Search Results (192)

Power-equipment fault detection is challenging in real-world inspections due to subtle defect cues and cluttered backgrounds. This paper proposes an improved YOLOv12-based framework for multi-class power defect detection. We introduce a Prior-Guided Region Attention (PG-RA) module and design a Lightweight Residual Efficient Layer Aggregation Network (LR-RELAN). In addition, we develop a Dual-Spectrum Adaptive Fusion Loss (DSAF Loss) function to jointly improve classification confidence and bounding box regression consistency, enabling more robust learning under complex scenes. To support defect-oriented visual analytics and system interpretability, the framework further constructs Region of Interest (ROI) pseudo point clouds from detection outputs and compares two denoising strategies, Statistical Outlier Removal (SOR) and Radius Outlier Removal (ROR). A Python-based graphical prototype integrates image import, defect detection, ROI pseudo point cloud construction, denoising, 3D visualization, and result archiving into a unified workflow. Experimental results demonstrate that the proposed method improves detection accuracy and robustness while maintaining real-time performance, and the ROI pseudo point cloud module provides an intuitive auxiliary view for defect-structure inspection in practical applications. Full article

Electromagnetic ultrasonic testing technology, owing to its couplant-free, high-temperature-resistant, and non-contact characteristics, exhibits unique advantages for thickness measurement in harsh industrial environments. However, its accuracy is fundamentally limited by inherent constraints in signal bandwidth and low signal-to-noise ratio. To address these challenges, this work proposes an electromagnetic ultrasonic thickness measurement method that integrates Adaptive Denoising with Bayesian Vector Autoregressive (AD-BVAR) spectral extrapolation. The approach employs Particle Swarm Optimization (PSO) and automatically determines the optimal parameters for Variational Mode Decomposition (VMD), followed by integration with Singular Value Decomposition (SVD) to achieve the adaptive denoising of signals. Subsequently, the BVAR model incorporating prior constraints performs robust extrapolation of the effective frequency band spectrum, ultimately achieving high measurement accuracy signal reconstruction. The experimental results demonstrate that on step blocks with thicknesses of 3 mm and 12.5 mm, the proposed method achieved significantly reduced error rates of 0.267% and 0.240%, respectively. This performance markedly surpasses that of the conventional Autoregressive (AR) method, which yielded errors of 0.767% and 0.560% under identical conditions, while maintaining stable performance across different thicknesses. Full article

In this paper, laser-assisted machining (LAM) of SiC ceramics was taken as the research object, and the different spectrum and energy spectrum characteristics and their changing trends of acoustic emission (AE) signals under processing states of brittleness, plasticity and thermal damage were analyzed. The numerical characterization of ceramic softening degree was indirectly realized by the energy spectrum characteristics of low-frequency band energy ratio, marking a methodological breakthrough in transitioning from qualitative analysis to quantitative detection for identifying plastic processing state. First, the surface morphology of the machined surface based on the single-factor experiment of laser power was analyzed, and three different processing states and ranges of laser power were determined, namely brittle state (0–185 W), plastic state (185–225 W) and thermal damage state (>225 W). Then, the wavelet packet denoising and spectrum analysis of AE signals under different processing states were carried out to obtain the corresponding frequency of the maximum amplitude and the amplitude change trend of the characteristic frequency (515 kHz) in the high-frequency domain. Finally, the energy spectrum analysis of acoustic emission signals was carried out, and the method of indirect characterization of ceramic softening degree by low-frequency band energy ratio was proposed. This paper provides a numerical characterization method and theoretical guidance for the detection and identification of the plastic processing state of ceramic laser-assisted cutting. Full article

