
This paper presents an experimental and numerical study of simplified aircushion-type platform models, mainly focusing on the internal free surface created by the entrapped air in the chamber. Since the elevation of this free surface significantly affects the motion of the floating platform, investigating its characteristics is of great importance. In this study, to measure and analyze the inner free-surface elevation, both mono- and partitioned-aircushion-type models were tested in regular waves. A conventional box-type floater without an air chamber was also tested under the same wave conditions for comparison. The inner free-surface elevation along the inner air chamber was measured under both fixed and moored conditions. The coupling effects between the inner free-surface elevation and the motion response of the platform are discussed. The inner free-surface profile was numerically reproduced using WAMIT, and the results were compared with experimental data. Results show that in the short-wavelength condition, the inner wave fields exhibit standing waves for both fixed and moored conditions. As the incident wavelength increases, the inner wave fields change into progressive waves for the moored condition, while sloshing patterns are observed for the fixed condition.
Unmanned surface vehicles (USVs) operating near other vessels can be affected by ship-to-ship hydrodynamic interactions. This study investigates a simulation-based trajectory-tracking framework that combines a hydrodynamic surrogate model (SM), nonlinear model predictive control (NMPC), and L1 adaptive feedback compensation. The surrogate is incorporated into the prediction model, while the L1 component is intended to compensate for residual uncertainty. A physical USV platform was constructed, and nominal dynamic parameters were identified from maneuvering tests, with reported velocity-fit values exceeding 70
To address the difficulty of acquiring deep‑sea hydrate mining riser vibration data, a spatiotemporal prediction model is developed. The model accounts for correlations among different positions at the same time, and vibration signals from shallow positions are used to predict those of deep‑sea risers. Within a temporal convolutional network (TCN) framework, the model integrates a dimension segment‑wise (DSW) mechanism and a two‑stage attention (TSA) mechanism to capture inter‑variable dependencies. The DSW embedding first segments the time series—length. covering different riser positions and both cross‑flow and in‑line displacements—into windows of varying lengths, outputting a two‑dimensional vector array. This array feeds into the TSA layer, which captures cross‑time and cross‑dimension dependencies in two stages: first modeling temporal dependencies among segments, then modeling dependencies across different dimensions. Leveraging dilated convolutions, the TCN backbone extracts long‑term and long‑range temporal dependencies. The extracted features are then weighted and passed to Kolmogorov-Arnold networks (KAN) for nonlinear fitting, thereby enhancing the model’s capacity to represent complex sequence relationships. Predicted in-line and cross-flow vibration displacements are compared with experimental data. The comparison indicates that the DT (DSW-TSA)-TCN-KAN model achieves accurate predictions across different working conditions, and its integrated modules contribute to the observed performance. Among the compared models, DT-TCN-KAN exhibits the highest prediction accuracy for mining riser vibration displacement.
In reliability engineering, accurately estimating the tails of probability density functions is critical for failure probability prediction, especially when working with limited data. Under such conditions, prediction accuracy often deteriorates. To address this, the fractional-order moment maximum entropy method enhanced by the maximum a posteriori estimation (FM-MEM+MAP) was developed, incorporating prior information through the maximum a posteriori (MAP) estimation based on fractional moments. This approach extracts implicit input–output mappings within the sample set and fuses them with observed data to enhance reliability prediction accuracy. The method’s performance was validated through numerical experiments and applied in time-domain wave simulations of ships, incorporating hydroelastic effects. By introducing uncertainty in parameters such as hull stiffness, the method facilitates probabilistic modeling of extreme wave loads. The resulting distribution model was demonstrated to be effective in ship reliability assessments using Monte Carlo simulation (MCS).
To address the complexity of flow field distribution and propeller performance variation of an underwater vehicle during pitching motion, a fluid–solid coupling calculation model is established. The flow field characteristics of the vehicle at different pitch angles are studied, and the reasons for changes in the velocity distribution over the propeller blades are explained from the perspective of flow field interference. The performance of the propeller under minimum hull after‑effect at various angles of attack is investigated, and the results under high and medium loading conditions are discussed. The effects of varying pitch angle on the flow field, pressure and velocity distributions, thrust coefficient, torque coefficient and open‑water efficiency of the underwater vehicle are revealed. The results show that the intensity and extent of turbulent kinetic energy generated by the vehicle during pitching motion are greater than those observed in straight‑line motion. When the vehicle is travelling in a straight line or at an oblique angle, pressure is primarily concentrated at the bow, whilst velocity is mainly distributed along the parallel mid‑body. At a high angle of attack, greater torque is required to rotate the propeller. Conversely, at a low angle of attack, the thrust generated by the propeller is greater than that produced at a high angle of attack. These research findings can serve as a reference for the development of long‑endurance underwater vehicles.