International Journal of Pure and Applied Mathematics

Volume 118 No. 19 2018, 1731-1750

ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version)
url: http://www.ijpam.eu
Special Issue
ijpam.eu

Detection of Faults Caused by Loose

Cable Connection Using Reflectometry
Taek-Hee Kim*1 , Yong-Seok Kim1 ,
Chong-Min Kim1 , Kil-Mok Shong1 , In-ho Ryu2
1
Electrical Safety Research Institute,
Korea Electrical Safety Corporation,
Wanju,55365 Korea
2
Department of IT System Engineering,
CHONBUK National University,
Junju,54896, Korea
[email protected] , [email protected] ,
[email protected] , [email protected] ,
[email protected]
Corresponding author*
Phone:+82-063-716-2363

February 4, 2018

Abstract
Background/Objectives: Electricfires caused by ca-
ble contact failure have led to a substantial amount of prop-
erty loss. A diagnostic technique for detecting cable contact
faults in advance is required to prevent electric fires and ac-
cidents.
Methods/Statistical analysis: A study was conducted
to detect the loose connection fault caused by a decreased
tightening torque of cable connection terminals. At loose
connection state below the reference torque, the poor con-
tact fault generated between the cable and the connection

1731
International Journal of Pure and Applied Mathematics Special Issue

terminals was measured using the STDR and SSTDR tech-

niques to detect the contact failure. The accuracy of the
reflection signal detection was improved by applying a ref-
erence signal elimination algorithm.
Findings: This study suggested a method to detect
cable faults in a loose connection state to prevent poor con-
tact fault in advance. Loose connection is regarded as a
soft fault that is difficult to distinguish, compared to other
palpable faults such as disconnection or short circuits. For
this reason, this study proved through experiments how the
poor contact fault can be detected at different tightening
torques. In addition, a suitable reflectometry according to
the state of the cable (live or dead) was selected and the de-
tection accuracy of reflection signal returning from the fault
point was improved by applying a reference signal elimina-
tion algorithm. The accuracy of the detection was verified
by checking the fault location and error rate of the detected
signal. As a result of such experiments, in a loose connec-
tion condition caused by a decreased tightening torque at a
dead-wire state, detecting complicated faults (poor contact
fault/disconnection fault) that may occur simultaneously in
a single cable was possible at the connection torque of using
STDR.
Improvements/Applications: Develop an optimal ref-
erence signal by applying various sequences, an algorithm
that can detect faults due to a fine change in connection
strength at the connection pointand an improved diagnos-
tic method that enables detection of cable faults in a more
diverse fault environment.
Key Words: Reflectometry, Poor contact, Loose con-
nection faults, Fault location, STDR, SSTDR.

1 Introduction
According to the fire statistics of the National Fire Protection Cen-
ter, the Ministry of Public Safety and Security, while the occupancy
rate of electric fire has stood still in the range of 20[%] for the past
10 years, the amount of property damage caused by electric fire has
almost doubled from KRW39,927 million in 2006 to KRW72,253
million in 2015 during the same period. This is because the num-

