Dissolved oxygen (DO) is a vital parameter in aquaculture, directly influencing fish health, growth and overall productivity. However, existing DO prediction methods largely rely on complex physical models that demand numerous parameter measurements and high computational cost. This study introduces a hybrid framework that integrates transfer learning (TL) with physics-guided neural networks (PGNNs) based on long short-term memory models (PG-LSTM-TL), to accurately predict pond DO levels from only 1 month of sparse and irregularly sampled nighttime data. The TL strategy leverages knowledge from a data-rich tidal estuary and adapts it to a data-scarce aquaculture setting, whereas the physics-guided component embeds two key domain constraints: The temperature-dependent DO solubility governed by Henry's law and the monotonic nighttime DO depletion trend. Unlike conventional TL approaches that primarily rely on fine-tuning or statistical feature alignment, we embed physics-guided constraints directly into the model learning stage. This integration mitigates negative transfer arising from environmental heterogeneity while preserving physically plausible multi-step DO trajectories. Results show that, for each daily cycle (9:00 p.m.-5:00 a.m.), the model uses just the first four hourly measurements to recursively forecast DO until dawn, achieving an overall mean absolute error (MAE) of 0.3593 mg/L and root mean square error (RMSE) of 0.4813 mg/L. This represents improvements of 35.5% (MAE) and 32.4% (RMSE) over conventional models. By combining physical insights with data efficiency, the proposed approach offers a practical and scalable solution for reliable DO prediction in small-scale aquaculture systems worldwide.

In pond aquaculture, maintaining stable Dissolved Oxygen (DO) concentrations is essential for preventing hypoxia, optimizing growth conditions, and ensuring sustainable operations. Therefore, DO prediction is a crucial aspect of intelligent aquaculture systems, directly influencing water quality, aquatic health, and overall productivity. Advances in sensor technology and data-driven modeling have significantly enhanced the ability to monitor and forecast DO levels, enabling proactive management strategies. This review presents a novel taxonomy for classifying DO prediction approaches in pond aquaculture, structured into three key areas: (1) Driven Factors of DO, examining environmental, biological, and operational influences on DO dynamics; (2) Predictive Models, methods ranging from statistical approaches to advanced deep learning, highlighting promising techniques such as physics-informed neural networks (PINNs) and transfer learning for data-scarce environments; and (3) Monitoring and Sensor Technologies, covering electrochemical and optical sensors, particularly fluorescence-based systems, integrated with Internet of Things (IoT) platforms for real-time assessment. By synthesizing these domains, the review identifies opportunities to enhance DO prediction accuracy and monitoring reliability, supporting intelligent aeration control, improved resource efficiency, and more resilient aquaculture operations.

Multi-label graph learning intends to capture the intrinsic complexity of real-world applications, where one sample is often related to multiple groups or consists of multiple objects. To date, a handful of multi-label graph learning methods exist, but none of them integrate training-time interpretation capability. While post-hoc graph explainers have been developed, they do not explicitly model label-dependent evidence sharing in multi-label graph learners, especially when label pairs are weakly or negatively associated. As a result, post-hoc approaches may miss how evidence should be shared or separated across different labels. This paper advances a new end-to-end self-explainable multi-label graph neural network (SEMGNN), which aims to simultaneously classify multi-labeled nodes and identify edges significantly contributing to each target node w.r.t. predicted labels. Different from post-hoc methods, SEMGNN jointly learns a predictor and a sparse edge-mask explainer within a unified framework and training objective. Label-label correlations are used to improve multi-label node classification and enhance individual label explanations, so that different labels of a node can be supported by distinct yet coherent structural and/or correlated evidence. Experiments and comparisons on synthetic and real-world multi-label networks, in social networking, entertainment, and life sciences, show that SEMGNN achieves competitive or improved predictive performance while providing more faithful and compact label-conditioned explanations.

