In view of the numerous and complex structures of ship machinery, equipment and the coupling of vibration transmission paths, a method of underwater radiated noise prediction based on BP (Back Propagation) neural network was proposed in this paper. A BP neural network based on gradient descent algorithm and Bayesian regularization algorithm was constructed respectively. Vibration data was taken as input, hull radiation noise was taken as output, and root mean square error (e_RMSE) and mean absolute error (e_MAE) were taken as evaluation indexes of model prediction accuracy. The results show that the generalization and robustness of Bayesian regularization BP neural network is better than that of gradient descent algorithm BP neural network. The error of Bayesian regularization is less than 3 dB, and the proposed method has good applicability in the field of ship radiation noise prediction.

An eccentric semi-submersible foundation was proposed considering the characteristic of wind energy direction concentration in actual environment. AQWA-Fast software was used to establish the floating wind coupling analysis model, and the kinematic response characteristics of symmetric and eccentric wind turbine were compared and analyzed under different incident angles of wind waves. The results show that the variation of wave incidence angle hardly affects the average motion response value and average power generation of the floating wind turbine. When the incident angles of wind and wave are the same, the eccentric platform has better motion performance in terms of sway, roll, and yaw degrees of freedom, and to some extent the mooring tension is reduced. When the wave incidence angle constantly changes, the eccentric design will improve the motion performance in terms of sway, roll, pitch, and yaw degrees of freedom. In addition, the eccentric floating wind turbine has better comprehensive power generation performance.

In order to investigate the hydrodynamic performance of ships in restricted channel of polar regions, a hybrid Green’s function method based on three-dimensional potential flow theory was established, in which, the free-surface Green’s function was taken as the kernel function in the fluid domain of channel. The ice sheet on both sides of the channel was modeled as a thin elastic plate. The ice-surface Green’s function, which automatically satisfies the ice-covered water surface, seabed and radiation conditions, was adopted in the fluid domain below the ice. The boundary integral equations were established with the two kernel functions above respectively. To improve the efficiency of computation, the influence coefficients related to the wave component of the ice-surface Green’s function were obtained based on its analytical integral over vertical line segment. On the basis of above method, a computer program was developed. By taking a barge with available computational data as the research object, the effects of control surface length and mesh density on the convergence of calculation were first discussed, and the reliability of the method and program was verified. Then further calculations and analysis were conducted on the hydrodynamic coefficients under different channel widths and ice thicknesses. From the numerical results the oscillatory phenomenon was observed in the hydrodynamic forces of the ship in water channel confined by ice sheets. And with the decrease of channel width and increase of ice thickness, the oscillation becomes stronger.

Numerical modelling based on Navier-Stokes equations and model experiment for studying liquid sloshing have the limits of low computational efficiency and high economic cost. Therefore, to predict the hydrodynamic pressure and wave height, the time-histories to numerical and experimental results were reconstructed in this paper through the neural network model. The total numerical and experimental pressures and free surface elevations were taken as training samples, and CNN, RNN and LSTM with strong repretational ability were used to reproduce the sloshing responses. The internal structural parameters of the neural network were systematically adjusted, besides, the errors and correlations between the predicted and actual values were analyzed. The results show that the error is lower than 4% and the correlations of both RNN and LSTM reach 0.88, which is in general superior to CNN, and that LSTM is optimal in predicting the long sequence data. Overall, three surrogate models can well predict the sloshing wave height and pressure, and are promising in the study of liquid sloshing.

In this paper, a fast prediction model was established for ship motion and load based on Gated Recurrent Neural Networks (GRU). GRU neural network is a concise and efficient recurrent neural network that captures the temporal information of training samples to establish a model for predicting unknown samples. The forecast model consisted of two independent GRU neural networks used to predict ship motion and load respectively. The historical ship pitch and heave data were jointly used as the input of the motion prediction model to predict the ship pitch and heave in the next few seconds. The motion prediction results were used as the input of the load prediction model to achieve the prediction of the vertical bending moment in the midship. The method was validated through model test data, and the results showed that the prediction results at different lead times were in good agreement with the test results in terms of amplitude and phase, verifying the feasibility of the established ship motion and load prediction model.

