Pub Date : 2026-01-24DOI: 10.1016/j.oceaneng.2026.124195
E. Baris Ondes , Cornel Sultan , James H. Vanzwieten
This paper presents an optimization framework for enhancing power generation in ocean current turbine (OCT) farms by arranging turbines within a defined spatial area. The turbines are anchored to the ocean floor and dynamically positioned to capture maximum energy from ocean currents. The optimization process accounts for turbine wake interactions, which can reduce efficiency if not properly managed. A Particle Swarm Optimization (PSO) algorithm is used to determine the turbine layout that maximizes the farm’s average power output within the constrained domain. By integrating a wake model into the optimization loop, the framework significantly improves the farm’s average power output, yielding power gains of 41–63% across arrays of 9, 16, and 25 turbines. This approach offers a fast and reliable solution for maximizing energy production, providing valuable insights into optimal turbine density and placement for future OCT farm designs.
{"title":"Power generation maximization framework with particle swarm optimization for ocean current turbine farms","authors":"E. Baris Ondes , Cornel Sultan , James H. Vanzwieten","doi":"10.1016/j.oceaneng.2026.124195","DOIUrl":"10.1016/j.oceaneng.2026.124195","url":null,"abstract":"<div><div>This paper presents an optimization framework for enhancing power generation in ocean current turbine (OCT) farms by arranging turbines within a defined spatial area. The turbines are anchored to the ocean floor and dynamically positioned to capture maximum energy from ocean currents. The optimization process accounts for turbine wake interactions, which can reduce efficiency if not properly managed. A Particle Swarm Optimization (PSO) algorithm is used to determine the turbine layout that maximizes the farm’s average power output within the constrained domain. By integrating a wake model into the optimization loop, the framework significantly improves the farm’s average power output, yielding power gains of 41–63% across arrays of 9, 16, and 25 turbines. This approach offers a fast and reliable solution for maximizing energy production, providing valuable insights into optimal turbine density and placement for future OCT farm designs.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124195"},"PeriodicalIF":5.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080318","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}
Pub Date : 2026-01-24DOI: 10.1016/j.oceaneng.2026.124302
Liu Yang , Yapeng Lyu , Luyao Li , Yue Ma , Qing Liu
Seafarer fatigue, stemming from monotonous navigational tasks and high work pressure, significantly increases accident risk. Existing studies often lack accuracy and reliability, relying on single-modal, simulated data. To fill this gap, this study conducted a 24-day real navigation experiment, collecting physiological (EEG, EDA, ECG) and psychological (Psych) data from 24 seafarers, yielding 212 labeled samples. Next, a total of 32-dimensional fatigue features were extracted from the multi-modal data, and a feature layer fusion strategy was proposed. Eight machine learning algorithms (including DT, KNN, SVM, ANN, RF, AdaBoost, XGBoost, and LightGBM) were then used to establish the multi-modal fatigue recognition model. The dataset was split 7:3 (train/test), with class imbalance corrected using SMOTE. Model performance was subsequently evaluated on the held-out test set using accuracy, precision, recall, and F1-score as primary indicators. A thorough comparison between single-modal, bi-modal, and multi-modal situations was conducted. The results indicated that the multi-modal approach (integrating EEG, EDA, ECG, and Psych) significantly outperforms other methods. The LightGBM model achieved a maximum accuracy of 95.93 %. This study contributes to more effective fatigue detection, enhancing seafarer management and navigation safety.
