Ocean Engineering最新文献_第9页

Iris time-frequency map visual feature-based cluster matching: A universal domain adaptation method for propulsion shafting fault diagnosis 基于虹膜时频图视觉特征的聚类匹配：一种推进轴系故障诊断的通用域自适应方法

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124450

Congyue Li , Guobin Li , Pengfei Xing , Dexin Cui , Yijin Sui , Hongpeng Zhang

The propulsion shafting is a vital component of ship power systems. Timely and accurate fault diagnosis is essential for ensuring navigational safety. Domain adaptation techniques have been widely applied in intelligent fault diagnosis. However, most existing methods overlook the critical impact of input representation quality on diagnostic performance and are confined to specific domain adaptation scenarios. In practical engineering, the label space relationships between domains are often unavailable, limiting the applicability of these methods. To address these issues, this study proposes a universal domain adaptation (UniDA) method, termed the source domain category anchor-guided cluster matching network. Specifically, the network utilizes iris time-frequency maps as input, which enhances the readability of the information. A similarity criterion is formulated to cluster features of the same type, subsequently matching them to the corresponding category anchors. Moreover, an inter-class representation decoupling constraint is designed to shape a more globally discriminative feature space. Further, a distance-based detection strategy is proposed to build reliable decision boundaries between common and private categories. Experimental results on the propulsion shafting dataset validate the effectiveness of the proposed method in handling diagnostic tasks involving domain and category shifts, outperforming other state-of-the-art methods. Additionally, visualization via gradient-weighted class activation mapping indicates that the network's decision-making is grounded in physically meaningful evidence, revealing the complementarity between interpretability and transferability.

推进轴系是船舶动力系统的重要组成部分。及时准确的故障诊断是保证航行安全的关键。领域自适应技术在智能故障诊断中得到了广泛的应用。然而，大多数现有方法忽略了输入表示质量对诊断性能的关键影响，并且局限于特定的领域适应场景。在实际工程中，领域之间的标签空间关系往往是不可用的，限制了这些方法的适用性。为了解决这些问题，本研究提出了一种通用领域自适应（UniDA）方法，称为源领域类别锚定引导聚类匹配网络。具体来说，该网络利用虹膜时频图作为输入，增强了信息的可读性。我们制定了一个相似标准来聚类相同类型的特征，然后将它们与相应的类别锚点进行匹配。此外，还设计了类间表示解耦约束，以形成更具全局判别性的特征空间。进一步提出了一种基于距离的检测策略，在公共和私有类别之间建立可靠的决策边界。在推进轴数据集上的实验结果验证了所提出的方法在处理涉及领域和类别转移的诊断任务方面的有效性，优于其他最先进的方法。此外，通过梯度加权类激活映射的可视化表明，网络的决策是基于物理上有意义的证据，揭示了可解释性和可转移性之间的互补性。

{"title":"Iris time-frequency map visual feature-based cluster matching: A universal domain adaptation method for propulsion shafting fault diagnosis","authors":"Congyue Li , Guobin Li , Pengfei Xing , Dexin Cui , Yijin Sui , Hongpeng Zhang","doi":"10.1016/j.oceaneng.2026.124450","DOIUrl":"10.1016/j.oceaneng.2026.124450","url":null,"abstract":"<div><div>The propulsion shafting is a vital component of ship power systems. Timely and accurate fault diagnosis is essential for ensuring navigational safety. Domain adaptation techniques have been widely applied in intelligent fault diagnosis. However, most existing methods overlook the critical impact of input representation quality on diagnostic performance and are confined to specific domain adaptation scenarios. In practical engineering, the label space relationships between domains are often unavailable, limiting the applicability of these methods. To address these issues, this study proposes a universal domain adaptation (UniDA) method, termed the source domain category anchor-guided cluster matching network. Specifically, the network utilizes iris time-frequency maps as input, which enhances the readability of the information. A similarity criterion is formulated to cluster features of the same type, subsequently matching them to the corresponding category anchors. Moreover, an inter-class representation decoupling constraint is designed to shape a more globally discriminative feature space. Further, a distance-based detection strategy is proposed to build reliable decision boundaries between common and private categories. Experimental results on the propulsion shafting dataset validate the effectiveness of the proposed method in handling diagnostic tasks involving domain and category shifts, outperforming other state-of-the-art methods. Additionally, visualization via gradient-weighted class activation mapping indicates that the network's decision-making is grounded in physically meaningful evidence, revealing the complementarity between interpretability and transferability.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124450"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170734","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Numerical investigations of flow fields around polyethylene terephthalate (PET) nets using Improved Delayed Detached Eddy Simulation 基于改进延迟分离涡流模拟的聚对苯二甲酸乙二醇酯（PET）网流场数值研究

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124441

Gang Wang , Zhi Wang , Chang-tao Guan , Zhao-qi Jiao , Jia-jun Hu , Lian-hui Wu , Xiao-yu Qu , Fu-kun Gui , De-jun Feng

