Search Results (346)

Fresnel high-concentration photovoltaic/thermal (HCPV/T) systems operating in sandy environments are susceptible to natural lens soiling, which attenuates the effective concentrated solar input and alters electrical and thermal performance. Natural-soiling comparison tests were conducted over 0–28 d, and an SD-CNN-BiLSTM-Attention model was developed to forecast cell-center temperature and electrical power 5, 10, and 20 min ahead. At a surface soiling density of 10.760 g·m⁻², current and electrical power decreased by 38.37% and 39.28%, respectively, relative to the concurrently operated clean-reference unit; cell-center temperature and water-tank temperature rise decreased by 7.74% and 15.13%. Thermal power also showed an overall downward trend, although the magnitude was affected by relatively large measurement uncertainty. Under grouped cross-validation, temperature RMSEs were 0.473, 0.515, and 0.555 °C at 5, 10, and 20 min, corresponding to reductions of 8.34%, 21.68%, and 44.07% relative to Persistence. Electrical-power RMSEs were 0.661, 0.618, and 0.640 W, with an 18.24% reduction relative to Persistence at 20 min. Ablation analysis showed a limited contribution from surface soiling density at 5 and 10 min but a clearer contribution at 20 min. These results support electrical and thermal performance assessment and short-term operational forecasting of Fresnel HCPV/T systems in sandy environments. Full article

Accurate continuous prediction of lower-limb joint angles from surface electromyography (sEMG) remains challenging because of the nonlinear, non-stationary, and subject-specific nature of sEMG signals, which can reduce robustness and lead to degraded prediction accuracy during highly dynamic gait phases. In this study, we propose RTGNet, a dual-branch deep learning framework for lower-limb joint-angle prediction from multichannel sEMG. The method constructs two feature views from the same sEMG stream: recurrence-plot (RP)-based representations for nonlinear texture characterization and time-series sequences for long-term temporal dependency modeling. These views are processed by a convolutional neural network (CNN) with a convolutional block attention module (CBAM) and a bidirectional long short-term memory network (BiLSTM), respectively, and integrated through an adaptive gated fusion mechanism. An enhanced Huber-TopK loss is further employed to emphasize samples with large prediction errors. Experiments on the SIAT-LLMD dataset under an offline cross-subject evaluation setting show that RTGNet achieves a mean absolute error (MAE) of 3.87°, a root mean square error (RMSE) of 5.25°, and an $R^2$ of 0.81 during walking, as well as an MAE of 4.53°, an RMSE of 6.47°, and an $R^2$ of 0.84 during stair ascent. The proposed framework outperforms temporal-only and RP-based baselines, and ablation results further support the effectiveness of the gated fusion strategy and CBAM attention. Overall, these results suggest that integrating recurrence texture and temporal dynamics is a promising strategy for sEMG-driven joint-angle prediction and provides a useful basis for future exoskeleton control-oriented studies. Full article

Heating load is jointly affected by meteorological conditions, system operating states, and historical evolution, exhibiting nonlinear, time-varying, and locally fluctuating characteristics. To improve short-term forecasting accuracy, this study proposes a particle swarm optimization (PSO)-based Trend–convolutional neural network (CNN)–bidirectional long short-term memory (BiLSTM)–Attention model. The model constructs trend-enhanced features from meteorological variables, operating parameters, temporal periodicity, historical lags, rolling statistics, differenced features, and exponentially weighted moving averages. CNN is used to extract local temporal features, BiLSTM captures bidirectional temporal dependencies, and Attention identifies key time steps. PSO further optimizes key hyperparameters. Two datasets are constructed from hourly heating-season operating data, and the proposed model is compared with BiLSTM, CNN-BiLSTM, CNN-BiLSTM-Attention, and Trend-CNN-BiLSTM-Attention models. The proposed model achieves the best performance on both datasets. For Dataset 1, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R²) are 0.628 GJ, 0.928 GJ, 9.308%, and 0.851, respectively; for Dataset 2, they are 0.3987 GJ, 0.5954 GJ, 10.32%, and 0.8709. These results indicate that trend-enhanced features and PSO improve forecasting performance and can support heating system operation scheduling. Full article