{"title":"From single-modal to multi-modal: How does multi-modal data integration enhance the precision of seafarer fatigue detection?","authors":"Liu Yang , Yapeng Lyu , Luyao Li , Yue Ma , Qing Liu","doi":"10.1016/j.oceaneng.2026.124302","DOIUrl":"10.1016/j.oceaneng.2026.124302","url":null,"abstract":"<div><div>Seafarer fatigue, stemming from monotonous navigational tasks and high work pressure, significantly increases accident risk. Existing studies often lack accuracy and reliability, relying on single-modal, simulated data. To fill this gap, this study conducted a 24-day real navigation experiment, collecting physiological (EEG, EDA, ECG) and psychological (Psych) data from 24 seafarers, yielding 212 labeled samples. Next, a total of 32-dimensional fatigue features were extracted from the multi-modal data, and a feature layer fusion strategy was proposed. Eight machine learning algorithms (including DT, KNN, SVM, ANN, RF, AdaBoost, XGBoost, and LightGBM) were then used to establish the multi-modal fatigue recognition model. The dataset was split 7:3 (train/test), with class imbalance corrected using SMOTE. Model performance was subsequently evaluated on the held-out test set using accuracy, precision, recall, and F1-score as primary indicators. A thorough comparison between single-modal, bi-modal, and multi-modal situations was conducted. The results indicated that the multi-modal approach (integrating EEG, EDA, ECG, and Psych) significantly outperforms other methods. The LightGBM model achieved a maximum accuracy of 95.93 %. This study contributes to more effective fatigue detection, enhancing seafarer management and navigation safety.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124302"},"PeriodicalIF":5.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036800","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}
Pub Date : 2026-01-24DOI: 10.1016/j.oceaneng.2026.124416
Lei Sun , Jun Wang , Zihao Li , Zilu Jiao , Yuxiang Ma
Reliable full directional spectra are critical for hull-load assessment, seakeeping, and route optimization. Among spectral characteristics, directional spreading has long been one of the most difficult quantities to predict accurately. This paper proposes a two-stage conditional diffusion framework that first synthesizes unit-energy spectral shapes via a cross-attention Denoiser and then regresses the total energy, thereby achieving an explicit decoupling of energy and shape to ensure physical consistency. Using a simulation dataset for evaluation, the study compare against two Generative Adversarial Networks (GAN) baselines. Results show that Diffusion Model (DM) achieves a marked improvement in mean wave directional spread (MAE from 0.111 to 0.030, from 0.237 to 0.934). Experimental results demonstrate that diffusion model also consistently outperforms GAN-based baselines across multiple complementary evaluation metrics. These gains indicate that the method provides a more reliable generator for directional spectra in engineering workflows.
{"title":"Spectra-Diffusion: A physics-consistent two-stage framework for directional wave spectrum prediction","authors":"Lei Sun , Jun Wang , Zihao Li , Zilu Jiao , Yuxiang Ma","doi":"10.1016/j.oceaneng.2026.124416","DOIUrl":"10.1016/j.oceaneng.2026.124416","url":null,"abstract":"<div><div>Reliable full directional spectra are critical for hull-load assessment, seakeeping, and route optimization. Among spectral characteristics, directional spreading has long been one of the most difficult quantities to predict accurately. This paper proposes a two-stage conditional diffusion framework that first synthesizes unit-energy spectral shapes via a cross-attention Denoiser and then regresses the total energy, thereby achieving an explicit decoupling of energy and shape to ensure physical consistency. Using a simulation dataset for evaluation, the study compare against two Generative Adversarial Networks (GAN) baselines. Results show that Diffusion Model (DM) achieves a marked improvement in mean wave directional spread <span><math><mrow><mo>(</mo><msub><mi>σ</mi><mi>s</mi></msub><mo>)</mo></mrow></math></span> (MAE from 0.111 to 0.030, <span><math><mrow><msup><mi>R</mi><mn>2</mn></msup></mrow></math></span> from 0.237 to 0.934). Experimental results demonstrate that diffusion model also consistently outperforms GAN-based baselines across multiple complementary evaluation metrics. These gains indicate that the method provides a more reliable generator for directional spectra in engineering workflows.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124416"},"PeriodicalIF":5.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036923","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}
Pub Date : 2026-01-23DOI: 10.1016/j.oceaneng.2026.124359
Lingzhen Li , Kaiming Bi , Xiao-Ling Zhao
Floating cities built on modular floating structures (MFSs) offer a promising solution to population growth and land scarcity, particularly in coastal regions. Although numerous studies have focused on the hydrodynamic behavior of MFSs through experimentation and simulation, a simple and effective static design model, particularly suitable for the early design stage, is still lacking. To bridge this gap, the current study develops an analytical model to evaluate the static behavior of finite MFSs with arbitrary floater numbers and sizes, addressing deflection, inclination, connector shear force, and connector bending moment. The model is constructed by superposing one infinite MFS and two semi-infinite MFSs, assuming rigid floaters and flexible connectors. Verification against numerical simulations confirms the accuracy of the finite MFS model. Notably, MFSs shorter than a quarter of their characteristic length can be approximated as rigid bodies. Furthermore, applications of the developed model to floating cities and bridges are demonstrated. The current study reveals the mechanism behind MFS static behavior.
