Advanced Engineering Informatics最新文献_第6页

Graph Neural Networks for building and civil infrastructure operation and maintenance enhancement 图神经网络用于建筑和民用基础设施的运行和维护改进

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102868

Sajith Wettewa, Lei Hou, Guomin Zhang

This systematic review, conducted within the PRISMA framework, investigates the disruptive capabilities of Graph Neural Networks (GNNs) in optimising Operations and Maintenance (OM) practices within the building and civil infrastructure domain. Addressing 5 research questions and encompassing 111 studies from 2014 to 2024, our study identifies the multifaceted applications of GNNs across different project stages from data enhancement to operational scenario enhancement. When considering integrated Facilities Management (FM) approaches, GNNs are employed for data enhancement purposes, leveraging techniques such as semantic enrichment of Building Information Modelling (BIM), various data imputation scenarios, and semantic segmentation of point clouds to enhance data quality and completeness. Operational scenarios involve the utilisation of GNN algorithms for anomaly detection, fault classification, system optimisation, and forecasting. Methodological optimisations crucial for GNN feasibility include feature engineering, architecture optimisation to balance complexity and overfitting risk, and the integration of Explainable Artificial Intelligence (XAI) methods to enhance model validity and trust. Physical principles integration through Physics-Informed Graph Neural Networks (PIGNNs) further enhances model explainability and validation. Future research directions focus on data interoperability enhancement, scalability improvements, and explainability enhancements. Automated graph generation and labelling, heterogeneous GNN models, supporting algorithms such as Long Short-Term Memory (LSTM) and reinforcement learning are proposed to overcome analysis limitations. Specific workflows targeting building performance-based semantic enrichment, building systems data imputation, and interdependency prediction are proposed in future directions. The review highlights the symbiotic relationship between GNN-based analysis and digital twin data analysis, emphasising the suitability of GNNs in addressing the demands of digital twin data analysis in the building and civil infrastructure domain.

本系统综述在 PRISMA 框架内进行，调查了图神经网络（GNN）在优化建筑和民用基础设施领域的运营和维护（OM）实践中的颠覆性能力。我们的研究解决了 5 个研究问题，涵盖了从 2014 年到 2024 年的 111 项研究，确定了图神经网络在从数据增强到操作场景增强等不同项目阶段的多方面应用。在考虑综合设施管理（FM）方法时，GNN 被用于数据增强目的，利用建筑信息模型（BIM）的语义丰富、各种数据估算方案和点云语义分割等技术来提高数据质量和完整性。运行场景包括利用 GNN 算法进行异常检测、故障分类、系统优化和预测。对 GNN 可行性至关重要的方法优化包括特征工程、架构优化以平衡复杂性和过拟合风险，以及集成可解释人工智能 (XAI) 方法以增强模型的有效性和可信度。通过物理信息图神经网络（PIGNN）整合物理原理，可进一步增强模型的可解释性和验证性。未来的研究方向主要集中在数据互操作性增强、可扩展性改进和可解释性增强等方面。为了克服分析的局限性，我们提出了自动图生成和标记、异构 GNN 模型、长短期记忆（LSTM）和强化学习等支持算法。在未来的发展方向中，还提出了针对基于建筑性能的语义丰富、建筑系统数据估算和相互依存预测的具体工作流程。综述强调了基于 GNN 的分析与数字孪生数据分析之间的共生关系，强调了 GNN 在满足建筑与民用基础设施领域数字孪生数据分析需求方面的适用性。

{"title":"Graph Neural Networks for building and civil infrastructure operation and maintenance enhancement","authors":"Sajith Wettewa, Lei Hou, Guomin Zhang","doi":"10.1016/j.aei.2024.102868","DOIUrl":"10.1016/j.aei.2024.102868","url":null,"abstract":"<div><div>This systematic review, conducted within the PRISMA framework, investigates the disruptive capabilities of Graph Neural Networks (GNNs) in optimising Operations and Maintenance (OM) practices within the building and civil infrastructure domain. Addressing 5 research questions and encompassing 111 studies from 2014 to 2024, our study identifies the multifaceted applications of GNNs across different project stages from data enhancement to operational scenario enhancement. When considering integrated Facilities Management (FM) approaches, GNNs are employed for data enhancement purposes, leveraging techniques such as semantic enrichment of Building Information Modelling (BIM), various data imputation scenarios, and semantic segmentation of point clouds to enhance data quality and completeness. Operational scenarios involve the utilisation of GNN algorithms for anomaly detection, fault classification, system optimisation, and forecasting. Methodological optimisations crucial for GNN feasibility include feature engineering, architecture optimisation to balance complexity and overfitting risk, and the integration of Explainable Artificial Intelligence (XAI) methods to enhance model validity and trust. Physical principles integration through Physics-Informed Graph Neural Networks (PIGNNs) further enhances model explainability and validation. Future research directions focus on data interoperability enhancement, scalability improvements, and explainability enhancements. Automated graph generation and labelling, heterogeneous GNN models, supporting algorithms such as Long Short-Term Memory (LSTM) and reinforcement learning are proposed to overcome analysis limitations. Specific workflows targeting building performance-based semantic enrichment, building systems data imputation, and interdependency prediction are proposed in future directions. The review highlights the symbiotic relationship between GNN-based analysis and digital twin data analysis, emphasising the suitability of GNNs in addressing the demands of digital twin data analysis in the building and civil infrastructure domain.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102868"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142532179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A conflict clique mitigation method for large-scale satellite mission planning based on heterogeneous graph learning 基于异构图学习的大规模卫星任务规划冲突聚类缓解方法

