Detailed machine-specific data are critical for accurate sustainability assessment and for supporting design decisions to reduce environmental impacts from manufacturing. However, obtaining, analyzing, and interpreting such fine-grained measurements can be challenging and inefficient. Existing methods for the above are time-consuming, do not fully capture process variability over time, and do not relate primary manufacturing data back to design decision-making. In this work, we propose a methodology to programmatically disaggregate process-level life cycle inventory data measurements, and relate it to both operations (i.e., activities or sub-processes) within the process and the geometric features created or affected by the process. We do this by leveraging the underlying machine code used to manufacture the part, in this case G-code, and by providing a scalable definition scheme for the corresponding operations, geometric features, and the relationship between them. Results can be used to generate targeted, actionable insights into process setup and product design improvements to address environmental impacts of manufacturing processes.
{"title":"Feature-centric allocation and visualization of primary manufacturing process life cycle inventory data","authors":"Teodor Vernica , Badrinath Veluri , Devarajan Ramanujan","doi":"10.1016/j.jmsy.2025.10.006","DOIUrl":"10.1016/j.jmsy.2025.10.006","url":null,"abstract":"<div><div>Detailed machine-specific data are critical for accurate sustainability assessment and for supporting design decisions to reduce environmental impacts from manufacturing. However, obtaining, analyzing, and interpreting such fine-grained measurements can be challenging and inefficient. Existing methods for the above are time-consuming, do not fully capture process variability over time, and do not relate primary manufacturing data back to design decision-making. In this work, we propose a methodology to programmatically disaggregate process-level life cycle inventory data measurements, and relate it to both operations (i.e., activities or sub-processes) within the process and the geometric features created or affected by the process. We do this by leveraging the underlying machine code used to manufacture the part, in this case G-code, and by providing a scalable definition scheme for the corresponding operations, geometric features, and the relationship between them. Results can be used to generate targeted, actionable insights into process setup and product design improvements to address environmental impacts of manufacturing processes.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 557-576"},"PeriodicalIF":14.2,"publicationDate":"2025-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145416703","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 : 2025-10-21DOI: 10.1016/j.jmsy.2025.10.005
Jianhao Lv, Jiahui Si, Ding Gao, Jinsong Bao
Existing AR-assisted Human–Robot Collaboration (HRC) systems passively respond to real-time information, lacking the ability to model, store, and leverage historical task knowledge in HRC scenarios, thus relying on replacing pre-programmed fixed-content modules for upgrades. To address this constraint, a historical visual question answering (VQA) framework with large language models is proposed. The unstructured visual frame is converted into structured information via structured visual representation, supported by a cross-modal interaction module and multi-component loss function to lay a structured foundation for storing historical experiences and subsequent reasoning. A temporally structured Memory Graph (MG) is constructed. Combined with large language models, historical VQA solves traditional VQA’s reliance on static images and lack of temporal continuity; An AR-assisted Human–Robot Interaction pipeline is designed for bidirectional transmission and visualization, integrating perception and reasoning results with AR to enable Human–Robot bidirectional communication. Quantitative and qualitative results show the method significantly outperforms in integrating historical and real-time information with supporting HRC VQA.
