Digital Twin For Underground Pipeline

Automation in Construction 144 (2022) 104617
Contents lists available at ScienceDirect
Automation in Construction
journal homepage: www.elsevier.com/locate/autcon
Brillouin fiber optic sensors and mobile augmented reality-based digital

twins for quantitative safety assessment of underground pipelines
Minghao Li, Xin Feng *, Yang Han
Faculty of Infrastructure Engineering, Dalian University of Technology, Dalian, China
A R T I C L E I N F O A B S T R A C T
Keywords: The impossibility of visual inspection and the complexity of combined loads hamper the quantitative assessment,
Underground pipelines lifetime prediction and control of underground pipelines during their lifecycles. A methodology based on mobile
Distributed fiber optic sensors augmented reality (MAR) and Brillouin fiber optic sensors (BFOSs) is presented to build a digital twin (DT) for
Digital twin
underground pipelines. Field experiments were carried out to demonstrate that the proposed method can
Mobile augmented reality (MAR)
Joint data–physics driven model
quantitatively assess and predict the structural safety of an underground pipeline from the DT. The results
Structural safety assessment demonstrate that the distributed sensor networks can measure important but unpredictable deformations (i.e.,
longitudinal bending and axial thermal strain), the joint data-physics driven model can estimate the structural
stress state more accurately than the common calculation model, and the MAR-based human–asset interaction
interface enables more intuitive, efficient, automated operation and maintenance (O&M). In the future, in-line
robotic systems and localized damage models should be further adopted for lifecycle O&M.
1. Introduction corporations would use DT technology by 2021, corresponding to a 10%

improvement in effectiveness [7]. Since the National Aeronautics and
Underground pipelines are widely used engineering assets, and their Space Administration (NASA) gave the first specific definition in 2012
safety management is easily neglected because they are not visible and [8], the DT concept has gradually been extended from the field of in
direct structural damage inspection is difficult [1]. However, with the dustrial manufacturing to construction engineering and from theoretical
rapid growth of underground infrastructure assets and the deterioration research to application practice [9–13]. At the asset scale, Lin et al. [14]
of existing structures over time, the increasing number of public emer proposed a DT-based collapse fragility assessment method for long-span
gencies caused by underground pipelines have brought safety manage cable-stayed bridges during strong earthquakes. Mohammadi and Taylor
ment to the forefront [2]. Underground pipelines have a complex spatial [15] introduced a smart city DT paradigm that revolutionized human
distribution and are controlled by different departments. In addition, –infrastructure interactions and urban planning through the Internet of
there are significant differences in the structural states of pipelines Things (IoT) and virtual reality (VR). Lu et al. [10,16] presented a sys
caused by specific operating conditions, such as foundation settlement tem architecture for DT technology that is specifically designed at the
and internal fluid pressure/temperature. Therefore, the digitalization building and city levels and developed a DT demonstrator at the West
and automation of safety management for underground pipelines is still Cambridge site of the University of Cambridge in the UK. Although each
a challenging issue. of these cases has its own unique definitions and requirements [5], a
As a next-generation technology that can integrate a series of successful DT paradigm relies on (1) the mutual interaction between the
advanced technologies (e.g., sensors, modeling and visualization), physical asset and its DT, which is related to high-performance sensor
emerging digital twin (DT) technology, which is considered to be an technologies and high-fidelity data communication; (2) DT modeling
efficient intelligence solution for digitalization and automation, uses approaches that should focus on both geometrical appearance and the
digital models to dynamically monitor, describe, simulate, predict and ability to carry out asset-specific analysis, prediction, optimization and
control physical counterparts in a timely manner [3–5]. Recently, DT decision-making support; and (3) a human–asset interaction interface
technology has drawn much attention from both the academic and in presented to the end users, which should improve the ability to under
dustrial communities [6]. Gartner predicted that 50% of large industrial stand the unique characteristics of an existing physical asset and be
* Corresponding author.
E-mail address: fengxin@dlut.edu.cn (X. Feng).
https://doi.org/10.1016/j.autcon.2022.104617
Received 3 March 2022; Received in revised form 5 October 2022; Accepted 7 October 2022
Available online 19 October 2022
0926-5805/© 2022 Elsevier B.V. All rights reserved.
M. Li et al. Automation in Construction 144 (2022) 104617
Table 1
Brief summary of modeling and visualization methods.
References Key technologies Key tools/algorithms Key contribution
Schall et al. (2012) AR, GIS, and 3D RTK GPS receiver; tablet PC Develops a field information system for pipelines, enabling planning,
[22] visualization construction and surveys to be executed efficiently on site.
Zhang et al. (2016) AR, BIM and GIS A spatial database using the GeoJSON format; CV Develops and compares two AR solutions (CV version and sensor
[23] algorithm; GPS; OpenGL 2.0 version) for the BIM and GIS of pipelines.
Soria et al. (2018) AR, GIS, CAD, 3D Google Tango; Unity3d engine; ARCore; Mapbox API; Manages and operates (CRUD) the spatial design information in a
[24] visualization Trimble R1 way that allows interactions with hidden underground elements.
Napolitano et al. VR, SHM data, sensor Kolor Panotour Pro; data integration and visualization A VR-based connection between SHM data and structures is
(2018) [25] networks established in an intuitive and interactive way.
Li et al. (2019) AR, GIS and database Unity3d engine; Mapbox API; mobile device Provides a client–server architecture to develop an AR system that
[26] saves time and costs in locating underground pipelines on site.
Zhi et al. (2019) GIS, Simulation models, Unity 3d engine; flooding simulation algorithm coupling Establishes a 3D dynamic visualization method of an urban drainage
[27] 3D visualization multisource heterogeneous data appearance model and dynamically simulates a flood disaster.
Fenais et al. (2020) AR, GIS and cloud-based Unity3d engine; ARKit; GPS or GNSS; cloud-based data Enables immersive mapping of underground utilities to guide site
[28,29] storage storage and management platform using the KML format construction while avoiding excavation damage to underground
pipelines.
Tan et al. (2021) OFDR, Data processing DFOS; MATLAB; algorithms for localization, mapping Develops promising approaches to detect, locate, quantify and
[30,31] and visualization and visualization of SHM data visualize pipe corrosion and concrete cracks using a DFOS.
AR = augmented reality; GIS = geographic information system; RTK GPS = real-time kinematic global positioning system; BIM = building information modeling; CV =
computer vision; CAD = computer-aided design; CRUD = create-read-update-delete; VR = virtual reality; SHM = structural health monitoring; GNSS = global
navigation satellite system; KML = keyhole markup language; OFDR = optical frequency domain reflectometry; and DFOS = distributed fiber optic sensor.
easier and more intuitive to operate. dynamic evolution of structural states (operation and maintenance
In pipeline engineering, Siemens proposed the Pipeline 4.0 concept (O&M) phase). During the O&M phase, which covers >80% of the
[17]. Pipeline 4.0 utilizes DT technology to explore innovative ap lifecycle, pipelines suffer from corrosion and load or operating condition
proaches to further reduce the operating expenses of pipeline systems changes, resulting in the dynamic evolution and deterioration of the
with no risk or cost. In the current phase, Pipeline 4.0 mainly focuses on structural state over time. Therefore, the current studies lack a modeling
developing DTs for pumping stations rather than pipelines. PetroChina and interaction approach for the dynamic evolution of the pipeline
has adopted digital recovery techniques, including aerial photogram structural state to guide lifecycle safety management.
metry, 3D laser scanning, 3D terrain construction, oblique photogram To address these challenges, a framework for DTs is presented in this
metry and digital 3D modeling, to build the DT of the China–Myanmar study for underground pipelines by using mobile augmented reality
oil and gas pipeline [18]. Stennikov et al. [19] presented an object- (MAR) and Brillouin fiber optic sensors (BFOSs). BFOSs, in addition to
oriented computer modeling technology to construct the DTs of pipe IoT-based pressure and temperature sensors, constitute distributed
line systems for online hydraulic calculations. Lieberman et al. [20] sensor networks, which are used to comprehensively sense the structural
discussed the potential benefits of using big data analytics and IoT stress states of underground pipelines. Then, a joint data-physics driven
principles to develop functional DTs of underground infrastructures and model is presented to achieve dynamic modeling of the evolution and
improve livability and resilience in modern cities. Carvalho [21] pre deterioration of specific asset structures. A MAR-based human–asset
sented a DT framework combining building information modeling (BIM) interaction interface is proposed to visualize geometric attributes, real-
and asset management (AM) techniques to support the maintenance and time monitoring data and structural safety assessment results. The
operation of water and wastewater utility assets. According to a litera feasibility of the proposed methodology for automated safety assessment
ture survey, pipeline DT technology is still in its infancy, and there are was validated by experimental investigations in a real pipe network.
few reports on using DTs for the structural safety assessment of under
ground pipelines. 2. Methodology development
The application of DTs to underground pipelines faces the following
challenges. (1) The sensor network technique underpins the mutual in Underground pipelines are not visible and are large-scale. These
teractions between DTs and hidden buried assets. Although nonde pipelines are subjected to complex combined loads and inevitably suffer
structive test (NDT)-based in-line inspection (ILI) tools, including from structural deterioration during long-term service. A framework of a
magnetic flux leakage (MFL), ultrasonic wall measurement (UT) and DT system based on MAR and BFOSs is proposed for quantitative safety
remote field eddy current (RFEC) tools, are commonly adopted in oil and assessment and O&M decision-making support. As shown in Fig. 1, the
gas pipelines, they are local inspection techniques rather than real-time DT of an underground pipeline has three main components: distributed
sensing techniques. However, these methods cannot measure the stress sensor networks oriented toward the physical pipeline, a joint data
states along a pipeline. The supervisory control and data acquisition –physics driven model oriented toward the DT, and a human–asset
(SCADA) system uses only sparse low-sampling transducers to monitor interaction interface based on MAR oriented toward safety inspectors.
the internal pressure, flow rate and temperature parameters of pipelines; Based on this framework, the methodology is presented in detail as
however, this system is unable to provide information regarding struc follows. It should be noted that only continuous metallic pipelines are
tural deterioration during a lifecycle. The comprehensive sensing of the investigated in this study, and jointed pipelines (e.g., bell-and-spigot
structural state, including both stress/deformation and deterioration in cast‑iron and concrete pipes) are beyond the scope of the proposed
the pipeline, is recognized as one of the most critical issues for pipeline approach.
DTs. (2) Human–asset interactions and visualization are vital for un
derground pipelines since they are not visible in the buried environ 2.1. Distributed sensor networks
ment. The state of the art is summarized in Table 1. Augmented reality
(AR) is considered to be a friendly human–asset interaction technique The sensing scheme for DTs depends on the structural failure modes
due to its fast communication, 3D visualization and beyond-reality and deterioration mechanism. In general, a combination of circum
interaction for underground pipelines. However, the current AR-based stances leads to pipeline failure. Pipeline breakage is likely to occur
modeling and visualization methods mainly focus on the geometric when operational and environmental loads are imposed on deteriorated
attribute information (design and construction phases) rather than the pipes with corrosion, degradation and distress. According to O'Day et al.
2
[32] and Rajani and Klenier [33], the main failure modes of pipelines flexural response (i.e., bending deformation) is dominant in the longi
can be categorized into circumferential and longitudinal breaks. tudinal strain caused by surface faulting. With the exception of geo
Circumferential breaks are mainly caused by longitudinal stress due to hazards and foundation defects, according to Li and Feng [49], the
thermal contraction, soil settlement or voids, and third-party interfer failure of oil and gas pipelines and heating pipelines can be caused by
ence. Longitudinal breaks are the result of excessive hoop stress caused excessive thermal stress. The thermal exchange process between the
by internal pressure, soil cover, frost load, structural deterioration, etc. internal hot fluid and the pipe commonly lasts only a few hours before
Current design codes clearly present a calculation model for hoop stress reaching a steady state; thus, we assume that thermal loading does not
from external loads (soil overburden and traffic load) and internal occur simultaneously with geohazards and foundation defects. If no
pressure [34,35]. The thermal axial stress in a continuous pipeline is also thermal action is identified from the monitoring data, the measurements
considered in the model. Rajani et al. [36] developed a pipe–soil inter of a single BFOS will only reflect the bending responses in the vertical
action analytical model for the longitudinal stress of jointed pipelines. plane. The proposed single-sensor configuration has the advantage of
This model accounts for temperature differentials and explains the easy installation. At the same time, a distributed fiber optic temperature
frequently observed occurrence of circumferential breaks in pipelines. sensor is placed in parallel along the strain sensor on the pipe crown.
Notably, the longitudinal bending stress due to foundation defects This sensor independently measures the temperature around the pipe
(settlements and voids) or third-party interference is not considered in line, so it can provide leakage detection [50] as well as thermal
the model because these events cannot be predicted in advance. During compensation for the strain sensor. The proposed sensing scheme pro
long-term service, the longitudinal bending actions induced by foun vides distributed real-time monitoring of the longitudinal strain/stress
dation defects and third-party interference may occur with unforeseen for a pipeline DT.
magnitudes and at arbitrary times and locations during the lifecycle. Hoop stress in a pipeline is created by the soil cover, traffic load, and
However, third-party interference and foundation defects are the second internal pressure. Even with BFOSs, distributed monitoring of hoop
and third greatest contributions to pipeline failures [37–39], stress is not cost-effective or practical for underground pipelines. In
respectively. previous research [34,36,46,51–53], theoretical models were provided
As indicated earlier, longitudinal bending stress may occur arbi for hoop stress in underground metallic pipelines with diameters <500
trarily along pipelines. The commonly used discrete point sensors, i.e., mm. These theoretical models are related to critical loads, such as in
strain gauges and fiber Bragg grating (FBG) sensors, can only measure ternal pressure, overburden loads and traffic loads. According to
the structural strain/stress in limited specific locations. This contradicts Ahammed and Melcher [34] and Mahmoodian and Li [51], the varia
the fact that the locations of the longitudinal actions on a pipeline tions in surface loads (soil overburden and traffic load) have low con
cannot be determined in advance. As a cutting-edge sensing technique, tributions to failure with increased exposure period. Attention should be
Brillouin backscattering-based fiber optic sensors can provide real given to abnormal events related to surface loads by routine inspection
distributed measurements for strain as well as temperature with a instead of real-time monitoring. The internal pressure, temperature
sensing range of up to a few tens of kilometers [40–43]. BFOSs are ideal differential, and continuous corrosion play significant roles in pipeline
candidates for monitoring pipelines due to their large spatial range and failures [34,54]. Thus, we propose a sensing scheme for the real-time
relatively small spatial resolution [44]. Inaudi and Glisic [45] reported monitoring of internal pressure and temperature. The IoT-based digi
several significant field application examples using BFOSs, including tal temperature sensor and pressure transmitter are placed at a certain
leakage detection for brine and gas pipelines and strain and deformation distance along the pipeline. By using the measured internal pressure and
monitoring for gas pipelines. Based on strain and temperature moni temperature data, the hoop stress can be accurately estimated by the
toring data from BFOSs and structural analysis of buried pipelines, Feng well-developed analytical model [34]. The effect of corrosion was also
et al. [46] proposed a methodology for pipeline structural performance considered in the stress analysis by introducing the corrosion model
assessment and applied it to practical cases. The capability of distributed [51]. With increasing elapsed service time, well-planned inspections
sensors to accurately measure the strain/stress state relies on their po should be conducted periodically, and the data from ILI tools can be
sitions on the pipeline. To monitor the longitudinal strain, mounting a utilized to calibrate and update the corrosion model. In conjunction with
single Brillouin fiber optic strain sensor on the pipe crown was proposed, BFOSs, digital pressure and temperature gauges are used in real-time
as shown in Fig. 1. Ni et al. [47] and Lan et al. [48] showed that the sensor networks for the pipeline DT.
Digital pressure transmitter