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102915

Xiaoen Feng, Minqiang Xu, Yuqing Li

For the large-scale and intensive demands of satellite remote sensing observation, the complexity of constraint relationships grows explosively with expansion of satellite task scale. How to efficiently deal with the complex and temporal varying constraint conflicts, and mine the implicit knowledge existing among satellite mission constraints, which is significant to enhance scheduling efficiency, however, is also a core difficulty in the satellite scheduling problem. In this paper, we propose a conflict clique mitigation method based on dynamic task-constrained heterogeneous graph learning to solve large-scale satellite mission scheduling. The method exploits the advantage of heterogeneous graphs to characterize multiple unstructured relationships, and projects the temporal-varying features of constraint conflicts to the spatial topology of multiple nodes and edges in a heterogeneous graph. Thus, a dynamic constraints heterogeneous graph model for satellite tasks based on sampling critical conflict cliques is developed. And an improved heterogeneous attention network with quadratic unconstrained binary optimization (HAN-QUBO) is proposed, which is able to deal with the heterogeneous graphs and attempts to represent the implicit principles of multiple constraints of satellite missions, so that the valuable strategies and experiences of conflict mitigation can be extracted. The simulation experiments demonstrate that the method can provide effective empirical guidance for multi-satellite scheduling, greatly relieve the pressure of cumbersome constraint conflict checking process for large-scale tasks. The average number of conflict resolutions has been reduced by about 73.48 % for EOSSPs with tens of thousands tasks, while the quality of solutions is maintained at the same time, which significantly improves the efficiency of multi-satellite scheduling.

对于大规模、高强度的卫星遥感观测需求，约束关系的复杂性随着卫星任务规模的扩大而爆炸式增长。如何有效处理复杂的、时变的约束冲突，挖掘卫星任务约束之间存在的隐含知识，对提高调度效率具有重要意义，但也是卫星调度问题的核心难点。本文提出了一种基于动态任务约束异构图学习的冲突簇缓解方法来解决大规模卫星任务调度问题。该方法利用异构图表征多种非结构关系的优势，将约束冲突的时变特征投射到异构图中多个节点和边的空间拓扑上。因此，建立了一种基于关键冲突群采样的卫星任务动态约束异构图模型。并提出了一种改进的二次无约束二元优化异构注意网络（HAN-QUBO），该网络能够处理异构图，并试图表示卫星任务多重约束的隐含原理，从而提取有价值的冲突缓解策略和经验。仿真实验证明，该方法可为多卫星调度提供有效的经验指导，大大缓解大规模任务约束冲突检查过程繁琐的压力。对于数万任务的 EOSSP，冲突解决的平均次数减少了约 73.48%，同时保持了解决方案的质量，显著提高了多卫星调度的效率。

{"title":"A conflict clique mitigation method for large-scale satellite mission planning based on heterogeneous graph learning","authors":"Xiaoen Feng, Minqiang Xu, Yuqing Li","doi":"10.1016/j.aei.2024.102915","DOIUrl":"10.1016/j.aei.2024.102915","url":null,"abstract":"<div><div>For the large-scale and intensive demands of satellite remote sensing observation, the complexity of constraint relationships grows explosively with expansion of satellite task scale. How to efficiently deal with the complex and temporal varying constraint conflicts, and mine the implicit knowledge existing among satellite mission constraints, which is significant to enhance scheduling efficiency, however, is also a core difficulty in the satellite scheduling problem. In this paper, we propose a conflict clique mitigation method based on dynamic task-constrained heterogeneous graph learning to solve large-scale satellite mission scheduling. The method exploits the advantage of heterogeneous graphs to characterize multiple unstructured relationships, and projects the temporal-varying features of constraint conflicts to the spatial topology of multiple nodes and edges in a heterogeneous graph. Thus, a dynamic constraints heterogeneous graph model for satellite tasks based on sampling critical conflict cliques is developed. And an improved heterogeneous attention network with quadratic unconstrained binary optimization (HAN-QUBO) is proposed, which is able to deal with the heterogeneous graphs and attempts to represent the implicit principles of multiple constraints of satellite missions, so that the valuable strategies and experiences of conflict mitigation can be extracted. The simulation experiments demonstrate that the method can provide effective empirical guidance for multi-satellite scheduling, greatly relieve the pressure of cumbersome constraint conflict checking process for large-scale tasks. The average number of conflict resolutions has been reduced by about 73.48 % for EOSSPs with tens of thousands tasks, while the quality of solutions is maintained at the same time, which significantly improves the efficiency of multi-satellite scheduling.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102915"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142578042","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

Dynamic airport gate assignment with improved Shuffled Frog-Leaping Algorithm and triangle membership function 利用改进的洗牌蛙跳算法和三角形成员函数实现动态机场登机口分配

