Professional Documents
Culture Documents
Article
Comparative Analysis of Ultrasonic and Traditional Gas-Leak
Detection Systems in the Process Industries: A Monte
Carlo Approach
Joon Hyuk Lee 1,2 , Youngsik Kim 3 , Inkwon Kim 4 , Seok Bum Hong 2, * and Hong Sik Yun 5, *
1 Fire Insurers Laboratories of Korea Affiliated with Korean Fire Protection Association, 1030,
Gyeongchung-daero, Ganam-eup, Yeoju 12661, Republic of Korea
2 Interdisciplinary Program for Crisis, Disaster and Risk Management, Sungkyunkwan University,
Suwon 16419, Republic of Korea
3 Stratio, Inc., Seongnam-si 13449, Republic of Korea
4 Sound Camera Business/Software Laboratory, SM Instruments, Inc., Daejeon 34109, Republic of Korea
5 School of Civil and Architectural Engineering, Sungkyunkwan University, Suwon 16419, Republic of Korea
* Correspondence: how098@skku.edu (S.B.H.); yoonhs@skku.edu (H.S.Y.)
Abstract: Gas leaks can cause disasters at process sites, including fires and explosions, and thus,
effective gas-leak detection systems are required. This study investigated the limitations of con-
ventional detectors and introduced an innovative ultrasonic sensor-based approach for continuous
monitoring. A new configuration for a stationary remote ultrasonic gas-leak monitoring system is
proposed. The selected material was 1-Butene. The detection probability was assessed through a
simulation based on a gas-leak scenario, detailing the selection criteria for leak sites and simulation
conditions. Computational fluid-dynamics (CFD) simulations were used to evaluate the detection
capability of the existing system, whereas Monte Carlo simulations were used to compare it with
the proposed ultrasonic system. The CFD simulation was performed by setting the lower detection
limit of the concentration-measurement-type gas detector to 600 ppm, and the leak-detection time
was approximately 8.895 s. A Monte Carlo simulation accounting for wind direction showed that
the existing system’s average detection probability was 3.41%. By integrating the existing and new
systems, the detection probability increased by 96.58% and 3.92% compared with that using only the
Citation: Lee, J.H.; Kim, Y.; Kim, I.; concentration-measurement and ultrasonic measurement methods, respectively. These rates provide
Hong, S.B.; Yun, H.S. Comparative valuable insights for future applications in oil refining and petrochemical industries.
Analysis of Ultrasonic and Traditional
Gas-Leak Detection Systems in the Keywords: oil refining and petrochemical industries; gas-leak detection; ultrasonic; fixed monitoring
Process Industries: A Monte Carlo system; computational fluid dynamics; Monte Carlo simulation; detection probability
Approach. Processes 2024, 12, 67.
https://doi.org/10.3390/pr12010067
cause lasting damage [9]. Early detection and rapid response to chemical leaks are necessary
to prevent major damage, and technologies to address these issues are being developed in
various ways. According to a statistical survey by the National Institute of Chemical Safety,
most chemical accidents have occurred on a large scale [10]. Fires caused by the leakage of
combustible gases in chemicals continue to occur. In 2022, fire damage caused by gas leaks
in Korea accounted for the largest amount of property loss in the past 10 years; apart from
property loss, these casualties continue to occur as shown in Table 1 [11].
• Concentration measurement: this is a traditional method that is widely used not only in
oil refining and petrochemical plants but also in other factories, offices, and residential
facilities. The gas-detection module determines whether the gas concentration in
the proposed environment exceeds a specified range. This includes the use of gas
detection systems to identify gases that may escape into the air [20];
• Thermal imaging: leaks can cause changes in temperature, and thermal cameras can
identify hotspots. Infrared optical gas imaging (IOGI) enables sensitive remote sensing
of toxic gases. This technique has been widely applied to leak detection, localisation,
and visualisation [21–23];
• Acoustic sensing: acoustic sensors detect sounds generated by leaks, which can be high-
frequency ultrasonic noises that are otherwise inaudible [21]. Recently, technologies
have been developed to detect gas leaks in pipeline systems using acoustic methods.
Wave propagation in pipelines is a major factor in acoustic methods when leaks occur,
and the friction between the pipe walls creates a pressure balance in the pipe. Further
tests and simulations must be performed to obtain rapid results and location estimates
for leaks using acoustic methods [24]. This method detects gas leaks by adhering a
sensor to a pipe, and some methods can detect leaks at a distance. The long-distance
measurement method uses equipment that can receive ultrasonic waves, such as a
microphone sensor, to measure gas leakage from a distance and even estimate the
location of the leak [25]. According to our recent interviews with workers at oil
refineries and petrochemical plants in Korea, the use of ultrasonic cameras to measure
distances is increasing. However, ultrasonic cameras in plants are portable and require
the expertise of skilled operators.
In previous studies, gas detection has primarily focused on traditional methods (pres-
sure, flow, concentration, thermal imaging, etc.). The introduction of ultrasonic detectors
capable of acoustic detection has been considered necessary. Although numerous studies
have been conducted on fire detection, there is a need for progressive research on the
detection of gas leaks that cause fires [26–30]. Currently, research on the detection of
1-butene leakage is lacking. Our study conducted simulations with 1-butene, which is
used in petrochemical processes, as the leak target, and thus, it serves as a good reference.
Several studies have been conducted on methods to ventilate combustible gases or enhance
safety management and response systems. However, ventilation and management alone
cannot completely eliminate the risk of explosions, and thus, advancements in the field of
detection are necessary [31–34].
Figure 1.
Figure Generation of
1. Generation of leakage
leakage signals
signals and
and measurement
measurement using
using sensor
sensor arrays.
arrays.
These signals
These signals are
are transmitted
transmitted through
through the
the air
air and
and reach
reach the
the microphone
microphone sensor
sensor array
array
in the ultrasonic camera. A sensor-array-based ultrasonic gas-leak location method
in the ultrasonic camera. A sensor-array-based ultrasonic gas-leak location method is thus is thus
used to
used todetect
detectand
andlocate
locatethe
thegenerated
generated ultrasonic
ultrasonic leaks.
leaks. TheThe location
location of leak
of the the leak
pointpoint
can
be determined using only the arrival-time difference between the leak point and theand
can be determined using only the arrival-time difference between the leak point sensorthe
sensor [39].
[39].
3.2. Differences between Ultrasonic and Traditional Methods
3.2. Differences between Ultrasonic and Traditional Methods
Gas concentration-measurement methods used in oil refineries and petrochemical
Gas
plants concentration-measurement
may methodsofused
have blind spots in the detection plantinleaks.
oil refineries andofpetrochemical
In the case oil refineries
plants may have blind spots in the detection of plant leaks. In the case of oil refineries
and petrochemical plants installed outside, there is a high possibility that the gas diffusion and
petrochemical
path is outsideplants installed range
the installation outside, there
of the is a high
detector owingpossibility thatair
to outdoor the gas diffusion
currents; in this
path is outside the installation range of the detector owing to outdoor air currents;
case, leak detection is extremely difficult. Particularly, gases that are lighter than air intend
this
to disperse into the atmosphere, making their accumulation and detection more difficult.
Additionally, traditional concentration-measurement methods can only detect certain gases.
As oil refineries and petrochemical plants handle various types of gas, it is difficult to detect
all of them using traditional methods, which only detect specific gas types.
The ultrasonic detection method allows for monitoring a space from a distance and is
suitable for detecting gas leaks in vast oil refineries and petrochemical plants, as explained
in Section 2.2. As the ultrasonic waves generated in the event of a gas leak are received
by the microphone sensor, indicating their location, the gas-leak site can be immediately
identified. The traditional method cannot specify the location even if a gas leak is detected;
therefore, additional confirmation of the cause of the surrounding leak is necessary. Ul-
to detect all of them using traditional methods, which only detect specific gas types.
The ultrasonic detection method allows for monitoring a space from a distance and
is suitable for detecting gas leaks in vast oil refineries and petrochemical plants, as ex-
plained in Section 2.2. As the ultrasonic waves generated in the event of a gas leak are
Processes 2024, 12, 67 received by the microphone sensor, indicating their location, the gas-leak site can be im-
5 of 26
mediately identified. The traditional method cannot specify the location even if a gas leak
is detected; therefore, additional confirmation of the cause of the surrounding leak is nec-
essary. detection
trasonic Ultrasonicdoes
detection does not
not require suchrequire such procedures,
procedures, therebya enabling
thereby enabling a rapid
rapid response.
response. In addition, regardless of the type of leaking gas, all gases that generate ultra-
In addition, regardless of the type of leaking gas, all gases that generate ultrasonic waves
sonic waves can be detected. Ultrasonic detection methods are currently developed
can be detected. Ultrasonic detection methods are currently developed only for portable only
for portable use; therefore, building a fixed monitoring system would be a more
use; therefore, building a fixed monitoring system would be a more effective gas-leak effective
gas-leak paradigm.
detection detection paradigm.
3.3.Development
3.3. DevelopmentofofaaFixed
FixedGas-Leak
Gas-LeakMonitoring
MonitoringSystem
SystemUtilising
UtilisingUltrasonic
UltrasonicTechnology
Technology
Figure2a2a
Figure shows
shows thethe
3D3D design
design andand prototype
prototype ofultrasonic
of the the ultrasonic
camera.camera. Ultrasonic
Ultrasonic cam-
cameras,
eras, whichwhich
are anare an important
important elementelement of gas-leak
of gas-leak monitoring
monitoring systems,
systems, are designed
are designed to
to min-
minimise
imise the product
the product size forsize for efficient
efficient space utilisation,
space utilisation, as shown asinshown
Figurein 2b.Figure 2b. The di-
The dimensions
ofmensions of this prototype
this prototype are are approximately148
approximately 148 mmmm× 188× mm 188× 60mm × 60× height
mm (width mm
(width × height
× thickness), and×itthickness),
weighs approximately
and it weighs1.6 kg.
approximately 1.6 kg.
