Journal of Manufacturing Systems最新文献_第4页

A spatio-temporal parallel ensemble learning approach for operation situation prediction in discrete manufacturing workshop 离散制造车间运行态势预测的时空并行集成学习方法

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-27 DOI: 10.1016/j.jmsy.2025.12.022

Sai Geng , Yu Guo , Weiwei Qian , Weiguang Fang , Shengbo Wang , Shaohua Huang , Xiaoyu Hou

In the complex and dynamic discrete manufacturing environment, accurate prediction of the workshop operation situation (WOS) is crucial to ensure on-time delivery of orders. However, the spatio-temporal (ST) coupling characteristics of the manufacturing process, dynamic fluctuations of workshop performance, and varying contributions of samples to the prediction model make WOS prediction more challenging. To address these issues, this paper proposes a ST parallel ensemble learning approach for WOS prediction. Specifically, based on workshop production data, a temporal data model and a dynamic graph model are constructed to comprehensively characterize the ST characteristics of the production process. Subsequently, this paper proposes a ST parallel ensemble learning method, named Adaboost-GLT, which integrates three ST weak learners (GCN, LSTM, and TGCN) to effectively capture the ST characteristics. Furthermore, a dynamic optimal selection mechanism is designed to adaptively select the best-performing weak learner at each stage, enabling the prediction method to evolve synchronously with the dynamic changes of the manufacturing process. Additionally, a sample weight updating strategy that takes into account sample timeliness and prediction error is introduced to improve the rationality of Adaboost-GLT's attention allocation to samples during training. Finally, the performance of Adaboost-GLT is experimentally validated on real workshop production datasets. The experimental results show that Adaboost-GLT can fully exploit the ST characteristics, effectively cope with the dynamic fluctuations of workshop performance, and thereby achieve high-precision prediction of WOS.

在复杂动态的离散制造环境下，准确预测车间运行状况是保证订单准时交货的关键。然而，制造过程的时空耦合特征、车间绩效的动态波动以及样本对预测模型的不同贡献使得WOS预测更具挑战性。为了解决这些问题，本文提出了一种用于WOS预测的ST并行集成学习方法。具体而言，以车间生产数据为基础，构建时序数据模型和动态图形模型，全面表征生产过程的ST特征。随后，本文提出了一种ST并行集成学习方法Adaboost-GLT，该方法集成了三个ST弱学习器（GCN、LSTM和TGCN）来有效捕获ST特征。此外，设计了动态最优选择机制，在每个阶段自适应地选择表现最好的弱学习者，使预测方法能够与制造过程的动态变化同步进化。此外，引入了考虑样本时效性和预测误差的样本权值更新策略，提高了Adaboost-GLT在训练过程中对样本的关注分配的合理性。最后，在实际车间生产数据集上对Adaboost-GLT的性能进行了实验验证。实验结果表明，Adaboost-GLT能够充分利用ST特性，有效应对车间性能的动态波动，从而实现对WOS的高精度预测。

{"title":"A spatio-temporal parallel ensemble learning approach for operation situation prediction in discrete manufacturing workshop","authors":"Sai Geng , Yu Guo , Weiwei Qian , Weiguang Fang , Shengbo Wang , Shaohua Huang , Xiaoyu Hou","doi":"10.1016/j.jmsy.2025.12.022","DOIUrl":"10.1016/j.jmsy.2025.12.022","url":null,"abstract":"<div><div>In the complex and dynamic discrete manufacturing environment, accurate prediction of the workshop operation situation (WOS) is crucial to ensure on-time delivery of orders. However, the spatio-temporal (ST) coupling characteristics of the manufacturing process, dynamic fluctuations of workshop performance, and varying contributions of samples to the prediction model make WOS prediction more challenging. To address these issues, this paper proposes a ST parallel ensemble learning approach for WOS prediction. Specifically, based on workshop production data, a temporal data model and a dynamic graph model are constructed to comprehensively characterize the ST characteristics of the production process. Subsequently, this paper proposes a ST parallel ensemble learning method, named Adaboost-GLT, which integrates three ST weak learners (GCN, LSTM, and TGCN) to effectively capture the ST characteristics. Furthermore, a dynamic optimal selection mechanism is designed to adaptively select the best-performing weak learner at each stage, enabling the prediction method to evolve synchronously with the dynamic changes of the manufacturing process. Additionally, a sample weight updating strategy that takes into account sample timeliness and prediction error is introduced to improve the rationality of Adaboost-GLT's attention allocation to samples during training. Finally, the performance of Adaboost-GLT is experimentally validated on real workshop production datasets. The experimental results show that Adaboost-GLT can fully exploit the ST characteristics, effectively cope with the dynamic fluctuations of workshop performance, and thereby achieve high-precision prediction of WOS.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 584-600"},"PeriodicalIF":14.2,"publicationDate":"2025-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836450","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}

引用次数: 0

Safe reinforcement learning with online filtering for fatigue-predictive human–robot task planning and allocation in production 基于在线过滤的疲劳预测人机任务规划与分配的安全强化学习

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-26 DOI: 10.1016/j.jmsy.2025.12.019

Jintao Xue, Xiao Li, Nianmin Zhang

Human–robot collaborative manufacturing, a core aspect of Industry 5.0, emphasizes ergonomics to enhance worker well-being. This paper addresses the dynamic human–robot task planning and allocation (HRTPA) problem, which involves determining when to perform tasks and who should execute them to maximize efficiency while ensuring workers’ physical fatigue remains within safe limits. The inclusion of fatigue constraints, combined with production dynamics, significantly increases the complexity of the HRTPA problem. Traditional fatigue–recovery models in HRTPA often rely on static, predefined hyperparameters. However, in practice, human fatigue sensitivity varies daily due to factors such as changed work conditions and insufficient sleep. To better capture this uncertainty, we treat fatigue-related parameters as inaccurate and estimate them online based on observed fatigue progression during production. To address these challenges, we propose PF-CD3Q, a safe reinforcement learning (safe RL) approach that integrates the particle filter with constrained dueling double deep Q-learning for real-time fatigue-predictive HRTPA. Specifically, we first develop PF-based estimators to track human fatigue and update fatigue model parameters in real-time. These estimators are then integrated into CD3Q by making task-level fatigue predictions during decision-making and excluding tasks that exceed fatigue limits, thereby constraining the action space and formulating the problem as a constrained Markov decision process (CMDP). Experimental results demonstrate that our PF-based estimators achieve high prediction accuracy and strong noise robustness, and that PF-CD3Q outperforms other algorithms across multiple performance metrics, significantly reducing the occurrence of overwork and adapting to unseen fatigue constraints after training. These findings validate the effectiveness of our approach under complex and dynamic production conditions, supporting both human well-being and the development of a more sustainable and efficient manufacturing paradigm.

