Journal of Manufacturing Systems最新文献

{"title":"Human-LLM collaboration in generative design for customization","authors":"Xingzhi Wang ,&nbsp;Zhoumingju Jiang ,&nbsp;Yi Xiong ,&nbsp;Ang Liu","doi":"10.1016/j.jmsy.2025.03.012","DOIUrl":"10.1016/j.jmsy.2025.03.012","url":null,"abstract":"<div><div>Generative design enables the rapid creation of diverse designs, making it a promising means for customization. However, due to the multidisciplinary knowledge required for operation, the full potential of generative design for customization (GDfC) remains under-explored. Recently, large language models (LLM) have attracted significant attention from designers. Unlike traditional text-based generative models, LLM’s expansive knowledge base and unique interaction capabilities offer clear advantages for assuming more proactive roles in GDfC. Against the background, this paper explores the potential of LLM in redefining GDfC. Based on the division of the generative design process, this paper identifies three human-LLM collaboration schemes to demonstrate the potential roles of LLM in GDfC. Additionally, this paper proposes a process framework based on the characteristics of required design knowledge, which aids designers in selecting the appropriate LLM performance enhancement strategy for their customization tasks. A case study of vehicle interior customization is presented to demonstrate the application of the proposed framework.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 425-435"},"PeriodicalIF":12.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143716329","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":"Real-time decision-making for Digital Twin in additive manufacturing with Model Predictive Control using time-series deep neural networks","authors":"Yi-Ping Chen ,&nbsp;Vispi Karkaria ,&nbsp;Ying-Kuan Tsai ,&nbsp;Faith Rolark ,&nbsp;Daniel Quispe ,&nbsp;Robert X. Gao ,&nbsp;Jian Cao ,&nbsp;Wei Chen","doi":"10.1016/j.jmsy.2025.03.009","DOIUrl":"10.1016/j.jmsy.2025.03.009","url":null,"abstract":"<div><div>Digital Twin – a virtual replica of a physical system enabling real-time monitoring, model updating, prediction, and decision-making – combined with recent advances in machine learning, offers new opportunities for proactive control strategies in autonomous manufacturing. However, achieving real-time decision-making with Digital Twins requires efficient optimization driven by accurate predictions of highly nonlinear manufacturing systems. This paper presents a simultaneous multi-step Model Predictive Control (MPC) framework for real-time decision-making, using a multivariate deep neural network, named Time-Series Dense Encoder (TiDE), as the surrogate model. Unlike conventional MPC models which only provide one-step ahead prediction, TiDE is capable of predicting future states within the prediction horizon in one shot (multi-step), significantly accelerating the MPC. Using Directed Energy Deposition (DED) additive manufacturing as a case study, we demonstrate the effectiveness of the proposed MPC in achieving melt pool temperature tracking to ensure part quality, while reducing porosity defects by regulating laser power to maintain melt pool depth constraints. In this work, we first show that TiDE is capable of accurately predicting melt pool temperature and depth. Second, we demonstrate that the proposed MPC achieves precise temperature tracking while satisfying melt pool depth constraints within a targeted dilution range (10%–30%), reducing potential porosity defects. Compared to Proportional–Integral–Derivative (PID) controller, the MPC results in smoother and less fluctuating laser power profiles with competitive or superior melt pool temperature control performance. This demonstrates the MPC’s proactive control capabilities, leveraging time-series prediction and real-time optimization, positioning it as a powerful tool for future Digital Twin applications and real-time process optimization in manufacturing.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 412-424"},"PeriodicalIF":12.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143705192","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":"Dynamic flexible flow shop scheduling via cross-attention networks and multi-agent reinforcement learning","authors":"Jinlong Zheng ,&nbsp;Yixin Zhao ,&nbsp;Yinya Li ,&nbsp;Jianfeng Li ,&nbsp;Liangeng Wang ,&nbsp;Di Yuan","doi":"10.1016/j.jmsy.2025.03.005","DOIUrl":"10.1016/j.jmsy.2025.03.005","url":null,"abstract":"<div><div>With the increasing uncertainty in production environments and changes in market demand, flexible and efficient scheduling solutions have become particularly critical. However, existing research mainly focuses on static scheduling or relatively simple dynamic scheduling problems, which are inadequate to address the complexities