Long slugs arriving in separators/slug catchers is a major flow assurance concern in the offshore oil production industry, potentially causing flooding and/or severe separation problems. The sizing of the receiving facilities is determined by the longest slugs, so the economic implications of slug length predictions can be substantial. Slugs may also over time cause serious fatigue issues in free-span pipe sections, as large load variations can drastically reduce the lifetime of the flange connections. In most laboratory experiments reported in the literature, slugs rarely become longer than around 30-40 pipe diameters, while in many oil production fields, slugs can be considerably longer. Consequently, there is a clear need to better understand how and why such long slugs appear in production systems, and in this paper we present results that shed some light on this matter. We present a unique set of two- and three-phase slug flow experiments conducted in a 766 meter long 8" pipe at 45 bara pressure. The first half of the pipe was horizontal, while the second half was inclined by 0.5 degrees. A total of ten narrow-beam gamma densitometers were mounted on the pipe to study flow evolution, and in particular slug length development. In addition, the average phase fractions were measured using two traversing gamma densitometers, and one 160 meter long section with shut-in valves. The pressure drop was also measured along the loop using a total of twelve pressure transmitters. The results show that the mean slug length initially increases with the distance from the inlet, but this increase slows down and the mean slug length typically reaches a value between 20 and 50 diameters at the outlet. At low flow rates, the slug length distributions tend to be extremely wide, sometimes with standard deviations approaching 100%. The longest slugs that we observed were over 250 pipe diameters (50 meters). At higher flow rates, the slug length distributions are generally narrower. The effect of the water cut on the slug length distribution is significant, but complex, and it is difficult to establish any general trends regarding this relationship. Finally, it was observed that slug flow often requires a very long distance to develop. Specifically, in most of the slug flow experiments, the flow regime 50 meters downstream of the inlet was not slug flow. The reported experiments are the first three-phase slug flow experiments ever conducted in a large-scale setup. By using a long, heavily instrumented pipe, we were able to study the evolution of slug length distributions over a long distance. We believe that these experiments can be of considerable value for developing tools for predicting slug lengths in multiphase transport systems, which is a critical matter for oil field operators.

A new machine learning (ML)/statistical-based methodology for conditioning and predicting production data for a well pad has been developed. Typically, data conditioning involves outlier detection, missing data, data imputation, and smoothing. Time-series production data prediction can be challenging because the target (wellbore oil production) depends on large-scale, high-dimensional data sets with unknown distributions and is influenced by missing data and outliers. Hence, data conditioning is key for accurate predictions. The current work is the first attempt at using ensemble ML and statistical techniques, such as multilayer perceptron (MLP), principal component analysis (PCA), and support vector regression (SVR), for well pad data conditioning using recently disclosed subsurface and production data from a field in the southern area of the Norwegian North Sea. The time-series forecasting based on large-scale, high-dimensional conditioned and cleaned data sets is also presented. The data with an oil production rate greater than 10 Sm 3 have been retained for data cleansing, which reduced the size of the production well data set by 14.9%. Outliers are detected using the z-score method. The missing values are predicted using a trained ML model on all available nonmissing data. The procedure first predicts the downhole missing values from all the wells, including the available neighboring wells, and then uses these features to predict other missing values for the well pad. In this paper, the two approaches implemented and compared for prediction of missing data are MLP and SVR, and PCA is performed to extract the most important data features. Production data with 12 related variables (i.e., dates, hours, temperature, pressure, etc.) are used to explore the complex nonlinearity of features and estimate wellbore oil production with ML and deep-learning models. Conventional SVR and MLP methods are implemented as the benchmark. During this work, more than 60% of the missing and abnormal data from the field data set are detected and imputed using advanced ML methods, such as MLP and SVR with radial basis function kernel. More than 6% of data are outliers and are removed using the z-score method. The modified SVR with time-series data structure and long short-term memory (LSTM) algorithms are used for the comparisons. An R-squared (R 2 ) of 98% is achieved for both the algorithms; however, LSTM has the lowest root mean square error (RMSE) results compared to SVR. Data conditioning is conventionally performed using statistical techniques, but here, an ensemble of ML techniques is used depending on the available data. This paper presents a new methodology to perform data conditioning and production prediction for a well pad using ML and neighboring well data. The ML algorithms used are highly efficient, as demonstrated by the results.

