Search Results (611)

Within the subway systems of megacities, individual passenger transfer experiences have long been marginalized due to an overemphasis on macro-level, systemic, and functional performance, positioning low transfer efficiency as a pervasive bottleneck in enhancing the overall network efficacy. Adopting an individual passenger perspective, this study takes the Changping Line of the Beijing Subway as an empirical case. By using walking speed to evaluate transfer efficiency and through field survey, behavioral experiment, and quantitative model analysis, this paper reveals the key built environment factors influencing transfer efficiency and their underlying impact mechanisms and also provides empirical evidence for the synergistic optimization of transfer efficiency and the built environment in megacity subway systems. The findings indicate that the built environment impacts transfer efficiency in macro-non-linear and micro-linear ways, specifically manifesting across six specific mechanisms: the geographic location mechanism, the pressure mechanism of high-density development, the spatial exclusivity mechanism of regional transport hubs, the topological penalty mechanism of transfer paths, the bottleneck constraint mechanism of node facilities, and the compensatory mechanism of information guidance. Furthermore, as a medium affecting transfer efficiency, the shaping of the built environment is essentially determined by the city’s subway planning and construction institutions, the external technical conditions of the particular stations, and localized tactical governance to manage the dynamic daily traffic mobility. Based on these findings, this study suggests that improving the transfer efficiency of megacity metro systems like the Changping Line should implement systemic strategies from four aspects: tailored TOD at the macro-spatial planning phase, the micro-spatial integration of indoor and outdoor built environments during the station design phase, differentiated collaborative governance to alleviate station-external intermodal transfer competitions during the operation phase, and digitally empowered transfer guidance to proactively manage transfer demand across three scenarios. Full article

Large language models (LLMs) built upon the Transformer architecture have achieved remarkable performance in natural language understanding, text generation and logical reasoning, while their internal working mechanisms remain poorly interpreted. This paper establishes a systematic mathematical analysis framework tailored for decoder-only Transformer LLMs, based on linear algebra, tensor analysis, probability theory, information theory, optimization dynamics and geometric deep learning. We conduct rigorous mathematical modeling and theoretical deduction on core modules including word embedding, position encoding, self-attention, feed-forward networks, training optimization and generalization reasoning, and explore the mathematical nature of semantic representation, contextual correlation, knowledge storage and logical inference within models. In this paper, we strictly distinguish between classic established Transformer theories and our original mathematical derivations and conclusions. Distinct from existing fragmented theoretical studies, this work presents six targeted novel contributions beyond conventional Transformer theories: (1) we construct the first full-process unified mathematical framework covering all core modules and the entire lifecycle of Transformer-based LLMs; (2) we provide strict mathematical proof to verify that single-head self-attention is essentially a kernel weighted average operation in reproducing kernel Hilbert space and derive the low-rank and sparse properties of attention weights; (3) we establish a high-dimensional non-convex optimization dynamics model for pre-training and mathematically prove that model training converges to flat local minima; (4) we derive a tighter upper bound of generalization error and quantify the quantitative relationship among model parameters, sequence length, training data scale and generalization performance; (5) we characterize the latent space as a low-curvature smooth Riemannian manifold and model logical reasoning as geometric transformation on this manifold; (6) we design multi-group controlled experiments on mainstream datasets to quantitatively validate all above theoretical conclusions. This paper further summarizes the inherent mathematical limitations of current Transformer LLMs and proposes feasible theoretical optimization paths, referring to state-of-the-art research published from 2021 to 2026. The outcomes of this research can provide solid mathematical theoretical support for improving model interpretability, optimizing network structures and boosting practical performance, and facilitate the transition of LLM research from empirical engineering practice to theory-driven development. Full article

Globally, wildfires are increasingly threatening forest ecosystems and human well-being, requiring proactive management strategies. Integrating wildfire mitigation with bioenergy production presents a dual opportunity to reduce fire risk while contributing to clean energy. This study builds upon previous work by incorporating updated annual heat load estimates from 32 off-grid communities in northern Canada to assess the amount of biomass at risk of wildfire that could be mobilized to meet local bioenergy needs. Our results reveal that energy consumption in the remote communities considered was previously significantly underestimated, with an average of 11,710 MWh per year, and a minimum and maximum of 1869 and 43,867 MWh per year, respectively. With the updated dataset, which includes both space heating and electricity energy usage, the average energy demand is approximately 300% higher than earlier estimates. Despite this substantial increase in energy consumption, the amount of biomass needed to meet local energy demand per year ranges from 352 to 8276 odt per year, representing only a small fraction (approximately 1.67% on average) of the total biomass identified as being at risk within a 10 km buffer. This corresponds to fuel treatment areas ranging from 4 to 222 hectares per year (around 51 ha on average), depending on the community. The results presented here, based on updated energy data, provide important insights into the operational feasibility of this approach. To be successful, implementation will require strong community leadership and collaboration with fire management agencies to design consistent and cost-effective fuel treatment strategies that are tailored to each community’s environmental conditions and energy needs. Full article