of actual production processes. This paper considers the dynamic flexible flow shop scheduling problem (DFFSP) characterized by diverse processes, complexity, and high flexibility, and proposes a multi-agent reinforcement learning algorithm based on cross-attention networks (MARL_CA). First, this paper proposes a novel state feature representation method, which represents the job processing data and the production Gantt chart as a state matrix, fully reflecting the environment state in the scheduling process. In addition, a cross-attention network is proposed to extract state features, enabling efficient discovery of complex relationships between jobs and machines, thereby enhancing the model's ability to understand intricate features. The model is trained using an independent proximal policy optimization (IPPO) based on the actor-critic method to help agents learn accurate and efficient scheduling strategies. Experimental results on a large number of static and dynamic scheduling instances demonstrate that the proposed algorithm outperforms traditional heuristic rules and other advanced algorithms, exhibiting strong learning efficiency and generalization capability.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 395-411"},"PeriodicalIF":12.2,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143697111","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":"Self-adaptive production performance monitoring framework under different operating regimes","authors":"Joao Paulo Jacomini Prioli ,&nbsp;Nur Banu Altinpulluk ,&nbsp;Jeremy L. Rickli ,&nbsp;Murat Yildirim","doi":"10.1016/j.jmsy.2025.02.011","DOIUrl":"10.1016/j.jmsy.2025.02.011","url":null,"abstract":"<div><div>Dynamic operational regimes in modern manufacturing systems generate a myriad of challenges for production performance monitoring applications. Heterogeneous data streams and fast production changeovers often complicate sensor information, leading to misinterpretation of systemic performance issues. Conventional methods address this problem by explicitly modeling these operational regimes. However, it requires significant engineering hours and expertise, constituting a substantial adoption barrier for small-to-medium enterprises (SMEs). This paper proposes a self-adaptive smart monitoring framework that autonomously discovers and accounts for operational regime changes to offer accurate predictions on systemic performance despite the complexities in continuous multi-sourced data acquisition and dynamic regime behavior of machines. Computational experiments tested the methodology using a predictive system in two manufacturing cells under dynamic operational regimes. The proposed framework outperforms benchmark policies commonly used in prediction models by improving prediction accuracy from 3% to 62%, along with a better convergence rate. The results demonstrated that the proposed framework can positively impact smart maintenance implementation for SMEs with limited resources.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 380-394"},"PeriodicalIF":12.2,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143697110","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 learning-enhanced digital twins for predictive analytics in battery pack assembly","authors":"Muhammad Waseem ,&nbsp;Changbai Tan ,&nbsp;Seog-Chan Oh ,&nbsp;Jorge Arinez ,&nbsp;Qing Chang","doi":"10.1016/j.jmsy.2025.03.007","DOIUrl":"10.1016/j.jmsy.2025.03.007","url":null,"abstract":"<div><div>The electric vehicle (EV) market is rapidly growing, with battery modules playing a central role in this transformation. However, optimizing production throughput in battery module assembly is challenging due to the complexity of multi-stage processes and bottlenecks that limit overall efficiency. Traditional solutions, such as direct shop floor adjustments, simulation models, and digital twins (DT), can be costly and less scalable. This study proposes a digital twin surrogate (DTS) model, integrating machine learning techniques—Linear Regression, Support Vector Regression, K-Nearest Neighbors, Random Forest Regression, Deep Neural Networks, XGBoost, and Long Short-Term Memory networks—to estimate throughput and predict future machine states. The impact of dataset size and aggregation methods on model performance is also examined, providing shop managers with insights into how production line variations affect throughput.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 344-355"},"PeriodicalIF":12.