sensors

Article
Evaluation of Commercial Corrosion Sensors for Real-Time
Monitoring of Pipe Wall Thickness under Various
Operational Conditions
Dong-Ho Shin 1 , Hyun-Kyu Hwang 1 , Heon-Hui Kim 2 and Jung-Hyung Lee 2, *

1 Graduate school, Mokpo National Maritime University, Mokpo 58628, Korea

2 Division of Marine Engineering, Mokpo National Maritime University, Mokpo 58628, Korea
* Correspondence: [email protected]; Tel.: +82-061-240-7205

Abstract: In this study, we investigated the performance and reliability of commercial corrosion
sensors for monitoring the integrity of piping systems in various fluid environments as an alternative
to ultrasonic transducers. To this end, we investigated pipes’ wall-thinning using commercial
electrical resistance (ER), linear polarization resistance (LPR), and ultrasonic transducer (UT) sensors
under various operating environments. A pilot-scale closed-loop test bed was built to simulate a real
pipeline flow situation, from which the sensor data were collected and analyzed. Experimental results
indicate that, in the case of the LPR sensor, it is challenging to accurately measure the corrosion rate
when a specific measure exceeds the threshold in a severe corrosion environment. In contrast, the ER
sensor could measure metal loss under all conditions and reflect the corresponding characteristics.
The metal loss (about 0.25 mm) of the real pipe after the experiment was confirmed to be equal to the
metal loss (0.254 mm) measured by the sensor. Furthermore, the regression analysis revealed a high
correlation between the results obtained from the ER and UT sensors. Thus, evaluating the remaining
Citation: Shin, D.-H.; Hwang, H.-K.; thickness of the piping system using the commercial ER sensor is deemed to be effective and reliable.
Kim, H.-H.; Lee, J.-H. Evaluation of
Commercial Corrosion Sensors for Keywords: corrosion; wall-thinning; ER and LPR sensor; UT sensor; electrochemical experiment
Real-Time Monitoring of Pipe Wall
Thickness under Various Operational
Conditions. Sensors 2022, 22, 7562.
https://doi.org/10.3390/s22197562 1. Introduction
Academic Editors: Nenad Gligoric, Piping systems play a vital role in many industries as a transport medium. In utility
Suparna De, Klaus Moessner industries, piping is used for the water supply, tanks, heating and cooling water, and
and Charmi Chande gas transfer. In chemical industries, especially in petroleum facilities, piping is applied
Received: 5 September 2022
to connect raw or refined chemicals to all components, such as the tanks and distillation
Accepted: 3 October 2022 facilities associated with each plant [1]. During the operation of the piping system, risks
Published: 6 October 2022 involve wall-thinning in the pipeline, which can lead to the leakage of hazardous chemicals
induced by a puncture or rupture in the piping system [2].
Publisher’s Note: MDPI stays neutral
Leakage occurs in these piping systems for various reasons. The American Water
with regard to jurisdictional claims in
Association (AWWA) classified the causes of leaks as poor or defective materials, excessive
published maps and institutional affil-
pressure, errors in valve operation (opening and closing), leaks caused by corrosion and
iations.
erosion, and accidental or intentional damage, which are considered to be continuous,
gradual, or unexpected [3]. Among the various causes, most pipe leakages occur due to
wall-thinning attributed to chemical and physical deterioration. The reason for this wall-
Copyright: © 2022 by the authors. thinning is corrosion and erosion, which causes unexpected leakage due to the continuous
Licensee MDPI, Basel, Switzerland. decrease in the thickness of pipe materials [3]. Corrosion and erosion reactions in pipes
This article is an open access article appear differently depending on the pipe’s intended use. The corrosion is the reaction
distributed under the terms and between a material and surrounding environment, and thus can be accelerated depending
conditions of the Creative Commons on the pH, operating temperature, and the presence or absence of chloride and sulfate ions.
Attribution (CC BY) license (https:// Generally, the pipe damage caused by corrosion appears to be larger and faster than other
creativecommons.org/licenses/by/ damage types [4].
4.0/).

Sensors 2022, 22, 7562. https://doi.org/10.3390/s22197562 https://www.mdpi.com/journal/sensors

Sensors 2022, 22, 7562 2 of 29

To date, most leak accidents have obtained attention from pipeline users or the mass
media, and there are often no programs to monitor and manage such accidents [5]. This
habituated negligence may lead to accidents due to pipe leaks, which have continued to
occur. In 2008, at the Kori Nuclear Power Plant Unit 3, which was in regular operation in
Korea, a leak occurred at the welding part of the drain valve on the B side of the steam
generator. The staff responded to the accident by manually stopping it [6,7].
Water distribution networks also possess a risk of leakage. The leak rate was reported
as 10.8% (about 720 million m3 ) of the total transported water nationwide. This leakage
could account for approximately USD 658 million of the annual production cost [8]. As
stated above, taking measures after a leak accident induces financial losses, environmental
issues, and human costs. To prevent such accidents, it is essential to have prognostics
and health management capabilities for integrity evaluations through condition-based
monitoring of the piping system.
Piping systems exhibit different pipe wall-thinning behaviors according to their vari-
ous uses and environments. In particular, corrosion behavior by the components contained
in the fluid appears to depend on industrial applications. Numerous studies have explored
the factors affecting pipe wall-thinning and sensor technology to monitor the condition of
the piping system and evaluate of the integrity of the piping system.
Diao et al. used four factors (salinity, pH, dissolved oxygen, and temperature) that
affect the corrosion and chemical composition of alloys as input data to develop a corrosion
rate (CR) prediction model for low alloys. The correlation between the input data and
CR was investigated based on machine learning algorithms and feature pre-processing
methods. Consequently, the multiple correlation coefficient (r2 ) of the actual CR value and
the predicted value was 0.9 or higher. They reported that this could be an excellent tool for
developing a CR prediction model algorithm through machine learning [9]. Sayed studied
the internal corrosion of piping systems as function of flow rate and reported that this
is one factor accelerating corrosion [10]. He also associated the failure mechanism of the
piping system with flow-enhanced corrosion.
Different types of sensors have been developed to directly evaluate piping system in-
tegrity. Among the sensors employed to monitor pipe integrity, corrosion sensors, acoustic
emission, and ultrasonic transducer (UT) sensors exhibit moderate performances when
monitoring the corrosion and wall-thinning of pipes [11]. The UT sensor employs an
operator to measure the remaining thickness of a pipe at a certain period in the field [12,13].
Measurements of the remaining pipe thickness using the UT sensor enable the determina-
tion of its remaining useful life. The sensor has some disadvantages, including a limited
measurement range according to the size of the sensor, and poor repeatability of measure-
ments depending on the skill level of the operator [14,15]. More importantly, measurements
cannot be achieved when the operator cannot access the target in real-time, such as in
a buried pipe [16], and a measurement error may be produced by the UT sensor when there
is fluid flow in the pipe [17].
Unlike the UT sensor, the corrosion sensor measures the CR of the surrounding
environment, which enables the lifetime of the pipe to be determined. This does not
require a skilled operator [18] and is the most widely used sensor for predicting the
pipeline conditions. The sensor converts the electric signal using the transmitter [19].
Representative corrosion sensors include electrical resistance (ER) and linear polarization
resistance (LPR) sensors. The ER sensor generates a signal corresponding to the metal loss;
the CR can be obtained by calculating the slope of the damage curve [18]. Compared to other
sensors, a conventional corrosion sensor can be applied to a relatively high-temperature
environment, which can be used to evaluate the effect of erosion [20,21]. The LPR sensor has
widespread use. It is convenient for the interpretation of data, as it directly calculates and
provides CR for the environment [18]. A growing number of the literature has investigated
ER and LPR sensors as techniques to monitor internal corrosion rate.
Sensors 2022, 22, 7562 3 of 29

Joosten et al. proposed a dual-element ER sensor for high-sensitivity corrosion moni-

toring technology and evaluated its stability in various environments. In their study, the
sensor showed an excellent performance in measuring the CR, and the influence of the
temperature was discussed [22]. Durrani et al. used a corrosion coupon and LPR sensor
to calculate the CR of a piping system. They claimed that, with an LPR sensor equipped
with a corroded electrode, over-predicted CR values from 20 to 44% could be induced [23],
indicating that measurement errors can be produced while the sensor is in use.
Previous studies have explored the feasibility of prototype ER or LPR sensors in limited
and controlled environments. Although several commercial sensors are available on the
market, few empirical studies have addressed their application to real-world operating
environments. To ensure the commercial viability of the corrosion sensors, the reliability
and measurement accuracy must be verified under different operating conditions in various
fluid environments. The proper selection of a sensor for pipeline integrity monitoring
requires consideration of the cost, availability, physical and chemical properties of the fluid,
and human resources for maintenance.
In this study, we investigated the performance and reliability of commercial corrosion
sensors for monitoring the integrity of the piping system in various fluid environments
as an alternative to the UT sensor. The electrochemical evaluation was conducted to
compare the measurement accuracy of the ER and LPR sensors compared to the UT sensor.
A comparative analysis of the data obtained from these sensors was performed to verify
the applicability of the corrosion sensors to monitoring wall thickness changes in pipes.
Furthermore, statistical methods, such as the regression analysis and Pearson correlation,
were employed to examine the influence of operating parameters on the sensor response.
The findings of this study suggest further development directions for corrosion sensors,
and is a useful reference for their industrial application.

2. Principle of Corrosion Sensor

2.1. Sensors for Piping System Condition and Wall-Thinning Monitoring
2.1.1. ER Sensor
The ER sensor measures an electric resistance value and transmits an output signal
as a damaged thickness value. This is similar to measuring the CR with the weight loss
method [24]. The CR measurement using the weight loss method is performed by exposing
the coupon to the actual corrosive environment or similar environments. This method
measures the damage by assessing the coupon’s weight loss after the experiment. Similarly,
the principle of the ER method is that when the sensor’s thickness is damaged by corrosion
or erosion, the shape of the sensor changes, and the measured ER increases [25]. The
change in the sensor resistance is expressed in the following equation, which is illustrated
in Figure 1 [26]. The resistance of the fresh sensing element (R1 ) before thickness reduction
is represented by t1 . As corrosion progresses, the thickness of the element is reduced from
t1 , which is the initial thickness, to t2 . Eventually, the thickness reduction (tc ) leads to an
increase in the resistance from R1 to R2 , as shown in the equation in Figure 1. The change
in the resistance of the ER sensor corresponds to a change in the metal loss.
The advantage of the ER method lies in the possibility of monitoring the damaged
thickness of the pipe in real-time by measuring the resistance value that changes over time
based on the initial resistance value. As shown in Figure 2, continuous measurements of
metal loss generate curves, where the slope gives the value of CR. Since metal losses under
the same environments may produce different CR over different periods of time, overall
CR can be determined by a linear fitting of the datapoints on the graph.
Sensors 2022,
Sensors 2022, 22,
22, 7562
7562 4 4of
of 31
29

Figure 1. Sensing principle of electric resistance (ER) sensor by thickness reduction.

The advantage of the ER method lies in the possibility of monitoring the damaged
thickness of the pipe in real-time by measuring the resistance value that changes over time
based on the initial resistance value. As shown in Figure 2, continuous measurements of
metal loss generate curves, where the slope gives the value of CR. Since metal losses under
the same environments may produce different CR over different periods of time, overall
CR can be determined by a linear fitting of the datapoints on the graph.
Figure 1.
Figure Sensing principle
1. Sensing principle of
of electric
electric resistance
resistance (ER)
(ER) sensor
sensor by
by thickness
thickness reduction.
reduction.

The advantage of the ER method lies in the possibility of monitoring the damaged
thickness of the pipe in real-time by measuring the resistance value that changes over time
based on the initial resistance value. As shown in Figure 2, continuous measurements of
metal loss generate curves, where the slope gives the value of CR. Since metal losses under
the same environments may produce different CR over different periods of time, overall
CR can be determined by a linear fitting of the datapoints on the graph.

Figure 2.
Figure An example
2. An exampleof
ofmetal
metalloss
lossversus
versustime curve
time and
curve a slope
and indicating
a slope corrosion
indicating rate.rate.
corrosion

Figure 3 illustrates the schematic diagram of a typical ER sensor [27]. The general
Figure 3 illustrates the schematic diagram of a typical ER sensor [27]. The general
operating method of the ER sensor involves transmitting a constant current to the exposed
operating method of the ER sensor involves transmitting a constant current to the exposed
and reference elements. Then, the resistance meter transmits the output value of the
specific resistance from the exposed element. If the exposed element is damaged by the
environment in which the ER sensor is placed, it transmits an increased output resistance.
In this 2.
Figure way, the resistance
An example output
of metal of thetime
loss versus exposed
curveelement is measured
and a slope indicating by a resistance
corrosion rate. meter
according to the thickness of the damaged element. The measured resistance calculates the
resistance
Figurecorresponding to the
3 illustrates the damaged
schematic thickness
diagram of through
a typicaltheERdedicated transmitter
sensor [27]. and
The general
transmits it as a voltage (1–5 V) or current (4–20 mA) signal to users. Users continuously
operating method of the ER sensor involves transmitting a constant current to the exposed
acquire the damaged thickness data of the ER sensor with the signal transmitted using
cific resistance from the exposed element. If the exposed element is damaged by the envi-
ronment in which the ER sensor is placed, it transmits an increased output resistance. In
this way, the resistance output of the exposed element is measured by a resistance meter
according to the thickness of the damaged element. The measured resistance calculates
the resistance corresponding to the damaged thickness through the dedicated transmitter
Sensors 2022, 22, 7562 5 of 29
and transmits it as a voltage (1–5 V) or current (4–20 mA) signal to users. Users continu-
ously acquire the damaged thickness data of the ER sensor with the signal transmitted
using a data acquisition (DAQ) system. Thus, a graph can be drawn and used to calculate
a data
the CR. acquisition
Nevertheless,(DAQ) system.
acquiring Thus, athickness
damaged graph can be over
data drawn and used
a certain to calculate
period the
is necessary
CR. Nevertheless, acquiring damaged thickness data over a certain period
to estimate the CR using the ER sensor. There is a disadvantage in secondary data pro- is necessary to
estimate the CR using the ER sensor. There is a disadvantage in secondary data
cessing being required after the experiment and data acquisition. Further, the thickness of processing
being
the ERrequired after theisexperiment
sensor element proportional and
todata acquisition.
its service Further, the
life. Therefore, thickness
users of the ERa
must employ
thick ER sensor element for long-term use. However, some reports claimed thatthick
sensor element is proportional to its service life. Therefore, users must employ a ER
the sen-
sitivity of the sensor may be degraded when a thick element is used [20]. According of
sensor element for long-term use. However, some reports claimed that the sensitivity to
the sensor may be degraded when a thick element is used [20]. According to Ref. [20], if
Ref. [20], if the element thickness of the sensor is manufactured in a thin form, it can be
the element thickness of the sensor is manufactured in a thin form, it can be applied to an
applied to an environment with a low CR. Li et al. reported that the highly sensitive thin
environment with a low CR. Li et al. reported that the highly sensitive thin film electric
film electric resistance sensor, developed in the form of a thin film, showed a good per-
resistance sensor, developed in the form of a thin film, showed a good performance when
formance when applied to various corrosive environments [28]. Therefore, the user must
applied to various corrosive environments [28]. Therefore, the user must accurately identify
accurately identify the environment in which the sensor is to be used and select the ele-
the environment in which the sensor is to be used and select the element thickness of the
ment thickness of the sensor accordingly.
sensor accordingly.

