In ship hull subassembly and intermediate assembly welding, the large number of structural components and frequent variations in structural layouts result in highly customized operations. Consequently, offline programming (OLP) results are rarely reusable across different workpieces, and repetitive modeling and path editing are required for each new assembly, which significantly prolongs the overall programming cycle—a challenge particularly prominent in shipbuilding. To address the resulting severely narrow and non-convex workspace, severe kinematic coupling, and mutual interference inherent in such customized tasks, this paper proposes a tailored spatial obstacle avoidance framework for a gantry-type dual-arm welding robot. The proposed framework integrates welding torch posture feasibility evaluation, bound-constrained Levenberg–Marquardt inverse kinematics solving, and hierarchical link-level collision detection. First, a default welding torch posture is constructed according to the weld seam direction, and a GJK-based posture evaluation and correction strategy is developed to ensure end-effector feasibility in constrained environments. Subsequently, the redundant inverse kinematics problem is formulated as a bound-constrained nonlinear least-squares optimization problem and solved using a trust-region Levenberg–Marquardt method combined with projection techniques. Finally, local collision detection and hierarchical configuration adjustment are performed at path corner and transition nodes according to different collision types. Simulation results show that 422, 372, and 362 key nodes are evaluated in Subassembly 1, Subassembly 2, and the combined assembly scenario, respectively. The method successfully obtains collision-free configurations for all nodes in the first two scenarios, whereas only a few transition nodes in the combined scenario require additional adjustment. Furthermore, PQArt offline simulation validates the continuity, kinematic feasibility, and collision-free performance of the planned trajectories. These results demonstrate that the proposed method effectively satisfies the spatial obstacle avoidance requirements of dual-arm welding robots in constrained shipbuilding environments and provides a practical reference for OLP and engineering applications.
The hydrodynamic behavior of a dual-vessel system is a key factor governing the local motion compatibility of side-by-side refuelling involving unmanned surface vehicles (USVs). In this study, the coupled motion responses of an S175 supply ship and a Wigley-III USV are investigated using ANSYS AQWA based on three-dimensional potential flow theory. The method is first validated against published experimental data for wave-excitation forces, added mass, and radiation damping, showing reasonable agreement in the dominant response trends. A parametric study is then carried out to examine the effects of wave heading, forward speed, USV size, and sea state on the heave, roll, and pitch responses of the two-vessel system. The results indicate that wave heading is the dominant operational factor. The most unfavorable conditions arise when the smaller USV is directly exposed to head seas, whereas a wave heading of about 150° significantly reduces USV motions because of the shielding effect of the larger supply ship. Speed variation mainly changes the frequency of encounter and may lead to resonance-related amplification when approaching the natural frequency range of the system. In addition, larger USVs show improved seakeeping performance, while the motion responses of the supply ship are less sensitive to changes in USV size. The present results provide preliminary hydrodynamic tendencies for screening candidate headings and spacings in side-by-side USV refuelling.
As digital transformation accelerates, the shipbuilding industry has shown growing interest in digital twins that link physical assets to digital models. Building a ship digital twin requires a piping digital model that integrates piping design information generated progressively during design; however, connecting information between the 2D piping model (e.g., P&ID drawing) and the 3D piping model remains difficult because they are created with different tools at different stages. Although approaches that transition to a unified common model framework from the early design stages have been proposed, they require a comprehensive overhaul of the existing design process and limit the reuse of accumulated design data. A method for post-hoc integration of information between two models created under the current workflow is thus needed. This paper presents a matching method for piping design information between 2D and 3D piping models using anchor nodes. The piping design information of both models was represented as a common piping graph model. And valves, equipment, and orifices—identifiable by their unique names—were represented as anchor nodes, reducing the matching problem between the two different piping models to a subgraph search at the piping model section level, which can be obtained by partitioning by these anchor nodes. The method was applied to test data and industrial data. For closed sections whose boundaries consist solely of anchor nodes, 100% matching accuracy was achieved on the test data. The sections that failed to match in the industrial data were attributable to actual design discrepancies between the 2D and 3D piping models, indicating that the method can also serve as a tool for detecting design errors between different piping models. These results confirm the feasibility of constructing a ship piping digital model while maintaining the existing design process. The proposed method is expected to help link the ship piping digital model to construction within current design workflows.
