Sensors & Actuators: A.

Physical 349 (2023) 114061

Contents lists available at ScienceDirect

Sensors and Actuators: A. Physical

journal homepage: www.journals.elsevier.com/sensors-and-actuators-a-physical

Pipe leakage detection using ultrasonic acoustic signals

Jae Cheol Lee, You Rak Choi, Jai Wan Cho *, 1
Department of Equipment & Structure Prediction Diagnosis Research, Korea Atomic Energy Research Institute, Daejeon, South Korea

A R T I C L E I N F O A B S T R A C T

Keywords: In the work reported in this paper, two ultrasonic leak detectors were designed and manufactured to detect pipe
Ultrasonic leak detection ASTM E1002-05 leaks remotely using a non-contact, non-destructive method. One ultrasonic leak detector module with a para­
nozzle bolic reflector (diameter 150 mm and focal length of 100 mm) and a second ultrasonic leak detection module
with a conical horn guide (length 100 mm, angle 9.5◦ ) were fabricated. After CAD design, the parabolic reflector
was processed by 3D printing manufacturing method. The performance of these devices was tested and evaluated
in an outdoor environment in accordance with the ASTM E1002–05 class II equipment verification standards.
The field verification tests were conducted about 20 times. According to the evaluation of the test results, the
average S/N (signal to noise) ratio of the ultrasonic leak detector with the parabolic reflector was about 4.97 dB,
and the average leak detection S/N ratio of the ultrasonic leak detector with a conical horn guide was about 1.89
dB

1. Introduction To verify its practical applicability, a saturated steam leakage mockup

test device was constructed at the Idemitsu Kosan Chiba (Japan) oil
If we assume that plumbing equipment corresponds to blood vessels, refinery. A neural network signal-processing technique was used to
then bleeding (that is, leakage of media from the pipes) must have a distinguish leakage and background noise from microphone signals. In
great influence on operation of the surrounding facility. Considering the the mechanical noise (approximately 90 dB) environment of the Ide­
potential for internal (within the structure) and environmental mitsu Kosan Chiba oil refinery during normal operation, a steam leak
(external) damage, the effect could be substantial. For example, the ejected from a nozzle with a pressure of 350 kPa was identified at a
leakage of highly corrosive fluids (acids, fluorine) and very dangerous distance of 16 m [3–7]. R.P. Cruz et al. achieved detection of a gas leak
substances such as Na have the potential for huge impacts on the envi­ with an aperture of 0.5 mm at a pressure difference of 100 kPa. An
ronment. In addition, an accident involving the leakage of a combustible omnidirectional condenser microphone with a frequency range of 50 Hz
gas under high pressure has the risk of causing secondary disasters to 20 kHz and a sensitivity of − 48 dB was used [8]. Lei Li et al. designed
(explosions, fires). For all these reasons, an effective system is needed for a virtual linear ultrasonic transducer array of only two transducers to
the early detection of pipe leaks. In general, when leakage of a fluid acquire leak signals. They used an ultrasonic transducer (FUS-40CR)
occurs due to defects (perforations, cracks) in high-pressure pipes, ul­ with a nominal frequency of 40 kHz and a sensitivity of − 46 dB to
trasonic waves are generated [1]. For example, when high-pressure gas detect compressed air under 700 kPa pressure, discharged from a
inside a pipe escapes through a break and flows into the air, the pressure 0.5 mm hole at a distance of 70 cm [9]. Mengjie Xu detected a gas leak
pulses and generates elastic waves. The released gas becomes turbulent signal emitted from a 0.1 mm hole at a 100 kPa gauge pressure at a
at the outlet and generates ultrasonic waves. When the leak rate of gas is distance of 500 mm by combining four ultrasonic sensors with a center
1x10− 3 std.cc/sec or above, it is known that the ultrasonic waves frequency of 40 kHz and a bandwidth of 6 kHz [10]. R.B. Santos et al.
generated due to turbulence can be detected [2]. An ultrasonic leakage achieved detection of gas leakage through a pore of 1 mm at a pressure
detection unit can detect such ultrasonic signals remotely (and whether difference of 0.6 MPa. An omnidirectional condenser microphone with a
a pipe is leaking) in a non-contact and non-destructive manner. frequency range of 16–20 kHz and a sensitivity of − 46 dB was used
In the late 2000 s, Tani et al. remotely detected ultrasonic leakage [11]. Jian Li et al. detected a flow rate of 40 cc/min leaking from a
signals using M/P (microphone) sensors in the band from 2 to 40 kHz. 0.1 mm hole at a distance of 20 cm using 18 micro electro-mechanical