To address the problems of severe feature coupling, difficult fault information extraction, and insufficient recognition accuracy for cylindrical roller bearings under multiple fault conditions, this paper proposes a multi-fault pattern recognition method based on a Northern Goshawk Optimization algorithm improved by refraction opposition-based learning and the sine–cosine algorithm (RSNGO). The RSNGO is used to optimize variational mode decomposition (VMD) and a convolutional neural network–bidirectional long short-term memory–self-attention (CNN–BiLSTM–SAT) network. First, RSNGO adaptively optimizes the number of decomposition modes and the penalty factor of VMD, and selects the optimal intrinsic mode function (IMF) components, from which time-domain statistical features are extracted to construct the sample set. Then, a CNN–BiLSTM–SAT diagnostic network is constructed, and RSNGO is employed to jointly optimize its key hyperparameters, including convolution kernel size, number of convolution kernels, number of BiLSTM hidden units, and initial learning rate. In this network, CNN extracts local features, BiLSTM models temporal dependencies, and the self-attention mechanism enhances the representation of critical fault features. Finally, the constructed feature samples are input into the optimized network to realize multi-fault pattern recognition of cylindrical roller bearings. Experimental results demonstrate that the proposed method effectively improves the separability and recognition accuracy of multi-fault features, exhibits strong robustness and generalization capability under complex operating conditions, and provides an effective solution for intelligent bearing fault diagnosis. Full article

Flight data-driven modeling has become an important approach for trajectory prediction, anomaly detection, and risk assessment in unmanned aerial vehicles and other aerial systems. Such data are usually high-dimensional, nonlinear, multirate, and non-stationary, especially during maneuvering flight, environmental disturbance, and mission-phase transitions. Traditional physics-based methods and shallow machine learning models often have limited adaptability in these conditions, while Long Short-Term Memory (LSTM) networks and their variants have shown strong potential for learning temporal dependencies from complex flight sequences. This paper reviews the development and application of LSTM-family models for flight data analysis, with attention to both methodological performance and resource-aware deployment. The reviewed models include basic LSTM, BiLSTM, CNN-LSTM, ConvLSTM, LSTM autoencoder, attention-enhanced LSTM, graph-based LSTM, and uncertainty-aware LSTM. Their applications are discussed in three main areas: flight trajectory prediction, anomaly detection, and risk assessment. Beyond prediction accuracy, this review also examines robustness to distribution shift, physical consistency, interpretability of anomalies, uncertainty estimation, and onboard implementation. A four-layer and ten-dimensional evaluation framework is presented across data characteristics, model performance, physical-mechanism consistency, and system-engineering constraints. The evidence shows that model size, runtime memory, computational cost, inference latency, energy consumption, and target hardware are often insufficiently reported, while direct quantitative onboard validation remains scarce. This gap highlights the need for standardized deployment reporting. Full article

Elevator door systems are critical and failure-prone subsystems in elevators, with vibration signals characterized by strong nonlinearity, non-stationarity, and complex temporal dependencies under varying operating conditions. Existing fault diagnosis methods often struggle to effectively model these characteristics, particularly in capturing cross-scale dependencies between local transient disturbances and global temporal evolution patterns. To address this issue, a cross-scale spatiotemporal feature learning framework integrating adaptive time-frequency decomposition and deep feature fusion is proposed. Variational mode decomposition (VMD) is applied to decompose nonlinear vibration signals into intrinsic mode functions, while fast Fourier transform (FFT) is used to extract global spectral information, forming a multi-channel time-frequency representation. The convolutional neural network (CNN) module is designed to extract local transient features associated with mechanical impacts, while the bidirectional long short-term memory (BiLSTM) module models long-term temporal dependencies. Furthermore, a cross-attention mechanism is introduced to dynamically fuse local and global representations, enhancing discriminative feature learning under complex operating conditions. Experimental results on a representative elevator door fault dataset show that the proposed approach attained an average accuracy of 99.18% and an average F1-score of 99.13%. Compared with conventional machine learning and neural network-based methods, the proposed framework demonstrates superior diagnostic accuracy and robust performance under the evaluated experimental conditions. The findings indicate that the proposed approach serves as an efficient strategy for cross-scale spatiotemporal modeling of nonlinear and non-stationary vibration signals, with considerable promise for intelligent health monitoring and prognostic maintenance of elevator door systems. Full article

