Opportunities and Challenges to Develop Digital Twins for Subsea Pipelines

: A vision of the digital twins of the subsea pipelines is provided in this paper, with a coverage of the current applications and the challenges of the digital twins in the design, construction, service life, and assessments of life extension. Digital twins are described as a paradigm combining multi-physics modelling with data-driven analytics, which are used to mirror the life of its corresponding twin. Realistic virtual models of structural systems are shown to bridge the gap between design and construction and to mirror the real and virtual worlds. Challenges in properly using the new tools and how to create accurate digital twins considering data acquired during the construction phase are discussed. The key opportunities for improved integrity management using the digital twin are data contextualization, standardization, automated anomaly detection, and learning through sharing. The collection, interpretation and sharing of data, and cyber-security are identiﬁed as some of the main challenges.


Introduction
Subsea pipelines are the flowlines connecting a subsea wellhead to a manifold or a platform, or the export lines (trunk lines) used as a long-distance transportation system for the oil and gas [1,2].Important considerations of the pipeline system are the efficient operation and the development of the system for future needs.Due to the low price of oil in recent years, oil and gas field operators are looking beyond the traditional operational maintenance strategy for the sake of reducing the downtime caused by the planned or sometimes unplanned preventive maintenance in the production field, thus reducing the operational cost (OPEX) [3,4].
In the digital solutions to subsea integrity management (SIM), the use and management of data bring benefits to the daily operations, e.g., the increased efficiency, optimization, reduction in cost, and safety [3].New technologies and associated processes are required for increased safety and better monitoring of the subsea assets.
Recent advancements in information and communication technologies, including cloud computing, high-performance processors, high dimensional visualization capabilities, internet of things (IoT), wearable technologies, additive manufacturing (AM), big data analytics (BDA), artificial intelligence (AI), autonomous robotic systems, submarine drones, and blockchain technology, have catalysed digital adoption across industries.These new technologies have facilitated the cyber-physical integration by which data can be collected, analysed, and visualized to make informed decisions and to serve as a basis for simulations to optimize operations [5].The concept of cyber-physical interaction and associated simulation is referred to as a Digital Twin.
The digital twin draws considerable interest as it can provide a cost-effective, reliable and intelligent maintenance strategy based on the machine learning (ML) algorithm [4] and assess extension life projects.The digital twin concept consists of three components: (1) the physical asset, (2) the virtual representation of the asset, and (3) the connection between the previous two components.The connection includes the information transferred from the physical asset to the digital twin and from the digital twin to the asset.A clear business purpose is a key principle in the development of digital twins to provide value [3].
The concept of the digital twin has also been used in offshore structures as platforms and subsea pipelines throughout the stages of construction and operation life.In this paper, the vision of the digital twin and its evolution, and the applications of the digital twin in subsea pipelines in design, construction and service life are reviewed.Further discussions and comments are also addressed.