Renal cell carcinoma (RCC) accounts for the majority of kidney cancers, with approximately 80,000 new diagnoses and over 14,000 deaths annually in the United States. Risk stratification is essential for prognostication, treatment selection, and clinical trial design across all disease stages. In localized and locally advanced RCC, pathological stage, histology, and grade remain the primary prognostic factors, while the International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) criteria serve as the standard risk stratification tool in the metastatic setting. However, current models rely predominantly on clinical and pathologic variables that act as indirect surrogates of tumor biology and do not account for the molecular heterogeneity inherent to RCC. This narrative review synthesizes and compares established and emerging risk stratification and prognostic models across all stages of RCC. Established models such as the IMDC criteria and the stage, size, grade, and necrosis (SSIGN) score demonstrate robust prognostic performance but are limited by their reliance on clinical and pathologic variables alone. Emerging biomarkers—including circulating tumor DNA, methylated DNA, artificial intelligence-based radiomics, and tissue-based molecular signatures—show promise for improving risk discrimination. The molecular heterogeneity of RCC underscores an urgent need for integrated molecular–clinical–pathologic prognostic tools tailored to specific histologic subtypes to enable more precise, individualized care. Full article

Modern agriculture in developing regions faces significant challenges due to labor scarcity and the health hazards associated with the manual application of chemical treatments. This study presents the design, development, and techno-economic evaluation of an experimental hexacopter unmanned ariel vehicle (UAV) platform specifically tailored for crop protection on fragmented, smallholder farmlands. The research aims to bridge the gap between expensive imported technology and the practical needs of small-scale farmers by providing a cost-effective, locally manufacturable solution. The methodology involved the integration of a modular spraying system and optimized control architecture into a high-stability hexacopter frame. Experimental evaluations focused on flight stability, payload capacity, and spray uniformity using water-sensitive media. The results indicate that the developed platform achieves high coverage efficiency while significantly reducing chemical waste compared to traditional manual methods. Furthermore, the economic analysis suggests that the operational costs are substantially lower than those of comparable imported systems, offering a favorable payback period within a few crop seasons. These findings demonstrate that an indigenous UAV spraying platform can enhance both operational safety and economic feasibility for smallholder agriculture. Full article

This paper proposes an intelligent time-series early warning framework based on a production LSTM–Transformer network for petrochemical refining processes. A cascaded encoder–decoder architecture is designed, where the LSTM extracts local temporal patterns and medium-term memory from noisy industrial data, while the Transformer models global dependencies and cross-unit interactions via multi-head self-attention. An adaptive feature fusion layer bridges the representational gap between the two networks. A multi-stage preprocessing pipeline tailored for refining MES data handles missing values, outliers, and mixed operating conditions. Using 120 variables from five units of a fluid catalytic cracking unit, the framework predicts the regenerator bed temperature up to 8 h (48 steps) ahead. Comparative experiments show that the production LSTM–Transformer achieves a mean MAE of 0.088, a mean RMSE of 0.113, and the lowest median MAPE of 19.91% among all models, outperforming standalone LSTM (MAE 0.095, MAPE 20.85%) and Transformer (MAE 0.088, MAPE 20.49%). Robustness analysis confirms stable performance under strong noise (down to 5 dB) and missing rates up to 50%, with a median MAE of 0.1027 across tags. This work provides an effective, end-to-end predictive early warning solution that balances accuracy, production importance coverage, and industrial robustness, offering a generalizable data-driven paradigm for process industries. Full article