{"title":"Historical visual question answering with large language model for Augmented Reality-assisted Human–Robot Collaboration","authors":"Jianhao Lv, Jiahui Si, Ding Gao, Jinsong Bao","doi":"10.1016/j.jmsy.2025.10.005","DOIUrl":"10.1016/j.jmsy.2025.10.005","url":null,"abstract":"<div><div>Existing AR-assisted Human–Robot Collaboration (HRC) systems passively respond to real-time information, lacking the ability to model, store, and leverage historical task knowledge in HRC scenarios, thus relying on replacing pre-programmed fixed-content modules for upgrades. To address this constraint, a historical visual question answering (VQA) framework with large language models is proposed. The unstructured visual frame is converted into structured information via structured visual representation, supported by a cross-modal interaction module and multi-component loss function to lay a structured foundation for storing historical experiences and subsequent reasoning. A temporally structured Memory Graph (MG) is constructed. Combined with large language models, historical VQA solves traditional VQA’s reliance on static images and lack of temporal continuity; An AR-assisted Human–Robot Interaction pipeline is designed for bidirectional transmission and visualization, integrating perception and reasoning results with AR to enable Human–Robot bidirectional communication. Quantitative and qualitative results show the method significantly outperforms in integrating historical and real-time information with supporting HRC VQA.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 546-556"},"PeriodicalIF":14.2,"publicationDate":"2025-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362159","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 : 2025-10-17DOI: 10.1016/j.jmsy.2025.10.003
Xuanhao Wen , Huajun Cao , Hongcheng Li , Weiwei Ge , Na Yang , Xiaohui Huang , Jin Zhou , Qiongzhi Zhang
The obligatory carbon neutrality targets drive manufacturers to improve energy performance for emission reduction. In this context, energy waste diagnosis in production has become increasingly critical. However, energy waste in manufacturing systems exhibits complex characteristics such as multi-source distributions, multi-form mediums and multi-variant influencers, resulting in a lack of comprehensive diagnosis methods. Inspired by effectiveness metrics for productivity waste diagnosis, this paper extends the concept of energy efficiency to energy effectiveness and proposes a novel energy waste diagnosis method. Firstly, it transcends the binary classification of energy consumption to establish a novel energy waste taxonomy. Next, a hierarchical framework of energy effectiveness metrics (indicators and dynamic benchmarks) is developed. These metrics are then quantified using data-driven approaches, such as meta-energy-blocks, to pinpoint the root-causes of waste. Finally, the method facilitates practical applications such as energy-saving potential estimation and waste visualization. An industrial case study on a die-casting unit demonstrates the method's effectiveness and practicality. The results revealed that actual energy consumption exceeded the ideal minimum by 11.5 times, indicating significant saving potential. Moreover, 37.2 % of energy was wasted due to managerial issues, with the method successfully identifying their specific root-causes for targeted improvements. The main novelty of the proposed method lies in its transferable effectiveness metric framework, which enables a comprehensive and in-depth diagnosis of diverse energy waste types, thereby bridging a critical gap in manufacturing energy management.
{"title":"From efficiency to effectiveness: A new method for diagnosing energy waste in manufacturing systems","authors":"Xuanhao Wen , Huajun Cao , Hongcheng Li , Weiwei Ge , Na Yang , Xiaohui Huang , Jin Zhou , Qiongzhi Zhang","doi":"10.1016/j.jmsy.2025.10.003","DOIUrl":"10.1016/j.jmsy.2025.10.003","url":null,"abstract":"<div><div>The obligatory carbon neutrality targets drive manufacturers to improve energy performance for emission reduction. In this context, energy waste diagnosis in production has become increasingly critical. However, energy waste in manufacturing systems exhibits complex characteristics such as multi-source distributions, multi-form mediums and multi-variant influencers, resulting in a lack of comprehensive diagnosis methods. Inspired by effectiveness metrics for productivity waste diagnosis, this paper extends the concept of energy efficiency to energy effectiveness and proposes a novel energy waste diagnosis method. Firstly, it transcends the binary classification of energy consumption to establish a novel energy waste taxonomy. Next, a hierarchical framework of energy effectiveness metrics (indicators and dynamic benchmarks) is developed. These metrics are then quantified using data-driven approaches, such as meta-energy-blocks, to pinpoint the root-causes of waste. Finally, the method facilitates practical applications such as energy-saving potential estimation and waste visualization. An industrial case study on a die-casting unit demonstrates the method's effectiveness and practicality. The results revealed that actual energy consumption exceeded the ideal minimum by 11.5 times, indicating significant saving potential. Moreover, 37.2 % of energy was wasted due to managerial issues, with the method successfully identifying their specific root-causes for targeted improvements. The main novelty of the proposed method lies in its transferable effectiveness metric framework, which enables a comprehensive and in-depth diagnosis of diverse energy waste types, thereby bridging a critical gap in manufacturing energy management.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 517-545"},"PeriodicalIF":14.2,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324716","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 : 2025-10-15DOI: 10.1016/j.jmsy.2025.09.016
