Reliability Engineering & System Safety最新文献_第2页

Robustness of disaster response systems: Hypergraph modelling, key local structures, and fortification methods 灾难响应系统的稳健性：超图建模、关键的局部结构和加固方法

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-07 DOI: 10.1016/j.ress.2026.112372

Chong Gao , Hui Jiang , Mang Li

Disaster response systems (DRS) operate in an uncertain environment and are continuously tested by external perturbations and internal failures. The ability to withstand disruptions and mitigate the risk of structural disintegration is a defining aspect of a resilient DRS. The current research on the resilience of DRS aims to reveal its robustness performance using metrics from social network analysis. However, the correspondence between these metrics and the underlying ability to maintain structural integrity is neither direct nor well-established. In this paper, we first construct faithful representations of DRS using hypergraph modelling. We demonstrate that hypergraph-based representations provide a principled basis for robustness analysis, revealing local structures key to structural integrity. We establish comprehensive robustness metrics and introduce a principled analysis framework to formalise these key local structures. We also perform a stability decomposition and obtain stability indicators that are used to identify the unstable local structures with weak internal coherence and fragile external embedding. Then we formulate the robustness-enhancing problem and develop fortification methods aiming at maximising the total robustness gain. Extensive simulations demonstrate that our approach substantially outperforms alternative strategies, including coreness-based and graph-based methods, across a wide range of fortification budgets. In addition, the proposed method has an efficient numerical implementation, and we validate it in large-scale synthetic hypergraphs.

灾害响应系统（DRS）在不确定的环境中运行，并不断受到外部扰动和内部故障的考验。抵御破坏和减轻结构解体风险的能力是弹性DRS的一个决定性方面。目前对DRS弹性的研究旨在利用社会网络分析的指标来揭示其鲁棒性。然而，这些指标与维护结构完整性的潜在能力之间的对应关系既不直接也不完善。在本文中，我们首先使用超图建模构造了DRS的忠实表示。我们证明了基于超图的表示为鲁棒性分析提供了原则基础，揭示了对结构完整性至关重要的局部结构。我们建立了全面的稳健性指标，并引入了一个原则性的分析框架来形式化这些关键的局部结构。我们还进行了稳定性分解，得到了稳定性指标，用于识别内部相干性弱、外部嵌入脆弱的不稳定局部结构。然后，我们提出了鲁棒性增强问题，并开发了旨在最大化总鲁棒性增益的强化方法。广泛的模拟表明，我们的方法大大优于其他策略，包括基于核心和基于图形的方法，在广泛的防御预算。此外，该方法具有高效的数值实现，并在大规模合成超图中进行了验证。

{"title":"Robustness of disaster response systems: Hypergraph modelling, key local structures, and fortification methods","authors":"Chong Gao , Hui Jiang , Mang Li","doi":"10.1016/j.ress.2026.112372","DOIUrl":"10.1016/j.ress.2026.112372","url":null,"abstract":"<div><div>Disaster response systems (DRS) operate in an uncertain environment and are continuously tested by external perturbations and internal failures. The ability to withstand disruptions and mitigate the risk of structural disintegration is a defining aspect of a resilient DRS. The current research on the resilience of DRS aims to reveal its robustness performance using metrics from social network analysis. However, the correspondence between these metrics and the underlying ability to maintain structural integrity is neither direct nor well-established. In this paper, we first construct faithful representations of DRS using hypergraph modelling. We demonstrate that hypergraph-based representations provide a principled basis for robustness analysis, revealing local structures key to structural integrity. We establish comprehensive robustness metrics and introduce a principled analysis framework to formalise these key local structures. We also perform a stability decomposition and obtain stability indicators that are used to identify the unstable local structures with weak internal coherence and fragile external embedding. Then we formulate the robustness-enhancing problem and develop fortification methods aiming at maximising the total robustness gain. Extensive simulations demonstrate that our approach substantially outperforms alternative strategies, including coreness-based and graph-based methods, across a wide range of fortification budgets. In addition, the proposed method has an efficient numerical implementation, and we validate it in large-scale synthetic hypergraphs.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112372"},"PeriodicalIF":11.0,"publicationDate":"2026-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174921","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}

引用次数: 0

A resilience enhancement approach for interdependent networks incorporating recovery coupling mechanisms 结合恢复耦合机制的相互依赖网络弹性增强方法

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-07 DOI: 10.1016/j.ress.2026.112375