Schematicdiagram
Figure3.3.Schematic
Figure diagram of
of ER
ER sensor
sensor and
and data
data acquisition
acquisition system.
system.
2.1.2. LPR Sensor
2.1.2. LPR Sensor
The LPR sensor employs the principle of the linear polarization resistance method [29].
The The
LPRLPRmethodsensor
canemploys
simply andthe quickly
principlecalculate
of the linear
a wide polarization
range of CRs resistance method
[30]. Thus, it is
[29]. The LPR method can simply and quickly calculate a wide range
used in various fields owing to the high accuracy of the measured values. As describedof CRs [30]. Thus, it
isabove,
used in various fields owing to the high accuracy of the measured values.
with the LPR method, the CR can be quickly calculated and monitored such that As described
above,
the userwith
can the LPR method,
immediately the CR
respond can be quickly
to emergencies [31].calculated and monitored
Further, because such that
the LPR method is
the useroncan
based an immediately
electrochemical respond to emergencies
process, it is generally[31]. Further,with
performed because
threethe LPR method
electrodes (the
isreference,
based oncounter,
an electrochemical
and working process, it is generally
electrode), and usesperformed with between
the correlation three electrodes (the
the current
reference,
density and counter, and working
the potential generatedelectrode), and uses the correlation
by the electrochemical between
reaction between thethe current
electrodes.
densityTheandLPRthe potential
sensor worksgenerated
by applying bya the
smallelectrochemical reaction the
voltage and measuring between
ratio ofthe elec-
voltage
trodes.
to current. When a power source is applied to the sensor electrodes, the current flows from
the working
The LPRelectrode to theby
sensor works counter electrode
applying a small in voltage
electrolyte
andsolution.
measuring To determine
the ratio ofpolariza-
voltage
tion
to resistance,
current. Whenthe currentsource
a power densityis and potential
applied to theshould
sensor first be determined.
electrodes, Therefore,
the current the
flows from
current density of the working electrode opposite the counter electrode is measured, and,
at the same time, the potential of the working electrode is measured against the reference
electrode. The measured current density value is inversely proportional to the ER value
according to Ohm’s law.
For example, assume that a low potentiodynamic polarization is applied between the
working and reference electrodes. Then, the current flows from the working electrode to
the counter electrode in the electrolyte solution. If the current density between the working
and counter electrodes is high, the electrical resistance of the working electrode is said
to be low. This means that the working electrode is easily ionized and the CR is fast [27].
Sensors 2022, 22, 7562 6 of 29

Using this principle, the LPR sensor calculates the polarization resistance (Rp ) from the
Sensors 2022, 22, 7562 linear slope of the relationship between the overpotential and current density by using 6 of 31
the overpotential within ± 20 mV at the maximum from the corrosion potential (Ecorr )
of the working electrode, measured against the reference electrode [27,32]. Furthermore,
by using the calculated Rp , the corrosion current density (icorr ) and CR are calculated in
the working electrode to the counter electrode in electrolyte solution. To determine polar-
real-time, and CR data can be acquired through a transmitter dedicated to the LPR sensor.
ization resistance, the current density and potential should first be determined. Therefore,
Consequently, the LPR sensor can only measure the CR when the electrochemical reaction
the current density of the working electrode opposite the counter electrode is measured,
(relationship between overvoltage and current) between the metal and aqueous solution has
and, at the same time, the potential of the working electrode is measured against the ref-
a linear relationship (Tafel behavior) [33]. Figure 4 shows a typical potential-current density
erence
graphelectrode.
for metallicThe measured
materials in acurrent
corrosive density
environmentvalue is[34]. inversely proportional
The relationship to thethe
between ER
value according to Ohm’s law.
potential-current density is linear near the Ecorr , which yields the icorr . This can be derived
For example,
according assume that aequation
to the Stern–Geary low potentiodynamic
and uses the mixed polarization is applied
potential betweenThe
theory [27,34]. the
working
relationship and between
referencethe electrodes. Then, and
current density the overpotential
current flowsinfrom the working
a solution electrode
following the Tafelto
the counter electrode in the
behavior is shown in the graph [35]. electrolyte solution. If the current density between the work-
ing and counter electrodes is high, the electrical resistance of the working electrode is said
to be low. This means that the working ia
Ea =electrode
β a (log is easily ) ionized and the CR is fast [27].(1)
Using this principle, the LPR sensor calculates the i corr polarization resistance (Rp) from the
linear slope of the relationship between the overpotential ic and current density by using
the overpotential within ± 20 mV at Ethe c =maximum
β c (log
icorr
)
from the corrosion potential (Ecorr(2)) of
the working electrode, measured against the reference electrode [27,32]. Furthermore, by
Ea = Ea − Ecorr , Ec = Ec − Ecorr (3)
using the calculated Rp, the corrosion current density (icorr) and CR are calculated in real-
time, and constant
β a : Tafel CR data of can be acquired
anodic reaction through
(slope of astraight
transmitter line indedicated to the LPR sensor.
E–logI curve).
β c : Tafel constant
Consequently, of cathodic
the LPR sensor reaction
can only(slopemeasure of straight
the CR when line inthe
E–logI curve).
electrochemical reaction
Ea : Tafel constant
(relationship betweenof anodic reaction
overvoltage and (slope of straight
current) between linethe
in E–logI
metal curve).
and aqueous solution
Ec :aTafel
has linear constant of cathodic
relationship (Tafelreaction
behavior) (slope
[33].ofFigure
straight4 line showsin E–logI curve).
a typical potential-current
density graph for metallic materials in a corrosive environment [34]. The relationship be-
tween the potential-current density ic
i app.isc linear
= ic −near
i a (logthe Ecorr ) , which yields the icorr. This (4) can
i corr
be derived according to the Stern–Geary equation and uses the mixed potential theory
[27,34].
i app. c : The relationship
Applied cathodic between the current density and overpotential in a solution fol-
current density.
i a : Current
lowing the Tafel behavior
density is shown
of anodic in the
oxidation graph [35].
reaction.
ic : Current density of cathodic reduction reaction.

Figure 4. Hypothetical linear polarization resistance curve and polarization resistance (Rp ).
Figure 4. Hypothetical linear polarization resistance curve and polarization resistance (Rp).

𝑖
𝐸 = 𝛽 (log ) (1)
𝑖
𝑖
Sensors 2022, 22, 7562 7 of 29

If Equations (1) and (2) are fixed in exponential form and then substituted into
Equation (4), Equation (5) is obtained.

− βε cc εa
i app. c = icorr (10 − 10 β a ) (5)

ε a : Anodic overpotential.
ε c : Cathodic overpotential.
Equation (5) approaches a straight line as the overpotential becomes close to zero
(Ecorr ). Hence, when it is differentiated, the equation for Rp is obtained as follows.

∆ε

dε βa βc B
Rp = − = = (6)
di app ε →0
∆i app ε →0
2.3 icorr ( β a + β c ) icorr

βa βc
B= (7)
2.3( β a + β c )
B = Tafel proportionality constant.
The CR is calculated according to the Faraday’s law of the icorr , calculated through
Equation (5), and Equation (8) is used to calculate the CR is as follows.

icorr
CR = K1 EW (8)
ρ

K1 = 3.27 × 10−3 , mm·g/µA·cm·yr

icorr = Corrosion current density
ρ = Density in g/cm3
EW = Equivalent weight
The advantage of the LPR sensor is that, compared to the ER sensor, the CR can
be directly monitored without additional data-processing. However, a disadvantage is
that it may be difficult to accurately measure the CR when the environment in which the
electrode is placed does not form a linear relationship between overpotential and current
density [34,36]. Hence, the CR measurement using the LPR sensor can only be accurately
measured when the electrochemical reaction of the electrode follows a constant reaction
rate mechanism (Tafel behavior). Further, although the performance may differ depending
on the maker, the LPR sensor has limitations in severely corrosive environments, as it can
generally measure up to a maximum CR of 200 mils penetration per year (MPY).
Sensors 2022, 22, 7562 Figure 5 illustrates a schematic diagram of the three electrode probes type of a com-
8 of 31
mercial LPR sensor [27]. The probe of the LPR sensor consists of three electrodes (working,
counter, and reference electrodes) made of the same metal. They can easily be assembled or
disassembled using screws [27].

Figure 5. Schematic diagram of LPR sensor with three removable electrodes (CE, RE and WE).
Figure 5. Schematic diagram of LPR sensor with three removable electrodes (CE, RE and WE).
3. Experimental Methods
3. Experimental
3.1. Test Bed forMethods
Simulating Piping System
3.1. TestABed
testfor
bedSimulating Piping
was designed andSystem
constructed to monitor wall-thinning damage to the pipe
byAcorrosion,
test bed as shown
was in Figure
designed and 6.constructed
The test bedtoconsisted
monitorofwall-thinning
a water tank, damage
strainer, cen-
to the
pipe by corrosion, as shown in Figure 6. The test bed consisted of a water tank, strainer,
centrifugal pump, valves, and metal piping to simulate the piping system. An air com-
pressor was connected to the water tank to ensure the smooth operation of the centrifugal
pump, and constant pressure was maintained using a pressure regulator. As the move-
3. Experimental Methods
3.1. Test Bed for Simulating Piping System
A test bed was designed and constructed to monitor wall-thinning damage to the
pipe by corrosion, as shown in Figure 6. The test bed consisted of a water tank, strainer,
Sensors 2022, 22, 7562 8 of 29
centrifugal pump, valves, and metal piping to simulate the piping system. An air com-
pressor was connected to the water tank to ensure the smooth operation of the centrifugal
pump, and constant pressure was maintained using a pressure regulator. As the move-
trifugal pump, valves, and metal piping to simulate the piping system. An air compressor
ment of the fluid generates frictional heat from the centrifugal pump, causing the fluid
was connected to the water tank to ensure the smooth operation of the centrifugal pump,
temperature
and constanttopressure
increase,wasthe temperature
maintained usingof the fluidregulator.
a pressure in the tank was
As the kept constant
movement of the using
an fluid
external chiller.
generates To measure
frictional heat from thethe
degree of corrosion
centrifugal damage,
pump, causing the grade
the fluid SS275to(carbon
temperature
steel) straight
increase, the pipe was used
temperature as fluid
of the the test pipe
in the tankfor thickness
was measurements
kept constant and analysis,
using an external chiller. and
To measure the degree of corrosion damage, the grade SS275 (carbon steel)
grade 304 stainless steel was used for the rest of the piping system. The carbon steel was straight pipe
was used
chosen as the
as the testtest pipe forbecause
material thicknessitmeasurements
is widely used andasanalysis, and grade
a pipeline 304 stainless
material in various in-
steel was used for the rest of the piping system. The carbon steel was chosen as the test
dustries due to its high strength, durability, moderate corrosion resistance, and cost-effec-
material because it is widely used as a pipeline material in various industries due to its
tiveness [37–39].durability, moderate corrosion resistance, and cost-effectiveness [37–39].
high strength,

Figure
Figure 6. Schematicdiagram
6. Schematic diagram of test
testbed
bedfor
forevaluating
evaluatingcommercial corrosion
commercial sensors
corrosion under various
sensors under various
operational conditions: 1: pressure sensor, 2: D.P sensor, 3: temperature sensor, 4: motor current
operational conditions: 1: pressure sensor, 2: D.P sensor, 3: temperature sensor, 4: motor current
sensor,
sensor, 5: ER
5: ER sensor,6:6:LPR
sensor, LPR sensor.
sensor.

The test bed was equipped with various sensors, including a pressure sensor (1),
differential pressure sensor (2), temperature sensor (3), pump motor current sensor (4), ER
sensor (5), and LPR sensor (6) to monitor the operating conditions and corrosion degree of
the piping system. Table 1 lists the measurement ranges and output signals of these sensors.
The pipeline had specially designed ports for the flush mounting of the corrosion
sensors. The corrosion sensors that were used are commercially available and their sensing
element was made of carbon steel so that the sensor’s response could be correlated with the
wall thinning of the carbon steel test pipe. Corrosion sensors (ER and LPR) can measure
the metal loss, and thus determine CR in real-time. As shown in Table 1, the ER sensor can
measure the metal loss in the range of 0–10 mils, and the LPR sensor was set to measure
the CR of 0–200 MPY.
Sensors 2022, 22, 7562 9 of 29

Table 1. Type of sensors and their properties.