This study addresses the challenge of efficiently planning compartment searches inside sunken large ships, where the numerous compartments complicate the process. A simulation-based approach is proposed, utilizing a mixed discrete event and discrete time simulation to model the search operation. The simulation considers key variables such as the number of diving sites, decompression chambers, and the order of compartment entry. Three case studies were conducted. The first simulated compartment searches based on individual dive plans, varying the number of decompression chambers (4 to 8) and dive order (max, min, random). Results indicated that using six or more chambers significantly reduced search time, with the longest dive time order (max) being the most efficient. The second case incorporated current speed variations, adjusting dive times accordingly. A new dive order (custom), combining short and long dives, was introduced. The simulation results showed that total search time did not decrease beyond six chambers, with the custom order proving most effective. The third case added search difficulty levels (high, medium, low) to the simulation, influencing diver movement speed within compartments. The simulation re-explored compartments if the total search time decreased beyond a threshold. The results indicated that certain compartments required further exploration. For each case, the simulation results were visualized using Gantt charts. The study provides a crucial reference for efficient search planning by simulating sunken hull compartment searches under varying conditions.
As the core load-bearing component of underwater hull structures, studying the buckling failure characteristics and their transformation laws of stiffened cylindrical shells under deep-water explosions is of great significance. In this paper, static and acoustic-structural coupling analysis methods were employed to conduct a numerical study on the dynamic buckling of underwater internally stiffened cylindrical shells subjected to the coupled action of deep-water explosion loads. The dynamic buckling modes of the internally stiffened cylindrical shell during the shock wave and two bubble pulsation stages were analyzed, and the buckling components were quantitatively assessed based on critical buckling displacement and the internal energy of the plate, shell, and ribs. Finally, the buckling evolution process of the internally stiffened cylindrical shell for each mode was examined. The research demonstrated that at the same water depth, as the explosion distance decreased, the primary damage area of the structure shifted from both ends to the middle. When the hydrostatic pressure was excessively high, damage occurred solely at both ends, and circumferential instability was likely to ensue. The buckling instability evolution laws of the internally stiffened cylindrical shell under shock wave loads exhibited similarities. During the first-stage bubble pulsation load, local buckling and aggravated deformation occurred in different parts for each mode. The second-stage bubble pulsation load further exacerbated the instability. Ribs tended to buckle in shallow water, whereas the plate-shell was more vulnerable in deep water.
Autonomous underwater vehicles (AUVs) are increasingly employed in marine exploration and reconnaissance, frequently in conjunction with submarines. However, the inherent uncertainty regarding the positional accuracy of the AUV and the submarine in underwater environments presents a significant risk of collision, particularly during recovery operations. This paper proposes an uncertainty aware guidance framework for recovering an AUV to a moving submarine. The AUV and submarine positions and covariances are estimated by an extended Kalman filter (EKF) based inertial navigation system. These covariances are used to inflate the hull geometries into configuration-space collision-risk zones. A three-dimensional extension of the time to closest point of approach (TCPA) and distance at closest point of approach (DCPA) is then mapped into a Mamdani fuzzy-logic system to produce a scalar collision-risk index. This index is coupled with a potential field guidance law whose repulsive component is padded by the EKF covariance, and with a risk dependent speed modulation that adjusts the AUV surge velocity. In a Monte Carlo campaign of 10,000 randomized encounters, the proposed fuzzy TCPA–DCPA metric achieved an area under the ROC curve of 0.92 and a Youden optimal decision threshold of 0.85, corresponding to a true-positive rate of 0.80 at a false-positive rate of 0.10. High-fidelity simulations in the Stonefish environment, including hydrodynamics and rigid-body collisions, show that conventional kinematic guidance and a naïve potential field result in collisions in quartering approaches. In contrast, the proposed uncertainty-aware potential field and speed-modulation strategy eliminate collisions and reduce the peak collision risk index from 0.88 to 0.69 while still enabling successful docking. These results indicate that explicitly incorporating navigation uncertainty into both collision-risk assessment and guidance substantially improves the safety of AUV–submarine recovery operations.