人机协同制造是工业5.0的一个核心方面，强调人体工程学，以提高工人的福祉。本文解决了动态人机任务规划和分配（HRTPA）问题，该问题涉及确定何时执行任务以及谁应该执行任务以最大化效率，同时确保工人的身体疲劳保持在安全范围内。考虑到疲劳约束和生产动态，HRTPA问题的复杂性大大增加。HRTPA中的传统疲劳恢复模型通常依赖于静态的、预定义的超参数。然而，在实践中，由于工作条件的变化和睡眠不足等因素，人类的疲劳敏感性每天都在变化。为了更好地捕捉这种不确定性，我们将疲劳相关参数视为不准确的，并根据生产过程中观察到的疲劳进展在线估计它们。为了解决这些挑战，我们提出了PF-CD3Q，这是一种安全强化学习（safe RL）方法，将粒子滤波器与约束决斗双深度q学习集成在一起，用于实时疲劳预测HRTPA。具体来说，我们首先开发了基于pf的估计器来跟踪人体疲劳并实时更新疲劳模型参数。然后，通过在决策过程中进行任务级疲劳预测并排除超出疲劳限制的任务，将这些估计器集成到CD3Q中，从而约束行动空间并将问题表述为约束马尔可夫决策过程（CMDP）。实验结果表明，基于pf的估计器具有较高的预测精度和较强的噪声鲁棒性，并且PF-CD3Q在多个性能指标上优于其他算法，显著减少了过度工作的发生，并在训练后适应了看不见的疲劳约束。这些发现验证了我们的方法在复杂和动态生产条件下的有效性，既支持人类福祉，又支持更可持续、更高效的制造范式的发展。

{"title":"Safe reinforcement learning with online filtering for fatigue-predictive human–robot task planning and allocation in production","authors":"Jintao Xue, Xiao Li, Nianmin Zhang","doi":"10.1016/j.jmsy.2025.12.019","DOIUrl":"10.1016/j.jmsy.2025.12.019","url":null,"abstract":"<div><div>Human–robot collaborative manufacturing, a core aspect of Industry 5.0, emphasizes ergonomics to enhance worker well-being. This paper addresses the dynamic human–robot task planning and allocation (HRTPA) problem, which involves determining when to perform tasks and who should execute them to maximize efficiency while ensuring workers’ physical fatigue remains within safe limits. The inclusion of fatigue constraints, combined with production dynamics, significantly increases the complexity of the HRTPA problem. Traditional fatigue–recovery models in HRTPA often rely on static, predefined hyperparameters. However, in practice, human fatigue sensitivity varies daily due to factors such as changed work conditions and insufficient sleep. To better capture this uncertainty, we treat fatigue-related parameters as inaccurate and estimate them online based on observed fatigue progression during production. To address these challenges, we propose PF-CD3Q, a safe reinforcement learning (safe RL) approach that integrates the <strong>p</strong>article <strong>f</strong>ilter with <strong>c</strong>onstrained <strong>d</strong>ueling <strong>d</strong>ouble <strong>d</strong>eep <strong>Q</strong>-learning for real-time fatigue-predictive HRTPA. Specifically, we first develop PF-based estimators to track human fatigue and update fatigue model parameters in real-time. These estimators are then integrated into CD3Q by making task-level fatigue predictions during decision-making and excluding tasks that exceed fatigue limits, thereby constraining the action space and formulating the problem as a constrained Markov decision process (CMDP). Experimental results demonstrate that our PF-based estimators achieve high prediction accuracy and strong noise robustness, and that PF-CD3Q outperforms other algorithms across multiple performance metrics, significantly reducing the occurrence of overwork and adapting to unseen fatigue constraints after training. These findings validate the effectiveness of our approach under complex and dynamic production conditions, supporting both human well-being and the development of a more sustainable and efficient manufacturing paradigm.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 561-583"},"PeriodicalIF":14.2,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836451","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}

引用次数: 0

Predicting aircraft assembly gaps considering structural deformation: A CGAN-based surrogate modeling approach 考虑结构变形的飞机装配间隙预测：基于cgan的代理建模方法

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-24 DOI: 10.1016/j.jmsy.2025.12.018

Haorui Sun , Yifan Zhang , Linbei Jiang , Shouguo Zheng , Qing Wang , Yinglin Ke

The precise assembly of aircraft structures remains a critical challenge in aerospace manufacturing, as assembly gaps are a primary factor undermining precision. This study proposes a surrogate model for the rapid and accurate prediction of assembly gaps in the presence of structural deformation. The methodology begins with creating a comprehensive assembly gap dataset that integrates both geometric deviations and structural deformations through assembly modeling and automated workflow. Building upon this dataset, an enhanced PointNet+ + (PNP) network is employed to extract and fuse multi-source assembly features from part shape point clouds and tooling movements. These fused features are then integrated with a generation network based on the Conditional Generative Adversarial Network (CGAN) architecture, reformulating gap prediction as a conditional generation task. The innovative integration of the two networks realizes an end-to-end pipeline, from initial assembly feature extraction to final assembly gap prediction. A representative wing-box structure was employed as a case study to validate the approach. The trained model efficiently predicts gap fields directly from multi-source assembly information. Experimental results demonstrate that the proposed model achieves prediction accuracy comparable to virtual assembly methods while significantly enhancing computational efficiency. These findings underscore the model’s efficacy, positioning it as a valuable tool for rapidly predicting gaps in aircraft assembly.

