Advanced Engineering Informatics最新文献_第5页

An LLM-based knowledge and function-augmented approach for optimal design of remanufacturing process 基于 LLM 的知识和功能增强方法，实现再制造流程的优化设计

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103206

Haiyang Zhang , Wei Yan , Huicong Hu , Xumei Zhang , Qingtao Liu , Hong Xia , Yingguang Zhang , Yuhao Lin

Remanufacturing that returns used products to a like-new condition, is essential for promoting the circular economy and reducing carbon emissions. The optimal design of remanufacturing process (ODRP), as a knowledge-intensive complex decision-making task, plays a vital role in the success of remanufacturing. However, insufficient utilization of remanufacturing knowledge, and the trade-offs among multi-objectives in decision-making scenarios make ODRP time-consuming and labor-intensive. With the development of next-generation artificial intelligence (AI) technologies, large language models (LLMs) provide an important enabling tool for complex decision-making tasks. However, existing LLMs still face significant challenges in ODRP due to a lack of remanufacturing knowledge and computational capabilities. To address this issue, an LLM-based approach augmented with knowledge and function is proposed in this paper. Firstly, based on the establishment of remanufacturing process database, a retrieval augmented generation (RAG)-based knowledge-augmented strategy is designed to retrieve failure information (e.g., failure form, failure degree, etc.) of returned products through the interaction with LLMs, and generate the feasible remanufacturing schemes. Secondly, a function-augmented mechanism with function learning is also proposed to calculate each objective value and combined assessed value of the generated remanufacturing schemes with LLMs, assisting process designers in designing optimal remanufacturing scheme and process parameters. Finally, the proposed approach is validated using a case study on automobile gearbox remanufacturing. The results indicate that the proposed knowledge-augmented strategy improves the average accuracy from 65% to 79% when using ChatGLM3-6B as the base LLMs. Additionally, the proposed function-augmented mechanism can calculate the minimum combined assessed value and make more realistic results for ODRP. Meanwhile, the proposed integrated approach provides a solution to knowledge-intensive and complex decision-making tasks, which has a broad application prospect.

{"title":"An LLM-based knowledge and function-augmented approach for optimal design of remanufacturing process","authors":"Haiyang Zhang , Wei Yan , Huicong Hu , Xumei Zhang , Qingtao Liu , Hong Xia , Yingguang Zhang , Yuhao Lin","doi":"10.1016/j.aei.2025.103206","DOIUrl":"10.1016/j.aei.2025.103206","url":null,"abstract":"<div><div>Remanufacturing that returns used products to a like-new condition, is essential for promoting the circular economy and reducing carbon emissions. The optimal design of remanufacturing process (ODRP), as a knowledge-intensive complex decision-making task, plays a vital role in the success of remanufacturing. However, insufficient utilization of remanufacturing knowledge, and the trade-offs among multi-objectives in decision-making scenarios make ODRP time-consuming and labor-intensive. With the development of next-generation artificial intelligence (AI) technologies, large language models (LLMs) provide an important enabling tool for complex decision-making tasks. However, existing LLMs still face significant challenges in ODRP due to a lack of remanufacturing knowledge and computational capabilities. To address this issue, an LLM-based approach augmented with knowledge and function is proposed in this paper. Firstly, based on the establishment of remanufacturing process database, a retrieval augmented generation (RAG)-based knowledge-augmented strategy is designed to retrieve failure information (e.g., failure form, failure degree, etc.) of returned products through the interaction with LLMs, and generate the feasible remanufacturing schemes. Secondly, a function-augmented mechanism with function learning is also proposed to calculate each objective value and combined assessed value of the generated remanufacturing schemes with LLMs, assisting process designers in designing optimal remanufacturing scheme and process parameters. Finally, the proposed approach is validated using a case study on automobile gearbox remanufacturing. The results indicate that the proposed knowledge-augmented strategy improves the average accuracy from 65% to 79% when using ChatGLM3-6B as the base LLMs. Additionally, the proposed function-augmented mechanism can calculate the minimum combined assessed value and make more realistic results for ODRP. Meanwhile, the proposed integrated approach provides a solution to knowledge-intensive and complex decision-making tasks, which has a broad application prospect.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103206"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143436482","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

Binocular vision-based pose monitoring technique for assembly alignment of precast concrete components

