Advanced Engineering Informatics最新文献_第8页

Automatic identification of integrated construction elements using open-set object detection based on image and text modality fusion

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

Advanced Engineering Informatics

Pub Date : 2025-01-06 DOI: 10.1016/j.aei.2024.103075

Ruying Cai , Zhigang Guo , Xiangsheng Chen , Jingru Li , Yi Tan , Jingyuan Tang

The application of object detection technology in the field of construction safety contributes significantly to on-site safety management and has already shown considerable progress. However, current research primarily focuses on detecting pre-defined classes annotated within single datasets. In-depth research in construction safety requires the detection of all influencing factors related to construction safety. The emergence of large language models offers new possibilities, and multimodal models that combine these with computer vision technology could break through the existing limitations. Therefore, this paper proposes the Grounding DINO multimodal model for the automatic detection of integrated construction elements, enhancing construction safety. First, this study reviews the literature to collect relevant datasets, summarizes their characteristics, and processes the data, including the processing of annotation files and the integration of classes. Subsequently, the Grounding DINO model is constructed, encompassing image and text feature extraction and enhancement, and a cross-modal decoder that fuses image and text features. Multiple dataset experimental strategies are designed to validate Grounding DINO’s capabilities in continuous learning, with a unified class system created based on integrated classes for model detection input text prompts. Finally, experiments involving zero-shot and fine-tuning evaluations, continuous learning validation, and effectiveness testing are conducted. The experimental results demonstrate the generalization capability and potential for continuous learning of the multimodal model.

{"title":"Automatic identification of integrated construction elements using open-set object detection based on image and text modality fusion","authors":"Ruying Cai , Zhigang Guo , Xiangsheng Chen , Jingru Li , Yi Tan , Jingyuan Tang","doi":"10.1016/j.aei.2024.103075","DOIUrl":"10.1016/j.aei.2024.103075","url":null,"abstract":"<div><div>The application of object detection technology in the field of construction safety contributes significantly to on-site safety management and has already shown considerable progress. However, current research primarily focuses on detecting pre-defined classes annotated within single datasets. In-depth research in construction safety requires the detection of all influencing factors related to construction safety. The emergence of large language models offers new possibilities, and multimodal models that combine these with computer vision technology could break through the existing limitations. Therefore, this paper proposes the Grounding DINO multimodal model for the automatic detection of integrated construction elements, enhancing construction safety. First, this study reviews the literature to collect relevant datasets, summarizes their characteristics, and processes the data, including the processing of annotation files and the integration of classes. Subsequently, the Grounding DINO model is constructed, encompassing image and text feature extraction and enhancement, and a cross-modal decoder that fuses image and text features. Multiple dataset experimental strategies are designed to validate Grounding DINO’s capabilities in continuous learning, with a unified class system created based on integrated classes for model detection input text prompts. Finally, experiments involving zero-shot and fine-tuning evaluations, continuous learning validation, and effectiveness testing are conducted. The experimental results demonstrate the generalization capability and potential for continuous learning of the multimodal model.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103075"},"PeriodicalIF":8.0,"publicationDate":"2025-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129695","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

You can be more trustworthy: A feature fusion reinforcement network for credible anti-noise fault diagnosis

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

Advanced Engineering Informatics

Pub Date : 2025-01-06 DOI: 10.1016/j.aei.2024.103056

Yuan Wei , Hongchong Peng , Mansong Rong , Xiaohui Gu , Xiangyan Chen

Significant research progress has been made in intelligent fault diagnosis algorithms. However, these methods face challenges such as noise interference and untrustworthy diagnostic results in industrial practice, which limit their performance in practical applications. This paper proposes a new feature fusion and reinforcement network combined with swin-transformer (Swin-FFRN) for noise-resistant and trustworthy diagnosis, which combines a global feature extraction network and a staged convolutional fusion operation for fine-grained fault feature extraction and noise suppression in 2D time–frequency maps. The Swin-FFRN is used to analyze 2D time–frequency map data of different mechanical faults in a low signal-to-noise ratio environment by introducing a channel attention mechanism and a spatial attention mechanism to strengthen the critical fault features that are strongly correlated with the classification of the faults so that the model focuses on the crucial features. Moreover, the noise immunity performance is evaluated using the latest methods on two different datasets, and intuitive visual interpretability is provided to show model credibility. The results show that the noise-resistant diagnostic accuracy of the proposed method is improved by 5.43% on average with respect to the SOTA method. By enhancing the key input features, the proposed method can give diagnostic results with a reasonable decision basis.

