Abstract This paper presents a research on the application of artificial neural networks (ANNs) to predict the seakeeping behavior of ships in head waves. The decisive input parameters of the ANNs are identified by analyzing the general equations governing the ship motions. Then, a ship database that considers all major merchant ship types and hull forms is set up and an in-house frequency domain 3D panel method is used to predict the heave and pitch motions of these ships in head waves at typical operational speeds, thus, establishing a motion database, which provides data to train the ANN networks. Several types of neural networks are explored and systematically trained. The developed network is applied to the prediction of the motions of several ships, which are not in the database, to demonstrate their efficiency in quickly and accurately predicting the seakeeping performance of typical merchant ships. INTRODUCTION The reliable prediction of the seakeeping behavior of ships in a seaway is a demanding task for naval architects and of great practical interest to ship owners/operators, as it affects both the design and operation of ships. Early assessment methods were based on potential flow theory. The first widely accepted method was the so-called strip-theory-based S.T.F. method (1970), whose performance is well documented in the literature and in relevant benchmark activities at ITTC (1978). Later on, more advanced methods of various levels of complexity, in both time domain and frequency domain, were developed. Some early examples refer to the works of Sclavounos and Nakos (1988), King et al. (1988), Lin and Yue (1990), Iwashita and Ohkusu (1992), Duan and Dai (1999), etc. Beck & Reed (2000) gave an extensive review of these methods and the benchmark study during the ITTC Workshop on Seakeeping (Kim, 2010) reported the performance of many of these methods. In parallel, sophisticated CFD tools based on free-surface RANSE methods were developed and underwent much progress with the fast development of hardware. Encouraging results have been presented by various researchers during recent international workshops (Larsson, et al., 2014; Shigunov et al., 2018) and CFD tools became a standard tool to assist practical ship design projects. As the computational resources are becoming inexpensive, full-scale viscous simulations are nowadays reported more often (e.g., Tezdogan et al., 2016).

In this study, geometric parameters of a hull such as prismatic coefficient, waterline length, waterplane area coefficient, and flare angle are considered as the main parameters related to the added resistance of ships in waves. The hull form is represented by cross-sectional curves and the free-form deformation (FFD) method is utilized for hull form variation. Non-dominated Sorting Genetic Algorithm-II (NSGA-II) is applied as a multi-objective optimization method to minimize both total resistance and speed loss of a ship in seaways. To estimate these objective functions, the calm water resistance and the added resistance due to winds are calculated by regression formulas, and the added resistance due to waves is predicted by a numerical method based on strip theory, momentum conservation method, and a modified asymptotic formula for the correction in short waves. Moreover, the speed loss is calculated using power-speed curves. It has been shown that the optimization solution with an obvious reduction of total resistance and speed loss in the sea condition of Beaufort 6 can be obtained. INTRODUCTION Recently, there has been a growing interest in operational efficiency and reduction of fuel consumption of ships in accordance with the regulation limiting the emission of greenhouse gases issued by the International Maritime Organization (IMO). Therefore, the shipbuilding industry began to incorporate the operational performance of ships in waves into the early-stage of the ship design. However, the typical ship designs have focused on optimizing only calm water resistance, so it is necessary to carry out multi-objective optimization of hull form considering environmental loads in actual seaways. There have been numerous studies on hull-form optimization considering the resistance of ships. It probably began by Weinblum (1959), and several studies were conducted since the 1970s to minimize calm water resistance in line with the improvement of numerical methods and computing power. Dawson (1977) proposed the modified Rankine source method to solve the wave resistance problem, and in many subsequent studies, the optimization of wave resistance utilizing this method was conducted. In addition, several interesting studies were carried out to optimize the calm water resistance (Nagamatsu et al., 1983; Papanikolaou et al., 1989; Larsson et al., 1992). Recently, studies on hydrodynamic optimization of hull form have been conducted using more sophisticated numerical methods and optimization techniques. Percival et al. (2001) minimized the calm water resistance of Wigley hull for three forward speeds. They distributed the control points over the entire hull and did not apply any constraints so the optimum hull was deformed impractically compared to the basis hull. However, this study was significant as it provided a sophisticated hull-form optimization methodology in terms of resistance. Choi (2015) also conducted a hull-form optimization to minimize the wave-making resistance of a container ship. Deformation was concentrated in the bow part and constraints are applied on displacement and wetted surface area to realize practical hull form variation. More recently, operational efficiency in waves is furthermore considered on hull-form optimization. Bolbot & Papanikolaou (2016) optimized the operational efficiency factors such as added resistance in waves and EEDI factor as the objective function. Yu et al. (2017) optimized both the wave-making resistance and the added resistance at wavelength ratio 0.5 of a 66,000 DWT bulk carrier by modifying a sectional area curve and section shape of the design load waterline. In the study from Jung & Kim (2018), the main dimensions of KVLCC2 tanker were modified and optimization was performed based on the total resistance and the speed reduction considering representative sea conditions. These recent studies have applied multi-objective optimization schemes because of the addition of optimization factors related to operational performance, and most of the hull deformations have been concentrated on the bow part, based on studies in which the front of the hull predominantly contributes to the added resistance in waves (Guo and Steen, 2011; Seo et al., 2013).