Victor Vantilborgh, Tom Lefebvre, Guillaume Crevecoeur
This paper proposes a generic low-cost method for probabilistic condition monitoring of industrial equipment. A computationally efficient recursive data-driven model is constructed that correlates real-time machine measurements with a quantity or several quantities of interest (QoIs), including artificial metrics such as the Remaining Useful Lifetime (RUL). To that end, a probabilistic state–space model (PSSM) is identified, based on a fully instrumented measurement set obtained from a limited set of experiments that can only be obtained in a specialized testing environment. The dataset contains both cheaply available sensory information as well as prohibitively expensive, invasive or artificially constructed sensory signals. To identify a Maximum Likelihood Estimate of the PSSM, we rely on the Expectation–Maximization (EM) algorithm and Sequential Monte Carlo (SMC) estimation techniques. During operation, only the vital, non-intrusive and cheap sensors are used. The PSSM is then repurposed to reconstruct the costly sensory signals, realizing an effective and general purpose virtual sensor. Our methodology demonstrates the capacity to robustly estimate unmeasured physical variables in real-time and artificially constructed prognostic QoIs, such as the RUL, even when working with an incomplete measurement array. We validate the presented methodology for condition monitoring on the C-MAPSS dataset and a solenoid valve (SV) use case. The presented tool has similar predictive capabilities as compared with other state-of-the-art RUL prognostic methods and furthermore provides uncertainty quantification and contextual information with respect to equipment health.
{"title":"Probabilistic state–space modeling for robust condition monitoring of industrial equipment","authors":"Victor Vantilborgh, Tom Lefebvre, Guillaume Crevecoeur","doi":"10.1016/j.jmsy.2025.09.016","DOIUrl":"10.1016/j.jmsy.2025.09.016","url":null,"abstract":"<div><div>This paper proposes a generic low-cost method for probabilistic condition monitoring of industrial equipment. A computationally efficient recursive data-driven model is constructed that correlates real-time machine measurements with a quantity or several quantities of interest (QoIs), including artificial metrics such as the Remaining Useful Lifetime (RUL). To that end, a probabilistic state–space model (PSSM) is identified, based on a fully instrumented measurement set obtained from a limited set of experiments that can only be obtained in a specialized testing environment. The dataset contains both cheaply available sensory information as well as prohibitively expensive, invasive or artificially constructed sensory signals. To identify a Maximum Likelihood Estimate of the PSSM, we rely on the Expectation–Maximization (EM) algorithm and Sequential Monte Carlo (SMC) estimation techniques. During operation, only the vital, non-intrusive and cheap sensors are used. The PSSM is then repurposed to reconstruct the costly sensory signals, realizing an effective and general purpose virtual sensor. Our methodology demonstrates the capacity to robustly estimate unmeasured physical variables in real-time and artificially constructed prognostic QoIs, such as the RUL, even when working with an incomplete measurement array. We validate the presented methodology for condition monitoring on the C-MAPSS dataset and a solenoid valve (SV) use case. The presented tool has similar predictive capabilities as compared with other state-of-the-art RUL prognostic methods and furthermore provides uncertainty quantification and contextual information with respect to equipment health.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 505-516"},"PeriodicalIF":14.2,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324717","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 : 2025-10-11DOI: 10.1016/j.jmsy.2025.10.002
Zheng Zhou , Yasong Li , Ruqiang Yan
Modeling degradation processes is essential for estimating industrial system performance and guiding maintenance decisions. Machine learning methods, due to the adaptability for diverse data types, show promise to model temporal evolution of degradation in an embedding space known as latent dynamics, especially for neural ordinary differential equations (NODE) with continuous-time property. However, inferring system states from observation data is an inverse problem, and NODEs often inherit ill-posedness further from their complex optimization landscape. We propose an invertible transfer function informed NODE to ensure stable latent dynamics, making the model robust to perturbations in observation data. First, a NODE describes the hidden degradation process, while an invertible Fourier neural operator maps between latent dynamics and observations. Error analysis reveals that stability is governed by data fidelity and the Lipschitz constant of the inverse mapping, forming the basis for our regularization technique. Additionally, we demonstrate that without ground truth degradation data, latent dynamics lack uniqueness, leading to infinite equivalent solutions. Tests on turbofan engine and battery datasets confirm improved robustness and performance in fault diagnosis and prognosis.
