Pub Date : 2026-04-01Epub Date: 2026-01-23DOI: 10.1016/j.cie.2026.111836
Yuheng Dang , Hengte Du , Xu Wang , Xing Pan
Multi-agent systems (MAS), as a representative complex system, have become crucial for analyzing cluster and heterogeneous behaviors in various domains such as biology, social science, military weapon and manufacturing. The MAS exhibits adaptability to environmental changes and can dynamically reconfigure its structure to enhance resilience while reducing vulnerability. However, existing research primarily focuses on proposing reconfiguration strategies to enhance resilience but lacks in-depth exploration of reconfigurable design and capability constraints. The study proposes a reconfigurable operation-loop network (RON) model for resilience analysis and reconfigurable design of MAS based on the operation loop. Subsequently, the performance measurement and resilience metric are presented for RON considering mission load. Furthermore, the mathematical model and optimization framework of reconfiguration are established with the consideration of reconfigurable attributes and the resilience objective. Finally, the feasibility, effectiveness, and superiority of the proposed models and metrics are illustrated through extensive experiments on case based on an emergency response system. Numerical results demonstrate that the performance metric considering mission load contributes to a more accurate assessment of RON resilience than conventional network metrics. This work could yield valuable insights for the reconfigurable and resilient design of MAS, while providing guidance and serving as a reference for future research efforts.
{"title":"Reconfigurable operation-loop network modeling and resilience optimization considering mission load","authors":"Yuheng Dang , Hengte Du , Xu Wang , Xing Pan","doi":"10.1016/j.cie.2026.111836","DOIUrl":"10.1016/j.cie.2026.111836","url":null,"abstract":"<div><div>Multi-agent systems (MAS), as a representative complex system, have become crucial for analyzing cluster and heterogeneous behaviors in various domains such as biology, social science, military weapon and manufacturing. The MAS exhibits adaptability to environmental changes and can dynamically reconfigure its structure to enhance resilience while reducing vulnerability. However, existing research primarily focuses on proposing reconfiguration strategies to enhance resilience but lacks in-depth exploration of reconfigurable design and capability constraints. The study proposes a reconfigurable operation-loop network (RON) model for resilience analysis and reconfigurable design of MAS based on the operation loop. Subsequently, the performance measurement and resilience metric are presented for RON considering mission load. Furthermore, the mathematical model and optimization framework of reconfiguration are established with the consideration of reconfigurable attributes and the resilience objective. Finally, the feasibility, effectiveness, and superiority of the proposed models and metrics are illustrated through extensive experiments on case based on an emergency response system. Numerical results demonstrate that the performance metric considering mission load contributes to a more accurate assessment of RON resilience than conventional network metrics. This work could yield valuable insights for the reconfigurable and resilient design of MAS, while providing guidance and serving as a reference for future research efforts.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111836"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081362","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 : 2026-04-01Epub Date: 2026-02-05DOI: 10.1016/j.cie.2026.111879
Qingwei Nie , Junsai Geng , Dunbing Tang , Changchun Liu
Computer Numerical Control (CNC) machine tools are the core equipment of modern manufacturing, which directly links their efficient operation to production efficiency and product quality. However, traditional methods mainly rely on experiential rules and historical data, which have limitations in handling multimodal data and dynamic knowledge updates. As a result, traditional methods struggle to address the dual challenges of frequent faults and the complexity of fault diagnosis. To address the above issues, this paper proposes a CNC machine tool fault diagnosis knowledge model driven by Knowledge Graph (KG) and Large Language Model (LLM), aiming to enhance the efficiency and accuracy of CNC machine tool fault diagnosis. First, multimodal data fusion techniques are used to integrate equipment operating parameters, fault cases, and maintenance experiences, among other multimodal data. Then, a fault knowledge graph is constructed by combining the design of industrial fault knowledge ontologies and dynamic knowledge update algorithms. In addition, the system integrates fine-tuned GPT-4o via a Retrieval-Augmented Generation (RAG) mechanism that incorporates the KG, retrieving relevant entities and paths to guide fault mode matching and causal reasoning algorithms while reducing hallucinations. This enables semantic understanding and causal inference of fault texts, facilitating intelligent fault diagnosis and root cause identification. Based on this, the system supports online diagnostic conversations on both web and Augmented Reality (AR) platforms, providing visualized graphs and maintenance decision-making recommendations. Finally, comparative and ablation experiments conducted in a real machine tool workshop demonstrated that the proposed method significantly outperforms traditional methods in terms of fault mode matching accuracy and diagnostic efficiency. This validates its feasibility and superiority, and provides an efficient and intelligent solution for industrial equipment reliability management.