Visualization of data and assessment results:
Digital temperature sensor
Mapping of structural state evolution and
deterioration between digital twins and
physical pipelines. Distributed sensor networks
Brillouin fiber optic strain sensor
Digital modeling Human-asset interaction Brillouin fiber optic temperature sensor
A dynamically updated asset-specific interface based on MAR
structural analysis model integrated
within real-time monitoring or Visualization of underground physical Digital temperature sensor
inspection data and decision-making pipelines through digital twins to guide
support. in-situ safety inspection and management. Digital pressure transmitter
Joint data-physics driven
structural analysis model
Visualization of geometric and physical attribute:
Reliability-based structural safety Mapping of geometric appearance, spatial location and

assessment model attribute information between digital twins and
physical pipelines.
Digital Twin Physical Pipeline
Fig. 1. Framework of the DT for an underground pipeline.
3
2.2. Digital modeling Poisson effect of hoop deformation under internal pressure.
Similarly, the hoop stress σ h is a result of internal pressure (σ ph), soil
A DT is a computational model or a set of coupled computation pressure (σ sh), and traffic load (σ Vh ) as follows:
models that evolves over time to persistently represent the structure,
σh (x) = σ ph (x) + σ sh (x) + σ Vh (x) (2)
behavior and context of a unique physical asset [55,56]. To fulfill this
goal, different models or model types can be adopted for different pur As indicated in Section 2.1, the longitudinal strain is monitored in
poses in DTs. In the construction sector, BIM or geographic information real time by BFOSs. Notably, the total longitudinal strain, including all
system (GIS)-based modeling approaches are adopted to establish visual components (internal pressure, temperature differential and longitudi
models at the building and city levels [57,58]. In addition to geometry nal bending), is measured. The problem is that the relationship between
visualization, data-driven modeling approaches have mostly been used thermal stress and strain under constrained underground conditions is
to map physical assets into a DT [16]. Smart data analysis using machine not simple. In contrast to bending stress, thermal stress is a result of the
learning (ML) and artificial intelligence (AI) play a major role in the axial movement of the pipeline being constrained under temperature
data-driven model [59]. However, the data-driven model has some differential actions; thus, thermal stress in the pipeline is not directly
downsides. The most serious limitation is the lack of interpretability. In calculated by the measured thermal strain. However, the axial thermal
particular, for safety-critical applications, black-box approaches are stress is the main contributor to circumferential breaks. Hence, the
unacceptable. Physics-based modeling is the classic approach to foremost issue for the digital model is to determine the occurrence and
analyzing, simulating and predicting the physical world. The major range of the thermal stress in the pipeline.
advantage of physics-based approaches is that they are generally less Under temperature differential actions, axial stress may be created
biased than data-driven models because they are mathematical de through restrained movement by the friction resistance between the
scriptions that characterize physical phenomena [60]. With the rapid pipe and soil. To prevent circumferential breaks caused by excessive
development of high-performance simulators, physics-based modeling axial thermal stress, devices such as compensators and sliding supports
approaches have been used to establish DT technology. The disadvan are commonly installed in the pipeline. Then, axial movement occurs at
tage of physics-based models is that they rely on parametric assump the unfixed end to release axial thermal stress. As a result, the pipeline is
tions. Therefore, they are mainly limited to the design phase of divided into two sections, i.e., the anchorage section and transition
engineering systems. section (in Fig. 2). The actual anchorage point xg cannot be determined
The overarching goal of this study is to develop digital models for from the physics-based model because pipe–soil friction has consider
data analysis, safety assessment, and health prediction of pipelines. As able uncertainty. In the proposed sensor networks, the temperatures
presented in Section 2.1, the actual longitudinal bending behavior in a inside and outside the pipeline can be measured by BFOSs and digital
pipeline cannot be determined by physics-based models in an a priori temperature sensors, respectively. By fusing these monitoring data, the
manner. On the other hand, hoop stress should be estimated by using a temperature differential ΔT ̃ values (where the tilde ̃• denotes the actual
structural analysis model with monitoring data (internal pressure and measurement in this study) are obtained along the pipeline. The tem
temperature). Neither pure physics-based nor pure data-driven ap perature differentials provide the signs of the possible axial movements.
proaches can individually model the stress state of pipelines. To close Once a temperature differential is detected, the difference in the
this gap, a joint data-physics-driven model is proposed for data analysis measured strains before and after temperature change is determined as
and structural state assessment in this study. This novel approach is the thermal strain along the pipeline. The section with strain changes is
intended to overcome the limitations of pure physics-based and pure the transition section, while the remaining part without changes can be
data-driven modeling methods. This model introduces monitoring data, seen as the anchorage section. According to the above analysis, the axial
i.e., longitudinal strain, internal pressure, and temperature differential thermal stress σ Tl (x) can be given by
data, into a well-understood physics-based model to accurately obtain ⎧
structural stress in a real-time manner. By utilizing real data, the new ⎪
⎨ ∼
− EαΔT (x), 0 ≤ x ≤ Lg (anchorage section)
modeling approach eliminates the shortfalls of parametric assumptions T
σl (x) = ∼ ∼T
(3)
⎪
and simultaneously improves the interpretability, accuracy and ⎩ − EαΔT (x) + E εl (x), Lg ≤ x ≤ L (transition section)
robustness of the model. Then, a reliability-based approach incorpo
rating the joint data-physics-based structural analysis model is proposed where E and α are the Young's modulus and thermal expansion coeffi
for online structural safety assessment. The presented modeling meth cient of the pipe material, respectively; ̃εTl is the thermal strain measured
odology allows DTs to dynamically update computational models inte within the transition section; and Lg and Lf are the lengths of the
grated with observational data so that they reflect the changing anchorage and transition sections identified by the measured strains,
structural state of physical pipelines. respectively.
If no temperature differentials are detected from the measured data,
2.2.1. Joint data-physics-driven structural analysis model then the variations in the measured strain are due to external actions
It is well known that the stress state is a direct indicator of structural such as foundation changes or third-party interference. This strain is
safety. In fact, pipe–soil interactions are complex under combined loads,
identified as longitudinal bending strain ̃εbl . Then, the longitudinal
which leads to a 3D stress state in the pipeline. The stress at an arbitrary
bending stress is directly calculated using Hooke's law with BFOS data.
location of the pipe wall includes three components, i.e., longitudinal
stress σl, hoop stress σ h, and radial stress σ r. According to references [34, b
σbl (x) = E • ̃ε l (x) (4)
61, 62], the radial stress can be neglected in the Tresca or Von Mises
failure criteria because the maximum principal stress mainly depends on With respect to the strain measurement, according to Kurashima
hoop and longitudinal stress for underground pipelines (Fig. 2). In et al. [63], external strain or temperature applied to the fiber leads to the
general, the effects of internal pressure, temperature differential, and proportionate Brillouin frequency shift (BFS) ΔvB, and temperature
bending actions on the longitudinal stress σ l are given as follows: compensation is required accordingly [46]:
σ l (x) = σ pl (x) + σ Tl (x) + σbl (x) (1) T