Understanding the vertical discrete electrofacies distributions in wells is a vital step to preserve the reservoir heterogeneity. Predicting the electrofacies distribution at all wells is commonly conducted manually or with the use of some graphing approaches, but recently different machine learning techniques have been adopted to categorize electrofacies. In this paper, two supervised machine-learning techniques were implemented to model electrofacies given well logging data for a well in order to predict the distributions in all other wells (classification) in a carbonate reservoir in a giant southern Iraqi Oil Field. The available data included open-hole and CPI well logging records in addition to the routine core analysis. The well discrete electrofacies distribution for the entire reservoir thickness has been obtained in our paper [OTC-29269-MS] using the Ward Hierarchical Clustering Analysis. For electrofacies classification, two supervised machine-learning techniques, K-Nearest Neighbors (KNN) and Random Forests (RF), were adopted to model the resulting electrofacies given the CPI well logging data for a well to predict at other wells that have missing data. These two supervised learning techniques were implemented as non-linear and non-parametric classifiers, which are imperative attribute due to the non-linearity of the electrofacies properties and the geological reservoir control. The results of this research illustrated that the reservoir electrofacies can be predicted through the use of the supervised learning techniques when well logging records and core data are available. The two adopted classification algorithms were analyzed and compared based on confusion table, transition probability matrix and total percent correct (TCP) of the identified electrofacies that reveal the accuracy of the classification. RF was observed to be the optimum approach as it led to better electrofacies classification in this carbonate reservoir than the KNN. The application of supervised machine learning techniques enhanced the accuracy and reduced the time spent in electrofacies classification. The two machine learning algorithms were implemented by R software, the most powerful statistical programming language.

Integration of renewable power sources into grids remains an active research and development area, particularly for less developed renewable energy technologies such as wave energy converters (WECs). WECs are projected to have strong early market penetration for remote communities, which serve as natural microgrids. Hence, accurate wave predictions to manage the interactions of a WEC array with microgrids is especially important. Recently developed, low-cost wave measurement buoys allow for operational assimilation of wave data at remote locations where real-time data have previously been unavailable. This work includes the development and assessment of a wave modeling framework with real-time data assimilation capabilities for WEC power prediction. The availability of real-time wave spectral components from low-cost wave measurement buoys allows for operational data assimilation with the Ensemble Kalman filter technique, whereby measured wave conditions within the numerical wave forecast model domain are assimilated onto the combined set of internal and boundary grid points while taking into account model and observation error covariances. The updated model state and boundary conditions allow for more accurate wave characteristic predictions at the locations of interest. Initial deployment data indicated that measured wave data from one buoy that were assimilated into the wave modeling framework resulted in improved forecast skill for a case where a traditional numerical forecast model (e.g., Simulating WAves Nearshore; SWAN) did not well represent the measured conditions. On average, the wave power forecast error was reduced from 73% to 43% using the data assimilation modeling with real-time wave observations.

Mooring line tension monitoring is required for permanently moored floating offshore platforms by some regional regulators and classification societies. This requirement is typically satisfied by installing physical sensors that directly measure the line tension. Experience shows these sensors have relatively short life compared to the platform operational life and consequently they need to be changed several times thereby increasing the operational expenses. It is also possible that changing the sensors in the field may not be feasible due to access and safety issues or it may be prohibitively expensive, which could lead to the platform operating without meeting the regulations. This paper presents a machine learning based model, which we call ‘virtual sensor’, for predicting the mooring line tensions based on the platform’s heading, horizontal position and six-degrees-of-freedom (6-dof) rigid body motions. The model’s development and testing are demonstrated with the help of data generated through numerical simulations of a permanently moored semi-submersible. When deployed in field, the inputs to the virtual sensor would be obtained from the global position system (GPS) and accelerometers. Both the GPS and accelerometer are cheaper to install and maintain, reliable and easy to replace. The neural network model is pre-trained using a dataset of 5000 static simulations and further fine-tuned with 48 dynamic simulation cases. Model performance on four mooring lines are presented in the study. The accuracy of the model was assessed by determining the percentage of predictions with errors within ±5% of the simulated mooring line tensions. Three of the mooring lines achieved accuracy greater than 90% and one mooring line achieved 77% accuracy. The relevant limitations of the study and future work are discussed in the paper.