Although wavelet neural networks (WNNs) combine the expressive capability of neural models with multiscale localization, there are currently few theoretical guarantees for their training. We investigate the weight decay (${\mathrm{L}}_2$ regularization) optimization dynamics of gradient descent (GD) for WNNs. Using explicit rates controlled by the spectrum of the regularized Gram matrix, we first demonstrate global linear convergence to the unique ridge solution for the feature regime when wavelet atoms are fixed and only the linear head is trained. Second, for fully trainable WNNs, we demonstrate linear rates in regions satisfying a Polyak–Łojasiewicz (PL) inequality and establish convergence of GD to stationary locations under standard smoothness and boundedness of wavelet parameters; weight decay enlarges these regions by suppressing flat directions. Third, we characterize the implicit bias in the over-parameterized neural tangent kernel (NTK) regime: GD converges to the minimum reproducing kernel Hilbert space (RKHS) norm interpolant associated with the WNN kernel with ${\mathrm{L}}_2$. In addition to an assessment process on synthetic regression, denoising, and ablations across λ and stepsize, we supplement the theory with useful recommendations on initialization, stepsize schedules, and regularization scales. Together, our findings give a principled prescription for dependable training that has broad applicability to signal processing applications and shed light on when and why ${\mathrm{L}}_2$-regularized GD is stable and quick for WNNs. Full article

Cognitive radio networks (CRNs) rely heavily on spectral sensing to detect primary user (PU) activity, yet detection at low signal-to-noise ratios (SNRs) remains a major challenge. Hence, a novel “Deep Wavelet Cyclostationary Independent Gaussian Markov Fourier Transform Dempster–Shafer Network” is proposed. When the signal waveform is submerged within the noise envelope and residual correlation emerges in the noise, it violates white Gaussian assumptions, leading to misidentification of signal presence. To resolve this, the Adaptive Continuous Wavelet Cyclostationary Denoising Autoencoder (ACWC-DAE) is introduced, in which the Adaptive Continuous Wavelet Transform (ACWT), Cyclostationary Independent Component Analysis Detection (CICAD), and Denoising Autoencoder (DAE) are introduced into the first hidden layer of a Deep Q-Network (DQN). It restores the bursty signal structure, separates the structured noise, and reconstructs clean signals, leading to accurate signal detection. Additionally, bursty and fading-affected primary user signals become fragmented and dip below the noise floor, causing conventional fixed-window sensing to fail in accumulating reliable evidence for detection under intermittent and low-duty-cycle conditions. Therefore, the Adaptive Gaussian Short-Time Fourier Transform Dempster–Shafer Model (AGSTFT-DSM) is incorporated into the second DQN layer, Adaptive Gaussian Mixture Hidden Markov Modeling (AGMHMM) tracks the hidden activity states, Adaptive Short-Time Fourier Transform (ASFT) resolves brief signal bursts, and Dempster–Shafer Theory (DST) fuses uncertain evidence to infer occupancy, thereby detecting an accurate user signal. The results obtained by the proposed model have a low error and detection time of 0.12 and 30.10 ms and a high accuracy of 97.8%, revealing the novel insight that adaptive wavelet denoising, along with uncertainty-aware evidence fusion, supports reliable spectrum detection under low-SNR conditions where existing models fail. Full article

Accurate short-term (10–20 days) natural gas load forecasting is crucial for the “tactical planning” of gas utilities, yet it faces significant challenges from high volatility, strong noise, and the high-dimensional multicollinearity of influencing factors. To address these issues, this paper proposes a novel hybrid forecasting framework: PCCA-ISSA-GRU. The framework first employs Principal Component Correlation Analysis (PCCA), which improves upon traditional PCA by incorporating correlation analysis to effectively select orthogonal features most relevant to the load, resolving multicollinearity. Concurrently, an Improved Singular Spectrum Analysis utilizes statistical criteria (skewness and kurtosis) to adaptively separate signals from Gaussian noise, denoising the historical load sequence. Finally, the dually optimized data is fed into a Gated Recurrent Unit (GRU) neural network for prediction. Validated on real-world data from a large city in Northern China, the PCCA-ISSA-GRU model demonstrated superior performance. For a 20-day forecast horizon, it achieved a Mean Absolute Percentage Error (MAPE) of 6.09%. Results show its accuracy is not only significantly better than single models (BPNN, LSTM, GRU) and classic hybrids (ARIMA-ANN), but also outperforms the state-of-the-art (SOTA) model, Informer, within the 10–20 days tactical window. This superiority was confirmed to be statistically significant by the Diebold–Mariano test (p < 0.05). More importantly, the model exhibited exceptional robustness, with its error increase during extreme weather scenarios (e.g., cold waves, rapid temperature changes) being substantially lower (+56.7%) than that of Informer (+109.2%). The PCCA-ISSA-GRU framework provides a high-precision, highly robust, and cost-effective solution for urban gas short-term load forecasting, offering significant practical value for critical operational decisions and high-risk scenarios. Full article

