Ocean surface is complex and difficult to predict accurately due to its random nature. Ocean surface waves in strong wind conditions have been widely studied for last few decades. Almost half of world’s winds are below 7.5 m/s and the physics of such winds contains a lot of uncertainties. The simulation of ocean waves is largely dependent on the driving winds force accuracy and source term parameterizations. However, low winds are often ignored on the perception of their lesser effect on overall results of existing models. It is important to understand the relative strength/ weaknesses of wave forecast models under low wind conditions from scientific perspective which should lead to improved wave forecast and wave-ocean-weather coupling capabilities. There are many critical thresholds involved in the initial generation and growth of wind waves whereas current parameterizations of wave models are mostly based on moderate – high wind conditions. Wave model’s performance, although not very prominent, contains bias under low winds conditions and these thresholds need to be embedded in current physics of wave forecast models for more accurate simulations. In this study, WAVEWATCH III (v6.07) wave forecast model with observation based source terms parameterizations (ST6 package) is used to simulate waves on a global scale. The model’s output is analyzed with a globally calibrated and cross validated global dataset of 13 altimeters to analyze its performance under low wind conditions. A relative error of −1 to 6 is observed in global significant wave heights simulated by WAVEWATCH III model compared to altimeter’s measured wave heights for wind speeds less than 5ms −1 .

Drilling practice has been evolving parallel to the developments in the oil and gas industry. Current supply and demand for oil and gas dictate search for hydrocarbons either at much deeper and hard-to-reach fields, or at unconventional fields, both requiring extended reach wells, long horizontal sections, and 3D complex trajectories. Cuttings transport is one of the most challenging problems while drilling such wells, especially at mid-range inclinations. For many years, numerous studies have been conducted to address modeling of cuttings transport, estimation of the concentration of cuttings as well as pressure losses inside the wellbores, considering various drilling variables having influence on the process. However, such attempts, either mechanistic or empirical, have many limitations due to various simplifications and assumptions made during the development stage. Fluid thixotropy, temperature variations in the wellbore, uncertainty in pipe eccentricity as well as chaotic motion of cuttings due to pipe rotation, imperfections in the wellbore walls, variations in the size and shape of the cuttings, presence of tool joints on the drillstring, etc. causes the modeling of the problem extremely difficult. Due to the complexity of the process, the estimations are usually not very accurate, or not reliable. In this study, data-driven models are used to address the estimation of cuttings concentration and frictional loss estimation in a well during drilling operations, instead of using mechanistic or empirical methods. The selected models include Artificial Neural Networks, Random Forest, and AdaBoost. The training of the models is determined using the experimental data regarding cuttings transport tests collected in the last 40 years at The University of Tulsa – Drilling Research Projects, which includes a wide range of wellbore and pipe sizes, inclinations, ROPs, pipe rotation speeds, flow rates, fluid and cuttings properties. The evaluation of the models is conducted using Root Mean Square Error, R-Squared Values, and P-Value. As the inputs of the data-driven models, independent drilling variables are directly used. Also, as a second approach, dimensionless groups are developed based on these independent drilling variables, and these dimensionless groups are used as the inputs of the models. Moreover, performance of the data-driven model results are compared with the results of a conventional mechanistic model. It is observed that in many cases, data-driven models perform significantly better than the mechanistic model, which provides a very promising direction to consider for real time drilling optimization and automation. It is also concluded that using the independent drilling variables directly as the model inputs provided more accurate results when compared with dimensional groups are used as the model inputs.

While performing fatigue reliability analysis of the butt-welded joints it is vital to estimate the Stress Concentration Factor (SCF) at these joints. A common approach adopted by industry to estimate the SCF at weld toes is to perform Finite Element Analysis (FEA) of the welded joints for different pipe sizes, flanges, valves etc. The SCF are calculated for each size by separately when required and are very time consuming. Although FEA is known for its accurate SCF calculation, but due to its high computational expense and time-consumption, SCF evaluation for different parameters makes the aforementioned method quite laborious. As an alternative response surface models (RSM) may be used for accurate estimation of SCF. The two basic steps in constructing a RSM are training and testing. The first corresponds to fitting a model to the intelligently chosen training points, while the second step involves comparing the predictions of the RSM to the actual response. This paper examines the applicability of 12 different RSMs for estimating SCF. The training and testing data is generated using FEA in ANSYS. In order to compare the accuracy of the RSMs, three metrics, namely, Root Mean Square Error (RMSE), Maximum Absolute Error (AAE), and Explained Variance Score (EVS) are used. A case study illustrating the applicability of the proposed approach is also presented.

