摘要 : This study introduces an accurate and efficient hybrid framework for fluid-structure interaction (FSI) that exploits the complementary strengths of smoothed particle hydrodynamics (SPH), the finite element method (FEM), and peridynamics (PD). SPH, a mesh-free Lagrangian method, is employed to model complex fluid flows characterized by large deformations and free surfaces, while solid domains are governed by a hybrid PD-FEM approach to reconcile computational efficiency with the ability to capture material failure. The main contribution is an element-wise adaptive FEM-PD transformation strategy, supported by a simplified partner ghost particle (PGP) coupling scheme for consistent information transfer across the PD-FEM interface. Driven by a critical bond stretch criterion, the scheme activates local FEM-PD transformation in regions where damage is expected to evolve. For fluid-solid interfaces, a binary ghost particle (BGP) scheme for SPH-FEM and a disguised ghost particle (DGP) scheme for SPH-PD are introduced. Both schemes accurately enforce interface conditions and provide consistent coupling at the multi-resolution interface. The feasibility and effectiveness of the proposed hybrid SPH-FEM-PD approach are demonstrated through benchmark problems involving structural dynamic deformation and fracture, hydrostatic problems, and FSI scenarios with structural failure.

摘要 : The existing low-light image fusion methods adopt step-by-step operation to improve the overall luminance of the fused image, which leads to the degradation of infrared salient information, modal incompatibility, and fails to address the issue of color deviation. To overcome the above drawbacks, a unified fusion framework is designed in this paper, that is, a knowledge distillation based scene enhancement and color preservation for infrared and low-light image fusion method (UIKDF). Specifically, a dual-stream teacher-student parallel architecture is constructed with the low-light image enhancement network as the teacher stream and an image fusion two-branch network as the student stream. The coupled feature distiller(CFD) is designed to group the teacher-enhanced knowledge and purposefully pull it into the same space as the visible branch of the student model, to ensure that the student model learns scene detail enhancement and is combined with logit distillation for increased efficiency. Secondly, a cascading semantic enhancement optimization mechanism (CSEM) is introduced into the infrared branch of the student network to capture the aggregation of multi-level salient information. In addition, a weighted cross-modality fusion module (WCMF) is developed to dynamically integrate the infrared salient information and the visible detail using the learning weighted graph to further improve the fusion performance. It is worth noting that in order to achieve simultaneous luminance enhancement and chrominance restoration, a transform domain loss function optimization scheme (TD-LossOpt) is proposed. The model is processed in the RGB space, but the loss function is calculated in the YCbCr space, where a gradient-weighted color loss is designed in the CbCr color channel to preserve the scene color. Compared with the existing state-of-the-art fusion methods, the proposed method has better performance in terms of both visual evaluation and objective evaluation on public datasets. The source code is available at https://github.com/lxq-jnu/UIKDF .

摘要 : In drone-based multi-object detection, densely distributed and heavily occluded targets often lead to fragmented feature representations and missed detections. To address these challenges, a Dense Receptive Occlusion Network (DRONet) is proposed for Unmanned Aerial Vehicle (UAV) imagery. DRONet builds on the RT-DETR framework and introduces three modules specifically designed for dense and occluded aerial scenes. Firstly, an Occlusion-Aware KANC Block (OAKB) is integrated into a ResNet18 backbone. By combining Kolmogorov–Arnold Networks with GRAM polynomial expansion, OAKB enhances the discrimination of fragmented and occluded targets while alleviating the latency bottleneck of vanilla KAN. Secondly, a Perceptual Spatial Integration (PSI) module is developed for the feature fusion network. PSI replaces simple concatenation with partial convolution and channel mixing across multiple scales to selectively strengthen small-object features in cluttered backgrounds. Finally, a Scalable Dilated Efficient Aggregation (SDEA) module aggregates deep features using multi-rate dilated convolutions and structural re-parameterization, enlarging the effective receptive field for small and occluded objects without increasing inference-time complexity. Experiments on the VisDrone and CARPK datasets demonstrate that DRONet achieves mAP50 scores of 50.1% and 98.7%, outperforming the RT-DETR baseline by 3.1% and 0.7%, and surpassing recent UAV-specific detectors while maintaining 60 FPS. These results confirm that DRONet improves both detection precision and stability in densely occluded drone scenes. The source code and pre-trained models will be made publicly available upon the acceptance of this manuscript.