* Corresponding author.
E-mail address: jwcho@kaeri.re.kr (J.W. Cho).
1
ORCID 0000-0001-8307-3536

https://doi.org/10.1016/j.sna.2022.114061

Available online 5 December 2022

0924-4247/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

system (MEMS) microphone arrays in the frequency band of 10–20 kHz range 20–80 kHz into 320 center frequencies fc with a unit bandwidth
[12]. S. Wang and X. Yao used a microphone (YSV5001) sensor in the (Δf). For 320 center frequencies, the FFT operation was performed using
range of 20 Hz to 20 kHz to detect the leakage from a 1 mm hole at a 2Δf as a window. Then, a 320 power spectrum was obtained for each of
pressure of 30 kPa [13]. Rui Xiao et al. used a microphone sensor (PCB the 320 center frequencies. In this paper, Δf was set to 200 Hz.
105B50) in the frequency range of 0.5 Hz to 40 kHz to detect pressur­
fc1st = 20kHz (1)
ized gas leaks from a 1 mm hole at 100 kPa [14].
During the work reported in this paper, two ultrasonic leak detectors
fc2nd = 20kHz + Δf (200Hz) = 20.2kHz (2)
were designed and manufactured to detect pipe leaks remotely using a
non-contact, non-destructive method. One ultrasonic leak detector fcnth = 20kHz + (n − 1)Δf (3)
module with a parabolic reflector (diameter 150 mm and focal length of
100 mm) and a second ultrasonic leak detection module with a gun-type The intensity distribution of the 320 FFT power spectra are received
conical horn were fabricated. When high-pressure gas leaks from a pipe, by a data logger (PC) through a serial interface (I/F), as shown in the
the audible sounds of 20 kHz or less, among all the acoustic frequencies block diagram of Fig. 1. A PC analyzes this and determines whether the
generated, are removed. An ultrasonic signal in the range of 20–80 kHz, pipe is leaking. Fig. 2 shows a photograph of the newly designed and
which is relatively easy to distinguish from background noises manufactured ultrasonic leak detector. The size of the new ultrasonic
(including mechanical noise) was processed. The ultrasonic sensor used leak detector is 67 × 40 mm. In the photographs shown in Fig. 2, the red
in the design and manufacture of the ultrasonic leak detector is a MEMS circle seen on the front face (top left) is a MEMS microphone. The
microphone. In addition, the performance of the newly designed and acoustic hole is located on the bottom of the chip. A microphone hole
manufactured ultrasonic leak detector was evaluated based on the ASTM (diameter 0.6 mm) was drilled in the PCB board to collect the ultrasonic
E1002–05 class II equipment verification criteria of the American So­ signal. The hole within the red dotted circle in the rear face of Fig. 2
ciety for Testing and Materials. In consideration of the need to apply (lower image) illustrates this.
leakage detection to piping facilities installed outdoors, a performance Fig. 3 shows the inherent response distribution of the ultrasonic leak
evaluation was also executed under the climatic conditions found in the detector developed in this work. In Fig. 3, the X axis represents fre­
outdoor environment and the ambient noise environment of the sur­ quency, and the Y axis represents the sensitivity. It has a peak sensitivity
rounding facilities. That is, the ultrasonic signal emitted from a nozzle around 30 kHz and was designed and manufactured to have relatively
having a diameter of 0.2 mm under a gauge pressure of about 70 kPa high response characteristics in the 20–40 kHz frequency range. Fig. 4