As polyethylene terephthalate (PET) nets are increasingly adopted for net systems in offshore aquaculture structures, the knowledge of its hydrodynamic characteristics remains limited. The unique hybrid smooth-twisted (HST) twines and hexagon net meshes indicate that the existing flow regimes over typical rigid or fabric nets cannot be referred to design or optimise PET nets effectively. This study compares the fluid flow properties, including time-averaged and instantaneous flow, turbulence distribution, and velocity reduction between stainless-steel net panels with fully-smooth (FS) twines and PET net panels with HST twines using Improved Delayed Detached Eddy Simulations (IDDES). It is revealed that structural features, particularly intersections and twine structures, significantly affect flow patterns. Compared to the isolated circular cylinder and the smooth twine, the double-twisted twine shows a significantly longer vortex formation length and higher base pressure. The twisting geometry alters the separation lines along the spanwise direction, resulting in three-dimensional separations. In comparison to the isolated circular cylinder and smooth twines, which effectively enhance near-wake flow stabilities, including the suppression of vortex shedding and the further spatial elongation of shear layers, as well as the decrease of turbulent kinetic energies. Moreover, the overall time-averaged velocity reduction behind the net panels with FS and HST twines drops to approximately 85 %–95 % of the freestream velocity, with similar reduction levels observed in both cases due to the strong three-dimensional mixing wake patterns. In addition, a velocity reduction formula suitable for PET net panels is proposed based on the Area-Averaged Velocity Reduction (AAVR) model and polynomial fitting. This study provides critical insights into the hydrodynamics as well as further structural optimisations of PET nets in offshore aquaculture structures.

随着聚对苯二甲酸乙二醇酯（PET）网越来越多地用于近海水产养殖结构的网系统，对其水动力特性的了解仍然有限。独特的混合平滑扭曲（HST）线和六边形网表明，现有的流动形式在典型的刚性或织物网不能有效地参考设计或优化PET网。本研究使用改进的延迟分离涡流模拟（IDDES）比较了不锈钢网板与HST网板之间的流体流动特性，包括时间平均和瞬时流动、湍流分布和速度降低。结果表明，结构特征，特别是交叉和缠绕结构，对流型有显著影响。与孤立圆柱和光滑线相比，双捻线的涡形成长度和基底压力明显增加。扭曲的几何形状改变了沿展向的分离线，导致三维分离。与孤立的圆柱和光滑的缠绕相比，它们有效地增强了近尾迹流动的稳定性，包括抑制了旋涡脱落和剪切层的进一步空间延伸，以及湍流动能的降低。此外，在有FS和HST缠绕的净板后面，总的时间平均速度下降到自由流速度的85% - 95%左右，由于强烈的三维混合尾迹模式，两种情况下观察到的下降水平相似。此外，基于面积平均速度衰减（AAVR）模型和多项式拟合，提出了适用于PET网板的速度衰减公式。这项研究为海洋水产养殖结构中PET网的流体动力学以及进一步的结构优化提供了重要的见解。

{"title":"Numerical investigations of flow fields around polyethylene terephthalate (PET) nets using Improved Delayed Detached Eddy Simulation","authors":"Gang Wang , Zhi Wang , Chang-tao Guan , Zhao-qi Jiao , Jia-jun Hu , Lian-hui Wu , Xiao-yu Qu , Fu-kun Gui , De-jun Feng","doi":"10.1016/j.oceaneng.2026.124441","DOIUrl":"10.1016/j.oceaneng.2026.124441","url":null,"abstract":"<div><div>As polyethylene terephthalate (PET) nets are increasingly adopted for net systems in offshore aquaculture structures, the knowledge of its hydrodynamic characteristics remains limited. The unique hybrid smooth-twisted (HST) twines and hexagon net meshes indicate that the existing flow regimes over typical rigid or fabric nets cannot be referred to design or optimise PET nets effectively. This study compares the fluid flow properties, including time-averaged and instantaneous flow, turbulence distribution, and velocity reduction between stainless-steel net panels with fully-smooth (FS) twines and PET net panels with HST twines using Improved Delayed Detached Eddy Simulations (IDDES). It is revealed that structural features, particularly intersections and twine structures, significantly affect flow patterns. Compared to the isolated circular cylinder and the smooth twine, the double-twisted twine shows a significantly longer vortex formation length and higher base pressure. The twisting geometry alters the separation lines along the spanwise direction, resulting in three-dimensional separations. In comparison to the isolated circular cylinder and smooth twines, which effectively enhance near-wake flow stabilities, including the suppression of vortex shedding and the further spatial elongation of shear layers, as well as the decrease of turbulent kinetic energies. Moreover, the overall time-averaged velocity reduction behind the net panels with FS and HST twines drops to approximately 85 %–95 % of the freestream velocity, with similar reduction levels observed in both cases due to the strong three-dimensional mixing wake patterns. In addition, a velocity reduction formula suitable for PET net panels is proposed based on the Area-Averaged Velocity Reduction (AAVR) model and polynomial fitting. This study provides critical insights into the hydrodynamics as well as further structural optimisations of PET nets in offshore aquaculture structures.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124441"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170738","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Motion prediction enhancement for offshore platforms under typhoon conditions using DynaSparse-ConvLSTM with dynamic sparse gating and multi-source fusion 基于动态稀疏门控和多源融合的DynaSparse-ConvLSTM增强台风条件下海上平台运动预测

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124444

Yang Chen , Lihao Yuan , Huoping Wang , Chao Luo , Zhi Zhou

To address key challenges in predicting offshore platform motion responses—including complex spatiotemporal coupling, underutilized motion periodicity, and limited real-time performance—this study proposes a dynamic sparse Convolutional Long Short-Term Memory (DynaSparse-ConvLSTM) architecture. The model introduces three major innovations: (i) a multi-source feature fusion mechanism that improves the representation of environmental dynamics and motion periodicity; (ii) a hierarchical feature-enhancement network with ReLU–Dropout activation designed to capture abrupt motion responses; and (iii) a dynamic sparse gating mechanism that reduces computational cost while maintaining predictive skill through single-layer, four-gate operations. Validation using typhoon field data from the “Lingshui 17-2” platform shows that the proposed model achieves a mean absolute error of 0.0237° under typhoon conditions—34.71 % lower than that of the RNN baseline—and obtains a coefficient of determination of 0.9117. Across all sea states, the median absolute error remains within 0.0091°–0.0273°, outperforming suboptimal models by at least 2.9 %. These results demonstrate the model's strong generalization capability and its potential as a practical prediction tool for optimizing offshore operations and supporting the development of next-generation marine equipment.