[Objective] Synthetic dye, a major contributor to global water pollution, poses substantial and irreversible threats to human health and ecological systems. Recently, water treatment technologies based on peroxymonosulfate (PMS) have emerged as a promising solution. The decomposition of PMS generates reactive oxygen species (ROS) with high redox potential, but the slow reaction rate limits its practical application in water remediation. To increase the PMS decomposition efficiency, a comprehensive experiment is designed to synthesize carbon-coated CeO2u2013Co3O4 composite catalysts. [Methods] In this comprehensive experiment, a hollow CeO2u2013Co3O4 precursor was initially prepared via the solvothermal method. Subsequently, a resorcinolu2013formaldehyde (RF) resin layer was applied to the precursor surfaces using the solu2013gel process, followed by carbonization to form a coreu2013shell structured material. Concurrently, by introducing silicon dioxide (SiO2) interlayers of different thickness values as templates, two types of carbon-coated Co3O4u2013CeO2 with a u201Crattleu201D structure were fabricated. The structural characteristics of the materials were analyzed using X-ray diffraction (XRD), transmission electron microscopy (TEM), scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), and N2 adsorptionu2013desorption techniques. The performance of the catalysts was evaluated using methylene blue (MB) as a model organic pollutant, and scavengers were employed to identify the predominant ROS involved in the reaction. [Results] The successful preparation of the designed materials was confirmed by TEM. Compared with the uncoated product (Co3O4u2013CeO2), the RF and SiO2 layers effectively inhibited grain growth during carbonization. The rattle-type composites exhibited significantly higher BET surface areas, particularly the one made with 0.75 mL tetraethyl orthosilicate or TEOS (Co3O4u2013CeO2@h-C-1, 222.1 m2/g). In contrast, the BET surface area of the coreu2013shell structured sample (Co3O4u2013CeO2@C, 24.8 m2/g) was even lower than that of Co3O4u2013CeO2 (55.8 m2/g) due to the sealing effect of dense carbon shell. The Co3O4u2013CeO2@h-C-1 demonstrated the best catalytic performance with 1O2 and SO4u00B7u2212 as the main ROS in MB degradation. This superiority is due to its highest BET surface area and unique rattle-type structure. Moreover, the small grain size of the metal oxides creates more defects and active sites on their surface, which is beneficial for PMS activation. [Conclusions] This comprehensive experiment simulates the entire scientific research process, encompassing the preparation, characterization, performance testing, and data analysis of carbon-coated Co3O4u2013CeO2 composites with different structures. This experimental project trains students to understand the structureu2013activity relationship of catalysts and the comprehensive application of knowledge from chemistry and material disciplines, enhancing students' experimental design capabilities and fostering scientific thinking and innovative awareness.

Development of UHTCs precursors with a high ceramic yield, low viscosity at high solid content, and low oxygen content of after pyrolysis is crucial for UHTCs-modified C/Cs composites fabricated by PIP process. In the present work, four types of TaC precursors with varying microstructures were designed and synthesized. The effects of the precursor microstructure on the morphology, size, ceramic yield and oxygen content after pyrolysis were investigated. The findings demonstrate that the three-dimensional network structure may lower the oxygen content of pyrolysis products and that the carbon source chelated structure can significantly increase the ceramic yield of TaC precursor. After pyrolysis at 1500 degrees C, the ceramic yield of the TaC precursor, TCP-4, was 51.80 wt%, and the oxygen content of the pyrolysis product was only 0.48 wt%. The TCP-4 ethanol solution at 90 wt% viscosity is 67.03 mPa S which meets the requirements of the PIP process. The formation thermodynamics of carbon source chelated three-dimensional network structure is analyzed. The present work not only provides a high-performance TaC precursor, but also offers a novel idea and approach for development of other ultra-high temperature ceramic precursors.

The mismatch between interfacial impedance at the hydrogel-electrical stimulation layer interface in conventional double-layer self-adhesive electrical stimulation wound dressings remains a critical challenge, limiting charge transfer efficiency, and therapeutic outcomes. Herein, this study introduces a photothermally activated pyroelectric-enhanced self-powered wound dressing designed to overcome this limitation through a synergistic tri-modal mechanism integrating photothermal, pyroelectric, and piezoelectric effects. The wound dressing comprises a dual-layer architecture: an outer layer of hydrophobic poly(vinylidene fluoride) (PVDF)/cotton-based electrostimulation film and an inner hydrophilic self-adhesive hydrogel layer. Upon NIR irradiation, the hydrogel layer undergoes localized photothermal heating, dynamically reducing interfacial impedance (≈10× increase in conductivity) and facilitating efficient charge migration across the interface. Concurrently, the NIR-induced photothermal effect activates pyroelectric polarization in the PVDF layer, which synergistically couples with piezoelectric output to generate an enhanced endogenous electric field (≈1.5× the electric field of piezoelectric-only effects). In vitro and in vivo studies showed that this dressing significantly promoted wound healing. Compared with the control group (on the 7th day), the inflammatory chemokine density reduced by 99.36×, the capillary density increased by 3.85×, resulting in a 2.59× enhancement in the wound healing rate. Consequently, the photothermally activated pyroelectric-enhanced self-powered wound dressing presents a highly sophisticated and effective therapeutic approach for accelerating wound healing.