was measured at a distance of 5 m or more. The leak rate (100 mL/min) shows the outline of the conical horn guide of the ultrasonic leak de­
according to ASTM E1002–05 class II device verification requirements tector in this work. The length of the horn guide l is 100 mm, the
was implemented experimentally. diameter d1 of the throat contacting the MEMS M/P through hole is
6 mm, and the outer diameter d2 of the conical horn guide is designed to
2. Design and manufacture of an ultrasonic leak detector be 40 mm. The acoustic field-of-view (FOV) of the conical horn guide is
calculated as 9.65 degrees as in Eq. (4). Fig. 5 is a conceptual diagram of
Fig. 1 shows a block diagram of the newly designed and manufac­ the horn guide of the parabolic reflector. The curved surface of the inner
tured ultrasonic leak detector. surface of the reflector has a parabolic structure. Ultrasonic waves
In the block diagram in Fig. 1, a MEMS microphone is shown as the (plane waves) incident on the reflector are focused on the MEME M/P
acoustic sensor. The frequency range of the microphone sensor is 100 Hz sensing area. In Fig. 5, l is the focal length, APR is the cross-sectional area
to 80 kHz, and the sensitivity of the sensor is − 38 dBV/Pa (10 dB at of the parabolic reflector, and Aλ is the cross-sectional area of the MEMS
100 Hz to 80 kHz). The amplifier stage amplifies the microphone sound M/P sensing area. Assuming that the ultrasonic wave incident on the
signal. The amplification unit was designed in three stages (1st stage: parabolic reflector are reflected and focused on the sensing area of the
60 dB, 2nd stage: 30 dB, 3rd stage: 14.6 dB), and the total gain is MEMS M/P, the gain GPR is as Eq. (5). In Eq. (5),η is the efficiency of the
104.6 dB. The audible frequency range below 20 kHz and the high fre­ aperture of the parabolic reflector. In general, η is between 50% and
quency range above 80 kHz are filtered out. The newly designed filter is 60%. In this work, a parabolic reflector with a focal length of 100 mm
a second-order Chebyshev high-pass filter (center frequency = 20 kHz, and a diameter of 150 mm was fabricated by 3D printing technique.
0.3 dB ripple) that cuts out the audible sound band below 20 kHz, and a Substituting this into Eq. (5), the gain of the parabolic reflector was
fourth-order Chebyshev low-pass filter (center frequency = 80 kHz, calculated to be about 31.87 dB. In the calculation of Eq. (5), the
0.3 dB ripple) that removes the high-frequency range above 80 kHz. The diameter of the sensing area (hole) of the MEMS M/P was set to 8.5 mm
A/D part converts 20–80 kHz analog signals collected by the micro­ corresponding to the wavelength of the 40 kHz.
( )
phone into digital signals with sampling frequency of 256 kHz. The d2 − d1
signal processing unit performs a fast Fourier transform (FFT) operation FOV conicalhornguide = arctan (4)
2l
and the block length (number of sampling points) for FFT processing was
set to 1024. The signal processing unit (SPU) splits the frequency in the

Fig. 1. Block diagram of the newly designed ultrasonic leak detection module.

2
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Fig. 2. Designed and manufactured ultrasonic leak detector boards.

Fig. 3. Inherent response characteristics of the new ultrasonic leak detector.

Fig. 4. Diagram of the conical horn guide.

