leckman’s study [12], based on the flight test results, a
derived from the LEWICE software, with T0=-4°C,
quarter-inch thick ice is sufficient to increase the stall
MVD=20 ȝm, LWC=0.5 g/m3, angle of attack 0° and speed
critical speed from 75 mph to 102 mph with undeployed
120 knots. Among them, the S&C and LEWICE
flap, while from 65 mph to 83 mph with deployed flap.
ice-induced performance degradations are most obvious, the
Based on the two-phase flow theory, Cao et al. [ 77 ]
maximum lift coefficient is decreased by 50% and the stall
developed an Eulerian method to simulate rime ice
angle of attack decreased by 9.5°. Hossain et al. [79]
accretions on an airfoil and analyzed the change of the
developed an envelope protection system to improve the
pressure coefficient along the iced NACA0012 airfoil. It can
envelope protection capabilities of an aircraft in icing
be seen from the results that the pressure distribution is
conditions. The system developed is capable of estimating
changed by the ice accretion.
the stall limits in icing conditions and predicting if the
aircraft will stall in the future. Merret et al. [106] considered
that changes in unsteady hinge moments were especially
effective in predicting stall. Cao et al. [80] summarized the
icing effects on the helicopter performance. Ice accretion on
main rotor has very bad effects on helicopter performance.
It affects the total lift, power etc. In addition, the effects of
ice accretions on the inlets, cooling bay inlets, tail rotor etc.
are also disadvantageous.
Table 13 Lift and stall angle of attack losses of different ice shapes [14]
Ice shape ο‫ܥ‬௟ ௠௔௫ (%) οߙ௦௧௔௟௟ (°)
Inter-cycle 30 2.3
Falled Boot 41 7.3
S&C 50 9.5
Fig. 39 Lift force coefficient ࡯࢒ as function of angle of attack and form LWEICE 50 9.5
of icing [78]

19
 Therefore, they opened the de-icing system to run for 11
minutes, but this did not succeed in preventing the accident.
At the same time, two other aircrafts in the same airspace
reported to the tower the icing phenomenon, indicating that
the day was a severe icing environment. Due to the poor
visibility, the pilot opened the autopilot system, which
allowed the crew not to notice the speed loss. In the
approach landing, the pilot put down the flaps and landing
gear, which further enhanced the drag and caused a stall.
When the flaps were deployed at 10°, the speed had
Fig. 41 Different ice shapes information [14]
dropped to a dangerous 243 km/h. But the pilot did not
follow the standard stall treatment process. The aircraft then
diverged and lost control, plunging into a substantial pitch
and yaw oscillation until it crashed.
5.2 Stability and controllability
Both the vertical and horizontal static stability and
dynamic stability can be affected by aircraft icing. From the
aerodynamic point of view, icing changes the aircraft
flowfield, making changes in the aerodynamic derivative.
From the dynamic point of view, icing changes the original
aircraft mass distribution. These changes will cause changes
in the aircraft stability. Maneuvering efficiency can be
significantly affected if the aircraft control surfaces are
frozen. Flight tests [84] have shown that pilots feel an
increase in lever strength, which may reach hundreds of
Fig. 42 Ice shapes effects on lift coefficient [14] pounds, and difficulty in handling. Rudder surface icing
Results of the wind tunnel tests by Addy [81] showed may cause the rudder surface stuck and the aircraft out of
that the presence of the ice caused significant performance control. Flap leading edge icing can lead to the air
degradation. A two-minute glaze ice accretion could reduce separation in advance, resulting in flap efficiency seriously
the maximum lift coefficient by 22% of that of the clean reduced.
wing.
For helicopters, icing will also reduce the lift of rotor. 5.2.1 Longitudinal stability and controllability
Taking into account the centrifugal force, Zhao [ 82 ] degradation
proposed a new numerical method for predicting 3-D ice
accretion on a helicopter rotor in hover. The results show Cao et al. [85] studied CH-47B tandem rotor helicopter
that ice accretion on the blades reduces the aerodynamic trim and flight characteristics in a rime icing condition. As
performance of the rotor in hover. Larger ice may lead to shown in Fig. 43, the iced pitching attitude angle ߠ
the decrement of rotor lift and increment of required power. becomes smaller and smaller with the increase of the
forward flight velocity ܺሶ , which means more instability
5.1.3 Flight speed reduction with the speed of a tandem helicopter in icing conditions.
They [86] also presented a method to predict the effects of
Due to the increased drag and reduced engine thrust rotor icing on the flight characteristics of a UH-60A
caused by icing, aircraft need more thrust to maintain level helicopter. The effects of icing on rotor force, torque and
flight. However, greater thrust makes fuel consumption flapping were incorporated in a nonlinear helicopter
increase. In order to maintain a relatively economical fuel dynamic model. Fig. 44 shows the effects of icing on
consumption rate, the pilot had to reduce the cruising speed helicopter pitch-attitude ș response at a forward velocity of
and height, resulting in hidden dangers to the flight safety ܺሶ ൌ ͷ ݇݊‫ ݏݐ݋‬in five different icing conditions. It can be
[30]. Pilots may not have noticed the aircraft speed loss and seen that at 5 knots forward velocity, rotor icing reduces the
after a period of time the aircraft airspeed reduces below the peak value of pitch-attitude response, which means an
critical stall speed and leads to a crash. The DHC-8 crash in obvious controllability degradation. Furthermore, the
Buffalo 2009 is such a case [83]. The accident day was cold hazardous effects on trims, stability, and controllability of
and foggy, with poor visibility. Before the contact loss, the UH-60A single rotor helicopter in icing/ice-free conditions
crew had found obvious ice on the wing and windshield. and within/without different types of wind field were
20
investigated in Ref. [87]. Fig. 45 shows the effects of efficiency loss. Severe cases lead to a complete stall, violent
downdraft and ice accretion on helicopter pitch channel changes in the steering force, aircraft suddenly falling, and
controllability with 1.0 cm step longitudinal control input. It in a very short period of time the aircraft front end hitting
also can be found that, ice accretion always reduces the the ground and crashing. In general, the only solution is to
peak value of the pitch-attitude response. close the flaps and pull the aircraft up, but the success rate
is not very high.

Fig. 43 Trim data of pitching attitude angle [85] Fig. 46 Typical tailplane icing stall crashing process [14]
Although tailplane icing may appear on any type of
aircraft, it is much more likely to appear on light general
aviation aircrafts than other aircrafts, for the following
reasons: First, General aviation aircrafts often fly at low
altitudes, which means they are exposed in icing conditions
much longer than large aircrafts. Second, General aviation
aircraft icing protection system often has only a simple
de-icing device, thus the airfoil performance is vulnerable to
the old ice residue in the de-icing gap and the new ice
accretion together. For most aircrafts, the pilot is unable to
see the tailplane and the tailplane de-icing system work.
Therefore, it is very important to install a device to warn the
Fig. 44 Effects of icing on responses of ș [86]
pilot the possibility of a tailplane stall, especially during the
approach and landing phases.
The Swedish-Soviet Union team [88] conducted both
flight and wind tunnel tests on a high-lift configuration
aircraft in the 1970s and 1980s. The second and third
reports describe in detail the effects of tailplane icing on the
vertical stability and controllability of aircraft. The NASA
Lewis Center [89] conducted flight tests on the DHC-6
Twin Otter aircraft and the results show that the vertical
stability is obviously degraded when the tailplane is frozen.
When the flaps deploy at 10°, this change becomes more
intense. High thrust coefficient and low angle of attack also
cause stability loss under icing conditions. The NASA/FAA
Fig. 45 Effects of icing on responses of ș, within or without downdraft Tailplane Icing Program [14-15] is a joint effort between
airflow [87] FAA and NASA to explore the characteristics of tailplane
For fixed wing aircraft, tailplane icing is not a new icing and to seek measures to reduce the icing risk. The
issue and there are sporadic reports of related incidents project tested four different ice shapes: S&C, LEWICE,
since the beginning of the 1950s. As of 1999, tailplane icing Inter-cycle, Failed-Boot, and studied the aircraft responses
has resulted in 16 accidents and 139 deaths [7]. Tailplane under different maneuvering inputs. The mechanism of
icing accidents often occur in the approach landing phase. icing-induced loss of elevator control effectiveness was
As the tailplane airfoil is often thinner than the wing, it is studied in detail by Baars et al. [90]. The results showed that
more prone to accrete severe ice. Coupled with the landing the ice accumulated on the tailplane was 3 to 6 times thicker
flap downwash, the tailplane flowfield under icing than that on the wing and the drag increase caused by icing
conditions is prone to separation, resulting in rudder was significant.
21
 The DHC-6 [13] is a model often used in tailplane the other side does not. Such an accident is the DHC-6 crash
icing research with the following reasons. First, previous on February 4, 2009. On the other hand, there is a
studies have shown that the DHC-6 is very sensitive to possibility that the ice ridge on the wings deteriorates the
tailplane icing. Second, the DHC-6 has been extensively flowfield, resulting in the formation of a negative pressure
researched and retrofitted by NASA, thus there is a large area on one side of the aileron. If the negative pressure is
database on the airplane. Third, the DHC-6 has experienced small, then the pilot will only feel the aileron response is
severe icing crashes, such as on October 13, 1978 and slow. Pilots in no assisted control aircrafts will feel an
March 12, 1985 [91]. Examples of tailplane icing related unbalanced force on the aileron, known as aileron snatch. If
cases include the 1977 Vickers crash [92] and the 1985 the pilot can adjust this unbalanced force, the aileron
DHC-6 crash [91]. The former is due to the No. 2 and No. 3 maneuver is still essentially effective. If the negative
engine output was insufficient, and the temperature of the pressure zone continues to develop, then the aileron may
heating de-icing system did not meet the requirements, suddenly deflect without pilot input, causing instability of
resulting in ice accretions on the horizontal stabilizer front the aircraft rolling. Edward Air Force Base has used a
edge. The latter was also because the anti-icing system did tanker to sprinkle water in the actual flight experiments to
not work properly and caused the tailplane icing. Both simulate the SLD icing conditions [99]. In the experiments,
crashes were very similar, tailplane stalling, pitching out of the aircraft showed a significant change in the unilateral
control and then almost vertical falling to the ground. Some aileron moment coefficient, as shown in Fig. 47, which may
tests were presented with a simulated DeHavilland DHC-6 lead to the above uncontrollable rolling.
Twin Otter by using an icing encounter flight simulator [93], It is worth noting that ice ridges formed behind the
specifically, a tailplane stall event during a steep descent de-icing system in this experiment, thus even if the de-icing
and a roll upset event during an emergency approach. Miller equipment is working properly, occurrence of accident may
[94] developed models for aircraft dynamics in the presence still not be prevented. For the top surface, the ice ridge
of icing that can ultimately lead to a model based method of location is often between 7% and 9% chord length, far
detecting icing, with excellent agreement between model behind the icing protection system, as shown in Fig. 48. Such
predictions and flight test data from the NASA Lewis Twin ice ridges are often jagged and discontinuous in the
Otter Icing Research Aircraft. Thomas et al. [95] from the spanwise direction, causing severe damage to the
NASA Glenn Research Center pointed out that the aerodynamic configuration. To make matters worse, when
simulated wing ice shapes significantly reduced the ‫ܥ‬௟ǡ௠௔௫ , the flaps open at 15°, the ice ridge will be moved further
while the simulated tail ice caused elevator control force back to 14% chord.
anomalies and tailplane stall when flaps were deflected 30ι
or greater. Lampton [96-97] developed a methodology and
simulation tool for preliminary safety and performance
evaluations of airplane dynamic response and climb
performance in icing conditions. Comparing results with the
flight data showed good agreement for the long duration
and gradual maneuvers investigated, and only relatively
simple data were needed to construct the models.
Furthermore, a methodology and simulation tool [98] was
developed for a preliminary yet accurate evaluation of
airplane dynamical response and stability and control
characteristics due to icing. It uses only basic mass
properties, configuration, and propulsion data, together with
known icing data obtained for similar configurations.

5.2.2 Rolling and overturn

The cause of rolling instability is often associated with Fig. 47 Effect of ice on aileron control with flaps retracted full aileron
asymmetric air separation due to residual ice on the wing deflection [100]
surface, which can lead to flight accidents in two ways. On The ATR-72 accident happened on October 31, 1994 in
the one hand, air separation results in an imbalance in the such a way [101]. At the time of accident, the ATR-72 was
lift between the wings, resulting in an additional rolling in the waiting line and ready to drop to 8000 ft. During the
moment. If the separation is very severe, the aircraft will descent, the airplane made a turn to the right and
fall into continuous roll-up until the overturning occurs, as immediately entered an uncontrollable roll to the right with
one side of the wing loses a significant amount of lift but almost 70°. The crew tried to correct it, but the plane rolled

22
again until it crashed. The cause of the accident was that the flight accidents in various flight phases, as shown in Fig. 49.
unilateral aileron automatically deflected upward due to the As can be seen from the figure, in all the phases of interest,
ice ridge formed behind the de-icing system. Because the the cruise and approach phases account for the two largest
autopilot was activated, the pilots were unable to sense the proportions, respectively about 1/3 of the flight accidents.
roll moment in real time, which made the accident very
sudden and difficult to remedy.

