Heterogeneous Multi-Sensor Fusion for AC Motor Fault Diagnosis via Graph Neural Networks

Electronics, Год журнала: 2025, Номер 14(10), С. 2005 - 2005

Опубликована: Май 15, 2025

Multi-sensor fault diagnosis, especially when using heterogeneous sensors, substantially enhances the accuracy of detection in asynchronous motors operating under high-interference conditions. A critical challenge multi-sensor diagnosis lies effectively fusing data from different sensors. Deep learning offers a promising solution by transforming into unified representation, thereby facilitating robust fusion. However, existing approaches often fail to fully exploit inter-sensor correlations and inherent prior physical knowledge. To address this limitation, we propose novel graph neural network-based model that emphasizes structure construction for information Our framework includes (1) multi-task enhanced autoencoder node feature extraction, enabling discriminative representation learning, particularly with sensor data; (2) an adjacency matrix builder integrated constraints improve generalization robustness model; (3) isomorphism network derive graph-level representations classification. experimental results demonstrate model’s effectiveness diagnosing faults, as it achieves superior performance compared conventional methods on two motor datasets.

Язык: Английский

---

A Comprehensive Survey of Multi-View Intelligent Fault Diagnosis Tailored to the Sensor, Machinery Equipment, and Industrial System Faults

Journal of Vibration Engineering & Technologies, Год журнала: 2025, Номер 13(5)

Опубликована: Май 29, 2025

Язык: Английский

---

Multiscale cyclic frequency demodulation-based feature fusion framework for multi-sensor driven gearbox intelligent fault detection

Knowledge-Based Systems, Год журнала: 2023, Номер 283, С. 111203 - 111203

Опубликована: Ноя. 17, 2023

Язык: Английский

---

Assessing Sensor Integrity for Nuclear Waste Monitoring Using Graph Neural Networks

Sensors, Год журнала: 2024, Номер 24(5), С. 1580 - 1580

Опубликована: Фев. 29, 2024

A deep geological repository for radioactive waste, such as Andra’s Cigéo project, requires long-term (persistent) monitoring. To achieve this goal, data from a network of sensors are acquired. This is subject to deterioration over time due environmental effects (radioactivity, mechanical the cell, etc.), and it paramount assess each sensor’s integrity ensure consistency enable precise monitoring facilities. Graph neural networks (GNNs) suitable detecting faulty in complex because they accurately depict physical phenomena that occur system take sensor network’s local structure into consideration predictions. In work, we leveraged availability experimental acquired Underground Research Laboratory (URL) train graph assessment integrity. The experiment considered work emulated thermal loading high-level waste (HLW) demonstrator cell (i.e., heating containment by nuclear waste). Using real URL layer was one novelties work. used model GNN inputted temperature field (at current past steps) returned state individual sensor, i.e., or not. other novelty lay application GraphSAGE which modified with elements Net framework detect sensors, up half being at once. proportion explained use distributed (optic fiber) on cell. GNNs trained were ultimately compared against standard classification methods (thresholding, artificial networks, demonstrated their effectiveness

Язык: Английский

---

Feature Enhancement via Linear Transformation and Its Application in Fault Diagnosis

IEEE Internet of Things Journal, Год журнала: 2024, Номер 11(12), С. 21895 - 21903

Опубликована: Март 18, 2024

The performance of neural networks is directly affected by the features obtained from backbones fault diagnosis networks. To obtain clear and improve networks, this paper constructs a new block based on linear transformation. Firstly, feature vector divided into decisive component an invalid component. Then, it worth noting that orthogonality these two components beneficial to model learning. According this, are extracted using spaces constructed relationships between four fundamental sub-spaces matrix. In sub-spaces, row space null employed extract useless component, respectively. Both implemented layers designed as encoder-decoder structure ensure existence space. spaces, constraint term proposed modify their weights. Lastly, cosine similarity input entirely. When incorporating some classic classifying they can achieve improved accuracy. Moreover, when comparing conventional spatial attention mechanisms, module demonstrates superior overall performance, including accuracy, antinoise ability, generalization ability.

Язык: Английский

---