飞机结构的精确装配仍然是航空航天制造中的一个关键挑战，因为装配间隙是影响精度的主要因素。本研究提出了一种替代模型，用于快速准确地预测存在结构变形的装配间隙。该方法首先创建一个全面的装配间隙数据集，该数据集通过装配建模和自动化工作流程集成了几何偏差和结构变形。在此数据集的基础上，采用增强的PointNet+ + (PNP)网络从零件形状点云和工装运动中提取和融合多源装配特征。然后将这些融合的特征与基于条件生成对抗网络（CGAN）架构的生成网络集成，将间隙预测重新定义为条件生成任务。两种网络的创新集成实现了从初始装配特征提取到最终装配间隙预测的端到端管道。以典型翼盒结构为例，对该方法进行了验证。训练后的模型可以直接从多源装配信息中有效地预测间隙场。实验结果表明，该模型的预测精度与虚拟装配方法相当，同时显著提高了计算效率。这些发现强调了该模型的有效性，将其定位为快速预测飞机装配间隙的有价值工具。

{"title":"Predicting aircraft assembly gaps considering structural deformation: A CGAN-based surrogate modeling approach","authors":"Haorui Sun , Yifan Zhang , Linbei Jiang , Shouguo Zheng , Qing Wang , Yinglin Ke","doi":"10.1016/j.jmsy.2025.12.018","DOIUrl":"10.1016/j.jmsy.2025.12.018","url":null,"abstract":"<div><div>The precise assembly of aircraft structures remains a critical challenge in aerospace manufacturing, as assembly gaps are a primary factor undermining precision. This study proposes a surrogate model for the rapid and accurate prediction of assembly gaps in the presence of structural deformation. The methodology begins with creating a comprehensive assembly gap dataset that integrates both geometric deviations and structural deformations through assembly modeling and automated workflow. Building upon this dataset, an enhanced PointNet+ + (PNP) network is employed to extract and fuse multi-source assembly features from part shape point clouds and tooling movements. These fused features are then integrated with a generation network based on the Conditional Generative Adversarial Network (CGAN) architecture, reformulating gap prediction as a conditional generation task. The innovative integration of the two networks realizes an end-to-end pipeline, from initial assembly feature extraction to final assembly gap prediction. A representative wing-box structure was employed as a case study to validate the approach. The trained model efficiently predicts gap fields directly from multi-source assembly information. Experimental results demonstrate that the proposed model achieves prediction accuracy comparable to virtual assembly methods while significantly enhancing computational efficiency. These findings underscore the model’s efficacy, positioning it as a valuable tool for rapidly predicting gaps in aircraft assembly.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 518-540"},"PeriodicalIF":14.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836453","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}

引用次数: 0

Incorporating scheduling autonomy of workers into flexible job shop scheduling: Learning and balancing decentralized task sequencing decisions with overall scheduling performance 将工人的调度自主权纳入灵活的作业车间调度：学习和平衡分散的任务排序决策与总体调度性能

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-24 DOI: 10.1016/j.jmsy.2025.12.020

Jan-Phillip Herrmann , Sven Tackenberg , Tharsika Pakeerathan Srirajan , Verena Nitsch

In manufacturing systems with a job shop organization, queues between workstations create an intermittent process flow, allowing workers to schedule tasks entering the queue based on their needs and preferences. The resulting scheduling autonomy of individual workers often leads to inefficiencies in the overall production process due to the loss of control. Companies are therefore increasingly using algorithmic scheduling systems to assign task sequences to workers, thereby drastically reducing their autonomy and negatively affecting their job performance and well-being. This paper extends the existing flexible job shop scheduling problem by sequencing preferences (FJSPSP) to incorporate a human-centered perspective by predicting workers’ task sequencing decisions using learning-to-rank (LTR) methods. By learning workers’ individual task sequencing preferences, it becomes possible to predict the processing sequence based on task characteristics. The scheduling algorithm for the FJSPSP presented in the paper incorporates workers’ learned sequencing preferences as constraints. Considering workers’ learned task sequencing decisions, the FJSPSP optimizes only task assignments to maintain workers’ autonomy over task sequences. The contributions of this paper are fourfold, namely, (1) presenting an approach to elicit sequencing decision datasets from workers, (2) demonstrating the successful prediction of humans’ and an actual worker’s task sequencing decisions with LTR, (3) formulating the FJSPSP variant that integrates workers’ sequencing preferences as constraints and proving its effectiveness in a simulation study, and (4) consolidating these steps into an explainable artificial intelligence (XAI)- and LTR-enabled sociotechnical system design framework. The paper closes with a discussion of the overall methodology and future research perspectives.

在具有作业车间组织的制造系统中，工作站之间的队列创建了一个间歇的流程流，允许工人根据自己的需要和偏好安排进入队列的任务。由此产生的单个工人的调度自主权往往导致由于失去控制而导致整个生产过程效率低下。因此，公司越来越多地使用算法调度系统为员工分配任务序列，从而大大降低了他们的自主权，并对他们的工作表现和幸福感产生了负面影响。本文将基于排序偏好（FJSPSP）的柔性作业车间调度问题扩展到以人为中心的视角，利用学习排序（LTR）方法预测工人的任务排序决策。通过了解工人的个体任务排序偏好，可以根据任务特征预测加工顺序。本文提出的FJSPSP调度算法将工人学习到的排序偏好作为约束。考虑到工人的学习任务排序决策，FJSPSP仅优化任务分配，以保持工人对任务序列的自主性。本文的贡献有四个方面，即(1)提出了一种从工人那里获取测序决策数据集的方法，(2)展示了用LTR成功预测人类和实际工人的任务测序决策，(3)制定了将工人的测序偏好作为约束的FJSPSP变体，并在模拟研究中证明了其有效性。(4)将这些步骤整合到可解释的人工智能（XAI）和ltr支持的社会技术系统设计框架中。论文最后讨论了总体方法和未来的研究前景。

