High-Order Neighbors Aware Representation Learning for Knowledge Graph Completion

IEEE Transactions on Neural Networks and Learning Systems, Journal Year: 2024, Volume and Issue: 36(3), P. 5273 - 5287

Published: April 22, 2024

As a building block of knowledge acquisition, graph completion (KGC) aims at inferring missing facts in graphs (KGs) automatically. Previous studies mainly focus on convolutional network (GCN)-based KG embedding (KGE) to determine the representations entities and relations, accordingly predicting triplets. However, most existing KGE methods suffer from limitations tail that are far away or even unreachable KGs. This limitation can be attributed related high-order information being largely ignored. In this work, we learning neighbors KGs improve performance prediction. Specifically, first introduce set new nodes called pedal augment for facilitating message passing between entities, effectively injecting into entity representation. Additionally, propose strength-guided neural networks aggregate neighboring representations. To address issue transmitting irrelevant higher order through nodes, which potentially hurt representation, further dynamically integrate aggregated representation each node with its corresponding self-representation. Extensive experiments have been conducted three benchmark datasets results demonstrate superiority our method compared strong baseline models.

Language: Английский

---

A Federated Learning Framework for Cloud–Edge Collaborative Fault Diagnosis of Wind Turbines

IEEE Internet of Things Journal, Journal Year: 2024, Volume and Issue: 11(13), P. 23170 - 23185

Published: April 25, 2024

In modern Internet of Things-enhanced wind power systems, most existing data-driven fault diagnosis approaches for turbines (WTs) are performed under a centralized paradigm that ignores data privacy. Recently, federated learning (FL) presented solution to enable edge WTs located at isolated sites collaboratively learn shared model without accessing local privacy-sensitive data. However, the practical issues label heterogeneity among clients and scarcity labeled still severely impede generation satisfactory model. To address these issues, we propose diagnostic knowledge-based FL framework (DKFLWT) collaborative distributed WTs. our DKFLWT framework, independently learned knowledge from each client, rather than parameters in conventional FL, is uploaded cloud server enrich client-specific information visible mitigate adverse effects on performance caused by heterogeneity. enhance overall efficiency develop two-stage, single-round training mechanism, which serves as universal platform can accommodate customized requirements users, implying convenient integration semi-supervised scenarios with limited Furthermore, spatio-temporal memory-enhanced autoencoder designed sufficiently exploit essential different patterns client. Experimental results demonstrate superior an improvement more 22.1% accuracy 37.2% against several compared methods all seriously heterogeneous scenarios.

Language: Английский

---

Human reliability analysis of offshore high integrity pressure protection system based on improved CREAM and HCR integration method

Ocean Engineering, Journal Year: 2024, Volume and Issue: 307, P. 118153 - 118153

Published: May 14, 2024

Language: Английский

---

Multidimensional information fusion and broad learning system-based condition recognition for energy pipeline safety

Knowledge-Based Systems, Journal Year: 2024, Volume and Issue: 300, P. 112259 - 112259

Published: July 17, 2024

Language: Английский

---

Methods and experiments for automatic control of surface back pressure based on Dung beetle optimizer-PID controller

Process Safety and Environmental Protection, Journal Year: 2024, Volume and Issue: 191, P. 360 - 374

Published: Aug. 30, 2024

Language: Английский

---

Intelligent full-stage stable fault diagnosis method for subsea production system

Ocean Engineering, Journal Year: 2024, Volume and Issue: 312, P. 119309 - 119309

Published: Sept. 20, 2024

Language: Английский

---

Fault diagnosis of regenerative thermal oxidizer system via dynamic uncertain causality graph integrated with early anomaly detection

Process Safety and Environmental Protection, Journal Year: 2023, Volume and Issue: 179, P. 724 - 734

Published: Sept. 23, 2023

Language: Английский

---

Three-model-driven fault diagnosis method for complex hydraulic control system: Subsea blowout preventer system as a case study

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 247, P. 123297 - 123297

Published: Jan. 23, 2024

Language: Английский

---

A critical review on machine learning applications in fiber composites and nanocomposites: Towards a control loop in the chain of processes in industries

Materials & Design, Journal Year: 2024, Volume and Issue: 245, P. 113247 - 113247

Published: Aug. 13, 2024

• Statistics and language processing tools were used to analysis papers. Fiber polymer composites largely in machine learning applications. Nanoparticles as reinforcement nonpolymeric matrices less frequent. Neural deep neural networks frequently used. Machine applications mostly occurred structure health monitoring. must be evaluated achieve correct use various fields. Their properties, performance, condition, integrity can quickly predicted optimized by (ML), after extensive training, compared with experiments conventional computational simulations. In this document, papers on ML fiber collected critically reviewed. It was revealed that kind environments have been primarily Supervised has more than unsupervised ML, whereas some specific semi–supervised ( e.g. , learning) or predictive control overlooked. Most successful the laboratory scale short term. Furthermore, deployment of addition, retroactive feedback from manufacturing polymers composite laminates structures neglected. Accordingly, a loop chain processes discussed. Additionally, statistics summarize analyze Finally, it proposed multiscale modeling using physics is potential approach advance predictions for future Therefore, physicochemical interactions (van der Waals electrostatic) nanoscale included.

Language: Английский

---

Fault Detection for Medium Voltage Switchgear Using a Deep Learning Hybrid 1D-CNN-LSTM Model

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 97574 - 97589

Published: Jan. 1, 2023

Medium voltage (MV) switchgear is a vital part of modern power systems, responsible for regulating the flow electrical and ensuring safety equipment personnel. However, can experience various types faults that compromise its reliability safety. Common in include arcing, tracking, corona, normal cases, mechanical faults. Accurate detection these essential maintaining MV switchgear. In this paper, we propose novel approach fault using hybrid model (1D-CNN-LSTM) both time domain (TD) frequency (FD). The proposed involves gathering dataset operation data pre-processing it to prepare training. then trained on dataset, performance evaluated testing phase. results phase demonstrate effectiveness detecting achieved 100% accuracy domains classifying Switchgear, including Additionally, 98.4% corona TD. study provides an effective efficient By learning spatial temporal features simultaneously, accurately classify TD FD. This has significant potential improve as well other industrial applications.

Language: Английский

---