Journal of Manufacturing Systems最新文献

{"title":"Machine node-enhanced graph contrastive learning with long-range prompt model for quality propagation in multistage manufacturing systems","authors":"Pei Wang ,&nbsp;Hai Qu ,&nbsp;Jinrui Liu","doi":"10.1016/j.jmsy.2025.05.018","DOIUrl":"10.1016/j.jmsy.2025.05.018","url":null,"abstract":"<div><div>More complex multistage manufacturing systems (MMSs) have become mainstream production processes. Deep learning can accurately predict product quality indicators and help improve product qualification rates. Existing single-stage models can only predict the quality of a single machine or a single stage, ignoring the propagation effects. Although some multistage models consider quality propagation effects, they simply aggregate the machine features within a stage. In addition, existing multistage models do not conform to the actual production process when utilizing machine space relationships. Traditional models with severe label data dependence are difficult to make full use of noise data, resulting in reduced prediction accuracy and lack of interpretability. To address the above problems, this paper proposes a machine node-enhanced graph contrastive learning method with long-range prompt (MNGCLP), extracting propagation effects in both pre-training and fine-tuning. Specifically, the graph data is first used to model the production relationship between machines and form a production process graph. Then, a machine-enhanced graph is designed in contrastive-based pre-training to better add production information to avoid violating the production process and reduce label dependency. Finally, the long-distance relationship between machine nodes captured in the original graph is used as prompts to fine-tune the pre-trained model. Experiments in public production data show that the proposed method outperforms traditional models and provides a reasonable explanation for the prediction results, validating the effectiveness of MNGCLP. Compared with supervised graph neural network, the proposed model improves the overall RMSE, MSE and MAE by 2.4 %, 4.8 % and 9.8 %, respectively. Compared with the contrastive graph neural network, the proposed model improves the overall RMSE, MSE and MAE by 2.5 %, 5.0 %, and 6.3 %, respectively.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 169-188"},"PeriodicalIF":12.2,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144138251","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}

{"title":"Spatial-temporal dual-stream fusion network with embedded knowledge: A process-generalized framework for tool wear monitoring","authors":"Zhixiang Chen ,&nbsp;Jin Zhang ,&nbsp;Fengze Qin ,&nbsp;Guibao Tao ,&nbsp;Duo Li ,&nbsp;Huajun Cao","doi":"10.1016/j.jmsy.2025.05.014","DOIUrl":"10.1016/j.jmsy.2025.05.014","url":null,"abstract":"<div><div>As a pivotal technology in intelligent computer numerical control (CNC) systems, tool wearing monitoring (TWM) significantly influences machining stability, product quality, and production efficiency. However, the complexity and randomness of the tool wear process pose significant challenges to traditional single-model methods. These methods often struggle to capture diverse data patterns, leading to low accuracy and poor processing generalization. This study proposes a Spatial-Temporal Dual Stream Fusion Network (STDSFnet) with embedded knowledge. The framework establishes strong correlations between multi-modal sensor data and tool wear states. The methodology consists of three key steps: (1) applying Spearman correlation analysis to select sensitive feature vectors, which helps reduce transient signal interference; (2) developing a dual-stream network to extract spatial and temporal wear-related information; and (3) integrating coordinated attention mechanisms and feature fusion modules to enhance the feature representation, thereby improve the TWM accuracy and processing generalization. Besides, to ensure that the trend of TWM results conforms to the actual wear pattern, domain knowledge is introduced to improve the reliability and interpretability of monitoring results. Experiments validate the effectiveness of STDSFnet on block-shaped stainless-steel workpieces and carbon fiber-reinforced polymer (CFRP) machining scenarios. Compared to baseline methods, STDSFnet reduces RMSE by 38.94 %–62.04 %, decreases MAE by 43.64 %–61.70 %, and improves R² by 4.96 %–28.59 %. These results confirm that spatial-temporal fusion with deep ensembles significantly boosts TWM accuracy and reliability.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 189-207"},"PeriodicalIF":12.2,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144146699","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}