Model-based prediction of vessel response is valuable for planning and execution of marine operations. Response-based operation criteria are expected to give less downtime and cheaper and safer operations than criteria based on wave height and wave period. At least this holds if the response model has sufficient accuracy. The accuracy can in principle be improved by tuning the model using measured inputs and outputs. It is envisioned that an advisory system for planning and execution of marine operations will contain a module for continuous model improvement based on on-site measurements of excitation and response. A premise for this is that the measurements be of sufficiently high quality. To test the potential of automatic model tuning an established numerical vessel model is subjected to tuning with high-precision data from a model test with wave disturbance only. This may give information of how well modelling can be achieved using tuning under favourable conditions and serve as a benchmark for tuning under noisy conditions. In addition the results may give suggestions for improvement of the mathematical model. A prototype tuning software is written in Matlab. The tuning principle is to minimize the output error, i.e. the difference between measured and simulated response, by adjusting the model's parameters. For the minimization a quasi-Newton method is used. The tuning software is tested with data from the model test and found to function as intended. Examples of tuning are given.

Uncontrollable kicks are the most costly events that occur while drilling for oil and gas. Formation water flow sometimes turns to kicks that lead to life threatening and environmentaly disastrous blowouts. Prediction of the possible abrupt pressure surges that characterize the subsurface geological setting before drilling sheds light on some of the challenges that may be encountered along the bore-hole trajectory. This will also help curtail human error during penetration of certain zones along the well-hole trajectory, and consequently reach the objective depth safely and with less nonproductive time (NPT). Before drilling, pore andfracture pressure predictions from seismic velocity are critical for assessing the economic feasibility and safety for the whole prospective trap. Integrating the seismic velocity drifts and the sequence stratigraphy semblance at the proposed location can point to the possible pressure transgressive intervals that can cause a sudden pressure surge (PS). Moreover, modifying the drilling tolerance window (DTW) to accommodate the expected hydrocarbon column in the prospective reservoirs reduces the potential of unexpected hard kicks at the shale – sand interface. This paper briefly discusses the impact of subsurface geopressure compartmentalization on seismic velocity drift and consequently on PS. It also examines the subsurface geological setting that can cause substantial pressure increase penetrating the lithological interfaces. Therefore, the pressure transgression and expected excess pressure in pay zones should be encompassed within the numerical algorithm of the predictive model before drilling. Monitoring the logging while drilling (LWD) data slopes in shale beds can successfully point to a possible kick ahead of the drill bit. Maneuvering the mud weight and casing program while drilling within the DTW based on the modified numerical pressure profile can achieve safe drilling. Examples from onshore and deep-water wells are shown. This paper covers several geological features that correlate to open bore-hole flows or kicks that sometimes develop to a blowout if the formation flow is not controlled by the right mud weight kill. Detecting these subsurface features and their associated seismic velocities before drilling can lead to safe drilling and avoiding NPT. Moreover, this paper sheds light on the potential to enhance drilling safety in advance, even in cases where managed pressure drilling (MPD) equipment is used.

