Pub Date : 2026-01-29DOI: 10.1016/j.ress.2026.112323
Wuyin Lin , Songming Yu , Xinran Yu , Yuxing Li , Cuiwei Liu
Integrating hydrogen into urban gas pipeline networks is a pivotal technology for energy transition yet poses critical safety threats, thus necessitating comprehensive risk assessment of hydrogen-blended natural gas pipelines. This study performs full quantitative risk assessment of leakage failure and accident evolution by proposing a novel framework that integrates causal inference (Bow-Tie analysis) with probabilistic machine learning (Bayesian networks), enabling systematic failure factor identification and dynamic accident progression simulation. Key findings indicate human factors and pipeline material degradation as primary triggers. The studied pipeline exhibits a low baseline failure probability, with dispersion emerging as the most likely consequence of leakage. Higher hydrogen blending ratios significantly elevate jet fire risk due to hydrogen’s low ignition energy, while hydrogen’s inherent buoyancy and high diffusivity notably mitigate the likelihood of flash fire and vapor cloud explosion. The case study verifies the model’s practicability, and macro-micro analyses provide holistic insights, offering a reliable method to guide pipeline safety and reliability improvement amid energy transition.
{"title":"Analysis of urban hydrogen-blended natural gas pipeline leak failure and accident evolution based on the combination of causal inference and probabilistic machine learning","authors":"Wuyin Lin , Songming Yu , Xinran Yu , Yuxing Li , Cuiwei Liu","doi":"10.1016/j.ress.2026.112323","DOIUrl":"10.1016/j.ress.2026.112323","url":null,"abstract":"<div><div>Integrating hydrogen into urban gas pipeline networks is a pivotal technology for energy transition yet poses critical safety threats, thus necessitating comprehensive risk assessment of hydrogen-blended natural gas pipelines. This study performs full quantitative risk assessment of leakage failure and accident evolution by proposing a novel framework that integrates causal inference (Bow-Tie analysis) with probabilistic machine learning (Bayesian networks), enabling systematic failure factor identification and dynamic accident progression simulation. Key findings indicate human factors and pipeline material degradation as primary triggers. The studied pipeline exhibits a low baseline failure probability, with dispersion emerging as the most likely consequence of leakage. Higher hydrogen blending ratios significantly elevate jet fire risk due to hydrogen’s low ignition energy, while hydrogen’s inherent buoyancy and high diffusivity notably mitigate the likelihood of flash fire and vapor cloud explosion. The case study verifies the model’s practicability, and macro-micro analyses provide holistic insights, offering a reliable method to guide pipeline safety and reliability improvement amid energy transition.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112323"},"PeriodicalIF":11.0,"publicationDate":"2026-01-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146098522","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-29DOI: 10.1016/j.ress.2026.112317
Wenbin Jiang , Wenkai Hu , Yupeng Li , Weihua Cao
As an effective alarm monitoring strategy, alarm event prediction helps mitigate the impact of alarm floods and the risk of industrial accidents by providing early warnings of potential future alarms, thereby allowing operators more time to take corrective action. However, in continuous industrial processes, varying operating conditions and abnormal states cause real-time fluctuations in alarm rates, posing challenges for existing methods to achieve satisfactory prediction performance. In view of such issues, this paper proposes a new alarm event prediction method adapting to variable alarm rates over long-term consecutive alarm monitoring periods using multi-dimensional sequence embedding and improved Informer. The contributions are threefold: 1) An adaptive alarm sequence segmentation strategy is designed to generate input alarm sequences adapting to alarm rates; 2) a multi-dimensional sequence embedding method based on both the alarm tags and time intervals is proposed to convert the textual alarm messages into numerical vectors; and 3) an Informer based alarm event prediction model is developed for precise and early alarm event prediction under alarm flood and non-flood periods. A case study based on the Vinyl Acetate Monomer public model is given to prove the effectiveness of the proposed method.
