Human reliability analysis plays a critical role in maintaining the safety and operational performance of complex industrial systems. However, conventional human reliability analysis data collection methods often suffer from limited granularity, static representations of human behavior, and heavy reliance on expert judgment, making them time-consuming and difficult to scale. To address these limitations, this paper proposes a novel, scenario-driven framework for the automated acquisition and estimation of human cognitive workload in industrial control settings. Leveraging fine-tuned large language models (LLMs) trained on authentic operational logs from high-temperature gas-cooled reactors, the proposed method simulates real-time workload dynamics across multiple roles, including reactor operators, shift supervisors, and secondary loop operator. The resulting system, termed Workload Estimation with LLMs and Agents (WELLA), integrates agent-based simulation and dynamic LLM reasoning to capture heterogeneous workload patterns and fluctuations during collaborative tasks. Experimental evaluations demonstrate that WELLA achieves superior prediction accuracy and adaptability compared to existing commercial LLM-based solutions. These findings highlight WELLA’s potential to enhance human reliability methodologies through scalable, data-driven workload modeling in complex socio-technical environments.
{"title":"An intelligent framework for automated human reliability data generation in complex industrial systems","authors":"Xingyu Xiao , Peng Chen , Qianqian Jia , Jiejuan Tong , Jun Zhao , Hongru Zhao , Jingang Liang , Haitao Wang","doi":"10.1016/j.cie.2026.111807","DOIUrl":"10.1016/j.cie.2026.111807","url":null,"abstract":"<div><div>Human reliability analysis plays a critical role in maintaining the safety and operational performance of complex industrial systems. However, conventional human reliability analysis data collection methods often suffer from limited granularity, static representations of human behavior, and heavy reliance on expert judgment, making them time-consuming and difficult to scale. To address these limitations, this paper proposes a novel, scenario-driven framework for the automated acquisition and estimation of human cognitive workload in industrial control settings. Leveraging fine-tuned large language models (LLMs) trained on authentic operational logs from high-temperature gas-cooled reactors, the proposed method simulates real-time workload dynamics across multiple roles, including reactor operators, shift supervisors, and secondary loop operator. The resulting system, termed Workload Estimation with LLMs and Agents (WELLA), integrates agent-based simulation and dynamic LLM reasoning to capture heterogeneous workload patterns and fluctuations during collaborative tasks. Experimental evaluations demonstrate that WELLA achieves superior prediction accuracy and adaptability compared to existing commercial LLM-based solutions. These findings highlight WELLA’s potential to enhance human reliability methodologies through scalable, data-driven workload modeling in complex socio-technical environments.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111807"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145927274","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 : 2026-03-01Epub Date: 2026-01-03DOI: 10.1016/j.cie.2025.111795
Maria Boluda-Prieto, Miguel A. Mateo-Casali, Francisco Fraile, Faustino Alarcon
In the era of Industry 4.0, advanced technologies are transforming production processes, although their adoption introduces challenges related to integration, security and scalability in industrial environments. This study investigates the preservation of industrial data continuity under stress conditions. A structured literature review was conducted in Scopus and Web of Science, following the PRISMA methodology, to identify the resilience requirements. This was followed by the design and implementation of a hybrid edge-to-cloud (E2C) microservices architecture. The approach was validated through controlled overload and saturation experiments in a hybrid cloud–edge environment, utilising orchestration, monitoring, and alert-management tools such as Kubernetes, Rancher, Prometheus, Grafana, and Node-RED. The proposed architecture combines edge node execution with centralised cloud coordination (E2C) to address these challenges. A central motivation for this design is the recognition that industrial data represents a strategic asset. Interruptions or losses in data flows directly affect process traceability and degrade the performance of AI-driven monitoring and optimisation models. Such losses also trigger reprocessing, waste and suboptimal operational decisions. Therefore, ensuring data continuity and availability becomes an essential requirement for competitiveness and industrial sustainability. In this context, the study’s main contribution is integrating orchestration, alert management and monitoring tools into a closed-loop system that triggers self-healing and self-correcting actions when predefined thresholds are exceeded. This approach avoids systematic network failures, enables autonomous recovery without manual intervention and keeps system operations and data available. The paper validates the self-healing and self-correcting mechanisms through overload and saturation experiments. Results demonstrate the feasibility of a zero-touch strategy that enhances efficiency and sustainability in industrial automation, supporting the transition towards Industry 5.0.
