Computers in Industry最新文献_第5页

An industrial defect detection method based on mixed noise synthesis 基于混合噪声合成的工业缺陷检测方法

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-10-02 DOI: 10.1016/j.compind.2025.104388

Aihua Ke , Jian Luo , Yaoxiang Yu , Le Li , Bo Cai

Deep learning-based methods have significantly reduced the cost of traditional manual quality inspection while enhancing accuracy and efficiency in industrial defect detection. As a result, these methods have become a prominent research focus in computer vision for intelligent manufacturing. They are increasingly applied in various production and operational contexts, including automated inspection, intelligent monitoring, and quality control. This paper presents a novel method called mixed noise synthesized defect detection, designed to identify multiple types of defects in industrial products. The proposed method employs a generative adversarial network architecture composed of a defect synthesizer, a defect discriminator, a feature extractor, and a multi-scale patch adaptor. By leveraging the feature extractor and multi-scale adaptor, the method effectively captures normal feature distributions and synthesizes realistic defect features through mixed noise synthesis, thereby significantly reducing reliance on labeled data. In addition, the defect discriminator uses a dual evaluation strategy that combines adversarial loss with Kullback–Leibler divergence to assess input features and quantify defect severity. Comprehensive experiments on benchmark anomaly detection datasets demonstrate that the method achieves high performance, with image-level and pixel-level area under the receiver operating characteristic curve scores of 99.8% and 99.4% for texture categories, and 96.7% and 98.3% for object categories, substantially outperforming state-of-the-art methods. The source code is publicly available at https://github.com/ah-ke/MNS-Defect.git.

基于深度学习的方法大大降低了传统人工质量检测的成本，同时提高了工业缺陷检测的准确性和效率。因此，这些方法已成为智能制造计算机视觉领域的一个突出研究热点。它们越来越多地应用于各种生产和操作环境，包括自动检测、智能监控和质量控制。本文提出了一种新的混合噪声综合缺陷检测方法，用于识别工业产品中多种类型的缺陷。该方法采用由缺陷综合器、缺陷鉴别器、特征提取器和多尺度贴片适配器组成的生成式对抗网络结构。该方法利用特征提取器和多尺度适配器，有效捕获正态特征分布，并通过混合噪声合成合成真实缺陷特征，从而显著降低对标记数据的依赖。此外，缺陷鉴别器采用对抗性损失与Kullback-Leibler散度相结合的双重评估策略来评估输入特征并量化缺陷严重程度。在基准异常检测数据集上进行的综合实验表明，该方法取得了较好的性能，纹理类和对象类的图像级和像素级接收器工作特征曲线下的面积得分分别为99.8%和99.4%，目标类的得分分别为96.7%和98.3%，大大优于现有方法。源代码可在https://github.com/ah-ke/MNS-Defect.git上公开获得。

{"title":"An industrial defect detection method based on mixed noise synthesis","authors":"Aihua Ke , Jian Luo , Yaoxiang Yu , Le Li , Bo Cai","doi":"10.1016/j.compind.2025.104388","DOIUrl":"10.1016/j.compind.2025.104388","url":null,"abstract":"<div><div>Deep learning-based methods have significantly reduced the cost of traditional manual quality inspection while enhancing accuracy and efficiency in industrial defect detection. As a result, these methods have become a prominent research focus in computer vision for intelligent manufacturing. They are increasingly applied in various production and operational contexts, including automated inspection, intelligent monitoring, and quality control. This paper presents a novel method called mixed noise synthesized defect detection, designed to identify multiple types of defects in industrial products. The proposed method employs a generative adversarial network architecture composed of a defect synthesizer, a defect discriminator, a feature extractor, and a multi-scale patch adaptor. By leveraging the feature extractor and multi-scale adaptor, the method effectively captures normal feature distributions and synthesizes realistic defect features through mixed noise synthesis, thereby significantly reducing reliance on labeled data. In addition, the defect discriminator uses a dual evaluation strategy that combines adversarial loss with Kullback–Leibler divergence to assess input features and quantify defect severity. Comprehensive experiments on benchmark anomaly detection datasets demonstrate that the method achieves high performance, with image-level and pixel-level area under the receiver operating characteristic curve scores of 99.8% and 99.4% for texture categories, and 96.7% and 98.3% for object categories, substantially outperforming state-of-the-art methods. The source code is publicly available at <span><span>https://github.com/ah-ke/MNS-Defect.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104388"},"PeriodicalIF":9.1,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145220527","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep learning-powered heating, ventilation, and air conditioning compressor fault diagnosis facing unseen domains and class imbalances 面向未知域和类不平衡的深度学习驱动的供暖、通风和空调压缩机故障诊断

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-30 DOI: 10.1016/j.compind.2025.104386