{"title":"You can be more trustworthy: A feature fusion reinforcement network for credible anti-noise fault diagnosis","authors":"Yuan Wei , Hongchong Peng , Mansong Rong , Xiaohui Gu , Xiangyan Chen","doi":"10.1016/j.aei.2024.103056","DOIUrl":"10.1016/j.aei.2024.103056","url":null,"abstract":"<div><div>Significant research progress has been made in intelligent fault diagnosis algorithms. However, these methods face challenges such as noise interference and untrustworthy diagnostic results in industrial practice, which limit their performance in practical applications. This paper proposes a new feature fusion and reinforcement network combined with swin-transformer (Swin-FFRN) for noise-resistant and trustworthy diagnosis, which combines a global feature extraction network and a staged convolutional fusion operation for fine-grained fault feature extraction and noise suppression in 2D time–frequency maps. The Swin-FFRN is used to analyze 2D time–frequency map data of different mechanical faults in a low signal-to-noise ratio environment by introducing a channel attention mechanism and a spatial attention mechanism to strengthen the critical fault features that are strongly correlated with the classification of the faults so that the model focuses on the crucial features. Moreover, the noise immunity performance is evaluated using the latest methods on two different datasets, and intuitive visual interpretability is provided to show model credibility. The results show that the noise-resistant diagnostic accuracy of the proposed method is improved by 5.43% on average with respect to the SOTA method. By enhancing the key input features, the proposed method can give diagnostic results with a reasonable decision basis.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103056"},"PeriodicalIF":8.0,"publicationDate":"2025-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129697","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

Target-driven dynamic coverage planning method for marsupial cluster system

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

Advanced Engineering Informatics

Pub Date : 2025-01-06 DOI: 10.1016/j.aei.2024.103071

Zhiyao Lu, Chongyu Liang, Chen Bai, Weichao Wu, Aigang Pan

Using marsupial unmanned cluster systems can significantly improve underwater unexploded ordnance (UXO) clearance through strategic planning. This study examines the planning method for these systems. Current geographic information databases provide limited insights on UXO targets, and neither coverage path planning (CPP) nor multi-robot task allocation (MRTA) alone can effectively tackle UXO clearance complexities. A target-driven planning approach is proposed to enhance the system’s performance by utilizing known target information while ensuring adequate area coverage. A multi-agent decision rule is proposed, focusing on pre-planning and agent empathy to assign new targets in scenarios with limited communication effectively. These two aspects form a target-driven dynamic coverage planning method, with simulation experiments designed to compare the time required for UXO clearance across various planning methods. The most important new thing that this study adds is a new planning method tailored to the marsupial cluster system. This method increases the effectiveness of removing underwater UXO by 0.86% to 8.96% when the target known rate is above 30.3%. In addition, the simulation results indicate a direct correlation between the utilization of known information and system efficiency improvements. The article can also further support that the more information is known, the more intelligent planning methods make sense.

