Pub Date : 2025-02-26DOI: 10.1016/j.compind.2025.104270
Heli Liu , Xiao Yang , Denis Politis , Huifeng Shi , Liliang Wang
The growing availability of metal forming data has driven a new era of data-centric approaches in digital manufacturing. This wealth of data enables the development of digitally enhanced metal forming processes and associated technologies. In this work, using the hot stamping data obtained from a cloud-based manufacturing database, the digital characteristics (DC), defined as the visualisation of a specific manufacturing process containing essential information spanning over the design, manufacturing, and application phases of the products, were unlocked for the hot stamping process. The complex contact conditions were successfully visualised by the hot stamping DC. Following this discovery, the performance of transient lubricant behaviours was evaluated under complex loading and constant loading conditions regarding coefficient of friction evolution and lubricant limit diagram (LLD), which is a digitally-enhanced approach to enable the quantitative evaluation of different lubricants. Results demonstrate that the efficacy of DC-enhanced methodology facilitates the insightful comprehension of transient tribological behaviours and offers great potential on customised lubricant development towards optimisation of hot stamping and metal forming processes.
{"title":"An evaluation scheme incorporating digital characteristics for transient tribological behaviours under complex loading conditions for the hot stamping process","authors":"Heli Liu , Xiao Yang , Denis Politis , Huifeng Shi , Liliang Wang","doi":"10.1016/j.compind.2025.104270","DOIUrl":"10.1016/j.compind.2025.104270","url":null,"abstract":"<div><div>The growing availability of metal forming data has driven a new era of data-centric approaches in digital manufacturing. This wealth of data enables the development of digitally enhanced metal forming processes and associated technologies. In this work, using the hot stamping data obtained from a cloud-based manufacturing database, the digital characteristics (DC), defined as the visualisation of a specific manufacturing process containing essential information spanning over the design, manufacturing, and application phases of the products, were unlocked for the hot stamping process. The complex contact conditions were successfully visualised by the hot stamping DC. Following this discovery, the performance of transient lubricant behaviours was evaluated under complex loading and constant loading conditions regarding coefficient of friction evolution and lubricant limit diagram (LLD), which is a digitally-enhanced approach to enable the quantitative evaluation of different lubricants. Results demonstrate that the efficacy of DC-enhanced methodology facilitates the insightful comprehension of transient tribological behaviours and offers great potential on customised lubricant development towards optimisation of hot stamping and metal forming processes.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104270"},"PeriodicalIF":8.2,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143488283","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-26DOI: 10.1016/j.compind.2025.104267
Francisco Ambrosio Garcia , Hendrik Devriendt , Hüseyin Metin , Merih Özer , Frank Naets
In process industry, plant operators often rely on their experience to choose suitable process settings that meet the productivity and quality goals. When these goals are not met, multiple changes to the settings might be necessary, which is time-consuming because each adjustment requires waiting for the new steady-state condition. A digital twin that quickly provides key performance indicators in steady-state as a function of these settings can speed up this task. The settings can be manually simulated before being adopted, or the digital twin can be integrated into an optimizer to automatically suggest optimal values to the operator, who ultimately makes the final decision. Despite advances in approaches to design such digital twins, most studies lack strategies to update the models when the plant behavior changes, and often overlook constraints and human-centric aspects of the plant operation. To address these gaps, we present a framework for training, tuning, and updating models for supporting the selection of process settings in continuous manufacturing. By directly mapping the steady-state conditions as a function of process settings, our approach enables informed decision-making and paves the way towards process optimization without requiring modifications to the plant control software, a crucial factor in established plants to ensure safety. We propose an interpretable model architecture, and a training process that incorporates both data and prior physical knowledge. Triggers detect deviations between the models’ predictions and the plant condition, in order to start model updates. The procedure for updating the models is tuned to perform consistently well in a variety of conditions, based on substantial simulations in historical data. To select the triggers, we balance technical and human aspects, by considering the trade-off between frequent model updates, increasing operator workload with frequent settings changes, versus how closely the models track the plant conditions. The framework is applied to five different stages of the fiberboard production process in a 1.4-year dataset, to predict key energy and quality-related variables as a function of process settings. The results show that the models, when connected to the data stream, are effectively updated when needed, show high sensitivity to the process settings and consistency with the available physical knowledge, making them well-suited to support the selection of process settings.