This study introduces a technique for identifying the parameters in the equation describing the heave/pitch motion of a ship by using only its measured response at sea. The parameters being identified are the direct and cross coupling damping and restoring parameters. These values are identified using a reinforcement learning technique. The proposed method would be suitable in identifying the direct and cross coupling damping and restoring parameters for a ship heaving and pitching under the action of unknown excitations effected by a realistic sea state. Published data, results of a commercial seakeeping package and recorded data for a ship's heave and pitch motions in random seas are used to test the accuracy and the validity of the method. It was shown that the method is reliable in the identification of the parameters of the coupled equation of the heave-pitch motions using only the measured response at sea. INTRODUCTION Predicting ship response to incident waves has been investigated since the 1960s (Salvensen, et al., 1970). Heave and pitch motions are the most studied among the six degrees of freedom of a ship in a seaway. The level of accuracy of predicted heave and pitch motions is higher than other degrees of freedom. Strip theory is the most popular method for analytical predictions of heave and pitch motions; however, strip theory often delivers inaccurate results in extreme sea states. Several other methods have been developed to predict vessel response to incident waves. Computational Fluid Dynamics or seakeeping computer codes based on potential flow theory (WAMIT, 2018) (ShipX, 2018) (Motions, 2018) play an increasingly important role in investigating and predicting the hydrodynamic performance of ships and offshore structures. structures. However, the linearization of the problem in most potential flow codes neglects several important effects. Firstly, pressure integration is performed over the time-average wetted surface, neglecting changes of the wetted surface due to waves and ship motions. Secondly, the influence of the changing wetted surface on the flow and hydrodynamic forces is not considered. While CFD simulations do not require the same simplifications, the computational cost is too great for many use cases. Recently, work on modelling the non-linear aspects of seakeeping using time-domain methods has become more common (He & Kashiwagi, 2014a) (Jiang, et al., 2015).

For a type of x-rudder underwater vehicle, a multi-objective control allocation algorithm for the distribution of the rudder angle of x-rudder is established. The algorithm considers the input saturation of the stern rudder, and can automatically calculate the rudder angle of the x-rudder according to the yaw and trim angular acceleration required by the underwater vehicle, so as to make the angular acceleration of the underwater vehicle which generated by the x-rudder as close to the target Yaw and trim angular acceleration as possible. In order to ensure the real-time control of the underwater vehicle, the solver of the rudder angle distribution formula should be as fast as possible, and the result should be as accurate as possible. Therefore, by analyzing the validity, celerity and accuracy of the sequential quadratic programming (SQP) method, the active set method and the trust region method, this paper selects the sequential quadratic programming as a suitable algorithm for the rudder angle assignment formula. INTRODUCTION Underwater vehicles are widely used in the field of ocean engineering. There are many types of underwater vehicle, which are usually divided into ROV with cable control and AUV with autonomous navigation. Compared with ROV, AUV has better adaptability because it is not limited by the cable and can perform longer and wider tasks. The use of rudder and propeller is a common control form of underwater vehicle, which is simple and reliable, and easy to operate. X-rudder is a new type of stern rudder. Its four rudders are arranged in X-shape. At present, the common x-rudders can be divided into the following three types: Fixed X-rudder. The rudder stock of this x-rudder can't be operated, in fact, it only plays the role of stabilizing wings. The motion control of this kind of underwater vehicles is usually completed by a series of thrusters. The function of the x-rudder is to provide additional stability to the underwater vehicle.

SPH-FEM conversion algorithm is proposed to simulate sloping structure and level ice interaction. This method combines the advantages of Finite Element Method (FEM) and Smoothed Particle Hydrodynamics (SPH), which provides a new idea for the research of ice engineering. In this paper, we used this method to simulate the interaction of sloping structure and level ice. Compared with the theoretical result by Croasdale 3D theory, this method predicted the icebreaking force correctly. Simultaneously, it accurately simulated the accumulation process of ice rubbles as the failure finite elements were replaced by SPH particles. Thus, the SPH-FEM conversion algorithm is considered an effective way and will be used in the numerical simulations of ice structures or icebreakers. INTRODUCTION The operation of ships or offshore platforms in the Arctic is often in hash environments, which mainly includes strong wind, high wave, sea ice, low temperature, icing and so on (Bridges et al., 2018; Dehghani-Sanij et al., 2017; Necci et al., 2019). Among them, the ice load on structures is the dominant environmental stress (Xu et al., 2015) and it can cause damage to offshore structures with serious consequences. For example, Vallinsgrunden lighthouse collapsed under several significant ice actions in 1979 (Bjerkås, 2007). Thus, some solutions were proposed to reduce or withstand ice load. Sloping structures are widely found in marine structures to reduce ice loads, such as wind turbines and bridge piers (Barker et al., 2005; Brown et al., 2010) because these reduce bending failure of ice and are likely to be exposed to a severe dynamic ice load (Matskevitch, 2002). Depending on the conditions, the interaction between sloping structure and ice may produce several phenomena. For example, the breaking ice will move up the cone and then clear around the structure in the case of a narrow sloping structure. As for a wide one, rubble accumulation may form in front of the structure (Paavilainen, 2011).