Hong Wang , Jun Lin , Zijun Zhang

Reliable fault diagnosis of compressors in heating, ventilation, and air conditioning (HVAC) systems is essential for enhancing their service reliability and energy conservation. However, heterogeneous working environments of HVAC compressors pose significant challenges for applying data-driven fault diagnosis methods. Domain generalization techniques have been developed to address data distribution discrepancies in cross-domain fault diagnosis. Yet, most existing methods assume that source domains have equal sizes and balanced class distributions. These assumptions limit their applicability to real-world scenarios that can encounter multiple levels of imbalance in both domain size and health status. Therefore, this work proposes a novel Adaptive Invariant Representation learning-based domain generalization Network (AIRNet) to enable a better HVAC compressor fault diagnosis performance in handling unseen domains and class imbalances. Specifically, AIRNet employs a probabilistic sampling strategy to adaptively extract balanced training samples from source domains, mitigating class imbalances and driving unbiased model learning. Furthermore, AIRNet integrates fault classification, metric learning, and domain adversarial training modules with a tailored data augmentation strategy, jointly enhancing its robustness and generalizability across unseen domains. These components collaborate to establish fault-discriminative and domain-invariant diagnostic boundaries while improving model resistance against unseen data distribution discrepancies. Extensive computational experiments on HVAC compressors demonstrate the superiority of AIRNet over state-of-the-art methods in addressing real-world industrial fault diagnosis challenges. Compared to the best-performing benchmark, AIRNet achieves an average performance gain of 1.11 % in total accuracy and 2.76 % in macro F1 score across all studied tasks. The code is available at https://github.com/ifuturekk/AIRNet.

对暖通空调系统中的压缩机进行可靠的故障诊断，对提高压缩机的运行可靠性和节能效果至关重要。然而，暖通空调压缩机的异构工作环境给数据驱动故障诊断方法的应用带来了巨大挑战。领域泛化技术是为了解决跨领域故障诊断中数据分布差异的问题而发展起来的。然而，大多数现有方法假设源域具有相等的大小和平衡的类分布。这些假设限制了它们对现实世界场景的适用性，这些场景可能会遇到域大小和健康状态的多级不平衡。因此，本研究提出了一种新的基于自适应不变表示学习的领域泛化网络（AIRNet），以在处理未知领域和类失衡时实现更好的HVAC压缩机故障诊断性能。具体来说，AIRNet采用概率抽样策略自适应地从源域提取平衡的训练样本，减轻类不平衡并驱动无偏模型学习。此外，AIRNet将故障分类、度量学习和领域对抗训练模块与定制的数据增强策略集成在一起，共同增强了其在未知领域的鲁棒性和泛化性。这些组件协作建立了错误判别和领域不变的诊断边界，同时提高了模型对看不见的数据分布差异的抵抗力。在暖通空调压缩机上进行的大量计算实验表明，在解决实际工业故障诊断挑战方面，AIRNet优于最先进的方法。与表现最好的基准测试相比，AIRNet在所有研究任务中实现了1.11 %的总准确率和2.76 %的宏观F1分数的平均性能增益。代码可在https://github.com/ifuturekk/AIRNet上获得。

{"title":"Deep learning-powered heating, ventilation, and air conditioning compressor fault diagnosis facing unseen domains and class imbalances","authors":"Hong Wang , Jun Lin , Zijun Zhang","doi":"10.1016/j.compind.2025.104386","DOIUrl":"10.1016/j.compind.2025.104386","url":null,"abstract":"<div><div>Reliable fault diagnosis of compressors in heating, ventilation, and air conditioning (HVAC) systems is essential for enhancing their service reliability and energy conservation. However, heterogeneous working environments of HVAC compressors pose significant challenges for applying data-driven fault diagnosis methods. Domain generalization techniques have been developed to address data distribution discrepancies in cross-domain fault diagnosis. Yet, most existing methods assume that source domains have equal sizes and balanced class distributions. These assumptions limit their applicability to real-world scenarios that can encounter multiple levels of imbalance in both domain size and health status. Therefore, this work proposes a novel Adaptive Invariant Representation learning-based domain generalization Network (AIRNet) to enable a better HVAC compressor fault diagnosis performance in handling unseen domains and class imbalances. Specifically, AIRNet employs a probabilistic sampling strategy to adaptively extract balanced training samples from source domains, mitigating class imbalances and driving unbiased model learning. Furthermore, AIRNet integrates fault classification, metric learning, and domain adversarial training modules with a tailored data augmentation strategy, jointly enhancing its robustness and generalizability across unseen domains. These components collaborate to establish fault-discriminative and domain-invariant diagnostic boundaries while improving model resistance against unseen data distribution discrepancies. Extensive computational experiments on HVAC compressors demonstrate the superiority of AIRNet over state-of-the-art methods in addressing real-world industrial fault diagnosis challenges. Compared to the best-performing benchmark, AIRNet achieves an average performance gain of 1.11 % in total accuracy and 2.76 % in macro F1 score across all studied tasks. The code is available at <span><span>https://github.com/ifuturekk/AIRNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104386"},"PeriodicalIF":9.1,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145220526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

From user-generated content to quality improvement: A multi-granularity analysis of customer satisfaction and attention in new energy vehicles using deep learning 从用户生成内容到质量提升：基于深度学习的新能源汽车客户满意度和关注度的多粒度分析

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-30 DOI: 10.1016/j.compind.2025.104380