{"title":"Invertible transfer function informed neural ODE to learn stable latent dynamics for degradation process modeling and remaining useful life prediction","authors":"Zheng Zhou , Yasong Li , Ruqiang Yan","doi":"10.1016/j.jmsy.2025.10.002","DOIUrl":"10.1016/j.jmsy.2025.10.002","url":null,"abstract":"<div><div>Modeling degradation processes is essential for estimating industrial system performance and guiding maintenance decisions. Machine learning methods, due to the adaptability for diverse data types, show promise to model temporal evolution of degradation in an embedding space known as latent dynamics, especially for neural ordinary differential equations (NODE) with continuous-time property. However, inferring system states from observation data is an inverse problem, and NODEs often inherit ill-posedness further from their complex optimization landscape. We propose an invertible transfer function informed NODE to ensure stable latent dynamics, making the model robust to perturbations in observation data. First, a NODE describes the hidden degradation process, while an invertible Fourier neural operator maps between latent dynamics and observations. Error analysis reveals that stability is governed by data fidelity and the Lipschitz constant of the inverse mapping, forming the basis for our regularization technique. Additionally, we demonstrate that without ground truth degradation data, latent dynamics lack uniqueness, leading to infinite equivalent solutions. Tests on turbofan engine and battery datasets confirm improved robustness and performance in fault diagnosis and prognosis.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 494-504"},"PeriodicalIF":14.2,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145266331","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}
Modern shopfloors generate large volumes of raw process monitoring data, which hold valuable rich information for data-driven machine learning applications. However, the utility of these data is typically limited by the intensive cost and effort required for manual labeling. To this end, this paper proposes a multi-level contrastive learning framework that leverages hierarchical contextual metadata to enable effective self-supervised learning (SSL) without reliance on extensive labeled data. The framework constructs semantic supervision signals at three levels of granularity, allowing the model to learn discriminative features that align with real manufacturing conditions. The proposed pipeline is validated on a resistance spot welding (RSW) dataset collected from a General Motors production line, with evaluations on two downstream tasks: expulsion detection and nugget size prediction. Experimental results show that linear probing on partially labeled data using the pretrained model achieves performance comparable to a fully supervised transformer-based model trained on the entire labeled dataset. This generalization capability enables the SSL framework to exploit value from unlabeled plant data, providing a scalable and efficient approach for deploying machine learning in industrial applications.
{"title":"Unleashing the power of unlabeled plant data: A hierarchical contrastive learning framework for dynamic manufacturing process monitoring","authors":"Xijia Zhao , Hassan Ghassemi-Armaki , Blair Carlson , Peng (Edward) Wang","doi":"10.1016/j.jmsy.2025.10.001","DOIUrl":"10.1016/j.jmsy.2025.10.001","url":null,"abstract":"<div><div>Modern shopfloors generate large volumes of raw process monitoring data, which hold valuable rich information for data-driven machine learning applications. However, the utility of these data is typically limited by the intensive cost and effort required for manual labeling. To this end, this paper proposes a multi-level contrastive learning framework that leverages hierarchical contextual metadata to enable effective self-supervised learning (SSL) without reliance on extensive labeled data. The framework constructs semantic supervision signals at three levels of granularity, allowing the model to learn discriminative features that align with real manufacturing conditions. The proposed pipeline is validated on a resistance spot welding (RSW) dataset collected from a General Motors production line, with evaluations on two downstream tasks: expulsion detection and nugget size prediction. Experimental results show that linear probing on partially labeled data using the pretrained model achieves performance comparable to a fully supervised transformer-based model trained on the entire labeled dataset. This generalization capability enables the SSL framework to exploit value from unlabeled plant data, providing a scalable and efficient approach for deploying machine learning in industrial applications.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 483-493"},"PeriodicalIF":14.2,"publicationDate":"2025-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145266687","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}