The paper introduces advanced flexible riser design fatigue analysis and describes an innovative method that allows the conversion of measured dynamic riser and vessel response into real time fatigue damage accumulation. This is achieved through a combination of field-proven riser monitoring system, machine learning, and global and local flexible riser stress analysis at both the tensile and pressure armour layers of a flexible riser. The innovative monitoring system outputs and benefits are described, being the input for the machine learning method used for fatigue tracking. Key aspects and benefits of the methodology, results and the efficiency increase in comparison to traditional methods are discussed and recommendations made on the multiple applications that are enabled through these. INTRODUCTION Unbonded flexible pipes are extensively used for riser systems deployed for subsea oil and gas exploration and production operations worldwide. These pipes were originally developed primarily as a shallow water solution for the transport of production, injection and export fluids. Nevertheless, flexible systems have also been developed for deepwater operations and now are a key enabling technology for exploration of both well-established offshore regions, such as Brazil and the North Sea, and other areas around the world, including, but not limited to, West Africa, Australia, South-East Asia, and China. As with any critical subsea structure, the industry has been investing in the development of methods and equipment to ensure safe operations in green and brown field developments (Elman and Alvim, 2008). Furthermore, there is an increasing demand for extending the system service life past the original design life. These aspects necessitate an understanding of the in-place structural condition to assess the remnant fatigue life. There are several limiting factors including the methods, the practical and financial aspects associated with inspection, monitoring and engineering assessment, which are discussed along the paper. The equipment and methods presented here provide an approach to address these limitations.

Arctic business is developing within the sectors of marine operations related to offshore oil and gas, mining, seafood, tourism, scientific expeditions, and world trade shipping. As offshore activities in the Arctic is a relatively new field, with only a handful relevant operations to draw experience from, and since full-scale trials are extremely expensive, there is an expressed need for more detailed and cost-efficient analysis and design review of concepts based on numerical simulations. Over the years, various numerical simulators, with different levels of technical readiness, have been developed to address issues related to Arctic Offshore Engineering. This paper focuses on the development of the Simulator for Arctic Marine Structures (SAMS), its basic theories and recent engineering applications. These include structural damage assessment for simulating the impact between a wave driven glacial ice feature with a semi-submersible structure; the decision-making support by ice load calculations on a shipwreck in the Arctic; and various Arctic vessel navigation simulations. The technical advancement in terms of multi-body dynamics, ice fracture mechanics and hydrodynamics that the simulator is built on is manifested. The presented engineering applications signify the importance, the versatility and maturity of SAMS in dealing with complex ice – structure interaction issues at different scales and at various levels of complexities. INTRODUCTION For Arctic structural design and most Arctic operation planning, ice load calculations are one of the key components. However, ice load calculation is complicated by various ice conditions and different ice failure modes. As Palmer et al. (1983) stated, ‘ it is no more likely that there should be a universal ice-force formula than that there should be a universal formula for the force on a body in a moving fluid, and in the present state of knowledge, it would be unwise to expect too much ’. Over the years, the common engineering practice, e.g., (ISO19906, 2019), in calculating ice load on a given structure takes several steps of presumptions: an assumption on the limiting mechanism and an assumption on the various possible failure modes (Løset et al., 2006, Sanderson, 1988, Palmer and Croasdale, 2013). These approaches often offer one design value and are handy for fast calculations, however, require high engineering experience in making sound judgements. As time steps into the new century, our computational power and tools are getting more advanced. In addition, our understanding towards the mechanics of ice – structure interactions are also ever growing thanks to more accumulated data and theoretical advancement. It is now at our disposal to 1) get rid of all those initially needed interaction scenario presumptions and start from the beginning with given ice condition and structure 2) to obtain time domain analysis of both the ice action and its effect on the structure; 3) in addition, these calculations can be so fast that an almost real-time domain analysis becomes possible, which opens door for simulations of Arctic marine operation. Given the background and targets, the Simulator for Arctic Marine Structures (SAMS) is developed over the past 8 years. This paper documents the theoretical background of SAMS and its recent applications.

Until 2019 the Rules of Russian Maritime Register of Shipping (RS) had only regulated design ice impact loads on inclined hull sides. These regulations are based on a hydrodynamic model of ship hull/ice interaction. At present large-size vessels operating in the Arctic seas have full hull lines with vertical sides extending over the entire hull entrance or even bulbous bows. Such vessels may have an ice class up to Arc7. The hydrodynamic model does not provide sufficiently simple analytical relations for class rules to estimate impact ice loads on vertical sides in way of forebody entrance in case of ship collisions with a floe or ice channel edge. This paper suggests a method for regulation of impact ice loads on vertical sides using the tools of experiment design theory in numerical simulations. The numerical experiment is ice load computation based on a physical hull/ice interaction model. The suggested method is implemented in application to the hydrodynamic hull/ice interaction model. Relationships of ice impact loads versus ice class, hullform parameters and ship mass are obtained in the form of regression formulas. Design relationships are developed to determine parameters of ice loads on vertical sides. These relationships are included in RS Rules of 2019. INTRODUCTION Modern ice class vessels may be designed with bulbous bows to ensure good seakeeping in open water. The current Rules of Russian Maritime Register of Shipping (RS Rules) allow this design for the ice-going vessels of Ice1 to Arc7 inclusive (RMRS, 2020). In particular, the bulbous bow is typical for double- acting ships (DAS), i.e. when the ship stern is designed for stern-first operation in ice conditions, while the bow is optimized for open-water sailing (Aleksandrov, Platonov and Tryaskin, 2018). Under the current RS Rules the design ice loads on ice belt in the forebody of an ice-class vessel are determined depending on the ice strengthening class and parameters of the hull form at the ice load waterline 1 . Ships with bulbous lines in way of the ice waterline (and often ballast waterline) will typically have a vertical side portion where the frame angle is very small or zero. Non-arctic vessels of Ice1-Ice3 may have a long part of vertical-side forebody extending from the fore perpendicular to the parallel middle body (Platonov and Tryaskin, 2019a). It was impossible to estimate ice loads according to current RS Rules for these ship design cases.