Zhaoguang Xu , Yifan Wu , Lin Tang , Shumeng Gui

Understanding customer satisfaction is crucial to improving product quality and ensuring the market competitiveness of new energy vehicles (NEVs). Although user-generated content (UGC)-based analysis offers a cost-effective alternative to traditional customer satisfaction surveys, existing studies have largely overlooked users’ fine-grained needs and rarely translated sentiment insights into actionable guidance for product improvement. To address this, we propose a novel Multi-Aspect Dynamic Knowledge Graph Convolutional Network to extract aspect-level customer perceptions from UGC. The model utilizes a scaled dependency matrix to filter redundant syntactic relations and captures semantic interactions across various aspects. It integrates a sentiment knowledge base with a cross-attention mechanism to enhance sentiment feature extraction. Leveraging the extracted sentiment, we develop a quantitative method to evaluate customer attention and satisfaction across multi-granularity indicators. Experiments on benchmark datasets show that our model outperforms most state-of-the-art methods. A case study of BYD NEVs based on 55,511 sentences from Autohome further validates its superiority, achieving 91.46% Macro-F1 and 91.41% accuracy. Furthermore, by incorporating a customized importance–performance analysis, we pinpoint high-attention aspects with low satisfaction, such as air conditioner and trunk size, which are subsequently integrated into a house of quality measure to support quality improvement. Our analysis further reveals a steady improvement in customer satisfaction across major aspects, despite temporary declines in certain years. We also observe a 14% decline in attention to battery range, alongside a 3.7% increase in vehicle space. These insights can help NEV manufacturers align their product quality improvement efforts with evolving customer expectations.

了解消费者满意度对于提高新能源汽车的产品质量和确保新能源汽车的市场竞争力至关重要。尽管基于用户生成内容（UGC）的分析为传统的客户满意度调查提供了一个具有成本效益的替代方案，但现有的研究在很大程度上忽略了用户的细粒度需求，并且很少将情感见解转化为产品改进的可操作指导。为了解决这个问题，我们提出了一种新的多方面动态知识图卷积网络，从UGC中提取方面级的客户感知。该模型利用缩放依赖矩阵来过滤冗余的句法关系，并捕获跨各个方面的语义交互。它将情感知识库与交叉注意机制相结合，增强了情感特征的提取。利用提取的情感，我们开发了一种定量方法来评估跨多粒度指标的客户关注和满意度。在基准数据集上的实验表明，我们的模型优于大多数最先进的方法。基于汽车之家55,511个句子的比亚迪新能源汽车案例进一步验证了其优势，实现了91.46%的Macro-F1和91.41%的准确率。此外，通过结合定制的重要性-性能分析，我们确定了高关注度和低满意度的方面，例如空调和后备箱大小，这些方面随后被集成到质量测量中，以支持质量改进。我们的分析进一步揭示了客户满意度在主要方面的稳步提高，尽管在某些年份暂时下降。我们还观察到，对电池续航里程的关注下降了14%，而对汽车空间的关注增加了3.7%。这些见解可以帮助新能源汽车制造商将他们的产品质量改进工作与不断变化的客户期望保持一致。

{"title":"From user-generated content to quality improvement: A multi-granularity analysis of customer satisfaction and attention in new energy vehicles using deep learning","authors":"Zhaoguang Xu , Yifan Wu , Lin Tang , Shumeng Gui","doi":"10.1016/j.compind.2025.104380","DOIUrl":"10.1016/j.compind.2025.104380","url":null,"abstract":"<div><div>Understanding customer satisfaction is crucial to improving product quality and ensuring the market competitiveness of new energy vehicles (NEVs). Although user-generated content (UGC)-based analysis offers a cost-effective alternative to traditional customer satisfaction surveys, existing studies have largely overlooked users’ fine-grained needs and rarely translated sentiment insights into actionable guidance for product improvement. To address this, we propose a novel Multi-Aspect Dynamic Knowledge Graph Convolutional Network to extract aspect-level customer perceptions from UGC. The model utilizes a scaled dependency matrix to filter redundant syntactic relations and captures semantic interactions across various aspects. It integrates a sentiment knowledge base with a cross-attention mechanism to enhance sentiment feature extraction. Leveraging the extracted sentiment, we develop a quantitative method to evaluate customer attention and satisfaction across multi-granularity indicators. Experiments on benchmark datasets show that our model outperforms most state-of-the-art methods. A case study of BYD NEVs based on 55,511 sentences from Autohome further validates its superiority, achieving 91.46% Macro-F1 and 91.41% accuracy. Furthermore, by incorporating a customized importance–performance analysis, we pinpoint high-attention aspects with low satisfaction, such as <em>air conditioner</em> and <em>trunk size</em>, which are subsequently integrated into a house of quality measure to support quality improvement. Our analysis further reveals a steady improvement in customer satisfaction across major aspects, despite temporary declines in certain years. We also observe a 14% decline in attention to battery range, alongside a 3.7% increase in vehicle space. These insights can help NEV manufacturers align their product quality improvement efforts with evolving customer expectations.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104380"},"PeriodicalIF":9.1,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Improving interoperability in robot digital twinning for facility management: An industry foundation class-represented RoboAvatar approach 提高设备管理中机器人数字孪生的互操作性：一个行业基础类代表的RoboAvatar方法

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-29 DOI: 10.1016/j.compind.2025.104384