̃ε bl (x) or ̃
εl (x) =
̃ B (x)
ΔvB − (dvB /dTB ) • ΔT
(5)
dvB /dε
where x is an arbitrary location along the pipeline and σpl , σTl ,
and are σ bl
the stress components corresponding to internal pressure, temperature where x denotes the location along the fiber; (dvB/dε) and (dvB/dTB)
differential and longitudinal bending, respectively. σpl is created by the denote the strain coefficient and temperature coefficient of the BFOS,
4
Fixed Support Underground Pipeline Compensator Traffic Load
σh xg (Anchorage Point)
σl Internal Pressure
Thermal Deformation
Soil Pressure
Compensator
Contraction Load resistance
Foundation Defects
Anchorage Section Lg Transition Section Lf Internal Pressure
x
0 Lg L
(a) (b)
Fig. 2. Underground pipeline stress analysis configuration: (a) longitudinal stress components and (b) hoop stress components.
respectively; and ΔT ̃ B (x) is the temperature data acquired by the and k and n are the corrosion growth coefficients. This model provides a
distributed fiber optic temperature sensor at location x. simplified but practically meaningful expression to capture corrosion
When internal pressure is applied to the pipeline, Poisson's effect on growth over time. In the initial phase of service, the parameters k and n
the longitudinal stress is determined as follows: can be estimated by field data or experience. With increasing exposure
time, the parameters should be validated and updated with inspection
σpl (x) =
̃
vPD (6) data.
2δ By incorporating a corrosion model, a digital model was established
for the real-time analysis of pipeline stress. Although it has a solid
where P ̃ is the measured internal pressure; D and δ are the internal
theoretical basis, the critical loads, i.e., internal pressure and tempera
diameter and wall thickness of the pipeline, respectively; and v is Pois ture, are both obtained from measured data. Most importantly, the
son's ratio. longitudinal bending and axial thermal strains/stresses are directly ob
By jointly utilizing strain and temperature measurements, the total tained from measurements. Clearly, this model is different from the
longitudinal stress σ l is finally derived by Eqs. (1), (3), (4) and (6). classical stress analysis model based on assumed parameters. The pro
According to references [34, 52], the components of hoop stress in posed approach combines physics-based modeling and data-driven
Eq. (2) are given by the following equations: modeling with monitoring data, which enables more informed
̃
PD decision-making and reliable prediction based on a highly interpretable
σph (x) = (7) model.
2δ
6km • Cd γB2d • Eδr 2.2.2. Reliability-based structural safety assessment

σsh (x) = (8) The proposed digital models involve a physics-based model and
Eδ3 + 24kd Pr̃ 3
monitoring data simultaneously. Several parameters in the models are
6km • Ic Ct F • Eδr derived from design values rather than actual values. The structural
σVh (x) = ( ) (9) deterioration process (i.e., corrosion) is associated with a stochastic
Le Eδ3 + 24kd Pr ̃ 3
phenomenon. As a result, uncertainty inevitably exists in material and
geometrical properties, the physical models of structural stress analysis,
where r is the internal radius of the pipeline; Cd is the coefficient of earth
and the monitoring sensors. Aleatory and epistemic uncertainties should
pressure; Bd is the width of the ditch; γ is the unit weight of the soil
be effectively reduced for structural safety assessment and O&M
backfill; km is the bending coefficient; kd is the deflection coefficient; Ic is
decision-making. In this section, a digital model for pipeline structural
the impact factor; Ct is a surface load coefficient; Le is the pipe effective
assessment and service life prediction based on time-dependent reli
length; and F is the surface wheel load.
ability is proposed to quantify the uncertainty during lifecycle service.
By substituting Eqs. (7) to (9) into Eq. (2), the total hoop stress is
The failure condition of pipelines needs to be defined in DTs. Both
obtained. Note that all components of hoop stress, as well as σpl , are
the Von Mises and Tresca failure criteria are extensively adopted to
related to the wall thickness δ. The structural deterioration of the
assess the structural failure of ductile metal pipelines [34], and fracture
pipeline is mainly due to corrosion. Corrosion induces the loss of pipe
toughness-based limit states have been proposed to compute the reli
wall thickness in a relatively uniform or localized way. Hence, corrosion
ability of pipelines with localized defects such as corrosion pits [51]. For
should be considered in the structural analysis model, especially for
urban pipeline networks, strength theory-based approaches (i.e., Von
lifecycle service. The loss of wall thickness due to local corrosion is
Mises and Tresca) are more appropriate because they seldom perform
commonly obtained by in-line NDT inspection in practical engineering.
inner inspections. Without loss of generality, the Tresca failure criterion
Once the NDT inspection is performed, the inspection results can be
(i.e., maximum shear stress criterion) is adopted to formulate the time-
adopted in the proposed framework; accordingly, the stress concentra
dependent reliability-based structural safety and prediction model.
tion and localized corrosion growth models should be introduced into
the stress calculation at the corrosion pit locations [51]. This study only σeq (x) = σ h (x) − σ l (x) > [σ] (11)
accounts for the structural deterioration caused by uniform corrosion.
Uniform corrosion models have been extensively investigated in many where σeq is the equivalent stress and [σ ] is the allowable stress of the
studies on underground and subsea pipelines [64]. A widely used power pipe material. The above equation expresses the relationship between
law model of corrosion [65] was adopted in this study. the load effects acting on the pipeline and the resistances.
The deterioration process during long-term service may impair the
δ = δ0 − ktn (10) safety and serviceability of pipelines. The probability of failure is
defined as the probability of violating the performance function defined
where δ0 is the initial wall thickness; t is the exposure period in years;
by Eq. (11). By assuming that σ eq(x) and [σ ] are statistically independent
5
random variables with respect to the exposure time t, the (instanta current performance of pipelines. Moreover, lifetime reliability-based
neous) probability of failure at location x is [66]. prediction enables informed decision-making in pipeline O&M for
{ }
∫∞ long-term service. The proposed approach provides effective ways for
PF (x, t) = P σeq (x, t) − [σ] < 0 = FR (u, x, t)fQ (u, x, t)du (12) DT to achieve successful lifecycle management.
0
where FR(u, x, t) and fQ(u, x, t) are the instantaneous cumulative proba 2.3. Human–asset interaction interface
bility distribution function (CPF) of the resistance and the instantaneous
probability density function (PDF) of the load effect, respectively. In DTs enable the real-time sensing, analysis, assessment, and predic
DTs, the instantaneous probability of failure is adopted to quantify the tion of the structural states of underground pipelines. Undoubtedly, a
current stress states along the pipeline by integrating the real-time more efficient human–asset interaction interface is required for DTs. The
monitoring data. As a result, the monitoring data can aid in proper major obstacle in interactions between the O&M operator and under
assessment, and the probabilistic indicator can handle the randomness ground pipeline is that the buried asset is not visible, which makes it
induced by the aleatory and epistemic uncertainties. The instantaneous difficult to effectively find hidden defects and respond quickly to
CPF/PDF of failure can be estimated using the first-order reliability emergencies (i.e., pipe leakages or breaks). The human–asset interaction
method (FORM) or Monte Carlo simulation (MCS). interface needs to have location-based accessibility and intuitive visu
In addition to real-time assessment, a pipeline DT should have the alization. AR can allow an operator to have real-time interactions with
capacity to predict the service performance over a service or inter hidden assets during O&M activities. Extensive investigations have
inspection period for proper lifecycle management. A performance in demonstrated that AR is more specifically appropriate for in situ appli
dicator is required to address uncertainties associated with time- cations (references [72–77]). The implementation of AR as a human
dependent deterioration development, i.e., corrosion growth [66,67]. –asset interaction interface can add to the long-term value of DTs. In this
A lifetime reliability indicator is introduced in this study to estimate the study, a MAR-based approach is proposed to manage and visualize the
probability of the time to failure. The time to failure of pipeline Tf is data, analysis results and prediction outcomes, which allows a high
treated as a random variable that defines the elapsed time from degree of interaction and immersive visualization with underground
installing the pipeline for operation until it fails for the first time pipelines and DTs.
[68,69]. The cumulative probability of failure F(x, Tf), the probability
that the pipeline at location x is not functioning at time t, is defined as 2.3.1. Visualization of geometric and physical attributes
( ) ( )
∫t Visualization of geometric appearances is vital for O&M of buried
F x, Tf = P x, Tf ≤ t = fT (x, u)du (13) assets. Geometric and physical attributes include the geometry, location,
and structural parameters of pipelines. Since pipelines are large-scale,
0
where fT(x, u) is the PDF of the time to failure derived by performing both precision and weight need to be considered for attribute
statistical analysis. modeling. Highly detailed models can fundamentally ensure precise
The lifetime reliability measure F(x, Tf) in Eq. (13) cannot be calcu descriptions, while coarse models can improve modeling efficiency and
lated analytically. MCS based on stress analysis and the corrosion prevent a single monolithic model from being used for a whole regional
growth model is utilized to calculate the PDF of time to failure, and then, pipeline system [78]. Hence, a dynamic hierarchical method based on
the time-dependent probabilistic performance profile is built for the the Unity3d game engine and Mapbox software development kits (SDKs)
pipeline. The probabilistic performance indicator for service life is is proposed to model the geometric attribute information of under
estimated, while the remainder of the lifetime is predicted through the ground pipeline systems at multiple spatial scales according to different
statistical information of the monitoring data and pipeline structural levels of detail (LODs). Thus, the model is divided into a system-level
analysis model. After obtaining the lifetime reliability profile, a pre submodel and a unit-level submodel (in Fig. 3(d) and Fig. 3(e), respec
dictive strategy based on the threshold is proposed to schedule the in tively). The system-level and unit-level models serve different levels of
spection or maintenance time. The decision-making function can be information requirements. For instance, in the process of structural
defined according to reliability-based assessment and prediction: safety assessment, the system-level model describes the association and
( ) ( ) coupling relationships between assets, and the unit-level model de
inspct
Tinspct = Tj if F x, Tj ≥ Fthres Tj = 1, 2, ⋯, Nd (14) scribes the more complex structural state information for each asset.
The system-level geometric attribute model (GAM) consists of pipe
where Tinspct is the time of the inspection; Tj is the elapsed time (unit: lines and related facilities (i.e., manholes, pumps, and valves).
year); Nd is the design service life; and Finspct
thres is the threshold of lifetime Regarding the GAM generation of support facilities, the prefabs (Fig. 3
reliability for the first inspection. Notably, the threshold is determined (c)) of a manhole, a pump and a valve are made in Unity3d with refer
according to the target reliability indicators for pipeline safety. Based on ence to their geometric shapes, and then, the geometric models of sup
Eq. (14), the first inspection is planned when the lifetime reliability port facilities are dynamically generated on Mapbox map layers in Unity
indicator reaches the predetermined threshold. This strategy is appro based on the prefabs and location (Lat, Lon) collections of cloud-based
priate for newly built structures. With the data collected by the in storage platforms. Finally, their unique identifiers are matched, and
spection, the parameters of the corrosion growth model can be basic attributes are added by programming in C#. For pipeline GAM
accordingly validated and updated. The Bayesian approach is commonly generation at the system level, based on the location {(Lat, 1, Lon, 1, Dep,
adopted for model updating [70,71]. In this study, we only investigate 1), ⋯, (Lat, s, Lon, s, Dep, s)} and geometry of each pipeline, mesh vertices of
the first inspection plan for newly built pipelines. the pipeline cross sections are generated at the s nodes by rotating a
According to the time-dependent reliability analysis, digital models point n times (360/n)◦ at a distance r (pipeline radius) in the direction
for the quantitative assessment and prediction of structural safety are normal to the pipeline segment; then, the mesh model is generated from
established for pipelines. In the mutual interaction between DTs and the mesh vertices, as shown in Fig. 3(d). Furthermore, the unique
monitoring data, the real-time reliability can be obtained to assess the pipeline identifiers and basic attributes {material, age, type, operation and
6
Fig. 3. Geometric attribute model: (a) 3D geometry model dynamic generation in Unity; (b) detailed view of 3D geometry models; (c) prefabs of a manhole, a pump
and a valve; (d) pipeline geometry model generation process at the system level; and (e) pipeline geometry model generation process at the unit level.
maintenance records} are added from the infrastructure-as-a-service ⎧