在工业4.0时代，先进技术正在改变生产流程，尽管它们的采用带来了与工业环境中的集成、安全性和可扩展性相关的挑战。本研究探讨在压力条件下工业数据连续性的保存。根据PRISMA方法，在Scopus和Web of Science上进行了结构化的文献综述，以确定弹性需求。接下来是混合型边缘到云（E2C）微服务架构的设计和实现。该方法在混合云边缘环境中通过控制过载和饱和实验进行了验证，该实验利用了编排、监控和警报管理工具，如Kubernetes、Rancher、Prometheus、Grafana和Node-RED。提出的架构将边缘节点执行与集中式云协调（E2C）相结合，以应对这些挑战。这种设计的核心动机是认识到工业数据代表着一种战略资产。数据流的中断或丢失直接影响流程的可追溯性，并降低人工智能驱动的监控和优化模型的性能。这种损失还会引发再处理、浪费和次优操作决策。因此，确保数据的连续性和可用性成为竞争力和工业可持续性的基本要求。在这种情况下，该研究的主要贡献是将编排、警报管理和监控工具集成到一个闭环系统中，该系统在超过预定义阈值时触发自我修复和自我纠正行动。这种方法避免了系统网络故障，无需人工干预即可实现自动恢复，并保持系统操作和数据可用。通过过载和饱和实验验证了自修复和自校正机制。结果表明，零接触战略的可行性，提高了工业自动化的效率和可持续性，支持向工业5.0的过渡。
{"title":"Resilient edge-to-cloud architecture with self-healing and self-correcting mechanisms for industrial data continuity","authors":"Maria Boluda-Prieto, Miguel A. Mateo-Casali, Francisco Fraile, Faustino Alarcon","doi":"10.1016/j.cie.2025.111795","DOIUrl":"10.1016/j.cie.2025.111795","url":null,"abstract":"<div><div>In the era of Industry 4.0, advanced technologies are transforming production processes, although their adoption introduces challenges related to integration, security and scalability in industrial environments. <em>This study investigates the preservation of industrial data continuity under stress conditions. A structured literature review was conducted in Scopus and Web of Science, following the PRISMA methodology, to identify the resilience requirements. This was followed by the design and implementation of a hybrid edge-to-cloud (E2C) microservices architecture. The approach was validated through controlled overload and saturation experiments in a hybrid cloud–edge environment, utilising orchestration, monitoring, and alert-management tools such as Kubernetes, Rancher, Prometheus, Grafana, and Node-RED. The proposed architecture</em> combines edge node execution with centralised cloud coordination (E2C) to address these challenges. A central motivation for this design is the recognition that industrial data represents a strategic asset. Interruptions or losses in data flows directly affect process traceability and degrade the performance of AI-driven monitoring and optimisation models. Such losses also trigger reprocessing, waste and suboptimal operational decisions. Therefore, ensuring data continuity and availability becomes an essential requirement for competitiveness and industrial sustainability. In this context, the study’s main contribution is integrating orchestration, alert management and monitoring tools into a closed-loop system that triggers self-healing and self-correcting actions when predefined thresholds are exceeded. This approach avoids systematic network failures, enables autonomous recovery without manual intervention and keeps system operations and data available. The paper validates the self-healing and self-correcting mechanisms through overload and saturation experiments. Results demonstrate the feasibility of a zero-touch strategy that enhances efficiency and sustainability in industrial automation, supporting the transition towards Industry 5.0.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111795"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145927778","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 : 2026-03-01Epub Date: 2025-12-19DOI: 10.1016/j.cie.2025.111756
N. Orkun Baycik , Canan G. Corlu , J. Gregory McDaniel , Alyssa Pierson
Illegal, unreported, and unregulated (IUU) fishing, representing about 26 million tons of fish caught annually, destroys our environment, economy, and society. The United Nations Food and Agriculture Organization recognizes IUU fishing as a global problem due to its impact on the environment and sustainability, global food security, and its association with other organized crimes. The goal of this paper is to present research opportunities that, when pursued, will benefit government agencies and non-profit organizations to tackle this complex societal problem. Through a comprehensive review of the literature and consultation with nonprofit organizations, we propose the first steps to develop new facility location and covering, network flow, and interdiction models to address IUU fishing. We highlight that collaborative efforts using operations research and analytics are necessary. The proposed models can contribute to practice by improving surveillance, understanding criminal operations, and disrupting illegal networks.