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103205

Lizhi Long , Jingjing Guo , Honghu Chu , Songyue Wang , Shaopeng Xu , Lu Deng

Vision-based pose monitoring for the assembly alignment of precast concrete components at long distances is often hindered by the small size of elliptical targets and cluttered construction backgrounds. To address the challenges in small ellipse identification and coordinate calculation, this study proposes a binocular vision-based pose monitoring technique for precast concrete components. First, an improved ellipse target detection algorithm is developed, integrating Omni-Dimensional Dynamic Convolution (ODConv) and a Coordinate Attention (CA) mechanism to enhance small target recognition under complex backgrounds. ODConv dynamically adjusts convolutional kernel shapes to capture multi-scale features while the CA mechanism embeds precise location information into channel attention. Additionally, The Normalized Wasserstein Distance (NWD) is employed to improve non-maximum suppression by considering both spatial distance and shape similarity, which is particularly beneficial for small and densely distributed targets. Second, an enhanced ellipse center extraction algorithm is introduced, utilizing image magnification and contrast enhancement for more accurate pixel coordinate extraction. The experiments demonstrated the superior performance of the proposed method with a precision of 97.6 %, a recall of 97.8 %, a mean average precision ([email protected]:0.95) of 70.5 %, and an inference speed of 35.8 FPS. This balance between accuracy and efficiency ensures the feasibility of real-time applications. Furthermore, field tests conducted at a hoisting site for precast concrete columns achieved positioning errors in the x and y directions within 5 mm over a monitoring range of 10 m, which confirmed the method’s practical reliability.

{"title":"Binocular vision-based pose monitoring technique for assembly alignment of precast concrete components","authors":"Lizhi Long , Jingjing Guo , Honghu Chu , Songyue Wang , Shaopeng Xu , Lu Deng","doi":"10.1016/j.aei.2025.103205","DOIUrl":"10.1016/j.aei.2025.103205","url":null,"abstract":"<div><div>Vision-based pose monitoring for the assembly alignment of precast concrete components at long distances is often hindered by the small size of elliptical targets and cluttered construction backgrounds. To address the challenges in small ellipse identification and coordinate calculation, this study proposes a binocular vision-based pose monitoring technique for precast concrete components. First, an improved ellipse target detection algorithm is developed, integrating Omni-Dimensional Dynamic Convolution (ODConv) and a Coordinate Attention (CA) mechanism to enhance small target recognition under complex backgrounds. ODConv dynamically adjusts convolutional kernel shapes to capture multi-scale features while the CA mechanism embeds precise location information into channel attention. Additionally, The Normalized Wasserstein Distance (NWD) is employed to improve non-maximum suppression by considering both spatial distance and shape similarity, which is particularly beneficial for small and densely distributed targets. Second, an enhanced ellipse center extraction algorithm is introduced, utilizing image magnification and contrast enhancement for more accurate pixel coordinate extraction. The experiments demonstrated the superior performance of the proposed method with a precision of 97.6 %, a recall of 97.8 %, a mean average precision ([email protected]:0.95) of 70.5 %, and an inference speed of 35.8 FPS. This balance between accuracy and efficiency ensures the feasibility of real-time applications. Furthermore, field tests conducted at a hoisting site for precast concrete columns achieved positioning errors in the x and y directions within 5 mm over a monitoring range of 10 m, which confirmed the method’s practical reliability.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103205"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143436484","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

Moving load induced dynamic response analysis of bridge based on physics-informed neural network

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103215

Yi-Fan Li , Wen-Yu He , Wei-Xin Ren , Ya-Hui Shao

It is crucial to calculate the dynamic response of bridge induced by moving load which is the main live load during operation. Physics-informed neural network (PINN) is powerful in calculating structural response induced by static load as it can provide the prior knowledge for neural network. This paper extends the PINN for dynamic response analysis of bridge subjected to moving load. Firstly, nondimensional partial differential equations of uniform and non-uniform bridges subjected to moving loads are derived. Then, the Dirac function is approximated by Gaussian function, and the corresponding sampling strategy is proposed. Thirdly, the Fourier embedding layer and causal weight are added in the deep neural network and loss function of PINN, respectively. Fourthly, the implementation procedures of the PINN based moving load induced dynamic response analysis method are provided accordingly. Finally, numerical experiments are conducted to verify the effectiveness and superiority proposed method. The results indicate that the moving load induced dynamic response can be obtained by PINN driven by physics (PINN-DP) when the bridge parameters are known, and the response of bridge with unknown parameters can be obtained by PINN driven by both physics and data (PINN-DPD) with small amount of monitored response. Besides, the sampling strategy and causal weights added in the PINN can improve the accuracy of the analyzed results.

