Search Results (15)

Epileptic seizure detection and classification from EEG recordings faces significant challenges due to extreme class imbalance. Analysis of the Temple University Hospital Seizure (TUSZ) dataset reveals imbalance ratios of 150:1 between common and rare seizure types, with high temporal heterogeneity (seizure durations of 1–1638 s). We propose a cascaded deep learning architecture with two specialized CNNs: a binary detector followed by a multi-class classifier. This approach decomposes the classification problem, reducing the maximum imbalance from 150:1 to manageable levels (9:1 binary, 5:1 type). The architecture implements a high-confidence filtering mechanism (threshold = 0.9), creating a 99.5% pure dataset for type classification, dynamic class-weighted optimization proportional to inverse class frequencies, and information flow refinement through progressive stages. Loss dynamics analysis reveals that our weighting scheme strategically redistributes optimization attention, reducing variance by 90.7% for majority classes while increasing variance for minority classes, ensuring all seizure types receive proportional learning signals regardless of representation. The binary classifier achieves 99.64% specificity and 98.23% sensitivity (ROC-AUC = 0.995). The type classifier demonstrates >99% accuracy across seven seizure categories with perfect (100%) classification for three seizure types despite minimal representation. Cross-dataset validation on the University of Bonn dataset confirms robust generalization (96.0% accuracy) for binary seizure detection. This framework effectively addresses multi-level imbalance in neurophysiological signal classification with hierarchical class structures. Full article

To address the limited generalization capability of data-driven fault diagnosis models caused by scarce gearbox fault samples in engineering practice, this paper proposes a hierarchical attention-guided data–knowledge dual-driven fusion network for intelligent fault diagnosis under few-shot conditions. Distinct from traditional single data-driven paradigms, this method breaks through the constraints of limited samples through the synergy of prior knowledge and monitoring data. First, domain knowledge of gearbox fault diagnosis is utilized to construct prior features of monitoring data. Second, a deep convolutional neural network is designed to hierarchically capture abstract features from monitoring data. Subsequently, a hierarchical attention module is proposed to realize adaptive fusion of prior features and abstract features through hierarchical feature weight allocation, generating highly discriminative fused features for accurate gearbox fault identification. Experimental results on gearbox fault data demonstrate that the proposed method achieves 0.9880 recognition accuracy with less than 10% of the training samples, significantly outperforming purely data-driven models such as MGAN and CNET, thus verifying its superior generalization ability to train despite data scarcity. This approach establishes a novel data–knowledge dual-driven fusion paradigm for intelligent fault diagnosis of mechanical equipment under few-shot conditions. Full article

Transformers have performed favorably in recent hyperspectral unmixing studies in which the self-attention mechanism possesses the ability to retain spectral information and spatial details. However, the lack of reliable prior information for correction guidance has resulted in an inadequate accuracy and robustness of the network. To benefit from the advantages of the Transformer architecture and to improve the interpretability and robustness of the network, a dual-branch network with prior information correction, incorporating a Transformer network (PICT-Net), is proposed. The upper branch utilizes pre-extracted endmembers to provide pure pixel prior information. The lower branch employs a Transformer structure for feature extraction and unmixing processing. A weight-sharing strategy is employed between the two branches to facilitate information sharing. The deep integration of prior knowledge into the Transformer architecture effectively reduces endmember variability in hyperspectral unmixing and enhances the model’s generalization capability and accuracy across diverse scenarios. Experimental results from experiments conducted on four real datasets demonstrate the effectiveness and superiority of the proposed model. Full article

Zinc has attracted significant attention as a versatile material with potential applications in various fields, particularly in biomedical engineering. Despite its desirable characteristics, such as biodegradability and biocompatibility, the inherently low mechanical strength of zinc has been a major limitation for its broader use in clinical applications. To address this issue and enhance its mechanical performance without compromising its biocompatibility, a novel composite material was developed by mixing zinc oxide (ZnO) with zinc (Zn). ZnO is widely recognized for its high chemical stability, non-toxicity, and antimicrobial properties, making it an excellent additive for biomedical materials. In this study, Zn-ZnO nanocomposites were fabricated by uniformly dispersing ZnO nanoparticles into molten zinc using an ultrasonic processor. The uniform distribution of ZnO nanoparticles within the zinc matrix was confirmed, and the resulting nanocomposites demonstrated remarkable improvements in mechanical properties. Specifically, the hardness and tensile strength of the Zn-ZnO nanocomposites were increased by approximately 90% and 160%, respectively, compared to pure zinc. To evaluate the biodegradation behavior of the materials, both pure zinc and Zn-ZnO nanocomposite samples were immersed in phosphate-buffered saline (PBS) at 37 °C, simulating physiological conditions. The degradation rate was assessed by measuring the weight loss of the material over time. The biodegradation rate of the Zn-ZnO nanocomposites was found to be nearly identical to that of pure zinc under identical conditions, indicating that the addition of ZnO did not adversely affect the degradability of the material. These findings suggest that Zn-ZnO nanocomposites offer a promising solution for biomedical applications by combining improved mechanical properties with maintained biodegradability and biocompatibility. Full article