{"title":"Industrial knowledge-enhanced fault diagnosis method: Integrating LLM and knowledge graph for fault reasoning and maintenance recommendation in CNC machine tools","authors":"Qingwei Nie , Junsai Geng , Dunbing Tang , Changchun Liu","doi":"10.1016/j.cie.2026.111879","DOIUrl":"10.1016/j.cie.2026.111879","url":null,"abstract":"<div><div>Computer Numerical Control (CNC) machine tools are the core equipment of modern manufacturing, which directly links their efficient operation to production efficiency and product quality. However, traditional methods mainly rely on experiential rules and historical data, which have limitations in handling multimodal data and dynamic knowledge updates. As a result, traditional methods struggle to address the dual challenges of frequent faults and the complexity of fault diagnosis. To address the above issues, this paper proposes a CNC machine tool fault diagnosis knowledge model driven by Knowledge Graph (KG) and Large Language Model (LLM), aiming to enhance the efficiency and accuracy of CNC machine tool fault diagnosis. First, multimodal data fusion techniques are used to integrate equipment operating parameters, fault cases, and maintenance experiences, among other multimodal data. Then, a fault knowledge graph is constructed by combining the design of industrial fault knowledge ontologies and dynamic knowledge update algorithms. In addition, the system integrates fine-tuned GPT-4o via a Retrieval-Augmented Generation (RAG) mechanism that incorporates the KG, retrieving relevant entities and paths to guide fault mode matching and causal reasoning algorithms while reducing hallucinations. This enables semantic understanding and causal inference of fault texts, facilitating intelligent fault diagnosis and root cause identification. Based on this, the system supports online diagnostic conversations on both web and Augmented Reality (AR) platforms, providing visualized graphs and maintenance decision-making recommendations. Finally, comparative and ablation experiments conducted in a real machine tool workshop demonstrated that the proposed method significantly outperforms traditional methods in terms of fault mode matching accuracy and diagnostic efficiency. This validates its feasibility and superiority, and provides an efficient and intelligent solution for industrial equipment reliability management.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111879"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190480","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 : 2026-04-01Epub Date: 2026-01-30DOI: 10.1016/j.cie.2026.111863
Zhi Liu , Wuyan Ding , Juan Tang , Xiao-Xue Zheng , Ching-Ter Chang
Remanufacturing supply chains (RSCs) play a critical role in promoting environmental sustainability. However, their operational and economic performance is often shaped by government-imposed reward-penalty mechanisms (RPMs). Building on the practical background of the electrical and electronic equipment industry, this study examines a multi-agent RSC comprising a manufacturer, a retailer, and a remanufacturer. By integrating non-cooperative and cooperative game theory, we develop five coalition models to examine how coalition structures influence equilibrium pricing and remanufacturing decisions under RPMs. The results show that although the grand coalition achieves the highest system profit and total sales, it may become unstable under certain parameter conditions. Then, we propose a novel coordination mechanism that incorporates an innovative coalition weight adjustment method derived from the characteristic function and integrates it with the classical value solution. Numerical experiments confirm that the coordination mechanism satisfies individual rationality and stabilizes cooperation. From a theoretical perspective, integrating RPM-driven coalition value variations into a cooperative game framework can provide a new analytical approach for examining coordination stability in multi-agent RSCs. Further results indicate that a stricter RPM leads to higher overall social welfare, and the impact of reward-penalty coefficients is more significant than that of target remanufacturing rates. Cooperation between remanufacturers and other supply chain members elevates the total sales quantity of products.
