sensors

Article
Railway Point-Operating Machine Fault Detection
Using Unlabeled Signaling Sensor Data
Pritesh Mistry *, Phil Lane and Paul Allen
School of Computing and Engineering, University of Huddersfield, Queensgate, Huddersfield HD1 3DH, UK;
[email protected] (P.L.); [email protected] (P.A.)
* Correspondence: [email protected]




Received: 31 March 2020; Accepted: 7 May 2020; Published: 9 May 2020


Abstract: In this study, we propose a methodology for the identification of potential fault occurrences
of railway point-operating machines, using unlabeled signal sensor data. Data supplied by Network
Rail, UK, is processed using a fast Fourier transform signal processing approach, coupled with the
mean and max current levels to identify potential faults in point-operating machines. The method
developed can dynamically adapt to the behavioral characteristics of individual point-operating
machines, thereby providing bespoke condition monitoring capabilities in situ and in real time.
The work described in this paper is not unique to railway point-operating machines, rather the data
pre-processing and methodology is readily applicable to any motorized device fitted with current
sensing capabilities. The novelty of our approach is that it does not require pre-labelled data with
historical fault occurrences and therefore closely resembles problems of the real world, with application
for smart city infrastructure. Lastly, we demonstrate the problems faced with handling such data and
the capability of our methodology to dynamically adapt to diverse data presentations.

Keywords: condition monitoring; signal processing; fast Fourier transform; railway point-operating
machines; turnout; fault detection; unlabeled data; smart sensors

1. Introduction
Railway point-operating machines (POM) are electro-mechanical devices that operate turnouts,
enabling a train to be directed from one track onto another via tapering rails, known as switch blades.
POMs are located at various points along a railway line and remotely operated to divert a train to the
normal (NR), or reverse (RN) direction [1]. The NR direction allows a train to pass straight through a
turnout, while the RN direction directs the train along the alternate path.
POM failures are considered critical failures of a rail network system. Signaling equipment and
turnout failures have been accounted for 55% of all railway infrastructure component failures [2],
leading to delays, costly repairs, and potentially hazardous situations [3]. Where delays are not kept to
within targets set out by regulators, substantial fines can be imposed onto the operator [3].
A POM (Figure 1) will naturally undergo a process of degradation as operational wear takes
hold, ultimately leading to complete failure. If progression to failure can be identified, preventative
maintenance can be scheduled, avoiding operational down time and costly repairs. The currently
practiced methods for POM maintenance include replacement after a fixed time period or beyond an
accumulated number of actuation operations. However, they do not consider environmental influences
such as track load or weather exposure [3]. A consequence of this may be that faults occur much sooner
than expected or appear sporadically without explanation. More dynamic detection methods are thus
sought to help monitor such occurrences.

Sensors 2020, 20, 2692; doi:10.3390/s20092692 www.mdpi.com/journal/sensors

 Early fault detection methods relied solely on threshold settings [4], but produced limited
success with a high false positive detection rate [5]. A thresholding method is unable to foresee a
potential fault developing, unless the threshold is set unnecessarily lower than optimal. It also adopts
a one-size-fits-all
Sensors 2020, 20, 2692 approach, and is not adaptable to the individual behavioral characteristics 2 of of
13
mechanical devices [3].

Figure
Figure 1.
1. Point-operating
Point-operating machine
machine (POM)
(POM) (https://en.wikipedia.org/wiki/Railroad_switch).
(https://en.wikipedia.org/wiki/Railroad_switch).

Statistical
Early faultapproaches
detection methodshave fared better,
relied solely [2,6–10] and provide
on threshold settings the added
[4], advantage
but produced of being
limited able
success
to function
with a high in realpositive
false time, without
detection therate
need[5].forA historically
thresholding labelled
methoddata of “good”
is unable and “bad”
to foresee examples.
a potential fault
Garcia et al.,unless
developing, used the the Kalman
threshold filter
is setapproach
unnecessarilyand compared
lower thannew data to
optimal. a reference
It also adopts apoint to identify
one-size-fits-all
faults. Their
approach, and relatively straight to
is not adaptable forward approach
the individual was ablecharacteristics
behavioral to detect the of majority
mechanicalof faults
devicesas they
[3].
occurred [2]. Statistical
Statistical approaches feature
have extraction of signal
fared better, data and
[2,6–10] followed
provide by clustering
the addedtechniques
advantagehave been
of being
successfully
able to function demonstrated
in real time, as without
a usefulthe faultneeddetection approach.
for historically A study
labelled datathat
of extracted
“good” and over ten
“bad”
statistical features
examples. Garcia etfrom theirthe
al., used signaling
Kalmandata, filter discovered
approach and thatcompared
many of new the faults
data toinvestigated
a reference point were
identifiable
to identify faults.by only tworelatively
Their features; the maximum
straight forward force and thewas
approach rootable
mean to square
detect the(RMS) [7]. Evidently
majority of faults
there
as theyare clear advantages
occurred [2]. Statistical for feature
using extraction
a statisticalof approach,
signal datanotably
followed they are well understood
by clustering techniques
endpoints,
have with relatively
been successfully fast computations
demonstrated as a usefuland therefore able toapproach.
fault detection highlight A attributes
study that of the signal
extracted
data that
over are not otherwise
ten statistical features fromvisible.their signaling data, discovered that many of the faults investigated
wereRecently
identifiable reported
by onlywork twohasfeatures;
covered the the vast field offorce
maximum machine and learning
the root techniques,
mean square with labelled
(RMS) [7].
data to train
Evidently there intelligent
are clearmodels
advantages to predict
for usingfuture failures.approach,
a statistical Although notablyhighly successful,
they are well theunderstood
downside
of such models
endpoints, withisrelatively
the requirement of labelled and
fast computations datatherefore
to train aablemodel, which is not
to highlight readilyofavailable
attributes the signal in
a real
data thattime
are not setting. Nevertheless,
otherwise visible. we shall briefly discuss some machine learning approaches
discovered
Recently from the literature
reported work has used in condition
covered the vastmonitoring
field of machineapplications.
learning techniques, with labelled
data For POMs
to train producing
intelligent modelssignalto data,
predict the predominant
future condition highly
failures. Although monitoring approach
successful, seems to be
the downside of
that of
such support
models is thevector machines
requirement (SVM).data
of labelled SVM to models are margin-based
train a model, discriminant
which is not readily available classifiers,
in a real
shown
time to perform
setting. well with
Nevertheless, wesuchshalldata.
brieflyFordiscuss
example, some using discrete
machine wavelet
learning transformdiscovered
approaches to extract
features
from the at different
literature levels
used of decomposition,
in condition monitoring SVM classifiers built with these features were highly
applications.
accurate in classifying
For POMs producing faults,
signalincluded
data, the being able to classify
predominant condition the monitoring
severity of approach
fault occurrenceseems to [11].
be
Vileiniskis
that of support et al., used
vector features(SVM).
machines extracted
SVMfrom models non-faulty POM actuations
are margin-based discriminant to develop
classifiers, decision
shown
boundaries
to perform well of a one-class
with suchSVM. data. They then tested
For example, their
using modelwavelet
discrete with unseen data to
transform tosee if thefeatures
extract actuation at
signals were abnormal, which fell outside of this boundary, and would
different levels of decomposition, SVM classifiers built with these features were highly accurate in be flagged as failures [3]. Bian
et al., combined
classifying faults, self-organizing
included being able maps to (SOM)
classify with SVM toofclassify
the severity degradation
fault occurrence [11]. states of POMs,
Vileiniskis et al.,
achieving
used features accuracies
extractedinfrom the non-faulty
region of POM 96 %actuations
[9]. Usingtoprincipal component
develop decision analysisof for
boundaries feature
a one-class
reduction,
SVM. TheyEker thenettested
al., developed
their model a SVMwithfault
unseendetection
data tomodel
see if that achieved 100
the actuation % accuracies
signals were abnormal, [10]. It
is worthwhile
which fell outside to note thatboundary,
of this studies reporting
and would highbeaccuracies
flagged aswere often[3].
failures foundBiantoetbe al.,using
combinedsmall
datasets or simulated
self-organizing maps (SOM)data towithbuild SVM models. Although
to classify this is common
degradation practice
states of POMs, in the domain,
achieving accuracies thesein
models
the region may of perform
96 % [9]. comparatively
Using principalwhen presented
component withfor
analysis larger volumes
feature of real-world
reduction, Eker et al., examples.
developed
a SVM fault detection model that achieved 100 % accuracies [10]. It is worthwhile to note that studies
reporting high accuracies were often found to be using small datasets or simulated data to build
models. Although this is common practice in the domain, these models may perform comparatively
when presented with larger volumes of real-world examples.
Sensors 2020, 20, 2692 3 of 13