Name of Sensor Measuring Range Output Signal

Pressure sensor for strainer inlet 0–100 kPa
Pressure sensor for strainer outlet 0–100 kPa
Pressure sensor for pump outlet 0–1 MPa
Differential pressure sensor for strainer 0–100 kPa
4–20 mA
Temperature sensor for fluid −50–250 ◦ C
Current transmitter for pump 0–5 A
LPR sensor 0–200 MPY
ER sensor 0–10 mils

Generally, as the corrosion reaction of the piping system occurs, or the chloride
concentration increases, metals or chloride are dissolved in the test solution, resulting in
a decrease in pH and an increase in electrical conductivity. To measure such changes, pH
and electrical conductivity were measured after sampling the test solution in a tank under
changing test conditions. The measured data were interpolated and compared with ER and
LPR sensor data.

3.2. Data Acquisition from Sensors

The DAQ system was developed using a low-cost embedded system (Raspberry Pi
4B) interfaced with an analog-digital converter (resolution: 24 bit, sample rate: 30 Ks) to
process the sensor signals (Figure 7). The system used a power supply module (24 VDC) for
sensors with a 4–20 mA current loop or a 0–5 V signal. Linux (Ubuntu 18.04) was chosen as
the operating system for the unit, and robot operating system (ROS2, Foxy) was chosen
as the middleware framework. The ROS was used as a platform for data acquisition as it
provides various functionalities and interfaces. In an ROS network, each sensor has its own
node, which is the basic unit of execution in ROS. The ROS nodes establish the network
by sending and receiving messages on a topic using a publish–subscribe communication
model. The real-time sensor data were transmitted to the PC via an LAN for monitoring
and then stored on the PC. The graphic user interface was built to display real-time data
Sensors 2022, 22, 7562 and operation using PyQT5, one of the cross-platform interface toolkit. All components 10 for
of 31
the DAQ system were packaged in a three-dimensional (3D)-printed enclosure designed
for wall mounting near the test bed.

Photoof
Figure7.7.Photo
Figure of the
the DAQ
DAQ system
system and
and the
theGUI
GUIfor
formonitoring
monitoringand
andcontrol of of
control sensor data.
sensor data.

3.3. Test Profile for Pipe Condition Monitoring

Figure 8 illustrates the test profile of this study. To monitor the conditions and ac-
quire data in various environments of the actual piping system, three test sets were ap-
Sensors 2022, 22, 7562 10 of 29
Figure 7. Photo of the DAQ system and the GUI for monitoring and control of sensor data.

3.3.
3.3. Test Profile for
Test Profile for Pipe
Pipe Condition
Condition Monitoring
Monitoring
Figure 8 illustrates the test profileofofthis
Figure 8 illustrates the test profile thisstudy.
study.
ToTo monitor
monitor thethe conditions
conditions andand ac-
acquire
quire data in various environments of the actual piping system, three test sets
data in various environments of the actual piping system, three test sets were applied in were ap-
plied in sequential
sequential order. Further,
order. Further, two failure
two failure scenarios scenarios were selected
were selected to realize
to realize the failure
the failure of the
of the piping system and the DAQ system that may occur during
piping system and the DAQ system that may occur during operation in the field,operation in theand
field,
the
and thesystem
piping piping data
system data
were were acquired
acquired simultaneously.
simultaneously.

Figure
Figure 8.
8. Proposed
Proposed test
test profile
profile with
with three
three programs
programsincluding
includingtwo
twofailure
failurescenarios
scenarios(P1:
(P1:NaCl
NaCl0%,
0%,
P2:
P2: NaCl
NaCl 1.75%,
1.75%, P3:
P3: NaCl
NaCl 3.5%,
3.5%, S1:
S1: failure
failureof
ofpiping
pipingsystem,
system,S2:
S2:failure
failureof
ofDAQ
DAQcommunication).
communication).

To analyze
analyze thethe environmental
environmentalfactors factorsaffecting
affectingthethewall-thinning
wall-thinningofof thethe piping
piping sys-
system,
tem, the NaCl
the NaCl concentration,
concentration, pump pump start/stop,
start/stop, and and
fluidfluid temperature
temperature werewere
chosenchosen as var-
as variables.
iables.
In the In thetest
first firstset
test(P1),
set (P1), the start/stop
the start/stop of of
thethe pumpwas
pump wasrepeated
repeatedin in the
the order of offluid
fluid
temperature
temperature 25, 25,30,
30,andand3535°C ◦ Catata achloride
chloride concentration
concentration of of
0%. This
0%. scenario
This scenariowaswas
used to
used
acquire the piping system data in a situation where the piping is
to acquire the piping system data in a situation where the piping is empty due to failure empty due to failure or
maintenance.
or maintenance. In the Infirst
the test
firstset scenario
test (S1), the
set scenario data
(S1), thewere
data acquired from various
were acquired from sensors
various
after
sensorsan after
emergency stop of stop
an emergency the pump
of the under
pump 0% chloride
under concentration
0% chloride and 35and
concentration 35 ◦ C
°C fluid
temperature.
fluid temperature.In the In second test settest
the second (P2),
setthe experiment
(P2), was performed
the experiment was performedunder under
the same the
conditions
same conditionsas in the
as first
in thetestfirst
set test
at a chloride concentration
set at a chloride of 1.75%.of
concentration In1.75%.
the second scenario
In the second
scenario
(S2), (S2), the connection
the connection cable between
cable between the sensor theand
sensor
the and
datathe data acquisition
acquisition device wasdevice was
inten-
intentionally
tionally disconnected
disconnected at a chloride
at a chloride concentration
concentration of 1.75%
of 1.75% and a and
fluidatemperature
fluid temperature
of 35
°C to ◦acquire
of 35 C to acquire
status status data according
data according to the failure
to the failure or poororconnection
poor connection of the sensor,
of the sensor, DAQ
DAQ device, and connection cable. In the third test set (P3), the test was performed with
3.5% chloride concentration, all other conditions being the same.

3.4. UT Sensor for Measuring Test Pipe Thickness

A UT sensor (Ultrasonic Thickness Gage, MTG6DL-TXC) was used to measure the
remaining thickness of the test pipe made of carbon steel after distinguishing it into
top, middle, and bottom sides, as shown in Figure 9. The thickness measurement was
repeated for a total of 21 points throughout the experiment. To confirm the accuracy
of the measurements obtained from the UT sensor, the average and standard deviation
of the measured data were calculated, and the average value of the calculated damaged
thickness was compared with the damaged thickness obtained with the ER sensor. After the
experiment, the test pipe was disassembled and cut using a fine cutting machine, and the
inner surface of the test pipe was observed with a 3D measuring laser confocal microscope
(OLS 5000, OLYMPUS, Tokyo, Japan) and a digital camera.
for a total of 21 points throughout the experiment. To confirm the accuracy of the meas-
urements obtained from the UT sensor, the average and standard deviation of the meas-
ured data were calculated, and the average value of the calculated damaged thickness was
compared with the damaged thickness obtained with the ER sensor. After the experiment,
the test pipe was disassembled and cut using a fine cutting machine, and the inner surface
Sensors 2022, 22, 7562 11 of 29
of the test pipe was observed with a 3D measuring laser confocal microscope (OLS 5000,
OLYMPUS, Tokyo, Japan) and a digital camera.

Figure 9. Illustration of remaining thickness measuring points on the test pipe with UT sensor.
Figure 9. Illustration of remaining thickness measuring points on the test pipe with UT sensor.
3.5. Electrochemical Test
3.5. Electrochemical Test
An electrochemical experiment was performed to compare the accuracy of commercial
An electrochemical experiment was performed to compare the accuracy of commer-
corrosion sensors (ER, LPR sensor). A potentiostat (Bio-Logic Science Instrument, FR/VCP)
cial corrosion sensors (ER, LPR sensor). A potentiostat (Bio-Logic Science Instrument,
was used for electrochemical experiment, and potentiodynamic polarization and LPR
FR/VCP) was used for electrochemical experiment, and potentiodynamic polarization and
experiments were performed to calculate the E , icorr , and CR. As shown in Figure 10,
LPR experiments were performed to calculate the Ecorr corr, icorr, and CR. As shown in Figure
thethe
10, electrochemical
electrochemical cell wascomposed
cell was composedofof a three-electrode
a three-electrode cell consisting
cell consisting of a working
of a working
electrode, a counter electrode (platinum mesh, 2 cm × 2 cm),
electrode, a counter electrode (platinum mesh, 2 cm × 2 cm), and a reference electrodeand a reference electrode
(Ag/AgCl electrode saturated by KCl). The working electrode used carbon steel, which is which
(Ag/AgCl electrode saturated by KCl). The working electrode used carbon steel,
is the
the samesame material
material as theas
testthe test
pipe of pipe ofbed.
the test the It
test
wasbed.mountedIt was mounted
with epoxy afterwith epoxy after
cutting
cutting to a size of 1 cm × 1 cm. To make the surface of the working
to a size of 1 cm × 1 cm. To make the surface of the working electrode as smooth as possi- electrode as smooth as
possible, the surface was polished using 2000 grit SiC paper. The test
ble, the surface was polished using 2000 grit SiC paper. The test solution was the same as solution was the same
the fluid used in the test bed, and the test conditions are summarized in Table 2. The po- 2. The
as the fluid used in the test bed, and the test conditions are summarized in Table
potentiodynamic
tentiodynamic polarization
polarization experiments
experiments were were conducted
conducted at a scanning
at a scanning rate ofrate of 1.0 mV/s
1.0 mV/s
Sensors 2022, 22, 7562 with
withapplied
appliedpotential
potentialfrom
from −0.25
−0.25 to 0.7 V based
to 0.7 on the
V based on open-circuit
the open-circuitpotential (OCP)
potential 31after after a
12 of (OCP)
arest
resttime
time of 900
900 s.
s. The
TheLPRLPRexperiment
experiment waswasperformed
performed at a at
scanning rate of
a scanning 10 mV/min
rate of 10 mV/min
from
from−30−30to to
30 30
mVmV based on OCP
based on OCPafterafter
the rest
the time of 900ofs.900 s.
rest time

Figure 10. Schematic diagram of three-electrode cell with WE, RE and CE and potentiostat for
Figure 10. Schematic diagram of three-electrode cell with WE, RE and CE and potentiostat for elec-
electrochemical
trochemical experiments.
experiments.

Table 2. Experimental conditions for the electrochemical tests to compare results of corrosion rate
with the commercial corrosion sensors.

No. 1 2 3 4 5 6 7 8 9
Sensors 2022, 22, 7562 12 of 29

Table 2. Experimental conditions for the electrochemical tests to compare results of corrosion rate
with the commercial corrosion sensors.

No. 1 2 3 4 5 6 7 8 9
NaCl (%) 0 1.75 3.5
Temp. (◦ C) 25 30 35 25 30 35 25 30 35

To calculate the Ecorr and icorr after the potentiodynamic polarization experiment, the
Tafel extrapolation method was employed in the range of ±0.25 V based on OCP, as shown
in Figure 11. The Tafel extrapolation method measures the Ecorr and icorr using a cathodic or
anodic polarization curve based on the mixed potential theory. In the obtained curve, when
the current density increases, a region changing to a straight line appears (i.e., a straight
Sensors 2022, 22, 7562 region on polarization curve). This is called a Tafel region, and the Ecorr and icorr can
13 of 31 be
measured by extrapolating the Tafel region and using the intersecting point. The CR can be
calculated by applying the measured icorr to Faraday’s law [40].

Figure 11. Determination of corrosion potential and current density using the Tafel extrapo-
Figure 11. Determination of corrosion potential and current density using the Tafel extrapolation
lation method on the Evans diagram for the oxidation and reduction reactions of metals in
method on the Evans diagram for the oxidation and reduction reactions of metals in corrosive envi-
corrosive environment.
ronment.

After the LPR experiment, the E , icorr , and Rp were calculated using an analysis tool
After the LPR experiment, the Ecorr,corr
icorr, and Rp were calculated using an analysis tool
(EC-Lab software V11.20, Bio-Logic, USA). LPR techniques measure the Ecorr and icorr using
(EC-Lab software V11.20, Bio-Logic, USA). LPR techniques measure the Ecorr and icorr using
electrochemical principles. Compared to the potentiodynamic polarization experiment,
electrochemical principles. Compared to the potentiodynamic polarization experiment,
LPR can perform measurements within a short time and in the low overpotential region
LPR can perform measurements within a short time and in the low overpotential region
within ±20 V from the Ecorr . The relationship between potential and current density is
within ±20 V from the Ecorr. The relationship between potential and current density is lin-
linear in Equation (9).
ear in Equation (9). ∆E βa βc
Rp = = (9)
∆i
∆𝐸 𝛽 𝛽( β a + β c )
2.3 icorr
𝑅 = = (9)
∆𝑖 requires
Polarization resistance measurement 2.3 𝑖 (𝛽
the +𝛽 )
determination of anodic and cathodic
Tafel constants,resistance
Polarization which aremeasurement
normally given in thethe
requires range of ±20 mV of
determination foranodic
most metals
and ca-and
alloys. In this study, we determined the Tafel constants within ±20 mV because any error
thodic Tafel constants, which are normally given in the range of ±20 mV for most metals
generated from Tafel constants can be neglected when evaluating the relative corrosion
andrate.
alloys. In this study,
To compare we according
the CRs determined
tothe Tafel
each method,±20
constants within
experimental themV
CR because any
was calculated
error generated
using the icorrfrom Tafelby
obtained constants
the LPRcan be neglected
method [41]. when evaluating the relative corro-
sion rate. To compare the CRs according to each experimental method, the CR was calcu-
lated using the icorr obtained by the LPR method [41].