Fig. 49 Percentage of occurrences by phase of flight [64]

Fig. 48 Typical ice ridge [102]

5.2.3 Yawing stability and controllability degradation 5.3.1 Take-off phase

It was pointed out that most aircraft icing accidents
Similar to the impact on tailplane, ice can also cause
took place during the take-off phase [104]. Aircraft in
yawing stability and controllability degradation. As
take-off phase is at lower speed and altitude, larger wing
mentioned earlier, lateral stability and controllability
angle of attack and more possible residual ice, thus is
performance changes generally manifest as the aircraft
vulnerable to the icing impact. This is why the FAA used
rolling response becomes slow and the rod force changes,
the pushover maneuver as a typical maneuver during the
which in severe cases will lead to rolling instability. While
tailplane icing program. The pushover maneuver is a good
the heading performance degradation will not affect flight
indication of overcorrection during take-off and climb
safety under normal circumstances, but in the case of
without the pilot being aware of ice. On the one hand, icing
unilateral engine failure may become a cause of disaster. At
reduces the aircraft performance, so that take-off glide
this time because the rudder needs to be used to overcome
distance and climb rates will be adversely affected. On the
the yaw moment caused by the asymmetric engine thrust,
other hand, residual ice and flap-induced downwash will
the deflection is much larger than in the normal flight. If
make the main wing and tailplane more easily stall.
there is no icing protection system on the vertical tail, the
CCAR-25 Appendix C defines the maximum take-off icing
aircraft may enter a situation similar to the longitudinal
conditions as LWC 0.35 g/m3, MED 20 ȝm, and ground
instability in the heading. However, because it will not
ambient temperature -9°C. The maximum take-off condition
directly cause the aircraft to hit the ground and crash, the
extends from the ground to 457 meters above the take-off
yawing stability and controllability degradation is not a
surface.
serious factor of disasters.
After take-off, the aircraft is more prone to ice
The results in Ref. [16] show that the static stability
derivative ‫ܥ‬௡ఉ of the DHC-6 can be decreased after being accretion during the initial climb phase. Because the low
speed tends to translate into a high angle of attack, which
frozen, especially in the case of zero thrust. The rudder
exposes the lower part of the aircraft and the wings to icing
efficiency derivative ‫ܥ‬௡ఋ௥ will be reduced by nearly 8% in
conditions. At the same time, the ice accretion may not be
the case of heavy glaze ice with tailplane. Cao et al. [103]
found by the pilot or can not be detected, so the pilot in the
used the NID control logic to study the flight envelope and
autopilot connected climbing process should be particularly
response of an iced aircraft. The results show that, for the
vigilant. Sibilski’s research [105] shows that a model of
iced aircraft, the flight envelope reduces apparently. The
icing leads to a very dangerous situation during a flight,
short period mode, the rolling mode and the Dutch-roll
especially when the aircraft passed to higher angles of
mode trend to be unstable. The characteristics of the rolling
attack, e.g., climbing flight. Deterioration of a plane’s flight
mode and Dutch-roll mode worsen rapidly.
properties is rapid and does not give any signs that are felt
5.3 In different flight phases by a pilot. Therefore, icing should be regarded as extremely
dangerous phenomenon. The Guide for Operation of
The literature [64] investigated the incidence of icing Aircraft Pilots at Low Temperature, published by the Civil
23
Aviation Administration of China (CAAC), does not
recommend the use of vertical speed (VS) mode to climb in
icing conditions. For aircraft without autothrottle in icing
conditions, the autopilot system will attempt to maintain
vertical speed regardless of speed, causing the aircraft to
suffer a potential stall risk. The pilot monitoring is critical to
ensure that the aircraft speed is not less than the minimum
speed under the current external environmental conditions
and configurations.

5.3.2 Cruise phase

For the modern airliner, the plane is often located
above the clouds in cruise phase, thus has lower icing
probability. But for general aviation aircraft, the flight often
locates below the clouds. As the height increases, the Fig. 51 Aircraft performance clean and iced (ࣁ=0 to 0.10) in a holding
temperature gradually decreases and it is also possible to turn and cruise. ࢻ vs time. V=230 ft/s and h=6560 ft. [21]
achieve icing conditions even in the subtropical range. At In addition, researches about aerodynamic performance
the same time, general aviation aircraft may take a long of large aircrafts under high-Reynolds number situations by
time to fly in the high LWC clouds, resulting in longer icing Hengst [107] and Bragg [108] indicated that the effect of
time. These can cause serious icing on the aircraft. underwing frost on aircraft take-off and climb performance
was small.

5.3.3 Approach or waiting phase

In this phase, the cause of accident is often the residual
ice in the cruise phase. Since the aircraft needs to turn
during the approach or waiting phase and the flaps need to
be deployed when approaching the airport, it is more likely
to induce stall than during the cruise phase. Another
noteworthy problem is that many aircrafts crashed during
this phase are with the automatic driving turned on, in
which case the pilot can not feel the changes in the aircraft
controllability until it is too late to remedy.

5.3.4 Landing phase

Fig. 50 Aircraft performance clean and iced (ࣁ=0 to 0.10) in a holding
Similar to the take-off phase, the aircraft is in the low
turn and cruise. Velocity vs time. V=230 ft/s and h=6560 ft. [21] altitude and low speed dangerous state during landing. The
Bragg et al. [21] analyzed the aircraft velocity-time effects of icing on the landing performance are with
response and the ߙ-time response in the cruise phase, as increasing landing speed, distance, time, and difficulty in
shown in Fig. 50 and Fig. 51. It can be seen from the figures, the tailplane trim. This is because the increase of the stall
after the aircraft cruise phase icing, the flight speed is speed increases the landing speed, so that the running time
reduced. In order to balance the gravity, the corresponding and distance are increased. If icing occurs on the tailplane, it
angle of attack will increase. Taking into account the may be difficult to trim and increase the lever force. On the
decreased stall angle of attack of the iced aircraft, the other hand, in order to reduce the landing speed, usually the
aircraft is prone to trigger a stall accident. The literature flaps need to be deployed, making the tailplane downwash
[106] pointed out that the aircraft aerodynamic performance more serious and resulting in additional nose-up pitching
degradation in the cruise phase is not easy to detect. The moment. Pilots in many crash accidents at the landing phase
corresponding aerodynamic deterioration is only noticeable suffer from spatial disorientation or similar symptoms,
when maneuvering like pulling up, is initiated. This is one because icing conditions often occur in severe weather
of the reasons for the frequent occurrence of cruise icing conditions such as fog and freezing rain. In this case, the
accidents. pilot will become rude and execute brutal manipulations,
which increased the crash probability.
Accident statistics show that most of the ice-related
accidents occur in the final phases of flight. The causes
24
include configuration change, low altitude, high load and
insufficient thrust setting. Loss of the aircraft control is also
a major cause. Icing contamination may result in wing stall,
icing contamination tailplane stall (ICTS), and roll
non-normal posture. Wing stall and roll capsizal can occur
in any flight phases. However, valid statistics show that
ICTS rarely occurs before approach and landing. Many
flight manuals have limitations on the maximum flap use in
icing conditions, as the tailplane is prone to stall on
approach flap settings. When there is ice on the wings and
the tailplanes, it is preferable to perform an undeployed flap
landing at a higher speed than in the normal approach.
Fig. 52 Variation in maximum lift with change in lift at fixed angle of
However, because of the higher approach speed, a longer
attack [106]
runway distance is required. Also, if ice is present, a greater
thrust should be used during landing leveling. Many
non-fatal icing accidents are attributed to stall at this phase.
It is also worth noting that though many articles
discuss have discussed ICTS, it does not mean ICTS is
more common than wing stall. In fact, data show that
accidents caused by wing stall due to icing contamination
are more common in icing conditions, but pilots often lack
training against ICTS to lower the danger.
5.4 Estimation methods of iced aerodynamic
parameters
The iced aerodynamic parameters are the key to Fig. 53 Variation of maximum lift coefficient with drag rise [106]
analyze the effects of icing on aircraft flight performance.
Cao et al. [ 109 ] elaborated on the acquirement of 5.4.2 Drag estimation
iced-aircraft aerodynamic parameters, including flight test,
According to icing wind tunnel experimental data, the
wind-tunnel experiment and numerical simulation. Some
increase of drag coefficient can be regarded as a piecewise
engineering algorithms are mainly introduced there. Based
function of icing time. When the icing time is less than 10
on the accumulation of a wealth of data obtained through a
minutes, the increase of drag coefficient can be expressed as
large number of wind tunnel and real fight tests for decades,
function of the freezing fraction n, collection efficiency E as
the icing effects have been simplified to a number of
well as other icing conditions. The increase of drag
practical engineering models. These models are simple to be
coefficient is generally described by the following formula
used, with clear physical meanings and acceptable accuracy
[19]:
and very suitable for predicting the overall performance
ο‫ܥ‬ௗ ൌ ‫ݖ‬ଵ ሺ‫ܣ‬஼ ‫ܧ‬ሻ݃ሺ݊ሻ                      ሺͷ െ ͳሻ
after icing.
‫ݖ‬ଵ  is a constant, ‫ܣ‬஼  is the cumulative coefficient, E is
5.4.1 Lift estimation the collection efficiency, n is the freezing fraction.
It can be seen that the increase of the drag coefficient is
According to the ice wind tunnel test results, at the directly proportional to the function of the freezing fraction
fixed angle of attack, the maximum lift coefficient and and the collection efficiency, as shown in Fig. 54.
change in lift coefficient is approximately linear, as shown The cumulative coefficient is expressed by the
in Fig. 52. This law can be used to predict stall boundary in following formula:
the design of icing envelope protection system. The ܸሺ‫ܥܹܮ‬ሻ‫ݐ‬
‫ܣ‬஼ ൌ                           ሺͷ െ ʹሻ
relationship between the maximum lift coefficient and drag ߩ௜௖௘ ܿ
rise after icing is similar, as shown in Fig. 53. It is available V is the flight speed, LWC is the liquid water content, t
to calculate the maximum lift coefficient if the change in the is the icing time, ߩ௜௖௘  is the icing density, c is the
lift or drag coefficient is known. Similarly, the two can also aerodynamic chord length of the airfoil. It is also noted in
be estimated by each other. some literatures that the chord length c is replaced with
twice the leading edge radius to obtain more accurate
results.

25
 icing serious factor, unrelated to the aircraft, only related to
the weather conditions, ݇஼ᇱ ಲ  is the aircraft icing factor
determined by factors such as the aircraft size, the flight
speed, the icing protection system working condition, and
the aircraft icing sensitivity, ‫ܥ‬ሺ஺ሻ௜௖௘ௗ  is the aerodynamic
derivative after icing.
ߟ௜௖௘  is a parameter based loosely on FAR-25 Appendix
C maximum continuous conditions on a 3 foot chord NACA
0012 airfoil (MVD=20 mm, ܸஶ =175 knots, LWC=0.65 g/m3,
t=10 min, T0=25 °F). ߟ௜௖௘  represents the ratio of the current
icing severity level to the maximum icing severity level, the
equation of which is defined as follows:
ߟ௜௖௘ ൌ
ο‫ܥ‬ௗ ሺܰ‫ʹͳͲͲ ܣܥܣ‬ǡ ܿ ൌ ͵ᇱ ǡ ܸ ൌ ͳ͹ͷ݇‫ݏݐ‬ǡ ܽܿ‫ݏ݊݋݅ݐ݅݀݊݋ܿ ݈ܽݑݐ‬ሻ
Fig. 54 Experimental and calculated values of the change in airfoil
drag coefficient due to ice accretion with AcE [19]
ο‫ܥ‬ௗೝ೐೑ ሺܰ‫ʹͳͲͲ ܣܥܣ‬ǡ ܿ ൌ ͵ᇱ ǡ ܿ‫ݐ݊݋‬ǡ ݉ܽ‫ݏ݊݋݅ݐ݅݀݊݋ܿ ݔ‬ሻ
According to the experimental data [110-111] in the ሺͷ െ ͷሻ
NASA icing wind tunnel, ݃ሺ݊ሻ should be a piecewise In order for Equation (5-4) to be applied on
function whose values are shown in Fig. 55. It can be seen non-NACA 0012 airfoils, for non-standard aircrafts, ݇஼ᇱ ಲ  is
that the drag increase is proportional to the freezing fraction expressed by the following formula:
when n<0.2, and reaches the extremum at n=0.2. When ߟ
݇஼ᇱ ಲ ൌ ݇                           ሺͷ െ ͸ሻ
n>0.2, the drag decreases with the increase of the freezing ߟ௜௖௘ ஼ಲ
fraction. The calculation methods of ߟ and ߟ௜௖௘ are basically
When the icing time is greater than 10 minutes, the the same, but the flight speed, chord length and airfoil are
drag increase can be expressed as: replaced with the specific aircraft. So ߟ can be used to
ο‫ܥ‬ௗ ൌ ‫ݖ‬ଶ ൫ͳ െ ݁ ௭యሺ௧ିଵ଴ሻ ൯ ൅ ο‫ܥ‬ௗ ሺ‫ ݐ‬ൌ ͳͲ݉݅݊ሻ  ሺͷ െ ͵ሻ express all kinds of aircraft icing information, also known
‫ݖ‬ଶ ǡ ‫ݖ‬ଷ  are constant calculated according to the previous as aircraft icing parameters. ݇஼ಲ  represents the variation of
formula at t=10min. the aerodynamic derivative ‫ܥ‬ሺ஺ሻ , expressing the sensitivity
of the derivative due to icing, which is constant for a given
aircraft. Then the aerodynamic derivative after icing can be
expressed as:
‫ܥ‬ሺ஺ሻ௜௖௘ௗ ൌ ൫ͳ ൅ ߟ݇஼ಲ ൯‫ܥ‬ሺ஺ሻ                 ሺͷ െ ͹ሻ
The relationships between ߟ and common icing
parameters were given by Pokhariyal [113], as shown in Fig.
56. It can be seen that, in a certain range, the larger MVD,
the larger ߟ and greater the icing risk. When the MVD is
less than 28 ȝm, the ambient temperature -3.9°C(25̧)
could cause the greatest icing harm. However, when the
MVD is larger than 28 ȝm, the ambient temperature
-6.7°C(20̧) could cause the greatest icing harm. Similarly,
Fig. 55 Experimental and calculated values of the change in airfoil
the icing hazard increases with increasing LWC and
drag coefficient due to ice accretion with freezing fraction, n [19]
increases most strongly at -6.7°C, as shown in Fig. 58. And
In addition, the drag generated by the natural wing as can be seen from Fig. 57, the damage is most severe when
roughness (rivets, deicing boots, etc.) should be accounted the temperature is about -5°C for different LWC.
for. Otherwise, the computed drag coefficient may be too
low [112].