(a) (b)
Figure2.2.Three-dimensional
Figure Three-dimensional (3D)
(3D) design
design of the
of the ultrasonic
ultrasonic camera
camera prototype
prototype and and actual
actual shape.
shape. (a)
(a) 3D
3D design of the prototype; (b) actual product of the developed prototype.
design of the prototype; (b) actual product of the developed prototype.
Becausemost
Because mostoiloilrefineries
refineries
andand petrochemical
petrochemical plants
plants areare installed
installed outdoors,
outdoors, the the cam-
camera
era has a waterproof design. It was developed and verified according
has a waterproof design. It was developed and verified according to IEC 60529 standards to IEC 60529 stand-
toards to achieve
achieve an IP ofan55IPorofhigher,
55 or which
higher,iswhich is the waterproofing
the highest highest waterproofing ratingcurrent
rating among among
current ultrasonic
ultrasonic cameras. cameras. This development
This development was intended
was intended to ensure to that
ensurethethat the could
device device
could withstand
withstand rain orwhen
rain or snow snowinstalled
when installed
outdoors.outdoors.
InInthis
thisstudy,
study,simulations
simulationswere wereperformed
performedusing
usinga ahighly
highlysensitive
sensitivegas-leak
gas-leakdetection
detection
sensorbased
sensor basedon onaa112-microphone
112-microphonearray. array.Laboratory-scale
Laboratory-scalegas-leak
gas-leaktests
testsusing
usingthis
thisultra-
ultra-
soniccamera
sonic cameraverified
verifiedthe thehigh
highsensitivity
sensitivityofofthe
thesensor
sensortotodetect
detectminute
minutegasgasleaks
leaksatataarate
rate
ofofless than 1.67 mL/s. When the leakage from an orifice was less than
less than 1.67 mL/s. When the leakage from an orifice was less than 1.67 mL/s, it was1.67 mL/s, it was
detected
detectedatataadistance
distanceofofapproximately
approximately5000 5000mm, withΦΦ0.2
mm,with 0.2mm,
mm,satisfying
satisfyingthetheASTM
ASTMEE
1002-11
1002-11standard
standard[40].
[40].TheTheresults
resultsofoftests
testsconducted
conductedatatan anoutdoor-demonstration
outdoor-demonstrationtest testsite
site
within
withinthe theFire
FireInsurers
InsurersLaboratories
LaboratoriesofofKoreaKorea(FILK)
(FILK)ininYeoju-si,
Yeoju-si,where
wheregas-leak
gas-leakfacilities
facilities
(vessels, pipes, etc.) were installed and demonstration tests performed, confirmed that gas
leakage can be detected from a distance using ultrasonic technology.
The portable ultrasonic gas-leak measurement equipment (camera) was used by
inspectors and workers to identify leaks by observing the heat map displayed on a screen
attached to the camera. The primary function of the camera in a fixed surveillance system is
to receive ultrasonic waves through a microphone sensor array, and in this case, it excludes
the camera screen, which is installed separately and connected to the camera using a wire.
Unlike portable devices, constant monitoring is possible inside a control room using a
monitor connected via a network. This system is currently in use in the gas-filled process
pipelines at a polymer manufacturing plant in the Yeosu Petrochemical Industrial Complex.
Constant surveillance in a control room that handles key management functions in oil
refineries and petrochemical plants enables a more rapid response to leaks. This study
predicts the detection probabilities of the conventional and ultrasonic gas-leak detectors
cludes the camera screen, which is installed separately and connected to the camera using
a wire. Unlike portable devices, constant monitoring is possible inside a control room us-
ing a monitor connected via a network. This system is currently in use in the gas-filled
process pipelines at a polymer manufacturing plant in the Yeosu Petrochemical Industrial
Processes 2024, 12, 67
Complex. Constant surveillance in a control room that handles key management functions
6 of 26
in oil refineries and petrochemical plants enables a more rapid response to leaks. This
study predicts the detection probabilities of the conventional and ultrasonic gas-leak de-
tectors and examines the feasibility of introducing a mixed detection system into oil refin-
and examines the feasibility of introducing a mixed detection system into oil refineries and
eries and petrochemical plants.
petrochemical plants.
4. Methods
4. Methods
Based on
Based on the
the test
test facilities
facilities of
of aa petrochemical
petrochemical plant,
plant, aa simulation
simulation was was performed
performed con-con-
sidering 1-butene,
sidering 1-butene,whichwhichisisa arawrawmaterial
materialasas well
well as as
an an intermediate
intermediate andand byproduct
byproduct in
in the
the actual process of the Yeosu Petrochemical Plant. Under the same
actual process of the Yeosu Petrochemical Plant. Under the same conditions, the detection conditions, the de-
tection performance
performance of the ultrasonic
of the ultrasonic monitoring
monitoring systemsystem described
described in Section
in Section 3 was3compared
was com-
with that of existing concentration-measurement methods for 1-butene leaks. To leaks.
pared with that of existing concentration-measurement methods for 1-butene compare To
compare and analyse the detection performance, the simulation utilised
and analyse the detection performance, the simulation utilised two tools: computational two tools: com-
putational
fluid fluid (CFD)
dynamics dynamics and (CFD)
systemand system (Monte
dynamics dynamics (Monte
Carlo Carlo simulation).
simulation). In addition, In the
ad-
dition, theperformance
detection detection performance of both technologies
of both technologies was derivedwas and
derived and analysed.
analysed. A com-
A comparison
parison between
between the existingthe method
existing and
method and the proposed
the proposed ultrasonicultrasonic
system will system will be
be a good a good
reference
reference when applying the technology to gas leakages in the future.
when applying the technology to gas leakages in the future. In addition, the combination In addition, the
combination of the two systems can be realistically applied to oil refineries
of the two systems can be realistically applied to oil refineries and petrochemical plants, and petro-
chemical
and plants, results
the analysis and thecan analysis
be usedresults can be used
as a reference forasactual
a reference for actual applications.
applications.
flowchart in Figure 3 illustrates the progress in the research on gas-leak detection
The flowchart
methods. ItItbegins
beginsbyby selecting
selectingthethescope of analysis,
scope which
of analysis, includes
which the traditional,
includes ultra-
the traditional,
sonic, and and
ultrasonic, combined
combined detection techniques.
detection Subsequently,
techniques. a CFD
Subsequently, simulation
a CFD was con-
simulation was
conducted
ducted on theon the existing
existing system
system by incorporating
by incorporating 3D 3D modelling
modelling andand numerical
numerical analysis.
analysis. In
In
thethe comparison
comparison phase,system-dynamics
phase, system-dynamicssimulations
simulationswere
wereusedusedto to design
design the variables
and parameters, which were further analysed using Monte Carlo simulations. Finally, the
integrated performance
performance of of the conventional
conventional and and ultrasonic
ultrasonic gas
gas detectors
detectors was evaluated to
derive the research
research outcomes.
outcomes.
study.
Figure 3. Flowchart of the study.
Figure 4. Initial design of the equipment layout plan for the simulation.
Figure 4. Initial design of the equipment layout plan for the simulation.
Based
Based on
on the
the initial
initial simulation
simulation design
design shown
shown in
in Figure
Figure 4,
4, modifications
modifications and
and additions
additions
were included
Processes 2024, 12, x FOR PEER REVIEW
were includedin
inthe
thedrawing
drawingby
byapplying
applyinga gas-leak
a gas-leak scenario,
scenario, as as shown
shown in the
in the 3D 3D 8draw-
of 27
drawings
ings in Figure
in Figure 5, which
5, which werewere
then then
used used forCFD
for the the CFD simulation.
simulation.
(a) (b)
Figure 5. Three-dimensional (3D) design for computational fluid dynamics (CFD) and system-dy-
Figure 5. Three-dimensional (3D) design for computational fluid dynamics (CFD) and system-
namics simulation. (a) 3D design for the simulation viewed from the top; (b) 3D design for the sim-
dynamics simulation. (a) 3D design for the simulation viewed from the top; (b) 3D design for the
ulation viewed from the top left.
simulation viewed from the top left.
The location was selected based on the outdoor-demonstration test site at the Fire
The location was selected based on the outdoor-demonstration test site at the Fire
Insurers Laboratories of Korea (FILK). The pressure inside the pipe where the leak oc-
Insurers Laboratories of Korea (FILK). The pressure inside the pipe where the leak occurred
curred was 2.2 bar, the leak diameter was 6.9 mm, and the leak rate was 0.00217 kg/s.
was 2.2 bar, the leak diameter was 6.9 mm, and the leak rate was 0.00217 kg/s. When the
When theperformed,
test was test was performed,
the average thewind
average wind direction
direction and speed and speed
were were approximately
approximately 1.2 m/s
1.2
from the south, and the weather was clear. The ultrasonic gas-leak detection time time
m/s from the south, and the weather was clear. The ultrasonic gas-leak detection was
was approximately
approximately 8 s, 8including
s, including a 3as3monitoring-video
s monitoring-videoprocessing
processingtime
timefor
for the
the integrated
integrated
surveillance
surveillance system
system andand aa 55 ss leak-determination
leak-determination time.
time. Simultaneously,
Simultaneously, ultrasonic
ultrasonic waves
waves
were generated and immediately received by the microphone sensor, and the leak was
immediately displayed on the screen. However, to determine whether it was a real leak or
a surrounding obstacle that could temporarily generate ultrasonic waves, such as wind,
the leak-detection decision time was arbitrarily set to 5 s. This is an expression of the in-
tention to identify a leak if the ultrasonic waves generated from the pipe continue for 5 s.
Processes 2024, 12, 67 8 of 26
were generated and immediately received by the microphone sensor, and the leak was
immediately displayed on the screen. However, to determine whether it was a real leak or
a surrounding obstacle that could temporarily generate ultrasonic waves, such as wind,
the leak-detection decision time was arbitrarily set to 5 s. This is an expression of the
intention to identify a leak if the ultrasonic waves generated from the pipe continue for
5 s. Depending on the site and situation, this time may be changed. For example, if the
inspector is skilled in judging an onsite situation, the time required to determine a leak
can be reduced or eliminated. Under these conditions, leak detection was possible up to
approximately 23 m, a distance exceeding 20 m. Figure 6 shows a photograph of the test site
Processes 2024, 12, x FOR PEER REVIEW 9 of 27
within the FILK where gas leaks were detected at a distance. The outdoor-demonstration
test site, shown in Figure 6, was constructed based on Figure 5.