Jiuyao Jiang, Jichao Li, Tianyang Lei, Kewei Yang

Modern infrastructure networks are expanding in both scale and interdependence. On the one hand, a localized disturbance can propagate through intra-layer connectivity and inter-layer dependency relations, thereby triggering cascading failures. On the other hand, under interdependence, a disrupted subsystem may draw resources and support from other subsystems to facilitate its own restoration, giving rise to recovery-coupling effects. To bolster resilience to unforeseen threats, we propose an Interdependent Network Resilience Enhancement Framework with Recovery Coupling Mechanisms (INRCM). We formulate a fifth-order tensor that represents a time-varying, two-layer system and jointly embeds recovery-coupling and cascading-failure dynamics. Building on this model, we derive a three-phase structural-resilience metric that evaluates performance in the pre-disturbance, disturbance, and recovery stages. Within INRCM, we develop a graph neural network-guided genetic algorithm as the optimization module to identify high-quality feasible recovery node sets at practical computational cost. Extensive experiments on synthetic and real-world networks across multiple scales, including larger-scale topologies, show that INRCM robustly accelerates post-disturbance recovery and consistently outperforms centrality-based heuristics in both convergence behavior and achieved resilience levels. The framework therefore offers actionable guidance for post-event restoration and for designing infrastructure systems with intrinsic resilience.

现代基础设施网络在规模和相互依存方面不断扩大。一方面，局部扰动可以通过层内连通性和层间依赖关系传播，从而引发级联故障。另一方面，在相互依赖的情况下，一个被破坏的子系统可能会从其他子系统获得资源和支持，以促进自身的恢复，从而产生恢复耦合效应。为了增强对不可预见威胁的弹性，我们提出了一个具有恢复耦合机制（INRCM）的相互依赖网络弹性增强框架。我们建立了一个五阶张量，它代表一个时变的两层系统，并共同嵌入恢复耦合和级联失效动力学。在此模型的基础上，我们推导了一个三相结构弹性度量，用于评估扰动前、扰动和恢复阶段的性能。在INRCM中，我们开发了一个图神经网络引导的遗传算法作为优化模块，以实际的计算成本识别高质量的可行恢复节点集。在多个尺度（包括更大规模拓扑）的合成和现实网络上进行的大量实验表明，INRCM稳健地加速了扰动后的恢复，并且在收敛行为和恢复水平上始终优于基于中心性的启发式。因此，该框架为灾后恢复和设计具有内在弹性的基础设施系统提供了可操作的指导。

{"title":"A resilience enhancement approach for interdependent networks incorporating recovery coupling mechanisms","authors":"Jiuyao Jiang, Jichao Li, Tianyang Lei, Kewei Yang","doi":"10.1016/j.ress.2026.112375","DOIUrl":"10.1016/j.ress.2026.112375","url":null,"abstract":"<div><div>Modern infrastructure networks are expanding in both scale and interdependence. On the one hand, a localized disturbance can propagate through intra-layer connectivity and inter-layer dependency relations, thereby triggering cascading failures. On the other hand, under interdependence, a disrupted subsystem may draw resources and support from other subsystems to facilitate its own restoration, giving rise to recovery-coupling effects. To bolster resilience to unforeseen threats, we propose an Interdependent Network Resilience Enhancement Framework with Recovery Coupling Mechanisms (INRCM). We formulate a fifth-order tensor that represents a time-varying, two-layer system and jointly embeds recovery-coupling and cascading-failure dynamics. Building on this model, we derive a three-phase structural-resilience metric that evaluates performance in the pre-disturbance, disturbance, and recovery stages. Within INRCM, we develop a graph neural network-guided genetic algorithm as the optimization module to identify high-quality feasible recovery node sets at practical computational cost. Extensive experiments on synthetic and real-world networks across multiple scales, including larger-scale topologies, show that INRCM robustly accelerates post-disturbance recovery and consistently outperforms centrality-based heuristics in both convergence behavior and achieved resilience levels. The framework therefore offers actionable guidance for post-event restoration and for designing infrastructure systems with intrinsic resilience.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112375"},"PeriodicalIF":11.0,"publicationDate":"2026-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174991","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}

引用次数: 0

Dynamic coupling reliability assessment of smart wind-photovoltaic-storage cyber-physical systems under multi-source uncertain information 多源不确定信息下智能风-光-储网络物理系统动态耦合可靠性评估

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-06 DOI: 10.1016/j.ress.2026.112368

Hongyan Dui , Rukun Wang , Wanyun Xia , Liudong Xing

With the increasing integration of wind, photovoltaic (PV), and energy storage systems, modern distributed energy systems rely on sensing, communication, and control infrastructures, evolving into smart cyber-physical systems. Under multi-source uncertainties including renewable variability, load fluctuations, communication degradation, and cyber attacks, existing methods struggle to capture coupled power–information dynamics and their joint reliability impacts. This study proposes a dynamic coupling reliability assessment framework for Wind-PV-Storage Cyber-Physical Systems (WPS-CPSs). A heterogeneous directed network is constructed to model structural relationships among generation, load, storage, and cyber components. Based on this model, a Power Flow Reliability Index (PFRI) and an Information Flow Reliability Index (IFRI) are developed to characterize physical-layer and cyber-layer performance, respectively. These indices are integrated via a coupling penalty mechanism to form the Dynamic Coupling Reliability Index (DCRI), enabling evaluation of cross-layer interactions. The framework is validated on a benchmark integrating a modified IEEE 14-bus system with engineering data from the Yancheng National Renewable Energy Demonstration Base. Results indicate that energy storage improves the lower reliability bound, Denial-of-Service and False Data Injection attacks cause distinct IFRI degradation, and DCRI captures coupling-penalty variations revealing both the benefits and risks of cyber-physical integration, offering guidance for resilient renewable-rich microgrids.