5.4.3 Arbitrary aerodynamic parameters estimation

Many methods and models have been developed to
estimate aircraft aerodynamic parameters in icing conditions.
An extensively used model proposed by Bragg [19] is
written as
‫ܥ‬ሺ஺ሻ௜௖௘ௗ ൌ ൫ͳ ൅ ߟ௜௖௘ ݇஼ᇱ ಲ ൯‫ܥ‬ሺ஺ሻ                   ሺͷ െ Ͷሻ
‫ܥ‬ሺ஺ሻ  is an arbitrary aerodynamic derivative, ߟ is the Fig. 56 Effect of MVD on ࣁ [113]

26
 derivatives and the icing severity factor was determined by
linear interpolation between the iced and clean aircraft
models based on Twin Otter aircraft. For both moderate and
severe icing encounters, reliable indications of
nonnegligible levels of icing severity are available by using
the ‫ ܪ‬ஶ NPFSI ID algorithm. Aykan et al. [116] studied
icing identification of the A340 model based on neural
networks and KF. The KF was used to increase state
measurement accuracy in order to increase the training
performance. An artificial neural network (ANN) was used
Fig. 57 Effect of static temperature on ࣁ [113]
as the identification technique. The neural network structure
was embodied with the aircraft estimated measurements as
the inputs the icing parameters as the outputs. The necessary
training and validation set for the neural network model of
the iced aircraft were obtained from the simulations for
various icing conditions. Based on observations of the five
main ice-affected parameters, ‫ܥ‬஽ഀ ,  ‫ܥ‬௅ഀ , ‫ܥ‬௅೜ , ‫ܥ‬ெഀ and
‫ܥ‬௅೜ , it is possible to identify if the airplane was under icing
fault. Aykan et al. [ 117 ] also presented a similar
identification method for a F16 model. Caliskan and
Hajiyev [118] systematically introduced icing identification
systems, including neural network and Kalman filter based
Fig. 58 Effect of LWC on ࣁ [113] icing identifications. They pointed that the NNs used for
both detection and classi¿cation were all multi-layer
5.5 Parameter identification feed-forward networks and drew a typical structure and the
As mentioned above, aircraft icing can lead to changes working principle of neural network, as shown in Fig. 59 and
in the aerodynamic parameters and poor aerodynamic Fig. 60. It is worth mentioning that, almost all neural
performance. Therefore, it is especially important to detect networks need to be built for specific models and use
aircraft icing. Parameter identification (ID) is a common existing data for training. Dong and Ai [119] provided an
and useful method for detecting and estimating flight inflight estimate of the aircraft dynamic parameters and
parameters after icing. There are many parameter icing location detection based on the ‫ ܪ‬ஶ parameter
identification methods for aircraft icing, including batch identification algorithm and the Probabilistic Neural
least-squares algorithm, Kalman Filter (KF), Neural Network (PNN) method. Both the timeliness and accuracy
Network (NN), combined NN/KF, ‫ ܪ‬ஶ algorithm, etc. of the ID framework were examined. A database
Using the NASA Twin Otter inflight icing research corresponding to different icing locations, severity levels,
aircraft, Melody et al. [114] investigated the performance of and disturbances/measurement noise paths was generated
a batch least-squares ID algorithm, an extended Kalman for the training and test work of the PNN detection net. The
filter method, and an ‫ ܪ‬ஶ ID algorithm in the context of simulation tests show that the detection network performs
icing detection. The icing detection was modeled as a very well, with a false alarm rate of 0.20% and a danger rate
simple detection threshold for each parameter, with the of 0.07% only.
threshold being the mean of the iced and clean parameters.
The simulation results demonstrate that accurate and
unambiguous icing indications are available in 3s for the
‫ ܪ‬ஶ algorithm in the presence of measurement noise, based
on the given baseline scenario. However, the batch
least-squares and the extended Kalman filter methods show
poor performance. Afterwards, considering the time-varying
nature of ice accretion, Melody et al. [115] investigated the
performance of an ‫ ܪ‬ஶ Noise-Perturbed Full-State
Fig. 59 Aircraft icing identi¿cation system [118]
Information (NPFSI) ID algorithm in the context of icing
detection during steady, level flight with turbulence. The
tailplane icing severity was captured by a severity factor, ߟҧ .
The relation between aircraft stability and control

27
 For each local model ‫ܯ‬௜ , a controller ‫ܥ‬௜ is designed. In a
MM design, an on-line procedure that determines the global
control action through a weighted combination of the
different controls needs to be developed. The control action
weighting is usually based on a bank of Kalman filters
actions, each of which is designed for one of the local
models. The probability Ͳ ൑ ߤ௜ ൑ ͳ of each model to be in
effect is computed on the basis of the residuals of the
Kalman filters. The control action is then computed as the
weighted combination:
ே ே

‫ݑ‬ሺ݇ሻ ൌ ෍ ߤ௜ ሺ݇ሻ‫ݑ‬௜ ሺ݇ሻ ǡ ෍ ߤ௜ ൌ ͳ       ሺͷ െ ͺሻ

Fig. 60 Structure of a feed-forward neural network with two hidden ௜ୀଵ ௜ୀଵ
layers [118] where ‫ݑ‬௜ ሺ݇ሻ is the control action produced by a controller
designed for the ݅-th local model.
5.6 Control techniques of aircraft icing
The Control Allocation (CA) method is used to
Aircraft performance will inevitably decline and some produce a desired set of forces and moments from a set of
component or sensor failures may also occur in icing actuators, as shown in Fig. 62. The output “y” can be a set of
conditions. At this time, the aircraft's control laws and loops desired moments and the CA block is used to select
need to be changed to maintain safe flight. This control appropriate setpoints for the actuators producing those
technique is called reconfigurable control, or Fault Tolerant moments.
Control (FTC). Reconfigurable control can be generally
classified as active and passive. A comparison between
active and passive FTC systems has been given by Jiang
and Yu [120]. A passive approach is relatively simple to
implement, but only can be feasible for a small number of
faults. On the contrary, with effective fault detection
methods, an active approach is more Àexible to deal with
different types of faults, including failure scenarios beyond
the design basis faults. A classification of reconfigurable
control was systematically introduced by Verhaegen et al.
Fig. 62 Control Allocation scheme [121]
[121], as shown in Fig. 61.
An approach involving Model Reference Adaptive
Control (MRAC) scheme through adaptive feedback
linearization via artificial neural network was primarily
developed by Calise et al. [122-128]. The approach splits
the dynamics of the aircraft into three Single-Input
Single-Output (SISO) subsystems, each of which has a
model reference adaptive controller for roll, pitch and yaw
motions. The scheme of a nonlinear adaptive output
feedback controller is shown in Fig. 63. The principles of
the feedback linearization on SISO systems can be referred
to in [121].
The Sliding Mode Control (SMC) method is a
nonlinear type of control methodology and a special case of
variable structure control. The proposed controller [129]
was setup in a two-loop cascade configuration, with the
ultimate goal of tracking a trajectory given by roll, pitch and
yaw angle setpoints. The outer-loop was designed using
standard robust SMC techniques. The SMC controller can
Fig. 61 Classification of reconfigurable flight control approaches [121] handle all structural failures modifying the dynamics of the
The Multiple Model (MM) method, based on a ¿nite plant less than the assumed uncertainty. The online
set of linear models ‫ܯ‬௜ , belongs to the class of projection adaptation of the boundary layer can handle partial loss of
based methods rather than to the on-line re-design methods. actuator surfaces, while avoiding limits and integrator

28
windup by reducing the tracking performance. However,
there are some limitations of the SMC controller. First, there
must be one and only one control surface for every
controlled variable and second, none of the control surfaces
can ever be lost. Second, the method may well result in an
excessively conservative controller in the non-failure
situation due to robust control [121].

Fig. 64 Model Reference Adaptive Control [121]