4.1.2. Gas-Leak
4.1.2. Gas-Leak Scenario
Scenario
On On1717July
July2021,
2021,a aleak
leakoccurred
occurredinina ahydrochloric
hydrochloricacid acidstorage
storagetank
tankatatthetheUlsan
Ulsan
Chemical
ChemicalPlant.
Plant.InInthis
thisaccident,
accident,the
thestored
storedhydrochloric
hydrochloric acidacid leaked owing to
leaked owing to the
the loosening
loosen-
ingof of
thethe
pipepipe flange
flange joint
joint at the
at the bottom
bottom of the
of the outdoor
outdoor hydrochloric
hydrochloric acid
acid storage
storage tank.
tank. The
The hydrochloric acid leaked from the flange joint, where the thread thickness
hydrochloric acid leaked from the flange joint, where the thread thickness decreased owing decreased
owing to long-term
to long-term chemical
chemical corrosion
corrosion of theof flange
the flange connection
connection bolt.bolt. Eleven
Eleven nearby
nearby resi-
residents
dents received
received treatment
treatment at hospital,
at the the hospital,
but but
somesome forests
forests andand crops
crops nearnear
thethe business
business sitesite
per-
perished. Expertshave
ished. Experts haveadvocated
advocatedthe theuse
useofofacid-resistant
acid-resistantmaterials
materialsin inprocesses
processesthat thathandle
han-
dle corrosive
corrosive substancesand
substances andsecure
secureworkspaces
workspacesfor formaintenance
maintenanceasaspreventive
preventivemeasures.
measures.In
Inaddition,
addition,regular
regularinspection
inspectionofofthe theflange
flangejoints
jointsisisimportant.
important.Similar
Similargasgasleaks
leakshave
haveoc-
curred frequently
occurred frequentlyininflanges
flangesduring
duringrecent
recentaccidents.
accidents.Table
Table2 2presents
presents the
thedata
datafrom
from Korea’s
Ko-
National
rea’s National FireFire
Agency,
Agency, showing
showing gas-leak
gas-leak accidents
accidents that have
that haveoccurred
occurredininpetrochemical
petrochem-
plants
ical plantsover
overthethe
past
pastfive years
five years[11].
[11].Examining
Examiningthe thecauses
causesof ofthe
theaccidents
accidents in in Table
Table 2,
2, it
it can be observed that most leaks occurred in pipelines (including pipes and valves), and a
a significant number occurred in joints such as flanges.
Leaking
Year Region Case of Accident Cause of Accident
Material
Jeonnam
2023 Facility failure RAW C9 + C Pipe burst
Yeosu
Ulsan Nam- Failure to comply with
2023 Acetonitrile Carelessness
gu safety standards
Processes 2024, 12, 67 9 of 26
Leaking
Processes Year Region
2024, 12, x FOR PEER REVIEW Case of Accident
Material
Cause of Accident 10 of 27
Many
Manypipes
pipesand
andconnecting
connectingequipment
equipmentininpetrochemical
petrochemicalplants
plantsleak
leakfrom
fromjoints,
joints,such
such
asasflanges, valves, and connectors. In [42], researchers selected a specific petrochemical
flanges, valves, and connectors. In [42], researchers selected a specific petrochemical
plant
plantininKorea,
Korea,analysed
analysedthe
thefrequency
frequencyofofleaks
leaksatatthat
thatplant,
plant,and
andfound
foundthat
thatthe
thefrequency
frequency
ofofflange
flange leaks was the highest among the pipeline joints. For a comparativeanalysis
leaks was the highest among the pipeline joints. For a comparative analysisofof
the
thedetection
detectionprobability
probabilityofofthe
theultrasonic
ultrasonicdetectors
detectorsconfigured
configuredininthe
theresearch
researcharea,
area,the
the
conventional gas leak and ultrasonic detectors were arranged as shown
conventional gas leak and ultrasonic detectors were arranged as shown in Figure 7.in Figure 7.
derived from the first law of thermodynamics, which links the heat transfer, work rate, and
kinetic energy in fluids.
→
∇·V = 0, (1)
→
dV → →
ρ = −∇ p + µ∇2 V + ρ g , (2)
dt
Processes 2024, 12, x FOR PEER REVIEW 12 of 27
dT
ρC p = ∇·(k∇ T ). (3)
dt
(a) (b)
Figure 8. Boundary conditions and mesh formation for the test environment. (a) Boundary condi-
Figure 8. Boundary conditions and mesh formation for the test environment. (a) Boundary conditions
tions of the flow area; (b) mesh formation for the test environment.
of the flow area; (b) mesh formation for the test environment.
Figure 9. Schematic of the Monte Carlo simulation procedure. (a) Example of random variables;
Figure 9. Schematic of the Monte Carlo simulation procedure. (a) Example of random variables; (b)
(b) example
example of Monte
of Monte CarloCarlo integration
integration and Monte
and Monte CarloCarlo sensitivity
sensitivity analysis.
analysis.
5. Simulation
5. Simulation
5.1. CFD
5.1. CFDAnalysis Conditions in CFD
5.1.1.
5.1.1. Analysis Conditions
Table 3 lists in CFD for the gas-leak flow analysis. A pressure-based solver
the conditions
wasTable
used3for
lists
thethetransient
conditions for thewhich
analysis, gas-leakis aflow analysis.
technique thatA simulates
pressure-based solver
the flow overwas
time
by dividing
used it into time
for the transient units and
analysis, whichconverging residuals.
is a technique The turbulence
that simulates the flowmodel
overemployed
time by
was theitk-omega
dividing into timeshear
unitsstress transport (SST),
and converging assuming
residuals. an air inletmodel
The turbulence speedemployed
of 1.2 m/s.wasThe
k-omega
the k-omega SST model
shear is atransport
stress hybrid model(SST),that combines
assuming an the
air free-flow-appropriate
inlet speed of 1.2 m/s.k-epsilon
The k-
and wall-analysis-suited
omega SST model is a hybrid k-omegamodelmodels. At the leakthe
that combines point, the inlet was set as ak-epsilon
free-flow-appropriate mass-flow
inlet
and aligned with the real
wall-analysis-suited test conditions
k-omega models.atAt 0.00217
the leakkg/s. For the
point, the inlet
initialisation,
was set as theapressure,
mass-
velocity,
flow and turbulent
inlet aligned with the kinetic energy
real test were set
conditions at to zero. kg/s.
0.00217 The calculation settings forthe
For the initialisation, the
analysis included a Courant number of one and an initial time step of
pressure, velocity, and turbulent kinetic energy were set to zero. The calculation settings 0.5 for the analysis.
for the analysis included a Courant number of one and an initial time step of 0.5 for the
analysis.
Processes 2024,12,
2024,
Processes 12,67x FOR PEER REVIEW 13 of 26
14 of 27
Table Conditionsfor
3. Conditions
Table 3. forgas-leak
gas-leakflow
flowanalysis.
analysis.
SetupSetup Specification
Specification
Program Fluent 2022 R2
Program Fluent 2022 R2
Solver Pressure-based solver
Solver Pressure-based solver
Time Time Transient
Transient
Turbulence Model
Turbulence Model k-omega SST
k-omega SST
Air Air
inletinlet Velocity inlet, 1.2
Velocity m/s 1.2 m/s
inlet,
Leak point
Leak inletinlet
point Mass-flow inlet, 1-butene
Mass-flow (C4H8),(C
inlet, 1-butene 0.00217 kg/s
4 H8 ), 0.00217 kg/s
Air Air
outlet
outlet Pressure outlet, 0 Pa gauge
Pressure outlet, 0 Pa gauge
Initialisation
Initialisation Pressure 0 Pa, Velocity
Pressure 0 m/s, 0Turbulent
0 Pa, Velocity kinetic energy
m/s, Turbulent m2/s2 0 m2 /s2
kinetic 0energy
Calculation Setting
Calculation Setting Adaptive, Max iteration/Time
Adaptive, step 100,
Max iteration/Time Global
step 100, Courant numbernumber
Global Courant 1, Initial 1,
0.5Initial
s, max0.5
1.5s,s,max
min1.5
0.1s,s min 0.1 s
SetupSetup Details Details
Initialisation
Initialisation Pressure 0 Pa, Velocity
Pressure 0 m/s, 0Turbulent
0 Pa, Velocity kinetic energy
m/s, Turbulent m2/s2 0 m2 /s2
kinetic 0energy
Calculation
Calculation Setting
Setting Adaptive,
Adaptive, Max iteration/Time
Max iteration/Time step
step 100, 100, Courant
Global Global Courant
numbernumber
1, Initial 1,
0.5Initial
s, max0.5
1.5s,s,max
min1.5
0.1s,s min 0.1 s
(a)
(b)
Figure 10. Visualisation of sensor detection due to leaked gas flow. (a) Leak range viewed from above;
(b) leak range and sensor detection viewed from the front.
Processes 2024, 12, x FOR PEER REVIEW 15 of 27
Processes 2024, 12, 67 Figure 10. Visualisation of sensor detection due to leaked gas flow. (a) Leak range viewed14from
of 26
above; (b) leak range and sensor detection viewed from the front.
The mass
massfraction
fractionofof 1-butene
1-butene detected
detected by gas-leak
by gas-leak sensor
sensor 7 was7approximately
was approximately
0.0006,
0.0006,
and theand the contour
contour graph,represents
graph, which which represents the from
the process process
thefrom thethe
leak to leak to the
smoke smoke
reaching
reaching
the sensor,the sensor, in
is shown is Figure
shown11. in Figure 11. A maximum
A maximum mass
mass fraction fraction
of 0.0006 wasof detected;
0.0006 was de-
when
tected; when
converted to aconverted to a concentration,
concentration, it wasppm).
it was 0.06% (~600 0.06% (~600 ppm).
