Photovoltaic Array Fault Diagnosis and Localization Method Based on Modulated Photocurrent and Machine Learning DOI Creative Commons
Yebo Tao, Tingting Yu, Jiayi Yang

et al.

Sensors, Journal Year: 2024, Volume and Issue: 25(1), P. 136 - 136

Published: Dec. 29, 2024

Photovoltaic arrays are exposed to outdoor conditions year-round, leading degradation, cracks, open circuits, and other faults. Hence, the establishment of an effective fault diagnosis system for photovoltaic is paramount importance. However, existing methods often trade off between high accuracy localization. To address this concern, paper proposes a identification localization approach based on modulated photocurrent machine learning. By irradiating different frequency-modulated light, method separates directly measures photoelectric conversion efficiency each panel, achieving both Through learning classification algorithms, current amplitude frequency panel identified achieve Compared methods, strengths lie in its ability high-speed high-accuracy by measuring only short-circuit current. Additionally, equipment cost low. The feasibility proposed demonstrated through practical experimentation. It determined that when utilizing neural network algorithm, speed meets measurement requirements (5800 obs/s), optimal (97.8%).

Language: Английский

Biologically inspired compound defect detection using a spiking neural network with continuous time–frequency gradients DOI
Zisheng Wang, Shaochen Li, Jianping Xuan

et al.

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 65, P. 103132 - 103132

Published: Jan. 24, 2025

Language: Английский

Citations

1
Defend Against Property Inference Attack for Flight Operations Data Sharing in FedMeta Framework DOI Creative Commons
Jin Lei,

Weiyun Li,

Meng Yue

et al.

Aerospace, Journal Year: 2025, Volume and Issue: 12(1), P. 41 - 41

Published: Jan. 11, 2025

Flight operations data play a central role in ensuring flight safety, optimizing operations, and driving innovation. However, these have become key target for cyber-attacks, are especially vulnerable to property inference attacks. Aiming at attacks shared application model training, we proposed FedMeta-CTGAN, novel approach that leverages federated meta-learning conditional tabular generative adversarial networks (CTGANs) protect data. Motivated by the need secure sharing aviation, as highlighted Federal Aviation Administration’s requirement ADS-B Out equipment on aircraft create situational awareness environment, our method aims prevent sensitive information leakage while maintaining performance. FedMeta-CTGAN exploits natural privacy-preserving properties of two-stage update meta-learning, using real train CTGAN synthetic fake query during meta-training. Comprehensive experiments operation dataset demonstrate effectiveness method. adapts quickly unbalanced data, achieving prediction accuracy 96.33%, reducing attacker’s AUC score 0.51 under Our contribution lies development efficient data-sharing solution which has potential revolutionize aviation industry.

Language: Английский

Citations

0
A MID-1DC+LRT Multi-Task Model for SOH Assessment and RUL Prediction of Mechanical Systems DOI Creative Commons

Hai Yang,

Xudong Yang, Dong Sun

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1368 - 1368

Published: Feb. 23, 2025

Predictive health management (PHM) plays a pivotal role in the maintenance of contemporary industrial systems, with evaluation state (SOH) and prediction remaining useful life (RUL) constituting its central objectives. Nevertheless, existing studies frequently approach these tasks isolation, overlooking their interdependence, predominantly concentrate on single-condition settings. While Transformers have demonstrated exceptional performance RUL prediction, substantial parameter requirements pose challenges to computational efficiency practical implementation. Further, multi-task learning (MTL) models often experience deterioration as result imbalanced weighting loss functions. To address challenges, MID-1DC+LRT model was proposed present study. The integrates multi-input data 1D convolutional neural network (1D-CNN) low-rank transformer (LRT) within an MTL framework. This processes high-dimensional sensor data, multi-condition indicator optimizing Transformer structure reduce complexity. A homoscedastic uncertainty-based method dynamically adjusts function weights, improving task collaboration generalization. results demonstrate that significantly outperformed methods SOH assessment under scenarios, demonstrating superior accuracy efficiency, especially complex dynamic environments.