The advantage of Model Predictive Control (MPC) is
that its ability to control multivariable systems and handle
constraints. However, this is held at the cost of a great deal
of computation. The MPC relies on an internal model of the
system, in other words, a reference fault model. The failures
including actuator and structural faults can be handled
naturally in a MPC framework via changes in the input
Fig. 63 Nonlinear Adaptive Output Feedback Controller [121]
constraints and internal model [131]. Actuator limit and rate
constraints can be written as:
The Eigenstructure Assignment (EA) method [130] to
‫ݑ‬௟ ൑ ‫ݑ‬௜ ሺ‫ݐ‬ሻ ൑ ‫ݑ‬௜௨
controller reconfiguration is an intuitive approach. The idea ቊ ௜௟                  ሺͷ െ ͳ͵ሻ
of the EA method is to place the eigenvalues of a linear ݀‫ݑ‬௜ ൑ ‫ݑ‬ሶప ሺ‫ݐ‬ሻ ൑ ݀‫ݑ‬௜௨
system using state feedback and then use any remaining for actuator inputs ‫ݑ‬௜ through ‫ݑ‬௠ . If actuator ݅ becomes
degrees of freedom to align the eigenvectors as accurately jammed at position ‫ݑ‬௜‫ כ‬, the constraints on input ݅ can be
as is possible. The eigenvalues determine the natural changed to:
‫ כݑ‬൑ ‫ݑ‬௜ ሺ‫ݐ‬ሻ ൑ ‫ݑ‬௜‫כ‬
frequency and damping of each mode while the ൜ ௜                     ሺͷ െ ͳͶሻ
eigenvectors control how much each mode contributes to a Ͳ ൑ ‫ݑ‬ሶప ሺ‫ݐ‬ሻ ൑ Ͳ
given output. Specifically, this method is exactly used to then the fault can be handled by the MPC controller
assign some of the most dominant eigenvalues while at the Due to the size and flight height of the Unmanned
same time minimizing the 2-norm of the difference between Aerial Vehicle (UAV), it is susceptible to icing. In recent
the corresponding eigenvectors years, a series of researches have been conducted on the
The model reference adaptive control (MRAC) method detection and control techniques of icing, especially for
can be effective for many types of structural failures. The UAVs. A two-unit system to deal with the fault caused by
plant output is forced to track a reference model. Consider aircraft icing was proposed by Tousi and Khorasani [132],
linear plants of the form: one for fault diagnosis and another for fault recovery.
‫ݔ‬ሶ ൌ ‫ ݔܣ‬൅ ‫ ݑܤ‬൅ ݀ Combined with the Eq. (5-4) presented in the previous
൜                         ሺͷ െ ͻሻ section, a robust sliding mode controller was designed. The
‫ ݕ‬ൌ ‫ݔܥ‬
where ‫ א ݔ‬Թ , ‫ א ݑ‬Թ௠ , ‫ א ݕ‬Թ௞ and a reference model of
௡ severity of icing can be predicted by a class of observations
the form: set in advance. Then this severity is used to modify the
‫ݕ‬ሶ ௗ ൌ ‫ܣ‬ௗ ‫ݕ‬ௗ ൅ ‫ܤ‬ௗ ‫                       ݎ‬ሺͷ െ ͳͲሻ controller to maintain safe flight. Given the aircraft model,
where ‫ݕ‬ௗ ‫ א‬Թ௞ and ‫ א ݎ‬Թ௞ . ‫ܣ‬ௗ (stable) and ‫ܤ‬ௗ are control law and hyperplane matrix, the state feedback
arbitrary square matrices. The state feedback of the form controller can be expressed. By incorporating the estimation
shown in Fig. 64 is considered: of the icing percentage on the aircraft structure and icing
‫ ݑ‬ൌ ‫ܥ‬଴ ‫ ݎ‬൅ ‫ܩ‬଴ ‫ ݔ‬൅ ‫                    ݒ‬ሺͷ െ ͳͳሻ severity from the diagnosis module, the controllers can be
where ‫ܥ‬଴ ‫ א‬Թ௞ൈ௞ , ‫ܩ‬଴ ‫ א‬Թ௞ൈ௡ and ‫ א ݒ‬Թ௞ are free modified in the recovery solution. Similarly, Rotondo et al.
controller parameters. The closed loop dynamics are then: [133] designed a Linear Parameter Varying (LPV) unknown
‫ݕ‬ሶ ൌ ሺ‫ ܣܥ‬൅ ‫ܩܤܥ‬଴ ሻ‫ ݔ‬൅ ‫ܥܤܥ‬଴ ‫ ݎ‬൅ ‫ ݒܤܥ‬൅ ‫  ݀ܥ‬ሺͷ െ ͳʹሻ input observer for the actuator fault and icing diagnosis of
The goal is now to make the closed loop dynamics UAV. The decision algorithm used information coming from
given by Eq. (5-12) match the desired dynamics of Eq. temporal and low-frequency residuals, considering the icing
(5-10). severity factor from the Eq. (5-4) at the same time. Seron et
al. [134] suggested that the Unknown Input Observer (UIO)
design exploited the change in equilibrium conditions,
29
caused by the icing effect, to identify a direction in the Alliance Icing Research Study (AIRS) [139] showed out
observer estimation error space that allows for aircraft icing that freezing drizzle, or supercooled large drops (SLD), was
detection and estimation. Hansen and Blake [135] and considered to be a dangerous icing condition and
Cristofaro et al. [136] separately designed a parameter approximately 80% of the SLD environments were assessed
adaptive estimator diagnosis scheme and a multiple model to have formed through a non-classical formation
adaptive estimation framework to detect UAV icing, using mechanism. In other words, most of the supercooled large
pitot tube, GPS and other sensor readings. In addition, droplets observed formed through condensation/coalescence
timely deicing is also necessary for the UAVs with ice without going through the ice phase. Due to the low
accretion. An electrically conductive carbon nanomaterial probability of encountering SLD weather conditions in
coating was designed by Sorensen, et al. [137] for an ice flight tests, numerical simulation method has played a more
protection system. This coating with low internal resistivity important role in related designs and demonstration works.
displayed electrically conductive characteristics suitable for In addition, the lack of experimental results is also an
a resistive heat source applied for thermal control of important factor to limit the research of SLD icing.
potential icing exposed UAV surface areas. Just a power
management system with a low-weight power source can be 6.1 Characteristics of SLD icing process
used to control the coating to deicing. SLD icing is significantly different from conventional
supercooled water droplet (diameter less than 40 ȝm) icing
5.7 Other icing effects
in many respects, such as water droplet spatial motion
Icing damages to the engine performance may exist in pattern, droplet-wall collision process, effect of water
the following forms: 1. Blocking the engine intake, resulting droplets on the wall boundary layer flow, etc. The main
in performance degradation. 2. Engine flameout. 3. reason is that some of the assumptions made for
Structural damage caused by engine inhalation of ice. conventional supercooled water droplets do not apply to
Engine flameout is a combination of many factors. Inlet SLD. Specifically, large water droplets in the movement
icing reduces the intake airflow and disturbs the flowfield. may have deformations, rupture and so on. After collision
The engine can not get enough oxygen to burn, resulting in with the wall, large water droplets may partially adhere to
rich oil. If the engine is in a low-speed state such as the wall and partially splash or bounce. Due to the limited
descending or waiting, a flameout may occur. cognitive level of the above process, the existing SLD icing
The sensors can be damaged by ice, resulting in simulation methods can only take simple empirical
reading errors and even complete failure. As a result, corrections based on the conventional water droplet models.
sensors for flight safety must be installed where they are Correction of deformation effect:
less prone to be iced and protected by icing protection The effect of large water droplets deformation is
systems. The most important sensors are the airspeed tube, considered by introducing the modified drag coefficient ‫ܥ‬஽ᇱ
gradienter and stall warning system. Considering the icing instead of the original drag coefficient ‫ܥ‬஽ [140]:
airspace is often with poor visibility, such as fog, any sensor ‫ܥ‬஽ᇱ ൌ ሺͳ െ ߦሻ‫ܥ‬ௗǡ௦௣௛ ൅ ߦ‫ܥ‬ௗǡௗ௜௦௞                ሺ͸ െ ͳሻ
failure could lead to aircraft crash. ‫ܥ‬ௗǡ௦௣௛ and ‫ܥ‬ௗǡௗ௜௦௞ are the drag coefficients of the
sphere and the disc, as follows:
6. SLD icing ʹͶ
‫ܥ‬ௗǡ௦௣௛ ൌ ͲǤ͵͸ ൅ ͷǤͶͺܴ݁ ି଴Ǥହ଻ଷ ൅         ሺ͸ െ ʹሻ
ܴ݁
͸Ͷ
As mentioned above, in recent years, the authorities ‫ܥ‬ௗǡௗ௜௦௞ ൌ ͳǤͳ ൅                       ሺ͸ െ ͵ሻ
have begun to pay attention to the aircraft icing problems in ߨܴ݁
ߦ is the eccentricity, expressed as function of the
SLD weather conditions and have developed corresponding
Weber number ܹ݁:
airworthiness regulations, such as the FAR-25 Appendix O ି଺
and the CS-25 Appendix O. In fact, since the 1994 ATR-72 ߦ ൌ ͳ െ ൫ͳ ൅ ͲǤͲͲ͹ξܹ݁൯             ሺ͸ െ Ͷሻ
crash in Indiana, USA, the SLD icing problem has attracted Correction of rupture effect:
wide attention. The rupture of large water droplets depends also on the
SLD icing generally occurs during aircraft encounters surface tension and aerodynamic forces. When droplet
freezing rain. The diameter of supercooled water droplets in Weber number is greater than the critical value, a large
the cloud is greater than 50 ȝm and exceeds the upper limit droplet will break into several small droplets. In the Euler
of the diameter defined in FAR-25 Appendix C. SLD icing method, the simulation of water droplet rupture is mainly
seriously affects the performance and controllability of through the introduction of additional control equations for
aircraft and threatens the flight safety. Therefore, a series of droplet diameter. The rupture process is accounted for in the
researches have been carried out on the characteristics of source term, that is, the source term gives the average
SLD icing. The results from a Canadian research [138] and change rate of droplet diameter over time in the rupture

30
process described by empirical formulas. The specific ‫ܭ‬ௐ
൒ ʹͲͲ                   ሺ͸ െ ͻሻ
formula is as follows [141]: ሺ•‹ ߠ௜ ሻଵǤଶହ
‫߲݀ ݀ܦ‬ ݀௦௧௔௕ െ ݀௢ where, ‫ܭ‬ௐ ൌ ξ‫ି כ ݂ܭ‬ଷΤ଼ , ݂ ‫  כ‬is the dimensionless droplet
ൌ ൅ ࢛ௗ ή ‫ †׏‬ൌ             ሺ͸ െ ͷሻ
‫ݐܦ‬ ߲‫ݐ‬ ܶ ଷ ௅ௐ஼ ଵΤଷ
݀௦௧௔௕ is the diameter at which the droplet finally impinging frequency,  ݂ ‫ כ‬ൌ ቀ ቁ , ߠ௜ is the
ଶ ௗ
reaches the steady state after rupture, and ݀௢ is the impinging angle, defined as the angle between the
diameter at which the water droplet begins to rupture. ݀௦௧௔௕ tangential direction of the wall surface and the direction of
can be given by the critical Weber number. Empirically, impinging.
critical Weber number ܹ݁௕ǡୡ୰୧୲ ൌ ͳʹ, as follows: The splashing mass ratio in an Eulerian framework,
ͳʹߪ defined as the ratio of the droplet mass leaving the surface
ܹ݁௕ǡୡ୰୧୲ ൌ ͳʹ ֜ ݀௦௧௔௕ ൌ    ሺ͸ െ ͸ሻ
ߩ௔ ܷஶ ௔ െ ࢛ௗ ȁଶ
ଶ ȁ࢛ after impinging to the incident mass, is given by:
௄ೈ
T is the time of dimensionless droplet rupture process, ‫ܥܹܮ‬௦ ଴Ǥ଴଴ଽଶ଴ଶ଺‫כ‬൬
ሺୱ୧୬ ఏ೔ ሻభǤమఱ
ିଶ଴଴൰
ൌ ͲǤ͹ሺͳ െ •‹ ߠ௜ ሻ ቈͳ െ ݁ ቉
determined by the following empirical formula: ‫ܥܹܮ‬௜
ܶൌ ሺ͸ െ ͳͲሻ
͸ǤͲͲͲሺܹ݁௕ െ ͳʹሻି଴Ǥଶହ  ͳǤʹ݁ ൅ ͳ ൑ ܹ݁௕ ൑ ͳǤͺ݁ ൅ ͳ The splashing tangential velocity ratio, defined as the
‫ۓ‬
ۖ ʹǤͶͷͲሺܹ݁௕ െ ͳʹሻା଴Ǥଶହ ͳǤͺ݁ ൅ ͳ ൑ ܹ݁௕ ൑ ͶǤͷ݁ ൅ ͳۗ ۖ ratio of the splashing droplet velocity component in the
ͳͶǤͳͲሺܹ݁௕ െ ͳʹሻି଴Ǥଶହ     ͶǤͷ݁ ൅ ͳ ൑ ܹ݁௕ ൑ ͵Ǥͷ݁ ൅ ʹ tangent plane of the collision point to the incident droplet
‫۔‬ ͵Ǥͷ݁ ൅ ʹ ൑ ܹ݁௕ ൑ ʹǤ͹݁ ൅ ͵ۘ
ۖ ͲǤ͹͸͸ሺܹ݁௕ െ ͳʹሻ
ା଴Ǥଶହ
ۖ tangential velocity component, is given by:
‫ە‬ ͷǤͷͲͲ ʹǤ͹݁ ൅ ͵ ൑ ܹ݁௕ ۙ ܸ௧ǡ௦
ൌ ͳǤͲ͹ͷ െ ͲǤͲͲʹͷߠ௜                ሺ͸ െ ͳͳሻ
ሺ͸ െ ͹ሻ ܸ௧ǡ௜
In addition, an experimental study [142] is presented The splashing normal velocity ratio, defined as the
on the deformation and breakup of water droplets in the ratio of the splashing droplet velocity component in the
vicinity of an incoming airfoil. It was found that, within the normal of the collision point to the incident water droplet
range of tested experimental conditions (the velocity of the normal velocity component, is given by:
incoming airfoils ranged between 50 m/s and 90 m/s, the ܸ௡ǡ௦
droplet diameter ranged between 364 ȝm to 1075 ȝm), the ൌ ͲǤ͵ െ ͲǤͲͲʹߠ௜                    ሺ͸ െ ͳʹሻ
ܸ௡ǡ௜
favored breakup mechanism was of the so-called “bag and The splashing water droplet diameter ratio, defined as
stamen” type. the ratio of the splashing water droplet diameter to the
Correction of effect on the wall: incident water droplet diameter, is given by:
As mentioned above, after collision with the wall, large ݀௦ ݀௦
water droplets may partially adhere to the wall and partially ൌ ͺǤ͹ʹ݁ ି଴Ǥ଴ଶ଼ଵ௄೎ ǡ     ͲǤͲͷ ൑ ൑ ͳ   ሺ͸ െ ͳ͵ሻ
݀௜ ݀௜
splash or rebound, i.e., the so called re-impingement effect,
as shown in Fig. 65. 6.2 Review of SLD icing effects on aircraft
As the diameter is larger than that of conventional
water droplets, SLD are with greater inertia and more
energy, thus the ice shapes and the icing severity of are
different from conventional water droplets. In addition, for
different airfoils, the SLD icing effects may also be
different.
Due to the deformation, rupture and rebounding of
SLD icing, the ice accretion on the upper surface behind the
de-icing system active area is likely to occur, may resulting
in ice protuberance along the spanwise direction [102].
Fig. 65 The physical phenomena occurring in large supercooled Johnson’s [100] wind tunnel measurements in 1940 showed
droplet impingement [143] that the maximum roll control power was reduced by 36%
Taking the LEWICE splashing/bouncing model as an due to ice accretion caused by full aileron deflection. In
example, the specific formula is given below [144-145]: 1947, Morris [146] reported the wind tunnel results of the
First, the Mundo number is introduced: simulated icing effect on the tailplane leading edge, which
ଵൗ
ଷ ଷ ହ ସ included an ice shape similar to that of SLD. In 1948,
ߩ௪ ݀ ܸ௡
‫ܭ‬ൌቆ ଶ ቇ                     ሺ͸ െ ͺሻ Thoren [147] conducted a 2-hour test flight in the freezing
ߪ ߤ௪
According to this, the judgment conditions of rain with a Lockheed P2V aircraft and observed the
splashing/bouncing can be given: backflow and icing phenomena behind the icing protection
area. In the Wyoming University study of the icing