Event-Related Potentials (ERPs) are low-amplitude neural responses elicited by sensory or cognitive stimuli, widely exploited as biomarkers in the early diagnosis of neurodevelopmental and neurodegenerative disorders such as autism spectrum disorder and Alzheimer’s disease, and as control signals in brain–computer interface (BCI) systems for severely disabled individuals. However, their extremely low signal-to-noise ratio (SNR) necessitates robust denoising, especially in real-time BCI applications where low latency, minimal computational overhead, and single-channel operation are critical constraints. While advanced offline methods like Independent Component Analysis (ICA) and wavelet-based thresholding offer effective denoising in multichannel settings, they are ill-suited for embedded, causal, and resource-constrained environments. To address this gap, we propose a Partial-Update Variable Step-Size LMS (PU-VSS-LMS) algorithm that complementarily combines dynamic step-size adaptation with a magnitude-driven partial-update strategy. Evaluated on synthetic ERP-like signals embedded in realistic EEG noise (SNR = 6 dB and 0 dB), PU-VSS-LMS achieves lower mean squared error (MSE: 0.0780 vs. 0.0850 at 6 dB) and higher output SNR (8.10 dB vs. 7.80 dB) than standard VSS-LMS, while outperforming ICA in waveform preservation and noise suppression. Importantly, it reduces computational load by 75% (updating only 4 of 16 coefficients), enabling faster execution without sacrificing accuracy. These results establish PU-VSS-LMS as a highly efficient and effective solution for real-time ERP denoising in embedded, single-channel biomedical systems. Full article

Fiber optic vibration identification has significant applications in engineering fields, like security surveillance and structural health assessment. However, present methods primarily depend on either temporal–frequency domain or image features simply, challenging the simultaneous consideration of both image attributes and the temporal dependencies of vibration signals. Consequently, the performance of fiber optic vibration recognition remains subject to improvement, and its effectiveness further diminishes under conditions of uneven data distribution. Therefore, this study integrates residual neural networks, long short-term memory networks, and diffusion denoising probabilistic models to propose a fiber optic vibration recognition method DR-LSTM, which incorporates both image and temporal features while ensuring high recognition accuracy across balanced and imbalanced data distributions. Firstly, features of the Mel spectrum image and temporal characteristics of fiber optic vibration events are extracted. Subsequently, specialized neural network models are developed for categories with scarce data to produce similar images for data augmentation. Finally, the retrieved composite characteristics are employed to train recognition models, thereby improving recognition accuracy. Experiments were performed on datasets from natural environment and anthropogenic vibration, including for both balanced and imbalanced data distributions. The results show that on the two balanced datasets, the proposed model achieves improvements in classification accuracy of at least 0.67% and 7.4% compared to conventional methods. In the two imbalanced datasets, the model’s accuracy exceeds that of conventional models by a minimum of 18.79% and 2.4%. This validates the effectiveness and feasibility of DR-LSTM in enhancing recognition accuracy and addressing issues with imbalanced data distribution. Full article