(IaaS) to rapidly generate lightweight GAMs. The system-level GAMs ⎪
⎪
⎪
⎪
⎪
dynamically constructed in the Unity3d engine are shown in Fig. 3(a) ⎪
⎪
⎪
⎪
and (b). ⎪
⎪
⎪
⎪
⎪ Color(1, 0, 0), y(x) ≥ y1st+ ;
For structural monitoring and assessment, the refined pipeline GAM ⎪
⎪
⎪ ( )
⎪
at the unit level should be constructed. The dynamic generation process ⎪
⎪
⎪
⎪ Color 1, 1 −
y − y2st+
, 0 , y1st+ > y(x) ≥ y2st+ ;
⎪
of the unit-level GAM is similar to that of the system-level GAM, but the ⎪
⎪
⎪ y2st+
⎪
pipeline geometry model at the system level is not sufficient to visualize ⎪
⎪
⎪
⎨
(
y
)
the structural monitoring data, resulting in an imprecise description of Color
y2st+
, 1, 0 , y2st+ > y(x) ≥ yO ;
(15)
the pipeline structural state. Therefore, mesh vertices of the pipeline ⎪
⎪
⎪
⎪ ( )
⎪ y
cross section are generated every Δd (Δd is the sampling interval of the ⎪
⎪
⎪ Color 0, 1, , yO > y(x) ≥ y2st− ;
⎪ y2st−
distributed sensors) from the starting node to achieve a high LOD for the ⎪
⎪
⎪
⎪ ( )
⎪
model in Fig. 3(e). ⎪
⎪
⎪ y − y2st−
⎪
⎪ Color 0, 1 − , 1 , y2st− > y(x) ≥ y1st− ;
⎪
⎪ y2st−
⎪
⎪
2.3.2. Visualization of real-time data and assessment results ⎪
⎪
⎪ Color(0, 0, 1), y1st− > y(x).
⎪
The DT should provide operators with real-time structural informa ⎪
⎪
⎪
⎩
tion (including monitoring raw strain data, stress analysis data, and
reliability-based assessment results) to facilitate responding actions or ( )
O&M decisions with greater accuracy, efficiency and safety. In addition where y(x) is the monitoring data and result data ̃εbl , ̃εTl σ bl , σ Tl , σ eq or
to visualizing the geometrical appearance, this study presents a 3D reliability PF at the distributed monitoring point x; y1st+ and y1st− are the
nephogram method based on mesh vertices to render the visualization of first upper and lower thresholds of the datasets, respectively; y2st+ and
the monitoring data and assessment results, as shown in Fig. 4. y2st− are the second upper and lower thresholds of the datasets,
According to the unit-level GAM, the sampling interval of the respectively; and yO is the zero point of the datasets. When y(x) is the
( )
distributed sensors is Δd, and thus, the mesh vertices of the pipeline monitoring data and the result data are ̃ εbl ,̃εTl σ bl , σTl , σeq , y1st+, y2st+, yO,
cross section correspond to the monitoring data of the distributed Bril y2st− and y1st− are taken as max{y(x)}Lx=0 , max{y(x)/2}Lx=0 , 0,
louin optical fiber sensors. The monitoring data are further processed
min{y(x)/2}Lx=0 and min{y(x)}Lx=0 . When y(x) is the reliability PF, y1st+,
through a joint data–physics driven model (Section 2.2), and the stress
or reliability data are visualized through Eq. (15), with the mesh vertices y2st+, yO, y2st− and y1st− are taken as max{PF }Lx=0 , μ + max{PF /2}Lx=0 , μ,
rendered to form a 3D nephogram: μ − min{PF }Lx=0 and min{PF }Lx=0 , respectively, where μ = mean{PF }Lx=0 .
Color(R, G, B) represents the mesh colors of the vertices in the Unity3d
7
Fig. 4. 3D nephogram form based on the rendering of mesh vertices.
shader in terms of red–green–blue (RGB) values. receivers are used to receive high-precision positioning coordinates
using the global positioning system fix data (GGA) format [85]. The
2.3.3. Mobile augmented reality visualization mobile app obtains precise positioning coordinates from GNSS receivers
AR is considered to be an effective and fast communication and in real time via a Bluetooth connection so that the Mapbox map layers
interaction tool that can visualize virtual objects [79,80]. However, are perfectly aligned with the physical world. In addition, ARCore/
current AR alignment methods, such as quick response (QR) markers ARKit's plane anchoring capability (the so-called ARPlaneAnchor object
[81], simultaneous localization and mapping (SLAM) [82,83] and global [28]) can be used to “seamlessly” overlay DTs with physical assets. This
positioning systems (GPSs) [28], have two main limitations: (1) the system continuously searches for feature points to identify the plane of
positioning accuracy of currently commercially available SLAM systems the surroundings and creates ARPlaneAnchor objects to anchor the
and GPSs is far from sufficient for practical engineering requirements, Mapbox map layers. Alignment of the orientation angle is carried out
and (2) the application of QR markers is limited to the tabletop level, far using a combination of the electronic compass of a mobile device and
from the city level. In addition, the computer vision (CV)-based method manual rotation of the map layers [24]. Location-based human–asset
[84] is affected by the environment and is mainly used for the appear interaction is achieved by the above approach.
ance of buildings rather than for hidden pipelines. With the rapid
development of AR-enabled software (e.g., ARCore, ARKit) and micro 3. Case study
global navigation satellite system (GNSS) surveying receivers with
centimeter-level location accuracy (e.g., Lite RTK from Qianxun Spatial A case study was performed to investigate the feasibility of the
Intelligence, Fig. 5(b)), AR-GIS technology can be used to achieve pre proposed pipeline DT for structural safety assessment. The study area is
cise mapping of spatial locations between DTs and physical assets in the approximately 83,800 m2, including gas, heating, water and drainage
field (Fig. 5). This solution features lightweight equipment, good pre pipelines as well as inspection manholes, valves and pumps. Fig. 6 and
cision, environmental robustness and large-scale applicability. Table 2 show the details of the pipeline system. The hardware and
As shown in Fig. 5, GNSSs (e.g., GPS or BeiDou) are used to send software configurations were first introduced in the DT. Then, MAR-
signals for navigation and location. Ground-based augmentation sys based immersive visualization and in situ interactions were obtained
tems are responsible for collecting and processing satellite data and for the pipelines at the system level. The selected pipeline was investi
providing calibration data to the GNSS receiver in real time. GNSS gated in detail to demonstrate the capability of the DT for real-time
(b) (a)
Pipeline system GIS

data
BDS/GPS
GGA 4G/5G/WiFi
Differential
Calibration
(c)
point cloud
micro-GNSS receiver
and mobile device
Underground pipelines
Fig. 5. Mobile augmented reality visualization: (a) the flow process; (b) Lite RTK micro-GNSS surveyor receiver; and (c) portable mobile devices.
8
safety assessment and lifetime prediction. respectively, and the measurement accuracy was ±10 με. These BFOSs
were arranged along HeatingPipe_2 from the fixed support to the
compensator, with a total measurement length of 57.7 m. The mea
3.1. Hardware and software configurations surement length of GasPipe_6 was 61.3 m, and the BFOSs were arranged
from the south fixed support to the north fixed support (Fig. 6). Thus,
3.1.1. Sensor networks and instrumentation distributed measurements could be obtained along the sensor path. To
The modeling parameters of the pipeline networks were collected monitor the internal pressure and temperature, SITRANS P DS III digital
from the design documents and as-built surveys. The digital models were pressure transmitters and DS18B20 digital temperature sensors were
then built at the system level. As the pilot demonstration at the unit level placed on the pipeline. The SITRANS P DS III digital pressure trans
for structural safety assessment, GasPipe_6 and HeatingPipe_2 (as shown mitters were connected to a personal computer (PC) through an analog-
in Fig. 6 and Table 2) were selected to have distributed sensor networks to-digital (A/D) converter, while the DS18B20 digital temperature
installed for real-time monitoring of the structural states. These pipeline sensors were connected to the PC via an RS-422 data interface. The
segments are located below roads with high pedestrian and vehicular sampling frequency was 1 Hz. All monitoring data were transmitted to
traffic. BFOSs were used to monitor the longitudinal strain along the the cloud platform.
pipeline segments. The commercially available NZS-DSS-C02 and NZS-
DTS-C05 were mounted on the pipeline to measure the strain and tem 3.1.2. DT platform
perature based on the sensor topology presented in Section 2.1. The The Unity3d game engine (version: Unity 2019.4) and C# pro
sensing fiber core of these sensors is a common SMF-28 single-mode gramming language (version: C#8.0. NET Core 3.0) were used for geo
optical fiber with a sensing fiber. Commercially available fiber optic metric modeling and joint data-physics driven modeling and were used
sensors were used in this study. The strain and temperature coefficients to launch the final application on a mobile device.
were provided by the vender. These coefficients were calibrated by
laboratory tests before field application. During sensor installation, the 3.1.3. Cloud-based data communication
pipe surface was cleaned to remove dirt and grease. Then, high- IaaS is the basic delivery model of cloud computing, and it offers a
performance epoxy resin was applied on the top and two sides of the high level of computing capacity (e.g., networking, servers and data
strain sensor to ensure that it was firmly bonded to the surface of the storage) to users [86]. Alibaba Cloud was selected as the IaaS cloud
pipe [49]. During this process, the strain sensor was slightly preten platform for uploading and sending the data collected from the sensor
sioned to keep it straight along the longitudinal axis of the pipe. Thus, networks. The flexible structured query language (SQL) of the relational
good strain transfer was achieved, and the readings of the strain sensor database, in combination with 4G/5G/WiFi communication technolo
represented the longitudinal strain of the pipeline. Neubrex NBX-6050A gies, was adopted to support data create-read-update-delete (CRUD)
was adopted for real-time data acquisition from the BFOSs. The spatial operations rapidly and securely.
resolution and sampling interval were set to 10 cm and 5 cm,
3.1.4. Mobile equipment
A HUAWEI P30 Pro smartphone was selected as the mobile device in
this study. The system configuration had the following components: 2 ×
Cortex-A76 2.6 GHz + 2 × Cortex-A76 1.92 GHz + 4 × Cortex-A55 1.8
GHz central processing units (CPUs), 8 GB random access memory
(RAM), and 128 GB read-only memory (ROM). The micro-GNSS receiver
was a centimeter-accurate Lite RTK with a 5 Hz update rate, ± (10 + 1
× 10− 6Dbs) mm flat accuracy and ± (15 + 1 × 10− 6Dbs) mm altitude
accuracy (Dbs is the distance between the measurement location and the
nearby base station). The Lite RTK was mounted on the back of the
HUAWEI P30 Pro to accurately track the position of the mobile phone in
real time.
3.2. System-level DT service
At the system level, the DT mainly provides real-time information

interaction and visualization of underground pipeline systems. In prac
tical applications, a system-level DT superimposes virtual pipeline
models on realistic scenarios, which enables the operator to see under
ground pipelines in the field. The operators can inspect the geometric
shapes and geographic locations of the underground pipelines in the
immersive perspective view. In this study, areas 1, 2, and 3 in Fig. 7 were
used as examples to demonstrate that system-level DTs can improve
human awareness of underground pipelines. The system-level geometry
models, as shown in Fig. 7 (a), were created in Unity3d, including GIS
data. The geometry models as well as the attribute data were then in
tegrated into the MAR-based application. When the operator started the
application software, the MAR-based service was launched on site. From
the perspective of the user, all the pipelines in the neighboring area were
displayed on the screen, as shown in Fig. 7(b). The current location was
registered by Lite RTK. The top of the screen displayed the geographical
location of the mobile device with longitude and latitude values of
121.52403◦ and 38.87922◦ , respectively. The error of horizontal posi
tioning was only 1.4 cm in this test (Fig. 7). The buried depth was shown
Fig. 6. Underground pipeline system plan of the study area. by selecting a specific pipeline. For instance, by clicking on the #2
9
Table 2
Details of the physical pipelines.
Pipeline_ID Type Location Attributes
(latitude, longitude, depth)
Material Construction Date
DrainagePipe_0 DN400 (38.87923,121.52324,0.5)…(38.88085,121.52231,0.5) HDPE pipe 08/10/2011