4. Results and Discussion

4.1. ER Sensor and LPR Sensor
4.1.1. ER Sensor

Effect of NaCl concentration

Sensors 2022, 22, 7562 13 of 29

4. Results and Discussion

4.1. ER Sensor and LPR Sensor
4.1.1. ER Sensor
ors 2022, 22, 7562 Effect of NaCl Concentration 14
Figure 12 illustrates a graph showing the data obtained using a conductivity meter,
pH meter, pressure sensor, and temperature sensor for comparison with the metal loss
value measured by the ER sensor.

Figure 12. Plot of raw data obtained from different sensors and metal loss from ER sensor.
Figure 12. Plot of raw data obtained from different sensors and metal loss from ER sensor.
In Figure 12, the electrical conductivity and pH data of stacks 1 and 2 are shown
by interpolating the data that were measured after collecting the fluid sample at regular
In Figureduring
intervals 12, the the electrical
experiment.conductivity and pH
Stack 3 is illustrated data
using theof stacks
NaCl 1 and 2 are
concentration as show
the experiment variables, and stacks 4, 5, and 6 are graphs showing
interpolating the data that were measured after collecting the fluid sample at regula the data that were
converted to the corresponding value after acquiring the current signal in real-time using
tervals during the experiment. Stack 3 is illustrated using the NaCl concentration a
the sensor installed in the test bed.
experiment variables,
As shown and12,
in Figure stacks
in the 4, 5, and
NaCl 6 are
0% test graphs
condition, showing
both the and
conductivity datametal
that were
vertedloss increased with an increase in experiment time and fluid temperature, whereas the using
to the corresponding value after acquiring the current signal in real-time
sensorpHinstalled
decreased. in This is attributed
the test bed. to the dissolution of metal and particles in the fluid
due to damage to the pipe. Generally, the metal ions are desorbed into the solution by
As shown in Figure 12, in the NaCl 0% test condition, both conductivity and m
an electrochemical reaction (corrosion), such that the conductivity increases and the pH of
loss increased
the solutionwith an increase
is lowered in experiment
by the hydrolysis reaction. time and fluid temperature,
This phenomenon whereas th
was more accelerated
decreased.
when NaCl Thisisisadded
attributed to the
to the fluid. dissolution
In the of metal
case of the second and and
third particles
programs, in
1.75the
andfluid du
3.5% of NaCl was added, respectively. Consequently, the conductivity,
damage to the pipe. Generally, the metal ions are desorbed into the solution pH, and metal loss by an
graph exhibited rapid fluctuations.
trochemical reaction (corrosion), such that the conductivity increases and the pH o
solution
First is lowered
Failure by the hydrolysis reaction. This phenomenon was more acceler
Scenario
when NaCl is added
Figure 13 showstothe
the fluid.
results In the case
of extracting of the corresponding
the section second and tothird programs,
the first failure 1.75
scenario in Figure 11, depicting the state of the ER sensor when the
3.5% of NaCl was added, respectively. Consequently, the conductivity, pH, and metalpiping system is
empty due to failure or ongoing repair. As shown in Figure 13, the ER sensor operates
graphnormally,
exhibited rapid fluctuations.
without errors, even if the pipe is empty. In particular, this showed a similar

First failure scenario

Figure 13 shows the results of extracting the section corresponding to the first fa
scenario in Figure 11, depicting the state of the ER sensor when the piping system is em
nsors 2022, 22, 7562 15 o

Sensors 2022, 22, 7562 14 of 29

ors 2022, 22, 7562 15

tendency to the state in which the fluid that does not flow was filled. Further, normal
operation was achieved even when experiments were performed consecutively without
special maintenance. In conclusion, the ER sensor does not affect the operating state of the
sensor itself, even if it is not immersed in the fluid.

Figure 13. (Extracted from Figure 12) Enlarged view of data showing the sensor responses dur
the first failure scenario (S1).

Effect of fluid flow

Figure
Figure 13.14 showsfrom
(Extracted theFigure
effect12)ofEnlarged
the fluid
viewflow
of dataon the pipe
showing wall-thinning
the sensor after
responses during the extract
the range
first
Figure corresponding
13.failure scenario from
(Extracted to the12)
(S1). Figure test conditions
Enlarged viewof of NaCl 0%. Thethe
data showing metal
sensorloss trends du
responses w
analyzed
the first for
failure
Effect the run or
scenario
of Fluid Flow stoppage period. When the pump of the piping system was r
(S1).
the metalFigure
loss increased by about 0.213 mil. During the pump stoppage, the metal loss w
14 shows the effect of the fluid flow on the pipe wall-thinning after extracting
increased
Effect by corresponding
ofrange
the fluid about
flow 0.018 mil. to theUsing the metal
test conditions loss values,
of NaCl 0%. The the
metalCRs
losscan bewere
trends calculated
pump run and
analyzed
Figure
for stoppage
the
14 shows
run or period,
stoppage
the effect ofand
period.were flow
the fluid
When 44.594
the
onMPY
pump of and wall-thinning
the
the pipe 34.698 MPY, respectively
piping system was run,
after extrac
highertheCRmetal
was lossnoted
increased by about
for the pump 0.213
runmil. During thewhich
conditions, pump stoppage,
might bethe metal loss to eros
attributed
the range corresponding
was increased by about 0.018to the
mil.test
Using conditions
the metal lossof values,
NaCl the 0%. The
CRs canmetal loss trends w
be calculated
being more dominant than corrosion under the test conditions of NaCl 0%. In addition
analyzed for the
for pump runstoppage
run and or stoppage
period,period.
and were When
44.594the
MPY pump of the
and 34.698 MPY,piping system was
respectively.
this, Awhen
higher NaCl,
CR was which
noted accelerates
for the pump metal
run corrosion,
conditions, which is present
might be in the to
attributed fluid, the effec
erosion
the metal loss increased by about 0.213 mil. During the pump stoppage, the metal loss
thought
beingtomore
be more significant
dominant [42,43].
than corrosion In addition,
under as the thickness
the test conditions of NaCl 0%.ofInthe diffusion la
addition
increased by about 0.018 mil. Using the metal loss values, the CRs can be calculated
to this, when NaCl, which accelerates metal corrosion, is present in
for dissolved oxygen present in the fluid decreases, the CR of the piping system increa the fluid, the effect
pump is run andtostoppage
thought period, and were
be more significant 44.594 MPY and 34.698 MPY, respectivel
[44–46]. To examine the possible [42,43]. In addition,
synergistic effect as
of the
thethickness
fluid flowof the diffusion
and the presence
higher CRfor
layer was notedoxygen
dissolved for thepresent
pumpinrun the conditions,
fluid decreases, which
the CR might
of the be attributed
piping system to ero
NaCl on the [44–46].
metal loss, the following twosynergistic
conditions were compared: flowNaCl 0% at 35
beingincreases
more dominant Tothan
examine the possible
corrosion under the test effect of the
conditions fluid
of NaCl and the
0%. In additio
and NaCl 3.5% at 35 °C. The two sets of data are shown in Figure 15.
this, presence
when NaCl, of NaCl on the metal loss, the following two conditions were compared: NaCl 0%
which accelerates metal corrosion, is present in the fluid, the effe
at 35 ◦ C and NaCl 3.5% at 35 ◦ C. The two sets of data are shown in Figure 15.
thought to be more significant [42,43]. In addition, as the thickness of the diffusion l
for dissolved oxygen present in the fluid decreases, the CR of the piping system incre
[44–46]. To examine the possible synergistic effect of the fluid flow and the presenc
NaCl on the metal loss, the following two conditions were compared: NaCl 0% at 3
and NaCl 3.5% at 35 °C. The two sets of data are shown in Figure 15.

Figure 14. (Extracted from Figure 12) Data showing the effect of the fluid flow on the pipe wall-
Figure 14. (Extracted
thinning from
under the test Figure
condition of 12)
NaClData
0%. showing the effect of the fluid flow on the pipe w
thinning under the test condition of NaCl 0%.

Figure 14. (Extracted from Figure 12) Data showing the effect of the fluid flow on the pipe w
loss due to the on/off action of the piping system. This could be attributed to the corros
reaction being insignificant in the NaCl 0% environment. In contrast, in the case of N
3.5% in Figure 15b, which is a severely corrosive environment, the fluid flow showed
opposite trend in terms of metal loss: a gradual increase when the piping system i
Sensors 2022, 22, 7562 15 of 29
operation, and a rapid decrease when the fluid flow is stopped.

(a)

(b)
FigureFigure 15. (Extracted
15. (Extracted from
from Figure 12)
Figure 12) Data
Datashowing synergistic
showing effect of
synergistic NaClof
effect and fluid and
NaCl flow fluid
in the flow in
presence and absence of NaCl: (a) NaCl 0% condition, (b) NaCl 3.5% condition.
presence and absence of NaCl: (a) NaCl 0% condition, (b) NaCl 3.5% condition.
Synergistic Effect of NaCl and Fluid Flow
CorrosionAs shown
rate in test
with Figure 15a, in the case of NaCl 0%, there was little difference in metal
parameters
loss due to the on/off action of the piping system. This could be attributed to the corrosion
Figure
reaction16a
beingillustrates thein comparison
insignificant of the CR with
the NaCl 0% environment. the test
In contrast, parameters
in the case of NaClafter ca
lating3.5%
the former
in Figureusing the metal
15b, which loss data
is a severely acquired
corrosive with thetheER
environment, sensor.
fluid For the data r
flow showed
anthe
ability, opposite
CR wastrendcalculated
in terms of metal loss:
for the a gradualin
variance increase
metalwhen
loss the piping system is
corresponding toina period
operation, and a rapid decrease when the fluid flow is stopped.
1440 min. Overall, as the experiment progressed, the CR increased and decreased depe
ing onCorrosion
whether thewith
Rate pump was operated. The CR appeared to rise with NaCl. Figure
Test Parameters
details theFigure
CR trends in thethe
16a illustrates early stage ofof the
comparison testwith
the CR forthe
further analysis.after
test parameters The analysis res
calculat-
ing the former using the metal loss data acquired with the ER sensor. For the
indicate that if the change in the variables is not significant (only the pH and conductiv data reliability,
the CR was calculated for the variance in metal loss corresponding to a period of 1440 min.
are changed), the CR increases to a specific value and then stabilizes or stagnates. T
Overall, as the experiment progressed, the CR increased and decreased depending on
exhibits a similar
whether tendency
the pump to the corrosion
was operated. of general
The CR appeared metals,
to rise with which
NaCl. Figurehave the stagnan
16b details
stabilizing characteristics of the CR, while forming an oxide film on the surface with
initial increase. Vasyliev studied the CRs of metals in the presence of flow rates [47]. T
Sensors 2022, 22, 7562 16 of 29

the CR trends in the early stage of the test for further analysis. The analysis results indicate
that if the change in the variables is not significant (only the pH and conductivity are
Sensors 2022, 22, 7562 17 of 31
changed), the CR increases to a specific value and then stabilizes or stagnates. This exhibits
a similar tendency to the corrosion of general metals, which have the stagnant or stabilizing
characteristics of the CR, while forming an oxide film on the surface with its initial increase.
experimental
Vasyliev studied results
the CRsshow that the
of metals CR presence
in the of metalsofexposed to a[47].
flow rates corrosive environment
The experimental
results show that the CR of metals exposed to a corrosive environment tendedperiod.
tended to decrease over time, and a constant CR is maintained over a certain This
to decrease
is in good agreement with the results derived in this study.
over time, and a constant CR is maintained over a certain period. This is in good agreement
with the results derived in this study.

Figure 16. Sensor data including corrosion rate calculated from ER sensor: (a) effects of fluid
Figure on
properties 16. corrosion
Sensor data including
rate, corrosion
(b) evolution rate calculated
of corrosion from ER
rate during thesensor:
test. (a) effects of fluid prop-
erties on corrosion rate, (b) evolution of corrosion rate during the test.
Second Failure Scenario
Second
Figurefailure scenarioan enlarged graph of the corresponding section, which analyzes
17 illustrates
the changeFigure 17 illustrates
in the CR duringanthe enlarged
failuregraph
of theof the corresponding
DAQ, which representssection,
the which
secondanalyzes
failure
the change
scenario. The DAQin thestores
CR during therecords
the last failure of the before
right DAQ, which represents
the failure, thesensor
as if the secondsignals
failure
scenario.
remain activeThe
evenDAQ stores
when the last records
communication right
fails before the
between the DAQ
failure,
andas sensors
if the sensor signals
(Figure 12).
remain active even when communication fails between the DAQ and
There was no change in the metal loss data during the period, so the CR decreased. How-sensors (Figure 12).
There
ever, when was nocommunication
the change in the metal loss datapast
is restored, during theare
data period, so theand
ignored, CR present
decreased. How-
data are
ever, when the communication is restored, past data are ignored, and
immediately acquired. Furthermore, if a line is drawn between two labeled points (a) and present data are
immediately acquired. Furthermore, if a line is drawn between two labeled points (a) and
(b), the missing data can be robustly interpolated (Figure 17). Consequently, the ER sensor
Sensors 2022, 22, 7562 18 of 3

ensors 2022, 22, 7562 18 of 3

Sensors 2022, 22, 7562 is expected to have relatively high reliability for use in piping system monitoring
17 of 29 becaus
it recovers the previous state immediately, despite DAQ-related issues.

is expected to have
(b), the missing relatively
data high reliability
can be robustly interpolatedfor use in17).
(Figure piping system monitoring
Consequently, the ER sensor becaus
is expected
it recovers theto have relatively
previous high reliability for
state immediately, use in DAQ-related
despite piping system monitoring
issues. because
it recovers the previous state immediately, despite DAQ-related issues.

