Search Results (9,188)

Ultra-precision thinning of hard and brittle materials like monocrystalline silicon demands high dynamic stability in thinning spindle. To address the challenge of accurately detecting subtle spindle chatter anomalies in industrial environments characterized by high noise and limited data, this paper proposes a physics-guided dual-stream attention fusion transfer network (PG-AFNet). First, a physics-guided signal preprocessing method was developed. Using variational mode decomposition (VMD) and continuous wavelet transform (CWT) masking, one-dimensional dynamic features and high-frequency regions of interest (ROIs) rich in transient impact features were extracted. Second, the PG-AFNet architecture was designed. By introducing an attention mechanism, it achieves deep integration of one-dimensional purely dynamic sequences with two-dimensional spatiotemporal visual textures to capture surface damage features caused by subtle vibrations. Finally, systematic validations were conducted using a real silicon wafer thinning dataset with 197 real samples. By overcoming small-sample limitations via physical augmentation, PG-AFNet achieved an 82.45% (86.64% after data augmentation) diagnostic accuracy, significantly outperforming traditional baselines. Furthermore, a large-scale cross-load validation on the diverse CWRU dataset yielded an exceptional 99.68% accuracy under mixed-load conditions, conclusively verifying the model’s robust domain generalization. Lastly, a rigorous ablation study explicitly quantified the indispensable contributions of the physics-guided dual-stream architecture and attention fusion. This research provides a feasible theoretical foundation for intelligent surface quality monitoring in semiconductor hard-brittle material processing. Full article

The reliability of high-frequency (HF) frequency selection technology relies on the prediction accuracy of the Maximum Usable Frequency of the ionospheric F2 layer (MUF-F2). To improve its short-term prediction performance, a novel hybrid deep learning prediction model is proposed, which achieves accurate modeling of the complex spatiotemporal variation patterns of MUF-F2 by integrating a feature enhancement mechanism, a dual-branch feature extraction structure, and a bidirectional temporal dependency capture network. The hybrid prediction model integrates the Channel Attention mechanism (CA), Dual-Branch Convolutional Neural Network (DCNN), and Bidirectional Long Short-Term Memory network (BiLSTM). The model is trained and validated using MUF-F2 data from 5 communication links over China during geomagnetically quiet periods and 4 during geomagnetic storm periods, with the difference in the number of links attributed to experimental constraints and the disruptive effects of geomagnetic storms. Its performance is evaluated via multiple metrics, and a comparative analysis is conducted with commonly used prediction models such as the Long Short-Term Memory (LSTM) network. Experimental results show that during geomagnetically quiet periods, the proposed model achieves lower prediction errors (Root Mean Square Error (RMSE) < 1.1 MHz, Mean Absolute Percentage Error (MAPE) < 3.8%) and a higher goodness of fit (coefficient of determination (R²) > 0.94), with the average error reduction across all links ranging 8 from 6.2% to 46.9% compared with the baseline model. Under geomagnetic storm disturbance conditions, the model still maintains robust prediction performance, with R² > 0.89 for all communication links, as well as RMSE < 0.6 MHz, Mean Absolute Error (MAE) < 0.4 MHz, and MAPE < 3.3%. The study demonstrates that the proposed CA-DCNN-BiLSTM model exhibits excellent prediction accuracy and anti-interference capability under different geomagnetic activity conditions, which can effectively improve the short-term prediction accuracy of MUF-F2 and provide more reliable technical support for HF communication frequency decision-making. Full article

CAR-T cell therapy has shown substantial efficacy in haematological malignancies, but its application to solid tumours remains limited by poor effector-cell infiltration, functional exhaustion, antigenic heterogeneity, and an immunosuppressive microenvironment. In this study, we develop a new spatiotemporal mathematical model of CAR-T therapy for solid tumours that integrates these resistance mechanisms within a single reaction–diffusion framework. The model is formulated as a system of partial differential equations describing functional and exhausted CAR-T cells, antigen-positive and antigen-low tumour subpopulations, and chemokine, immunosuppressive, and hypoxic fields. Steady-state analysis and finite-difference simulations showed that therapeutic outcome is governed by the interplay between CAR-T cell infiltration, exhaustion, and antigen escape. The model reproduces partial tumour regression followed by residual tumour persistence, therapy-driven enrichment of antigen-low cells, and reduced efficacy under stronger immunosuppressive and hypoxic conditions. In the combination therapy scenario considered here, repeated simulated CAR-T cell administration together with attenuation of the suppressive microenvironment improves tumour control. The proposed model provides a mechanistic basis for analysing resistance and for future optimisation studies of CAR-T therapy in solid tumours. Full article