Shape analysis is also a common approach to condition monitoring problems, and more relevant
to this study than the use of classifiers alone. Shape analysis primarily focuses on understanding the
signal shape profiles of normal and abnormal actuation events, attempting to identify subtle differences
caused by the aging of an instrument. These differences can then be modelled for fault identification.
Sa et al., used a shapelet algorithm which extracts a subsequence of the time series called a shapelet.
The shapelet attempts to minimize the in-class distance while maximizing the between-class distance of
examples in the dataset. They used these extracted subsequences with a decision tree model, producing
accuracies of 97 % [1]. In a similar study, Garcia et al., compared the expected shapes of signals and
used a vector autoregressive moving-average model (VARMA) to forecast the outcome of new signals.
If the new signal resulted in a signal beyond the expectation of the model a fault was considered to
have been detected [12].
In this study, we use real-world unlabeled data collected from POMs to identify potential fault
occurrences using a fast Fourier transform and curve fitting approach. It builds on our previously
reported work in this area [13]. We corroborate our methodology against more traditional thresholding
approaches to evaluate the effectiveness of our proposed method.
The remainder of this article is arranged as follows: Section 2 details the data used in this study
and the pre-processing steps leading to actuation extraction. Section 3 presents the methodology
of this work, more specifically, feature extract from actuation signal data, fast Fourier transform,
and curve fitting, which are all integral parts of the proposed methodology. The results of this work
with interpretation of findings are detailed in Section 4. Lastly the work is concluded in Section 5.

2. Signal Sensor Data

In this study, we use electrical current draw data collected by sensors attached to POMs across
three regions of the United Kingdom (UK); London North-Eastern (LNE), London North-Western
(LNWN) and Sussex. Data have been collected and supplied by Network Rail, UK, and contains the
records of 331 POMs, whose actuations have been recorded over a period from April 2018 to June 2018.
The data for each region consists of several thousand csv (comma-separated values) files, each file
containing a daily log of all the actuation operations performed by any particular POM in a given
direction (NR or RN). Each file contains the current draw in amperes (amps or A) against a time stamp
recorded in milliseconds (ms) as a continuous sequential log for any given day. Depending on the
physical location of a particular POM, an instrument on a busy railway line will naturally contain
many more actuations per day than a POM located on an infrequently used railway line. The daily log
files which make up the dataset are therefore populated with a varying number of actuations according
to frequency of POM use.
Table 1 shows the number of files and POMs in each of the three regions initially available from
the dataset. The data are considered to be unlabeled as no indication of fault occurrences within the
logged events or otherwise are available.

Table 1. Files present in the dataset for each region.

Region No. of Files No. of POMs

LNE 15,802 103
LNWN 17,081 129
Sussex 12,807 99

An example of a subset of data extracted for point-operating machine POM-1 moving in the NR
direction is shown in Table 2. These data were recorded on the 1st June 2018, with the initial data entry
beginning at 02:28:05.527 in the morning. The table shows the first five records (Row0-Row4) and
the last five records (Row242-Row246) of the first actuation. The first five records of the subsequent
actuation are also shown (Row247-Row251). This example illustrates the continuous sequential way the
data are recorded. The frequency of which the data are recorded during an actuation is approximately
Sensors 2020, 20, 2692 4 of 13

10 ms. Once an actuation has executed, data recording stops until the next operation of the POM is
actioned. Since the frequency of data records in any given actuation is approximately 10 ms, we can
distinguish between two sequential actuations by noting the time difference between each record.
Row246 and Row247 of Table 2 represent the end of one actuation (Row246) and the beginning of the
next actuation (Row247) for point-operating machine POM-1 in the NR direction. The time stamp for
these records are recorded as 2018-06-01 02:28:07.977 (Row246) and 2018-06-01 05:30:57.793 (Row247).
This time difference equates to many minutes and indicates that those two records cannot belong to the
same actuation event. Lastly, we can see from this data that for any given actuation, the current draw
recording begins at a value of 0.00 A (zero amps) and ends with a recorded value of 0.00 A. This is seen
for the records at Row0 (start of the first actuation) and Row246 (end of the first actuation), then again
at Row247 (start of second actuation).

Table 2. Current (A) draw signal data for point-operating machine POM-1 moving in the NR direction.

Row No. Date Time POM-1 NR Current

Row0 2018-06-01 02:28:05.527 0.00
Row1 2018-06-01 02:28:05.537 13.18
Row2 2018-06-01 02:28:05.547 19.43
Row3 2018-06-01 02:28:05.557 18.04
Row4 2018-06-01 02:28:05.567 16.15
.. .. ..
. . .
Row242 2018-06-01 02:28:07.937 0.82
Row243 2018-06-01 02:28:07.947 0.81
Row244 2018-06-01 02:28:07.957 0.80
Row245 2018-06-01 02:28:07.967 0.76
Row246 2018-06-01 02:28:07.977 0.00
Row247 2018-06-01 05:30:57.793 0.00
Row248 2018-06-01 05:30:57.803 0.81
Row249 2018-06-01 05:30:57.813 16.22
Row250 2018-06-01 05:30:57.823 19.37
Row251 2018-06-01 05:30:57.833 17.55
.. .. ..
. . .