随着风能、光伏（PV）和储能系统的日益融合，现代分布式能源系统依赖于传感、通信和控制基础设施，向智能网络物理系统发展。在包括可再生可变性、负荷波动、通信退化和网络攻击在内的多源不确定性下，现有方法难以捕捉耦合的电力信息动态及其联合可靠性影响。本文提出了风电-光伏-储能网络物理系统（wps - cps）的动态耦合可靠性评估框架。构建了一个异构定向网络来模拟生成、负载、存储和网络组件之间的结构关系。在此模型的基础上，分别建立了描述物理层和网络层性能的潮流可靠性指标（PFRI）和信息流可靠性指标（IFRI）。这些指标通过耦合惩罚机制集成，形成动态耦合可靠性指数（DCRI），从而能够评估跨层相互作用。结合改进的IEEE 14总线系统和盐城国家可再生能源示范基地的工程数据，对该框架进行了基准验证。结果表明，储能提高了低可靠性界限，拒绝服务和虚假数据注入攻击导致明显的IFRI退化，DCRI捕获耦合惩罚变化，揭示了网络物理集成的好处和风险，为弹性可再生资源丰富的微电网提供了指导。

{"title":"Dynamic coupling reliability assessment of smart wind-photovoltaic-storage cyber-physical systems under multi-source uncertain information","authors":"Hongyan Dui , Rukun Wang , Wanyun Xia , Liudong Xing","doi":"10.1016/j.ress.2026.112368","DOIUrl":"10.1016/j.ress.2026.112368","url":null,"abstract":"<div><div>With the increasing integration of wind, photovoltaic (PV), and energy storage systems, modern distributed energy systems rely on sensing, communication, and control infrastructures, evolving into smart cyber-physical systems. Under multi-source uncertainties including renewable variability, load fluctuations, communication degradation, and cyber attacks, existing methods struggle to capture coupled power–information dynamics and their joint reliability impacts. This study proposes a dynamic coupling reliability assessment framework for Wind-PV-Storage Cyber-Physical Systems (WPS-CPSs). A heterogeneous directed network is constructed to model structural relationships among generation, load, storage, and cyber components. Based on this model, a Power Flow Reliability Index (PFRI) and an Information Flow Reliability Index (IFRI) are developed to characterize physical-layer and cyber-layer performance, respectively. These indices are integrated via a coupling penalty mechanism to form the Dynamic Coupling Reliability Index (DCRI), enabling evaluation of cross-layer interactions. The framework is validated on a benchmark integrating a modified IEEE 14-bus system with engineering data from the Yancheng National Renewable Energy Demonstration Base. Results indicate that energy storage improves the lower reliability bound, Denial-of-Service and False Data Injection attacks cause distinct IFRI degradation, and DCRI captures coupling-penalty variations revealing both the benefits and risks of cyber-physical integration, offering guidance for resilient renewable-rich microgrids.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112368"},"PeriodicalIF":11.0,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174952","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}

引用次数: 0

Analysis of systems with dependent components through a variance-based index and regression importance signature 通过基于方差的指标和回归重要性特征对具有相关成分的系统进行分析

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-06 DOI: 10.1016/j.ress.2026.112357

Antonio Di Crescenzo , Giulia Pisano , Alfonso Suárez-Llorens

This paper extends a recent regression importance index in reliability theory to subgroups of components, introducing the regression importance signature. This measure identifies, for any fixed number k of components, the subgroup of size k whose joint state most significantly influences system lifetime variability. In addition to this extension, we investigate new properties of the index that enhance understanding of how component reliability and structural role jointly affect system importance. Unlike traditional single-component indices, the measure captures heterogeneity and interactions, reducing them to intuitive behaviours in series or parallel systems and revealing subsystems or dependencies that cannot be detected by individual measures. General results for dependent components are provided, with conditions for comparing individual components and subgroups.