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102888

Hsien-Pin Hsu , Wan-Fang Yang , Tran Thi Bich Chau Vo

The rapid development of the air transportation industry has increased air traffic, posing challenges to the task of airport gate assignment (AGA) for flights. Most past studies have solved the AGA problem (AGAP) using deterministic models, which are incapable of dealing with uncertainty and dynamic conditions at airports. Thus, this research employs fuzzy theory and proposes a triangular membership function to handle flight uncertainty in the AGAP. In addition, an improved metaheuristic, termed the improved Shuffled Frog-Leaping Algorithm (ISFLA), is proposed to circumvent the computationally intractable problems commonly faced by exact approaches when handling large instances. In this research, the AGAP is first formulated as a stochastic Mixed-Integer Linear Programming (MILP) model, with stochastic flight lateness and earliness considered. The objective of this model is to minimize the total cost, which consists of three sub-costs: passenger walking distances, non-preferred gate (NPG) assignments for planes, and fuzzy idle times of gates. These three sub-costs correspond to the major concerns of passengers, airlines, and airports, respectively. The cooperation between the ISFLA and the triangular membership function demonstrates their capability to effectively handle big AGAP instances. Furthermore, the experimental results show that the ISFLA outperforms the standard SFLA, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Firefly Algorithm (FA).

航空运输业的快速发展增加了航空交通量，给航班的机场登机口分配（AGA）任务带来了挑战。以往的研究大多采用确定性模型解决登机口分配问题（AGAP），无法应对机场的不确定性和动态条件。因此，本研究采用了模糊理论，并提出了三角成员函数来处理 AGAP 中的航班不确定性。此外，还提出了一种改进的元启发式算法，即改进的洗牌蛙跳算法（ISFLA），以规避精确算法在处理大型实例时通常面临的计算棘手问题。在这项研究中，AGAP 首先被表述为一个随机混合整数线性规划（MILP）模型，并考虑了随机航班延迟和提前的问题。该模型的目标是使总成本最小化，总成本由三个子成本组成：乘客步行距离、飞机的非首选登机口（NPG）分配和登机口的模糊空闲时间。这三个子成本分别对应乘客、航空公司和机场的主要关注点。ISFLA 与三角阶乘函数之间的合作表明，它们能够有效地处理大型 AGAP 实例。此外，实验结果表明，ISFLA 优于标准 SFLA、遗传算法（GA）、粒子群优化（PSO）和萤火虫算法（FA）。

{"title":"Dynamic airport gate assignment with improved Shuffled Frog-Leaping Algorithm and triangle membership function","authors":"Hsien-Pin Hsu , Wan-Fang Yang , Tran Thi Bich Chau Vo","doi":"10.1016/j.aei.2024.102888","DOIUrl":"10.1016/j.aei.2024.102888","url":null,"abstract":"<div><div>The rapid development of the air transportation industry has increased air traffic, posing challenges to the task of airport gate assignment (AGA) for flights. Most past studies have solved the AGA problem (AGAP) using deterministic models, which are incapable of dealing with uncertainty and dynamic conditions at airports. Thus, this research employs fuzzy theory and proposes a triangular membership function to handle flight uncertainty in the AGAP. In addition, an improved metaheuristic, termed the improved Shuffled Frog-Leaping Algorithm (ISFLA), is proposed to circumvent the computationally intractable problems commonly faced by exact approaches when handling large instances. In this research, the AGAP is first formulated as a stochastic Mixed-Integer Linear Programming (MILP) model, with stochastic flight lateness and earliness considered. The objective of this model is to minimize the total cost, which consists of three sub-costs: passenger walking distances, non-preferred gate (NPG) assignments for planes, and fuzzy idle times of gates. These three sub-costs correspond to the major concerns of passengers, airlines, and airports, respectively. The cooperation between the ISFLA and the triangular membership function demonstrates their capability to effectively handle big AGAP instances. Furthermore, the experimental results show that the ISFLA outperforms the standard SFLA, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Firefly Algorithm (FA).</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102888"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142586520","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

FedITA: A cloud–edge collaboration framework for domain generalization-based federated fault diagnosis of machine-level industrial motors FedITA：基于领域泛化的机器级工业电机联合故障诊断云边协作框架

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102853

Yiming He, Weiming Shen

Adequate samples are necessary for establishing a high-performance supervised learning model for intelligent fault diagnosis. Startup companies may only have normal devices and therefore there exists extreme class imbalance of training samples. Lack of faulty devices makes it difficult to independently establish supervised learning. The ideal aggregated training using raw data from multiple client sources may lead to potential conflicts of interest, making it difficult to implement. In addition, individual difference caused by manufacturing inconsistencies and dynamic testing environments is a special interference for machine-level industrial motors, which is more significant in the information flow of multiple client sources. This article proposes a federated iterative learning algorithm (FedITA) as a cloud–edge collaboration framework for domain generalization-based federated fault diagnosis of machine-level industrial motors. The proposed FedITA utilizes progressive training and iterative weight updates to enhance secure interaction between different clients, effectively reducing the risk of overfitting caused by extreme class imbalance. A hybrid perception mechanism is implemented by developing complementary perception modules and integrated into a hybrid perception field network (HPFNet) as a recommended global federated model. The proposed method and model are performed on real production line signals and can achieve mean cross-machine F1-score of 96.50% in limited communication.

