Search Results (379)

This study proposes a nondestructive method for evaluating egg freshness using laser Doppler vibrometry. A vibration measurement platform was constructed to collect vibration signals from intact eggs representing four freshness grades. The vibration signals were processed using multiplicative scatter correction, standard normal variate transformation, moving average filtering, Savitzky–Golay smoothing, and first-order Savitzky–Golay differentiation. Principal component analysis, successive projections algorithm, and competitive adaptive reweighted sampling were then used for feature extraction and dimensionality reduction. Six classification models were developed and compared, including support vector machine, k-nearest neighbor, random forest, naive Bayes, discriminant analysis, and linear discriminant analysis. The model based on moving average filtering, successive projections algorithm, and discriminant analysis using measurements from three egg locations achieved a classification accuracy of 96.7%. The model based on moving average filtering, competitive adaptive reweighted sampling, and random forest using blunt-end measurements also achieved an accuracy of 96.7%. These results demonstrate that laser Doppler vibrometry combined with machine learning can distinguish eggs with different freshness levels, providing a promising nondestructive approach for egg-quality evaluation. Full article

Background: Mitogen-activated protein kinase 3 (MAPK3/ERK1) plays a central role in cellular proliferation, inflammation, apoptosis, and survival signalling and has been implicated in oral squamous cell carcinoma (OSCC), periodontitis, oral lichen planus, and other chronic oral inflammatory diseases. The present study employed an integrated computational workflow combining machine learning (ML)-based quantitative structure–activity relationship (QSAR) modelling, molecular docking, density functional theory (DFT), molecular dynamics (MD) simulation, and MM-GBSA analysis to identify and characterise potent MAPK3 inhibitors. Methods: A curated dataset of 907 experimentally validated MAPK3 inhibitors was retrieved from the ChEMBL database and processed using molecular descriptors and Morgan fingerprints. Multiple ML algorithms were evaluated under scaffold-based validation, with Light Gradient Boosting Machine (LightGBM) demonstrating the best predictive performance. Results: The final model achieved strong classification capability with ROC-AUC values of 0.898 and 0.926. Feature importance analysis revealed that local structural motifs captured by fingerprint descriptors played dominant roles in MAPK3 inhibitory activity. The top-ranked compounds were subjected to molecular docking, where compounds 58324148 and 137531515 exhibited strong binding affinities of −11.9 and −11.0 kcal/mol, respectively. DFT calculations demonstrated favourable electronic properties with low HOMO–LUMO energy gaps, while MD simulations confirmed stable receptor–ligand interactions throughout 200 ns trajectories. MM-GBSA analysis further supported strong binding stability dominated by van der Waals interactions. Conclusions: Overall, the integrated computational framework successfully identified promising MAPK3 inhibitor candidates with potential therapeutic relevance for oral inflammatory and proliferative diseases. Full article

