sensors

Article
Detection of Aircraft Touchdown Using Longitudinal
Acceleration and Continuous Wavelet Transformation
Jerzy Bakunowicz 1 and Paweł Rzucidło 2, *
1 Frenchay Campus, The University of the West of England, Coldharbour Lane, Bristol BS16 1QY, UK;
[email protected]
2 Department of Avionics and Control Systems, Faculty of Mechanical Engineering and Aeronautics,
Rzeszów University of Technology, al. Powstańców Warszawy 12, 35-959 Rzeszów, Poland
* Correspondence: [email protected]




Received: 10 November 2020; Accepted: 15 December 2020; Published: 17 December 2020


Abstract: The paper presents a methodology enabling the detection of aircraft touchdowns based on
data obtained from accelerometers attached to the structural parts of the airframe in the cockpit or
passenger compartment. Precise determination of the moment and place of touchdown of the main
landing gear is challenging when analysing parameters such as height, flight speed and rate of
descent. During the tests of the I-31T aircraft, it turned out that vibrations of the aircraft structure
caused by the contact of the front and main landing gear with the ground have a repetitive character.
In particular, this applies to longitudinal acceleration. The use of continuous wavelet analysis
(CWT) allowed finding unique periodic features of the landing phenomenon that distinguish it from
other forms of vibration occurring in individual flight phases. Ground and flight observations of
experimental aeroplane MP-02 Czajka verified the proposed method of virtual touchdown detection.
The results presented in this paper justify that this method may find broader application, especially
for the light aircraft class.

Keywords: virtual sensing; touchdown; detection; aeroplane; light aircraft; landing gear; continuous
wavelet; accelerometers; acceleration; MEMS

1. Introduction
The unambiguous set down of a touchdown point and time of the landing seems to be a
relatively simple task when observing the process from outside of the aeroplane at a fixed position.
It becomes manageable even more when using dedicated optical and recording devices. However,
the process becomes complicated if the measurement takes place directly on board of a commercially
operating aeroplane or, ultimately, when analysing data collected from flight recorders in the paper [1]
touchdown point detection of an A320 airliner is determined based on comparison of data recorded
from two radio altimeters and an accelerometer. Small aeroplanes are not commonly equipped with
specialized sensors and most of the recorded data onboard comes from Global Navigation Satellite
System (GNSS) receivers. On the basis of such data, it is only possible to determine the approximate
landing site [2]. An additional complication comes with the required accuracy both in space and time
domains. Furthermore, the limited number and quality of data available render the task challenging.
The undercarriages of transport and corporate aeroplanes possess weight on wheels (WOW) sensors,
which provide information about the touchdown of a particular leg. There are almost no such solutions
in normally operated sport, tourist aeroplanes and gliders. Experimental application of dynamometer
sensors in a PZL-104 Wilga aeroplane is presented in paper [3]. The same team presented advanced
research with the use of GNSS/INS (inertial navigation system) and an optical sensor for determining
general aviation (GA) aeroplane take-off and landing performance on a grassy airfield in the work [4].

Sensors 2020, 20, 7231; doi:10.3390/s20247231 www.mdpi.com/journal/sensors

 Sensors
Sensors 2020,20,
2020, 20,7231
x FOR PEER REVIEW 2 of 2625
2 of

(inertial navigation system) and an optical sensor for determining general aviation (GA) aeroplane
The use ofand
take-off onboard cameras
landing and image
performance on analysis
a grassysystems
airfield may
in theinwork
the future enable
[4]. The use precise measurements
of onboard cameras
for theimage
and needsanalysis
of normal operation
systems may in of the
small aviation,
future enablebut such measurements
precise works are currently for theatneeds
an experimental
of normal
operation
level [5]. of small aviation, but such works are currently at an experimental level [5].
Theunmistakable
The unmistakabledetermination
determination of of the
the moment when when thethe wheels
wheelscontact
contactthetherunway
runwaysurface
surfaceisis
crucialfor
crucial forautomatic
automaticcontrol
controlsystems
systems andand active
active dumpers [6]. [6]. The
Thesignal
signalenables
enableseffective
effectivebraking
brakingofof
theaircraft.
the aircraft.ItItmay
mayalso
also increase
increase the
the crew
crew situational
situational awareness,
awareness, such suchas asthe
thewarning
warningand andalerting
alerting
systemthat
system thatbraking
brakingisisnot
notpossible
possible under
under given
given meteorological
meteorological conditions,
conditions,for forexample,
example,ininRunway
Runway
OverrunPrevention
Overrun Prevention System
System (ROPS).
(ROPS). TheTheexactexact determination
determination of touchdown
of the the touchdown location
location is of
is of interest
interest to the manufacturers of transport aircraft. The precise detection of the touchdown
to the manufacturers of transport aircraft. The precise detection of the touchdown location and time is location
and
not time is
trivial, not trivial,
according to according
measurementsto measurements
performed on performed
the A350on the A350[7].
aeroplane aeroplane [7].
Designers and operators of remotely controlled aerial
Designers and operators of remotely controlled aerial vehicles face similar vehicles face similar Information
problems. problems.
about the contact between the aircraft undercarriage and the ground may be necessarynecessary
Information about the contact between the aircraft undercarriage and the ground may be to change
to operating
the change themode operating
of themode
controlof system
the control
or tosystem or to
adapt the adapt theparameters
operating operating of parameters
the controlof laws.
the
control
The laws.restrictions
stringent The stringent restrictions
on the dimensions onandthe masses
dimensions and masses
of unmanned of unmanned
aircraft mean that aircraft mean
the installation
that the installation of measuring systems (e.g., miniature WOW equivalents) is often out of the
of measuring systems (e.g., miniature WOW equivalents) is often out of the question. In these cases,
question. In these cases, a virtual sensor seems to be a sensible solution. They can be defined as
a virtual sensor seems to be a sensible solution. They can be defined as sensors measuring at a location
sensors measuring at a location where the sensor should physically be placed, but it is not. Examples
where the sensor should physically be placed, but it is not. Examples of such solutions from various
of such solutions from various fields are presented in the works [8–10]. Virtual measurement
fields are presented in the works [8–10]. Virtual measurement methods enable the determination of
methods enable the determination of specific quantities based on measurements from other
specific quantities based on measurements from other physically available sensors. This work focuses
physically available sensors. This work focuses on this class of sensor solutions, and the motivation
on this class of sensor solutions, and the motivation for it were the results obtained during flight tests
for it were the results obtained during flight tests of an experimental light turbine-powered aircraft
of an experimental light turbine-powered aircraft I-31T [11].
I-31T [11].
2. The Idea of a Touch-Down Virtual Sensing
2. The Idea of a Touch-Down Virtual Sensing
2.1. Genesis
2.1. Genesis
The authors of this work have encountered the problem of determining the exact moment
The authors of this work have encountered the problem of determining the exact moment and
and touchdown point by analysing data gathered during the test campaign of the I-31T turboprop
touchdown point by analysing data gathered during the test campaign of the I-31T turboprop
experimental light aeroplane in the Efficient Systems and Propulsion for Small Aircraft (ESPOSA)
experimental light aeroplane in the Efficient Systems and Propulsion for Small Aircraft (ESPOSA)
project [11]—Figure 1A. During one of the landings, the pilot unexpectedly experienced the shimmy
project [11]—Figure 1A. During one of the landings, the pilot unexpectedly experienced the
vibrations of the front landing gear. The authors presented the analysis of this occurrence in detail in
shimmy vibrations of the front landing gear. The authors presented the analysis of this occurrence
the 2017 paper [12]. One of the problems identified that required solution was the determination of
in detail in the 2017 paper [12]. One of the problems identified that required solution was the
the exact time of a touchdown by airframe vibrations analysis. The information about altitude and
determination of the exact time of a touchdown by airframe vibrations analysis. The information
speeds is, obviously, not sufficient as presented in Figures 2A and 3A. Depending on the wind speed
about altitude and speeds is, obviously, not sufficient as presented in Figures 2A and 3A.
and
Depending ontouchdown
direction, may and
the wind speed occur over a relatively
direction, touchdownwide
mayrange
occurofover
speeds, having wide
a relatively only the speed
range of
relative to the ground in the available data sets.
speeds, having only the speed relative to the ground in the available data sets.

Figure1.1.I-31T
Figure I-31T aeroplane
aeroplane (A) and flight data
data measurement
measurementsystem
system(B).
(B).
Sensors 2020, 20, 7231 3 of 25
Sensors 2020,
Sensors 2020, 20,
20, xx FOR
FOR PEER
PEER REVIEW
REVIEW 33 of
of 26
26

Figure
Figure 2.
Figure2. Time
2.Time plots
Timeplots ofofGS
plotsof GS and
GSand HH
andH GPS (A), ax (B) and az (C) during landing; flight No.03, ESPOSA.
GPS (A), ax (B)
(A), andand
ax (B) az (C) during
az (C) landing;
during flight
landing; No.03,
flight ESPOSA.
No.03, ESPOSA.
GPS

Figure 3.
Figure 3. Time
Time plots
plots of GS
GS and
and HGPS
GPS (A), ax (B) and az (C) during landing; flight No. 07, ESPOSA.
Figure 3. Time plots ofofGS and HH (A), ax (B) and az (C) during landing; flight No. 07, ESPOSA.
GPS (A), ax (B) and az (C) during landing; flight No. 07, ESPOSA.