{"title":"Nation’s fight against illegal fishing: Research opportunities in operations research and management science","authors":"N. Orkun Baycik , Canan G. Corlu , J. Gregory McDaniel , Alyssa Pierson","doi":"10.1016/j.cie.2025.111756","DOIUrl":"10.1016/j.cie.2025.111756","url":null,"abstract":"<div><div>Illegal, unreported, and unregulated (IUU) fishing, representing about 26 million tons of fish caught annually, destroys our environment, economy, and society. The United Nations Food and Agriculture Organization recognizes IUU fishing as a global problem due to its impact on the environment and sustainability, global food security, and its association with other organized crimes. The goal of this paper is to present research opportunities that, when pursued, will benefit government agencies and non-profit organizations to tackle this complex societal problem. Through a comprehensive review of the literature and consultation with nonprofit organizations, we propose the first steps to develop new facility location and covering, network flow, and interdiction models to address IUU fishing. We highlight that collaborative efforts using operations research and analytics are necessary. The proposed models can contribute to practice by improving surveillance, understanding criminal operations, and disrupting illegal networks.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111756"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841683","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 : 2026-03-01Epub Date: 2025-12-17DOI: 10.1016/j.cie.2025.111736
Zijian Ling, Youlong Yang, An Huang
Social network group decision-making (SNGDM) provides valuable support for describing the opinion exchange in the decision-making process by using the connected social relationships among decision makers (DMs). With the expansion of social media, DMs are interconnected through various types of links. In this cases, interaction of DMs are no longer confined to single-type binary relationships but exhibit complex multiplexing and high-order dynamic characteristics. To this end, this study develops a consensus model based on multilayer network for improving the reliability of decision-making. First, we construct an attributed multilayer network by utilizing multiple social relationships and decision information, in which attributes serve as auxiliary information to establish additional exotic connectivity patterns. Then, the natural interaction of DMs shows a specific high-order correlation, where some activities occurring over the links of a layer depend on the dynamics of certain links on other layers. We propose an interactive joint random walk model to map this co-evolution into an activity-driven network dynamics process. To accurately capture hidden collective structure, state-based non-columnar communities and physical-based overlapping communities are detected. The reinforcement effects generated in these two types of communities can identify influential nodes and communities, guiding decision aggregation to reach higher consensus level. Finally, a numerical example is presented, and simulation experiments and comparative analysis are performed to validate the effectiveness and superiority of proposed model.