The development of high-temperature electromagnetic wave (EMW) absorbers balancing conductive loss and polarization loss remains a key challenge for aerospace applications. Here in this study, we propose a crystal phase engineering strategy to prepare MXene-derived-TiC/SiC (M-TiC/SiC) composite fibers via precise electrospinning and controlled pyrolysis. During this process, Ti3C2Tx nanosheets undergo in-situ confined transformation into oriented TiC nanocrystals embedded within the fiber matrix, creating tunable heterogeneous interfaces among crystalline TiC and SiC, amorphous SiOxCy, and turbostratic carbon. This compositiondependent design simultaneously optimizes conductive networks and polarization relaxation. Low Ti3C2Tx loading (TSF-3) yields incomplete conductive pathways, while excessive loading (TSF-9) causes impedance mismatch. The optimized specimen (TSF-7) achieves exceptional EMW absorption performance, with a minimum reflection loss (RLmin) of -67.18 dB at 9.92 GHz and a maximum effective absorption bandwidth (EAB) of 2.72 GHz. Dielectric analysis confirms interfacial polarization as the dominant loss mechanism, driven by interfacial charge accumulation-separation and defect-induced dipole interactions. Crucially, the material maintains stability below 600 degrees C, retaining an RLmin of -62.27 dB after oxidation. Gradual attenuation occurs at higher temperatures, establishing an operational limit below 1000 degrees C. This work sets an example for designing thermally stable EMW absorbers through heterogeneous interface engineering, offering transformative potential for applications in extreme environments.

Numerous social recommendation models leverage user-item rating data (i.e., collaborative domain) and social connections (i.e., social domain) to obtain users' preferences for delivering personalized recommendations. However, existing algorithms for Top-N recommendation fail to fully exploit the diversity of the rating data, thereby overlooking certain implicit user preferences. Additionally, it is imperative to acknowledge the unique characteristics inherent in each object, necessitating careful consideration of such distinctions when designing recommendation algorithms. In this paper, we propose multi-relation graph contrastive learning with an adaptive strategy (MCLA) for social recommendation. Specifically, we divide the user-item rating data and the multiple social data into distinct relations, treating each as an independent graph. Graph Neural Networks (GNNs) are then employed to aggregate messages from each relation separately. Furthermore, we use an adaptive fusion strategy to fuse user embeddings in the collaborative domain and social domain by utilizing the user's characteristics. To alleviate the data sparsity problem after relation splitting, we present a simple yet effective adaptive noise contrastive learning method. Experiments on several real-world datasets demonstrate the effectiveness of the proposed MCLA. Both source code and benchmark data are available via https://github.com/chenai1018/MCLA.

Magnesium phosphate bone cement (MPC) is one of the most important candidates for bone implant materials due to its properties such as no need for sintering, self-curing formation, high early compressive strength and rapid degradation. However, traditional methods for creating porous structures for MPC, such as foaming or incorporating degradable materials, often result in closed pores or pores predominantly localized on the surface, severely limiting its comprehensive degradation and effective formation of new bone. In this study, porous MPC scaffolds with directional layered channels were successfully prepared using innovative directional freeze-drying and in situ hydration techniques. The key findings of the study were that the incorporation of degradable acidic PVA staple fibers not only significantly improved the mechanical properties of the porous MPC scaffolds, but also effectively addressed the issue of high alkaline cell toxicity upon degradation. When the fiber content is 3 wt%, the compressive strength of the porous scaffold increases from 4.67 to 8.83 MPa, which was attributed to both the high tensile strength displayed by PVA fibers and the presence of multiple fiber forms within the scaffold. And the acidification of PVA fibers during degradation resulting in a decrease in the pH value of the SBF from 8.93 to 7.67, which reduced the cytotoxicity of the scaffolds. Furthermore, through a 7-day microzone in situ hydration process, the porous MPC scaffold with a Mg/P molar ratio of 1.45:1 transformed into KMgPO4 & sdot;6H2O. The pH value of simulated body fluid (SBF) showed an increasing trend with the rising Mg/P molar ratio following degradation of porous MPC scaffolds. The conclusions of the study showed that the prepared porous PVAf/MPC scaffolds with oriented hierarchical channels exhibit non-cytotoxicity and provide an optimal environment for cell survival and proliferation, and demonstrated great potential applications in the field of biodegradable bone implantation materials.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

