Search Results (10,007)

This paper presents a dual-structured convolutional neural network (CNN) for image classification, which integrates two parallel branches: CNN-A with spatial attention and CNN-B with channel attention. The spatial attention module in CNN-A dynamically emphasizes discriminative regions by aggregating channel-wise information, while the channel attention mechanism in CNN-B adaptively recalibrates feature channel importance. The extracted features from both branches are fused through concatenation, enhancing the model’s representational capacity by capturing complementary spatial and channel-wise dependencies. Extensive experiments on a 12-class image dataset demonstrate the superiority of the proposed model over state-of-the-art methods, achieving 98.06% accuracy, 96.00% precision, and 98.01% F1-score. Despite a marginally longer training time, the model exhibits robust convergence and generalization, as evidenced by stable loss curves and high per-class recognition rates (>90%). The results validate the efficacy of dual attention mechanisms in improving feature discrimination for complex image classification tasks. Full article

Stable adhesion on non-magnetic, steep, and irregular underwater surfaces (e.g., concrete dams with cracks or biofilms) remains a challenge for inspection robots. This study develops a novel adsorption mechanism based on the synergistic operation of a Venturi-ejector and a composite suction cup. The mechanism utilizes the Venturi effect to generate stable negative pressure via hydrodynamic entrainment and innovatively adopts a composite suction cup—comprising a rigid base and a dual-layer EPDM sponge (closed-cell + open-cell)—to achieve adaptive sealing, thereby reliably applying the efficient negative-pressure generation capability to rough underwater surfaces. Theoretical modeling established the quantitative relationship between adsorption force (F) and key parameters (nozzle/throat diameters, suction cup radius). CFD simulations revealed optimal adsorption at a nozzle diameter of 4.4 mm and throat diameter of 5.8 mm, achieving a peak simulated F of 520 N. Experiments demonstrated a maximum F of 417.9 N at 88.9 W power. The composite seal significantly reduced leakage on high-roughness surfaces (Ra ≥ 6 mm) compared to single-layer designs. Integrated into an inspection robot, the system provided stable adhesion (>600 N per single adsorption device) on vertical walls and reliable operation under real-world conditions at Balnetan Dam, enabling mechanical-arm-assisted maintenance. Full article

This paper presents the application of conic programming methods to the limit analysis of plane rigid-plastic problems in structural and geotechnical engineering. The approach is based on the formulation of yield criteria as second-order cone constraints and on the dual optimization problem, which directly provides collapse mechanisms and limit loads. Two benchmark examples are investigated. The first concerns a deep beam under uniform top pressure, analyzed with linear and quadratic finite elements. The results confirm the ability of the method to reproduce realistic collapse mechanisms and demonstrate the effect of mesh refinement and element type on convergence. The second example addresses the ultimate bearing capacity of a strip footing on cohesive-frictional soil. The numerical implementation was carried out in MATLAB using CVX with MOSEK as the solver, which ensures practical applicability and efficient computations. Different soil models are considered, including Mohr–Coulomb and two Drucker–Prager variants, and the results are compared with the classical Terzaghi solution. Additional elastoplastic FEM simulations carried out in a commercial program are also presented. The comparison highlights the differences between rigid-plastic optimization and incremental elastoplastic analyses, showing that both conservative and liberal estimates of bearing capacity can be obtained. The study shows that conic programming is an efficient and flexible framework for limit analysis of plane rigid-plastic problems, providing engineers with complementary tools for assessing ultimate loads, while also ensuring good computational efficiency. Full article

Introduction: In minimally invasive, off-clamp partial nephrectomy (ocMIPN), the reconstructive strategy profoundly influences functional outcomes. Traditional nephrometry scores aid preoperative planning but do not directly inform the choice of closure technique. This dual-institutional study aimed primarily to identify preoperative CT-derived parameters predictive of renorrhaphy versus a sutureless approach, and secondarily to compare perioperative and functional outcomes between these techniques. Methods: We retrospectively analyzed 201 consecutive ocMIPN cases performed using a standardized off-clamp technique by two experienced surgical teams across robotic platforms and conventional laparoscopy. Preoperative CT scans were centrally reviewed to quantify morphometric features, including contact surface area (CSA), tumor radius, and Gerota’s fascia thickness. Univariable and multivariable logistic regression models—one restricted to radiologic variables and one expanded with RENAL score terms—were generated to identify independent predictors. Perioperative outcomes, renal functional metrics, and Trifecta rates were compared between cohorts. Results: Among the 201 patients, 101 (50.2%) underwent sutureless reconstruction and 100 (49.8%) renorrhaphy. Cohorts were comparable at baseline except for tumor size (3.1 vs. 3.6 cm; p = 0.04). In multivariable analysis, CSA > 15 cm² (OR 3.93; 95% CI 1.26–12.26; p = 0.02) and tumor radius (OR 1.14 per mm; 95% CI 1.01–1.29; p = 0.04) consistently predicted renorrhaphy, while Gerota’s fascia < 10 mm emerged as significant only in the expanded specification (OR 0.08; 95% CI 0.01–0.70; p = 0.02). Integration with RENAL improved predictive performance (ΔAUC 0.06; NRI 0.14; IDI 0.07), and the final model demonstrated strong discrimination (AUC 0.81) with satisfactory calibration. Perioperative outcomes, postoperative renal function, and Trifecta achievement were similar between groups (all p ≥ 0.21). Conclusions: A concise set of CT-derived morphologic markers—CSA, tumor radius, and perinephric fascia thickness—anticipated reconstructive strategy in ocMIPN and augmented the discriminatory power of RENAL nephrometry. When anatomy was favorable, sutureless repair was not associated with statistically significant differences in perioperative safety or renal function, although the study was not powered for formal equivalence testing. These findings support the integration of radiologic markers into preoperative planning frameworks for nephron-sparing surgery. Full article