(a) (b)
(c) (d)
(e) (f)
Figure 11. Contour graph of the process until 1-butene reaches gas-leak detector number 7 after a
Figure 11. Contour graph of the process until 1-butene reaches gas-leak detector number 7 after a
leak: (a) 1 s; (b) 2 s; (c) 3 s; (d) 4 s; (e) 5 s; (f) ≥6 s.
leak: (a) 1 s; (b) 2 s; (c) 3 s; (d) 4 s; (e) 5 s; (f) ≥6 s.
The trends
The trends shown
shownininFigure
Figure1111correspond
correspond to to
thethe results
results detected
detected overover
timetime by gas-
by gas-leak
leak detector 7 (sensor 7). Figure 12 presents the results of detecting the mass
detector 7 (sensor 7). Figure 12 presents the results of detecting the mass fraction of 1-butenefraction of
1-butene over time, with an average of 0.00036 and a maximum value of
over time, with an average of 0.00036 and a maximum value of 0.00061 over approximately 0.00061 over ap-
proximately
9.5 9.5 s. Thevalue
s. The maximum maximum value when
of 0.00061, of 0.00061, when to
converted converted to concentration,
concentration, was 0.06%,wasas
0.06%, as mentioned previously, which equates to approximately 600
mentioned previously, which equates to approximately 600 ppm. The sensor range used forppm. The sensor
range usedwas
modelling for (0–2000)
modelling ppm,wasand
(0–2000)
the 600ppm,
ppmand the 600
detected in ppm detected in
the simulation wastheconsidered
simulationa
was considered a level that could be reliably detected. Moreover, the remaining
level that could be reliably detected. Moreover, the remaining gas-leak detectors (numbers gas-leak
detectors (numbers 1–6 and 8) detected nearly zero volumes, largely
1–6 and 8) detected nearly zero volumes, largely because of the constant southward because of the con-
winds.
stantwind
This southward
preventswinds. Thisfrom
the gas wind prevents
reaching thethe gas fromthereby
detectors, reaching the detectors,
affecting thereby
their detection
affecting their
capabilities. Thisdetection
phenomenon capabilities. This phenomenon
was influenced was influenced
more by environmental conditionsmorethan
by
by the shortcomings of the detectors. The detection time showed a significant increase,
from 3.5 to 5 s, after which the rate of increase slowed and stabilised at approximately 6.5 s.
1-Butene was detected at 8.875 s when the concentration reached 600 ppm.
environmental conditions than by the shortcomings of the detectors. The detection time
Processes 2024, 12, 67
showed a significant increase, from 3.5 to 5 s, after which the rate of increase slowed and
15 of 26
stabilised at approximately 6.5 s. 1-Butene was detected at 8.875 s when the concentration
reached 600 ppm.
Figure12.
Figure 12.Time-based
Time-basedgraph
graphofofthe
themass
massfraction
fractionofof1-butene.
1-butene.
5.2.
5.2.Monte
MonteCarlo
CarloSimulation
Simulation
In
In this study,we
this study, weemployed
employedaaMonteMonteCarlo
Carlosimulation
simulationtotoanalyse
analysethe
thedetection
detection prob-
prob-
abilities
abilities of the concentration-measurement, ultrasonic, and combined gas-leakdetection
of the concentration-measurement, ultrasonic, and combined gas-leak detection
systems.
systems.TheTheobjective
objective of
ofutilising
utilisingaaMonte
MonteCarlo
Carlosimulation
simulation was
was to
to approximate
approximate thethe de-
de-
tection
tectionprobabilities
probabilitiesofofthese
thesespecific
specificsensors
sensorsbybyincorporating
incorporatingrandom
randomvariables,
variables,such
suchas
as
wind
winddirection,
direction,totoreduce
reducethe
therisk
riskofofleak
leakincidents.
incidents.This
Thismethodology
methodologyprovides
providesaadetailed
detailed
understanding
understanding of ofthe
theperformance
performanceof ofconcentration-measurement,
concentration-measurement, ultrasonic,
ultrasonic, and
and com-
com-
bined gas-leak detection systems under various environmental conditions,
bined gas-leak detection systems under various environmental conditions, particularly particularly
concerning
concerningwindwindimpact.
impact.
5.2.1. Gas-Leak Detection Probability Prediction Model
5.2.1. Gas-Leak Detection Probability Prediction Model
Table 4 lists the factors of the gas-leak detection probability prediction model. The in-
Table 4 lists the factors of the gas-leak detection probability prediction model. The
put parameters comprise the wind direction, random seed, and seed value, which served as
input parameters comprise the wind direction, random seed, and seed value, which
random variables in the Monte Carlo simulations. These factors are broadly categorised as
served as random variables in the Monte Carlo simulations. These factors are broadly cat-
common, concentration-measurement, ultrasonic, and combined gas-leak detection factors.
egorised as common, concentration-measurement, ultrasonic, and combined gas-leak de-
The parameters include the maximum detection standard, minimum detection standard,
tection factors. The parameters include the maximum detection standard, minimum de-
and azimuth, which represent the maximum and minimum values of the wind direction
tection standard, and azimuth, which represent the maximum and minimum values of
and azimuth of the sensors relative to the leak point, respectively. Sensor detection proba-
the wind direction and azimuth of the sensors relative to the leak point, respectively. Sen-
bilities 1, 2, and 3 represent the detection probabilities of the concentration-measurement,
sor detection
ultrasonic, andprobabilities 1, 2, and
combined gas-leak 3 represent
detectors, the detection
respectively. probabilities
The Poisson of the concen-
distribution, a prob-
tration-measurement, ultrasonic, and combined gas-leak detectors, respectively.
ability distribution based on the likelihood of certain events occurring within a specific The time
Pois-
son distribution, a probability distribution based on the likelihood of certain
period [53–56], was applied to the combined detector model in our study. Here, ‘events’ events occur-
ring to
refer within a specific
instances wheretime period
smoke [53–56], was
is undetected. applied toin
Specifically, the combined
the combineddetector
detectormodel
model,in
our study. Here, ‘events’ refer to instances where smoke is undetected. Specifically,
which includes the ultrasonic detector, we assume that events of gas-leak nondetection in the
combined
occur detector
once every 24 model,
h. This which includes
statistical the ultrasonic
approach detector, we and
aids in understanding assume that events
predicting the
of gas-leak nondetection occur once every 24 h. This statistical
detection capabilities of the combined system under various conditions. approach aids in under-
standing and predicting
We defined the detection
the maximum detectioncapabilities
standard asof the
the upper
combinedlimitsystem under
at which the various
sensor
conditions.
detection was considered reliable. The sensor detection probability measures the likelihood
of accurately detecting a gas leak under various conditions. The role of the Poisson
distribution, particularly in its application to the combined detector model, was also
elucidated. This statistical approach was employed to model the frequency of gas-leak
non-detection events based on the average occurrence rate in the combined detector system.
Figure 13 shows the azimuthal configuration. It shows the azimuth (A) and distance
(D, m) between sensors 1 and 8 and leak points 1 and 2. The azimuth from leak point 1 to
sensors 1–8 ranges from 3.60 to 334.81, while that from leak point 2 to sensors 1–8 ranges
Factor Category Symbol Value Range Unit
Wind direction W ≥0 Degree
Common Input
Random seed R 0–1 -
Maximum detection standard Max 0–360 Degree
Processes 2024, 12, 67 Minimum detection standard Parameter Min 0–360 Degree
16 of 26
Gas-leak detector
Azimuth A 0–360 Degree
Sensor detection probability 1 Output SD1 0–100 %
Poisson
from 30.73 (distribution)
to 343.50. Variable
If the wind direction and azimuth P fall within a≥0certain range,
- the
Ultrasonic detector
algorithm processesprobability
Sensor detection them as if 2the sensorOutput
detects a gas SD2
leak. 0–100 %
Combined detector Sensor detection probability 3 Output SD3 0–100 %
Table 4. List of parameters for gas-leak detection probability prediction model.
We defined the maximum detection standard as the upper limit at which the sensor
Factor Category Symbol Value Range Unit
detection was considered reliable. The sensor detection probability measures the likeli-
Common
Wind
hood direction detecting a gas leak under various
of accurately Input
W conditions.≥The
0 Degree
role of the Poisson
Random seed R 0–1 -
distribution, particularly in its application to the combined detector model, was also elu-
Maximum detection
cidated. standard approach was employedMax
This statistical 0–360
to model the frequency Degreenon-
of gas-leak
Minimum
detection events based on the average occurrence rate in the combined detectorDegree
detection standard Parameter Min 0–360 system.
Gas-leak detector
Azimuth A 0–360 Degree
Figure 13 shows the azimuthal configuration. It shows the azimuth (A) and distance
Sensor detection probability 1 Output SD1 0–100 %
(D, m) between sensors 1 and 8 and leak points 1 and 2. The azimuth from leak point 1 to
Poisson (distribution) Variable P ≥0 -
Ultrasonic detector sensors 1–8 probability
ranges from
Sensor detection 2 3.60 toOutput
334.81, while that from leak point
SD2 2 to sensors 1–8
0–100 % ranges
from 30.73 to 343.50. If the wind direction and azimuth fall within a certain range, the
Combined detector Sensor detection probability 3 Output SD3 0–100 %
algorithm processes them as if the sensor detects a gas leak.
(a) (b)
Figure 13.
Figure 13. Azimuth
Azimuth and
and distance
distance between
between the
the leak
leak point
point and
and the
the sensors.
sensors. (a)
(a) Information
Information between
between
leak point 1 and sensors; (b) information between leak point 2 and the sensors.
leak point 1 and sensors; (b) information between leak point 2 and the sensors.