( )2 and manufactured ultrasonic leak detector, the ASTM (American Society

πD
GPR = η (5) for Testing and Materials) E1002–05 Class II equipment verification
λ
criterion was used. It is a verification requirement for class II equipment
to detect ultrasonic sound signals from leaks of gas (dry nitrogen)
3. Experimental results
emitted from a nozzle (orifice) with a diameter of 0.2 mm under a gauge
pressure of 70 kPa at a distance of 5 m. The gas leak rate recommended
In this work, for evaluation of the performance of the newly designed

3
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Fig. 5. Diagram of the parabolic reflector horn guide.

in the ASTM E1002–05 document is 1.5std.atm.cm3 /sec. Fig. 6 shows the the ultrasonic leak detector with a parabolic reflector, and ULD#2 is the
outdoor experimental setup based on the ASTM E1002–05 Class II ultrasonic leak detector with a conical horn guide. Fig. 7 is a picture of
criteria. The outdoor verification experiment assumed leakage detection the outdoor test scene. The pressure regulator attached to the nitrogen
of pipes installed on the roadside of the Yeosu Industrial Complex. The cylinder was adjusted so that the discharge gas flow rate of the leakage
nitrogen gas contained in a bombe passed through a tube (6 mm in mock-up nozzle (diameter 0.1 mm) was 100 mL/min. As shown in the
diameter, ~ 10 m long) and was discharged through a nozzle (0.1 mm test setup, this was checked by watching the flow-gauge scale attached
diameter) attached to the pipe surface. next to the leak nozzle. The 0.1 mm nozzle used was a commercial
A pressure gauge was mounted near the nozzle, and the pressure of product. A commercial 0.1 mm nozzle (PC6-M5M-0.1) was attached by
the gas discharged through the 0.1 mm nozzle was measured. In addi­ adding a 5 mm tap on a cylindrical aluminum pipe with a thickness of
tion, the prototype ultrasonic leak detector was placed at a distance of ~ 3 mm and a diameter of 150 mm. Measurements were carried out in two
5 m (line-of-sight) distance from the nozzle. Measurement data for the ways. First, background noise was measured. With the bombe valve
ultrasonic leak detector was stored in a PC data logger (located inside closed, environmental noise in the direction of the 0.1 mm nozzle
the container sheltering the equipment) through an RS-232 C interface attached to the pipe surface was measured. Then, the ultrasonic leakage
cable. In Fig. 6, ULD#1, located on the roof of an outdoor container, is sound was measured. The valve attached to the bombe was opened, and

Fig. 6. Experimental setup for evaluation of the performance of the ultrasonic leak detection unit.

4
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Fig. 7. Photograph of the outdoor test setup.

the nitrogen gas discharge regulator pressure was adjusted so that the leakage ultrasonic energy from a 0.1 mm nozzle at a flow rate of 100 cc/
flow gauge scale near the 0.1 mm nozzle attached to the pipe surface min at a distance of ~ 6.5 m. Fig. 8 is the intensity measured at 38 kHz
was about 100 cc/min. Each leak measurement was performed for about by ultrasonic leak detector ULD #1 with a parabolic reflector. Fig. 9 is
two hours, and the rest of the time was spent on background noise the 32 kHz intensity component of ULD#2 measurements with a
measurement. 100 mm conical horn guide. As shown in Fig. 8, in the case of an ul­
Figs. 8 and 9 show measurements of outdoor leak signal by ultrasonic trasonic leak detector with a parabolic reflector, the background signal
leak detectors (ULD#1 with parabolic reflector and ULD#2 with conical and the leak signal are clearly distinguished. In most leak tests, it can be
horn guide). In the figures, the X-axis represents the number of mea­ seen that the minimum intensity of the leakage signal is greater than the
surements, and the Y-axis is the ultrasonic energy at 38 kHz and 32 kHz. maximum intensity of the background noise. On the other hand, in the
From 5 PM on November 6th to 9 AM on November 9th, more than case of ULD #2 with conical horn guide, the minimum intensity of the
20,000 measurements were taken at sampling intervals of ten seconds. leakage signal is less than the maximum intensity of the background
In the figures, the blue distribution curve is the raw signal, and the red noise. It can be explained that the sensitivity of ULD #2 with conical
color is the RMS of the measurement signal. horn guide is lower than that of ULD #1 with parabolic reflector. In the
maximum sensitivity frequency component, the signal-to-noise ratio
1∑ n
UERMS = UEi (6) (SNR) of ULD #1 with parabolic reflector is 4.176 dB and the SNR of
n i=1 ULD #2 with conical horn guide is 1.494 dB. In Eqs. 7 and 8, LERMS is the
As shown in Figs. 8 and 9, 4 leak measurement tests were performed ultrasonic leak energy emitted at 100 cc/min from a 0.1 mm nozzle. The
for about 2.5 days. Leak measurement tests are the results of measuring term UERMS is the natural noise of the test environment. As a result of a