Amid ongoing educational reforms and the rapid advancement of the knowledge economy, innovative teaching behaviors are not only closely related to teachers’ professional growth and students’ academic achievement but are also regarded as the key driving force for the evolution of the educational system. Consequently, identifying effective ways to foster teachers’ innovative teaching behaviors has become a central concern in educational psychology and management. Grounded in the Job Demands–Resources framework, this study developed and tested a chained mediation model using survey data from 1163 Chinese elementary and secondary school teachers. The model examines how teachers’ emotional competence fosters innovative teaching behaviors and elucidates the underlying mechanisms. The results revealed that (1) emotional competence significantly and positively predicted innovative teaching behaviors, and (2) teaching efficacy and work vitality served not only as independent mediators but also as sequential mediators in this relationship. These findings extend the understanding of the antecedents of teachers’ innovative behaviors from an emotional perspective, demonstrating that emotional competence, as a critical psychological resource, can be transformed into innovative teaching behaviors through dual “cognitive–motivational” and “energy–motivational” pathways. This study offers both theoretical insights and practical implications for advancing teaching innovation by strengthening teachers’ emotional competence, teaching efficacy, and work vitality. Full article

The autonomous navigation of inspection robots in complex forest environments heavily relies on accurate trunk detection. However, existing detection models struggle to achieve both high accuracy and real-time performance on resource-constrained edge devices. To address this challenge, this study proposes a lightweight algorithm named YOLOv11-TrunkLight. The core innovations of the algorithm include (1) a novel StarNet_Trunk backbone network, which replaces traditional residual connections with element-wise multiplication and incorporates depthwise separable convolutions, significantly reducing computational complexity while maintaining a large receptive field; (2) the C2DA deformable attention module, which effectively handles the geometric deformation of tree trunks through dynamic relative position bias encoding; and (3) the EffiDet detection head, which improves detection speed and reduces the number of parameters through dual-path feature decoupling and a dynamic anchor mechanism. Experimental results demonstrate that compared to the baseline YOLOv11 model, our method improves detection speed by 13.5%, reduces the number of parameters by 34.6%, and decreases computational load (FLOPs) by 39.7%, while the average precision (mAP) is only marginally reduced by 0.1%. These advancements make the algorithm particularly suitable for deployment on resource-constrained edge devices of inspection robots, providing reliable technical support for intelligent forestry management. Full article

Background: Bone health in youth is influenced by both modifiable factors, such as nutrition and physical activity, and non-modifiable factors, such as biological maturation and heredity. Understanding how these elements interact to predict body composition may enhance the effectiveness of early interventions. Importantly, the integration of both supervised and unsupervised machine learning models enables a data-driven exploration of complex relationships, allowing for accurate prediction and subgroup discovery. Methods: This cross-sectional study examined 94 male athletes during the developmental period. Anthropometric, performance, and nutritional data were collected, and bone parameters were assessed using dual-energy X-ray absorptiometry (DXA). Three supervised machine learning models (Random Forest, Gradient Boosting, and Support Vector Regression) were trained to predict Total Body-Less Head (TBLH) values. Nested cross-validation assessed model performance. Unsupervised clustering (K-Means) was also applied to identify dietary intake profiles (calcium, protein, vitamin D). SHAP analysis was used for model interpretability. Results: The Random Forest model yielded the best predictive performance (R² = 0.71, RMSE = 0.057). Weight, height, and handgrip strength were the most influential predictors. Clustering analysis revealed two distinct nutritional profiles; however, t-tests showed no significant differences in TBLH or regional BMD between the clusters. Conclusions: Machine learning, both supervised for accurate prediction and unsupervised for nutritional subgroup discovery, provides a robust, interpretable framework for assessing adolescent bone health. While dietary intake clusters did not align with significant differences in bone parameters, this finding underscores the multifactorial nature of skeletal development and highlights areas for further exploration. Full article