1732
International Journal of Pure and Applied Mathematics Special Issue

ber of electric energy users has surged year by year thanks to the
economic growth and enhanced living standard of people, exposing
themselves to an increasing risk of electric fire and property dam-
age. According to the statistics on electricity accidents (by the type
of electric equipment) in 2015, among the total 7,760 electric fires
in Korea, fires caused by wiring and/or wiring devices accounted for
21.3[%] (1,654 cases); by the cause of electric fire, poor cable con-
tact triggered a total of 748 fires that left 18 casualties and property
damage of KRW 3,402,208,000, the second largest amount after a
short-circuit caused by insulation deterioration1 . Accidents caused
by poor cable contact lead to power outages and fire, which trigger
enormous human and economic damage such as property damage,
information loss, and production disruption. For this reason, a
diagnostic technology needs to be developed to detect poor cable
fault and prevent a huge economic loss and/or personal injury due
to electric fire in advance.
For diagnosis of poor contact fault at the site, however, they
mainly rely on a method of detecting cable deterioration using the
heat generated by a cable connection point or an arc phenomenon,
visual inspection of the inspector, and/or tightening of cable termi-
nal bolts. With these conventional methods, it is difficult to detect
poor contact fault before any abnormal issue arises. Recently, a
new cable diagnosis method using reflectometry has been applied
in the field, but it could only detect temporary palpable faults such
as disconnection fault or short-circuit fault; with soft faults such
as poor contact fault, it showed a high detection error rate or a
difficulty in detecting faults2,3 .
Poor contact fault is divided into loose connection fault caused
by a drop in the torque at the connection point of the cable and
intermittent fault that repeats connection and disconnection at the
connection point due to external influences and physical forces4−6 .
A diagnostic method for identifying intermittent fault and detect-
ing the exact faulty locations had been suggested in a precedent
study of the author 7 . This time, a study was conducted to detect
loose connection fault (one of poor contact faults) caused by a drop
in the cable connection tightness (torque). In order to realize the
loose connection fault of the cable and circuit breaker, the tighten-
ing torque of 6.0 [Nm] (corresponding to the connection terminal
screw diameter of 8 [mm]) was applied in the experiment as the

1733
International Journal of Pure and Applied Mathematics Special Issue

reference; a diagnostic method was proposed to detect the contact

failure through reflectometry in a loose connection state under the
reference torque of 6 [Nm] 8,9 . To improve the performance of the
cable fault detection, the STDR/SSTDR technique and the ref-
erence signal elimination algorithm were used to more accurately
detect the reflection signal from the faulty point; the accuracy of
detection was verified by checking the experimental results of fault
point and detection error rate 10 .

2 STDR/SSTDR and Reference signal

elimination algorithm
2.1 STDR/SSTDR
Sequence Time-Domain Reflectometry (STDR) and Spread Spec-
trum Time-Domain Reflectometry (SSTDR) algorithms are known
for a strong resistance against noise environment and an excellent
fault location detection performance. By adopting the spectrum
technology used in digital communications, both techniques use
the signals applied to the cable as sequences and solve issues of the
conventional Time Domain Reflectometry (TDR) such as signal at-
tenuation and noise 11−13 . When a cable failure is detected in a
live wire condition, SSTDR that uses a sine-wave reference signal
was more advantageous than STDR using square-wave signal due
to distortion of waveform7 .
STDR/SSTDR determine the types of faults (e.g. disconnection
and short-circuit) by i) applying a pseudo noise (PN) sequence with
good auto-correlation properties to the cable as in Figure 1, ii)
calculating the maximum values of the coefficients through cross-
correlation with the reflection signal from the fault location, iii)
detecting the fault location based on the time difference between
the maximum values, and iv) analyzing phases of the reflection
signals 13 . SSTDR differs from STDR in that it modulates the PN
sequence with good autocorrelation properties using a sine-wave
signal and applies such a modulated signal to the cable.

1734
International Journal of Pure and Applied Mathematics Special Issue

Fig. 1. The principle of STDR/SSTDR

In STDR/SSTDR, the reference signals(t)using the sequence

c = [c, c1 , · · · , c(N −1) ], ci ∈ [−1, 1] with length N is generated by
multiplying the pseudo noise signal and the carrier signal as in the
following Equation (1).
N
X −1
s(t) = cn P Tc (t − nTs ) (1)
n=0

In STDR, the signal multiplied by the sequence is divided into 1

and 0 in Equation(2). In SSTDR, it is a sine-wave signal as shown
in Equation (3).
 