Accurate tracking of maritime obstacles is essential to ensuring safe ship navigation. To this end, a variety of sensors—such as AIS (Automatic Identification System), RADAR (RAdio Detection And Ranging), LiDAR (Light Detection And Ranging), and cameras—are widely employed. Data from these sensors is used to estimate the state of surrounding obstacles. However, each sensor has unique characteristics, including its detection range and measurement frequency, making it challenging to achieve stable and precise tracking with a single sensor. Therefore, a sensor fusion method for obstacle tracking is required to compensate for each sensor's strengths and weaknesses. In this study, we proposed a track-to-track fusion method to improve the tracking accuracy of maritime obstacles by integrating RADAR and camera data. The RADAR data were processed using the CA-CFAR (Cell Averaging-Constant False Alarm Rate) algorithm, and the camera data were processed using YOLOv11 (You Only Look Once v11) to detect maritime obstacles. The EKF (Extended Kalman Filter) was then applied to the detection values from both sensors to track the maritime obstacles. These tracking results were then integrated using a track-to-track fusion method applying the WMA (Weighted Moving Average) algorithm. Finally, the proposed method was validated across a range of maritime encounter scenarios using a virtual simulator that emulates the maritime environment. As a result, the track-to-track fusion method recorded mean tracking errors of 9.91 m (95.81%) for position, 10.88° (71.10%) for COG (Course Over Ground), and 1.34 knots (79.68%) for SOG (Speed Over Ground), relative to the single-sensor-based tracking method (100%). Therefore, the track-to-track fusion method achieved reductions of 4.19% in position, 28.90% in COG, and 20.32% in SOG, significantly improving performance compared to the single-sensor-based tracking method. These results indicate that the track-to-track fusion method effectively enhances the reliability of maritime obstacle tracking.
This study evaluates the crashworthiness of a Duct-type Reinforced Concrete Structure (DRCS) for protecting submarine cables against stock anchor drop impacts, utilizing both full-scale testing and numerical analysis. A full-scale test with a 1.27-ton anchor dropped from an equivalent height of 1.5 m was conducted to obtain strain histories and damage patterns. To establish a reliable numerical methodology, Coupled Euler-Lagrange (CEL)-based Fluid-Structure Interaction (FSI) simulations were performed. Four concrete constitutive models—ISO, WCM, RHT, and CSCM—were critically compared to identify the most suitable model for reproducing experimental results. The comparison revealed that the Continuous Surface Cap Model (CSCM) demonstrated the highest accuracy, with an average strain error of approximately 14% and a standard deviation of error ratio of 0.02, effectively capturing the strain rate effects and local damage behavior observed in the test. The validated numerical model exhibited an average error range of 14–27% across key crashworthiness indicators, which is within the acceptable range for concrete dynamics. The findings confirm that the proposed numerical approach using the CSCM model provides a robust framework for the safety design and assessment of submarine cable protection structures.
Sloshing can increase boil–off gas (BOG) generation in partially filled tanks by renewing near–wall thermal boundary layers and deforming the liquid–vapor interface. This study quantifies the corresponding boil–off–rate (BOR) amplification in a rectangular water–steam system maintained at a nominal pressure of 2.0 bar, with saturation and outer–boundary temperatures of 120 and 160 °C, respectively. A volume–of–fluid (VOF) unsteady Reynolds–averaged Navier–Stokes (URANS) solver is coupled with conjugate heat conduction through a modeled solid insulation layer, Rohsenow–type weak nucleate wall boiling, and a volumetric interfacial evaporation closure. Hydrodynamics are benchmarked against an isothermal sway–excited rectangular–tank experiment, and static heat ingress is assessed using Nusselt–number comparisons and boiling–regime checks. Twenty forced–sway cases cover filling ratios of 20–80% and excitation periods of 0.9–1.7 s. Near the dominant sloshing response, the time–averaged BOR reaches 2.8–4.0 times the static value, while the longest–period cases yield amplification factors of 1.09–1.57. The wetted–wall pathway contributes 62–93% of the dynamic BOR. Dynamic–to–static wetted–wall and free–surface area ratios organize the observed BOR amplification, and a ridge–regularized surrogate achieves R2 = 0.894 and MAPE = 11.5% within the investigated design space.
Safe autonomous berthing requires precise estimation of the relative distance and heading angle of a ship with respect to the quay wall. However, the performance of quay wall detection using shipborne 3D LiDAR (three-dimensional Light Detection and Ranging) may be significantly degraded under very near range conditions owing to the loss of planar information caused by the limited vertical field of view (FOV) and the geometric interference induced by fenders attached to the quay wall front. Considering these practical limitations in real port environments, this paper proposes a robust algorithm for quay wall detection and berthing state estimation. In the proposed framework, the 3D point cloud is projected onto a 2D (two-dimensional) plane, after which an azimuth-based 2D scan is performed to mitigate the limitation of the FOV. Moreover, the actual quay wall line is robustly identified by eliminating fender interference on the basis of the geometric parallelism of port structures and differences in point distribution. The proposed algorithm was experimentally evaluated through lateral berthing tests conducted at Ulsan Port using the autonomous test ship Haeyang Nuri. Under near-range conditions, the proposed algorithm achieved the lowest distance estimation error among the evaluated methods, with RMSED=0.0605m, while maintaining stable relative angle estimation with a maximum angular deviation of 0.7530°. These results suggest that the proposed method is well suited for robust quay wall detection and relative position estimation in full-scale berthing operations at real ports.