Current deep learning-based dense target detection models face dual challenges in industrial scenarios: high computational complexity leading to insufficient inference efficiency on mobile devices, and missed/false detections caused by dense small targets, high inter-class similarity, and complex background interference. To address these issues, this paper proposes DBA-YOLO, a lightweight model based on YOLOv10, which significantly reduces computational complexity through model compression and algorithm optimization while maintaining high accuracy. Key improvements include the following: (1) a C2f PA module for enhanced feature extraction, (2) a parameter-refined BIMAFPN neck structure to improve small target detection, and (3) a DyDHead module integrating scale, space, and task awareness for spatial feature weighting. To validate DBA-YOLO, we constructed a real-world dataset from cigarette package images. Experiments on SKU-110K and our dataset show that DBA-YOLO achieves 91.3% detection accuracy (1.4% higher than baseline), with mAP and mAP75 improvements of 2–3%. Additionally, the model reduces parameters by 3.6%, balancing efficiency and performance for resource-constrained devices. Full article

This study investigates the uniaxial compression failure of magnesium-based wood-like material (MWM) prisms (100 × 100 × 300 mm³) using digital image correlation (DIC). The results revealed an average compressive strength of 8.76 MPa and a dominant failure mode with Y-shaped or inclined penetrating cracks. A novel piecewise constitutive model was established, combining a quartic polynomial and a rational fraction, demonstrating high fitting accuracy. Critically, the proportional limit was identified to be very low (20–35% of peak stress), attributed to early-stage damage from fiber–matrix interfacial defects. DIC analysis quantitatively distinguished dual crack initiation modes, pure mode I (occurring at ≈100% peak load) and mixed mode I/II (initiating earlier at 90.02% peak load), demonstrating that tensile shear coupling accelerates failure. These findings provide critical mechanistic insights and a reliable model for optimizing MWM in sustainable construction. Future work will explore the material’s behavior under multiaxial loading. Full article

To address the challenges of frequent path conflicts and prolonged computation times in path planning for large-scale multi-UAV operations within urban low-altitude airspace, this study proposes a conflict-free path planning method integrating 3D Jump Point Search (JPS) and an incremental update mechanism. A hierarchical algorithmic architecture is employed: the lower level utilizes the 3D-JPS algorithm for efficient single-UAV path planning, while the upper level implements a conflict detection and resolution mechanism based on a dual-objective cost function and incremental updates for multi-UAV coordination. Large-scale UAV path planning simulations were conducted using a 3D grid model representing urban low-altitude airspace, with performance comparisons made against traditional methods. The results demonstrate that the proposed algorithm significantly reduces the number of path search nodes and exhibits superior computational efficiency for large-scale UAV path planning. Specifically, under high-density scenarios of 120 UAVs per square kilometer, the proposed DOCBS + IJPS method can reduce the conflict-free path planning time by 35.56% compared to the traditional CBS + A* conflict search and resolution algorithm. Full article

Bitcoin’s extreme price volatility presents significant challenges for investors and traders, necessitating accurate predictive models to guide decision-making in cryptocurrency markets. This study compares the performance of machine learning approaches for Bitcoin price prediction, specifically examining XGBoost gradient boosting, Long Short-Term Memory (LSTM), and GARCH-DL neural networks using comprehensive market data spanning December 2013 to May 2025. We employed extensive feature engineering incorporating technical indicators, applied multiple machine and deep learning models configurations including standalone and ensemble approaches, and utilized cross-validation techniques to assess model robustness. Based on the empirical results, the most significant practical implication is that traders and financial institutions should adopt a dual-model approach, deploying XGBoost for directional trading strategies and utilizing LSTM models for applications requiring precise magnitude predictions, due to their superior continuous forecasting performance. This research demonstrates that traditional technical indicators, particularly market capitalization and price extremes, remain highly predictive in algorithmic trading contexts, validating their continued integration into modern cryptocurrency prediction systems. For risk management applications, the attention-based LSTM’s superior risk-adjusted returns, combined with enhanced interpretability, make it particularly valuable for institutional portfolio optimization and regulatory compliance requirements. The findings suggest that ensemble methods offer balanced performance across multiple evaluation criteria, providing a robust foundation for production trading systems where consistent performance is more valuable than optimization for single metrics. These results enable practitioners to make evidence-based decisions about model selection based on their specific trading objectives, whether focused on directional accuracy for signal generation or precision of magnitude for risk assessment and portfolio management. Full article