{"title":"Moving load induced dynamic response analysis of bridge based on physics-informed neural network","authors":"Yi-Fan Li , Wen-Yu He , Wei-Xin Ren , Ya-Hui Shao","doi":"10.1016/j.aei.2025.103215","DOIUrl":"10.1016/j.aei.2025.103215","url":null,"abstract":"<div><div>It is crucial to calculate the dynamic response of bridge induced by moving load which is the main live load during operation. Physics-informed neural network (PINN) is powerful in calculating structural response induced by static load as it can provide the prior knowledge for neural network. This paper extends the PINN for dynamic response analysis of bridge subjected to moving load. Firstly, nondimensional partial differential equations of uniform and non-uniform bridges subjected to moving loads are derived. Then, the Dirac function is approximated by Gaussian function, and the corresponding sampling strategy is proposed. Thirdly, the Fourier embedding layer and causal weight are added in the deep neural network and loss function of PINN, respectively. Fourthly, the implementation procedures of the PINN based moving load induced dynamic response analysis method are provided accordingly. Finally, numerical experiments are conducted to verify the effectiveness and superiority proposed method. The results indicate that the moving load induced dynamic response can be obtained by PINN driven by physics (PINN-DP) when the bridge parameters are known, and the response of bridge with unknown parameters can be obtained by PINN driven by both physics and data (PINN-DPD) with small amount of monitored response. Besides, the sampling strategy and causal weights added in the PINN can improve the accuracy of the analyzed results.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103215"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429068","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

Multivariate failure prognosis of cutting tools under heterogeneous operating conditions

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103198

Zhenggeng Ye , Le Wang , Hui Yang , Zhiqiang Cai

Failure risk prognosis is indispensable to predict the remaining useful life (RUL) of cutting tools, thereby improving the timely maintenance and boosting the productivity of manufacturing systems. However, the heterogeneity of working conditions is holding back this target. Traditional methods do not discern lifetime data from heterogeneous working conditions but rather aggregate these data for parameter estimation. As such, most of the existing methods become inflexible and cannot adequately handle dynamic and heterogeneous working conditions. Therefore, this paper presents a novel knowledge-driven prognostic framework to integrate the physical feature-based classification model of homogeneous working conditions with the failure risk prognosis of RUL. This new framework effectively identifies and categorizes various types of working conditions with a similarity-evaluation method. Further, a multivariate model integrating lifetime variabilities under homogeneous conditions and real-time prior information is proposed for fault risk and RUL prognosis. This work provides a novel prognostic approach for future risks even with the uncertainty of working conditions. Finally, a case study with degradation datasets of milling insert in the machining center is performed to evaluate and validate the effectiveness of the proposed framework.

{"title":"Multivariate failure prognosis of cutting tools under heterogeneous operating conditions","authors":"Zhenggeng Ye , Le Wang , Hui Yang , Zhiqiang Cai","doi":"10.1016/j.aei.2025.103198","DOIUrl":"10.1016/j.aei.2025.103198","url":null,"abstract":"<div><div>Failure risk prognosis is indispensable to predict the remaining useful life (RUL) of cutting tools, thereby improving the timely maintenance and boosting the productivity of manufacturing systems. However, the heterogeneity of working conditions is holding back this target. Traditional methods do not discern lifetime data from heterogeneous working conditions but rather aggregate these data for parameter estimation. As such, most of the existing methods become inflexible and cannot adequately handle dynamic and heterogeneous working conditions. Therefore, this paper presents a novel knowledge-driven prognostic framework to integrate the physical feature-based classification model of homogeneous working conditions with the failure risk prognosis of RUL. This new framework effectively identifies and categorizes various types of working conditions with a similarity-evaluation method. Further, a multivariate model integrating lifetime variabilities under homogeneous conditions and real-time prior information is proposed for fault risk and RUL prognosis. This work provides a novel prognostic approach for future risks even with the uncertainty of working conditions. Finally, a case study with degradation datasets of milling insert in the machining center is performed to evaluate and validate the effectiveness of the proposed framework.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103198"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430139","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