Literature Review 2.1. Digital Twin
A digital twin is defined as one virtual representation of a system (or an asset) that can be used to calculate the states of the system and to make the information available.The integrated models and data of the digital twin can provide decision support over the life cycle of the system or asset.The idea to use a twin model can be dated back to the 1970s when two identical space vehicles were built in NASA's Apollo program to allow for mirroring of the conditions of the space vehicle [6].
Although the concept of digital twins was first put forward in 1991 [7], the model of the digital twin was first introduced in 2002 as a concept for product lifecycle management (PLM).After its initial names of mirrored spaces model (MSM) and information mirroring model (IMM), the model was finally referred to as the Digital Twin in 2011 [8].
The digital twin has been applied to a wide range of industries including aerospace [9,10], automotive [11,12], healthcare [13][14][15][16], manufacturing [17][18][19], and smart city [20,21].Some early applications of the digital twin can be found in NASA's spacecraft [22][23][24] and in the US Air Force's jet fighters [25].More recently, world-class vendors such as Dassault Systèmes, PTC, and Siemens use the digital twin concept in their PLM.The digital twin model was also proposed for the robust deployment of IoT [26].Aiming at developing digital twins for all built cars, TESLA enabled synchronous data transmission between their cars and the factory [27].From 2017 to 2019, the Gartner listed the digital twin as being among the top 10 strategic technology trends and indicated that billions of things would have their digital twins within 3-5 years [28][29][30].
Although the concept of the digital twin is not new, it was more a descriptive one and lacked auxiliary technologies in the first few years [31][32][33].Figure 1 shows the number of results obtained by searching "digital twin" as a "topic" in the Web of Science database, indicating a tremendous increase of research interest in the digital twin both in industry and academia, especially since 2017/2018.In terms of the country of study, China is the leading country in publications on the digital twin, followed by Germany and the USA.
Even though the digital twin technologies are of interest to companies in the industry, a major part of the studies (about 79%) are conducted by researchers in academia, according to a survey [34].Most academic research focuses on improving modelling techniques rather than optimizing data and implementing digital twins.Only 6% of the articles are initiated in the industry; the remaining are collaborative studies between academia and industry.There are few connections between industry and academia, particularly due to commercial secrecy.To meet the new requirement from applications, Tao et al. [35] presented an extended five-dimension digital twin model, adding data and services to the initial threedimensional digital twin concept.Figure 2 shows the key technologies for modelling each dimension of the digital twin.Qi et al. [36] classified the tools for the service applications of digital twins into platform service tools, simulation service tools, optimization service tools, and diagnostic and prognosis service tools.A list of the tools for each category is shown in Figure 3.To meet the new requirement from applications, Tao et al. [35] presented an extended five-dimension digital twin model, adding data and services to the initial three-dimensional digital twin concept.Figure 2 shows the key technologies for modelling each dimension of the digital twin.Qi et al. [36] classified the tools for the service applications of digital twins into platform service tools, simulation service tools, optimization service tools, and diagnostic and prognosis service tools.A list of the tools for each category is shown in Figure 3.To meet the new requirement from applications, Tao et al. [35] presented an extended five-dimension digital twin model, adding data and services to the initial threedimensional digital twin concept.Figure 2 shows the key technologies for modelling each dimension of the digital twin.Qi et al. [36] classified the tools for the service applications of digital twins into platform service tools, simulation service tools, optimization service tools, and diagnostic and prognosis service tools.A list of the tools for each category is shown in Figure 3.   ANSYS Twin Builder is one of the multi-physics simulation packages used to add physical realism to the digital twins [27].It contains extensive application-specific libraries and features third-party tool integration and can enable engineers to quickly build, validate and deploy the digital models of physical assets at an appropriate level of detail.
As with all new concepts, there are also obstacles to the further application of the digital twin.For instance, it is sometimes difficult to collect the most important data from thousands of sensors that track vibration, temperature, environmental conditions, force, speed, or power.In addition, the data can be spread among different owners in various formats.Consequently, the digital twin may fail to echo what is going on in the real world, leading to poor decisions made by the managers accordingly [37].In this regard, further research is needed for improved data collection and processing methods to implement the communication interface between the digital and the physical twins.In terms of standardization, the development of universal platforms and tools are also required for further applications of digital twins.

Subsea Pipelines and Application of Digital Twin
The first oil pipeline is widely believed to have been installed in the 1860s in the USA to transport crude oil [38].Since then, subsea pipelines have become the most economical means of efficiently transporting crude oil, natural gas, and other products from offshore installations for the exploitation of subsea reservoirs.Figure 4 illustrates different uses of subsea pipelines associated with platforms and wells, including the infield flowlines and export pipelines.
In a harsh sea environment, sufficient structural strength is necessary for designing, analysing, and operating subsea pipelines to guarantee the safety and integrity of subsea pipelines [38,39] throughout their life span.ANSYS Twin Builder is one of the multi-physics simulation packages used to add physical realism to the digital twins [27].It contains extensive application-specific libraries and features third-party tool integration and can enable engineers to quickly build, validate and deploy the digital models of physical assets at an appropriate level of detail.
As with all new concepts, there are also obstacles to the further application of the digital twin.For instance, it is sometimes difficult to collect the most important data from thousands of sensors that track vibration, temperature, environmental conditions, force, speed, or power.In addition, the data can be spread among different owners in various formats.Consequently, the digital twin may fail to echo what is going on in the real world, leading to poor decisions made by the managers accordingly [37].In this regard, further research is needed for improved data collection and processing methods to implement the communication interface between the digital and the physical twins.In terms of standardization, the development of universal platforms and tools are also required for further applications of digital twins.