本文将可靠性理论中最新的回归重要性指标推广到部件子群，引入回归重要性签名。对于任何固定数量的k个组件，该测量方法确定大小为k的子组，其联合状态最显著地影响系统寿命变异性。除此之外，我们还研究了该指标的新属性，以增强对组件可靠性和结构角色如何共同影响系统重要性的理解。与传统的单组分指数不同，该方法捕捉异质性和相互作用，将它们简化为串联或并行系统中的直观行为，并揭示单个方法无法检测到的子系统或依赖关系。提供了相关组件的一般结果，并提供了比较单个组件和子组的条件。

引用次数: 0

SCADA data-driven failure rate and reliability prediction for offshore wind turbines SCADA数据驱动的海上风力涡轮机故障率和可靠性预测

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-06 DOI: 10.1016/j.ress.2026.112378

Xiangyu Kong , Ruishu Huang , He Li , C. Guedes Soares

A data-driven model is proposed for the failure rate prediction of offshore wind turbines using the Supervisory Control And Data Acquisition (SCADA) data from onshore and offshore wind farms. The wind turbines are first decomposed into multiple components according to maintenance records. An adaptive weighting algorithm is then developed to assess the relative contributions of reliability-influencing factors to failure rate conversion. Subsequently, a failure rate prediction model is proposed for offshore wind turbines based on transforming onshore device failure rates. The result shows that the failure rate of offshore wind turbines is approximately 23% higher than that of onshore wind turbines. Comparative results confirm that the proposed method generates lower estimation errors than existing approaches.

利用陆上和海上风电场的监控和数据采集（SCADA）数据，提出了一种数据驱动的海上风力发电机故障率预测模型。根据维护记录，首先将风力涡轮机分解为多个部件。然后提出了一种自适应加权算法来评估可靠性影响因素对故障率转换的相对贡献。在此基础上，提出了一种基于陆上设备故障率变换的海上风力发电机故障率预测模型。结果表明，海上风力机的故障率比陆上风力机高约23%。对比结果表明，该方法比现有方法产生的估计误差更小。

引用次数: 0

Optimizing post-disaster road restoration with reinforcement learning: A traveler-behavior-aware approach 用强化学习优化灾后道路恢复：一种旅行者行为感知方法

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-06 DOI: 10.1016/j.ress.2026.112371

Maryam Babaee , Namrata Saha , Frank Mediavilla Ponce , Shabnam Rezapour , M. Hadi Amini

This paper introduces an AI-driven method to optimize short-term road network restoration in disaster-affected areas, aiming to maximize post-disaster traffic acceleration. Our approach considers travelers' behavior, gradual adaptation to network changes, limited recovery resources, and uncertainties in recovery times. Addressing these complexities requires a stochastic approach for uncertainties, sequential decision-making for resource management, and a model-free technique for simulating traveler adaptation.

To tackle these challenges, we develop the Traveler-Adaptive Restoration Mechanism (TARM), integrating Reinforcement Learning (RL), the Markov Decision Process (MDP), and optimization-based day-to-day traffic simulation. The method is evaluated on Sioux Falls' road network under tornado scenarios based on historical data. Results highlight the influence of travelers’ route choices and the speed of restoration information dissemination on optimal policies.

Findings reveal that accelerating the road restoration process by increasing restoration resources does not necessarily enhance the traffic movement efficiency in disaster-affected communities during the disaster response period. Furthermore, we demonstrate that contrary to exiting studies, shortening restoration period is not an appropriate measure of efficiency for post-disaster restoration operations. In fact, reducing the restoration period may adversely impact traffic movements during the response phase in post-disaster situations.

本文介绍了一种人工智能驱动的受灾地区短期路网修复优化方法，旨在最大限度地提高灾后交通加速速度。我们的方法考虑了旅行者的行为、对网络变化的逐渐适应、有限的恢复资源以及恢复时间的不确定性。解决这些复杂性需要一种随机方法来解决不确定性，需要一种连续的资源管理决策，需要一种无模型的技术来模拟旅行者的适应。为了应对这些挑战，我们开发了旅行者自适应恢复机制（TARM），整合了强化学习（RL）、马尔可夫决策过程（MDP）和基于优化的日常交通模拟。基于历史数据，以苏福尔斯市龙卷风情景下的路网为例对该方法进行了评价。结果显示出行者路线选择和恢复信息传播速度对最优策略的影响。研究结果表明，通过增加道路恢复资源来加快道路恢复进程并不一定能提高受灾社区在灾害响应期间的交通运行效率。此外，我们证明了与现有研究相反，缩短恢复周期并不是灾后恢复作业效率的适当衡量标准。事实上，在灾后情况下，缩短恢复时间可能会对响应阶段的交通流动产生不利影响。