DrainagePipe_1 DN400 (38.88009,121.52231,0.5)…(38.88156,121.52227,0.5) HDPE pipe 14/10/2011
GasPipe_0 DN300 (38.87921,121.52324,1.0)…(38.88004,121.52321,1.0) Seamless steel pipe 15/06/2015
GasPipe_2 DN300 (38.88004,121.52234,1.0), (38.87999,121.52147,1.0) Seamless steel pipe 20/06/2015
HeatingPipe_0 DN350 (38.88084,121.52324,1.6)…(38.88084,121.52147,1.6) Seamless steel pipe 20/08/2017
HeatingPipe_1 DN350 (38.87921,121.52482,1.6), (38.87921,121.52428,1.6) Seamless steel pipe 04/08/2017
HeatingPipe_2 DN350 (38.87921,121.52428,1.6), (38.87920,121.52362,1.6) Seamless steel pipe 09/08/2017
HeatingPipe_3 DN350 (38.87920,121.52362,1.6)…(38.88084,121.52147,1.6) Seamless steel pipe 18/08/2017
WaterPipe_0 DN350 (38.87924,121.52482,1.2)…(38.87922,121.52326,1.2) Ductile iron pipe 08/10/2011
WaterPipe_1 DN350 (38.87922,121.52326,1.2), (38.88005,121.52322,1.2) Ductile iron pipe 15/10/2011
heating pipeline, the selected pipeline was highlighted with a flashing the longitudinal strain were collected with BFOSs. After trench backfill
edge (left diagrams of Fig. 7(c)). The pipe ID and depth were then shown and road pavement application (from August 21, 2017, to September 28,
as HeatPipe_2 and − 1.6 m, respectively. By moving the device, the 2017), bending deformation occurred because of the differential set
positions were tracked, and the pipelines were accordingly displayed in tlement along the pipeline segment. Then, the bending strains were
the field of view of the screen. From Fig. 7(b), the pipelines and their monitored by comparing the current data with the baseline data. When
related facilities were correctly localized and immersively visualized in the operator enters the “Query Date” as “2017_09_28” and selects the
each area. The relative locations among the pipelines could be pin data type, the raw data (e.g., longitudinal strain or temperature) or the
pointed from the perspective of the operator. In underground con derived data (e.g., bending strain or equivalent stress) are displayed in
struction or rehabilitation actions, the existing pipelines can be the MAR-based DT. The visualization is presented in two ways: a 3D
accurately mapped in the field by using this MAR-based location service, nephogram and a 2D curve. In Fig. 8 (a) and (c), the bending strains are
which is a promising tool for preventing underground pipeline strikes. displayed in nephogram form. The different colors represent the
In addition to in situ mapping of underground pipelines, the system- different amplitudes of bending strain. The distribution of bending strain
level DT enables information querying in a friendly, interactive way. The along the pipeline segment is displayed in an intuitive way. According to
attribute data can be queried by clicking the pipeline. The nominal the bending strain nephogram, the high strain sector in the pipeline can
diameter, material, type, and age are displayed on the screen. The O&M be easily localized with the aid of MAR. Furthermore, the curves of
records, including content, time and operator, are shown to support bending strain and temperature as well as a magnified diagram are also
rational operation. The detailed information provided by the DTs is shown in the DT. The temperatures were uniformly distributed with an
magnified in the bottom right diagram of Fig. 7(c). The MAR-based amplitude of 23 ◦ C. This indicates that no leakage occurred along the
interaction enhances human perception and information extraction pipeline. However, the bending behavior was very complicated, with
from the virtual world to the real world, thus improving access to and tensile and compressive bending alternatively occurring in the pipeline.
utilization of information. By utilizing multisource heterogeneous data, Two peaks appeared at locations of 2.4 m and 56.3 m with amplitudes of
pipeline DTs can guide various real-life project tasks, such as commu 625 με and 938 με, respectively. Based on the O&M records of the
nication among different teams, project management and unbiased de system-level DT service in Fig. 7 (c), during the construction phase, the
cision-making. pipeline is subject to bending deformation due to the overburden loads,
unevenness of the pipe trench and foundation settlement [49]. Such a
3.3. Unit-level DT service large bending strain and its complicated distribution cannot be pre
dicted in advance. The monitoring data demonstrate that the proposed
As mentioned in Section 3.1, the pipeline with ID HeatingPipe_2 was sensing scheme is able to provide the bending strain for the pipeline DT
selected as the testbed to investigate the unit-level DT. The geometry, in real time. Moreover, MAR-based data visualization can add the values
material and load parameters of the pipeline are listed in Table 3. The of monitoring data for use and interpretation.
pipeline segment is subjected to altered loads and has a thermal The pipeline experienced a heating and boosting process from
compensator. In comparison with the water and gas pipelines, this October 28, 2017, to November 4, 2017. The internal temperature
pipeline suffers from large temperature differentials. The pipe–soil increased from 23 ◦ C to 57 ◦ C, as measured by the DS18B20 digital
interaction is more complicated during long-term service. The unit-level temperature sensors, while the internal pressure rose to 1.0 MPa, as
DT of the pipeline segment was established by integrating the real-time monitored by a pressure transmitter. At the same time, longitudinal
monitoring data with the joint data–physics driven models proposed in strain was monitored by the BFOSs. The longitudinal strain caused by
Section 2.2. In the MAR-based application, the unit-level DT is entered the process is shown by the nephogram and curves in the DT (in Fig. 8
by clicking the DigitalTwin button (see Fig. 7(c)). At this level, the (b) and (d)). The measured strain did not present a clear bending feature,
monitoring data and stress analysis results are visualized for in situ data but instead a two-section shape. By fusing the internal temperature and
query, which makes the preliminary assessment of the pipeline segment pressure data, this was recognized as axial thermal strain because a
intuitive. Then, the reliability-based structural safety assessment and thermal compensator (i.e., bellows) was set in the right end of the
lifetime prediction are obtained from the DT to support O&M decision- pipeline segment. The position of the bellows could be identified by
making. All the assessments are performed in the MAR environment in using the system-level DT. For comparison, the longitudinal strain was
real time. measured 72 h later. As shown in Fig. 8(d), the longitudinal strain after
72 h was almost a flat line (i.e., dark blue line) with very small fluctu
3.3.1. Data visualization and real-time assessment ations. No bending or axial deformation occurred during the normal
HeatingPipe_2 was installed on August 9, 2017. The baseline data of operation phase. The amplitudes of the dark red curve (T1) before 21.77
10
Fig. 7. Location-based service for system-level DTs: (a) aerial view of the pipeline system in Unity; (b) in situ view for the inspector; and (c) comprehensive in
formation query for underground pipelines in the field.
m were approximately 0. Thus, this part could be identified as the primary stress required to balance the external loads, resulting in an
anchorage section, while the measured strains from 21.77 m to 57.70 m allowable stress that is never permitted to be higher than σy on the
demonstrated a gradually increasing trend that could be approximately outside wall of the pipe [87]. As shown in Fig. 8 (e), the peak longitu
fitted by a straight line, and this part was seen as the transition section. dinal bending stress of HeatingPipe_2 was 193 MPa, which did not reach
Notably, the complex soil–pipe interaction (induced by uneven bedding, the yield strength (235 MPa) of the pipe material. According to the
imperfect friction, etc.) was nonuniform along the pipeline, leading to ASME Code [87] and Chinese Code [88], thermal stress is classified as
T1 not being an ideal linear curve in the transition section. To the best of the secondary stress developed by the self-constraint of the adjacent
our knowledge, field observations of axial thermal deformation in un material or the structure, and it should be analyzed using shakedown
derground pipelines have rarely been reported in the literature. It is theory. Therefore, a range of stress variation larger than the yield
evident that thermal deformation cannot be obtained without BFOSs. strength can occur in the case of self-limiting thermal deformation [89].
When the bending strain and axial thermal strain were obtained, all According to the design code [88], the allowable value [σ] in Eq. (11)
the monitoring data (including the internal pressure and temperature should be min{2σy, σb} = 375 MPa. As shown in Fig. 8 (f), the maximum
differential) and the design parameters were input into the joint data equivalent stress was 262 MPa at 2.36 m in the anchorage section, while
–physics driven stress analysis models. The longitudinal and hoop the maximum value was 295 MPa at 56.77 m in the transition section.
stresses at arbitrary locations in the monitoring segment were calcu The peak equivalent stresses were all less than the allowable value. The
lated. To assess structural safety, the longitudinal bending stress is the pipeline segment can be preliminarily determined to be in a safe state.
11
Table 3 A comparison of the JC method and MCS method for computing the
Random variables and corresponding parameters. real-time reliability is shown in Fig. 9(b). This demonstrates that
Type of Symbol Description Mean value Coefficient of compared with the MCS method, the JC method demonstrates accept
variables variation able accuracy of the calculation results, with an average error of 0.18%,
Real time/ but saves significant time costs (MCS: 5.9 h; JC: 1.7 s) and makes real-
Lifetime
time assessment possible. Precise online assessment can be achieved in
E
Modulus of
206 (GPa) 0.033
pipeline DTs for early warning and timely control of emergencies.
elasticity
Material
σy Yield strength 235 (MPa) 0.05
σb Tensile strength 375 (MPa) 0.05
3.3.2. Service life prediction and detection/maintenance strategy
D Nominal diameter 350 (mm) 0.01 For long-term service, predictive O&M must be performed with the
δ0 Wall thickness 9 (mm) 0.05 aid of a pipeline DT. With increasing service time, corrosion plays an
Geometry
Le
Pipe effective
– 0.03 increasingly important role in structural deterioration. Based on real-
length
time monitoring and a corrosion growth model, the lifetime reliability
Bd Width of ditch – 0.05
Calculation of the pipeline was analyzed. As previously mentioned, the most
Cd 1.32 0.2 dangerous locations in the anchorage section and transition section were
coefficient
Ic Impact factor 1.25 0.2 at 2.36 m and 56.77 m, respectively, as obtained from the real-time
Surface load assessment. Since these locations are more susceptible to damage that
Ct 0.12 0.15
coefficient
Thermal expansion 12.6 × 10− 6 would lead to pipeline failure, they were selected for lifetime reliability-
α 0.1 based failure analysis. By performing an MCS with 250,000 samples, the
coefficient ( C− 1)
◦
v Poisson's ratio 0.3 0.023 evolutionary process of the PDF of the time to arrive at the limit state of
Coefficients k Multiplier constant 0.226 0.56 allowable stress is plotted in Fig. 10. The PDF of the time, as well as the
Exponential
n 0.53 0.26 mean and standard deviation of time t, needed to reach various equiv
constant
Bending moment alent stresses can be obtained from Fig. 10(a). The evolution processes
km 0.235 0.15 with and without bending stress are totally different. In Fig. 10(b), the
coefficient
kd
Deflection
0.108 0.15
PDFs of the time to failure at the most dangerous locations in the
coefficient anchorage and transition sections are obtained, and they have a higher
− 5
1.89 × 10
γ Unit weight of soil 0.1 probability density in the interval [0, 50] than that without accounting
(N/mm3)
Wheel load of 41.2 × 103 for the bending stress. The probability of failure for a specific service
Loading F 0.1
traffic (N) duration (e.g., 50 years) can be estimated by integrating the PDF from
P
̃ Internal pressure – 0.05/0.1 0 to this time period.
Temperature Based on the lifetime reliability analysis, the DT presents a lifetime
ΔT
̃ – 0.05/0.15
Monitoring differential
performance profile of the pipeline throughout its service life, as shown
Longitudinal
̃εL
bending strain
– 0.01/0.01 in Fig. 11. Then, the predictive O&M strategy can be recommended. In
this study, only the first inspection schedule is investigated. The basis of
Note: The random variables all obey a normal distribution, and the statistics are
the inspection plan is threshold-based. The threshold that triggered the
given by [34,51,53].
first inspection was assumed to be 10− 2 (similar to that in [51,91]).
When the CPF reached 10− 2, the corresponding time of the first in
However, Fig. 8(f) shows that the equivalent stress derived from the spection was 9 years. If the bending stress was neglected, then the
joint data–physics driven model was greater than that from the common counterpart increased to 17.2 years. There was an 8.2-year difference.
calculation model. The commonly used method underestimates the The DT provides a more prudent decision basis from the monitoring
stress because the actual bending strain cannot be obtained. The results data. Before the first inspection, once the changes are detected by the
based on in situ monitoring indicate that the actual stress state of the distributed sensor networks, the structural state should be reassessed by
pipeline might be dominated by the bending actions due to differential using the new data, and the lifetime reliability should be updated
settlements. Therefore, the BFOS-based DT can achieve a more accurate accordingly. When the first inspection is completed, the inspection data
estimation of the real stress state for the underground pipeline. of wall thinning are used to validate the corrosion growth model. Sub
The uncertainties related to the models, parameters, and sensors sequently, the Bayesian method can be adopted to update the parame
were considered in the reliability-based structural safety assessment. For ters (i.e., k and n) in the corrosion growth model. From the perspective of
the DT, the reliability computation was performed by using the FOSM- the lifecycle, the DT can dynamically assess and predict the structural
based JC method (i.e., the improved first-order second-moment state of the pipeline in a self-evolving manner.
method), and the analysis results are shown in Fig. 9(a) with a reliability
index nephogram. The MAR-based interaction interface showed the real-
4. Conclusions
time reliability index of the pipeline. Most of the reliability indices along
the pipeline were >6.0. The reliability indices at the peak stress loca
DTs pertain to sensing, assessment, prediction and control of struc
tions were reduced to 4.74 and 3.51; however, these values were greater
tural states through the real-time mutual interaction between the
than the target reliability index of 3.2 specified by the design code [90].
physical asset and its DT. This paper presents BFOS- and MAR-based
The real-time reliability showed that the pipeline was globally safe after
methodologies to develop DTs for underground pipelines. Sensor net
the heating and boosting process. The reliability-based method provides
works integrating BFOSs and IoT-based digital sensors were proposed to
structural safety assessment in a probabilistic sense by accounting for
enable real-time monitoring of the structural states of underground
the inevitable uncertainties.
pipelines. In the framework of digital modeling, a joint data–physics
12
Fig. 8. AR curves and 3D nephogram of the monitoring data and result data: (a) 3D nephogram of bending strain ̃εbl ; (b) 3D nephogram of measured thermal strain
εTl ; (c) curves of bending strain ̃εbl ; (d) curves of measured thermal strain ̃εTl ; (e) curves of longitudinal bending stress σbl and axial thermal stress σTl ; and (f) curves of
̃
equivalent stress σeq.
driven model was proposed to assess the structural stress state in a Based on the case study, the following conclusions can be drawn:
distributed manner, and reliability-based structural safety assessment
models were built for lifetime prediction and informed O&M decisions. • The proposed sensor networks play an important role for the DT. The
A MAR-based human–asset interaction interface was developed to pro unpredictable deformations, e.g., longitudinal bending strain and
vide real-time interactions between the operator and the hidden pipe axial thermal strain, along the pipeline were measured. The results of
lines. A series of field experiments were carried out to investigate the the case study showed that the bending response caused by over
feasibility of the proposed methodology. burden loads, the unevenness of the pipe trench and foundation
13
10 10
9
9
8
8
7
7
6
Reliability index β
Error(%)
6 5
4
5
3
MCS method (5.9h) JC method (1.7s)
4
Error Avg
2
3
1
0.18
2
0 5 10 15 20 25 30 35 40 45 50 55
Distance (m)
(a) (b)
Fig. 9. Real-time safety assessment: (a) AR graph and 3D nephogram and (b) comparison between the JC method and MCS method (sample size is 250,000).
Fig. 10. Lifetime measures of the #2 heating pipeline: (a) time-dependent equivalent stress with PDFs of time when σ eq(t)= 200 MPa and 300 MPa and (b) PDF of the
time to failure.
settlement reached 80% of the allowable value (i.e., yield strength), than that of the common calculation model that cannot account for
which could be a main failure factor for the underground pipeline. the actual structural stress state.
The measurements of larger-than-expected bending stress induced • A MAR-based approach was proposed to manage and visualize the
by several conditions during the construction and early service data, analysis results and prediction outcomes of pipelines. In the
phases offer field evidence for updating the common assumptions of human–asset interface, both virtual and real worlds coexist, and
current calculation models. immersive visualization allows hidden underground pipelines to be
• The joint data–physics driven model can accurately simulate the seen in the field. The case results found that operators experienced a
actual stress state in the pipeline by integrating the monitoring data new and improved human–asset interaction in which pipeline loca
and the well-understood structural analysis theory. The case results tion distributions and dangerous locations of structural states were
showed that the equivalent stress derived from the joint data–physics visualized to assist with construction and maintenance. The pro
driven model, benefiting from the integration of monitoring data, posed human–asset interaction interface can achieve easier, more
was greater than that from the common calculation model. intuitive and more efficient O&M for underground pipelines.
Reliability-based assessment and prediction can address the un
certainties associated with the models, parameters and sensors to The proposed DT provides a promising tool for the lifecycle man
produce more rational results and optimize the inspection schedule. agement of underground pipelines. However, the present study requires
The case study found that the corresponding time of the first in the installation of BFOSs on the pipeline, and thus, it is limited to newly
spection predicted by the DT was 9 years, which was 8.2 years earlier built pipelines. The proposed stress calculation models are applicable to
14
0.50
σeq(t) without σbl
0.45 σeq(t) with σbl in the anchorage section
σeq(t) with σbl in the transition section
Cumulative Probability of Failure Value F(t)