要为智能故障诊断建立高性能的监督学习模型，就必须有足够的样本。初创公司可能只有正常设备，因此训练样本的类别极不平衡。由于缺乏故障设备，很难独立建立监督学习模型。使用来自多个客户来源的原始数据进行理想的聚合训练可能会导致潜在的利益冲突，从而难以实施。此外，制造不一致性和动态测试环境造成的个体差异是机器级工业电机的特殊干扰，在多客户端源的信息流中更为显著。本文提出了一种联合迭代学习算法（FedITA）作为云边协作框架，用于基于领域泛化的机器级工业电机联合故障诊断。所提出的 FedITA 利用渐进式训练和迭代权重更新来加强不同客户端之间的安全交互，有效降低了因极端类不平衡而导致的过拟合风险。通过开发互补感知模块实现了混合感知机制，并将其集成到混合感知场网络（HPFNet）中，作为推荐的全局联合模型。所提出的方法和模型在真实生产线信号上进行了验证，并在有限的通信条件下实现了平均跨机 F1 分数 96.50%。

{"title":"FedITA: A cloud–edge collaboration framework for domain generalization-based federated fault diagnosis of machine-level industrial motors","authors":"Yiming He, Weiming Shen","doi":"10.1016/j.aei.2024.102853","DOIUrl":"10.1016/j.aei.2024.102853","url":null,"abstract":"<div><div>Adequate samples are necessary for establishing a high-performance supervised learning model for intelligent fault diagnosis. Startup companies may only have normal devices and therefore there exists extreme class imbalance of training samples. Lack of faulty devices makes it difficult to independently establish supervised learning. The ideal aggregated training using raw data from multiple client sources may lead to potential conflicts of interest, making it difficult to implement. In addition, individual difference caused by manufacturing inconsistencies and dynamic testing environments is a special interference for machine-level industrial motors, which is more significant in the information flow of multiple client sources. This article proposes a federated iterative learning algorithm (FedITA) as a cloud–edge collaboration framework for domain generalization-based federated fault diagnosis of machine-level industrial motors. The proposed FedITA utilizes progressive training and iterative weight updates to enhance secure interaction between different clients, effectively reducing the risk of overfitting caused by extreme class imbalance. A hybrid perception mechanism is implemented by developing complementary perception modules and integrated into a hybrid perception field network (HPFNet) as a recommended global federated model. The proposed method and model are performed on real production line signals and can achieve mean cross-machine F1-score of 96.50% in limited communication.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102853"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142417122","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 novel semi-supervised prediction modeling method based on deep learning for flotation process with large drift of working conditions 基于深度学习的新型半监督预测建模方法，适用于工况漂移较大的浮选工艺

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102934

Fanlei Lu, Weihua Gui, Liyang Qin, Xiaoli Wang, Jiayi Zhou

Deep neural networks have been broadly utilized for soft sensing modeling for the process performance which is significant for process control but cannot be measured online. However, the popular deep learning models still cannot adapt to large drift of working conditions in the process industry, which causes the model accuracy to become worse and worse with the time go on. Moreover, the cost of acquiring sufficient labeled data is very high. Therefore, in this study, a semi-supervised deep learning method called dynamic multi-scale selective kernel network (DMS-Sknet) with novel loss function is proposed by taking the flotation process as the case. In DMS-SKnet, multiscale features are extracted from froth images by using multi-scale dilated convolution kernel, and then fused with other process data in time series. A channel attention module with soft attention is designed to learn the important relationships between multi-scale feature maps and process features. Finally, based on the semi-supervised Mean-teacher (MT) learning framework, a new loss function is proposed, in which temporal distance is considered to improve the generalization ability and the long-term accuracy of the network. The experimental results using industrial flotation process data show that this method can effectively improve the grade prediction accuracy after a long period of significant changes in the working conditions.

深度神经网络已被广泛用于过程性能的软传感建模，这对过程控制意义重大，但却无法在线测量。然而，目前流行的深度学习模型仍然无法适应流程工业中的大量工作条件漂移，这导致模型的准确性随着时间的推移变得越来越差。此外，获取足够标注数据的成本也非常高。因此，本研究以浮选工艺为例，提出了一种具有新颖损失函数的半监督深度学习方法--动态多尺度选择核网络（DMS-Sknet）。在 DMS-SKnet 中，使用多尺度扩张卷积核从浮选图像中提取多尺度特征，然后与时间序列中的其他过程数据融合。为了学习多尺度特征图与过程特征之间的重要关系，设计了一个具有软注意力的通道注意力模块。最后，基于半监督平均教师（MT）学习框架，提出了一种新的损失函数，其中考虑了时间距离，以提高网络的泛化能力和长期精度。使用工业浮选工艺数据的实验结果表明，该方法能有效提高工况长期显著变化后的品位预测精度。

{"title":"A novel semi-supervised prediction modeling method based on deep learning for flotation process with large drift of working conditions","authors":"Fanlei Lu, Weihua Gui, Liyang Qin, Xiaoli Wang, Jiayi Zhou","doi":"10.1016/j.aei.2024.102934","DOIUrl":"10.1016/j.aei.2024.102934","url":null,"abstract":"<div><div>Deep neural networks have been broadly utilized for soft sensing modeling for the process performance which is significant for process control but cannot be measured online. However, the popular deep learning models still cannot adapt to large drift of working conditions in the process industry, which causes the model accuracy to become worse and worse with the time go on. Moreover, the cost of acquiring sufficient labeled data is very high. Therefore, in this study, a semi-supervised deep learning method called dynamic multi-scale selective kernel network (DMS-Sknet) with novel loss function is proposed by taking the flotation process as the case. In DMS-SKnet, multiscale features are extracted from froth images by using multi-scale dilated convolution kernel, and then fused with other process data in time series. A channel attention module with soft attention is designed to learn the important relationships between multi-scale feature maps and process features. Finally, based on the semi-supervised Mean-teacher (MT) learning framework, a new loss function is proposed, in which temporal distance is considered to improve the generalization ability and the long-term accuracy of the network. The experimental results using industrial flotation process data show that this method can effectively improve the grade prediction accuracy after a long period of significant changes in the working conditions.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102934"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142659118","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 task-oriented theil index-based meta-learning network with gradient calibration strategy for rotating machinery fault diagnosis with limited samples 基于任务导向 Theil 指数的元学习网络与梯度校准策略，用于有限样本的旋转机械故障诊断