{"title":"Target-driven dynamic coverage planning method for marsupial cluster system","authors":"Zhiyao Lu, Chongyu Liang, Chen Bai, Weichao Wu, Aigang Pan","doi":"10.1016/j.aei.2024.103071","DOIUrl":"10.1016/j.aei.2024.103071","url":null,"abstract":"<div><div>Using marsupial unmanned cluster systems can significantly improve underwater unexploded ordnance (UXO) clearance through strategic planning. This study examines the planning method for these systems. Current geographic information databases provide limited insights on UXO targets, and neither coverage path planning (CPP) nor multi-robot task allocation (MRTA) alone can effectively tackle UXO clearance complexities. A target-driven planning approach is proposed to enhance the system’s performance by utilizing known target information while ensuring adequate area coverage. A multi-agent decision rule is proposed, focusing on pre-planning and agent empathy to assign new targets in scenarios with limited communication effectively. These two aspects form a target-driven dynamic coverage planning method, with simulation experiments designed to compare the time required for UXO clearance across various planning methods. The most important new thing that this study adds is a new planning method tailored to the marsupial cluster system. This method increases the effectiveness of removing underwater UXO by 0.86% to 8.96% when the target known rate is above 30.3%. In addition, the simulation results indicate a direct correlation between the utilization of known information and system efficiency improvements. The article can also further support that the more information is known, the more intelligent planning methods make sense.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103071"},"PeriodicalIF":8.0,"publicationDate":"2025-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129696","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

A supervised contrastive learning method based on online complement strategy for long-tailed fine-grained fault diagnosis

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

Advanced Engineering Informatics

Pub Date : 2025-01-04 DOI: 10.1016/j.aei.2024.103079

Zhiqian Zhao , Yinghou Jiao , Yeyin Xu , Runchao Zhao

As industrial automation and intelligence advance, equipment complexity rises, leading to diverse fault patterns. In fine-grained fault diagnosis, sample scarcity causes a significant long-tail effect, where main fault categories dominate. High intra-class variance and inter-class similarity in fine-grained categories impede the performance of traditional supervised contrastive learning, particularly for underrepresented tail categories in feature space. To address the above problems, a novel supervised contrast learning method for long-tailed fine-grained fault diagnosis, OC-SupCon, is proposed to improve the feature representations through the online complement strategy. Supervised contrastive learning is used as the model framework to ensure that each batch contains the inherent features of all fine-grained categories by introducing a class-centered prototype. Then, data augmentation is dynamically complemented by assessing the neighborhood sparsity of the samples to reduce the unfavorable influence on the features of the tail categories. Finally, the dominance of the head category is mitigated by balancing the gradient contributions of different fine-grained categories. In addition, Logit compensation technique is used in the classifier branch to adjust the category boundaries, and the class center prototypes are dynamically updated during the training process. The experimental results show that the proposed method exhibits significant performance in long-tailed fine-grained fault diagnosis tasks compared to existing state-of-the-art methods. The code is available from https://github.com/zhiqan/OC-Supcon.

{"title":"A supervised contrastive learning method based on online complement strategy for long-tailed fine-grained fault diagnosis","authors":"Zhiqian Zhao , Yinghou Jiao , Yeyin Xu , Runchao Zhao","doi":"10.1016/j.aei.2024.103079","DOIUrl":"10.1016/j.aei.2024.103079","url":null,"abstract":"<div><div>As industrial automation and intelligence advance, equipment complexity rises, leading to diverse fault patterns. In fine-grained fault diagnosis, sample scarcity causes a significant long-tail effect, where main fault categories dominate. High intra-class variance and inter-class similarity in fine-grained categories impede the performance of traditional supervised contrastive learning, particularly for underrepresented tail categories in feature space. To address the above problems, a novel supervised contrast learning method for long-tailed fine-grained fault diagnosis, OC-SupCon, is proposed to improve the feature representations through the online complement strategy. Supervised contrastive learning is used as the model framework to ensure that each batch contains the inherent features of all fine-grained categories by introducing a class-centered prototype. Then, data augmentation is dynamically complemented by assessing the neighborhood sparsity of the samples to reduce the unfavorable influence on the features of the tail categories. Finally, the dominance of the head category is mitigated by balancing the gradient contributions of different fine-grained categories. In addition, Logit compensation technique is used in the classifier branch to adjust the category boundaries, and the class center prototypes are dynamically updated during the training process. The experimental results show that the proposed method exhibits significant performance in long-tailed fine-grained fault diagnosis tasks compared to existing state-of-the-art methods. The code is available from <span><span>https://github.com/zhiqan/OC-Supcon</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103079"},"PeriodicalIF":8.0,"publicationDate":"2025-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129693","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