0.40
0.35
0.30
0.25
0.20
0.15 8.2 YRS
0.10
tdet = 9 YRS
tdet = 14.1 YRS
0.05
F(t)=10-2
0.00
tdet = 17.2 YRS
0 10 20 30 40 50
Time (Years)
Fig. 11. Cumulative probability of failure and probabilistic maintenance planning.
continuous thin-walled metallic pipelines with diameters <500 mm, and [5] M. Liu, S. Fang, H. Dong, C. Xu, Review of digital twin about concepts,
technologies, and industrial applications, J. Manuf. Syst. 58 (2021) 346–361,
the scope of this study does not include bell-and-spigot pipes. Moreover,
https://doi.org/10.1016/j.jmsy.2020.06.017.
the current study only concerns the decision basis for the first inspection [6] F. Tao, Q. Qi, Make more digital twins, Nature 573 (7775) (2019) 490–491,
according to the uniform corrosion model. In the future, in-line robotic https://doi.org/10.1038/d41586-019-02849-1.
systems and localized damage models should be further adopted for [7] Gartner, Prepare for the Impact of Digital Twins, from. https://www.gartner.com/
smarterwithgartner/prepare-for-the-impact-of-digital-twins, 2017. Accessed date:
lifecycle O&M by accounting for more complicated deterioration sce 25 April 2019.
narios (i.e., corrosion pits, cracks, etc.). [8] M. Shafto, M. Conroy, R. Doyle, E. Glaessgen, C. Kemp, J. LeMoigne, L. Wang,
Modeling, Simulation, Information Technology & Processing Roadmap,
Technology Area 11, Washington, in: https://sbir.nasa.gov/content/modeling
Declaration of Competing Interest -simulation-information-technology-and-processing, 2012.
[9] C. Ye, L. Butler, B. Calka, M. Iangurazov, Q. Lu, A. Gregory, M. Girolami,
The authors declare that they have no known competing financial C. Middleton, A Digital Twin of Bridges for Structural Health Monitoring,
Proceedings of the 12th International Workshop on Structural Health Monitoring,
interests or personal relationships that could have appeared to influence Stanford University, 2019, pp. 3–10, https://doi.org/10.12783/shm2019/32287.
the work reported in this paper. [10] Q. Lu, A.K. Parlikad, P. Woodall, G. Don Ranasinghe, X. Xie, Z. Liang,
E. Konstantinou, J. Heaton, J. Schooling, Developing a digital twin at building and
city levels: case study of West Cambridge campus, J. Manag. Eng. 36 (3) (2020)
Data availability 1–19, https://doi.org/10.1061/(asce)me.1943-5479.0000763.
[11] I. Errandonea, S. Beltrán, S. Arrizabalaga, Digital twin for maintenance: a literature
Data will be made available on request. review, Comput. Ind. 123 (2020) 1–14, https://doi.org/10.1016/j.
compind.2020.103316.
[12] T.R. Wanasinghe, L. Wroblewski, B.K. Petersen, R.G. Gosine, L.A. James, O. De
Acknowledgments Silva, G.K.I. Mann, P.J. Warrian, Digital twin for the oil and gas industry: overview,
research trends, opportunities, and challenges, IEEE Access 8 (2020)
This work was supported by the National Natural Science Foundation 104175–104197, https://doi.org/10.1109/access.2020.2998723.
[13] C. Boje, A. Guerriero, S. Kubicki, Y. Rezgui, Towards a semantic construction
of China (Grant No. 52079024) and the Fundamental Research Funds for digital twin: directions for future research, Autom. Constr. 114 (2020) 1–16,
the Central Universities (Grant No. DUT20LAB133). https://doi.org/10.1016/j.autcon.2020.103179.
[14] K. Lin, Y.-L. Xu, X. Lu, Z. Guan, J. Li, Digital twin-based collapse fragility
assessment of a long-span cable-stayed bridge under strong earthquakes, Autom.
References Constr. 123 (2021) 1–13, https://doi.org/10.1016/j.autcon.2020.103547.
[15] N. Mohammadi, J.E. Taylor, Smart City Digital Twins, 2017 IEEE Symposium
[1] D. Liu, J. Chen, S. Li, W. Cui, An integrated visualization framework to support Series on Computational Intelligence (SSCI), 2017, pp. 1–5, https://doi.org/
whole-process management of water pipeline safety, Autom. Constr. 89 (2018) 10.1109/SSCI.2017.8285439.
24–37, https://doi.org/10.1016/j.autcon.2018.01.010. [16] Q. Lu, X. Xie, A.K. Parlikad, J.M. Schooling, Digital twin-enabled anomaly
[2] T. Zheng, Z. Liang, L. Zhang, S. Tang, Z. Cui, Safety assessment of buried natural detection for built asset monitoring in operation and maintenance, Autom. Constr.
gas pipelines with corrosion defects under the ground settlement, Eng. Fail. Anal. 118 (2020) 1–16, https://doi.org/10.1016/j.autcon.2020.103277.
129 (2021) 1–14, https://doi.org/10.1016/j.engfailanal.2021.105663. [17] E. Fowler, T. Salazar, Think smart, pump smart, World Pipelines 7 (2018) 63–66.
[3] M. Grieves, J. Vickers, Digital twin: Mitigating unpredictable, undesirable https://www.worldpipelines.com/magazine/world-pipelines/july-2018.
emergent behavior in complex systems, in: F.-J. Kahlen, S. Flumerfelt, A. Alves [18] M. Xiong, L. Gu, Z. Wu, Y. Deng, S. Li, Y. Zou, Z. Li, C. Feng, Construction and
(Eds.), Transdisciplinary Perspectives on Complex Systems: New Findings and application of digital twin in the in-service oil and gas pipeline, Oil and Gas Stor.
Approaches, Springer International Publishing, Cham, 2017, pp. 85–113, https:// Transportat. 38 (5) (2019) 503–509. https://kns.cnki.net/kcms/detail/13.1093.te.
doi.org/10.1007/978-3-319-38756-7_4. 20190214.0851.002.html.
[4] T.H.J. Uhlemann, C. Lehmann, R. Steinhilper, The digital twin: realizing the cyber- [19] V.A. Stennikov, A. Alekseev, E. Mikhailovsky, N.I. Voropai, S.V. Alekseenko, S.
physical production system for industry 4.0, Procedia CIRP 61 (2017) 335–340, P. Filippov, N.A. Yusifbeyli, B. Sereter, P. Changwei, F.J. Lin, M. Negnevitsky,
https://doi.org/10.1016/j.procir.2016.11.152. C. Rehtanz, J.Y. Yoon, Development of information and computing technologies
for modelling of pipeline and hydraulic systems, in: E3S Web of Conference:
ENERGY-21- Sustainable Development & Smart Management, Sept. 7, 2020-Sept.
15
11, 2020, EDP Sciences, Irkutsk, Russia, 2020, pp. 1–9, https://doi.org/10.1051/ [45] D. Inaudi, B. Glisic, Long-range pipeline monitoring by distributed fiber optic
e3sconf/202020902002. sensing, J. Press. Vessel. Technol. 132 (1) (2009) 1–9, https://doi.org/10.1115/
[20] J. Lieberman, A. Leidner, G. Percivall, C. Ronsdorf, Using big data analytics and IoT 1.3062942.
principles to keep an eye on underground infrastructure, in: 2017 IEEE [46] X. Feng, Y. Han, Z. Wang, H. Liu, Structural performance monitoring of buried
International Conference on Big Data (Big Data), Dec. 11, 2017-Dec. 14, 2017, pipelines using distributed fiber optic sensors, J. Civ. Struct. Heal. Monit. 8 (3)
IEEE, Boston, MA, U.S, 2017, pp. 4592–4601, https://doi.org/10.1109/ (2018) 509–516, https://doi.org/10.1007/s13349-018-0286-3.
BigData.2017.8258503. [47] P. Ni, I.D. Moore, W.A. Take, Distributed fibre optic sensing of strains on buried
[21] W.O. Carvalho, BIM and AM to manage critical and relevant water and wastewater full-scale PVC pipelines crossing a normal fault, Géotechnique 68 (1) (2018) 1–17,
utilities assets, in: E. Toledo Santos, S. Scheer (Eds.), Proceedings of the 18th https://doi.org/10.1680/jgeot.16.P.161.
International Conference on Computing in Civil and Building Engineering, Aug. 18, [48] H. Lan, G. Ma, N.A. Hoult, I.D. Moore, Stress concentrations due to simulated
2020-Aug. 20, 2020, Springer International Publishing, Cham, 2020, pp. 697–720, corrosion pits in buried metal pipes under longitudinal bending, J. Civ. Struct.
https://doi.org/10.1007/978-3-030-51295-8_49. Heal. Monit. 12 (4) (2022) 785–796, https://doi.org/10.1007/s13349-021-00522-
[22] G. Schall, S. Zollmann, G. Reitmayr, Smart Vidente: advances in mobile augmented 8.
reality for interactive visualization of underground infrastructure, Pers. Ubiquit. [49] M. Li, X. Feng, Multisensor data fusion-based structural health monitoring for
Comput. 17 (7) (2012) 1533–1549, https://doi.org/10.1007/s00779-012-0599-x. buried metallic pipelines under complicated stress states, J. Civ. Struct. Heal.
[23] X. Zhang, Y. Han, D. Hao, Z. Lv, ARGIS-based outdoor underground pipeline Monit. (2022) 1–13, https://doi.org/10.1007/s13349-022-00609-w.
information system, J. Vis. Commun. Image Represent. 40 (2016) 779–790, [50] M. Nikles, B.H. Vogel, F. Briffod, S. Grosswig, F. Sauser, S. Luebbecke, A. Bals,
https://doi.org/10.1016/j.jvcir.2016.07.011. T. Pfeiffer, Leakage detection using fiber optics distributed temperature
[24] G. Soria, L.M. Ortega Alvarado, F.R. Feito, Augmented and virtual reality for monitoring, in: Vol. 5384, Proceedings of SPIE - The International Society for
underground facilities management, J. Comput. Inf. Sci. Eng. 18 (4) (2018) 1–9, Optical Engineering, 2004, pp. 18–25, https://doi.org/10.1117/12.540270.
https://doi.org/10.1115/1.4040460. [51] M. Mahmoodian, C.Q. Li, Reliability-based service life prediction of corrosion-
[25] R. Napolitano, A. Blyth, B. Glisic, Virtual environments for visualizing structural affected cast Iron pipes considering multifailure modes, J. Infrastruct. Syst. 24 (2)
health monitoring sensor networks, data, and metadata, Sensors (Basel) 18 (1) (2018) 1–8, https://doi.org/10.1061/(asce)is.1943-555x.0000417.
(2018) 1–14, https://doi.org/10.3390/s18010243. [52] A. Amirat, A. Mohamed-Chateauneuf, K. Chaoui, Reliability assessment of
[26] M.C. Li, S. Amerudin, Z. Mohamed Yusof, Development of augmented reality underground pipelines under the combined effect of active corrosion and residual
pipeline Visualiser (Arpv) application for visualising underground water pipeline, stress, Int. J. Press. Vessel. Pip. 83 (2) (2006) 107–117, https://doi.org/10.1016/j.
Int. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. XLII-4 (W16) (2019) 365–373, ijpvp.2005.11.004.
https://doi.org/10.5194/isprs-archives-XLII-4-W16-365-2019. [53] R. Sadiq, B. Rajani, Y. Kleiner, Probabilistic risk analysis of corrosion associated
[27] G. Zhi, Z. Liao, W. Tian, X. Wang, J. Chen, A 3D dynamic visualization method failures in cast iron water mains, Reliab. Eng. Syst. Saf. 86 (1) (2004) 1–10,
coupled with an urban drainage model, J. Hydrol. 577 (2019) 1–14, https://doi. https://doi.org/10.1016/j.ress.2003.12.007.
org/10.1016/j.jhydrol.2019.123988. [54] B. Rajani, A. Abdel-Akher, Re-assessment of resistance of cast iron pipes subjected
[28] A. Fenais, S.T. Ariaratnam, S.K. Ayer, N. Smilovsky, Integrating geographic to vertical loads and internal pressure, Eng. Struct. 45 (2012) 192–212, https://doi.
information systems and augmented reality for mapping underground utilities, org/10.1016/j.engstruct.2012.06.019.
Infrastructures 4 (4) (2019) 1–17, https://doi.org/10.3390/ [55] AIAA Digital Engineering Integration Committee, Digital Twin: Definition & Value,
infrastructures4040060. American Institute of Aeronautics and Astronautics (AIAA) and Aerospace
[29] A. Fenais, S.T. Ariaratnam, N. Smilovsky, Assessing the accuracy of an outdoor Industries Association (AIA). https://www.aia-aerospace.org/report/digital-twin-
augmented reality solution for mapping underground utilities, J. Pipeline Syst. paper, 2020.
Eng. Pract. 11 (3) (2020) 1–9, https://doi.org/10.1061/(ASCE)PS.1949- [56] M.G. Kapteyn, J.V.R. Pretorius, K.E. Willcox, A probabilistic graphical model
1204.0000474. foundation for enabling predictive digital twins at scale, Nat. Computat. Sci. 1 (5)
[30] X. Tan, L. Fan, Y. Huang, Y. Bao, Detection, visualization, quantification, and (2021) 337–347, https://doi.org/10.1038/s43588-021-00069-0.
warning of pipe corrosion using distributed fiber optic sensors, Autom. Constr. 132 [57] L. Ding, Y. Zhou, B. Akinci, Building information modeling (BIM) application