{"title":"Multi-agent remanufacturing supply chain coordination under reward-penalty mechanism: A novel cooperative game approach","authors":"Zhi Liu , Wuyan Ding , Juan Tang , Xiao-Xue Zheng , Ching-Ter Chang","doi":"10.1016/j.cie.2026.111863","DOIUrl":"10.1016/j.cie.2026.111863","url":null,"abstract":"<div><div>Remanufacturing supply chains (RSCs) play a critical role in promoting environmental sustainability. However, their operational and economic performance is often shaped by government-imposed reward-penalty mechanisms (RPMs). Building on the practical background of the electrical and electronic equipment industry, this study examines a multi-agent RSC comprising a manufacturer, a retailer, and a remanufacturer. By integrating non-cooperative and cooperative game theory, we develop five coalition models to examine how coalition structures influence equilibrium pricing and remanufacturing decisions under RPMs. The results show that although the grand coalition achieves the highest system profit and total sales, it may become unstable under certain parameter conditions. Then, we propose a novel coordination mechanism that incorporates an innovative coalition weight adjustment method derived from the characteristic function and integrates it with the classical <span><math><mi>τ</mi></math></span> value solution. Numerical experiments confirm that the coordination mechanism satisfies individual rationality and stabilizes cooperation. From a theoretical perspective, integrating RPM-driven coalition value variations into a cooperative game framework can provide a new analytical approach for examining coordination stability in multi-agent RSCs. Further results indicate that a stricter RPM leads to higher overall social welfare, and the impact of reward-penalty coefficients is more significant than that of target remanufacturing rates. Cooperation between remanufacturers and other supply chain members elevates the total sales quantity of products.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111863"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190422","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 : 2026-04-01Epub Date: 2026-01-22DOI: 10.1016/j.cie.2026.111842
Pei-Yang Wu , Ren-Yong Guo , Ying-En Ge
This study investigates the carriage parking problem in the dynamic autonomous non-stop rail transit (DANRT) system, with a particular focus on the movement behaviors of passengers. A cell transmission model (CTM) is formulated to depict the movement behaviors of passengers in the DANRT system and the interactions between passengers. The parameters in the CTM are calibrated by using a set of video recordings and reproducing the arching phenomenon of passengers. To optimize carriage parking schemes, a swarm intelligence-based heuristic algorithm is proposed, where the CTM is embedded into the evaluation process to dynamically assess passenger moving efficiency during each iteration. We conduct a set of numerical experiments to evaluate the effect of algorithm parameters on algorithm performance and the influence of passenger behaviors on passenger waiting times. The results demonstrate that the algorithm can further reduce the theoretical minimum total passenger waiting time obtained without considering passenger movement behaviors and interactions by about 6%. Additionally, overall system efficiency reaches its maximum when the frequency of carriage re-selection behavior of passengers remains at a moderate level. Moreover, it is essential to deliberately designate carriages for passengers to improve the travel efficiency of passengers in the DANRT system.
{"title":"Optimization of carriage parking based on simulation of passenger dynamics in the dynamic autonomous non-stop rail transit system","authors":"Pei-Yang Wu , Ren-Yong Guo , Ying-En Ge","doi":"10.1016/j.cie.2026.111842","DOIUrl":"10.1016/j.cie.2026.111842","url":null,"abstract":"<div><div>This study investigates the carriage parking problem in the dynamic autonomous non-stop rail transit (DANRT) system, with a particular focus on the movement behaviors of passengers. A cell transmission model (CTM) is formulated to depict the movement behaviors of passengers in the DANRT system and the interactions between passengers. The parameters in the CTM are calibrated by using a set of video recordings and reproducing the arching phenomenon of passengers. To optimize carriage parking schemes, a swarm intelligence-based heuristic algorithm is proposed, where the CTM is embedded into the evaluation process to dynamically assess passenger moving efficiency during each iteration. We conduct a set of numerical experiments to evaluate the effect of algorithm parameters on algorithm performance and the influence of passenger behaviors on passenger waiting times. The results demonstrate that the algorithm can further reduce the theoretical minimum total passenger waiting time obtained without considering passenger movement behaviors and interactions by about 6%. Additionally, overall system efficiency reaches its maximum when the frequency of carriage re-selection behavior of passengers remains at a moderate level. Moreover, it is essential to deliberately designate carriages for passengers to improve the travel efficiency of passengers in the DANRT system.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111842"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045177","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 : 2026-04-01Epub Date: 2026-01-10DOI: 10.1016/j.cie.2026.111808
Avi Herbon , Simone Zanoni
Relatively few studies in the field of inventory management of perishables focus on preservation efforts, and even fewer have considered the opposite challenge: accelerating product aging. This issue is particularly relevant for goods like wine and cheese, where perceived quality initially increases over time. In this study, we develop an analytical model to evaluate the economic trade-off between investing in technologies that accelerate aging—thus shifting demand to earlier periods—and the associated implementation costs.