Scanned point cloud registration for localization of aircraft access panel and complementary frame with weak features

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103209

Ruchen Chen , Jun Yang , Runfeng Xiao , Yang Hui , Aiming Xu , Qiang He , Zhengjie Xue , Pengpo Guo

The quality of trimming access panels on aircraft skin directly affects flight safety and aerodynamic characteristics of aircraft. It is crucial to obtain the trim allowance accurately in a single attempt. However, access panels and their complementary frames are thin-walled components with random surface curvature and weak features, making precise localization challenging. We propose a new efficient and precise localization framework that aligns the scanned point clouds of the access panel and complementary frame to their standard model. We design a contour inflection point feature (CIF) descriptor that facilitates feature retrieval and matching during the alignment process and addresses issues of weak features leading to matching errors. Additionally, we propose a proportional segmented weighted ICP (PSW-ICP) method for precise alignment, which overcomes the problem of local optima in the alignment process due to contour differences. Experiments with multiple types of access panels demonstrate that the proposed registration method significantly outperforms existing algorithms in terms of accuracy and efficiency, achieving a mean localization error of less than 0.07 mm. This provides valuable guidance for the digital assembly of aircraft skin.

{"title":"Scanned point cloud registration for localization of aircraft access panel and complementary frame with weak features","authors":"Ruchen Chen , Jun Yang , Runfeng Xiao , Yang Hui , Aiming Xu , Qiang He , Zhengjie Xue , Pengpo Guo","doi":"10.1016/j.aei.2025.103209","DOIUrl":"10.1016/j.aei.2025.103209","url":null,"abstract":"<div><div>The quality of trimming access panels on aircraft skin directly affects flight safety and aerodynamic characteristics of aircraft. It is crucial to obtain the trim allowance accurately in a single attempt. However, access panels and their complementary frames are thin-walled components with random surface curvature and weak features, making precise localization challenging. We propose a new efficient and precise localization framework that aligns the scanned point clouds of the access panel and complementary frame to their standard model. We design a contour inflection point feature (CIF) descriptor that facilitates feature retrieval and matching during the alignment process and addresses issues of weak features leading to matching errors. Additionally, we propose a proportional segmented weighted ICP (PSW-ICP) method for precise alignment, which overcomes the problem of local optima in the alignment process due to contour differences. Experiments with multiple types of access panels demonstrate that the proposed registration method significantly outperforms existing algorithms in terms of accuracy and efficiency, achieving a mean localization error of less than 0.07 mm. This provides valuable guidance for the digital assembly of aircraft skin.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103209"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143436483","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

FD-LLM: Large language model for fault diagnosis of complex equipment FD-LLM：用于复杂设备故障诊断的大型语言模型

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103208

Lin Lin, Sihao Zhang, Song Fu, Yikun Liu

In complex equipment fault diagnosis, traditional deep learning-based fault diagnosis methods usually require special design and training of “one model for one scenario,” the features of different fault categories overlap seriously, making misdiagnosis easy to occur. The Multimodal Large Language Model (MM-LLM) demonstrates strong multimodal understanding and logical reasoning abilities. This article attempts to use MM-LLM for complex equipment fault diagnosis to yield higher diagnostic accuracy. However, existing MM-LLMs lack domain-specific knowledge and modal alignment training for engineering time-series data, limiting their effectiveness in industrial fault diagnosis tasks. This article proposes a new fault diagnosis method based on MM-LLM in response to the above issues. First, by conducting modal alignment training on the description text of engineering data and equipment operation status in the feature space, the ability of the LLM to understand time-series data modalities is activated. Second, a fuzzy semantic embedding method is proposed to address the difficulty of identifying engineering data with pattern aliasing in the feature space. In addition, supplementary fault diagnosis background knowledge is introduced into MM-LLM by a learnable prompt embedding. Finally, the LORA method is used to fine-tune the LLM. Experimental results show that the proposed method can achieve higher fault classification accuracy.

