Pub Date : 2025-01-29DOI: 10.1016/j.aei.2025.103139
Wenhao Hu , Yun Wang , Yuchen He , Lijuan Qian , Dongping Zhang , Yongchang Meng
The soft sensing technique for mass customization (MC) process is always challenging due to its modular production, customized assembly and multi-scale. This paper proposes a soft sensing technique for MC process based on a heterogeneous causal graph attention network (HCGAT). Firstly, a priori causal additive model (PCAM) is proposed, which utilizes the mechanism and data to construct a causal skeleton jointly. On this basis, an additive noise model (ANM) together with scoring mechanism is designed to obtain the reliable causal graph. Secondly, self-learned weight parameter matrices are applied to the data block in each scale, enabling the mapping of distinct dimensional information onto a common dimension. Finally, a unique quality prediction framework is carried out to tackle the soft sensing modeling in MC process where specific combinations of different accessories are encoded for product classification. The performance of the proposed method is evaluated on a numerical simulation dataset and an electric automobile manufacturing dataset where the experimental results show the superiority of the methods in the efficacy and accuracy of quality prediction.
{"title":"Soft sensing technique for mass customization based on heterogeneous causal graph attention networks","authors":"Wenhao Hu , Yun Wang , Yuchen He , Lijuan Qian , Dongping Zhang , Yongchang Meng","doi":"10.1016/j.aei.2025.103139","DOIUrl":"10.1016/j.aei.2025.103139","url":null,"abstract":"<div><div>The soft sensing technique for mass customization (MC) process is always challenging due to its modular production, customized assembly and multi-scale. This paper proposes a soft sensing technique for MC process based on a heterogeneous causal graph attention network (HCGAT). Firstly, a priori causal additive model (PCAM) is proposed, which utilizes the mechanism and data to construct a causal skeleton jointly. On this basis, an additive noise model (ANM) together with scoring mechanism is designed to obtain the reliable causal graph. Secondly, self-learned weight parameter matrices are applied to the data block in each scale, enabling the mapping of distinct dimensional information onto a common dimension. Finally, a unique quality prediction framework is carried out to tackle the soft sensing modeling in MC process where specific combinations of different accessories are encoded for product classification. The performance of the proposed method is evaluated on a numerical simulation dataset and an electric automobile manufacturing dataset where the experimental results show the superiority of the methods in the efficacy and accuracy of quality prediction.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103139"},"PeriodicalIF":8.0,"publicationDate":"2025-01-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137044","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}
Object-goal navigation is a task in embodied AI where an agent is navigated to a specified object within unfamiliar indoor scenarios. This task is crucial for engineering activities such as training agents in 3D simulated environments and deploying these models in actual mobile robots. Extensive research has been conducted to develop various navigation methods, including end-to-end reinforcement learning and modular map-based approaches. However, fully enabling an agent to perceive and understand the environment, and to navigate towards a target object as efficiently as humans, remains a considerable challenge. In this study, we introduce a data-driven and modular map-based approach, trained on a dataset incorporated with common-sense knowledge of object-to-room relationships extracted from a Large Language Model (LLM), aiming to enhance the efficiency of object-goal navigation. This approach enables the agent to seek the target object in rooms where it is commonly found (e.g., a bed in a bedroom, a couch in a living room), according to LLM-based common-sense knowledge. Additionally, we employ the multi-channel Swin-Unet architecture for multi-task learning, integrating multimodal sensory inputs to effectively extract meaningful features for spatial comprehension and navigation. Results from the Habitat simulator show that our framework surpasses the baseline by an average of 10.6% in the Success-weighted by Path Length (SPL) efficiency metric. Real-world demonstrations confirm that our method can effectively navigate multiple rooms in the object-goal navigation task. For further details and real-world demonstrations, please visit our project webpage (https://sunleyuan.github.io/ObjectNav).