{"title":"Physics-informed digital twin design for supporting the selection of process settings in continuous manufacturing, with a focus in fiberboard production","authors":"Francisco Ambrosio Garcia , Hendrik Devriendt , Hüseyin Metin , Merih Özer , Frank Naets","doi":"10.1016/j.compind.2025.104267","DOIUrl":"10.1016/j.compind.2025.104267","url":null,"abstract":"<div><div>In process industry, plant operators often rely on their experience to choose suitable process settings that meet the productivity and quality goals. When these goals are not met, multiple changes to the settings might be necessary, which is time-consuming because each adjustment requires waiting for the new steady-state condition. A digital twin that quickly provides key performance indicators in steady-state as a function of these settings can speed up this task. The settings can be manually simulated before being adopted, or the digital twin can be integrated into an optimizer to automatically suggest optimal values to the operator, who ultimately makes the final decision. Despite advances in approaches to design such digital twins, most studies lack strategies to update the models when the plant behavior changes, and often overlook constraints and human-centric aspects of the plant operation. To address these gaps, we present a framework for training, tuning, and updating models for supporting the selection of process settings in continuous manufacturing. By directly mapping the steady-state conditions as a function of process settings, our approach enables informed decision-making and paves the way towards process optimization without requiring modifications to the plant control software, a crucial factor in established plants to ensure safety. We propose an interpretable model architecture, and a training process that incorporates both data and prior physical knowledge. Triggers detect deviations between the models’ predictions and the plant condition, in order to start model updates. The procedure for updating the models is tuned to perform consistently well in a variety of conditions, based on substantial simulations in historical data. To select the triggers, we balance technical and human aspects, by considering the trade-off between frequent model updates, increasing operator workload with frequent settings changes, versus how closely the models track the plant conditions. The framework is applied to five different stages of the fiberboard production process in a 1.4-year dataset, to predict key energy and quality-related variables as a function of process settings. The results show that the models, when connected to the data stream, are effectively updated when needed, show high sensitivity to the process settings and consistency with the available physical knowledge, making them well-suited to support the selection of process settings.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"168 ","pages":"Article 104267"},"PeriodicalIF":8.2,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143488703","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}
Pub Date : 2025-02-26DOI: 10.1016/j.compind.2025.104269
Teng Zhang , Fangyu Peng , Zhao Yang , Xiaowei Tang , Rong Yan
Despite the extensive use of robots in numerous fields, condition-sensitive robotic machining errors represent a significant obstacle to their high-precision implementation. Prediction-based compensatory control represents a crucial approach to enhancing robot accuracy. The extant machining error prediction methods are beset with shortcomings, including inadequate feature extraction, limited generalizability with respect to working conditions, and the squandering of knowledge. Therefore, the influence mechanisms of robot errors by different working conditions and spatial ontology properties are explored in this paper. A spatial-temporal dual-view error prediction model is constructed for a single condition. Moreover, an innovative unsupervised generalized prediction strategy of machining error for new conditions under the historical task knowledge distillation of Multi-Teacher-Single-Student (MTSS) is proposed. This strategy enables the extraction and reuse of knowledge at three levels: teacher-teaching, student-learning, and generalized expansion. It also ensures the high-precision, lightweight, and high-efficiency prediction of machining error for unseen conditions. The proposed method was validated on constructed complex part inner wall features. The minimum mean absolute error (MAE) indicator for single condition prediction is 0.005 mm, which is a significantly more accurate result than other methods under comparison. Furthermore, the average MAE of unsupervised generalization for new conditions is 0.019 mm, which meets the practical application requirements. Furthermore, the distilled model complexity is reduced by 75 %, and the average inference efficiency is enhanced by over 95 %. This provides the potential for lightweight online deployment. The proposed method offers a robust foundation for prediction-based error online compensation, which is anticipated to facilitate the expansion of robots in high-precision scenarios.