Digital twins have attracted significant attention across different domains for decades. In the maritime and the energy industries, digital twins have been mainly used for system condition monitoring, project visualization, crew training, real-time decision making/support, and predictive maintenance based on onsite measurement data from onboard sensors. Such a digital twin normally presumes the vessel’s operational condition by assistance from sensors and engineering judgement. However, a vessel’s operational condition and loading state may shift quite often due to the frequently changing operational scenarios, tasks, and environmental conditions. In addition, the true vessel state (e.g., inertia distribution) may deviate from the intended one according to planning due to possible engineering errors. Even though there are sensors helping to monitor vessel condition such as draft monitoring systems and ballast systems, several important vessel parameters are difficult to measure directly, e.g., moment of inertia, center of gravity, and nonlinear hydrodynamic damping. This paper proposes a framework for monitoring vessel condition and providing decision support based on quantitative risk assessment, through a vessel state observer which is able to self-tune the important but uncertain vessel parameters by utilizing the available prior knowledge, vessel measurements, and information about the associated sea states. The tuned vessel parameters improve the information about the real-time vessel condition and consequently assist to improve the prediction accuracy of vessel seakeeping performance in the near future for the emerging wave conditions. Furthermore, the tuned results and the response prediction can then be applied to a decision support system, quantitatively evaluating potential risk and providing suggestions. The framework consists of 5 modules, i.e., wave data acquisition and processing, vessel data acquisition and processing, vessel seakeeping model tuning, real-time vessel motion and critical structural response prediction, and risk awareness and avoidance. Details of each module are described in the paper. The proposed framework can also assist in the development of autonomous ships.

This paper investigates the error incurred by predicting the wave-induced heave forces on a submerged body using a mathematically simple flat face bow and stern rather than the typical curved ones. We examine both theoretically and experimentally how the heave force changes when flat end faces replace hemispheric end caps. We consider bodies with various length-to-diameter ratios to identify the influence that the ratio of the end cap length to the total body length plays on the heave forces. We also consider bodies with forward speed to identify if heave forces are affected by leading edge separation around the blunt front face which is not present for the hemispheric end cap. The theoretical predictions are from an existing potential flow analytical solution while the experimental results were collected in a towing tank with wavemaking capability. The theoretically predicted percent increase in the heave forces caused by the flat end face was confirmed by the experimental results. Finally, the theoretical predictions also showed that the percent increase in the heave force is independent of the forward speed of the body. This was also confirmed with the experimental results.

Reliable sea ice concentration (SIC) information assists the safe and energy-efficient ship navigation along the Northern Sea Route (NSR). In particular, the accurate SIC forecast is a top priority. This study proposes a statistical interpolation method to reduce the errors induced by the traditional interpolation method. An auto-regressive integrated moving average (AR/MA) model is developed based on reanalysis data. The AR/MA model can be used for short-term SIC forecasts along the NSR. Model validation has been conducted through a specially designed cross-validation. The route availability is estimated according to the SIC forecast. The results indicate that the specified NSR will be open for shipping from 2021 to 2024. The work also indicates the feasibility of the proposed statistical models to assist NSR shipping management.

Due to the quality and error of the data itself, historical automatic identification system (AIS) data was insufficient used to predict navigation risk at sea, but it adequately used to train decision-making neural networks. This paper presents a real AIS ship navigation environment with a rule-based and a neural-based decision processes with frame motion and training the decision network using a deep reinforcement learning algorithm. Rule-based decision-making has several applications in the field of adaptive systems, expert systems, and decision support systems, it also including general ship navigation which regulated by the convention on the international regulations for preventing collisions at sea (COLREGs). However, if someone intend to achieve full unmanned ship navigation without any remote control at the open sea, a rule-based decision-making system cannot be implemented alone. With the growing amount of data, complex sea environment, different collision scenarios, the agent-based decision has become an important role in transportation. For ships, combined rule-based and neural-based decision-making is the only option. It has become progressively challenging to satisfy autonomous decision-making development requirements. This study uses deep reinforcement learning to evaluate the performance of decision-making efficiency under different AIS data input shapes. The results show that the decision neural network trained with AIS data has good robustness and a high ability to achieve collision avoidance. Furthermore, using the same methodology, include instructive guidance for processing radar, camera, ENC, etc., respond to different risk perception tasks in different scenarios. It has important implications for fully unmanned navigation.