Approaching precipitation forecast refers to the prediction of precipitation within a short time scale, which is usually regarded as a spatiotemporal sequence prediction problem based on radar echo maps. However, due to its reliance on single-image prediction, it lacks good capture of sudden severe convective events and physical constraints, which may lead to prediction ambiguities and issues such as false alarms and missed alarms. Therefore, this study dynamically combines meteorological elements from surface observations with upper-air reanalysis data to establish complex nonlinear relationships among meteorological variables based on multisource data. We design a Residual Spatiotemporal Convolutional Network (ResSTConvNet) specifically for this purpose. In this model, data fusion is achieved through the channel attention mechanism, which assigns weights to different channels. Feature extraction is conducted through simultaneous three-dimensional and two-dimensional convolution operations using a pure convolutional structure, allowing the learning of spatiotemporal feature information. Finally, feature fitting is accomplished through residual connections, enhancing the model’s predictive capability. Furthermore, we evaluate the performance of our model in 0–3 h forecasting. The results show that compared with baseline methods, this network exhibits significantly better performance in predicting heavy rainfall. Moreover, as the forecast lead time increases, the spatial features of the forecast results from our network are richer than those of other baseline models, leading to more accurate predictions of precipitation intensity and coverage area. Full article

Edible coatings have attracted the attention of researchers in recent years due to their degradability, safety, non-toxicity, low cost, good preservation effect, and other advantages. To prepare a new edible film with good mechanical and barrier properties, carboxymethyl chitosan (CMCS) and gelatin (GL) were selected as the film-forming matrix in this experiment, and glycerol, CaCl₂, Tween-20, and ascorbic acid (AA) have been added as plasticizers, crosslinking agents, surfactants, and antioxidants. Crosslinking agents and antioxidants first, the film was prepared by the casting method, and single factor tests were used to compare the effects of different CMCS: GL (w:w), glycerol, CaCl₂, Tween-20, and AA on mechanical properties (Tensile Strength (TS), Elongation at break (EAB)) and barrier properties (Water Vapor Permeability (WVP), Oxygen Permeability (OP)). Then, the weighting of each performance index is determined by a combination of principal component analysis and the comprehensive membership evaluation method. The formula for calculating the overall rating of edible film performance was determined. Finally, the manufacturing process of edible film with better performance was optimized by a response surface test. The results showed that the influence of each factor on the performance of the edible film was as follows: Glycerol addition > CaCl₂ addition > CMCS:GL, Tween-20, and AA had no significant influence on the performance of the edible film. When calculating the overall edible film property score, the weights of TS, EAB, WVP, and OP were 0.251, 0.068, 0.334, and 0.347, respectively. The optimal formulation for an edible film based on CMCS-GL with better properties than pure CMCS and GL film was CMCS:GL = 2:1, with the addition of 1% glycerol, 2% CaCl₂, 0.1% Tween-20, and 2% AA. The TS, EAB, OP, and WVP of the film obtained with this formula were: 16.28 MPa, 71.46%, 1.39 × 10⁻¹² g·cm/(cm²·s·Pa), 5.10 × 10⁻¹¹ cm³·cm/(m²·s·Pa), respectively. This study suggests that CMCS-GL-based edible coatings can be used as a new food packaging material. Full article

Hydrogels containing hydrophobic materials have attracted great attention for their potential applications in drug delivery and biosensors. This work presents a kneading-dough-inspired method for dispersing hydrophobic particles (HPs) into water. The kneading process can quickly mix HPs with polyethyleneimine (PEI) polymer solution to form “dough”, which facilitates the formation of stable suspensions in aqueous solutions. Combining with photo or thermal curing processes, one type of HPs incorporated PEI-polyacrylamide (PEI/PAM) composite hydrogel exhibiting good self-healing ability, tunable mechanical property is synthesized. The incorporating of HPs into the gel network results in the decrease in the swelling ratio, as well as the enhancement of the compressive modulus by more than five times. Moreover, the stable mechanism of polyethyleneimine-modified particles has been investigated using surface force apparatus, where the pure repulsion during approaching contributes to the good stability of the suspension. The stabilization time of the suspension is dependent on the molecular weight of PEI: the higher the molecular weight is, the better the stability of the suspension will be. Overall, this work demonstrates a useful strategy to introduce HPs into functional hydrogel networks. Future research can be focused on understanding the strengthening mechanism of HPs in the gel networks. Full article