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102870

Mingzhe Mu, Hongkai Jiang, Xin Wang, Yutong Dong

In industrial scenarios, rotating machinery operates in harsh environments under complex and variable conditions, which leads to a scarcity of available data. This brings challenges to intelligent model-based rotating machinery fault diagnosis. For this issue, a task-oriented theil index-based meta-learning network with gradient calibration strategy (TTIMN-GCS) is proposed for rotating machinery fault diagnosis with limited samples. Firstly, a fine-grained feature learner (FGFL) is designed to extract high-dimensional fine-grained fault information from limited samples. The FGFL is modeled after the human recognition process of fine-grained objects, enhancing distinguishing between fault categories with subtle differences. Secondly, a task inequality metric named task-oriented theil index is developed to acquire more competitive update rules from limited samples, which creatively frees the initial performance of the meta-FGFL from being overly tied to specific tasks. Finally, a gradient calibration strategy is proposed to adjust the optimization trajectory of TTIMN-GCS, which facilitates the diagnostic model evolution toward robust generalization performance. Four diagnostic cases on several datasets are designed, and the diagnostic accuracies under the 5-shot setting reach 98.18 %, 96.68 %, 94.60 %, and 93.90 %, respectively, which are better than other state-of-the-art methods. Experimental results exhibit that the TTIMN-GCS has a remarkable capability to identify new fault categories from a few samples and is potentially promising for engineering applications.

在工业场景中，旋转机械在复杂多变的恶劣环境中运行，导致可用数据稀缺。这给基于模型的智能旋转机械故障诊断带来了挑战。针对这一问题，我们提出了一种基于任务导向 Theil 索引的元学习网络与梯度校准策略（TTIMN-GCS），用于样本有限的旋转机械故障诊断。首先，设计了一个细粒度特征学习器（FGFL），用于从有限样本中提取高维细粒度故障信息。细粒度特征学习器仿照人类对细粒度对象的识别过程，增强了对具有细微差别的故障类别的区分能力。其次，开发了一种名为 "面向任务的 Theil 指数 "的任务不平等度量，以便从有限的样本中获取更具竞争力的更新规则，从而创造性地将元 FGFL 的初始性能从与特定任务的过度绑定中解放出来。最后，还提出了梯度校准策略来调整 TTIMN-GCS 的优化轨迹，从而促进诊断模型向稳健的泛化性能演进。在多个数据集上设计了四个诊断案例，在 5 次拍摄设置下的诊断准确率分别达到 98.18%、96.68%、94.60% 和 93.90%，优于其他先进方法。实验结果表明，TTIMN-GCS 具有从少量样本中识别新故障类别的卓越能力，在工程应用中具有潜在的发展前景。

{"title":"A task-oriented theil index-based meta-learning network with gradient calibration strategy for rotating machinery fault diagnosis with limited samples","authors":"Mingzhe Mu, Hongkai Jiang, Xin Wang, Yutong Dong","doi":"10.1016/j.aei.2024.102870","DOIUrl":"10.1016/j.aei.2024.102870","url":null,"abstract":"<div><div>In industrial scenarios, rotating machinery operates in harsh environments under complex and variable conditions, which leads to a scarcity of available data. This brings challenges to intelligent model-based rotating machinery fault diagnosis. For this issue, a task-oriented theil index-based meta-learning network with gradient calibration strategy (TTIMN-GCS) is proposed for rotating machinery fault diagnosis with limited samples. Firstly, a fine-grained feature learner (FGFL) is designed to extract high-dimensional fine-grained fault information from limited samples. The FGFL is modeled after the human recognition process of fine-grained objects, enhancing distinguishing between fault categories with subtle differences. Secondly, a task inequality metric named task-oriented theil index is developed to acquire more competitive update rules from limited samples, which creatively frees the initial performance of the meta-FGFL from being overly tied to specific tasks. Finally, a gradient calibration strategy is proposed to adjust the optimization trajectory of TTIMN-GCS, which facilitates the diagnostic model evolution toward robust generalization performance. Four diagnostic cases on several datasets are designed, and the diagnostic accuracies under the 5-shot setting reach 98.18 %, 96.68 %, 94.60 %, and 93.90 %, respectively, which are better than other state-of-the-art methods. Experimental results exhibit that the TTIMN-GCS has a remarkable capability to identify new fault categories from a few samples and is potentially promising for engineering applications.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102870"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442569","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

Hybrid physics-embedded recurrent neural networks for fault diagnosis under time-varying conditions based on multivariate proprioceptive signals 基于多变量本体感觉信号的混合物理嵌入式递归神经网络，用于时变条件下的故障诊断