{"title":"Multi-dimensional sequence embedding and improved Informer for prediction of industrial alarm events","authors":"Wenbin Jiang , Wenkai Hu , Yupeng Li , Weihua Cao","doi":"10.1016/j.ress.2026.112317","DOIUrl":"10.1016/j.ress.2026.112317","url":null,"abstract":"<div><div>As an effective alarm monitoring strategy, alarm event prediction helps mitigate the impact of alarm floods and the risk of industrial accidents by providing early warnings of potential future alarms, thereby allowing operators more time to take corrective action. However, in continuous industrial processes, varying operating conditions and abnormal states cause real-time fluctuations in alarm rates, posing challenges for existing methods to achieve satisfactory prediction performance. In view of such issues, this paper proposes a new alarm event prediction method adapting to variable alarm rates over long-term consecutive alarm monitoring periods using multi-dimensional sequence embedding and improved Informer. The contributions are threefold: 1) An adaptive alarm sequence segmentation strategy is designed to generate input alarm sequences adapting to alarm rates; 2) a multi-dimensional sequence embedding method based on both the alarm tags and time intervals is proposed to convert the textual alarm messages into numerical vectors; and 3) an Informer based alarm event prediction model is developed for precise and early alarm event prediction under alarm flood and non-flood periods. A case study based on the Vinyl Acetate Monomer public model is given to prove the effectiveness of the proposed method.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112317"},"PeriodicalIF":11.0,"publicationDate":"2026-01-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146098524","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-28DOI: 10.1016/j.ress.2026.112301
Chunhui Guo, Zhenglin Liang
Many real-world systems experience both natural degradation and random shocks, with degradation often assessed using only partial information. When both factors are considered, the underlying degradation process may carry a high risk of transitioning rapidly to a more severe state, making the interpretation of partial observations particularly challenging. To address this challenge, we formulate partially observable systems under combined natural degradation and random shock effects as a partially observable continuous-time Markov model. Based on this model, we introduce a risk-informed inspection and maintenance policy that schedules inspections according to a predefined risk threshold, aiming to reduce costs. We demonstrate that the optimal maintenance approach follows a control-limit policy, applied at decision epochs determined by the evolving risk profile. Leveraging this structural insight, we design a tailored Transformer-augmented Deep Q-Network algorithm to effectively optimize the inspection and maintenance policy under partial observation, which is regarded as a novel and online algorithm for the Partially Observable Markov Decision Process with a multi-dimensional continuous state space. The proposed approach is validated through a case study involving lithium-ion battery maintenance. The results reveal that our approach achieves an average reduction of 57.4% in inspection costs compared to traditional periodic inspection schemes.
{"title":"Transformer-augmented deep Q-network-based risk-informed maintenance policy for partially observable systems under combined degradation and random shock effects","authors":"Chunhui Guo, Zhenglin Liang","doi":"10.1016/j.ress.2026.112301","DOIUrl":"10.1016/j.ress.2026.112301","url":null,"abstract":"<div><div>Many real-world systems experience both natural degradation and random shocks, with degradation often assessed using only partial information. When both factors are considered, the underlying degradation process may carry a high risk of transitioning rapidly to a more severe state, making the interpretation of partial observations particularly challenging. To address this challenge, we formulate partially observable systems under combined natural degradation and random shock effects as a partially observable continuous-time Markov model. Based on this model, we introduce a risk-informed inspection and maintenance policy that schedules inspections according to a predefined risk threshold, aiming to reduce costs. We demonstrate that the optimal maintenance approach follows a control-limit policy, applied at decision epochs determined by the evolving risk profile. Leveraging this structural insight, we design a tailored Transformer-augmented Deep Q-Network algorithm to effectively optimize the inspection and maintenance policy under partial observation, which is regarded as a novel and online algorithm for the Partially Observable Markov Decision Process with a multi-dimensional continuous state space. The proposed approach is validated through a case study involving lithium-ion battery maintenance. The results reveal that our approach achieves an average reduction of 57.4% in inspection costs compared to traditional periodic inspection schemes.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112301"},"PeriodicalIF":11.0,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079556","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-28DOI: 10.1016/j.ress.2026.112305
Guoqing Yang, Hongye Yuan, Wenshuai Yang, Ruru Jia
The suddenness and the casualties’ uncertainty of natural disasters urgently require a fair and robust network design for the medical supply distribution and the injured evacuation to reduce their post-disaster impact. This study establishes a distributionally robust chance-constrained model for medical supplies allocation in last-mile relief networks, with the objective of minimizing the worst-case Conditional Value-at-Risk (CVaR) of supply shortages. The distribution of severely injured casualties is characterized via a scenario-wise ambiguity set, thereby the proposed model is reformulated as a mixed-integer linear programming problem for tractability. Numerical experiment based on Wenchuan earthquake derives several important findings. First, total supplies and raw materials exhibit analogous effects—increasing either reduces shortage levels initially, but further reductions are constrained by the other factor; Second, in response to high risks, the tendency is to build additional medical stations rather than expanding the scale of existing ones to disperse risk. Conversely, when risk is low, scaling up existing medical stations is preferred over establishing temporary facilities; Finally, under out-of-sample data fluctuations, the CVaR model demonstrates stronger robustness than the sample average approximation model, with consistently smaller standard deviations and superior stability.