Addressing growing electromagnetic pollution, carbon materials, with low density, high specific surface area, designability, and excellent dielectric properties, are strong candidates for electromagnetic wave absorption. However, their high conductivity causes poor impedance matching, severely limiting applications. By combining them with BN materials that have a low dielectric constant and controlling their morphology and interfacial bonding at the micro-nano scale, it is expected to resolve this issue. Therefore, by constructing a carbon nano-particle coating layer on the surface of boron nitride nanosheets (BNNS), cactus-like BNNS@C composites were successfully prepared. The results indicated that the carbon nanoparticles were uniformly distributed on the surface of BNNS. Importantly, this BNNS@C composites exhibit relatively low graphitization degree, high defect content, and nitrogen doping characteristics. And at the thickness of 1.0 mm, the cactus-like BNNS@C composites achieved a minimum reflection loss value (RLmin) of-35.54 dB and an effective absorption bandwidth (EAB) of 2.86 GHz (15.14-18.00 GHz). This benefit was due to the conduction loss, dipole polarization loss, and the multiple reflection losses within the heterogeneous interfaces of the cactus-like BNNS@C composites. Therefore, this study provides a reliable reference for the preparation of lightweight and thin high-efficiency microwave absorbing composite materials.

This study explores the integration of artificial intelligence (AI) and finite element analysis (FEA) in spine surgery, highlighting their complementary roles across preoperative planning, intraoperative execution, and postoperative outcome prediction. The synergy between AI and FEA is reshaping modern spine care by improving biomechanical modeling, enhancing surgical precision, and enabling personalized treatment strategies. In the preoperative phase, AI-augmented FEA supports the design of patient-specific surgical plans, optimizing implant placement and simulating mechanical responses under various loading conditions. Intraoperatively, AI enables real-time image-guided navigation, robotic assistance, and automated anatomical recognition, reducing the risk of surgical error. Postoperatively, predictive models built on FEA simulations and patient data assist in tracking recovery, forecasting complications, and informing rehabilitation protocols. Together, these technologies contribute to a data-driven paradigm shift toward precision spine surgery. As intelligent feedback systems, digital twins, and autonomous surgical platforms continue to evolve, AI–FEA integration is poised to play a transformative role in delivering safer, more efficient, and individualized spine care.

Pure BN fibers exhibit limitations in electromagnetic wave absorption due to their low dielectric constant and magnetic permeability. While some existing heterogeneous absorbing materials can enhance absorption capacity, the complexity and high density of these material structures restrict further development. In this study, BN-NiOVOx fiber composite with a high-density heterogeneous interfacial structure was developed by hydrothermal method combined with calcination. Regulation of calcination temperature at 600 degrees C led to the formation of VOxloaded NiO nanoflower spheres and a multiphase composition on BN fiber surfaces, effectively increasing the heterogeneous interface density. Performance tests demonstrate excellent electromagnetic wave absorption capabilities, with a maximum absorption bandwidth of up to 4 GHz and a peak reflection loss greater than 20 dB. This study offers new insights into the design of high-performance composite wave-absorbing materials and provides robust support for the practical application of electromagnetic pollution protection technologies.

This paper proposes the application of explanation methods to enhance the interpretability of graph neural network (GNN) models in fault location for power grids. GNN models have exhibited remarkable precision in utilizing phasor data from various locations around the grid and integrating the system’s topology, an advantage rarely harnessed by alternative machine learning techniques. This capability makes GNNs highly effective in identifying fault occurrences in power grids. Despite their greater performance, these models can encounter criticism for their "black box" nature, which conceals the reasoning behind their predictions. Lack of transparency significantly hinders power utility operations, as interpretability is crucial to building trust, accountability, and actionable insights. This research presents a comprehensive framework that systematically evaluates state-of-the-art explanation strategies, representing the first use of such a framework for Graph Neural Network models for defect location detection. By assessing the strengths and weaknesses of different explanatory methods, it identifies and recommends the most effective strategies for clarifying the decision-making processes of GNN models. These recommendations aim to improve the transparency of fault predictions, allowing utility providers to better understand and trust the models’ output. The proposed framework not only enhances the practical usability of GNN-based systems but also contributes to advancing their adoption in critical power grid applications.

The PZS surface can provide excellent photothermal deicing performance and corrosion resistance for aluminum alloys.