{"title":"Enhancing multimodal-input object goal navigation by leveraging large language models for inferring room–object relationship knowledge","authors":"Leyuan Sun , Asako Kanezaki , Guillaume Caron , Yusuke Yoshiyasu","doi":"10.1016/j.aei.2025.103135","DOIUrl":"10.1016/j.aei.2025.103135","url":null,"abstract":"<div><div>Object-goal navigation is a task in embodied AI where an agent is navigated to a specified object within unfamiliar indoor scenarios. This task is crucial for engineering activities such as training agents in 3D simulated environments and deploying these models in actual mobile robots. Extensive research has been conducted to develop various navigation methods, including end-to-end reinforcement learning and modular map-based approaches. However, fully enabling an agent to perceive and understand the environment, and to navigate towards a target object as efficiently as humans, remains a considerable challenge. In this study, we introduce a data-driven and modular map-based approach, trained on a dataset incorporated with common-sense knowledge of object-to-room relationships extracted from a Large Language Model (LLM), aiming to enhance the efficiency of object-goal navigation. This approach enables the agent to seek the target object in rooms where it is commonly found (e.g., a bed in a bedroom, a couch in a living room), according to LLM-based common-sense knowledge. Additionally, we employ the multi-channel Swin-Unet architecture for multi-task learning, integrating multimodal sensory inputs to effectively extract meaningful features for spatial comprehension and navigation. Results from the Habitat simulator show that our framework surpasses the baseline by an average of 10.6% in the Success-weighted by Path Length (SPL) efficiency metric. Real-world demonstrations confirm that our method can effectively navigate multiple rooms in the object-goal navigation task. For further details and real-world demonstrations, please visit our project webpage (<span><span>https://sunleyuan.github.io/ObjectNav</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103135"},"PeriodicalIF":8.0,"publicationDate":"2025-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137043","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}
Pub Date : 2025-01-25DOI: 10.1016/j.aei.2025.103125
Yu Ren , Lun Zhao , Yongtao Zhang , Yiyao Liu , Jinfeng Yang , Haigang Zhang , Baiying Lei
The detection of prohibited items is an extremely time-sensitive task, yet the complex convolutional neural network (CNN) model has a slow inference speed, which is not conducive to deployment and application in actual security inspection scenarios. Knowledge distillation is a key technology for improving the performance of lightweight models. However, most knowledge distillation methods do not perform differentiated distillation of the foreground and background. In addition, the structural misalignment in heterogeneous networks hinders the effective transfer of knowledge. These factors limit the generalization of knowledge distillation in X-ray image analysis. To solve these problems, we propose a method based on feature knowledge distillation, called XFKD, which aims to improve the detection performance of lightweight models for prohibited items in X-ray images. Specifically, XFKD consists of Local Distillation (LD) and Global Distillation (GD). LD uses mask attention to guide the student network to focus on key knowledge, enhancing its learning capacity. GD learns and reconstructs the relationships between global features from the teacher network, and then transfers to the student network. Furthermore, to weaken the impact of structural differences of heterogeneous networks on knowledge transfer, the features obtained by the teacher network are used as supervised “input” with prior knowledge, not just “target” is transferred to the student network to improve imitation ability. To verify the effectiveness and generalization of XFKD, experiments were carried out on two X-ray security inspection image datasets (SIXray, OPIXray) and COCO datasets. The results show that XFKD performs well in knowledge distillations of various homogeneous and heterogeneous networks, RetinaNet (ResNet101-ResNet50) and YOLOv4 (CSPDarkNet53-MobileNetV3) with XFKD strategy achieve 81. 25% mAP and 76. 32% mAP in the SIXray dataset, which is 7.1% and 1.89% higher than the baseline, respectively. XFKD can improve the detection performance of lightweight models. Our code is available at https://github.com/RY-97/XFKD.