{"title":"Distributionally robust fairness-based last-mile relief network optimization with casualty uncertainty","authors":"Guoqing Yang, Hongye Yuan, Wenshuai Yang, Ruru Jia","doi":"10.1016/j.ress.2026.112305","DOIUrl":"10.1016/j.ress.2026.112305","url":null,"abstract":"<div><div>The suddenness and the casualties’ uncertainty of natural disasters urgently require a fair and robust network design for the medical supply distribution and the injured evacuation to reduce their post-disaster impact. This study establishes a distributionally robust chance-constrained model for medical supplies allocation in last-mile relief networks, with the objective of minimizing the worst-case Conditional Value-at-Risk (CVaR) of supply shortages. The distribution of severely injured casualties is characterized via a scenario-wise ambiguity set, thereby the proposed model is reformulated as a mixed-integer linear programming problem for tractability. Numerical experiment based on Wenchuan earthquake derives several important findings. First, total supplies and raw materials exhibit analogous effects—increasing either reduces shortage levels initially, but further reductions are constrained by the other factor; Second, in response to high risks, the tendency is to build additional medical stations rather than expanding the scale of existing ones to disperse risk. Conversely, when risk is low, scaling up existing medical stations is preferred over establishing temporary facilities; Finally, under out-of-sample data fluctuations, the CVaR model demonstrates stronger robustness than the sample average approximation model, with consistently smaller standard deviations and superior stability.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112305"},"PeriodicalIF":11.0,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-28DOI: 10.1016/j.ress.2026.112304
Dingrong Tan , Xiaoda Shen , Ye Deng , Jun Wu
Many infrastructure systems are modeled as spatially embedded networks, where the topology is constrained by geometry and distance costs. A central problem is to identify spatial regions whose node/edge removal causes the largest drop in a specified network function under a fixed perturbation budget, with applications from disease prevention to congestion mitigation. However, existing regional identification models struggle to accurately and directly describe the true extent of network damage, and most approaches fail to seamlessly integrate geographic information with network topology, resulting in poor precision when identifying vulnerable regions. In this paper, we first introduce a virtual node model that more effectively captures network damage through a granularity-enhancement mechanism. Furthermore, we propose a deep learning framework (SNDM-VN) based on graph neural networks, which is trained with supervised learning on a large set of small synthetic spatial networks and accurately identifies vulnerable regions in previously unseen real-world networks. Extensive experiments demonstrate that SNDM-VN significantly outperforms baseline methods in vulnerable region detection tasks. Through large-scale data-driven learning, the proposed framework effectively integrates topological and spatial features to accurately identify vulnerable regions that could severely compromise network reliability – something traditional methods find difficult. Our results provide accurate region-level identification and extend the scope of deep learning applications in spatial network analysis.