Nonlinear state estimation in electric power systems remains challenging under mixed-measurement conditions due to the coexistence of legacy SCADA and PMU data with markedly different reliability levels, the sensitivity of classical Gauss–Newton-type methods to heterogeneous noise and numerical conditioning, and the increasing complexity of large-scale grids. To address these issues, this paper proposes ST-ResGAT, a spatio-temporal residual graph attention framework for nonlinear state estimation under heterogeneous sensing conditions. The proposed method models the problem on an augmented heterogeneous factor graph, employs a reliability-aware heterogeneous graph attention mechanism with residual propagation to adaptively fuse measurements of different quality, and further refines the graph-based estimates through a lightweight LSTM post-processing module that exploits short-term temporal continuity. All datasets are generated using pandapower on the IEEE 30-bus, IEEE 118-bus, and IEEE 1354-bus benchmark systems to ensure full reproducibility of the experimental pipeline. Experimental results show that the proposed method consistently achieves lower estimation errors than WLS, DNN, GAT, and PINN baselines across all three systems, while also exhibiting more compact node-level error distributions and stronger spatial consistency. Multi-seed ablation studies further indicate that residual propagation, reliability-aware attention, and temporal refinement play complementary roles across different system scales. Robustness experiments additionally show that, under random measurement exclusion as well as bias, Gaussian, and mixed corrupted-measurement settings, ST-ResGAT exhibits smooth and progressive degradation, including on the newly added large-scale IEEE 1354-bus benchmark. These results suggest that the proposed framework is a promising direction for data-driven state estimation under controlled mixed-measurement benchmark conditions. Full article

Retinal detachments following blunt ocular trauma are challenging to predict due to the complex and transient biomechanical responses of the globe. This study combines an in vitro weight-drop experiment and finite element analysis (FEA) to evaluate the mechanical pathways leading to traumatic retinal detachment and to predict the spatial likelihood of injury. In the in vitro model, a cylindrical weight was impacted onto freshly enucleated mouse eyes (16 weeks old) supported on a rigid metal plate. Following impact, the eyes were sectioned and stained using hematoxylin and eosin (H&E) for histological assessment. A finite element model of a mouse eye, including the cornea, sclera, lens, zonule, vitreous body, aqueous humor, and retina, was reconstructed from the histological section and used to simulate the whole sequence of compression and rebound following the blunt impact. The simulation demonstrated that the lens retained a high momentum. It generated an alternating compressive (up to −6.57 × 10⁻³ MPa) and tensile (up to 1.62 × 10⁻³ MPa) radial stress at the posterior pole and sustained compressive stress at the peripheral region (up to −3.12 × 10⁻³ MPa) and tensile-compressive stress variation at the equatorial region of the retina. In addition, the regions experiencing tensile stress overlapped with the region exhibiting retinal detachment in the in vitro experiment. These findings highlight the spatiotemporal mapping of biomechanical stress to predict traumatic retinal detachment following blunt impact and provide an understanding of early biomechanical response following ocular trauma. Full article

Non-cohesive earth dams are widely distributed in natural and semi-engineering scenarios, and overtopping-induced breaches are their most catastrophic failure mode. Accurate prediction of the overtopping failure process and breach evolution is critical for risk assessment, emergency management, and dam design optimization. In this study, an improved 3D numerical method is developed to simulate the coupled hydrodynamic–erosion–breach evolution processes of non-cohesive earth dams. The model based on the finite volume method integrates three core modules: a hydrodynamic module based on the Reynolds-Averaged Navier–Stokes equations with the Volume of Fluid method for free surface tracking, a dam material erosion module considering particle entrainment and transport mechanisms of non-cohesive soils, and a breach development module coupling erosion and gravitational collapse. To validate the model, two levels of verification are conducted: first, a classic benchmark dam break case is employed to confirm the feasibility of the hydrodynamic and breach evolution algorithms; second, published flume experimental data of non-cohesive earth dam overtopping failures are adopted to evaluate the model accuracy in predicting breach hydrographs and spatiotemporal evolution of breach geometry. The results demonstrate that the proposed model accurately reproduces the key characteristics of overtopping failure with high fidelity. The predicted breach flow rates and flow depths are in excellent agreement with experimental observations, with relative errors less than 5% for both peak discharge and time to peak. Consequently, this study provides a reliable numerical tool for detailed simulation of non-cohesive earth dam breaches and offers scientific support for emergency management. Full article