31
aerodynamic effects, Cooper et al. [148] encountered twice and Whalen et al. [157-159]. Considering the equivalent
30 to 300 ȝm water droplets icing with the King Air aircraft thermal power from anti-ice systems, a method for ice shape
in the flight test process with exceptionally large flight prediction in the presence of anti-ice situations was
performance loss. In 1996, the effects of large-droplet ice proposed by Cao et al. [160]. The predicted results show
accretion on aircraft control were reviewed by Bragg [99]. that ice accretion may happen during a long time flight in
Ashenden and Marwitz [11] analyzed the icing conditions of icing weather conditions, even if the anti-ice system is
several King Air aircrafts to determine the effect of various activated, especially in the rear of the anti-ice region. If the
icing conditions on flight performance and found that MVD becomes larger, the ice ridge will be more destructive
freezing drizzle icing leading to the largest flight to the flowfield. Furthermore, for larger diameter water
performance loss in all icing conditions. Ashenden et al. droplets, how to accurately predict the trajectory of water
[ 149 ] found similar results in a low-Reynolds number droplets becomes more important. Oleskiw [161] showed
two-dimensional wind tunnel icing experiment, opening the that the response of two LWC estimation instruments
de-icing system in freezing drizzle conditions leads to more dropped off as MVDs increased beyond 100 ȝm. Sor [162]
severe aerodynamic performance losses. The SLD icing presented a theoretical model to predict water droplet
effect on the aircraft control surfaces was preliminary trajectories in the flow past an airfoil. The droplets’ initial
studied in tailplane stall report by Trunov and diameters are in the range from 550 ȝm to 1050 ȝm. It
Ingelman-Sundberg [150]. Due to the downwash caused by could be observed from the results that, in general, there is a
the wing flap deflection and the ice on the tailplane upper reasonably good agreement between the measured and
part, the maximum lift decreases and the stall angle of computed droplet trajectories.
attack increases, leading to tailplane stall. Bragg [151] For now, because the SLD icing weather conditions are
mentioned that, for NACA 0012 airfoil, the SLD icing too complex for wind tunnel simulation, thus study on the
conditions could result in ice ridges on the airfoil rear SLD icing effects on the flight performance mainly depends
surface, causing a large amount of separated air bubbles, on numerical simulation. The lack of experimental data is
leading to airfoil premature stall, great lift loss, drag the key issue in the study. In addition, the microscopic
increase and pitch moment change. The NASA/FAA/NCAR physical mechanism of SLD icing is also different from the
supercooled large droplet icing flight research program [152] conventional water droplet mechanism. Study on its real
during the winter of 1996-97 indicated that a considerable mechanism also needs to be developed.
investment of financial resources and time was required to
obtain quality SLD flight data. So numerical simulation is a 7. Analysis of typical aircraft icing accidents
very convenient method of SLD research. In 1999, Dunn
[153] simulated ice accretion under the SLD icing condition. The principles of selecting a typical accident case are
He studied the aerodynamic effect of SLD icing on NACA as follows:
23012 airfoil using NSU2D code to predict the pitch 1. Typicality. Aircraft structure icing causes serious
moment and hinge moment, which laid a foundation for damage to the normal flight in this case, and is the main
studying SLD icing effects on aileron or rudder control. Lee cause of the final crash. Engine icing, spatial orientation
[ 154 ] used the experimental method to study the disorder caused by icing environment such as fog or
aerodynamic effect of simulated SLD ice accretion on a freezing rain, and other reasons are not considered.
modified NACA 23012 airfoil. An integrated experimental 2. Disastrousness. The accidents leading to heavy
and computational investigation was conducted by Bragg et casualties are selected. The NTSB generally pays
al. [155] to determine the effect of simulated ridge ice insufficient attention to light small aircraft and no serious
shapes on airfoil aerodynamics, which may form aft of casualties accident. This kind of aircraft often lacks of
protected surfaces in SLD conditions. It was found that recording equipment, the investigation report of which is
large separation bubbles form downstream of the simulated usually relatively simple, not enough to fully reflect the
ridge ice accretions severely degraded the airfoil accident. Therefore, all the selected accidents resulted in
performance. This degradation was primarily a function of heavy casualties, thus caused the attention of relevant
ice shape size and location and nearly independent of departments, and were investigated deeply.
Reynolds number and ice shape geometry. For the 3. Data integrity. Accident reports with complete
forwardloaded NACA 23012m airfoil, the loss of meteorological parameters are selected.
performance included an 80% loss of ‫ܥ‬௟ǡ௠௔௫ for an upper 4. Supplementary explanation. Since both LWC and
surface ice shape location of ‫ ݔ‬Τܿ ൌ ͲǤͳʹ. In 2001, the MVD require specialized equipment for accurate
effect of runback and ‘‘ridge’’ ice accretions caused by SLD measurements, they are seldom mentioned in existing
was systematically summarized by Lynch [ 156 ]. accident investigation reports. Therefore, these two
Furthermore, a number of valuable researches about the parameters are only given in a few accident cases.
SLD and runback ice accretions were provided by Broeren
32
7.1 Statistics of aircraft icing accidents status of icing accident aircraft include LWC, MVD,
temperature, angle of attack, speed, altitude, icing time,
The following accident data are mainly from the flight status and crash type. The details are shown in Table 14
accident investigation reports issued by the National and Table 15.
Transportation Safety Board (NTSB), the Air Accidents
Investigation Branch (AAIB) and other air transport safety
departments. The main parameters to describe the crash
Table 14 Crash status of icing accident aircraft, part I
Aircraft model LWC (g/m3) MVD (ȝm) T (°C) ߙ (°) V (kts) H (ft)
SR22 — SLD -6 — — 3600
EMB-500 — — -1 — 88 Landing soon
SPORTSMAN
— — -1 — 65 6500
GS-2
Beech BE58 — Freezing rain -2 — — 6600
DASh-8
— — -7~1 11 131 2300
(DHC-8)
JS41 — — -4 — 170 6000
ATR 42ϋ320 — — -8 — 125 900
Cessna 500 — — — — — —
Saab 340B — — — — — —
Cessna 560 — — -3 — 96 6100
ATR-42 — Probable SLD -15~-5 — 160~155 12000
Crash after
Cessna 208B — Freezing rain -5 — —
take-off
EMB-120RT 0.5~0.8 10~30 -3 10 156 4000
ATR-72 0.45 70 -2 5 175~180 About 10000
Crash after
Fokker F28 — — Ground icing — —
take-off
Crash after
Fokker F28 — — Ground icing — —
take-off

British 0.3 (in Over 1000 (250 160, then

-2~-5 — 15600
Aerospace ATP simulation) in simulation) dropped to 142

Table 15 Crash status of icing accident aircraft, part II

Aircraft model Icing time (min) Flight status Crash type Data source
SR22 — Approach Under investigation NTSB ID: CEN15FA040 [163]

Approach to
EMB-500 at least 15 Roll after stall AAR1601 [164]
landing
SPORTSMAN
— Descent Loss of control until impact NTSB ID: CEN14FA032 [165]
GS-2
Loss of control due to
freezing rain and severe
Beech BE58 — Climbing NTSB ID: CEN13FA130 [166]
mixed icing condtions until
impact
DASh-8 (DHC-8) — Approach Stall due to low speed AAR1001 [167]
JS41, en-route, North West of
JS41 — Climbing Tailplane stall Aberdeen UK, 2008 (HF GND WX
LOC) [168]

33
 Approach to
ATR 42ϋ320 — Stall due to low speed AAR1102 [169]
landing
No crash
Uncommanded roll during Lessons from Icing Accidents and
Cessna 500 — —
landing Incidents [170]
Struck wing on runway
No crash
Lessons from Icing Accidents and
Saab 340B — — Lost 5000 feet altitude
Incidents [170]
Nearly inverted
Cessna 560 — Descent Roll after stall AAR0702 [171]
AT43, en-route, Folgefonna Norway,
ATR-42 — Climbing Rolling instability
2005 [172]
C208, vicinity Pelee Island Canada,
Cessna 208B — Take-off Stall and overload
2004 (WX HF GND LOC) [173]
EMB-120RT 45 Approach Wing stall AAR9804 [174]
ATR-72 17.4 Holding pattern Rolling instability AAR9601 [175]
Hitting a tree due to
Fokker F28 — Take-off Air Ontario Flight 1363 [176]
low climb rate
Fokker F28 — Take-off Stall AAR9302 [177]

British Aerospace 4/1992 British Aerospace ATP,

2 (in simulation) Climbing Stall
ATP G-BMYK, 11 August 1991 [178]

seen that the aircraft icing accidents occurred between -2

7.2 Analysis of parameters in aircraft icing and -6°C and mostly at -5°C, as shown in Fig. 68.
accidents LWC and MVD:
Based on these reports, some consistency and Unfortunately, because general aviation aircraft often
relevance can be concluded from the aircraft icing accident lacks measurement and recording equipment, it is very
conditions, though the conditions are complex and cover a difficult to fully reproduce the LWC and MVD when the
wide range. accident happens. As the accident investigation requires a
Ground temperature: lot of manpower and resources, in general, the investigation
The ground temperature of ten representative fatal report usually does not contain the above two data unless
aircraft icing accidents in recent years is shown in Fig. 66. It causing many casualties. Ground measurement equipment is
can be seen that most of the ground temperature values at also a solution, but its accuracy is too low to be acceptable.
the accident time is about 0°C. The points of higher ground This situation has not reached a breakthrough till now.
temperature on the right side are because the aircraft was
flying high and the accident occurred far away from the
ground. According to ATR ALL WEATHER OPERATIONS
BROCHURE, when the outside air temperature (OAT) is
less than 5°C and the humidity is appropriate, the aircraft
surface may be contaminated with ice during taxi or take-off.
At this time, because the aerodynamic force is not
significant, the ice accretion not only gathers on the wing
leading edge, but also contaminates the area beyond the
de-icing system. Similar results were presented in another
literature [64], as shown in Fig. 67.
Flight ambient temperature:
Eight cases are selected as the typical aircraft icing
accidents. The models of these cases numbered 1 to 8 are
ATR-72 [175], Cessna 560 [171], EMB-120RT [174],
British Aerospace ATP [178], ATR-42 [172], Cessna 208B Fig. 66 Ground temperature of different accidents
[173], JS41 [168], and Beech BE58 [166]. It can be clearly
34
 Due to the different aircraft models and flight
conditions, the flight speeds when accident occurs can not
be simply contrasted. However, some basic laws can still be
reflected by these data. First, all the accident aircraft flight
speeds were below 200 knots, a relatively low speed for
normal flight. In this speed range, the aerodynamic heat is
not sufficient to melt the ice. Second, due to the low flight
speed, which is often close to the stall speed in the flight
manual, the drag increase and wing stall angle of attack
decrease caused by icing make that any pilot manipulation
can easily lead to stall crash. It is often recommended in the
recent flight manuals that, under icing meteorological
Fig. 67 Percentage of occurrences by ground temperature range [64] conditions, a suitable safety margin should be maintained
with the minimum level flight speed specified in the
manual.
Icing time:
For a particular accident, icing time is the time it takes
to travel through the icing conditions. Different cloud sizes
lead directly to difference icing time. Airworthiness
regulations often use the time taken to cross the standard
clouds. The standard stratiform cloud and cumuliform cloud
ranges are 17.4 nautical miles and 2.6 nautical miles, at the
current altitude and cruising speed. The icing research is
either based on a uniform time determined by experiment
equipment and arrangements, or the time of the accident, as
shown in Table 15.
Fig. 68 Flight ambient temperature of different accidents Angle of attack:
Taking into account that only a fraction of the
Even fortunately leaving a record, different accidents is caused by wing stall, the wing local angle of
measurement methods can bring different results. In the attack does not have a special representative significance.
case of the ATR-72 accident, when the accident happened, However, in the EMB-120RT and DASh-8 (DHC-8) two
the KLOT WSR-88D Doppler radar observation data wing stall accidents, the wing local angle of attack had
showed LWC distribution between 0.01 and 0.7 g/m3. reached or exceeded 10°. The EMB-120RT used a type of
According to the data and ICE4A from NTSB, the LWC at airfoil that had a high proportion in icing accidents, i.e., the
the accident height was estimated to be 0.74 g/m3. However, NACA 230XX series airfoil, whose root airfoil is NACA
according to the aircraft icing forecast manual published by 23018 and tip airfoil is NACA 23012. The accident
the US Air Force, the estimated result was about 0.59 g/m3. percentage for the NACA 230XX series airfoil is shown in
A personally estimated result of one expert from the NCAR Fig. 35.
was 0.3 to 1.0 g/m3. The result given by the French accident Flight altitude:
statistics office was 0.3 to 0.7 g/m3. For the same accident, According to the NTSB accident database and other
the MVD forecast result was between 100 and 2000 ȝm by accident investigation reports, the vertical extent of ice
Doppler radar, belonging to freezing drizzle or freezing rain occurrence is very wide, ranging from sea level to over
defined by FAA. It is noteworthy that it is difficult to 10000 ft. The altitude of thirteen aircraft icing crash
predict when and where freezing rain or freezing drizzle accidents is shown in Fig. 69. It can be seen that all
forms by the existing technologies, and it is difficult to accidents occurred below 22000 ft, consistent with the FAA
detect even if it occurs. Also because of the large droplet Aircraft Icing Handbook [58]. Most of them occurred at
diameter, the icing area for the freezing rain on the wing is altitudes below 10000 ft, further proving the previous
much larger than that for the conventional droplets. conclusion on the flight altitude. For general aviation
Therefore, conventional leading-edge de-icing equipment aircraft, the icing possibility exists throughout its flight
tends to fail in this situation. It can be seen from the envelope because it can not climb sufficiently high to avoid
meteorological parameters in Table 14 that the SLD icing high water content clouds.
condition is a serious disaster condition in aircraft icing and
urgently needs to be studied.
Flight speed:
35
 7) A spatial orientation disorder or the similar symptom
due to low visibility.
Icing crash type:
As can be seen from Table 16, stall-related accidents
account for the majority, including the wing stall and
control surface stall. Due to the vision limit, the pilot
basically can not get aware of whether ice accretion is
present on the tailplane or not. Therefore, unless there are
special evidences to distinct the wing or control surface stall,
most reports do not distinct them specially.
In Accident 1, 4, 5, 6, 9, 12 and 15, the pilot played a
decisive role in the accident. The direct cause of accident 1
is the misuse of the autopilot. The direct cause of accident 4
and 12 is that the pilot did not pay attention to the airspeed
indicator, resulting in the airspeed below the safety standard.
Fig. 69 Accident altitude of different aircraft models
The direct cause of Accident 5, 6 and 9 is the violation of
Flight phase: the take-off regulation, of which the first two were caused
It has been shown in Table 15 that icing accidents can
by the residual ice on the wings before take-off and the last
occur at all flight phases. The main cause of accident at was due to carrying staff and goods beyond the allowable
take-off phase is often associated with the residual ice, such
weight range. The direct cause of accident 13 is the
as the Fokker F28 accident. The main cause of accident at approach without turning on the airplane’s wing and
climbing phase is usually that the pilot only focuses on
horizontal stabilizer deice system in the icing condition.
maintaining the vertical speed and neglects the horizontal NTSB pointed out that most of the common accidents were
speed reduction, resulting in too low aircraft speed.
due to the loss of control (LOC) in instrument
Accidents occurring at approach and landing phase are often meteorological conditions (IMC), e.g., Accident 14 and 15.
caused by the ice accretion in the cruise and descent phases.
Table 16 Examples of the icing-caused crashes
Most accidents at take-off and climbing phase can be
Accident
avoided, if the pilots carefully check the aircraft before Aircraft model Crash type
number
taking off, follow the flight manual to deice, pay close
ATR-72 1 Rolling instability
attention to the indicators at climbing and maintain a high
Cessna 560 2 Roll after stall
speed. Flight risks at other phases, such as cruise and
landing, are mainly dependent on the natural environment EMB-120RT 3 Wing stall
and the conditions of the aircraft itself, which are sometimes DASh-8
4 Stall due to low speed
beyond the pilot control range and result in irreparable (DHC-8)
accidents. Hitting a tree due to low climb
Fokker F28 5
Pilot manipulation: rate
It can be seen from a large number of accident Fokker F28 6 Stall
investigation reports that aircraft icing accident is often the British
7 Stall
result of a combination of many factors, in which the pilot Aerospace ATP
incorrect manipulation is the most fatal factor. In these ATR-42 8 Rolling instability
accident cases, the incorrect manipulations are summarized Cessna 208B 9 Stall and overload
as: JS41 10 Tailplane stall
1) No timely attention on the speed and climb rate SR22 11 Under investigation
changes. ATR 42ϋ320 12 Stall due to low speed
2) Still using autopilots in high-risk icing areas.
EMB-500 13 Roll after stall
3) Over or rough manipulations.
SPORTSMAN
4) No timely opening of the de-icing equipment, or 14 Loss of control until impact
GS-2
driving an aircraft without the icing airworthiness
Loss of control due to freezing
certification into the icing areas.
Beech BE58 15 rain and severe mixed icing
5) Opening flaps when landing or approach with ice
condtions until impact
accretions.
6) No timely pushing rod and closing flaps to remedy Cessna 500 16 No crash
when there is a sign of stall, such as rapidly decreasing Saab 340B 17 No crash
climb rate or reducing pitch manipulation efficiency.