为了解决预测海上平台运动响应的关键挑战，包括复杂的时空耦合、未充分利用的运动周期性和有限的实时性能，本研究提出了一种动态稀疏卷积长短期记忆（DynaSparse-ConvLSTM）架构。该模型引入了三个主要创新：(i)一种多源特征融合机制，改善了环境动力学和运动周期性的表示；（ii）具有ReLU-Dropout激活的分层特征增强网络，旨在捕获突然运动响应；（iii）动态稀疏门控机制，通过单层四门操作降低计算成本，同时保持预测技能。利用“陵水17-2”平台的台风场数据验证，该模型在台风条件下的平均绝对误差为0.0237°，比RNN基线的平均绝对误差低34.71%，确定系数为0.9117。在所有海况下，绝对误差中位数保持在0.0091°-0.0273°之间，比次优模型至少高出2.9%。这些结果表明，该模型具有强大的泛化能力，具有作为优化海上作业和支持下一代海洋设备开发的实用预测工具的潜力。

{"title":"Motion prediction enhancement for offshore platforms under typhoon conditions using DynaSparse-ConvLSTM with dynamic sparse gating and multi-source fusion","authors":"Yang Chen , Lihao Yuan , Huoping Wang , Chao Luo , Zhi Zhou","doi":"10.1016/j.oceaneng.2026.124444","DOIUrl":"10.1016/j.oceaneng.2026.124444","url":null,"abstract":"<div><div>To address key challenges in predicting offshore platform motion responses—including complex spatiotemporal coupling, underutilized motion periodicity, and limited real-time performance—this study proposes a dynamic sparse Convolutional Long Short-Term Memory (DynaSparse-ConvLSTM) architecture. The model introduces three major innovations: (i) a multi-source feature fusion mechanism that improves the representation of environmental dynamics and motion periodicity; (ii) a hierarchical feature-enhancement network with ReLU–Dropout activation designed to capture abrupt motion responses; and (iii) a dynamic sparse gating mechanism that reduces computational cost while maintaining predictive skill through single-layer, four-gate operations. Validation using typhoon field data from the “Lingshui 17-2” platform shows that the proposed model achieves a mean absolute error of 0.0237° under typhoon conditions—34.71 % lower than that of the RNN baseline—and obtains a coefficient of determination of 0.9117. Across all sea states, the median absolute error remains within 0.0091°–0.0273°, outperforming suboptimal models by at least 2.9 %. These results demonstrate the model's strong generalization capability and its potential as a practical prediction tool for optimizing offshore operations and supporting the development of next-generation marine equipment.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124444"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170672","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Skin friction reduction on smooth and rough surfaces using air lubrication 使用空气润滑减少光滑和粗糙表面的皮肤摩擦

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124464

Jonathan L. Carter, Victoria M. Hynes, J. Travis Hunsucker

Air lubrication systems have been shown to reduce skin friction drag in both laboratory and full-scale settings, but there has been little research into the interactions between air lubrication systems and rough surfaces. These surfaces are ubiquitous to all in-service ships. A flow channel and floating element balance were designed and assembled to address 1) whether the presence of additional surface roughness significantly affects the amount of air needed for significant skin friction drag reduction and 2) whether the percent drag reduction differs between roughened and smooth surfaces after the optimal air injection rate is obtained. Three different surfaces were tested: a smooth acrylic and two roughened acrylic panels. The results show that air lubrication has the potential to decrease the frictional drag coefficients by as much as 30 % at laboratory scale bulk Reynolds numbers (Re_H ∼2.8 × 10⁴ to 3.3 × 10⁴). However, roughened surfaces required higher air injection to reach similar drag reductions. The practical implication is that the increased air injection rates will result in a decrease of ALS efficiency as the ship's hull becomes more fouled and/or coating roughness increases.

在实验室和全尺寸环境中，空气润滑系统已被证明可以减少表面摩擦阻力，但很少有关于空气润滑系统与粗糙表面之间相互作用的研究。这些表面无处不在，所有在役船舶。设计并组装了一个流道和浮动元件平衡器，以解决以下问题：1)额外表面粗糙度是否会显著影响显著减少表面摩擦阻力所需的空气量；2)在获得最佳空气喷射速率后，粗糙表面和光滑表面的阻力减少百分比是否不同。测试了三种不同的表面：光滑的丙烯酸和两个粗糙的丙烯酸面板。结果表明，在实验室尺度的体积雷诺数（ReH ~ 2.8 × 104至3.3 × 104）下，空气润滑有可能将摩擦阻力系数降低多达30%。然而，粗糙的表面需要更高的空气注入才能达到类似的阻力降低。实际意义是，随着船体变得更脏和/或涂层粗糙度增加，增加的空气喷射速率将导致ALS效率降低。

