Journal of Manufacturing Systems最新文献

{"title":"A4PS: Agentic AI-assisted advanced planning and scheduling with large language models for smart manufacturing","authors":"Mingxing Li ,&nbsp;Qu Zhou ,&nbsp;Wanshan Li ,&nbsp;Ting Qu ,&nbsp;Maolin Yang ,&nbsp;Pingyu Jiang","doi":"10.1016/j.jmsy.2026.01.003","DOIUrl":"10.1016/j.jmsy.2026.01.003","url":null,"abstract":"<div><div>Advanced Planning and Scheduling (APS) for manufacturing systems is becoming more complex against the backdrop of intelligent transformation and increasing demand for customisation. In real-world APS applications subject to multi-source dynamics, objective alterations, constraints removals/additions, algorithm upgrades are inevitable. Such structural changes of APS, requiring seamless coordination among experts such as production managers, modelling engineers, algorithm developers, are often lengthy and less flexible. This poses new challenges in cross-domain/inter-process coordination and rapid multi-disciplinary knowledge integration/reuse. This paper proposes a novel <strong><em>A</em></strong><em>gentic</em> <strong><em>A</em></strong><em>I-</em><strong><em>A</em></strong><em>ssisted</em> <strong><em>APS</em></strong> (A⁴PS) framework, utilising Large Language Models (LLMs) and agents to assist modification/update processes of APS. Firstly, a multi-agentic AI-enabled workflow is designed following standard operating procedure of APS to facilitate the cross-domain agent coordination. Secondly, a multi-step knowledge augmentation method is proposed to endow LLM agents with specialised APS knowledge. Thirdly, a Retrieval-Augmented Generation (RAG) and Chain of Thought (CoT)-enhanced method is developed for knowledge use and interaction. Experiments are conducted with an APS dataset which is created based on classical APS cases and manufacturing researchers. Compared with basic LLMs, A⁴PS exhibited substantially superior performance across both basic and complex cases in metrics such as modelling task success rate, absolute percentage error of solution results, optimisation algorithm code logic completion rate, and code executability rate. Case study demonstrates that A⁴PS enables LLMs to coordinate, learn APS knowledge, and imitate experts in the reasoning process, and ultimately realise APS assistance using natural language. This work proposes a novel solution that uses LLMs and agentic AI to assist APS modification/update process, contributing to AI-driven smart manufacturing in Industry 4.0.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 207-226"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978607","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 collaborative process parameter recommender system for fleets of networked manufacturing machines — with application to 3D printing","authors":"Sicong Guo ,&nbsp;Weishi Wang ,&nbsp;Chenhuan Jiang ,&nbsp;Mohamed Elidrisi ,&nbsp;Myungjin Lee ,&nbsp;Harsha V. Madhyastha ,&nbsp;Raed Al Kontar ,&nbsp;Chinedum E. Okwudire","doi":"10.1016/j.jmsy.2025.12.028","DOIUrl":"10.1016/j.jmsy.2025.12.028","url":null,"abstract":"<div><div>Fleets of networked manufacturing machines of the same type, that are collocated or geographically distributed, are growing in popularity. An excellent example is the rise of 3D print farms, which consist of multiple networked 3D printers operating in parallel, enabling faster production and efficient mass customization. However, optimizing process parameters across a fleet of manufacturing machines, even of the same type, remains a challenge due to machine-to-machine variability. Traditional trial-and-error approaches are inefficient, requiring extensive testing to determine optimal process parameters for an entire fleet. In this work, we introduce a machine learning-based collaborative recommender system that optimizes process parameters for each machine in a fleet by modeling the problem as a sequential matrix completion task. Our approach leverages spectral clustering and alternating least squares to iteratively refine parameter predictions, enabling real-time collaboration among the machines in a fleet while minimizing the number of experimental trials. We validate our method using a mini 3D print farm consisting of ten 3D printers for which we optimize acceleration and speed settings to minimize surface roughness and printing time, thus maximizing print quality and productivity. Our approach achieves significantly faster convergence to optimal process parameters relative to a comparable non-collaborative technique.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 22-33"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145882953","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":"The end-to-end real-time scheduling method for hybrid flow shop based on heterogeneous cooperative multi-agent deep reinforcement learning","authors":"Siqi Liu ,&nbsp;Haiping Zhu ,&nbsp;Liezheng Shen ,&nbsp;Yinda Li ,&nbsp;Junbao Zhang","doi":"10.1016/j.jmsy.2026.02.011","DOIUrl":"10.1016/j.jmsy.2026.02.011","url":null,"abstract":"<div><div>Personalized customization increases the complexity of the hybrid flow shop production environment, which is characterized by the presence of variable discrete events. The buffer capacities of machines and the transportation times between processing stages and the warehouse influence the