Figure 17. (Extracted from Figure 16) An enlarged view of data showing the sensor responses durin
the second failure scenario (S2).

Effect
Figure of (Extracted
Figure
17. fluid temperature
17. (Extracted fromFigure
from Figure 16)
16) An
Anenlarged
enlarged view of data
view showing
of data the sensor
showing responses
the sensor during during
responses
the
the secondsecond failure
failure scenario
scenariothe (S2).
(S2).graph used to analyze the CR trend as fluid temperatur
Figure 18 shows
Effect of
changes. Fluid
The CRTemperature
first gradually increased as the experiment was conducted at a constan
Effect of Figure temperature
fluid
fluid temperature, 18 shows andthethen
graph used to analyze
stabilized the CRThis
over time. trendtrend
as fluid temperature
(arrow changes.
in red) was repeate
The CR
Figure 18
throughout first gradually
the shows increased
the graph
test period, as
indicatingthe
usedexperiment
to the
that was
analyze conducted
CR showedthe CR at a constant
trendvariation,
a cyclic fluid tem-an initia
as fluid with
temperatur
perature, and then stabilized over time. This trend (arrow in red) was repeated throughout
changes.
increase The CR first
followed by agradually
subsequent increased as the The
stabilization. experiment
temperature was conducted
variation led at to
a constan
a chang
the test period, indicating that the CR showed a cyclic variation, with an initial increase
fluid
in the temperature,
corrosion
followed and then
sensor response.
by a subsequent stabilized over
As aThe
stabilization. time.
result, This trend
the CRvariation
temperature gradually (arrow in red)
led increased was
to a change with repeated
in thethe flui
throughout
temperature the test
(line inperiod,
pale indicating
blue). The that the
analysis CR showed
clearly a cyclic
demonstrates
corrosion sensor response. As a result, the CR gradually increased with the fluid tempera- variation,
that both with
the an initia
tempera
increase
tureture followed
and(line
NaCl in pale by a subsequent
blue).
concentration of thestabilization.
The analysis clearly Theas temperature
demonstrates
fluid, as well thethat
flowboth thevariation
rate, led and
temperature
have to a chang
a significant effe
in
ontheNaCl concentration
thecorrosion
CR of thesensor of the fluid,
pipingresponse. as well as the flow rate, have a significant
system. As a result, the CR gradually increased with the fluideffect on the CR
of the piping system.
temperature (line in pale blue). The analysis clearly demonstrates that both the tempera
ture and NaCl concentration of the fluid, as well as the flow rate, have a significant effec
on the CR of the piping system.

Figure
Figure 18.18. (Extractedfrom
(Extracted from Figure
Figure 16)
16)An
Anenlarged view
enlarged of data
view showing
of data the variation
showing in corrosion
the variation in corrosio
rate with fluid temperature.
rate with fluid temperature.
Determination of Corrosion Rate for Each Individual Test Condition
Determination
Figure 18.Table of corrosion
(Extracted
3 shows from
the Figurerate
16)for
wall-thinningAneach individual
enlarged
and CR view oftest
datacondition
corresponding showing
to the testthe variationafter
conditions in corrosion
rate with fluid
converting temperature.
the ER the
Table 3 shows sensor data that wereand
wall-thinning obtained during the experiment
CR corresponding to the into
test the wall-
conditions afte
thinning amount.
converting the ER sensor data that were obtained during the experiment into the wal
The wall-thinning
Determination of corrosion amount
rate and CR determination
for each individual testusing the ER sensor indicate that
condition
thinning amount.
the wall-thinning amount and CR increased with the increase in the temperature and NaCl
Table 3 shows
concentration. the wall-thinning
In addition and
to the flow of the CRits
fluid, corresponding to the
temperature likewise test conditions
accelerates metal afte
converting
thinning the
and ER sensor
the CR usingdata
NaCl that were
[35]. In obtained
the case of NaClduring the
of 0%, the experiment
wall-thinning into the wall
increased
thinning amount.
by a factor of 11 times (0.037 mil → 0.424 mil) as the temperature rose. In contrast, in the
case of NaCl of 1.75 and 3.5%, the wall-thinning increased by 1.5 (0.816 mil → 1.271 mil)
and 1.6 times (1.581 mil → 2.539 mil), respectively. Thus, when NaCl is not present, the
Sensors 2022, 22, 7562 18 of 29

influence of temperature on the wall-thinning is relatively significant, but when NaCl is

present, the effect of temperature is rapidly reduced. This suggests that temperature is
a main factor influencing the wall-thinning when chloride is not present in the fluid. The
temperature effect was less pronounced in the chloride environment, since chloride plays
a vital role in the corrosion of ferrous metals.

Table 3. Results of calculation of metal loss and corrosion rate with ER sensor with different experi-
mental conditions.

Time (min) 0–10,407 10,408–19,023 19,024–28,127

NaCl (%) 0 1.75 3.5
Temp. (◦ C) 25 30 35 25 30 35 25 30 35
Wall thinning (mil) 0.037 0.239 0.424 0.816 1.092 1.271 1.581 2.001 2.539
Sum of Wall thinning (mil) 0.7 3.179 6.121
MPY 7.653 49.691 62.424 179.301 232.277 255.366 314.644 419.014 526.845
Average of MPY 42.588 223.409 418.688

Further, the CR was calculated using the wall-thinning values and time. The ob-
tained CR was less than 100 MPY under the condition of 0% NaCl, while the obtained
CR was greater than 150 MPY under NaCl concentrations of 1.75 and 3.5%. It was no-
ticeable that the CR increased by about 5.3 times (42.588 MPY → 223.409 MPY) when
NaCl concentration increased from 0% to 1.75%. This trend became saturated even when
NaCl concentration doubled from 1.75% to 3.5%, showing that CR increased by 1.87 times
(223.409 MPY → 2418.688 MPY). The increase in the CR was greatest at the experiment
temperature condition of 25 ◦ C.

Comparison of Temperature and Chloride Concentration Effects

Figure 19 illustrates the graph describing the wall-thinning data according to the
NaCl concentrations and the fluid temperatures in Table 3, to analyze their correlation
and influence on pipe wall-thinning. Figure 19a shows the increase in metal loss with the
fluid temperature variables in each NaCl condition, whereas Figure 19b shows the rise
in metal loss with NaCl concentrations under the fluid temperature condition. Under an
Sensors 2022, 22, 7562 increased NaCl concentration, the changes in metal loss were much greater than 20 ofthose
31
induced by increased fluid temperature. This indicates that the NaCl concentration is
a more determinant factor for metal loss compared with fluid temperature.

(a) (b)
Figure19.
Figure 19.Comparison
Comparison ofof metal
metal loss
loss of
of ER
ER sensor
sensor under
under different
differentNaCl
NaClconcentrations
concentrationsand fluid
and fluid
temperatures: (a) NaCl concentration, (b) fluid temperature.
temperatures: (a) NaCl concentration, (b) fluid temperature.
4.1.2. LPR Sensor
Figure 20 compares the data obtained through each type of sensor with the LPR sen-
sor data. The current signal was obtained through the LPR sensor, which results from the
calculation of the CR.
(a) (b)
Figure 19. Comparison of metal loss of ER sensor under different NaCl concentrations and fluid
temperatures: (a) NaCl concentration, (b) fluid temperature.
Sensors 2022, 22, 7562 19 of 29

4.1.2. LPR Sensor

Figure 20 compares the data obtained through each type of sensor with the LPR sen-
4.1.2. LPR Sensor
sor data. The current signal was obtained through the LPR sensor, which results from the
Figure
calculation of20 compares
the CR. the data obtained through each type of sensor with the LPR sensor
data. The current signal was obtained through the LPR sensor, which results from the
calculation of the CR.

Figure 20. Plot of raw data obtained from different sensors and the corrosion rate calculated from
LPR 20.
Figure sensor.
Plot of raw data obtained from different sensors and the corrosion rate calculated from
LPR sensor.
As shown in Figure 20, under the NaCl 0% test condition, when the fluid temperature
is 25 ◦ C, 4 MPY is obtained, and as the fluid temperature increases, the CR is measured
As shown in Figure 20, under the NaCl 0% test condition, when the fluid temperature
to be about 20 MPY. In contrast, in the case of a solution containing 1.75% NaCl, about
is 25
190°C,
MPY 4 MPY
was is obtained,
measured and
at 25 asand
◦ C, the200
fluid
MPYtemperature increases,
was measured the
at 30 ◦ C or CR is measured
higher. In the to
be case
about 20 MPY. In contrast, in the case of a solution containing 1.75% NaCl,
of the LPR sensor used in this study, the measurement range is 0–200 MPY. Therefore, about 190
MPY CRswas measured
measured aboveat 200
25 °C,
MPY and
are200 MPY was
unreliable, measured
including at 30 °C or
the measured higher.
NaCl 1.75%Inand
the case
fluid temperature of 30 ◦ C. As described above, the LPR sensor is convenient, as the signal
output during measurement indicates the CR. Nevertheless, it has limitations, as it cannot
be used in a severe corrosion environment due to its fixed measurement range.

4.1.3. Comparison of Corrosion Rate by the Corrosion Sensor and the Electrochemical
Experiment (Lab Scale)
Figure 21 compares the CRs obtained using ER and LPR sensors, and the CR calculated
by electrochemical experiment. The lab-scale electrochemical experiments, including
potentiodynamic polarization and linear polarization, were conducted under the same test
conditions as the corrosion sensors. The results of the experiment are described in Figures
S1–S5 in the Supplementary Materials.
Considering the similarity of the experimental methods, the ER sensor and the po-
tentiodynamic polarization experiment were compared, whereas the LPR sensor was
compared with the linear polarization resistance experiment. Furthermore, the CRs mea-
sured during the stoppage of the piping system were compared. As shown in the graph, in
the absence of chloride, the CR measured using the ER and LPR sensors was similar to the
CR measured using the electrochemical experiment. However, in the presence of chloride,
Figures S1-S5 in the Supplementary Materials.
Considering the similarity of the experimental methods, the ER sensor and the po-
tentiodynamic polarization experiment were compared, whereas the LPR sensor was
compared with the linear polarization resistance experiment. Furthermore, the CRs meas-
ured during the stoppage of the piping system were compared. As shown in the graph, in
Sensors 2022, 22, 7562 20 of 29
the absence of chloride, the CR measured using the ER and LPR sensors was similar to the
CR measured using the electrochemical experiment. However, in the presence of chloride,
the ER and LPR sensors showed a significantly higher CRs. Thus, the CR measured using
the ER and LPR
a commercial sensors showed
corrosion a significantly
sensor and higher CRs.
the CR calculated usingThus, thescale
the lab CR measured
may differ using
de-
apending
commercial corrosion sensor and the CR calculated using the lab scale may differ
on the various experimental parameters. Therefore, corrosion damage can be fur- depend-
ing
theron the various
accelerated experimental
using parameters.
various types Therefore,
of corrosion corrosion damage
and environmental can be such
variables, further
as
accelerated using various types of corrosion and environmental variables, such as
galvanic corrosion and stress corrosion cracking. Fluid pressure, which may occur in the galvanic
corrosion
operating and stress corrosion
environment of thecracking. Fluid pressure,
actual piping which
system, may alsomay occur
cause in the operating
corrosion fatigue.
environment of the actual piping system, may also cause corrosion fatigue.
Therefore, UT thickness measurements were used to find a method that better matches Therefore, UT
thickness measurements were used to find a method that better matches the CR of the real
the CR of the real piping system.
piping system.

(a) (b)
Figure 21.
Figure 21. Comparison
Comparison ofof corrosion
corrosion rates
rates obtained
obtained different
different methods
methods and
and experimental
experimental conditions:
conditions:
(a) ER sensor versus potentiodynamic polarization experiment (lab. test) and (b) LPR sensor versus
(a) ER sensor versus potentiodynamic polarization experiment (lab. test) and (b) LPR sensor versus
linear polarization resistance (lab. Test).
linear polarization resistance (lab. Test).
4.2. UT Measurements
4.2.1. Result of UT thickness
Thicknessmeasurements
Measurementsof oftest
Testpipe
Pipe
Sensors 2022, 22, 7562 The remaining thickness
thickness ofof the
the test
test pipe
pipewaswasdetermined
determinedusing
usingUT UTthickness
thickness 31mea-
22 of meas-
surements to compare the accuracy of the corrosion
urements to compare the accuracy of the corrosion damage and rate damage and rate obtained from the
corrosion sensor and at the lab scale. The measured values were converted
corrosion sensor and at the lab scale. The measured values were converted to a metal loss to a metal loss
of compared
compared
the test pipewith
with the
thedamaged
damaged
increased thickness
with thethickness
test time caused
at all by
caused bythe
theERERsensor.
measurement The
sensor.
pointsThedamaged 22).thickness
damaged
(Figure thickness
Partic- of
the test
ularly, thepipe increased
increased with the
damaged test time
thickness wasat shown
all measurement points
in the order (Figure
of bottom 22). Particularly,
> middle > top
theAs
side. increased
stated in damaged thicknessthe
the introduction, wasmeasurement
shown in theof order of bottom thickness
the remaining > middle > oftop
theside.
test As
pipe does not always indicate a constant trend, depending on the operator’s skill level.does
stated in the introduction, the measurement of the remaining thickness of the test pipe
not always indicate
Nevertheless, the finalameasured
constant trend, depending
value showed on theresult
a similar operator’s
to theskill level. reduction
thickness Nevertheless,
the final measured value
measured using the ER sensor. showed a similar result to the thickness reduction measured using
the ER sensor.