In this paper, based on the idea of two-parameter model of bond-based peridynamics (BB-PD), the formulas of bond force is derived, and the tension and simple shear of plate are simulated by the derived form of bond force. Then the time step and the corresponding parameters of the model under the quasi-static algorithm and dynamic relaxation algorithm are derived,respectively. The two algorithms are applied to the simulations, respectively. By conducted two typical simulations It is proved that the two-parameter model can solve the problem of fixed Poisson's ratio, and the correctness of formulas of bond force and parameters in calculation. INTRODUCTION When dealing with the crack problems, the peridynamics (PD) is calculated by the integral equation, which avoids the shortcomings of discontinuity equation of the traditional continuum mechanics at the crack (Silling,2000; Silling and Askari,2005). However, BB-PD is only related to bulk modulus when simulating linear elastic materials, so that the Poisson's ratio is fixed to 1/ 3 for plane stress problems, and 1/ 4 for plane strain and 3D problems. It is impossible to accurately simulate problems with the arbitrary Poisson's ratios. The fixed Poisson's ratio has so much limits, when solving the practical problem. Therefore, the state-based peridynamic was proposed (Silling, Epton, Weckner, Xu, and Askari,2007). There is no problem of the fixed Poisson's ratio in the state base. However, the state-based peridynamic is more complicated than the bond-based peridynamic. Therefore, domestic and foreign researchers have done a lot of research on solving the problem of fixed Poisson's ratio. Gerstle, Sau and Silling (2007) proposed the micro-polar model, which adds a pair of moments based on the original bond force. However, the model needs to process a large amount of calculation in actual use, and it does not give a clear potential energy function. Prakash and Seidel (2015) proposed the twoparameter linear elastic peridynamic model. Two coefficients of springs were introduced to describe the displacements in the normal and tangential directions to solve the fixed Poisson's ratio problem. Similarly, Zhu and Ni (2017) considered the shear deformation in the BB-PD model of the bond. By introducing the rotation effect of the bond, the bond force between the two material points can be decomposed into normal and tangential components, and the energy of the bond can also be divided into tensile energy and shear energy. In fact, the twoparameter models proposed by Prakash and Seidel (2015) and Zhu and Ni (2017) are the same models in theory, except that the definitions of the potential energy functions are different, which resulted in the different final forms of coefficients. Diana and Casolo (2019) summarized several theoretical hypotheses of bond and proposed corresponding solutions. And a micro-polar model considering shear deformation of bond and failure criteria for shear deformation was proposed.

This paper presents a target tracking control method for a biomimetic underwater vehicle named RobCutt-II. The RobCutt-II is propelled by double undulatory fins which are distributed on both sides. In order to implement target tracking and observation in a certain distance, we propose a method by decoupling the target tracking control problem into distance, heading and depth control problems. Based on this idea, we design a target tracking controller composed of a navigation system, a distance controller, a heading controller, a depth controller and a switching logic controller. The simulation results verify the effectiveness of the proposed method on the biomimetic underwater vehicle. INTRODUCTION Underwater vehicles play an important role in marine geological surveys and underwater rescue. Unlike traditional underwater vehicles driven by axial propellers, biomimetic underwater vehicles (BUVs) are propelled by undulatory fins for high mobility in complex underwater environments. Moreover, high stability, low noise, and small environmental disturbance are the advantages of biomimetic underwater vehicles compared with traditional underwater vehicles. A variety of biomimetic underwater vehicles have been developed by scientists around the world. A BUV was designed by Delft University of Technology in the Netherlands (Simons, Bergers, Henrion, Hulzenga, Jutte, Pas, Van Schravendijk, Vercruyssen and Wilken, 2009). An undulatory fin propelled underwater vehicle was designed by Osaka University (Rahman, 2012). Zhou et al. from Nanyang Technological University developed a CPG-controlled underwater vehicle mimicking devil fish (Zhou and Low, 2012). In our previous work, we designed a biomimetic underwater vehicle propelled by undulatory fins named RobCutt-II (Wang, Wang and Wang, 2016). By means of different undulatory modes of the double fins, the RobCutt-II generates forces in the forward direction, vertical direction and moment to change the heading angle. However, the RobCutt-II cannot generate effective lateral force, which is considered as an underactuated system. Many interesting method and results have been presented for underactuated systems. For example, Aguiar et al. proposed a control method for an underwater vehicle without lateral and vertical control (Aguiar and Pascoal, 2007). Yang et al. designed a robust trajectory tracking controller for marine surface vessels (Yang, Du, Liu, Guo and Abraham, 2014). Zhou et al. designed a path following method using the extended state observer and the neural network. (Peng, Wang and Han, 2019). In some underwater observation tasks, the underactuated underwater vehicles are required to track and to keep a certain distance from the target for observing. Existing researches pay little attention to this issue.