The prediction of equipment failures typically falls along two paths: statistical analysis of past failures, or using sensor data to monitor the condition of equipment in the field. This study instead takes a novel approach by implementing machine learning on the data within maintenance systems (which are typically used to monitor and track maintenance activities/workflows). This implementation serves to map the correlations between maintenance data and failures, to predict failures by only monitoring an organization's maintenance system. This involved identifying maintenance system variables or attributes that normally represent specific steps in the maintenance process, but could potentially provide indirect information on the reliability of assets or the risk of future failure (e.g. work orders created/completed, work order types, original equipment manufacturer recalls/upgrades, inventory rotation rates, procedure changes, turnaround events, etc.). Machine learning is then used to determine correlations between failures and system variables (which results in an optimized model). The resulting model can then be implemented on current data to predict future failures. Although largely designed around workflow management, the underlying metadata from an organization's maintenance system can indeed be transformed to provide valuable insights and quantify risk. Machine Learning is able to predict failures with a meaningful degree of accuracy. Even if the transparency of model logic is insufficient to take specific actions to mitigate failures, it is enough to dynamically assign a level of risk to individual assets, and even predict risk for future time periods. When interpreting the optimized machine learning model with methods such as feature impact charts, it is also possible to identify the strongest correlations between failures and different Computerized Maintenance Management System (CMMS) derived variables (variables that were designed and transformed from actual historical data for the purpose of model building). Insights gained from this additional analysis, such as the true impact that preventive maintenance delays have on risk, offer great potential for understanding the behavior of equipment and maintenance organizations.

Computational modeling (CFD) has made great strides in the last 20 years in fighting and overcoming its "expensive" stigma. The oil & gas industry has embraced the use of CFD to assess everything from the design of equipment, optimization of process flows, and hazards modeling of fire and explosion events all in an effort to strive for the goal of inherently safer designs. In current offshore facility design, CFD modeling, to assess the consequences of vapor cloud explosions has been established as common practice. For most major projects, CFD explosion modeling is used heavily in Feed and Detailed (later) stages of design. At these stages hundreds of scenarios are simulated and combined with extensive probabilistic and statistical calculations to define design loads of critical systems and assess risk. The primary advantage of using CFD is that it incorporates more details of the design and provides more realistic and design specific analysis. In the case of explosions, CFD provides valuable data for design, particularly explosion data that is significantly more accurate than that from simple open field models. Despite this adoption within the design process and decades of advances, CFD modeling practitioners still admit that there are increases of cost and time, when compared to the ever-present phenomenological approaches. This has kept many design processes from utilizing CFD in the earliest stages; namely Concept and pre-Feed. The arguments have traditionally been along the lines of: The design is progressing and changing too fast to allow time for CFD models to be built., There is insufficient model detail to construct viable CFD models., The early stage concerns can be captured "well enough" with simple approaches., etc. Modeling representations and approaches can be simplified and streamlined, while still providing the advantages of CFD. Variability in geometric representations can be captured with physics representations rather than needing to create predictive CAD geometry. These types of simplifications can increase the ability to simply change and update models and conduct early stage sensitivities in projects, while still capturing critical explosion effects. The purpose of this paper will be to show how these simplified approaches can be used in an effective and efficient manner in early stages of design. That the traditional arguments against early adoption of CFD can be countered, with the end result being greater early stage information and influence on the design. Allowing effective change at a time in the process when it is not cost prohibitive and other controls have yet to prohibit design changes. The result being a continued improvement in the goal of inherently safer design.

Novel methods were required to address the high operating temperatures of the Appomattox flowlines. Appomattox is predicted to operate as high as 365F, requiring thermal mitigation along the length of the flowline and anchoring at the base of the riser. Existing thermal mitigation methods were screened but were found unacceptable. Alternatively, a system of multiple buoyancy modules was found to be highly efficient and reliable to address the thermal expansion along the length of the flowline. Also, a unique rigid anchoring system was designed to restrain the riser base. Appomattox began production in Q2 of 2019. A flowline survey performed in Q3 2019 demonstrated good alignment with the analysis predictions for buckling locations, shapes and buckle magnitudes.