Background: DNA methylation is a key epigenetic mechanism whose dynamics are closely linked to ageing. Modern epigenetic clocks predict biological age based on the average methylation level. The concept of “epigenetic rejuvenation” posits that at early stages of development, the biological age of the embryo may decrease, reaching a minimum (“ground zero”) at the gastrulation stage. However, standard averaging methods may not account for important rearrangements in the internal structure of methylation. Objective: To apply the apparatus of information theory and topological data science to the analysis of scNMT-seq data and to test whether DNA methylation entropy decreases from stage E4.5 to E6.5, which would correspond to an approach towards the biological zero state. Methods: Publicly available scNMT-seq data (GSE121690) were analyzed. Five entropy measures were calculated for each cell (Shannon, Renyi, Tsallis, LZ-complexity, local gradient entropy (entropy of variations in the smoothed histogram of the methylation distribution), and persistent entropy (PE)—a topological complexity measure). For the five-dimensional entropy feature space, a Rips complex was constructed, and persistence diagrams H_0 and H_1 were computed. Results: All five entropy measures decreased significantly, with LZ complexity showing the largest relative reduction (−28.4%) and the strongest independent signal (partial r = −0.181). Among all the complexity measures studied, LZ complexity exhibited the largest relative reduction, underscoring its heightened sensitivity to the progressive ordering of the epigenetic landscape. Notably, the ternary encoding of LZ complexity showed strong correlation with Shannon entropy (r = 0.71), indicating that algorithmic complexity, when accounting for partial methylation states, aligns closely with statistical entropy while retaining sensitivity to spatial order. The consistency of results across binary and ternary encodings confirms the robustness of LZ complexity as a structural biomarker. Persistent entropy confirmed the general dynamics (decrease from 15.91 to 14.89, p = 0.01). Topological analysis of the multidimensional space revealed a qualitative reorganization: at stage E6.5, stable cyclic structures (H₁) emerge, while at E4.5 the space is dominated by a single large-scale cycle. Null model validation confirmed that the observed H₁-cycles are genuine topological features rather than random fluctuations. Comprehensive topological characterization showed that normalized persistent entropy increases from 0.846 to 0.882 (p < 0.001), while maximum persistence decreases from 0.446 to 0.218 (p < 0.001), reflecting a transition from a homogeneous state to structured diversification—multiple, evenly distributed cycles corresponding to distinct cell lineages. Consistent with this, regional disorder (RE/RD) at the single-cell level decreases from E4.5 to E6.5 (RE: −25.5%, RD: −27.4%, p < 10⁻¹³), while global entropy also decreases, together painting a picture of epigenetic rejuvenation as ordered consolidation at the whole-genome scale. An SVM model trained on 15 entropy and structural features achieved stage classification with an accuracy of 93.4% and AUC of 0.981, confirming the diagnostic potential of the approach. Conclusions: The decrease in DNA methylation entropy from E4.5 to E6.5 corresponds to an approach to “ground zero”—the point of minimum biological age in embryogenesis—and supports the hypothesis of a link between decreasing entropy and epigenetic rejuvenation. The addition of topological analysis reveals the hidden organization of epigenetic diversity, showing that ordering does not lead to homogenization but is accompanied by the formation of distinguishable cell lineages. Full article

Direction-of-arrival (DOA) estimation is a core research topic in array signal processing, and two-dimensional (2-D) DOA estimation can jointly acquire the azimuth and elevation of incoming signals, which bears great practical value. Traditional subspace and sparse reconstruction algorithms are plagued by heavy computation and deteriorated accuracy under imperfect array manifolds, low signal-to-noise ratios (SNRs) and insufficient snapshots. To enhance estimation robustness and inference speed simultaneously, this paper presents a dual-branch convolutional neural network (CNN) for 2-D DOA estimation based on uniform rectangular arrays. The network takes the sample covariance matrix of array received data as input. A shared feature encoder with residual blocks and channel-attention modules extracts common spatial features, followed by two prediction heads with independent parameters for elevation and azimuth estimation. Because each branch has a 61-dimensional output while two sources may be simultaneously present, the angle estimation is formulated as multi-label classification using sigmoid outputs and weighted binary cross-entropy. Simulations covering diverse SNRs, snapshot counts, angular intervals and off-grid cases verify that the proposed network obtains smaller root mean square errors than methods with multiple signal classification (MUSIC), estimation of signal parameters via rotational invariance techniques (ESPRIT) and ordinary CNN methods, with millisecond-level inference latency. This framework offers an efficient, high-precision real-time 2-D DOA estimation scheme for complicated electromagnetic scenes. Full article