The model incorporates a heterogeneous market, where consumers differ in their sensitivity to price and perceived quality. We derive conditions ensuring the uniqueness of the optimal “effort window”—the time reduction required to reach peak perceived quality. Using a numerical illustration, we explore how consumer heterogeneity, cycle length, and initial product quality influence both profitability and optimal strategy.
Our findings show that accelerating aging is more beneficial when consumers are relatively homogeneous, while in highly heterogeneous markets, such investment may prove uneconomical. Additionally, cycle length plays a critical role in determining profitability, emphasizing the need to integrate inventory policy with technological investment. These results provide actionable insights for practitioners and managers in industries where product maturity affects demand, including wine, luxury goods, and electronics, where model cycles and innovation timing influence demand.
{"title":"Speeding up wine aging vs. implementation costs","authors":"Avi Herbon , Simone Zanoni","doi":"10.1016/j.cie.2026.111808","DOIUrl":"10.1016/j.cie.2026.111808","url":null,"abstract":"<div><div>Relatively few studies in the field of inventory management of perishables focus on preservation efforts, and even fewer have considered the opposite challenge: accelerating product aging. This issue is particularly relevant for goods like wine and cheese, where perceived quality initially increases over time. In this study, we develop an analytical model to evaluate the economic trade-off between investing in technologies that accelerate aging—thus shifting demand to earlier periods—and the associated implementation costs.</div><div>The model incorporates a heterogeneous market, where consumers differ in their sensitivity to price and perceived quality. We derive conditions ensuring the uniqueness of the optimal “<em>effort window</em>”—the time reduction required to reach peak perceived quality. Using a numerical illustration, we explore how consumer heterogeneity, cycle length, and initial product quality influence both profitability and optimal strategy.</div><div>Our findings show that accelerating aging is more beneficial when consumers are relatively homogeneous, while in highly heterogeneous markets, such investment may prove uneconomical. Additionally, cycle length plays a critical role in determining profitability, emphasizing the need to integrate inventory policy with technological investment. These results provide actionable insights for practitioners and managers in industries where product maturity affects demand, including wine, luxury goods, and electronics, where model cycles and innovation timing influence demand.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111808"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045175","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 : 2026-04-01Epub Date: 2026-01-29DOI: 10.1016/j.cie.2026.111832
Muhammad Waqas , Fatemeh Sogandi , Ali Yeganeh , Songhua Xu
Monitoring time between events has become increasingly important in statistical process control, especially in applications where event timing provides more informative insights than direct measurement of quality characteristics. Traditional approaches for monitoring univariate and multivariate time-between-events data often depend on parametric assumptions and conventional statistical control charts, which can be inadequate when the underlying distributions are unknown, complex, or subject to change. In this study, we address these limitations by developing a monitoring framework based on the Gumbel’s bivariate exponential distribution, tailored for real-world applications involving two dependent variables. Recognizing the challenges posed by parameter estimation and distributional assumptions, we extend our model to include both parametric and nonparametric structures. Moreover, conventional statistical control charts are found to exhibit reduced performance in nonparametric settings, particularly in detecting complex and unknown process changes. To address this limitation, a machine learning–based control chart is proposed, which incorporates an artificial neural network enhanced by an attention mechanism. In this framework, statistical features derived from the Gumbel’s bivariate exponential process are imported as memory-based input features. The attention mechanism is employed to guide the model in focusing on the most relevant temporal dependencies, thereby enhancing its sensitivity to subtle shifts. This hybrid approach is designed to improve the early detection of out-of-control conditions and reduce the risk of nonconforming products or harmful events in dynamic environments. Through extensive Monte Carlo simulations, encompassing shifts in scale, dependency parameters, and various nonparametric distributional changes (with the underlying process type known), the detection capability of the proposed method has been evaluated. The results show that the proposed method generally provides faster time-to-signal for OOC conditions across most scenarios, although its superiority is not universal. As a practical application, an employee communication network within a company, analogous to social network monitoring, is considered, representing a novel context for time between events-based surveillance. It is demonstrated that the proposed method can effectively detect unnatural or anomalous communication patterns between employees, highlighting its potential for identifying irregularities in networked environments.