OnOnthe
On theother
the otherhand,
other hand,the
hand, theinformation
the information on
information on airspeed
on airspeed only only gives
gives aaa general
only gives general impression
generalimpression
impressiononon the
onthe landing
thelanding
landing
process.
process. Most
Most runways
runways and
and
process. Most runways and landing landing
landing sites
sites are
are not
not perfectly
perfectly flat
flat and
and level.
level. The
The differences
differences
perfectly flat and level. The differences between between
between
theextremes
the
the extremesofof
extremes of the
thethe runway
runway
runway may may
may reach
reach
reach up
upup to several
to several
to several tens
tenstens of meters.
of meters.
of meters. Another
Another
Another problem
problem
problem associated
associated
associated with
with
with
the the detection
the detection
detection of a touchdown
of a touchdown
of a touchdown point in
pointapproach
point in a classic a classic
in a classic approach
is the approach is the height
is the measurement.
height or altitude or altitude
height or altitude
In this
measurement. In
measurement. In this case,
case, barometric
barometric andand satellite
satellite measurement methods methods are are not
not sufficiently
sufficiently
case, barometric andthis satellite measurement methods aremeasurement
not sufficiently precise in the time and space
precise in
precise in the time
time and
and space
space domain, as discussed in [13].
domain, as the
discussed in [13]. domain, as discussed in [13].
The task
The task in question
question required
required a solution
solution based
based on data data recorded by by the autonomous
autonomous flight
The task ininquestion required aa solution based on on data recorded
recorded by the the autonomousflight flight
parameter recorder
parameterrecorder
recorder on board the I-31T aeroplane [12,14]. There was no connection between thethe
parameter onon board
board thethe I-31T
I-31T aeroplane
aeroplane [12,14].
[12,14]. There
There was was no connection
no connection betweenbetween
the device
device (Figure
device (Figure 1B)
1B) and
and onboard
onboard installations.
installations. It It also
also had
had an an independent
independent powerpower supply
supply and
and GNSS
GNSS
antenna system [15,16]. The system recorded the
antenna system [15,16]. The system recorded the following data:following data:
Sensors 2020, 20, 7231 4 of 25

(Figure 1B) and onboard installations. It also had an independent power supply and GNSS antenna
system
Sensors [15,16].
2020, 20, xThe
FORsystem recorded the following data:
PEER REVIEW 4 of 26

• ax , ay , az accelerations; sampling frequency 50 Hz,

• a , a , a accelerations; sampling frequency 50 Hz,
•• angular
angularratesratesp, p,
q, r;
q,sampling
r; samplingfrequency 50 Hz,
frequency 50 Hz,
•• Euler Φ, Θ, Ψ,
Euler angles , ,  ; sampling frequencyHz,
angles sampling frequency 50 50 Hz,
•• static
staticpressure
pressure inside
insidethe
thenon-pressurised
non-pressurised cockpit ps ;psampling
cockpit frequency
; sampling frequency10 10
HzHz
•• navigation data
navigation data GPS φ  , GPS
, λ  , H
, H GPS , TT, GS (latitude, longitude, altitude, truetrack,
, TT, GS (latitude, longitude, altitude, true track,and
andground
ground
speed);
speed);sampling
samplingfrequency
frequency 1010
Hz,
Hz,
• • Global
GlobalPositioning
Positioning System
System (GPS)
(GPS) time; sampling
time; sampling frequency
frequency 1010
Hz.Hz.

2.2.
2.2.Continuous
ContinuousWavelet
WaveletTransformation
Transformation(CWT)
(CWT)
The
Themost
moststraightforward
straightforwardanswer answertotothetheresearch
researchquestion
questionwould
wouldseemseemtotobe bethe
theanalysis
analysisof ofaz𝑎
acceleration
accelerationtimetimeseries,
series,because
becauseitsitsrapid
rapidincrease
increaseshould
shouldbebeexpected
expectedatatthethetime
timeofofa atouchdown
touchdown
(Figure
(Figure2C).
2C).This
Thisapproach
approachisiscorrect
correctiningeneral.
general.However,
However,it itisisnot
notsufficient
sufficientbecause
becausethe thetouchdown
touchdown
process
processcancanbebecarried
carriedoutoutsmoothly,
smoothly,without
withoutananunmistakable
unmistakableaccentaccent(Figure
(Figure3C).
3C).OnOnthe theother
otherhand,
hand,
the accelerationaxa takes
theacceleration takesthethenegative
negativevalue
value when
when the tyre tyre touches
touchesthetherunway
runwaysurface.
surface.Therefore
Therefore the
the analysis of the a looks promising (Figures 2B and 3B).
analysis of the x looks promising (Figures 2B and 3B). Unfortunately, Unfortunately, a x was a was not measured
not measured directly
ondirectly
the undercarriage leg. A setleg.
on the undercarriage of micro-electro-mechanical
A set of micro-electro-mechanicalsystem system
(MEMS) accelerometers
(MEMS) [17] was
accelerometers [17]
attached to thetofloor
was attached in theincockpit,
the floor closeclose
the cockpit, to the centre
to the of gravity.
centre The
of gravity. incipient
The incipient touchdown
touchdowneffect effect
manifests
manifestsasasvibrations
vibrationstransmitted
transmittedby bythe
the aircraft’s
aircraft’s structure
structure to the sensors.
sensors. Direct
Directinterpretation
interpretationof
rawaax and aaz time-series
ofraw time-seriesmight
mightintroduce
introducesignificant
significantinaccuracies,
inaccuracies,asasinin[7,12].
[7,12].The
Thecharacteristic
characteristic
phenomenon
phenomenonaccompanying
accompanyingthe thetouchdown
touchdown of of the
the landing gear (both
(both main
main andand front
frontin inthe
thecase
caseof
ofI-31T)
I-31T)isisthe
the formation
formation of of longitudinal
longitudinal oscillations
oscillations (recorded
(recorded mainly in the the form of ax changes;
form of a changes;
Figures
Figures2B2B and
and 3B).
3B).These
Theseoscillations
oscillationsareare
visible if the
visible selected
if the timetime
selected series of recorded
series parameters
of recorded are
parameters
processed with continuous wavelet analysis (CWT). The authors applied Morlet
are processed with continuous wavelet analysis (CWT). The authors applied Morlet wavelet as the wavelet as the basic
wavelet (Figure 4A).
basic wavelet (FigureThis4A).
wavelet
Thisenables
waveletprecise
enableslocation
precise during shock,
location seismic,
during short-term
shock, seismic,and highly
short-term
damped oscillations
and highly damped[18]. oscillations [18].

Figure 4. Example of Morlet wavelet (A) and Gaussian wavelet (B).

Figure 4. Example of Morlet wavelet (A) and Gaussian wavelet (B).
The continuous wavelet transformation of the function f(t) is determined by Equation (1).
The continuous wavelet transformation of the function f(t) is determined by Equation (1). The
The function ψ∗ab ∗ (t) is the basic wavelet and is described by Equation (2). The parameter a is a scale,
function ψ (t) is the basic wavelet and is described by Equation (2). The parameter a is a scale,
while b is a wavelet shift.
while b is a wavelet shift.
Z +∞
CWT(a, b) = f(t)·ψab ∗ (t)dt. (1)
−∞
∗
CWT(a, b) = 1 f(t) ∙ψ
t−b (t)dt. (1)
∗
ψab (t) = √ ψ , (2)
a a
1 wavelet.
∗ mother
where ψ(t) is a continuous function called the t b
ψ (t) = ψ , (2)
√a a
where ψ(t) is a continuous function called the mother wavelet.
The normalisation 1/√a in the equation ensures that the integral energy given by each
translated and dilated wavelet is independent of the scale parameter a [19]. It also guarantees that
Sensors 2020, 20, 7231 5 of 25

Sensors 2020, 20, x FOR PEER REVIEW 5 of 26

√
The normalisation 1/ a in the equation ensures that the integral energy given by each translated
the
andwavelet
dilated transform
wavelet is at each scale aofisthe
independent comparable
scale parameterdirectly to each
a [19]. other
It also and to the
guarantees transforms
that of
the wavelet
different
transform time seriesscale
at each [20]. aThe mother wavelet
is comparable should
directly satisfy
to each the and
other so-called,
to the admissibility
transforms ofcondition
different
given by Equation (3).
time series [20]. The mother wavelet should satisfy the so-called, admissibility condition given by
Equation (3). |Ψ(f)|
(f) df < ∞,
Z ∞  2 (3)
Ψ|f|
df < ∞, (3)
−∞ |f|
where Ψ(f)—Fourier transform of ψ(t).
where Ψ(f) Fourier
Fourier transformtransform ψ(t).
and itsofmodifications are base on harmonic waves, while the functions in
Fourier transform and its modifications
CWT are families of wavelets (e.g., Morlet coarse are base on harmonic
wavelets (Figurewaves,
4A), Gauss while(Figure
the functions in CWT
4B), Mexican
are families
Hat, Meyer’sofinfinite
wavelets (e.g.,wavelets,
regular Morlet coarse wavelets
Daubechies (Figure 4A),
orthogonal Gauss and
wavelets (Figure 4B), Mexican
compact Hat,
bi-wavelets
Meyer’s infinite regular wavelets, Daubechies orthogonal wavelets and compact
based on B-splines). CWT transform enables the full reconstruction of function f(t) by inverse bi-wavelets based on
B-splines).
transform (4).CWT transform enables the full reconstruction of function f ( t ) by inverse transform (4).

11 da ∙ db
Z +∞ Z +∞
da·db
f(t)
f(t)== CWT(a,
CWT(a,b) b)ψψab ∗∗((t)
t) 2 . . (4)
(4)
CCψ −∞ −∞ a a

Continuous wavelet
Continuous wavelet analysis
analysis enables
enables precise
precise location
locationinin time
time of
of fast-changing
fast-changingphenomena.
phenomena.
Decreasingthe
Decreasing scalea aincreases
thescale increasesthe thefrequency
frequency carrier
carrier of of
thethe wavelet,
wavelet, but but
this this is connected
is connected to
to the
the simultaneous increase of its time carrier. The resolution of the transformation
simultaneous increase of its time carrier. The resolution of the transformation relative to time andrelative to time and
frequency is dependent
frequency dependenton onthe scale
the factor
scale factor a. aAs
a. As result of the of
a result scaling process,process,
the scaling the time the
and time
frequency
and
are divided
frequency areschematically in Figurein
divided schematically 5A. According
Figure to the Heisenberg
5A. According ambiguity
to the Heisenberg principle,
ambiguity it is not
principle,
itpossible to obtain
is not possible toany resolution
obtain in the time
any resolution and
in the frequency
time domaindomain
and frequency simultaneously.
simultaneously.