社会网络群体决策（Social network group decision， SNGDM）利用决策者之间的关联社会关系，为描述决策过程中的意见交换提供了有价值的支持。随着社交媒体的扩展，dm通过各种类型的链接相互连接。在这种情况下，dm的相互作用不再局限于单一类型的二元关系，而是表现出复杂的多路复用和高阶动态特性。为此，本研究开发了基于多层网络的共识模型，以提高决策的可靠性。首先，我们利用多种社会关系和决策信息构建了一个带有属性的多层网络，其中属性作为辅助信息来建立额外的外部连接模式。然后，dm的自然相互作用显示出特定的高阶相关性，其中在一层的链接上发生的一些活动依赖于其他层上某些链接的动态。我们提出了一个交互式联合随机行走模型，将这种共同进化映射为一个活动驱动的网络动力学过程。为了准确捕获隐藏的集体结构，检测基于状态的非柱状群落和基于物理的重叠群落。这两类群体产生的强化效应可以识别有影响的节点和群体，引导决策聚集达到更高的共识水平。最后给出了一个数值算例，并进行了仿真实验和对比分析，验证了所提模型的有效性和优越性。
{"title":"An activity-driven temporal multilayer network framework to support consensus in group decision making","authors":"Zijian Ling, Youlong Yang, An Huang","doi":"10.1016/j.cie.2025.111736","DOIUrl":"10.1016/j.cie.2025.111736","url":null,"abstract":"<div><div>Social network group decision-making (SNGDM) provides valuable support for describing the opinion exchange in the decision-making process by using the connected social relationships among decision makers (DMs). With the expansion of social media, DMs are interconnected through various types of links. In this cases, interaction of DMs are no longer confined to single-type binary relationships but exhibit complex multiplexing and high-order dynamic characteristics. To this end, this study develops a consensus model based on multilayer network for improving the reliability of decision-making. First, we construct an attributed multilayer network by utilizing multiple social relationships and decision information, in which attributes serve as auxiliary information to establish additional exotic connectivity patterns. Then, the natural interaction of DMs shows a specific high-order correlation, where some activities occurring over the links of a layer depend on the dynamics of certain links on other layers. We propose an interactive joint random walk model to map this co-evolution into an activity-driven network dynamics process. To accurately capture hidden collective structure, state-based non-columnar communities and physical-based overlapping communities are detected. The reinforcement effects generated in these two types of communities can identify influential nodes and communities, guiding decision aggregation to reach higher consensus level. Finally, a numerical example is presented, and simulation experiments and comparative analysis are performed to validate the effectiveness and superiority of proposed model.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111736"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841614","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 : 2026-03-01Epub Date: 2026-01-13DOI: 10.1016/j.cie.2026.111825
Gwang Kim , Youngchul Shin , Yoonjea Jeong
In this study, we address the challenge of reliable monitoring using unmanned aerial vehicles (UAVs) to minimize the sum of travel costs associated with monitoring activities over a specified period. UAV systems are prone to failures caused by uncertainties and unforeseen factors. These disruptions can interfere with system operations, thereby affecting overall performance. The model considers uncertainties related to UAV failure and aims to minimize additional losses incurred due to these uncertainties. We formulate the problem as a two-stage programming model, consisting of here-and-now decisions in the first stage and recourse decision in the second stage. We utilize the sample average approximation (SAA) approach to address the reliable monitoring problem with UAV failure. A solution methodology based on the scenario decomposition technique is employed to enhance the computational efficiency of the SAA method. In addition, numerical experiments are conducted to evaluate statistical estimates of the model bounds using SAA problems and to assess the performance of the proposed algorithm.
{"title":"Scenario decomposition approach for mobile multi-agent monitoring under failure","authors":"Gwang Kim , Youngchul Shin , Yoonjea Jeong","doi":"10.1016/j.cie.2026.111825","DOIUrl":"10.1016/j.cie.2026.111825","url":null,"abstract":"<div><div>In this study, we address the challenge of reliable monitoring using unmanned aerial vehicles (UAVs) to minimize the sum of travel costs associated with monitoring activities over a specified period. UAV systems are prone to failures caused by uncertainties and unforeseen factors. These disruptions can interfere with system operations, thereby affecting overall performance. The model considers uncertainties related to UAV failure and aims to minimize additional losses incurred due to these uncertainties. We formulate the problem as a two-stage programming model, consisting of <em>here-and-now</em> decisions in the first stage and <em>recourse decision</em> in the second stage. We utilize the sample average approximation (SAA) approach to address the reliable monitoring problem with UAV failure. A solution methodology based on the scenario decomposition technique is employed to enhance the computational efficiency of the SAA method. In addition, numerical experiments are conducted to evaluate statistical estimates of the model bounds using SAA problems and to assess the performance of the proposed algorithm.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111825"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977748","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 : 2026-03-01Epub Date: 2025-12-30DOI: 10.1016/j.cie.2025.111780
Jin Yi , Hanhai Shou , Xinyu Li , Wuzhuang Zhou , Chih-Hsing Chu
Five-axis flank milling is an advanced machining technology widely employed for producing complex geometries, where tool path planning directly influences surface quality and machining efficiency. However, the mathematical nature of the problem presents a large-scale global optimization challenge, involving the search for optimal solutions within a high-dimensional decision space. To address this issue, this paper provides a review of eight state-of-the-art large-scale global optimization algorithms and conducts a comparative evaluation of their performance on representative test geometries. Analyses of the test results reveal key factors that determine the computational performance of the algorithms from multiple perspectives. These insights further inform the design of advanced evolutionary optimization algorithms for effective tool path planning in five-axis flank milling.