Fig. 8. Ultrasonic (38 kHz) signal measurements (ULD #1 with Parabolic Reflector).

5
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Fig. 9. Ultrasonic (32 kHz) signal measurements (ULD #2 with conical horn guide).

comparative SNR calculation at maximum sensitivity frequency, the frequencies of the two leak detectors are different for the same leak. The
sensitivity of ULD#1 with the parabolic reflector (38 kHz) was about 2.8 ultrasonic leak detector with a parabolic reflector has the maximum
times higher than that of ULD #2 with the conical horn guide (32 kHz). sensitivity at 38 kHz, and the ultrasonic leak detector with the conical
⃒ ⃒ horn guide shows the maximum sensitivity at 32 kHz. The shape of the
⃒ LERMS ⃒
SNRULD#1,f =38kHz = 10log⃒⃒ ⃒ = 4.176dB (7) reflector (parabolic curved surface) was fabricated around a 40 kHz
UERMS ⃒ULD#1,f =38kHz
reference frequency. When the signal of the 40 kHz reference frequency
⃒ ⃒ is incident on the reflector, it is focused on one point of the sensor while
⃒ LERMS ⃒
SNRULD#2,f =32kHz = 10log⃒⃒ ⃒ = 1.494dB (8) having almost the same phase. On the other hand, when the incident
UERMS ⃒ULD#2,f =32kHz frequency is out of the reference frequency, distortion is generated and
Figs. 10 and 11 show the measurement results of the two types of focused on various surfaces of the sensor. It is thought that the difference
ultrasonic leak detectors newly designed and manufactured. In these from the conical horn guide is due to the difference in the shape of the
figures, the X-axis represents the frequency, and the Y-axis represents reflector, the installation position, and the observation FOV (field-of-
the ultrasonic energy. The black distribution curve shows the natural view). The fact that the maximum sensitivity frequency varies depend­
outdoor-noise measurement results over 2.5 days. The red distribution ing on the horn attached to the ultrasonic leak detector needs further
curve is a measurement of the leakage ultrasound discharged at 100 cc/ analysis.
min from a 0.1 mm nozzle. Fig. 10 is the measurement result of the ULD The outdoor measurement experiment was performed 19 times.
#1 with parabolic reflector, and Fig. 11 is the measurement result of the Tables 1 and 2 summarize the root mean square (RMS), Xrms , energy E,
ULD #2 with conical horn guide. From the measurement results shown variance Xs and SNR of the two types of ultrasonic leak detectors
in Figs. 10 and 11, ULD #1 with the parabolic reflector shows very good designed and manufactured in this study. The RMS, energy, variance,
response characteristics; enough to distinguish leak signals (on a graph) and SNR were defined as follows.
with the naked eye. It can be seen that the ULD #2 with conical horn 1∑ n ⃒ ⃒
guide also has sufficient sensitivity to distinguish the leakage signal from Xrms = ⃒UEleak − UEbackgroundnoise ⃒
i
(9)
n i=1
the background noise. In Figs. 10 and 11, the maximum sensitivity

Fig. 10. Ultrasonic intensity distribution in an outdoor environment (ULD #1 with parabolic reflector).

6
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Fig. 11. Ultrasonic intensity distribution in an outdoor environment (ULD #2 with conical horn guide).