{"title":"Skin friction reduction on smooth and rough surfaces using air lubrication","authors":"Jonathan L. Carter, Victoria M. Hynes, J. Travis Hunsucker","doi":"10.1016/j.oceaneng.2026.124464","DOIUrl":"10.1016/j.oceaneng.2026.124464","url":null,"abstract":"<div><div>Air lubrication systems have been shown to reduce skin friction drag in both laboratory and full-scale settings, but there has been little research into the interactions between air lubrication systems and rough surfaces. These surfaces are ubiquitous to all in-service ships. A flow channel and floating element balance were designed and assembled to address 1) whether the presence of additional surface roughness significantly affects the amount of air needed for significant skin friction drag reduction and 2) whether the percent drag reduction differs between roughened and smooth surfaces after the optimal air injection rate is obtained. Three different surfaces were tested: a smooth acrylic and two roughened acrylic panels. The results show that air lubrication has the potential to decrease the frictional drag coefficients by as much as 30 % at laboratory scale bulk Reynolds numbers (Re<sub>H</sub> ∼2.8 × 10<sup>4</sup> to 3.3 × 10<sup>4</sup>). However, roughened surfaces required higher air injection to reach similar drag reductions. The practical implication is that the increased air injection rates will result in a decrease of ALS efficiency as the ship's hull becomes more fouled and/or coating roughness increases.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124464"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Determinants of casualty severity in maritime accidents: Evidence from text mining of 581 investigation reports, 2014–2023 海上事故伤亡严重程度的决定因素：来自581份调查报告的文本挖掘证据，2014-2023

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124456

Dongge Liu , Liang Cheng , Bo Ai , Rui Xie , Xidong Wang , Hengshuai Shang

This study develops a reproducible pipeline that transforms long-form accident reports into structured evidence on the determinants of casualty severity in China sea. Using 581 investigation reports issued by China's Maritime Safety Administration from 2014 to 2023, we combine domain-adapted MBERT models for named entity recognition and text classification with an ordered logistic regression and marginal-effects analysis to estimate how accident type, ship type, human, management, and environment factors shape casualty severity. The MBERT-BiLSTM-CRF and MBERT-Softmax components deliver high-accuracy extraction of key elements, after which ordered logistic regression and proportional-odds modeling yields interpretable effects at scale. The results show that collision (OR = 2.246, p < 0.001) and sinking (OR = 3.626, p = 0.014) markedly elevate casualty severity; fishing vessels (OR = 4.213, p = 0.029) display heightened casualty sensitivity; low visibility (OR = 4.774, p = 0.048) act as a critical situational trigger; Among human factors, alcohol use and fatigue among human factors (OR = 5.396, p = 0.049) yield the largest marginal effects; and absence of safety warnings (OR = 2.506, p = 0.019) is the most consequential management deficiencies. The proportional-lines test supports model validity. Mapping these findings into practice, we recommend targeted controls along three pathways: engineering measures for high-kinetic-energy and loss-of-stability scenarios, operational measures for detection-and-response constraints, and managerial measures that strengthen risk communication and alarm systems. The framework offers a reusable route from textual evidence to causal quantification and provides a ranked basis for regulators and operators to prioritize high-marginal-effect interventions that can most efficiently reduce casualties.

本研究开发了一个可复制的管道，将长篇事故报告转化为关于中国海上伤亡严重程度决定因素的结构化证据。利用2014年至2023年中国海事局发布的581份调查报告，我们将用于命名实体识别和文本分类的领域适应MBERT模型与有序逻辑回归和边际效应分析相结合，以估计事故类型、船型、人、管理和环境因素如何影响事故严重程度。MBERT-BiLSTM-CRF和MBERT-Softmax组件提供高精度的关键元素提取，之后有序逻辑回归和比例赔率建模产生大规模的可解释效果。结果表明，碰撞（OR = 2.246, p < 0.001）和下沉（OR = 3.626, p = 0.014）显著提高了伤亡严重程度；渔船（OR = 4.213, p = 0.029）表现出更高的伤亡敏感性；低能见度（OR = 4.774, p = 0.048）是关键情境触发因素；在人为因素中，饮酒和疲劳的边际效应最大（OR = 5.396, p = 0.049）；缺乏安全警告（OR = 2.506, p = 0.019）是最严重的管理缺陷。比例线检验支持模型的有效性。将这些发现应用到实践中，我们建议采取以下三种有针对性的控制措施：针对高动能和失稳情景的工程措施，针对检测和响应约束的操作措施，以及加强风险沟通和警报系统的管理措施。该框架提供了从文本证据到因果量化的可重用路径，并为监管机构和运营商提供了一个排序基础，以优先考虑能够最有效地减少伤亡的高边际效应干预措施。