Matched field processing (MFP) is usually performed in an ocean waveguide with a vertical line array to determine a source position. Aiming to improve depth and horizontal distance resolution of MFP, this paper introduces an optimization method that applies the Lucy-Richardson deconvolution algorithm to Bartlett MFP. Point spread function (PSF) in the deconvolution algorithm is established based on norm mode model (NM) and fast field analysis. Examples are presented to illustrate deconvolved MFP (DMFP), and DMFP yields a low sidelobe level better than Bartlett MFP. Compared with MVDP, it retains the robustness on environment mismatch. INTRODUCTION Vertical line arrays have been used for target localization in waveguide environment based on matched field processing and other methods (Sazontov and Malekhanov, 2015). MFP is performed by matching the data received by array elements with the simulated data of all possible target locations. The Bartlett MFP, similar to plane wave beamforming, is a popular and conventional method in which the received complex vector on various sensors and simulated vectors depending on location parameters are processed by linear correlation. When the simulated vector coincides with the received data of a given target, all elements of target vector are coherently summed and the Bartlett MFP gets the peak of the output in ambiguity diagram. Suppose noise is uncorrelated between the receivers and is hence incoherently summed. Consequently, the signal-to-noise ratio (SNR) at the ambiguity diagram is enhanced compared with the input SNR. The advantages of the Bartlett MFP are its robustness against signal mismatch between the assumed and the actual data (such as when there are measurement and approximation errors for waveguide parameters), and small number of samples required for a reliable estimate of the target location. The disadvantage of the Bartlett MFP is the high sidelobe level making localization of a weak signal difficult in the presence of a loud interferer. Many other methods with high resolution have been proposed based on the inverse of the signal vector covariance matrix to overcome the deficit of the Bartlett MFP. Well-known methods use the minimum-variance distortionless processor (MVDP), and eigenvector-based MUSIC algorithms (Kolev and Georgiev, 2007). These algorithms produce low sidelobe levels and narrow beamwidths, and hence allows for better interference rejection and localization of targets with a small location separation. At the same time, these algorithms are known to be sensitive to the signal mismatches. Signal mismatch can cause part of the signal to be misidentified as noise and consequently suppressed by the algorithm. To diminish the sensitivity, the MVDP algorithm is modified under additional constraints, such as the white noise gain constraint or distributed sources constraints, to produce a wider beam in the target location while keeping the sidelobe level low. The second problem with these high-resolution algorithms is that the covariance matrix of the signal is usually not known and is replaced by the sample covariance matrix. Assuming the signal is stationarity in time domain, the sampled covariance matrix requires many samples of signals to assure rapid convergence and minimum power loss in the beam output. The high-resolution algorithms usually have poor performance under certain conditions and if one were to ignore the non-stationarity and still average the sampled covariance matrix over many samples, the matched field processor's ability to reject interference is much decreased. The MUSIC algorithm has the same problem requiring a sufficient number of snapshots to obtain reliable estimates of the signal and noise eigenvectors. It also needs prior knowledge of the number of sources to separate the signal vectors from the noise vectors (Zhang and Gao, 2013).

Based on a new embedded operating system, a control system for a torpedo type AUV is designed. The AUV has a horizontal rudder, a vertical rudder, two horizontal thrusters, two vertical thrusters, and a main thruster as the execution mechanism. The control algorithm uses integral separation fuzzy PID control algorithm based on engineering experience. Based on this algorithm, the course controller of AUV is designed, and numerical simulation tests are performed. The test results show that the AUV control system is reliable and the control effect is good. INTRODUCTION In recent years, with the deepening of the development of marine resources, autonomous underwater vehicle technology has also made great progress. It has the advantages of low cost, small size, and strong flexibility. It is widely used in deep-sea resource surveys and other fields. AUVs are usually highly nonlinear and coupled, and are affected by parametric perturbations and current disturbances during navigation. Therefore, AUVs require high hardware and software reliability as a guarantee when performing tasks in the water. A new embedded operating system has excellent performance and high reliability and this paper designs AUV control system based on the embedded operating system.E.H.Mamdani (1974) first used fuzzy logic in the control of the boiler, marking the birth of fuzzy control. Practice has proved that fuzzy control has better control effect for the system with uncertain parameters. Chen(2013) Applied the fuzzy PID control method to UUV, and achieved good control effect on the model uncertainty UUV. Based on the past practical engineering experience, this paper uses the integral separation fuzzy PID control algorithm to design the AUV actuator controller. This article first briefly introduces the characteristics of this embedded operating system, and then introduces the design architecture of the torpedo AUV control system, and models and analyzes it. The control effect of course controller is verified by numerical simulation.