The formation of (bio)molecular condensates via liquid–liquid phase separation in cells has received increasing attention, as these aggregates play important functional and regulatory roles within biological systems. However, the majority of studies focused on the behavior of pure systems in bulk solutions, thus neglecting confinement effects and the interplay between the numerous molecules present in cells. To better understand the physical mechanisms driving condensation in cellular environments, we perform molecular simulations of binary polymer mixtures in spherical droplets, considering both monodisperse and polydisperse molecular weight distributions for the longer polymer species. We find that confinement induces a spatial separation of the polymers by length, with the longer ones moving to the droplet center. This partitioning causes a distinct increase in the local polymer concentration near the droplet center, which is more pronounced in polydisperse systems. Consequently, the confined systems exhibit liquid–liquid phase separation at average polymer concentrations where bulk systems are still in the one-phase regime. Full article

To learn discriminative features, hyperspectral image (HSI), containing 3-D cube data, is a preferable means of capturing multi-head self-attention from both spatial and spectral domains if the burden in model optimization and computation is low. In this paper, we design a dual multi-head contextual self-attention (DMuCA) network for HSI classification with the fewest possible parameters and lower computation costs. To effectively capture rich contextual dependencies from both domains, we decouple the spatial and spectral contextual attention into two sub-blocks, SaMCA and SeMCA, where depth-wise convolution is employed to contextualize the input keys in the pure dimension. Thereafter, multi-head local attentions are implemented as group processing when the keys are alternately concatenated with the queries. In particular, in the SeMCA block, we group the spatial pixels by evenly sampling and create multi-head channel attention on each sampling set, to reduce the number of the training parameters and avoid the storage increase. In addition, the static contextual keys are fused with the dynamic attentional features in each block to strengthen the capacity of the model in data representation. Finally, the decoupled sub-blocks are weighted and summed together for 3-D attention perception of HSI. The DMuCA module is then plugged into a ResNet to perform HSI classification. Extensive experiments demonstrate that our proposed DMuCA achieves excellent results over several state-of-the-art attention mechanisms with the same backbone. Full article

Ocean current (OC) prediction plays an important role for carrying out ocean-related activities. There are plenty of studies for OC prediction with deep learning to pursue better prediction performance, and the attention mechanism was widely used for these studies. However, the attention mechanism was usually combined with deep learning models rather than purely used to predict OC, or, if it was purely used, did not further optimize the attention weight. Therefore, a deep learning model based on weighted pure attention mechanism is proposed in this paper. This model uses the pure attention mechanism, introduces a weight parameter for the generated attention weight, and moves more attentions from other elements to the key elements based on weight parameter setting. To our knowledge, it is the first attempt to use the weighted pure attention mechanism to improve the OC prediction performance, and it is an innovation for OC prediction. The experiment results indicate that the proposed model can fully take advantage of the strengths from the pure attention mechanism; it can further optimize the pure attention mechanism and significantly improve the prediction performance, and is reliable for OC prediction with high performance for a wide time range and large spatial scope. Full article

A number of health-promoting properties of Stevia rebaudiana Bertoni and its glycosides, including the antihyperglycemic activity, have been found. The mechanisms of the antidiabetic action of stevia have not been fully understood. The aim of this study was to evaluate the effects of supplementary steviol glycosides on high-fat fed streptozotocin-induced diabetic rats with particular attention to lipid metabolism. The experiment was conducted on 70 male Wistar rats, of which 60 were fed a high-fat diet for 8 weeks followed by intraperitoneal injection of streptozotocin, to induce type 2 diabetes. Afterwards, rats were divided into six groups and fed a high-fat diet supplemented with pure stevioside or rebaudioside A, at two levels (500 or 2500 mg/kg body weight (b.w.)) for 5 weeks. Three additional groups: diabetic untreated, diabetic treated with metformin, and healthy, served as respective controls. Blood and dissected internal organs were collected for hematological, biochemical, and histopathological tests. It was found that dietary supplementation with steviol glycosides did not affect blood glucose, insulin, and insulin resistance indices, antioxidant biomarkers, but normalized hyperlipidemia and affected the appetite, as well as attenuated blood liver and kidney function indices, and reduced tissular damage in diabetic rats. Steviol glycosides normalize lipid metabolism and attenuate internal organs damage in diabetes. Full article