{"title":"Feature knowledge distillation-based model lightweight for prohibited item detection in X-ray security inspection images","authors":"Yu Ren , Lun Zhao , Yongtao Zhang , Yiyao Liu , Jinfeng Yang , Haigang Zhang , Baiying Lei","doi":"10.1016/j.aei.2025.103125","DOIUrl":"10.1016/j.aei.2025.103125","url":null,"abstract":"<div><div>The detection of prohibited items is an extremely time-sensitive task, yet the complex convolutional neural network (CNN) model has a slow inference speed, which is not conducive to deployment and application in actual security inspection scenarios. Knowledge distillation is a key technology for improving the performance of lightweight models. However, most knowledge distillation methods do not perform differentiated distillation of the foreground and background. In addition, the structural misalignment in heterogeneous networks hinders the effective transfer of knowledge. These factors limit the generalization of knowledge distillation in X-ray image analysis. To solve these problems, we propose a method based on feature knowledge distillation, called XFKD, which aims to improve the detection performance of lightweight models for prohibited items in X-ray images. Specifically, XFKD consists of Local Distillation (LD) and Global Distillation (GD). LD uses mask attention to guide the student network to focus on key knowledge, enhancing its learning capacity. GD learns and reconstructs the relationships between global features from the teacher network, and then transfers to the student network. Furthermore, to weaken the impact of structural differences of heterogeneous networks on knowledge transfer, the features obtained by the teacher network are used as supervised “input” with prior knowledge, not just “target” is transferred to the student network to improve imitation ability. To verify the effectiveness and generalization of XFKD, experiments were carried out on two X-ray security inspection image datasets (SIXray, OPIXray) and COCO datasets. The results show that XFKD performs well in knowledge distillations of various homogeneous and heterogeneous networks, RetinaNet (ResNet101-ResNet50) and YOLOv4 (CSPDarkNet53-MobileNetV3) with XFKD strategy achieve 81. 25% mAP and 76. 32% mAP in the SIXray dataset, which is 7.1% and 1.89% higher than the baseline, respectively. XFKD can improve the detection performance of lightweight models. Our code is available at <span><span>https://github.com/RY-97/XFKD</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103125"},"PeriodicalIF":8.0,"publicationDate":"2025-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137042","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}
Pub Date : 2025-01-25DOI: 10.1016/j.aei.2025.103137
Jianzhou Wang, Menggang Kou, Runze Li, Yuansheng Qian, Zhiwu Li
Accurate ultra-short-term regional wind power forecasting is crucial for real-time power grid dispatching and frequency regulation. However, recent works have primarily focused on designing complex model structures, often overlooking computational efficiency or the balance between efficiency and prediction accuracy, which limits practical applications. This paper aims to improve data quality, reduce computational cost, and enhance prediction accuracy by proposing a spatiotemporal prediction method for large-scale wind farms. To address the parameter sensitivity issue of the density-based spatial clustering of applications with noise (DBSCAN) and to enhance its anomaly detection capability and robustness, we propose an improved version of DBSCAN, applied to wind turbine data anomaly detection. Simultaneously, leveraging the high-performance advantage of the lightweight gradient boosting machine, abnormal data are quickly corrected. Using spectral clustering based on graph theory, we optimally partition the data graph to form wind farm clusters. Subsequently, a two-layer adaptive graph convolutional recurrent neural network (AGCRN) is employed to capture complex spatiotemporal correlations between wind turbines in each cluster. Finally, the regional total power forecast is obtained by summing the forecast outputs of all clusters. Through numerical simulations using measured data from Dataset 1 (134 wind turbines) and Dataset 2 (200 wind turbines), the results indicate that the proposed data preprocessing scheme can achieve at least a 50 % improvement in the model. By forecasting in clusters, the mean absolute error (MAE) can be reduced by 45.84 %, training time shortened by 70.84 %, and GPU memory saved by 94.04 %. Compared with advanced models such as Transformer variants and TimeNet, the multi-layer AGCRN achieves the highest prediction accuracy, exceeding 85 %.
