Explore the vast range of titles available.

Bolt looseness may occur on wind turbine (WT) blades exposed to operational and environmental variability conditions, which sometimes can cause catastrophic consequences. Therefore, it is necessary to monitor the loosening state of WT blade root bolts. In order to solve this problem, this paper proposes a method to monitor the looseness of blade root bolts using the sensors installed on the WT blade. An experimental platform was first built by installing acceleration and strain sensors for monitoring bolt looseness. Through the physical experiment of blade root bolts' looseness, the response data of blade sensors is then obtained under different bolt looseness numbers and degrees. Afterwards, the sensor signal of the blade root bolts is analyzed in time domain, frequency domain, and time‐frequency domain, and the sensitivity features of various signals are extracted. So the eigenvalue category as the input of the state discrimination model was determined. The LightGBM (light gradient boosting machine) classification algorithm was applied to identify different bolt looseness states for the multi‐domain features. The impact of different combinations of sensor categories and quantities as the data source on the identification results is discussed, and a reference for the selection of sensors is provided. The proposed method can discriminate four bolt states at an accuracy of around 99.8% using 5‐fold cross‐validation.

Purpose Schizophrenia (SZ) is a severe mental disorder. Using electroencephalogram (EEG) signals for objective and accurate recognition of SZ is critical for timely clinical intervention. However, due to the highly non‐stationary nature of EEG signals and the complex spatial correlations of neural activities, existing recognition methods still face key challenges, including incomplete feature extraction, limited computational efficiency, and insufficient modeling of long‐range temporal dependencies. Methods To address these issues, this paper proposes a multi‐domain EEG feature fusion network (MDEFusion). By synergistically introducing a half‐split crossover sparse autoencoder (HCSAE) and a bidirectional attention long short‐term memory (BALSTM) network, the model achieves efficient fusion of multi‐domain features and the modeling of long‐range temporal dependencies. Specifically, MDEFusion constructs an improved SAE with a half‐split crossover mechanism to perform nonlinear cross‐fusion of features across the time, frequency, and spatial domains. This effectively compresses high‐dimensional redundant information while simultaneously enhancing cross‐domain feature interaction capabilities. Furthermore, the BALSTM network is utilized to strengthen the contextual correlation between forward and backward sequences. This enables the precise capture of subtle yet critical pathological dynamic features in EEG signals, effectively mitigating the difficulty of stably modeling long‐period temporal dependencies. Results Experimental results demonstrate that, compared with state‐of‐the‐art methods, MDEFusion achieves an accuracy of 93.1% on the RepOD dataset and 94.6% on the NNCI dataset. Conclusion This paper provides an efficient and reliable EEG analysis tool for the auxiliary diagnosis of schizophrenia, demonstrating significant application value for clinical decision support systems. This paper proposes a multi‐domain EEG feature fusion network. By synergistically introducing a half‐split crossover sparse autoencoder and a bidirectional attention long short‐term memory network, the model achieves efficient fusion of multi‐domain features and modeling of long‐range temporal dependencies.

Nonlinear systems exhibit strong coupling and time‐varying parameters; these characteristics make it difficult for traditional mechanism‐based modeling methods to adapt to uncertain loads and complex working conditions. The prediction accuracy of vibration responses is thus limited. In order to handle these issues, this study proposes a big data‐driven improved back propagation neural network (BPNN) method for high‐precision prediction of vibration responses in nonlinear systems. This method first collects full‐condition vibration data through multi‐source sensors; it extracts multi‐domain (MD) features from the Time Domain (TD), Frequency Domain (FD), and Time‐Frequency Domain (TFD) to construct a high‐dimensional input dataset. Second, a three‐layer BPNN structure with input, hidden, and output layers is designed; meanwhile, the Rectified Linear Unit (ReLU) activation function is adopted to solve the vanishing gradient problem. Finally, adaptive learning rate and momentum terms are introduced to optimize the model training mechanism; this modification improves training speed and stability. Based on the public Nonlinear Mechanical Vibration Dataset (NMVD), a three‐level comparison system (a benchmark model, an improved model, and mainstream methods) is established. Experimental results show that the improved BP model converges in 1373 iterations; its convergence speed is increased by 50% compared with the traditional BP model. The Mean Squared Error (MSE) is reduced to 0.0032, the Coefficient of Determination (R2) reaches 0.9745, and the residual standard deviation is reduced by 39.87%. Thus, the proposed model shows remarkable advantages in both prediction accuracy and training stability. This study solves the technical limitations of traditional methods in predicting vibration responses of nonlinear systems via MD feature fusion and optimization of the BP algorithm training mechanism. This provides an efficient engineering approach for vibration analysis of complex nonlinear systems; it also offers data‐driven technical support for equipment vibration control and fault early warning. This study utilizes big data and the BP algorithm to predict vibrations in nonlinear systems. These results demonstrate that the improved BP model increases convergence speed by 50%, reduces the mean squared error by 64.04% to 0.0032, and achieves a coefficient of determination of 0.9745.