{"title":"UGP-KD: An unsupervised generalized prediction framework for robot machining quality under historical task knowledge distillation for new tasks","authors":"Teng Zhang , Fangyu Peng , Zhao Yang , Xiaowei Tang , Rong Yan","doi":"10.1016/j.compind.2025.104269","DOIUrl":"10.1016/j.compind.2025.104269","url":null,"abstract":"<div><div>Despite the extensive use of robots in numerous fields, condition-sensitive robotic machining errors represent a significant obstacle to their high-precision implementation. Prediction-based compensatory control represents a crucial approach to enhancing robot accuracy. The extant machining error prediction methods are beset with shortcomings, including inadequate feature extraction, limited generalizability with respect to working conditions, and the squandering of knowledge. Therefore, the influence mechanisms of robot errors by different working conditions and spatial ontology properties are explored in this paper. A spatial-temporal dual-view error prediction model is constructed for a single condition. Moreover, an innovative unsupervised generalized prediction strategy of machining error for new conditions under the historical task knowledge distillation of Multi-Teacher-Single-Student (MTSS) is proposed. This strategy enables the extraction and reuse of knowledge at three levels: teacher-teaching, student-learning, and generalized expansion. It also ensures the high-precision, lightweight, and high-efficiency prediction of machining error for unseen conditions. The proposed method was validated on constructed complex part inner wall features. The minimum mean absolute error (MAE) indicator for single condition prediction is 0.005 mm, which is a significantly more accurate result than other methods under comparison. Furthermore, the average MAE of unsupervised generalization for new conditions is 0.019 mm, which meets the practical application requirements. Furthermore, the distilled model complexity is reduced by 75 %, and the average inference efficiency is enhanced by over 95 %. This provides the potential for lightweight online deployment. The proposed method offers a robust foundation for prediction-based error online compensation, which is anticipated to facilitate the expansion of robots in high-precision scenarios.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"168 ","pages":"Article 104269"},"PeriodicalIF":8.2,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143488702","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}
Pub Date : 2025-02-24DOI: 10.1016/j.compind.2025.104265
Fajia Wan, Guo Zhang, Zeteng Li
Surface defect detection is a research hotspot in the field of computer vision. Due to the complex characteristics of metal surfaces and the multitude of industrial defects, it remains a challenging task. In order to meet the requirement of accurate identification of surface defects on copper strips and plates in industrial quality control, we propose a computer vision-based dual-branch features fusion neural network named as DFSDNet. We gather defect samples to construct the KUST-DET dataset of surface defects on copper strips and plates to support the training and evaluation of detection models. Experiments on the KUST-DET dataset demonstrate that DFSDNet-s achieved a mean Average Precision (mAP) of 88.53%, while maintaining low computational complexity and low parameters, and the model achieves a good balance between detection precision and computational efficiency. In addition, the mAP on the NEU-DET dataset is 75.67%, showcasing its good defect detection performance. Experiments have shown that DFSDNet is an effective surface defect detection model with great potential in other metal industry applications.
{"title":"DFSDNet: A dual-branch multi-scale feature fusion network for surface defect detection of copper strips and plates","authors":"Fajia Wan, Guo Zhang, Zeteng Li","doi":"10.1016/j.compind.2025.104265","DOIUrl":"10.1016/j.compind.2025.104265","url":null,"abstract":"<div><div>Surface defect detection is a research hotspot in the field of computer vision. Due to the complex characteristics of metal surfaces and the multitude of industrial defects, it remains a challenging task. In order to meet the requirement of accurate identification of surface defects on copper strips and plates in industrial quality control, we propose a computer vision-based dual-branch features fusion neural network named as DFSDNet. We gather defect samples to construct the KUST-DET dataset of surface defects on copper strips and plates to support the training and evaluation of detection models. Experiments on the KUST-DET dataset demonstrate that DFSDNet-s achieved a mean Average Precision (mAP) of 88.53%, while maintaining low computational complexity and low parameters, and the model achieves a good balance between detection precision and computational efficiency. In addition, the mAP on the NEU-DET dataset is 75.67%, showcasing its good defect detection performance. Experiments have shown that DFSDNet is an effective surface defect detection model with great potential in other metal industry applications.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104265"},"PeriodicalIF":8.2,"publicationDate":"2025-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143479623","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}
In high-speed milling, chatter detection plays an important role in ensuring surface quality and safe machining. Traditionally, chatter detection is performed by manually setting the feature threshold, which is unreliable. In this paper, an intelligent chatter detection method is proposed based on deep learning. The proposed method is featured by automatic chatter detection based on multi-channel features, and it is applicable in different milling conditions. To adaptively obtain the chatter signal and avoid the problem of modal mixing, the successive variational mode decomposition method is first used to extract the chatter frequency components without selecting parameters. Then, multi-channel features are extracted from the reconstructed chatter signal, and sensitive features strongly related to the milling chatter are selected based on mutual information metric. Next, a novel multi-channel feature fusion network, composed of the gated attention mechanism, ResNet module, CapsNet module, and classification module, is constructed to mine feature information and implement chatter detection. Finally, the signal data are acquired through a series of milling experiments. The identification performance of the model is evaluated in three scenarios, and an average accuracy of 0.9887 is achieved. In addition, ablation experiments and comparative studies with other detection methods are performed. The results show that the proposed method can improve the accuracy and generalization of chatter detection.