{"title":"Analytical model for static behavior of modular floating structures (MFSs) with arbitrary floater numbers and sizes","authors":"Lingzhen Li , Kaiming Bi , Xiao-Ling Zhao","doi":"10.1016/j.oceaneng.2026.124359","DOIUrl":"10.1016/j.oceaneng.2026.124359","url":null,"abstract":"<div><div>Floating cities built on modular floating structures (MFSs) offer a promising solution to population growth and land scarcity, particularly in coastal regions. Although numerous studies have focused on the hydrodynamic behavior of MFSs through experimentation and simulation, a simple and effective static design model, particularly suitable for the early design stage, is still lacking. To bridge this gap, the current study develops an analytical model to evaluate the static behavior of finite MFSs with arbitrary floater numbers and sizes, addressing deflection, inclination, connector shear force, and connector bending moment. The model is constructed by superposing one infinite MFS and two semi-infinite MFSs, assuming rigid floaters and flexible connectors. Verification against numerical simulations confirms the accuracy of the finite MFS model. Notably, MFSs shorter than a quarter of their characteristic length can be approximated as rigid bodies. Furthermore, applications of the developed model to floating cities and bridges are demonstrated. The current study reveals the mechanism behind MFS static behavior.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124359"},"PeriodicalIF":5.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036731","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}
Pub Date : 2026-01-23DOI: 10.1016/j.oceaneng.2026.124335
Mingji Zhao , Zhengjun Han , Guijie Shi , Deyu Wang
The structural integrity of ship grillages under compressive loads is critical for overall hull girder strength, necessitating accurate prediction of their ultimate bearing capacity and failure mechanisms. Facing the demand of the lightweight design and reliability of ship structure, this study carried out the longitudinal in-plane compression ultimate strength test on an AH36 steel I-type mezzanine plate frame, which reached a measured peak load of 6975.27 kN and identified its failure modes. A full-size complete shell-element model was used for the numerical study, and the ultimate load from the finite element analysis was 7307.41 kN (+4.76 %), which, however, failed to reflect the cracking phenomenon in the joint area observed in the test. To address this, a coupled shell-solid model was developed to ensure the computational efficiency and accuracy of the results. In this model, the extended NH-GTN (Nahshon-Hutchinson Gurson-Tvergaard-Needleman) damage model with shear modification is introduced into the weld elements. The time sequence parameter ρ and the critical margin CLI are proposed to quantify the sequential relationship between weld cracking and overall stability failure. The study on longitudinal continuity of the core layer shows that the ultimate strength decreases by about 5.75 % compared to the ideal case when complete discontinuity exists along the longitudinal direction. The coupled model can reproduce the evolution of "overall buckling followed by weld cracking" and reveals that cracking occurs after the peak load (i.e., ρ > 1, 0<CLI<1). Accordingly, a simplified criterion is proposed for the peak damage ratio. The results of this study provide a quantitative basis for evaluating the ultimate load capacity of I-type sandwich panels and optimizing connection details.