Subsea Pipelines and Application of Digital Twin
The first oil pipeline is widely believed to have been installed in the 1860s in the USA to transport crude oil [38].Since then, subsea pipelines have become the most economical means of efficiently transporting crude oil, natural gas, and other products from offshore installations for the exploitation of subsea reservoirs.Figure 4 illustrates different uses of subsea pipelines associated with platforms and wells, including the infield flowlines and export pipelines.Pipeline design is affected by multiple factors identified during the early stages (i.e., conceptualization, front end engineering and design stage, and detailed engineering phase).These factors include the site selection, route survey, local environmental conditions, material selection, wall-thickness design, pipeline protections, and the budget of the project [40,41].During the pipelay process, the pipelines are deposited from installation vessels, with new pipe segments welded to form the pipelines.A proper pipe-lay is required to avoid excessive bending stresses that may cause fractures and buckling.The low temperature on the seafloor can sometimes cause a global contraction of the pipeline, while heat coming up from the reservoir fluid may induce local thermal expansion of the pipeline [42,43].Temperature or pressure changes in the operation process can also cause "pipeline walking" [44,45] in the case of improper restraints of the pipelines.Corrosion and erosion, which occur due to chemical attacks and abrasion from the internal fluids, are one of the major limiting factors in the continued operation of subsea pipelines [46][47][48][49][50].These factors necessitate the routine inspection of subsea pipelines and the use of monitoring technologies.
According to the B31G code [51] of the American Society of Mechanical Engineers (ASME), the geometry of the corrosion pits can be idealized as elliptical shapes, and a bulging factor can be applied when considering the defect geometry.The class society Det Norske Veritas (DNV) headquartered in Norway published the standard ST-F101 [52] and also one document of recommended practice RP-F101 [53] for pipeline applications.The American Petroleum Institute (API) Specification 5L [54] includes the requirements for manufacturing seamless and welded steel pipes in the transportation of oil and gas.One current challenge for subsea pipeline inspection is the often-deep waters.
More recently, sandwich pipes (SPs) were proposed as an effective alternative to the pipe-in-pipe (PIP) system for ultra-deepwater applications [55][56][57].In the SPs, a polyester foam material with low density and heat-conducting ability is filled between two metal pipes, providing high structural resistance with thermal insulation capability [58][59][60][61].
As reported by DNV [3], the recent digitalization requires the reduced cost of sensor technology and computational power, as well as cloud storage and computing.In a harsh sea environment, sufficient structural strength is necessary for designing, analysing, and operating subsea pipelines to guarantee the safety and integrity of subsea pipelines [38,39] throughout their life span.
Pipeline design is affected by multiple factors identified during the early stages (i.e., conceptualization, front end engineering and design stage, and detailed engineering phase).These factors include the site selection, route survey, local environmental conditions, material selection, wall-thickness design, pipeline protections, and the budget of the project [40,41].During the pipelay process, the pipelines are deposited from installation vessels, with new pipe segments welded to form the pipelines.A proper pipe-lay is required to avoid excessive bending stresses that may cause fractures and buckling.The low temperature on the seafloor can sometimes cause a global contraction of the pipeline, while heat coming up from the reservoir fluid may induce local thermal expansion of the pipeline [42,43].Temperature or pressure changes in the operation process can also cause "pipeline walking" [44,45] in the case of improper restraints of the pipelines.Corrosion and erosion, which occur due to chemical attacks and abrasion from the internal fluids, are one of the major limiting factors in the continued operation of subsea pipelines [46][47][48][49][50].These factors necessitate the routine inspection of subsea pipelines and the use of monitoring technologies.
According to the B31G code [51] of the American Society of Mechanical Engineers (ASME), the geometry of the corrosion pits can be idealized as elliptical shapes, and a bulging factor can be applied when considering the defect geometry.The class society Det Norske Veritas (DNV) headquartered in Norway published the standard ST-F101 [52] and also one document of recommended practice RP-F101 [53] for pipeline applications.The American Petroleum Institute (API) Specification 5L [54] includes the requirements for manufacturing seamless and welded steel pipes in the transportation of oil and gas.One current challenge for subsea pipeline inspection is the often-deep waters.
More recently, sandwich pipes (SPs) were proposed as an effective alternative to the pipe-in-pipe (PIP) system for ultra-deepwater applications [55][56][57].In the SPs, a polyester foam material with low density and heat-conducting ability is filled between two metal pipes, providing high structural resistance with thermal insulation capability [58][59][60][61].
As reported by DNV [3], the recent digitalization requires the reduced cost of sensor technology and computational power, as well as cloud storage and computing.Figure 5 illustrates the main technologies and enablers in digitalization for SIM: digital worker + support, inspection + data collection, and the analysis.
illustrates the main technologies and enablers in digitalization for SIM: digital worker + support, inspection + data collection, and the analysis.DNV define in the RP-A204 [62] six levels of capability (stages of the evolution) of digital twins, as illustrated in Figure 6.Real-time data streams are not required in Level 0 (standalone) but are in Level 1 (descriptive capability).Level 2 (diagnostic capability) provides support to monitor the conditions and detect the fault, together with troubleshooting.Health and condition indicators are further enriched in Level 3 (predictive) to support prognostic capabilities.Level 4 (prescriptive capability) can be used to provide recommended actions based on the predictions.In Level 5 (autonomous capability), the users can determine the functional element to perform actions or make decisions regarding the system.The digitalization of the data exchanges of the subsea pipelines plays an important role in avoiding the errors of the data during manual transcriptions.In order to simplify the data exchange between parties at the different stages of projects, the pipeline data exchange format (PDEF) collaborative initiative was proposed.It is an open-source joint DNV define in the RP-A204 [62] six levels of capability (stages of the evolution) of digital twins, as illustrated in Figure 6.Real-time data streams are not required in Level 0 (standalone) but are in Level 1 (descriptive capability).Level 2 (diagnostic capability) provides support to monitor the conditions and detect the fault, together with troubleshooting.Health and condition indicators are further enriched in Level 3 (predictive) to support prognostic capabilities.Level 4 (prescriptive capability) can be used to provide recommended actions based on the predictions.In Level 5 (autonomous capability), the users can determine the functional element to perform actions or make decisions regarding the system.The key enablers for digitalization in subsea integrity management [3].
DNV define in the RP-A204 [62] six levels of capability (stages of the evolution) of digital twins, as illustrated in Figure 6.Real-time data streams are not required in Level 0 (standalone) but are in Level 1 (descriptive capability).Level 2 (diagnostic capability) provides support to monitor the conditions and detect the fault, together with troubleshooting.Health and condition indicators are further enriched in Level 3 (predictive) to support prognostic capabilities.Level 4 (prescriptive capability) can be used to provide recommended actions based on the predictions.In Level 5 (autonomous capability), the users can determine the functional element to perform actions or make decisions regarding the system.The digitalization of the data exchanges of the subsea pipelines plays an important role in avoiding the errors of the data during manual transcriptions.In order to simplify the data exchange between parties at the different stages of projects, the pipeline data exchange format (PDEF) collaborative initiative was proposed.It is an open-source joint industry project (JIP) involving many companies in the pipeline industry, aiming at developing a standard format for the exchange of the data and describing the data used in the design of subsea pipelines [3].