{"title":"Train small, deploy large: Scaling multi-agent reinforcement learning for multi-stage manufacturing lines","authors":"Kshitij Bhatta,&nbsp;Qing Chang","doi":"10.1016/j.jmsy.2025.04.017","DOIUrl":"10.1016/j.jmsy.2025.04.017","url":null,"abstract":"<div><div>We present a novel control framework using Multi Agent Reinforcement learning (MARL) that is scalable in the number of workstations in a multi-stage manufacturing line. We show that the dynamics of any production line, regardless of size, can be decoupled into three fundamental expressions. These expressions capture the dynamics of (1) the first workstation, (2) all intermediate workstations, and (3) the last workstation. This decoupling, combined with observation engineering enables training a characteristic 3-workstation, 2-buffer model using MARL methods, which can then generalize to production lines with <span><math><mi>w</mi></math></span> workstations with arbitrary cycle times, buffer capacities and reliability models. A numerical study is then conducted to validate the framework.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 155-168"},"PeriodicalIF":12.2,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134807","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}

{"title":"Coarse-to-fine vision-based welding spot anomaly detection in production lines of body-in-white","authors":"Weijie Liu,&nbsp;Jie Hu,&nbsp;Jin Qi","doi":"10.1016/j.jmsy.2025.05.003","DOIUrl":"10.1016/j.jmsy.2025.05.003","url":null,"abstract":"<div><div>Computer vision-assisted methods for weld quality inspection enable rapid and automated surface defect detection through image data. However, the application of Computer Vision-Assisted Inspection (CVAI) in real-world production lines faces substantial, long-term challenges due to complex environments, imbalanced data samples, real-time processing demands, and safety requirements. Our paper proposes a novel two-stage Coarse-to-Fine Anomaly Detection (CTFAD) framework, which integrates the YOLOv8 network architecture for initial detection with an ensemble of neural networks for fine-grained classification. Additionally, we introduce a voting-based algorithm for improved decision-making accuracy. Experimental results on real-world datasets demonstrate that, compared to standard end-to-end methods, CTFAD enhances detection accuracy and operational efficiency. Our contributions include (1) proposing the CTFAD pipeline for weld anomaly detection, (2) establishing voting-based classification module to increase system robustness and generalization, and (3) developing an integrated weld detection system encompassing data acquisition, processing, analysis, and anomaly alerting. Our code is available at <span><span>https://github.com/wj-liu0730/ctfad-jms</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 144-154"},"PeriodicalIF":12.2,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144115901","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}

{"title":"Robotic aero-engine pipe grasping posture and motion planning method in multi-stage processing based on multi-objective optimization","authors":"Bo Zhou ,&nbsp;Jibin Zhao ,&nbsp;Tianyu Zhang ,&nbsp;Renbo Xia ,&nbsp;Hongfeng Wang ,&nbsp;Junwei Wang ,&nbsp;Long Gao","doi":"10.1016/j.jmsy.2025.05.011","DOIUrl":"10.1016/j.jmsy.2025.05.011","url":null,"abstract":"<div><div>In the aviation industry, the manufacturing of aero-engine bending pipes consists of multiple processes, e.g., straight pipe bending, allowance removal, precision chamfering, and inspection. Currently, the production of aero-engine bending pipes is carried out mainly manually, and uncontrollable human factors can greatly affect the manufacturing accuracy and efficiency. This paper introduces a grasping posture and motion planning method. It can efficiently and accurately perform multi-stage precision manufacturing tasks on aero-engine bending pipe production lines. First, we provide a gravity deformation model for flexible bending pipes and verify its accuracy through finite element simulations and experiments. It can be used to compensate accurately for the motion of bending pipes entering the processing hole along a straight path. We propose a coordinate transformation method and an accurate calibration method. It can achieve precise transformation of the original posture of the computer-aided design (CAD) model and the postures at the processing stations, e.g., the outlet of the clamping mold of the bending machine, the processing holes of the residual removal and the flat end chamfer. Next, we propose converting the linear paths into NURBS interpolation paths and using the S-shaped acceleration/deceleration (ACC/DEC) feed rate planning method. It can convert the velocity, acceleration, and jerk into the corresponding angular velocity, acceleration, and jerk of each joint in the joint space to ensure that processing and inspection tasks can be performed under appropriate kinematic constraints. Then, an improved NSGA-II algorithm is proposed, which can solve the trade-off minimization problem of processing time, energy consumption, and vibration by incorporating multi-objective optimization functions and motion constraints. The simulation and experimental results demonstrated the superiority of optimization method in terms of processing accuracy, precision, stability, and efficiency. The performance of the improved NSGA-II algorithm is compared with that of other multi-objective optimization algorithms, and the superior performance of the algorithm is verified.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 117-143"},"PeriodicalIF":12.2,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099869","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}