36
8. Summary and conclusions diameter between 40 and 500 microns) often leads to
fatal accidents. Freezing drizzle droplet can be 10 times
The foregoing content is a systematic and as long as normal droplets in diameter, and about 1000
comprehensive review of the supercooled droplet icing times greater in volume and weight, even much greater
issues on aircraft. The primary intent is to define the range for freezing rain droplet. When the supercooled cloud
of possible consequences occurring in natural flight contains rain droplets, an impact layer will be formed,
conditions, especially in the most severe icing conditions resulting in an extremely fast and dangerous increase in
that are easily encountered. The full-scale flight test stall speed and drag, as well as lateral control
technique has not been used herein to either promote or help abnormality. It is more dangerous that greater inertia
correlate existing test results regarding the scarce testing and impact efficiency of the rain droplets will result in
results and the intrinsic limitations of this method such as ice accretion beyond the icing boundary of the normal
risk and inaccuracy acquisition of icing parameters due to droplet, sometimes beyond the protection area of the
the variability of natural environment. anti-icing or de-icing system. In this situation, as the
Generally, six parts have been made to identify current conventional anti-icing or de-icing methods fail, the
state-of-the-art research achievements of icing effects on aircraft will be under high risk of crash.
aircraft aerodynamics in this review. The first part analyzes At present, SLD icing is one of the hottest research
the causes of ice accretion on aircraft. The second part directions in the field of aircraft icing, especially on
introduces the types and severity of aircraft icing. The third SLD icing mechanism, aerodynamic and flight
part discusses the main parameters that dominate a specific performance changes caused by SLD icing and ice
icing condition. The fourth part synthesizes the current shape formed under the co-working of the icing
research achievements on the effects of icing on aircraft conditions of SLD and the anti-icing system. In doing
aerodynamics and flight mechanics. The fifth part briefly these, wind-tunnel experiment and numerical
introduces the characteristics and effects of the recently simulation are expected to cooperate to simulate the
particularly focused SLD icing issues. The last part icing conditions of supercooled large droplet.
summarizes and analyzes the past aircraft icing accidents Lastly, after reviewing the extensive issues of aircraft
and their specific causes. The following summarizes some icing, it is concluded that important lessons usually learned
important conclusions by the aforementioned three by one generation with aspect to the various serious
categories. consequences of aircraft icing have not been well
1) Ice accretion on aircraft is classified into rime ice, glaze disseminated or accepted by the later generations, resulting
ice and mixed ice based on its property. Because of its in many icing-induced aviation accidents which seemingly
streamlined or spear-like shape, rime ice has limited could have been avoided if enough attention had been paid
effect on the flowfield of aircraft, so the harm is less to. The review presented here is intended to attract people’s
than the latter two. Most glaze ice has double or eyes to the detrimental effects of icing on aircraft flight
multiple horns, and it has the greatest impact on the safety since it is inevitable for an airplane to encounter icing
flowfield and aerodynamic performance. Mixed ice is in flight, especially a long-duration at the stage of cruise.
one of the most common icing forms in reality. It has a Some measures are necessary to deal with problems of
mediate impact on aircraft aerodynamic performance. aircraft flight safety in icing conditions, especially these
2) When the aircraft is cruising or approaching, accidents SLD icing conditions. In addition, continuing education is
are most likely to occur under the state of icing. necessarily conducted to broadly utilize the profound
Therefore, in the research of flight states, these two lessons in the past and spread the knowledge of icing on
should be attached greater importance to. Meanwhile, aircraft aerodynamics and flight mechanics in the future.
the angle of flap and the control inputs of pilot also
have significant impacts on flight safety under state of Fundings
icing.
3) Conditions that possibly lead to severe icing are: Financial support was provided by the National
specific temperature (about -4°C) and altitude, Natural Science Foundation of China (grant number
supercooled water droplets with larger diameter (20 to 11702014).
25 Ɋ for continuous icing conditions), high liquid
water content (determined by historical data along with References
the droplet diameter), a period of flight under icing 1. Chang L. (2010). Aircraft Icing and Aviation Safety. Aer-
condition (at least 15 minutes). onautical Science and Technology, (5), 12-14.
4) Icing of SLD caused by freezing rain (droplet diameter 2. Board, T. S. (1986). Determining the effects of weather in
larger than 500 microns) and freezing drizzle (droplet aircraft accident investigations.
3. Nicholson, J. R., & Jafferis, W. (1988). Atmospheric sci-
37
 ences program at nasa Kennedy space center. D. (2007). Aircraft Performance Sensitivity to Icing
4. Mandel, E. (1989, January). Severe weather-Impact on Cloud Conditions. In 45th AIAA Aerospace Sciences
aviation and FAA programs in response. In 27th Aero- Meeting and Exhibit (pp. 2007-0086).
space Sciences Meeting (p. 794). 21. Bragg, M. B., Basar, T., Perkins, W. R., Selig, M. S.,
5. Forbes, G., Hosler, C., Klemp, J., Krider, E., & Mcginley, Voulgaris, P. G., Melody, J. W., & Sarter, N. B. (2002).
J. (1989). Weather support for the space program. Aero- Smart icing systems for aircraft icing safety. AIAA Paper,
space Sciences Meeting. 813.
6. Ferguson, D., & Radke, J. (1993, August). System for 22. COLE, J., & SAND, W. (1991). Statistical study of air-
adverse weather landing. In Aircraft Design, Systems, and craft icing accidents. In AIAA, Aerospace Sciences
Operations Meeting (p. 3980). Meeting, 29 th, Reno, NV (p. 1991).
7. Elshamy, M. (1995). Effect of atmospheric processes on 23. Gent, R. W., Dart, N. P., & Cansdale, J. T. (2000). Aircraft
launch decisions. Journal of Spacecraft and Rockets, icing. Philosophical Transactions of the Royal Society of
32(5), 801-805. London A: Mathematical, Physical and Engineering Sci-
8. Bragg, M. B., Perkins, W. R., Sarter, N. B., Basar, T., ences, 358(1776), 2873-29
Voulgaris, P. G., Gurbacki, H. M., ... & McCray, S. A. 24. Bragg, M. B. (1982). Rime ice accretion and its effect on
(1998). An interdisciplinary approach to inflight aircraft airfoil performance. NTIS, SPRINGFIELD, VA. 1982.
icing safety. AIAA Paper, (98-0095), 12-15. 25. Addy, H. E., Potapczuk, M. G., & Sheldon, D. W. (1997).
9. Preston, G. M., & Blackman, C. C. (1948). Effects of ice Modern airfoil ice accretions. National Aeronautics and
formations on airplane performance in level cruising Space Administration.
flight. 26. Shin, J., & Bond, T. H. (1992). Results of an icing test on
10. Bragg, M., Basar, T., Perkins, W., Loth, E., Sarter, N., a NACA 0012 airfoil in the NASA Lewis Icing Research
Selig, M., Sivier, K., Voulgaris, P., and Wickens, C., Tunnel. National Aeronautics and Space Administration.
"Smart Icing Systems Year 1 Interim Report," Smart Icing 27. Shin, J., & Bind, T. H. Experimental and computational
Systems Project, University of Illinois at Urba- ice shapes and resulting drag increase for a NACA 0012
na-Champaign, 1998. airfoil. 1992. NASA technical memorandum, 105743.
11. Ashenden, R., & Marwitz, J. D. (1997). Turboprop air- 28. Wilder R W. A theoretical and experimental means to
craft performance response to various environmental predict ice accretion shapes for evaluating aircraft han-
conditions. Journal of aircraft, 34(3), 278-287. dling and performance characteristics [J]. AGARD Air-
12. Leckman, P. R. (1971). Qualification of light aircraft for craft Icing 20 p(SEE N 79-15036 06-05), 1978.
flight in icing conditions (No. 710394). SAE Technical 29. Czernkovich, N. (2004). Understanding in-flight icing. In
Paper. Transport Canada Aviation Safety Seminar (pp. 1-21).
13. Ratvasky, T. P., & Ranaudo, R. J. (1993). Icing Effects on 30. Vukits, T. (2002, January). Overview and risk assessment
Aircraft Stability and Control Determined from Flight of icing for transport category aircraft and components. In
Data. Preliminary Results, NASA TM 105977. 40th AIAA Aerospace Sciences Meeting & Exhibit (p.
AIAA-93-0398, January. 811).
14. Ratvasky, T. P., VanZante, J. F., & Riley, J. T. (1999). 31. Heinrich, Ross, Zumwalt, Provorse,Padmanabhan,
NASA/FAA tailplane icing program overview. Thompsom, and Riley, Aircraft Icing Handbook, (Vol-
15. https://www1.grc.nasa.gov/facilities/hangar/ umes 1-3), FAA Technical Report DOT/FAA/CT-88/8-1,
16. Ranaudo, R. J., Mikkelsen, K. L., McKnight, R. C., Ide, R. Sept 1993.
F., Reehorst, A. L., Jordan, J. L., ... & Platz, S. J. (1986). 32. Wright, W. B. (2002). User Manual for the NASA Glenn
The measurement of aircraft performance and stability Ice Accretion Code LEWICE. Version 2.2. 2.
and control after flight through natural icing conditions. 33. Mikkelsen, K. L., McKnight, R. C., Ranaudo, R. J., &
17. Addy Jr, H. E. (2000). Ice accretions and icing effects for Perkins, P. J. (1985). Icing flight research: aerodynamic
modern airfoils (No. NASA-E-12228). NATIONAL effects of ice and ice shape documentation with stereo
AERONAUTICS AND SPACE ADMINISTRATION photography. AIAA paper, 85-0468.
CLEVELAND OH GLENN RESEARCH CENTER. 34. Kind, R. J., Potapczuk, M. G., Feo, A., Golia, C., & Shah,
18. Papadakis, M., Yeong, H. W., Wong, S. C., Vargas, M., & A. D. (1998). Experimental and computational simulation
Potapczuk, M. (2003, January). Aerodynamic perfor- of in-flight icing phenomena. Progress in Aerospace Sci-
mance of a swept wing with ice accretions. In 41 st AIAA ences, 34(5), 257-345.
Aerospace Sciences Meeting & Exhibit, Reno, NV. 35. Cao, Y., & Huang, J. (2014). New method for direct nu-
19. Bragg, M. B., Hutchison, T., Merret, J., Oltman, R., & merical simulation of three-dimensional ice accretion.
Pokhariyal, D. (2000). Effect of ice accretion on aircraft Journal of Aircraft, 52(2), 650-659.
flight dynamics. AIAA paper, 360, 2000. 36. 36 Cao, Y., & Hou, S. (2016). Extension to the Myers
20. Campbell, S. E., Broeren, A. B., Bragg, M. B., & Miller, Model for Three-dimensional Glaze Icing Calculation on
38
 Aircraft Surface. AIAA Journal of Aircraft, Vol. 53, No. 1, aircraft icing environments with supercooled large drops
January–February, pp.106-116 for application to commercial aircraft certification. Jour-
37. Cao, Y., Huang, J., & Yin, J. (2016). Numerical simula- nal of Applied Meteorology and Climatology, 51(2),
tion of three-dimensional ice accretion on an aircraft wing. 265-284.
International Journal of Heat and Mass Transfer, 92, 54. Hu, T., Li, Y., Lv, H., & Tian, B. (2014). Study on Air-