Automated Creation of the Digital Twin during Construction
The oil and gas industry has been adopting digital twins of asset life cycle management (ALCM) in recent years.Subsea pipelines are unique when compared to other major infrastructure assets due to the fact that they are buried immediately during construction and are used to transport hazardous and explosive contents at high pressures.These facts result in an enormous responsibility for all phases of a pipeline asset's life cycle, including construction, operation, and long-term integrity management.
To accurately trace the condition of a pipeline asset at any time in its lifecycle to previous environmental conditions, operational conditions or events during its life, a single source of truth representing the entire meter-by-meter condition of the asset (i.e., a digital twin), must be created.The creation of the pipeline digital twin during pipeline construction can lead to a full representation of the asset from creation to decommissioning, which construction management, quality control, engineering and integrity can all refer to [63].The goal is to create the data required by inspectors during construction and capture it in such a manner that it can also be easily or instantly accessed and used by the operator during construction for quality management and operations and maintenance post-commissioning.
Field trials were completed to test and evaluate workflows and sensor platforms for the creation of digital twins for pipelines in 2017 [63].After pipeline stringing, welding, and lowering, the open ditch and lowered pipe were scanned in the final resting position prior to backfilling.The laser imaging, detection, and ranging (LiDAR) system and downwardfacing camera were suspended above the open ditch on the side-mounted boom.The system has a measurement rate of up to 700,000 measurements per second, high accuracy fibre optic gyro (FOG) inertial measurement units (IMU), and a high-end survey-grade global positioning system (GPS).The trials resulted in highly accurate pipeline centrelines, weld locations, depth to cover (DoC), and ditch geometry capture in digital formats.The resulting point clouds contain about 6.2 million highly-accurate points over a scan distance of 370 m.The weld location was added to the geographic information system (GIS) based on the imagery interpretation.
In addition to the remote sensing method, techniques of magnetic flux leakage and acoustic detection are commonly used to detect the cracks of subsea pipelines.One of the limitations of the LiDAR inspection is the assumption used in the algorithm of the software.Since visualization is one key feature of digital transformation, methods to obtain an accurate and useable result from the massive amount of measurement data need to be developed and improved in the further application of digital twins.
A pipeline design automation was introduced based on the cloud-based digital twin McDermott SubseaXD [64].SubseaXD is collaborating with Dassault Systemes/Simulia to leverage their 3D Experience platform with a smart, collaborative PLM platform.The web-based graphical user interface (GUI) worked as an integrated system producing a 3D digital field diagram, together with all pipeline design calculations in one digital platform.Various calculations, including wall thickness calculations based on API/DNV/ASME code check, on-bottom stability analysis, pipeline span analysis, pipeline end expansion analysis, out of straightness analysis, and pipeline buckling analysis are performed sequentially and systematically in the cloud using the metadata information (i.e., pipe data, soil, environment) available through Python-based API from the digital field data.Abaqus and Orcaflex are integrated with the SubseaXD, which can be used for detailed finite element analysis (FEA).It was stated that the automated pipeline design can save hours with fewer errors, thus saving on cost.
Based on the study in [63], the following recommendations for further digital twin creations were provided:

•
Refine the hardware used, and thus the field execution and subsequent data processing workflow.

•
Compare the digital twin centreline to the in-line inspection and weld location surveys, and seek to explain variances.

•
Make the results more readily available using cloud services and mobile communications to better expose the data to decision-makers.

•
Investigate the schedule and cost savings in greater detail.

•
Garner more operator feedback on the potential value from a construction quality control or quality assurance perspective.

•
Examine long-term impacts on the integrity and general operations and maintenance, as well as on failure investigations.