{"title":"Determinants of casualty severity in maritime accidents: Evidence from text mining of 581 investigation reports, 2014–2023","authors":"Dongge Liu , Liang Cheng , Bo Ai , Rui Xie , Xidong Wang , Hengshuai Shang","doi":"10.1016/j.oceaneng.2026.124456","DOIUrl":"10.1016/j.oceaneng.2026.124456","url":null,"abstract":"<div><div>This study develops a reproducible pipeline that transforms long-form accident reports into structured evidence on the determinants of casualty severity in China sea. Using 581 investigation reports issued by China's Maritime Safety Administration from 2014 to 2023, we combine domain-adapted MBERT models for named entity recognition and text classification with an ordered logistic regression and marginal-effects analysis to estimate how accident type, ship type, human, management, and environment factors shape casualty severity. The MBERT-BiLSTM-CRF and MBERT-Softmax components deliver high-accuracy extraction of key elements, after which ordered logistic regression and proportional-odds modeling yields interpretable effects at scale. The results show that collision (OR = 2.246, p < 0.001) and sinking (OR = 3.626, p = 0.014) markedly elevate casualty severity; fishing vessels (OR = 4.213, p = 0.029) display heightened casualty sensitivity; low visibility (OR = 4.774, p = 0.048) act as a critical situational trigger; Among human factors, alcohol use and fatigue among human factors (OR = 5.396, p = 0.049) yield the largest marginal effects; and absence of safety warnings (OR = 2.506, p = 0.019) is the most consequential management deficiencies. The proportional-lines test supports model validity. Mapping these findings into practice, we recommend targeted controls along three pathways: engineering measures for high-kinetic-energy and loss-of-stability scenarios, operational measures for detection-and-response constraints, and managerial measures that strengthen risk communication and alarm systems. The framework offers a reusable route from textual evidence to causal quantification and provides a ranked basis for regulators and operators to prioritize high-marginal-effect interventions that can most efficiently reduce casualties.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124456"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Data-driven reduced-order prediction modeling for cylinder vortex-induced vibrations in the subcritical Reynolds number regime 亚临界雷诺数下圆柱涡激振动的数据驱动降阶预测建模

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124358

Jingyuan Wang , Li Zou , Guoqing Jin , Zongbing Yu , Maoyue Lin

Vortex-induced vibration is a common flow–structure interaction problem in ocean engineering and typically requires substantial computational effort when analyzed using full-order numerical simulations. In such cases, efficient and stable tools for systematic analysis and short-range prediction of structural responses remain limited. Based on experimental data obtained from a rigid cylinder with two degrees of freedom, this study develops a data-driven reduced-order prediction framework to reconstruct both the structural response and wake-flow evolution, while enabling motion prediction under small variations in system parameters. By integrating proper orthogonal decomposition and dynamic mode decomposition, a reduced-order model is constructed to achieve a reduced-order reconstruction of structural dynamics and vorticity fields across multiple scales. The framework is further extended by incorporating structural control equations and Tikhonov regularisation, resulting in a closed-loop predictive model applicable across neighbouring operating conditions. Results show that a 90 % energy truncation criterion allows accurate capture of the dominant structural dynamics, with reconstruction accuracy quantified using root-mean-square errors. The reduced-order model reproduces the primary wake structures observed in experiments, while the predictive framework preserves consistent response characteristics across different mass ratios and captures local variations in wake spatial scales, indicating its applicability to cross-condition prediction.

涡激振动是海洋工程中常见的流-结构相互作用问题，在全阶数值模拟中需要大量的计算量。在这种情况下，用于系统分析和短期预测结构反应的有效和稳定的工具仍然有限。基于两自由度刚性圆柱体的实验数据，本研究开发了一个数据驱动的降阶预测框架，用于重建结构响应和尾流演变，同时实现系统参数小变化下的运动预测。通过适当的正交分解和动力模态分解相结合，构建了降阶模型，实现了结构动力学和涡度场的多尺度降阶重构。该框架通过结合结构控制方程和Tikhonov正则化进一步扩展，从而形成一个适用于相邻操作条件的闭环预测模型。结果表明，90%的能量截断准则可以准确捕获主要结构动力学，重建精度使用均方根误差量化。降阶模型再现了实验中观察到的初级尾流结构，而预测框架保留了不同质量比下的一致响应特征，并捕获了尾流空间尺度的局部变化，表明其适用于交叉条件预测。

{"title":"Data-driven reduced-order prediction modeling for cylinder vortex-induced vibrations in the subcritical Reynolds number regime","authors":"Jingyuan Wang , Li Zou , Guoqing Jin , Zongbing Yu , Maoyue Lin","doi":"10.1016/j.oceaneng.2026.124358","DOIUrl":"10.1016/j.oceaneng.2026.124358","url":null,"abstract":"<div><div>Vortex-induced vibration is a common flow–structure interaction problem in ocean engineering and typically requires substantial computational effort when analyzed using full-order numerical simulations. In such cases, efficient and stable tools for systematic analysis and short-range prediction of structural responses remain limited. Based on experimental data obtained from a rigid cylinder with two degrees of freedom, this study develops a data-driven reduced-order prediction framework to reconstruct both the structural response and wake-flow evolution, while enabling motion prediction under small variations in system parameters. By integrating proper orthogonal decomposition and dynamic mode decomposition, a reduced-order model is constructed to achieve a reduced-order reconstruction of structural dynamics and vorticity fields across multiple scales. The framework is further extended by incorporating structural control equations and Tikhonov regularisation, resulting in a closed-loop predictive model applicable across neighbouring operating conditions. Results show that a 90 % energy truncation criterion allows accurate capture of the dominant structural dynamics, with reconstruction accuracy quantified using root-mean-square errors. The reduced-order model reproduces the primary wake structures observed in experiments, while the predictive framework preserves consistent response characteristics across different mass ratios and captures local variations in wake spatial scales, indicating its applicability to cross-condition prediction.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124358"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hydrodynamic shielding by distributed standalone perforated cage systems for fish farming 用于养鱼的分布式独立穿孔网箱系统的水动力屏蔽