{"title":"Ultimate strength assessment of I-type sandwich panel considering coupled buckling and weld fracture failure","authors":"Mingji Zhao , Zhengjun Han , Guijie Shi , Deyu Wang","doi":"10.1016/j.oceaneng.2026.124335","DOIUrl":"10.1016/j.oceaneng.2026.124335","url":null,"abstract":"<div><div>The structural integrity of ship grillages under compressive loads is critical for overall hull girder strength, necessitating accurate prediction of their ultimate bearing capacity and failure mechanisms. Facing the demand of the lightweight design and reliability of ship structure, this study carried out the longitudinal in-plane compression ultimate strength test on an AH36 steel I-type mezzanine plate frame, which reached a measured peak load of 6975.27 kN and identified its failure modes. A full-size complete shell-element model was used for the numerical study, and the ultimate load from the finite element analysis was 7307.41 kN (+4.76 %), which, however, failed to reflect the cracking phenomenon in the joint area observed in the test. To address this, a coupled shell-solid model was developed to ensure the computational efficiency and accuracy of the results. In this model, the extended NH-GTN (Nahshon-Hutchinson Gurson-Tvergaard-Needleman) damage model with shear modification is introduced into the weld elements. The time sequence parameter <em>ρ</em> and the critical margin <em>CLI</em> are proposed to quantify the sequential relationship between weld cracking and overall stability failure. The study on longitudinal continuity of the core layer shows that the ultimate strength decreases by about 5.75 % compared to the ideal case when complete discontinuity exists along the longitudinal direction. The coupled model can reproduce the evolution of \"overall buckling followed by weld cracking\" and reveals that cracking occurs after the peak load (i.e., <em>ρ</em> > 1, 0<<em>CLI</em><1). Accordingly, a simplified criterion is proposed for the peak damage ratio. The results of this study provide a quantitative basis for evaluating the ultimate load capacity of I-type sandwich panels and optimizing connection details.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124335"},"PeriodicalIF":5.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036733","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}
Pub Date : 2026-01-23DOI: 10.1016/j.oceaneng.2026.124383
Jinbo Chen , Zhuo Wang , Chao Tang , Zhaolong Han , Xing Tao , Yunliang Shao , Xiaoni Wu
Torpedo anchors offer a fast and cost-effective solution for floating wind turbine moorings. However, the application of torpedo anchors outside Brazil remains limited due to insufficient design guidance and uncertainty in model parameters. Therefore, the key objective of this paper is to critically assess and provide appropriate model parameters for practical design, including the hydrodynamic and geotechnical models. The paper presents the hydrodynamic model, the penetration model, and the holding capacity model separately with both existing data in the literature and new data. The key findings are: (1) an equivalent system drag coefficient of 1.0 can be used in the hydrodynamic model for predicting the anchor impact velocity in water; (2) a soil strain-rate parameter of 0.123 and an added mass coefficient of 2.0 can be used in the penetration model for predicting the anchor final embedment; (3) a closed-form failure envelope under inclined loading is assessed to be appropriate for the anchor holding capacity. Uncertainties observed in laboratory tests and field installations during the various anchor design stages are covered by the suggested low to high estimates of the model parameters for practical design. The paper ends with discussions and recommendations for current practice and future studies.