A marine riser usually works in the form of multiple riser clusters. The hydrodynamic interference characteristics between adjacent risers are obviously seen, and the interference can accelerate the fatigue damage. Therefore, it is meaningful to investigate the vortex-shedding patterns and the hydrodynamic interference characteristics of twin cylinders in order to ensure safe operation of risers. This study was aimd to parametrically investigate flows around twin moving cylinders at a staggered arrangement by a ghost cell method, in which, the incompressible Navier-Stokes equations were solved on a Cartesian staggered grid by using an in-house developed time semi-implicit finite difference method. A ghost cell method was used to enforce the no-slip boundary conditions. A radial basis iso-surface function was used to track the moving boundary implicitly and identify the properties of background grid. Based on the present numerical model, flows around twin forced moving cylinders at a staggered arrangement were simulated. Vortex patterns and force coefficients were analyzed under different gaps and oscillation frequencies. Typical interference phenomena such as synchronized, deflected and merged vortex patterns were observed. The results can provide theoretical guidance for the arrangement optimization of the riser clusters.

In order to understand the motion characteristics of amphibious aircraft planing in head waves, in this paper, the Chinese numerical tank-Cartesian grid finite difference method was used to study the planing motion response characteristics of an amphibious aircraft in waves under different wave conditions. The effects of wavelength and wave height on heave and pitch response of the aircraft and the vertical overload of mid-ship and stem were summarized. Based on the step-by-step method, the Navier-Stokes equation was discretized in space. The immersed boundary method was used to capture the boundary of complex objects and accurately simulate its large motion. At the same time, the THINC/SW algorithm was used to effectively simulate the violent motion of free surface. The results show that the wavelength and wave height have a great influence on the planing motion response of the amphibious aircraft. With the increase of planing speed, the pitch resonance wavelength increases, and the heave resonance wavelength decreases first and then increases, while the linear correlation between the heave and pitch response of the aircraft, the vertical overload of the mid-ship and stem and the wave height will also be weakened.

Aiming at the initial stress field generated by welding of ring-ribbed cylindrical shell structures with initial geometric defects after shape correction, this paper presents the study on the influence of initial stress field on the strength and stability of ring-ribbed cylindrical shell structures under the premise of considering geometric nonlinearity caused by large deformation. The finite element model of ring-ribbed cylindrical shell with initial geometric defects was constructed by using ANSYS software and the initial stress field was calculated. The strength and stability of ring-ribbed cylindrical shell with initial geometric defects and stress field were solved by arc length method. The comparison between the experimental results and the calculated results verifies the validity of the model. The analysis shows that the ultimate bearing capacity of the ring-ribbed cylindrical shell under hydrostatic external pressure is slightly reduced, and the regularity of the instability waveform is reduced, but the failure position remains unchanged.

In order to improve the propulsion performance of contra-rotating propellers and reduce the cavitation effect, an end plate was applied to the contra-rotating propeller. The cavitation performance and propulsion performance of the end-plate contra-rotating propeller were analyzed. The RANS method with Schnerr-Sauer cavitation model was used for analyzing. Then the propulsion performance of contra-rotating propellers composed of conventional skewed propellers and end-plate propellers were checked for comparison. It is found that the addition of an end-plate makes the contra-rotating propeller possess better anti-cavitation performance. The sheet cavitation range is reduced by about 59% in the bollard state (J=0), and the cavitation is delayed when J=0.1. In addition, under the condition of low advance speed, the end-plate contra-rotating propeller shows a higher propulsion efficiency by 0.9%~3.1% than the conventional one. The open water performance of the end-plate contra-rotating propeller was tested at different rotation speeds, and the data were in good agreement with the simulation results considering the cavitation model. In this study, an end plate is innovatively applied to the contra-rotating propeller, which is suitable for the propulsion and operation requirements of low-speed submersibles.

The additional strengthening effect caused by the non-proportionality of loading-path under cyclic loading is an important factor that shortens fatigue life of material. To solve this issue, the plane of maximum shear strain amplitude was treated as the critical plane, and the non-proportionality of material and loading-path were both considered. A new non-proportionality factor was introduced to quantify the impact of non-proportionality loading on fatigue life of material based on the equivalent strain model and critical interface theory. Secondly, the damage mechanism and fatigue failure mode of the specimen were also considered, the maximum normal stress on the critical plane was adopted to characterize its contribution to fatigue failure. On this basis, a multiaxial fatigue life prediction model was proposed by combining the non-proportional factors. Finally, the proposed model was verified by using the fatigue test data of four materials under multiaxial loading, and the prediction results were compared with five proposed models. The results show that the proposed model can effectively improve the fatigue life prediction accuracy under non-proportional loads compared with the existing models.