Junjie Chen , Weisheng Lu , Xiang Ji , Yonglin Fu

With its bi-directional information flow, a digital twin offers the potential to enhance predictability and controllability of robots for facility management (FM). The implementation of FM involves frequent robot-building interactions, necessitating information exchanges between a robot digital twin (RDT) and a building information model (BIM). However, such information exchanges are prohibited by the different data formats used by the RDT and BIM. Our recent study has proven the viability of industry foundation class (IFC) in digitally representing robots as Avatars, and seamlessly integrating the resulting RoboAvatars into BIM-based software. Building upon that, this paper explores how the IFC-represented RoboAvatars can be used to improve interoperability of RDTs for FM. A lab experiment was conducted with an indoor trash picking robot. It demonstrates effectiveness of IFC-based RDTs in FM via the freely exchangeable robot-building information. The robot movements can be mirrored with high granularity within a BIM context. Information from BIM can be directly retrieved to trigger robot movements remotely. The research contributes to the field of FM robotics by providing the world’s first methodology to directly develop and deploy RDTs in a mainstream BIM-based environment.

通过双向信息流，数字孪生体提供了增强设备管理（FM）机器人的可预测性和可控性的潜力。FM的实施涉及频繁的机器人与建筑的交互，需要机器人数字孪生体（RDT）和建筑信息模型（BIM）之间的信息交换。但是，由于RDT和BIM使用的数据格式不同，这种信息交换是不允许的。我们最近的研究已经证明了工业基础类（IFC）在数字化表示机器人为阿凡达方面的可行性，并将由此产生的RoboAvatars无缝集成到基于bim的软件中。在此基础上，本文探讨了如何使用ifc代表的RoboAvatars来提高FM的rdt的互操作性。以室内垃圾捡捡机器人为实验对象进行了实验研究。通过自由交换的机器人制造信息，证明了基于ifc的rdt在FM中的有效性。机器人的运动可以在BIM上下文中以高粒度进行镜像。来自BIM的信息可以直接检索，从而远程触发机器人的动作。该研究通过提供世界上第一个在基于bim的主流环境中直接开发和部署rdt的方法，为FM机器人领域做出了贡献。

{"title":"Improving interoperability in robot digital twinning for facility management: An industry foundation class-represented RoboAvatar approach","authors":"Junjie Chen , Weisheng Lu , Xiang Ji , Yonglin Fu","doi":"10.1016/j.compind.2025.104384","DOIUrl":"10.1016/j.compind.2025.104384","url":null,"abstract":"<div><div>With its bi-directional information flow, a digital twin offers the potential to enhance predictability and controllability of robots for facility management (FM). The implementation of FM involves frequent robot-building interactions, necessitating information exchanges between a robot digital twin (RDT) and a building information model (BIM). However, such information exchanges are prohibited by the different data formats used by the RDT and BIM. Our recent study has proven the viability of industry foundation class (IFC) in digitally representing robots as Avatars, and seamlessly integrating the resulting RoboAvatars into BIM-based software. Building upon that, this paper explores how the IFC-represented RoboAvatars can be used to improve interoperability of RDTs for FM. A lab experiment was conducted with an indoor trash picking robot. It demonstrates effectiveness of IFC-based RDTs in FM via the freely exchangeable robot-building information. The robot movements can be mirrored with high granularity within a BIM context. Information from BIM can be directly retrieved to trigger robot movements remotely. The research contributes to the field of FM robotics by providing the world’s first methodology to directly develop and deploy RDTs in a mainstream BIM-based environment.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104384"},"PeriodicalIF":9.1,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Segmentation, correction, and classification of abnormal sensor data in mechanical engineering based on multi-task learning 基于多任务学习的机械工程异常传感器数据分割、校正与分类

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-29 DOI: 10.1016/j.compind.2025.104387

Xirui Chen, Hui Liu

Rolling bearings and hydraulic internal pumps are the two most commonly used fault diagnosis devices in mechanical engineering. However, harsh industrial environments not only harm their health but also the sensors used for monitoring. Abnormal sensor data problems are common in practice and significantly affect data-based fault detection methods. Therefore, this study jointly investigates the anomaly detection of sensor data and the fault detection of engineering components. The related issues are divided into three tasks: classification, correction, and segmentation of abnormal sensor data. A multi-task learning framework based on the teacher-student structure is then proposed to fulfill these tasks in one shot. The designed feature corrector corrects abnormal representations, while the correction attention guides the classifier to focus on the normal parts. A semantic segmentation model is integrated to achieve novel and comprehensive anomaly detection. The proposed multi-task framework is validated using rolling bearing and hydraulic pump datasets. The experimental results show that the jointly trained models outperform those that are trained independently.

滚动轴承和液压内泵是机械工程中最常用的两种故障诊断装置。然而，恶劣的工业环境不仅损害了他们的健康，也损害了用于监测的传感器。传感器数据异常问题在实际应用中很常见，严重影响基于数据的故障检测方法。因此，本研究将传感器数据的异常检测与工程部件的故障检测结合起来进行研究。相关问题分为三个任务：异常传感器数据的分类、校正和分割。在此基础上，提出了一种基于师生结构的多任务学习框架，以一次性完成这些任务。所设计的特征校正器对异常表示进行校正，而校正注意引导分类器关注正常部分。结合语义分割模型，实现新颖、全面的异常检测。使用滚动轴承和液压泵数据集验证了所提出的多任务框架。实验结果表明，联合训练的模型优于独立训练的模型。