Figure 5. Symbolic
Figure Symbolicdivision
divisionof of
thethe
time-frequency planeplane
time-frequency in the wavelet transform
in the wavelet (A) and for
transform comparison
(A) and for
in the short-time
comparison in theFourier transform
short-time Fourier(STFT) (B). (STFT) (B).
transform

Wavelet analysis
Wavelet analysis gives
gives better
better frequency
frequency resolution
resolution and
and worse
worse time
time resolution
resolution for
for large
large scale
scale
values. For
values. For small
small scale valuesa we
scalevalues 𝑎 wehave the the
have opposite situation.
opposite For comparison,
situation. in short
For comparison, in time
shortFourier
time
Fourier transformation (STFT), the resolution in the time and frequency domain is constant,itmaking
transformation (STFT), the resolution in the time and frequency domain is constant, making difficult
itindifficult
many cases to isolate
in many cases short-term processes processes
to isolate short-term (Figure 5B).
(Figure 5B).
2.3. Proposed Method and Algorithm
2.3. Proposed Method and Algorithm
Figure 6 shows the results of continuous wavelet transform of a and a accelerations in a 50 s
Figure 6 shows the results of continuous wavelet transform of a xand a z accelerations in a 50 s
sequence. The touchdown occurred after 25 s relative to the starting time from the beginning of
sequence. The touchdown occurred after 25 s relative to the starting time from the beginning of the
the presented sequence. This landing was characterised by an evident touchdown of the main landing
presented sequence. This landing was characterised by an evident touchdown of the main landing
gear, visible on the CWT scalograms of the a and a variables. As mentioned earlier, touchdown
gear, visible on the CWT scalograms of the a x and a z variables. As mentioned earlier, touchdown
detection based on the analysis of the a variable will not always be useful. In the case of a smooth and
detection based on the analysis of the az variable will not always be useful. In the case of a smooth
delicate touchdown, the vibrations along this axis will have negligibly small amplitude. An example
and delicate touchdown, the vibrations along this axis will have negligibly small amplitude. An
of such a landing is presented in Figure 7. Continuous wavelet analysis enables accurate touchdown
example of such a landing is presented in Figure 7. Continuous wavelet analysis enables accurate
detection, perfectly visible on the ax plot (Figure 7A). The amplitude of az vibrations, in this case,
touchdown detection, perfectly visible on the a plot (Figure 7A). The amplitude of a vibrations,
in this case, is comparable to the amplitude of vibrations coming from other sources (Figure 7B)—such
as the engine or the rolling of the plane on the runway surface. In both cases presented, the landing
 Sensors 2020, 20, 7231 6 of 25

is comparable to the amplitude of vibrations coming from other sources (Figure 7B)—such as the engine
or the rolling of the plane on the runway surface. In both cases presented, the landing might have
looked like the front gear touched the runway almost precisely at the same time as the main (so-called
three points landing). Alternatively, the aeroplane’s nose lowered very gently. Therefore the front
wheel impact on the runway is barely visible. In many other flights, it was possible to identify
the moment when the aeroplane’s nose lowered, and the front wheel touched the ground. An example
of such a landing is shown in Figure 8. Vibrations along the x-axis, associated with the contact of
individual wheels (main landing gear and front wheel), are visible both on the time series of ax and
the corresponding CWT transform. Interestingly, this landing also did not mark any changes in
the course of az and the corresponding CWT scalogram. Practical absence of ax oscillations during
the flight (in particular in the range of the frequencies analysed) becomes an additional argument in
favour of using acceleration analysis in the longitudinal axis to identify the touchdown of the aircraft.
Sensors 2020, 20, x FOR PEER REVIEW 6 of 26
The oscillations of the variable az in flight are visible in Figures 2C, 3C, 6B and 7B.
Therefore, the reader may ask the question: “what is the point of using CWT since one can
might have looked like the front gear touched the runway almost precisely at the same time as the
easily recognise the touchdown phenomenon on ax time histories?” However, the CWT charts allow
main (so-called three points landing). Alternatively, the aeroplane’s nose lowered very gently.
precise location of the phenomenon in time, with accuracy close to the resolution of the transform.
Therefore the front wheel impact on the runway is barely visible. In many other flights, it was possible to
The touchdown of the aeroplane causes a rapid change in the acceleration ax . A step function can
identify the moment when the aeroplane’s nose lowered, and the front wheel touched the ground. An
be mapped using an infinite number of harmonic components. Assuming such nature of the a
example of such a landing is shown in Figure 8. Vibrations along the 𝑥-axis, associated with the contact x
during touchdown, one can expect its trace in all frequency components of the signal. The CWT
of individual wheels (main landing gear and front wheel), are visible both on the time series of a and
resolution for scale equal to 1 is close to the Nyquist frequency (limit frequency) of the discrete signal.
the corresponding CWT transform. Interestingly, this landing also did not mark any changes in the
Finding the phenomenon trace for smaller scales allows for much more accurate time analysis of
course of a and the corresponding CWT scalogram. Practical absence of a oscillations during the
CWT (Figure 5A) than in methods based on Fourier transform, in which the accuracy of location
flight (in particular in the range of the frequencies analysed) becomes an additional argument in favour
of phenomena in time is constant, independent of the signal frequency and equal to the width of
of using acceleration analysis in the longitudinal axis to identify the touchdown of the aircraft. The
the analysing window (Figure 5B).
oscillations of the variable a in flight are visible in Figures 2C, 3C, 6B and 7B.

Figure 6. Continuous
Figure wavelet
6. Continuous transform
wavelet of aof
transform x (A) andand
ax (A) az (B); flight
az (B); No.No.
flight 03, ESPOSA.
03, ESPOSA.
Sensors 2020,
Sensors 20,20,
2020, x FOR
7231 PEER REVIEW 7 of7 of
2625
Figure 6. Continuous wavelet transform of ax (A) and az (B); flight No. 03, ESPOSA.
Therefore, the reader may ask the question: “what is the point of using CWT since one can
easily recognise the touchdown phenomenon on a time histories?” However, the CWT charts
allow precise location of the phenomenon in time, with accuracy close to the resolution of the
transform. The touchdown of the aeroplane causes a rapid change in the acceleration a . A step
function can be mapped using an infinite number of harmonic components. Assuming such nature
of the a during touchdown, one can expect its trace in all frequency components of the signal. The
CWT resolution for scale equal to 1 is close to the Nyquist frequency (limit frequency) of the discrete
signal. Finding the phenomenon trace for smaller scales allows for much more accurate time analysis
of CWT (Figure 5A) than in methods based on Fourier transform, in which the accuracy of location
of phenomena in time is constant, independent of the signal frequency and equal to the width of the
analysing window (Figure 5B).
Interpretation of time series can be cumbersome and requires an individual approach and
assessment of the phenomenon. The ability to assess not only the amplitude but also the frequency
of structural vibrations occurring during the landing of the aircraft immediately (free vibrations and
vibrations coming from the ground after the touchdown) favours using a continuous wavelet
transform. These frequencies are in the range of 5–16 Hz for the I-31T aeroplane (Figures 6–8). Figure
8 shows a typical landing. The touchdown of the front wheel followed touchdown of the main
landing gear, and both are visible on plots. In Figure 9, a detailed analysis of this process was carried
out for 4 s. Oscillations of a acceleration related to the touchdown of the main landing gear are
relatively short in duration, extinguishing very rapidly after 0.5 s. This promises a good starting
point for a virtual sensor
Figure
Figure characterised
7.7.Continuous
Continuous bytransform
wavelet
wavelet quick response
transformofofaaxx(A)time.
(A) andaazz (B);
and (B); flight
flight No.
No. 07, ESPOSA.
ESPOSA.

Figure
Figure8. 8.
Time plots
Time ofof
plots GS, HH
GS, GPS (A),
GPS ax a(B)
(A), and
x (B) continuous
and wavelet
continuous transform
wavelet of of
transform ax a(C); flight
x (C); No.
flight No.
01b, ESPOSA.
01b, ESPOSA.

Interpretation of time series can be cumbersome and requires an individual approach and
assessment of the phenomenon. The ability to assess not only the amplitude but also the frequency
of structural vibrations occurring during the landing of the aircraft immediately (free vibrations and
vibrations coming from the ground after the touchdown) favours using a continuous wavelet transform.
These frequencies are in the range of 5–16 Hz for the I-31T aeroplane (Figures 6–8). Figure 8 shows
a typical landing. The touchdown of the front wheel followed touchdown of the main landing gear,
and both are visible on plots. In Figure 9, a detailed analysis of this process was carried out for 4 s.
Oscillations of ax acceleration related to the touchdown of the main landing gear are relatively short
 Sensors 2020, 20, 7231 8 of 25

in duration, extinguishing very rapidly after 0.5 s. This promises a good starting point for a virtual
sensor
Sensors characterised
2020, 20, x FOR PEERby quick response time.
REVIEW 8 of 26

Figure
Figure9. 9.
Accurate
Accurate location of of
location thethe
touchdown based
touchdown onon
based thethe
value ax a(A)
value and
x (A) continuous
and wavelet
continuous wavelet
transform
transform (B); flight
(B); No.
flight 01b,
No. ESPOSA.
01b, ESPOSA.