{"title":"Benchmarking state-of-the-art large-scale global optimization algorithms for tool path planning in five-axis flank milling: A comparative performance analysis","authors":"Jin Yi , Hanhai Shou , Xinyu Li , Wuzhuang Zhou , Chih-Hsing Chu","doi":"10.1016/j.cie.2025.111780","DOIUrl":"10.1016/j.cie.2025.111780","url":null,"abstract":"<div><div>Five-axis flank milling is an advanced machining technology widely employed for producing complex geometries, where tool path planning directly influences surface quality and machining efficiency. However, the mathematical nature of the problem presents a large-scale global optimization challenge, involving the search for optimal solutions within a high-dimensional decision space. To address this issue, this paper provides a review of eight state-of-the-art large-scale global optimization algorithms and conducts a comparative evaluation of their performance on representative test geometries. Analyses of the test results reveal key factors that determine the computational performance of the algorithms from multiple perspectives. These insights further inform the design of advanced evolutionary optimization algorithms for effective tool path planning in five-axis flank milling.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111780"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885434","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}
Fully automated workshops equipped with robots are trending in the manufacturing industry because of the rapid development in automation. This study addresses the flexible job shop scheduling problem with robot constraints to minimize the makespan. To this end, a mixed integer linear programming (MILP) and constraint programming (CP) models are developed, to achieve optimal solutions for small-scale instances. Further, a novel CP-assisted evolutionary algorithm with deep Q-network (EA-DQN-CP) is proposed to solve the problem effectively given its NP-hard nature, because this algorithm can comprehensively utilize advantages of the CP model, evolutionary algorithm (EA) and deep Q-network (DQN). EA-DQN-CP includes two stages: (1) A DQN-assisted EA (EA-DQN) is designed for obtaining a high-quality solution efficiently, wherein the DQN helps EA to select suitable search operators. (2) The CP model is used to optimize the obtained solution of EA-DQN. Experimental results demonstrate the effectiveness of MILP model, CP model, and EA-DQN-CP.
{"title":"A novel Constraint Programming-Assisted evolutionary algorithm with deep Q-network for flexible job shop scheduling problem with robot constraints","authors":"Weiyao Cheng , Leilei Meng , Yushuai Zhang , Chaoyong Zhang , Biao Zhang , Hongyan Sang","doi":"10.1016/j.cie.2025.111793","DOIUrl":"10.1016/j.cie.2025.111793","url":null,"abstract":"<div><div>Fully automated workshops equipped with robots are trending in the manufacturing industry because of the rapid development in automation. This study addresses the flexible job shop scheduling problem with robot constraints to minimize the makespan. To this end, a mixed integer linear programming (MILP) and constraint programming (CP) models are developed, to achieve optimal solutions for small-scale instances. Further, a novel CP-assisted evolutionary algorithm with deep Q-network (EA-DQN-CP) is proposed to solve the problem effectively given its NP-hard nature, because this algorithm can comprehensively utilize advantages of the CP model, evolutionary algorithm (EA) and deep Q-network (DQN). EA-DQN-CP includes two stages: (1) A DQN-assisted EA (EA-DQN) is designed for obtaining a high-quality solution efficiently, wherein the DQN helps EA to select suitable search operators. (2) The CP model is used to optimize the obtained solution of EA-DQN. Experimental results demonstrate the effectiveness of MILP model, CP model, and EA-DQN-CP.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111793"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885572","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 : 2026-03-01Epub Date: 2025-12-18DOI: 10.1016/j.cie.2025.111769
YongKyung Oh , Jiin Kwak , Sungil Kim