{"title":"Optimizing post-disaster road restoration with reinforcement learning: A traveler-behavior-aware approach","authors":"Maryam Babaee , Namrata Saha , Frank Mediavilla Ponce , Shabnam Rezapour , M. Hadi Amini","doi":"10.1016/j.ress.2026.112371","DOIUrl":"10.1016/j.ress.2026.112371","url":null,"abstract":"<div><div>This paper introduces an AI-driven method to optimize short-term road network restoration in disaster-affected areas, aiming to maximize post-disaster traffic acceleration. Our approach considers travelers' behavior, gradual adaptation to network changes, limited recovery resources, and uncertainties in recovery times. Addressing these complexities requires a stochastic approach for uncertainties, sequential decision-making for resource management, and a model-free technique for simulating traveler adaptation.</div><div>To tackle these challenges, we develop the Traveler-Adaptive Restoration Mechanism (TARM), integrating Reinforcement Learning (RL), the Markov Decision Process (MDP), and optimization-based day-to-day traffic simulation. The method is evaluated on Sioux Falls' road network under tornado scenarios based on historical data. Results highlight the influence of travelers’ route choices and the speed of restoration information dissemination on optimal policies.</div><div>Findings reveal that accelerating the road restoration process by increasing restoration resources does not necessarily enhance the traffic movement efficiency in disaster-affected communities during the disaster response period. Furthermore, we demonstrate that contrary to exiting studies, shortening restoration period is not an appropriate measure of efficiency for post-disaster restoration operations. In fact, reducing the restoration period may adversely impact traffic movements during the response phase in post-disaster situations.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112371"},"PeriodicalIF":11.0,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174992","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}

引用次数: 0

Deep transfer learning based on cross-domain subsequence alignment and feature contribution interpretation for remaining useful life prediction 基于跨域子序列比对和特征贡献解释的深度迁移学习剩余使用寿命预测

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-05 DOI: 10.1016/j.ress.2026.112370

Tongshan Liu, Yiming Li, Zhihao Hu, Congjie Fu, Guiqiu Song

Deep transfer learning enables the transfer of knowledge learned from a source domain to a target domain, thereby alleviating the few-shot problem in remaining useful life (RUL) prediction tasks. However, due to differences in operating conditions and working scenarios, a significant domain shift commonly exists between the source and target domains. Existing cross-domain transfer learning still faces the following challenges: difficulties in cross-domain subsequence alignment, insufficient robustness under few-shot conditions, and limited interpretability of model decisions, which collectively degrade the stability and reliability of prediction results. To address these issues, this paper proposes a RUL transfer prediction method based on cross-domain subsequence alignment and feature contribution interpretation to optimize the cross-domain transfer process. First, a multi-scale subsequence alignment strategy is developed to select the most informative source domain subsequences, thereby reducing the risk of negative transfer. Second, an interpretable cross-domain transfer learning model is constructed by embedding a self-attention mechanism and a Bayesian linear layer into the domain adaptation framework, which enhances robustness in cross-domain prediction and mitigates overfitting under few-shot conditions in the target domain. Finally, a lightweight analysis method for identifying critical time steps and feature contributions is established to further improve the interpretability of the decision-making process. Experimental results on the IEEE PHM 2012 bearing dataset and the XJTU-SY bearing dataset demonstrate the effectiveness of the proposed method, achieving an RMSE of 0.06, an MAE of 0.05, and an R² of 0.96.

深度迁移学习能够将从源领域学习到的知识转移到目标领域，从而缓解了剩余使用寿命（RUL）预测任务中的少射问题。然而，由于操作条件和工作场景的差异，源域和目标域之间通常存在显著的域转移。现有的跨域迁移学习仍然面临着以下挑战：跨域子序列对齐困难，少镜头条件下鲁棒性不足，模型决策的可解释性有限，这些共同降低了预测结果的稳定性和可靠性。针对这些问题，本文提出了一种基于跨域子序列对齐和特征贡献解释的规则语言迁移预测方法，以优化跨域迁移过程。首先，提出了一种多尺度子序列对齐策略，选择信息量最大的源域子序列，从而降低了负迁移的风险。其次，通过在领域自适应框架中嵌入自注意机制和贝叶斯线性层，构建了可解释的跨领域迁移学习模型，增强了跨领域预测的鲁棒性，缓解了目标领域在少数条件下的过拟合问题；最后，建立了一种用于识别关键时间步长和特征贡献的轻量级分析方法，以进一步提高决策过程的可解释性。在IEEE PHM 2012轴承数据集和XJTU-SY轴承数据集上的实验结果表明了该方法的有效性，RMSE为0.06，MAE为0.05，R²为0.96。