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683288","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 generic digital twin model construction strategy for cross-field implementations with comprehensiveness, operability and scalability","authors":"Xiaojun Liu ,&nbsp;Chongxin Wang ,&nbsp;Feixiang Wang ,&nbsp;Xiaoli Qiu ,&nbsp;Fengyi Feng ,&nbsp;Yang Sun","doi":"10.1016/j.jmsy.2025.02.020","DOIUrl":"10.1016/j.jmsy.2025.02.020","url":null,"abstract":"<div><div>In recent years, the prominence of Digital Twins as pivotal tools in digitization and intelligence has sparked widespread interest. However, the diversity of Digital Twin applications has led to a plethora of evolving technologies, standards, and building methods. These varying terms and frequent incompatibilities necessitate a unified approach to characterize and craft Digital Twin models. This endeavor aims not only to streamline construction processes but also to ensure the reusability and collaboration of Digital Twin models across diverse scenarios. This work proposes Digital twin model building strategy (DTBS), deriving four key processes for constructing digital twin models from the perspective of application scenarios: based on physical entities, twin service requirements, physical data, and entity requirements. Subsequently, by defining the application scenarios and employing suitable strategies, the building of digital twin models is accomplished. The DTBS serves as the core strategy for the building of digital twin models, guiding the complete construction process of digital twin models. The DTBS aims to achieve three objectives: comprehensiveness (encompassing all stages of digital twin model building), operability (with low thresholds for researchers and practitioners), and scalability (encompassing not just one scenario, but multiple domains). Additionally, through case studies, the effectiveness of the Digital twin model building strategy in practical engineering contexts is expounded upon. This strategy's strength lies in its ability to maintain scalability while also demonstrating comprehensiveness and operability.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 366-379"},"PeriodicalIF":12.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683290","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":"Life cycle assessment and energy characterization of friction surfacing deposition of aluminum alloys","authors":"Simone Amantia ,&nbsp;Kirill Kalashnikov ,&nbsp;Gianni Campatelli ,&nbsp;Livan Fratini ,&nbsp;Giuseppe Ingarao","doi":"10.1016/j.jmsy.2025.03.008","DOIUrl":"10.1016/j.jmsy.2025.03.008","url":null,"abstract":"<div><div>In this work, an experimental investigation of Friction Surfacing Deposition (FSD) using the 2000-series heat-treatable aluminum alloy was performed including the environmental impact characterization of the process. The effect of main controlling process parameters and their interactions on energy demand during the single layer deposition was evaluated. A full Life Cycle Assessment (LCA) analysis was conducted for layer-by-layer deposition and a comparison of FSD with the Cold Metal Transfer Wire Arc Additive Manufacturing (WAAM) was performed for a specific wall-shaped sample production. It was observed that the FSD process is characterized by lower processing energy than WAAM, but also by a much higher amount of material scrap connected to undeposited parts of consumable tools such as a flash. To assess the possibility of reducing the material waste during FSD, the comparative LCA analysis was expanded to study the impact of the deposited layer length. It was shown that the FSD method can be a more environmentally friendly process when the deposition of at least 450-mm-long layer using a unique tool is required.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 356-365"},"PeriodicalIF":12.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-objective collaborative optimization of green disassembly planning and recovery option decision considering the learning effect","authors":"Yibing Li ,&nbsp;Wenxia Zhu ,&nbsp;Jun Guo ,&nbsp;Kaipu Wang ,&nbsp;Liang Gao","doi":"10.1016/j.jmsy.2025.03.011","DOIUrl":"10.1016/j.jmsy.2025.03.011","url":null,"abstract":"<div><div>With the continuous improvement of environmental awareness, the recovery of end-of-life products has received widespread attention. Rational decision-making on the recovery options of product parts is an effective way to achieve environmental goals. Meanwhile, manual disassembly is very important in the recycling process, and the learning effect of workers has a great influence on disassembly. Therefore, a collaborative selective disassembly planning and end-of-life products recovery option decision model considering the learning effect is proposed. The objective is to minimize disassembly time, and carbon emissions and maximize disassembly profit. To obtain a high-quality disassembly scheme, an improved multi-objective genetic algorithm based on Q-learning is proposed. To improve the quality of the initial solution, a three-layer encoding strategy including disassembly sequence, disassembly decision sequence, and recovery option decision