Touchdown
Touchdown detection
detection directly
directly based
based on on
the the values
values of theof wavelet
the wavelet transform
transform coefficients
coefficients of theof
acceleration 𝑎 requires
the acceleration ax requires
analysis of the obtained scalograms. In practice, it could be reduced to to
analysis of the obtained scalograms. In practice, it could be reduced aa
qualitativevisual
qualitative visual interpretation
interpretation ofof thetheimage,
image,asas mentioned
mentioned above. However,
above. However,it is possible to develop
it is possible to
functions that implement automatic touchdown detection in the analysed
develop functions that implement automatic touchdown detection in the analysed data sets. The data sets. The authors
focused
authors on relationships
focused that allow
on relationships that determining the moment
allow determining of touchdown
the moment based on
of touchdown the recorded
based on the
acceleration
recorded acceleration x only. a
values avalues Other
only. flight parameters,
Other if necessary,
flight parameters, can provide
if necessary, canadditional support for
provide additional
the method.
support for theThe verification
method. of the technique
The verification used threeused
of the technique different
threevirtual signals
different (VSig).
virtual signals (VSig).
TheThefirst
first virtual signal
signal(VSig 1 ),1),
(VSig described
described by the
by Formula
the Formula (5), assumes that extreme
(5), assumes that CWT
extreme coefficients
CWT
will occur at
coefficients theoccur
will time ofat landing.
the time The maximum
of landing. Thevalues
maximum soughtvalues
were divided
sought into
weretwo ranges:into
divided 1–31 and
two
ranges: 1–31 and 32–256, respectively. By observing a series of landings, it was established (Table 1)of
32–256, respectively. By observing a series of landings, it was established (Table 1) that the weight
thethe
that components
weight ofwith frequencies corresponding
the components with frequencies to the 32–256 scalestoisthe
corresponding much higher
32–256 thanisfor
scales lower
much
scale values.
higher than forThe CWT
lower coefficient
scale values.reached
The CWT its maximum
coefficientvalue in the
reached itsrange 1–31 invalue
maximum landing No. range
in the 15 only.
Therefore,
1–31 an empirical
in landing No. 15 only.weight factor of
Therefore, an4empirical
was adopted weight in formula:
factor of 4 was adopted in formula:
(b) = max CWT(a, b) +
VSig VSig max max
4 ∙ CWT(a,
1 (b∈) = max CWT(a, b∈) + 4·CWTb).
(a, b). (5)(5)
a∈<1–31> a∈<32–256>

The second
Table relationship
1. The coefficients (6), as well
calculated asdetection
using the thirdalgorithms
(7), assumes
in 22that during
landings of the touchdown,
MP-02 aeroplane.the sum
of the scale factor values will reach the local extreme.
Max. Value
Landing of
of CWT VSig1 256 VSig2 VSig3
No. MP-02 Remarks
X
Coeff.VSig2 (Peak
b) =/Delay
0.512· PeakCWT (a, b). Peak /Delay
/Delay (6)
(GPS Time)
(Freq. Scale) a=32
T&G, concrete
17 May 2017 #1 256
1 0.60 (255) 2.58/0.09 s X 4.34/0.07 s 3.41/0.06 s 26
10:48:31.9 VSig3 (b) = 0.373· CWT(a, b). (7)
EPRJ
a=1
T&G, concrete
17 May 2017 #2
2 0.30 (55) 1.45/0.05 s 2.49/0.05 s 2.02/0.05 s 26
10:54:21
EPRJ
17 May 2017 #3 T&G, grass 26
3 0.20 (249) 1.04/0.10 s 1.86/0.09 s 1.18/0.07 s
11:00:39.7 EPRJ
T&G, concrete
17 May 2017 #4
4 0.20 (255) 0.93/0.08 s 2.75/0.06 s 2.24/0.06 s 26
11:05:14.5
EPRJ
T&G, concrete
12 July 2017 #1
5 0.70 (24) 0.84/0.09 s 0.32/0.09 s 0.85/0.09 s 26 *
11:59:46.7
EPRJ
Sensors 2020, 20, 7231 9 of 25

Table 1. The coefficients calculated using detection algorithms in 22 landings of MP-02 aeroplane.

Max. Value
Landing of
of CWT VSig1 VSig2 VSig3
No. MP-02 Remarks
Coeff. Peak /Delay Peak /Delay Peak /Delay
(GPS Time)
(Freq. Scale)
17 May 2017 #1 T&G, concrete 26
1 0.60 (255) 2.58/0.09 s 4.34/0.07 s 3.41/0.06 s
10:48:31.9 EPRJ
17 May 2017 #2 T&G, concrete 26
2 0.30 (55) 1.45/0.05 s 2.49/0.05 s 2.02/0.05 s
10:54:21 EPRJ
17 May 2017 #3 T&G, grass 26
3 0.20 (249) 1.04/0.10 s 1.86/0.09 s 1.18/0.07 s
11:00:39.7 EPRJ
17 May 2017 #4 T&G, concrete 26
4 0.20 (255) 0.93/0.08 s 2.75/0.06 s 2.24/0.06 s
11:05:14.5 EPRJ
12 July 2017 #1 T&G, concrete 26 *
5 0.70 (24) 0.84/0.09 s 0.32/0.09 s 0.85/0.09 s
11:59:46.7 EPRJ
12 July 2017 #2 T&G, concrete 26
6 0.45 (184) 2.02/0.04 s 3.79/0.05 s 3.02/0.00 s
12:05:15.9 EPRJ
12 July 2017 #3 T&G, grass 26
7 0.32 (255) 1.32/0.16 s 2.25/0.10 s 1.84/0.39 s
12:10:50.2 EPRJ
12 July 2017 #4 LDG, concrete 26
8 0.27 (140) 1.15/0.05 s 1.82/0.00 s 1.46/0.00 s
12:16:13.5 EPRJ
12 July 2017 #5 T&G, concrete 26
9 0.18 (254) 0.88/0.04 s 1.60/0.04 s 1.38/0.00 s
13:23:22.2 EPRJ
12 July 2017 #6 LDG, concrete 26
10 0.25 (56) 1.13/0.03 s 1.74/0.02 s 1.57/0.03 s
13:53:26.3 EPRJ
24 November
T&G, grass 26
11 2017 #1 0.22 (100) 1.14/0.06 s 2.07/0.05 s 1.24/0.06 s
EPRJ
12:28:13.1
24 November
T&G, grass 26
12 2017 #2 0.22 (50) 1.40/0.01 s 3.58/0.09 s 2.99/0.09 s
EPRJ
12:33:41.4
24 November
T&G, grass 26
13 2017 #3 0.18 (235) 0.81/0.02 s 3.59/0.04 s 2.72/0.04 s
EPRJ
12:39:29.9
24 Nov 2017 #4 T&G, grass 26
14 0.35 (48) 1.48/0.05 s 2.41/0.09 s 1.94/0.09 s
12:45:05.7 EPRJ
24th November
T&G, grass 26
15 2017 #5 0.21 (30) 0.95/0.09 s 1.74/0.07 s 1.51/0.02 s
EPRJ
12:50:52.1
24 November
LDG, concrete 26
16 2017 #6 0.24 (40) 1.02/0.04 s 1.92/0.02 s 1.54/0.02 s
EPRJ
12:55:19.6
11 May 2018 #1 T&G, concrete 08
17 0.30 (56) 1.30/0.03 s 2.15/0.02 s 1.72/0.02 s
08:42:58.3 EPRJ
11 May 2018 #2 T&G, concrete 26
18 0.35 (52) 1.4/0.03 s 2.19/0.02 s 1.74/0.04 s
08:48:13.3 EPRJ
11 May 2018 #3
19 0.25 (52) 1.25/0.01 s 1.65/0.01 s 1.63/0.01 s LDG, Laszki 77
09:10:05.3
27 July 2018 #1 T&G, concrete 08
20 0.70 (256) 2.70/0.02 s 4.92/0.01 s 3.89/0.01 s
14:56:53.6 EPRJ
27 July 2018 #2 T&G, concrete 08
21 0.55 (256) 2.40/0.04 s 4.71/0.02 s 3.59/0.04 s
15:02:11.9 EPRJ
27 July 2018 #3 LDG, concrete 08
22 0.30 (108) 1.35/0.07 s 2.36/0.03 s 1.72/0.04 s
15:07:12.1 EPRJ
*—runway touched by left main wheel only, LDG—landing, T&G—touch and go, EPRJ—Rzeszów-Jasionka
airport code.

Formula (6) takes into account the sum of CWT for the 32–256 scale range, while Formula (7)
takes into account the entire range of frequencies. The scaling coefficients 0.512 and 0.373 have been
chosen deliberately, so that the standard deviations of noise (σin ) of the virtual signals (6) and (7) have
the same values with VSig1 (5). The coefficients were calculated using the measurement database
from research flights of I-31T. To determine whether a touchdown has been detected at a given time,
 Sensors 2020, 20, x FOR PEER REVIEW 10 of 26

VSig (b) = 0.373 ∙ CWT(a, b). (7)

Sensors 2020, 20, 7231 10 of 25

Formula (6) takes into account the sum of CWT for the 32–256 scale range, while Formula (7)
takes into account the entire range of frequencies. The scaling coefficients 0.512 and 0.373 have been
the relationship (8) should be used, which compares the values returned by the virtual signals with
chosen deliberately, so that the standard deviations of noise (σin) of the virtual signals (6) and (7)
the sensitivity threshold.
have the same values with VSig1 (5). The coefficients were calculated using the measurement
database from research flights of I-31T. To determine whether a touchdown has been detected at a
1 if VSigi (b) ≥ Thri
(8) should be (used,
given time, the relationshipDETECTION b) = which compares the values
. returned by the virtual (8)
0 otherwise
signals with the sensitivity threshold.
The threshold values enable detection
(b)of=the VSig (b) ≥ Thr
1 if smoothest landing, because of the signal to(8) noise
DETECTION .
0 otherwise on one wheel followed by climbs) was
ratio (SNR) level. Incomplete landing No. 5 (short touchdown
excludedThe threshold
from values(Table
the analyses enable1).
detection of the smoothest
The individual landing,
SNR values becausesignals
for virtual of the signal to noise
were determined
ratio (SNR) level. Incomplete landing No. 5 (short touchdown
using the measurement data and according to the dependency [21]: on one wheel followed by climbs) was
excluded from the analyses (Table 1). The individual SNR values for virtual signals were determined
Ai
using the measurement data and according to the dependency [21]:
SNRi = . (9)
A σin
SNR = . (9)
σ
Calculated values of SNR for VSig1 –VSig3 are: SNR1 = 9.34 dB, SNR2 = 12.19 dB and
SNR3 = Calculated values of SNR
10.59 dB. Detection thresholds 1–VSig
for VSigwere 3 are: SNR1 =
determined 9.34 dB, SNR
reducing 2 = 12.19
the value of dB and SNR
virtual 3 = 10.59
signal weakest
peak (Ai ) by 10% and rounding to the nearest tenth: Thr1 = 0.8, Thr2 = 1.5 and Thr3 = 1.
dB. Detection thresholds were determined reducing the value of virtual signal weakest peak (A i) by

10% and rounding to the nearest tenth: Thr1 = 0.8, Thr2 = 1.5 and Thr3 = 1.
3. Verification of the Methodology
3. Verification of the Methodology
3.1. Testing Platform
3.1. Testing Platform
A different experimental aeroplane was used to verify the methodology; a flying laboratory
based onAadifferent experimental
serial ultralight aeroplane
design was used
MP-02 Czajka to verify
(Figure 10). the methodology;
It was a flying
equipped with laboratory
a control system
based on a serial ultralight design MP-02 Czajka (Figure 10).
implemented as part of the terrain flying surveyor (LOT—latajacy It was ˛ obserwator terenu system
equipped with a control in Polish)
implemented as part of the terrain flying surveyor (LOT—latający obserwator terenu in Polish)
project [22]. The avionics installed onboard included, among others, integrated Dynon Avionics D700.
project [22]. The avionics installed onboard included, among others, integrated Dynon Avionics
The system allows recording of many flight parameters, navigation data as well as the parameters of
D700. The system allows recording of many flight parameters, navigation data as well as the
the engine. The sampling rates are 4 Hz in the emergency flight recorder mode and up to 16 Hz in
parameters of the engine. The sampling rates are 4 Hz in the emergency flight recorder mode and up
the to
user-programmed mode.
16 Hz in the user-programmed mode.