Table 2 represents a typical example of the data files contained within the larger dataset available
for this study. The frequency of data records is 10 ms, although for some files, individual records that
exceeded a 10 ms frequency is observed.

2.1. Pre-Processing
Where real-world data is concerned, it is seldom free of error. As such, a significant element of
pre-processing was required. For this study, all our data pre-processing and analysis was performed
using Knime analytics platform [14] and R Statistical package [15]. Where sensors are concerned,
missing records are common errors due to sensor malfunctions and other equipment failures. To begin,
all 45,690 files were checked for errors. Files that contained missing values or erroneous values which
resulted in logged data not akin to that shown in Table 2 were removed from the dataset. Most of the
errors identified involved data files which were entirely absent of any recorded data.
Although the csv files in our dataset were labelled with the operating point machine name and
date of data collection, we discovered several csv files in different folders with identical file names.
Unable to ascertain the validity of these files, all files with duplicate file names were also removed
from the dataset.
Sensors 2020, 20, 2692 5 of 13

For sensor data of this nature or similar, many different types of errors can present themselves in
the final dataset. As such, the data cleansing process is often unique to the task at hand. Since our
dataset was large, removal of erroneous files was not an issue. For smaller datasets, an appropriate
method for handling erroneous data would be more appropriate. The initial pre-processing step
described, reduced the dataset from 45,690 files to 23,389 files, as shown in Table 3.

Table 3. Remaining files after removal or erroneous files.

Region No. of Files No. of POMs

LNE 6533 97
LNWN 9715 105
Sussex 7141 91

2.2. Actuation Extraction

Once the data was cleansed of errors, the remaining 23,389 files were further processed.
The methodology proposed in this paper requires each individual actuation to be processed separately.
This is required so that features can be extracted from each actuation. In this study, the direction of
actuation, either NR or RN, were not treated separately. Rather the actuation for a given machine
in either direction was simply treated as a sequential movement operation performed by the POM.
For this initial study, the assumption was made that the occurrence of a fault would be apparent
regardless of the direction of movement.
For our dataset, the frequency of current draw measurements was typically 10 ms between records
as seen in Table 2. On occasion a value greater than 10 ms, generally between 20-100, is observed.
Given these observations, it was decided that a timestamp difference of greater than 500 ms would be
used as the threshold value for distinguishing between adjacent actuations in the sequential data records.
As an example, records in Row0-Row246 of Table 2 all show a time difference of 10 ms between
subsequent records. However, the time difference between Row246 (2018-06-01 02:28:07.977) and
Row247 (2018-06-01 05:30:57.793) of Table 2 is greater than 500 ms. This difference of greater than 500 ms
indicates the end of one actuation and the start of the subsequent actuation. It is also worthy of mention
that all actuation events begin with a current reading of 0.00 A (zero amps) and end with a current
reading of 0.00 A. This can be seen in Table 2 for records Row0 (start of first actuation) and Row246
(end of first actuation), and once again for Row247, which is the start of the subsequent actuation.
A typical current draw signal of a single actuation event for point-operating machine POM-1 is
shown in Figure 2. This event was recorded on the 01st June 2018 with the instrument moving in the
NR direction. Some of the data points for this actuation are shown in Table 2. In this paper, we describe
two phases of an actuation event—the push-out phase and the swing phase. These phases are used
in our proposed methodology to extract features from the actuation event. In other published work,
Vileiniskis et al. [3], further defines the push-out phase to consist of two stages—the motor start-up
and the unlocking stages. The swing phase also consists of two stages—the switch rail movement and
the locking stages. Other studies also describe the actuation signal similarly [5,9,16]. For the purpose
of this study, we will keep with the two phases shown in Figure 2 namely the push-out phase and the
swing phase.
 Sensors 2020, 20, x FOR PEER REVIEW 6 of 14

2. Extraction of the mean current draw and time duration of the swing phase.

Sensors 2020, 3. x FOR

FastPEER
Fourier Transformanalysis of the time domain signal into frequency domain
Sensors 2020, 20,20,
2692 REVIEW 6 of
6 of 1314
representations, followed by goodness of fit analysis of a linear fitted cure.
2. Extraction of the mean current draw and time duration of the swing phase.

3. Fast Fourier Transform analysis of the time domain signal into frequency domain
representations, followed by goodness of fit analysis of a linear fitted cure.

Figure 2. Current draw signal of a single actuation event for point-operating machine POM-1, moving
Figure 2. Current draw signal of a single actuation event for point-operating machine POM-1, moving
in the NR direction, recorded on the 01st June 2018. The push-out phase (red) and swing phase (blue)
in the NR direction, recorded on the 01st June 2018. The push-out phase (red) and swing phase (blue)
of the actuation event are shown.
of the actuation event are shown.
3. Method
To isolate
Figure the push-out
2. Current phase fromactuation
the swing phase for any given actuation, a fixed time
After isolation ofdraw signal
actuation of a single
signals for each event
POM, for point-operating
as described above, machine POM-1,
the faultmoving
detection
duration could be used as a cut-off point to separate the two phases.
in the NR direction, recorded on the 01st June 2018. The push-out phase (red) and swing phase (blue) However, the current draw
methodology proposed in this study involves the extraction of three principal features from these
signalofproduced by POMs
the actuation event are shown. are not only unique between different instruments but also change
signals. These features are then used to develop our fault detection capabilities. The three features
significantly for the same POM over time. Figure 3 shows the current draw signal of two different
extracted
POMsTomoving are as follows:
isolateinthe thepush-out
NR direction. phaseThe from signals produced
the swing phase arefor visibly different
any given in several
actuation, ways,time
a fixed for
example,
1.duration
Extractionthe maximum
couldofbe theused current
maximumas a cut-off value
current (peak
point current)
drawto(peakseparate is noticeably
the and
current) two time different
phases. for
However,
duration the two
of thethe instruments.
current
push-out draw
phase.
The
2.signal signal produced
produced by by the
POMs twoare POMs
not after
only the
unique
Extraction of the mean current draw and time duration of the swing phase.1000 ms
between point are also
different noticeably
instruments different.
but also change
For any single theactuation event, thetime.
current draw signal theis represented by two arrays, I and T, of
3.significantly
Fast Fourier for Transform
same POM analysis over of the timeFigure
domain3 shows
signal intocurrent
frequency draw signal
domain of two different
representations,
equal
POMs size,
movingn. Where theI =NR[i1…i n] and T The = [t1…tn], I represents the
are current draw array, while Tways,
denotes
followed byingoodness direction.
of fit analysis ofsignals a linear produced
fitted cure. visibly different in several for
the time array.
example, The start current
the maximum and endvalue points of the
(peak push-out
current) phase signal
is noticeably can therefore
different for the two be instruments.
denoted by
(tThe To
pstart , iisolate
signal the(tpush-out
pstart) produced
and pend,by
ipend )phase
the two POMs from the
respectively, swing
and
after phase
similarly
the 1000 ms for
the any
start
point given
and
are actuation,
end
also pointsa fixed
noticeably thetime
ofdifferent. duration
swing phase
could
signalbe For used
can be as
any a cut-off
denoted
single bypoint
(tsstartto
actuation separate
, isstart
event, ) and the
(tsendtwo
the current phases.
, isend However,
) respectively.
draw signal the current
is represented bydraw signal produced
two arrays, I and T, of
byequal
POMs size,aren.not onlyI unique
Where = [i1…in]betweenand T = different
[t1…tn], I instruments
represents the but also change
current significantly
draw array, while T for the
denotes
same POM over time. Figure 3 shows the current draw signal
the time array. The start and end points of the push-out phase signal can therefore be denoted of two different POMs moving in theby
NR direction.
(tpstart The(tsignals
, ipstart) and produced
pend, ipend ) respectively, are visibly different in
and similarly theseveral
start andways, end forpoints
example, theswing
of the maximumphase
current
signal value can be(peakdenotedcurrent) is noticeably
by (tsstart , isstart) and different
(tsend, isendfor the two instruments. The signal produced by
) respectively.
the two POMs after the 1000 ms point are also noticeably different.