Traffic congestion remains a major challenge in developed countries, disrupting mobility and affecting economic and social activities. Among its various types, non-recurrent congestion — caused by unexpected events such as accidents, maintenance, or debris — remains difficult to predict due to its irregular spatio-temporal dynamics. While existing models effectively forecast recurrent traffic, they are less applicable to non-recurrent events characterized by abrupt and anomalous patterns. This study presents a pattern-based framework that integrates the weighted K-nearest neighbor (WK-NN) algorithm with dynamic time warping (DTW) for similarity-based prediction of non-recurrent congestion impact. The framework estimates speed drop ratios (SDRs) and propagates the predicted effects to neighboring road segments, enabling a network-level assessment of disruption. By identifying historical patterns most similar to the current incident, the proposed approach enhances interpretability and traceability for operational use. We evaluate the method using 2780 real-world traffic incident records combining data from the Korean National Police Agency and NAVER Corporation. Experimental results demonstrate that the proposed framework achieves consistent and competitive performance compared with benchmark machine learning and deep learning models. These findings suggest the framework’s potential for supporting practical decision-making in traffic control centers through timely and interpretable congestion impact forecasts.
{"title":"Predicting non-recurrent congestion impact: A pattern-based approach for speed drop ratio prediction using weighted K-nearest neighbors","authors":"YongKyung Oh , Jiin Kwak , Sungil Kim","doi":"10.1016/j.cie.2025.111769","DOIUrl":"10.1016/j.cie.2025.111769","url":null,"abstract":"<div><div>Traffic congestion remains a major challenge in developed countries, disrupting mobility and affecting economic and social activities. Among its various types, non-recurrent congestion — caused by unexpected events such as accidents, maintenance, or debris — remains difficult to predict due to its irregular spatio-temporal dynamics. While existing models effectively forecast recurrent traffic, they are less applicable to non-recurrent events characterized by abrupt and anomalous patterns. This study presents a pattern-based framework that integrates the weighted K-nearest neighbor (WK-NN) algorithm with dynamic time warping (DTW) for similarity-based prediction of non-recurrent congestion impact. The framework estimates speed drop ratios (SDRs) and propagates the predicted effects to neighboring road segments, enabling a network-level assessment of disruption. By identifying historical patterns most similar to the current incident, the proposed approach enhances interpretability and traceability for operational use. We evaluate the method using 2780 real-world traffic incident records combining data from the Korean National Police Agency and NAVER Corporation. Experimental results demonstrate that the proposed framework achieves consistent and competitive performance compared with benchmark machine learning and deep learning models. These findings suggest the framework’s potential for supporting practical decision-making in traffic control centers through timely and interpretable congestion impact forecasts.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111769"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841693","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 : 2026-03-01Epub Date: 2025-12-20DOI: 10.1016/j.cie.2025.111773
Huiting Li , Jiapeng Zhang , Xiaodi Chen , Haoxin Guo , Jianhua Liu , Cunbo Zhuang
The advancement of Industry 5.0 has driven a growing body of research that examines the impact of human factors on production processes. However, studies that simultaneously consider multiple types of human factors remain scarce. In this study, a comprehensive set of human factors, including workers’ skill proficiency, fatigue levels, interpersonal dynamics, and work experience, is incorporated into the assembly scheduling framework. Based on these considerations, the multi-objective scheduling problem in complex product assembly shops with parallel teams is investigated, with optimization objectives including makespan, transportation time, total waiting time, and team workload imbalance. To address this problem, an improved non-dominated sorting genetic algorithm is proposed. The algorithm features enhancement strategies, such as a destruction-reconstruction approach for optimizing the initial population and an improved evolutionary process. The proposed algorithm is evaluated against alternative algorithms using four case studies derived from actual production scenarios. The results demonstrate that the proposed method achieves superior solution quality and efficiency.
