Search Results (63)

Wind shear at the Hong Kong International Airport (HKIA) poses a significant safety risk due to terrain-induced airflow disruptions near the runways. Accurate assessment is essential for safeguarding aircraft during take-off and landing, as abrupt changes in wind speed or direction can compromise flight stability. This study introduces a hybrid framework for short-term wind shear prediction based on data collected from Doppler LiDAR systems positioned near the central and south runways of the HKIA. These systems provide high-resolution measurements of wind shear magnitude along critical flight paths. To predict wind shear more effectively, the proposed framework integrates a signal processing technique with a deep learning strategy. It begins with optimized variational mode decomposition (OVMD), which decomposes the wind shear time series into intrinsic mode functions (IMFs), each capturing distinct temporal characteristics. These IMFs are then modeled using bidirectional gated recurrent units (BiGRU), with hyperparameters optimized via the Tree-structured Parzen Estimator (TPE). To further enhance prediction accuracy, residual errors are corrected using Extreme Gradient Boosting (XGBoost), which captures discrepancies between the reconstructed signal and actual observations. The resulting OVMD–BiGRU–XGBoost framework exhibits strong predictive performance on testing data, achieving R² values of 0.729 and 0.926, RMSE values of 0.931 and 0.709, and MAE values of 0.624 and 0.521 for the central and south runways, respectively. Compared with GRUs, LSTM, BiLSTM, and ResNet-based baselines, the proposed framework achieves higher accuracy and a more effective representation of multi-scale temporal dynamics. It contributes to improving short-term wind shear prediction and supports operational planning and safety management in airport environments. Full article

Electric submersible progressive-cavity pumps (ESPCPs) deliver high lifting efficiency but are prone to failure in the high-temperature, high-pressure, and multiphase down-hole environment, leading to production losses and elevated maintenance costs. To achieve reliable condition recognition under these noisy and highly imbalanced data constraints, we fuse deep residual feature learning, ensemble decision-making, and generative augmentation into a unified diagnosis pipeline. A class-aware TimeGAN first synthesizes realistic minority-fault sequences, enlarging the training pool derived from 360 field records. The augmented data are then fed to a CNN backbone equipped with ResNet blocks, and its deep features are classified by a Random-Forest head (CNN-ResNet-RF). Across five benchmark architectures—including plain CNN, CNN-ResNet, GRU-based, and hybrid baselines—the proposed model attains the highest overall validation accuracy (≈97%) and the best Macro-F1, while the confusion-matrix diagonal confirms marked reductions in the previously dominant misclassification between tubing-leakage and low-parameter states. These results demonstrate that residual encoding, ensemble voting, and realistic data augmentation are complementary in coping with sparse, noisy, and class-imbalanced ESPCP signals. The approach therefore offers a practical and robust solution for the real-time down-hole monitoring and preventive maintenance of ESPCP systems. Full article

As the most basic mechanical components, bearing troubleshooting is essential to ensure the safe and reliable operation of rotating machinery. Bearing fault diagnosis is challenging due to the scarcity of bearing fault diagnosis samples and the susceptibility of fault signals to external noise. To address these issues, a ResNet-CACNN-BiGRU-SDPA bearing fault diagnosis method based on time–frequency bi-domain and feature fusion is proposed. First, the model takes the augmented time-domain signals as inputs and reconstructs them into frequency-domain signals using FFT, which gives the signals a bi-directional time–frequency domain receptive field. Second, the long sequence time-domain signal is processed by a ResNet residual block structure, and a CACNN method is proposed to realize local feature extraction of the frequency-domain signal. Then, the extracted time–frequency domain long sequence features are fed into a two-layer BiGRU for bidirectional deep global feature mining. Finally, the long-range feature dependencies are dynamically captured by SDPA, while the global dual-domain features are spliced and passed into Softmax to obtain the model output. In order to verify the model performance, experiments were carried out on the CWRU and JNU bearing datasets, and the results showed that the method had high accuracy under both small sample size and noise perturbation conditions, which verified the model’s good fault-feature-learning capability and noise immunity performance. Full article

Facial expressions involve dynamic changes, and facial expression recognition based on static images struggles to capture the temporal information inherent in these dynamic changes. The resultant degradation in real-world performance critically impedes the integration of facial expression recognition systems into intelligent sensing applications. Therefore, this paper proposes a facial expression recognition method for image sequences based on the fusion of dual neural networks (ResNet and residual bidirectional GRU—Res-RBG). The model proposed in this paper achieves recognition accuracies of 98.10% and 88.64% on the CK+ and Oulu-CASIA datasets, respectively. Moreover, the model has a parameter size of only 64.20 M. Compared to existing methods for image sequence-based facial expression recognition, the approach presented in this paper demonstrates certain advantages, indicating strong potential for future edge sensor deployment. Full article