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102851

Rourou Li, Tangbin Xia, Feng Luo, Yimin Jiang, Zhen Chen, Lifeng Xi

Accurate fault diagnosis for industrial robots is imperative to improve their availability. Proprioceptive signals collected by intrinsic sensors of robot joint servo drive systems provide a nonintrusive and promising way for practical in-situ diagnosis. However, they generally exhibit significant non-stationarity owing to time-varying operation conditions and limited sampling frequencies constrained by system hardware, which poses challenges in fault signature identification. Thus, a hybrid physics-embedded recurrent neural network is proposed for robot fault diagnosis under variable operation conditions based on proprioceptive signals. It embeds robot governing ordinary differential equations (ODE) as an inductive bias to account for known dynamics. Concurrently, tailored neural networks (NN) are leveraged to compensate for unmodeled dynamics residuum and unmeasurable health states, efficiently extending the hypothesis space. Hereinto, system status-represented latent space inferred from observations is comprehensively regularized by state reconstruction, fault classification, and Fisher discrimination losses to promote state representability and class distinguishability. Furthermore, a bilinear layer-based NN is constructed to statistically model intrinsic nonlinearities simplified away by physical models. Finally, the model-based and data-driven components are synergistically integrated by a differentiable ODE solver to form an end-to-end trainable framework. The superiority of the presented method is illustrated through the simulated and in-situ industrial robot datasets.

要提高工业机器人的可用性，就必须对其进行精确的故障诊断。机器人关节伺服驱动系统的固有传感器收集的感知信号为实际现场诊断提供了一种非侵入式的可行方法。然而，由于运行条件随时间变化，系统硬件限制了有限的采样频率，这些信号通常表现出明显的非稳态性，这给故障特征识别带来了挑战。因此，本文提出了一种基于本体感觉信号的混合物理嵌入式递归神经网络，用于可变运行条件下的机器人故障诊断。它嵌入了机器人治理常微分方程（ODE）作为归纳偏置，以考虑已知动态。同时，利用定制的神经网络（NN）来补偿未建模的动态残留和不可测量的健康状态，从而有效地扩展了假设空间。在此基础上，通过状态重构、故障分类和费雪分辨损失对从观测结果中推断出的系统状态表示潜空间进行全面正则化，以提高状态可表示性和类别可区分性。此外，还构建了基于双线性层的 NN，以便对物理模型所简化的内在非线性进行统计建模。最后，基于模型的部分和数据驱动的部分通过一个可微分的 ODE 求解器协同整合，形成一个端到端的可训练框架。通过模拟和现场工业机器人数据集，说明了所介绍方法的优越性。

{"title":"Hybrid physics-embedded recurrent neural networks for fault diagnosis under time-varying conditions based on multivariate proprioceptive signals","authors":"Rourou Li, Tangbin Xia, Feng Luo, Yimin Jiang, Zhen Chen, Lifeng Xi","doi":"10.1016/j.aei.2024.102851","DOIUrl":"10.1016/j.aei.2024.102851","url":null,"abstract":"<div><div>Accurate fault diagnosis for industrial robots is imperative to improve their availability. Proprioceptive signals collected by intrinsic sensors of robot joint servo drive systems provide a nonintrusive and promising way for practical in-situ diagnosis. However, they generally exhibit significant non-stationarity owing to time-varying operation conditions and limited sampling frequencies constrained by system hardware, which poses challenges in fault signature identification. Thus, a hybrid physics-embedded recurrent neural network is proposed for robot fault diagnosis under variable operation conditions based on proprioceptive signals. It embeds robot governing ordinary differential equations (ODE) as an inductive bias to account for known dynamics. Concurrently, tailored neural networks (NN) are leveraged to compensate for unmodeled dynamics residuum and unmeasurable health states, efficiently extending the hypothesis space. Hereinto, system status-represented latent space inferred from observations is comprehensively regularized by state reconstruction, fault classification, and Fisher discrimination losses to promote state representability and class distinguishability. Furthermore, a bilinear layer-based NN is constructed to statistically model intrinsic nonlinearities simplified away by physical models. Finally, the model-based and data-driven components are synergistically integrated by a differentiable ODE solver to form an end-to-end trainable framework. The superiority of the presented method is illustrated through the simulated and in-situ industrial robot datasets.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102851"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142417046","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

Massive-Scale construction dataset synthesis through Stable Diffusion for Machine learning training 通过稳定扩散合成大规模建筑数据集，用于机器学习训练

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102866

Sungkook Hong , Byungjoo Choi , Youngjib Ham , JungHo Jeon , Hyunsoo Kim

Advancements of artificial intelligence (AI)-driven image generation provide opportunities to address a problem in machine learning applications that have suffered from a lack of domain-specific training data. This study explores the feasibility of employing synthesized images (SIs) generated through Stable Diffusion as training data for construction. This study aims to examine the potential of Stable Diffusion in construction, and the performance of convolutional neural network (CNN) models trained exclusively on SIs. A total of 82.01% of images synthesized are suitable for representing construction tasks. The CNN model trained on preprocessed SIs (with context-based labeling results) exhibited a classification accuracy of 89.09%. The CNN model trained solely on raw SIs (synthesized images without context-based labeling results) achieved a successful classification rate of 86.51% for the images. This study presents the viability of SIs as a training dataset and introduces context-based labeling through object detection techniques, enhancing the performance of estimation models.