Update of the Digital Twin with Information Acquired in Inspections
Notable offshore production regions include the Gulf of Mexico (GoM), the North Sea, Brazil, West Africa, the Persian Gulf, Atlantic Canada, the Gulf of Thailand, the East and the South China Sea, the Caspian Sea, and the Southern and Western Australia [65].The GoM has the greatest number of offshore pipelines installed, followed by the North Sea.From 1952 to 2017, more than 72,000 km of pipelines have been installed in the GoM, and about 42,000 km of them are still active.The North Sea has the second-largest pipeline network with approximately 45,000 km of pipeline installed since 1966 [66].
However, the issues and obstacles make the inspection and monitoring of pipelines a challenging task.Problems may occur throughout the life of the pipe, and the environment of service is full of potential dangers [65].In the use of digital twin technology in offshore structures, the twin model must be updated during the service life with the data of corrosion and other structural degradation gathered during service life inspections and could be updatable according to the accidental damages.
A pipe segment is often coated during the manufacturing process to protect it against corrosion or abrasion.Additional layers of coating can be added, depending on the requirements.Some subsea pipelines have outer concrete coatings for protection against corrosion and impacts and weight stabilization [67,68].PIP designs may also be implemented with additional layers for protection and thermal insulation [69][70][71].In addition to the inspection of the primary metallic body, the coating layers may also require inspections for damage and debonding from the pipelines.Depending on the task and the technology used in the inspection, part of the pipelines may need to be stripped of coatings to be fully inspected.
Based on an investigation by The US Department of Transportation, the major failures of subsea pipelines were categorized into five categories: mechanical, operational, corrosion, natural hazards, and third party.All of these possible failures necessitate the routine inspection of the pipeline and the use of permanent monitoring technologies.Nowadays, one of the challenges for subsea pipeline inspection is the extreme water depths.Many well proved inspection technologies cannot be delivered to the pipeline without costly equipment and procedures.In the cases of multi-layered pipelines, the cost of stripping away the coatings for routine inspections in deepwater is impractical.In-line inspection (ILI) may be used to inspect the pipeline and the inner from the inside.Pigging is one of the ILI techniques in which devices referred to as "pigs" or scrapers are inserted into pipelines to perform inspection activities.Pigging can be conducted on a variety of pipeline sizes without having to stop the flow of material through the line.
Given the increasing capacity of satellite links, the potential of cloud computing and deep learning (DL) algorithms, the digital twin models with millimetre precision move the subsea asset management into a new era, enabling engineers to incorporate less margin into their remediation advice which translates into more targeted maintenance.A solution to processing high-resolution data collected from pipeline inspections was presented by the geo-data specialist company Fugro using in-house Remote Observation, Automatic Modelling, Economic Simulations (Roames) technology [71].The resultant product is a web-based service that enables pipeline inspection data to be uploaded to the secure cloud environment in near real-time, processed using ML, verified by experts onshore and visualized in an intuitive 4D web viewer.This approach can greatly reduce the cost of infrastructure management practices, lower the risk exposure, and contribute to the extended life of the pipeline.
The project was executed which employed a Kongsberg Hugin autonomous underwater vehicle (AUV) operated at an average of 3 m altitude and 4.5 kt [72].The point cloud was acquired with a Reson 7125 multi-beam echosounder (MBES).The pipeline position was automatically detected using a convolution neural network (CNN), which was trained to detect pipeline profiles from a cross-section of the point cloud.Ground truth data was generated from historical surveys, where the pipeline position was placed manually by a qualified data processor.The ground truth contains a pipe diameter marker, representing the position.Using the pipe and terrain model, the free-span and burial events can then be computed automatically.The web-based Roames Pipe Inspection tool can be used to inspect and adjust machine-learning-based point classification and pipeline positioning.The neural network-based pipe modelling positions the pipe within the correct location relative to the pipe points.The results of the neural network quantitative analysis showed a 97.2% accuracy in detecting pipe burials.The observed mean error in pipeline position was 0.46 cm.
While these new developments provide the operators with the desired deliverable, the large volume of information presents new challenges for inspection contractors and their onboard data processors working in a remote environment.Ongoing technological advancements in satellite communications present new opportunities for data to be transferred off the vessel in near real-time.This allows data processing to be performed securely in the cloud environment and validated by a global team of experts in offices around the world.
It is noted that the construction of the digital twins shall be started at the design phase of the offshore structure.The digital twin model must be updated with as-built and as-installed conditions and could be updatable in a fast way for accidental damages, permitting a quick evaluation of the asset safety and providing information.