Monitoring the integrity of mooring lines on floating offshore platforms is one of the key factors in ensuring safe and productive offshore operations. Sensors, such as inclinometers, compressive load cells or strain sensors, can be used to monitor the inclination angles or tensions on mooring lines. An alternative method using only dry monitoring systems, such as DGPS (Differential Global Positioning System), Gyrocompass and/or IMU (Inertial Measurement / Motion Unit), can also be used to monitor the integrity of mooring lines. This method uses the measured motions and positions of a vessel without any information on the environmental conditions to detect mooring line failure. The detection of mooring line failure is based on detecting shifts in low-frequency periods and mean yaw angles as a function of vessel position, mass and added mass. The proposed method utilizes Artificial Neural Network (ANN) to recognize and classify patterns. The training of an ANN model requires examples/data associated with intact mooring lines and broken mooring line(s). Examples/data of broken mooring line(s) are practically available only from numerical simulations. Therefore, it is important to address these two key topics: (1) Is the real behavior of the floating offshore platform sufficiently aligned with numerical simulations? and (2) The effect of the accuracy of monitoring equipment on the performance of an ANN-based system. The first topic is reviewed briefly with its possible solution including some sensitivity tests, and this paper focuses on addressing the second topic. A system architecture is discussed in this paper along with the accuracy of the monitoring equipment. As an example, an ANN model has been trained to detect a broken mooring line of a spread-moored FPSO. This ANN model has been tested on its performance in dealing with a range of possible errors associated with the monitoring equipment. Furthermore, the tests have been carried out for a combination of variables that are not included in the ANN training, such as: vessel draft (mass), sea state conditions and directions. This paper presents the results of the tests for various variable sensitivities, which cover vessel positions, mean yaw angles and vessel drafts. These are essentially testing the tolerance of a trained ANN model against error or noise in the data. The results show that a trained ANN model can be error/noise tolerant.

There is an ever-increasing amount of data being recorded in oilfield operations. During drilling a well a large number of parameters is being monitored and saved, often reaching several hundreds. We are seemingly monitoring everything, from basic parameters such as Weight on Bit, Torque, and Rate of Penetration (ROP), to the exhaust temperature of engine no. 3. Unfortunately, the quality of collected data does not match the quantity. Critical sensors, such as gamma and inclination, are often lagging many meters behind the bit. Despite best efforts, sensors stop working, hard drives corrupt files, and data mud pulse telemetry uplinks fail. Methods of infilling data spanning many meters or minutes are necessary. We present a novel approach that enables reliable prediction of data lagging behind the bit through deep neural networks by merging trend-based prediction with traditional neural network approach. We were able to predict continuous inclination data in a curved section of a well with an average absolute error of only 0.4 degrees up to 20 meters from last known value.

Atmospheric and ocean datasets using numerical modeling allied to data assimilation are valuable tools for planning and maintaining marine activities, particularly due to their spatial coverage and resolution. However, even modern analyses and reanalyses present critical errors where cyclonic winds are often underestimated in some locations, leading to issues in wind-wave climate hindcasts and forecasts. This work aims to evaluate two of the newest datasets available regarding their ability to reproduce extratropical cyclone tracks associated with extreme waves in the Atlantic Ocean. The analysis is focused on the storm track position and cyclone intensity, two important features that control the wave climate. The datasets used are the ECMWF’s ERA5 and NCEP’s CFSv2. The cyclones are identified and compared between 2011 and 2018 using the relative vorticity at 850hPa and the intensity is measured through the 10-meters wind speed. The results show that the differences in track density exist mainly in coastal areas and they can be related to the coarser resolution of CFSv2 compared to ERA5. CFSv2 presents a higher number of cyclones associated with extreme wave events in the South and North Atlantic, corresponding to 87.1% and 89.2% of the cyclones identified, while ERA5 presents 66.6% and 59.5% respectively. The maximum intensity distribution shows that CFSv2 tends to produce more intense cyclones than ERA5, even for identical storms in the datasets. The distances of matching cyclones in the two datasets are around 0.5°.