1, 0 ≤ t < Tc
P Tc (t) = (2)
0, etc
 
 1 
cos(2πfc t), 0 ≤ t < Tc , fc =
P Tc (t) = Tc (3)
 0, etc 
Where,Tc is the chip duration of the PN sequence having a value
of 1 or -1 and fc is the frequency of the carrier wave. If the signal
of Equation (1) is applied to the cable, reflection occurs after a
certain time delay at the characteristic impedance change point of
the cable. The reflected signal may contain normal noise depending
on the cable test environment and can be calculated as in Eq. (4).
N
X −1
r(t) = ak s(t − Tk ) + g(t) (4)
n=0

1735
International Journal of Pure and Applied Mathematics Special Issue

Where,ak is the magnitude of the reflection signal ak s(t − Tk )

in relation to the reference signal,is the delay time from when the
reference signal is reflected at the cable fault location, and g(t) is
the noise signal.
Next, as shown in Equation (5), the reference signal s(t) is
shifted by time τ to analyze the time cross-correlation with the re-
flection signal r(t) and to get the maximum value of the correlation
coefficient. Then, the time taken for the signal to be reflected back
from the fault point is calculated using the difference between the
maximum values of correlation coefficients of reference/reflection
signals. The fault location of the cable is calculated using Equa-
tion (6).
Z
1 Ts
Ctime,sr (τ ) = s(t)r(t − τ )dt (5)
Ts 0
Vp × τD
d= (6)
2a
Where, Ts denotes a time width or cycle of the reference signal
s(t) , Vp means a velocity of propagation (VOP) of a signal applied
to the power cable, and τD indicates a time from when a signal is
applied to the cable to when it is reflected back.
Then, the time taken for the signal to be reflected back from the
fault point is calculated using the difference between the maximum
values of correlation coefficients of reference/reflection signals. The
distance to the cable fault location, d is calculated using Equa-
tion(6).

2.2 Applying reference signal elimination algo-

rithm
By measuring the cable fault detection using the existing SSTDR,
it was possible to detect the fault location because the correlation
value of reflection signal, in the case of disconnection fault, was big
as shown in Figure 2 (a). In the case of contact failure as shown
in Figure 2 (b), however, it was difficult to determine the fault
location due to the small correlation value of reflection signal. For
this reason, this study used a reference signal elimination algorithm
to improve the fault detection performance in case of poor contact
fault.

1736
International Journal of Pure and Applied Mathematics Special Issue

Fig. 2. Detecting cable fault using the existing SSTDR (a)

disconnection fault (b) poor contact fault

In the reference signal elimination algorithm, after the reference

signal is applied to the cable through STDR/ SSTDR, the maxi-
mum point of the reference signal is obtained first, and then the
reference signal is eliminated; through the time-correlation analysis
of the signal reflected from the fault location, the maximum point of
the reflection signal is detected. The time difference and the phase
between the maximum values of the detected reference/reflection
signals are analyzed to determine the fault location and the fault
type3 .

1737
International Journal of Pure and Applied Mathematics Special Issue

Fig. 3.Technique of Reference Signal Elimination

The proposed method proceeds in the order in Figure 3. In
the first time-correlation analysis step, the maximum point of the
reference signal is obtained by locating its maximum value τ1 us-
ing Equation (5). In the reference elimination step, the reference
signal s(t) is removed from the reflection signals r(t) , to produce
b(t) = r(t) − s(t − τ1 ) . In the second time-correlation step, a time
cross-correlation analysis with b(t) and s(t) using Equation(7) is
conducted to obtain the maximum correlation value τ2 of the re-
flection signal. As the maximum value of the correlation coefficient
is close to 1 due to the normalization effect, the fault classification
is clarified and the measurement error is reduced.
Z
1 Ts
Ctime,sr (τ ) = s(t)b(t − τ )dt (7)
Ts 0
Finally, in the fault detection step, the difference between the
maximum correlation coefficients of reference/reflection signals τD =
τ2 − τ1 is calculated. The distance to the cable fault location is ob-
tained using Equation(6).