Maintenance Planning Based on the Digital Twin
In recent years, the digital twins have been implemented in different industrial sectors, and in the design, production, manufacturing, and maintenance of the subsea assets.Among these application areas, maintenance is one of the most researched applications, as the execution of maintenance tasks may have a great influence on the business.
Different maintenance strategies might be used in the decision making, namely reactive maintenance, preventive maintenance, condition-based maintenance, predictive maintenance, and prescriptive maintenance [73][74][75][76][77][78][79][80][81][82][83][84].Predictive maintenance combines conditioning monitoring integrating with an ML-based decision support system and can enhance economic efficiency and availability [4].It has the capability to determine when to perform the maintenance based on the real conditions of the subsea pipelines.Once in place, these capabilities could reduce the unplanned maintenance downtime events and thus optimize the OPEX.
Figure 7 shows a digital twin system for predictive maintenance.One of the components is a computational model of the asset which is normally a finite element (FE) model.The computational model of the subsea pipeline is updated based on different field sensor data such as the motion sensor/accelerometer, the subsea strain gauge, the acoustic doppler current profiler (ADCP), the wave radar, the subsea pressure sensor, and the subsea temperature sensor (see Figure 8).Both the data-driven and FE-based models can be used to predict the remaining fatigue life (RFL) of the pipelines based on the stress range data, and are further used in the decision-making process.Knowledge of the RFL can enable efficient maintenance planning and avoid unpredicted shutdowns.Another component of the digital twin is the IoT/sensor system installed in the physical asset.IoT brings together low-cost sensors, cloud computing, and BDA in subsea pipeline systems where robustness, reliability, and security are highly desired.The third component is the data analytics to find the insight between the measured sensor data and apply a machine-learning algorithm to find the RFL based on the measured strain gauge data.The hydrodynamic load can be measured in real-time using the field sensor system and fed into the digital model.Wireless data loggers are connected to the IoT-based systems which transmit the data to store in the cloud for online analytics, visualization, and reporting.The data stored in the cloud is accessible onshore or onboard for data analytics.
Figure 9 illustrates the predictive maintenance model based on sensor data analytics and ML algorithms.The predictive maintenance schedule is estimated by using a system of artificial neural network (ANN).The memory blocks of a long short-term memory (LSTM) are used for the layers of a recurrent neural network (RNN).Another component of the digital twin is the IoT/sensor system installed in the physical asset.IoT brings together low-cost sensors, cloud computing, and BDA in subsea pipeline systems where robustness, reliability, and security are highly desired.The third component is the data analytics to find the insight between the measured sensor data and apply a machine-learning algorithm to find the RFL based on the measured strain gauge data.The hydrodynamic load can be measured in real-time using the field sensor system and fed into the digital model.Wireless data loggers are connected to the IoT-based systems which transmit the data to store in the cloud for online analytics, visualization, and reporting.The data stored in the cloud is accessible onshore or onboard for data analytics.
Figure 9 illustrates the predictive maintenance model based on sensor data analytics and ML algorithms.The predictive maintenance schedule is estimated by using a system of artificial neural network (ANN).The memory blocks of a long short-term memory (LSTM) are used for the layers of a recurrent neural network (RNN).Another component of the digital twin is the IoT/sensor system installed in the physical asset.IoT brings together low-cost sensors, cloud computing, and BDA in subsea pipeline systems where robustness, reliability, and security are highly desired.The third component is the data analytics to find the insight between the measured sensor data and apply a machine-learning algorithm to find the RFL based on the measured strain gauge data.The hydrodynamic load can be measured in real-time using the field sensor system and fed into the digital model.Wireless data loggers are connected to the IoT-based systems which transmit the data to store in the cloud for online analytics, visualization, and reporting.The data stored in the cloud is accessible onshore or onboard for data analytics.
Figure 9 illustrates the predictive maintenance model based on sensor data analytics and ML algorithms.The predictive maintenance schedule is estimated by using a system of artificial neural network (ANN).The memory blocks of a long short-term memory (LSTM) are used for the layers of a recurrent neural network (RNN).

Fast Assessment of Failures and Inspection Planning
In the technically demanding submarine environment, subsea pipelines are susceptible to various damage threats, which may lead to catastrophic failures of the disastrous economic and environmental variety.For instance, the waves and currents can lead to a scouring of the soil underneath the pipelines and free span problems where a segment of the pipeline becomes unsupported except at the two ends of the free span length.Free span lengths above the allowable design limit can result in fatigue damage through vortex-induced vibrations (VIV) [65, [85][86][87].Corrosion and erosion can occur from chemical attacks and abrasion due to the internal fluids containing abrasive sand particles travelling at high velocities.Corrosions, when combined with tensile stresses, can lead to stress cracking and leakage in subsea pipelines [46][47][48][49][50][51][52]65,88].
To detect and correct malfunctions of the physical assent, the digital twin applies ML, DL, and AI algorithms.It was revealed in a survey that asset monitoring and maintenance is the most anticipated application area for the digital twin [5].The fitness-for-service (FFS) of the physical asset is continuously monitored by the digital twin to identify potential failures.The big-data analytics capabilities of the digital twin can monitor the asset and send warnings to the responsible parties.
Digital twins provide high-fidelity accurate models and keep updating through the lifecycle of the pipelines with gathered data from sensors and inspections.Thus, they can reproduce the current state of the pipelines in the virtual space.Comparisons between digital twin simulated data and collected data can help determine the failure mode.One advantage of the state monitoring by the digital twin is that users can monitor the product state from any remote location through the unique identifier incorporated by the digital twin [31].
A computer vision-based digital twin model for real-time corrosion inspection was proposed in [89].The CNN algorithm was used for the automated corrosion identification and classification from the remotely operated vehicles (ROV) images and ILI data.Based on the corrosion assessment by the digital twin, predictive and prescriptive maintenance strategies are recommended.
During the service life of subsea pipelines, failure mechanisms such as external/internal sheath damage, fatigue damage, or corrosion may arise.Therefore, high OPEX is consumed to confirm the fitness of the system [90].