{"title":"Intelligent chatter detection in high-speed milling using successive variational mode decomposition and a multi-channel feature fusion network","authors":"Liangshi Sun , Xianzhen Huang , Jiatong Zhao , Zhiyuan Jiang , Fusheng Jiang","doi":"10.1016/j.compind.2025.104266","DOIUrl":"10.1016/j.compind.2025.104266","url":null,"abstract":"<div><div>In high-speed milling, chatter detection plays an important role in ensuring surface quality and safe machining. Traditionally, chatter detection is performed by manually setting the feature threshold, which is unreliable. In this paper, an intelligent chatter detection method is proposed based on deep learning. The proposed method is featured by automatic chatter detection based on multi-channel features, and it is applicable in different milling conditions. To adaptively obtain the chatter signal and avoid the problem of modal mixing, the successive variational mode decomposition method is first used to extract the chatter frequency components without selecting parameters. Then, multi-channel features are extracted from the reconstructed chatter signal, and sensitive features strongly related to the milling chatter are selected based on mutual information metric. Next, a novel multi-channel feature fusion network, composed of the gated attention mechanism, ResNet module, CapsNet module, and classification module, is constructed to mine feature information and implement chatter detection. Finally, the signal data are acquired through a series of milling experiments. The identification performance of the model is evaluated in three scenarios, and an average accuracy of 0.9887 is achieved. In addition, ablation experiments and comparative studies with other detection methods are performed. The results show that the proposed method can improve the accuracy and generalization of chatter detection.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104266"},"PeriodicalIF":8.2,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143453179","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}
Pub Date : 2025-02-18DOI: 10.1016/j.compind.2025.104264
Juan Luis Ramos Villalon , Luis de la Torre , Zhongcheng Lei , Wenshan Hu , Hugo Tadashi Kussaba , Victoria Lemieux
Cyber-physical systems (CPSs) is a general concept that encompasses a wide variety of systems. Depending on their nature, application, and accessibility needs and restrictions, CPSs can differ a lot from each other. This paper proposes a classification of CPSs based on their accessibility needs and restrictions and, more importantly, presents an approach to create a decentralised and worldwide common access management framework for CPSs for non-critical infrastructures that are meant to be shared and accessed remotely (i.e., offered as a service). The presented solution uses a permissionless blockchain, existing fungible tokens, and a combination of smart contracts based on nonfungible token standards/proposals to enable CPS owners to manage secure, flexible access without centralised oversight. In addition, the proposed framework provides built-in mechanisms for: (i) charging for the use of CPSs, (ii) availability calendar configuration, (iii) worldwide visibility, (iv) easy integration with authentication/authorisation methods, and (v) access control flexibility.