{"title":"Segmentation, correction, and classification of abnormal sensor data in mechanical engineering based on multi-task learning","authors":"Xirui Chen, Hui Liu","doi":"10.1016/j.compind.2025.104387","DOIUrl":"10.1016/j.compind.2025.104387","url":null,"abstract":"<div><div>Rolling bearings and hydraulic internal pumps are the two most commonly used fault diagnosis devices in mechanical engineering. However, harsh industrial environments not only harm their health but also the sensors used for monitoring. Abnormal sensor data problems are common in practice and significantly affect data-based fault detection methods. Therefore, this study jointly investigates the anomaly detection of sensor data and the fault detection of engineering components. The related issues are divided into three tasks: classification, correction, and segmentation of abnormal sensor data. A multi-task learning framework based on the teacher-student structure is then proposed to fulfill these tasks in one shot. The designed feature corrector corrects abnormal representations, while the correction attention guides the classifier to focus on the normal parts. A semantic segmentation model is integrated to achieve novel and comprehensive anomaly detection. The proposed multi-task framework is validated using rolling bearing and hydraulic pump datasets. The experimental results show that the jointly trained models outperform those that are trained independently.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104387"},"PeriodicalIF":9.1,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Zero-shot printed circuit board defect detection via optical flow and reconstruction guidance 基于光流和重构制导的零射印刷电路板缺陷检测

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-25 DOI: 10.1016/j.compind.2025.104355

Xinghang Yin , Shuxia Wang , Yue Wang , Peng Wang , Yongxu Liu , Tianle Shen , Hengjie Qiao

Deep learning is widely used in printed circuit board (PCB) defect detection, owing to its excellent performance. Different types and styles of PCBs exist, for application in different fields and scenarios, making it necessary to fine-tune model on unseen PCB types to maintain detection performance. Few-shot learning methods reduce the cost of data collection and annotation, as they require fewer samples. Under ideal circumstances and with standardized electronic components, image differencing techniques can highlight defects by comparing test images with defect-free reference images, making them category-agnostic, generalizable, and highly interpretable. However, they require careful image preprocessing and parameter selection, and fail if the images are misaligned. To address this issue, while preserving the generalizability of image differencing, we propose a method for PCB defect detection by simulating image differencing using a neural network comprising a shared encoder and three decoders for different tasks: (1) The flow decoder outputs an optical flow displacement field to align image pairs and guides the encoder to learn pixel correspondence relationships, (2) The reconstruction decoder guides the encoder to focus on perceiving the discrepancies between images. (3) The mask decoder locates defective areas with significant visual discrepancies between images. We train the network exclusively on synthetic data and then test it on the publicly available datasets, DeepPCB, PCBS, and MVTec AD, achieving results comparable to that of supervised learning with numerous real samples. Ablation experiments demonstrate that optical flow and reconstruction guidance can effectively enhance the robustness of the network.

深度学习以其优异的性能在印刷电路板缺陷检测中得到了广泛的应用。不同类型和风格的PCB存在，用于不同的领域和场景，因此有必要在未见过的PCB类型上微调模型以保持检测性能。Few-shot学习方法减少了数据收集和注释的成本，因为它们需要较少的样本。在理想的情况下，使用标准化的电子元件，图像区分技术可以通过比较测试图像和无缺陷的参考图像来突出缺陷，使它们与类别无关、可推广和高度可解释。然而，它们需要仔细的图像预处理和参数选择，如果图像不对齐就会失败。为了解决这一问题，在保持图像差异的普遍性的同时，我们提出了一种PCB缺陷检测方法，通过使用一个由共享编码器和三个不同任务的解码器组成的神经网络来模拟图像差异：(1)流解码器输出光流位移场对图像对进行对齐，引导编码器学习像素对应关系；(2)重构解码器引导编码器专注于感知图像之间的差异。(3)掩码解码器定位图像间视觉差异较大的缺陷区域。我们专门在合成数据上训练网络，然后在公开可用的数据集，DeepPCB， PCBS和MVTec AD上进行测试，取得了与具有大量真实样本的监督学习相当的结果。烧蚀实验表明，光流和重构引导可以有效地增强网络的鲁棒性。

{"title":"Zero-shot printed circuit board defect detection via optical flow and reconstruction guidance","authors":"Xinghang Yin , Shuxia Wang , Yue Wang , Peng Wang , Yongxu Liu , Tianle Shen , Hengjie Qiao","doi":"10.1016/j.compind.2025.104355","DOIUrl":"10.1016/j.compind.2025.104355","url":null,"abstract":"<div><div>Deep learning is widely used in printed circuit board (PCB) defect detection, owing to its excellent performance. Different types and styles of PCBs exist, for application in different fields and scenarios, making it necessary to fine-tune model on unseen PCB types to maintain detection performance. Few-shot learning methods reduce the cost of data collection and annotation, as they require fewer samples. Under ideal circumstances and with standardized electronic components, image differencing techniques can highlight defects by comparing test images with defect-free reference images, making them category-agnostic, generalizable, and highly interpretable. However, they require careful image preprocessing and parameter selection, and fail if the images are misaligned. To address this issue, while preserving the generalizability of image differencing, we propose a method for PCB defect detection by simulating image differencing using a neural network comprising a shared encoder and three decoders for different tasks: (1) The flow decoder outputs an optical flow displacement field to align image pairs and guides the encoder to learn pixel correspondence relationships, (2) The reconstruction decoder guides the encoder to focus on perceiving the discrepancies between images. (3) The mask decoder locates defective areas with significant visual discrepancies between images. We train the network exclusively on synthetic data and then test it on the publicly available datasets, DeepPCB, PCBS, and MVTec AD, achieving results comparable to that of supervised learning with numerous real samples. Ablation experiments demonstrate that optical flow and reconstruction guidance can effectively enhance the robustness of the network.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104355"},"PeriodicalIF":9.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fault diagnosis technology of aero-engine rotors based on meta-action theory driven by machine learning for reliability improvement 基于机器学习驱动元作用理论的航空发动机转子故障诊断技术