Efficient resource recovery from end-of-life (EOL) products requires well-structured disassembly processes. A disassembly line with multiple workstations provides a systematic framework, assigning specific tasks to each station. The disassembly line balancing problem (DLBP) aims to optimize task allocation to improve performance indicators such as cycle time. However, in practical scenarios, the states of EOL product components are often uncertain during the disassembly process. To tackle this challenge, we propose a DLBP with uncertain component states (DLB-UCS) model, which incorporates component state changes as dynamic factors during EOL product disassembly. Unlike conventional DLBP models, DLB-UCS supports real-time adaptation to uncertain changes, making it more consistent with industrial conditions. To solve this problem, we develop a causal inference-based dynamic multi-objective evolutionary algorithm (CI-DMOEA) that simultaneously minimizes the total disassembly cycle time and the number of robotic units required. In particular, a causal feature selection technique based on conditional independence testing is used for efficient initial population generation, enhancing adaptability in dynamic environments. Extensive comparative experiments on eight disassembly scenarios against three state-of-the-art dynamic multi-objective optimization algorithms show that the proposed CI-DMOEA demonstrates superior performance and responsiveness.
{"title":"Causal inference-based dynamic optimization for disassembly line balancing with uncertain component states","authors":"Yilin Fang, Zhiyao Li, Kai Huang, Junyufeng Chen, Ling Gui, Xinyi Chen","doi":"10.1016/j.jmsy.2025.09.018","DOIUrl":"10.1016/j.jmsy.2025.09.018","url":null,"abstract":"<div><div>Efficient resource recovery from end-of-life (EOL) products requires well-structured disassembly processes. A disassembly line with multiple workstations provides a systematic framework, assigning specific tasks to each station. The disassembly line balancing problem (DLBP) aims to optimize task allocation to improve performance indicators such as cycle time. However, in practical scenarios, the states of EOL product components are often uncertain during the disassembly process. To tackle this challenge, we propose a DLBP with uncertain component states (DLB-UCS) model, which incorporates component state changes as dynamic factors during EOL product disassembly. Unlike conventional DLBP models, DLB-UCS supports real-time adaptation to uncertain changes, making it more consistent with industrial conditions. To solve this problem, we develop a causal inference-based dynamic multi-objective evolutionary algorithm (CI-DMOEA) that simultaneously minimizes the total disassembly cycle time and the number of robotic units required. In particular, a causal feature selection technique based on conditional independence testing is used for efficient initial population generation, enhancing adaptability in dynamic environments. Extensive comparative experiments on eight disassembly scenarios against three state-of-the-art dynamic multi-objective optimization algorithms show that the proposed CI-DMOEA demonstrates superior performance and responsiveness.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 470-482"},"PeriodicalIF":14.2,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145266688","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 : 2025-10-04DOI: 10.1016/j.jmsy.2025.09.019
Guofu Ding , Mingyuan Liu , Haojie Chen , Jian Zhang , Shuying Wang , Jiaxiang Xie
Digital Twin (DT) has become a key enabling technology for intelligent upgrading in discrete manufacturing shops. However, existing research primarily focuses on specific production applications, lacking consideration of the continuity across the overall production operation cycle. Such decentralized research leads to model fragmentation, data silos, and logical inconsistencies among multiple application scenarios such as pre-production planning, in-production execution, and post-production analysis, making effective integration and collaboration difficult. To address these challenges, this paper proposes a Seven-Element virtual reconstruction theory that enables consistent modeling of production elements, organizational forms, and execution logic. Based on this theory, a DT shop construction and operation framework centered on unified production logic is developed to support seamless integration and collaboration of various production applications. Additionally, operation methods for three core production applications of DT shop execution, simulation, and monitoring are systematically developed, establishing an overall technical system throughout the entire production cycle driven by a unified model. Corresponding DT industrial software systems are developed to support engineering implementation of the proposed methods. Validation through an actual shop floor demonstrates that the proposed method effectively achieves model unification and data fusion across multiple application scenarios, enhances both effectiveness and consistency of DT shop construction and operation.