Language: Английский

Citations

0
Federated learning based on dynamic hierarchical game incentives in Industrial Internet of Things DOI
Y. A. Tang, Lina Ni, Jufeng Li

et al.

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 65, P. 103214 - 103214

Published: Feb. 24, 2025

Language: Английский

Citations

0
Neuromorphic computing-enabled generalized machine fault diagnosis with dynamic vision DOI
Changhao Liu, Xiang Li, X Chen

et al.

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 65, P. 103300 - 103300

Published: April 4, 2025

Language: Английский

Citations

0
Hydropower projects investment estimation artificial intelligence model by similarity measure-based few-shot machine learning DOI
Mengnan Shi, Xinyu Qu,

Hongtao Li

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 153, P. 110891 - 110891

Published: April 21, 2025

Language: Английский

Citations

0
A motor bearing fault diagnosis model based on multi-adversarial domain adaptation DOI Creative Commons
Xinming Liu, Ruiming Zhang,

Jin-Ping Li

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Nov. 23, 2024

In response to the weakened capability of feature transfer and parameter distribution alignment across domains due significant differences in data collected by different devices, this paper constructs a motor bearing fault diagnosis model based on multi-adversarial domain adaptation. Initially, an improved residual network is employed as extraction module enhance capabilities. It then incorporates Selective Kernel Network (SKNet) implement attention mechanisms convolutional kernels, Global Context (GCNet) effectively utilize global contextual information for re-weighting channels. Additionally, uses multi-kernel maximum mean discrepancy measure between classes, establishing dynamic adjustment factor conjunction with multiple discriminators modulate importance marginal conditional distributions. Ultimately, proposed was applied experiments operating conditions demonstrating excellent diagnostic results.

Language: Английский

Citations

2
DouN-GNN:Double nodes graph neural network for few-shot learning DOI
Yan Zhang, Xudong Zhou, Nian Wang

et al.

Neurocomputing, Journal Year: 2024, Volume and Issue: unknown, P. 128970 - 128970

Published: Nov. 1, 2024

Language: Английский

Citations

0
Research on bearing fault diagnosis based on a multimodal method DOI Creative Commons
Hao Chen, Shengjie Li, Xi Lu

et al.

Mathematical Biosciences & Engineering, Journal Year: 2024, Volume and Issue: 21(12), P. 7688 - 7706

Published: Jan. 1, 2024

As an essential component of mechanical systems, bearing fault diagnosis is crucial to ensure the safe operation equipment. However, vibration data from bearings often exhibit non-stationary and nonlinear features, which complicates diagnosis. To address this challenge, paper introduces a novel multi-scale time-frequency statistical features fusion model (MTSF-FM). Specifically, method first employs continuous wavelet transform generate images, capturing local global signal at different scales. Contrast enhancement techniques are then used improve visual quality these images. Next, extracted images using geometry group network obtain deep image modalities. In parallel, 13 key original in domain. Convolutional neural networks employed for feature extraction. Experimental results demonstrate that MTSF-FM achieves accuracies 98.5% 95.1% on two public datasets. These findings highlight effectiveness analyzing complex propose

Language: Английский

Citations

0
Knowledge Distillation‐Based Zero‐Shot Learning for Process Fault Diagnosis DOI Creative Commons
Yi Liu, Jiajun Huang, Mingwei Jia

et al.

Advanced Intelligent Systems, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Data‐driven deep learning is effective in diagnosing known faults, but not so well when new or unknown faults occur. With as a zero‐shot problem, this article proposes method for detecting and isolating based on knowledge distillation within teacher–student framework. Process data image are equivalent their spatiotemporal dimensions, convolutional neural networks selected the teacher model, pretrained data. Information under both normal fault conditions then effectively extracted from process by well‐trained model. Subsequently, used to transfer only of model student When an arises, there exist differences between information Contributions variables calculated quantifying these through gradients, thereby fault. Finally, compared with series baseline methods two state‐of‐the‐art methods, proposed improves diagnosis accuracy 3.08% 26.13% Tennessee Eastman 3.48% 41.45% sour water treatment process. Additionally, physical consistency isolation visually assessed.

Language: Английский

Citations

0