Bunkering simultaneous operations (SIMOPS) enhance operational efficiency by enabling bunkering and unloading work to proceed concurrently. However, the reduced clearance between the vessel and the quay, as well as between adjacent vessels, increases the likelihood of gap flow and gap resonance. Such hydrodynamic interactions may lead to pronounced amplification of wave elevations at specific locations, thereby imposing elevated loads on the mooring system. In particular, during bunkering simultaneous operations, two gaps one between the vessel and the quay, and another between the two vessels are formed, resulting in more complex flow characteristics and an increased influence of nonlinear effects. To quantitatively elucidate the wave amplification phenomena induced by gap flow under such conditions, this study establishes a two-dimensional CFD-based numerical wave tank and evaluates the wave elevation RAOs across a range of hull geometries, drafts, and spatial configurations. First, under a conventional unloading work condition involving a single floating body, the gap resonance characteristics between the floating body and the quay wall were examined for a range of hull geometries including rectangular and bilge-rounded cross sections and draft conditions. The shift in resonance frequency and the variation in response amplitude associated with changes in the midship sectional area coefficient (CM) were compared, thereby elucidating the influence of hull form and draft on the resonance behavior. Subsequently, a bunkering simultaneous operation scenario involving two vessels was considered, and the wave elevation amplification occurring within both the vessel-quay wall gap and the vessel-vessel gap was investigated concurrently. By systematically varying the draft of the outer vessel and analyzing the interaction of wave responses within the two gaps, it was found that the resonance frequency shifts and response amplitude variations become considerably more complex in the combined gap environment.
As carbon pricing expands and decarbonization targets become stricter, wind propulsion is being reconsidered for commercial shipping. This review summarizes the global deployment of wind-propelled ships and development trends. A multi-source data integration approach, drawing on peer-reviewed papers, public vessel databases, and industry and association reports, was used to build a ship-level inventory of vessels equipped with wind propulsion systems (WPS) by the end of 2025. The dataset includes 96 ships, and 90 were still in service by the end of 2025. Based on this inventory, this review describes current applications by operating regions, ship types, retrofit compared with new build routes, and whether the wind device serves as primary or auxiliary propulsion. Five main wind propulsion systems, including Flettner rotors, suction wing sails, rigid wing sails, soft sails, and kite sails, are then compared in terms of adoption trends and key integration features. In addition, this review also brings together the main technical challenges, including the efficiency of wind energy use, safety and reliability, operational and environmental factors, and economic considerations and commercial viability. Finally, policy measures, classification, and industry guidance in major regions are outlined to explain the main drivers and obstacles to wider adoption.
This paper presents a novel Advanced Uncertainty-Driven Integrated Mechanical (AUDIM) framework for the reliability-based optimization of submarine pressure hulls. The methodology integrates multi-source uncertainties material variability, manufacturing tolerances, and operational loading fluctuations into a holistic probabilistic design process. It employs a hierarchical Bayesian framework coupled with adaptive Gaussian process surrogate modeling to directly quantify reliability indices. The implementation follows a four-stage process: (1) uncertainty landscape mapping via global sensitivity analysis; (2) adaptive Kriging (AK-MCS) metamodeling for computationally efficient nonlinear finite element analysis; (3) failure probability estimation using the first-order reliability method (FORM) combined with importance sampling; and (4) multi-objective robust design optimization via the sequential optimization and reliability assessment (SORA) algorithm within an NSGA-II optimizer. When applied to a deep-diving submarine pressure hull with a baseline weight of 51 tons, the AUDIM framework achieved a 4.6% weight reduction while maintaining a system buckling reliability index of β = 4.82, exceeding classification society requirements. Sensitivity analysis revealed that external pressure variability contributes 42.3% to displacement variance, offering clear guidance for targeted quality control. The results demonstrate that probabilistic design enables scientifically balanced trade-offs between weight, cost, and safety, providing naval architects with a transparent, computationally efficient decision-support tool for next-generation submarine design.
