A novel semi-supervised learning rolling bearing fault diagnosis method based on SNNGAN DOI
Zhi Qiu,

Shanfei Fan,

Haibo Liang

et al.

Measurement Science and Technology, Journal Year: 2024, Volume and Issue: 35(8), P. 086135 - 086135

Published: May 16, 2024

Abstract In practical industrial environments, rotating machinery typically operates under normal conditions. As a result, the signals collected are primarily signals. This imbalance in sample data diminishes effectiveness of fault diagnosis. To address this issue, paper produces novel semi-supervised diagnosis approach based on Siamese neural network combined with generative adversarial (SNNGAN) to enhance classification accuracy. Firstly, vibration subjected continuous wavelet transformation obtain time–frequency representations, which utilized for pre-training convolutional encoders generator and discriminator. Subsequently, cosine similarity algorithm is employed ensure quality generated samples. For data, set threshold. Those surpassing threshold assigned their corresponding labels added original set. Otherwise, those falling below transformed back into vectors through an inverse transform then serve as input create new Finally, experiments conducted newly balanced four imbalanced experiments, results demonstrate that SNNGAN outperforms other methods average accuracy, G-mean, F1 score, accuracy values 0.919, 0.948, 0.927, 0.953 respective datasets. Therefore, exhibits outstanding performance conditions imbalance.

Language: Английский

Ternary Precursor Centrifuge Rolling Bearing Fault Diagnosis Based on Adaptive Sample Length Adjustment of 1DCNN-SeNet DOI Open Access
Feng Xu, Zhen Sui,

Jiangang Ye

et al.

Processes, Journal Year: 2024, Volume and Issue: 12(4), P. 702 - 702

Published: March 29, 2024

To address the issues of uneven sample lengths in centrifuge machine bearings ternary precursor, inaccurate fault feature extraction, and insensitivity important channels rolling bearings, a bearing diagnosis method based on adaptive length adjustment one-dimensional convolutional neural network (1DCNN) squeeze-and-excitation (SeNet) is proposed. Firstly, by controlling cumulative variance contribution rate principal component analysis algorithm, achieved, reducing data with to same dimensionality for various classes. Then, 1DCNN extracts local features from signals through convolution-pooling operations, while SeNet introduces channel attention mechanism which can adaptively adjust importance between different channels. Finally, 1DCNN-SeNet model compared four classic models experimental CWRU dataset. The results indicate that proposed exhibits high diagnostic accuracy demonstrating good adaptability generalization capabilities.

Language: Английский

Citations

4
Automatic sleep stage classification using deep learning: signals, data representation, and neural networks DOI Creative Commons
Peng Liu, Wei Qian, Hua Zhang

et al.

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 57(11)

Published: Sept. 23, 2024

Language: Английский

Citations

4
Ensemble of convolution neural networks on heterogeneous signals for sleep stage scoring DOI Creative Commons
Enrique Fernández-Blanco, Carlos Fernández-Lozano, Alejandro Pazos

et al.

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: April 2, 2025

Language: Английский

Citations

0
Meta-narrative review: the impact of music therapy on sleep and future research directions DOI Creative Commons

Qiaoqiao Gou,

Meihui Li,

Xiaoyu Wang

et al.

Frontiers in Neurology, Journal Year: 2025, Volume and Issue: 15

Published: Jan. 7, 2025

Sleep is essential to human health, yet 27% of the global population suffers from sleep issues, which often lead fatigue, depression, and impaired cognitive function. While pharmacological treatments exist, non-pharmacological approaches like music therapy have shown promise in enhancing quality. This review, analyzing 27 studies with various experimental paradigms, confirms that significantly improves subjective quality, largely by alleviating anxiety regulating mood through perceptual pathways. However, effects on objective measures remain inconclusive, suggesting individual differences may play a significant role. Future research should focus refining intervention designs integrate both assessments better elucidate physiological psychological mechanisms therapy. Key recommendations include personalized selection, development age-appropriate interventions, minimization external interferences maximize therapeutic outcomes. Additionally, incorporating variables status, lifestyle, environmental factors offer more comprehensive understanding therapy's long-term adaptability effectiveness for diverse populations. review offers critical directions practical support future applications health.