Multi-scale dynamic graph mutual information network for planet bearing health monitoring under imbalanced data

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

Advanced Engineering Informatics

Pub Date : 2025-01-04 DOI: 10.1016/j.aei.2024.103096

Wenbin Cai , Dezun Zhao , Tianyang Wang

In engineering, imbalanced data collected from planet bearings causes most intelligent models to shrink the decision boundary of minor classes and degrade diagnostic accuracy. Different from these models under the assumption of data balance, graph-based methods focus on the relationship between data to alleviate the issue of data imbalance, but they have restrictions on single-feature propagation and only rely on the feature extraction capability of convolutional operations. As such, a multi-scale dynamic graph mutual information network (MDGMIN) is proposed for the health monitoring of planet bearings with imbalanced data. First, a dual spatial–temporal graph generation algorithm is designed to construct dynamic and distance graphs via the gated convolution in the temporal dimension and the cosine similarity and Top-k sorting mechanism in the spatial dimension. Second, multi-scale dynamic edge graph convolutional layers are constructed to extract specific and similar features, and they are weighted fused via an attention mechanism. Finally, mutual information learning is developed to foster the model in capturing graph features in-depth through commonality and discrepancy constraints, and a new loss-driven function based on two constraints is proposed to update the training objective. Experimental analysis on an imbalanced planet bearing dataset verifies that the developed MDGMIN reaches the diagnostic accuracy of 92.80%, exceeding that of state-of-the-art methods on the dataset with an imbalanced ratio of 20:1. In addition, the generalizability of the MDGMIN is validated in another bearing dataset from the planetary gearbox.

{"title":"Multi-scale dynamic graph mutual information network for planet bearing health monitoring under imbalanced data","authors":"Wenbin Cai , Dezun Zhao , Tianyang Wang","doi":"10.1016/j.aei.2024.103096","DOIUrl":"10.1016/j.aei.2024.103096","url":null,"abstract":"<div><div>In engineering, imbalanced data collected from planet bearings causes most intelligent models to shrink the decision boundary of minor classes and degrade diagnostic accuracy. Different from these models under the assumption of data balance, graph-based methods focus on the relationship between data to alleviate the issue of data imbalance, but they have restrictions on single-feature propagation and only rely on the feature extraction capability of convolutional operations. As such, a multi-scale dynamic graph mutual information network (MDGMIN) is proposed for the health monitoring of planet bearings with imbalanced data. First, a dual spatial–temporal graph generation algorithm is designed to construct dynamic and distance graphs via the gated convolution in the temporal dimension and the cosine similarity and Top-k sorting mechanism in the spatial dimension. Second, multi-scale dynamic edge graph convolutional layers are constructed to extract specific and similar features, and they are weighted fused via an attention mechanism. Finally, mutual information learning is developed to foster the model in capturing graph features in-depth through commonality and discrepancy constraints, and a new loss-driven function based on two constraints is proposed to update the training objective. Experimental analysis on an imbalanced planet bearing dataset verifies that the developed MDGMIN reaches the diagnostic accuracy of 92.80%, exceeding that of state-of-the-art methods on the dataset with an imbalanced ratio of 20:1. In addition, the generalizability of the MDGMIN is validated in another bearing dataset from the planetary gearbox.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103096"},"PeriodicalIF":8.0,"publicationDate":"2025-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129627","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

Neural ODE powered model for bearing remaining useful life predictions with intra- and inter-domain shifts