{"title":"Incorporating scheduling autonomy of workers into flexible job shop scheduling: Learning and balancing decentralized task sequencing decisions with overall scheduling performance","authors":"Jan-Phillip Herrmann , Sven Tackenberg , Tharsika Pakeerathan Srirajan , Verena Nitsch","doi":"10.1016/j.jmsy.2025.12.020","DOIUrl":"10.1016/j.jmsy.2025.12.020","url":null,"abstract":"<div><div>In manufacturing systems with a job shop organization, queues between workstations create an intermittent process flow, allowing workers to schedule tasks entering the queue based on their needs and preferences. The resulting scheduling autonomy of individual workers often leads to inefficiencies in the overall production process due to the loss of control. Companies are therefore increasingly using algorithmic scheduling systems to assign task sequences to workers, thereby drastically reducing their autonomy and negatively affecting their job performance and well-being. This paper extends the existing flexible job shop scheduling problem by sequencing preferences (FJSPSP) to incorporate a human-centered perspective by predicting workers’ task sequencing decisions using learning-to-rank (LTR) methods. By learning workers’ individual task sequencing preferences, it becomes possible to predict the processing sequence based on task characteristics. The scheduling algorithm for the FJSPSP presented in the paper incorporates workers’ learned sequencing preferences as constraints. Considering workers’ learned task sequencing decisions, the FJSPSP optimizes only task assignments to maintain workers’ autonomy over task sequences. The contributions of this paper are fourfold, namely, (1) presenting an approach to elicit sequencing decision datasets from workers, (2) demonstrating the successful prediction of humans’ and an actual worker’s task sequencing decisions with LTR, (3) formulating the FJSPSP variant that integrates workers’ sequencing preferences as constraints and proving its effectiveness in a simulation study, and (4) consolidating these steps into an explainable artificial intelligence (XAI)- and LTR-enabled sociotechnical system design framework. The paper closes with a discussion of the overall methodology and future research perspectives.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 541-560"},"PeriodicalIF":14.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836452","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}

引用次数: 0

Semi-supervised and unsupervised learning for health indicator extraction from guided waves in aerospace composite structures 航空复合材料结构导波健康指标提取的半监督和无监督学习

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-19 DOI: 10.1016/j.jmsy.2025.12.014

James Josep Perry , Pablo Garcia-Conde Ortiz , George Konstantinou , Cornelie Vergouwen, Edlyn Santha Kumaran, Morteza Moradi

Health indicators (HIs) are central to diagnosing and prognosing the condition of aerospace composite structures, enabling efficient maintenance and operational safety. However, extracting reliable HIs remains challenging due to variability in material properties, stochastic damage evolution, and diverse damage modes. Manufacturing defects (e.g. disbonds) and in-service incidents (e.g. bird strikes) further complicate this process. This study presents a comprehensive data-driven framework that learns HIs via two learning approaches integrated with multi-domain signal processing. Because ground-truth HIs are unavailable, a semi-supervised and an unsupervised approach are proposed: (i) a diversity deep semi-supervised anomaly detection (Diversity-DeepSAD) approach augmented with continuous auxiliary labels used as hypothetical damage proxies, which overcomes the limitation of prior binary labels and enables modelling of intermediate degradation, and (ii) a degradation-trend-constrained variational autoencoder (DTC-VAE), in which the monotonicity criterion is embedded via an explicit trend constraint. Guided waves with multiple excitation frequencies are used to monitor single-stiffener composite structures under fatigue loading. Time, frequency, and time–frequency representations are explored, and per-frequency HIs are fused via unsupervised ensemble learning to mitigate frequency dependence and reduce variance. Using fast Fourier transform features, the models achieved fitness scores of 81.6% (Diversity-DeepSAD) and 92.3% (DTC-VAE), indicating improved monotonicity and consistency over existing baselines. The proposed history-independent framework, supported by prognostic metrics–guided Bayesian optimisation and excitation frequency-agnostic HI fusion, enables the estimation of more robust HIs for aeronautical composite structures.

健康指标（HIs）是诊断和预测航空航天复合材料结构状况、实现高效维护和操作安全的核心。然而，由于材料特性的可变性、随机损伤演变和不同的损伤模式，提取可靠的HIs仍然具有挑战性。制造缺陷（如脱销）和在役事故（如鸟撞）使这一过程进一步复杂化。本研究提出了一个全面的数据驱动框架，该框架通过两种学习方法与多域信号处理相结合来学习HIs。由于真实HIs不可用，提出了半监督和无监督的方法：(i)多样性深度半监督异常检测（diversity - deepsad）方法增强了连续辅助标签作为假设损伤代理，克服了先前二元标签的限制，并能够对中间退化进行建模；（ii）退化趋势约束的变分自编码器（DTC-VAE），其中单调性标准通过明确的趋势约束嵌入。采用多激励频率导波对单加筋复合材料结构进行疲劳监测。研究了时间、频率和时频表示，并通过无监督集成学习将每频率HIs融合，以减轻频率依赖性并减少方差。使用快速傅里叶变换特征，模型的适应度得分为81.6% （Diversity-DeepSAD）和92.3% (DTC-VAE)，表明与现有基线相比，模型的单调性和一致性得到了改善。所提出的历史无关框架，由预测指标引导的贝叶斯优化和激励频率不可知的HI融合支持，能够对航空复合材料结构进行更稳健的HI估计。

{"title":"Semi-supervised and unsupervised learning for health indicator extraction from guided waves in aerospace composite structures","authors":"James Josep Perry , Pablo Garcia-Conde Ortiz , George Konstantinou , Cornelie Vergouwen, Edlyn Santha Kumaran, Morteza Moradi","doi":"10.1016/j.jmsy.2025.12.014","DOIUrl":"10.1016/j.jmsy.2025.12.014","url":null,"abstract":"<div><div>Health indicators (HIs) are central to diagnosing and prognosing the condition of aerospace composite structures, enabling efficient maintenance and operational safety. However, extracting reliable HIs remains challenging due to variability in material properties, stochastic damage evolution, and diverse damage modes. Manufacturing defects (e.g. disbonds) and in-service incidents (e.g. bird strikes) further complicate this process. This study presents a comprehensive data-driven framework that learns HIs via two learning approaches integrated with multi-domain signal processing. Because ground-truth HIs are unavailable, a semi-supervised and an unsupervised approach are proposed: (i) a diversity deep semi-supervised anomaly detection (Diversity-DeepSAD) approach augmented with continuous auxiliary labels used as hypothetical damage proxies, which overcomes the limitation of prior binary labels and enables modelling of intermediate degradation, and (ii) a degradation-trend-constrained variational autoencoder (DTC-VAE), in which the monotonicity criterion is embedded via an explicit trend constraint. Guided waves with multiple excitation frequencies are used to monitor single-stiffener composite structures under fatigue loading. Time, frequency, and time–frequency representations are explored, and per-frequency HIs are fused via unsupervised ensemble learning to mitigate frequency dependence and reduce variance. Using fast Fourier transform features, the models achieved fitness scores of 81.6% (Diversity-DeepSAD) and 92.3% (DTC-VAE), indicating improved monotonicity and consistency over existing baselines. The proposed history-independent framework, supported by prognostic metrics–guided Bayesian optimisation and excitation frequency-agnostic HI fusion, enables the estimation of more robust HIs for aeronautical composite structures.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 468-492"},"PeriodicalIF":14.2,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786670","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}