{"title":"FD-LLM: Large language model for fault diagnosis of complex equipment","authors":"Lin Lin, Sihao Zhang, Song Fu, Yikun Liu","doi":"10.1016/j.aei.2025.103208","DOIUrl":"10.1016/j.aei.2025.103208","url":null,"abstract":"<div><div>In complex equipment fault diagnosis, traditional deep learning-based fault diagnosis methods usually require special design and training of “one model for one scenario,” the features of different fault categories overlap seriously, making misdiagnosis easy to occur. The Multimodal Large Language Model (MM-LLM) demonstrates strong multimodal understanding and logical reasoning abilities. This article attempts to use MM-LLM for complex equipment fault diagnosis to yield higher diagnostic accuracy. However, existing MM-LLMs lack domain-specific knowledge and modal alignment training for engineering time-series data, limiting their effectiveness in industrial fault diagnosis tasks. This article proposes a new fault diagnosis method based on MM-LLM in response to the above issues. First, by conducting modal alignment training on the description text of engineering data and equipment operation status in the feature space, the ability of the LLM to understand time-series data modalities is activated. Second, a fuzzy semantic embedding method is proposed to address the difficulty of identifying engineering data with pattern aliasing in the feature space. In addition, supplementary fault diagnosis background knowledge is introduced into MM-LLM by a learnable prompt embedding. Finally, the LORA method is used to fine-tune the LLM. Experimental results show that the proposed method can achieve higher fault classification accuracy.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103208"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430140","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

SR-FABNet: Super-Resolution branch guided Fourier attention detection network for efficient optical inspection of nanoscale wafer defects

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-18 DOI: 10.1016/j.aei.2025.103200

Leisheng Chen , Kai Meng , Hangying Zhang , Junquan Zhou , Peihuang Lou

In-line optical inspection of nanoscale defects in patterned wafers is crucial for yield management and control in advanced semiconductor production. Current industrial inspection methods primarily rely on die-to-die or die-to-database comparisons. However, due to the continuously shrinking process nodes, optical methods suffer from limited optical resolution and efficiency loss, presenting significant challenges. Therefore, there is an urgent need for efficient and precise optical inspection methods to detect nanoscale physical defects from low-resolution optical diffraction patterns. To address this gap, we propose a super-resolution (SR) branch guided Fourier attention detection network for efficient optical inspections of nanoscale wafer defects. The network establishes a SR branch to guide the utilization of high-resolution image information for defect detection. Moreover, we introduce a novel Fourier Attention Block (FAB) to enhance the model’s capability of discerning defects from sophisticatedly-designed background patterns by exploring the distribution of information in the frequency domain. Additionally, knowledge distillation strategy is also incorporated to improve the deployment efficiency and generalization capability of our model. A series of experiments show that the proposed method achieves an optimal mAP of 96.1 % for the task of patterned wafer defect detection, which is 2.4 % higher than that of YOLOv11s with only a slight inference time increase (10.0 ms to 10.8 ms), achieving a good balance between detection efficiency and accuracy. Meanwhile, the proposed algorithm also works well for complex scenarios even under limited data training. The proposed method provides a strong tool for intelligent and end-to-end defect inspection of patterned wafers.

{"title":"SR-FABNet: Super-Resolution branch guided Fourier attention detection network for efficient optical inspection of nanoscale wafer defects","authors":"Leisheng Chen , Kai Meng , Hangying Zhang , Junquan Zhou , Peihuang Lou","doi":"10.1016/j.aei.2025.103200","DOIUrl":"10.1016/j.aei.2025.103200","url":null,"abstract":"<div><div>In-line optical inspection of nanoscale defects in patterned wafers is crucial for yield management and control in advanced semiconductor production. Current industrial inspection methods primarily rely on die-to-die or die-to-database comparisons. However, due to the continuously shrinking process nodes, optical methods suffer from limited optical resolution and efficiency loss, presenting significant challenges. Therefore, there is an urgent need for efficient and precise optical inspection methods to detect nanoscale physical defects from low-resolution optical diffraction patterns. To address this gap, we propose a super-resolution (SR) branch guided Fourier attention detection network for efficient optical inspections of nanoscale wafer defects. The network establishes a SR branch to guide the utilization of high-resolution image information for defect detection. Moreover, we introduce a novel Fourier Attention Block (FAB) to enhance the model’s capability of discerning defects from sophisticatedly-designed background patterns by exploring the distribution of information in the frequency domain. Additionally, knowledge distillation strategy is also incorporated to improve the deployment efficiency and generalization capability of our model. A series of experiments show that the proposed method achieves an optimal mAP of 96.1 % for the task of patterned wafer defect detection, which is 2.4 % higher than that of YOLOv11s with only a slight inference time increase (10.0 ms to 10.8 ms), achieving a good balance between detection efficiency and accuracy. Meanwhile, the proposed algorithm also works well for complex scenarios even under limited data training. The proposed method provides a strong tool for intelligent and end-to-end defect inspection of patterned wafers.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103200"},"PeriodicalIF":8.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430141","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