{"title":"Graph neural network-based identification of vulnerable regions in spatial complex networks via virtual node model","authors":"Dingrong Tan , Xiaoda Shen , Ye Deng , Jun Wu","doi":"10.1016/j.ress.2026.112304","DOIUrl":"10.1016/j.ress.2026.112304","url":null,"abstract":"<div><div>Many infrastructure systems are modeled as spatially embedded networks, where the topology is constrained by geometry and distance costs. A central problem is to identify spatial regions whose node/edge removal causes the largest drop in a specified network function under a fixed perturbation budget, with applications from disease prevention to congestion mitigation. However, existing regional identification models struggle to accurately and directly describe the true extent of network damage, and most approaches fail to seamlessly integrate geographic information with network topology, resulting in poor precision when identifying vulnerable regions. In this paper, we first introduce a virtual node model that more effectively captures network damage through a granularity-enhancement mechanism. Furthermore, we propose a deep learning framework (SNDM-VN) based on graph neural networks, which is trained with supervised learning on a large set of small synthetic spatial networks and accurately identifies vulnerable regions in previously unseen real-world networks. Extensive experiments demonstrate that SNDM-VN significantly outperforms baseline methods in vulnerable region detection tasks. Through large-scale data-driven learning, the proposed framework effectively integrates topological and spatial features to accurately identify vulnerable regions that could severely compromise network reliability – something traditional methods find difficult. Our results provide accurate region-level identification and extend the scope of deep learning applications in spatial network analysis.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112304"},"PeriodicalIF":11.0,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146189764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-28DOI: 10.1016/j.ress.2026.112302
Meng Li, Yu-Rong Song, Bo Song, Guo-Ping Jiang
Urban transportation systems are essential for sustaining urban growth and ensuring efficient resource allocation. Existing studies primarily focus on evaluating network resilience after system disturbances, with insufficient attention paid to the response mechanisms during disturbances and the enhancement of resilience afterward. Therefore, we propose a cascading failure model that considers passenger transfer impedance, and design a recovery priority strategy for failed nodes to maximize the resilience of the urban transportation interdependent network (UTIN). Specifically, based on traffic sensing data, we construct a station-centric UTIN to assess structural resilience under various disruption scenarios and different transfer distances. By combining impedance function and flow redistribution, passenger behavior and node load update are considered. Additionally, the recovery priority strategy for failed nodes is discussed. The results indicate: 1) UTINs with longer transfer distances exhibit stronger resistance to risks. When considering impedance costs, the optimal transfer distance is 800 m. 2) During cascading failure propagation, optimizing flow distribution effectively lowers the critical capacity threshold required for system stability, thereby enhancing network resilience. 3) During the recovery phase, different recovery strategies exhibit significant differences in their effectiveness in restoring system resilience. The research findings provide valuable references for disaster prevention, emergency response, and post-disaster recovery in urban transportation systems.
{"title":"Resilience assessment and enhancement of urban transportation interdependent network under cascading failure","authors":"Meng Li, Yu-Rong Song, Bo Song, Guo-Ping Jiang","doi":"10.1016/j.ress.2026.112302","DOIUrl":"10.1016/j.ress.2026.112302","url":null,"abstract":"<div><div>Urban transportation systems are essential for sustaining urban growth and ensuring efficient resource allocation. Existing studies primarily focus on evaluating network resilience after system disturbances, with insufficient attention paid to the response mechanisms during disturbances and the enhancement of resilience afterward. Therefore, we propose a cascading failure model that considers passenger transfer impedance, and design a recovery priority strategy for failed nodes to maximize the resilience of the urban transportation interdependent network (UTIN). Specifically, based on traffic sensing data, we construct a station-centric UTIN to assess structural resilience under various disruption scenarios and different transfer distances. By combining impedance function and flow redistribution, passenger behavior and node load update are considered. Additionally, the recovery priority strategy for failed nodes is discussed. The results indicate: 1) UTINs with longer transfer distances exhibit stronger resistance to risks. When considering impedance costs, the optimal transfer distance is 800 m. 2) During cascading failure propagation, optimizing flow distribution effectively lowers the critical capacity threshold required for system stability, thereby enhancing network resilience. 3) During the recovery phase, different recovery strategies exhibit significant differences in their effectiveness in restoring system resilience. The research findings provide valuable references for disaster prevention, emergency response, and post-disaster recovery in urban transportation systems.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112302"},"PeriodicalIF":11.0,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146098523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-27DOI: 10.1016/j.ress.2026.112300