Language: Английский

Citations

0
Enhancing sleep stage classification through simultaneous time–frequency tokenization DOI
Qiaoli Zhou, Shurui Li,

Xiyuan Ye

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107553 - 107553

Published: Feb. 20, 2025

Language: Английский

Citations

0
LVR: A language and vision fusion method for rice diseases segmentation under complex environment DOI
Tianrui Zhao,

Honglin Zhou,

Miying Yan

et al.

European Journal of Agronomy, Journal Year: 2025, Volume and Issue: 168, P. 127599 - 127599

Published: March 13, 2025

Language: Английский

Citations

0
Supervised autoregressive eXogenous Networks with Fractional Grünwald–Letnikov finite differences: Tumor Evolution and Immune Responses under Therapeutic Influence fractals model DOI
Hassan Raza,

Muhammad Junaid Ali Asif Raja,

Rikza Mubeen

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 107, P. 107871 - 107871

Published: March 30, 2025

Language: Английский

Citations

0
Transforming physical fitness and exercise behaviors in adolescent health using a life log sharing model DOI Creative Commons
Shanshan Wang, J. H. Liu

Frontiers in Public Health, Journal Year: 2025, Volume and Issue: 13

Published: April 4, 2025

Introduction This study investigates the potential of a deep learning-based Life Log Sharing Model (LLSM) to enhance adolescent physical fitness and exercise behaviors through personalized public health interventions. Methods We developed hybrid Temporal–Spatial Convolutional Neural Network-Bidirectional Long Short-Term Memory (TS-CNN-BiLSTM) model. model integrates temporal, textual, visual features from multimodal life log data (exercise type, duration, intensity) classify predict activity behaviors. Two datasets, Geo-Life (with location data) Time-Life (without data), were constructed evaluate impact spatial information on classification performance. The utilizes CNNs for local feature extraction BiLSTM networks capture temporal dynamics, maintaining user privacy. Results TS-CNN-BiLSTM achieved an average accuracy 99.6% across eight types, outperforming state-of-the-art methods by 1.9–4.4%. Temporal identified as crucial detecting recurring behavioral trends periodic patterns. Discussion These findings demonstrate efficacy integrating with learning accurate classification. high supports its developing promotion strategies, including tailored interventions, incentives, social support mechanisms, engagement in activities advance education management.

Language: Английский

Citations

0
MtCLSS: Multi-Task Contrastive Learning for Semi-Supervised Pediatric Sleep Staging DOI
Yamei Li, Shengqiong Luo, Haibo Zhang

et al.

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2022, Volume and Issue: 27(6), P. 2647 - 2655

Published: Oct. 10, 2022

The continuing increase in the incidence and recognition of children's sleep disorders has heightened demand for automatic pediatric staging. Supervised stage algorithms, however, are often faced with challenges such as limited availability physicians data heterogeneity. Drawing upon two quickly advancing fields, i.e., semi-supervised learning self-supervised contrastive learning, we propose a multi-task strategy recognition, abbreviated MtCLSS. Specifically, signal-adapted transformations applied to electroencephalogram (EEG) recordings full night polysomnogram, which facilitates network improve its representation ability through identifying transformations. We also introduce an extension loss function, thus adapting setting. In this way, proposed framework learns not only task-specific features from small amount supervised data, but extracts general signal transformations, improving model robustness. MtCLSS is evaluated on real-world dataset promising performance (0.80 accuracy, 0.78 F1-score 0.74 kappa). examine generality well-known public dataset. experimental results demonstrate effectiveness EEG based staging very labeled scenarios.

Language: Английский

Citations

17
LVF: A language and vision fusion framework for tomato diseases segmentation DOI

Yang Hu,

Jiale Zhu, Guoxiong Zhou

et al.

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 227, P. 109484 - 109484

Published: Oct. 9, 2024

Language: Английский

Citations

3