{"title":"An attention-based machine learning control chart for monitoring Gumbel’s bivariate time between events: application to early anomaly detection in employee communication networks","authors":"Muhammad Waqas , Fatemeh Sogandi , Ali Yeganeh , Songhua Xu","doi":"10.1016/j.cie.2026.111832","DOIUrl":"10.1016/j.cie.2026.111832","url":null,"abstract":"<div><div>Monitoring time between events has become increasingly important in statistical process control, especially in applications where event timing provides more informative insights than direct measurement of quality characteristics. Traditional approaches for monitoring univariate and multivariate time-between-events data often depend on parametric assumptions and conventional statistical control charts, which can be inadequate when the underlying distributions are unknown, complex, or subject to change. In this study, we address these limitations by developing a monitoring framework based on the Gumbel’s bivariate exponential distribution, tailored for real-world applications involving two dependent variables. Recognizing the challenges posed by parameter estimation and distributional assumptions, we extend our model to include both parametric and nonparametric structures. Moreover, conventional statistical control charts are found to exhibit reduced performance in nonparametric settings, particularly in detecting complex and unknown process changes. To address this limitation, a machine learning–based control chart is proposed, which incorporates an artificial neural network enhanced by an attention mechanism. In this framework, statistical features derived from the Gumbel’s bivariate exponential process are imported as memory-based input features. The attention mechanism is employed to guide the model in focusing on the most relevant temporal dependencies, thereby enhancing its sensitivity to subtle shifts. This hybrid approach is designed to improve the early detection of out-of-control conditions and reduce the risk of nonconforming products or harmful events in dynamic environments. Through extensive Monte Carlo simulations, encompassing shifts in scale, dependency parameters, and various nonparametric distributional changes (with the underlying process type known), the detection capability of the proposed method has been evaluated. The results show that the proposed method generally provides faster time-to-signal for OOC conditions across most scenarios, although its superiority is not universal. As a practical application, an employee communication network within a company, analogous to social network monitoring, is considered, representing a novel context for time between events-based surveillance. It is demonstrated that the proposed method can effectively detect unnatural or anomalous communication patterns between employees, highlighting its potential for identifying irregularities in networked environments.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111832"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190413","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}
The flexible job shop scheduling problem is a foundational NP-hard challenge in smart manufacturing. While deep reinforcement learning (DRL) combined with graph neural networks has shown promise in learning adaptive policies, existing methods often rely on greedy selection or simple sampling at inference, which constrains their long-horizon planning and generalization capabilities. In this paper, we propose a budget-scalable, inference-time Hybrid Monte Carlo Tree Search (Hybrid-MCTS) framework that effectively enhances the performance of pre-trained DRL policies without requiring any retraining. The framework employs a policy-guided two-stage refinement process with iteration-wise budget scheduling: Stage I utilizes entropy-adaptive greedy screening to quickly establish per-action incumbents and prune the search space; Stage II performs a diversity-aware iterative simulation using a modified upper confidence bound to explore high-potential trajectories while preserving outcome variability. To optimize efficiency, the procedure incorporates an early-stopping mechanism based on consecutive non-improving iterative rounds. This design enables multi-step lookahead and provides a tunable quality–time trade-off, where an increased simulation budget consistently yields non-inferior solutions. Extensive experiments on public flexible job shop scheduling problem benchmarks and two representative DRL schedulers demonstrate consistent improvements over greedy and policy-sampling inference. Notably, the framework enables a standard scheduler to outperform an exact algorithm by 4.04% on large-scale instances while maintaining a comparable computational budget. Its inference-only design enables straightforward integration in high-precision scheduling scenarios.