Figure 14
Figure 14 shows
shows the the causality
causality map
map for
for the
the parameters
parameters of of the
the gas-leak
gas-leak detection
detection prob-
prob-
ability model. The purpose of the causality map is to illuminate
ability model. The purpose of the causality map is to illuminate the overall flow the overall flow of
of the
the
system and the causal relationships between factors for the analysis
system and the causal relationships between factors for the analysis of sensor detection of sensor detection
probability. The
probability. The map
map delineates
delineates the
the flow
flow of
of model
model parameters
parameters and and their
their interrelationships,
interrelationships,
where aa plus
where plus sign
sign (+)
(+) indicates
indicates direct
direct proportionality
proportionalityand andaaminus
minussignsign((−)
−) indicates
indicates inverse
inverse
proportionality between variables. The figure emphasises the interactions
proportionality between variables. The figure emphasises the interactions between the between the
various sensor detection probabilities. In the model, sensor detection probability
various sensor detection probabilities. In the model, sensor detection probability 1 (SD1) 1 (SD1)
is influenced by wind direction, which is modelled using the Monte Carlo method with a
random seed.
random seed. The
The random
randomseed seedprovided
provideda afixed
fixedstarting
startingpoint
pointinin
thethe simulations,
simulations, ensur-
ensuring
ing that
that the same
the same sequence
sequence of random
of random numbers
numbers was produced,
was produced, whichwhich was crucial
was crucial for thefor the
repro-
reproducibility
ducibility of theofsimulation
the simulation outcomes.
outcomes. Different
Different randomrandomseedsseeds
werewere applied
applied to ato a nor-
normal
mal distribution
distribution that awas
that was partaof part
theof the wind-direction
wind-direction parameters,
parameters, allowing allowing for various
for various Monte
Carlo simulations. The Poisson distribution, a statistical tool that predicts the probability
of a given number of events occurring in a fixed interval of time or space, was used to
determine sensor detection probability 2 (SD2). This distribution was incorporated into the
model to simulate the detection probability of ultrasonic detectors using the Monte Carlo
method and random seeding. The causality map also includes the maximum and minimum
detection standards, which are likely thresholds for the sensor detection capabilities. These
thresholds, along with the azimuth, which represents the orientation of the sensor relative
to the leak source, determine the likelihood of detecting a gas leak.
to determine sensor detection probability 2 (SD2). This distribution was incorporated into
the model to simulate the detection probability of ultrasonic detectors using the Monte
Carlo method and random seeding. The causality map also includes the maximum and
minimum detection standards, which are likely thresholds for the sensor detection capa-
Processes 2024, 12, 67 bilities. These thresholds, along with the azimuth, which represents the orientation of the
17 of 26
sensor relative to the leak source, determine the likelihood of detecting a gas leak.
Figure 14.
Figure Causalmap
14. Causal map of
of gas-leak
gas-leak detection
detection probability
probability model
model parameters.
parameters.
The description
The description provided
provided in in the
the text
text also
also mentions
mentions the the use
use of
of normal
normal distribution
distribution
formulas and
formulas and Poisson
Poissondistributions
distributionsin inMonte
MonteCarlo
Carlosimulations,
simulations,which whichare
arecommon
commonmeth-
meth-
ods for simulating random events and are especially useful in
ods for simulating random events and are especially useful in complex systems wherecomplex systems where
exact solutions
exact solutionsare
are not
not feasible.
feasible. This
Thisfigure
figurerepresents
representsthe thecomplex
complexinterplay
interplayofofthe
thestatis-
statis-
tical and physical parameters that collectively form the backbone of the
tical and physical parameters that collectively form the backbone of the studied gas-leak studied gas-leak
detection model.
detection model.
Equation (4) pertains
Equation (4) pertains to
to wind
wind direction.
direction. ‘Normal
‘Normal ()’()’ is
is aa random-number
random-number generation
generation
algorithm based on the normal distribution, ‘m’ is the statistical mean of wind direction,
algorithm based on the normal distribution, ‘m’ is the statistical mean of wind direction,
and ‘st’ signifies the standard deviation. The mean ‘m’ was determined to be 157.93 based
and ‘st’ signifies the standard deviation. The mean ‘m’ was determined to be 157.93 based
on the study area, the Fire Insurer Laboratories of Korea (FILK), and the standard deviation
on the study area, the Fire Insurer Laboratories of Korea (FILK), and the standard devia-
was calculated to be 124.92, both of which were then input into the Monte Carlo model.
tion was calculated to be 124.92, both of which were then input into the Monte Carlo
Equation (5) relates to the SD1 estimates for gas-leak detectors and represents the percentage
model. Equation (5) relates to the SD1 estimates for gas-leak detectors and represents the
ratio of detected occurrences at the final time of 720 h. ‘W ’ refers to the wind direction
percentage ratio of detected occurrences at the final time oft720 h. ‘Wt’ refers to the wind
over time, ‘Min’ is the minimum detection standard, and ‘Max’ is the maximum detection
direction over time, ‘Min’ is the minimum detection standard, and ‘Max’ is the maximum
standard. ‘Min’ and ‘Max’ are depicted as values of plus or minus 5–10, based on the wind
detection standard. ‘Min’ and ‘Max’ are depicted as values of plus or minus 5–10, based
angles shown in Figure 13a,b.
on the wind angles shown in Figure 13a,b.
360, 360, i𝑖𝑓fW𝑊 ≥ 360360
WW == 0,0, W
𝑊 ≤ 00 (4)
(4)
Normal((m,
Normal st,RR),
m, st, ), else
else
5.2.2.
5.2.2. Prediction
Prediction Results
Resultsand
andComparative
ComparativeAnalysis
Analysisofof Detection
DetectionProbability
Probability
Figure 15 shows the sensitivity and windrose graphs
Figure 15 shows the sensitivity and windrose graphs for the windfor the wind direction.
direction. The The
ver-
vertical span of the graph represents the sensitivity and indicates the statistical
tical span of the graph represents the sensitivity and indicates the statistical range (5–95%) range
(5–95%) for the wind direction. At the final timemin,
of 720 ◦
for the wind direction. At the final time of 720 themin, the average
average was 144.43
was 144.43°, and the, and
sen-
the sensitivity was 342.47 ◦ . Windrose is primarily used in meteorology and indicates the
sitivity was 342.47°. Windrose is primarily used in meteorology and indicates the direction
direction in which
in which wind wind isBased
is blowing. blowing. Based
on the on the observation
Janghowon Janghowon station,
observation station,wind
the average the
average wind speed was 1.6 m/s, and the average wind direction
speed was 1.6 m/s, and the average wind direction was north (NNW). was north (NNW).
Wind direc tion_ AWS (95 Perc entile) Wind direc tion_ AWS (90 Perc entile) Wind direc tion_ AWS (75 Perc entile)
Wind direc tion_ AWS (50 Perc entile) Wind direc tion_ AWS (25 Perc entile) Wind direc tion_ AWS (10 Perc entile)
Wind direc tion_ AWS (5 Perc entile)
400
300
200
100
0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours )
(a) (b)
Figure 15. Wind-velocity prediction results and windrose graph. (a) Wind-velocity prediction results;
(b) windrose graph.
Figure 16 shows a sensitivity graph of the detection rates of the eight sensors when
gas leaked from leak point 1. The average detection probabilities for sensors 1–8 over 720 h
were 2.92%, 2.64%, 2.54%, 3.89%, 2.19%, 1.80%, 11.85%, and 2.83%, respectively. The vertical
span in Figure 16 represents the sensitivity and shows the statistical range of detection
probabilities, covering values between 5% and 95% of the total percentage. The sensitivities,
Processes 2024, 12, x FOR PEER REVIEW 20 of 27
Detection probability_leak1_sensor 1 (95 Percentile) Detection probability_leak1_sensor 1 (90 Percentile) Detection probability_leak1_sensor 2 (95 Percentile) Detection probability_leak1_sensor 2 (90 Percentile)
Detection probability_leak1_sensor 1 (75 Percentile) Detection probability_leak1_sensor 1 (50 Percentile) Detection probability_leak1_sensor 2 (75 Percentile) Detection probability_leak1_sensor 2 (50 Percentile)
Detection probability_leak1_sensor 1 (25 Percentile) Detection probability_leak1_sensor 1 (10 Percentile) Detection probability_leak1_sensor 2 (25 Percentile) Detection probability_leak1_sensor 2 (10 Percentile)
Detection probability_leak1_sensor 1 (5 Percentile) Detection probability_leak1_sensor 2 (5 Percentile)
% %
10 10
8 8
6 6
4 4
2 2
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
(a) (b)
Detection probability_leak1_sensor 3 (95 Percentile) Detection probability_leak1_sensor 3 (90 Percentile) Detection probability_leak1_sensor 4 (95 Percentile) Detection probability_leak1_sensor 4 (90 Percentile)
Detection probability_leak1_sensor 3 (75 Percentile) Detection probability_leak1_sensor 3 (50 Percentile) Detection probability_leak1_sensor 4 (75 Percentile) Detection probability_leak1_sensor 4 (50 Percentile)
Detection probability_leak1_sensor 3 (25 Percentile) Detection probability_leak1_sensor 3 (10 Percentile) Detection probability_leak1_sensor 4 (25 Percentile) Detection probability_leak1_sensor 4 (10 Percentile)
Detection probability_leak1_sensor 3 (5 Percentile) Detection probability_leak1_sensor 4 (5 Percentile)
% %
10 10
8 8
6 6
4 4
2 2
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours) Time (Hours)
(c) (d)
Detection probability_leak1_sensor 5 (95 Percentile) Detection probability_leak1_sensor 5 (90 Percentile) Detection probability_leak1_sensor 6 (95 Percentile) Detection probability_leak1_sensor 6 (90 Percentile)
Detection probability_leak1_sensor 5 (75 Percentile) Detection probability_leak1_sensor 5 (50 Percentile) Detection probability_leak1_sensor 6 (75 Percentile) Detection probability_leak1_sensor 6 (50 Percentile)
Detection probability_leak1_sensor 5 (25 Percentile) Detection probability_leak1_sensor 5 (10 Percentile) Detection probability_leak1_sensor 6 (25 Percentile) Detection probability_leak1_sensor 6 (10 Percentile)
Detection probability_leak1_sensor 5 (5 Percentile) Detection probability_leak1_sensor 6 (5 Percentile)
% %
10 5
8 4
6 3
4 2
2 1
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 20 40 60 80 100 120 140 160 180 200 220 240260 280 300 320 340360 380 400420440460480500 520 540 560 580 600 620 640660 680 700 720
Processes 2024,12,
Processes2024, 12,x67
FOR PEER REVIEW 21 of 27
20 of 26
20 8
15 6
20 8
0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours)
15 6 Time (Hours)
(g) (h)
10 4
Figure 16. Analysis results of concentration-measurement type gas-leak detection probability (from
5 leak 1 to the indicated sensor): (a)2 leak 1 to sensor 1; (b) leak 1 to sensor 2; (c) leak 1 to sensor 3; (d)
leak 1 to sensor 4; (e) leak 1 to sensor 5; (f) leak 1 to sensor 6; (g) leak 1 to sensor 7; (h) leak 1 to sensor
0
0 40 80 120 160 200 240 280 320 360 8. 400 440 480 520 0 560 600 640 680 720
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours)
Time (Hours)
(g) Figure 17 shows a sensitivity graph of the detection(h) rates of the eight sensors when
gas leaked from leak point 2. The average detection rates for sensors 1–8 over 720 h were
Figure 16. Analysis results ofofconcentration-measurement type gas-leak detection probability (from
3.44%,16.