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-25 DOI: 10.1016/j.compind.2025.104381

Yulong Li , Junfa Li , Hui Long , Shutao Wen , Minghui Gu , Hongwei Wang

The intricate structure of electromechanical products presents significant challenges in fault diagnosis, and conventional methods frequently fail to capture the correlation between time-domain and frequency-domain features of fault vibration signals. Moreover, these methods typically rely on extensive training datasets and demonstrate limited generalization capabilities. To overcome these limitations, this paper introduces a fault analysis framework based on the meta-action unit (MAU) to streamline fault diagnosis processes in electromechanical products. An integrated model comprising Fast Fourier Transform (FFT), Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (Bi-GRU), Transformer and Attention mechanisms, which designated as the FFT-CNN-Bi-GRU-Transformer-Attention model, was developed to enhance the extraction and representation of vibration signal features, thereby improving model robustness and accuracy. The methodology involves several sequential processes. Initially, fault signals were collected using the MAU and transformed from the time-domain to frequency-domain via FFT. Subsequently, a CNN was employed to automatically extract salient features from the frequency-domain signals. Bi-GRU was then applied to process these features in both forward and backward directions, thus enriching the expressiveness of the data representation. To facilitate efficient parallel computation, the Transformer mechanism was incorporated to refine the output from the Bi-GRU, while the Attention mechanism was used to capture intricate fault features and patterns, significantly enhancing the model’s diagnostic performance. The proposed method was validated using an aero-engine rotor unit as a test case, achieving an accuracy of 98.16 %. Comparative analyses with conventional fault diagnosis techniques underscore the clear advantages of the proposed method. This method provides a foundation for accurate fault identification and timely maintenance of aero-engine rotors, as well as other electromechanical products with analogous structural characteristics.

机电产品复杂的结构给故障诊断带来了很大的挑战，传统的方法往往无法捕捉故障振动信号的时域和频域特征之间的相关性。此外，这些方法通常依赖于广泛的训练数据集，并且泛化能力有限。为了克服这些局限性，本文引入了一种基于元动作单元（MAU）的故障分析框架，以简化机电产品的故障诊断过程。基于快速傅里叶变换（FFT）、卷积神经网络（CNN）、双向门控循环单元（Bi-GRU）、变压器和注意力机制，建立了FFT-CNN-Bi-GRU-变压器-注意力模型，增强了对振动信号特征的提取和表征，从而提高了模型的鲁棒性和准确性。该方法涉及几个连续的过程。首先利用MAU采集故障信号，通过FFT将故障信号从时域变换到频域。随后，利用CNN从频域信号中自动提取显著特征。然后应用Bi-GRU对这些特征进行正向和反向处理，从而丰富了数据表示的表达性。为了提高并行计算的效率，引入了Transformer机制来细化Bi-GRU的输出，而使用了Attention机制来捕获复杂的故障特征和模式，显著提高了模型的诊断性能。以某航空发动机转子单元为例，对该方法进行了验证，准确率达到98.16 %。与传统故障诊断技术的对比分析表明，该方法具有明显的优越性。该方法为航空发动机转子及其他具有类似结构特征的机电产品的准确故障识别和及时维修提供了基础。

{"title":"Fault diagnosis technology of aero-engine rotors based on meta-action theory driven by machine learning for reliability improvement","authors":"Yulong Li , Junfa Li , Hui Long , Shutao Wen , Minghui Gu , Hongwei Wang","doi":"10.1016/j.compind.2025.104381","DOIUrl":"10.1016/j.compind.2025.104381","url":null,"abstract":"<div><div>The intricate structure of electromechanical products presents significant challenges in fault diagnosis, and conventional methods frequently fail to capture the correlation between time-domain and frequency-domain features of fault vibration signals. Moreover, these methods typically rely on extensive training datasets and demonstrate limited generalization capabilities. To overcome these limitations, this paper introduces a fault analysis framework based on the meta-action unit (MAU) to streamline fault diagnosis processes in electromechanical products. An integrated model comprising Fast Fourier Transform (FFT), Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (Bi-GRU), Transformer and Attention mechanisms, which designated as the FFT-CNN-Bi-GRU-Transformer-Attention model, was developed to enhance the extraction and representation of vibration signal features, thereby improving model robustness and accuracy. The methodology involves several sequential processes. Initially, fault signals were collected using the MAU and transformed from the time-domain to frequency-domain via FFT. Subsequently, a CNN was employed to automatically extract salient features from the frequency-domain signals. Bi-GRU was then applied to process these features in both forward and backward directions, thus enriching the expressiveness of the data representation. To facilitate efficient parallel computation, the Transformer mechanism was incorporated to refine the output from the Bi-GRU, while the Attention mechanism was used to capture intricate fault features and patterns, significantly enhancing the model’s diagnostic performance. The proposed method was validated using an aero-engine rotor unit as a test case, achieving an accuracy of 98.16 %. Comparative analyses with conventional fault diagnosis techniques underscore the clear advantages of the proposed method. This method provides a foundation for accurate fault identification and timely maintenance of aero-engine rotors, as well as other electromechanical products with analogous structural characteristics.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104381"},"PeriodicalIF":9.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Transfer learning of state-based potential games for process optimization in decentralized manufacturing systems 分散制造系统过程优化中基于状态势能博弈的迁移学习