This study investigates predictive modeling of failure states in manufacturing systems using artificial intelligence in the context of sustainable maintenance and Industry 4.0. The proposed methodological framework is based on historical operational and maintenance data from a single manufacturing device, encompassing multiple process and operational signals such as vibrations, temperature, electric current, and operational logs. The aim is to predict failure within a short-term horizon to support maintenance planning. The article compares Random Forest and XGBoost algorithms at different prediction horizons (8 h and 16 h) to identify the trade-off between classification accuracy and lead time for maintenance planning. Model outputs are analyzed using explainable artificial intelligence and transformed into a risk index compatible with the FMEA methodology. The practical contribution of the proposed approach is illustrated through a scenario-based what-if assessment of potential sustainability impacts, particularly in terms of estimated reductions in unplanned downtime, material waste, and energy consumption. The results point to the potential of integrating AI-supported maintenance as a tool for increasing the reliability and sustainability of manufacturing systems. The novelty of the proposed framework lies in the integration of predictive maintenance, explainable artificial intelligence, replay-based maintenance assessment, dynamic FMEA risk assessment, and sustainability impact quantification into a unified decision-support framework. Full article

To effectively extract fault characteristics from complex vibration signals and improve the diagnostic performance of deep learning networks, this paper introduces a wind turbine blade fault diagnosis method that combines Multi-scale Enhanced Hierarchical Fuzzy Entropy (MEHFE), Isolation Forest, and the Grey Wolf Optimization (GWO) algorithm for optimizing the Gated Recurrent Unit (GRU). Initially, the MEHFE algorithm is applied to decompose and reconstruct three-directional vibration signals at the blade root, thereby extracting “scale-frequency” dual-dimensional features that represent the evolution of fault frequency structure and complexity across multiple scales. Subsequently, Isolation Forest is employed to assess and filter feature importance, constructing an optimal feature subset to mitigate redundancy and noise interference. Finally, the optimal features are fed into the GRU network for fault pattern recognition, and the GWO algorithm is utilized to adaptively optimize network hyperparameters, thereby enhancing classification accuracy and noise resilience. Simulation experiments on typical wind turbine blade faults reveal that when GRU serves as the classifier, the diagnostic accuracy of MEHFE exceeds 76%. After feature optimization with Isolation Forest and network parameter optimization with GWO, the diagnostic accuracy surpasses 93%, demonstrating notable advantages in both classification capability and stability. Even under conditions of noise interference, the accuracy remains above 90%. The research substantiates that the proposed method can effectively extract pattern information indicative of blade structural damage from vibration data, achieving high fault recognition accuracy and robustness. Full article

Passive acoustic localization in complex shallow waters requires algorithms tailored to specific operational constraints. This paper investigates the adaptability, computational efficiency, and statistical performance boundaries of five matched-field processing (MFP) methods—Bartlett, Minimum Variance Distortionless Response (MVDR), Multiple Signal Classification (MUSIC), Reduced Covariance Matrix (RCM), and Rank and Trace Minimization (RTM)—using the Elba-93 sea trial dataset. Error metrics and processing complexities are systematically evaluated across stationary and maneuvering target scenarios. Rigorous non-parametric statistical tests reveal distinct operational boundaries: under stationary conditions dominated by systemic environmental mismatch, energy-based processors guarantee reliable baseline stability. Conversely, under snapshot-deficient dynamic conditions tracking a receding target, standard high-resolution subspace methods become highly vulnerable to trajectory jumps. In such highly dynamic scenarios, adaptive energy-based processors (specifically MVDR) exhibit the most stable tracking continuity and lowest numerical peak errors. Simultaneously, the operational adaptability of subspace methods is improved via covariance matrix reconstruction (CMR). Specifically, the RCM technique effectively decouples unstructured sensor noise, mitigating maximum trajectory deviations and providing a balanced trade-off between computational efficiency and robustness. Statistical evaluations confirm the fundamental performance boundaries in static environments, while highlighting sample-size limitations in highly dynamic scenarios, thereby establishing a realistic, evidence-based benchmark for marine engineering applications. Full article