overall production process. Therefore, the research problem addressed in this paper is the real-time scheduling problem for hybrid flow shops with limited buffers and transportation time. To effectively tackle this challenging scheduling problem, this paper proposes a discrete-event-driven heterogeneous cooperative multi-agent deep reinforcement learning (HC-MADRL) framework. Within this framework, two types of heterogeneous scheduling agents are designed: the Job Agent (JA) and Stage-Machine Agents (SMAs). The JA employs an integrated LSTM-Pointer Network (LPN) architecture to select the most appropriate job based on the current job sequence states in real time. In contrast, the SMAs utilize a Transformer-based weighted network (TWN) incorporating a machine mask mechanism to assign the selected job to the optimal machine according to the state of its processing stage. Then, eight real-time job states and five generic stage-machine states are designed for the JA and SMAs, respectively. Meanwhile, a multi-level reward mechanism with reward shaping is proposed to guide the agents in learning to optimize the scheduling objective. Based on this, this paper presents an independent multi-agent proximal policy optimization (I-MAPPO) training algorithm that integrates two critic networks for heterogeneous multi-agent systems. Lastly, the experimental results demonstrate that the proposed method exhibits superior scheduling performance and real-time responsiveness compared with four benchmark real-time scheduling methods: PDRs, DQN, PPO, MAPPO and other two advanced multi-agent schedulers.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 487-512"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188652","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 Human–Robot collaborative framework for draping of advanced composite materials","authors":"Matteo Terreran ,&nbsp;Alberto Gottardi ,&nbsp;Giorgio Nicola ,&nbsp;Dominik Zielinski ,&nbsp;Christoph Frommel ,&nbsp;Matthias Rehm ,&nbsp;Jakob Sabban ,&nbsp;Giuseppe Salemi ,&nbsp;Dieter Christ ,&nbsp;Christian Eitzinger ,&nbsp;Stefano Ghidoni ,&nbsp;Enrico Villagrossi","doi":"10.1016/j.jmsy.2026.01.018","DOIUrl":"10.1016/j.jmsy.2026.01.018","url":null,"abstract":"<div><div>Manual draping carbon fibre and fibreglass plies is a critical yet labour-intensive step in composite manufacturing, where production efficiency and part quality strongly depend on operator skill and repeatability. This paper addresses the challenges of introducing Human–Robot Collaboration (HRC) to support draping operations while maintaining flexibility, safety, and industrial feasibility. The system featured intuitive human–robot interaction through a human action recognition system, a context reasoner, and a human–robot collaborative transport module designed for manipulating planar deformable objects. The framework was validated on two robotic workcells addressing three industrial use cases from the aerospace, automotive and maritime sectors. An extensive experimental campaign, culminating in the production of multi-layer components, demonstrated substantial gains in efficiency and operator support: process time for large components was reduced by 50%, while operator workload decreased by 40%. The robot’s precision in material placement also cut ply draping time per layer from 12% in manual operations to 6%, contributing to lower physical and cognitive demands on operators (workload scores averaging below 30%). The study shows that the benefits of HRC in draping operations are highly context-dependent: partial automation yields limited gains for small/medium components, whereas for large components it significantly improves performance by reducing cycle time, enhancing product quality, and minimising ergonomic strain on operators. Finally, the paper presents a critical analysis of the proposed solution, highlighting its drawbacks and suggesting improvements.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 455-472"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146187920","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":"Automated extraction of comprehensive digital twin models for smart manufacturing systems","authors":"Atieh Khodadadi ,&nbsp;Sanja Lazarova-Molnar","doi":"10.1016/j.jmsy.2026.01.012","DOIUrl":"10.1016/j.jmsy.2026.01.012","url":null,"abstract":"<div><div>Manufacturing systems involve multiple, often conflicting, objectives referred to as performance indicators, including production efficiency, resource utilization, energy consumption, carbon emissions, and waste reduction, which correspond to different dimensions of the system, such as time, energy consumption, and waste generation aspects. Digital Twins have emerged as a powerful tool, integrating data-driven simulation and analysis of complex systems, such as manufacturing systems. Process Mining (PM), along with data analysis, enables the automatic discovery of executable discrete-event simulation models directly from production event logs. These data-driven models are the key to enabling near-real-time Digital Twins of discrete-event systems. Stochastic Petri Nets (SPNs) offer a robust and intuitive modeling formalism well-suited for representing the extracted models derived from PM, particularly in the context of manufacturing systems. However, standard SPNs face