{"title":"Virtual reconstruction-based digital twin shop floor: theoretical methodology, industrial software, and applications","authors":"Guofu Ding , Mingyuan Liu , Haojie Chen , Jian Zhang , Shuying Wang , Jiaxiang Xie","doi":"10.1016/j.jmsy.2025.09.019","DOIUrl":"10.1016/j.jmsy.2025.09.019","url":null,"abstract":"<div><div>Digital Twin (DT) has become a key enabling technology for intelligent upgrading in discrete manufacturing shops. However, existing research primarily focuses on specific production applications, lacking consideration of the continuity across the overall production operation cycle. Such decentralized research leads to model fragmentation, data silos, and logical inconsistencies among multiple application scenarios such as pre-production planning, in-production execution, and post-production analysis, making effective integration and collaboration difficult. To address these challenges, this paper proposes a Seven-Element virtual reconstruction theory that enables consistent modeling of production elements, organizational forms, and execution logic. Based on this theory, a DT shop construction and operation framework centered on unified production logic is developed to support seamless integration and collaboration of various production applications. Additionally, operation methods for three core production applications of DT shop execution, simulation, and monitoring are systematically developed, establishing an overall technical system throughout the entire production cycle driven by a unified model. Corresponding DT industrial software systems are developed to support engineering implementation of the proposed methods. Validation through an actual shop floor demonstrates that the proposed method effectively achieves model unification and data fusion across multiple application scenarios, enhances both effectiveness and consistency of DT shop construction and operation.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 424-444"},"PeriodicalIF":14.2,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145220749","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 : 2025-10-04DOI: 10.1016/j.jmsy.2025.09.017
Jiaming Zhang , Jiewu Leng , Xuming Lai , Libin Lin , Linshan Ding , Lei Yue
With the continuous advancement of intelligent manufacturing, the reconfigurable manufacturing system (RMS) has become an important development direction for modern manufacturing industry by virtue of its high degree of flexibility and reconfigurable characteristics. As a concrete realization form of RMS, reconfigurable automated production line (RAPL) provides an effective technical path to cope with diversified and individualized market demands. In this study, a multi-constraint mathematical model is constructed around the cell configuration and balance optimization problem in RAPL, taking into account the different production line organization methods and cell service modes. Multi-objectives are established involving the minimization of the cycle time, the smoothing index among the manufacturing cells, and the total number of machines of the RAPL. Recognizing the collaborative interaction between mobile robots and machines, a specific theoretical cycle time derivation method is proposed for this RAPL system, and a general-purpose simulation model is designed to support the evaluation and optimization of multiple configuration schemes, thereby verifying the accuracy of the derivation model (with an error of only 1.5 %). To overcome the inefficiency and trial-and-error nature of manual methods, a multi-objective chaotic evolutionary algorithm (MOCEO) is developed. MOCEO demonstrates superior performance and stability, achieving high-quality solutions in a single run and outperforming classical algorithms such as NSGA-II and SPEA2 in hypervolume (HV), distance (GD) and other metrics. The proposed approach provides reliable decision-making support, enabling efficient and effective configuration and balancing of RAPL systems.