Many machine learning and deep learning algorithms based on Artificial Neural Networks (ANNs) have been implemented on software platforms for handwritten digit and character recognition. However, an ANN is difficult to deploy on an embedded platform based on a Central Processing Unit (CPU) because of its large computation, complex structure, and frequent memory access. However, Field Programmable Gate Array (FPGA) devices facilitate this task and offer the capability to design fully customizable hardware architectures. Additionally, they provide high flexibility and high parallel computations based on parallel processing techniques, and they contain sufficient on-chip Digital Signal Processing (DSP) blocks useful for complicated multiplications. In this paper, we present a detailed FPGA-based implementation of a handwritten digit recognition system based on a Multi-Layer Perceptron (MLP) model. The internal modules of the network are designed using the VHSIC Hardware Description Language (VHDL) to achieve a high-level optimization on the hardware platform, and the functionality is simulated and tested using Vivado ISIM Tools. The system has been characterized to reach acceptable performance compared to previous approaches. After implementing the whole neural network on a Xilinx Pynq-Z2 board, it occupies in the device 20758 LUTs, 4426 FFs, 3.50 blocks of random-access memory (BRAM), and 42 DSPs. It reaches an execution time of 2.192 µs to recognize a handwritten number, while consuming only 0.36 Watts, and it achieves a classification accuracy of 97%. Additionally, the proposed architecture can be easily scaled on different FPGA devices thanks to its regularity. Therefore, it offers more portability of the architecture and can be used on different real embedded applications. Full article

Background: Chronic exposure to addictive substances induces persistent alterations in neural dynamics, reflecting maladaptive brain plasticity. While such changes are well documented using neuroimaging techniques, their electrophysiological signatures—particularly those derived from nonlinear EEG complexity—remain insufficiently explored across diverse substance use profiles. This preliminary study aims to investigate whether nonlinear EEG complexity measures can serve as sensitive biomarkers of maladaptive plasticity in substance use disorder (SUD) across multiple substance categories. Methods: A total of 350 participants were included and categorized into seven groups (n = 50 each): six substance use groups (cannabis, heroin, heroin–cannabis, methamphetamine–cannabis, methamphetamine–heroin, and multi-drug) and one control group without a diagnosis of substance use disorder. Resting state EEG signals were recorded using an eight-channel system. Four nonlinear features, Largest Lyapunov Exponent (LLE), Fractal Dimension (FD), Hurst Exponent (HE), and Kolmogorov Complexity (KC) were extracted. Statistical analysis was performed using two-way ANOVA, and classification was conducted using the K Nearest Neighbour (KNN) algorithm. Results: Significant group differences (p < 0.05) were observed across all nonlinear features. Control participants without a diagnosis of substance use disorder consistently exhibited higher complexity values compared to substance use groups, indicating reduced neural dynamical variability associated with the history of sustained substance uses over multiple years. Region wise analysis revealed that frontal and central cortical areas linked to motor planning and sensorimotor integration were particularly affected. The KNN classifier achieved an accuracy of 98.4%, sensitivity of 100%, and specificity of 96.8%. Conclusions: Nonlinear EEG complexity measures provide a robust electrophysiological marker of substance induced maladaptive brain plasticity. The observed reduction in complexity reflects impaired neural adaptability, particularly within motor control networks. These findings highlight the potential of EEG based complexity metrics for objective assessment, classification, and neurorehabilitation monitoring in substance use disorders. Full article

Aiming at the problems with traditional transformer winding deformation detection, requiring power outages, low signal-to-noise ratios for online monitoring, and insufficient feature extraction, this paper proposes a live monitoring and intelligent diagnosis method based on pulse-coupled injection. At the hardware level, a semi-ring capacitive coupling sensor is developed and designed, which realizes non-contact injection of high-frequency pulse signals and high-SNR extraction without a power outage. The reliability of the system under complex working conditions is verified by field experiments on multiple actual 110 kV transformers. At the algorithm level, an innovative MSCNN–Transformer–PGA deep composite model fused with prior electromagnetic physical knowledge is constructed and combined with the transformer equivalent circuit model. The model uses a multi-scale convolution to extract local details of frequency response signals, adopts Transformer to establish the global sequence dependence, and introduces a Physics-Guided Attention mechanism (PGA) to adaptively focus on the key fault physical frequency bands. The experimental results show that the proposed method effectively overcomes electromagnetic noise interference, and the fault classification accuracy of single-modal pulse frequency response data reaches 97.6%, providing a high-precision online monitoring solution for the safe operation and maintenance of transformers. Full article