Figure 3. Current draw signal of actuation events for point-operating machines POM-2 (peak current
≈ 20A) and POM-3 (peak current ≈ 15A) in the NR direction.

Figure
Figure3.3.Current
Currentdraw signal
draw of actuation
signal events
of actuation for point-operating
events machines
for point-operating POM-2 (peak
machines POM-2 (peak≈current
current 20A)
and POM-3 (peak current ≈ 15A) in the NR direction.
≈ 20A) and POM-3 (peak current ≈ 15A) in the NR direction.
Sensors 2020, 20, 2692 7 of 13

For any single actuation event, the current draw signal is represented by two arrays, I and T,
of equal size, n. Where I = [i1 . . . in ] and T = [t1 . . . tn ], I represents the current draw array, while T
denotes the time array. The start and end points of the push-out phase signal can therefore be denoted
by (tpstart , ipstart ) and (tpend , ipend ) respectively, and similarly the start and end points of the swing
phase signal can be denoted by (tsstart , isstart ) and (tsend , isend ) respectively.
To automate the separation of the push-out phase from the swing phase, we employed a gradient
of the slope method, calculated using pair-wise sequential adjacent records. For current draw, i,
and time, t, then;
[i] − [i − 1]
gradient ( gradn ) = (1)
[t] − [t − 1]
The aim of the gradient of the slope method is to identify point (tsstart , isstart ) in the signal data.
For any given actuation, the signal begins with a positive gradient from point (tpstart , ipstart ) up to the
peak current point (denoted here as (tpeak , ipeak ), see Figure 3). After the peak current, the gradient
turns negative as the current draw value descends. At some point during this descend the gradient
becomes non-negative (i.e., grad ≥ 0.00), which we deem to be the start of the swing phase. To identify
the start of the swing phase (tsstart , isstart ), an array, G, of length n, is represented by

1, if ( gradn ≥ 0 and gradn−1 < 0)

(
G= (2)
0, otherwise

The index position at which the first occurrence of “1” appears in array G, therefore, corresponds
to the index position of array, I, and array, T, whose values correspond to the coordinates for (tsstart ,
isstart ). This therefore identifies the point along the actuation signal which corresponds to the beginning
of the swing phase.
Upon separation of the push-out phase from the swing phase, the following features are extracted.
Peak current: The peak current (tpeak , ipeak ) of a signal is found in the push-out phase of an
actuation (see Figure 3). The peak current is taken as the maximum single current draw value found in
the push-out phase.
Push-out phase time duration: The push-out duration is taken as the time difference in milliseconds
(ms) between tpend and tpstart .
Mean current of the swing phase: The mean current draw of the swing phase is taken as the
arithmetic mean current value of the swing phase.
Swing phase time duration: The swing duration is taken as the time difference in milliseconds
(ms) between tsend and tsstart .

3.1. Fast Fourier Transform Analysis

The time series signal data of an actuation event of a POM is shown in Figure 2. Fast Fourier
transform (FFT) facilitates the analysis of such signals in the time domain, to deconstruct components
into a frequency domain representation. It is then possible to analyze the different frequencies present
in the original signal using the frequency domain components that are represented as discrete values
or bins (discrete Fourier transform (DFT)). The frequency domain is useful at showing if noise or jitter
is contained within your time domain signal, which is helpful when attempting to identify spurious
actuation signals from unlabeled data. The principles of FFT are well explained in the literature [17,18]
and so not explained further. The details of the application of FFT for this study is discussed herein.
Typically, each single actuation event in the dataset resulted in approximately 250 sample points.
To ensure all actuation events were processed equally, a zero padding was added to the end of
every actuation signal which resulted in all profiles containing 512 samples and allowing quicker
computation time. A sliding window of size 512, together with a step size of 256 was used with the
input signal. The Hann windowing function was applied to the signal, this minimizes spectral leakage,
allowing the FFT to extract spectral data, and prevents it from producing the wrong frequencies.
Sensors 2020, 20, 2692 8 of 13

Sensors 2020, 20, x FOR PEER REVIEW 8 of 14

FFT assumes a finite dataset, which can be thought of as a single period of a periodic signal. If data are
theofamplitude
not of the discontinuous
periodic circular data at the
topologies, windowing boundaries
using the Hann offunction,
the signal. A FFTthe
reduces of the signal domain
amplitude of the
data shown indata
discontinuous Table 2, for
at the point-operating
boundaries machine
of the signal. POM-1
A FFT moving
of the signalin the NRdata
domain direction,
shownisinplotted
Table 2,as
a frequency
for domain
point-operating plot inPOM-1
machine Figure moving
4. in the NR direction, is plotted as a frequency domain plot
in Figure 4.