Tim Bastek , Jens Denecke , Jürgen Schmidt
Gas pipeline failure continues to be a serious hazard for people in the vicinity of gas pipelines, particularly given the increase in urban development and aging infrastructure. This study critically reviews the current state and potential of data-driven approaches in pipeline integrity management systems (PIMS) for most critical threats. In addition to a purely theoretical discussion, three illustrative case studies are used to highlight the main limitations in the following areas: a) third-party damage assessment, b) the quality of in-line-Inspection (ILI) data and c) machine learning-based external corrosion evaluation. A quantitative risk analysis was performed to analyze shortcomings in context of current prevention practices. Research gaps lie in the evaluation of probability of failure insufficiently dependent on the gas pipeline location but in practice on pipeline design. A new GIS-based, probabilistic approach was proposed to assess TPD using available environmental data. Secondly, published ILI data was analyzed, which reveals a large amount of corrosion detected over pipeline route, but low replicability from one ILI run to another - limiting usage in PIMS and data driven modelling. Thirdly, a hybrid support vector regression model was trained to predict external corrosion, but its performance proved unstable: prediction accuracy dropped by 27% during cross-validation, highlighting the practical risks of model overfitting. This study highlights the need for more robust, context-sensitive models and outlines potential advancements to improve pipeline safety and system reliability using data-driven strategies.
{"title":"Future directions for data-driven approaches in pipeline integrity management: Risk assessment, in-line inspection, and machine learning","authors":"Tim Bastek , Jens Denecke , Jürgen Schmidt","doi":"10.1016/j.ress.2026.112300","DOIUrl":"10.1016/j.ress.2026.112300","url":null,"abstract":"<div><div>Gas pipeline failure continues to be a serious hazard for people in the vicinity of gas pipelines, particularly given the increase in urban development and aging infrastructure. This study critically reviews the current state and potential of data-driven approaches in pipeline integrity management systems (PIMS) for most critical threats. In addition to a purely theoretical discussion, three illustrative case studies are used to highlight the main limitations in the following areas: a) third-party damage assessment, b) the quality of in-line-Inspection (ILI) data and c) machine learning-based external corrosion evaluation. A quantitative risk analysis was performed to analyze shortcomings in context of current prevention practices. Research gaps lie in the evaluation of probability of failure insufficiently dependent on the gas pipeline location but in practice on pipeline design. A new GIS-based, probabilistic approach was proposed to assess TPD using available environmental data. Secondly, published ILI data was analyzed, which reveals a large amount of corrosion detected over pipeline route, but low replicability from one ILI run to another - limiting usage in PIMS and data driven modelling. Thirdly, a hybrid support vector regression model was trained to predict external corrosion, but its performance proved unstable: prediction accuracy dropped by 27% during cross-validation, highlighting the practical risks of model overfitting. This study highlights the need for more robust, context-sensitive models and outlines potential advancements to improve pipeline safety and system reliability using data-driven strategies.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112300"},"PeriodicalIF":11.0,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-27DOI: 10.1016/j.ress.2026.112303
Jie Jiang, Yifan Yang, Yudi Chen
The significant interplay between the physical and societal functioning of infrastructure underscores the need for mitigation strategies that balance both the physical and societal considerations facing frequently increasing and intensifying hazards. However, existing models typically either focus on assessing physical or societal impacts separately or incorporate the social vulnerability index as modification parameters in resilience-driven restoration objective functions, which fail to orchestrate and integrate societal ramifications explicitly into the strategy formulation process and evaluate the efficacy of physical and societal strategies in an equivalent degree of detail at neighborhood-level within a community. To fill this gap, this paper develops an integrated mathematical model for the formulation and prioritization of mitigation strategies in the flooding hazard preplanning stage, with the objective of alleviating physical performance degradation and the loss of residents’ capabilities to meet their diverse societal needs. The strength of the model lies in its fine-grained physical co-simulation model for generating proactive flooding scenarios, its capacity for multi-dimensional strategies formulation that incorporates waterlogging characteristics, link-level traffic performance index (TPI), residents' adapted routing behavior quantified by betweenness accessibility (BA), and its ability to evaluate both physical and societal efficacy in an integrated manner to relieve hazard-induced impacts.