In this paper, the numerical model of level ice is developed in the framework of finite elements and the water is considered by SPH method, the numerical method is established by adopting the contact algorithm to calculate the contact forces between finite elements and water particles. Then the icebreaking process of the polar tanker is calculated and its navigation in open water simulated by CFD numerical simulation based on viscous theory. The resistance of a polar tanker under multiple load cases is predicted. INTRODUCTION The melting speed of ice in Arctic region is accelerating because of the changes in climate, which makes the marine transportation in Arctic region possible. The polar tanker will play an important role and its full route includes both ice area and open water area. And the demand of the polar tanker with self-icebreaking ability is increasing recently. For a polar tanker, the comprehensive performance in both of the ice area and the open water area should be considered in the design phase. In the resistance prediction of ice-breaking, finite element method is a commonly used numerical calculation methods. Compared with the oversimplification of the model in the empirical formula and the long period and high cost of the model test, it can take into account the effects of sea ice failure modes and more ship form parameters. The Cohesive Zone Model (CZM) has found great attention in the applied numerical fracture literature to solve problems of multiple cracking brittle and quasi-brittle solids during impacts and failure loading. Then the CZM was extended into sophisticated numerical algorithms in the framework of finite elements, termed the Cohesive Element Model (CEM), for the simulation of ice-structure interactions, (Konuk et al, 2009). Arne Gürtner et al. (2009) used the numerical method to simulate the action to a lighthouse, the simulation results indicate that the numerical method captures many of the qualitative observation as well as quantitatively derives comparable global ice loads to the lighthouse to those of the selected ice event. Wang et al. (2008) developed a collision model of crushable ice based on nonlinear dynamic finite element and analyze the ice load of a FLNG ship using commercial code DYTRAN. Yang et al. (2008) adopted the method of fluid-structure interaction and established the nonlinear finite element model of collision between ship and ocean platform by taking sea ice as medium.

A single kind of wheel or track underwater walking operation robots may do not suit deep-sea complex terrain conditions. To solve these problems, a wheel/track convertor is presented in this paper, the convertor adapts to various terrains and it has both advantages of wheel and track walking system. A control algorithm is designed independently for safe changing the convertor mode. The dynamic simulation analysis demonstrates that the wheel/track convertor can transform the wheel/track mode dynamically, it can walk smoothly on two kinds of different terrains. INTRODUCTION There are a lot of complex and mixed abominable terrains in the deep sea (Liu, Hu, Zhang, Dai, and Yang, 2016). A single kind of wheel or track underwater walking operation robots may not be able to fully adapt to these complex terrains (Inoue, Takagi, and Shiosawa, 2010). Therefore, in order to fit the requirements of walking speed and crossing obstacles as far as possible, the currently underwater walking robots commonly use a fixed track walking mechanism. However, the speed of the fixed track mechanism can not be effectively improved in hard ground and a single wheel mechanism may not move quickly in soft ground. To solve these problems, a metamorphic mechanism for wheel and track conversion has been presented (Paillat, Lucidarme, and Hardouin, 2008). These deformation mechanisms can be turned into triangular tracks walking mechanisms from wheel mechanisms. Through the rapid conversion of the wheel/track, the deformation mechanisms can adapt to a variety of soft and hard grounds, which greatly improves the ability of robots to overcome obstacles (Cui, Gao, Guo, and Dong, 2016; Peng, Zhou, Hu, and Ou, 2014; Qu and Zhong, 2013). However, this kind of metamorphic mechanisms also can not be adapted to these hard or soft roads with complicated terrains. In order to solve these problems, this article is presented a wheel/track convertor, which has two modes of wheel and quadrangular track. The shape of the track mode can be adjusted in real-time according to the current road conditions.

In water level prediction, Tidal level is a part that satisfies tidal laws, and the residual water level caused by environmental factors is nondeterministic and time-varying. The difficulty of its prediction lies in the inability to quantify short-term meteorological factors such as wind, air pressure, and precipitation, and the increase and decrease of water caused by seasonal climate factors. This is the main reason that restricts the accuracy of traditional water level prediction models. With the development of artificial intelligence, neural networks and fuzzy logic reasoning are gradually applied to marine hydrologic factor prediction. Empirical modal decomposition is a method for handling nonstationary signals that can decompose the different scale fluctuations in the signal step by step. Optimization of residual water level sample data by using empirical modal decomposition method can effectively improve the approximation and prediction ability of the model in machine learning. Based on the above analysis, this paper uses the combined model of Empirical Modal Decomposition and Long Short-Term Memory to separate the data from the residual water level sequence and use machine learning training and prediction to improve the short-term prediction accuracy of the residual water level. Then the predicted residual water level and the harmonic analysis water level are superimposed, that is, the predicted water level. The results show that this method can improve the prediction accuracy of tidal water level to a certain extent and can be used in tidal level prediction. INTRODUCTION Tide is the one of significant dynamic phenomena in about port and harbor activities, and it is also an important factor affecting marine economic activities. With the development of the world economy, trade between countries is becoming more and more frequent. As one of the important transportation modes, maritime shipping occupies a dominant position in the world's transportation industry due to its large capacity and low cost. In recent years, China's shipping industry has made great progress. The leading role of shipping in foreign trade and transportation has become increasingly prominent. The number and types of ships entering and leaving the port have continued to increase. Real-time and accurate water level prediction can provide safety guarantee for large ships entering and leaving ports and diving operations, and also provide a basis for decision-making for shipping establishment. Taking advantage of the strong time correlation of water level changes, accurate water level prediction can reduce the number of tide detection stations and expand the prediction range, thereby saving costs.