引用次数: 0

Industrial Value Symbiont in the context of Industrial Symbiosis: Retrospect and prospect 产业共生背景下的产业价值共生：回顾与展望

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-19 DOI: 10.1016/j.jmsy.2025.12.003

Weinan Sha , Xinguo Ming , Zhihua Chen , Xianyu Zhang , Jiapeng You

Industrial Symbiosis (IS) is regarded as a key enabler of the Circular Economy; its objective is to establish closed resource cycles and promote sustainable production through intensive inter-industrial exchanges of energy, materials, water, and by-products/waste. However, there exists a lack of value-oriented propositions in IS practices within industrial parks before realizing the circular economy vision, and a significant gap remains between the existing IS production system and the evolutionary prospects of a mature industrial ecosystem. To provide prospective insights for the next IS evolution, in this paper, a very first discussion of the Industrial Value Symbiont (IVS) is proposed by retrospecting its ongoing evolutionary paradigm. Then, this paper analyzes several frontier concepts and definitions related to IVS, and presents a proper connotation. To refine its intrinsic constituent diversification and delineate its operational essence, the conceptual framework of IVS is further constructed. Finally, pathways for implementing IVS towards IS 5.0, together with related enablers, digital initiatives, and potential strategies, are discussed. Barriers, challenges, and future research directions of IVS are concluded, respectively. We expect that this work may serve as a cornerstone resource for advancing the evolution and development of IS, offering guidance for value-driven IS in an underexplored research domain.

工业共生（IS）被认为是循环经济的关键推动者；其目标是建立封闭的资源循环，并通过工业间能源、材料、水和副产品/废物的密集交换促进可持续生产。然而，在实现循环经济愿景之前，工业园区内的信息系统实践缺乏以价值为导向的主张，现有的信息系统生产系统与成熟的产业生态系统的演进前景之间存在较大差距。为了提供对下一个IS进化的前瞻性见解，本文通过回顾其正在进行的进化范式，提出了对工业价值共生体（IVS）的第一次讨论。然后，分析了与IVS相关的几个前沿概念和定义，提出了IVS的正确内涵。为了细化其内在成分的多元化，勾画其运作本质，进一步构建了IVS的概念框架。最后，讨论了实现IVS到IS 5.0的途径，以及相关的推动因素、数字倡议和潜在战略。总结了IVS的障碍、挑战和未来的研究方向。我们期望这项工作可以作为推动IS进化和发展的基石资源，在一个尚未开发的研究领域为价值驱动的IS提供指导。

{"title":"Industrial Value Symbiont in the context of Industrial Symbiosis: Retrospect and prospect","authors":"Weinan Sha , Xinguo Ming , Zhihua Chen , Xianyu Zhang , Jiapeng You","doi":"10.1016/j.jmsy.2025.12.003","DOIUrl":"10.1016/j.jmsy.2025.12.003","url":null,"abstract":"<div><div>Industrial Symbiosis (IS) is regarded as a key enabler of the Circular Economy; its objective is to establish closed resource cycles and promote sustainable production through intensive inter-industrial exchanges of energy, materials, water, and by-products/waste. However, there exists a lack of value-oriented propositions in IS practices within industrial parks before realizing the circular economy vision, and a significant gap remains between the existing IS production system and the evolutionary prospects of a mature industrial ecosystem. To provide prospective insights for the next IS evolution, in this paper, a very first discussion of the Industrial Value Symbiont (IVS) is proposed by retrospecting its ongoing evolutionary paradigm. Then, this paper analyzes several frontier concepts and definitions related to IVS, and presents a proper connotation. To refine its intrinsic constituent diversification and delineate its operational essence, the conceptual framework of IVS is further constructed. Finally, pathways for implementing IVS towards IS 5.0, together with related enablers, digital initiatives, and potential strategies, are discussed. Barriers, challenges, and future research directions of IVS are concluded, respectively. We expect that this work may serve as a cornerstone resource for advancing the evolution and development of IS, offering guidance for value-driven IS in an underexplored research domain.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 493-517"},"PeriodicalIF":14.2,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786610","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}

引用次数: 0

MCMAS: Causality-driven collaborative optimization in low-carbon industrial parks with large language models-empowered multi-agent systems MCMAS：基于大型语言模型的多智能体系统的低碳工业园区因果关系驱动协同优化

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-19 DOI: 10.1016/j.jmsy.2025.12.015

Tao Wu , Jie Li , Qiang Liu , Jinsong Bao

Low-carbon smart industrial parks face challenges such as the high volatility of renewable energy, highly uncertain load demands, and the pronounced dynamic complexity of multi-energy coupled systems. However, current scheduling systems heavily rely on expert knowledge and statistical correlations, which generally suffer from insufficient accuracy, low efficiency, and poor transparency in decision-making. To address these issues, this paper proposes an energy causality-driven digital twins multi-agent scheduling system, named MCMAS, aimed at achieving high-precision, reliable, and interpretable multi-energy coordinated optimization under complex operating conditions. Initially, a multi-energy causal dynamic model is proposed, which initializes the causal model by integrating conditional mutual information with domain knowledge and applies causal intervention to effectively eliminate spurious correlations, thereby accurately characterizing the high-fidelity causal topology of the system. Subsequently, a causality cognition-driven multi-agent collaborative decision-making mechanism is designed, where a cooperative reward function, integrating local rewards, upstream penalties, and downstream incentives, guides global cooperative strategies of agents, thereby enhancing the economic efficiency and reliability of the scheduling system. Finally, a large language model-driven dual-cross evaluation mechanism is designed, which integrates process mining and counterfactual causal inference to conduct dual-cross validation of scheduling strategies, thereby quantifying confidence levels and enhancing the interpretability of decision-making schemes. Comparative experiments conducted in a representative smart industrial park in Shanghai demonstrate that, compared with benchmark models such as MPC and QMIX-MAS, MCMAS reduces total operating costs by approximately 37.04 %, decreases carbon emissions by 45.19 %, and improves the Sharpe Ratio by 37.3 %. The results indicate that MCMAS can effectively coordinate multi-energy supply and dynamic production loads across different scenarios, reducing operational costs and carbon emissions.