{"title":"A decentralised approach to cyber-physical systems as a service: Managing shared access worldwide through blockchain standards","authors":"Juan Luis Ramos Villalon , Luis de la Torre , Zhongcheng Lei , Wenshan Hu , Hugo Tadashi Kussaba , Victoria Lemieux","doi":"10.1016/j.compind.2025.104264","DOIUrl":"10.1016/j.compind.2025.104264","url":null,"abstract":"<div><div>Cyber-physical systems (CPSs) is a general concept that encompasses a wide variety of systems. Depending on their nature, application, and accessibility needs and restrictions, CPSs can differ a lot from each other. This paper proposes a classification of CPSs based on their accessibility needs and restrictions and, more importantly, presents an approach to create a decentralised and worldwide common access management framework for CPSs for non-critical infrastructures that are meant to be shared and accessed remotely (i.e., offered as a service). The presented solution uses a permissionless blockchain, existing fungible tokens, and a combination of smart contracts based on nonfungible token standards/proposals to enable CPS owners to manage secure, flexible access without centralised oversight. In addition, the proposed framework provides built-in mechanisms for: (i) charging for the use of CPSs, (ii) availability calendar configuration, (iii) worldwide visibility, (iv) easy integration with authentication/authorisation methods, and (v) access control flexibility.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104264"},"PeriodicalIF":8.2,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143428034","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}
Pub Date : 2025-02-18DOI: 10.1016/j.compind.2025.104263
Daniel Pakkala, Niko Känsäkoski, Tapio Heikkilä, Jere Backman, Pekka Pääkkönen
Fostered by the recent development in artificial intelligence technologies, digitalization in industries is proceeding towards intelligent automation of various physical work processes with autonomous robotic applications, in dynamic and non-deterministic environments, and in collaboration with human workers. The article presents an explorative case study on designing a cognitive situation-adaptive Autonomous Mobile Robotics (AMR) application for material hauling, in a simulated underground mining context. The goal of the research is to synthesize and present new design knowledge for improving situation-adaptation capabilities of AMR applications, which are increasingly required as the operational environments for the AMRs become dynamic, non-deterministic, and include people working on the same area with the robots. The research applies design science research methodology, and evaluates the results empirically via a prototype system, which is demonstrated in laboratory setting simulating an underground tunnel network. As an outstanding contribution, the results contribute a novel, nascent, and empirically evaluated design approach, which proposes three design aspects combining design and engineering activities across the systems engineering, knowledge engineering, computer science and robotics disciplines. Empirical evaluation is made via design, development, and demonstration of a system architecture and prototype system of a cognitive situation-adaptive AMR application, which is used in synthesis and evaluation of the design approach. The three design aspects proposed by the approach are 1) Context of operation, 2) Knowledge-driven behaviour, and 3) Knowledge driven operation. Also design challenges, future research and development needs, and innovation potential on designing of cognitive situation-adaptive AMR applications for industrial use are identified and discussed.
{"title":"On design of cognitive situation-adaptive autonomous mobile robotic applications","authors":"Daniel Pakkala, Niko Känsäkoski, Tapio Heikkilä, Jere Backman, Pekka Pääkkönen","doi":"10.1016/j.compind.2025.104263","DOIUrl":"10.1016/j.compind.2025.104263","url":null,"abstract":"<div><div>Fostered by the recent development in artificial intelligence technologies, digitalization in industries is proceeding towards intelligent automation of various physical work processes with autonomous robotic applications, in dynamic and non-deterministic environments, and in collaboration with human workers. The article presents an explorative case study on designing a cognitive situation-adaptive Autonomous Mobile Robotics (AMR) application for material hauling, in a simulated underground mining context. The goal of the research is to synthesize and present new design knowledge for improving situation-adaptation capabilities of AMR applications, which are increasingly required as the operational environments for the AMRs become dynamic, non-deterministic, and include people working on the same area with the robots. The research applies design science research methodology, and evaluates the results empirically via a prototype system, which is demonstrated in laboratory setting simulating an underground tunnel network. As an outstanding contribution, the results contribute a novel, nascent, and empirically evaluated design approach, which proposes three design aspects combining design and engineering activities across the systems engineering, knowledge engineering, computer science and robotics disciplines. Empirical evaluation is made via design, development, and demonstration of a system architecture and prototype system of a cognitive situation-adaptive AMR application, which is used in synthesis and evaluation of the design approach. The three design aspects proposed by the approach are 1) Context of operation, 2) Knowledge-driven behaviour, and 3) Knowledge driven operation. Also design challenges, future research and development needs, and innovation potential on designing of cognitive situation-adaptive AMR applications for industrial use are identified and discussed.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104263"},"PeriodicalIF":8.2,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143428035","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-14DOI: 10.1016/j.compind.2025.104262
Yuchen Liang , Yuqi Wang , Weidong Li , Duc Truong Pham , Jinzhong Lu
Faults occurring during machining processes can severely impact productivity and product quality. Deep learning models have been actively used to develop fault diagnosis approaches. However, it is challenging for industries to adopt the approaches due to their inability to adapt to varying machining conditions. To address the issue, a novel diagnostic approach is designed based on a hybrid convolutional neural network (CNN)-long short-term memory (LSTM) model and an incremental transfer learning strategy. Based on the incremental transfer learning, the CNN-LSTM model can acquire knowledge from previous machining conditions (source domain) and effectively apply it to new conditions (target domain). In the diagnostic approach, instance-based transfer learning, knowledge-based transfer learning, and incremental transfer learning are combined to improve the training efficiency and overcome the issue of forgetting previously learned knowledge. The CNN-LSTM-attention model is designed as a supplementary model when the data complexity is high. Experimental results show that the approach increased the average training accuracy from 88.63 % to 97.10 %, and required training datasets were reduced by 96.97 %. In addition, the incremental transfer learning reduced false detections for 71.24 %.