{"title":"Budget-scalable inference-time hybrid MCTS for enhancing DRL-based flexible job shop scheduling","authors":"Yuzhi Zhang, Shidu Dong, Ting Wen, Zhenfang Yuan, Jianfeng Xiao, Zhuo Diao","doi":"10.1016/j.cie.2026.111897","DOIUrl":"10.1016/j.cie.2026.111897","url":null,"abstract":"<div><div>The flexible job shop scheduling problem is a foundational NP-hard challenge in smart manufacturing. While deep reinforcement learning (DRL) combined with graph neural networks has shown promise in learning adaptive policies, existing methods often rely on greedy selection or simple sampling at inference, which constrains their long-horizon planning and generalization capabilities. In this paper, we propose a budget-scalable, inference-time Hybrid Monte Carlo Tree Search (Hybrid-MCTS) framework that effectively enhances the performance of pre-trained DRL policies without requiring any retraining. The framework employs a policy-guided two-stage refinement process with iteration-wise budget scheduling: Stage I utilizes entropy-adaptive greedy screening to quickly establish per-action incumbents and prune the search space; Stage II performs a diversity-aware iterative simulation using a modified upper confidence bound to explore high-potential trajectories while preserving outcome variability. To optimize efficiency, the procedure incorporates an early-stopping mechanism based on consecutive non-improving iterative rounds. This design enables multi-step lookahead and provides a tunable quality–time trade-off, where an increased simulation budget consistently yields non-inferior solutions. Extensive experiments on public flexible job shop scheduling problem benchmarks and two representative DRL schedulers demonstrate consistent improvements over greedy and policy-sampling inference. Notably, the framework enables a standard scheduler to outperform an exact algorithm by 4.04% on large-scale instances while maintaining a comparable computational budget. Its inference-only design enables straightforward integration in high-precision scheduling scenarios.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111897"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190363","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 : 2026-04-01Epub Date: 2026-01-30DOI: 10.1016/j.cie.2026.111862
Qingli Zhao , Zhi-Ping Fan
Blockchain technology, as a cutting-edge way to provide customers with credible product traceability information, has been increasingly adopted in e-commerce supply chains. Nonetheless, it is crucial to consider whether adopting this technology is beneficial for the members of an e-commerce supply chain. This study investigates a hybrid sales scenario where a manufacturer distributes products through both reselling and agency channels on an e-commerce platform. We analyze the specific conditions under which the manufacturer adopts blockchain technology in the agency and/or reselling channels, as well as the impacts of such adoption on the e-commerce supply chain. The results indicate that the adoption of blockchain technology does not necessarily lead to an increase in product prices or an expansion of consumer demand. Its impacts on product prices and consumer demand depend on the blockchain technology adoption cost and the percentage fee in the agency channel. Additionally, when the blockchain adoption cost is low (high), it is beneficial for the manufacturer to adopt (not adopt) blockchain technology in both the reselling and agency channels. When the cost is moderate, it is beneficial for the manufacturer to only adopt the technology in the reselling channel. In particular, adopting blockchain technology can enable the supply chain members to achieve the Pareto improvement if the blockchain adoption cost is low.
{"title":"Blockchain technology adoption decisions of a manufacturer under a hybrid sales scenario","authors":"Qingli Zhao , Zhi-Ping Fan","doi":"10.1016/j.cie.2026.111862","DOIUrl":"10.1016/j.cie.2026.111862","url":null,"abstract":"<div><div>Blockchain technology, as a cutting-edge way to provide customers with credible product traceability information, has been increasingly adopted in e-commerce supply chains. Nonetheless, it is crucial to consider whether adopting this technology is beneficial for the members of an e-commerce supply chain. This study investigates a hybrid sales scenario where a manufacturer distributes products through both reselling and agency channels on an e-commerce platform. We analyze the specific conditions under which the manufacturer adopts blockchain technology in the agency and/or reselling channels, as well as the impacts of such adoption on the e-commerce supply chain. The results indicate that the adoption of blockchain technology does not necessarily lead to an increase in product prices or an expansion of consumer demand. Its impacts on product prices and consumer demand depend on the blockchain technology adoption cost and the percentage fee in the agency channel. Additionally, when the blockchain adoption cost is low (high), it is beneficial for the manufacturer to adopt (not adopt) blockchain technology in both the reselling and agency channels. When the cost is moderate, it is beneficial for the manufacturer to only adopt the technology in the reselling channel. In particular, adopting blockchain technology can enable the supply chain members to achieve the Pareto improvement if the blockchain adoption cost is low.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111862"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190420","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 : 2026-04-01Epub Date: 2026-01-23DOI: 10.1016/j.cie.2026.111835
Marta Rinaldi , Eric H. Grosse
The COVID-19 pandemic has affected the global economy, increasing the focus on resilience and survivability in supply chains (SCs). This study addresses these topics and advances understanding of how strategic decisions can enhance supply chain viability under adverse conditions. It integrates empirical case studies and simulation modeling to investigate two SC adaptation strategies and their impact on SC performance. A dynamic supplier selection process based on scalability and substitution was implemented, and varied thresholds and time horizons were tested and evaluated. This study makes significant contributions to the literature by integrating the constructs of resilience and efficiency, offering valuable insights for academics and practitioners. The results demonstrate how a combination of scalability and substitution can enhance SC viability while maintaining SC performance. Integrating simulation results with real-world case study data enriches the literature by bridging the gap between theory and practice and providing actionable managerial insights to enhance both strategic planning and operational viability in SCs.