34-54. worthiness Problems of Operating in Supercooled Large
38. Schweikhard, W. G., & Kohlman, D. L. (1982). Flight Drops Icing Conditions for Transport Category Airplanes.
Test Principles and Practices. University of Kansas. Procedia Engineering, 80, 467-478.
39. Politovich, M. K. (2000). Predicting glaze or rime ice 55. FAA. Data and Analysis for the Development of an En-
growth on airfoils. Journal of aircraft, 37(1), 117-121. gineering Standard for Supercooled Large Drop Condi-
40. https://www.faasafety.gov/files/gslac/FTB/InflightIcing/m tions, DOT/FAA/AR-09/10, March, 2009.
ixed.jpg 56. Marwitz, J., Politovich, M., Bernstein, B., Ralph, F.,
41. Aeronautical Information Manual (AIM), updated annu- Neiman, P., Ashenden, R., & Bresch, J. (1997). Meteoro-
ally; Federal Aviation Administration, Washington, DC logical conditions associated with the ATR72 aircraft ac-
20590. cident near Roselawn, Indiana, on 31 October 1994. Bul-
42. Jeck, R. K. (2001). A history and interpretation of aircraft letin of the American Meteorological Society, 78(1),
icing intensity definitions and FAA rules for operating in 41-52.
icing conditions. FEDERAL AVIATION 57. Yuan K. (2008). Aircraft flight characteristics in condi-
ADMINISTRATION ATLANTIC CITY NJ AIRPORT tions of icing and windshear. Beihang University.
AND AIRCRAFT SAFETY RESEARCH AND 58. Heinrich, A., Ross, R., Zumwalt, G., Provorse, J., &
DEVELOPMENT. Padmanabhan, V. (1991). Aircraft Icing Handbook. Vol-
43. Zhang C. (2000). Flight Meteorology. China Meteorolog- ume 2. GATES LEARJET CORP WICHITA KS.
ical Press. 59. Lewis, W., & Bergrun, N. R. (1952). A probability analy-
44. L.V. Mitchell. Aircraft Icing-A New Look. Aerospace sis of the meteorological factors conducive to aircraft ic-
Safety, Dec. 1964, pp. 9-11, Published by the U.S. Air ing in the United States (No. NACA-TN-2738).
Force. NATIONAL AERONAUTICS AND SPACE
45. Jacobs, E. N. (1934). Airfoil section characteristics as af- ADMINISTRATION WASHINGTON DC.
fected by protuberances. 60. Bragg, M. B., Gregorek, G. M., & Lee, J. D. (1986). Air-
46. Lee, S., Kim, H. S., & Bragg, M. B. (2000). Investigation foil aerodynamics in icing conditions. Journal of Aircraft,
of factors that influence iced-airfoil aerodynamics. In 23(1), 76-81.
AIAA, Aerospace Sciences Meeting and Exhibit, 38 th, 61. Cao, Y., Ma, C., Zhang, Q., & Sheridan, J. (2012). Nu-
Reno, NV. merical simulation of ice accretions on an aircraft wing.
47. Kim, H. S., & Bragg, M. B. (1999). Effects of lead- Aerospace Science and Technology, 23(1), 296-304.
ing-edge ice accretion geometry on airfoil performance. 62. Cao, Y., Huang, J., Xu, Z., & Yin, J. (2016). Insight into
AIAA paper, 99-3150. rime ice accretion on an aircraft wing and corresponding
48. Cao, Y., Yuan, K., & Li, G. (2011). Effects of ice geome- effects on aerodynamic performance. The Aeronautical
try on airfoil performance using neural networks predic- Journal, 120(1229), 1101-1122.
tion. Aircraft Engineering and Aerospace Technology, 63. Gadebusch, J. A. (2012). Computational 3-D Icing Re-
83(5), 266-274. sults for Scaled DLR-F6 Geometry. AIAA Journal, 2,
49. Bragg, M. B., Broeren, A. P., & Blumenthal, L. A. (2005). 911-927.
Iced-airfoil aerodynamics. Progress in Aerospace Scienc- 64. Zeppetelli, D., & Habashi, W. G. (2012). In-Flight Icing
es, 41(5), 323-362. Risk Management Through Computational Fluid Dynam-
50. Regulations, F. A. (2013). Part 25-Airworthiness stand- ics-Icing Analysis. Journal of Aircraft, 49(2), 611-621.
ards: Transport category airplanes. Federal Aviation Ad- 65. R. Kirchner. (1983). Aircraft icing roughness features and
ministration (FAA), USA. its effect on the icing process. AIAA Paper, (83-0111).
51. Serke, D. J., Reehorst, A. L., & Politovich, M. K. (2010, 66. Vargas, M., & Reshotko, E. (1998). Physical mechanisms
October). Supercooled large drop detection with NASA's of glaze ice scallop formations on swept wings.
Icing Remote Sensing System. In Remote Sensing (pp. 67. Cao, Y., Wu, Z., & Xu, Z. (2014). Effects of rainfall on
782705-782705). International Society for Optics and aircraft aerodynamics. Progress in Aerospace Sciences, 71,
Photonics. 85-127.
52. Brahimi, M. T., Tran, P., Chocron, D., Tezok, F., & Para- 68. Cao, Y., Chen, K., & Sheridan, J. (2008). Flowfield simu-
schivoiu, I. (1997). Effect of supercooled large droplets lation and aerodynamic performance analysis of complex
on ice accretion characteristics. AIAA Paper, (97-0306). iced airfoils with hybrid multi-block grid. Proceedings of
53. Cober, S. G., & Isaac, G. A. (2012). Characterization of the Institution of Mechanical Engineers, Part G: Journal
39
 of Aerospace Engineering, 222(3), 417-422. 86. Cao, Y., Li, G., & Hess, R. A. (2012). Helicopter flight
69. Xu Z. (2013). Aircraft flight characteristics in icing con- characteristics in icing conditions. The Aeronautical
ditions. Beihang University. Journal, 116(1183), 963-979.
70. Valarezo, W. O., Lynch, F. T., & McGhee, R. J. (1993). 87. Cao, Y., Li, G., & Sheridan, J. (2014). Airflow hazard
Aerodynamic performance effects due to small lead- prediction for helicopter flight in icing condition. Pro-
ing-edge ice(roughness) on wings and tails. Journal of ceedings of the Institution of Mechanical Engineers, Part
aircraft, 30(6), 807-812. G: Journal of Aerospace Engineering, 228(1), 147-154.
71. Khodadoust, A., Dominik, C., Shin, J., & Miller, D. 88. Trunov, O. K., & Ingelman-Sundberg, M. (1985). On the
(1995). Effect of In-Flight Ice Accretion on the Perfor- Problem of Horizontal Tail Stall Due to Ice. Report JR-3,
mance of a Multi-Element Airfoil. The Swedish Soviet Working Group on Scien-
72. Papadakis, M., Alansatan, S., & Wong, S. C. (2000). tific-Technical Cooperation in the Field of Flight Safety, 1,
Aerodynamic characteristics of a symmetric NACA sec- 15-18.
tion with simulated ice shapes. In 38th Aerospace Sci- 89. Ranaudo, R. J., Mikkelsen, K. L., McKnight, R. C., &
ences Meeting and Exhibit (p. 98). Perkins Jr, P. J. (1984). Performance degradation of a
73. Gurbacki, H., & Bragg, M. (2000). Sensing aircraft icing typical twin engine commuter type aircraft in measured
effects by flap hinge moment measurement. In 17th Ap- natural icing conditions.
plied Aerodynamics Conference (p. 3149). 90. Baars, W. J., Stearman, R. O., & Tinney, C. E. (2010). A
74. Gurbacki, H. M., & Bragg, M. B. (2002). Unsteady aero- Review on the Impact of Icing on Aircraft Stability and
dynamic measurements on an iced airfoil. AIAA paper, Control. Journal of Aeroelasticity and Structural Dynam-
241, 2002. ics, 2(1).
75. Mirzaei, M., Ardekani, M. A., & Doosttalab, M. (2009). 91. ANC85FA054. Washington, DC. NTSB
Numerical and experimental study of flow field charac- 92. Sanz, Michael (2006). Linjeflyg – ett folkflyg från start
teristics of an iced airfoil. Aerospace science and tech- till landning (in Swedish). Alt om Hobby. pp. 167–168.
nology, 13(6), 267-276. ISBN 91-7243-038-9.
76. Cebeci, T. (1995). Effect of ice on airfoil stall at high 93. Sehgal, B., Deters, R. W., & Selig, M. S. (2002). Icing
Reynolds numbers. AIAA journal, 33(7), 1351-1352. encounter flight simulator. AIAA Paper, 817.
77. Cao, Y., Zhang, Q., & Sheridan, J. (2008). Numerical 94. Miller, R., & Ribbens, W. (1999, January). The effects of
simulation of rime ice accretions on an airfoil using an icing on the longitudinal dynamics of an icing research
Eulerian method. The Aeronautical Journal, 112(1131), aircraft. In 37th Aerospace Sciences Meeting and Exhibit
243-249. (p. 636).
78. Sibilski, K. (1997). Some thoughts on mathematical mod- 95. Ratvasky, T. P., Blankenship, K., Rieke, W., & Brinker, D.
els for aircraft accidents simulation. Aviation Safety H. J. (2003). Iced aircraft flight data for flight simulator val-
Soekha (eds.), VSP Publishing Company, Utrecht, Holan- idation.
dia. 96. Lampton, A., & Valasek, J. (2007). Prediction of icing ef-
79. Hossain, K. N., Sharma, V., Bragg, M. B., & Voulgaris, P. fects on the dynamic response of light airplanes. Journal
G. (2003). Envelope protection and control adaptation in of guidance, control, and dynamics, 30(3), 722-732.
icing encounters. AIAA Paper, 25, 2003. 97. Lampton, A., & Valasek, J. (2008). Prediction of icing ef-
80. Cao, Y., & Chen, K. (2010). Helicopter icing. The Aero- fects on the coupled dynamic response of light airplanes.
nautical Journal, 114(1152), 83-90. Journal of guidance, control, and dynamics, 31(3), 656.
81. Addy Jr, H. E., Broeren, A. P., Zoeckler, J. G., & Lee, S. 98. Lampton, A., & Valasek, J. (2012). Prediction of icing ef-
(2003). A wind tunnel study of icing effects on a business fects on the lateral/directional stability and control of light
jet airfoil. AIAA paper, 727, 2003. airplanes. Aerospace Science and Technology, 23(1),
82. Zhao, G. Q., Zhao, Q. J., & Chen, X. (2016). New 3-D ice 305-311.
accretion method of hovering rotor including effects of 99. Bragg, M. B. (1996). Aircraft aerodynamic effects due to
centrifugal force. Aerospace Science and Technology, 48, large droplet ice accretions. AIAA paper, 932.
122-130. 100. Johnson, C. L. (1940). Wing loading, icing and associated
83. Aviation Accident Report AAR-10-01. Washington, DC. aspects of modern transport design. Journal of the Aero-
NTSB nautical Sciences, 8(2), 43-54.
84. Yu C. (1998). Flying test research of the icing and its ef- 101. In-flight Icing Encounter and Loss of Control Simmons
fects on flight performance for Y12-II aircraft. Flight Airlines, d.b.a. American Eagle Flight 4184 Avions de
Dynamics. Transport Regional (ATR) Model 72-212, N401AM,
85. Cao, Y., Li, G., & Zhong, G. (2010). Tandem helicopter Roselawn, Indiana October 31, 1994; Volume 1:
trim and flight characteristics in the icing condition. (NTSB/AAR-96-01)
Journal of Aircraft, 47(5), 1559-1569. 102. National Transportation Safety Board, “Icing Tanker Test
40
 Factual Report”, Docket No: SA-512, Exhibit No: 13B, detection and identification of aircraft icing and reconfig-
DCA95MA001, Washington D.C., 1995. urable control. Progress in Aerospace Sciences, 60, 12-34.
103. Cao, Y., & Yuan, K. (2007). Aircraft flight characteristics 119. Dong, Y., & Ai, J. (2013). Research on inflight parameter
in conditions of windshear and icing. The Aeronautical identification and icing location detection of the aircraft.
Journal, 111(1115), 41-49. Aerospace Science and Technology, 29(1), 305-312.
104. Petty, K. R., & Floyd, C. D. (2004, October). A statistical 120. Jiang, J., & Yu, X. (2012). Fault-tolerant control systems:
review of aviation airframe icing accidents in the US. In A comparative study between active and passive ap-
Proceedings of the 11th Conference on Aviation, Range, proaches. Annual Reviews in control, 36(1), 60-72.
and Aerospace Hyannis. 121. Verhaegen, M., Kanev, S., Hallouzi, R., Jones, C.,
105. Sibilski, K., Lasek, M., Ladyzynska-Kozdras, E., & Maciejowski, J., & Smail, H. (2010). Fault tolerant flight
Maryniak, J. (2004). Aircraft Climbing Flight Dynamics control-a survey. Fault Tolerant Flight Control, 47-89.
With Simulated Ice Accretion. AIAA Paper, 4948. 122. Idan, M., Johnson, M., Calise, A. J., & Kaneshige, J.
106. Merret, J., Hossain, K., & Bragg, M. B. (2002). Envelope (2001). Intelligent aerodynamic/propulsion flight control