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-25 DOI: 10.1016/j.compind.2025.104376

Steve Yuwono , Dorothea Schwung , Andreas Schwung

This paper presents a novel online transfer learning approach in state-based potential games (TL-SbPGs) for distributed self-optimization in manufacturing systems. The approach targets practical industrial scenarios where knowledge sharing among similar players enhances learning in large-scale and decentralized environments. TL-SbPGs enable players to reuse learned policies from others, which improves learning outcomes and accelerates convergence. To accomplish this goal, we develop transfer learning concepts and similarity criteria for players, which offer two distinct settings: (a) predefined similarities between players and (b) dynamically inferred similarities between players during training. The applicability of the SbPG framework to transfer learning is formally established. Furthermore, we present a method to optimize the timing and weighting of knowledge transfer. Experimental results from a laboratory-scale testbed show that TL-SbPGs improve production efficiency and reduce power consumption compared to vanilla SbPGs.

提出了一种新的基于状态的潜在博弈在线迁移学习方法，用于制造系统的分布式自优化。该方法针对实际的工业场景，其中类似参与者之间的知识共享可以增强大规模和分散环境中的学习。tl - sbpg使玩家能够重用从其他人那里学习到的策略，从而提高学习效果并加速收敛。为了实现这一目标，我们为球员开发了迁移学习概念和相似性标准，它们提供了两种不同的设置：(a)预定义的球员之间的相似性；(b)在训练期间动态推断球员之间的相似性。SbPG框架在迁移学习中的适用性正式确立。此外，我们还提出了一种优化知识转移时间和权重的方法。实验结果表明，与普通sbpg相比，tl - sbpg提高了生产效率，降低了功耗。

引用次数: 0

Enhanced cross-domain signal and physics-based interpretation for fault diagnosis of aircraft brake control valve under limited onboard signal acquisition 有限机载信号采集条件下飞机刹车控制阀故障诊断的增强跨域信号和物理解释

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-25 DOI: 10.1016/j.compind.2025.104378

Lin Lin , Yin Chen , Wenhui He , Song Fu , Shiwei Suo

Sensor data collection from commercial aircraft faces challenges such as incomplete datasets, difficulty in assessing sensor significance, and inability to detect anomalous time points, leading to issues like ambiguous brake control valve faults categories. To address these, a new modeling framework is proposed to improve fault mode distinguishability through high-dimensional mapping. This framework uses Variational Autoencoders for training, combining reconstruction error and latent space similarity. It trains low-dimensional sensor data in two rounds, gradually approximating the target domain and synthesizing high-dimensional samples, enhancing cross-domain feature representation. Additionally, a time-adaptive weight allocation mechanism in a Bidirectional Long Short-Term Memory highlights critical signals, while a multi-head spatial attention mechanism reduces irrelevant signals. Experimental results show that the proposed fault diagnosis approach for brake control valves, utilizing aircraft onboard sensor data, achieves over 96 % in accuracy, precision, recall, and F1-score, outperforming the best performance of six classical network models by approximately 5 %.

商用飞机的传感器数据收集面临着数据集不完整、传感器重要性评估困难、无法检测异常时间点等挑战，从而导致制动控制阀故障类别模糊等问题。为了解决这些问题，提出了一种新的建模框架，通过高维映射提高故障模式的可分辨性。该框架使用变分自编码器进行训练，结合重构误差和潜在空间相似度。它分两轮训练低维传感器数据，逐步逼近目标域，合成高维样本，增强跨域特征表示。此外，双向长短期记忆中的时间自适应权重分配机制突出了关键信号，而多头空间注意机制则减少了无关信号。实验结果表明，基于机载传感器数据的制动控制阀故障诊断方法在准确率、精密度、召回率和f1分数方面均达到了96% %以上，比六种经典网络模型的最佳性能高出约5 %。