This study investigates the vibration and radiated noise of submerged structures in polynya regions, where ice patterns altered by global warming affect hydroacoustic interactions. A theoretical framework integrating a multiple condensed transfer function (MCTF) for vibration analysis of conical-cylindrical-spherical shells with internal substructures, and a modified boundary element method (MBEM) for acoustic modeling with discontinuous ice-water boundaries, is proposed. Forced vibration tests and underwater acoustic measurements validate the methods. Results show that polynya geometry distorts near-field transmission loss and directivity, with larger polynyas amplifying sound pressure via phase-reversed reflections. Clustered polynyas cause interference-induced heterogeneity. The study highlights limitations of homogeneous ice-cover assumptions and underscores the need for polynya-specific acoustic models.
Ship Pipeline Routing Design (SPRD) aims to generate pipeline layouts by considering various objectives and constraints within a constrained 3D space. Given the spatial complexity and large scale of pipelines, the design process is difficult, time-consuming, and requires significant manual effort. For multi-pipeline comprehensive routing, traditional methods primarily adopt a two-step strategy combining path-search and collaborative optimization; however, both efficiency and optimization need improvement. This paper presents a deep reinforcement learning-based framework for ship pipeline routing design (DRL-SPRD), which employs multiple agents to independently perceive the environment and make decisions, with each agent corresponding to an individual pipeline, to achieve end-to-end comprehensive SPRD. It introduces an architecture where action sequences of other agents assist in collaborative design, and realizes intelligent design of multi-pipeline comprehensive routing through a sequential-alternating decision-making mode. Additionally, to address the inefficiency of traditional grid-based representation methods, a spatial representation method incorporating prior topological information is proposed. This method uses constructed topological relationships to replace fixed-step selection, thereby achieving efficient spatial representation. Finally, comparative experiments are conducted to verify the effectiveness and efficiency, followed by validation of the model’s performance in complex design environments using practical engineering cases, which demonstrate that the model exhibits high feasibility and advancement.
Ship docking is a frequent and crucial operation during a vessel's life cycle, serving various purposes such as construction, maintenance, inspection, modification, and repair. The docking conditions, where the ship is borne by supports, require determining the number and placement of supports based on the ship and on-site environment to prevent damage to both the hull and the supports. Traditionally, docking operations relied heavily on the experience of on-site experts. However, as ships have grown larger, ship types have diversified, and hull structures have become more complex, the mass distribution has become considerably complicated, exposing limitations of experience-based docking arrangement work. Consequently, shipyards have implemented various evaluation methods to ensure safety during docking. Although CAE software and computing power have been improved, analysis work still requires significant time and effort. Finite element analysis (FEA), which is used in most shipyards to assess the safety of docking conditions, has high accuracy but requires significant computational cost. This study proposed a new embedding method capable of representing support arrangements for docking conditions and a novel neural network structure based on physical constraints to develop an AI surrogate model capable of performing conventional FEA. To verify the accuracy of the AI surrogate model, the total predicted reaction force and support reaction forces were compared with FEA results. The developed neural network structure demonstrated reliable prediction results by exactly guaranteeing zero reaction force at unsupported locations. The developed surrogate model exhibited a total reaction force error of 1.95% and an average relative reaction force error of 3.48% for the entire test set, along with a prediction time of 1.9 ms per single case, achieving real-time performance. The developed AI surrogate model is expected to assist in determining the optimal support arrangement for specific on-site conditions by evaluating various support layout candidates.
This study addresses a block-scheduling problem in shipyard painting processes by incorporating detour-path constraints and group completion-time requirements. A truncated branch-and-bound (B&B) algorithm guided by simulated annealing (SA) is proposed. In the methodology, SA is first used to rapidly identify a high-quality initial solution, which is then refined through the B&B procedure. The performance of the proposed algorithm is evaluated by comparing its solution quality and computation time with those obtained from a MILP model. Furthermore, a case study based on a Korean shipyard is conducted to validate the robustness of the algorithm. By incorporating practical operational factors observed in real shipyards, this study provides a foundation for developing automated scheduling systems for shipyard painting operations.