(a) (b) (c)

Figure 22. Variations in metal loss with time converted from the remaining thickness of test pipe
Figure 22. Variations in metal loss with time converted from the remaining thickness of test pipe
measured by UT sensor for three different regions of the test pipe: (a) top side, (b) middle side and
measured by UT sensor for three different regions of the test pipe: (a) top side, (b) middle side and (c)
(c) bottom side.
bottom side.
4.2.2. Internal surface analysis of test pipe after the corrosion experiment
To analyze the degree and tendency of damage according to the location of the test
pipe, the test pipe was cut after the test, and the surface was examined with a 3D micro-
scope. A summary of the analysis using a 3D microscope is provided in Figure S7.
Sensors 2022, 22, 7562 21 of 29

4.2.2. Internal Surface Analysis of Test Pipe after the Corrosion Experiment
To analyze the degree and tendency of damage according to the location of the test
pipe, the test pipe was cut after the test, and the surface was examined with a 3D microscope.
A summary of the analysis using a 3D microscope is provided in Figure S7.
Figure 23 shows the results of the remaining thickness, measured after cutting the test
pipe. For accurate thickness measurements, the oxides formed on the surface were removed
using a physical method. Carbon steel generally forms corrosion oxides or hydrates on the
surface through electrochemical reactions with various ions, such as oxygen and chloride.
The corrosion oxides formed on the surface exhibit passivation characteristics and inhibit
the corrosion of metals from solutions. These oxides are known as oxide or passivation
films. However, corrosion is not protected, as the oxide film is not densely formed in
Sensors 2022, 22, 7562 environments where chloride is present. As a result of 3D microscopic analysis in Figure 23 of 31
S7, corrosion oxides or hydrates were observed at a relatively high density at the bottom
side of the test pipe, and the surface roughness value was similarly large.

Figure
Figure23.
23.Actual
Actualremaining
remainingthickness
thicknessmeasurements
measurementsfor
forthree
threedifferent
differenttest
testpipe
pipepositions
positionsafter
experiment.
after experiment.

In contrast,analysis;
4.2.3. Statistical the lowest amountstandard
Average, of corrosion oxide or and
deviation, hydrate was observed
polynomial on the top
regression
side, and the surface roughness value was the smallest. Thus, the tendency of corrosion
Figure
oxide 24a–c plot
and hydrate the average
formation and standard
is different depending deviation of the
on the point at data
whichtothereveal
pipethe devia-
makes
tions andwith
contact trends
the in the The
fluid. UT density
measurement data at different
of the corrosion product ismeasurement
considered topositions.
be high, asThe
solid
corrosion oxides and hydrates are formed in the upper direction due to the positionalgra-
line in a green color is the ninth-order polynomial fitting using a commercial
phing software (Origin
characteristics Pro 9),side,
of the bottom and and
the dotted line in purple
the desorption color indicates
phenomenon the linear
is relatively small. fit-
ting of the damaged
However, the densitythickness
of the topcorresponding to the
side is considered range
to be of chloride
relatively low as concentration
corrosion oxides vari-
ables. The curve
and hydrates growfitting
in thewith a ninth-order
downward polynomial
direction, revealed
and desorption of thethat the occurs
product final damaged
more
actively. had
thickness When measuring
a similar the remaining
tendency thickness of
to the corrosion the test
sensor as pipe, the remaining
the experiment thickness
time, and chlo-
of the
ride bottom sideincreased.
concentration was measured to be the smallest,
In particular, and
as a result ofthe
thetop sidefitting
linear was estimated to be
for each chloride
the largest.
concentration section, the slope of the damaged thickness graph was found to increase
with the chloride concentration.
4.2.3. Statistical Analysis; Average, Standard Deviation, and Polynomial Regression
Figure 24a–c plot the average and standard deviation of the data to reveal the devia-
tions and trends in the UT measurement data at different measurement positions. The solid
line in a green color is the ninth-order polynomial fitting using a commercial graphing
software (Origin Pro 9), and the dotted line in purple color indicates the linear fitting of
the damaged thickness corresponding to the range of chloride concentration variables.
The curve fitting with a ninth-order polynomial revealed that the final damaged thick-
ness had a similar tendency to the corrosion sensor as the experiment time, and chloride
concentration increased. In particular, as a result of the linear fitting for each chloride
concentration section, the slope of the damaged thickness graph was found to increase with
the chloride concentration.

(a) (b) (c)

Figure 24. Polynomial regression analysis of metal loss data obtained from UT sensor after calculat-
ing average and standard deviation: (a) top side, (b) middle side and (c) bottom side.

4.2.4. Linear regression analysis and adjusted coefficient of determination

As discussed, a measurement error is inevitably produced when measuring the pipe
ting of the damaged thickness corresponding to the range of chloride concentration vari-
ables. The curve fitting with a ninth-order polynomial revealed that the final damaged
thickness had a similar tendency to the corrosion sensor as the experiment time, and chlo-
ride concentration increased. In particular, as a result of the linear fitting for each chloride
Sensors 2022, 22, 7562 concentration section, the slope of the damaged thickness graph was found to increase 22 of 29
with the chloride concentration.

(a) (b) (c)

Figure 24. Polynomial regression analysis of metal loss data obtained from UT sensor after calculat-
Figure 24. Polynomial regression analysis of metal loss data obtained from UT sensor after calculating
ing average and standard deviation: (a) top side, (b) middle side and (c) bottom side.
average and standard deviation: (a) top side, (b) middle side and (c) bottom side.

4.2.4.
4.2.4.Linear
Linearregression
Regression analysis and
Analysis adjusted
and Adjusted coefficient of determination
Coefficient of Determination
AsAsdiscussed,
discussed,a ameasurement
measurementerror errorisisinevitably
inevitablyproduced
producedwhenwhenmeasuring
measuringthe thepipe
pipe
thickness
thicknessusing
usingthetheUT
UTsensor.
sensor.InInthis
thisstudy,
study,however,
however,thickness
thicknessmeasurements
measurementsby byUTUT
sensors
sensorsindicated
indicatedaa tendency followthe
tendency to follow thewall-thinning
wall-thinningtrend,
trend,and
andthethe terminal
terminal metal
metal loss
loss corresponded
corresponded to actual
to the the actual
pipepipe thickness
thickness in Figure
in Figure 23. Based
23. Based on these
on these results,results, a re-
a regression
gression
analysisanalysis was performed
was performed betweenbetween
the damage the damage
thicknessthickness
measured measured
by the ERby the ER
sensor andsen-
UT
sor and UT measurement
measurement data, as
data, as shown in shown
Figure in25.Figure 25. To pre-process
To pre-process the UT measurement
the UT measurement data, the
data,
datathe
weredata were calculated
calculated as theand
as the average average
maximum and maximum damage
damage values values
of the data of the dataat
measured
all measurement
measured points on the
at all measurement upper,
points onmiddle,
the upper,andmiddle,
bottom sides at the same
and bottom sides time,
at theand
sameare
described
time, and arein described
Figure 25a,b.
in Figure 25a,b.
This analysis shows that the R2adj value (adjusted coefficient of determination) was
about 0.848 and 0.854, respectively, which can be interpreted as a similar trend existing
between the ER sensor and UT measurement data. Consequently, it is possible to apply a
commercial ER sensor to evaluate the remaining thickness of the piping system, and this
is considered to have high accuracy when calculating the CR.

Figure 25. Linear regression analysis and adjusted coefficient determination for metal loss by ER and
UT sensor: (a) metal loss versus average metal loss by UT sensor and (b) metal loss versus maximum
metal loss by UT sensor.
14 214
Metal loss This analysis shows that the R
value (adjusted coefficient of determination) was
Metal loss adj
12 about 0.848 and 0.854, respectively,
Linear fit for metal loss which can
Linear be
fit for interpreted
metal loss as a similar trend existing
Average of metal loss by UT, mil

12
Maximum metal loss by UT, mil

between the ER sensor and UT measurement data. Consequently, it is possible to apply

R2adj = 0.85401
10 R2adj = 0.84857
a commercial ER sensor to evaluate the
10 remaining thickness of the piping system, and this
is considered to have high accuracy when calculating the CR.
8 8

Equation y = a + b*x
6 6 No Weighting
Equation y = a + b*x Weight
Weight No Weighting Residual Sum 37.86758
Residual Sum 40.75637 of Squares
4 of Squares 4 Pearson's r 0.92507
Pearson's r 0.92215 0.85401
Adj. R-Square
Adj. R-Square 0.84857
Value Standard Erro
2 Value Standard Error 2
Intercept 5.59928 0.09694
Intercept 4.03804 0.10057
?$OP:A=1 Slope 0.51938 0.02341
Slope 0.52739 0.02428
Sensors 2022, 22, 7562 23 of 29

4.3. Correlation Coefficient Analysis of Metal Loss and CR Considering Various Factors
Correlation analysis is used to determine the correlation or intensity of correlation
between two continuous variables [48]. The intensity of correlation can be quantified by
calculating a correlation coefficient. Therefore, the correlation between two variables can
be determined using the calculated correlation coefficient, and the correlation intensity
can be evaluated to rank correlations for various variables [49]. The Pearson correlation
coefficient, Spearman correlation coefficient, and Kendall’s tau correlation coefficient are
typically used as ways to calculate the correlation coefficient [50].
In this study, the Pearson correlation coefficient was employed to identify the variables
that correlate with the features of metal loss and CR. The calculated results are shown in
Figure 26. Two datasets were integrated using time as a pre-processing process so that
the obtained pump pressure and pump motor load data were among the experimental
variables used in this study for Pearson correlation coefficient analysis. Moreover, using the
calculated flow and quantity of electric charge, the correlation coefficient was calculated and
the correlation was analyzed. The lighter color indicates a significant correlation coefficient,
Sensors 2022, 22, 7562 25 of 31
and the darker color indicates a small correlation coefficient. Generally, a correlation
coefficient value greater than ±0.8 indicates a strong correlation.

Pearsoncorrelation
Figure 26. Pearson correlationcoefficient
coefficientmap
mapfor
forcorrelation
correlationanalysis
analysisbetween
betweenmetal
metalloss
loss(corrosion
(corrosion
rate) and various
various fluid
fluid parameters.
parameters. The
ThePearson
Pearsoncorrelation
correlationcoefficient
coefficientindicating
indicatingstrong
strongcorrelation
correlation
≥±0.8)isismarked
((≥±0.8) markedwith
withaablack
blackdashed
dashedline.
line.

As shown in the pink area of Figure 26, NaCl, conductivity (cond.), pH, flow, and
5. Discussion
quantity of electric charge (EC) were found to have a strong correlation with metal loss and
5.1. Various Factors Affecting Corrosion and Wall-Thinning of Piping System
CR. In particular, as the pH decreased, the metal loss and CR showed a negative correlation
One ofand
coefficient, the primary concerns
the remaining for metallic
variables had apipelines involve internal
positive correlation damage from cor-
coefficient.
rosion, leading to wall-thinning. There are several different environmental factors affect-
5. Discussion
ing corrosion, including temperature, conductivity, pH, flow rate and fluid pressure. In
this study, NaCl
5.1. Various Factorsconcentrations, fluidand
Affecting Corrosion flow and temperature
Wall-Thinning were
of Piping considered to be the
System
mostOnesignificant
of the primary concerns for metallic pipelines involve internalis damage
factors regarding the corrosion of the ER sensor, which the identical
from
material used in the piping system.
corrosion, leading to wall-thinning. There are several different environmental factors
As shown
affecting in Figure
corrosion, 12, metal
including loss in the
temperature, ER sensor increases
conductivity, withand
pH, flow rate NaCl concentra-
fluid pressure.
tions.
In thisNaCl is NaCl
study, a major factor accelerating
concentrations, fluid the
flowelectrochemical
and temperature corrosion reaction ofto
were considered metals.
be the
In particular, it accelerates corrosion damage by self-propagating and autocatalytic mech-
anisms, as the hydration reaction of chloride ions lowers the pH. Therefore, when NaCl is
added, the metal loss of the ER sensor undergoes a rapid increase.
The fluid flow is another important factor affecting the internal damage to a pipe. In
general, the fluid flow accelerates wall-thinning by affecting pipe erosion and corrosion
Sensors 2022, 22, 7562 24 of 29