{"title":"Deposition sequence optimization for minimizing substrate plate distortion using the simplified WAAM simulation","authors":"Xiao Fan Zhao,&nbsp;Hannes Panzer,&nbsp;Avelino Zapata,&nbsp;Felix Riegger,&nbsp;Siegfried Baehr,&nbsp;Michael F. Zaeh","doi":"10.1016/j.jmsy.2025.05.010","DOIUrl":"10.1016/j.jmsy.2025.05.010","url":null,"abstract":"<div><div>Wire arc additive manufacturing (WAAM) is a viable alternative to conventional machining or other additive manufacturing technologies, especially for the production of large, thin-walled components. However, finding the optimal deposition sequence for a minimal substrate plate distortion is challenging due to the vast number of possible sequences. The present study tackles this challenge by exploring three distinct objective functions for predicting distortion using the simplified WAAM simulation (SWS) – a semi-analytical model for the time-efficient estimation of thermal histories in WAAM parts. Using the SWS together with three temperature-based objective functions, distortion scores were calculated for each deposition sequence of a four-sectioned wall geometry. A subset of deposition sequences was then simulated using an experimentally validated thermomechanical finite element (FE) simulation. The correlation between the simulated distortion from the FE model and the distortion score from each objective function was analyzed. The results implied a strong and definitive statistical correlation between the substrate plate distortion and one particular objective function which considers the thermal eccentricity. Subsequently, the wall geometry, together with an additional A-shaped geometry, was manufactured using the best, the worst, and a third deposition sequence. After the WAAM process, the substrate plate distortions were measured using a 3D scanner. The scan results validated the prior optimization, indicating the highest distortion for the worst sequence, the lowest distortion for the best sequence, and a level of distortion in between those extremes for the third sequence. The findings of this article can be utilized for the preliminary selection of deposition sequences of WAAM parts.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 103-116"},"PeriodicalIF":12.2,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099868","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}

{"title":"Digital twin modeling of the robotic gluing system for predicting the quality of glue lines and optimizing gluing parameters","authors":"Ruihao Kang ,&nbsp;Junshan Hu ,&nbsp;Mingyu Li ,&nbsp;Qi Zhang ,&nbsp;Xingtao Su ,&nbsp;Zhengping Li ,&nbsp;Wei Tian","doi":"10.1016/j.jmsy.2025.05.008","DOIUrl":"10.1016/j.jmsy.2025.05.008","url":null,"abstract":"<div><div>Digital twin (DT) technology is changing the current pattern of intelligent manufacturing, it makes up for the shortcomings of process parameter optimization methods to improve real-time and predictability. This paper developed DT models for the robotic gluing system to predict the quality (width and thickness) of glue lines and optimize gluing parameters (trajectory and extrusion speeds). The DT framework based on the geometric, physical, behavioral, and rule models is constructed to monitor and optimize the gluing parameters in real-time. An improved backpropagation neural network (BPNN) prediction model based on whale optimization algorithm (WOA) is established to predict the width and thickness of glue lines from historical and real-time data, while simultaneously enabling real-time calculation of the cross-sectional area of glue lines. A multi-objective optimization model constructed using non-dominated sorting genetic algorithm (NSGA-II) is used to optimize the gluing parameters. The DT prototype of the robotic gluing system has been developed and verified experimentally. The position calibration of the geometric model is used to correct the gluing trajectory before gluing, and the position errors of the gluing points are within ± 0.5 mm. The gluing trajectory is designed to test the effectiveness of the adaptive optimization of gluing parameters. The prediction errors of the width and thickness of the glue line are controlled between ± 0.5 mm and ± 0.3 mm, individually. After parameter optimization, the width and thickness of the glue line at the corner are reduced by 4.53 % and 7.54 %, respectively, thus avoiding glue accumulation. This reduction solves the problem of poor consistency in the quality of glue lines and verifies the feasibility of integrated monitoring, prediction, and optimization based on the DT model.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 1074-1092"},"PeriodicalIF":12.2,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144088906","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}