{"title":"Adaptive fault diagnosis of machining processes enabled by hybrid deep learning and incremental transfer learning","authors":"Yuchen Liang , Yuqi Wang , Weidong Li , Duc Truong Pham , Jinzhong Lu","doi":"10.1016/j.compind.2025.104262","DOIUrl":"10.1016/j.compind.2025.104262","url":null,"abstract":"<div><div>Faults occurring during machining processes can severely impact productivity and product quality. Deep learning models have been actively used to develop fault diagnosis approaches. However, it is challenging for industries to adopt the approaches due to their inability to adapt to varying machining conditions. To address the issue, a novel diagnostic approach is designed based on a hybrid convolutional neural network (CNN)-long short-term memory (LSTM) model and an incremental transfer learning strategy. Based on the incremental transfer learning, the CNN-LSTM model can acquire knowledge from previous machining conditions (source domain) and effectively apply it to new conditions (target domain). In the diagnostic approach, instance-based transfer learning, knowledge-based transfer learning, and incremental transfer learning are combined to improve the training efficiency and overcome the issue of forgetting previously learned knowledge. The CNN-LSTM-attention model is designed as a supplementary model when the data complexity is high. Experimental results show that the approach increased the average training accuracy from 88.63 % to 97.10 %, and required training datasets were reduced by 96.97 %. In addition, the incremental transfer learning reduced false detections for 71.24 %.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104262"},"PeriodicalIF":8.2,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419124","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}
Pub Date : 2025-02-10DOI: 10.1016/j.compind.2025.104256
Marvin Herchenbach , Sven Weinzierl , Sandra Zilker , Erik Schwulera , Martin Matzner
In recent years, there has been a remarkable shift from automated plants to intelligent production in the industrial context, accelerated by technologies such as artificial intelligence (AI). The ultimate goal is an autonomous plant that is capable of self-regulation and self-optimization. In electronics production, the first approaches have been proposed for deriving and adjusting machine parameters for solder paste printing in the surface-mount technology production of printed circuit boards. However, these approaches are often static and perform reactive actions since they are either based on expert systems or data-driven models. To reach a dynamic optimization, this work proposes a methodology, called adaptive AI-based causal control, allowing offline and online optimization. Following the principles of the Design for Six Sigma method, customer-oriented key performance indicators were derived, that aimed at a stable soldering process by focusing on the spread of the solder volume and a dedicated overall spread metric. The offline optimization (open-loop control) is based on a surrogate model approach to find optimal initial printing parameters. The online optimization (closed-loop control) employs a data-driven model predictive control to adjust the printing parameters dynamically. In addition, to consider the causal effects of the control variables in the online optimization, a causal graph is exploited in the predictive controller. Regarding the effectiveness of the open-loop control, our evaluation reveals a reduction in spread by 11.3% in production. Furthermore, in terms of the efficacy of the closed-loop control, we obtain a reduction in volume range by 16.7% in a simulated setting of the predictive controller. Thereby, the integration of a causal inference component based on a generated causal graph, achieving a recall of 76.9% by considering process knowledge identified with domain experts, accounts for about 2.8% of the recall.