{"title":"Ultra-short-term wind power forecasting jointly driven by anomaly detection, clustering and graph convolutional recurrent neural networks","authors":"Jianzhou Wang, Menggang Kou, Runze Li, Yuansheng Qian, Zhiwu Li","doi":"10.1016/j.aei.2025.103137","DOIUrl":"10.1016/j.aei.2025.103137","url":null,"abstract":"<div><div>Accurate ultra-short-term regional wind power forecasting is crucial for real-time power grid dispatching and frequency regulation. However, recent works have primarily focused on designing complex model structures, often overlooking computational efficiency or the balance between efficiency and prediction accuracy, which limits practical applications. This paper aims to improve data quality, reduce computational cost, and enhance prediction accuracy by proposing a spatiotemporal prediction method for large-scale wind farms. To address the parameter sensitivity issue of the density-based spatial clustering of applications with noise (DBSCAN) and to enhance its anomaly detection capability and robustness, we propose an improved version of DBSCAN, applied to wind turbine data anomaly detection. Simultaneously, leveraging the high-performance advantage of the lightweight gradient boosting machine, abnormal data are quickly corrected. Using spectral clustering based on graph theory, we optimally partition the data graph to form wind farm clusters. Subsequently, a two-layer adaptive graph convolutional recurrent neural network (AGCRN) is employed to capture complex spatiotemporal correlations between wind turbines in each cluster. Finally, the regional total power forecast is obtained by summing the forecast outputs of all clusters. Through numerical simulations using measured data from Dataset 1 (134 wind turbines) and Dataset 2 (200 wind turbines), the results indicate that the proposed data preprocessing scheme can achieve at least a 50 % improvement in the model. By forecasting in clusters, the mean absolute error (MAE) can be reduced by 45.84 %, training time shortened by 70.84 %, and GPU memory saved by 94.04 %. Compared with advanced models such as Transformer variants and TimeNet, the multi-layer AGCRN achieves the highest prediction accuracy, exceeding 85 %.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103137"},"PeriodicalIF":8.0,"publicationDate":"2025-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136944","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}
Pub Date : 2025-01-24DOI: 10.1016/j.aei.2025.103132
Zisheng Wang , Shaochen Li , Jianping Xuan , Tielin Shi
Compound defects frequently arise, posing serious threats to the reliability and safety of machines as their structures become increasingly complex. Traditional approaches primarily rely on deep learning algorithms based on artificial neural networks. However, these methods cannot fully replicate the information transmission functions of human brain neurons, resulting in limited biological interpretability and reduced reliability in practice. To address this challenge, this paper introduces a biologically inspired approach for compound defect detection using a spiking neural network with an improved pooling layer. The proposed method integrates a specially designed spiking neuron with a wavelet packet pooling mechanism (WPPM), forming the WPPM-spiking neural network (WPPM-SNN) model. This model employs spiking layers enhanced by a wavelet gradient, enabling it to adeptly extract nuanced features from preprocessed samples with compound defects. Specifically, WPPM simulates the wavelet transform within the pooling layers, based on mathematical analysis. This model was studied on both laboratorial and engineering verifications, achieving composite defect accuracies of 99% and 84.79%, respectively. Compared with popular deep models, the proposed model demonstrated accuracy improvements of 6.92% and 5.87%, respectively. Those empirical results on multiple evaluation metrics clearly demonstrate that the WPPM-SNN model significantly outperforms popular multi-label learning techniques in detecting compound defects.