Figure 4. Fast Fourier transform, frequency domain plot of the time domain signal data shown in
Figure 4. Fast Fourier transform, frequency domain plot of the time domain signal data shown in
Table 2 for point-operating machine POM-1, moving in the NR direction.
Table 2 for point-operating machine POM-1, moving in the NR direction.
3.2. Curve Fitting
3.2. Curve Fitting
For actuation events to be compared, it is necessary to process the frequency domain plots
produced For (see
actuation
Figureevents to be compared,
4). Anomalies in actuation it is events
necessary can tothenprocess the frequency
be identified domain
to indicate plots
possible
produced
fault occurrence (see Figure
in the 4).
POM.Anomalies
An attempt in actuation
to fit second eventsand canthird
then order
be identified to indicate
polynomials to thepossible
curve
fault occurrence in the POM. An attempt to fit second and third
proved unsuccessful, although other studies have been more successful with their data [16]. Thus, order polynomials to the curve
proved unsuccessful, although other studies have been more successful
the frequency domain signal was further processed by taking the log10 values of both the Amplitude with their data [16]. Thus,
thethe
and frequency
Frequency domain
values signal was further
produced from the processed
FFT analysis.by taking the log10 values of both the Amplitude
andLogthe10Frequency values produced from the FFT analysis.
calculations produce a more linear relationship between the Amplitude and Frequency
domain, Log 10 calculations
allowing a linearproduce
curve to abemore fittedlinear
and the relationship
goodness of between the Amplitude
fit calculated. and Frequency
The frequency domain
domain,
data used allowing
in Figure a4 linear
has beencurvelogto
10
be fitted
transformed and the
and goodness
fitted withof fit
a calculated.
linear curve The
usingfrequency
the R domain
statistical
data used in Figure 4 has been log 10 transformed
package [15]. A plot of this data is shown in Figure 5 below. and fitted with a linear curve using the R statistical
package
After [15].
curveAfitting,
plot ofthethisgoodness
data is shown
of fit isinassessed
Figure 5asbelow. the residual standard error (RSE). This makes
After curve fitting, the goodness of fit
for the final feature extracted for our fault detection analysis. is assessed as the residual standard error (RSE). This
makes for the final feature extracted for our fault
These observed residuals are subsequently used as an estimation detection analysis. parameter of the variability in
These observed
the samples. When theresiduals
RSE is 0 are subsequently
(zero), the curve fits used theasdata
an estimation
perfectly, in parameter of the variability
the real world, this wouldin
the samples. When the RSE is 0 (zero), the curve fits the data
likely be due to overfitting. The RSE in this study becomes useful as a measure of variabilityperfectly, in the real world, thisin
would
the
likely be due to overfitting. The RSE in this study becomes useful
data, which can imply current draw variability in the actuation event of a POM. Thus, a high RSE valueas a measure of variability in the
data,
for one which
actuation canevent
imply current
may indicatedrawfaultvariability
occurrence in when
the actuation
compared event
to anofactuation
a POM.withThus, a highRSE
a lower RSE
value. In this study, we used an arbitrary RSE threshold value of ≥0.3 to serve as a guide to indicate a
value for one actuation event may indicate fault occurrence when compared to an actuation with
lower
error in RSE value. Insignal.
the actuation this study, we used
In practice thisan arbitrary
value wouldRSE needthreshold
to be setvalue of ≥0.3 depending
dynamically to serve as on a guide
the
to indicate error in the actuation signal. In practice this value would
baseline value of each individual POM. For this study, a value of ≥0.3 was used across all instruments. need to be set dynamically
depending on the baseline value of each individual POM. For this study, a value of ≥0.3 was used
across all instruments.
Sensors 2020, 20, 2692 9 of 13
Sensors 2020, 20, x FOR PEER REVIEW 9 of 14

Figure 5. Linear curve fitted to the log10 transformations of the FFT analysis outputs for point-operating
Figure 5. Linear curve fitted to the log10 transformations of the FFT analysis outputs for point-
machine POM-1 in the NR direction.
operating machine POM-1 in the NR direction.
4. Results
4. Results
In this section, we discuss the outcomes of the methodology described above, which is based on
In this section,
the proposition that anweabnormally
discuss thefunctioning
outcomes ofPOM, the methodology
can be identifieddescribed
from theabove, which
current is based
draw signalon
thethat
data proposition
it produces.that Identifying
an abnormally functioning
irregular patterns, POM, can be identified
are indicative of likelyfrom
signstheof current
wear ordraw signal
imminent
data that it produces. Identifying irregular patterns, are indicative
failure [1]. The data used in this study are unlabeled data, here we compare actuation events fromof likely signs of wear or imminent
failure
POMs that[1].have
The been
data processed
used in this bystudy are unlabeled
the method described data,
andhereshow we compare
how potentialactuation events from
fault occurrences
POMs
can that haveThese
be identified. been processed
candidatesby the method
would then bedescribed
flagged up and
forshow
reviewhow and potential fault occurrences
investigation.
canOurbe identified. These candidates would then be flagged up for review
feature extraction approach used traditional methods of mean current and max current draw and investigation.
that haveOurbeen feature
usedextraction
and widely approach
reported used traditional
in other studiesmethods of mean current
[16]. Furthermore, and maxa current
we undertook more
draw that
specific FFT have been to
approach used and frequency
extract widely reporteddomaininfeatures
other studies
which[16].
wereFurthermore,
transformed we intoundertook
the linear a
more
form specific
before FFT approach
calculating the RSEtoofextract
the curvefrequency
fitted. The domain
RSE is features
used aswhich were transformed
an additional into of
feature as part the
linear
our form before calculating the RSE of the curve fitted. The RSE is used as an additional feature as
analysis.
partToofdemonstrate
our analysis. the FFT approach, two examples of two different POMs are shown below. Figures 6
To
and 7, show demonstrate
the currentthe FFTsignal
draw approach, two examples of
of point-operating two different
machine POM-4POMs are shown
and POM-5 below. Figure
respectively.
6 andFigureFigure 7, show
6 shows two the current
actuation drawfor
events signal of point-operating
point-operating machinemachine POM-4 and
POM-4 moving in thePOM-5
RN
respectively.
direction recorded on the 2018-03-13 and 2018-05-17 (Figure 6A,B). Following the method proposed
above,Figure
the RSE6 values
shows calculated
two actuation from events
these two for actuation
point-operating machine POM-4
events following FFT log10 moving in the RN
transformation
direction
and recorded
linear curve on the
fitting, 2018-03-13
suggest that the and 2018-05-17
actuation (Figureat6A
produced theand Figure
later 6B). Following
date (Figure the method
6) is indicative of
proposed above, the RSE values calculated from these two actuation events
potential failure or error. We can see from the highlighted areas (red ellipse) where the variability in the 10 following FFT log
transformation
signal occurs between and linear
the twocurve fitting, suggest
individual events. that the actuation
Furthermore, producederror
the potential at the later (Figure
signal date (Figure
6B)
6) is indicative of potential failure or error. We can see from the highlighted
produces a higher RSE value of 0.256 (Figure 6C) compared to the earlier actuation, which produces a areas (red ellipse) where
thevalue
RSE variability
of 0.092 in(Figure
the signal6D).occurs
The fitted between
curve the two individual
of Figure 6C shows how events.
theFurthermore, the potential
FFT log10 transformation
oferror signal (Figure
the frequency 6B) produces
and magnitude domaina higher RSE value
data points, of 0.256
fit much more (Figure
closely6C) compared
to the to the earlier
curve resulting in a
actuation, which produces a RSE
much lower RSE value than that of Figure 6D. value of 0.092 (Figure 6D). The fitted curve of Figure 6C shows how
theLikewise,
FFT log10intransformation of the frequency and magnitude domain
Figure 7 we present the actuation events for point-operating machine POM-5 moving data points, fit much more
inclosely
the NR todirection,
the curve resulting
with the in a much
events lower RSE
recorded on twovalue than that
separate of Figure
dates. Once6D. again, we show the
variability in the actuation signals (Figure 7A,B) and the subsequent curve fitting following FFT log10
transformation (Figure 7C,D). The actuation event which results in a greater degree of variability
produces a higher RSE value of 0.288 (Figure 7D), compared to the event from an earlier date,
which produced a RSE value of 0.060 (Figure 7C).
Sensors 2020, 20, 2692 10 of 13
Sensors 2020, 20, x FOR PEER REVIEW 10 of 14