As vibratory driving is increasingly used to install offshore foundations, especially for offshore windfarms, an accurate soil model is key to reliable pile driving predictions and thus to improved project planning and execution. During the pile driving, the soil properties, and thus the soil parameters, do not remain constant. For vibratory driving, this effect is generally simplified by the application of the Beta Method, which was first proposed by Jonker in 1987[ 1 ]. While this simplified approach could be justified based on the computing capabilities at that time, a more sophisticated soil model should be used as part of pile driving simulations to ensure that the results are more accurate. In this paper, the state-of-the-art soil model for vibratory driving will be described and then illustrated through two brief case studies where the pile driving simulation results are compared with the actual installation records. The paper will show that while the predictions are not perfect, they forecast the pile driving process realistically. To further enhance prediction capabilities, the paper will suggest that pile driving with vibratory hammers is monitored and that the recorded data be used for post-processing to gain a more in-depth understanding of the soil behavior.

Integrating discrete facies classification into the estimation of formation permeability is a crucial step to improve reservoir characterization and to preserve heterogeneity quantification. Therefore, it is essential to obtain the most accurate estimation of permeability in non-cored intervals in order to attain realistic reservoir characterization and modeling. In our most recent paper [OTC-30906-MS], the electrofacies classifications have been conduced for a well from a carbonate reservoir in a Giant Southern Iraqi oil field. The same predicted discrete electrofacies distribution was included in this paper along with well logging interpretations to model and predict the reservoir core permeability for all wells. The well logging interpretations that were included in permeability modelling are neutron porosity, shale volume, and water saturation as a function of depth. The regression and machine learning approaches adopted for permeability modelling are multiple linear regression (MLR), smooth generalized additive Modeling (SGAM) and Random Forest (RF) Algorithm. The classified electrofacies were considered as a discrete independent variable in the core permeability modelling to provide different model fits given each electrofacies type in order to capture the different permeability variances. The matching visualization between the observed and predicted core permeability, the computed root mean square prediction error and adjusted squared R were considered as validation and accuracy tools to compare between the three modelling approaches. Since there are too many intervals with missing core permeability measurements, the modelling was first adopted on the intervals that have permeability readings (known subset). The prediction was then conducted given the same known permeability intervals in addition to the entire dataset (full dataset). The root mean square prediction error and adjusted squared R for the Random Forest were significantly better than in both MLR and SGAM for the known subset and full dataset. It can be concluded that combining the electrofacies in one permeability model has accurate, fast and an automation procedure of prediction for other wells. The two machine learning algorithms were implemented by R software, the most powerful statistical programming language.

Gas hydrate formation is considered one of the major problems facing the oil and gas industry as it poses a significant threat to the production, transportation and processing of natural gas. These solid structures can nucleate and agglomerate gradually so that a large cluster of hydrate is formed, which can clog flow lines, chokes, valves, and other production facilities. Thus, an accurate predictive model is necessary for designing natural gas production systems at safe operating conditions and mitigating the issues induced by the formation of hydrates. In this context, a thermodynamic model for gas hydrate equilibrium conditions and cage occupancies of N 2 + CH 4 and N 2 + CO 4 gas mixtures at different compositions is proposed. The van der Waals-Platteeuw thermodynamic theory coupled with the Peng-Robinson equation of state and Langmuir adsorption model are employed in the proposed model. The experimental measurements generated using a cryogenic sapphire cell for the pressure and temperature ranges of (5-25) MPa and (275.5-292.95) K, respectively, were used to evaluate the accuracy of this model. The resulting data show that increasing nitrogen mole percentage in the gas mixtures results in decreasing of equilibrium hydrate temperatures. The deviations between the experimental and predictions are discussed. Furthermore, the cage occupancies for the gas mixtures in hydrate have been evaluated. The results demonstrate an increase in the cage occupancy for both the small and large cavities with pressure.