{"title":"An assessment of model parameters for offshore torpedo anchors","authors":"Jinbo Chen , Zhuo Wang , Chao Tang , Zhaolong Han , Xing Tao , Yunliang Shao , Xiaoni Wu","doi":"10.1016/j.oceaneng.2026.124383","DOIUrl":"10.1016/j.oceaneng.2026.124383","url":null,"abstract":"<div><div>Torpedo anchors offer a fast and cost-effective solution for floating wind turbine moorings. However, the application of torpedo anchors outside Brazil remains limited due to insufficient design guidance and uncertainty in model parameters. Therefore, the key objective of this paper is to critically assess and provide appropriate model parameters for practical design, including the hydrodynamic and geotechnical models. The paper presents the hydrodynamic model, the penetration model, and the holding capacity model separately with both existing data in the literature and new data. The key findings are: (1) an equivalent system drag coefficient of 1.0 can be used in the hydrodynamic model for predicting the anchor impact velocity in water; (2) a soil strain-rate parameter of 0.123 and an added mass coefficient of 2.0 can be used in the penetration model for predicting the anchor final embedment; (3) a closed-form failure envelope under inclined loading is assessed to be appropriate for the anchor holding capacity. Uncertainties observed in laboratory tests and field installations during the various anchor design stages are covered by the suggested low to high estimates of the model parameters for practical design. The paper ends with discussions and recommendations for current practice and future studies.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124383"},"PeriodicalIF":5.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036818","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}
Pub Date : 2026-01-23DOI: 10.1016/j.oceaneng.2026.124348
Shibo Wu , Baoping Cai , Yiqin Fu , Yixin Zhao , Xiaoyan Shao , Chuntan Gao
Dynamic positioning (DP) system failures during deep-sea operations can lead to severe accidents, including blowouts and environmental damage. Performing reliable failure mode and effects analysis (FMEA) is therefore essential for ensuring the safety of offshore assets. However, traditional FMEA struggles with the inherent complexity and uncertainty of DP systems in the marine environment. To address these gaps, this paper proposes an enhanced T-spherical fuzzy FMEA model integrating systems theory and artificial intelligence techniques. The systems theory process analysis method is introduced to transcend the component-centric perspective, and effectively capture system-level hidden failure modes. The artificial intelligence enhanced risk perception data fusion and weight allocation method is constructed based on T-spherical fuzzy theory, to address cognitive uncertainty and overcome subjectivity drawbacks. A robust ranking and classification framework combining the alternative-by-alternative comparison method with K-means clustering to prevent ranking reversal and enable automatic risk grading. A case study of riserless light well intervention vessels shows that the systemic failures and control loop failures, rather than isolated hardware failures, constitute the primary risks in modern DP systems. Furthermore, the results indicate that the proposed model exhibits strong robustness and practical applicability, providing a valuable decision-support tool for safety management of complex deep-sea equipment.
{"title":"A novel failure mode and effects analysis model enhanced with systems theory and artificial intelligence for dynamic positioning systems in offshore operations","authors":"Shibo Wu , Baoping Cai , Yiqin Fu , Yixin Zhao , Xiaoyan Shao , Chuntan Gao","doi":"10.1016/j.oceaneng.2026.124348","DOIUrl":"10.1016/j.oceaneng.2026.124348","url":null,"abstract":"<div><div>Dynamic positioning (DP) system failures during deep-sea operations can lead to severe accidents, including blowouts and environmental damage. Performing reliable failure mode and effects analysis (FMEA) is therefore essential for ensuring the safety of offshore assets. However, traditional FMEA struggles with the inherent complexity and uncertainty of DP systems in the marine environment. To address these gaps, this paper proposes an enhanced T-spherical fuzzy FMEA model integrating systems theory and artificial intelligence techniques. The systems theory process analysis method is introduced to transcend the component-centric perspective, and effectively capture system-level hidden failure modes. The artificial intelligence enhanced risk perception data fusion and weight allocation method is constructed based on T-spherical fuzzy theory, to address cognitive uncertainty and overcome subjectivity drawbacks. A robust ranking and classification framework combining the alternative-by-alternative comparison method with K-means clustering to prevent ranking reversal and enable automatic risk grading. A case study of riserless light well intervention vessels shows that the systemic failures and control loop failures, rather than isolated hardware failures, constitute the primary risks in modern DP systems. Furthermore, the results indicate that the proposed model exhibits strong robustness and practical applicability, providing a valuable decision-support tool for safety management of complex deep-sea equipment.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124348"},"PeriodicalIF":5.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036922","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}
Pub Date : 2026-01-23DOI: 10.1016/j.oceaneng.2026.124404
Yixiao Luan , Xiaowei Tang , Yubin Ren , Yuxia Hu , Xingxing Wang , Zaijin You
Suction caissons serve as a common foundation solution in deep-sea offshore engineering, where ensuring accurate uplift resistance predictions is critical to structural integrity and economic feasibility. This paper develops a prediction framework based on a modified stacking model, which synergizes engineering knowledge with data-driven methods under drained sand conditions. A range of empirical models and machine learning algorithms were first systematically evaluated to pinpoint the most effective individual predictors. The highest-performing empirical expression was subsequently embedded as an additional model input to enhance transparency and improve predictive outcomes. Prediction results reveal that the stacking models achieve marked improvements over single machine learning models, with maximum reduction in mean absolute percentage error (MAPE) of 14.43 %, and increase in the coefficient of determination (R2) of 7.28 %, depending on the adopted sampling strategy. Among the various learning methods, deep neural networks (DNNs) consistently exhibited superior performance compared to tree-based algorithms. The developed hybrid approach captures complex nonlinear dependencies linked to caisson geometry and addresses the shortcomings of single-model predictions. By fusing physical principles with machine learning, the framework strengthens both predictive power and model robustness, offering a reliable and interpretable solution for uplift capacity estimation of suction caissons.