To address the issues of poor differentiation capability for high-order signals and low average recognition rates in existing communication modulation recognition techniques, this paper first performs denoising using an entropy-based dynamic Singular Value Decomposition (SVD) method and proposes a three-channel convolutional gated recurrent units (GRU) model combined with an improved SE attention mechanism for automatic modulation recognition.The model denoises in-phase/quadrature (I/Q) signals using the SVD method to enhance signal quality. By combining one-dimensional (1D) convolutional and two-dimensional (2D) convolutional, it employs a three-channel approach to extract spatial features and capture local correlations. GRU is utilized to capture temporal sequence features so as to enhance the perception of dynamic changes. Additionally, an improved SE block is introduced to optimize feature representation, adaptively adjust channel weights, and improve classification performance. Experiments on the RadioML2016.10a dataset show that the model has a maximum classification recognition rate of 92.54%. Compared with traditional CNN, ResNet, CLDNN, GRU2, DAE, and LSTM2, the average recognition accuracy is improved by 5.41% to 8.93%. At the same time, the model significantly enhances the differentiation capability between 16QAM and 64QAM, reducing the average confusion probability by 27.70% to 39.40%. Full article

Women exhibit marked physiological transformations in pregnancy, mandating regular and holistic assessment. Maternal and fetal vitality is governed by a spectrum of clinical, demographic, and lifestyle factors throughout this critical period. The existing maternal health monitoring techniques lack precision in assessing pregnancy-related risks, often leading to late interventions and adverse outcomes. Accurate and timely risk prediction is crucial to avoid miscarriages. This research proposes a deep learning framework for personalized pregnancy risk prediction using the NFHS-5 dataset, and class imbalance is addressed through a hybrid NearMiss-SMOTE approach. Fifty-one primary features are selected via the LASSO to refine the dataset and enhance model interpretability and efficiency. The framework integrates a multimodal model (NFHS-5, fetal plane images, and EHG time series) along with two core architectures. ResRNN-Net further combines Bi-LSTM, CNNs, and attention mechanisms to capture sequential dependencies. MultiScaleFusion-Net leverages GRU and multiscale convolutions for effective feature extraction. Additionally, TabNet and MLP models are explored to compare interpretability and computational efficiency. SHAP and Grad-CAM are used to ensure transparency and explainability, offering both feature importance and visual explanations of predictions. The proposed models are trained using 5-fold stratified cross-validation and evaluated with metrics including accuracy, precision, recall, F1-score, and ROC–AUC. The results demonstrate that MultiScaleFusion-Net balances accuracy and computational efficiency, making it suitable for real-time clinical deployment, while ResRNN-Net achieves higher precision at a slight computational cost. Performance comparisons with baseline machine learning models confirm the superiority of deep learning approaches, achieving over 80% accuracy in pregnancy complication prediction. Full article

Advanced Metering Infrastructure (AMI), as a critical data collection and communication hub within the smart grid architecture, is highly vulnerable to network intrusions due to its open bidirectional communication network. A significant challenge in AMI traffic data is the severe class imbalance, where existing methods tend to favor majority class samples while neglecting the detection of minority class attacks, thereby undermining the overall reliability of the detection system. Additionally, current approaches exhibit limitations in spatiotemporal feature extraction, failing to effectively capture the complex dependencies within network traffic data. In terms of global dependency modeling, existing models struggle to dynamically adjust key features, impacting the efficiency and accuracy of intrusion detection and response. To address these issues, this paper proposes an innovative hybrid deep learning model, AS-TBR, for AMI intrusion detection in smart grids. The proposed model incorporates the Adaptive Synthetic Sampling (ADASYN) technique to mitigate data imbalance, thereby enhancing the detection accuracy of minority class samples. Simultaneously, Transformer is leveraged to capture global temporal dependencies, BiGRU is employed to model bidirectional temporal relationships, and ResNet is utilized for deep spatial feature extraction. Experimental results demonstrate that the AS-TBR model achieves an accuracy of 93% on the UNSW-NB15 dataset and 80% on the NSL-KDD dataset. Furthermore, it outperforms baseline models in terms of precision, recall, and other key evaluation metrics, validating its effectiveness and robustness in AMI intrusion detection. Full article