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

Advanced Engineering Informatics

Pub Date : 2025-01-04 DOI: 10.1016/j.aei.2024.103077

Tao Hu, Zhenling Mo, Zijun Zhang

In bearing remaining useful life (RUL) predictions, current domain adaptation (DA) and domain generalization (DG) methods are typically concerned with mitigating inter-domain shifts (DSs)—a type of DSs existing across the bearing degradation data sequences. Yet, intra-DSs along the bearing degradation data sequences, which are another type of DSs governing inter-DSs, have not attracted sufficient attention, thus hindering the applicability of existing methods. Moreover, many existing DG methods are developed based on multi-source domains, while bearing RUL predictions in reality often expect models of single-source DG capability. This study investigates the potential of the neural ordinary differential equation (ODE) for filling the aforementioned research gaps, leading to a novel neural ODE powered modeling (NOMI) scheme. First, the ODE characteristic of time invariance is utilized to address intra-DSs for learning time-invariant latent features from a single source bearing degradation data domain. Then, the gained time consistency could reduce heterogeneous intra-DS patterns, thereby decreasing inter-DSs and promoting model generalizability. The designed ODE module can be conveniently employed under DA and DG scenarios. Additionally, with a further gradient manipulation technique, the proposed model can be trained efficiently. Theoretical analyses demonstrate the benefits of intra-domain minimization for solving the data distribution problem. The experimental results based on multiple bearing datasets also verify the superiority of our proposed method compared with state-of-the-art approaches.

{"title":"Neural ODE powered model for bearing remaining useful life predictions with intra- and inter-domain shifts","authors":"Tao Hu, Zhenling Mo, Zijun Zhang","doi":"10.1016/j.aei.2024.103077","DOIUrl":"10.1016/j.aei.2024.103077","url":null,"abstract":"<div><div>In bearing remaining useful life (RUL) predictions, current domain adaptation (DA) and domain generalization (DG) methods are typically concerned with mitigating inter-domain shifts (DSs)—a type of DSs existing across the bearing degradation data sequences. Yet, intra-DSs along the bearing degradation data sequences, which are another type of DSs governing inter-DSs, have not attracted sufficient attention, thus hindering the applicability of existing methods. Moreover, many existing DG methods are developed based on multi-source domains, while bearing RUL predictions in reality often expect models of single-source DG capability. This study investigates the potential of the neural ordinary differential equation (ODE) for filling the aforementioned research gaps, leading to a novel neural ODE powered modeling (NOMI) scheme. First, the ODE characteristic of time invariance is utilized to address intra-DSs for learning time-invariant latent features from a single source bearing degradation data domain. Then, the gained time consistency could reduce heterogeneous intra-DS patterns, thereby decreasing inter-DSs and promoting model generalizability. The designed ODE module can be conveniently employed under DA and DG scenarios. Additionally, with a further gradient manipulation technique, the proposed model can be trained efficiently. Theoretical analyses demonstrate the benefits of intra-domain minimization for solving the data distribution problem. The experimental results based on multiple bearing datasets also verify the superiority of our proposed method compared with state-of-the-art approaches.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103077"},"PeriodicalIF":8.0,"publicationDate":"2025-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129694","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

LLMOA: A novel large language model assisted hyper-heuristic optimization algorithm

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

Advanced Engineering Informatics

Pub Date : 2025-01-03 DOI: 10.1016/j.aei.2024.103042

Rui Zhong , Abdelazim G. Hussien , Jun Yu , Masaharu Munetomo

This work presents a novel approach, the large language model assisted hyper-heuristic optimization algorithm (LLMOA), tailored to address complex optimization challenges. Comprising two essential components – the high-level component and the low-level component – LLMOA leverages the LLM (i.e., Gemini) with prompt engineering in its high-level component to construct optimization sequences automatically and intelligently. Furthermore, we propose novel elite-based local search operators as low-level heuristics (LLHs), which draw inspiration from the proximate optimality principle (POP). These local search operators cooperated with well-known mutation and crossover operators from differential evolution (DE), at a total of ten efficient and versatile search operators, forming the whole LLHs. To assess the competitiveness of LLMOA, we conducted comprehensive numerical experiments across CEC2014, CEC2020, CEC2022, and ten engineering optimization problems, benchmarking against eleven state-of-the-art optimizers. Our experimental findings and statistical analyses underscore the powerfulness and effectiveness of LLMOA. Moreover, ablation experiments reveal the pivotal role of integrating the LLM Gemini and prompt engineering as the high-level component. Conclusively, this study provides a feasible avenue to introduce LLM to the evolutionary computation (EC) community. The research’s source code is available for download at https://github.com/RuiZhong961230/LLMOA.