In seismic data processing, noise not only affects velocity analysis and seismic migration, but also causes potential risks in post-stack processing because of the artifacts. The singular value decomposition (SVD) method based on the time domain and the frequency domain is effective for noise suppression, but it is very sensitive to singular value selection. This paper proposes a method of adaptive SVD denoising in both time and frequency domains (ASTF), with three steps. Firstly, two Hankel matrices are constructed in the time domain and frequency domain, respectively. Secondly, the parameters of the reconstruction matrix are adaptively selected based on the singular value second-order difference spectrum. Finally, the weights of these two matrices are learned through ternary search. Experiments were carried out on synthetic data and field data to prove the effectiveness of ASTF. The results show that this method can effectively suppress noise. Full article

Rolling element bearings are commonly used in rotating machines. Bearing fault detection and diagnosis play a critical role in machine operations to recognize bearing faults at their early stage and prevent machine performance degradation, improve operation quality, and reduce maintenance costs. Although many fault detection techniques are proposed in the literature for bearing condition monitoring, reliable bearing fault detection remains a challenging task in this research and development field. This study proposes an enhanced Teager–Kaiser (eTK) technique for bearing fault detection and diagnosis. Vibration signals are used for analysis. The eTK technique is novel in two aspects: Firstly, an empirical mode decomposition analysis is suggested to recognize representative intrinsic mode functions (IMFs) with different frequency components. Secondly, an eTK denoising filter is proposed to improve the signal-to-noise ratio of the selected IMF features. The analytical signal spectrum analysis is conducted to identify representative features for bearing fault detection. The effectiveness of the proposed eTK technique is verified by experimental tests corresponding to different bearing conditions. Full article

Automatic modulation classification (AMC) of received unknown signals is critical in modern communication systems, enabling intelligent signal interception and spectrum management. In this paper, we propose a wavelet-based spectrum convolutional neural network (WS-CNN) model that integrates signal processing techniques with deep learning to achieve robust classification under challenging conditions, including noise, fading, and Doppler effects. The WS-CNN model is based on wavelet analysis and a convolutional neural network (CNN). Specifically, the proposed wavelet analysis, including wavelet threshold denoising, median filtering, and continuous wavelet transformation, is used for signal preprocessing to extract features and generate a compact 2D diagram. The 2D diagram is subsequently fed into the CNN for classification. The simulation results show that the proposed WS-CNN model achieves higher classification rates across a wide range of signal-to-noise ratios (SNRs) compared with existing methods. Full article

Ice loads continue to pose challenges to the structural safety of offshore wind turbines (OWTs), while the rapid development of offshore wind power in cold regions is enabling the deployment of OWTs in deeper waters. To accurately simulate the dynamic response of an OWT under combined ice–wind loading, this paper proposes a Discrete Element Method–Wind Turbine Integrated Analysis (DEM-WTIA) framework. The framework can synchronously simulate discontinuous ice-crushing processes and aeroelastic–structural dynamic responses through a holistic turbine model that incorporates rotor dynamics and control systems. To address the issue of insufficient prediction accuracy for dynamic responses, we introduced a multivariate time series forecasting method that integrates a secondary decomposition strategy with a hybrid prediction model. First, we developed a parallel signal processing mechanism, termed Adaptive Complete Ensemble Empirical Mode Decomposition with Improved Singular Spectrum Analysis (CEEMDAN-ISSA), which achieves adaptive denoising via permutation entropy-driven dynamic window optimization and multi-feature fusion-based anomaly detection, yielding a noise suppression rate of 76.4%. Furthermore, we propose the F-Transformer prediction model, which incorporates a Frequency-Enhanced Channel Attention Mechanism (FECAM). By integrating the Discrete Cosine Transform (DCT) into the Transformer architecture, the F-Transformer mines hidden features in the frequency domain, capturing potential periodicities in discontinuous data. Experimental results demonstrate that signals processed by ISSA exhibit increased signal-to-noise ratios and enhanced fidelity. The F-Transformer achieves a maximum reduction of 31.86% in mean squared error compared to the standard Transformer and maintains a coefficient of determination (R²) above 0.91 under multi-condition coupled testing. By combining adaptive decomposition and frequency-domain enhancement techniques, this framework provides a precise and highly adaptable ultra-short-term response forecasting tool for the safe operation and maintenance of offshore wind power in cold regions. Full article