(2021) 1–13, https://doi.org/10.1016/j.autcon.2021.103953. framework: the process of expanding from 3D to computable nD, Autom. Constr.
[31] X. Tan, A. Abu-Obeidah, Y. Bao, H. Nassif, W. Nasreddine, Measurement and 46 (2014) 82–93, https://doi.org/10.1016/j.autcon.2014.04.009.
visualization of strains and cracks in CFRP post-tensioned fiber reinforced concrete [58] P. Conejos, F. Martínez Alzamora, M. Hervás Carot, J. Alonso Campos,
beams using distributed fiber optic sensors, Autom. Constr. 124 (2021) 1–13, Development and Use of a Digital Twin for the Water Supply and Distribution
https://doi.org/10.1016/j.autcon.2021.103604. Network of Valencia (Spain), 17th International Computing & Control for the
[32] D.K. O’Day, R. Weiss, S. Chiavari, D. Blair, Water Main Evaluation for Water Industry Conference (CCWI), Sept. 1, 2019-Sept. 4, 2019, pp. 2–4. Exeter, U.
Rehabilitation/Replacement, Denver, CO: American Water Works Association K, https://www.researchgate.net/publication/340236600_Development_and_Use_
Research Foundation; Cincinnati, US-EPA, OH, 1987. https://nepis.epa.gov. of_a_Digital_Twin_for_the_Water_Supply_and_Distribution_Network_of_Valencia_
[33] B. Rajani, Y. Kleiner, Comprehensive review of structural deterioration of water Spain.
mains: physically based models, Urban Water 3 (3) (2001) 151–164, https://doi. [59] J.M.D. Delgado, L.J. Butler, I. Brilakis, M.Z.E.B. Elshafie, C.R. Middleton,
org/10.1016/S1462-0758(01)00032-2. Structural performance monitoring using a dynamic data-driven BIM environment,
[34] M. Ahammed, R.E. Melchers, Probabilistic analysis of underground pipelines J. Comput. Civ. Eng. 32 (3) (2018) 1–17, https://doi.org/10.1061/(ASCE)
subject to combined stresses and corrosion, Eng. Struct. 19 (12) (1997) 988–994, CP.1943-5487.0000749.
https://doi.org/10.1016/S0141-0296(97)00043-6. [60] A. Rasheed, O. San, T. Kvamsdal, Digital twin: values, challenges and enablers from
[35] AWWA, C101–67(R1977): ANSI/AWWA Standard for Thickness Design of Cast- a modeling perspective, IEEE Access 8 (2020) 21980–22012, https://doi.org/
iron Pipe, American Water Works Association, Denver, Colorado, 1997. 10.1109/ACCESS.2020.2970143.
https://www.awwa.org. [61] S. Tesfamariam, B. Rajani, R. Sadiq, Possibilistic approach for consideration of
[36] B. Rajani, C. Zhan, S. Kuraoka, Pipe-soil interaction analysis of jointed water uncertainties to estimate structural capacity of ageing cast iron water mains, Can.
mains, Can. Geotech. J. 33 (3) (1996) 393–404, https://doi.org/10.1139/t96-061. J. Civ. Eng. 33 (8) (2006) 1050–1064, https://doi.org/10.1139/l06-042.
[37] M. Almahakeri, A. Fam, I.D. Moore, Experimental investigation of longitudinal [62] R. Amaya-Gómez, M. Sánchez-Silva, E. Bastidas-Arteaga, F. Schoefs, F. Muñoz,
bending of buried steel pipes pulled through dense sand, J. Pipeline Syst. Eng. Reliability assessments of corroded pipelines based on internal pressure – a review,
Pract. 5 (2) (2014) 1–10, https://doi.org/10.1061/(ASCE)PS.1949-1204.0000141. Eng. Fail. Anal. 98 (2019) 190–214, https://doi.org/10.1016/j.
[38] Guidelines for the Seismic Design of Oil and Gas Pipeline Systems, American engfailanal.2019.01.064.
Society of Civil Engineers (ASCE), New York, 1984. ISBN: 978-0-87262-428-3. [63] T. Kurashima, M. Tateda, Thermal effects on the Brillouin frequency shift in
[39] D.G. Honegger, R.W. Gailing, D.J. Nyman, Guidelines for the Seismic Design and jacketed optical silica fibers, Appl. Opt. 29 (15) (1990) 2219–2222, https://doi.
Assessment of Natural Gas and Liquid Hydrocarbon Pipelines, 2002 4th org/10.1364/AO.29.002219.
International Pipeline Conference, Vol. 4th International Pipeline Conference, [64] M. Xie, Z. Tian, A review on pipeline integrity management utilizing in-line
Parts A and B, 2002, pp. 563–570, https://doi.org/10.1115/IPC2002-27330. inspection data, Eng. Fail. Anal. 92 (2018) 222–239, https://doi.org/10.1016/j.
[40] F. Ansari, Practical implementation of optical fiber sensors in civil structural health engfailanal.2018.05.010.
monitoring, J. Intell. Mater. Syst. Struct. 18 (8) (2007) 879–889, https://doi.org/ [65] V. Kucera, E. Mattsson, Atmospheric corrosion, in: F. Mansfeld (Ed.), Corrosion
10.1177/1045389X06075760. Mechanisms, Marcel Dekker, New York, 1987. ISBN: 978-1-003-06567-8.
[41] Z. Fang, H. Su, F. Ansari, Modal analysis of structures based on distributed [66] B.R. Ellingwood, Risk-informed condition assessment of civil infrastructure: state
measurement of dynamic strains with optical fibers, Mech. Syst. Signal Process. of practice and research issues, Struct. Infrastruct. Eng. 1 (1) (2005) 7–18, https://
159 (2021) 1–15, https://doi.org/10.1016/j.ymssp.2021.107835. doi.org/10.1080/15732470412331289341.
[42] Y. Yao, M. Yan, Y. Bao, Measurement of cable forces for automated monitoring of [67] D.M. Frangopol, Life-cycle performance, management, and optimisation of
engineering structures using fiber optic sensors: a review, Autom. Constr. 126 structural systems under uncertainty: accomplishments and challenges, Struct.
(2021) 1–14, https://doi.org/10.1016/j.autcon.2021.103687. Infrastruct. Eng. 7 (6) (2011) 389–413, https://doi.org/10.1080/
[43] X. Tan, Y. Bao, Q. Zhang, H. Nassif, G. Chen, Strain transfer effect in distributed 15732471003594427.
fiber optic sensors under an arbitrary field, Autom. Constr. 124 (2021) 1–13, [68] M. Rausand, A. Barros, A. Hoyland, System Reliability Theory: Models, Statistical
https://doi.org/10.1016/j.autcon.2021.103597. Methods, and Applications, 3rd edition, Wiley, New York, 1994. ISBN: 978–1–119-
[44] B. Glisic, Comparative study of distributed sensors for strain monitoring of 37395-7, https://www.wiley.com/en-us/System+Reliability+Theory%3A+Mo
pipelines [J], Geotechn. Eng. (J. Southeast Asia Geotechn. Soc. SEAGS) 50 (2) dels%2C+Statistical+Methods%2C+and+Applications%2C+3rd+Edition
(2019) 28–35. https://collaborate.princeton.edu/en/publications/comparative -p-9781119373520.
-study-of-distributed-sensors-for-strain-monitoring-of. [69] M. Soliman, D.M. Frangopol, Reliability and remaining life assessment of fatigue
critical steel structures: integration of inspection and monitoring information, in:
16
Proceedings of Structures Congress 2013, May 2, 2013-May 4, 2013, Pittsburgh, Health Monit. 20 (4) (2020) 1980–1996, https://doi.org/10.1177/
PA. U.S, 2013, pp. 709–720, https://doi.org/10.1061/9780784412848.063. 1475921720977017.
[70] D.M. Liu, Frangopol, Utility and information analysis for optimum inspection of [81] M. Chu, J. Matthews, P.E.D. Love, Integrating mobile building information
fatigue-sensitive structures, J. Struct. Eng. 145 (2) (2019) 1–12, https://doi.org/ modelling and augmented reality systems: an experimental study, Autom. Constr.
10.1061/(ASCE)ST.1943-541X.0002257. 85 (2018) 305–316, https://doi.org/10.1016/j.autcon.2017.10.032.
[71] Y. Sahraoui, R. Khelif, A. Chateauneuf, Maintenance planning under imperfect [82] R. Napolitano, Z. Liu, C. Sun, B. Glisic, Combination of image-based documentation
inspections of corroded pipelines, Int. J. Press. Vessel. Pip. 104 (2013) 76–82, and augmented reality for structural health monitoring and building pathology,
https://doi.org/10.1016/j.ijpvp.2013.01.009. Frontiers in built, Environment 5 (2019) 1–14, https://doi.org/10.3389/
[72] H.-L. Chi, S.-C. Kang, X. Wang, Research trends and opportunities of augmented fbuil.2019.00050.
reality applications in architecture, engineering, and construction, Autom. Constr. [83] T. Bailey, H. Durrant-Whyte, Simultaneous localization and mapping (SLAM): part
33 (2013) 116–122, https://doi.org/10.1016/j.autcon.2012.12.017. II, IEEE Robot. Automat. Magaz. 13 (3) (2006) 108–117, https://doi.org/10.1109/
[73] H.W. Chuah, Why and Who Will Adopt Extended Reality Technology? Literature MRA.2006.1678144.
Review, Synthesis, and Future Research Agenda, Available at SSRN, 2018, https:// [84] H. Bae, M. Golparvar-Fard, J. White, High-precision vision-based mobile
doi.org/10.2139/ssrn.3300469. Accessed date: December 13. augmented reality system for context-aware architectural, engineering,
[74] G.A. Kumar, A.K. Patil, T.W. Kang, Y.H. Chai, Sensor fusion based pipeline construction and facility management (AEC/FM) applications, Visualizat. Eng. 1
inspection for the augmented reality system, Symmetry 11 (10) (2019) 1–19, (1) (2013) 1–13, https://doi.org/10.1186/2213-7459-1-3.
https://doi.org/10.3390/sym11101325. [85] W. Li, Y. Han, Y. Liu, C. Zhu, Y. Ren, Y. Wang, G. Chen, Real-time location-based
[75] E.M. Wetzel, W.Y. Thabet, The use of a BIM-based framework to support safe rendering of urban underground pipelines, ISPRS Int. J. Geo Inf. 7 (1) (2018) 1–17,
facility management processes, Autom. Constr. 60 (2015) 12–24, https://doi.org/ https://doi.org/10.3390/ijgi7010032.
10.1016/j.autcon.2015.09.004. [86] P.X.W. Zou, P. Lun, D. Cipolla, S. Mohamed, Cloud-based safety information and
[76] X. Wang, P.E.D. Love, M.J. Kim, C.-S. Park, C.-P. Sing, L. Hou, A conceptual communication system in infrastructure construction, Saf. Sci. 98 (2017) 50–69,
framework for integrating building information modeling with augmented reality, https://doi.org/10.1016/j.ssci.2017.05.006.
Autom. Constr. 34 (2013) 37–44, https://doi.org/10.1016/j.autcon.2012.10.012. [87] Boiler and Pressure Vessel Code, American Society of Mechanical Engineers
[77] V. Gomez-Jauregui, C. Manchado, J. Del-Castillo-Igareda, C. Otero, Quantitative (ASME), New York, 2012, https://doi.org/10.1115/1.859865.ch8. ISBN: 978-0-
evaluation of overlaying discrepancies in mobile augmented reality applications 79185-986-5.
for AEC/FM, Adv. Eng. Softw. 127 (2019) 124–140, https://doi.org/10.1016/j. [88] Technical specification for directly buried hot-water heating pipeline in city, in:
advengsoft.2018.11.002. CJJ/T 81–2013, China, 2013, ISBN: 1511223799, 2013. Accessed July 26, 2013
[78] Y. Deng, J.C.P. Cheng, C. Anumba, Mapping between BIM and 3D GIS in different from, http://www.mohurd.gov.cn/info/public.
levels of detail using schema mediation and instance comparison, Autom. Constr. [89] M. Staat, M. Heitzer, Numerical Methods for Limit and Shakedown Analysis:
67 (2016) 1–21, https://doi.org/10.1016/j.autcon.2016.03.006. Deterministic and Probabilistic Problems, John von Neumann Institute for
[79] D.D.L. Mascareñas, J.P. Ballor, O.L. McClain, M.A. Mellor, C.-Y. Shen, B. Bleck, Computing, Central Institute for Applied Mathematics, 2003. ISBN: 3–00–010001-
J. Morales, L.-M.R. Yeong, B. Narushof, P. Shelton, E. Martinez, Y. Yang, 6, http://hdl.handle.net/2128/2926.
A. Cattaneo, T.A. Harden, F. Moreu, Augmented reality for next generation [90] General principles on reliability for structures, ISO2394. https://www.iso.
infrastructure inspections, Struct. Health Monit. 20 (4) (2020) 1957–1979, https:// org/obp/ui/#iso:std:iso:2394:ed-3:v1:en, 1998. (Accessed 16 October 2022).
doi.org/10.1177/1475921720953846. [91] N.M. Okasha, D.M. Frangopol, A. Decò, Integration of structural health monitoring
[80] D. Maharjan, M. Agüero, D. Mascarenas, R. Fierro, F. Moreu, Enabling in life-cycle performance assessment of ship structures under uncertainty, Mar.
human–infrastructure interfaces for inspection using augmented reality, Struct. Struct. 23 (3) (2010) 303–321, https://doi.org/10.1016/j.marstruc.2010.07.004.
17