人工智能（AI）驱动的图像生成技术的进步为解决机器学习应用中因缺乏特定领域的训练数据而造成的问题提供了机会。本研究探讨了使用通过稳定扩散生成的合成图像（SI）作为建筑培训数据的可行性。本研究旨在考察稳定扩散技术在建筑领域的应用潜力，以及完全基于 SIs 训练的卷积神经网络（CNN）模型的性能。合成的图像中共有 82.01% 适合用于表现建筑任务。在预处理 SI（基于上下文的标注结果）上训练的 CNN 模型的分类准确率为 89.09%。仅根据原始 SI（没有基于上下文的标注结果的合成图像）训练的 CNN 模型的图像分类成功率为 86.51%。本研究介绍了将 SI 作为训练数据集的可行性，并通过对象检测技术引入了基于上下文的标注，从而提高了估算模型的性能。

{"title":"Massive-Scale construction dataset synthesis through Stable Diffusion for Machine learning training","authors":"Sungkook Hong , Byungjoo Choi , Youngjib Ham , JungHo Jeon , Hyunsoo Kim","doi":"10.1016/j.aei.2024.102866","DOIUrl":"10.1016/j.aei.2024.102866","url":null,"abstract":"<div><div>Advancements of artificial intelligence (AI)-driven image generation provide opportunities to address a problem in machine learning applications that have suffered from a lack of domain-specific training data. This study explores the feasibility of employing synthesized images (SIs) generated through Stable Diffusion as training data for construction. This study aims to examine the potential of Stable Diffusion in construction, and the performance of convolutional neural network (CNN) models trained exclusively on SIs. A total of 82.01% of images synthesized are suitable for representing construction tasks. The CNN model trained on preprocessed SIs (with context-based labeling results) exhibited a classification accuracy of 89.09%. The CNN model trained solely on raw SIs (synthesized images without context-based labeling results) achieved a successful classification rate of 86.51% for the images. This study presents the viability of SIs as a training dataset and introduces context-based labeling through object detection techniques, enhancing the performance of estimation models.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102866"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142417048","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

Human–machine cooperative decision-making and planning for automated vehicles using spatial projection of hand gestures 利用手势的空间投影实现自动驾驶汽车的人机协同决策和规划

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102864

Yiran Zhang , Zhongxu Hu , Peng Hang , Shanhe Lou , Chen Lv

Significant challenges in perception, prediction, and decision-making within self-driving systems remain inadequately addressed. Concurrently, the advancement of autonomous driving technologies reduces driver engagement, inadvertently eroding their proficiency. Integrating human cognitive flexibility and experiential insight with the machine’s precision and reliability offers a promising approach for the transitional phase towards fully automated driving. This study presents a human-machine collaboration approach to enhance the highly automated vehicles’ high-level flexibility and personalization attribute without the need for passengers’ prior driving experience. Firstly, we propose a tactical human–vehicle collaboration framework leveraging the hand-landmark extraction algorithm and augmented visual feedback. The proposed vision-based interface projects the gesture onto the ground and feeds it back to the driver through the augmented reality head-up display (AR-HUD) for intuitive interaction. The projection offers strategic decision-making guidance and planning recommendations for the vehicle. Utilizing these suggestions, the automation algorithm efficiently manages the remaining tasks, including collision avoidance and adherence to traffic regulations. This approach minimizes the driver’s engagement in routine driving tasks and negates the need for driving skills. Incorporating cooperative game theory, the methodology optimally balances personalization with system robustness. Finally, we compare our approach with conventional manual driving schemes that both can assist the self-driving car in avoiding unknown obstacles and reaching the personalized goal. Results demonstrate that the proposed decision-making and planning collaboration scheme significantly reduces human physical burdens without compromising driving performance and driver mental workloads.

自动驾驶系统在感知、预测和决策方面面临的重大挑战仍未得到充分解决。同时，自动驾驶技术的进步降低了驾驶员的参与度，无意中削弱了他们的熟练程度。将人类认知的灵活性和经验洞察力与机器的精确性和可靠性相结合，为实现全自动驾驶的过渡阶段提供了一种前景广阔的方法。本研究提出了一种人机协作方法，以增强高度自动驾驶汽车的高级灵活性和个性化属性，而无需乘客事先具备驾驶经验。首先，我们提出了一个战术人车协作框架，利用手势标记提取算法和增强视觉反馈。所提出的基于视觉的界面可将手势投射到地面上，并通过增强现实平视显示器（AR-HUD）反馈给驾驶员，从而实现直观的交互。投影为车辆提供战略决策指导和规划建议。利用这些建议，自动驾驶算法可有效管理其余任务，包括避免碰撞和遵守交通法规。这种方法最大限度地减少了驾驶员在日常驾驶任务中的参与，并消除了对驾驶技能的需求。结合合作博弈理论，该方法在个性化与系统稳健性之间实现了最佳平衡。最后，我们将我们的方法与传统的手动驾驶方案进行了比较，两者都能帮助自动驾驶汽车避开未知障碍并达到个性化目标。结果表明，所提出的决策和规划协作方案大大减轻了人类的体力负担，同时又不影响驾驶性能和驾驶员的脑力劳动负荷。