sequence is designed. Four search strategies are designed as actions for Q-learning, and the state is constructed based on population fitness. This way can enable the algorithm to dynamically adjust the optimization search strategy during the iterative process. Then, the accuracy and effectiveness of the algorithm are verified by two test cases. Next, the proposed model and algorithm are applied to a real refrigerator disassembly case. The results show that due to the learning effect, the efficiency of the disassembly can be increased by 31.66 %, the cost can be reduced by 30.44 %, and the carbon emissions can be reduced by 30.07 %. In addition, carbon emissions can be reduced by 34.82 % by co-optimizing disassembly planning and recovery option decisions.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 324-343"},"PeriodicalIF":12.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683287","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":"Task allocation and scheduling to enhance human–robot collaboration in production line by synergizing efficiency and fatigue","authors":"Fan Zeng ,&nbsp;Changxiang Fan ,&nbsp;Shouhei Shirafuji ,&nbsp;Yusheng Wang ,&nbsp;Masahiro Nishio ,&nbsp;Jun Ota","doi":"10.1016/j.jmsy.2025.03.006","DOIUrl":"10.1016/j.jmsy.2025.03.006","url":null,"abstract":"<div><div>Introducing robots to assist humans in production lines can reduce human fatigue, but efficiency should also not be overlooked. Therefore, task allocation and scheduling, which determine who performs tasks and when they start and finish, should consider both efficiency and fatigue in human–robot collaboration. Efficiency needs to be maximized while fatigue needs to be minimized, necessitating a compromise solution to balance these conflicting objectives. Task allocation guided by multiple objectives is computationally more complex. Furthermore, the production line, with its numerous components and tasks, typically has a larger search space, especially in scenarios involving multiple humans and robots. This complexity makes it challenging for most current human–robot task allocation methods to effectively address such problems. Thus, a new task allocation and scheduling method to balance efficiency and fatigue is proposed in this paper. It reallocates initial sequential human actions to all the humans and robots, obtains locally optimal solutions by multi-heuristics search with efficiency and fatigue synergized, and a fast-converging greedy search is then employed to refine these locally optimal solutions to approach the global optimum. What is more, the proposed method was applied to a laboratory-constructed production line and extended to more complex scenarios involving four different setups, as well as the scalability experiment, demonstrating superior task allocation and scheduling capabilities in balancing the efficiency and fatigue of complex scenarios.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 309-323"},"PeriodicalIF":12.2,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CAM as a Service with dynamic toolpath generation ability for process optimization in STEP-NC compliant CNC machining","authors":"Kaiyao Zhang,&nbsp;Wenlei Xiao,&nbsp;Xiangming Fan,&nbsp;Gang Zhao","doi":"10.1016/j.jmsy.2025.03.004","DOIUrl":"10.1016/j.jmsy.2025.03.004","url":null,"abstract":"<div><div>The next generation of STEP-NC technology needs to achieve more intelligent process optimization. Currently, the calculation method of toolpath length in process optimization algorithms hinders the flexibility and adaptability of algorithm applications. Process optimization needs to generate toolpath based on dynamic process parameter combinations automatically. To address this issue, this paper deploys CAM on the cloud based on the STEP-NC edge-cloud collaboration system, enabling the automatic generation of toolpath through interaction with the process parameter optimization process. Building on this, a non-dominated sorting genetic algorithm III with CAM as a service (NSGAIII-CaaS) for process optimization is proposed. Additionally, a process optimization method for machining feature elements is introduced. Finally, the proposed method is applied to optimize process parameters for three features of a typical part from COMAC, targeting machining cost and machining time. The feasibility of the proposed method’s application in manufacturing enterprises is verified. Using the optimized process parameters for machining features, the cost is reduced by over 70%, efficiency is improved by 70%, and redundant toolpath in machining features are optimized.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"80 ","pages":"Pages 294-308"},"PeriodicalIF":12.2,"publicationDate":"2025-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143683140","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}