To address the limitations of insufficient feature representation, large model size, and high deployment cost in network traffic classification, a lightweight classification framework based on multi-teacher knowledge distillation is proposed. The framework consists of two heterogeneous teacher networks and a compact student network to enable end-to-end traffic classification under constrained computational resources. The teacher networks incorporate complementary spatio-temporal modeling strategies, including a bidirectional temporal convolutional network (BiTCN) enhanced with attention mechanisms and convolutional neural network (CNN), and a parallel spatio-temporal fusion architecture integrating bidirectional long short-term memory (BiLSTM) and CNN. Knowledge from the teacher ensemble is distilled into a lightweight CNN-based student network through soft-target supervision, leading to improved generalization capability with significantly reduced model complexity. Experimental results demonstrate that effective knowledge transfer is achieved while reducing model parameters by more than 80%, and performance gains of about 1–3% are obtained compared with baseline methods. These results indicate strong potential for practical deployment in resource-constrained network environments. Full article

Millimeter-wave radar has gradually gained popularity as a sensor mode for Human Activity Recognition (HAR) in recent years because it preserves the privacy of individuals and is resistant to environmental conditions. Nevertheless, the fast inference of high-dimensional and sparse 4D radar data is still difficult to perform on low-resource edge devices. Current models, including 3D Convolutional Neural Networks and Transformer-based models, are frequently plagued by extensive parameter overhead or quadratic computational complexity, which restricts their applicability to edge applications. The present paper attempts to resolve these issues by introducing RadarSSM as a lightweight spatiotemporal hybrid network in the context of radar-based HAR. The explicit separation of spatial feature extraction and temporal dependency modeling helps RadarSSM decrease the overall complexity of computation significantly. Specifically, a spatial encoder based on depthwise separable 3D convolutions is designed to efficiently capture fine-grained geometric and motion features from voxelized radar data. For temporal modeling, a bidirectional State Space Model is introduced to capture long-range temporal dependencies with linear time complexity $O\left(T\right)$, thereby avoiding the quadratic cost associated with self-attention mechanisms. Extensive experiments conducted on public radar HAR datasets demonstrate that RadarSSM achieves accuracy competitive with state-of-the-art methods while substantially reducing parameter count and computational cost relative to representative convolutional baselines. These results validate the effectiveness of RadarSSM and highlight its suitability for efficient radar sensing on edge hardware. Full article

The world evolves continuously across space and time. Massive volumes of data are generated through sensing, simulation, remote observation, and human activities, capturing dynamic processes in environmental, social, economic, and engineered systems. Critical insights are embedded within these large-scale spatiotemporal datasets. Spatiotemporal Data Science provides a conceptual and methodological framework for analyzing such data by integrating spatiotemporal thinking, computational infrastructure, artificial intelligence, and domain knowledge. The field advances methods for data acquisition, harmonization, modeling, visualization, and decision support, enabling applications in natural disaster response, public health, climate adaptation, infrastructure resilience, and geopolitical analysis. By leveraging emerging technologies—including generative Artificial Intelligence (AI), large-scale cloud platforms, Graphics Processing Unit (GPU) acceleration, and digital twin systems—Spatiotemporal Data Science enables scalable, interoperable, and solution-oriented research and innovation. It represents a critical frontier for scientific discovery, engineering advancement, technological innovation, education, and societal benefit. Spatiotemporal Data Science is a transdisciplinary field that studies and models dynamic phenomena across space and time by integrating spatial theory, temporal reasoning, artificial intelligence, and scalable computational infrastructure. It enables the development of adaptive, predictive, and increasingly autonomous systems for understanding and managing complex real-world processes. Full article