Table 1 1∑ n ⃒ ⃒
Xs = (⃒UEleak − UEbackgroundnoise ⃒i − Xrms )2 (10)
Summary of the feature calculation results (ULD #1 with parabolic reflector). n i=1
Leak Test RMS Variance Energy SNR B/G Measurements √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
05:51–07:25, 0.0509 0.0046 0.079 4.768 07:25–18:14 28 Oct 1∑ n ⃒ ⃒
E= ⃒UEleak − UEbackgroundnoise ⃒2 (11)
28 Oct n i=1 i

05:22–06:55, 0.0389 0.0023 0.062 4.222 18:20 28 Oct - 05:22

29 Oct 31 Oct ∑
n
17:22–18:50, 0.0365 0.0020 0.058 4.052 |UEleak |i
30 Oct SNR = 10log ∑ i=1
(12)
n ⃒ ⃒
06:00–07:30, 0.0476 0.0032 0.074 4.541 07:47 31 Oct - 05:52 3 ⃒UEbackgroundnoise ⃒
31 Oct Nov i=1
i
07:19–09:30, 0.0526 0.0035 0.079 5.463 05:58 3 Nov - 16:24 6
3 Nov Nov In Tables 1 and 2, the first column indicates the date and time of the
07:10–09:27, 0.0489 0.0031 0.074 5.241 leak detection test, and the last column indicates the background noise
4 Nov
measurement time interval. According to the calculation results of the
05:43–07:46, 0.0385 0.0023 0.060 4.541
5 Nov
statistical features of the leak detection tests performed 19 times (with
17:23–18:53, 0.0397 0.0024 0.063 4.543 17:22 6 Nov - 09:23 9 parabolic reflector) and 18 times (with the conical horn guide), the
6 Nov Nov average SNR of ULD#1 with the parabolic reflector was 4.97 dB, and the
06:04–07:52, 0.0425 0.0027 0.067 4.739 average SNR of ULD#2 with the conical horn guide was 1.89 dB. For
7 Nov
outdoors, the ultrasonic leak detectors designed and manufactured in
05:35–07:29, 0.0517 0.004 0.081 5.319
8 Nov this paper can be said to satisfy the ASTM E1002–05 class II equipment
05:13–07:30, 0.0474 0.0031 0.073 5.060 verification criteria.
9 Nov
05:14–07:13, 0.0517 0.0035 0.078 5.453 09:23 9 Nov - 17:59 11 4. Conclusions
10 Nov Nov
05:15–07:25, 0.0524 0.0035 0.079 5.495
11 Nov An ultrasonic leakage detector was designed and manufactured to
05:16–07:20, 0.0560 0.0041 0.085 5.884 17:59 11 Nov - 05:12 detect pipe leakage remotely. This involved design of a sound focusing
12 Nov 15 Nov horn and a parabolic reflector. Next, an ultrasonic leak detector module
16:54–18:48, 0.0519 0.0036 0.079 5.641
with a parabolic reflector (focal length 100 mm, diameter 150 mm) was
12 Nov
15:19–17:20, 0.0377 0.0020 0.058 4.681 fabricated, along with a second ultrasonic leak detection module with a
13 Nov conical horn guide (length 100 mm, angle 0.166 radian). The perfor­
05:26–07:15, 0.0468 0.0029 0.071 5.322 mance of these devices was tested and evaluated in an outdoor envi­
14 Nov ronment in accordance with the ASTM E1002–05 class I&II equipment
15:40–17:40, 0.0433 0.0026 0.067 5.087
14 Nov
verification standards. The flow rate of 100 cc/min required by the
05:27–07:37, 0.0349 0.0016 0.054 4.338 05:15 15 Nov - 05:57 verification standard was implemented experimentally. The ultrasonic
15 Nov 16 Nov leak signal emitted at a flow rate of 100 cc/min from a 0.1 mm diameter
nozzle was tested at a distance of about 6.5 m. The tests were performed
about 20 times. As a result of the test evaluation, the average SNR of the
ultrasonic leak detector with the parabolic reflector was 4.97 dB, and
the average SNR of the ultrasonic leak detector with the conical horn
guide was 1.89 dB. The two ultrasonic leak detectors designed and
manufactured in this study satisfied ASTM E1002–05 class II equipment
verification requirements.
As a result of visiting the plant site, there is rarely a single pipe, and
multiple pipes are installed over long distances. In this case, with a