{"title":"Deep transfer learning based on cross-domain subsequence alignment and feature contribution interpretation for remaining useful life prediction","authors":"Tongshan Liu, Yiming Li, Zhihao Hu, Congjie Fu, Guiqiu Song","doi":"10.1016/j.ress.2026.112370","DOIUrl":"10.1016/j.ress.2026.112370","url":null,"abstract":"<div><div>Deep transfer learning enables the transfer of knowledge learned from a source domain to a target domain, thereby alleviating the few-shot problem in remaining useful life (RUL) prediction tasks. However, due to differences in operating conditions and working scenarios, a significant domain shift commonly exists between the source and target domains. Existing cross-domain transfer learning still faces the following challenges: difficulties in cross-domain subsequence alignment, insufficient robustness under few-shot conditions, and limited interpretability of model decisions, which collectively degrade the stability and reliability of prediction results. To address these issues, this paper proposes a RUL transfer prediction method based on cross-domain subsequence alignment and feature contribution interpretation to optimize the cross-domain transfer process. First, a multi-scale subsequence alignment strategy is developed to select the most informative source domain subsequences, thereby reducing the risk of negative transfer. Second, an interpretable cross-domain transfer learning model is constructed by embedding a self-attention mechanism and a Bayesian linear layer into the domain adaptation framework, which enhances robustness in cross-domain prediction and mitigates overfitting under few-shot conditions in the target domain. Finally, a lightweight analysis method for identifying critical time steps and feature contributions is established to further improve the interpretability of the decision-making process. Experimental results on the IEEE PHM 2012 bearing dataset and the XJTU-SY bearing dataset demonstrate the effectiveness of the proposed method, achieving an RMSE of 0.06, an MAE of 0.05, and an R² of 0.96.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112370"},"PeriodicalIF":11.0,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174767","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}

引用次数: 0

Joint optimization of maintenance and spare parts management in upstream–downstream systems under quality control 在质量控制下，对上下游系统的维修和备件管理进行联合优化

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-05 DOI: 10.1016/j.ress.2026.112355

Zha Yang , Ren Lina , Ding Jiajun , Kong Xiangyuan

This paper studies an upstream–downstream serial production system in which maintenance, quality control, and spare-parts management are commonly optimized in isolation, and proposes a joint optimization strategy to coordinate their coupled dynamics. To sustain production continuity, a three-layer maintenance mechanism is developed: (i) within-lot quality-triggered preventive replacement, (ii) between-lot degradation-driven preventive maintenance, and (iii) post-failure corrective maintenance. For quality assurance and demand buffering, we adopt a “two inspections and one inventory” structure, consisting of an in-process inspection after the upstream stage, a final inspection after the downstream stage, and a capped finished-goods inventory. Spare-parts provisioning is differentiated: the upstream stage is controlled by an (s, S) spare-parts pool, while the downstream stage applies a dual-mode ordering policy conditioned on the real-time finished-goods inventory level. The resulting multi-scenario cost model is optimized using discrete-event simulation, response surface methodology, and a genetic algorithm. Numerical experiments yield an average cost rate of 351.97 $/day, with savings of 7.08 $/day versus Model 1 (without quality control), 10.51 $/day versus Model 2 (without differentiated spare-parts provisioning), and 15.58 $/day versus the baseline Model 3, demonstrating improved coordination and cost-effectiveness.

本文研究了维修、质量控制和备件管理通常孤立优化的上下游串行生产系统，提出了一种协调其耦合动态的联合优化策略。为了维持生产连续性，开发了三层维护机制：(i)批次内质量触发的预防性更换，（ii）批次间退化驱动的预防性维护，以及（iii）故障后纠正维护。为了保证质量和缓冲需求，我们采用“两检一存”的结构，即上游阶段结束后进行中检，下游阶段结束后进行终检，成品库存上限。备品备件供应差异化：上游阶段由（s, s）备品备件池控制，下游阶段采用以实时成品库存水平为条件的双模式订购策略。使用离散事件模拟、响应面方法和遗传算法对所得到的多场景成本模型进行优化。数值实验得出的平均成本率为351.97美元/天，与模型1（没有质量控制）相比节省7.08美元/天，与模型2（没有差异化的备件供应）相比节省10.51美元/天，与基线模型3相比节省15.58美元/天，显示出改进的协调和成本效益。

{"title":"Joint optimization of maintenance and spare parts management in upstream–downstream systems under quality control","authors":"Zha Yang , Ren Lina , Ding Jiajun , Kong Xiangyuan","doi":"10.1016/j.ress.2026.112355","DOIUrl":"10.1016/j.ress.2026.112355","url":null,"abstract":"<div><div>This paper studies an upstream–downstream serial production system in which maintenance, quality control, and spare-parts management are commonly optimized in isolation, and proposes a joint optimization strategy to coordinate their coupled dynamics. To sustain production continuity, a three-layer maintenance mechanism is developed: (i) within-lot quality-triggered preventive replacement, (ii) between-lot degradation-driven preventive maintenance, and (iii) post-failure corrective maintenance. For quality assurance and demand buffering, we adopt a “two inspections and one inventory” structure, consisting of an in-process inspection after the upstream stage, a final inspection after the downstream stage, and a capped finished-goods inventory. Spare-parts provisioning is differentiated: the upstream stage is controlled by an (<em>s, S</em>) spare-parts pool, while the downstream stage applies a dual-mode ordering policy conditioned on the real-time finished-goods inventory level. The resulting multi-scenario cost model is optimized using discrete-event simulation, response surface methodology, and a genetic algorithm. Numerical experiments yield an average cost rate of 351.97 $/day, with savings of 7.08 $/day versus Model 1 (without quality control), 10.51 $/day versus Model 2 (without differentiated spare-parts provisioning), and 15.58 $/day versus the baseline Model 3, demonstrating improved coordination and cost-effectiveness.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112355"},"PeriodicalIF":11.0,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174954","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}