Water planning in semi-arid regions depends on climate information that resolves both seasonal timing and topographic gradients. This study evaluated 15 CMIP6 models over Zacatecas, Mexico, and produced a 1 km historical dataset for 1985–2014 by statistically refining bias-corrected daily fields from NEX-GDDP-CMIP6. Downscaling was referenced to the CHELSA climatology: temperature was refined using an elevation-informed hybrid spline approach, whereas rainfall was downscaled with geographically weighted regression (GWR) to represent orographic gradients. The resulting fields were assessed against two independent observational baselines: an automated INIFAP network (2004–2014) and a conventional CONAGUA network (1985–2014). For temperature, BCC-CSM2-MR showed the highest performance, with a Pearson correlation of R = 0.94 for both T_max and T_min. A consistent network-dependent bias pattern was identified: the downscaled models overestimated the diurnal temperature range relative to INIFAP but underestimated it relative to CONAGUA, highlighting the influence of instrumentation and observational protocols on model evaluation. For rainfall, ACCESS-ESM1-5 reproduced the seasonal cycle and dominant orographic patterns, with a correlation of R = 0.611 despite the intrinsic stochasticity of semi-arid rainfall. The resulting 1 km fields provide a spatially consistent baseline for regional applications, including stochastic weather generation and impact models in complex semi-arid regions. Full article

Ocean temperature and salinity structures are crucial in understanding ocean circulation and heat–salt transport processes. However, the high cost and limited spatiotemporal coverage of in situ observations make it difficult to reconstruct high-resolution three-dimensional temperature–salinity (T-S) fields. To address these limitations and the strong spatiotemporal heterogeneity of T-S structures in the South China Sea (SCS), the Smart Synthetic Ocean Profile (SSOP) model is proposed, which is a two-stage machine learning-based inversion framework for reconstructing subsurface T-S profiles from satellite surface data. The framework integrates localized training, adaptive model selection, and an error correction strategy. Using climate-state grids with a consistent spatiotemporal resolution as a baseline, multiple candidate regression models are independently trained for each grid point–depth layer–month combination, and the optimal model is selected through performance validation to generate initial T-S profiles. An error correction module is then introduced to refine temperature profile deviations, improving profile consistency and overall accuracy. Experiments using three independent observational periods from the SCS show that SSOP reliably reconstructs vertical T-S structures, particularly in the upper ocean and thermocline. Comparisons with in situ observations indicate that SSOP achieves improved accuracy relative to the Modular Ocean Data Assimilation System and climatology. Full article

Climate change poses significant challenges to global agricultural systems, exerting profound impacts on crop yields and water resource management, which are particularly pronounced in cold-region rice-growing systems. This study employed the AquaCrop model to assess the spatiotemporal distribution characteristics of rice yield, crop water requirement (ET_c), irrigation water requirement (Ir), and water use efficiency (WUE) in Heilongjiang Province under the RCP4.5 and RCP8.5 scenarios from 2021 to 2080. The results indicate that the average rice yield in Heilongjiang Province will increase by approximately 2% to 5%, with more significant gains observed in the western and southern regions. However, climate warming will cause ET_c to increase by 3–7%, leading to a rise in Ir of about 5–12%, which is particularly pronounced under the RCP8.5 scenario. Compared to RCP4.5, the yield under RCP8.5 will increase by 1–2%, but the increase in Ir will be more significant. Despite these changes, WUE remains within a relatively constrained range (approximately 1.55~1.75 kg·m⁻³), as yield increases are largely offset by corresponding rises in ET_c. Overall, the findings reveal a pronounced yield–water trade-off in cold-region rice systems under future climate scenarios, indicating that yield gains may be accompanied by heightened Ir. Full article