Extending the geohazards assessment throughout the overburden section above the main reservoir is a relatively recent practice. Traditionally, a geohazards assessment for the riserless section of wells has been a key element of pre-drill studies in offshore well planning. Potential geohazards such as seafloor/buried faults, gumbos, gas hydrates, shallow gas and shallow water flow are routinely investigated using seismic reflectors, seismic amplitude, extent of well-known problematic stratigraphic units, and offset wells drilling data. As a result, the understanding of the regional distribution of geohazards has increased significantly. Although the common geohazards as mentioned above are well known, the assessment has been challenging with deeper depths. The loss of seismic resolution with depth affects the interpretation of the stratigraphy, structure, pore pressure and fracture gradient. The uncertainty in the depth and inclination of key marker events including stratigraphic horizons, chronostratigraphic ties, etc. might lead to a wide margin of possible pore pressure and fracture gradient (PPFG) estimates. The velocity-effective stress transformations used for pre-drill PPFG from low resolution seismic data is also less reliable, compounding the error resulting from predicted stratigraphic markers (horizons). Seismic interpretations with low resolution are inadequate to identify thin over-pressured zones. The paper presents an integrated workflow that maximizes the predictability of geohazards for the entire reservoir overburden section. A variety of seismic volumes including amplitude reflection, amplitude versus offset (AVO), seismic inversion, seismic velocity, coherence data, etc. allows for the optimization of interpretations such as stratigraphy, structure, and rock properties. A detailed geologic model with advanced seismic processing techniques provides a high-resolution understanding of structure and stratigraphy, seismic attribute distributions, and spatial velocity variations. The model is useful to identify key faults, leak points, sealing intervals, and trapping mechanisms. Understanding the stratigraphic facies assists in mapping the intervals of pressure generation and retention zones. Considering these limitations, offset well data is integrated when available and utilized to characterize seismic facies and rock properties in sparse data environments. These data are then correlated with seismic reflection and velocity data to develop a well-constrained geologic model. Multiple types of seismic volumes with various frequencies, coverages, and penetration provide better control and understanding of the riserless section. This may include AVO and inversion volumes, which are not commonly used in a shallow geohazards assessment. Finally, a fully integrated geohazards assessment, from the seafloor to the main reservoir, results in an optimized drilling program and is developed to minimize the impact of geohazards and drilling risks along the wellbore trajectory.

The performance of a transient multiphase flow simulator was evaluated using high-quality field data measured in a large diameter gas-condensate offshore production system during different operating conditions (pigging, production shut-in and restart, production ramp-up, quasi steady state). Production data included measurements recorded from more than 120 sensors (pressure, temperature, flow rate, monoethylene glycol (MEG) fraction, slug catcher volume) during a three-year production time span. After thorough data screening and analysis, 19 distinct subsets encompassing both steady state conditions and at least one pigging event per subset were selected as benchmarks for validating simulation results. For each subset, a simulation model was developed to account for ambient and operating conditions, fluid properties, bathymetry profiles, thermal insulation and pipe burial, and hydrate inhibitor tracking (MEG). A comprehensive statistical analysis comparing predicted and measured pressure drop, pigged liquid volumes, outlet temperatures, and rich MEG mass fractions over the three-year time period demonstrated that the simulator predictions were in good agreement with the field data. The analysis included an uncertainty assessment of the measured production flow rates and volumes to better estimate the simulator accuracy. It was found that the simulator predicted both the cumulated rich MEG and condensate volumes received at the delivery point within the measurement uncertainty. The predicted rich MEG fractions at delivery were also in good agreement with the measurements, underlining the reliability of the inhibitor tracking module of the simulator. The pig travel time and the total liquid volume displaced by the pig during each pigging event were the main parameters considered to evaluate the accuracy of the simulator; both were predicted within a 10% error margin. Multiphase flow simulators are often developed and tuned against experimental data recorded in small to medium diameter scale pilots. The opportunity to validate a transient multiphase flow simulator against measured operating conditions in large diameter pipelines over an unprecedented three-year time span is valuable, not only for quantifying the performance of the simulator, but also for assessing its scalability.