Fault signals in rotating machinery typically manifest as long time-series data embedded with local high-frequency impulses. Traditional deep learning methods often struggle to simultaneously capture these transient local impacts and model long-term global degradation features. To address this challenge, this paper proposes a novel GCU-SAM enhanced Transformer for intelligent fault diagnosis. The proposed network integrates a Gated Convolutional Unit (GCU) with a Self-Attention Mechanism (SAM). By introducing the GCU as a local inductive bias prior to the global attention module, the model dynamically captures and purifies local impulse responses via reset and update gates. Subsequently, a cascaded multi-head self-attention mechanism models long-sequence global evolution trends, forming an integrated framework for local fine-grained perception and global correlation modeling. Validated on the CWRU bearing and SEU gearbox datasets, the proposed architecture achieves superior diagnostic performance with an average F1-score of 99.01%. Compared to representative baselines, including 1D-CNN, BiLSTM, and the standard Transformer, the GCU-SAM significantly boosts diagnostic accuracy and effectively overcomes the early-stage optimization oscillations inherent in pure attention mechanisms. By achieving a deep multi-scale fusion of local abrupt changes and global degradation trends, the model exhibits exceptional feature-clustering discriminative capability and convergence stability, providing a robust and highly accurate solution for the intelligent fault diagnosis of complex rotating machinery. Full article

Network Intrusion Detection Systems (NIDS) face a trade-off between detection accuracy and computing efficiency, particularly in the edge environment with strict real-time requirements and limited resources. Current approaches rely on complicated models, which are computationally demanding, or simple ones, which sacrifice detection performance. To deal with this issue, we propose a lightweight cloud-edge cooperative two-stage NIDS architecture, which separates the real-time detection and detailed classification. At the edge, a decision tree based on feature selection is used for rapid binary classification by using only the top 10 most informative features, thus efficiently screening out abnormal traffic with minimum processing cost. Meanwhile, the cloud server identifies attack classification accurately by using a hybrid CNN-BiLSTM-Attention model to capture the spatial structures, temporal relationships, and semantic relevance. This hierarchical design effectively balances detection performance and system efficiency. Experiments conducted on UNSW-NB15, NSL-KDD, and CIC-IDS2017 datasets indicate that our suggested scheme can obtain competitive performance both at the edge and in the cloud. The edge model obtains binary classification accuracy of 86.04%, 95.27%, and 99.01%, respectively, with very low processing cost (less than 100 FLOPs per sample). The cloud model achieves multi-class accuracy of 92.23%, 97.18%, and 98.66%, respectively, with AUC values higher than 0.98. The hierarchical cloud–edge collaborative design provides an efficient and accurate solution for intrusion detection under resource-restricted situations. Full article