{"title":"Multi-objective scheduling for complex assembly shops considering multiple human factors","authors":"Huiting Li , Jiapeng Zhang , Xiaodi Chen , Haoxin Guo , Jianhua Liu , Cunbo Zhuang","doi":"10.1016/j.cie.2025.111773","DOIUrl":"10.1016/j.cie.2025.111773","url":null,"abstract":"<div><div>The advancement of Industry 5.0 has driven a growing body of research that examines the impact of human factors on production processes. However, studies that simultaneously consider multiple types of human factors remain scarce. In this study, a comprehensive set of human factors, including workers’ skill proficiency, fatigue levels, interpersonal dynamics, and work experience, is incorporated into the assembly scheduling framework. Based on these considerations, the multi-objective scheduling problem in complex product assembly shops with parallel teams is investigated, with optimization objectives including makespan, transportation time, total waiting time, and team workload imbalance. To address this problem, an improved non-dominated sorting genetic algorithm is proposed. The algorithm features enhancement strategies, such as a destruction-reconstruction approach for optimizing the initial population and an improved evolutionary process. The proposed algorithm is evaluated against alternative algorithms using four case studies derived from actual production scenarios. The results demonstrate that the proposed method achieves superior solution quality and efficiency.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111773"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841694","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 : 2026-03-01Epub Date: 2025-12-17DOI: 10.1016/j.cie.2025.111760
Kuan Xu , Zhiguang Cao , Chenlong Zheng , Lindong Liu
In this study, we propose a reinforcement learning-based adaptive variable neighborhood search (RL-AVNS) method designed for effectively solving the Vehicle Routing Problem with Multiple Time Windows (VRPMTW). Unlike traditional adaptive approaches that rely solely on historical operator performance, our method integrates a reinforcement learning framework to dynamically select neighborhood operators based on real-time solution states and learned experience. We introduce a fitness metric that quantifies customers’ temporal flexibility to improve the shaking phase, and employ a transformer-based neural policy network to intelligently guide operator selection during the local search. Extensive computational experiments are conducted on realistic scenarios derived from the replenishment of unmanned vending machines, characterized by multiple clustered replenishment windows. Results demonstrate that RL-AVNS significantly outperforms traditional variable neighborhood search (VNS), adaptive VNS (AVNS), and state-of-the-art learning-based heuristics, achieving substantial improvements in solution quality and computational efficiency across various instance scales and time window complexities. Particularly notable is the algorithm’s capability to generalize effectively to problem instances not encountered during training, underscoring its practical utility for complex logistics scenarios.
{"title":"Learning to search for vehicle routing with multiple time windows","authors":"Kuan Xu , Zhiguang Cao , Chenlong Zheng , Lindong Liu","doi":"10.1016/j.cie.2025.111760","DOIUrl":"10.1016/j.cie.2025.111760","url":null,"abstract":"<div><div>In this study, we propose a reinforcement learning-based adaptive variable neighborhood search (RL-AVNS) method designed for effectively solving the Vehicle Routing Problem with Multiple Time Windows (VRPMTW). Unlike traditional adaptive approaches that rely solely on historical operator performance, our method integrates a reinforcement learning framework to dynamically select neighborhood operators based on real-time solution states and learned experience. We introduce a fitness metric that quantifies customers’ temporal flexibility to improve the shaking phase, and employ a transformer-based neural policy network to intelligently guide operator selection during the local search. Extensive computational experiments are conducted on realistic scenarios derived from the replenishment of unmanned vending machines, characterized by multiple clustered replenishment windows. Results demonstrate that RL-AVNS significantly outperforms traditional variable neighborhood search (VNS), adaptive VNS (AVNS), and state-of-the-art learning-based heuristics, achieving substantial improvements in solution quality and computational efficiency across various instance scales and time window complexities. Particularly notable is the algorithm’s capability to generalize effectively to problem instances not encountered during training, underscoring its practical utility for complex logistics scenarios.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"213 ","pages":"Article 111760"},"PeriodicalIF":6.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841690","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