{"title":"An integrated approach to evaluating and prioritizing socio-physical flooding mitigation planning to enhance resilience in a community","authors":"Jie Jiang, Yifan Yang, Yudi Chen","doi":"10.1016/j.ress.2026.112303","DOIUrl":"10.1016/j.ress.2026.112303","url":null,"abstract":"<div><div>The significant interplay between the physical and societal functioning of infrastructure underscores the need for mitigation strategies that balance both the physical and societal considerations facing frequently increasing and intensifying hazards. However, existing models typically either focus on assessing physical or societal impacts separately or incorporate the social vulnerability index as modification parameters in resilience-driven restoration objective functions, which fail to orchestrate and integrate societal ramifications explicitly into the strategy formulation process and evaluate the efficacy of physical and societal strategies in an equivalent degree of detail at neighborhood-level within a community. To fill this gap, this paper develops an integrated mathematical model for the formulation and prioritization of mitigation strategies in the flooding hazard preplanning stage, with the objective of alleviating physical performance degradation and the loss of residents’ capabilities to meet their diverse societal needs. The strength of the model lies in its fine-grained physical co-simulation model for generating proactive flooding scenarios, its capacity for multi-dimensional strategies formulation that incorporates waterlogging characteristics, link-level traffic performance index (TPI), residents' adapted routing behavior quantified by betweenness accessibility (BA), and its ability to evaluate both physical and societal efficacy in an integrated manner to relieve hazard-induced impacts.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112303"},"PeriodicalIF":11.0,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146189802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-27DOI: 10.1016/j.ress.2026.112295
Huanhuan Hu , Pan Wang , Fukang Xin , Zheng Zhang , Haihe Li , Jiahua Zhang
The failure probability concerning specified design parameters, termed the failure probability function (FPF), is essential in reliability-based design. Conventional methods require high computational costs for complex systems due to repeated expensive simulations. Although single-loop methods with active learning Kriging (AK) have been proposed to reduce these costs, their efficiency remains limited by suboptimal sampling and inaccurate kernel density estimation (KDE). To address these challenges, this work introduces a novel multi-purpose K-nearest neighbor (KNN) framework integrated with an enhanced AK in an augmented space, termed the SL-AK-KNN method. The method leverages the adaptive capabilities of KNN in two key aspects: (1) as a spatial-information-guided learning function that improves both global and local efficiency of AK by exploring and exploiting sample density variations across different regions, and (2) as an adaptive nonparametric density estimator for approximating the conditional joint probability density function (PDF), thereby mitigating KDE’s edge region inaccuracies without relying on kernel functions and fixed bandwidth. It is intuitively well-suited for exploratory analysis of unknown density distributions. Numerical examples demonstrate that the proposed framework significantly reduces computational costs while enhancing FPF estimation accuracy, enabling robust reliability design for the engineering applications of the bracket structure and hydraulic pipeline system.