摘要 : Fractures introduce significant heterogeneity into rock masses, resulting in complex hydro-mechanical coupling behaviors. By integrating the fracture seepage model with classical Biot's consolidation theory, an extended peridynamic framework is proposed capable of modeling the response of fractured rock masses under multi-physical condition. A spring-like peridynamic interface model is developed to simulate both sealed and open fractures in rock-mass, and an adaptive dynamic relaxation method is adopted to update the deformation field, resolving the large gap in explicit time steps between the deformation and seepage fields and enhancing computational efficiency. Two benchmark examples are presented to validate the accuracy of the proposed model in simulating fracture deformation and coupled processes in fractured rock. The influences of interface type, length, and inclination angle on the consolidation behavior are systematically analyzed, and the effects of interface strength and fluid exchange coefficient on the deformation and seepage response of layered rock masses are investigated.

摘要 : Photocatalytic CO 2 reduction is a pivotal technology for mitigating the energy crisis and achieving carbon neutrality. The formation of heterojunctions and the introduction of defects are two commonly employed methods for enhancing photocatalytic efficiency. In this study, a novel S-scheme heterojunction photocatalyst with oxygen vacancies (V O ), hm-C 4 N 3 /V O -Bi 2 WO 6 (CN/V O -BWO), was rationally designed and successfully synthesized. The synergistic integration of V O and S-scheme heterojunction established a rapid interfacial charge transfer channel between V O -BWO and hm-CN, significantly enhancing the production rates of CO (13.97 μmol g −1 h −1 ) and CH 4 (4.97 μmol g −1 h −1 ) with an apparent quantum yield (AQY) of 0.09 % (400 nm). Theoretical calculations, in-situ XPS and KPFM jointly demonstrate that the internal electric field (IEF) formed by the S-scheme heterojunction enables nearly unimpeded transport of photo-generated electrons and reduces the energy barrier of the rate-determining step reaction. And the V O functions as an electron transport bridge, enhancing the IEF and continuously transferring photoexcited electrons from the V O -BWO interface to the hm-CN surface, thereby participating in photocatalytic CO 2 reduction reactions. This work presents an innovative design strategy for constructing S-scheme heterojunction photocatalysts containing V O , and offering a new idea for the efficient utilization of CO 2 .

摘要 : Janus membranes have attracted extensive attention due to their broad applications in multiple fields. However, their practical utility is often hampered by low positive liquid transport rates (less than 1 μL/s), which severely limits their practical applications in real-world scenarios such as baby urine drainage and diabetic wound exudate management. Herein, a Janus membrane with a large pore array enabling ultrafast unidirectional liquid transport was demonstrated. Specifically, we utilized commercially available hydrophobic cotton fabrics to construct Janus membranes through femtosecond laser microprocessing method. The as-prepared large-pore Janus membrane (LPJM) with pore size larger than 0.3 mm exhibited ultrafast liquid transport speed of > 20 μL/s, capable of transporting 4 μL water completely in < 0.2 s in the positive direction, which was about 45 times faster than the conventional Janus membrane (CJM) made of same material. The LPJM maintains this exceptional performance over at least 15 permeation cycles, demonstrating robust reusability. Furthermore, the large transport channels of LPJM facilitate a continuous, high-flow-rate liquid passage of up to 3000 μL/min with minimal residual liquid retention. This work provides critical inspiration for developing high-performance Janus membranes capable of efficient fluid handling in demanding real-world scenarios.

摘要 : To explore the feasibility of enhancing the electroactivity of coffee ground biochar (CBC)-based carbon mats through controlled incorporation of reduced graphene oxide (rGO), various carbon mats consisting of powdered activated carbon (PAC), CBC, and rGO were fabricated and applied for the removal of representative antibiotics via peroxymonosulfate (PMS) activation under continuous filtration. The effects of the operational modes, water matrices, and target pollutant types on the performance of cathodic electroactive membranes were systematically investigated under limited carbon loading. Results indicated that non-radical singlet oxygen attack and surface-confined oxidation were the dominant pathways for target pollutant removal by CBC/rGO electroactive membranes. The membranes exhibited distinct oxidative selectivity, with a higher oxidation capacity for electron-donating group-bearing sulfonamides (sulfamethoxazole: 96.2% removal) than for electron-withdrawing group-bearing fluoroquinolones (levofloxacin: 62.6% removal) over a short contact time within the carbon mats, as evidenced by the removal efficiencies of both target pollutants and dissolved organic carbon. Density functional theory calculations and liquid chromatography-mass spectrometry analysis revealed that antibiotic degradation involves hydroxylation, C–N/C–O bond cleavage, and aromatic ring opening, which are closely correlated with the structural and electronic properties of the target pollutants. Notably, Ecological Structure-Activity Relationships toxicity predictions demonstrated that several degradation intermediates posed higher developmental and mutagenic toxicity risks than their parent compounds, thereby highlighting the necessity of regulating the oxidative selectivity in electroactive membranes and developing effective post-treatment processes for filtrates to mitigate such toxic risks.