Figure10.
Figure 10.The
Theexperimental
experimental MP-02b
MP-02b“Czajka”
“Czajka”aeroplane.
aeroplane.

In contrast
In contrast to to
thethe datasets
data setsobtained
obtained during
during the
the I-31T
I-31T tests,
tests,the
thedata
datafrom
fromthethe
D700 on on
D700 MP-02
MP-02
recorder
recorder include
include information
information on indicated
on indicated air speed
air speed (IAS), (IAS),
true airtrue air (TAS)
speed speedand
(TAS) andrevolutions
engine engine
per minute (RPMs). Unfortunately, the a acceleration has a quantisation level not
per revolutions
minute (RPMs). Unfortunately, the az acceleration has a quantisation level not higher than 0.1,
higher than 0.1, which, in practice, makes it impossible to analyse this parameter quantitatively. The
which, in practice, makes it impossible to analyse this parameter quantitatively. The D700 system
D700 system does not allow the acquisition of a acceleration. Additional, custom made, onboard
does not allow the acquisition of ax acceleration. Additional, custom made, onboard equipment
includes the miniature PCDL-01 recorder (Figure 11) and a network of small measuring modules [23],
in particular:

• PCDA-01 air data computer;

• satellite navigation receiver PCGP-01;
• PCIM-01 inertial measurement system.
 Sensors 2020, 20,
equipment x FOR PEER
includes theREVIEW 11 of 26
miniature PCDL-01 recorder (Figure 11) and a network of small measuring
modules [23], in particular:
equipment includes the miniature PCDL-01 recorder (Figure 11) and a network of small measuring
•modules
PCDA-01
[23], inairparticular:
data computer;
Sensors• 2020,satellite
20, 7231navigation receiver PCGP-01; 11 of 25
•• PCDA-01 air data computer;
PCIM-01 inertial measurement system.
• satellite navigation receiver PCGP-01;
• PCIM-01 inertial measurement system.

Figure
Figure 11. 11. Flight
Flight datameasurement
data measurement system
system (A)
(A)and
andground
groundsurvey system
survey (B). (B).
system

The devices Figure 11. Flight data measurement system (A) and ground survey system (B).
operate
The devices operate in theinCANaerospace
the CANaerospace standard standard
[24,25], [24,25],
enabling enabling measurement
measurement and
and recording
recording of inertial quantities with a frequency of 1 kHz. Navigation data is acquired and recorded
of inertialThequantities
devices with a frequency
operate in the of 1 kHz. Navigation
CANaerospace data[24,25],
is acquired and recorded at a frequency
at a frequency of 10 Hz, while aerometric data is standard
at a frequency of enabling
100 Hz. Both measurement
the D700 and and
of 10 recording
Hz, whileofaerometric
inertial data
quantities is at
with a frequency
a frequency ofof 100
1 Hz.
kHz. Both
Navigation the D700
data and
is PCDL-01
acquired and systems
recorded allow
PCDL-01 systems allow synchronisation of recorded data with the standard GPS time. In practice,
synchronisation
at a means
this frequencyof recorded
of 10 Hz, data
the possibility while with
of mutual
the
aerometric standard
data is GPS
synchronisation at of time.sets
a data
frequency In practice,
of 100 Hz.
obtained
thisBoth
from
means
twothe
the
D700 possibility
independent and
of mutual
PCDL-01 synchronisation
systems allow of data sets
synchronisation obtained
of from
recorded two
data independent
with
measurement and recording systems. The D700 system has 1/16 of a second accuracy. The PCDL-01 the measurement
standard GPS time. Inand recording
practice,
this means
systems.
and The D700
PCGP-01 thesystems
possibility
systemhave hasofbeen
mutual
1/16 of asynchronisation
designedsecond accuracy.
to achieve of data
Thesets
accuracy obtained
PCDL-01 and
in synchronisationfrom two
withindependent
PCGP-01 a systems
GPS time have
measurement
been standard
designedclose and recording
to achieve
to 0.001 accuracy
s. systems.
in synchronisation with a GPS time standard close toPCDL-01
The D700 system has 1/16 of a second accuracy. The 0.001 s.
and
During PCGP-01
During systems
the tests,
the tests, have
the the been
recorder’s
recorder’s designed
locationto
location achieve
was
was asascloseaccuracy
close the in
to the
to synchronisation
centre
centre of
ofgravity
gravityasas with a GPS
possible
possible (15time
cmcm in
(15
standard close to 0.001 s.
frontinoffront of the centre
the centre of gravity
of gravity in theinlongitudinal
the longitudinal axis, axis, as shown
as shown in in Figure11A).
Figure 11A).The The aircraft
aircraft also
also had
During the tests, the tires
recorder’s location wastoas seeclose to the
the wheels
centre of gravity as possible (15 12).
cm
whitehad
in
white
markers
front of
markers
the
the on
on centre the
tires
of
so that
gravity
so it
that
in
is iteasier
the
is easier
to see
longitudinal
when
when
axis, as
the
shown
wheels
in
touch
touch
Figure
thethe
11A).
ground
ground
The
(Figure
aircraft
(Figure
also
12).
The exact moment of touchdown was identified by video analysis (the image in the ground station
The exact moment
had white of touchdown
markers with
on theGPS tirestime was
so that identified
it is easier by video analysis (the image in the ground station
was synchronized as well as theto see when
onboard the wheels
recorder). touch
White the ground
markers were(Figure
applied12). to
was synchronized
The exact moment with of GPS time aswas
touchdown well as the onboard
identified by video recorder).
analysis (the White
image markers
in the werestation
ground applied to
the tires in order to facilitate the assessment of the time when the wheels came into contact with the
the tires
was in order
synchronized
ground.
to facilitate
The obtained GPSthe
with contact time assessment
timeas well as the
determines
of onboard
the timerecorder).
the moment
when the
“0” for
wheels
White
the markers
came
data from
into
were contact
theapplied
onboard to with
the ground.
the tires The
in obtained
order to contact
facilitate the time
assessmentdetermines
of the the
time moment
when the
recorder (Figure 12B). For this purpose, a precise GPS receiver was used, updating the navigation“0”
wheels for the
came data
into from
contact the
with onboard
the
ground.
recorder
message The
(Figure obtained
with12B). Forcontact
a frequency this
of 20 time
purpose,
Hz. The determines
amarkers
preciseof the themoment
GPS receiver
current “0”
GPS wasfor the
used,
time anddata from
updating
the system thethe
timeonboard
navigation
of the
recorder
message
ground with (Figure 12B).
a frequency
recorder For this
(ruggedofmobile20 Hz.purpose, a precise
The markers
computer) GPS receiver
of thetocurrent
are applied was
individual used,
GPS updating
timeframes
video and the the navigation
system
obtained fromtime of
message
the video with a frequency
camera (Figureof11B).20 Hz. Thesolution
This markersmakesof the current
it possibleGPS to time and the system
determine time of the
the characteristic
the ground recorder (rugged mobile computer) are applied to individual video frames obtained
ground recorder (rugged mobile computer) areaircraft
applied to individual videoupframes
to 0.05 obtained from
fromphenomena occurring
the video camera during
(Figure the landing
11B). of an
This solution makes with itan accuracyto
possible ofdetermine s.the characteristic
the video camera (Figure 11B). This solution makes it possible to determine the characteristic
phenomena occurring during the landing of an aircraft with an accuracy of up to 0.05 s.
phenomena occurring during the landing of an aircraft with an accuracy of up to 0.05 s.

Figure 12. Time-lapse recording of MP-02 touchdown based on the image from the camera and GPS time:
just before (A) and just after touchdown (B).
Figure
Figure 12. Time-lapse
12. Time-lapse recording of
recording of MP-02
MP-02touchdown based
touchdown on the
based onimage from the
the image camera
from theand GPS time:
camera and GPS
time:just
justbefore
before(A)(A)
andand
just after touchdown
just after (B).
touchdown (B).