{"title":"Biologically inspired compound defect detection using a spiking neural network with continuous time–frequency gradients","authors":"Zisheng Wang , Shaochen Li , Jianping Xuan , Tielin Shi","doi":"10.1016/j.aei.2025.103132","DOIUrl":"10.1016/j.aei.2025.103132","url":null,"abstract":"<div><div>Compound defects frequently arise, posing serious threats to the reliability and safety of machines as their structures become increasingly complex. Traditional approaches primarily rely on deep learning algorithms based on artificial neural networks. However, these methods cannot fully replicate the information transmission functions of human brain neurons, resulting in limited biological interpretability and reduced reliability in practice. To address this challenge, this paper introduces a biologically inspired approach for compound defect detection using a spiking neural network with an improved pooling layer. The proposed method integrates a specially designed spiking neuron with a wavelet packet pooling mechanism (WPPM), forming the WPPM-spiking neural network (WPPM-SNN) model. This model employs spiking layers enhanced by a wavelet gradient, enabling it to adeptly extract nuanced features from preprocessed samples with compound defects. Specifically, WPPM simulates the wavelet transform within the pooling layers, based on mathematical analysis. This model was studied on both laboratorial and engineering verifications, achieving composite defect accuracies of 99% and 84.79%, respectively. Compared with popular deep models, the proposed model demonstrated accuracy improvements of 6.92% and 5.87%, respectively. Those empirical results on multiple evaluation metrics clearly demonstrate that the WPPM-SNN model significantly outperforms popular multi-label learning techniques in detecting compound defects.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103132"},"PeriodicalIF":8.0,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136942","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}
Pub Date : 2025-01-24DOI: 10.1016/j.aei.2025.103122
Huining Pei , Mingzhe Xu , Xinyu Liu , Hao Gong , Guiyang Li , Zhonghang Bai
The use of new technologies to extract users’ needs and provide them with personalized services has become a development trend. Rely on cloud platform data, to recommend personalized services to users more accurately, the emotional preferences of users are combined with their long- and short-term interests. Moreover, a personalized cloud platform service recommendation model, called the multi-directional aggregate graph convolutional network (MGCN) is proposed. First, the importance of online text analysis in the field of cloud service platforms is summarized, as is the lack of research on the acquisition of user needs based on online text. Second, the relevant theoretical knowledge and models of online text acquisition are explored. Third, a personalized cloud platform service recommendation model is proposed based on the optimized GCN models to analyze the emotional preferences and long- and short-term interests of users. Finally, the feasibility of the improved methodology are verified using the relevant Q&A data of the top 100 most active users on three professional cloud service platforms in China. The findings provide new concepts for the front-end construction of cloud service platforms.
{"title":"A user demand acquisition method for cloud services based on user sentiment analysis and long- and short-term preferences","authors":"Huining Pei , Mingzhe Xu , Xinyu Liu , Hao Gong , Guiyang Li , Zhonghang Bai","doi":"10.1016/j.aei.2025.103122","DOIUrl":"10.1016/j.aei.2025.103122","url":null,"abstract":"<div><div>The use of new technologies to extract users’ needs and provide them with personalized services has become a development trend. Rely on cloud platform data, to recommend personalized services to users more accurately, the emotional preferences of users are combined with their long- and short-term interests. Moreover, a personalized cloud platform service recommendation model, called the multi-directional aggregate graph convolutional network (MGCN) is proposed. First, the importance of online text analysis in the field of cloud service platforms is summarized, as is the lack of research on the acquisition of user needs based on online text. Second, the relevant theoretical knowledge and models of online text acquisition are explored. Third, a personalized cloud platform service recommendation model is proposed based on the optimized GCN models to analyze the emotional preferences and long- and short-term interests of users. Finally, the feasibility of the improved methodology are verified using the relevant Q&A data of the top 100 most active users on three professional cloud service platforms in China. The findings provide new concepts for the front-end construction of cloud service platforms.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103122"},"PeriodicalIF":8.0,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136974","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}
Pub Date : 2025-01-23DOI: 10.1016/j.aei.2025.103134
Xingming Liao , Chong Chen , Zhuowei Wang , Ying Liu , Tao Wang , Lianglun Cheng
With the rapid deployment of industrial robots in manufacturing, the demand for advanced maintenance techniques to sustain operational efficiency has become crucial. Fault diagnosis Knowledge Graph (KG) is essential as it interlinks multi-source data related to industrial robot faults, capturing multi-level semantic associations among different fault events. However, the construction and application of fine-grained fault diagnosis KG face significant challenges due to the inherent complexity of nested entities in maintenance texts and the severe scarcity of annotated industrial data. In this study, we propose a Large Language Model (LLM) assisted data augmentation approach, which handles the complex nested entities in maintenance corpora and constructs a more fine-grained fault diagnosis KG. Firstly, the fine-grained ontology is constructed via LLM Assistance in Industrial Nested Named Entity Recognition (assInNNER). Then, an Industrial Nested Label Classification Template (INCT) is designed, enabling the use of nested entities in Attention-map aware keyword selection for the Industrial Nested Language Model (ANLM) data augmentation methods. ANLM can effectively improve the model’s performance in nested entity extraction when corpora are scarce. Subsequently, a Confidence Filtering Mechanism (CFM) is introduced to evaluate and select the generated data for enhancement, and assInNNER is further deployed to recall the negative samples corpus again to further improve performance. Experimental studies based on multi-source corpora demonstrate that compared to existing algorithms, our method achieves an average F1 increase of 8.25 %, 3.31 %, and 1.96 % in 5%, 10 %, and 25 % in few-shot settings, respectively.