{"title":"A synergistic approach: multi-purpose K-nearest neighbor and active learning Kriging for efficient failure probability function estimation","authors":"Huanhuan Hu , Pan Wang , Fukang Xin , Zheng Zhang , Haihe Li , Jiahua Zhang","doi":"10.1016/j.ress.2026.112295","DOIUrl":"10.1016/j.ress.2026.112295","url":null,"abstract":"<div><div>The failure probability concerning specified design parameters, termed the failure probability function (FPF), is essential in reliability-based design. Conventional methods require high computational costs for complex systems due to repeated expensive simulations. Although single-loop methods with active learning Kriging (AK) have been proposed to reduce these costs, their efficiency remains limited by suboptimal sampling and inaccurate kernel density estimation (KDE). To address these challenges, this work introduces a novel multi-purpose K-nearest neighbor (KNN) framework integrated with an enhanced AK in an augmented space, termed the SL-AK-KNN method. The method leverages the adaptive capabilities of KNN in two key aspects: (1) as a spatial-information-guided learning function that improves both global and local efficiency of AK by exploring and exploiting sample density variations across different regions, and (2) as an adaptive nonparametric density estimator for approximating the conditional joint probability density function (PDF), thereby mitigating KDE’s edge region inaccuracies without relying on kernel functions and fixed bandwidth. It is intuitively well-suited for exploratory analysis of unknown density distributions. Numerical examples demonstrate that the proposed framework significantly reduces computational costs while enhancing FPF estimation accuracy, enabling robust reliability design for the engineering applications of the bracket structure and hydraulic pipeline system.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112295"},"PeriodicalIF":11.0,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079651","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-01-27DOI: 10.1016/j.ress.2026.112290
Hongzhen Wang , Zhengjiang Liu , Xiang-Yu Zhou , Lianbo Li , Shanshan Fei , Xinjian Wang
The assessment of multi-ship collision risk situation holds important theoretical value and practical significance for enhancing waterborne vessel safety supervision and ensuring safe navigation. However, maritime multi-ship navigation risks are often influenced by the coupled influence of hydro-meteorological conditions and multi-ship navigation situations, exhibiting significant uncertainty and fuzziness. In order to address those gaps, this study aims to propose a collision risk assessment method for multi-ships. First, a dual-dimensional evaluation indicator system integrating hydro-meteorological factors and multi-ship characteristics was constructed, accompanied by six calculation methods for indicator values, providing an operational basis for accurate risk assessment. Subsequently, game theory was employed to integrate weighting results derived from the best-worst method and the extension correlation function method, so as to mitigate the one-sidedness of a single weighting approach. Finally, based on the designed indicator interval grades, a finite interval cloud generator was constructed to characterize the fuzziness and uncertainty of the indicators, thereby achieving a precise quantitative rating of multi-ship collision risk. Validation through four groups of multi-ship potential encounter scenarios in the Bohai Sea of China shows that the proposed method can accurately distinguish the risk levels of different scenarios. Moreover, the variance of the evaluation results is 1 to 4.17 times that of the traditional extension cloud model, indicating higher confidence and sensitivity. The method provides objective and precise technical support for navigation situation monitoring in multi-ship potential encounter scenarios.
{"title":"Multi-ship collision risk situation assessment based on finite interval cloud model","authors":"Hongzhen Wang , Zhengjiang Liu , Xiang-Yu Zhou , Lianbo Li , Shanshan Fei , Xinjian Wang","doi":"10.1016/j.ress.2026.112290","DOIUrl":"10.1016/j.ress.2026.112290","url":null,"abstract":"<div><div>The assessment of multi-ship collision risk situation holds important theoretical value and practical significance for enhancing waterborne vessel safety supervision and ensuring safe navigation. However, maritime multi-ship navigation risks are often influenced by the coupled influence of hydro-meteorological conditions and multi-ship navigation situations, exhibiting significant uncertainty and fuzziness. In order to address those gaps, this study aims to propose a collision risk assessment method for multi-ships. First, a dual-dimensional evaluation indicator system integrating hydro-meteorological factors and multi-ship characteristics was constructed, accompanied by six calculation methods for indicator values, providing an operational basis for accurate risk assessment. Subsequently, game theory was employed to integrate weighting results derived from the best-worst method and the extension correlation function method, so as to mitigate the one-sidedness of a single weighting approach. Finally, based on the designed indicator interval grades, a finite interval cloud generator was constructed to characterize the fuzziness and uncertainty of the indicators, thereby achieving a precise quantitative rating of multi-ship collision risk. Validation through four groups of multi-ship potential encounter scenarios in the Bohai Sea of China shows that the proposed method can accurately distinguish the risk levels of different scenarios. Moreover, the variance of the evaluation results is 1 to 4.17 times that of the traditional extension cloud model, indicating higher confidence and sensitivity. The method provides objective and precise technical support for navigation situation monitoring in multi-ship potential encounter scenarios.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"271 ","pages":"Article 112290"},"PeriodicalIF":11.0,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079655","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号