The navigation in restricted waters imposes several challenges when compared to open sea navigation. Smaller dimensions, higher traffic density and the dynamics of obstacles such as sandbanks are examples of contributors to the difficulty. Due to these aspects, local experienced maritime pilots go onboard in order to support the ship’s crew with their skills and specific regional knowledge. Despite these efforts, several accidents still occur around the world. In order to contribute to a better understanding of the events composing accidental sequences, this paper presents a hybrid modelling specific for restricted waters. The main techniques used are the fault tree analysis and event tree analysis. The former provides a framework to investigate the causes, while the latter allows modelling the sequence of actions necessary to avoid an accident. The models are quantified using statistical data available in the literature and a prospective human performance model developed by the Technique for Early Consideration of Human Reliability (TECHR). The results include combined estimates of human error probabilities and technical failure probabilities, which can be used to inform the causation factor for a waterway risk analysis model. In other words, given that the ship encounters a potential accidental scenario while navigating, the proposed models allow computing the failure probability that of the evasive actions sequence. The novelty of this work resides on the possibility of explicitly considering dynamicity and recovery actions when computing the causation factor, what is not a typical feature of similar works. The results obtained were compared with several results available in the literature and have been shown to be compatible.

Specific engineering tools are used to design Floating Offshore Wind Turbines (FOWT). These so-called aero-hydro-servo-elastic solvers simulate the coupled behaviour of the turbine subjected to wind with the floater motion due to waves, including elasticity of the whole structure. The implemented hydrodynamic forces rely on a strong Oil&Gas background and include potential flow theory and empirical laws, such as Morison forces. The undergoing study aims at re-evaluating the validity range of such theories, when applied to FOWT. To do so, CFD simulations will be run to model wave propagation and interaction with a FOWT floater. Hydrodynamic forces will be extracted from the CFD simulations and compared to current hydrodynamic theories. A fixed cylinder in regular second order deep water waves (steepness of 0.9) is simulated and results are validated against experiments [1]. This basic first case implemented with Open-FOAM using waves2Foam library allows to master regular wave generation and interaction with a rather simple structure, running multiple simulations. Convergence (mesh refinement, time step) and parameterization (numerical schemes, turbulence models) studies are carried out to ensure controlled wave generation. An accurate Numerical Wave Tank (NWT) is finally obtained. However, the resolution of air/water interface with Volume Of Fluid (VOF) MULES method seems to be responsible for extreme air velocities on crests resulting in wave damping. This phenomena is solved by decreasing time step. Hydrodynamic forces on the cylinder match experiments with an error below 3%. As the flow is turbulent ( Re = 10 5 ), a turbulence model is included in the simulation giving results rather close to the ones obtained without turbulence model.

Simulation is increasingly used as an engineering tool in the maritime industry. Currently, there is a particular focus on digital twins , where a supplier of equipment should also supply a simulation model of the equipment. Today, the person configuring the simulator must manually connect all required signals between these models, interpreting the meaning of each signal name to determine which signal can be connected together (e.g., connecting “pressure out” to “pressure in”, while ensuring that their units match). This is a tedious and error-prone process. In this paper, we present a method for automatically connecting variables of different models while complying with the FMI standard in co-simulations. This also includes a method for unit conversions. The methods for both automatic connections and unit conversions are based on ontologies. The key idea is to build an ontology, based on existing ontologies, which enables structuring model input and output variables into groups and hierarchies, and assign them a defined type. This enables reasoning about connections between models, and simplifies the model connections. The focus in this paper is to present the ontology, how FMI interface signals can be connected to the ontology, and how this can be used to simplify model connections.