Population aging has become a major demographic process in modern societies, with its course varying considerably across space. This study examined the scale and dynamics of population aging across Poland’s voivodeships in 2015–2024 and identified its regional patterns. The analysis used data from Statistics Poland’s Local Data Bank for 16 voivodeships and included indicators capturing age composition, demographic dependency, and fertility. The analysis was conducted for 16 Polish voivodeships using data from Statistics Poland’s Local Data Bank for 2015–2024 and indicators describing age structure, demographic dependency, and fertility. An analysis of changes in indicator values over time and Kohonen self-organizing maps (SOM) were applied in two model variants, differing in the measure of population aging adopted. To ensure a consistent direction of interpretation, the variables were appropriately transformed and then standardized. The results indicate spatial variation in the level of population aging and differing dynamics of change during the study period. Four regional profiles were identified, reflecting different patterns of indicators describing age structure, demographic burden, and fertility. Kohonen self-organizing maps were used as an exploratory tool to organize voivodeships according to the similarity of their demographic profiles and to describe changes in their profile assignment over time. From the perspective of sustainability, the identified profiles make it possible to capture territorially differentiated demographic conditions that may be relevant to healthcare, long-term care, regional labor markets, social services, and family policy. The results may support sustainable regional development by providing a basis for designing public policy tailored to the specific characteristics of individual voivodeships. Thus, the study links a multidimensional typology of demographic aging with the need for socially sustainable regional policy. The results suggest that SOM can serve as a useful exploratory tool for visualizing and classifying regional demographic aging profiles. Full article

Accurate and robust detection of leaf diseases is a key enabler for precision agriculture and large-scale crop health monitoring. Despite the strong generalization of modern one-stage detectors (e.g., YOLOv8), two domain-specific challenges remain: (i) weak or blurry lesion boundaries hinder precise localization, and (ii) low color contrast between diseased and healthy tissues forces models to rely on subtle texture patterns rather than salient shapes. To tackle these challenges, we reframe the core agricultural disease detection task as the identification of “asymmetric morphological anomalies” and propose a domain-tailored enhancement framework. First, we introduce an Edge Enhancement Module (EEM) that explicitly strengthens boundary-aware representations. Inspired by the natural symmetry of healthy leaves, our EEM is specifically designed to capture symmetry-breaking boundary discontinuities and localized asymmetric edges caused by disease lesions. Our method enhances edge and texture cues that are indicative of disease lesions, which often exhibit local asymmetries and boundary discontinuities. The EEM includes a Differential Normalized Pooling Block (DNPB) that highlights edge responses through discrepancies between max pooling and average pooling, which also models cross-group edge correlations. Second, the Lightweight Texture-Sensitive Feature Enhancement (LTSFE) mechanism amplifies texture-discriminative channels under low-contrast conditions by leveraging complementary global statistics and efficient channel mixing, all with negligible computational overhead. We evaluated our method on a self-constructed dataset of 106,434 images with 225,640 annotations covering diverse crops. Experiments show that the proposed method achieves state-of-the-art accuracy (81.54% mAP@0.5:0.95) while maintaining real-time inference (142 FPS), consistently outperforming strong baselines. Ablations confirm the effectiveness and complementarity of EEM and LTSFE, demonstrating that domain-specific architectural design, inspired by biological symmetry, can substantially improve agricultural vision systems. Full article

Unmanned Aerial Vehicle (UAV) image object detection holds significant application value in the low-altitude economy, traffic monitoring, intelligent agriculture, and disaster rescue. However, due to the top-down perspective, UAV images typically suffer from challenges such as small target scales, dense object distribution, severe occlusions, and complex backgrounds. These issues often limit the recall and localization accuracy of general-purpose detectors when they are directly applied to UAV small-object detection scenarios. To address these aforementioned challenges, this paper proposes an Adaptive Dynamic Aggregation YOLO network, termed ADA-YOLO. The novelty of ADA-YOLO lies in its highly efficient combinatorial design specifically tailored for UAV small object detection, while retaining the efficient backbone of YOLOv8, we systematically reconstruct the neck and detection head to improve accuracy. Specifically, a high-resolution P2 detection branch is incorporated to construct a P2–P5 multi-scale prediction structure. Furthermore, the lightweight DySample dynamic upsampling module is adopted to replace traditional upsampling methods, and an Adaptive Spatial Feature Fusion (ASFF) mechanism is introduced to alleviate semantic conflicts and noise interference during multi-scale feature fusion. This synergistic combination explicitly addresses multi-scale representation challenges and enhances small-object detection performance in complex scenes. Comparative experiments with the baseline YOLOv8n on the VisDrone2019 dataset demonstrate that ADA-YOLO achieves an improvement of 11.3% in mAP@0.5 and 8.2% in mAP@0.5:0.95. The improved model achieves these performance gains with a modest parameter increase and acceptable computational complexity. Finally, ablation experiments further validate the effectiveness of each individual module and their synergistic gains. Full article