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124389

S. Salman, K.G. Vijay

As sustainable ocean development progresses, marine systems are evolving to meet the demands of multifunctional marine operations. Distributed configurations are increasingly adopted in aquaculture due to their inherent wave-damping and load-mitigating characteristics. This study investigates gravity wave interactions with distributed array of multiple standalone modular porous cages through a combined experimental and numerical approach. The boundary value problem is formulated within the framework of linearized potential flow theory and solved using Dual Boundary Element Method, defining a nonlinear quadratic pressure-drop model across the porous interfaces. Experimental results are quantitatively validated against the present numerical solutions using the Nash-Sutcliffe Efficiency metric, and corroborated through comparison with data reported in existing literature. A systematic parametric study explores the effects of key design variables, including number of modules, porosity, wave height, and both symmetric and asymmetric inter-modular spacing, across shallow, intermediate, and deep water regimes. In deep-water conditions, the leeward modules experience reductions of approximately 5% in horizontal forces and 15% in vertical forces, attributable to shielding effects from the seaward side. In intermediate depths, inter-module resonance induces nonlinear scattering and significant force disparities, particularly at higher porosity. Asymmetric layouts with extended trailing gaps show improved shielding across various conditions. These findings provide practical guidance for designing distributed aquaculture systems and related gridded wave energy mitigation strategies, emphasizing the importance of cage configuration and porosity in enhancing hydrodynamic performance and structural safety.

随着海洋可持续发展的推进，海洋系统正在不断发展，以满足多功能海洋作业的需求。由于其固有的减波和减载特性，分布式配置在水产养殖中越来越多地被采用。本文采用实验与数值相结合的方法研究了重力波与多个独立模块化多孔笼分布阵列的相互作用。边界值问题在线性化势流理论框架内建立，采用双边界元法求解，定义了多孔界面上的非线性二次压降模型。实验结果使用Nash-Sutcliffe效率度量对目前的数值解进行了定量验证，并通过与现有文献报道的数据进行了比较。系统的参数化研究探讨了关键设计变量的影响，包括模块数量、孔隙度、波高、对称和非对称模块间距，适用于浅水、中水和深水区域。在深水条件下，由于向海侧的屏蔽作用，下风模块的水平力降低了约5%，垂直力降低了15%。在中等深度，模间共振引起非线性散射和显著的力差，特别是在高孔隙率下。延长尾隙的非对称布局在各种条件下都具有更好的屏蔽效果。这些发现为分布式养殖系统的设计和相关的网格化波浪能缓解策略提供了实用指导，强调了笼形和孔隙度对提高水动力性能和结构安全性的重要性。

{"title":"Hydrodynamic shielding by distributed standalone perforated cage systems for fish farming","authors":"S. Salman, K.G. Vijay","doi":"10.1016/j.oceaneng.2026.124389","DOIUrl":"10.1016/j.oceaneng.2026.124389","url":null,"abstract":"<div><div>As sustainable ocean development progresses, marine systems are evolving to meet the demands of multifunctional marine operations. Distributed configurations are increasingly adopted in aquaculture due to their inherent wave-damping and load-mitigating characteristics. This study investigates gravity wave interactions with distributed array of multiple standalone modular porous cages through a combined experimental and numerical approach. The boundary value problem is formulated within the framework of linearized potential flow theory and solved using Dual Boundary Element Method, defining a nonlinear quadratic pressure-drop model across the porous interfaces. Experimental results are quantitatively validated against the present numerical solutions using the Nash-Sutcliffe Efficiency metric, and corroborated through comparison with data reported in existing literature. A systematic parametric study explores the effects of key design variables, including number of modules, porosity, wave height, and both symmetric and asymmetric inter-modular spacing, across shallow, intermediate, and deep water regimes. In deep-water conditions, the leeward modules experience reductions of approximately 5% in horizontal forces and 15% in vertical forces, attributable to shielding effects from the seaward side. In intermediate depths, inter-module resonance induces nonlinear scattering and significant force disparities, particularly at higher porosity. Asymmetric layouts with extended trailing gaps show improved shielding across various conditions. These findings provide practical guidance for designing distributed aquaculture systems and related gridded wave energy mitigation strategies, emphasizing the importance of cage configuration and porosity in enhancing hydrodynamic performance and structural safety.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124389"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170673","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A hybrid model based on a dual-layer decomposition framework and LSTM-Informer for significant wave height prediction 基于双层分解框架和LSTM-Informer的混合波高预测模型

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124412

Shuqing Xiong, Wenxiu Yang, Chunhui Dong

An accurate forecast of significant wave height (SWH) is essential for ocean engineering projects, maritime navigation, and the exploitation of marine energy. However, due to the non-linearities and non-smoothness of ocean waves, conventional models often underperform in SWH forecasting tasks. This paper proposes a hybrid model that combines STL decomposition, variational mode decomposition (VMD), long short-term memory (LSTM), and Informer. This study uses datasets from three buoy stations to compare the performance of several models, including LSTM, Transformer, EMD-LSTM, EMD-Transformer, VMD-LSTM, and SMD-LSTM-Informer (our proposed model), at different forecast durations: 3-h, 6-h, 12-h, 24-h, 36-h, and 48-h ahead. The research on forecasting performance across three stations indicates that the SMD-LSTM-Informer model outperforms others, achieving an average R² value of 0.87 over 48-h ahead. In addition, ablation experiments are conducted using a fourth dataset to demonstrate the effectiveness and robustness of the SMD-LSTM-Informer model. The results indicate the SMD-LSTM-Informer model exhibits superiority in both short- and long-term SWH forecasting tasks.