3 Experiments and Results

3.1 Test method of connection fault detection
To detect cable contact failure based on STDR/SSTDR, a system
consisting of an arbitrary waveform generator (AWG, a PXI type

1738
International Journal of Pure and Applied Mathematics Special Issue

from National Instrument Corp.), digital oscilloscope, and a con-

trol PC was prepared. The arbitrary waveform generator used in
the experiment supports a 16-bit resolution and a sampling rate of
200 MS/s, and the digital oscilloscope features a sampling rate of
5 GS/s and a bandwidth of 1.5 GHz. A program to control the
arbitrary waveform generator and digital oscilloscope and trans-
mit/receive signals from the control PC was developed using LAB-
VIEW,MATLAB was used for data measurement and signal anal-
ysis.
Among the most frequently used FR-CV cables, the CV-2c6sq
cable was selected for indoor and underground cables under low
pressure 600 [V]. For the signal cable for transmitting and receiv-
ing signals to and from the cable, RG58 cable (1.33[m] in length)
with T-connector was used. An experiment to detect a loose con-
nection, one of the poor contact faults, caused by a drop in the bolt
tightening strength (torque) of the contact area between the cable
and breaker was conducted as shown in Figure 4.

Fig. 4. Configuration of test system detecting loose connection

fault

The poor contact faults can be divided into i) loose connection

fault where the cable connection is not made properly due to a de-
creased tightening torque of the contact terminal at the connection
point of the cable and ii) intermittent connection fault where the
connection and disconnection are repeated at the connection site.
In this study, a test of a poor contact that may arise due to a re-
duced bolt tightening torque of the connection area between the
cable and circuit breaker was carried out after cable installation. A
wiring breaker was used for the connection point as shown in Figure
5 (a), according to IEC 60947-7-1 Standardsas shown in table 1, a
torque driver as in Figure 5 (b) was used to detect contact failure

1739
International Journal of Pure and Applied Mathematics Special Issue

at loose connection state (with 6 [Nm] or lower of reference torque

value).

Table 1. Terminal tightening torque regulations

Fig. 5.(a) Access terminal (b) Loose connection

Table 2 shows the test conditions for contact failure detection

caused by a degradation of cable connection strength using STDR
/SSTDR.

Table 2. Experimental conditions of poor contact fault

3.2 Experimental results and considerations

The CV 2c6sq cable at a dead-wire state was connected each cir-
cuit breaker on the front/rear part of the connection point (breaker)
as in Figure 4, and the tightening torque of the connection bolts

10

1740
International Journal of Pure and Applied Mathematics Special Issue

was less than the reference value of 6[Nm]. In addition, PN se-

quence with good autocorrelation properties was applied to the
cable through STDR/SSTDR, which features an excellent fault lo-
cation detection capability, and a reference signal elimination algo-
rithm designed to enhance the reflection signal detection accuracy
was adopted to detect poor contact caused by a looseconnection
fault.
Figure 6 shows the result of the fault detection test using STDR
/SSTDR; for the test, the 120 [m] cable (front) and the 60 [m] cable
(rear) of the CV 2c6sq cable were connected to the circuit breakers,
the connection tightening torque was set to be 1∼2 [Nm] at the 120
[m] point, and a fault due to disconnection was generated at the
point of 180 [m]. Experimental results in Figure 6 (a) indicate the
connection fault error rate of 0.14[%] at 119.83 [m] and disconnec-
tion fault error rate of 0.83[%] at 178.50 [m]; the error rates are
calculated using the distance in the Equation(8) and Equation(9)
in the MATLAB program. Figure 6 (b) shows the connection fault
error rate of 0.78 [%] at 119.07 [m].
159
(1.905 × 108 ) × { 125×106}
D= − 1.33 = 119.83[m] (8)
2
120 − 119.83
Errorrates[%] = = 0.14[%] (9)
120 × 100