Alloys in the V-Si-B system are a new and promising class of light-weight refractory metal materials for high temperature applications. Presently, the main attention is focused on three-phase alloy compositions that consist of a vanadium solid solution phase and the two intermetallic phases V₃Si and V₅SiB₂. Similar to other refractory metal alloys, a major drawback is the poor oxidation resistance. In this study, initial pack-cementation experiments were performed on commercially available pure vanadium and a three-phase alloy V-9Si-5B to achieve an oxidation protection for this new type of high temperature material. This advance in oxidation resistance now enables the attractive mechanical properties of V-Si-B alloys to be used for high temperature structural applications. Full article

Fatigue is one of the main causes for in service failure of mechanical components and structures. With the development of new materials, such as high strength aluminium or titanium alloys with different microstructures from steels, materials no longer have a fatigue limit in the classical sense, where it was accepted that they would have ‘infinite life’ from 10 million (107) cycles. The emergence of new materials used in critical mechanical parts, including parts obtained from metal additive manufacturing (AM), the need for weight reduction and the ambition to travel greater distances in shorter periods of time, have brought many challenges to design engineers, since they demand predictability of material properties and that they are readily available. Most fatigue testing today still uses uniaxial loads. However, it is generally recognised that multiaxial stresses occur in many full-scale structures, being rare the occurrence of pure uniaxial stress states. By combining both Ultrasonic Fatigue Testing with multiaxial testing through Single-Input-Multiple-Output Modal Analysis, the high costs of both equipment and time to conduct experiments have seen a massive improvement. It is presently possible to test materials under multiaxial loading conditions and for a very high number of cycles in a fraction of the time compared to non-ultrasonic fatigue testing methods (days compared to months or years). This work presents the current status of ultrasonic fatigue testing machines working at a frequency of 20 kHz to date, with emphasis on multiaxial fatigue and very high cycle fatigue. Special attention will be put into the performance of multiaxial fatigue tests of classical cylindrical specimens under tension/torsion and flat cruciform specimens under in-plane bi-axial testing using low cost piezoelectric transducers. Together with the description of the testing machines and associated instrumentation, some experimental results of fatigue tests are presented in order to demonstrate how ultrasonic fatigue testing can be used to determine the behaviour of a steel alloy from a railway wheel at very high cycle fatigue regime when subjected to multiaxial tension/torsion loadings. Full article

With the more serious threats from complex electromagnetic environments, composites composed of conductive or semiconductive fillers and polymeric matrices could exhibit excellent nonlinear I-V characteristics, and have drawn significant attention in the field of overvoltage protection. In this research, graphene nanoplatelets (GNPs) are decorated by ZnO and mixed into an epoxy resin (ER) matrix via solution blending to prepare composites. A characterization analysis and the I-V measurement results of the GNPs/ER composites indicate that ZnO nanoparticles are well bonded with GNPs and exhibit obvious nonlinear I-V behavior under proper applied voltage with high nonlinear coefficients. The switching threshold voltage and nonlinear coefficients could be controlled by adjusting the weight ratio of GNPs and ZnO of the filler. Moreover, compared with the poor recoverability of pure GNP-filled ER in previous research, the GNP-ZnO/ER composites exhibited excellent reversibility of nonlinear I-V behavior under multiple repetitive I-V measurements. And compared with different composites, the sample with a 1:8 weight ratio of GO to Zn(Ac)₂ presents the smallest variation of switching threshold voltage at 158 V, with a standard deviation of 1.27% from among 20 measurements, which indicates the best reversibility. Finally, the conducting mechanism of the reversible nonlinear I-V characteristic is investigated and analyzed. Full article

Rigid metallic fiber structures made from a variety of different metals and alloys have been investigated mainly with regard to their functional properties such as heat transfer, pressure drop, or filtration characteristics. With the recent advent of aluminum and magnesium-based fiber structures, the application of such structures in light-weight crash absorbers has become conceivable. The present paper therefore elucidates the mechanical behavior of rigid sintered fiber structures under quasi-static and dynamic loading. Special attention is paid to the strongly anisotropic properties observed for different directions of loading in relation to the main fiber orientation. Basically, the structures show an orthotropic behavior; however, a finite thickness of the fiber slabs results in moderate deviations from a purely orthotropic behavior. The morphology of the tested specimens is examined by computed tomography, and experimental results for different directions of loading as well as different relative densities are presented. Numerical calculations were carried out using real structural data derived from the computed tomography data. Depending on the direction of loading, the fiber structures show a distinctively different deformation behavior both experimentally and numerically. Based on these results, the prevalent modes of deformation are discussed and a first comparison with an established polymer foam and an assessment of the applicability of aluminum fiber structures in crash protection devices is attempted. Full article