{"title":"Human–machine cooperative decision-making and planning for automated vehicles using spatial projection of hand gestures","authors":"Yiran Zhang , Zhongxu Hu , Peng Hang , Shanhe Lou , Chen Lv","doi":"10.1016/j.aei.2024.102864","DOIUrl":"10.1016/j.aei.2024.102864","url":null,"abstract":"<div><div>Significant challenges in perception, prediction, and decision-making within self-driving systems remain inadequately addressed. Concurrently, the advancement of autonomous driving technologies reduces driver engagement, inadvertently eroding their proficiency. Integrating human cognitive flexibility and experiential insight with the machine’s precision and reliability offers a promising approach for the transitional phase towards fully automated driving. This study presents a human-machine collaboration approach to enhance the highly automated vehicles’ high-level flexibility and personalization attribute without the need for passengers’ prior driving experience. Firstly, we propose a tactical human–vehicle collaboration framework leveraging the hand-landmark extraction algorithm and augmented visual feedback. The proposed vision-based interface projects the gesture onto the ground and feeds it back to the driver through the augmented reality head-up display (AR-HUD) for intuitive interaction. The projection offers strategic decision-making guidance and planning recommendations for the vehicle. Utilizing these suggestions, the automation algorithm efficiently manages the remaining tasks, including collision avoidance and adherence to traffic regulations. This approach minimizes the driver’s engagement in routine driving tasks and negates the need for driving skills. Incorporating cooperative game theory, the methodology optimally balances personalization with system robustness. Finally, we compare our approach with conventional manual driving schemes that both can assist the self-driving car in avoiding unknown obstacles and reaching the personalized goal. Results demonstrate that the proposed decision-making and planning collaboration scheme significantly reduces human physical burdens without compromising driving performance and driver mental workloads.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102864"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442731","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

Fault diagnosis of mobile robot based on dual-graph convolutional network with prior fault knowledge 基于先验故障知识的双图卷积网络的移动机器人故障诊断

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

Advanced Engineering Informatics

Pub Date : 2024-10-01 DOI: 10.1016/j.aei.2024.102865

Longda Zhang , Fengyu Zhou , Peng Duan , Xianfeng Yuan

Effective integration of multi-sensor measurements is crucial for mobile robot fault diagnosis. However, in multi-sensor relationship modeling, existing methods often neglect the impact of different fault types and fail to consider the relations among data samples. To address these issues, a novel dual-graph convolutional network with prior fault knowledge (FKDGCN) is proposed. Specifically, we construct multi-sensor topological graphs based on prior fault knowledge, which effectively consider the impact of fault categories on sensor correlations. Subsequently, sample affinity graphs are constructed based on the temporal relationship and data similarity, and a sample correlation feature extraction module (SCFEM) is designed to capture the interdependence among data samples. Eventually, a novel dual-graph convolutional network is proposed to fuse multi-sample features and multi-sensor spatial–temporal features, in which more comprehensive fault information can be extracted. The effectiveness of FKDGCN is thoroughly validated on datasets collected from a real robot fault diagnosis test bench. Experimental results indicate that FKDGCN achieves outstanding diagnosis performance compared to state-of-the-art methods, with an average accuracy of over 98% on the balanced dataset and over 90% on two imbalanced datasets.

有效整合多传感器测量数据对于移动机器人故障诊断至关重要。然而，在多传感器关系建模中，现有方法往往忽略了不同故障类型的影响，也没有考虑数据样本之间的关系。为了解决这些问题，我们提出了一种具有先验故障知识的新型双图卷积网络（FKDGCN）。具体来说，我们基于先验故障知识构建了多传感器拓扑图，有效地考虑了故障类别对传感器相关性的影响。随后，我们根据时间关系和数据相似性构建了样本亲和图，并设计了样本相关性特征提取模块（SCFEM）来捕捉数据样本之间的相互依存关系。最后，提出了一种新颖的双图卷积网络来融合多样本特征和多传感器时空特征，从而提取出更全面的故障信息。FKDGCN 的有效性在真实机器人故障诊断测试台收集的数据集上得到了充分验证。实验结果表明，与最先进的方法相比，FKDGCN 实现了出色的诊断性能，在平衡数据集上的平均准确率超过 98%，在两个不平衡数据集上的平均准确率超过 90%。

{"title":"Fault diagnosis of mobile robot based on dual-graph convolutional network with prior fault knowledge","authors":"Longda Zhang , Fengyu Zhou , Peng Duan , Xianfeng Yuan","doi":"10.1016/j.aei.2024.102865","DOIUrl":"10.1016/j.aei.2024.102865","url":null,"abstract":"<div><div>Effective integration of multi-sensor measurements is crucial for mobile robot fault diagnosis. However, in multi-sensor relationship modeling, existing methods often neglect the impact of different fault types and fail to consider the relations among data samples. To address these issues, a novel dual-graph convolutional network with prior fault knowledge (FKDGCN) is proposed. Specifically, we construct multi-sensor topological graphs based on prior fault knowledge, which effectively consider the impact of fault categories on sensor correlations. Subsequently, sample affinity graphs are constructed based on the temporal relationship and data similarity, and a sample correlation feature extraction module (SCFEM) is designed to capture the interdependence among data samples. Eventually, a novel dual-graph convolutional network is proposed to fuse multi-sample features and multi-sensor spatial–temporal features, in which more comprehensive fault information can be extracted. The effectiveness of FKDGCN is thoroughly validated on datasets collected from a real robot fault diagnosis test bench. Experimental results indicate that FKDGCN achieves outstanding diagnosis performance compared to state-of-the-art methods, with an average accuracy of over 98% on the balanced dataset and over 90% on two imbalanced datasets.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"62 ","pages":"Article 102865"},"PeriodicalIF":8.0,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142446802","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