{"title":"LLMOA: A novel large language model assisted hyper-heuristic optimization algorithm","authors":"Rui Zhong , Abdelazim G. Hussien , Jun Yu , Masaharu Munetomo","doi":"10.1016/j.aei.2024.103042","DOIUrl":"10.1016/j.aei.2024.103042","url":null,"abstract":"<div><div>This work presents a novel approach, the large language model assisted hyper-heuristic optimization algorithm (LLMOA), tailored to address complex optimization challenges. Comprising two essential components – the high-level component and the low-level component – LLMOA leverages the LLM (i.e., Gemini) with prompt engineering in its high-level component to construct optimization sequences automatically and intelligently. Furthermore, we propose novel elite-based local search operators as low-level heuristics (LLHs), which draw inspiration from the proximate optimality principle (POP). These local search operators cooperated with well-known mutation and crossover operators from differential evolution (DE), at a total of ten efficient and versatile search operators, forming the whole LLHs. To assess the competitiveness of LLMOA, we conducted comprehensive numerical experiments across CEC2014, CEC2020, CEC2022, and ten engineering optimization problems, benchmarking against eleven state-of-the-art optimizers. Our experimental findings and statistical analyses underscore the powerfulness and effectiveness of LLMOA. Moreover, ablation experiments reveal the pivotal role of integrating the LLM Gemini and prompt engineering as the high-level component. Conclusively, this study provides a feasible avenue to introduce LLM to the evolutionary computation (EC) community. The research’s source code is available for download at <span><span>https://github.com/RuiZhong961230/LLMOA</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103042"},"PeriodicalIF":8.0,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129691","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

Product carbon emissions estimation method in the early design stage based on multi-perspective similarity matching of design scenarios

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

Advanced Engineering Informatics

Pub Date : 2025-01-03 DOI: 10.1016/j.aei.2024.103094

Lin Kong , Yanyan Nie , Liming Wang , Fangyi Li , Lirong Zhou , Geng Wang , Haiyang Lu , Xingyuan Xiao , Weitong Liu , Yan Ma

Carbon emissions estimation of design schemes during the early design stage enables thorough consideration of environmental issues from the source, which holds critical significance for carbon reduction and emission mitigation. Nevertheless, the scarcity of life cycle inventory data, coupled with the intricacies involved in the collection, presents a formidable challenge to conducting precise carbon emissions assessment. To address this issue, this research proposes a product carbon emissions estimation method in the early design stage based on multi-perspective similarity matching of design scenarios, which utilizes the idea of knowledge reuse through case-based reasoning. Specifically, the case-based reasoning framework encompassing case base construction, case retrieval, reuse, and revision has been outlined, which standards the procedure for obtaining the most similar case. Moreover, the design scenario is defined to comprehensively describe all life cycle activities that influence product carbon emissions, and the design scenario-based multi-layer model is constructed that encompasses the product’s lifecycle-related design information pertinent to carbon emissions, along with its intricate interrelationships, serving as the input information for precise case retrieval. Subsequently, a multi-perspective similarity matching strategy that integrates both the attribute and correlation information of design scenarios is developed, which accurately identifies the most similar case in the case base, enabling the efficient reuse of historical data. An example of the wind turbine gearbox is given as an example, the results indicating that the proposed carbon emission estimation method aligns most closely with actual machining conditions, achieving a minimal error of 2.75%, thereby unequivocally validating its effectiveness and reliability. This work provides designers with a targeted strategy for obtaining carbon emissions during the early design stage, thereby facilitating optimized decision-making for low-carbon design.