Integrated registration and utility of mobile AR Human-Machine collaborative assembly in rail transit 轨道交通中移动 AR 人机协作装配的综合注册和实用性

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-17 DOI: 10.1016/j.aei.2025.103168

Jiu Yong , Jianguo Wei , Xiaomei Lei , Yangping Wang , Jianwu Dang , Wenhuan Lu

In the Industry 5.0 and digital transformation stage, with human–machine collaboration at the core, an AR human–machine collaboration system can be used to enhance the perception ability of business personnel towards assembly interaction scenarios and objects, effectively improving the efficiency and quality of assembly. However, existing research on the effectiveness of AR human–machine collaborative assembly in more open and flexible rail transit scenarios is uncertain and limited. This article studies the integrated registration and utility of mobile AR human–machine collaborative assembly in rail transit. By proposing a mobile AR End-to-end Integrated Registration Network AR-EIRNet, which is based on AR assembly objects multiscale feure detection, pose inference optimization, and virtual-real fusion rendering, the virtual-real fusion registration application of rail transit AR human–machine collaborative assembly is realized, and then conduct in-depth analysis of AR-EIRNet performance based on RGB-synthesized enhanced images of rail transit, and the effectiveness of AR human–machine collaborative assembly is comprehensively and deeply evaluated via the virtual-real interaction behaviour of the ZD6 switch machine and a subjective questionnaire survey. The experimental results show that in open and complex rail transit scenarios, the registration accuracy of AR-EIRNet reaches 80%, and the assembly time and error count are reduced by 38% and 43% respectively, demonstrating AR-EIRNet has strong robustness and generalizability. AR handheld are more in line with operating habits, but the lack of immersion and interactive experience limits the significant improvement in complex assembly effects. AR glasses with multimodal perception can increase the cognitive load, but moderate cognitive pressure and multimodal interaction are beneficial for stimulating autonomous assembly motivation. The organic combination of two types of AR interactions can effectively meet the practical needs of adaptive, refined, and efficient AR human–machine collaborative digital assembly operations in open and complex rail transit scenarios.

{"title":"Integrated registration and utility of mobile AR Human-Machine collaborative assembly in rail transit","authors":"Jiu Yong , Jianguo Wei , Xiaomei Lei , Yangping Wang , Jianwu Dang , Wenhuan Lu","doi":"10.1016/j.aei.2025.103168","DOIUrl":"10.1016/j.aei.2025.103168","url":null,"abstract":"<div><div>In the Industry 5.0 and digital transformation stage, with human–machine collaboration at the core, an AR human–machine collaboration system can be used to enhance the perception ability of business personnel towards assembly interaction scenarios and objects, effectively improving the efficiency and quality of assembly. However, existing research on the effectiveness of AR human–machine collaborative assembly in more open and flexible rail transit scenarios is uncertain and limited. This article studies the integrated registration and utility of mobile AR human–machine collaborative assembly in rail transit. By proposing a mobile AR End-to-end Integrated Registration Network AR-EIRNet, which is based on AR assembly objects multiscale feure detection, pose inference optimization, and virtual-real fusion rendering, the virtual-real fusion registration application of rail transit AR human–machine collaborative assembly is realized, and then conduct in-depth analysis of AR-EIRNet performance based on RGB-synthesized enhanced images of rail transit, and the effectiveness of AR human–machine collaborative assembly is comprehensively and deeply evaluated via the virtual-real interaction behaviour of the ZD6 switch machine and a subjective questionnaire survey. The experimental results show that in open and complex rail transit scenarios, the registration accuracy of AR-EIRNet reaches 80%, and the assembly time and error count are reduced by 38% and 43% respectively, demonstrating AR-EIRNet has strong robustness and generalizability. AR handheld are more in line with operating habits, but the lack of immersion and interactive experience limits the significant improvement in complex assembly effects. AR glasses with multimodal perception can increase the cognitive load, but moderate cognitive pressure and multimodal interaction are beneficial for stimulating autonomous assembly motivation. The organic combination of two types of AR interactions can effectively meet the practical needs of adaptive, refined, and efficient AR human–machine collaborative digital assembly operations in open and complex rail transit scenarios.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103168"},"PeriodicalIF":8.0,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143421233","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