有效波高（SWH）的准确预报对海洋工程、海上航行和海洋能源开发具有重要意义。然而，由于海浪的非线性和非平滑性，传统模型在SWH预测任务中往往表现不佳。本文提出了一种结合STL分解、变分模态分解、长短期记忆（LSTM）和信息源的混合模型。本研究使用来自三个浮标站的数据集，比较了LSTM、Transformer、EMD-LSTM、EMD-Transformer、VMD-LSTM和smd -LSTM- inforformer（我们提出的模型）在不同预测时间（提前3小时、6小时、12小时、24小时、36小时和48小时）下的性能。对3个台站的预测性能研究表明，SMD-LSTM-Informer模型的预测性能优于其他模型，提前48 h的平均R2值为0.87。此外，利用第四个数据集进行了烧蚀实验，以验证SMD-LSTM-Informer模型的有效性和鲁棒性。结果表明，SMD-LSTM-Informer模型在短期和长期SWH预测任务中都表现出优越性。

{"title":"A hybrid model based on a dual-layer decomposition framework and LSTM-Informer for significant wave height prediction","authors":"Shuqing Xiong, Wenxiu Yang, Chunhui Dong","doi":"10.1016/j.oceaneng.2026.124412","DOIUrl":"10.1016/j.oceaneng.2026.124412","url":null,"abstract":"<div><div>An accurate forecast of significant wave height (SWH) is essential for ocean engineering projects, maritime navigation, and the exploitation of marine energy. However, due to the non-linearities and non-smoothness of ocean waves, conventional models often underperform in SWH forecasting tasks. This paper proposes a hybrid model that combines STL decomposition, variational mode decomposition (VMD), long short-term memory (LSTM), and Informer. This study uses datasets from three buoy stations to compare the performance of several models, including LSTM, Transformer, EMD-LSTM, EMD-Transformer, VMD-LSTM, and SMD-LSTM-Informer (our proposed model), at different forecast durations: 3-h, 6-h, 12-h, 24-h, 36-h, and 48-h ahead. The research on forecasting performance across three stations indicates that the SMD-LSTM-Informer model outperforms others, achieving an average R<sup>2</sup> value of 0.87 over 48-h ahead. In addition, ablation experiments are conducted using a fourth dataset to demonstrate the effectiveness and robustness of the SMD-LSTM-Informer model. The results indicate the SMD-LSTM-Informer model exhibits superiority in both short- and long-term SWH forecasting tasks.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124412"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170674","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Effects of the slope-acceleration on the wake structure of a wall-mounted rectangular cylinder in three-dimensional flow 三维流动中坡加速度对壁挂式矩形圆柱尾迹结构的影响

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124497

Zengshun Chen (陈增顺) , Peng Hu (胡鹏) , Yemeng Xu (许叶萌) , Zhihang Zhao (赵智航) , Yunfei Fu (付云飞) , Zheyu Zhang (张哲宇)

Wall boundary conditions significantly influence the flow around finite-length three-dimensional prismatic structures, particularly in offshore engineering where these structures face complex wind fields. This study combines wind tunnel experimental data and large eddy simulation (LES) methods to investigate the time-averaged and instantaneous wake evolution of a wall-mounted square prism (height-to-width ratio 8:1) on 45° sloped terrain with varying slope-top distances. Compared to flat terrain, the sloped terrain accelerated the flow, significantly altering the aerodynamic and flow field characteristics at the bottom and upper-middle parts of the prism. This caused the disappearance of the windward stagnation point, dissipation of the bottom horseshoe vortex, and a shift in the pressure distribution on the side bottom to a "quasi-two-dimensional" form. The tip vortex significantly strengthened and extended downward, forming a double-layer vortex system. As the prism moved further from the slope crest, the wake gradually reverted to the flow mode of flat terrain, evolving from a dipole structure to a quadrupole structure. This study clarifies the interference mechanism of sloped terrain on the flow around wall-mounted blunt bodies, emphasizing terrain effects on vortex structure evolution and offering theoretical insights for optimizing aerodynamic design in offshore engineering under complex terrain conditions.

壁面边界条件显著影响有限长三维棱柱结构的流动，特别是在海上工程中，这些结构面临复杂的风场。本研究结合风洞实验数据和大涡模拟（LES）方法，研究了45°倾斜地形上不同坡顶距离的壁挂式方形棱镜（高宽比为8:1）的时均和瞬时尾迹演化。与平坦地形相比，倾斜地形加速了流动，显著改变了棱镜底部和中上部分的气动特性和流场特性。这导致迎风滞止点消失，底部马蹄形涡耗散，侧底部压力分布向“准二维”形式转移。叶尖涡明显增强并向下扩展，形成双层涡系统。随着棱柱离坡顶越远，尾迹逐渐恢复到平坦地形的流动模式，由偶极子结构演变为四极子结构。本研究阐明了斜坡地形对壁挂钝体绕流的干扰机理，强调了地形对旋涡结构演变的影响，为复杂地形条件下海洋工程气动优化设计提供了理论依据。