Figure 6.6. Actuation

Figure Actuation events
events of
of point-operating
point-operating machine
machine POM-4
POM-4 inin the
the RN
RNdirection.
direction. The
The events
events are
are
recorded on two separate days—2018-03-13 (A) and 2018-05-17 (B)—with their
recorded on two separate days—2018-03-13 (A) and 2018-05-17 (B)—with their FFT log10 FFT log10 transformation
and linear curveand
transformation fromlinear
each curve
event from
directly
eachbelow
event(C,D). Thebelow
directly red ellipse
(C and(A,B)
D). highlights the area
The red ellipse of the
(A and B)
Sensors 2020,
two 20, x FOR
actuation PEER that
events REVIEW
show variability in the signal, which is reflected in the observed RSE 11 of 14
values.
highlights the area of the two actuation events that show variability in the signal, which is reflected
in the observed RSE values.

Likewise, in Figure 7 we present the actuation events for point-operating machine POM-5
moving in the NR direction, with the events recorded on two separate dates. Once again, we show
the variability in the actuation signals (Figure 7A and Figure 7B) and the subsequent curve fitting
following FFT log10 transformation (Figure 7C and Figure 7D). The actuation event which results in
a greater degree of variability produces a higher RSE value of 0.288 (Figure 7D), compared to the
event from an earlier date, which produced a RSE value of 0.060 (Figure 7C).
Similar observations are made throughout the dataset for other POMs where potential faults can
be flagged using the method proposed. Interestingly, the signal profiles of point-operating machine
POM-4 (Figure 6) are very different when compared visually to that of point-operating machine
POM-5 (Figure 7). The areas of the actuation signal where variability is found is clearly not fixed
between different POMs. We also observe that the max peak current for point-operating machine
POM-4 for both the actuation events (Figure 6A and Figure 6B) is approximately 19 A. The same is
observed for point-operating machine POM-5 (Figure 7A and Figure 7B). Therefore, we cannot
exclusively rely on single events such as max peak current as error indicators. Thus, the method
proposed in this study allows the entire signal to be processed accounting for instrument variation.
Figure
Figure
For each 7. Actuation
7. Actuation
instrument events
events
in theof
of point-operating
point-operating
dataset, we have machine
machine POM-5
POM-5
processed thein the
the NR
NR direction.
inentirety direction. The events
The
of the actuation events are
are that
events
recorded
recorded on
ontwo separate
two days—2018-04-12
separate days—2018-04-12(A) and 2018-05-16
(A) and (B)—with
2018-05-16 their FFT
(B)—with log transformation
their
10 FFT log
have been recorded. It is, therefore, possible to observe the RSE values for each of these events 10 to
and linear curveand
transformation from eachcurve
linear eventfrom
directly
eachbelow
event(C,D). Thebelow
directly red ellipse
(C and (A,B)
D). highlights
The red the area
ellipse (A of the
and B)
produce a characteristic profile for a given machine. Plotting the RSE against time, for each
two actuation
highlights the events
area ofthat
the show
two variabilityevents
actuation in the signal, which is reflected in signal,
the observed RSE values.
instrument quickly identifies actuations in the that show
instrument’s variability
timeline in the
where which
it has is reflected
potentially failed
in
or about the observed
to fail. RSE values.
It also helps identify the baseline
Similar observations are made throughout the RSE value
dataset for for
otherany givenwhere
POMs instrument.
potential Tofaults
illustrate
can
this point,
be flagged two
using contrasting
the method examples
proposed. have been selected
Interestingly, the and
signal are shown
profiles in
of Figure 8 below.
point-operating It shows
machine
This again demonstrates that different POMs will all have their individual characteristics and
the calculated
POM-4 (Figure RSE
6) are values for point-operating
very different machines POM-6 ofand POM-3 plotted sequentially.
therefore a methodology that can when compared
be adaptive visually
is more to that
suited to the point-operating
type of challenge machine
addressedPOM-5 in
From
(Figure these plots, it can be readily observed that point-operating machine
7). The areas of the actuation signal where variability is found is clearly not fixed between POM-6, produced RSE
this study.
values that
different are very
POMs. We consistent
also observetothatonethe another
max (1943current
peak actuations). This RSE value, of approximately
Interestingly, point-operating machine POM-3, resultedfor in point-operating
two RSE values machine of ≥0.3, whichPOM-4 for
is the
0.04,the
both seems to be the
actuation eventsbaseline
(Figure value
6A,B)(red
is line) for this machine.
approximately 19 A. The In contrast,
same is point-operating
observed for machine
point-operating
threshold value used in this study to indicate potential fault via the methodology proposed.
POM-3, produced
machine POM-5 RSE values that were much less consistent (1674 actuations), but nevertheless maxa
In line with (Figure
the more 7A,B). Therefore,
conventional we cannot
approaches ofexclusively
signal faultrely on single
detection, twoevents such as
additional useful
baseline
peak value
current of approximately 0.90 could be inferred.
features wereasextracted
error indicators.
from ourThus, the method
actuation proposed
signal data; 1) Thein maximum
this study allows
currentthe drawentire signal
(peak to be
current)
processed
of the push-outaccounting
phasefor and instrument
2) The mean variation.
current draw of the swing phase. We can use these features
to validate our methodology proposed in this study. Plotting the mean current, the peak current and
the RSE on a scatter plot produces some interesting trends between different POMs.
To illustrate this, two examples have been selected on the basis that their processed actuation
events resulted in at least one RSE value of ≥0.3. We selected such instruments so that a plot of the
peak current (push-out phase) against the mean current (swing phase) of the actuation events could
 transformation and linear curve from each event directly below (C and D). The red ellipse (A and B)
highlights the area of the two actuation events that show variability in the signal, which is reflected
in the observed RSE values.