Background: Cardiovascular diseases (CVDs) are among the leading causes of death globally. In order to treat CVDs successfully in the early stages, it is crucial to diagnose them in time. The ECG is one of the most common and non-invasive methods to detect heart rhythms and to diagnose arrhythmias. However, the analysis of ECG recordings manually requires a lot of time and experience because the morphology of ECG signals and the characteristics of their waveforms are very complex and show large overlaps between different types of arrhythmias. So far, various approaches for automated analysis of ECG signals have been developed, mostly based on deep learning (DL). In general, these methods are able to analyze ECG signals automatically and to detect different types of arrhythmias. Most approaches, however, are based on a purely data-driven feature learning and do not pay attention to the morphology-sensitive temporal structure of ECG signals, which is important for a discriminative diagnosis of arrhythmias. Methods: In this paper, we propose an Attention-Based Morphology-Aware (ABMA) framework to leverage multilead ECG signals in conjunction with automatically computed physiological features using a hybrid deep learning architecture. ABMA leverages multi-scale convolutional neural networks to learn local morphology features, and bidirectional long short-term memory (BiLSTM) networks to model temporal rhythms in ECG signals. We designed an ABMA module that incorporates a morphology scoring network (MSN) in order to (1) estimate the morphology-aware importance of different ECG segments and (2) learn the temporal importance of ECG features. The learned attention weights enable learning to focus on key sections of ECG signals without predefined boundaries or manual annotation of fiducial points. To understand the contribution of each individual component of the framework, we performed an extensive ablation study, where we removed the handcrafted feature branch, the ABMA module, the MSN, and the multi-head attention mechanism, one at a time, and compared the results against a fixed set of experimental configurations. Results: To assess the performance of the proposed framework in three-class classification between sinus rhythm (SA), atrial fibrillation (AFIB), and ventricular tachycardia (VT), we employed a stratified 10-fold cross-validation protocol. Our approach achieved a mean accuracy of 95.18 ± 1.18%, followed by a corresponding weighted F1-score of 95.19 ± 1.18% and a macro F1-score of 94.66 ± 1.35%. Notably, the performance of the proposed complete ABMA framework considerably outperformed the baseline CNN–BiLSTM architecture. Furthermore, in the primary evaluation metrics (i.e., accuracy, F1-score), the complete framework showed statistically significant improvements against the baseline through paired two-sided t-tests (p < 0.001). The ablation study indicated that each architectural component contributed positively to the overall classification performance, with the complete ABMA framework outperforming all reduced variants. Conclusions: The framework was evaluated by stratified cross-validation on a publicly available dataset. Our framework outperformed the baseline CNN–BiLSTM model as well as the respective ablation models in terms of classification performance. The findings from the current study are based on a retrospective analysis and therefore future studies using an independent external dataset, from multiple centers, or as part of a prospective clinical study are necessary in order to establish the generalizability and clinical utility of the proposed framework. The ABMA framework is currently viewed as a very promising research framework for intelligent ECG analysis, but it is not yet a clinically validated diagnostic tool. Full article

Accurate wind power forecasting is imperative for ensuring grid stability and facilitating the large-scale integration of renewable energy—both central pillars of the global energy transition and the Dual Carbon strategic goals. However, existing methods often fail to fully capture the spatial heterogeneity and interdependencies among individual turbines, limiting their effectiveness for sustainable grid operation. To address this gap, this paper proposes an ultra-short-term wind power forecasting framework that incorporates explicit multi-dimensional spatial features. At the feature level, a 12-dimensional spatial feature system is constructed to quantify the microscale topology of wind farms. These static spatial attributes are seamlessly fused with dynamic temporal data using a dimensionality-balance factor strategy. Finally, a hybrid deep learning network comprising a multi-scale CNN, a multi-layer BiLSTM, and a multi-head self-attention mechanism is developed to capture complex spatiotemporal patterns. Experimental results on three real-world datasets show that the proposed method significantly outperforms baseline models, reducing the Mean Absolute Percentage Error by up to 11.09% and improving the coefficient of determination R² up to 0.9120. By improving forecast accuracy and robustness, the method directly supports more reliable grid dispatching, reduces curtailment of wind energy, and thus contributes to the sustainable utilization of renewable resources. These findings demonstrate that incorporating explicit spatial correlation effectively enhances the accuracy and robustness of ultra-short-term wind power forecasting, providing robust decision support for power grid dispatching and advancing the sustainability of modern power systems. Full article