{"title":"Effects of the slope-acceleration on the wake structure of a wall-mounted rectangular cylinder in three-dimensional flow","authors":"Zengshun Chen (陈增顺) , Peng Hu (胡鹏) , Yemeng Xu (许叶萌) , Zhihang Zhao (赵智航) , Yunfei Fu (付云飞) , Zheyu Zhang (张哲宇)","doi":"10.1016/j.oceaneng.2026.124497","DOIUrl":"10.1016/j.oceaneng.2026.124497","url":null,"abstract":"<div><div>Wall boundary conditions significantly influence the flow around finite-length three-dimensional prismatic structures, particularly in offshore engineering where these structures face complex wind fields. This study combines wind tunnel experimental data and large eddy simulation (LES) methods to investigate the time-averaged and instantaneous wake evolution of a wall-mounted square prism (height-to-width ratio 8:1) on 45° sloped terrain with varying slope-top distances. Compared to flat terrain, the sloped terrain accelerated the flow, significantly altering the aerodynamic and flow field characteristics at the bottom and upper-middle parts of the prism. This caused the disappearance of the windward stagnation point, dissipation of the bottom horseshoe vortex, and a shift in the pressure distribution on the side bottom to a \"quasi-two-dimensional\" form. The tip vortex significantly strengthened and extended downward, forming a double-layer vortex system. As the prism moved further from the slope crest, the wake gradually reverted to the flow mode of flat terrain, evolving from a dipole structure to a quadrupole structure. This study clarifies the interference mechanism of sloped terrain on the flow around wall-mounted blunt bodies, emphasizing terrain effects on vortex structure evolution and offering theoretical insights for optimizing aerodynamic design in offshore engineering under complex terrain conditions.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124497"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170613","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Towards operational safety in maritime transportation: a neurophysiological workload measurement using deep learning 迈向海上运输的操作安全：使用深度学习的神经生理工作量测量

IF 5.5 2区工程技术 Q1 ENGINEERING, CIVIL

Ocean Engineering

Pub Date : 2026-02-02 DOI: 10.1016/j.oceaneng.2026.124399

Ruonan Jiang , Shiqi Fan

Human factors account for 70 %–90 % of maritime accidents, with mental workload (MWL) being a significant risk element. Current assessments of seafarer MWL lack objective neurophysiological measures, restricting accurate monitoring and proactive safety measures. This study employed electroencephalography (EEG) data from 10 crew members during simulated navigation tasks to introduce an EEG-based MWL measurement framework for simulated navigation tasks, aimed at monitoring cognitive states and supporting maritime safety management. Three innovations are proposed: (1) a maritime-specific EEG index indicating task-related cognitive demand; (2) a neurobehavioral link between EEG metrics and observed operational errors, showing how high MWL relates to operational safety; (3) a hybrid convolutional neural network and bidirectional long short-term memory (CNN-BiLSTM) model for classifying MWL states. The model provides objective assessments by identifying key EEG bands with Shapley Additive Explanations (SHAP). In a cross-subject validation, the model achieved an AUC of 0.94, demonstrating its ability to generalize across different seafarers. These findings illustrate how EEG-derived workload assessment can inform data-driven crew management and training strategies, providing a neurophysiological foundation that may support enhancing maritime safety from a human factor perspective.

人为因素占海上事故的70% - 90%，其中精神负荷（MWL）是一个重要的风险因素。目前对海员MWL的评估缺乏客观的神经生理指标，限制了准确的监测和主动的安全措施。本研究利用10名船员在模拟导航任务中的脑电图（EEG）数据，介绍了一种基于脑电图的模拟导航任务MWL测量框架，旨在监测认知状态，为海上安全管理提供支持。本文提出了三个创新点：(1)基于任务相关认知需求的海事EEG指标；(2) EEG指标与观察到的操作错误之间存在神经行为联系，表明MWL高与操作安全有关；(3)基于卷积神经网络和双向长短期记忆（CNN-BiLSTM）混合模型的MWL状态分类。该模型通过Shapley加性解释（Shapley Additive explanatory， SHAP）识别关键EEG波段，提供客观评价。在跨主题验证中，该模型的AUC为0.94，表明其具有跨不同海员的泛化能力。这些发现说明了脑电图衍生的工作量评估如何为数据驱动的船员管理和培训策略提供信息，为从人为因素的角度支持加强海上安全提供神经生理学基础。

{"title":"Towards operational safety in maritime transportation: a neurophysiological workload measurement using deep learning","authors":"Ruonan Jiang , Shiqi Fan","doi":"10.1016/j.oceaneng.2026.124399","DOIUrl":"10.1016/j.oceaneng.2026.124399","url":null,"abstract":"<div><div>Human factors account for 70 %–90 % of maritime accidents, with mental workload (MWL) being a significant risk element. Current assessments of seafarer MWL lack objective neurophysiological measures, restricting accurate monitoring and proactive safety measures. This study employed electroencephalography (EEG) data from 10 crew members during simulated navigation tasks to introduce an EEG-based MWL measurement framework for simulated navigation tasks, aimed at monitoring cognitive states and supporting maritime safety management. Three innovations are proposed: (1) a maritime-specific EEG index indicating task-related cognitive demand; (2) a neurobehavioral link between EEG metrics and observed operational errors, showing how high MWL relates to operational safety; (3) a hybrid convolutional neural network and bidirectional long short-term memory (CNN-BiLSTM) model for classifying MWL states. The model provides objective assessments by identifying key EEG bands with Shapley Additive Explanations (SHAP). In a cross-subject validation, the model achieved an AUC of 0.94, demonstrating its ability to generalize across different seafarers. These findings illustrate how EEG-derived workload assessment can inform data-driven crew management and training strategies, providing a neurophysiological foundation that may support enhancing maritime safety from a human factor perspective.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124399"},"PeriodicalIF":5.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170670","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0