{"title":"Enhancing supply chain viability through dynamic adaptation: a simulation-based approach to scalability and substitution strategies","authors":"Marta Rinaldi , Eric H. Grosse","doi":"10.1016/j.cie.2026.111835","DOIUrl":"10.1016/j.cie.2026.111835","url":null,"abstract":"<div><div>The COVID-19 pandemic has affected the global economy, increasing the focus on resilience and survivability in supply chains (SCs). This study addresses these topics and advances understanding of how strategic decisions can enhance supply chain viability under adverse conditions. It integrates empirical case studies and simulation modeling to investigate two SC adaptation strategies and their impact on SC performance. A dynamic supplier selection process based on scalability and substitution was implemented, and varied thresholds and time horizons were tested and evaluated. This study makes significant contributions to the literature by integrating the constructs of resilience and efficiency, offering valuable insights for academics and practitioners. The results demonstrate how a combination of scalability and substitution can enhance SC viability while maintaining SC performance. Integrating simulation results with real-world case study data enriches the literature by bridging the gap between theory and practice and providing actionable managerial insights to enhance both strategic planning and operational viability in SCs.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111835"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190358","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 : 2026-04-01Epub Date: 2026-01-27DOI: 10.1016/j.cie.2026.111869
Anshuman Kumar, S.P. Sarmah
Semiconductor manufacturing systems are subject to significant operational uncertainties stemming from fluctuating customer demand, variable supplier lead times and machine-level disruptions. This study presents a multi-stage stochastic programming framework that integrates production planning, procurement scheduling, inventory control and emissions management within a unified decision-making model. The framework explicitly incorporates environmental regulations through periodic emission thresholds and tool cleaning constraints while accounting for sourcing risks through supplier classification and diversification. By modelling uncertainty through a scenario-based approach, the proposed method enables both anticipatory and adaptive decisions that enhance system robustness. The model is implemented using mixed-integer programming techniques and validated through computational experiments based on empirically motivated scenarios. Results demonstrate improved cost efficiency, service level adherence, and regulatory compliance compared to deterministic baselines. Sensitivity analysis highlights key trade-offs, showing that stricter emission caps can increase total costs, while supplier diversification helps mitigate disruption risks. The results underscore the value of stochastic programming in capturing the complex interdependencies in semiconductor supply chains and provide a rigorous decision-support tool for managing uncertainty in high-precision manufacturing systems.
{"title":"A multi-stage stochastic model for sustainable semiconductor manufacturing","authors":"Anshuman Kumar, S.P. Sarmah","doi":"10.1016/j.cie.2026.111869","DOIUrl":"10.1016/j.cie.2026.111869","url":null,"abstract":"<div><div>Semiconductor manufacturing systems are subject to significant operational uncertainties stemming from fluctuating customer demand, variable supplier lead times and machine-level disruptions. This study presents a multi-stage stochastic programming framework that integrates production planning, procurement scheduling, inventory control and emissions management within a unified decision-making model. The framework explicitly incorporates environmental regulations through periodic emission thresholds and tool cleaning constraints while accounting for sourcing risks through supplier classification and diversification. By modelling uncertainty through a scenario-based approach, the proposed method enables both anticipatory and adaptive decisions that enhance system robustness. The model is implemented using mixed-integer programming techniques and validated through computational experiments based on empirically motivated scenarios. Results demonstrate improved cost efficiency, service level adherence, and regulatory compliance compared to deterministic baselines. Sensitivity analysis highlights key trade-offs, showing that stricter emission caps can increase total costs, while supplier diversification helps mitigate disruption risks. The results underscore the value of stochastic programming in capturing the complex interdependencies in semiconductor supply chains and provide a rigorous decision-support tool for managing uncertainty in high-precision manufacturing systems.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"214 ","pages":"Article 111869"},"PeriodicalIF":6.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146190357","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