{"title":"Multi-objective chaotic evolutionary-based cell configuration and load balancing for reconfigurable production lines","authors":"Jiaming Zhang , Jiewu Leng , Xuming Lai , Libin Lin , Linshan Ding , Lei Yue","doi":"10.1016/j.jmsy.2025.09.017","DOIUrl":"10.1016/j.jmsy.2025.09.017","url":null,"abstract":"<div><div>With the continuous advancement of intelligent manufacturing, the reconfigurable manufacturing system (RMS) has become an important development direction for modern manufacturing industry by virtue of its high degree of flexibility and reconfigurable characteristics. As a concrete realization form of RMS, reconfigurable automated production line (RAPL) provides an effective technical path to cope with diversified and individualized market demands. In this study, a multi-constraint mathematical model is constructed around the cell configuration and balance optimization problem in RAPL, taking into account the different production line organization methods and cell service modes. Multi-objectives are established involving the minimization of the cycle time, the smoothing index among the manufacturing cells, and the total number of machines of the RAPL. Recognizing the collaborative interaction between mobile robots and machines, a specific theoretical cycle time derivation method is proposed for this RAPL system, and a general-purpose simulation model is designed to support the evaluation and optimization of multiple configuration schemes, thereby verifying the accuracy of the derivation model (with an error of only 1.5 %). To overcome the inefficiency and trial-and-error nature of manual methods, a multi-objective chaotic evolutionary algorithm (MOCEO) is developed. MOCEO demonstrates superior performance and stability, achieving high-quality solutions in a single run and outperforming classical algorithms such as NSGA-II and SPEA2 in hypervolume (HV), distance (GD) and other metrics. The proposed approach provides reliable decision-making support, enabling efficient and effective configuration and balancing of RAPL systems.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 445-469"},"PeriodicalIF":14.2,"publicationDate":"2025-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145220748","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 : 2025-09-29DOI: 10.1016/j.jmsy.2025.09.014
Liping Ma , Zhengjie Yang , Hongjuan Yan , Dehu Gao , Xurong Gong
In precision assembly scenarios such as aerospace and automotive engineering, the random starting positions of internal and external threads pose a significant challenge. While achieving specified tightening torque ranges is critical for sealing integrity, precisely controlling the final orientation of threaded connections remains difficult for varying thread pairings. This study proposes a framework integrating visual feature extraction with pre-trained large language models (LLMs) to enable controlled assembly of randomly aligned threads. Using the directional tightening process of hydraulic cylinder barrels and pipe fittings as a case study, the method’s feasibility is validated: First, computer vision techniques extract thread assembly features; then, servo-driven tightening devices perform directional tightening experiments on different fittings, with results recorded. Through structured prompt engineering, assembly parameters, visual features, and experimental outcomes are input into the LLM, the gasket thickness and thread phase are regarded as the controlled input variables, while the collaborative condition judgment of tightening torque and end orientation serves as the output variables. Results demonstrate that pre-trained LLMs, unlike traditional deep learning methods, not only adapt to raw data but also accurately predict directional tightening outcomes for randomly selected shims without requiring additional training. This work provides a novel approach for applying LLMs in precision assembly.
{"title":"Controlled assembly of random threads based on large language models","authors":"Liping Ma , Zhengjie Yang , Hongjuan Yan , Dehu Gao , Xurong Gong","doi":"10.1016/j.jmsy.2025.09.014","DOIUrl":"10.1016/j.jmsy.2025.09.014","url":null,"abstract":"<div><div>In precision assembly scenarios such as aerospace and automotive engineering, the random starting positions of internal and external threads pose a significant challenge. While achieving specified tightening torque ranges is critical for sealing integrity, precisely controlling the final orientation of threaded connections remains difficult for varying thread pairings. This study proposes a framework integrating visual feature extraction with pre-trained large language models (LLMs) to enable controlled assembly of randomly aligned threads. Using the directional tightening process of hydraulic cylinder barrels and pipe fittings as a case study, the method’s feasibility is validated: First, computer vision techniques extract thread assembly features; then, servo-driven tightening devices perform directional tightening experiments on different fittings, with results recorded. Through structured prompt engineering, assembly parameters, visual features, and experimental outcomes are input into the LLM, the gasket thickness and thread phase are regarded as the controlled input variables, while the collaborative condition judgment of tightening torque and end orientation serves as the output variables. Results demonstrate that pre-trained LLMs, unlike traditional deep learning methods, not only adapt to raw data but also accurately predict directional tightening outcomes for randomly selected shims without requiring additional training. This work provides a novel approach for applying LLMs in precision assembly.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"83 ","pages":"Pages 392-404"},"PeriodicalIF":14.2,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145220736","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