3.2. Selection and Accuracy of the MEMS Sensors Used

The sensors for the I-31T aircraft were selected due to other experimental studies that were
underway, before the authors took up the topic of virtual touchdown detection. Verification was
carried out on the MP-02 aircraft, on which accelerometers with a frequency of 1 kHz were intentionally
installed. The MEMS class of sensors was selected, as they are very often used in modern integrated
 Sensors 2020, 20, x FOR PEER REVIEW 12 of 26

Sensors3.2. Selection
2020, 20, 7231and Accuracy of the MEMS Sensors Used 12 of 25
The sensors for the I-31T aircraft were selected due to other experimental studies that were
underway, before the authors took up the topic of virtual touchdown detection. Verification was
systems for indicating
carried out on theand recording
MP-02 flight
aircraft, parameters.
on which In the future,
accelerometers with this may allow
a frequency of the method
1 kHz were to be
widely used in practice.
intentionally installed. The MEMS class of sensors was selected, as they are very often used in
In orderintegrated
modern to check measurement uncertainty
systems for indicating of the sensors,
and recording root mean square
flight parameters. (RMS)this
In the future, values
maywere
calculated
allow for
the recordings 106 samples
method to beofwidely used inin static state. Results obtained for accelerometers installed in
practice.
axes x andInzorder to check measurement uncertainty of the sensors, root mean square (RMS) values were
are respectively:
calculated for recordings of 106 samples in static state. Results obtained for accelerometers installed
• 0.0019
in axes and
x and0.0052 for I-31T measurement system (sampling frequency of 50 Hz),
z are respectively:
• 0.0028
•
and 0.0041 for PCIM-01 inertial measurement unit (sampling frequency of 1000 Hz).
0.0019 and 0.0052 for I-31T measurement system (sampling frequency of 50 Hz),
• 0.0028 and
Stochastic 0.0041 for PCIM-01
characteristics inertialwere
of sensors measurement unit (sampling
also investigated frequency
using Allanofvariance
1000 Hz). [17] and
generalized methodcharacteristics
Stochastic of wavelet moments
of sensors(GMWM)
were also [12,26]. Figure
investigated using13 Allan
presents empirical
variance wavelet
[17] and
generalized method of wavelet 6
moments (GMWM) [12,26]. Figure 13 presents
variance (WV) of measured data (10 observations per parameter). The GMWM results allowed empirical wavelet
variance
estimating (WV) of measured
parameters datanoise
of process (106 observations
which wereper parameter).
compared withThewavelet
GMWMvariance
results allowed
values for
estimating parameters of process noise which were compared with wavelet variance values for
optimized sensor models and real sensors. Plots presented in Figure 13 allow us to understand what
optimized sensor models and real sensors. Plots presented in Figure 13 allow us to understand what
kind of processes are contributing to the overall error model. Error model of accelerometer X installed
kind of processes are contributing to the overall error model. Error model of accelerometer X
on MP-02 can be specified using a combination of first-order autoregressive process (AR1) and random
installed on MP-02 can be specified using a combination of first-order autoregressive process (AR1)
walk and
process (RW).
random walkAccelerometer X installed onboard
process (RW). Accelerometer theonboard
X installed I-31T is
thecharacterized by an additional
I-31T is characterized by an
second-order
additionalautoregressive process.
second-order autoregressive process.

Figure 13. Wavelet

Figure variance
13. Wavelet calculated
variance calculatedfor 1066 observations
for10 observations ofofaa : measurement
x:xmeasurement system
system installed
installed on on
the board of I-31T
the board (A),(A),
of I-31T PCIM-01
PCIM-01inertial
inertialmeasurement unit (B).
measurement unit (B).

3.3. Analysis of Sample

3.3. Analysis Data
of Sample Data
By analysing
By analysing thethe dataset
data setrecorded
recorded bybythetheD700 system
D700 (Figure
system 14), it14),
(Figure is possible to identify
it is possible the
to identify
touchdown time on the a plot (0 s of relative time). The changes in vertical velocity VS (Figure
the touchdown time on the az plot (0 s of relative time). The changes in vertical velocity VS (Figure 14C) 14C)
play play a secondary
a secondary rolerole in this
in this analysisdue
analysis due toto the
the delays
delays resulting
resultingfrom
from thethe
design of the
design of aircraft
the aircraft
variometer (the process of pressure equalisation by capillary tube in a classic mechanical variometer
variometer (the process of pressure equalisation by capillary tube in a classic mechanical variometer or
algorithms using integrated pressure and inertial measurements in electronic variometers). Information
about barometric altitude, GPS altitude (Figure 14A), Euler angles (Figure 14D) as well as TAS and
IAS speeds (Figure 14B) are of secondary importance in the process of identifying the touchdown
moment. The engine revolutions per minute, also shown in Figure 14E, can be helpful in the analysis
of az acceleration. It enables distinguishing of structural vibrations caused by the engine from the ones
coming from external forces (e.g., ground reaction, atmospheric gusts). A rough comparison of engine
Sensors 2020, 20, 7231 13 of 25

revs recorded by the D700 with the results of continuous wavelet analysis of ax and az accelerations
from the PCIM-01 device (Figure 15) highlights potential possibilities in using correlated data sets from
these two systems.
Sensors 2020, 20, x FOR PEER REVIEW 14 of 26

Figure
Figure 14. Flight
14. Flight parameters
parameters recorded
recorded during
during MP-02
MP-02 landing
landing No.No. 4 (12
4 (12 JulyJuly 2017,
2017, paved
paved runway);
runway); recorder
recorder D700, fs = 16 Hz; Hbaro and HGPS (A), IAS and TAS (B), VS (C), Pitch and Roll (D), RPM (E), az (F).
D700, fs = 16 Hz; Hbaro and HGPS (A), IAS and TAS (B), VS (C), Pitch and Roll (D), RPM (E), az (F).
Sensors 2020, 20, 7231 14 of 25
Sensors 2020, 20, x FOR PEER REVIEW 15 of 26

Figure
Figure 15.15.Flight
Flightparameters
parametersGS GS(A),
(A),aax x(B)
(B)and
and aazz (D)
(D) recorded
recorded during
during MP-02
MP-02 landing
landingNo.No.44(12
(12July
July
2017, paved runway) and continuous wavelet transform of the acceleration values a
2017, paved runway) and continuous wavelet transform of the acceleration values ax (C) and az (E); x (C) and az (E);

PCDL-01
PCDL-01 recorder; s = 1 kHz.
recorder;fs f= 1 kHz.

The accelerations spectrum subjected to wavelet analysis has a lower limit of 7 Hz (Figure 15C,E).
Below this value, there are no significant phenomena in the touchdown detection process,
and the slow-changing phenomena associated with aircraft motion are not intense. The moment
of a touchdown on the concrete runway is visible on both time series (Figure 15B,D) as well as
on charts showing the modified wavelet coefficients of transformations of ax (Figure 15C) and az
Sensors 2020, 20, 7231 15 of 25

(Figure 15E). This event corresponds to the value of ax amplitude exceeding 0.2 (interpreting CWT
results), with the maximum frequencies in the range of 15–30 Hz. A detailed analysis of ax acceleration
time series in the vicinity of the touchdown point (−1 s to 3 s relative time, Figures 16 and 17) show
that ax oscillations reach an amplitude close to 0.3, while the az amplitude exceeds 0.5 in the 0–0.4 s
period. Vibrations with frequencies exceeding 30 Hz are associated with the engine. The CWT
plot of ax demonstrates the ground reaction with a mid-frequency of about 10–15 Hz from 1st to
10th second of relative time (Figure 15C). Beyond this point, the aeroplane speed decreased during
the roll-out and the recorded vibrations transmitted through the fuselage to the measurement system
were extinguished.
In order to compare the data from a classic flight recorder operating at 16 Hz (D700) and a
high-frequency recorder, the data from both devices in the vicinity of the touchdown point were
presented (Figures 16 and 17, respectively). The observed changes of aircraft height and speed in both
cases do not even allow determining of the approximate touchdown point. The plot of the vertical
speed on the standard avionics equipment suggests that the touchdown could occur just before or
after 1 s or more (VS parameter oscillates around 0). Interpretation of changes in a pitch angle brought
similar conclusions. Both signals are not reliable enough to predict the touchdown moment.
The analysis of the vertical acceleration component az allows identifying of the touchdown with
an accuracy of 0.2 s (referring to the time series shown in Figure 16F) and with an accuracy exceeding
0.05 s (Figure 17D). In this particular case, both signals ax and az carry sufficient information for
unambiguous detection of the event. However, the observed phenomenon in the case of ax is much
more intense. Both components, ax and az , were recorded at the frequency of 1 kHz with a similar
resolution in the time domain.
Figures 18–21 present data recorded during two landings followed by immediate take-offs
(so-called touch and goes). Figures 18 and 19 refer to a landing on a grass runway. Figure 18 shows
data 50 s ahead and 50 s after the touchdown point. The CWT allowed isolating vibrations coming
from the propulsion unit and ground reactions. In this particular case, ground reaction, associated
with the touchdown of the main gear, is strongly manifested. Therefore, touchdown moment is easy to
identify at time series as well as the results of ax and az wavelet analysis.
Figure 19 shows a narrow segment of data limited to 4 s (starting at 1 s ahead of touchdown).
They can be used to distinguish the touchdown of the main and front gear respectively. CWT analysis
of ax brings the most accurate result.
A similar analysis was also carried out for the touch and go on the concrete runway. The diagrams
presented in Figure 20 indicate the different nature of vibrations coming from the ground reactions.
In particular, the z-axis vibration amplitude is much smaller. However, the touchdown moment is
still clearly visible. It is possible to distinguish the touchdown of the main and front gear (Figure 21).
In this case, CWT analysis of the ax variable gives the best results. For variable az it is not possible to
capture the touchdown of the front undercarriage. This finding shows the potential for the touchdown
detection of every undercarriage leg separately (formulation of specific procedures and criteria goes
beyond this study). However, the authors included it in future research.
In addition to qualitative studies that prove the concept, quantitative detection algorithms enable
designing post-processing software or onboard hardware. The authors proposed three solutions of
virtual signals, which were tested using data from 22 flights of the MP-02 aeroplane.
The threshold values were determined for MP-02 (for other types this value must be calculated
individually). Figure 22 presents the peak values of VSig1–3 obtained in the analysis of 22 landings.
Table 1 contains information about the detection delay for each of the signals as well as the maximum
values of CWT coefficients occurring at the time of the touchdown along with the corresponding
frequency scale. One landing only (number 5) was not detectable using all three algorithms. However,
analysis of video recording showed that during the touch-and-go only symbolic contact of the left main
landing gear tyre with the runway surface occurred. In this case, the touchdown was rudimentary and
was not detected (because that is how the thresholds were chosen).
Sensors 2020, 20, 7231 16 of 25
Sensors 2020, 20, x FOR PEER REVIEW 16 of 26

Figure16.
Figure 16.Touch
Touchand
andgogo MP-02
MP-02 landing
landing No.No. 4 (12
4 (12 JulyJuly 2017,
2017, paved
paved runway);
runway); recorder
recorder fs = 16
D700,D700, fs Hz;
= 16
Hz; H and H (A), IAS and TAS (B), VS (C), Pitch and Roll (D), RPM (E),
Hbaro and HGPS (A), IAS and TAS (B), VS (C), Pitch and Roll (D), RPM (E), az (F).
baro GPS az (F).
Sensors 2020, 20, 7231 17 of 25
Sensors 2020, 20, x FOR PEER REVIEW 17 of 26

Figure 17.17.Flight
Figure Flightparameters
parametersGS
GS(A),
(A),aaxx (B)
(B) and azz (D)
(D) and
andCWT
CWTofofaax x(C)
(C)and
andaza(E)
z (E) (MP-02
(MP-02 landing
landing
No.4, 1212
No.4, July
July 2017,
2017,paved
pavedrunway).
runway).