引用次数: 0

A hybrid neural network-based concrete gravity dam seismic response prediction method quantifying reservoir water level uncertainty 基于混合神经网络的混凝土重力坝地震反应预测方法，量化水库水位不确定性

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-05 DOI: 10.1016/j.ress.2026.112366

Bo Liu , Qiang Xu , Jianying Xing , Jianyun Chen , Mingming Wang , Jing Li , Tianran Zhang

Accurate real-time prediction of the seismic response of gravity dams is critical for their safety assessment, however, the seismic response characteristics of gravity dams are highly variable due to variations in the gravity dam-reservoir-foundation system (e.g., reservoir water level) and seismic characteristics (e.g., impulse characteristics). To address these challenges, this study proposes a hybrid neural network model (SW-HNN) that integrates a new system feature attention mechanism and a unique wavelet decomposition-based impulse identification module to predict the impulse seismic response of gravity dams, considering reservoir water level variability. The model effectively captures the complex interrelations between system characteristics of the gravity dam-reservoir-foundation system, ground motion impulse characteristics, and dam responses. To further enhance model performance, an improved balanced sampling technique is developed for ground motion datasets, which enriches the feature set and mitigates the influence of imbalanced feature distributions. The required datasets for model training and validation are generated through nonlinear time-history analyses of gravity dam-reservoir-foundation systems with varying reservoir water levels. Experimental results confirm the accuracy and robustness of the SW-HNN model. The validity and superiority of the SW-HNN model are further verified by ablation analysis and comparison experiments. Additionally, the SW-HNN is an interpretable deep learning model capable of ranking system features by evaluating changes in its internal parameters.

准确实时预测重力坝的地震反应对其安全性评价至关重要，然而，由于重力坝-水库-基础系统（如水库水位）和地震特征（如冲击特性）的变化，重力坝的地震反应特征变化很大。为了解决这些挑战，本研究提出了一种混合神经网络模型（SW-HNN），该模型集成了一种新的系统特征注意机制和一种独特的基于小波分解的脉冲识别模块，用于在考虑水库水位变化的情况下预测重力坝的脉冲地震响应。该模型有效地捕捉了重力坝-水库-基础系统的系统特性、地震动脉冲特性和大坝响应之间的复杂相互关系。为了进一步提高模型的性能，提出了一种改进的平衡采样技术，丰富了地面运动数据的特征集，减轻了不平衡特征分布的影响。模型训练和验证所需的数据集是通过对具有不同水库水位的重力坝-水库-基础系统的非线性时程分析生成的。实验结果验证了SW-HNN模型的准确性和鲁棒性。通过烧蚀分析和对比实验，进一步验证了SW-HNN模型的有效性和优越性。此外，SW-HNN是一个可解释的深度学习模型，能够通过评估其内部参数的变化来对系统特征进行排名。

{"title":"A hybrid neural network-based concrete gravity dam seismic response prediction method quantifying reservoir water level uncertainty","authors":"Bo Liu , Qiang Xu , Jianying Xing , Jianyun Chen , Mingming Wang , Jing Li , Tianran Zhang","doi":"10.1016/j.ress.2026.112366","DOIUrl":"10.1016/j.ress.2026.112366","url":null,"abstract":"<div><div>Accurate real-time prediction of the seismic response of gravity dams is critical for their safety assessment, however, the seismic response characteristics of gravity dams are highly variable due to variations in the gravity dam-reservoir-foundation system (e.g., reservoir water level) and seismic characteristics (e.g., impulse characteristics). To address these challenges, this study proposes a hybrid neural network model (SW-HNN) that integrates a new system feature attention mechanism and a unique wavelet decomposition-based impulse identification module to predict the impulse seismic response of gravity dams, considering reservoir water level variability. The model effectively captures the complex interrelations between system characteristics of the gravity dam-reservoir-foundation system, ground motion impulse characteristics, and dam responses. To further enhance model performance, an improved balanced sampling technique is developed for ground motion datasets, which enriches the feature set and mitigates the influence of imbalanced feature distributions. The required datasets for model training and validation are generated through nonlinear time-history analyses of gravity dam-reservoir-foundation systems with varying reservoir water levels. Experimental results confirm the accuracy and robustness of the SW-HNN model. The validity and superiority of the SW-HNN model are further verified by ablation analysis and comparison experiments. Additionally, the SW-HNN is an interpretable deep learning model capable of ranking system features by evaluating changes in its internal parameters.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112366"},"PeriodicalIF":11.0,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174951","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}