Solar-driven interfacial evaporation (SIE) has emerged as a transformative, off-grid technology that confines heat at the air–liquid interface, enabling high-efficiency vapor generation for decentralized water purification. Here, we present a comprehensive and critical review of the field, tracing its evolution from fundamental photothermal principles to integrated multifunctional systems. We first elucidate the thermodynamics of interfacial heat localization and the resultant enhancement in evaporation efficiency. We then systematically analyze material innovation strategies—including broadband-absorbing photothermal agents and tailored evaporator architectures—designed to overcome persistent challenges such as salt crystallization, fouling, and thermal losses. Moving beyond freshwater production, we highlight emerging pathways for extending SIE platforms toward water–energy cogeneration, selective resource recovery, and zero-liquid-discharge wastewater treatment. We further identify and objectively assess the key bottlenecks that currently hinder the transition from laboratory-scale prototypes to real-world deployment, with a focus on long-term material robustness under harsh environments, adaptability to fluctuating water chemistries, and techno-economic viability. Finally, we outline forward-looking research directions, including stimulus-responsive smart evaporators, elucidation of multi-field coupling mechanisms, and the establishment of standardized performance evaluation protocols. This review aims to provide both a tutorial for newcomers and a critical assessment for experienced researchers, offering a balanced perspective on the current state-of-the-art and a roadmap for translating SIE from academic research into sustainable, impactful technologies. Full article

Indonesia, located along the equatorial belt, has consistently high solar irradiance, offering strong potential for renewable energy development. However, limited availability of high-resolution solar radiation data constrains accurate system design. This study aims to evaluate eight empirical models for estimating hourly solar radiation under tropical conditions and identify the most suitable approach for data-scarce regions. The novelty lies in a comparative assessment tailored to Indonesia’s tropical climate and its application to sustainable energy planning. Model performance is assessed using MBE, RMSE, and R² against measured data. The results identify the most accurate model, which serves as the basis for developing a modified model that better represents local atmospheric characteristics. The proposed model improves estimation accuracy and supports more reliable solar resource assessment for sustainable energy applications in tropical regions. These findings support improved solar resource assessment and contribute to more reliable and sustainable solar energy system development in tropical regions. Full article

Enhancing oil and gas production performance is essential for maintaining the economic sustainability of petroleum enterprises and meeting the increasing global energy requirements. In this context, subsurface production optimization constitutes a fundamental component of strategic reservoir management, directly affecting critical decisions such as well location design and the regulation of operational parameters. Nevertheless, conventional reservoir optimization approaches are frequently constrained by high computational costs and limited optimization effectiveness. To overcome these limitations, evolutionary algorithms have gained considerable attention for addressing complex optimization tasks, owing to their gradient-free nature and strong capability for parallel exploration. This paper proposes a chaotic exploitation orthogonal learning fruit fly optimization algorithm (COFOA) tailored for global optimization and oil and gas production optimization. Specifically, we integrate a chaotic exploitation mechanism and an orthogonal learning strategy to improve the balance between exploration and exploitation. Following the population update in FOA, the chaotic exploitation mechanism is first applied to help the population escape local optima and enhance search efficiency. Subsequently, the orthogonal learning strategy is employed to strengthen the algorithm’s exploitation capability. To evaluate the performance of the improved FOA, extensive experiments were conducted on benchmark functions from IEEE CEC 2017 and IEEE CEC 2022, including ablation studies, scalability tests and comparisons with state-of-the-art algorithms. The results demonstrate that the proposed FOA significantly outperforms competing algorithms in optimizing reservoir production. COFOA demonstrates consistent performance superiority over all compared algorithms in terms of mean NPV. Specifically, it achieves improvements of approximately 2.35% to 16.23% compared with existing methods. Notably, COFOA outperforms strong competitors such as mSCA and BLPSO by 2.35% and 3.81%, respectively, while achieving more significant gains over algorithms such as SCADE (15.31%) and CCMSCSA (16.23%). Even when compared with relatively competitive methods like HGWO and CCMWOA, COFOA still maintains performance improvements of 4.79% and 6.12%, respectively. These results clearly demonstrate the superior optimization capability of COFOA in terms of maximizing NPV under complex reservoir conditions. Full article