This again demonstrates that different POMs will all have their individual characteristics and
Sensors 2020, 20, 2692 11 of 13
therefore a methodology that can be adaptive is more suited to the type of challenge addressed in
this study.
Interestingly,
For point-operating
each instrument machine
in the dataset, POM-3,
we have resultedthe
processed in entirety
two RSEof values of ≥0.3, which
the actuation eventsisthat
the
threshold
have been value used It
recorded. inis,
this study to possible
therefore, indicate potential
to observe fault
the via
RSEthe methodology
values for each proposed.
of these events to
produce a characteristic profile for a given machine. Plotting the RSE against time, foradditional
In line with the more conventional approaches of signal fault detection, two useful
each instrument
features were extracted from our actuation signal data; 1) The maximum current draw (peak
quickly identifies actuations in the instrument’s timeline where it has potentially failed or about to fail. current)
ofalso
It the push-out phasethe
helps identify and 2) The RSE
baseline meanvalue
current
for draw of theinstrument.
any given swing phase. ToWe can usethis
illustrate these features
point, two
to validate our methodology proposed in this study. Plotting the mean current,
contrasting examples have been selected and are shown in Figure 8 below. It shows the calculated the peak current and
the RSE on a scatter plot produces some interesting trends between different
RSE values for point-operating machines POM-6 and POM-3 plotted sequentially. From these plots,POMs.
it canTo
beillustrate this, twothat
readily observed examples have beenmachine
point-operating selectedPOM-6,
on the basis that their
produced processed
RSE values thatactuation
are very
events resulted in at least one RSE value of ≥0.3. We selected such instruments
consistent to one another (1943 actuations). This RSE value, of approximately 0.04, seems so that a plot of the
to be the
peak current
baseline value(push-out
(red line)phase) against
for this the mean
machine. currentpoint-operating
In contrast, (swing phase) ofmachine
the actuation
POM-3, events could
produced
be differentiated by RSE.
RSE values that were much less consistent (1674 actuations), but nevertheless a baseline value of
approximately 0.90 could be inferred.

Figure 8.
Figure 8. Sequential RSE
RSE values
values plots
plots for
for point-operating
point-operatingmachine
machinePOM-6 (n== 1943)
POM-6(n 1943)and
andPOM-3 (n==
POM-3(n
1674). Possible
1674). Possible baseline
baseline for
for each
each instrument
instrument shown
shown in
in red.
red.

This again demonstrates that different POMs will all have their individual characteristics and
therefore a methodology that can be adaptive is more suited to the type of challenge addressed in
this study.
Interestingly, point-operating machine POM-3, resulted in two RSE values of ≥0.3, which is the
threshold value used in this study to indicate potential fault via the methodology proposed.
In line with the more conventional approaches of signal fault detection, two additional useful
features were extracted from our actuation signal data; 1) The maximum current draw (peak current)
of the push-out phase and 2) The mean current draw of the swing phase. We can use these features to
validate our methodology proposed in this study. Plotting the mean current, the peak current and the
RSE on a scatter plot produces some interesting trends between different POMs.
To illustrate this, two examples have been selected on the basis that their processed actuation
events resulted in at least one RSE value of ≥0.3. We selected such instruments so that a plot of the
peak current (push-out phase) against the mean current (swing phase) of the actuation events could be
differentiated by RSE.
Figure 9 shows a scatter chart for point-operating machine POM-3 and POM-7. The peak current
(A) of the push-out phase is plotted against the mean current (A) of the swing phase, with the data
points colored (red or blue) according to their RSE value. Again, we see that the pattern of the scatter
plots for the two POMs are quite different. Point-operating machine POM-7 contains data that is much
more tightly clustered together and generally of higher peak current and mean current values than
that of point-operating machine POM-3. What we do see in common however is the locations of the
data points of RSE value ≥0.3. For both point-operating instruments, the data points which meet the
threshold of RSE ≥0.3 are pushed out to the right of the major cluster.
 plots for the two POMs are quite different. Point-operating machine POM-7 contains data that is
much more tightly clustered together and generally of higher peak current and mean current values
than that of point-operating machine POM-3. What we do see in common however is the locations of
the data points of RSE value ≥0.3. For both point-operating instruments, the data points which meet
the threshold of RSE ≥0.3 are pushed out to the right of the major cluster.
Sensors 2020, 20, 2692 12 of 13

Figure9.9.Mean
Figure Meancurrent
currentagainst
againstPeak
Peakcurrent
currentofofactuation
actuationsignal
signaldata
datafor
forpoint-operating
point-operatingmachine
machine
POM-3
POM-3andandPOM-7.
POM-7.Data
Datapoints
pointsofofRSE values≥0.3
RSEvalues (red),<0.3
≥0.3(red), <0.3(blue).
(blue).