{"title":"Product carbon emissions estimation method in the early design stage based on multi-perspective similarity matching of design scenarios","authors":"Lin Kong , Yanyan Nie , Liming Wang , Fangyi Li , Lirong Zhou , Geng Wang , Haiyang Lu , Xingyuan Xiao , Weitong Liu , Yan Ma","doi":"10.1016/j.aei.2024.103094","DOIUrl":"10.1016/j.aei.2024.103094","url":null,"abstract":"<div><div>Carbon emissions estimation of design schemes during the early design stage enables thorough consideration of environmental issues from the source, which holds critical significance for carbon reduction and emission mitigation. Nevertheless, the scarcity of life cycle inventory data, coupled with the intricacies involved in the collection, presents a formidable challenge to conducting precise carbon emissions assessment. To address this issue, this research proposes a product carbon emissions estimation method in the early design stage based on multi-perspective similarity matching of design scenarios, which utilizes the idea of knowledge reuse through case-based reasoning. Specifically, the case-based reasoning framework encompassing case base construction, case retrieval, reuse, and revision has been outlined, which standards the procedure for obtaining the most similar case. Moreover, the design scenario is defined to comprehensively describe all life cycle activities that influence product carbon emissions, and the design scenario-based multi-layer model is constructed that encompasses the product’s lifecycle-related design information pertinent to carbon emissions, along with its intricate interrelationships, serving as the input information for precise case retrieval. Subsequently, a multi-perspective similarity matching strategy that integrates both the attribute and correlation information of design scenarios is developed, which accurately identifies the most similar case in the case base, enabling the efficient reuse of historical data. An example of the wind turbine gearbox is given as an example, the results indicating that the proposed carbon emission estimation method aligns most closely with actual machining conditions, achieving a minimal error of 2.75%, thereby unequivocally validating its effectiveness and reliability. This work provides designers with a targeted strategy for obtaining carbon emissions during the early design stage, thereby facilitating optimized decision-making for low-carbon design.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103094"},"PeriodicalIF":8.0,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129626","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

A rollout heuristic-reinforcement learning hybrid algorithm for disassembly sequence planning with uncertain depreciation condition and diversified recovering strategies

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

Advanced Engineering Informatics

Pub Date : 2025-01-03 DOI: 10.1016/j.aei.2024.103082

Yaping Ren , Zhehao Xu , Yanzi Zhang , Jiayi Liu , Leilei Meng , Wenwen Lin

Disassembly is one of the crucial aspects of green manufacturing. For the end-of-life products, an effective disassembly sequence planning method can enhance recovery value and mitigate the negative consequences of resource depletion and waste generation. However, both the uncertainty of product depreciation condition and the NP-hard characteristics (including the determination of disassembly sequences and the selection of recovering strategies of subassemblies) of the disassembly sequence planning results in difficulties to determine the optimal/near-optimal disassembly solutions. To address these challenges, this work establishes an extended Petri net that considers diversified recovering strategies of each subassembly caused by uncertain product depreciation condition. Then, a rollout heuristic-reinforcement learning hybrid algorithm that integrates a rollout decision rule into the reinforcement learning procedure is proposed to rapidly find the high-quality disassembly solutions based on the extended Petri net, in which the uncertainty of disassembly information is tackled by training disassembly samples and the global exploration capability of the learning procedure is significantly improved by using the rollout decision rule. Finally, three products with different complexities and sizes are used to verify the performance of the proposed algorithm, and the experimental results indicate that our proposed rollout heuristic-reinforcement learning hybrid algorithm can efficiently compute the high-quality disassembly sequences under various disassembly environments.