The future knowledge of the waves and force is indispensable for the model identification and the real-time control of ocean engineering devices. In order to effectively control the motion of the offshore structures in a real-time manner, it is required to have an accurate and efficient prediction of the waves. Machine learning has been widely applied in ocean engineering field as it offers compromise between prediction accuracy and computational cost. The present study focuses on wave-force prediction of offshore structures based on deep machine learning algorithms. A novel wave-force prediction model is proposed, which makes full use of the efficient processing characteristics of Long Short-Term Memory Recurrent Neural Network (LSTM RNN) and Nonlinear Autoregressive Exogenous Feedback Neural Network (NARX FNN) for time series data processing. The relationship between the wave height and the wave height is non-causal and nonlinear which need future wave height knowledge for current wave excitation force. Therefore, The LSTM RNN is firstly utilized for multi-step prediction of the time series of wave elevation. The NARX FNN is used to address the model system identification between the wave heights and the wave force. Then, the LSTM RNN is further applied to predict the future force of offshore structures for the real-time control of the structure motions. After that, the proposed deep machine learning algorithm is utilized for wave-force prediction based on the experimental data obtained in Kelvin Hydrodynamic Laboratory and the optimal horizon can be specified for the test model by comparing the performance of different prediction horizons. The results indicate that LSTM-NARX model can successfully predict the time series of the waves and force. 1 INTRODUCTION Wave energy as a dense and stable renewable energy resource is forecast to have the potential to supply 10% of European electricity needs or to generate the equivalent of up to 20% of UK electricity; about half today's total renewable generation. WECs convert the oscillation of kinetic and potential energy carried by ocean gravity waves to electrical energy that can be delivered to the electrical grid through a mechanism known as power take-off system (PTO) (Anderlini,2019) . Wave energy drives two or more parts move relatively then energy is captured by hydraulic mechanic or direct drive. However, there is no wave energy converter reaching commercial stage due to its high levelised cost of energy (LCOE). There is one of the ways to move a step forward by improving the power absorption efficiency under real-time control. By controlling the force exerted from PTO system, such as latching control, it is possible to tune the velocity of WEC with the excitation force of incoming wave for achieving the maximum energy absorption. The wave excitation force is regarded as the combination of incident component and diffraction component from the view of linear potential theory (LPT).

We numerically investigate the flow control problem of the flow passing a stationary cylinder at a fixed Reynold number 500 using two attached control cylinders with different rotation rates. Compared to the traditional uniform (lattice) sampling method, we developed an active learning strategy based on Gaussian Process Regression (GPR), drastically reducing the number of simulations and accelerating the scientific findings. We also discussed the effects of parameters on different hydrodynamic coeffcients, and verified the feasibility of this strategy. The mechanism of this asymmetric flow control model was also further studied by analyzing flow patterns. INTRODUCTION Fatigues and vibrations due to dynamic marine environment, like Vortex-Induced Vibration(VIV), are fairly severe that the analysis of fluid and structure interaction (FSI) for offshore structures are important. Since almost all of these structures have crucial components such as piles and stacks that could be modeled as cylinders in fluid domain, we narrow our focus on the flow around the cylinder, one representative type of bluff body, where there exists well-known studies having figured out the mechanism of oscillations (Sarpkaya, 1979) and the crucial effect of Reynold number (Pastò, 2008) of this FSI problem, as well as different wake modes of two side-by-side circular cylinders at low Reynolds numbers (Sangmo, 2003) and fluid characteristics of rotating cylinders next to a wall (Cheng et al., 2007; Rao et al., 2011). Almost all these studies aim at how to enhance the hydrodynamic properties of target bodies, promoting a vivid flow control field consisting of various active (Jahanmiri, 2010) and passive control strategies (Kumar et al., 2008), with related studies (Sakamoto et al., 1994) on their mechanism and the nature of the controlled wake. However, though extra energy is needed, active flow control, with Moving Surface Boundary-Layer Control (Modi, 1997) as the most frequently used method, has been regarded to be more robust and flexible than its counterpart. One model incorporating this methodology is to use two attached rotating cylinders (Muddada et al., 2010), which is also the prototype of our improved model. Specifically, With an objective of drag reduction and focusing on the mean drag coeffcient ( C D MEAN ), this model could achieve higher energy effciency of underwater vehicles, while their maneuverability could be improved with the mean lift coeffcient ( C L MEAN ) studied. Similarly, When this method used in wave energy generation with Root-Mean-Square (RMS) values of the drag and the lift force ( C D RMS , C L RMS ) targeted, the productivity of energy harvest device could be improved by magnifying vibration amplitudes.