摘要 : With the rapid advancement of 3D printing, concrete is required to bear significant stress loads during its early-age period. The high stress creep failure behavior is directly associated with the structural stability and build ability requirements of 3D printed concrete elements during layer deposition. Nonlinear creep manifests in concrete subjected to high stresses. This process results in the accumulation of damage, which culminates in creep failure of concrete materials. However, limited information exists on this phenomenon in alkali-activated slag concrete. In this study, the creep failure of alkali-activated slag concrete under high uniaxial stress was investigated, and the creep failure characteristics of alkali-activated slag concrete were obtained. Additionally, the effects of loading age and stress ratio were investigated. The failure mechanism, creep coefficient, and nominal Poisson's ratio of alkali-activated slag concrete specimens were investigated. Findings indicated that as age increases, the influence of loading age on the nonlinear creep of alkali-activated slag concrete progressively diminishes. Stress levels significantly affect the nonlinear creep of alkali-activated slag concrete. Specifically, when the stress level reached 0.95 f c, the alkali-activated slag concrete specimen failed after sustaining the load for a few seconds. During the creep phase, the circumferential creep coefficient and nominal Poisson's ratio of the specimen exhibited a substantial increase, with the circumferential creep coefficient markedly exceeding the axial creep coefficient. Furthermore, the specimens exhibited significant lateral expansion upon destruction. Non-destructive and crack analyses were conducted to identify the creep-failure mechanism. The creep-failure study indicated that the propagation of microcracks inside the specimen influenced the nonlinear creep and creep-failure characteristics of alkali-activated slag concrete specimens subjected to high sustained compressive stress. The deterioration of the adhesive substance between particles progressively culminated in extensive fractures, which resulted in creep failure.

摘要 : This study investigates the optimization of interfacial adhesion between degradable silk fibroin (SF) coatings and polyurethane (PU) substrates through controlled surface roughness engineering. By systematically varying substrate roughness (P180–P800 grit), we demonstrate that intermediate roughness (P400–P600, Sa ≈ 0.9 μm) maximizes adhesion via mechanical interlocking, achieving a 40% reduction in fracture area compared to extreme roughness levels. Fourier transform infrared spectroscopy (FT-IR) confirms hydrogen bonding between SF's amide groups and PU's urethane linkages, while Scanning Electron Microscope (SEM) reveals crack-free coatings due to controlled drying (10°C, 50% RH). White Light Interferometry (WLI) quantifies roughness gradients, with P600 substrates exhibiting optimal stress redistribution, 78.7μN critical delamination loads in nanoscratch tests and 1.7 MPa the adhesion strength in peel stress test. These findings establish a tripartite adhesion mechanism—combining mechanical interlocking, wettability optimization, and defect minimization—thereby providing a robust solution to the critical challenge of coating delamination in neural implants.

摘要 : While organic phosphorus (OP) is increasingly identified as a critical driver of eutrophication due to its high bioavailability, traditional remediation strategies predominantly target inorganic phosphate (IP). Herein, a novel calcium‐lanthanum peroxide modified halloysite nanocomposite (CLPH) was developed for the synchronous removal of IP and OP from aqueous systems. Batch experiments revealed superior adsorption capacities for both phosphate (P i , 65.29 mg-P/g) and Myo -inositol hexakisphosphate (IHP, employed as a model OP, 59.40 mg-P/g), surpassing the hydroxide-based analogue (CLH). Furthermore, CLPH demonstrated excellent selectivity in the presence of competing ions and natural organic matter. Mechanistic analysis indicated that P removal is governed by electrostatic attraction, inner-sphere complexation, and surface precipitation. Crucially, IHP removal was found to involve partial catalytic hydrolysis into P i followed by subsequent capture, validating the reactive sequestration mechanism. In real waster matrices, CLPH reduced total P in surface water to below 0.02 mg-P/L within 180 min, and lowered P concentrations in raw sewage and secondary effluent to below 0.30 and 0.05 mg-P/L, respectively, achieving over 98% removal of dissolved OP. These results position CLPH as a promising bifunctional material for mitigating eutrophication through the reactive sequestration of IP and OP.