低碳智能产业园面临着可再生能源的高波动性、负载需求的高度不确定性以及多能耦合系统的动态复杂性等挑战。然而，目前的调度系统严重依赖于专家知识和统计相关性，普遍存在准确性不足、效率低下和决策透明度差的问题。针对这些问题，本文提出了一种能量因果驱动的数字孪生多智能体调度系统MCMAS，旨在实现复杂工况下高精度、可靠、可解释的多能量协同优化。首先，提出了一个多能因果动态模型，该模型通过整合条件互信息和领域知识来初始化因果模型，并通过因果干预有效地消除虚假关联，从而准确表征系统的高保真因果拓扑。在此基础上，设计了因果认知驱动的多智能体协同决策机制，通过整合局部奖励、上游惩罚和下游激励的合作奖励函数，指导智能体的全局合作策略，从而提高调度系统的经济效率和可靠性。最后，设计了一种大型语言模型驱动的双交叉评估机制，将过程挖掘和反事实因果推理相结合，对调度策略进行双交叉验证，量化置信水平，增强决策方案的可解释性。在上海某具有代表性的智慧工业园区进行的对比实验表明，与MPC、QMIX-MAS等基准模型相比，MCMAS总运营成本降低约37.04 %，碳排放降低45.19 %，夏普比提高37.3 %。结果表明，MCMAS可以有效协调多种能源供应和不同场景下的动态生产负荷，降低运行成本和碳排放。

{"title":"MCMAS: Causality-driven collaborative optimization in low-carbon industrial parks with large language models-empowered multi-agent systems","authors":"Tao Wu , Jie Li , Qiang Liu , Jinsong Bao","doi":"10.1016/j.jmsy.2025.12.015","DOIUrl":"10.1016/j.jmsy.2025.12.015","url":null,"abstract":"<div><div>Low-carbon smart industrial parks face challenges such as the high volatility of renewable energy, highly uncertain load demands, and the pronounced dynamic complexity of multi-energy coupled systems. However, current scheduling systems heavily rely on expert knowledge and statistical correlations, which generally suffer from insufficient accuracy, low efficiency, and poor transparency in decision-making. To address these issues, this paper proposes an energy causality-driven digital twins multi-agent scheduling system, named MCMAS, aimed at achieving high-precision, reliable, and interpretable multi-energy coordinated optimization under complex operating conditions. Initially, a multi-energy causal dynamic model is proposed, which initializes the causal model by integrating conditional mutual information with domain knowledge and applies causal intervention to effectively eliminate spurious correlations, thereby accurately characterizing the high-fidelity causal topology of the system. Subsequently, a causality cognition-driven multi-agent collaborative decision-making mechanism is designed, where a cooperative reward function, integrating local rewards, upstream penalties, and downstream incentives, guides global cooperative strategies of agents, thereby enhancing the economic efficiency and reliability of the scheduling system. Finally, a large language model-driven dual-cross evaluation mechanism is designed, which integrates process mining and counterfactual causal inference to conduct dual-cross validation of scheduling strategies, thereby quantifying confidence levels and enhancing the interpretability of decision-making schemes. Comparative experiments conducted in a representative smart industrial park in Shanghai demonstrate that, compared with benchmark models such as MPC and QMIX-MAS, MCMAS reduces total operating costs by approximately 37.04 %, decreases carbon emissions by 45.19 %, and improves the Sharpe Ratio by 37.3 %. The results indicate that MCMAS can effectively coordinate multi-energy supply and dynamic production loads across different scenarios, reducing operational costs and carbon emissions.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 451-467"},"PeriodicalIF":14.2,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786671","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}

引用次数: 0

An extended graphical method with reinforcement learning for distributed job shop scheduling problems with transfers 基于强化学习的分布式作业车间调度问题的扩展图解方法

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-18 DOI: 10.1016/j.jmsy.2025.12.007

Lin Huang , Donglin Wang , Shikui Zhao

Modern manufacturing often requires collaboration across multiple factories due to the dispersion of specialized equipment and differences in processing capacity. In such scenarios, jobs must be processed across different sites, and cross-factory transfers significantly impact scheduling performance. This paper studies the distributed job shop scheduling problem with transfers and proposes an Extended Graphical Method with Reinforcement Learning (EGMRL) to effectively address it. The main innovations of EGMRL are fourfold. First, a transfer-zone is introduced into the graphical method, enabling explicit modeling of both intra-factory and inter-factory transportation times. Second, a layered path search algorithm is developed to accelerate path exploration, thereby improving computational efficiency while maintaining accuracy. Third, a Q-learning–based adaptive strategy dynamically guides job deletion and reinsertion according to the inter-factory state, enhancing adaptability across different problem scales. Finally, a tabu search module is integrated as a local improvement strategy to refine factory-level schedules and prevent premature convergence. Comprehensive experiments on 240 extended benchmark instances and a real-world engineering case study demonstrate that EGMRL consistently outperforms four competitive algorithms in terms of solution quality and stability. Furthermore, the results suggest that the extended graphical method provides promising new solution approaches for tackling scheduling problems with other practical constraints, such as worker–machine collaboration and sequence-dependent setup times.