{"title":"Large language model assisted fine-grained knowledge graph construction for robotic fault diagnosis","authors":"Xingming Liao , Chong Chen , Zhuowei Wang , Ying Liu , Tao Wang , Lianglun Cheng","doi":"10.1016/j.aei.2025.103134","DOIUrl":"10.1016/j.aei.2025.103134","url":null,"abstract":"<div><div>With the rapid deployment of industrial robots in manufacturing, the demand for advanced maintenance techniques to sustain operational efficiency has become crucial. Fault diagnosis Knowledge Graph (KG) is essential as it interlinks multi-source data related to industrial robot faults, capturing multi-level semantic associations among different fault events. However, the construction and application of fine-grained fault diagnosis KG face significant challenges due to the inherent complexity of nested entities in maintenance texts and the severe scarcity of annotated industrial data. In this study, we propose a Large Language Model (LLM) assisted data augmentation approach, which handles the complex nested entities in maintenance corpora and constructs a more fine-grained fault diagnosis KG. Firstly, the fine-grained ontology is constructed via LLM Assistance in Industrial Nested Named Entity Recognition (assInNNER). Then, an Industrial Nested Label Classification Template (INCT) is designed, enabling the use of nested entities in Attention-map aware keyword selection for the Industrial Nested Language Model (ANLM) data augmentation methods. ANLM can effectively improve the model’s performance in nested entity extraction when corpora are scarce. Subsequently, a Confidence Filtering Mechanism (CFM) is introduced to evaluate and select the generated data for enhancement, and assInNNER is further deployed to recall the negative samples corpus again to further improve performance. Experimental studies based on multi-source corpora demonstrate that compared to existing algorithms, our method achieves an average F1 increase of 8.25 %, 3.31 %, and 1.96 % in 5%, 10 %, and 25 % in few-shot settings, respectively.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103134"},"PeriodicalIF":8.0,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136973","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}
Pub Date : 2025-01-22DOI: 10.1016/j.aei.2025.103136
Shijun Xie , Changqing Shen , Dong Wang , Juanjuan Shi , Weiguo Huang , Zhongkui Zhu
In the rapidly evolving industrial environment, bearings may develop new fault types, posing significant challenges to deep learning-based intelligent fault diagnosis models. These models often suffer from catastrophic forgetting when encountering unknown fault types, resulting in performance degradation. Lifelong learning strategies offer a solution by enabling models to retain old knowledge while acquiring new information. However, traditional replay-based lifelong learning methods typically involve risks of privacy leakage and escalating storage costs. To address these issues, this study proposes a novel lifelong learning method called lifelong learning based on dual distillation (LLDD), which integrates a dual-distillation mechanism comprising dataset distillation and feature distillation, and introduces an equiangular basis vector (EBV) classifier. The dataset distillation technique compresses the dataset of each task into a small number of synthetic data that capture the essential information of the task, serving as replay exemplars. This approach reduces reliance on original data and storage costs. Feature distillation ensures that the model’s representations do not deviate significantly from previous ones. The proposed method effectively prevents an increase in the number of model parameters during the lifelong learning process by incorporating the EBV classifier, thereby maintaining model complexity stability. The performance of LLDD is validated on two bearing diagnosis cases with incremental fault types. Results demonstrate that the proposed method surpasses other lifelong learning methods in performance and memory efficiency.