To address the issue of motion artifact interference faced by wearable ECG monitoring devices in dynamic environments, this paper proposes an adaptive motion artifact removal framework based on improved Variational Mode Decomposition (VMD). By designing a parameter self-adjustment mechanism and a multi-feature fusion mode selection strategy, the algorithm’s adaptability to non-stationary ECG signals and noise separation accuracy are enhanced. Experiments on the MIT-BIH Arrhythmia Database demonstrate that the improved VMD algorithm outperforms traditional wavelet transform, Recursive Least Squares (RLS), and conventional VMD methods in multiple performance metrics. Specifically, the signal-to-noise ratio (SNR) is improved by 5.17 dB, the Percentage Root Mean Squared Difference (PRD) is reduced to 49.13%, the correlation coefficient is increased to 0.88, and high real-time processing capability (Real-Time Processing Ratio, RTR = 22.5) is maintained, meeting the low-latency requirements of wearable devices. Moreover, case studies on pathological recordings (e.g., Wolff–Parkinson–White syndrome and third-degree atrioventricular block) reveal that the improved VMD better preserves clinically significant features such as delta waves and dissociated P waves. Furthermore, a downstream arrhythmia classification task using a CWT-CNN classifier achieves 91.67% accuracy on denoised heartbeats, which is 2.67 percentage points higher than that on raw noisy signals (89.00%), confirming the practical benefit of the proposed preprocessing for AI-based diagnosis. This study provides an effective processing solution for improving the signal quality of wearable ECG monitoring. Full article

To address the challenges of extracting mechanical fault features in Gas Insulated Switchgear (GIS) under complex operating conditions and the insufficient diagnostic accuracy associated with traditional one-dimensional time-series signals, this paper proposes a GIS fault-classification method based on composite dimensional upscaling images of vibration signals and the Vision Transformer (ViT) algorithm. This method first employs a sliding window slicing strategy to segment the raw long-sequence vibration signals into multiple overlapping time segments. Then, it utilizes the Gramian Angular Summation Field (GASF), Gramian Angular Difference Field (GADF), and Markov Transition Field (MTF) to perform composite dimensional upscaling on these segmented signals, projecting the resulting features into a three-channel RGB composite two-dimensional image. Subsequently, the global self-attention mechanism of the Vision Transformer (ViT) processes the dimensionally upscaled data to achieve the fault classification of the GIS equipment. Experimental results demonstrate that, compared to single-channel ViT variants, Convolutional Neural Networks (CNN), and Residual Networks (ResNet), the proposed algorithm achieves the highest overall performance in the training set experiments, and the superiority of this method is verified through ablation studies and comparative experiments. Furthermore, the average accuracy of the algorithm on the testing set reaches 95.63%, proving the reliability and accuracy of the proposed method. Full article

Recent studies on EEG-based automated depression detection have primarily depended on complex deep learning models. While these methods improve classification performance, their practical application is limited by high computational complexity, challenging training processes, and poor interpretability. This paper proposes an efficient method for depression recognition, which extracts multi-domain features from preprocessed EEG signals and selects the most discriminative feature subset by integrating the rapid preliminary screening capability of RankSearch with the interactive optimization ability of the Genetic Algorithm (GA). Our approach first eliminates redundant features efficiently through RankSearch, then deeply explores inter-feature relationships via GA, significantly enhancing classification performance while maintaining feature-level interpretability. Using the optimized feature subset, we evaluate performance with multiple machine learning classifiers (Decision Tree, KNN, Random Forest, SVM, XGBoost). Experiments on the public HUSM dataset demonstrate superior performance under rigorous cross-validation (accuracy = 95.08%, sensitivity = 95.99%, specificity = 94.30%, F1-score = 95%, AUC = 0.9514), with feature importance analysis further confirming interpretability. Compared to existing models, our method achieves lower computational complexity and higher clinical practicality, offering a more efficient technical solution for objective depression diagnosis. Full article