由于专业设备的分散和加工能力的差异，现代制造业通常需要跨多个工厂进行协作。在这种情况下，作业必须跨不同的站点进行处理，并且跨工厂的传输会显著影响调度性能。本文研究了带有传输的分布式作业车间调度问题，提出了一种扩展图形化强化学习方法（EGMRL）来有效地解决该问题。EGMRL的主要创新有四个方面。首先，在图形化方法中引入了一个运输区域，使工厂内和工厂间运输时间的显式建模成为可能。其次，提出了一种分层路径搜索算法，加快路径搜索速度，在保证精度的同时提高了计算效率。第三，基于q学习的自适应策略根据工厂间状态动态引导作业的删除和重新插入，增强了不同问题尺度的适应性。最后，将禁忌搜索模块集成为局部改进策略，以细化工厂级计划并防止过早收敛。在240个扩展基准实例上的综合实验和一个实际工程案例研究表明，EGMRL在解决方案质量和稳定性方面始终优于四种竞争算法。此外，结果表明，扩展图形方法为解决具有其他实际约束的调度问题提供了有希望的新解决方法，例如工人-机器协作和顺序相关的设置时间。

{"title":"An extended graphical method with reinforcement learning for distributed job shop scheduling problems with transfers","authors":"Lin Huang , Donglin Wang , Shikui Zhao","doi":"10.1016/j.jmsy.2025.12.007","DOIUrl":"10.1016/j.jmsy.2025.12.007","url":null,"abstract":"<div><div>Modern manufacturing often requires collaboration across multiple factories due to the dispersion of specialized equipment and differences in processing capacity. In such scenarios, jobs must be processed across different sites, and cross-factory transfers significantly impact scheduling performance. This paper studies the distributed job shop scheduling problem with transfers and proposes an Extended Graphical Method with Reinforcement Learning (EGMRL) to effectively address it. The main innovations of EGMRL are fourfold. First, a transfer-zone is introduced into the graphical method, enabling explicit modeling of both intra-factory and inter-factory transportation times. Second, a layered path search algorithm is developed to accelerate path exploration, thereby improving computational efficiency while maintaining accuracy. Third, a Q-learning–based adaptive strategy dynamically guides job deletion and reinsertion according to the inter-factory state, enhancing adaptability across different problem scales. Finally, a tabu search module is integrated as a local improvement strategy to refine factory-level schedules and prevent premature convergence. Comprehensive experiments on 240 extended benchmark instances and a real-world engineering case study demonstrate that EGMRL consistently outperforms four competitive algorithms in terms of solution quality and stability. Furthermore, the results suggest that the extended graphical method provides promising new solution approaches for tackling scheduling problems with other practical constraints, such as worker–machine collaboration and sequence-dependent setup times.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 426-450"},"PeriodicalIF":14.2,"publicationDate":"2025-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786605","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}

引用次数: 0

Tool wear condition diagnosis using ensemble learning with regularized fusion of domain knowledge and physical information for Ti-6Al-4V milling 基于集成学习的Ti-6Al-4V铣削刀具磨损状态诊断

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-18 DOI: 10.1016/j.jmsy.2025.12.010

Yan Xu , Li Li , Guanghui Lang , Yang Luo , Junhua Zhao , Congbo Li

Tool wear is a critical factor influencing the cost of machining products, especially in the milling of complex Ti-6Al-4V components, where tool wear is significantly accelerated. Therefore, it is essential to develop accurate methods for diagnosing tool wear. This paper proposes an ensemble learning model that integrates domain knowledge and physical information through regularized fusion for the diagnosis of tool wear conditions. Initially, a physics-based model was developed to relate milling forces to the width of tool flank wear (VB) by analyzing the characteristics of milling forces under conditions of progressive tool wear. Following this, a stacking ensemble framework was utilized to combine the selected base models. Domain-specific knowledge, encompassing different stages of tool wear, as well as physical information, including the force-VB relationship, were integrated through the formulation of the loss function. Moreover, a two-stage feature reduction methodology integrating Spearman's rank correlation analysis (Spearman's ρ) with Principal Component Analysis (PCA) was introduced to improve the relevance and compactness of the features. Subsequently, milling experiments were performed on a machining center to assess the effectiveness and practical applicability of the proposed approach.

刀具磨损是影响加工产品成本的关键因素，特别是在复杂Ti-6Al-4V部件的铣削中，刀具磨损明显加速。因此，开发准确的刀具磨损诊断方法至关重要。提出了一种将领域知识和物理信息通过正则化融合相结合的集成学习模型，用于刀具磨损状态的诊断。首先，通过分析刀具渐进磨损条件下铣削力的特征，建立了铣削力与刀面磨损宽度（VB）之间的物理模型。接着，利用一个堆叠集成框架来组合所选的基本模型。特定领域的知识，包括刀具磨损的不同阶段，以及物理信息，包括力- vb关系，通过损失函数的公式集成。此外，引入了一种结合Spearman秩相关分析（Spearman’s ρ）和主成分分析（PCA）的两阶段特征约简方法，以提高特征的相关性和紧凑性。随后，在加工中心进行了铣削实验，以评估该方法的有效性和实用性。

{"title":"Tool wear condition diagnosis using ensemble learning with regularized fusion of domain knowledge and physical information for Ti-6Al-4V milling","authors":"Yan Xu , Li Li , Guanghui Lang , Yang Luo , Junhua Zhao , Congbo Li","doi":"10.1016/j.jmsy.2025.12.010","DOIUrl":"10.1016/j.jmsy.2025.12.010","url":null,"abstract":"<div><div>Tool wear is a critical factor influencing the cost of machining products, especially in the milling of complex Ti-6Al-4V components, where tool wear is significantly accelerated. Therefore, it is essential to develop accurate methods for diagnosing tool wear. This paper proposes an ensemble learning model that integrates domain knowledge and physical information through regularized fusion for the diagnosis of tool wear conditions. Initially, a physics-based model was developed to relate milling forces to the width of tool flank wear (<em>VB</em>) by analyzing the characteristics of milling forces under conditions of progressive tool wear. Following this, a stacking ensemble framework was utilized to combine the selected base models. Domain-specific knowledge, encompassing different stages of tool wear, as well as physical information, including the force-<em>VB</em> relationship, were integrated through the formulation of the loss function. Moreover, a two-stage feature reduction methodology integrating Spearman's rank correlation analysis (Spearman's ρ) with Principal Component Analysis (PCA) was introduced to improve the relevance and compactness of the features. Subsequently, milling experiments were performed on a machining center to assess the effectiveness and practical applicability of the proposed approach.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 383-402"},"PeriodicalIF":14.2,"publicationDate":"2025-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786609","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}