引用次数: 0

Active learning Kriging with functional dimension reduction for reliability analysis of stochastic dynamical systems 基于功能降维的主动学习Kriging随机动力系统可靠性分析

IF 11 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Reliability Engineering & System Safety

Pub Date : 2026-02-05 DOI: 10.1016/j.ress.2026.112360

Zhouzhou Song , Chao Dang , Marcos A. Valdebenito , Matthias G.R. Faes

In engineering applications, it is important to evaluate the reliability of dynamical systems under various uncertainties arising from materials, manufacturing processes, and external excitations. Surrogate models are widely employed to enable efficient reliability analysis in complex, computationally intensive engineering problems. However, building high-accuracy surrogate models for estimating dynamic systems’ reliability with limited computational resources remains a significant challenge. This paper presents a novel active learning Kriging method based on functional dimension reduction (AKFDR) for the efficient estimation of first-passage failure probabilities of stochastic dynamical systems. In this approach, Kriging surrogate models are constructed in a latent functional space obtained through functional dimension reduction, enabling probabilistic predictions of dynamic responses. By leveraging the prediction uncertainty and the concept of trajectory misclassification probability (TMP), a new learning function incorporating a weighted correlation criterion is then developed to guide the selection of the best next sample for model enhancement. Furthermore, an error-based stopping criterion is proposed to judge the convergence of the active learning process. The final surrogate model is then used to estimate the first-passage failure probability via Monte Carlo simulation. Through three numerical examples of varying dimensionality and complexity, it is shown that the proposed method is efficient and accurate for first-passage probability evaluation of stochastic dynamical systems.

在工程应用中，评估动力系统在由材料、制造过程和外部激励引起的各种不确定性下的可靠性是很重要的。代理模型被广泛应用于复杂的、计算密集型的工程问题中，以实现高效的可靠性分析。然而，在有限的计算资源下建立高精度的替代模型来估计动态系统的可靠性仍然是一个重大挑战。提出了一种基于泛函降维（AKFDR）的主动学习Kriging方法，用于随机动力系统首路失效概率的有效估计。在这种方法中，Kriging代理模型在通过功能降维获得的潜在功能空间中构建，从而实现动态响应的概率预测。利用预测的不确定性和轨迹误分类概率（TMP）的概念，建立了一个新的学习函数，结合加权相关准则来指导选择最佳的下一个样本进行模型增强。此外，提出了一种基于误差的停止准则来判断主动学习过程的收敛性。通过蒙特卡罗模拟，利用最终的代理模型来估计首通道失效概率。通过三个不同维数和复杂度的数值算例，表明该方法对于随机动力系统的首通概率评估是有效和准确的。

{"title":"Active learning Kriging with functional dimension reduction for reliability analysis of stochastic dynamical systems","authors":"Zhouzhou Song , Chao Dang , Marcos A. Valdebenito , Matthias G.R. Faes","doi":"10.1016/j.ress.2026.112360","DOIUrl":"10.1016/j.ress.2026.112360","url":null,"abstract":"<div><div>In engineering applications, it is important to evaluate the reliability of dynamical systems under various uncertainties arising from materials, manufacturing processes, and external excitations. Surrogate models are widely employed to enable efficient reliability analysis in complex, computationally intensive engineering problems. However, building high-accuracy surrogate models for estimating dynamic systems’ reliability with limited computational resources remains a significant challenge. This paper presents a novel active learning Kriging method based on functional dimension reduction (AKFDR) for the efficient estimation of first-passage failure probabilities of stochastic dynamical systems. In this approach, Kriging surrogate models are constructed in a latent functional space obtained through functional dimension reduction, enabling probabilistic predictions of dynamic responses. By leveraging the prediction uncertainty and the concept of trajectory misclassification probability (TMP), a new learning function incorporating a weighted correlation criterion is then developed to guide the selection of the best next sample for model enhancement. Furthermore, an error-based stopping criterion is proposed to judge the convergence of the active learning process. The final surrogate model is then used to estimate the first-passage failure probability via Monte Carlo simulation. Through three numerical examples of varying dimensionality and complexity, it is shown that the proposed method is efficient and accurate for first-passage probability evaluation of stochastic dynamical systems.</div></div>","PeriodicalId":54500,"journal":{"name":"Reliability Engineering & System Safety","volume":"273 ","pages":"Article 112360"},"PeriodicalIF":11.0,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174995","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}

引用次数: 0