Noise contamination is a common challenge in the analysis of time series data, where stochastic perturbations can obscure deterministic dynamics and complicate the interpretation of signals from chaotic and physiological systems. Reliable identification of noise regimes and their intensity is therefore essential for robust analysis of dynamical and biomedical signals, where incorrect attribution of stochastic perturbations can lead to misleading interpretations of system behavior. For this reason, the present study examines the role of complexity-based descriptors for identifying stochastic perturbations in time series and analyzes how these metrics respond to different noise regimes across heterogeneous dynamical systems. A supervised learning approach based on complexity descriptors was developed to analyze controlled perturbations in multiple signal types. Gaussian, pink, and low-frequency noise disturbances were injected at predefined intensity levels into the Rössler and Lorenz chaotic systems, the Hénon map, and synthetic electrocardiogram signals, while AR(1) processes were used for validation on inherently stochastic signals. From these systems, eighteen entropy-based, fractal, statistical, and singular value decomposition-based complexity metrics were extracted from either raw signals or reconstructed phase spaces. These features were used to perform three classification tasks that capture different aspects of noise characterization, including detecting the presence of noise, identifying the perturbation type, and discriminating between different noise intensities. In addition to predictive modeling, the study evaluates the complexity profiles and feature relevance of the metrics under varying perturbation regimes. The results show that no single complexity metric consistently discriminates noise regimes across all systems. Instead, system-specific relevance patterns emerge. Under given experimental constraints (data partitioning, machine learning algorithm, etc.), Approximate Entropy provides the strongest discrimination for the Lorenz system and the Hénon map, the Coefficient of Variation, Sample and Permutation Entropy dominate classification for ECG signals, and the Condition Number and Variance of first derivative together with Fisher Information are most informative for the Rössler system. Across all datasets, the proposed framework achieves an average accuracy of 99% for noise presence detection, 98.4% for noise type classification, and 98.5% for noise intensity classification. These findings demonstrate that complexity metrics capture structural and statistical signatures of stochastic perturbations across a diverse set of dynamic systems. Full article

The vibration signals of Electro-Hydrostatic Actuators (EHAs) exhibit strong non-linearity and non-stationarity, particularly under complex coupling mechanisms, making the extraction of intrinsic fault features computationally challenging. Conventional deep learning approaches often lack mathematical interpretability and struggle to decouple superimposed fault signatures from incomplete datasets. To address these issues, this paper proposes the Enhanced Continuous Wavelet Transform Capsule Network (ECWTCN), an intelligent decoupled diagnosis framework designed for multiscale signal analysis. The architecture integrates a wavelet-kernel convolution layer to extract physically interpretable time–frequency features across multiple scales, effectively capturing transient impulses associated with incipient faults. Furthermore, a novel maximized aggregation routing algorithm is introduced to optimize the dynamic routing process, enhancing global feature aggregation. A distinct advantage of the ECWTCN is its capability to generalize distinct fault patterns, enabling the identification of unseen compound faults by training exclusively on normal and single-fault samples. Comparative experiments show that the proposed method delivers strong multi-label classification performance under operating condition A, achieving a Subset Accuracy of 93.7% and a Label Ranking Average Precision of 0.998. Complexity analysis further confirms the method’s efficiency in terms of FLOPs and parameter size. This work presents a robust, lightweight, and mathematically interpretable solution for the analysis of complex signals in high-reliability equipment. Full article