11

1741
International Journal of Pure and Applied Mathematics Special Issue

Fig. 6. CV 2c6sq 120m + 60m cable connection (a) STDR (b)

SSTDR

12

1742
International Journal of Pure and Applied Mathematics Special Issue

Fig. 7. CV 2c6sq 152m + 200m cable connection (a) STDR (b)

SSTDR

13

1743
International Journal of Pure and Applied Mathematics Special Issue

Fig. 8. CV 2c6sq 200m + 60m cable connection (a) STDR (b)

SSTDR

14

1744
International Journal of Pure and Applied Mathematics Special Issue

Fig. 9. CV 2c6sq 200m + 200m cable connection (a) STDR (b)

SSTDR

Through experiments on the connection fault detection that

may occur at a relatively high tightening torque of 3 [Nm] or higher
at the connection area, detecting connection failure caused by loose
connection at 120 [m] was impossible as indicated in Figure 10, and
only disconnection fault that may occur in the whole cable was de-
tected. Figure 10 (a) and Figure 10 (b) show the disconnection
fault error rates of 0.41 [%] at 179.26 [m] and of 1.25 [%] at 177.74
[m], respectively.

15

1745
International Journal of Pure and Applied Mathematics Special Issue

Fig. 10. CV 2c6sq 120m + 60m cable connection (a) STDR (b)
SSTDR

At the cable and connection terminal area, an experiment of

poor contact fault detection, which is caused by a decreased tight-
ening torque and subsequent loose connection, was carried out; the
tightening torque of the connection terminal (breaker) was 6.0 [Nm]
or less, which corresponds to the terminal screw diameter of 8 [mm].
From the experiment, it was found that detecting the connection
fault at the tightening torque of 3 [Nm] (or higher) was not possible
as in Figure 10. Table 3 shows the error rates of connection failure
location/detection in the range of connection tightening torque of
1∼2 [Nm].
As a result, STDR, rather than SSTDR, was more effective in
detecting multiple cable faults such as poor contact fault and dis-
connection fault simultaneously in the same cable at a dead-wire
state; with SSTDR, detecting poor contact fault on the front area
was possible but detecting disconnection fault that may occur on

16

1746
International Journal of Pure and Applied Mathematics Special Issue

the rear side simultaneously was difficult. As a result of STDR

fault detection in Figure 7 (1), the error rate of contact failure de-
tection was relatively high. In a precedent study, SSTDR using
the sine-wave signal with less distortion of the waveform due to
noises in a live wire condition was found to be more advantageous
than STDR in detecting the poor contact fault7 . The conclusions
obtained from the above experimental results: Diagnosis with a sig-
nal measurement method selected according to the cable condition
(live or dead) is more helpful in detecting the exact fault location
and complicated faults (poor contact fault/disconnection fault) in
a single cable. The reasons why detecting a poor contact fault at a
relatively high tightening torque of 3 [Nm] (or higher) is impossible
include a low sampling rate, wrong connection method, or other
various factors.

Table 3. Error rate and fault location for poor contact fault

4 Conclusion
A loose connection condition, in which the cable connection is not
properly made due to a reduced cable connection tightening torque,
leads to poor contact fault as evidenced in this study. This study
suggested a method to detect cable faults in a loose connection
state to prevent poor contact fault in advance. Loose connection
is regarded as a soft fault that is difficult to distinguish, compared
to other palpable faults such as disconnection or short circuits. For
this reason, this study proved through experiments how the poor
contact fault can be detected at different tightening torques. In
addition, a suitable reflectometry according to the state of the ca-
ble (live or dead) was selected and the detection accuracy of re-