The use of liquefied natural gas (LNG) has been recognized as an effective alternative fuel for marine propulsion, evidently, a growing number of vessels, including passenger ships, is already running on such a fuel. While LNG bunkering can be performed in several ways, depending on transferred volumes and managerial considerations, the time spent for LNG bunkering is regarded to be a key factor to ensure the cost-effectiveness of such operation, since a minimization of bunkering duration at the berth will make port infrastructures available for more vessels. As a result, passenger embarkment is simultaneously conducted with ship bunkering, which may increase the potential risk to individuals both onboard and nearby. Given this background, this paper is to investigate the potential risk of passenger vessels with the identification of credible scenarios for port operations that can be carried out simultaneously with LNG bunkering. An approach of quantitative risk assessment is applied to determine the risk levels pertinent to proposed scenarios. For frequency analysis, the technique of the fault tree analysis is adopted to integrate the influence of human error and management policy to the likelihood of unwanted events. Consequence analysis is conducted in aids of a commercial software to simulate the impact of LNG dispersion and fires. Research findings have shown that simultaneous operations have a negative impact both on frequency and consequence of accidental scenarios arising from LNG bunkering, taking the risk beyond lower acceptance criteria. Finally, a quantification of the risk increase is proposed to help stakeholders identify criticalities and reduce the risk contributions of simultaneous operations.

In this paper, an end-to-end nonlinear model reduction methodology is presented based on the convolutional recurrent autoencoder networks. The methodology is developed in the context of overall data-driven reduced order model framework proposed in the paper. The basic idea behind the methodology is to obtain the low dimensional representations via convolutional neural networks and evolve these low dimensional features via recurrent neural networks in time domain. The high dimensional representations are constructed from the evolved low dimensional features via transpose convolutional neural networks. With an unsupervised training strategy, the model serves as an end to end tool which can evolve the flow state of the nonlinear dynamical system. The convolutional recurrent autoencoder network model is applied on the problem of flow past bluff bodies for the first time. To demonstrate the effectiveness of the methodology, two canonical problems namely the flow past plain cylinder and the flow past side-by-side cylinders are explored in this paper. Pressure and velocity fields of the unsteady flow are predicted in future via the convolutional recurrent autoencoder model. The performance of the model is satisfactory for both the problems. Specifically, the multiscale nature and the gap flow dynamics of the side-by-side cylinders are captured by the proposed data-driven model reduction methodology. The error metrics, the normalized squared error and the normalized reconstruction error are considered for the assessment of the data-driven framework.

The study on flow around a hydrofoil has been performed using various methods experimentally and numerically. Here we use a purely data-driven model using deep neural networks (DNNs) to reconstruct the flow fields. Its results are also compared with that obtained by traditional methods. The datasets of flow fields around a static hydrofoil obtained from computational fluid dynamics (CFD) are used for training the DNNs and then the trained data-driven model is utilized to make reconstructions and predictions. 9 different physics informed loss functions, which contain the prediction error and the error measuring the violation of the conservation law, are proposed to train the parameters of DNNs. Effects of the proposed loss functions on reconstruction of velocity field and pressure field are analyzed. Lift and drag forces predicted during the training time are also analyzed. However, the data-driven model based on DNNs with the optimum loss function works well in the interpolation but fails in the extrapolation. The reasons causing the errors are discussed.

This study was built upon previous works conducted by the authors in a series of numerical studies on submarine hydrodynamics and is aimed at enhancing the accuracy of computational fluid dynamics (CFD) application processes, which estimate the hydrodynamic performance of underwater vehicles for steady translation conditions in the horizontal and vertical planes. In an earlier work, the computed straight-ahead resistance of a submarine agreed with those of experiments within a comparison error of 2%. However, a maximum comparison error of approximately 20% was obtained for sway force under a steady translation condition. The Defense Advanced Research Projects Agency (DARPA) Suboff submarine model was adopted as a benchmark, and the computational modeling was based on the Reynolds-averaged Navier–Stokes (RANS) turbulence model for steady simulations. The curvature correction approach was tested to improve the computation of circumferential flow around the cylindrical hull, in particular. The dominant maneuvering coefficients were calculated using the computed forces and moments as a function of the yaw and pitch angles along with simplified equations of motion by fitting a curve to the plots. The hydrodynamic forces and moments exerted on the stern plane were individually computed using a locally refined mesh around the tail section. It was confirmed that the curvature correction approach improved the computational accuracy for the steady translation conditions, and general trends were captured over the tested yaw and pitch angles. However, some data points had notable comparison errors. Some of the estimated maneuvering coefficients agreed well between the CFD simulation and the experiments, whereas others had considerable comparison errors. The individually computed forces and moments exerted on the stern plane that had attack angles were inconsistent with those obtained in experiments. Those comparison errors may have been amplified by the complexity of configuration and arisen from differences in the experiments, such as the presence of a free surface and supporting strut to mount the hull to a carriage, and, perhaps, the geometrical differences owing to machining accuracy. The investigation of flow field at the propeller plane revealed that the wake distributions inside the nozzle were significantly affected by the angled stern planes; the reduced velocity area was expanded and shifted. Furthermore, harmonic analysis of the wake fraction was conducted, and several primary n th harmonics were observed, which were associated with the struts and stern planes. This result suggests the risk of higher noise levels associated with the number of blades.