Single-photon Light Detection and Ranging (LiDAR), which is capable of detecting single-photon signals, has developed rapidly in the field of long-range imaging. Due to the long detection range and limited laser power, the accumulated signal photons of single-photon LiDAR are extremely sparse. Meanwhile, the dark current counts, backscattering noise, and background noise of the single-photon detector are significant, resulting in an extremely low signal-background ratio of the detection data. However, existing algorithms struggle to accomplish the depth reconstruction on data with extremely low signal-to-background ratio (SBR). To address the challenges of complex spatiotemporal correlation and feature sparsity in long-range single-photon imaging depth reconstruction, we design a deep reconstruction algorithm based on a classification formulation, specifically tailored for single-echo detection scenarios. We propose a wavelet denoising preprocessing module and a four-dimensional attention module to learn the spatiotemporal correlations of the photon-counting cube data. Sawtooth-arranged dilated convolutions are utilized during the pixel-wise denoising process to extract sparse features, and non-local total variation regularization combined with cross-entropy is introduced as a joint loss function. For depth reconstruction of data with an SBR of 1:100, the root-mean-square error is less than 0.022 m, which is 66.72% lower than that of the best baseline algorithm. It also achieves promising depth reconstruction results on data with an SBR of 1:300. Full article

Sensor-based human activity recognition (HAR) plays a fundamental role in healthcare monitoring, sports analytics, and ambient-assisted living. Although deep learning has substantially advanced HAR performance, two practical issues still limit its real-world deployment: (i) the distribution shift caused by changes in users or sensor placements can degrade generalization, and (ii) the quadratic $\mathcal{O}\left(L^2\right)$ complexity of standard self-attention hinders efficient long-sequence modeling on resource-constrained wearable devices. To address these issues, we propose DSHformer, which is an accuracy-oriented HAR framework that combines compact channel–temporal encoding with locality-sensitive hashing (LSH)-based attention. Specifically, DSHformer (i) employs a low-parameter patch-based graph-attention encoder to jointly model latent relationships among sensor channel–temporal dynamics; (ii) introduces a trainable prototype pool together with a multi-layer decomposition network to improve intra-class compactness and inter-class separability via prototype alignment; and (iii) introduces a decomposition-stable LSH-based attention mechanism tailored for HAR, whose core design couples prototype-guided feature decomposition with locality-sensitive hashing to ensure that semantically related tokens remain consistently grouped in the same hash bucket even after decomposition-induced attenuation. The mechanism thereby operates at $\mathcal{O}(LlogL)$ attention complexity on longer sensor sequences. Extensive experiments on five public benchmarks (WISDM, UCI-HAR, PAMAP2, Opportunity, and UniMiB-SHAR) show that DSHformer achieves accuracies of 98.6%, 93.7%, 98.4%, 88.5%, and 96.6%, respectively, achieving competitive or superior performance compared with both Transformer variants and HAR-specific baselines under the adopted benchmark protocols. Ablation studies further confirm the complementary contribution of each component. Full article

Tool wear significantly impacts the productivity and efficiency of sheet metal stamping operations, particularly in high-volume progressive die applications. This study presents an iterative calibration framework for Archard’s wear model, tailored to industrial stamping processes. The proposed methodology integrates finite element simulations with experimentally measured wear data obtained from production components, enabling data-driven calibration of the wear coefficient $K_{sim}$. The framework achieves high predictive accuracy, with deviations of 1.4–3.7% between simulated and optically measured wear depths and localization, after more than 15 million strokes. Rapid convergence is obtained within two to three calibration cycles, significantly reducing computational effort while maintaining physical fidelity. The simulation setup incorporates detailed modelling of contact pressure, sliding velocity, and stress distribution, validated using optical surface measurement systems and coordinate-based metrology. Beyond the specific industrial case, the framework demonstrates transferability to other sheet metal forming processes, such as bending, blanking, and coining, by leveraging physically based parameter adaptation across comparable pressure–velocity regimes. The approach enables predictive wear modeling in data-scarce environments and supports early-stage tool design workflows. Overall, the proposed methodology bridges the gap between empirical calibration and generalized simulation, contributing both methodological rigour and practical applicability to manufacturing science. Full article