Figure 3.60%,
Analysis 4.18%, results4.65%, 2.58%, 2.24%, 1.61%, and type
concentration-measurement 2.84%, respectively.
gas-leak detectionThe vertical(from
probability span
leak 1 to the indicated sensor): (a) leak 1 to sensor 1; (b) leak 1 to sensor 2; (c) leak 1 to sensor 3; (d)
in Figure
leak
leak
1 to the
1 to sensor 17 represents
indicated
4; (e) leak 1 to the
sensor): sensitivity
sensor
(a) leak 1
5; (f) leak and
to
1 toshows
sensor
sensorthe
1; (b) statistical
leak
6; (g)
1 to range7;of(h)
sensor
leak 1 to sensor
2; (c) detection
leak 1 to proba-
sensor
leak 1 to sensor
3;
8.bilities, covering values between 5% and 95% of the total percentage. The corresponding
(d) leak 1 to sensor 4; (e) leak 1 to sensor 5; (f) leak 1 to sensor 6; (g) leak 1 to sensor 7; (h) leak 1 to
sensitivities
sensor 8. were 2.50%, 2.22%, 2.36%, 2.36%, 1.66%, 1.80%, 1.53%, and 1.94%.
Figure 17 shows a sensitivity graph of the detection rates of the eight sensors when
gasDetection
Detection probability_leak2_sensor 1 (95 Percentile)
Detection probability_leak2_sensor 1 (75 Percentile) leaked
Detection from leak point
Percentile)2. The average
probability_leak2_sensor 1 (90 Percentile)
probability_leak2_sensor 1 (50 detection
Detection probability rates
_leak2_sensor 2 for sensors
Detection probability_leak2_sensor 2 (95 Percentile)
(75 Percentile) 1–8_leak2_sensor
over 720 h were
Detection probability_leak2_sensor 2 (90 Percentile)
Detection probability 2 (50 Percentile)
Detection probability_leak2_sensor 1 (25 Percentile) Detection probability_leak2_sensor 1 (10 Percentile) Detection probability_leak2_sensor 2 (25 Percentile) Detection probability_leak2_sensor 2 (10 Percentile)
Detection probability_leak2_sensor 1 (5 Percentile)3.44%, 3.60%, 4.18%, 4.65%, 2.58%, 2.24%, 1.61%,
Detection probability and 22.84%,
_leak2_sensor (5 Percentile)respectively. The vertical span
% %
10
in Figure 17 represents the sensitivity and shows the statistical range of detection proba- 10
bilities, covering values between 5% and 95% of the total percentage. The corresponding
8 8
sensitivities were 2.50%, 2.22%, 2.36%, 2.36%, 1.66%, 1.80%, 1.53%, and 1.94%.
6 6
Detection probability_leak2_sensor 1 (95 Percentile) Detection probability_leak2_sensor 1 (90 Percentile) Detection probability_leak2_sensor 2 (95 Percentile) Detection probability_leak2_sensor 2 (90 Percentile)
Detection probability_leak2_sensor 1 (75 Percentile) Detection probability_leak2_sensor 1 (50 Percentile) Detection probability_leak2_sensor 2 (75 Percentile) Detection probability_leak2_sensor 2 (50 Percentile)
Detection probability_leak2_sensor 1 (25 Percentile) Detection probability_leak2_sensor 1 (10 Percentile) Detection probability_leak2_sensor 2 (25 Percentile) Detection probability_leak2_sensor 2 (10 Percentile)
4 4 Detection probability_leak2_sensor 2 (5 Percentile)
Detection probability_leak2_sensor 1 (5 Percentile)
% %
10 10
2 2
8 8
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
(a) (b)
4 Detection probability_leak2_sensor 3 (95 Percentile) Detection probability_leak2_sensor 3 (90 Percentile) 4 Detection probability_leak2_sensor 4 (95 Percentile) Detection probability_leak2_sensor 4 (90 Percentile)
Detection probability_leak2_sensor 3 (75 Percentile) Detection probability_leak2_sensor 3 (50 Percentile) Detection probability_leak2_sensor 4 (75 Percentile) Detection probability_leak2_sensor 4 (50 Percentile)
Detection probability_leak2_sensor 3 (25 Percentile) Detection probability_leak2_sensor 3 (10 Percentile) Detection probability_leak2_sensor 4 (25 Percentile) Detection probability_leak2_sensor 4 (10 Percentile)
Detection probability_leak2_sensor 3 (5 Percentile) Detection probability_leak2_sensor 4 (5 Percentile)
2 % 2%
15 15
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
0 0
10 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 10 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
(c) (d)
5 5
Figure 17. Cont.
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
(c) (d)
Processes2024,
Processes 12,67
2024,12, x FOR PEER REVIEW 22 of 27
21 of 26
Detection probability_leak2_sensor 5 (95 Percentile) Detection probability_leak2_sensor 5 (90 Percentile) Detection probability_leak2_sensor 6 (95 Percentile) Detection probability_leak2_sensor 6 (90 Percentile)
Detection probability_leak2_sensor 5 (75 Percentile) Detection probability_leak2_sensor 5 (50 Percentile) Detection probability_leak2_sensor 6 (75 Percentile) Detection probability_leak2_sensor 6 (50 Percentile)
Detection probability_leak2_sensor 5 (25 Percentile) Detection probability_leak2_sensor 5 (10 Percentile) Detection probability_leak2_sensor 6 (25 Percentile) Detection probability_leak2_sensor 6 (10 Percentile)
Detection probability_leak2_sensor 5 (5 Percentile) Detection probability_leak2_sensor 6 (5 Percentile)
% %
10 10
8 8
6 6
4 4
2 2
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
8 8
6 6
4 4
2 2
0 0
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720 0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
(g) (h)
Figure 17.
Figure 17. Analysis
Analysisresults
resultsof
ofconcentration-measurement
concentration-measurementtype typegas-leak
gas-leakdetection
detectionprobability
probability (leak
(leak 2
2 to the indicated sensor): (a) leak 2 to sensor 1; (b) leak 2 to sensor 2; (c) leak 2 to sensor 3; (d) leak
to the indicated sensor): (a) leak 2 to sensor 1; (b) leak 2 to sensor 2; (c) leak 2 to sensor 3; (d) leak 2 to
2 to sensor 4; (e) leak 2 to sensor 5; (f) leak 2 to sensor 6; (g) leak 2 to sensor 7; (h) leak 2 to sensor 8.
sensor 4; (e) leak 2 to sensor 5; (f) leak 2 to sensor 6; (g) leak 2 to sensor 7; (h) leak 2 to sensor 8.
Figure 18
Figure 18 shows
shows the
the sensitivity
sensitivityofofthe
thedetection
detectionrate ofof
rate thethe
ultrasonic sensor.
ultrasonic TheThe
sensor. an-
alysed detection probability calculated from the average sensitivity was 96.07%,
analysed detection probability calculated from the average sensitivity was 96.07%, and theand the
sensitivity was 2.36%.
sensitivity was 2.36%.
Figure 19 compares the analysed detection probabilities for concentration-measurement,
ultrasonic, Detection probability_ulltrasonic (95 Percentile )
and combined gas-leak detectors.Detection
Detection probability_ulltrasonic (75 Percentile )
Whenprobability_ulltrasonic
converged to (90 Percentile )
99.99%, the ultrasonic
Detection probability_ulltrasonic (50 Percentile )
Detection
detector converged probability_ulltrasonic
to 96.07%, (25
and Percentile
the ) Detection
average probability_ulltrasonic
detection probability (10 Percentile
for )
each sensor of
Detection probability_ulltrasonic (5 Percentile )
the% concentration-measurement type gas-leak detector was 3.78% for leak point 1 and
100
3.04% for leak point 2. Upon analysing leak point 1, which exhibited a higher detection
probability, it was determined that the efficiency of the ultrasonic detector was approxi-
99
mately 25.42 times that of the concentration-measurement gas-leak detector. This factor was
calculated by dividing the detection probability of the former by that of the latter. Similarly,
98
the efficiency of the combined detector was approximately 26.45 times greater than that
of the concentration-measurement gas-leak detector, which was determined by dividing
97
the detection probability of the former by that of the latter. The average of leak points 1
and
96
2 was 3.41%; in this case, the ultrasonic detector was approximately 28.17 times more
efficient than the concentration-measurement gas-leak detector, and the combined detector
was
95 approximately 29.32 times more efficient.
94
93
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours)
2 to the indicated sensor): (a) leak 2 to sensor 1; (b) leak 2 to sensor 2; (c) leak 2 to sensor 3; (d) leak
2 to sensor 4; (e) leak 2 to sensor 5; (f) leak 2 to sensor 6; (g) leak 2 to sensor 7; (h) leak 2 to sensor 8.