With the rapid advancement of electronic information technology, the issue of electromagnetic radiation has become increasingly severe. It is urgent to develop high efficiency microwave absorber with excellent microwave absorbing performance. The construction of complex and diversified hierarchical porous structure is regarded as a promising way to enhance electromagnetic wave absorption. In this study, leveraging the porous architecture of melamine foam, a BN@C composite material featuring a hierarchical pore structure comprising larger and smaller pores, both residing in the micrometer regime, was synthesized via a high-temperature-assisted sol-gel method, using polyvinyl alcohol as the carbon precursor and boron nitride (BN) as the filler. By optimizing the molar ratio of raw materials, BN@C composites exhibiting superior microwave absorption performance were successfully obtained. When the matching thickness was set at 2.5mm and 1.5mm, respectively, the minimum reflection loss (RLmin) reached -29.12dB, and the maximum effective absorption bandwidth (EAB) was 5.44GHz, demonstrating a broad application prospect in the realm of microwave absorption. This research not only provides novel insights into enhancing the microwave absorption capabilities of carbon-based materials but also lays a material foundation for the development of electromagnetic radiation protection and stealth technologies.

In this work, low-cost and positively-charged AgO-based composites have been successfully synthesized using commercial MgO nanoparticles as support and charge modifier. These AgO-based composites were composed of monoclinic AgO, cubic AgCl, cubic MgO and hexagonal Mg(OH)2. The AgO content in the composites increased with increasing oxidation time. The composites possessed nearly hexagonal sheet-like shape with the median size in the range of 150-180 nm, and increased specific surface area as well as abundant positive charges in a wide pH range. Importantly, these positively-charged AgO-based composites exhibited enhanced bactericidal activity against both S. aureus and E. coli. The maximum bactericidal rates against S. aureus and E. coli. were 99.96 % and 99.98 %, respectively. The enhanced antibacterial activity of AgO-based composites was caused by the synergistic effects from the electrostatic interaction, lager specific surface area and positive charge. Accordingly, the AgO-based composites would be useful for the fast and effective sterilization of bacteria.

In the field of dentistry, the process of resorption and atrophy of the alveolar bone is frequently observed foll,owing tooth extraction, a phenomenon that can significantly impact the recovery process of patients undergoing dental implant treatment. Bone powder has emerged as a promising solution for the reconstruction of alveolar bone, offering a protective barrier and a means to replenish defective bone tissue by filling the voids with this material. Hydroxyapatite (HA) bone powder has garnered significant attention in research endeavors focused on bone restoration materials, owing to its distinctive advantages. In this study, a novel approach was employed, utilizing yeast as a template for synthesizing HA bone powder with a high specific surface area and mesoporous structure. The performance and formation mechanism of HA bone powder were thoroughly investigated, with the objective of providing a more effective solution for bone repair. The experimental findings demonstrated that the particle characteristics of HA precursors with varying fermentation times exhibited variability, and the sintering temperature influenced the crystallinity, particle size, and pore structure of porous HA. The most optimal samples were identified through a comparative analysis with Bio-Oss bone powders. The rate of change in mineralization quality and the rate of release of degraded calcium and phosphorus ions exhibited variation in accordance with different fermentation times. The binding mode and energy difference between phenylalanine and Ca2+ were revealed by density flooding theory and provided heterogeneous nucleation sites for HA nucleation, which ultimately led to the formation of a porous structure. However, the study is not without its limitations, and the process can be further optimized in the future to explore the long-term performance and biocompatibility of the material in vivo. This would expand the scope of application and provide better solutions for dental and alveolar bone problems.

Cellulose-based conductive gels represent a unique platform for integrating intelligent electronic devices seamlessly into daily life due to their excellent flexibility, adjustable three-dimensional (3D) structure, and sustainability. Mechanical strength and conductivity, as two key parameters, play significant roles in this process. Nevertheless, transferring excellent mechanical properties and conductivity to 3D gels simultaneously poses numerous challenges due to their inherent conflict in typical cases. The advancements in functionalizing crosslinking networks at the single cellulosic material level and within the constructed cellulose-based 3D matrix have fundamentally altered their utility. This review provides a systematic and in-depth understanding of designing advanced crosslinking networks in developing cellulose-based conductive gels with superior mechanical strength and conductivity. Here, we introduce the advantages of cellulose in designing conductive gels and the component effect of the gels on mechanical and conductive properties. Then, we systematically summarize the importance and design methods of crosslinking network engineering in balancing these features theoretically. Furthermore, fabrication strategies for achieving superior mechanical strength and enhanced conductivity through structural optimization of cellulose-derived crosslinking networks are investigated, with particular emphasis on interfacial engineering and functional integration mechanisms. We further review the compatibility of crosslinking networks and other key properties (self-healing and low-temperature tolerance). We also discuss advanced analysis methods of structure-performance relationship for developing novel cellulose-based conductive gels with superior physicochemical characteristics. Finally, we introduce potential applications and highlight key technologies to broaden the application prospects of cellulose-based conductive gels for smart wearable devices.