{"title":"Feature points classification of computerized numerical control finishing tool path based on graph neural network","authors":"Jiejun Xie ,&nbsp;Pengcheng Hu ,&nbsp;Yingbo Song ,&nbsp;Xin Liu","doi":"10.1016/j.jmsy.2025.05.015","DOIUrl":"10.1016/j.jmsy.2025.05.015","url":null,"abstract":"<div><div>In Computer Numerical Control (CNC) machining, surface defects occur due to inaccurate recognition of feature points in the finishing tool path and unsmooth feed rate planning by the CNC system. Therefore, accurately identifying feature points of tool path is crucial for high-speed and high-precision CNC machining. Existing algorithms often rely on simple, manually set thresholds and do not consider cross directional geometric information of tool path, leading to poor performance in feature point recognition. This paper presents a new method for identifying and classifying feature points using a graph neural network (GNN) by aggregating geometric features from both the feed and cross directions of the tool path to automatically and accurately identify feature points in finishing tool paths. The method begins by creating a graph-based representation of the tool path, which provides detailed geometric information for Cutter Location (CL) points. It also introduces an algorithm for identifying cross directional points related to CL points and a spatial convolution method that combines feed and cross directional geometric features, based on which a Feature Point-Graph Neural Network (FP-GNN) is constructed. Extensive testing shows that the FP-GNN model performs exceptionally well in classifying tool path feature points, surpassing existing methods. As a direction application of the proposed method, physical machining examples are conducted, demonstrating that optimizing feed rates in the cross direction—based on identified feature points—improves the continuity of the feed rate in both directions, enhancing the surface machining quality.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 75-102"},"PeriodicalIF":12.2,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099867","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}

{"title":"A survey on XR remote collaboration in industry","authors":"Peng Wang ,&nbsp;Huan Yang ,&nbsp;Mark Billinghurst ,&nbsp;Shuang Zhao ,&nbsp;Yao Wang ,&nbsp;Zhou Liu ,&nbsp;Yi Zhang","doi":"10.1016/j.jmsy.2025.05.016","DOIUrl":"10.1016/j.jmsy.2025.05.016","url":null,"abstract":"<div><div>This paper provides a survey of eXtended Reality (XR) remote collaboration in industry. The field of XR remote collaboration is currently at a critical point as collaborative XR systems are becoming more prevalent in industry. Moreover, research on XR remote collaboration in industry is a compelling and evolving field of study, especially because XR technology has recently reached a level of maturity that enables researchers and practitioners to use its capabilities to enhance remote collaboration, rather than focusing solely on the development of the underlying technology. However, to our knowledge, there has yet to be a comprehensive survey on XR remote collaboration in industry. Thus, this paper presents a systematic review of literature published between 2019 and 2024 in this domain. We identified a total of 161 papers, with more than 57 % published in the last three years, and all relevant studies are discussed in detail from fundamental requirements, industrial applications, toolkits/platforms, and system evaluations. Our findings indicate that XR remote collaboration holds great potential across various industrial domains, including design (e.g., design review and co-design), manufacturing (e.g., assembly and inspection), service (e.g., maintenance and monitoring), and training. Since XR remote collaboration is being implemented in real industrial environments, this paper aims to provide a comprehensive academic roadmap and valuable insights into the cutting-edge of XR remote collaboration in industry. This work will also serve as a resource for both current and future researchers who are interested in collaborative XR systems and applications in industry.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 49-74"},"PeriodicalIF":12.2,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099650","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}

{"title":"SFRGNN-DA: An enhanced graph neural network with domain adaptation for feature recognition in structural parts machining","authors":"Xiaohu Zheng ,&nbsp;Hongbo Chen ,&nbsp;Fangzhou He ,&nbsp;Xiaojia Liu","doi":"10.1016/j.jmsy.2025.05.005","DOIUrl":"10.1016/j.jmsy.2025.05.005","url":null,"abstract":"<div><div>Optimizing the recognition of machining features in structural parts is vital for enhancing the efficiency of NC machining planning and ensuring quality control. However, the inherent complexity and stringent precision requirements of these parts often render existing feature recognition methods inadequate for accurately identifying model features. To address this challenge, a novel graph neural network model (SFRGNN) is introduced. The methodology begins with a specialized feature extraction module that captures both geometric and topological properties of the parts, providing a comprehensive basis for further analysis. Following this, SFRGNN integrates a graph neural network with a Spatial Self-Attention (SSA) module, a configuration designed to enhance the extraction of high-level semantic information crucial for accurately distinguishing machining features. This network architecture allows SFRGNN to interpret complex feature relationships with improved precision. Additionally, an enhanced domain adaptation module (DA) is incorporated to improve SFRGNN’s generalization capabilities and performance in machining feature recognition. Numerous experiments on different data sets confirmed that SFRGNN achieved excellent accuracy in identifying real-world structural part features and demonstrated enhanced performance, which will be helpful for subsequent process planning for part features in real-world scenarios.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"81 ","pages":"Pages 16-33"},"PeriodicalIF":12.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144072253","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}