7
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Table 2
Summary of feature calculation results (ULD #2 with conical horn guide).
Leak Test RMS Variance Energy SNR B/G Measurements

05:51–07:25, 0.02228 0.00111 0.040086 2.336239 07:25–18:14 28 Oct

28 Oct
05:22–06:55, 0.02143 0.001087 0.03932 2.317686 18:20 28 Oct –
29 Oct 05:22 31 Oct
17:22–18:50, 0.013224 0.000467 0.025339 1.569014
30 Oct
07:19–09:30, 0.02009 0.000937 0.036609 2.181826 05:57 3 Nov –
3 Nov 18:48 5 Nov
07:10–09:27, 0.018618 0.000842 0.034481 2.054245
4 Nov
05:43–07:46, 0.019911 0.000962 0.036863 2.166498
5 Nov
17:23–18:53, 0.011349 0.000349 0.021862 1.311952 17:22 6 Nov –
6 Nov 09:23 9 Nov
06:04–07:52, 0.01278 0.000466 0.025092 1.456004
7 Nov
05:35–07:29, 0.013286 0.000504 0.026084 1.505397
8 Nov
05:13–07:30, 0.012235 0.000337 0.022071 1.398482
9 Nov
05:14–07:13, 0.017057 0.000678 0.031131 2.195762 09:23 9 Nov –
10 Nov 17:59 11 Nov
05:15–07:25, 0.01902 0.000844 0.034729 2.390307
11 Nov
05:16–07:20, 0.019811 0.0009 0.035955 2.242951 17:59 11 Nov -
12 Nov 18:59 14 Nov
16:54–18:48, 0.018323 0.000792 0.033587 2.112853
12 Nov
15:19–17:15, 0.012845 0.000445 0.024689 1.578991
13 Nov
05:26–07:15, 0.01729 0.0007 0.031606 2.010712
14 Nov
15:19–17:20, 0.014184 0.000518 0.026823 1.717478
14 Nov
05:27–07:37, 0.011878 0.000338 0.021899 1.386835 05:15 15 Nov –
15 Nov 05:57 16 Nov