challenges in incorporating dimensions beyond time, such as energy consumption and waste generation. This limitation often results in suboptimal decision-making and reduced system efficiency. In this paper, we propose a Comprehensive Digital Twin (CDT) framework that employs Multi-Flow Process Mining (MFPM) to automatically extract Multidimensional Stochastic Petri Nets (MDSPNs) as underlying models of manufacturing systems. To support the modeling and simulation of extracted MDSPNs, we introduce and utilize our tool, MDPySPN. The CDT framework supports multi-objective decision-making for various performance indicators of the system. Through an illustrative case study of hot forging process chains, we showcase the development of CDT for time, energy consumption, and waste generation dimensions. We further illustrate the utilization of CDT to analyze and support decision-making to enhance the case study system according to its objectives.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 287-306"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146078234","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":"Machine vision defect segmentation and geometric measurement for real time quality monitoring in friction stir welding","authors":"Naveen Loganathan,&nbsp;Joel Andersson,&nbsp;Vivek Patel","doi":"10.1016/j.jmsy.2026.01.007","DOIUrl":"10.1016/j.jmsy.2026.01.007","url":null,"abstract":"<div><div>Weld quality in friction stir welding (FSW) is difficult to maintain because rapid changes in heat input and material flow can generate transient surface defects during welding. These defects cannot be detected in real time using conventional inspection approaches, resulting in increased inspection time and higher production cost. Real-time visual monitoring is therefore required to support stable and efficient production. This study investigates whether modern convolutional neural network (CNN) models can provide reliable, in-situ segmentation of FSW surface defects together with accurate geometric measurements during welding. A multi-class dataset of weld-surface video frames was created and annotated for flash, burrs, voids, galling, tool interaction, and weld-zone regions. Several CNN-based segmentation models were evaluated, and a lightweight architecture suitable for real-time deployment was selected and integrated with a high-dynamic-range industrial camera on the FSW setup. The system performs continuous segmentation and extracts weld width and defect area from live video at approximately 25 frames per second. Quantitative validation against optical-microscope measurements demonstrated near microscope-level accuracy, with sub-millimetre weld-width deviations and defect-area errors below 6 %. These results demonstrate that real-time visual segmentation can provide reliable weld-quality monitoring in FSW, support early defect detection, and establish a practical foundation for future automated process-control strategies in manufacturing environments.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 175-192"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978549","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":"Computer-aided manufacturing in reality: augmented reality-enabled precision motion planning for robotic repair of complex shapes","authors":"Chih-Hsing Chu ,&nbsp;Cong-Yun Lai ,&nbsp;Chih-Husan Shih","doi":"10.1016/j.jmsy.2026.01.006","DOIUrl":"10.1016/j.jmsy.2026.01.006","url":null,"abstract":"<div><div>Augmented Reality (AR) is increasingly used as a user interface for robot motion planning in manufacturing environments. However, the accuracy of 3D pointing (waypoint input) in most AR applications is insufficient for tasks that demand high precision. This study introduces the concept of Computer-Aided Manufacturing in Reality (CAMiR) to enable high-precision motion planning and path computation for robotic welding and grinding operations. Human interactions within the AR scene, intended for visualization purposes, are separated from spatial pointing, used for defining robot movements. The points manually specified using a self-tracking tool are aligned with the measurements obtained from an external laser profiler through 3D registration, thereby reducing positional deviations relative to the workpiece. Geometric modeling functions are accessible through AR interfaces to construct geometric elements from the aligned points, which define processing regions and generate corresponding paths. A prototyping system implementing CAMiR allows operators to plan and simulate robot motion directly on real parts without the need to reconstruct CAD models. Evaluation results on test workpieces indicate that the pointing accuracy is significantly improved compared with previous studies and the ground surface exhibits superior geometric precision. The repair of complex 3D defects using robotic motions generated by the system verifies the effectiveness of the proposed approach and highlights its potential to enhance the practical applicability of AR interfaces in high-precision manufacturing tasks.