{"title":"A new lifelong learning method based on dual distillation for bearing diagnosis with incremental fault types","authors":"Shijun Xie , Changqing Shen , Dong Wang , Juanjuan Shi , Weiguo Huang , Zhongkui Zhu","doi":"10.1016/j.aei.2025.103136","DOIUrl":"10.1016/j.aei.2025.103136","url":null,"abstract":"<div><div>In the rapidly evolving industrial environment, bearings may develop new fault types, posing significant challenges to deep learning-based intelligent fault diagnosis models. These models often suffer from catastrophic forgetting when encountering unknown fault types, resulting in performance degradation. Lifelong learning strategies offer a solution by enabling models to retain old knowledge while acquiring new information. However, traditional replay-based lifelong learning methods typically involve risks of privacy leakage and escalating storage costs. To address these issues, this study proposes a novel lifelong learning method called lifelong learning based on dual distillation (LLDD), which integrates a dual-distillation mechanism comprising dataset distillation and feature distillation, and introduces an equiangular basis vector (EBV) classifier. The dataset distillation technique compresses the dataset of each task into a small number of synthetic data that capture the essential information of the task, serving as replay exemplars. This approach reduces reliance on original data and storage costs. Feature distillation ensures that the model’s representations do not deviate significantly from previous ones. The proposed method effectively prevents an increase in the number of model parameters during the lifelong learning process by incorporating the EBV classifier, thereby maintaining model complexity stability. The performance of LLDD is validated on two bearing diagnosis cases with incremental fault types. Results demonstrate that the proposed method surpasses other lifelong learning methods in performance and memory efficiency.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103136"},"PeriodicalIF":8.0,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143136975","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}
Pub Date : 2025-01-21DOI: 10.1016/j.aei.2024.103105
Xin Liu , Zijuan Yang , Lin Gong , Minxia Liu , Xi Xiang , Zhenchong Mo
Diverse design requirements and the high dependency on artistic knowledge of designers make determining harmonious color schemes for web interface design challenging, calling for high-quality automatic color scheme generation. Yet, current studies are often limited to either data-driven approaches or art theories. In this paper, a conditional generative adversarial network (CGAN)-based color scheme generation method, CS-Ganerator, is proposed by integrating both knowledge and data to enable the automatic generation of color schemes for web interface design. Initially, an improved K-Means clustering algorithm is proposed and used to extract color scheme instances from a large image dataset with diverse themes. Subsequently, a CGAN model augmented with knowledge modules is employed to learn the underlying color and thematic relationships under aesthetic principles, enabling the generation of thematic color schemes. The generated schemes are then evaluated and filtered for harmony based on color theory, and categorized by warmth, darkness, and gradient to realize customized color preferences. The experimental results validate that the proposed CS-Ganerator can effectively generate diverse color schemes that highly match with the specific theme. The data and code are available at https://github.com/mzzdxg/CS-Ganerator.