Accurate structural damage identification under limited data availability and measurement noise remains a persistent challenge in structural health monitoring (SHM). This study proposes TLCA-RATNet, a transfer learning-enhanced residual attention temporal network for vibration-based damage state classification under noisy and small sample conditions. RATNet integrates adaptive threshold residual denoising, residual attention, and bidirectional gated recurrent unit (BiGRU)-based temporal modeling to suppress noise, emphasize damage-sensitive features, and capture global temporal dependencies. The current implementation is formulated as a single-task classifier, in which local feature refinement and global temporal representation are jointly optimized end-to-end through a unified damage classification objective. Transfer learning further initializes the target domain model using knowledge learned from a data-rich source structure, while regularized fine-tuning reduces overfitting on limited target samples. Experiments were conducted on a six-story lumped-mass shear structure, a three-story physical frame, and the IASC-ASCE SHM benchmark structure, using 10–145 training samples per damage class and additive-noise conditions ranging from 0 to 30 dB signal-to-noise ratio. On Dataset 2 at 5 dB, TLCA-RATNet achieved an accuracy of 89.86%, exceeding LSTM and CNN-BiGRU by 9.94 and 13.10 percentage points, respectively. On Dataset 3 at 0 dB, it achieved 86.00% accuracy, outperforming CNN-BiGRU by 10.72 percentage points. In the limited sample transfer experiment on Dataset 2, transfer learning increased the accuracy from 93.05% to 100.00%, representing a gain of 6.95 percentage points over training from scratch. These results indicate that TLCA-RATNet provides a data-efficient and noise-robust approach for damage state screening and rapid model adaptation in SHM applications with scarce labeled data and noisy measurements. Full article

Neurodevelopmental disorders (NDDs) are associated with impairments in communication, behavior, and social interaction, making accurate diagnosis clinically challenging. Autism Spectrum Disorder (ASD), a major NDD, often exhibits atypical speech patterns characterized by altered prosody and reduced emotional expressiveness. The study proposes a hybrid dual-path framework for ASD detection from emotional speech using two strategies: PCA–GMM-based acoustic modeling and a CNN–BiLSTM–Attention architecture for spectral–temporal feature learning. The proposed framework captures probabilistic, spectral, and temporal speech characteristics for robust ASD classification. Acoustic analysis demonstrated clear separability between ASD and non-ASD speech, while the deep learning framework achieved stable and reliable performance across multiple emotional conditions. Experimental evaluation achieved 98.3% accuracy, AUC values ranging from 0.9699 to 0.9864, and F1-scores up to 0.9891. The findings highlight the potential of AI-driven speech analysis as a scalable and non-invasive tool for early ASD screening and predictive healthcare applications. Full article

Respiratory diseases are a major global health challenge. However, identification of respiratory diseases is often limited by subjectivity, environmental noise and inter-clinician variability. This study presents an explainable multimodal deep learning framework for recording-level multiclass classification of respiratory audio signals. The proposed system integrates two complementary representations—a spectro-temporal encoder based on a CNN–BiLSTM-attention architecture and a handcrafted acoustic-feature encoder capturing acoustic descriptors commonly used in respiratory-audio analysis, including MFCCs, zero-crossing rate, spectral centroid, spectral bandwidth, chroma, RMS energy, and spectral rolloff features. These branches are combined through late-stage fusion to leverage both data-driven representation learning and domain-informed acoustic cues. The proposed model was trained and internally evaluated on the Asthma Detection Dataset Version 2, comprising five respiratory categories: bronchial disease, asthma, COPD, healthy, and pneumonia. Mono conversion, resampling to 16 kHz, 100–2000 Hz band-pass filtering, amplitude normalisation, fixed 4 s trimming or zero-padding, training-only augmentation, handcrafted-feature extraction, mel-spectrogram generation, quality control auditing, and stratified recording-level partitioning have been applied in the pre-processing steps. Across five repeated experiments with different random seeds, the proposed hybrid model achieved a mean held-out recording-level test accuracy of $0.9099\pm 0.0163$, balanced accuracy of $0.8936\pm 0.0152$, macro F1-score of $0.8937\pm 0.0177$, macro ROC–AUC of $0.9867\pm 0.0010$, and macro PR–AUC of $0.9489\pm 0.0044$. Conventional machine learning baseline comparisons showed that the proposed model achieved stronger internal accuracy, balanced accuracy, macro recall, macro F1-score, and macro ROC–AUC than classical machine learning algorithms trained on handcrafted acoustic features, although Random Forest remained competitive in macro PR–AUC. Ablation analysis shows that the deep spectro-temporal branch was the primary contributor to predictive performance, while the handcrafted branch provided complementary interpretable acoustic information rather than consistently improving all classification metrics. Explainability was incorporated using Grad-CAM and Integrated Gradients for spectrogram-based interpretation and SHAP for handcrafted-feature attribution. Domain-shift evaluation on the ICBHI Respiratory Sound Database and a COPD-focused cohort revealed substantial dataset shift effects, including poor healthy-case recognition on ICBHI and seed-dependent COPD recognition in the COPD-focused cohort. Identifier-aware sensitivity analyses showed lower performance than the main recording-level split, suggesting that subject-like or source-level overlap may inflate internal performance estimates. The findings should be interpreted as promising internal held-out recording-level algorithmic performance with limited external transfer, rather than evidence of readiness for clinical use. Full article