Digital Twin For Underground Pipeline

Uploaded by

Copyright:

Available Formats

Digital Twin For Underground Pipeline

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Digital Twin For Underground Pipeline

Uploaded by

Copyright:

Available Formats

Automation in Construction 144 (2022) 104617

Contents lists available at ScienceDirect

Brillouin fiber optic sensors and mobile augmented reality-based digital

1. Introduction corporations would use DT technology by 2021, corresponding to a 10%

Digital pressure transmitter

Reliability-based structural safety Mapping of geometric appearance, spatial location and

Digital Twin Physical Pipeline

Fig. 1. Framework of the DT for an underground pipeline.

σ l (x) = σ pl (x) + σ Tl (x) + σbl (x) (1) T

Fixed Support Underground Pipeline Compensator Traffic Load

Anchorage Section Lg Transition Section Lf Internal Pressure

6km • Cd γB2d • Eδr 2.2.2. Reliability-based structural safety assessment

maintenance records} are added from the infrastructure-as-a-service ⎧

Fig. 4. 3D nephogram form based on the rendering of mesh vertices.

Pipeline system GIS

3.2. System-level DT service

At the system level, the DT mainly provides real-time information

DrainagePipe_0 DN400 (38.87923,121.52324,0.5)…(38.88085,121.52231,0.5) HDPE pipe 08/10/2011

Cumulative Probability of Failure Value F(t)

0.15 8.2 YRS

Fig. 11. Cumulative probability of failure and probabilistic maintenance planning.

You might also like