{"title":"Prediction model development of suction caisson bearing capacity under drained condition employing modified stacking method","authors":"Yixiao Luan , Xiaowei Tang , Yubin Ren , Yuxia Hu , Xingxing Wang , Zaijin You","doi":"10.1016/j.oceaneng.2026.124404","DOIUrl":"10.1016/j.oceaneng.2026.124404","url":null,"abstract":"<div><div>Suction caissons serve as a common foundation solution in deep-sea offshore engineering, where ensuring accurate uplift resistance predictions is critical to structural integrity and economic feasibility. This paper develops a prediction framework based on a modified stacking model, which synergizes engineering knowledge with data-driven methods under drained sand conditions. A range of empirical models and machine learning algorithms were first systematically evaluated to pinpoint the most effective individual predictors. The highest-performing empirical expression was subsequently embedded as an additional model input to enhance transparency and improve predictive outcomes. Prediction results reveal that the stacking models achieve marked improvements over single machine learning models, with maximum reduction in mean absolute percentage error (<em>MAPE</em>) of 14.43 %, and increase in the coefficient of determination (<em>R</em><sup>2</sup>) of 7.28 %, depending on the adopted sampling strategy. Among the various learning methods, deep neural networks (DNNs) consistently exhibited superior performance compared to tree-based algorithms. The developed hybrid approach captures complex nonlinear dependencies linked to caisson geometry and addresses the shortcomings of single-model predictions. By fusing physical principles with machine learning, the framework strengthens both predictive power and model robustness, offering a reliable and interpretable solution for uplift capacity estimation of suction caissons.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124404"},"PeriodicalIF":5.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036732","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}
Pub Date : 2026-01-22DOI: 10.1016/j.oceaneng.2026.124363
He Lan , Guiqin Xue
Minimizing casualties in maritime accidents is the paramount objective of maritime safety management. This study utilizes machine learning techniques to quantify human contributions to very serious maritime accidents. Based on 174 reports of very serious maritime accidents, human factors involved in the accidents are systematically identified using Grounded Theory and the improved HFACS framework, and then a database of human factors in very serious accidents is established. Then, Association Rule is introduced into the LightGBM development process, and the developed model accuracy reaches 85.94 %. SHAP analysis further reveals the different impacts of human factors on very serious maritime accidents. Failure to follow the rules in sight of one another, failure to take effective collision avoidance action early, inadequate safety management, poor competence, and insufficient manning are identified as important factors leading to very serious maritime accidents. These findings provide useful references for prioritizing maritime safety interventions in high-risk scenarios.