引用次数: 0

Physics-informed multi-task heterogeneous transfer learning network for robust milling tool wear and RUL prediction under anomalous data conditions 基于物理信息的多任务异构迁移学习网络用于铣刀磨损和异常数据条件下的RUL预测

IF 14.2 1区工程技术 Q1 ENGINEERING, INDUSTRIAL

Journal of Manufacturing Systems

Pub Date : 2025-12-18 DOI: 10.1016/j.jmsy.2025.12.011

Jinrui Liu , Pei Wang , Kesong Zhou , Qinfeng Wang , Hui Wang

Milling is widely employed in precision component manufacturing for aerospace and other high-precision industries. Indirect tool wear monitoring via machining signals provides critical references for timely tool replacement in real-world, ensuring production quality and cost efficiency. However, due to practical constraints (e.g., sensor costs), some machining signal channels may be missing, while sensor failures (e.g., poor contact or damage) can lead to abnormal signals in other channels. Consequently, historical models trained on intact data often fail when confronted with such heterogeneous data. To address this issue, this paper proposes a physics-informed multi-task heterogeneous transfer learning network for robust milling tool wear prediction under anomalous data conditions. The framework employs channel-specific autoencoders, jointly pre-trained via inter-channel adaptation on intact historical data, along with physics-informed attention, to homogenize heterogeneous data. Furthermore, a Frank-Wolfe solver based multi-task loss weighting strategy, inter-domain adaptation, and soft/hard physical constraints guide the predictor's learning process, ensuring consistency with prior physical knowledge and enabling parallel prediction of multi-blade wear. Finally, the max predicted wear is utilized to estimate remaining useful life. Comparative studies with state-of-the-art (SOTA) models in related domains validate the superior accuracy and robustness of the proposed method. Extensive experiments on two public datasets further demonstrate that the proposed method achieves high generalizability. Taking PHM2010 dataset as an example, compared to the best-performing comparative model, the MAE of the proposed model tested on normal or heterogeneous data are 13.8 % and 25.8 % lower, respectively. And the degradation rates of MAE, RMSE and R2 are 0.15, 3.12, and 3.86 % points lower, respectively, when tested on simulated channel-failure data.

铣削广泛应用于航空航天和其他高精度行业的精密部件制造。通过加工信号对刀具磨损进行间接监测，为实际生产中及时更换刀具提供了重要依据，保证了生产质量和成本效益。然而，由于实际限制（如传感器成本），一些加工信号通道可能会缺失，而传感器故障（如接触不良或损坏）可能导致其他通道信号异常。因此，在完整数据上训练的历史模型在面对异构数据时往往会失败。为了解决这一问题，本文提出了一种基于物理的多任务异构迁移学习网络，用于异常数据条件下铣刀磨损的鲁棒预测。该框架采用通道特定的自编码器，通过通道间适应对完整的历史数据进行预训练，以及物理信息的关注，以均匀化异构数据。此外，基于Frank-Wolfe求解器的多任务损失加权策略、域间适应和软/硬物理约束指导了预测器的学习过程，确保了与先前物理知识的一致性，并实现了多叶片磨损的并行预测。最后，利用最大预测磨损来估计剩余使用寿命。通过与相关领域最先进的SOTA模型的比较研究，验证了该方法的准确性和鲁棒性。在两个公共数据集上的大量实验进一步证明了该方法具有较高的泛化性。以PHM2010数据集为例，与性能最好的对比模型相比，本文提出的模型在正常数据和异构数据上的MAE分别降低了13.8 %和25.8 %。在模拟信道失效数据上测试，MAE、RMSE和R2的退化率分别降低了0.15、3.12和3.86个点。

{"title":"Physics-informed multi-task heterogeneous transfer learning network for robust milling tool wear and RUL prediction under anomalous data conditions","authors":"Jinrui Liu , Pei Wang , Kesong Zhou , Qinfeng Wang , Hui Wang","doi":"10.1016/j.jmsy.2025.12.011","DOIUrl":"10.1016/j.jmsy.2025.12.011","url":null,"abstract":"<div><div>Milling is widely employed in precision component manufacturing for aerospace and other high-precision industries. Indirect tool wear monitoring via machining signals provides critical references for timely tool replacement in real-world, ensuring production quality and cost efficiency. However, due to practical constraints (e.g., sensor costs), some machining signal channels may be missing, while sensor failures (e.g., poor contact or damage) can lead to abnormal signals in other channels. Consequently, historical models trained on intact data often fail when confronted with such heterogeneous data. To address this issue, this paper proposes a physics-informed multi-task heterogeneous transfer learning network for robust milling tool wear prediction under anomalous data conditions. The framework employs channel-specific autoencoders, jointly pre-trained via inter-channel adaptation on intact historical data, along with physics-informed attention, to homogenize heterogeneous data. Furthermore, a Frank-Wolfe solver based multi-task loss weighting strategy, inter-domain adaptation, and soft/hard physical constraints guide the predictor's learning process, ensuring consistency with prior physical knowledge and enabling parallel prediction of multi-blade wear. Finally, the max predicted wear is utilized to estimate remaining useful life. Comparative studies with state-of-the-art (SOTA) models in related domains validate the superior accuracy and robustness of the proposed method. Extensive experiments on two public datasets further demonstrate that the proposed method achieves high generalizability. Taking PHM2010 dataset as an example, compared to the best-performing comparative model, the MAE of the proposed model tested on normal or heterogeneous data are 13.8 % and 25.8 % lower, respectively. And the degradation rates of MAE, RMSE and R2 are 0.15, 3.12, and 3.86 % points lower, respectively, when tested on simulated channel-failure data.</div></div>","PeriodicalId":16227,"journal":{"name":"Journal of Manufacturing Systems","volume":"84 ","pages":"Pages 403-425"},"PeriodicalIF":14.2,"publicationDate":"2025-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145786672","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}

引用次数: 0