This paper proposes a human activity recognition (HAR) method for frequency-modulated continuous wave (FMCW) radar sensors. The method utilizes a multi-domain feature attention fusion network (MFAFN) model that addresses the limitation of relying on a single range or velocity feature to describe human activity. Specifically, the network fuses time-Doppler (TD) and time-range (TR) maps of human activities, resulting in a more comprehensive representation of the activities being performed. In the feature fusion phase, the multi-feature attention fusion module (MAFM) combines features of different depth levels by introducing a channel attention mechanism. Additionally, a multi-classification focus loss (MFL) function is applied to classify confusable samples. The experimental results demonstrate that the proposed method achieves 97.58% recognition accuracy on the dataset provided by the University of Glasgow, UK. Compared to existing HAR methods for the same dataset, the proposed method showed an improvement of about 0.9–5.5%, especially in the classification of confusable activities, showing an improvement of up to 18.33%.

Unmanned aerial vehicle (UAV) imagery detection faces challenges in preserving small object features during multi-level downsampling, handling angle and altitude-dependent variations in aerial scenes, achieving accurate localization in dense environments, and performing real-time detection. To address these limitations, we propose HSF-DETR, a lightweight transformer-based detector specifically designed for UAV imagery. First, we design a hybrid progressive fusion network (HPFNet) as the backbone, which adaptively modulates receptive fields to capture multi-scale information while preserving fine-grained details critical for small object detection. Second, building upon features extracted by HPFNet, we develop MultiScaleNet, which enhances feature representation through dual-layer optimization and cross-domain feature learning, significantly improving the model’s capability to handle complex aerial scenarios with diverse object orientations. Finally, to address spatial–semantic alignment challenges, we devise a position-aware align context and spatial tuning (PACST) module that ensures effective feature calibration through precise alignment and adaptive fusion across scales. This hierarchical architecture is complemented by our novel AdaptDist-IoU loss with dynamic weight allocation, which enhances localization accuracy, particularly in dense environments. Extensive experiments using standard detection metrics (mAP50 and mAP50:95) on the VisDrone2019 test dataset demonstrate that HSF-DETR achieves superior performance with 0.428 mAP50 (+5.4%) and 0.253 mAP50:95 (+4%) when compared with RT-DETR, while maintaining real-time inference (69.3 FPS) on an NVIDIA RTX 4090D GPU with only 15.24M parameters and 63.6 GFLOPs. Further validation across multiple public remote sensing datasets confirms the robust generalization capability of HSF-DETR in diverse aerial scenarios, offering a practical solution for resource-constrained UAV applications where both detection quality and processing speed are crucial.

Precise identification of active jamming in complex electromagnetic environments remains critically challenging for cognitive radar systems. Current methods often exhibit limitations in insufficient feature extraction and underutilization of jamming signals, leading to substantial performance degradation, particularly in low jamming-to-noise ratio (JNR) scenarios. To address these challenges, we propose a novel framework based on a multi-domain fusion network, MDFNet, to recognize 12 types of active jamming signals. MDFNet improves the recognition robustness under varying JNR conditions through a two-stage fusion of complementary features from pulse compression time–frequency (PC-TF) and range-Doppler (RD) domain images. Specifically, a novel dual-modal feature fusion (DMFF) module integrates PC-TF and RD features, while a decision fusion strategy leverages their distinctive characteristics. Experiments on typical jamming dataset demonstrate that MDFNet achieves an overall recognition accuracy of 96.05%. Notably, at a JNR of −20 dB, MDFNet outperforms the existing fusion methods by 12.86–18.19%. In summary, our proposed method significantly enhances the jamming recognition capability of cognitive radar systems in complex environments.