As a new concept of special AUV, the Ocean Bottom Flying Node (OBFN) can well complete the tasks of offshore oil exploration, water cruising and monitoring. However, due to the limitation of their own space, the OBFN can only carry a small amount of energy. In order to improve their navigation efficiency and better complete their operational tasks, the OBFN is required to have good hydrodynamic performance. In this paper, the drag reduction performance and volume space are used as optimization parameters, and the OBFN configuration is optimized through optimization theory. Taking into account the internal space volume, a better drag reduction configuration is obtained. 1. INTRODUCTION In order to meet the needs of large-scale network deployment in seismic exploration applications, Harbin engineering University develops a new concept special Autonomous Underwater Vehicle (AUV)—Ocean Bottom Flying Node (OBFN), which combines Ocean Bottom Node (OBN) and AUV techniques. Equipped with multi-functional sensors, it can also meet multi-task requirements such as long-term sitting, sea cruise, marine resource exploration, and submarine seismic monitoring. The original OBFN model is shown in Fig. 1. The OBFN only relies on its own energy when sailing underwater. Limited by its internal space, the OBFN can only carry a small amount of fuel, so it is necessary to improve navigation efficiency and reduce energy consumption. Since the large number of internal parts and the complicated external molding, the optimization design of OBFN is a huge challenge. At present, the traditional optimization methods are more from the personal experience of the designer, which is not only inefficient, but also often cannot obtain the global optimal solution. In recent years, breakthroughs have been made in Computational Fluid Dynamics (CFD) and modern optimization theory, which has greatly accelerated the development of hydrodynamic optimization. AUV hydrodynamic optimization presents the characteristics of multidisciplinary cross-cutting and multi-objective optimization. A large number of scholars have underwater vehicle optimization.

Based on the work by Chen (1993) on the Green function in water of finite depth in which two depth-effect functions are defined, we have performed new analyses and obtained new formulations. The analytical logarithmic terms are extracted from the depth-effect functions. Special numerical algorithms based on integration along complex contours have been developed for numerical computations with an absolute error of order 10 −8 . Furthermore, the formulation is designed that the depth-effect functions are fully smooth varying and then approximated by Chebychev polynomials. Not only the Green's function itself, but also its first-order derivatives and second-order derivatives are numerically evaluated and approximated independently by polynomials. The analytical terms plus polynomials are then used to enhance the effciency and capability of our in-house software HydroStar. Unlike previous studies, the analytical part of the Green's function contains explicitly the logarithmic singularity when the depth-scaled wavenumber tends to zero. Usually included in the "smooth" part of approximation, this singular term is then extracted and the remaining part without singularity is fully smooth and much better approximated with consistent accuracy of O(10 −6 ) and less partitions. Furthermore, the added-mass and damping coeffcients around a hemisphere are given as an example of benchmark results. Finally, the mathematical and physical aspects of the Green's function for wave frequency tending to zero are discussed. INTRODUCTION Being important for hydrodynamic analysis of floating bodies in general and some renewable devices installed in shallow water, in particular, the Green function of wave radiation and diffraction in water of finite depth is re-analysed here. Lots of studies have been done for the Green function of diffraction and radiation in deepwater. As summarized in Xie et al. (2018), there are more than 20 formulations developed for different domains of applications. Unlike the Green function in deep water, however, only few studies on its formulation and effcient computations presented in classical work by John (1950), Endo (1982), Newman (1985, 1992) and Chen (1993). Recently, an increasing interest seems rebounding as shown in the work by Shan et al. (2018), Liu et al. (2018) and Mackay (2019).

It has been proved that injecting bubbles into a turbulent boundary layer can reduce the frictional drag on the plate. Many studies have simulated this problem using Euler-Lagrange method to accurately calculate the motion of bubbles. However, bubble coalescence and breakup have not been fully considered in the Eulerian-Lagrangian simulation of bubble drag reduction problem. The main goal of the present study is to develop bubble breakup and coalescence modules in an Euler-Lagrange code. Simulation of bubbles injected into a turbulent boundary layer is performed. And by means of these two modules, the bubble size distribution is analyzed in detail. INTRODUCTION Using bubbles to reduce skin frictional resistance of ship or pipe has been a focus for researchers for a long time. These kinds of practical problems can be simplified to injecting bubbles into a turbulent boundary layer. Hassan et al. (2005) injected bubbles into the channel flow by electrolysis of water and obtained obvious drag reduction effect. In their next experiment, Ortiz-Villafuerte and Hassan (2006) investigated a turbulent boundary layer laden with microbubbles and adopted particle tracking velocimetry technique to analyze the effects of bubbles on turbulent coherent structures. Results showed that the velocity fluctuation in the flow direction is strengthened while the velocity fluctuation perpendicular to the plate is weakened. Murai et al. (2006, 2007) conducted similar experiments and measured the Reynolds stress distribution in the channel. It was found that the Reynolds shear stress decreased effectively near the wall surface, then quickly recovered as the flow develops to the core of the channel and increased outside the original boundary layer. The Reynolds shear stress near the wall is an important part of the turbulence friction resistance. Therefore, the reduction of Reynolds shear stress near the wall was regarded as an important drag reduction mechanism. Above experimental studies have shown that only bubbles entering the turbulent boundary layer can contribute to drag reduction. There are obvious fluid velocity fluctuation and strong shear stress in turbulent boundary layer, which can lead to the breakup and coalescence of bubbles. Bubble breakup and coalescence lead to the continuous change of bubble size distribution, which has an important effect on drag reduction. This phenomenon also puts forward higher requirements for the numerical simulation algorithm.