Ongoing advancements in machine learning show great promise for automating medical data interpretation, potentially saving valuable time in life-threatening situations. One such area is the analysis of electrocardiograms (ECGs). In this study, we investigate the effectiveness of using a DenseNet121 encoder with three decoder architectures: Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and a Transformer-based approach. We utilize these models to generate automated ECG reports from the publicly available PTB-XL dataset. Our results show that the DenseNet121 encoder paired with a GRU decoder yields higher performance than previously achieved. It achieves a METEOR (Metric for Evaluation of Translation with Explicit Ordering) score of 72.19%, outperforming the previous best result of 55.53% from a ResNet34-based model that used LSTM and Transformer components. We also discuss several important design choices, such as how to initialize decoders, how to use attention mechanisms, and how to apply data augmentation. These findings offer valuable insights into creating more robust and reliable deep learning tools for ECG interpretation. Full article

Despite significant advances in deep learning-based image captioning, many state-of-the-art models still struggle to balance visual grounding (i.e., accurate object and scene descriptions) with linguistic coherence (i.e., grammatical fluency and appropriate use of non-visual tokens such as articles and prepositions). To address these limitations, we propose a hybrid image captioning framework that integrates handcrafted and deep visual features. Specifically, we combine local descriptors—Scale-Invariant Feature Transform (SIFT) and Bag of Features (BoF)—with high-level semantic features extracted using ResNet50. This dual representation captures both fine-grained spatial details and contextual semantics. The decoder employs Bahdanau attention refined with an Attention-on-Attention (AoA) mechanism to optimize visual-textual alignment, while GloVe embeddings and a GRU-based sequence model ensure fluent language generation. The proposed system is trained on 200,000 image-caption pairs from the MS COCO train2014 dataset and evaluated on 50,000 held-out MS COCO pairs plus the Flickr8K benchmark. Our model achieves a CIDEr score of 128.3 and a SPICE score of 29.24, reflecting clear improvements over baselines in both semantic precision—particularly for spatial relationships—and grammatical fluency. These results validate that combining classical computer vision techniques with modern attention mechanisms yields more interpretable and linguistically precise captions, addressing key limitations in neural caption generation. Full article

Wearable sensors for human activity recognition (HAR) have gained significant attention across multiple domains, such as personal health monitoring and intelligent home systems. Despite notable advancements in deep learning for HAR, understanding the decision-making process of complex models remains challenging. This study introduces an advanced deep residual network integrated with a squeeze-and-excitation (SE) mechanism to improve recognition accuracy and model interpretability. The proposed model, ConvResBiGRU-SE, was tested using the UCI-HAR and WISDM datasets. It achieved remarkable accuracies of 99.18% and 98.78%, respectively, surpassing existing state-of-the-art methods. The SE mechanism enhanced the model’s ability to focus on essential features, while gradient-weighted class activation mapping (Grad-CAM) increased interpretability by highlighting essential sensory data influencing predictions. Additionally, ablation experiments validated the contribution of each component to the model’s overall performance. This research advances HAR technology by offering a more transparent and efficient recognition system. The enhanced transparency and predictive accuracy may increase user trust and facilitate smoother integration into real-world applications. Full article

Introduction: Ensuring vehicular cybersecurity is a critical challenge due to the increasing connectivity of modern vehicles, and traditional centralised learning approaches for intrusion detection pose significant privacy risks, as they require sensitive data to be shared from multiple vehicles to a central server. Objective: The aim of this study is therefore to develop an in-vehicle intrusion detection system (IVIDS) that integrates federated learning (FL) with neural networks, enabling decentralised and privacy-preserving detection of cyberattacks in vehicular networks. The proposed system extends previous research by detecting a broader range of attacks (eight types) and exploring different deep learning architectures. Methods: This study employs an extended version of the publicly available VeReMi dataset to train and evaluate multiple neural network architectures, including Multilayer Perceptrons (MLPs), Gated Recurrent Units (GRUs), and Long Short-Term Memory (LSTM) networks. Federated learning is utilised to enable collaborative model training across multiple vehicles without sharing raw data. Various data preprocessing techniques and differential privacy mechanisms are also explored. Results and Conclusions: The experimental results demonstrate that LSTM networks outperform both MLP and GRU architectures in classifying vehicular cyberattacks. The best LSTM model, trained with two previous message lags and standard normalisation, achieved a classification accuracy of 96.75% in detecting eight types of attacks, surpassing previous studies, and demonstrating the potential of applying neural networks designed to work with time series data. Full article