{"title":"Enhanced cross-domain signal and physics-based interpretation for fault diagnosis of aircraft brake control valve under limited onboard signal acquisition","authors":"Lin Lin , Yin Chen , Wenhui He , Song Fu , Shiwei Suo","doi":"10.1016/j.compind.2025.104378","DOIUrl":"10.1016/j.compind.2025.104378","url":null,"abstract":"<div><div>Sensor data collection from commercial aircraft faces challenges such as incomplete datasets, difficulty in assessing sensor significance, and inability to detect anomalous time points, leading to issues like ambiguous brake control valve faults categories. To address these, a new modeling framework is proposed to improve fault mode distinguishability through high-dimensional mapping. This framework uses Variational Autoencoders for training, combining reconstruction error and latent space similarity. It trains low-dimensional sensor data in two rounds, gradually approximating the target domain and synthesizing high-dimensional samples, enhancing cross-domain feature representation. Additionally, a time-adaptive weight allocation mechanism in a Bidirectional Long Short-Term Memory highlights critical signals, while a multi-head spatial attention mechanism reduces irrelevant signals. Experimental results show that the proposed fault diagnosis approach for brake control valves, utilizing aircraft onboard sensor data, achieves over 96 % in accuracy, precision, recall, and F1-score, outperforming the best performance of six classical network models by approximately 5 %.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104378"},"PeriodicalIF":9.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SmartARW: A text-aware smart mobile industrial augmented reality (AR) wiring assembly system SmartARW：文本感知智能移动工业增强现实（AR）布线装配系统

IF 9.1 1区计算机科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers in Industry

Pub Date : 2025-09-24 DOI: 10.1016/j.compind.2025.104379

Tienong Zhang , Wei Fang , Lixi Chen , Qiankun Zhang , Hao Hu , Jiapeng Bi

Augmented reality (AR) has demonstrated its potential by delivering intuitive guidance on the workbench directly, alleviating the operators’ mental load for traditional paper-based wiring assembly tasks while ensuring procedural correctness. Nevertheless, current industrial AR wiring applications mainly focus on superimposing the virtual models with the commercial AR glass, lacking practical adaptability due to ergonomic issues derived from their weight, and human intervention is also necessary to monitor the current wiring harness and activate ongoing procedural progress. To bridge this gap for real-world applications, this article proposes human-centric SmartARW: a text-aware smart mobile industrial AR wiring assembly system. Firstly, a mobile AR wiring assembly system is established with the off-the-shelf tablet and sensor module, followed by accurate system calibration among different components and self-contained markerless motion tracking. Then, a closed-loop confirming strategy enabled lightweight text-aware network is proposed and integrated into the mobile AR system, as well as lots of shop-floor datasets are prepared and augmented to improve the context-aware performance, thus the status of the ongoing wiring activity can be perceived accurately even encountered challenge scenarios while activating the corresponding AR instructions automatically. Finally, extensive quantitative and qualitative experiments are carried out to illustrate that the proposed SmartARW has strong usability and can provide superior performance for human-centric smart wiring assembly action.

增强现实（AR）通过直接在工作台上提供直观的指导，减轻了操作员对传统纸质布线组装任务的心理负担，同时确保了程序的正确性，从而展示了其潜力。然而，目前的工业AR布线应用主要集中在将虚拟模型与商业AR玻璃叠加，由于其重量引起的人体工程学问题，缺乏实际适应性，并且人工干预也是必要的，以监测当前的线束并激活正在进行的程序进度。为了在实际应用中弥补这一差距，本文提出了以人为中心的SmartARW：一种文本感知的智能移动工业AR布线装配系统。首先，利用现成的平板电脑和传感器模块建立移动AR布线装配系统，然后在不同组件之间进行精确的系统校准，并进行独立的无标记运动跟踪。然后，提出了一个闭环确认策略支持的轻量级文本感知网络，并将其集成到移动AR系统中，并准备和增强了大量的车间数据集，以提高上下文感知性能，从而在遇到挑战场景时也能准确地感知正在进行的布线活动的状态，同时自动激活相应的AR指令。最后，进行了大量的定量和定性实验，以说明所提出的SmartARW具有很强的可用性，并且可以为以人为中心的智能布线装配动作提供优越的性能。

{"title":"SmartARW: A text-aware smart mobile industrial augmented reality (AR) wiring assembly system","authors":"Tienong Zhang , Wei Fang , Lixi Chen , Qiankun Zhang , Hao Hu , Jiapeng Bi","doi":"10.1016/j.compind.2025.104379","DOIUrl":"10.1016/j.compind.2025.104379","url":null,"abstract":"<div><div>Augmented reality (AR) has demonstrated its potential by delivering intuitive guidance on the workbench directly, alleviating the operators’ mental load for traditional paper-based wiring assembly tasks while ensuring procedural correctness. Nevertheless, current industrial AR wiring applications mainly focus on superimposing the virtual models with the commercial AR glass, lacking practical adaptability due to ergonomic issues derived from their weight, and human intervention is also necessary to monitor the current wiring harness and activate ongoing procedural progress. To bridge this gap for real-world applications, this article proposes human-centric SmartARW: a text-aware smart mobile industrial AR wiring assembly system. Firstly, a mobile AR wiring assembly system is established with the off-the-shelf tablet and sensor module, followed by accurate system calibration among different components and self-contained markerless motion tracking. Then, a closed-loop confirming strategy enabled lightweight text-aware network is proposed and integrated into the mobile AR system, as well as lots of shop-floor datasets are prepared and augmented to improve the context-aware performance, thus the status of the ongoing wiring activity can be perceived accurately even encountered challenge scenarios while activating the corresponding AR instructions automatically. Finally, extensive quantitative and qualitative experiments are carried out to illustrate that the proposed SmartARW has strong usability and can provide superior performance for human-centric smart wiring assembly action.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"173 ","pages":"Article 104379"},"PeriodicalIF":9.1,"publicationDate":"2025-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158583","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0