most significant factors regarding the corrosion of the ER sensor, which is the identical
material used in the piping system.
As shown in Figure 12, metal loss in the ER sensor increases with NaCl concentrations.
NaCl is a major factor accelerating the electrochemical corrosion reaction of metals. In par-
ticular, it accelerates corrosion damage by self-propagating and autocatalytic mechanisms,
as the hydration reaction of chloride ions lowers the pH. Therefore, when NaCl is added,
the metal loss of the ER sensor undergoes a rapid increase.
The fluid flow is another important factor affecting the internal damage to a pipe. In
general, the fluid flow accelerates wall-thinning by affecting pipe erosion and corrosion
reactions. First, pipe erosion refers to a phenomenon that causes the continuous wall-
thinning of metal pipes due to mechanical and physical interactions between the fluid flow
and the metal pipe surface [51]. Furthermore, the mechanism of erosion that can occur in
the piping system varies according to the pressure changes and erosion behavior of the
fluid, such as cavitation erosion, flashing erosion, liquid impingement erosion, and solid
particle erosion [52]. Second, the corrosion reaction accelerates the wall-thinning of the
piping system, as the electrochemical reaction rate changes according to the fluid flow.
This phenomenon is known as flow-accelerated corrosion (FAC) and has been studied by
numerous researchers. The FAC mechanism is as follows [53]. Generally, the thickness of
the mass transfer boundary layer is reduced when there is fluid flow. This facilitates the
diffusion of Fe2+ ions and generation of corrosion products due to the corrosion reaction
of the metal, leading to an increase in the CR of the metal [54,55]. In the particular case
of carbon steel, which cannot form a relatively strong and dense passivation film, the CR
significantly increases as the fluid flow velocity increases [56]. Thus, when the flow velocity
of the piping system is fast, the corrosion reaction is activated according to the decrease in
the thickness of the mass transfer boundary layer between the inner surface of the piping
and the solution.
The potential synergistic effects on CR between NaCl concentration and fluid flow
were studied. The CR appears differently as the electrochemical reaction rate changes
depending on the flow velocity of the piping system, and the flow velocity can be considered
to affect the CR of the piping system. In addition, physical damage, such as erosion, may
occur if flow velocity is present. However, according to the experimental results in this
study, in the absence of NaCl, the damage caused by the flow velocity was insignificant.
Combining these results, the increase in the CR has a more significant effect on the damage
of the pipe than the effect of physical erosion in the flow velocity conducted in this study.
Nevertheless, the damage caused by physical erosion cannot be ignored at higher flow
velocities. In particular, when two factors work together rather than each independently
affecting the pipe, the wall-thinning of the piping system can be accelerated by exerting
a synergistic effect [57]. Therefore, the flow velocity is an essential factor to be considered
when evaluating the remaining useful life of the piping system [58,59].
Aside from the NaCl and fluid flow, the temperature also increases the CR measured by
the ER sensor (Figure 18). This is because, when the temperature increases, the CR rises by
enhancing the activity of the various ions participating in the electrochemical reaction and
changing the free energy that determines the degree of reaction between metal and fluid [60].
When the fluid is at room temperature, the effect on metal loss according to the increase in
temperature is insignificant, as, in the relatively low-temperature environment, erosion is
more dominant than corrosion [61,62]. In the case of the flow velocity applied in this study,
the damage due to erosion was minimal, the temperature of the controlled variable was low,
and the change rate was not significant, so that the effect of the temperature was considered
negligible. In contrast, because NaCl is a major factor accelerating metal corrosion, even
a small amount can have a relatively significant effect. From Figures 14 and 19, it can
be seen that NaCl, start/stop, and fluid temperature, which are the test variables that
were selected in this study, correlate with each other and affect the pipe wall-thinning.
However, as the effects of start/stop and fluid temperature on pipe wall-thinning differ
Sensors 2022, 22, 7562 25 of 29

depending on the presence or absence of NaCl, that NaCl was evaluated to have the most
significant influence.

5.2. Reliability of ER Sensor

The reliability of the ER sensor was evaluated by comparing the results obtained from
the cross-sectional observation and thickness measurements that were carried out mechan-
ically (ruler) and ultrasonically for the damaged pipe. The cross-sectional observation
demonstrated the different amounts of damage produced in the different positions and
orientations of the pipe.
The results of the cross-sectional analysis indicated that the bottom side experiences
the most damage due to corrosion and erosion. According the results of 3D analysis
and thickness measurements, the electrochemical reaction occurs most actively on the
bottom side compared to the top, resulting in high-density oxides and hydrates and
greater corrosion damage. Corrosion oxides and hydrates accumulate due to the positional
characteristics of the bottom side, and the influence of gravity further accelerates erosion
and corrosion by fluid.
The remaining thickness measurement showed that the top side had the largest remain-
ing thickness, while the bottom side had the smallest remaining thickness. These results
nearly correspond to the UT sensor measurement result shown in Figure 22. Furthermore,
the average value of the damaged thickness for the three positions showed similar results
to the damaged sensing element thicknesses of the ER and LPR sensors in Figure S6.
The UT measurement method is generally considered to be suitable for measuring
the damaged thickness of the piping system. However, as in the ninth-order polynomial
fitting, the trends in the measured data is inconsistent, with numerous large deviations.
This is because pipe thickness measurements made with a UT sensor may generate dif-
ferent measurement accuracies depending on the surroundings, fluid flow, and operator
measurement skill, as stated in the introduction. Given the disadvantages of the UT sensor,
the ER sensor should be a more reasonable choice than the conventional UT sensor when
determining the metal loss and CR of pipes.

5.3. Correlation Coefficient Analysis

The correlation coefficient analysis was carried out between the operating parmeters
and ER sensor, and a revealed strong relationship was revealed between the ER sensor
and fluid parameters such as NaCl concentration, conductivity, pH, flow and quantity of
electric chage. This was consistent with the results that were derived by analyzing the
factors affecting the corrosion damage of the piping system according the electrochemical
reaction obtained by measuring the metal loss and CR using the ER sensor in Section 4.1.
The conductivity and pH are factors affected by NaCl. The pH showed a relatively low
correlation coefficient, which can be attributed to the relatively small range of changes due
to the characteristics of the pH data. This showed similar characteristics in terms of the
variable of the fluid temperature. In general, the temperature is considered to be a major
factor influencing the corrosion of metals and aqueous solutions [35,63]. However, because
the fluid temperature variable used in this study has a relatively small range, it was not
considered to significantly affect the corrosion damage. In contrast, the NaCl concentration
and flow showed a relatively large correlation, as they had a variable range that could
sufficiently affect the metal corrosion.

6. Conclusions
Internal pipe wall-thinning and the CR were measured in various pipe usage environ-
ments using a custom-built test bed equipped with various sensors. The characteristics of
the CR change were analyzed according to the presence or absence of NaCl concentrations,
temperature, and flow velocity, and the results were compared with those of the lab scale
electrochemical setup.
Sensors 2022, 22, 7562 26 of 29

The damaged thickness and CR measurements obtained using a commercial ER sensor

indicate that the CR increases with NaCl concentrations, fluid temperature, and flow
velocity. Notably, the ER sensor’s wall thickness determination strongly correlates with the
UT sensor’s measurements. A direct measurement of the remaining thickness of the pipe
after the test confirmed that the thickness of the real pipe corresponds to the measurements
of both ER and UT sensors. Contrary to expectations, the LPR sensor did not yield the
expected CR in a corrosive environment containing NaCl. The LPR sensor reading reached
its maximum value under a NaCl concentration of 1.75% at temperatures above 30 ◦ C,
indicating that the accuracy of CR may have been compromised. This is attributed to
the following measuring methods of the sensor: the ER sensor is based on the physical
principle of measuring changes in the electrical resistance due to metal loss, whereas the
LPR sensor is based on the electrochemical principles in which the polarization resistance
is obtained.
The CR sensing accuracy was evaluated by comparing the CR response from the
corrosion sensors and the reference CR obtained from a lab-scale electrochemical exper-
iment. As expected, the results of the electrochemical evaluation showed a remarkably
low CR compared to that of the ER sensor. The reason for this is clear: there was little
flow velocity in the lab scale experiment, which would not lead to a change in the mass
transfer boundary layer between the working electrode and the aqueous solution; thus,
the corrosion reaction was suppressed. However, both the ER sensor and electrochemical
evaluation demonstrated a similar trend to NaCl concentration and fluid temperature.
The inner surface of the pipe was optically observed after the test to confirm the
accuracy of the metal loss in the ER sensor. The remaining pipe thickness was also measured
at regular intervals using a UT sensor, and a regression analysis was performed using the
metal loss values from the ER and UT sensors. The analysis of thickness variations revealed
that the metal loss measured by the ER sensor is in good agreement with both the wall-
thinning value of the actual pipe and the value measured by the UT sensor. In the regression
analysis, a strong correlation was found between the metal loss values from the ER sensor
and those from UT sensors. Consequently, the commercial ER sensor was found to have
high reliability to monitor the metal loss and evaluate the remaining pipe thickness.
To analyze the correlation of factors affecting the wall-thinning of the pipe, the corre-
lation coefficient was calculated using the operating parameters and the sensor response
from the ER sensor. The NaCl amount and flow showed a high correlation coefficient of 0.9
or more. However, the temperature had a lower correlation coefficient than other variables,
possibility due to the limited range of the test.
The results of this investigation demonstrated that the ER sensor has acceptable
reliability for monitoring the wall thickness of industrial pipelines. However, the sensing
element of 10 mil thickness only had a short lifetime (approximately 2 weeks), although
the test was conducted under highly accelerated and harsh conditions for the sensor.
Thus, the thickness of the sensing element should be considered under harsh operating
conditions, particularly in chloride-containing environments. In addition, there are several
other aspects of implementation, such as material, wall thickness, and the coating of the
target pipe.
To date, the ER sensor has been used successfully in corrosion monitoring, but few stud-
ies have reported any relationship between operating parameters and pipe wall-thickness.
The results of this work not only provide useful information for the determination of CR
using an ER sensor under various operating parameters in the piping system, in addition
to new insights regarding the use of ER sensor as a pipe wall-thickness measuring method,
which provides an alternative to the conventional UT sensor.
Further studies are required to follow-up our pilot research: A pipe lifetime prediction
algorithm should be developed using a single metallic sensing element to extend its
application to the integrity monitoring of various metallic pipes using machine learning.
A machine learning model can be trained with the labeled dataset to establish the pipe’s
lifetime prediction model. The training data consist of the pipeline operating data as input
Sensors 2022, 22, 7562 27 of 29

and the corrosion sensor response (i.e., metal loss or CR) as target output. If the model is
trained, it could predict the remaining thickness of the pipeline, which is the output of the
model. To explore the feasibility of a long-term of use of the sensor, more resistant metals
and alloys, such as stainless steel, can be used as the sensing element. Data from different
sensing materials can be correlated, based on the corresponding experimental data. Future
work may benefit from the present findings and data.
As presented in this study, the ER sensor is considered to be the most reasonable and
suitable sensor for monitoring the wall-thinning of pipes for industrial applications, and
can provide an alternative to UT sensor to obtain the pipe’s wall thickness. It should be
noted that the implementation of an ER sensor requires a correct understanding of the
electrochemical reaction mechanisms that occur under various operating conditions.

Supplementary Materials: The following supporting information can be downloaded at: https://
www.mdpi.com/xxx/s1, References [64–66] are cited in the supplementary materials. Figure S1:
Potentiodynamic polarization curves of the carbon steel under different fluid temperatures and
NaCl concentrations: (a) NaCl 0%, (b) NaCl 1.75% and (c) NaCl 3.5%.; Figure S2: Comparison of
electrochemical corrosion parameters of the carbon steel with different fluid temperatures and NaCl
concentrations: (a) corrosion potential and (b) corrosion current density.; Figure S3: Linear polariza-
tion resistance curves of the carbon steel with different fluid temperatures and NaCl concentrations:
(a) NaCl 0%, (b) NaCl 1.75% and (c) NaCl 3.5%.; Figure S4: Comparison of electrochemical corro-
sion parameters of the carbon steel obtained from the linear polarization resistance with different
fluid temperatures and NaCl concentrations: (a) corrosion potential, (b) corrosion current density
and (c) polarization resistance.; Figure S5: Comparison of corrosion rates obtained from different
electrochemical evaluation methods under different fluid temperatures and NaCl concentrations:
(a) potentiodynamic polarization and (b) linear polarization resistance.; Figure S6: Photographs
of (a) sensing element of ER sensor and (b) sensing element of LPR sensor after the test.; Figure
S7: Surface analysis of the inner side of the test pipe using 3D laser confocal microscope.; Table S1:
Results of calculation of corrosion rate after potentiodynamic polarization and linear polarization
resistance experiment.
Author Contributions: Conceptualization, J.-H.L. and D.-H.S.; methodology, J.-H.L. and D.-H.S.;
software, H.-H.K.; validation, J.-H.L., D.-H.S. and H.-H.K.; formal analysis, D.-H.S. and H.-K.H.;
investigation, D.-H.S. and H.-K.H.; resources, D.-H.S.; data curation, J.-H.L.; writing—original
draft preparation, D.-H.S.; writing—review and editing, J.-H.L.; visualization, D.-H.S. and H.-K.H.;
supervision, J.-H.L.; project administration, J.-H.L.; funding acquisition, J.-H.L. All authors have read
and agreed to the published version of the manuscript.
Funding: This work was supported by the ‘Autonomous Ship Technology Development Program
(20011164, Development of Performance Monitoring and Failure Prediction and Diagnosis Technology
for Engine System of Autonomous Ships)’ funded by the Ministry of Trade, Industry & Energy
(MOTIE, Korea).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data are not publicly available due to the project requirements.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Paolacci, F.; Reza, M.S.; Bursi, O.S.; Gresnigt, A.M.; Kumar, A. Main issues on the seismic design of industrial piping systems and
components. In Proceedings of the ASME 2013 Pressure Vessels and Piping Conference, Paris, France, 14–18 July 2013; pp. 1–10.
2. Khattak, M.A.; Zareen, N.; Mukhtar, A.; Kazi, S.; Jalil, A.; Ahmed, Z.; Jan, M.M. Root cause analysis (RCA) of fractured ASTM
A53 carbon steel pipe at oil & gas company. Case Stud. Eng. Fail. Anal. 2016, 7, 1–8. [CrossRef]
3. Mpesha, W. Leak Detection in Pipes by Frequency Response Method; University of South Carolina: Columbia, SC, USA, 1999; pp. 1–5.
4. Sereda, P.J. Weather Factors Affecting Corrosion of Metals; ASTM Special Technical Publication: Philadelphia, PA, USA, 1974;
Volume 558, pp. 7–22.
5. Mpesha, W.; Chaudhry, M.H.; Gassman, S.L. Leak detection in pipes by frequency response method using a step excitation.
J. Hydraul. Res. 2002, 40, 55–62. [CrossRef]
Sensors 2022, 22, 7562 28 of 29