{"title":"Assessment of human contribution to very serious maritime accidents based on machine learning techniques","authors":"He Lan , Guiqin Xue","doi":"10.1016/j.oceaneng.2026.124363","DOIUrl":"10.1016/j.oceaneng.2026.124363","url":null,"abstract":"<div><div>Minimizing casualties in maritime accidents is the paramount objective of maritime safety management. This study utilizes machine learning techniques to quantify human contributions to very serious maritime accidents. Based on 174 reports of very serious maritime accidents, human factors involved in the accidents are systematically identified using Grounded Theory and the improved HFACS framework, and then a database of human factors in very serious accidents is established. Then, Association Rule is introduced into the LightGBM development process, and the developed model accuracy reaches 85.94 %. SHAP analysis further reveals the different impacts of human factors on very serious maritime accidents. Failure to follow the rules in sight of one another, failure to take effective collision avoidance action early, inadequate safety management, poor competence, and insufficient manning are identified as important factors leading to very serious maritime accidents. These findings provide useful references for prioritizing maritime safety interventions in high-risk scenarios.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124363"},"PeriodicalIF":5.5,"publicationDate":"2026-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036683","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}
Pub Date : 2026-01-22DOI: 10.1016/j.oceaneng.2026.124360
Jiancai Gao , Yuhong Wang , Ying Li , Chenyang Zhang , Jian Wang , Haixiao Liu
For various types of gravity installed anchor (GIA) that rely on free fall to penetrate the seabed, it is vital before installation to accurately evaluate the velocity that the anchor impacts the seabed. However, challenges are arising when facing emerging engineering developments: (1) the GIA has to be released in air to ensure a sufficient falling distance when the water is not deep enough; (2) to enhance the penetration of GIAs in sandy seabed, auxiliary techniques are considered to apply such as adding extra mass to the anchor; (3) compared to conventional GIAs in deep waters, the GIA with smaller size is more often an option for offshore floating applications, such as renewable energy developments. To deal with the varieties and complexities in evaluating the impact velocity, a systematic study is performed to explore the effects of different factors, including the anchor type, the release heights both in water and in air, the added mass, and the anchor scale. Three typical types of GIAs, namely the finless, T98 and OMNI-Max anchors, are selected in the present study. By combing theoretical and computational fluid dynamics (CFD) analyses, a unified explicit expression of the falling velocity of GIAs is derived in terms of multiple factors, which can be simply and quickly used to calculate the impact velocity of GIAs for various applications.
{"title":"The impact velocity of gravity installed anchors released in air at different height, added mass and scale conditions","authors":"Jiancai Gao , Yuhong Wang , Ying Li , Chenyang Zhang , Jian Wang , Haixiao Liu","doi":"10.1016/j.oceaneng.2026.124360","DOIUrl":"10.1016/j.oceaneng.2026.124360","url":null,"abstract":"<div><div>For various types of gravity installed anchor (GIA) that rely on free fall to penetrate the seabed, it is vital before installation to accurately evaluate the velocity that the anchor impacts the seabed. However, challenges are arising when facing emerging engineering developments: (1) the GIA has to be released in air to ensure a sufficient falling distance when the water is not deep enough; (2) to enhance the penetration of GIAs in sandy seabed, auxiliary techniques are considered to apply such as adding extra mass to the anchor; (3) compared to conventional GIAs in deep waters, the GIA with smaller size is more often an option for offshore floating applications, such as renewable energy developments. To deal with the varieties and complexities in evaluating the impact velocity, a systematic study is performed to explore the effects of different factors, including the anchor type, the release heights both in water and in air, the added mass, and the anchor scale. Three typical types of GIAs, namely the finless, T98 and OMNI-Max anchors, are selected in the present study. By combing theoretical and computational fluid dynamics (CFD) analyses, a unified explicit expression of the falling velocity of GIAs is derived in terms of multiple factors, which can be simply and quickly used to calculate the impact velocity of GIAs for various applications.</div></div>","PeriodicalId":19403,"journal":{"name":"Ocean Engineering","volume":"351 ","pages":"Article 124360"},"PeriodicalIF":5.5,"publicationDate":"2026-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036920","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}