5. Conclusions
5. Conclusion
In this study, we have proposed a methodology for the identification of potential fault occurrences
In this study, we have proposed a methodology for the identification of potential fault
from current and time signal data of POMs. Data acquired from Network Rail, UK, was cleansed and
occurrences from current and time signal data of POMs. Data acquired from Network Rail, UK, was
pre-processed, with subsequent work focusing on the exploration of a potential condition monitoring
cleansed and pre-processed, with subsequent work focusing on the exploration of a potential
methodology suitable for motorized devices fitted with current sensing capabilities.
condition monitoring methodology suitable for motorized devices fitted with current sensing
The developed method combines mean and max current level thresholds for the push-out and
capabilities.
swing phases with an FFT-based approach which is dynamically suited to the individualistic nature of
The developed method combines mean and max current level thresholds for the push-out and
different POMs. Many studies have been reported in the literature that process labelled data, using a
swing phases with an FFT-based approach which is dynamically suited to the individualistic nature
multitude of approaches, to develop condition monitoring capabilities. The novelty in the method we
of different POMs. Many studies have been reported in the literature that process labelled data, using
propose is that it uses a large volume of real-world unlabeled data. As such we are not pre-aware of
a multitude of approaches, to develop condition monitoring capabilities. The novelty in the method
fault occurrences within the data.
we propose is that it uses a large volume of real-world unlabeled data. As such we are not pre-aware
It is apparent from our efforts that observable differences are evident from the actuation signals
of fault occurrences within the data.
of perceived good and poorly functioning POMs. We exploit these behavioral differences to develop
It is apparent from our efforts that observable differences are evident from the actuation signals
fault sensing capabilities to flag instruments where malfunction could be imminent. This work
of perceived good and poorly functioning POMs. We exploit these behavioral differences to develop
highlights the uniqueness of mechanized motorized devices, and the requirement to move away from
fault sensing capabilities to flag instruments where malfunction could be imminent. This work
a one-size-fits-all approach. Our methodology can consider each instrument on a case by case basis
highlights the uniqueness of mechanized motorized devices, and the requirement to move away from
and therefore monitor at the individual level. The method is unaffected by uncontrollable factors
a one-size-fits-all approach. Our methodology can consider each instrument on a case by case basis
such as environmental conditions or varying operating characteristics which will naturally vary from
and therefore monitor at the individual level. The method is unaffected by uncontrollable factors
machine to machine.
such as environmental conditions or varying operating characteristics which will naturally vary from
A limitation not addressed in this study was the identification of unique RSE threshold values,
machine to machine.
whereby an initial value of ≥0.3, was selected based on the data sets being studied in the earlier stages
A limitation not addressed in this study was the identification of unique RSE threshold values,
of the work. However, as our latter results demonstrate (see Figure 8), some instruments may require
whereby an initial value of ≥0.3, was selected based on the data sets being studied in the earlier stages
much lower thresholds, with the potential to overlook faults if this is not considered. In future work
of the work. However, as our latter results demonstrate (see Figure 8), some instruments may require
we will look at addressing this dynamic thresholding issue by using some form of machine learning
much lower thresholds, with the potential to overlook faults if this is not considered. In future work
approach. It may be more appropriate to use a clustering technique to identify “outliers” in the scatter
we will look at addressing this dynamic thresholding issue by using some form of machine learning
plots of peak current against mean current, such techniques would accommodate for the unique
approach. It may be more appropriate to use a clustering technique to identify “outliers” in the scatter
behavioral characteristics of each instrument. In the field we envisage this machine learning to be
plots of peak current against mean current, such techniques would accommodate for the unique
realized by grouping machines of the same type and using the analytics engine to establish a threshold
behavioral characteristics of each instrument. In the field we envisage this machine learning to be
value relative to the performance of the machines in a particular group.
It is also worthy of note that the direction of the point movement, regarding it being either the
out-swing (NR) or the return (RN) motion was not considered separately in this study. It would be
worthwhile separating the actuations of the two directional movements to see if this produces more
conclusive results. In other reported studies, a difference with respect to directional movement have
been reported [2]. Unfortunately, although timestamps were an integral part of our data records,
we could not use this feature more effectively since we did not know when, in the history of a POM,
where failures or repairs had taken place.
In summary, this work has demonstrated a feasible approach to identify potentially failing POMs,
using data processing techniques that could readily be integrated into a wider remote condition
monitoring architecture.
Sensors 2020, 20, 2692 13 of 13

Author Contributions: The work described in this article is a collaborative effort from all the authors.
Conceptualisation, P.M., P.L., and P.A.; Data Curation, P.M.; Formal Analysis, P.M.; Investigation, P.M., P.L., and
P.A.; Methodology, P.M., P.L., and P.A.; Software, P.M.; Validation, P.M.; Visualisation, P.M.; Preparing draft, review
& editing, P.M., and P.A. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Acknowledgments: The authors would like to thank Fiona Lynch, Network Rail, UK, for supplying the dataset
for this study.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Sa, J.; Choi, Y.; Chung, Y.; Kim, H.-Y.; Park, D.; Yoon, S. Replacement Condition Detection of Railway Point
Machines Using an Electric Current Sensor. Sensors 2017, 17, 263. [CrossRef] [PubMed]
2. Márquez, F.P.; Schmid, F.; Collado, J.C. A reliability centered approach to remote condition monitoring.
A railway points case study. Reliab. Eng. Syst. Saf. 2003, 80, 33–40. [CrossRef]
3. Vileiniskis, M.; Remenyte-Prescott, R.; Rama, D. A fault detection method for railway point systems. Proc. Inst.
Mech. Eng. Part F J. Rail Rapid Transit 2015, 230, 852–865. [CrossRef]
4. Alwadie, A. The Decision making System for Condition Monitoring of Induction Motors Based on Vector
Control Model. Machines 2017, 5, 27. [CrossRef]
5. García Márquez, F.P.; Roberts, C.; Tobias, A.M. Railway point mechanisms: Condition monitoring and fault
detection. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2009, 224, 35–44. [CrossRef]
6. Bolbolamiri, N.; Sanai, M.S.; Mirabadi, A. Time-domain stator current condition monitoring: Analyzing point
failures detection by Kolmogorov-Smirnov (KS) test. Int. J. Electr. Comput. Energ. Electron. Commun. Eng.
2012, 6, 587–592.
7. McHutchon, M.A.; Staszewski, W.J.; Schmid, F. Signal Processing for Remote Condition Monitoring of
Railway Points. Strain 2005, 41, 71–85. [CrossRef]
8. Atamuradov, V.; Medjaher, K.; Camci, F.; Dersin, P.; Zerhouni, N. Railway Point Machine Prognostics Based
on Feature Fusion and Health State Assessment. IEEE Trans. Instrum. Meas. 2018, 68, 2691–2704. [CrossRef]
9. Bian, C.; Yang, S.; Huang, T.; Xu, Q.; Liu, J.; Zio, E. Degradation detection method for railway point machines.
arXiv 2018, arXiv:1809.02349.
10. Eker, Ö.; Camci, F.; Kumar, U. SVM Based Diagnostics on Railway Turnouts. Int. J. Perform. Eng. 2012, 8,
289–298.
11. Asada, T.; Roberts, C.; Koseki, T. An algorithm for improved performance of railway condition monitoring
equipment: Alternating-current point machine case study. Transp. Res. Part C Emerg. Technol. 2013, 30, 81–92.
[CrossRef]
12. García, F.P.; Pedregal, D.J.; Roberts, C. Time series methods applied to failure prediction and detection. Reliab.
Eng. Syst. Saf. 2010, 95, 698–703. [CrossRef]
13. Mistry, P.; Lane, P.; Allen, P.; Al-Aqrabi, H.; Hill, R. Condition Monitoring of Motorised Devices for Smart
Infrastructure Capabilities. In Smart City and Informatization; Springer: Singapore, 2019; pp. 392–403.
14. Knime. Available online: https://www.knime.com/ (accessed on 20 February 2020).
15. R. Available online: https://www.r-project.org/ (accessed on 20 February 2020).
16. Chamroukhi, F.; Same, A.; Aknin, P.; Antoni, M. Switch mechanism diagnosis using a pattern recognition
approach. In Proceedings of the 2008 4th IET International Conference on Railway Condition Monitoring,
Derby, UK, 18–20 June 2008; pp. 1–4.
17. Cochran, W.T.; Cooley, J.W.; Favin, D.L.; Helms, H.D.; Kaenel, R.A.; Lang, W.W.; Maling, G.C.; Nelson, D.E.;
Rader, C.M.; Welch, P.D. What is the fast Fourier transform? Proc. IEEE 1967, 55, 1664–1674. [CrossRef]
18. Oberst, U. The Fast Fourier Transform. SIAM J. Control Optim. 2007, 46, 496–540. [CrossRef]

© 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/).