Relation extraction in the equipment domain often exhibits asymmetric patterns, where entities participate in multiple overlapping relations that break the expected structural symmetry of semantic associations. Such asymmetry increases task complexity and reduces extraction accuracy in conventional approaches. To address this issue, we propose a symmetry- and asymmetry-aware dual-path retrieval and in-context learning-based large language model. Specifically, the BGE-M3 embedding model is fine-tuned for domain-specific adaptation, and a multi-level retrieval database is constructed to capture both global semantic symmetry at the sentence level and local asymmetric interactions at the relation level. A dual-path retrieval strategy, combined with Reciprocal Rank Fusion, integrates these complementary perspectives, while task-specific prompt templates further enhance extraction accuracy. Experimental results demonstrate that our method not only mitigates the challenges posed by overlapping and asymmetric relations but also leverages the latent symmetry of semantic structures to improve performance. Experimental results show that our approach effectively mitigates challenges from overlapping and asymmetric relations while exploiting latent semantic symmetry, achieving an F1-score of 88.53%, a 1.86% improvement over the strongest baseline (GPT-RE). Full article

Accurate carbon emission estimation across all operational stages of container terminals is essential for advancing low-carbon development in the transportation sector and designing effective emission reduction pathways. This study develops a two-layer carbon accounting framework that integrates vessel berthing–waiting and terminal operations, tailored to the operational characteristics of Shanghai Port container terminals. The Ship Traffic Emission Assessment Model (STEAM) is applied to estimate emissions during berthing, while a bottom-up method is employed for mobile-mode container handling operations. Targeted mitigation strategies—such as shore power adoption, operational optimization, and “oil-to-electricity” or “oil-to-gas” transitions—are evaluated through comparative analysis. Results show that vessels generate substantial emissions during erthing, which can be significantly reduced (by over 60%) through shore power usage. In terminal operations, internal transport trucks have the highest emissions, followed by straddle carriers, container tractors, and forklifts; in stacking, tire cranes dominate emissions. Comprehensive comparisons indicate that “oil-to-electricity” can reduce total emissions by approximately 39%, while “oil-to-gas” can achieve reductions of about 73%. These findings provide technical and policy insights for supporting the green transformation of container terminals under the national dual-carbon strategy. Full article

With the rapid development of urbanization and tourism in China, increasing attention has been paid to the protection and utilization of historical and cultural heritage, while tourists’ demands for travel experiences have gradually shifted towards in-depth cultural perception. This paper selects Beijing Fayuan Temple Historic and Cultural District as the research case, and adopts methods such as the LDA (Latent Dirichlet Allocation) topic model, collection and analysis of online text data, and field research to explore the current situation of pedestrian space in Fayuan Temple District and its optimization strategies from the perspective of tourists’ perception. The study found that the dimensions of tourists’ perception of the pedestrian space in Fayuan Temple District mainly include six aspects: historical buildings and relics, tour modes and transportation, natural landscapes and environment, historical figures and culture, residents’ life and activities, and tourists’ experiences and visits. By integrating online text data, questionnaire surveys, and on-site behavioral observations, the study constructed a “physical environment-cultural experience-behavioral network” three-dimensional IPA (Importance–Possession Analysis) evaluation model, and analyzed and evaluated the high-frequency perception elements in tourists’ spontaneous evaluations. Based on the current situation evaluation of the pedestrian space in Fayuan Temple District, this paper puts forward optimization strategies for the perception of pedestrian space from the aspects of block space, transportation usage, landscape ecology, digital technology, and cultural symbol translation. It aims to promote the high-quality development of historical blocks by improving and optimizing the pedestrian space, and achieve the dual goals of cultural inheritance and utilization of tourism resources. Full article

Effective production control requires aligning strategic planning with real-time execution under dynamic and stochastic conditions. This study proposes a scalable dual-agent Deep Reinforcement Learning (DRL) framework for the joint optimisation of Work-In-Process (WIP) control and job sequencing in flow-shop environments. A strategic DQN agent regulates global WIP to meet throughput targets, while a tactical DQN agent adaptively selects dispatching rules at the machine level on an event-driven basis. Parameter sharing in the tactical agent ensures inherent scalability, overcoming the combinatorial complexity of multi-machine scheduling. The agents coordinate indirectly via a shared simulation environment, learning to balance global stability with local responsiveness. The framework is validated through a discrete-event simulation integrating agent-based modelling, demonstrating consistent performance across multiple production scales (5–15 machines) and process time variabilities. Results show that the approach matches or surpasses analytical benchmarks and outperforms static rule-based strategies, highlighting its robustness, adaptability, and potential as a foundation for future Hierarchical Reinforcement Learning applications in manufacturing. Full article