Freak waves, abnormally large waves, that occur in the open-ocean can cause significant damage to offshore structures and vessels. In this paper, we attempt to numerically reproduce the experiments of McAllister et al., (2019, J. Fluid Mech. [1]), to investigate the potential properties of the Draupner freak wave [2] in more detail. We use a Smoothed Particle Hydrodynamics (SPH) method to solve the full-3D Navier-Stokes equations. This Lagrangian method is able to recreate wave breaking, and has the potential to fully reproduce these experiments with the aim of providing further insight into properties of the waves created such as their kinematics and geometry. We compare time histories of water surface elevation produced numerically using four different particle sizes with experimentally-obtained data. We find good agreement in the time domain, with r 2 (coefficient of determination) values between experimental and numerical data of over 0.94 the error in maximum wave height was less than 5 % for the finest particle size (over 100 million particles). We also numerically reproduce wave breaking observed in the experiments, where jet formation and breaking phenomena are qualitatively similar in appearance.

Residual stresses are internal self-equilibrating stresses that remain in the component even after the removal of external load. The aforementioned stress when superimposed by the operating stresses on the offshore piping, enhance the chances of fracture failure of the components. Thus, it is vital to accurately estimate the residual stresses in topside piping while performing their fitness for service (FFS) evaluation. In the present work, residual stress profiles of girth welded topside sections of P91 pipes piping are estimate using a machine learning approach. The training and testing data for machine learning is acquired from experimental measurements database by Veqter, UK. Twelve different machine learning algorithms, namely, multi-linear regression (MLR), Random Forest (RF), Gaussian process regression (GPR), support vector regression (SVR), Gradient boosting (GB) etc. have been trained and tested. In order to compare the accuracy of the algorithms, four metrics, namely, Root Mean Square Error (RMSE), Estimated Variance Score (EVS), Maximum Absolute Error (AAE), and Coefficient of Determination (R^2) are used. Gradient boosting algorithm gives the best prediction of the residual stress, which is then used to estimate the residual stress for the simulated input parameter space. In the future work authors shall utilize the residual stress predictions from Gradient boosting algorithm to train the Bayesian Network, which can then be used for estimating less conservative through-thickness residual stresses distribution over a wide range of pipe geometries (radius to thickness ratio) and welding parameters (based on heat input). Furthermore, besides topside piping, the proposed approach finds its potential applications in structural integrity assessment of offshore structures, and pressure equipment’s girth welds.

Accurate prediction of the fatigue strength of steels is vital, due to the extremely high cost (and time) of fatigue testing and the often fatal consequences of fatigue failures. The work presented in this paper is an extension of the previous paper submitted to OMAE 2019. The main objective of this manuscript is to utilize Artificial Intelligence (AI) to predict fatigue strength, based on composition and process parameters, using the fatigue dataset for carbon and low alloy steel available from the National Institute of Material Science (NIMS) database, MatNavi. A deep learning framework Keras is used to build a Neural Network (NN), which is trained and tested on the data set obtained from MatNavi. The fatigue strength values estimated using NN are compared to the values predicted by the gradient boosting algorithm, which was the most accurate model in the OMAE 2019 paper. The comparison is done using metrics such as root mean square error (RMSE), Mean Absolute Error (MAE), Coefficient of Determination (R 2 ) and Explained Variance Score (EVS). Thereafter, the trained NN model is used to make predictions of fatigue strength for the simulated data (1 million samples) of input parameters, which is then used to generate conditional probability tables for the Bayesian Network (BN). The main advantage of using BN over previously used machine learning algorithms is that BN can be used to make both forward and backward propagation during the Bayesian inference. A case study illustrating the applicability of the proposed approach is also presented. Furthermore, a dashboard is developed using PowerBI, which can be used by practicing engineers to estimate fatigue strength based on composition and process parameters.