protection and atmospheric disturbances in icing encoun- for flight safety: A nonlinear adaptive approach. In Amer-
ters. AIAA Paper, 814, 2002. ican Control Conference, 2001. Proceedings of the 2001
107. Van Hengst, J., & Boer, J. N. (1991). The effect of (Vol. 4, pp. 2918-2923). IEEE.
hoar-frosted wings on the Fokker 50 take-off characteris- 123. Johnson, E. N., & Calise, A. J. (2001). Neural network
tics. AGARD CP, 496. adaptive control of systems with input saturation. In
108. Bragg, M. B., Heinrich, D. C., Valarezo, W. O., & American Control Conference, 2001. Proceedings of the
McGhee, R. J. (1994). Effect of underwing frost on a 2001 (Vol. 5, pp. 3527-3532). IEEE.
transport aircraft airfoil at flight Reynolds number. Jour- 124. Idan, M., Johnson, M., & Calise, A. J. (2002). Hierar-
nal of Aircraft, 31(6). chical approach to adaptive control for improved flight
109. Cao, Y., Wu, Z., Su, Y., & Xu, Z. (2015). Aircraft flight safety. Journal of Guidance Control and Dynamics, 25(6),
characteristics in icing conditions. Progress in Aerospace 1012-1020.
Sciences, 74, 62-80. 125. Calise, A. J., Hovakimyan, N., & Idan, M. (2001). Adap-
110. Olsen, W., Shaw, R., & Newton, J. (1984). Ice shapes and tive output feedback control of nonlinear systems using
the resulting drag increase for a NACA 0012 airfoil. Na- neural networks. Automatica, 37(8), 1201-1211.
tional Aeronautics and Space Administration. 126. Calise, A. J., Lee, S., & Sharma, M. (2000, August). De-
111. Bowden, D. T. (1956). Effect of Pneumatic De-icers and velopment of a reconfigurable flight control law for the
Ice Formations on Aerodynamic Characteristics of a Air- X-36 tailless fighter aircraft. In AIAA guidance, naviga-
foil. tion, and control conference (pp. 2000-3940).
112. Cebeci, T., & Kafyeke, F. (2003). Aircraft icing. Annual 127. Wise, K. A., Brinker, J. S., Calise, A. J., Enns, D. F.,
review of fluid mechanics, 35(1), 11-21. Elgersma, M. R., & Voulgaris, P. (1999). Direct adaptive
113. Pokhariyal, D., Bragg, M. B., Hutchison, T., & Merret, J. reconfigurable flight control for a tailless advanced fight-
(2001). Aircraft flight dynamics with simulated ice accre- er aircraft. International Journal of Robust and Nonlinear
tion. AIAA Paper, 541. Control, 9(14), 999-1012.
114. Melody, J. W., Baúar, T., Perkins, W. R., & Voulgaris, P. G. 128. AJ, C., Lee, S., & Sharma, M. (1998). Direct adaptive
(2000). Parameter identification for inflight detection and reconfigurable control of a tailless fighter aircraft.
characterization of aircraft icing. Control Engineering 129. Shtessel, Y., Buffington, J., & Banda, S. (1999). Multiple
Practice, 8(9), 985-1001. timescale flight control using reconfigurable sliding
115. Melody, J. W., Hillbrand, T., Baúar, T., & Perkins, W. R. modes. Journal of Guidance, Control, and Dynamics,
(2001). H҄ parameter identification for inflight detec- 22(6), 873-883.
tion of aircraft icing: The time-varying case. Control En- 130. Liu, G. P., & Patton, R. (1998). Eigenstructure assignment
gineering Practice, 9(12), 1327-1335. for control system design. John Wiley & Sons, Inc..
116. Aykan, R., Hajiyev, C., & Çaliúkan, F. (2005). Kalman 131. Mignone, D. (2002). Control and estimation of hybrid
filter and neural network-based icing identification ap- systems with mathematical optimization (Doctoral dis-
plied to A340 aircraft dynamics. Aircraft Engineering and sertation).
Aerospace Technology, 77(1), 23-33. 132. Tousi, M. M., & Khorasani, K. (2009, March). Fault di-
117. Aykan, R., Hajiyev, C., & Caliskan, F. (2005, August). agnosis and recovery from structural failures (icing) in
Aircraft icing detection, identification and reconfigurable unmanned aerial vehicles. In Systems Conference, 2009
control based on Kalman filtering and neural networks. In 3rd Annual IEEE (pp. 302-307). IEEE.
AIAA Atmospheric Flight Mechanics Conference and 133. Rotondo, D., Cristofaro, A., Johansen, T. A., Nejjari, F., &
Exhibit, San Francisco, California. Puig, V. (2015, September). Icing detection in unmanned
118. Caliskan, F., & Hajiyev, C. (2013). A review of in-flight aerial vehicles with longitudinal motion using an LPV
41
 unknown input observer. In Control Applications (CCA), 147. Thoren, R. L. (1948). Icing Flight Tests on the Lockheed
2015 IEEE Conference on (pp. 984-989). IEEE. P2V. ASME paper, (48-SA), 41.
134. Seron, M. M., Johansen, T. A., De Doná, J. A., & Cris- 148. Cooper, W. A., Sand, W. R., Veal, D. L., & Politovich, M.
tofaro, A. (2015, November). Detection and estimation of K. (1984). Effects of icing on performance of a research
icing in unmanned aerial vehicles using a bank of un- airplane. Journal of Aircraft, 21(9), 708-715.
known input observers. In Control Conference (AUCC), 149. Ashenden, R., Lindberg, W., & Marwitz, J. (1996).
2015 5th Australian (pp. 87-92). IEEE. Two-dimensional NACA 23012 airfoil performance deg-
135. Hansen, S., & Blanke, M. (2014). Diagnosis of airspeed radation by super-cooled cloud, drizzle, and rain drop ic-
measurement faults for unmanned aerial vehicles. IEEE ing. AIAA Paper, (96-0870).
Transactions on Aerospace and Electronic Systems, 50(1), 150. Trunov, O. K., & Ingelman-Sundberg, M. (1985). On the
224-239. Problem of Horizontal Tail Stall Due to Ice. Report JR-3,
136. Cristofaro, A., Johansen, T. A., & Aguiar, A. P. (2015, Ju- The Swedish Soviet Working Group on Scien-
ly). Icing detection and identification for unmanned aerial tific-Technical Cooperation in the Field of Flight Safety, 1,
vehicles: Multiple model adaptive estimation. In Control 15-18.
Conference (ECC), 2015 European (pp. 1651-1656). 151. Bragg, M. B. (1996, August). Aerodynamics of super-
IEEE. cooled-large-droplet ice accretions and the effect on air-
137. Sorensen, K. L., Helland, A. S., & Johansen, T. A. (2015, craft control. In Proceedings of the FAA International
March). Carbon nanomaterial-based wing temperature Conference on Aircraft Inflight Icing (Vol. 2, pp.
control system for in-flight anti-icing and de-icing of un- 387-399).
manned aerial vehicles. In Aerospace Conference, 2015 152. Miller, D., Ratvasky, T., Bernstein, B., McDonough, F., &
IEEE (pp. 1-6). IEEE. Strapp, J. W. (1998). NASA/FAA/NCAR Supercooled
138. Isaac, G. A., Cober, S. G., Strapp, J. W., Korolev, A. V., Large Droplet Icing Flight Research: Summary of Winter
Tremblay, A., & Marcotte, D. L. (2001). Recent Canadian 1996-1997 Flight Operations.
research on aircraft in-flight icing. Canadian Aeronautics 153. Dunn, T. A., Loth, E., & Bragg, M. B. (1999). Computa-
and Space Journal, 47(3), 213-221. tional investigation of simulated large-droplet ice shapes
139. Isaac, G. A., Cober, S. G., Strapp, J. W., Hudak, D., Rat- on airfoil aerodynamics. Journal of aircraft, 36(5),
vasky, T. P., Marcotte, D. L., & Fabry, F. (2001, January). 836-843.
Preliminary results from the Alliance Icing Research 154. Lee, S., & Bragg, M. B. (1999). Experimental investiga-
Study (AIRS). In AIAA 39th Aerospace Sci. Meeting and tion of simulated large-droplet ice shapes on airfoil aero-
Exhibit (pp. 2001-0393). dynamics. Journal of Aircraft, 36(5), 844-850.
140. Wang, C., Chang, S., & Wu, H. (2014). Lagrangian Ap- 155. Bragg, M. B., & Loth, E. (2000). Effects of large-droplet
proach for Simulating supercooled large droplets’ Im- ice accretion on airfoil and wing aerodynamics and con-
pingement Effect. Journal of Aircraft, 52(2), 524-537. trol (No. DOT/FAA/AR-00/14). ILLINOIS UNIV AT
141. Honsek, R., Habashi, W. G., & Aubé, M. S. (2008). Eu- URBANA DEPT OF AERONAUTICAL AND
lerian modeling of in-flight icing due to supercooled large ASTRONAUTICAL ENGINEERING.
droplets. Journal of aircraft, 45(4), 1290-1296. 156. Lynch F T, Khodadoust A. Effects of ice accretions on
142. García-Magariño, A., Sor, S., & Velazquez, A. (2015). aircraft aerodynamics [J]. Progress in Aerospace Sciences,
Experimental characterization of water droplet defor- 2001, 37(8): 669-767.
mation and breakup in the vicinity of a moving airfoil. 157. Broeren, A. P., LaMarre, C. M., Bragg, M. B., & Lee, S.
Aerospace Science and Technology, 45, 490-500. (2005). Characteristics of SLD ice accretions on airfoils
143. Huang, J., Nie, S., Cao, Y., Yao, Y., & Yao, J. (2016). and their aerodynamic effects. AIAA Paper, 75.
Multistep Simulation for Three-dimensinal Ice Accretion 158. Whalen, E. A., Broeren, A. P., Bragg, M. B., & Lee, S.
on an Aircraft Wing. In AIAA Modeling and Simulation (2005). Characteristics of runback ice accretions on air-
Technologies Conference (p. 1918). foils and their aerodynamic effects. AIAA paper, 1065.
144. Wright, W. B., & Potapczuk, M. G. (2004). 159. Whalen, E. A., Broeren, A. P., & Bragg, M. B. (2006).
Semi-empirical modeling of SLD physics. Considerations for Aerodynamic Testing of Scaled Run-
145. Bilodeau, D. R., Habashi, W. G., Fossati, M., & Baruzzi, back Ice Accretions. AIAA Paper, 260.
G. S. (2015). Eulerian Modeling of Supercooled Large 160. Cao, Y., Zhong, G., & Ma, C. (2011). Numerical simula-
Droplet Splashing and Bouncing. Journal of Aircraft, tion of ice accretion prediction on multiple element airfoil.
52(5), 1611-1624. Science China Technological Sciences, 54(9), 2296-2304.
146. Morris, D. E. (1952). Designing to Avoid Dangerous Be- 161. Oleskiw, M. M. (2001). A review of 65 years of aircraft
haviour of an Aircraft Due to the Effects of Control Hinge in-flight icing research at NRC. Canadian Aeronautics
Moments of Ice on the Leading Edge of the Fixed Surface. and Space Journal, 47(3), 259-268.
HM Stationery Office. 162. Sor, S., García-Magariño, A., & Velazquez, A. (2016).
42
 Model to predict water droplet trajectories in the flow
past an airfoil. Aerospace Science and Technology, 58,
26-35.
163. http://www.ntsb.gov/about/employment/_layouts/ntsb.avi
ation/brief2.aspx?ev_id=20141107X75136&ntsbno
=CEN15FA040&akey=1
164. Aviation Accident Report AAR-16-01. Washington, DC.
NTSB
165. https://app.ntsb.gov/pdfgenerator/ReportGeneratorFile.as
hx?EventID=20131030X04941&AKey=1&RType=Final
&IType=FA
166. https://app.ntsb.gov/pdfgenerator/ReportGeneratorFile.as
hx?EventID=20130111X25405&AKey=1&RType=Final
&IType=FA
167. Aviation Accident Report AAR-10-01. Washington, DC.
NTSB
168. AAIB Bulletin: 10/2009, G-MAJV, EW/C2008/04/02
169. Aviation Accident Report AAR-11-02. Washington, DC.
NTSB
170. Weener, E. (2011). Lessons from Icing Accidents and In-
cidents. Experimental Aircraft Association.
171. Aviation Accident Report AAR-07-02. Washington, DC.
NTSB
172. REPORT ON THE SERIOUS INCIDENT OVER
GLACIER FOLGEFONNA, NORWAY ON 14.
SEPTEMBER 2005 WITH ATR 42-320, LN-FAO,
OPERATED BY COAST AIR AS
173. Aviation Investigation Report: Loss of Control Georgian
Express Ltd. Cessna 208B Caravan C-FAGA Pelee Island,
Ontario 17 January 2004 Report Number A04H0001
174. Aviation Accident Report AAR-98-04. Washington, DC.
NTSB
175. Aviation Accident Report AAR-96-01. Washington, DC.
NTSB
176. Moshansky, V. P. (1992). Commission of inquiry into the
Air Ontario crash at Dryden, Ontario. Minister of Supply
and Services Canada.
177. Aviation Accident Report AAR-93-02. Washington, DC.
NTSB
178. Report No: 4/1992. Report on the incident to British Aer-
ospace ATP, G-BMYK 10 miles north of Cowly, near
Oxford on 11 August 1991

43