{"title":"A methodology for adaptive AI-based causal control: Toward an autonomous factory in solder paste printing","authors":"Marvin Herchenbach , Sven Weinzierl , Sandra Zilker , Erik Schwulera , Martin Matzner","doi":"10.1016/j.compind.2025.104256","DOIUrl":"10.1016/j.compind.2025.104256","url":null,"abstract":"<div><div>In recent years, there has been a remarkable shift from automated plants to intelligent production in the industrial context, accelerated by technologies such as artificial intelligence (AI). The ultimate goal is an autonomous plant that is capable of self-regulation and self-optimization. In electronics production, the first approaches have been proposed for deriving and adjusting machine parameters for solder paste printing in the surface-mount technology production of printed circuit boards. However, these approaches are often static and perform reactive actions since they are either based on expert systems or data-driven models. To reach a dynamic optimization, this work proposes a methodology, called adaptive AI-based causal control, allowing offline and online optimization. Following the principles of the Design for Six Sigma method, customer-oriented key performance indicators were derived, that aimed at a stable soldering process by focusing on the spread of the solder volume and a dedicated overall spread metric. The offline optimization (open-loop control) is based on a surrogate model approach to find optimal initial printing parameters. The online optimization (closed-loop control) employs a data-driven model predictive control to adjust the printing parameters dynamically. In addition, to consider the causal effects of the control variables in the online optimization, a causal graph is exploited in the predictive controller. Regarding the effectiveness of the open-loop control, our evaluation reveals a reduction in spread by 11.3% in production. Furthermore, in terms of the efficacy of the closed-loop control, we obtain a reduction in volume range by 16.7% in a simulated setting of the predictive controller. Thereby, the integration of a causal inference component based on a generated causal graph, achieving a recall of 76.9% by considering process knowledge identified with domain experts, accounts for about 2.8% of the recall.</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104256"},"PeriodicalIF":8.2,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The degradation of industrial systems is a natural and often unavoidable process. Hidden Markov Models (HMMs) are used for state-based bearing degradation analysis. A challenge arises because bearings can deteriorate in multiple ways, depending on crack locations. To address this, a Multi-Branch Hidden Markov Model (MB-HMM) was developed to handle multiple deteriorations. However, MB-HMM primarily uses simulated data where deterioration is known in advance. In contrast, real-world sensors collect data with uncertainties, potentially causing false alarms and impacting the First Predicting Time (FPT). We used the FEMTO-bearing dataset, which includes continuous monitoring until failure, with unknown fault locations and varying degradation levels. This study presents a comprehensive preprocessing framework and employs the Extended Multi-Branch HMM (EMB-HMM). Our experimental analysis shows that the proposed strategy significantly enhances the Signal-to-Noise Ratio (SNR). The active branch is defined based on prior and posterior probabilities, with the branch's prior probability and topology linked to the four fault frequencies of the bearing. The EMB-HMM outperforms other models in state prediction, featuring four branches and five hidden states. It improves state sequence accuracy, predicts degradation levels and FPT, and achieves zero false alarms for Fake Fault (FF).
{"title":"Contribution to estimating the level of bearing degradation using a Multi-Branch Hidden Markov Model approach","authors":"Indrawata Wardhana , Amal Gouiaa-Mtibaa , Pascal Vrignat , Frédéric Kratz","doi":"10.1016/j.compind.2025.104254","DOIUrl":"10.1016/j.compind.2025.104254","url":null,"abstract":"<div><div>The degradation of industrial systems is a natural and often unavoidable process. Hidden Markov Models (HMMs) are used for state-based bearing degradation analysis. A challenge arises because bearings can deteriorate in multiple ways, depending on crack locations. To address this, a Multi-Branch Hidden Markov Model (MB-HMM) was developed to handle multiple deteriorations. However, MB-HMM primarily uses simulated data where deterioration is known in advance. In contrast, real-world sensors collect data with uncertainties, potentially causing false alarms and impacting the First Predicting Time (FPT). We used the FEMTO-bearing dataset, which includes continuous monitoring until failure, with unknown fault locations and varying degradation levels. This study presents a comprehensive preprocessing framework and employs the Extended Multi-Branch HMM (EMB-HMM). Our experimental analysis shows that the proposed strategy significantly enhances the Signal-to-Noise Ratio (SNR). The active branch is defined based on prior and posterior probabilities, with the branch's prior probability and topology linked to the four fault frequencies of the bearing. The EMB-HMM outperforms other models in state prediction, featuring four branches and five hidden states. It improves state sequence accuracy, predicts degradation levels and FPT, and achieves zero false alarms for Fake Fault (FF).</div></div>","PeriodicalId":55219,"journal":{"name":"Computers in Industry","volume":"167 ","pages":"Article 104254"},"PeriodicalIF":8.2,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143349239","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号