Traffic emissions serve as one of the most significant sources of atmospheric PM2.5 pollution in developing countries, driven by the prevalence of aging vehicle fleets and the inadequacy of regulatory frameworks to mitigate emissions effectively. This study presents a Hybrid Population-Based Training (PBT)–ResNet framework for classifying traffic-related PM2.5 levels into hazardous exposure (HE) and acceptable exposure (AE), based on the World Health Organization (WHO) guidelines. The framework integrates ResNet architectures (ResNet18, ResNet34, and ResNet50) with PBT-driven hyperparameter optimization, using data from Open-Seneca sensors along the Nairobi Expressway, combined with meteorological and traffic data. First, analysis showed that the PBT-tuned ResNet34 was the most effective model, achieving a precision (0.988), recall (0.971), F1-Score (0.979), Matthews Correlation Coefficient (MCC) of 0.904, Geometric Mean (G-Mean) of 0.962, and Balanced Accuracy (BA) of 0.962, outperforming alternative models, including ResNet18, ResNet34, and baseline approaches such as Feedforward Neural Networks (FNN), Bidirectional Long Short-Term Memory (BiLSTM), Bidirectional Gated Recurrent Unit (BiGRU), and Gene Expression Programming (GEP). Subsequent feature importance analysis using a permutation-based strategy, along with SHAP analysis, revealed that humidity and hourly traffic volume were the most influential features. The findings indicated that medium to high humidity values were associated with an increased likelihood of HE, while medium to high traffic volumes similarly contributed to the occurrence of HE. Full article

This study proposes an enhanced network intrusion detection model, 1D-TCN-ResNet-BiGRU-Multi-Head Attention (TRBMA), aimed at addressing the issues of incomplete learning of temporal features and low accuracy in the classification of malicious traffic found in existing models. The TRBMA model utilizes Temporal Convolutional Networks (TCNs) to improve the ResNet18 architecture and incorporates Bidirectional Gated Recurrent Units (BiGRUs) and Multi-Head Self-Attention mechanisms to enhance the comprehensive learning of temporal features. Additionally, the ResNet network is adapted into a one-dimensional version that is more suitable for processing time-series data, while the AdamW optimizer is employed to improve the convergence speed and generalization ability during model training. Experimental results on the CIC-IDS-2017 dataset indicate that the TRBMA model achieves an accuracy of 98.66% in predicting malicious traffic types, with improvements in precision, recall, and F1-score compared to the baseline model. Furthermore, to address the challenge of low identification rates for malicious traffic types with small sample sizes in unbalanced datasets, this paper introduces TRBMA (BS-OSS), a variant of the TRBMA model that integrates Borderline SMOTE-OSS hybrid sampling. Experimental results demonstrate that this model effectively identifies malicious traffic types with small sample sizes, achieving an overall prediction accuracy of 99.88%, thereby significantly enhancing the performance of the network intrusion detection model. Full article

With the widespread deployment of photovoltaic (PV) power stations, timely identification and rectification of module defects are crucial for extending service life and preserving efficiency. PV arrays, subjected to severe outside circumstances, are prone to defects exacerbated by dust accumulation, potentially leading to complex compound faults. The resemblance between individual and compound faults sometimes leads to misclassification. To address this challenge, this paper presents a novel hybrid deep learning model, ResGRU, which integrates a residual network (ResNet) with bidirectional gated recurrent units (BiGRU) to improve fault diagnostic accuracy. Additionally, a Squeeze-and-Excitation (SE) module is incorporated to enhance relevant features while suppressing irrelevant ones, hence improving performance. To further optimize inter-class separability and intra-class compactness, a center loss function is employed as an auxiliary loss to enhance the model’s discriminative capacity. This proposed method facilitates the automated extraction of fault features from I-V curves and accurate diagnosis of individual faults, partial shading scenarios, and compound faults under varying levels of dust accumulation, hence aiding in the formulation of efficient cleaning schedules. Experimental findings indicate that the suggested model achieves 99.94% accuracy on pristine data and 98.21% accuracy on noisy data, markedly surpassing established techniques such as artificial neural networks (ANN), ResNet, random forests (RF), multi-scale SE-ResNet, and other ResNet-based approaches. Thus, the model offers a reliable solution for accurate PV array fault diagnosis. Full article

With the rapid advancement of communication technologies, wireless networks have not only transformed people’s lifestyles but also spurred the development of numerous emerging applications and services. Against this backdrop, research on Wi-Fi-based human activity recognition (HAR) has become a hot topic in both academia and industry. Channel State Information (CSI) contains rich spatiotemporal information. However, existing deep learning methods for human activity recognition (HAR) typically focus on either temporal or spatial features. While some approaches do combine both types of features, they often emphasize temporal sequences and underutilize spatial information. In contrast, this paper proposes an enhanced approach by modifying residual networks (ResNet) instead of using simple CNN. This modification allows for effective spatial feature extraction while preserving temporal information. The extracted spatial features are then fed into a modifying GRU model for temporal sequence learning. Our model achieves an accuracy of 99.4% on the UT_HAR dataset and 99.24% on the NTU-FI HAR dataset. Compared to other existing models, RGANet shows improvements of 1.21% on the UT_HAR dataset and 0.38% on the NTU-FI HAR dataset. Full article