Figure 18 shows the sensitivity of the detection rate of the ultrasonic sensor. The an-
Processes 2024, 12, 67 alysed detection probability calculated from the average sensitivity was 96.07%, and 22 ofthe
26
sensitivity was 2.36%.
Detec tion probability_ ulltras onic (95 Perc entile) Detec tion probability_ ulltras onic (90 Perc entile)
Detec tion probability_ ulltras onic (75 Perc entile) Detec tion probability_ ulltras onic (50 Perc entile)
Processes 2024, 12, x FOR PEER REVIEW Detec tion probability_ ulltras onic (25 Perc entile) Detec tion probability_ ulltras onic (10 Perc entile) 23 of 27
Detec tion probability_ ulltras onic (5 Perc entile)
%
100
Detec tion probability_ ulltras onic (Average) Detec tion probability_ leak1_ average (Average)
Detec tion probability_ leak2_ average (Average) Detec tion probability_ duet_ average (Average)
%
100
50
0 40 80 120 160 200 240 280 320 360 400 440 480 520 560 600 640 680 720
Time (Hours )
Figure
Figure 19.
19. Comparative
Comparative graph
graph of
of the
the detection
detection rates
rates for
for the
the concentration-measurement,
concentration-measurement, ultrasonic,
ultrasonic,
and combined gas-leak detectors.
and combined gas-leak detectors.
These probabilities are based on simulations targeting the verification test site built
within the
the Fire
Fire Insurers
InsurersLaboratories
LaboratoriesofofKorea
Korea (FILK)
(FILK)and andareare
limited
limitedto cases where
to cases the
where
structures
the structuresof piping and vessels
of piping are simple,
and vessels and alland
are simple, the all
expected leakageleakage
the expected areas can be seen
areas can
by
be the
seenultrasonic camera incamera
by the ultrasonic one direction.
in one Petrochemical plants have plants
direction. Petrochemical complex installations
have complex
with multiplewith
installations pipes and structures,
multiple pipes andmaking it difficult
structures, to monitor
making all leaks
it difficult from one
to monitor alldirec-
leaks
from Therefore,
tion. one direction. Therefore,
a gas-leak a gas-leak
detection methoddetection
must bemethod
selected must
andbeapplied
selected and applied
efficiently, ac-
efficiently,
cording to according to the
the structure andstructure andof
conditions conditions
each areaof each area
within eachwithin each petrochemical
petrochemical plant. The
plant.
model Thecan bemodel can be appropriately
appropriately utilised
utilised if the if therange
detection detectionof therange of the ultrasonic
ultrasonic camera is
camera is clearly identified according to the characteristics of the area, and a
clearly identified according to the characteristics of the area, and a comparative simulation comparative
simulation
of the area is ofperformed.
the area is performed.
6. Discussion
Monte Carlo simulations confirmed that the leakage detection system using ultra-
sonic waves has a higher probability of detection performance than the existing concen-
tration-measurement methods. However, it is realistically impossible to remove all the
Processes 2024, 12, 67 23 of 26
6. Discussion
Monte Carlo simulations confirmed that the leakage detection system using ultrasonic
waves has a higher probability of detection performance than the existing concentration-
measurement methods. However, it is realistically impossible to remove all the gas-leak
detectors using the concentration-measurement method installed in oil refineries and
petrochemical plants and to replace them with ultrasonic detection methods. Instead, it
is more realistic to maintain the existing system and add an ultrasonic detection system.
When the two systems were combined, the detection probability was 96.58% higher than
that attained when the concentration-measurement method was used alone, and 3.92%
higher than that when the ultrasonic measurement method was used alone. If the two
systems are used interchangeably, the concept of failure safety, which can detect gas leaks
in another system even if one system breaks down, can be introduced into the workplace.
Safety against failure is a basic and important concept. This study suggests a direction for
advanced gas-leak detection systems in future oil refining and petrochemical plants. In
addition, applications in places where large amounts of gas are used, such as workplaces,
buildings, residential facilities, and sales facilities in other industries, as well as in the
oil refining and petrochemical industries, can be considered. However, it is difficult to
apply it uniformly to all spaces. First, it is necessary to determine the suitability of each
space by identifying its characteristics and leaking materials and performing simulations.
In addition, optimal placement of ultrasonic and concentration-measurement gas-leak
detectors is required, considering the characteristics of each region. The optimal placement
includes a discussion of camera placement issues when the expected gas-leak area is
obscured by pipes or structures. If a fixed ultrasound camera can rotate up, down, left,
and right, it can detect more areas, and the development of this technology will increase
its utility. Ultrasonic detection technology is more prone to misidentification owing to
ambient noise than because of the ultrasonic waves generated by leakage. Further research
is required to overcome the misperceptions caused by ambient noise. Additional research
requires the incorporation of artificial intelligence, and the use of convolutional neural
networks may be a solution.
7. Conclusions
In this study, we propose an ultrasonic gas-leak monitoring system for quick and accu-
rate detection of gas leaks, which is one of the major risks in oil refining and petrochemical
plants. We built a demonstration test facility that simulated oil refining and petrochemical
plants and conducted tests to verify the system. To compare and analyse this verified ultra-
sonic gas-leak monitoring system with existing systems, detection performance parameters,
such as detection time and detection probability, were calculated and quantified through
CFD and Monte Carlo simulations.
(1) A fixed ultrasonic gas-leak monitoring system was proposed. The performance of the
ultrasonic camera was enhanced, including securing the detection performance in the
case of microleakage and strengthening the waterproofing function. To confirm the
performance of the ultrasonic camera, a laboratory-scale test was conducted following
the ASTM E 1002-11 standard, and gas leakage was detected even when the camera
was 5 m away from the leak source under a microleakage condition of 1.67 mL/s. The
waterproofing performance of the ultrasonic camera exhibited an IP of 55 or higher;
(2) We individually visited oil refining and petrochemical plants to review drawings and
conduct field investigations. Layouts and 3D drawings were created through field and
literature research. Based on these drawings, a demonstration test facility simulating
oil refining and petrochemical plants within the FILK was built. The literature related
to gas leaks and frequencies was investigated, and a test scenario was established
through an analysis of accident cases;
(3) The performance of the ultrasonic gas-leak monitoring system was verified through
demonstration tests that reflected the environment of oil refining and petrochemical
plants. It was confirmed that gas-leak detection was possible up to approximately
Processes 2024, 12, 67 24 of 26
23 m under the conditions of internal piping pressure of 2.2 bar, leak diameter of
6.9 mm, and leak rate of 0.00217 kg/s. The ultrasonic gas-leak detection time was
approximately 8 s, including 3 s for the integrated surveillance system monitoring
image processing time and 5 s for leak determination;
(4) Using CFD, we visually expressed the flow owing to gas diffusion and the phenomenon
of reaching a concentration that could be detected by concentration-measuring gas de-
tectors, calculated the gas-leak detection time, and compared it with that of an ultrasonic
gas-leak monitoring system. The CFD results indicated that the gas-leak detection time
of the existing system was 8.895 s. This is slightly longer than the 8 s leak detection time
of the ultrasonic gas-leak monitoring system;
(5) As a result of the Monte Carlo simulation, it was calculated that, when the two systems
were combined, the detection probability was 96.58% higher than that achieved when
the concentration-measurement method was used alone and 3.92% higher than that
attained when the ultrasonic measurement method was used alone. The combination
of the two systems exhibited the best performance.
References
1. Gómez-Mares, M.; Muñoz, M.; Casal, J. Axial Temperature Distribution in Vertical Jet Fires. J. Hazard. Mater. 2009, 172, 54–60.
[CrossRef]
2. Ogle, R.A.; Megerle, M.V.; Morrison, D.R.; Carpenter, A.R. Explosion Caused by Flashing Liquid in a Process Vessel. J. Hazard.
Mater. 2004, 115, 133–140. [CrossRef]
3. Zhang, Y.; Cui, Z.; Wang, M.; Liu, B.; Fan, X.; Tian, W. An Energy-Efficiency Prediction Method in Crude Distillation Process
Based on Long Short-Term Memory Network. Processes 2023, 11, 1257. [CrossRef]
4. Jonathan, E.; Sugiarto, S. Risk Analysis of Gas Dispersion, Fire and Explosion Due to Gas Pipeline Leak at Onshore Receiving
Facility of PT XYZ in Muara Karang Using Aloha Software 5.4. Int. J. Eng. Bus. Soc. Sci. 2023, 1, 186–198. [CrossRef]
5. Dadkani, P.; Noorzai, E.; Ghanbari, A.H.; Gharib, A. Risk Analysis of Gas Leakage in Gas Pressure Reduction Station and Its
Consequences: A Case Study for Zahedan. Heliyon 2021, 7, e06911. [CrossRef]
6. Satterfield, T. Advocacy after Bhopal: Environmentalism, Disaster, New Global Orders. Kim Fortun J. Anthropol. Res. 2003, 59,
284–286. [CrossRef]
7. da Silva, I.G.S.; Pappalardo, J.R.; da Rocha e Silva, N.M.P.; Converti, A.; de Almeida, F.C.G.; Sarubbo, L.A. Treatment of Motor
Oil-Contaminated Soil with Green Surfactant Using a Mobile Remediation System. Processes 2023, 11, 1081. [CrossRef]
8. French McCay, D.P.; Whittier, N.; Ward, M.; Santos, C. Spill Hazard Evaluation for Chemicals Shipped in Bulk Using Modeling.
Environ. Model. Softw. 2006, 21, 156–169. [CrossRef]
9. Singh, S.; Chakma, S.; Alawa, B.; Kalyanasundaram, M.; Diwan, V. Identification, Characterization, and Implications of
Microplastics in Soil—A Case Study of Bhopal, Central India. J. Hazard. Mater. Adv. 2023, 9, 100225. [CrossRef]
10. NICS Integrated Chemical Information System (ICIS). Available online: https://www.icis.me.go.kr/ (accessed on 4 July 2023).
11. National Fire Agency National Fire Data System (NFDS). Available online: https://www.nfds.go.kr (accessed on 3 July 2023).