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 193-206"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979109","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 double-sided bundle auction mechanism for collaborative additive manufacturing","authors":"Mingyue Sun ,&nbsp;Jinpeng Li ,&nbsp;Jiaxin Fan ,&nbsp;Mengdi Zhang ,&nbsp;Zhiheng Zhao ,&nbsp;George Q. Huang","doi":"10.1016/j.jmsy.2026.02.006","DOIUrl":"10.1016/j.jmsy.2026.02.006","url":null,"abstract":"<div><div>Collaborative manufacturing has become an essential strategy for improving resource utilization and profitability in the additive manufacturing (AM) industry. In this study, we propose an additive manufacturing collaboration platform (AMCP) to facilitate bilateral resource exchange in decentralized AM environments. To address the challenges posed by the two-way, many-to-many exchange of heterogeneous AM resources in AMCP, we design a five-phase double-sided bundle auction (DBA) mechanism. This mechanism is proven to satisfy incentive compatibility, individual rationality, budget balance, and allocative efficiency, ensuring fair and efficient collaboration among self-interested participants. To optimize the winner determination problem in DBA, which arises from overlapping transactions and flexible bundle bidding, we reformulate it as a maximum weight independent set problem in graph theory. Based on this reformulation, we develop a graph-based adaptive neighborhood search algorithm that balances computational efficiency and solution quality. The experimental results demonstrate that the proposed mechanism is both feasible and robust. It achieves superior social welfare, cooperation rates, and scalability when compared to traditional approaches. Additionally, sensitivity analysis reveals the robustness and resilience of the proposed mechanism, even in the presence of strategic manufacturers.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 473-486"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146187921","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":"Wafer defect semantic reasoning via a three-stage retrieval-augmented system in semiconductor manufacturing","authors":"Xinting Liao ,&nbsp;Jie Zhang ,&nbsp;Youlong Lyu ,&nbsp;Junliang Wang","doi":"10.1016/j.jmsy.2025.12.026","DOIUrl":"10.1016/j.jmsy.2025.12.026","url":null,"abstract":"<div><div>Wafer defect detection is a crucial part of semiconductor manufacturing, requiring both high generalization across distribution shifts and interpretable explanations for diverse engineering roles. Existing deep learning methods often fall short in semantic reasoning and domain transferability. To address these challenges, we propose WaferDSR-RAG, a novel retrieval-augmented framework that integrates a bimodal wafer defect knowledge graph (BWDKG) with a three-stage semantic adaptation strategy, comprising defect visual-semantic alignment, defect-relevant knowledge retrieval and screening, and role-adaptive explanation generation. This design allows the system to dynamically adapt to unseen defect distributions and generate expert-aligned explanations tailored to different engineering responsibilities. Extensive experiments on both in-distribution and cross-fab wafer maps demonstrate that WaferDSR-RAG consistently outperforms state-of-the-art baselines in both detection accuracy and explanation quality. Compared to GPT-4o and Gemini 1.5 Pro, our method generates more semantically accurate and practically useful explanations for engineers with different roles, as validated by automatic metrics and domain expert evaluations. WaferDSR-RAG represents a scalable solution for wafer defect detection and reasoning in real-world semiconductor production.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 269-286"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146034981","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 physics-informed embodied intelligence framework for smart garment manufacturing","authors":"Shufei Li ,&nbsp;Ao Dong ,&nbsp;Ray Wai Man Kong ,&nbsp;Cheng Liu","doi":"10.1016/j.jmsy.2026.01.015","DOIUrl":"10.1016/j.jmsy.2026.01.015","url":null,"abstract":"<div><div>Embodied intelligence systems are becoming integral to smart manufacturing, driven by the human-centric, sustainable, and resilient vision of Industry 5.0. While robots have achieved remarkable success in manipulating rigid objects, robust handling of deformable fabrics in garment manufacturing remains challenging, particularly for pick-and-place operations required in sewing processes. In production environments, robots often struggle to adapt gripping strategies to different fabric properties and deformation behavior, limiting smart manufacturing development. To address these challenges, this paper proposes a physics-informed embodied intelligence framework tailored for smart garment manufacturing. The system integrates task-aware planning, enabled by vision–language understanding of the surrounding environment, with physics-informed manipulation, which incorporates material property inference and finite element analysis. This integration enables robots to perform garment manufacturing tasks with greater adaptability, precision, and intelligence. Experiments conducted in real-world scenarios validate the effectiveness of the proposed framework in supporting flexible and intelligent production processes.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"85 ","pages":"Pages 307-317"},"PeriodicalIF":14.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146078236","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}