Cross-session variability in neural signal statistics often leads to distribution shifts that substantially degrade the generalization performance of motor decoding models. To address this challenge, this study proposes a CNN-enhanced Transformer with distribution alignment (CNN-Trans-DA), which integrates local temporal feature extraction with global sequence modeling and a lightweight domain adaptation strategy. A one-dimensional convolutional module is introduced prior to the Transformer encoder to capture local firing patterns, while multi-head attention is used to learn long-range temporal dependencies. For cross-session adaptation, we design a joint distribution alignment method that simultaneously minimizes marginal and conditional distribution discrepancies through maximum mean discrepancy (MMD) and correlation alignment (CORAL). Using four-direction monkey fingertip movement data, the proposed CNN-Trans-DA model achieves an accuracy of 84.64 percent under single-session decoding, outperforming the standard Transformer (81.22%), CNN (80.78%), and BiLSTM (79.29%). In cross-session transfer experiments, the model maintains an 82.42 percent decoding accuracy, showing clear improvements over models without distribution alignment. These results demonstrate that the proposed approach can effectively handle session-induced distribution drift and achieve robust motor intention decoding, providing a scalable and generalizable solution for cross-session brain–machine interface applications. Full article

Extreme and sudden temperature fluctuations observed as a result of global climate change increase the environmental pressure on energy transmission infrastructure. These meteorological changes significantly increase the risk of failure for porcelain insulators, which exhibit low thermal resistance and are susceptible to sudden arcing and surface deformations. In this study, a hybrid CNN–BiLSTM–SE architecture augmented with the Squeeze-and-Excitation attention mechanism is proposed using surface leakage current signals to diagnose healthy, cracked, and broken structural conditions in three-unit porcelain insulators. The SE block in the architecture dynamically rescales feature maps from CNN layers on a channel-by-channel basis. Thus, it highlights the signal characteristic that is dominant for fault diagnosis just before the BiLSTM units learn temporal dependencies. Leakage current data were obtained under an experimental setup at 60 kV for 15 different conditions covering all possible combinations of healthy, cracked, and broken insulator units. The raw signals were preprocessed with the Savitzky–Golay filter to suppress noise while preserving the diagnostic waveform morphology. 24 features covering time-domain statistics, frequency-domain spectral characteristics, and wavelet-domain energy components were extracted and used as model inputs. The CNN–BiLSTM–SE architecture achieved a classification accuracy of 93.83%, surpassing the standalone CNN (88.89%), BiLSTM (87.65%), and CNN–BiLSTM (91.36%) models, as well as classical machine-learning baselines (SVM: 87.65%, Random Forest: 90.12%, Boosted Trees: 87.65%). Full article