6. Korea Institute of Nuclear Safety. Reports of Accident and Breakdown in the Nuclear Power Plant; 080606K3; Korea Institute of Nuclear
Safety: Daejeon, Korea, 2008.
7. Korea Institute of Nuclear Safety. Reports of Accident and Breakdown in the Nuclear Power Plant; 170327K4; Korea Institute of Nuclear
Safety: Daejeon, Korea, 2018.
8. Lee, C.W.; Yoo, D.G. Development of leakage detection model and its application for water distribution networks using RNN-
LSTM. Sustainability 2021, 13, 9262. [CrossRef]
9. Diao, Y.; Yan, L.; Gao, K. Improvement of the machine learning-based corrosion rate prediction model through the optimization
of input features. Mater. Des. 2021, 198, 109326. [CrossRef]
10. El-Sayed, M.H. Flow enhanced corrosion of water injection pipelines. Eng. Fail. Anal. 2015, 50, 1–6. [CrossRef]
11. Reddy, M.S.B.; Ponnamma, D.; Sadasivuni, K.K.; Aich, S.; Kailasa, S.; Parangusan, H.; Ibrahim, M.; Eldeib, S.; Shehata, O.; Ismail, M.; et al.
Sensors in advancing the capabilities of corrosion detection: A review. Sens. Actuators A Phys. 2021, 332, 113086. [CrossRef]
12. Agarwala, V.S.; Reed, P.L.; Ahmad, S. Corrosion detection and monitoring—A Review. In Proceedings of the CORROSION 2000
NACE-00271, Orlando, FL, USA, 26–31 March 2000.
13. Cohn, M.J.; Norton, J.W. Case studies of pulsed eddy current to measure wall loss in feedwater piping and heater shells.
In Proceedings of the ASME Pressure Vessels and Piping Conference, Chicago, IL, USA, 27–31 July 2008; pp. 721–731.
14. Oh, S.B.; Cheong, Y.M.; Kim, D.J.; Kim, K.M. On-Line Monitoring of Pipe Wall Thinning by a High Temperature Ultrasonic
Waveguide System at the Flow Accelerated Corrosion Proof Facility. Sensors. 2019, 19, 1762. [CrossRef]
15. Calleja, T.V.; Su, D.; Bruijn, F.D.; Miro, J.V. Learning spatial correlations for Bayesian fusion in pipe thickness mapping.
In Proceedings of the 2014 IEEE International Conference on Robotics and Automation (ICRA), Hong Kong, China,
31 May–7 June 2014.
16. Kim, Y.H. Inspection of Pipeline. J. Korean Soc. Nondestruct. Test. 1998, 18, 121–132.
17. Cheong, Y.M.; Kim, K.M.; Kim, D.J. High-temperature ultrasonic thickness monitoring for pipe thinning in a flow-accelerated
corrosion proof test facility. Nucl. Eng. Technol. 2017, 49, 1463–1471. [CrossRef]
18. Wright, R.F.; Lu, P.; Devkota, J.; Lu, F.; Ziomek, M.M.; Ohodnicki, P.R., Jr. Corrosion Sensors for Structural Health Monitoring of
Oil and Natural Gas Infrastructure: A Review. Sensors 2019, 19, 3964. [CrossRef]
19. Liu, Z.; Kleiner, Y. State-of-the-art review of technologies for pipe structural health monitoring. IEEE Sens. J. 2012, 12,
1987–1992. [CrossRef]
20. Kouril, M.; Prosek, T.; Scheffel, B.; Dubois, F. High sensitivity electrical resistance sensors for indoor corrosion monitoring.
Corros. Eng. Sci. Technol. 2013, 48, 282–287. [CrossRef]
21. Prosek, T.; Le Bozec, N.; Thierry, D. Application of automated corrosion sensors for monitoring the rate of corrosion during
accelerated corrosion tests. Mater. Corros. 2014, 65, 448–456. [CrossRef]
22. Joosten, M.W.; Koks, J.; Humble, P.G.; Btakset, T.J.; Keilty, D.M. Internal corrosion monitoring of subsea production flowlines-Probe
design and testing. In Proceedings of the CORROSION Conference, San Diego, CA, USA, 22–27 March 1998.
23. Durrani, F.; Wesley, R.; Srikandarajah, V.; Eftekhari, M.; Munn, S. Predicting Corrosion rate in Chilled HVAC Pipe Network:
Coupon vs Linear Polarisation Resistance method. Eng. Fail. Anal. 2020, 109, 104261. [CrossRef]
24. Mckenzie, M.; Vassie, P.R. Use of weight loss coupons and electrical resistance probes in atmospheric corrosion tests. Br. Corros. J.
1985, 20, 117–124. [CrossRef]
25. Boving, K.G. Chapter 8. ER Probe. In NDE Handbook: Non-Destructive Examination Methods for Condition Monitoring; Butterworth-
Heinemann: Oxford, UK, 2014; pp. 81–87. [CrossRef]
26. Jung, S.W. Characteristics of Thin Film Electric Resistance Sensors for Measurement of Corrosion Rates. Ph.D. Thesis, Kookmin
University, Seoul, Korea, 2005.
27. Yang, L.; Chiang, K.T. On-line and real-time corrosion monitoring techniques of metals and alloys in nuclear power plants
and laboratories. In Understanding and Mitigating Ageing in Nuclear Power Plants. Materials and Operational Aspects of Plant Life
Management (Plim); Woodhead Publishing: Sawston, UK, 2010; Volume 14, pp. 417–455. [CrossRef]
28. Li, S.Y.; Kim, Y.G.; Jung, S.; Song, H.S.; Lee, S.M. Application of steel thin film electrical resistance sensor for in situ corrosion
monitoring. Sens. Actuators B Chem. 2007, 120, 368–377. [CrossRef]
29. Colony, C.H. An Evaluation of Corrosion Sensors for the Monitoring of the Main Cables of the Anthony Wayne Bridge. Master’s
Thesis, The University of Toledo, Toledo, OH, USA, 2016.
30. Agrawal, A.K. Corrosion monitoring. Encycl. Mater. Sci. 2000, 47, 26–29. [CrossRef]
31. Nazir, M.H.; Saeed, A.; Khan, Z.A. Electrochemical corrosion failure analysis of large complex engineering structures by using
micro-LPR sensors. Sens. Actuators B Chem. 2018, 268, 232–244. [CrossRef]
32. Campos-Silva, I.E.; Rodríguez-Castro, G.A. 18-Boriding to improve the mechanical properties and corrosion resistance of steels.
Thermochem. Surf. Eng. 2015, 28, 2399–2410. [CrossRef]
33. Mansfeld, F. The Polarization Resistance Technique for Measuring Corrosion Currents. Adv. Corros. Sci. Technol. 1976, 6,
163–262. [CrossRef]
34. Walter, G.W. Problems arising in the determination of accurate corrosion rates from polarization resistance measurements.
Corros. Sci. 1977, 17, 983–993. [CrossRef]
35. Jones, D.A. Principles and Prevention of Corrosion, 2nd ed.; Pearson-Prentice Hall: Upper Saddler River, NJ, USA, 1992; pp. 190–193.
Sensors 2022, 22, 7562 29 of 29

36. Nazir, M.H.; Khan, Z.A.; Saeed, A.; Stokes, K. A predictive model for life assessment of automotive exhaust mufflers subject to
internal corrosion failure due to exhaust gas condensation. Eng. Fail. Anal. 2016, 63, 43–60. [CrossRef]
37. Mikael, W.B.; Jan, B.A.; Lars, W.A.; Mads, B.; Curt, J.C.; Niels-Jorgen, R.N. Design and Installation of Marine Pipelines; Wiley-
Blackwell: Hoboken, NJ, USA, 2009; pp. 27–35.
38. Rahmani, K.; Jadidian, R.; Haghtalab, S. Evaluation of inhibitors and biocides on the corrosion, scaling and biofouling control of
carbon steel and copper–nickel alloys in a power plant cooling water system. Desalination 2016, 393, 174–185. [CrossRef]
39. Mingxuan, Y.; Suoying, H.; Ming, G.; Mengfei, X.; Jiayu, M.; Xiang, H.; Kamel, H. Comparative study on the cooling performance
of evaporative cooling systems using seawater and freshwater. Int. J. Refrig. 2020, 121, 23–32. [CrossRef]
40. ASTM G102-89; Standard Test Method for Calculation of Corrosion Rates and Related Information from Electrochemical
Measurements. ASTM International: West Conshohocken, PA, USA, 2004.
41. ASTM G59-97; Standard Test Method for Conducting Potentiodynamic Polarization Resistance Measurements. ASTM Interna-
tional: West Conshohocken, PA, USA, 2003.
42. Ajmal, T.S.; Arya, S.B.; Udupa, K.R. Effect of hydrodynamics on the flow accelerated corrosion (FAC) and electrochemical
impedance behavior of line pipe steel for petroleum industry. Int. J. Press. Vessel Pip. 2019, 174, 42–53. [CrossRef]
43. Zhang, X.; Xiao, K.; Dong, C.; Wu, J.; Li, X.; Huang, Y. In situ Raman spectroscopy study of corrosion products on the surface of
carbon steel in solution containing Cl− and SO4 2− . Eng. Fail. Anal. 2011, 18, 1981–1989. [CrossRef]
44. Tamura, H. The role of rusts in corrosion and corrosion protection of iron and steel. Corros. Sci. 2008, 50, 1872–1883. [CrossRef]
45. Sander, A.; Berghult, B.; Broo, A.E.; Johansson, E.L.; Hedberg, T.A. Iron corrosion in drinking water distribution systems—The
effect of pH, calcium and hydrogen carbonate. Corros. Sci. 1996, 38, 443–455. [CrossRef]
46. Hodgkiess, T.; Vassiliou, G. Complexities in the erosion corrosion of copper-nickel alloys in saline water. Desalination 2005, 183,
235–247. [CrossRef]
47. Vasyliev, G.S. The influence of flow rate on corrosion of mild steel in hot tap water. Corros. Sci. 2015, 98, 33–39. [CrossRef]
48. Gogtay, N.J.; Thatte, U.M. Principles of Correlation Analysis. J. Assoc. Phys. India 2017, 65, 78–81.
49. Yan, L.; Diao, Y.; Lang, Z.; Gao, K. Corrosion rate prediction and influencing factors evaluation of low-alloy steels in marine
atmosphere using machine learning approach. Sci. Technol. Adv. Mater. 2020, 21, 359–370. [CrossRef] [PubMed]
50. De Winter, J.C.F.; Gosling, S.D.; Potter, J. Comparing the Pearson and Spearman Correlation Coefficients Across Distributions and
Sample Sizes: A Tutorial Using Simulations and Empirical Data. Psychol. Methods 2016, 21, 273–290. [CrossRef] [PubMed]
51. Totten, G.E. ASM Handbook: Volume 18 Friction, Lubrication, and Wear Technology; ASM International Handbook Committee:
West Conshohocken, PA, USA, 2017; pp. 381–382.
52. Crokett, H.M.; Horowitz, J.S. Erosion in Nuclear Piping Systems. J. Press. Vessel Technol. 2010, 132, 024051. [CrossRef]
53. Chexal, B.; Horowitz, J.; Dooley, B. Flow-Accelerated Corrosion in Power Plants; Revision 1. No. EPRI-TR-106611-R1; Electric Power
Research Inst.: Palo Alto, CA, USA; Electricite de France: Paris, France; Siemens AG Power Generation: Munich, Germany, 1998.
54. Giddey, S.; Cherry, B.; Lawson, F.; Forsyth, M. Stability of oxide films formed on mild steel in turbulent flow conditions of alkaline
solutions at elevated temperatures. Corros. Sci. 2001, 43, 1497–1517. [CrossRef]
55. Kain, V. Flow accelerated corrosion: Forms, mechanisms and case studies. Procedia Eng. 2014, 86, 576–588. [CrossRef]
56. Denpo, K.; Ohawa, H. Fluid Flow Effects on CO2 Corrosion Resistance of Oil Well Materials. Corrosion 1993, 49, 442–449. [CrossRef]
57. Xu, Y.; Zhang, Q.; Zhou, Q.; Gao, S.; Wang, B.; Wang, X.; Huang, Y. Flow accelerated corrosion and erosion-corrosion behavior of
marine carbon steel in natural seawater. NPJ Mater. Degrad. 2021, 56, 1–13. [CrossRef]
58. Wael, H.A.; Mufaiu, M.B.; Meamer, E.N.; Abdelsalam, A.S.; Hassan, M.B. Experimental investigation of flow accelerated corrosion
under two-phase flow conditions. Nucl. Eng. Des. 2014, 267, 34–43. [CrossRef]
59. Muhammadu, M.M.; Sheriff, J.M.; Hamzahb, E. Effect of flow pattern at pipe bends on corrosion behaviour of low carbon steek
and its challenges. J. Teknol. Sci. Eng. 2013, 63, 55–65. [CrossRef]
60. Gurten, A.A.; Keles, H.; Bayol, E.; Kandemirli, F. The effect of temperature and concentration on the inhibition of acid corrosion
of carbon steel by newly synthesized Schiff base. J. Ind. Eng. Chem. 2015, 27, 68–78. [CrossRef]
61. Levy, A.V.; Yan, J.; Patterson, J. Elevated Temperature Erosion of Steels. Wear 1986, 108, 43–60. [CrossRef]
62. Wellman, R.G.; Nicholls, J.R. High temperature erosion–oxidation mechanisms, maps and models. Wear 2004, 256, 907–917. [CrossRef]
63. Melchers, R.E. Effect of Temperature on the Marine Immersion Corrosion of Carbon Steels. Corrosion 2002, 58, 768–782. [CrossRef]
64. Dougall, B.M. Effect of Chloride Ion on the Localized Breakdown of Nickel Oxide Films. J. Electrochem. Soc. 1979, 126,
919–925. [CrossRef]
65. Clayton, C.R.; Lu, Y.C. A Bipolar Model of the Passivity of Stainless Steel: The Role of Mo Addition. J. Electrochem. Soc. 1986, 133,
2465–2473. [CrossRef]
66. ASTM G1-03; Standard Test Method for Preparing, Cleaning, and Evaluating Corrosion Test Specimens. ASTM International:
West Conshohocken, PA, USA, 2003.