Lost circulation is one of the frequent challenges encountered in the well drilling and completion process, which can not only increase well construction time and operational cost but also pose great risk to the formation. However, choosing the most useful treatments may still be a problem due to the complexity of the drilling and geological condition. In this paper, machine-learning algorithms and big data technology are employed to mine and analyze drilling data of wells in South China Sea where lost circulation is severe. Geological characteristics, drilling fluids property parameters and operational drilling parameters are both considered. Moreover, an artificial neural network is employed to conduct supervised learning. The four metrics: accuracy, precision, f 1 score and recall are used to evaluate the model. The trained artificial neural network model is employed to predict the lost circulation risks. To train and test the proposed model, drilling operation parameters, geological parameters and drilling property parameters are collected for lost circulation events for 50 drilled wells over past two years in South China Sea. The trained model is excellent with the most important evaluation metrics, attaining an accuracy up to 92%, with f 1 score, recall and precision up to 89% similarly. This suggests that the model have a good generalization ability and can be applied to other fields. Data analysis through an artificial neural network is carried out to develop a lost circulation prediction system model. This methodology can predict six lost circulation risks, each is defined according to drilling mud loss rate. This is one of the first attempts to predict lost circulation using data-analytics and artificial intelligence. The proposed intelligent lost circulation prediction method can assist the drilling engineer to choose the optimal drilling parameters prior to drilling and avoid lost circulation events.

This paper describes a live fatigue prediction methodology comprising measured motion response, maritime environment and process data for a Floating Production Storage and Offloading vessel (FPSO) moored in 700m water depth offshore Brazil. The measured data is utilized to improve traditional time domain dynamic analysis models, along with Machine Learning (ML) techniques. The resul of this is significant reduction in uncertainties, enabling live riser fatigue predictions and providing a basis for life extension and improved accuracy of riser and vessel response analysis. The methodology consists of using a combination of autonomous and online motion response sensors directly installed on the riser and interfacing FPSO structures. The measured environmental data, FPSO and riser response data are utilized in a ML environment to build more realistic riser response and fatigue prediction models. As FPSO heading is important for vessel dynamics, especially roll, and the vessel dynamics are a key factor in the riser dynamics at this field, the first focus was directed towards predicting vessel heading relative to swell. The heading model developed by ML showed good agreement and was used as a key tool in a traditional fatigue analysis using OrcaFlex & BFLEX. This analysis was based on historical sea states from the last two years (from EU's Copernicus Marine Environment Monitoring Service). The results show that the fatigue analysis from the design phase is conservative and life time extension is achievable. As the fully instrumented measurement campaign ended after 4 months, the work focused on utilizing all the captured data to give improved insight and develop both traditional simulation and ML-models. For future fatigue predictions based on the developed "fatigue counter", the ambition is to maintain good accuracy with less instrumentation. In the present phase, FPSO and riser response data from a 4-month campaign have been used to establish a ‘correlation’ between riser behavior, environmental data and FPSO heading and motion. Calibration of a traditional numerical model is performed using measurement data along with a direct ‘waves to fatigue’ prediction based on modern ML techniques. This illustrates enabling technologies based on combination of data streams from multiple data sources and superior data accessibility. The correlations established between different field data allow the development of a "live" riser fatigue model presenting results in online dashboards as an integrated part of the riser Integrity Management (IM) system. All relevant stakeholders are provided with necessary information to ensure safe and extended operation of critical elements of the FPSO. The paper illustrates the power and applicability of modern numerical techniques, made possible by combining data from 6 different streaming data sources, ranging from satellites to clamp-on motion sensors.