In response to the current issues of one-sided effective feature extraction and low classification accuracy in multi-class motor imagery recognition, this study proposes an Electroencephalogram (EEG) signal recognition method based on multi-domain feature fusion and optimized multi-kernel extreme learning machine. Firstly, the EEG signals are preprocessed using the Improved Comprehensive Ensemble Empirical Mode Decomposition (ICEEMD) algorithm combined with the Pearson correlation coefficient to eliminate noise and interference. Secondly, multivariate autoregressive (MVAR) model, wavelet packet decomposition, and Riemannian geometry methods are used to extract features from the time domain, frequency domain, and spatial domain, respectively, to construct a joint time-frequency-space feature vector. Subsequently, kernel principal component analysis (KPCA) is employed to fuse and reduce the dimensionality of the joint features, resulting in a reduced-dimensional fused feature vector. Finally, these feature vectors are input into a Radius-incorporated multi-kernel extreme learning machine (RIO-MKELM) for classification. The experimental results indicate that through multi-domain feature fusion and the incorporation of radius in a multi-kernel extreme learning machine, feature selection can be performed more effectively, eliminating redundant or irrelevant features and retaining the most useful information for classification. This approach enhances classification accuracy and other evaluation metrics, with the final classification accuracy reaching 95.49%, sensitivity at 97.88%, specificity at 98.12%, recall at 97.88%, and F1 Score at 96.67%. The findings of this study are of significant importance for the development and practical application of brain-computer interface (BCI) systems.

Wind energy, as a key link in renewable energy, has seen its penetration in the power grid increase in recent years. In this context, accurate and reliable short-term wind power prediction is particularly important for the real-time scheduling and operation of power systems. However, many deep learning-based methods rely on the relationship between wind speed and wind power to build a prediction model. These methods tend to consider only the temporal features and ignore the spatial and frequency domain features of the wind power variables, resulting in poor prediction accuracy. In addition to this, existing power forecasts for wind farms are often based on the wind farm level, without considering the impact of individual turbines on the wind power forecast. Therefore, this paper proposes a wind power prediction model based on multi-feature domain learning (MFDnet). Firstly, the model captures the similarity between turbines using the latitude, longitude and wind speed of the turbines, and constructs a turbine group with similar features as input based on the nearest neighbor algorithm. On this basis, the Seq2Seq framework is utilized to achieve weighted fusion with temporal and spatial features in multi-feature domains through high-frequency feature extraction by DWT. Finally, the validity of the model is verified with data from a wind farm in the U.S. The results show that the overall performance of the model outperforms other wind farm power prediction algorithms, and reduces MAE by 25.5% and RMSE by 20.6% compared to the baseline persistence model in predicting the next hour of wind power.

In recent years, unmanned aerial vehicle (UAV) technology has advanced rapidly, leading to its widespread deployment. However, this proliferation has been accompanied by a rise in unauthorized “black flight”, which poses a series of security risks to low-altitude airspace. Therefore, it is imperative to develop effective drone detection and identification techniques for airspace security management. This paper presents a radio frequency (RF)-based drone recognition method via variational mode decomposition (VMD) and multi-domain feature fusion. First, the collected RF signals exchanged between drones and their controllers are preprocessed using VMD. Subsequently, a multi-domain feature extraction method is introduced, which extracts time-domain, frequency-domain and time–frequency-domain features from the modes after VMD. To reduce feature dimensionality, a two-stage feature selection scheme based on ReliefF is then proposed. Finally, a support vector machine (SVM) is constructed for UAV classification. Experimental results on the open-source CardRF dataset show that the proposed method achieves superior performance compared to existing schemes, with an average identification accuracy of over 74.7% at SNRs greater than −10 dB.

Aiming at the difficulties in extracting effective features and low classification accuracy in the current multi-class motor imagery recognition, this paper proposes a multi-class motor imagery recognition method based on the combination of multi-domain feature fusion and twin support vector machine (TWSVM). First, the Autoregressive (AR) model, the bispectrum analysis method, and the common spatial pattern method are used to extract the features of the signal in temporal domain, frequency domain, and space domain, and construct a joint feature; then use the kernel principal component analysis to fuse the joint feature, the fusion features are generated by extracting the principal components whose cumulative contribution rate is more than 95%; Finally, the fusion features are sent to TWSVM optimized by bat algorithm for classification of the EEG, obtain an average recognition rate of 92.38%, which provides an effective method for multi-class motor imagery recognition, which will greatly promote in practical application based on BCI.