Figures 18–21 present data recorded during two landings followed by immediate take-offs
(so-called touch and goes). Figures 18 and 19 refer to a landing on a grass runway. Figure 18 shows
data 50 s ahead and 50 s after the touchdown point. The CWT allowed isolating vibrations coming
from the propulsion unit and ground reactions. In this particular case, ground reaction, associated
with the touchdown of the main gear, is strongly manifested. Therefore, touchdown moment is easy
to identify at time series as well as the results of a and a wavelet analysis.
 Sensors 2020, 20, x FOR PEER REVIEW 18 of 26

Figure 19 shows a narrow segment of data limited to 4 s (starting at 1 s ahead of touchdown).

They
Sensors can20,be
2020, used to distinguish the touchdown of the main and front gear respectively. CWT
7231 18 of 25
analysis of a brings the most accurate result.

Figure
Figure 18.Flight
18. Flight parameters
parametersGS GS(A), ax a
(A), (B) andand
x (B) az (D)
az recorded duringduring
(D) recorded MP-02 MP-02
landing landing
No. 3 andNo.
take-off
3 and
take-off (12 July 2017, unpaved runway 26 EPRJ) and continuous wavelet transform of the ax a(C)
(12 July 2017, unpaved runway 26 EPRJ) and continuous wavelet transform of the ax (C) and z (E);
and
az PCDL-01 recorder;
(E); PCDL-01 fs = 1 kHz.
recorder; fs = 1 kHz.
Sensors 2020, 20, 7231 19 of 25
Sensors 2020, 20, x FOR PEER REVIEW 19 of 26

Figure
Figure 19.19.Flight
Flightparameters
parametersGS GS(A),
(A),aaxx (B)
(B) and
and aazz (D)
(D) recorded during MP-02
recorded during MP-02 landing
landingNo.No.3 3andand
take-off
take-off (12(12 July
July 2017,
2017, unpavedrunway
unpaved runway26 26EPRJ)
EPRJ)and
andcontinuous
continuous wavelet
wavelet transform
transformofofthe
theaax x(C)
(C)and
and
z (E);
az a(E); recorder;fsfs== 11 kHz.
PCDL-01recorder;
PCDL-01 kHz.
Sensors 2020, 20, 7231 20 of 25
Sensors 2020, 20, x FOR PEER REVIEW 20 of 26

Figure
Figure20.20.Flight
Flightparameters
parameters GS GS (A),
(A), ax (B)
(B) and
and aazz (D)
(D) recorded
recordedduring
duringMP-02
MP-02landing
landingNo. No.2 and
2 and
take-off
take-off(12
(12July
July2017,
2017,paved
paved runway)
runway) and
and continuous wavelettransform
continuous wavelet transformofofthe
theaxax(C)
(C)and
andaza(E);
z (E);
PCDL-01 recorder;fsfs== 1 kHz.
PCDL-01recorder;
Sensors 2020, 20, 7231 21 of 25
Sensors 2020, 20, x FOR PEER REVIEW 21 of 26

Sensors 2020, 20, x FOR PEER REVIEW 22 of 26

not possible to capture the touchdown of the front undercarriage. This finding shows the potential
for the touchdown detection of every undercarriage leg separately (formulation of specific
procedures and criteria goes beyond this study). However, the authors included it in future research.
In addition to qualitative studies that prove the concept, quantitative detection algorithms
enable designing post-processing software or onboard hardware. The authors proposed three
solutions of virtual signals, which were tested using data from 22 flights of the MP-02 aeroplane.
The threshold values were determined for MP-02 (for other types this value must be calculated
individually). Figure 22 presents the peak values of VSig1–3 obtained in the analysis of 22 landings.
Table 1 contains information about the detection delay for each of the signals as well as the
maximum values of CWT coefficients occurring at the time of the touchdown along with the
corresponding frequency scale. One landing only (number 5) was not detectable using all three
algorithms.
FigureFigure However,
21. Flight
21. Flight analysis
parameters
parameters GSGSof video
(A),
(A), axax(B)recording
(B) and a
and azz (D)showed
(D) recordedthat
recorded during
during
during MP-02 the touch-and-go
landing
MP-02 landingNo. No. only
2 and2 and
symbolic
take-off (12 July 2017, paved runway 26 EPRJ) and continuous wavelet transform of the ax (C) and athe
contact
take-off (12 of
July the left
2017, main
paved landing
runway 26 gear
EPRJ) tyre
and with the
continuous runway
waveletsurface occurred.
transform of the a x In
(C)this
and case,
az
z
touchdown was
(E); PCDL-01
(E); PCDL-01 rudimentary
f =fs 1=and
recorder;
recorder; was not detected (because that is how the thresholds were chosen).
1 kHz.
kHz.
s

A similar analysis was also carried out for the touch and go on the concrete runway. The
diagrams presented in Figure 20 indicate the different nature of vibrations coming from the ground
reactions. In particular, the 𝑧-axis vibration amplitude is much smaller. However, the touchdown
moment is still clearly visible. It is possible to distinguish the touchdown of the main and front gear
(Figure 21). In this case, CWT analysis of the a variable gives the best results. For variable a it is

Figure Peak
22.22.
Figure values
Peak valuesofofVSig
VSig1–3 determinedfor
1–3 determined for2222 MP-02
MP-02 flights.
flights.

Figure 23 presents flight speed (GS) and altitude (H ) time histories recorded during four
circuits performed in a sequence during one flight. They were compared with the values returned by
the VSig1-3. The chart (B) shows that for three landings the Vsig1-3 outputs exceeded the established
thresholds in the vicinity of the touchdown. For the case of an aborted touch-and-go at about 400 s,
these values are slightly lower than the threshold value. This is the case when the left main landing
Sensors 2020, 20, 7231 22 of 25

Figure 22. Peak values of VSig1–3 determined for 22 MP-02 flights.

Figure 23 presents flight speed (GS) and altitude (HGPS ) time histories recorded during four
Figure 23 presents
circuits performed flight speed
in a sequence (GS)one
during andflight.
altitudeThey
(H were
) timecompared
histories recorded
with the during
valuesfourreturned by
circuits performed in a sequence during one flight. They were compared with the values returned by
the VSig1-3 . The chart (B) shows that for three landings the Vsig1-3 outputs exceeded the established
the VSig1-3. The chart (B) shows that for three landings the Vsig1-3 outputs exceeded the established
thresholds in the vicinity of the touchdown. For the case of an aborted touch-and-go at about 400 s,
thresholds in the vicinity of the touchdown. For the case of an aborted touch-and-go at about 400 s,
these these
values are are
values slightly lower
slightly lowerthan thethreshold
than the threshold value.
value. This This
is theis thewhen
case casethewhen the left
left main main landing
landing
gear tyre
gear only
tyre touched the ground.
only touched In addition
the ground. In additionto to
thethe
landings,
landings,oneonemore
moreevent
event exceeded
exceeded thethe threshold
threshold value—take-off
value—take-off run. However, run.the
However,
developedthe developed
algorithms algorithms can automatically
can automatically exclude
exclude thisthis
event from a
event from a set of landings. Additional analysis of a changes in the vicinity of the
set of landings. Additional analysis of ax changes in the vicinity of the touchdown point is required in touchdown
point is required in order to obtain reliable automatic interpretation. The a decreases after the
order to obtain reliable automatic interpretation. The ax decreases after the touchdown and during
touchdown and during the roll-out, while after lift-off and during the climb its value increases. It is
the roll-out, while in
also beneficial after
thislift-off
case to and
referduring
to otherthe climb
flight its value
parameters, increases.
such as verticalItspeed,
is alsoground
beneficial
speedin this case
to refer
andtopitch
other flight parameters, such as vertical speed, ground speed and pitch rate.
rate.

Figure 23. Automatic touchdown detection in four circuit patterns; (A) flight parameters and (B) virtual
signals, detection thresholds and detected peaks.

4. Conclusions
The presented methodology is based on the analysis of the flight test campaign results.
Several different landing cases gave significant sets of data. Although most of the landings recorded
resembled typical load distribution patterns, as described in the requirements, there were many
which did not bring unambiguous information about the landing process. During the tests, it turned
out that the situation occurs when the signal from the accelerometer installed in the vertical axis of
the aircraft (z-axis) is not a sufficient source of information to detect the touchdown of a light aircraft.
The results were in line with both expectations and previous practical experiences. This situation
may occur if the touchdown is very smooth and delicate, without an unmistakable ascent of the main
landing gear impact on the runway surface. The algorithm of the virtual touchdown sensor, using data
from a three-axis accelerometer, mandatorily takes into account the analysis of vibration frequencies
in the longitudinal axis of the aircraft. Recording of the specific level of vibration amplitude in
the predefined frequency range may define a prerequisite for touchdown flagging. Continuous wavelet
transformation proved to be helpful in the process of analysing airframe vibrations. It allows precise
locating of oscillatory phenomena in the time domain. In the CWT analysis presented in this paper,
the authors decided to use the complex Morlet wavelet, which is very well suited for the analysis of
impact phenomena. With information about the aeroplane speed as well as the altitude above ground
level (AGL), one can impose additional conditions for the virtual touchdown sensor, thus eliminating
any possible false alarms coming from in-flight phases other than landing.
The analyses used data recorded during eighteen landings of the I-31T turboprop light aeroplane,
twenty-two of the MP-02 ultralight aeroplane, as well as during eighteen experimental flights of
the PW-6U glider (archival data from the Advanced In-flight Measurements—AIM2 project [27],
Sensors 2020, 20, 7231 23 of 25

not presented in this work). In each case, unambiguous touchdown detection was possible based on
the results of CWT analysis of the ax acceleration, as the primary source.
The proposed method may be particularly useful for the detection of aircraft touchdown (both in
time and space) during off-line bulk data analysis. Preliminary results indicate that on this basis, it is
also possible to develop an onboard virtual sensor for touchdown detection, although the authors are
aware of the limitations. In order to detect vibrations at specific frequencies, it is necessary to collect
an appropriate number of samples, which in turn translates into delays. In the case of the proposed
algorithms, using 256 scale levels and with a sampling frequency of 1000 Hz, the detection delay
can reach 0.25 s. Taking into account other flight parameters, it is possible to reduce the virtual
signal algorithm to 32 scale levels, which will allow a possible reduction to 0.03 s. It should also be
remembered that we analyse vibrations recorded by the measuring system attached to the structure
of the aircraft fuselage, not vibrations within the landing gear. This results in additional delays in
the virtual sensor system.
The main goal of this work was not to develop a solution that could compete with classical
measuring systems. The authors, however, intended to create an algorithm for a virtual sensor that
would enable accurate detection of the time and place of touchdown, based on the recorded typical set
of flight parameters. This goal was achieved with regard to the off-line analysis. In the future, building
a test device is planned that will be able detection of selected events, including touchdown, in a real
physical system.