Fast Assessment of Failures and Inspection Planning
In the technically demanding submarine environment, subsea pipelines are susceptible to various damage threats, which may lead to catastrophic failures of the disastrous economic and environmental variety.For instance, the waves and currents can lead to a scouring of the soil underneath the pipelines and free span problems where a segment of the pipeline becomes unsupported except at the two ends of the free span length.Free span lengths above the allowable design limit can result in fatigue damage through vortexinduced vibrations (VIV) [65, [85][86][87].Corrosion and erosion can occur from chemical attacks and abrasion due to the internal fluids containing abrasive sand particles travelling at high velocities.Corrosions, when combined with tensile stresses, can lead to stress cracking and leakage in subsea pipelines [46][47][48][49][50][51][52]65,88].
To detect and correct malfunctions of the physical assent, the digital twin applies ML, DL, and AI algorithms.It was revealed in a survey that asset monitoring and maintenance is the most anticipated application area for the digital twin [5].The fitness-for-service (FFS) of the physical asset is continuously monitored by the digital twin to identify potential failures.The big-data analytics capabilities of the digital twin can monitor the asset and send warnings to the responsible parties.
Digital twins provide high-fidelity accurate models and keep updating through the lifecycle of the pipelines with gathered data from sensors and inspections.Thus, they can reproduce the current state of the pipelines in the virtual space.Comparisons between digital twin simulated data and collected data can help determine the failure mode.One advantage of the state monitoring by the digital twin is that users can monitor the product state from any remote location through the unique identifier incorporated by the digital twin [31].
A computer vision-based digital twin model for real-time corrosion inspection was proposed in [89].The CNN algorithm was used for the automated corrosion identification and classification from the remotely operated vehicles (ROV) images and ILI data.Based on the corrosion assessment by the digital twin, predictive and prescriptive maintenance strategies are recommended.
During the service life of subsea pipelines, failure mechanisms such as external/internal sheath damage, fatigue damage, or corrosion may arise.Therefore, high OPEX is consumed to confirm the fitness of the system [90].
Based on the reliability assessments [91][92][93], the risk-based inspection planning methodology has been used for the integrity management of subsea pipelines.The steps of the methodology include (1) data gathering, (2) development of risk criteria, (3) probability of failure (PoF) and consequence of failure (CoF), and (4) risk evaluation and enhanced inspection strategy [94][95][96][97][98][99].Bayesian network (BN) and genetic algorithm (GA) were used to develop a framework for the inspection decision-making for subsea pipelines [100].Using digital twins, risk target data analysis and risk estimation by prognostic and ML techniques were performed [101,102].