{"title":"A rollout heuristic-reinforcement learning hybrid algorithm for disassembly sequence planning with uncertain depreciation condition and diversified recovering strategies","authors":"Yaping Ren , Zhehao Xu , Yanzi Zhang , Jiayi Liu , Leilei Meng , Wenwen Lin","doi":"10.1016/j.aei.2024.103082","DOIUrl":"10.1016/j.aei.2024.103082","url":null,"abstract":"<div><div>Disassembly is one of the crucial aspects of green manufacturing. For the end-of-life products, an effective disassembly sequence planning method can enhance recovery value and mitigate the negative consequences of resource depletion and waste generation. However, both the uncertainty of product depreciation condition and the NP-hard characteristics (including the determination of disassembly sequences and the selection of recovering strategies of subassemblies) of the disassembly sequence planning results in difficulties to determine the optimal/near-optimal disassembly solutions. To address these challenges, this work establishes an extended Petri net that considers diversified recovering strategies of each subassembly caused by uncertain product depreciation condition. Then, a rollout heuristic-reinforcement learning hybrid algorithm that integrates a rollout decision rule into the reinforcement learning procedure is proposed to rapidly find the high-quality disassembly solutions based on the extended Petri net, in which the uncertainty of disassembly information is tackled by training disassembly samples and the global exploration capability of the learning procedure is significantly improved by using the rollout decision rule. Finally, three products with different complexities and sizes are used to verify the performance of the proposed algorithm, and the experimental results indicate that our proposed rollout heuristic-reinforcement learning hybrid algorithm can efficiently compute the high-quality disassembly sequences under various disassembly environments.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 103082"},"PeriodicalIF":8.0,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129690","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

A self-supervised masked spatial distribution learning method for predicting machinery remaining useful life with missing data reconstruction

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

Advanced Engineering Informatics

Pub Date : 2025-01-02 DOI: 10.1016/j.aei.2024.102938

Ben Niu , Yi Xiao , Qinge Xiao , Yang Liu , Tao Peng , Zhile Yang

Accurately predicting the remaining useful life (RUL) of machines is vital for assessing machine health and minimizing economic losses resulting from downtime in sensor-equipped machines. However, real-world applications often encounter challenges such as rapid production cycles and unstable network conditions, inevitably leading to significant amounts of missing data. This challenges data-driven machinery RUL prediction, as conventional deep learning methods may struggle with missing data, impacting prediction accuracy. To address the issue, a missing data reconstruction method based on self-learning of mask spatial distribution is proposed. The structured spatial distribution characteristics of the mask, learned by the autoencoder, serve as self-supervised information for the imputation network to improve the data reconstruction performance. Meanwhile, a multi-task learning-enhanced prediction network architecture with adaptive weight adjustment is designed, defining tasks by RUL prediction under different data reconstruction accuracies. After pre-training on multiple tasks, the prediction network’s learning efficiency benefits from incorporating both common and task-specific rules for feature extraction from similar reconstructed data distributions. The proposed method is evaluated through ablation and comparative tests on application scenarios and standard datasets. Experimental results show that the proposed algorithm performs competitively against state-of-the-art data reconstruction algorithms on these test suites.

{"title":"A self-supervised masked spatial distribution learning method for predicting machinery remaining useful life with missing data reconstruction","authors":"Ben Niu , Yi Xiao , Qinge Xiao , Yang Liu , Tao Peng , Zhile Yang","doi":"10.1016/j.aei.2024.102938","DOIUrl":"10.1016/j.aei.2024.102938","url":null,"abstract":"<div><div>Accurately predicting the remaining useful life (RUL) of machines is vital for assessing machine health and minimizing economic losses resulting from downtime in sensor-equipped machines. However, real-world applications often encounter challenges such as rapid production cycles and unstable network conditions, inevitably leading to significant amounts of missing data. This challenges data-driven machinery RUL prediction, as conventional deep learning methods may struggle with missing data, impacting prediction accuracy. To address the issue, a missing data reconstruction method based on self-learning of mask spatial distribution is proposed. The structured spatial distribution characteristics of the mask, learned by the autoencoder, serve as self-supervised information for the imputation network to improve the data reconstruction performance. Meanwhile, a multi-task learning-enhanced prediction network architecture with adaptive weight adjustment is designed, defining tasks by RUL prediction under different data reconstruction accuracies. After pre-training on multiple tasks, the prediction network’s learning efficiency benefits from incorporating both common and task-specific rules for feature extraction from similar reconstructed data distributions. The proposed method is evaluated through ablation and comparative tests on application scenarios and standard datasets. Experimental results show that the proposed algorithm performs competitively against state-of-the-art data reconstruction algorithms on these test suites.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"64 ","pages":"Article 102938"},"PeriodicalIF":8.0,"publicationDate":"2025-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143129741","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