single ultrasonic leak detector, the leakage On/Off of the observation References
zone (the area of pipes that the ultrasonic leak detector sees) can be
specified. However, it is not possible to specify which pipe is leaking out [1] A.A. Pollock, S.-Y.S. Hsu, Leak detection using acoustic emission, J. Acoust. Emiss.
Vol. 1 (No. 4) (1982) 237–243.
of several pipes. In order for the ultrasonic leak detector developed in [2] Alan S. Bandes, Detect Leaks with Ultrasound”, Chemical Engineering, [online]
this study to have plant applicability, it is thought that a solution to this Available: Dec. 2002, http://www.che.com.
is needed. In a future study, we intend to devise a solution that can [3] Tetsuji Tani, Toru Nagasako, Yasunari Fujimoto, Tadashi Iokibe, Proposal and
Application of Chaos Information Criteria for High-pressure Gas Leak Detection in
specify the leaking pipe among several pipes. Petroleum Refining Plant, pp387–390, 21st Fuzzy System Symposium, Sept. 2005.
[4] Tetsuji Tani, Toru Nagasako, Yasunari Fujimoto, Tadashi Iokibe, and Toru
Author statements Yamaguchi, Chaos Information Criteria to Detect High-pressure Gas Leak in
Petroleum Refining Plant, SICE-ICASE Int. Joint Conf. pp5415–5418, Oct 2006.
[5] Tetsuji Tani, Yasunari Fujimoto, Toru Yamaguchi, “Experiments of high-pressure
Jae Cheol Lee, Yu Rak Choi and Jai Wan Cho have reviewed and gas leak detection system applied chaos information criteria and its results
approved the final version of manuscript being submitted. Authors (Japanese)”, 23rd Fuzzy System Symposium, Aug. 2007.
[6] Tetsuji Tani, Yasunari Fujimoto, Toru Yamaguchi, High-pressure Gas Leak
warrant that the research article is author’s original work, hasn’t Detection Applied Chaos Analysis to Acoustic Sound Data Transformed by Inverse
received prior publication and isn’t under consideration for publication Filter (Japanese), 24th Fuzzy System Symposium, Sep. 2008.
elsewhere. [7] Tetsuji Tani, Toru Nagasako, Yasunari Fujimoto, Toru Yamaguchi, Detection of
high-pressure gas leak in petroleum refining plant applying chaos information
criteria (Japanese), Jpn. Soc. Fuzzy Theory Intell. Inform. Vol.21 (No.1) (2009)
Declaration of Competing Interest 77–89, pp77-89.
[8] R.P. Cruz, F.V. Silva, A.M. Fileti, et al., Machine learning and acoustic method
applied to leak detection and location in low-pressure gas pipelines, Clean.
The authors declare that they have no known competing financial Technol. Environ. Policy (2020) 1–12.
interests or personal relationships that could have appeared to influence [9] Lei Li, K. Yang, X. Bian, et al., A gas leakage localization method based on a virtual
the work reported in this paper. ultrasonic sensor array, Sensors 19 (14) (2019).
[10] Mengjie Xu, Tao Wang, Study on gas leakage localization method based on
ultrasonic sensor area array, Int. Conf. Adv. Intell. Mechatron. (2017) 136–141.
Data Availability [11] R.B. Santos, E.O. De Sousa, F.V. Da Silva, S.L. Da Cruz, A.M.F. Fileti, Detection and
on-line prediction of leak magnitude in a gas pipeline using an acoustic method
and neural network data processing, Braz. J. Chem. Eng. 31 (1) (2014) 145–153.
Data will be made available on request. [12] Jian Li, et al., High-sensitivity gas leak detection sensor based on a compact
microphone array, Measurement Vol.174 (2021).
Acknowledgements [13] Shen Wang, Xuefeng Yao, Aeroacoustics measurement of the gas leakage rate for
single hole, Rev. Sci. Instrum. Vol. 91 (2020).
[14] Rui Xiao, et al., Leak detection of gas pipelines using acoustic signals based on
This work was supported by a National Research Council of Science wavelet transform and Support Vector Machine, Measurement Vol. 146 (2019)
& Technology (NST) grant by the Korea government (MSIP) (No. CRC- 479–489.
15-05-ETRI)

8
J.C. Lee et al. Sensors and Actuators: A. Physical 349 (2023) 114061

Jae Cheol Lee Graduated from the Department of Electronics Engineering, Kyungpook Research Division, Korea Atomic Energy Research Institute. His research topic is the
National University, Master’s Degree. Currently, he is working in the Smart Structural development of AI technology for the application of nuclear power plants.
Safety and Prognosis Research Division, Korea Atomic Energy Research Institute. His
research topics are the design/manufacturing of ultrasonic leak sensors and the develop­
Jai Wan Cho Graduated from the Department of Electronics Engineering, Kyungpook
ment of sensors with built-in AI inference engine.
National University, Master’s Degree. Currently, he is working in the Smart Structural
Safety and Prognosis Research Division, Korea Atomic Energy Research Institute. His
Yu Rak Choi Graduated from the Department of Computer Science, Chungnam National research topic is the development of sensor system V&V(verification and validation)
University, Ph. D.Currently, he is working in the Smart Structural Safety and Prognosis technology for nuclear power plant