Modern wireless communication systems increasingly rely on Automatic Modulation Classification (AMC) to enhance reliability and adaptability, especially in the presence of severe signal degradation. However, despite significant progress driven by deep learning, many AMC models still struggle with high computational overhead, suboptimal performance under low-signal-to-noise conditions, and limited interpretability, factors that hinder their deployment in real-time, resource-constrained environments. To address these challenges, we propose the Convolutional Random Fourier Features with Denoising Thresholding Network (CRFFDT-Net), a compact and interpretable deep kernel architecture that integrates Convolutional Random Fourier Features (CRFFSinCos), an automatic threshold-based denoising module, and a hybrid time-domain feature extractor composed of CNN and GRU layers. Our approach is validated on the RadioML 2016.10A benchmark dataset, encompassing eleven modulation types across a wide signal-to-noise ratio (SNR) spectrum. Experimental results demonstrate that CRFFDT-Net achieves an average classification accuracy that is statistically comparable to state-of-the-art models, while requiring significantly fewer parameters and offering lower inference latency. This highlights an exceptional accuracy–complexity trade-off. Moreover, interpretability analysis using GradCAM++ highlights the pivotal role of the Convolutional Random Fourier Features in the representation learning process, providing valuable insight into the model’s decision-making. These results underscore the promise of CRFFDT-Net as a lightweight and explainable solution for AMC in real-world, low-power communication systems. Full article

Efficient spectrum utilization is critical for meeting the growing data demands of modern wireless communication networks. Automatic Modulation Classification (AMC) plays a key role in enhancing spectrum efficiency by accurately identifying modulation schemes in received signals—an essential capability for dynamic spectrum allocation and interference mitigation, particularly in cognitive radio (CR) systems. With the increasing deployment of smart edge devices, such as IoT nodes with limited computational and memory resources, there is a pressing need for lightweight AMC models that balance low complexity with high classification accuracy. In this study, we propose a low-complexity, lightweight deep learning (DL) AMC model optimized for resource-constrained edge devices. We introduce a dual-path deep residual shrinkage network (DP-DRSN) with garrote thresholding for effective signal denoising, and we designed a compact hybrid CNN-LSTM architecture comprising only 27,072 training parameters. The proposed model achieved average classification accuracies of 61.20%, 63.78%, and 62.13% on the RML2016.10a, RML2016.10b, and RML2018.01a datasets, respectively, demonstrating a strong balance between model efficiency and classification performance. These results highlight the model’s potential for enabling accurate and efficient AMC on edge devices with limited resources, despite not surpassing state-of-the-art accuracy owing to its deliberate emphasis on computational efficiency. Full article

Wavelength modulation-tunable diode laser absorption spectroscopy (WM-TDLAS) is a critical tool for gas detection. However, noise in second harmonic signals degrades detection performance. This study presents a hybrid denoising algorithm combining Empirical Mode Decomposition (EMD) and wavelet adaptive thresholding to enhance WM-TDLAS performance. The algorithm decomposes raw signals into intrinsic mode functions (IMFs) via EMD, selectively denoises high-frequency IMFs using wavelet thresholding, and reconstructs the signal while preserving spectral features. Simulation and experimental validation using the CH₄ absorption spectrum at 1654 nm demonstrate that the system achieves a threefold improvement in detection precision (0.1181 ppm). Allan variance analysis revealed that the detection capability of the system was significantly enhanced, with the minimum detection limit (MDL) drastically reduced from 2.31 ppb to 0.53 ppb at 230 s integration time. This approach enhances WM-TDLAS performance without hardware modification, offering significant potential for environmental monitoring and industrial safety applications. Full article