12. Seo, W.; Im, S.; Lee, G. Characteristics of the Received Signal of an Ultrasonic Sensor Installed in a Chamber with Micro-Leakage.
Mech. Sci. 2021, 12, 1051–1060. [CrossRef]
Processes 2024, 12, 67 25 of 26
13. Li, J.; Wan, J.; Wang, T.; Yuan, G.; Jurado, M.J.; He, Q. Leakage Simulation and Acoustic Characteristics Based on Acoustic Logging
by Ultrasonic Detection. Adv. Geo-Energy Res. 2022, 6, 181–191. [CrossRef]
14. Naranjo, E. The Monitor-Monitoring Large Hydrogen Sulfide Releases: Why Sensor Recovery Time Is Important. Available
online: https://www.damonhydro.com (accessed on 26 November 2023).
15. Ramya, V. Embedded System for Hazardous Gas Detection and Alerting. Int. J. Distrib. Parallel Syst. 2012, 3, 287–300. [CrossRef]
16. Jarvis, R.; Cawley, P.; Nagy, P.B. Current Deflection NDE for the Inspection and Monitoring of Pipes. NDT E Int. 2016, 81, 46–59.
[CrossRef]
17. Fang, J.; Xiang, J.; Ma, L.; Liu, H.; Wang, C.; Liang, S. Gas-Driven Endoscopic Robot for Visual Inspection of Corrosion Defects
Inside Gas Pipelines. Processes 2023, 11, 1098. [CrossRef]
18. Wang, T.K.; Lin, Y.H.; Shen, J.Y. Developing and Implementing an AI-Based Leak Detection System in a Long-Distance Gas
Pipeline. Adv. Technol. Innov. 2022, 7, 169–180. [CrossRef]
19. Adegboye, M.A.; Karnik, A.; Fung, W.K. Numerical Study of Pipeline Leak Detection for Gas-Liquid Stratified Flow. J. Nat. Gas.
Sci. Eng. 2021, 94, 104054. [CrossRef]
20. Ishak, A.J.; Mahmood, S.N.; Hussain, A.S.T. GSM Based Gas Leak Monitoring System. Period. Eng. Nat. Sci. 2019, 7, 670–678.
[CrossRef]
21. Meribout, M.; Khezzar, L.; Azzi, A.; Ghendour, N. Leak Detection Systems in Oil and Gas Fields: Present Trends and Future
Prospects. Flow Meas. Instrum. 2020, 75, 101772. [CrossRef]
22. Kang, R.; Liatsis, P.; Kyritsis, D.C. Emission Quantification via Passive Infrared Optical Gas Imaging: A Review. Energies 2022,
15, 3304. [CrossRef]
23. Olbrycht, R. A Novel Method for Sensitivity Modelling of Optical Gas Imaging Thermal Cameras with Warm Filters. Quant.
Infrared Thermogr. J. 2022, 19, 331–346. [CrossRef]
24. Adnan, N.F.; Ghazali, M.F.; Amin, M.M.; Hamat, A.M.A. Leak Detection in Gas Pipeline by Acoustic and Signal Processing—A
Review. IOP Conf. Ser. Mater. Sci. Eng. 2015, 100, 012013. [CrossRef]
25. Li, J.; Li, Y.; Huang, X.; Ren, J.; Feng, H.; Zhang, Y.; Yang, X. High-Sensitivity Gas Leak Detection Sensor Based on a Compact
Microphone Array. Measurement 2021, 174, 109017. [CrossRef]
26. Diaconu, B.M. Recent Advances and Emerging Directions in Fire Detection Systems Based on Machine Learning Algorithms. Fire
2023, 6, 441. [CrossRef]
27. Martinka, J.; Rantuch, P.; Martinka, F.; Wachter, I.; Štefko, T. Improvement of Heat Release Rate Measurement from Woods Based
on Their Combustion Products Temperature Rise. Processes 2023, 11, 1206. [CrossRef]
28. Mahalingam, A.; Naayagi, R.T.; Mastorakis, N.E. Design and Implementation of an Economic Gas Leakage Detector. Recent. Res.
Appl. Electr. Comput. Eng. 2012, 3.
29. Khan, M.M. Sensor-Based Gas Leakage Detector System. Eng. Proc. 2020, 2. [CrossRef]
30. Meribout, M. Gas Leak-Detection and Measurement Systems: Prospects and Future Trends. IEEE Trans. Instrum. Meas. 2021, 70.
[CrossRef]
31. Kim, S.-R.; Moon, H.-S.; Jeong, P.-H. Optimal Ventilation Design for Flammable Gas Leaking from Gas Box Used in Semiconductor
Manufacturing: Case Study on Korean Semiconductor Industry. Fire 2023, 6, 432. [CrossRef]
32. Ou, J.; Wang, X.; Ming, Y.; Sun, X. Study on the Influence of Ventilation Speed on Smoke and Temperature Characteristics of
Complex Underground Spaces. Fire 2023, 6, 436. [CrossRef]
33. Reddy, D.B.M. Gas Leakage with Auto Ventilation and Smart Management System Using IOT. Int. J. Sci. Res. Eng. Manag. 2022, 6.
[CrossRef]
34. Hadlock, C.R.; Woolford, S.W. Optimizing Management of Emergency Gas Leaks: A Case Study in Business Analytics. J. Bus.
Anal. 2019, 2, 88–99. [CrossRef]
35. Bandes, A.S. Operations & Maintenance: Detect Leaks with Ultrasound. Chem. Eng. 2002, 109.
36. Freiberger, K. Development and Evaluation of Source Localization Algorithms for Coincident Microphone Arrays; Graz University of
Technology: Graz, Austria, 2010.
37. Dibiase, J.H. A High-Accuracy, Low-Latency Technique for Talker Localization in Reverberant Environments Using Microphone Arrays;
Brown University: Providence, RI, USA, 2000.
38. SM Instruments Inc. Ultrasonic Camera. Available online: https://www.smins.co.kr (accessed on 29 June 2023).
39. Xu, M.; Tao, W. Study on Gas Leakage Localization Method Based on Ultrasonic Sensor Area Array. In Proceedings of the
IEEE/ASME International Conference on Advanced Intelligent Mechatronics, Munich, Germany, 3–7 July 2017; AIM: Cranberry
Township, PA, USA, 2017.
40. ASTM International ASTM E1002-11, Standard Practice for Leaks Using Ultrasonics. Available online: http://www.astm.org/
Standards/ (accessed on 21 June 2023).
41. KFPA KFS (Korea Fire Safety) 700-2018 Standards: Standard on Fire Protection for Petroleum and Petrochemical Plants. Available
online: https://kfs.kfpa.or.kr/ (accessed on 21 June 2023).
42. Ku, C.; An, S.; Ma, B. Analysis Characteristic of Non-Point Source in Petrochemical. KIGAS 2022, 26. [CrossRef]
43. KOSHA Technical Guidelines for Installation and Maintenance of Gas Leak Detector (Kosha Guide P-166-2020). Available online:
https://www.kosha.or.kr/ (accessed on 21 June 2023).
Processes 2024, 12, 67 26 of 26
44. Rong, Y.; Peng, J.; Gao, J.; Zhang, X.; Li, X.; Pan, X.; Chen, J.; Chen, S. Numerical Investigation on the Liquid Hydrogen Leakage
and Protection Strategy. Processes 2023, 11, 1173. [CrossRef]
45. Li, G.; Yang, X.; Dai, G. CFD Simulation of Effects of the Configuration of Gas Distributors on Gas-Liquid Flow and Mixing in a
Bubble Column. Chem. Eng. Sci. 2009, 64, 5104–5116. [CrossRef]
46. Tauseef, S.M.; Rashtchian, D.; Abbasi, S.A. CFD-Based Simulation of Dense Gas Dispersion in Presence of Obstacles. J. Loss Prev.
Process Ind. 2011, 24, 371–376. [CrossRef]
47. Grumbach, F.; Müller, A.; Reusch, P.; Trojahn, S. Robustness Prediction in Dynamic Production Processes—A New Surrogate
Measure Based on Regression Machine Learning. Processes 2023, 11, 1267. [CrossRef]
48. Pokorádi, L. Monte-Carlo Simulation-Based Accessibility Analysis of Temporal Systems. Symmetry 2022, 14, 303. [CrossRef]
49. Bird, G.A. Monte carlo simulation of gas flows. Annu. Rev. Fluid. Mech. 1978, 114, 11–31. [CrossRef]
50. Kim, J.; Smit, B. Efficient Monte Carlo Simulations of Gas Molecules inside Porous Materials. J. Chem. Theory Comput. 2012, 8,
2336–2343. [CrossRef]
51. Qazi, A.; Simsekler, M.C.E. Risk Assessment of Construction Projects Using Monte Carlo Simulation. Int. J. Manag. Proj. Bus.
2021, 14, 1202–1218. [CrossRef]
52. Ming, X.; Liang, Q.; Dawson, R.; Xia, X.; Hou, J. A Quantitative Multi-Hazard Risk Assessment Framework for Compound
Flooding Considering Hazard Inter-Dependencies and Interactions. J. Hydrol. 2022, 607, 127477. [CrossRef]
53. El-Dawoody, M.; Eliwa, M.S.; El-Morshedy, M. An Extension of the Poisson Distribution: Features and Application for Medical
Data Modeling. Processes 2023, 11, 1195. [CrossRef]
54. Sahai, H.; Khurshid, A. Confidence Intervals for the Mean of a Poisson Distribution: A Review. Biom. J. 1993, 35, 857–867.
[CrossRef]
55. Mijburgh, P.A.; Visagie, I.J.H. An Overview of Goodness-of-Fit Tests for the Poisson Distribution. S. Afr. Stat. J. 2020, 54, 207–230.
[CrossRef]
56. Bidounga, R.; Batsindila Nganga, P.C.; Niéré, L.; Mizère, D. A Note on the (Weighted) Bivariate Poisson Distribution. Eur. J. Pure
Appl. Math. 2021, 14, 192–203. [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.