Forest soil conservation is pivotal for controlling soil erosion and ensuring ecological security. Taking the Qilian Mountains Area in China as the research region, this study used ArcMap 10.8 software to process data for six prefecture-level cities in the area from 2008 to 2023. The Revised Universal Soil Loss Equation (RUSLE) model was applied to quantify the forest soil conservation amount and evaluate its ecosystem service value (ESV). Their spatiotemporal variations and dynamic evolution patterns were analyzed, alongside the influence of soil organic matter (OM) and nitrogen (N), phosphorus (P), and potassium (K) contents. The results showed that the average contents of OM, N, P and K in the forest soils of the Qilian Mountains Area were 24.22 g·kg⁻¹, 1.54 g·kg⁻¹, 0.70 g·kg⁻¹, and 19.96 g·kg⁻¹, respectively, with significant regional heterogeneity. Haibei Tibetan Autonomous Prefecture (HBTAP) had the highest while Jinchang City (JC) had the lowest. From 2008 to 2023, the average annual forest soil conservation amount and its ESV of the region were 1.749 × 10⁹ tons and 2.0444 × 10¹⁰ yuan, respectively, both showing a fluctuating trend of initial increase followed by a decrease. Spatially, HBTAP ranked first in average annual forest soil conservation amount per unit area and ESV. Jiuquan City (JQ) had the lowest forest soil conservation amount per unit area, and JC the lowest ESV. Forest soil conservation and its ESV in the region were affected by the contents of soil nutrients (OM and N, P, K elements), vegetation types and quality, topography, climate, and human activities (including ecological governance), which collectively intensified the spatiotemporal heterogeneity. These findings provide a theoretical basis for precise regional ecological protection and differentiated restoration strategies in arid regions. Full article

The plain river network region is faced with ecological and environmental challenges such as insufficient hydrological connectivity and degradation of ecosystem services under the influence of urbanization and human activities, and therefore attention needs to be paid to river network changes in this region and the synergistic benefits of natural–social–economic multidimensional factors. This study took the Lixiahe region, a typical plain river network region, as the research object, using Mann–Kendall, spatial autocorrelation analysis, random forest, multiple validation and Granger causality test of key drivers to analyze the spatiotemporal evolution of its river network from 2013 to 2025 and quantify driving mechanisms from natural, social and economic factors. The results showed that: (1) From 2013 to 2025, the Lixiahe Plain river network region tended to be trunk and artificial, with the number and connectivity of river networks showing an upward trend while the curvature of river network decreased significantly. (2) The Global Moran’s I index of the Lixiahe Plain river network decreased from 0.612 to 0.534, indicating a continued weakening of spatial agglomeration in the water area and exhibiting characteristics of edge fragmentation. (3) Random forest analysis showed that socioeconomic factors dominated recent river network change in the Lixiahe Plain. Economic factors mainly influenced quantity-related indicators, while social factors were more important for meander degree and connectivity in several ecologically sensitive counties. Multilevel validation demonstrated the robustness and generalization ability of the model. Granger causality analysis further indicated that GDP, road network density, freshwater aquaculture area, and agricultural output statistically preceded changes in key hydrological indicators. These findings suggest that river network management in plain river network regions should move beyond quantity-based engineering expansion and adopt a multi-indicator, spatially differentiated approach. Integrating river quantity, morphology, and connectivity into management can better support the balance between socioeconomic development and ecological protection and promote the sustainable optimization of river network. Full article

Effective wildfire management demands seamless integration of real-time detection and long-term spread forecasting. This paper proposes a novel power-efficient UAV–satellite hybrid pipeline that synergizes the agility of UAVs with the scale of satellite intelligence. The system begins with a dashboard-guided, multi-UAV detection module that scores fire likelihood from historical satellite data and enables scalable, energy-efficient deployment with low-latency onboard processing. This aerial component ensures persistent surveillance and reliable ignition detection, supported by a Dual LoRa (Long Range) communication scheme for robust and low-power connectivity. It achieves an F1-score of 97.4% while minimizing power consumption to extend operational flight times. Following detection, the pipeline transitions to a dynamic perimeter-prediction phase utilizing a custom Canadian boreal dataset. We employ a Squeeze-and-Excitation Residual U-Net (SE-ResUNet) to model spatiotemporal fire propagation based on static terrain and dynamic environmental features. The model was validated using a dynamic simulation framework that evaluates temporal consistency and convergence behavior against final cumulative burned-area masks, effectively addressing the absence of daily ground truth. Under these conditions, the model achieves a recall of 84% and an AUC of 0.97, demonstrating a strong capability to delineate active fire fronts. By coupling dashboard-driven UAV sensing with satellite-based predictive modeling, this work establishes a modular, foundational framework to support data-scarce forecasting in modern wildfire management. Full article