Harnessing unsupervised learning for retrieving CAD assembly models from public datasets

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-17 DOI: 10.1016/j.aei.2025.103182

Yixuan Li , Jie Zhang , Jiazhen Pang , Ya Yao

Retrieving assembly models from public datasets can yield enriched outcomes and broaden the spectrum of insights. However, public datasets often present unique challenges, such as variance in quality and granularity of assembly models, lack of standardized methods for organizing and labeling, which hinder efficient and accurate retrieval. To address these issues, this paper presents a robust two-step retrieval method tailored for CAD assembly models from public datasets. The first phase utilizes hierarchical clustering in an unsupervised learning framework to systematically organize CAD assembly models. Each assembly model is represented by a feature vector that encapsulates geometrical and topological features derived from its Boundary Representation (B-rep), and reflects hierarchical relationships among parts and components. These feature vectors serve as the basis for systematic indexing via hierarchical clustering, grouping models based on similarity measurement. Each cluster’s centroid, representing the collective feature vector, facilitates efficient and targeted retrieval. In the second phase, the query model is directly compared to cluster centroids, enabling rapid identification of similar assembly collections. To enhance precision within identified clusters, we introduce a fine-grained retrieval technique that integrates Optimal Subsequence Bijection (OSB) with Maximum Mean Discrepancy (MMD). Evaluations on a heterogeneous dataset demonstrate that our method not only streamlines dataset organization but also effectively addresses quality variations, significantly improving retrieval efficiency across extensive collections.

{"title":"Harnessing unsupervised learning for retrieving CAD assembly models from public datasets","authors":"Yixuan Li , Jie Zhang , Jiazhen Pang , Ya Yao","doi":"10.1016/j.aei.2025.103182","DOIUrl":"10.1016/j.aei.2025.103182","url":null,"abstract":"<div><div>Retrieving assembly models from public datasets can yield enriched outcomes and broaden the spectrum of insights. However, public datasets often present unique challenges, such as variance in quality and granularity of assembly models, lack of standardized methods for organizing and labeling, which hinder efficient and accurate retrieval. To address these issues, this paper presents a robust two-step retrieval method tailored for CAD assembly models from public datasets. The first phase utilizes hierarchical clustering in an unsupervised learning framework to systematically organize CAD assembly models. Each assembly model is represented by a feature vector that encapsulates geometrical and topological features derived from its Boundary Representation (B-rep), and reflects hierarchical relationships among parts and components. These feature vectors serve as the basis for systematic indexing via hierarchical clustering, grouping models based on similarity measurement. Each cluster’s centroid, representing the collective feature vector, facilitates efficient and targeted retrieval. In the second phase, the query model is directly compared to cluster centroids, enabling rapid identification of similar assembly collections. To enhance precision within identified clusters, we introduce a fine-grained retrieval technique that integrates Optimal Subsequence Bijection (OSB) with Maximum Mean Discrepancy (MMD). Evaluations on a heterogeneous dataset demonstrate that our method not only streamlines dataset organization but also effectively addresses quality variations, significantly improving retrieval efficiency across extensive collections.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103182"},"PeriodicalIF":8.0,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430137","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

SBDNet: A deep learning-based method for the segmentation and quantification of fatigue cracks in steel bridges

IF 8 1区工程技术 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Advanced Engineering Informatics

Pub Date : 2025-02-17 DOI: 10.1016/j.aei.2025.103186

Xiao Wang , Qingrui Yue , Xiaogang Liu

Employing deep learning to automate the processing of fatigue crack images in steel structure bridges is a cutting-edge research frontier in damage assessment and safe operation. However, existing methods lack pixel-level segmentation accuracy and quantitative metrics for model generalization. This paper introduces Steel Bridge Damage Networks (SBDNet), which is designed for high-precision pixel-level segmentation and quantification of fatigue cracks. We established metrics to measure domain differences between training and test sets and validated SBDNet on a fatigue crack image dataset, comparing its performance with state-of-the-art models. Results show that SBDNet achieves an average IoU of 76.8% and a crack geometric quantification error of less than 3%, exhibiting robust generalization. The proposed method enhances damage detection efficiency and provides quantitative references for maintenance decision-making.

引用次数: 0