Life Extension Assessments
With the increasing maturity of the oil and gas industry, the requirement to operate a subsea pipeline beyond the design life is becoming commonplace [103].The life extension of the subsea pipelines opens up many development opportunities.If the pipelines can be reused for future developments, significant capital expenditure reduction can be achieved [104,105].
The process of a life extension assessment considering consolidated guidelines [106-109] is summarized as follows: (1) Definition of the operational context and premises for an extended operational period, and identification of new threats.(2) Assessment of current condition, functionality and integrity of the system (Diagnostic).The basic premise of the life-extension process is similar regardless of which guideline is applied.Figure 10 displays a typical life extension process.The detailed descriptions of the key steps in the process can be found in [104].
A digital twin can be of great value to support ALCM and hence support optimal management of the asset lifetime [110].The life extension assessment process clearly maps into the classification and implementation levels of the digital twin.
A digital twin supported by the system of systems concepts was proposed in [110] to represent an aid for model-based condition assessment, and the estimation of RUL contributes with prescriptive capabilities to the identification of measures related to conditionbased or predictive maintenance policies, optimal operation, and control to extend equipment lifetime.In this regard, the life cycle losses and costs can be reduced.Following the standards and guidelines for life extension evaluations, a risk picture can be automated and integrated within levels 2 and 3 of the digital twin to guarantee the visualization of current and future risks.This can then serve as inputs into level 4 to identify the risk-based extension measures and to visualize the mitigated risk picture.
Another digital twin concept was proposed to provide an accurate estimation of the true fatigue life of assets to unlock potential fatigue life and ultimately extend the life of assets in the oil and gas industry [111].The digital twin concept was divided into four tiers that allow for unlocking the RFL of the subsea asset: (1) High-resolution modelling of the asset (RB-FEA).
(2) Update the model to reflect real-world conditions (Digital Twin).Fluid flow assurance issues including hydrates, wax deposition, asphaltenes and naphthenates, slugging, scales, corrosion, and erosion, bring a critical operational challenge to the production and transportation of pipelines [114].Computational fluid dynamics (CFD) can be applied, combined with different models (e.g., CSMHyK rheological model, Eulerian-Eulerian CFD-model, population balance model), to provide accurate analysis in terms of flow assurance [115].The models can be integrated into commercial CFD packages such as FLUENT, STAR-CD, TransAT, and STAR-CCM+ for such analyses.DNV developed hydraulic modelling software Synergi Gas for the optimization and sim- The reduced basis finite element analysis (RB-FEA) technology by Akselos is faster than conventional FEA in which higher accuracy is ensured by using posterior accuracy indicators and automated model enrichment [112].Among the existing case studies [111,113], the Akselos digital twin is used for the life extension of subsea assets.
Fluid flow assurance issues including hydrates, wax deposition, asphaltenes and naphthenates, slugging, scales, corrosion, and erosion, bring a critical operational challenge to the production and transportation of pipelines [114].Computational fluid dynamics (CFD) can be applied, combined with different models (e.g., CSMHyK rheological model, Eulerian-Eulerian CFD-model, population balance model), to provide accurate analysis in terms of flow assurance [115].The models can be integrated into commercial CFD packages such as FLUENT, STAR-CD, TransAT, and STAR-CCM+ for such analyses.DNV developed hydraulic modelling software Synergi Gas for the optimization and simulation of gas distribution and transmission networks.In the further development of digital twins in subsea pipelines, flow assurance and fluid composition tracking need to be taken into consideration.
Table 1 lists the reviewed papers on the application of digital twins on different areas of subsea pipelines.It reveals that the maintenance and manufacturing of the pipelines are two top-ranked applications of digital twins.It can also be seen that half of the publications are journal papers.Given that the investigations and discussions on digital twins were mainly published in international conferences in the early 2010s, as reported in [5,34], more researchers intend to publish their research in journals in the coming years.social impact, it was revealed that digital twin technologies can result in the redistribution of the workplace without much impact on employment [112].
The influence of digital twins also relies on the quantitative changes in new technologies.The following steps are suggested to make the research and development of digital twins more coherent: unify data and model standards; create a public database for sharing data and models; develop products and services to help digital twins become easier to build and use; develop universal platforms and tools for digital twin applications; and establish forums for practitioners and researchers.

Figure 1 .
Figure 1.The number of published results about digital twins in the Web of Science.

Figure 1 .
Figure 1.The number of published results about digital twins in the Web of Science.

Figure 1 .
Figure 1.The number of published results about digital twins in the Web of Science.

Figure 3 .
Figure 3.The tools for services applications of digital twins.Reproduced from [36], with permission from Elsevier, 2022.

Figure 3 .
Figure 3.The tools for services applications of digital twins.Reproduced from [36], with permission from Elsevier, 2022.

Figure 6 .
Figure 6.The capability levels of the digital twins at different stages of the evolution [62].
J. Mar.Sci.Eng.2022, 10, x FOR PEER REVIEW 6 of 19illustrates the main technologies and enablers in digitalization for SIM: digital worker + support, inspection + data collection, and the analysis.

Figure 6 .
Figure 6.The capability levels of the digital twins at different stages of the evolution [62].The digitalization of the data exchanges of the subsea pipelines plays an important role in avoiding the errors of the data during manual transcriptions.In order to simplify the data exchange between parties at the different stages of projects, the pipeline data exchange format (PDEF) collaborative initiative was proposed.It is an open-source joint

Figure 6 .
Figure 6.The capability levels of the digital twins at different stages of the evolution [62].

Figure 8 .
Figure 8.A field sensor system and the data-driven model [4].

Figure 8 .
Figure 8.A field sensor system and the data-driven model [4].

Figure 8 .
Figure 8.A field sensor system and the data-driven model [4].

Figure 9 .
Figure 9.A predictive maintenance model based on sensor data analytics [4].

Figure 9 .
Figure 9.A predictive maintenance model based on sensor data analytics [4].

( 3 )
Fatigue calculations based on continuous monitoring.(4) Statistical correlation between sea states and fatigue damage.Retrospective fatigue calculation.

Table 1 .
A categorical review of the applications of digital twins on subsea pipelines.