The first phase of the Chain FEARS (Finite Element Analysis of Residual Strength) Joint Industry Project (JIP) aimed to develop guidance for the determination of a rational discard criteria for mooring chains subject to severe pitting corrosion which, based on current code requirements, would otherwise require immediate removal and replacement. Critical to the ability to establish rational discard criteria, is the ability to accurately predict the residual strength of degraded chain, and to have as a benchmark for loss in strength, an accurate estimate of the chain in its as-new condition. With a correlated FEA method for residual strength prediction and a benchmark for as-new condition capacity, it would then be possible to establish a theoretical relationship between different types of degradation and mooring chain capacity loss, from which rational discard criteria would be derived. To this end the Chain FEARS JIP first developed a Finite Element Analyses (FEA) residual capacity assessment method to accurately predict the residual strength of degraded chains. A number of assessments were carried out to establish the sensitivity of the Predicted Break Load (PBL) to both engineering parameters such as friction coefficient, and numerical modelling techniques. The developed method was validated by the correlation of the PBL against a number of physical break tests. This paper presents a review of the break strength test data of pitting corrosion degraded chain links. The FEA modelling methodology based Predicted Break Load (PBL) are compared with the test data Actual Break Load (ABL) along with the sensitivity of engineering parameters and numerical model modelling techniques on predictions. The developed FEA method accurately predicts the location of the ‘failure’ within the chain string and the ductile necking failure mode, determined to be the prevalent mode of failure for the chain links samples considered in this study. The degree of correlation between PBL and ABL confirms that accurate prediction of the effects of corrosion degradation consequent on uniform and large pitting corrosion can be accurately predicted by use of the Finite Element Method. The developed FEA method was also employed to establish a benchmark for the strength capacity of as-new condition links as presented in [ 1 ], the basis for assessing the relationship between corrosion degradation and residual chain link capacity [ 2 ] and a basis for a multi-axial fatigue assessment method to establish the fatigue capacity of as-new and degraded chain links [ 3 , 4 , 5 ].

Deepwater exploration and production presents some of the industry's most complex challenges, requiring huge capital investment and long-term commitment. Subsea fields face exceptional challenges during drilling, cementing, and operations. Any downtime means project delay and lost production. Consequently, it is important to have real-time data from high-fidelity measurement-while-drilling (MWD) and logging-while-drilling (LWD) tools during drilling operations. Mud pulse telemetry is a widely used method to transmit MWD and LWD data to the surface. Availability of this real-time data is important for the economic success of the drilling operation. Consequently, it is important to maintain operational telemetry systems that provide fast and reliable data rates for downhole drilling. However, telemetry system failure due to sand erosion can be costly and cause downtime and maintenance costs to the operators. Hence, there is a constant need for accurate prediction of the location and erosion rate of these downhole systems. In this study, we build on our earlier work, where we presented a comprehensive investigation on erosion of mud pulse telemetry tools consisting of numerical simulations and field data. In the current work, we take the numerical analysis further: we predict the location of the high erosion rates and model equipment topological changes due to sand erosion and its impact to fluid flow. The importance of this step is to capture the effect of changing geometry on the erosion rate, enabling us to estimate the reliability of the tool accurately. In addition, we investigate the structural integrity of the tool and the effect that sand erosion has on it. Structural integrity analysis of the eroded tool geometry is performed using finite element analysis (FEA). For model validation, simulation results will be compared with erosion patterns from field tests.

Accurate forecasts of coastal erosion are essential for the effective management (operation and protection) of critical infrastructure such as gas terminals and shallow-buried nearshore pipelines, preventing the costly losses of production associated with storm damage or exposure. Traditionally, these predictions were the preserve of computationally-expensive, morphodynamic simulations of the three-dimensional structure of the beach surface, however recent developments in reduced-complexity ‘equilibrium’ models have been shown to skilfully hindcast coastal change in cross-shore and long-shore transport dominated environments more accurately, over much longer time-scales. The simplicity and stability of these models – expressed as a function of the incident wave power and the relative equilibrium in dimensionless fall velocity – make them particularly appropriate for assessing the current ‘health’ of the coastline in actionable terms, while unlocking their potential use in forecast mode. Here, we present such a system, forced by data from the Met Office Wave Ensemble Prediction System, capable of providing real-time probabilistic forecasts of important coastal indices (e.g. beach volume and shoreline position) out to seven days ahead. The system is calibrated using an extended Kalman Filter and becomes more accurate over time as it assimilates more observational measurements. Once calibrated, tests on unseen data from the University of Plymouth coastal monitoring station at Perranporth, UK, during Winter 2017/18 confirm it can accurately predict the impact of an extreme storm sequence on coastal erosion and subsequent recovery. This promises the potential for a new coastal management tool, able to be applied to other vulnerable locations.