Accurate estimation of wind and current loads is essential for the maneuvering and station-keeping performance of modern vessels. Existing empirical approaches, such as Fujiwara's regression and DNV-ST-0111, are widely used during the early stages of design; however, their reliability for multihull vessels, particularly those with asymmetric superstructures, has not yet been fully verified. This study evaluates these methods using a 19.7 m Crew Transfer Vessel (CTV) catamaran as a case study. Surge (Cwx/Ccx), sway (Cwy/Ccy), and yaw moment (Cwn/Ccn) coefficients are computed over drift angles from 0 degrees to 180 degrees using steady RANS double-body simulations, complemented by URANS multiphase simulations to assess free-surface effects. Numerical accuracy is verified through mesh convergence and ITTC-based uncertainty analysis, yielding GCI values below 1-2%. The results show that Fujiwara's method significantly overestimates the longitudinal wind load (Cwx) by up to similar to 70% due to its inability to capture asymmetric separation and three-dimensional flow interactions, while also underpredicting stern-wind behavior. Although lateral forces and yaw trends are generally reproduced, notable deviations occur near peak loading conditions. For current loads, DNV-ST-0111 predicts overall trends reasonably well but underestimates yaw moments due to neglect of tunnel flow interaction and asymmetric pressure distribution. Free-surface effects are found to be negligible at the considered low Froude numbers, supporting the use of the double-body approach for efficient load prediction. Overall, this study provides a quantitative assessment of the limitations of widely used empirical approaches for multihull vessels and demonstrates the importance of resolving three-dimensional flow interactions for accurate load prediction. The results show CFD as a reliable tool for estimating environmental loads on catamaran configurations and provide a validated dataset that can support the development of improved prediction models. The CFD-based coefficients presented in this study provide a reference dataset for future model development and practical design applications.
Ship subdivision optimization is challenging because it involves a high-dimensional mixed-integer design space and requires repeated strength evaluations. To address these challenges in preliminary ship design, this study develops a two-level parametric modeling strategy based on an Inner Shell Control Entity (ISCE). This strategy provides a unified and flexible description of inner-shell geometry and transverse bulkhead arrangements, enabling the automatic generation of three-dimensional compartment configurations suitable for strength, hydrostatic, and stability analyses. Building on this parametric representation, a multi-objective optimization framework combining NSGA-II with a Dueling Double Deep Q-Network (D3QN) is proposed to improve the search efficiency of ship subdivision design. The optimization model aims to maximize total cargo volume while minimizing still-water bending moment under constraints of geometric feasibility, compartment capacity allocation, hydrostatics and stability, and longitudinal strength. The evolutionary process is formulated as a Markov decision process. A D3QN agent adaptively selects crossover and mutation probability combinations according to evolutionary state indicators, thereby achieving an effective balance between exploration and exploitation. Case studies on a 55,000 DWT product oil tanker and a 50,000 DWT bulk carrier demonstrate that the proposed framework attains Pareto-optimal solutions comparable to those obtained by conventional NSGA-II and multi-objective particle swarm optimization (MOPSO), while significantly accelerating convergence. Specifically, the number of generations required to reach 99% of the maximum hypervolume is reduced by 83.9% and 77.0%, respectively, compared with standard NSGA-II. Furthermore, the D3QN agent trained on the tanker case is directly transferred to the bulk-carrier case without retraining, demonstrating cross-ship-type generalization capability. The optimized subdivision schemes provide cargo-volume adjustment ranges of 5.7% and 8.0% without altering the principal dimensions, confirming the practical applicability of the proposed method in supporting engineering trade-offs between payload capacity and longitudinal strength.
Achieving both stability and adaptability is a central challenge for multibody autonomous underwater vehicles (AUVs) operating in complex marine environments. This paper develops a unified mathematical modeling framework that integrates hydrostatic regulation and morphological reconfiguration. Three mechanisms are formulated: buoyancy adjustment for quasi-static equilibrium control, aperture-angle variation for dynamic angular morphing, and linkage-length modulation for geometric-spatial morphing. Numerical simulations based on the derived nonlinear dynamics evaluate their effects under roll disturbances. Results show that compact morphs yield rapid but oscillatory recovery, extended morphs enhance stability at the cost of slower convergence, and intermediate settings provide balanced performance. Phase-space trajectories and posture recovery diagrams confirm the distinct yet complementary roles of the three mechanisms. The study paves the way for the dynamics analysis and control of variable-morphing multibody systems, offering insights into stability-adaptability trade-offs and providing a foundation for future work on fluid-structure interaction, optimization, and digital-twin applications in underwater robotics.