Author Contributions: Conceptualization, J.B. and P.R.; methodology, J.B. and P.R.; software, P.R.; validation, J.B.
and P.R.; resources, J.B. and P.R.; data curation, P.R.; writing—original draft preparation, J.B.; writing—review
and editing, J.B. and P.R.; visualization, P.R.; supervision, J.B. and P.R. All authors have read and agreed to
the published version of the manuscript.
Funding: Elements of this scientific work have been financed from: project ESPOSA (Efficient Systems and
Propulsion for Small Aircraft), co-funded by the European Commission in 7th Framework Program (grant
agreement 284859), Polish scientific funds for years 2010–2012 as a development project No. O R00 0116 11 and
Polish financial sources for science in the period of 2011–2015 granted for the international co-funded project.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
Θ pitch Euler angle (deg);
Φ roll Euler angle (deg);
Ψ yaw Euler angle (deg);
λGPS GPS longitude (deg);
φGPS GPS latitude (deg);
σn standard deviation of virtual signal noise;
Ψ (f) Fourier transform of ψ(t);
ψ(t) mother wavelet;
ψab ∗ basic wavelet;
a scale factor;
ax longitudinal acceleration related to g0 ;
ay lateral acceleration related to g0 ;
az vertical acceleration related to g0 ;
A virtual signal peak
b wavelet shift;
f(t) generic function;
fs sampling time (s);
g0 standard acceleration due to gravity (9.80665 m·s−2 );
GS ground speed (m·s−1 );
HGPS GPS altitude (m);
IAS indicated airspeed (m·s−1 );
p roll angular velocity (deg s−1 );
Sensors 2020, 20, 7231 24 of 25

ps static atmospheric pressure (Pa);

q pitch angular velocity (deg s−1 );
r yaw angular velocity (deg s−1 );
SNR signal to noise ratio;
t time (s);
TAS true airspeed (m·s−1 );
TT true track (m·s−1 );
VS vertical speed (m·s−1 );

References
1. Soares, P.F.M. Flight Data Monitoring and its Application on Algorithms for Precursor Detection. Master’s
Thesis, Instituto Superior Técnico Lisbon, Lisbon, Portugal, 2014.
2. Masiulionis, T.; Stankūnas, J. Review of equipment of flight analysis and development of interactive
aeronautical chart using Google Earth’s software. Transport 2017, 33, 580–588. [CrossRef]
3. Pytka, J.; Budzyński, P.; Łyszczyk, T.; Józwik, J.; Michałowska, J.; Tofil, A.; Błażejczak, D.; Laskowski, J.
Determining Wheel Forces and Moments on Aircraft Landing Gear with a Dynamometer Sensor. Sensors 2019,
20, 227. [CrossRef] [PubMed]
4. Pytka, J.; Budzyński, P.; Józwik, J.; Michałowska, J.; Tofil, A.; Łyszczyk, T.; Błażejczak, D. Application of
GNSS/INS and an Optical Sensor for Determining Airplane Takeoff and Landing Performance on a Grassy
Airfield. Sensors 2019, 19, 5492. [CrossRef] [PubMed]
5. Pieniazek, J. Vision based airplane localization for automatic approach. In Proceedings of the 2017 IEEE
International Workshop on Metrology for AeroSpace (MetroAeroSpace), Padua, Italy, 21–23 June 2017;
pp. 424–427. [CrossRef]
6. Luong, Q.V.; Jang, D.-S.; Hwang, J. Intelligent Control Based on a Neural Network for Aircraft Landing Gear
with a Magnetorheological Damper in Different Landing Scenarios. Appl. Sci. 2020, 10, 5962. [CrossRef]
7. De Rivals, G. Landing Trajectory Computation. In Proceedings of the EOFDM 3rd Conference, Cologne,
Germany, 6 February 2014.
8. Hirsenkorn, N.; Hanke, T.; Rauch, A.; Dehlink, B.; Rasshofer, R.; Biebl, E. Virtual sensor models for real-time
applications. Adv. Radio Sci. 2016, 14, 31–37. [CrossRef]
9. Inamdar, S.; Allemang, R.J.; Phillips, A.W. Estimation of Frequency and Damping of a Rotating System Using
MEOT and Virtual Sensor Concept. Special Topics in Structural Dynamics 2016; Springer: Berlin, Germany, 2016;
Volume 6, pp. 23–35.
10. Dolega, B.; Kopecki, G.; Tomczyk, A. Possibilities of using software redundancy in low cost aeronautical
control systems. In Proceedings of the 2016 IEEE Metrology for Aerospace (MetroAeroSpace), Florence, Italy,
22–23 June 2016; pp. 33–37. [CrossRef]
11. Piwek, K.; Wiśniowski, W. Small air transport aircraft entry requirements evoked by FlightPath 2050.
Aircr. Eng. Aerosp. Technol. 2016, 88, 341–347. [CrossRef]
12. Bakunowicz, J.; Rzucidło, P. Measurement and analysis of standard flight parameters based on MEMS
sensors. J. Sens. Sens. Syst. 2017, 6, 211–222. [CrossRef]
13. Szpunar, R.; Rzucidło, P. Application of ASG-EUPOS network for on-board control system of a small
unmanned aircraft. Prz. Elektrotech. 2012, 88, 154–158.
14. Rzucidlo, P.; Kopecki, G.H.; DeGroot, K.; Kucaba-Pietal, A.; Smusz, R.; Szewczyk, M.; Szumski, M. Data
acquisition system for PW-6U in flight boundary layer mapping. Aircr. Eng. Aerosp. Technol. 2016, 88, 572–579.
[CrossRef]
15. Kopecki, G.H.; Rzucidlo, P. Integration of optical measurement methods with flight parameter measurement
systems. Meas. Sci. Technol. 2016, 27, 054003. [CrossRef]
16. Boden, F.; de Groot, K.; Meyer, J.; Meyer, R.; Bakunowicz, J.; Rzucidło, P.; Smusz, R.; Szewczyk, M.; Szumski, M.
Flow and Structure Deformation Research of a Composite Glider in Flight Conditions. In Proceedings of
the 46th Society of Flight Test Engineers International Symposium Breaking Barriers, Lancaster, CA, USA,
14−18 September 2015.
17. Adamcik, F. Error analysis of the inertial measurement unit/inercinio matavimu˛ bloko klaidu˛ analizė. Aviation
2011, 15, 5–10. [CrossRef]
Sensors 2020, 20, 7231 25 of 25

18. Goupillaud, P.; Grossmann, A.; Morlet, J. Cycle-octave and related transforms in seismic signal analysis.
Geoexploration 1984, 23, 85–102. [CrossRef]
19. Dziedziech, K.; Staszewski, W.J.; Uhl, T. Wavelet-Based Frequency Response Function: Comparative Study
of Input Excitation. Shock. Vib. 2014, 2014, 502762. [CrossRef]
20. Torrence, C.; Compo, G.P. A practical guide to wavelet analysis. Bull. Am. Meteorol. Soc. 1998, 79, 61–78.
[CrossRef]
21. Bentz, C.M.; Baudzus, L.; Krummrich, P.M. Signal to Noise Ratio (SNR) Enhancement Comparison of
Impulse-, Coding- and Novel Linear-Frequency-Chirp-Based Optical Time Domain Reflectometry (OTDR)
for Passive Optical Network (PON) Monitoring Based on Unique Combinations of Wavelength Selective
Mirrors. Photonics 2014, 1, 33–46. [CrossRef]
22. Basmadji, F.; Gruszecki, J.; Kordos, D.; Rzucidlo, P. Development of ground station for a terrain
observer-hardware in the loop simulations. In Proceedings of the AIAA Modeling and Simulation
Technologies Conference, Minneapolis, MN, USA, 13−16 August 2012.
23. Grzybowski, P.; Klimczuk, M.; Rzucidlo, P. Distributed measurement system based on CAN data bus.
Aircr. Eng. Aerosp. Technol. 2018, 90, 1249–1258. [CrossRef]
24. Stock, M. CANaerospace-Interface Specification for Airborne CAN Applications V 1.7; Stock Flight Systems:
Berg/Farchach, Germany, 2006; p. 12.
25. Rzucidlo, P. Monitoring and data acquisition system for experimental general aviation fly-by-wire aircraft.
Aviation 2006, 10, 16–20. [CrossRef]
26. Guerrier, S.; Skaloud, J.; Stebler, Y.; Victoria-Feser, M.-P. Wavelet-Variance-Based Estimation for Composite
Stochastic Processes. J. Am. Stat. Assoc. 2013, 108, 1021–1030. [CrossRef] [PubMed]
27. Boden, F.; Lawson, N.; James, W.; Tatam, R.P. Special feature on Advanced ln-flight Measurement Techniques
AIM2. Meas. Sci. Technol. 2017, 28, 1–2. [CrossRef]

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).