{"title":"Intelligent color scheme generation for web interface color design based on knowledge − data fusion method","authors":"Xin Liu , Zijuan Yang , Lin Gong , Minxia Liu , Xi Xiang , Zhenchong Mo","doi":"10.1016/j.aei.2024.103105","DOIUrl":"10.1016/j.aei.2024.103105","url":null,"abstract":"<div><div>Diverse design requirements and the high dependency on artistic knowledge of designers make determining harmonious color schemes for web interface design challenging, calling for high-quality automatic color scheme generation. Yet, current studies are often limited to either data-driven approaches or art theories. In this paper, a conditional generative adversarial network (CGAN)-based color scheme generation method, CS-Ganerator, is proposed by integrating both knowledge and data to enable the automatic generation of color schemes for web interface design. Initially, an improved K-Means clustering algorithm is proposed and used to extract color scheme instances from a large image dataset with diverse themes. Subsequently, a CGAN model augmented with knowledge modules is employed to learn the underlying color and thematic relationships under aesthetic principles, enabling the generation of thematic color schemes. The generated schemes are then evaluated and filtered for harmony based on color theory, and categorized by warmth, darkness, and gradient to realize customized color preferences. The experimental results validate that the proposed CS-Ganerator can effectively generate diverse color schemes that highly match with the specific theme. The data and code are available at <span><span>https://github.com/mzzdxg/CS-Ganerator</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103105"},"PeriodicalIF":8.0,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137068","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}
Pub Date : 2025-01-21DOI: 10.1016/j.aei.2024.103097
Liming Xiao , Tao Fang , Guangquan Huang , Muhammet Deveci
Design concept evaluation (DCE) is crucial at the new product development stage, and effective evaluation can validate the feasibility of the new product and reduce project risks. Many valuable DCE methods have been introduced to identify the optimal one among several design concepts. However, previous methods have some drawbacks, such as the manipulation of multiple uncertainties, the determination of weights, and the identification of the best concept. To solve these problems, this paper develops a novel DCE model based on q-rung orthopair fuzzy rough (q-ROFR) sets, which employs the fuzzy-weighted zero inconsistency (FWZIC) and combined compromise solution (CoCoSo) methods. First, an evaluation environment for q-ROFR sets is provided by integrating the advantages of q-rung orthopair fuzzy sets and rough approximations, and some novel q-ROFR Einstein aggregation operators are introduced to aggregate the information to handle uncertainties more effectively. On this basis, the FWZIC method is adopted to determine the weights of criteria more reliably, and the CoCoSo method is utilized to evaluate the performance of the alternatives to determine the optimal solution more accurately and flexibly. Finally, a case study regarding the DCE of horizontal machining centers, sensitivity analysis, and comparisons are presented to validate the superiority of the proposed model. Results show that the proposed method is effective and can identify the best concepts more reliably.
{"title":"An integrated design concept evaluation method based on fuzzy weighted zero inconsistency and combined compromise solution considering inherent uncertainties","authors":"Liming Xiao , Tao Fang , Guangquan Huang , Muhammet Deveci","doi":"10.1016/j.aei.2024.103097","DOIUrl":"10.1016/j.aei.2024.103097","url":null,"abstract":"<div><div>Design concept evaluation (DCE) is crucial at the new product development stage, and effective evaluation can validate the feasibility of the new product and reduce project risks. Many valuable DCE methods have been introduced to identify the optimal one among several design concepts. However, previous methods have some drawbacks, such as the manipulation of multiple uncertainties, the determination of weights, and the identification of the best concept. To solve these problems, this paper develops a novel DCE model based on <em>q</em>-rung orthopair fuzzy rough (<em>q</em>-ROFR) sets, which employs the fuzzy-weighted zero inconsistency (FWZIC) and combined compromise solution (CoCoSo) methods. First, an evaluation environment for <em>q</em>-ROFR sets is provided by integrating the advantages of <em>q</em>-rung orthopair fuzzy sets and rough approximations, and some novel <em>q</em>-ROFR Einstein aggregation operators are introduced to aggregate the information to handle uncertainties more effectively. On this basis, the FWZIC method is adopted to determine the weights of criteria more reliably, and the CoCoSo method is utilized to evaluate the performance of the alternatives to determine the optimal solution more accurately and flexibly. Finally, a case study regarding the DCE of horizontal machining centers, sensitivity analysis, and comparisons are presented to validate the superiority of the proposed model. Results show that the proposed method is effective and can identify the best concepts more reliably.</div></div>","PeriodicalId":50941,"journal":{"name":"Advanced Engineering Informatics","volume":"65 ","pages":"Article 103097"},"PeriodicalIF":8.0,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137096","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号