17

1747
International Journal of Pure and Applied Mathematics Special Issue

flection signal returning from the fault point was improved by ap-
plying a reference signal elimination algorithm. The accuracy of
the detection was verified by checking the fault location and error
rate of the detected signal. As a result of such experiments, in a
loose connection condition caused by a decreased tightening torque
at a dead-wire state, detecting complicated faults (poor contact
fault/disconnection fault) that may occur simultaneously in a sin-
gle cable was possible at the connection torque of 1∼2 [Nm] using
STDR.
Going forward, this study aims to develop an optimal reference
signal by applying various sequences, an algorithm that can detect
faults due to a fine change in connection strength at the connection
point (tightening torque of 3 [Nm] or higher), and an improved
diagnostic method that enables detection of cable faults in a more
diverse fault environment.
Acknowledgment
This work was supported by Civil-Military Technology Cooper-
ation Program (15-CM-MA-14).

References
[1] Korea Electrical Safety Corporation, “A Statistical Analysis
on the Electrical Accident”, 2016, www.esps.or.kr/statistics/
internalElecConst/overAllStatistics/elecFireMot.do.

[2] Jeong-chay Jeon, Taek-hee Kim, Jae-Geun Yoo, “Fault Detec-

tion of Low-voltage Cable using Time-Frequency Correlation
in SSTDR”, KIEE, Vol. 64, No. 3, pp. 498-504, March 2015.

[3] Jeong-chay Jeon, Taek-hee Kim, “Cable Fault Detection Im-

provement of STDR Using Reference Signal Elimination”,
KAIS, Vol. 17, No. 3, pp.450-456, March, 2016.

[4] Xia Zhou, Thomas Schoepf, “Detection and Formation Process

of Overheated Electrical Joint due to Faulty Connection”, 26th
International Conference on Electrical Contacts (ICEC 2012),
pp. 288-295, Sep, 2012.

[5] Jim Livie, “The Application of On-Line Travelling Wave Tech-

niques in the Location of Intermittent Faults on low Voltage

18

1748
International Journal of Pure and Applied Mathematics Special Issue

Underground Cables”, 2008 IET 9th International Conference

in Developments in Power System Protection (DPSP 2008),
pp. 714-719, May. 2008.

[6] Maud Fchet, Nicolas Ravot, “Soft Fault Detection in Cables

using the Cluster Time-Frequency Domain Reflectometry”,
IEEE Electromagnetic Compatibility Magazine, Vol 2, Issue
1, pp 54-69, 1st Quarter, 02 May 2013.

[7] Taek-hee Kim, Jeong-chay, “Diagnosis of Poor contact Fault

in the Power Cable Using SSTDR”, KIEE, Vol. 65, No. 8, pp.
1442-1449, Aug 2016.

[8] IEC 60947-7-1, Standards for Low-voltage switchgear and

controlgear-Part 7-1:Ancillary equipment-Terminal blocks for
copper conductors.

[9] Jeong-jae Han, So-ryoung Park, “Performance Comparison

and Improvement of STDR/SSTDR Schemes Using Various
Sequences”, KICS, Vol. 39, No. 11, pp.637-644, Nov, 2014.

[10] Paul Smith, “Spread spectrum time domain reflectometry”,

Ph. D. dissertation, Dept. Elect. Comput.Eng., Utah Univ.,
Logan, 2003.

[11] P. Smith; C. Furse, and J. Gunther,“ Analysis of spread spec-

trum time domain reflectometry for wire fault location”, IEEE
Sensors J., Vol. 5, No. 6, pp. 1469-1478, Dec. 2005.

[12] Cynthia Furse, Paul Smith, Mehdi Safavi and Chet Lo,
“Feasibility of Spread Spectrum Sensors for Location of Arcs
on Live Wires”, IEEE Sensors Journal, Vol. 5, No. 6, pp. 1445-
1450, Dec. 2005.

[13] Chirag R. Sharma, Cynthia Furse and Reid R. Harrison, “Low-

Power STDR CMOS Sensor for Location Faults in Aging Air-
craft Wiring”, IEEE Sensors Journal, Vol.7, No.1, pp. 43-50,
Jan. 2007.

19

1749
1750