Abstract:Multi-view data often suffer from feature space heterogeneity and noise interference, which severely affect the stability and accuracy of subspace clustering. To address the limitations of existing methods in handling view dimension differences and low-quality view impacts, this paper proposes a decoupled projection and shared structural consistency multi-view subspace clustering method (DP-SC-MVC). The model designs independent linear projection matrices for each view to achieve feature decoupling and dimension alignment. It simultaneously learns a shared representation across views and imposes low-rank constraints to maintain structural consistency. A dynamic weighting mechanism is introduced to adaptively adjust the contribution of each view, suppressing the interference from inferior views. Additionally, -norm regularization is incorporated to enhance robustness against noise and outliers. The unified optimization framework is solved using the augmented Lagrange multiplier method and alternating direction minimization. Experimental results on multiple multi-view benchmark datasets demonstrate that the proposed method significantly outperforms existing mainstream multi-view subspace clustering methods in terms of clustering performance. The DP-SC-MVC method effectively addresses the issues of feature space heterogeneity and noise, demonstrating strong practical value and broad application prospects.

Abstract:To address the issues of input delay and actuator saturation in the operation of high-speed trains, a new dynamic surface funnel control scheme was proposed, which ensures the stability of the control system, eliminates the influence of delays, and enables the train to have ideal tracking performance without saturation. Firstly, a longitudinal dynamics model of the high-speed trains with input delay and saturation was established through force analysis. Then, based on the new dynamic surface control framework, a novel dynamic compensation and constraint mechanism was proposed to address the dual constraints of time delays and saturation, which ensures the tracking error was strictly confined within a dynamic funnel boundary and achieves asymptotic tracking without saturation. Finally, the CRH380A high-speed train was employed as the controlled object for simulation verification. The results show that the designed controller effectively suppresses the input delay effect of the train and remains stable when the input saturation occurs, and the train quickly recovers the ideal tracking performance without saturation. The proposed method can effectively compensate for the input delay effect and constrain the input saturation, effectively improving the operational safety and reliability of high-speed trains.

Abstract:To address the issue of muck spewing during earth pressure balance (EPB) shield tunneling in water?rich sandy strata, this study systematically investigates the evolution of permeability characteristics of conditioned muck and the critical conditions for spewing. A pressurized permeation apparatus capable of simultaneously simulating soil pressure and water pressure at various burial depths was developed. Using this setup, the evolution of muck permeability under the coupled effects of multiple factors-including permeation pressurePw, confining pressurePs, foam injection ratio (FIR), types of conditioning agents, and particle gradation-was systematically examined. The results reveal that foam conditioning exhibits distinct “critical injection zones” and “over?conditioning zones,” with an optimal FIRof approximately 5%–8%; bentonite slurry demonstrates more stable impermeability under pressurized conditions. Based on the experimental data, a response surface model incorporating the stress ratio Pw/Ps, FIR, and the permeability coefficient ratio k/k0 was established, and a critical stress ratio Pw/Ps= 0.5 was identified. Furthermore, a multi?dimensional critical criterion for spewing centered on “permeability coefficient-water head-flow rate” was proposed, leading to the development of a tiered early?warning system based on the concept of a dynamic critical permeability coefficient. Finally, a differentiated prevention and control strategy that matches “stratum–conditioning agent–parameters” was formulated. The research outcomes provide a systematic solution, from mechanism to application, for controlling spewing risks in EPB shield tunneling through water?rich sandy strata.

Abstract:In recent years, with the widespread application of unmanned aerial vehicle technology, frequent unauthorized "black flying" incidents have posed a severe threat to low-altitude airspace security, making the research of counter-UAV detection and perception technologies highly urgent. However, conventional single-sensor detection methods face significant challenges in accuracy and robustness due to the difficulty of capturing micro-targets, dense obstacles, and complex electromagnetic and meteorological disturbances in low-altitude scenarios. To address this issue, this paper provides a comprehensive review of the key algorithms and cutting-edge advancements in low-altitude counter-UAV detection and perception. First, it systematically summarizes the evolution of vision-based object detection algorithms, focusing on the innovative applications of traditional Convolutional Neural Networks, the ultra-lightweight YOLO series, and Transformer architectures with global context perception capabilities in detecting "low, slow, and small" UAVs. Second, it analyzes the detection advantages and technical bottlenecks of non-visual sensors, including radar, acoustic, and radio frequency, when combined with deep learning technologies. Finally, addressing the demands for all-weather detection in complex environments, the paper deeply explores the current development of multimodal fusion algorithms, such as RGB-infrared and multispectral fusion. This paper concludes by summarizing the current challenges in perception robustness and generalization capabilities of core counter-UAV algorithms, and anticipates future critical research directions including visual large models, air-ground multi-agent coordination, and multi-source information fusion, aiming to provide useful references for the technological evolution and engineering practice of low-altitude counter-UAV systems.

Abstract:To address the challenges of large model size, high computational complexity, and notable accuracy degradation when deploying railway track obstacle detection models on edge devices, this paper proposes a lightweight re-parameterized network, YOLOv11s-Slim-Rep. The network achieves synergistic optimization of accuracy and efficiency through two core improvements: adjusting the network width scaling factor from 0.50 to 0.35, resulting in a 44.8% reduction in model size and a 45.4% decrease in parameter count; and replacing all stride=2 downsampling convolutional layers with a RepConvCustom module, which is fused into a single convolution via structural re-parameterization, incurring zero additional inference overhead. On two railway track obstacle datasets (NewData and Railway), compared to the YOLOv11s baseline, the mAP50 decreases by only approximately 1.1% while achieving nearly 45% model compression. Compared to YOLOv11n, it improves mAP50 by 4.7% in complex scenarios, with a cross-dataset mAP50 fluctuation of only 0.6%. The proposed method maintains high detection accuracy while significantly compressing the model, demonstrating strong potential for edge deployment.

Abstract:Ice accretion on catenary cantilever assemblies makes ice removal difficult and complicates the determination of effective de-icing paths for complex tubular components. To address this issue, a mechanical de-icing method based on the impact of a cylindrical polytetrafluoroethylene (PTFE) rod is proposed. A finite element model of an iced German-type aluminum alloy cantilever assembly was established, including the horizontal arm, inclined arm, cantilever support, and positioning tube. The model was used to analyze local compressive damage, crack propagation, and ice–substrate interfacial debonding under impact loading. Based on the single-impact de-icing range of each component, an axial impact spacing strategy with a 50% overlap ratio was adopted, and an alternating upper-lower impact path was planned. Full-diameter axially segmented tests and overall path-based de-icing tests were then conducted to validate the proposed method. The results showed that the relative error in de-icing area between the simulation and experiments ranged from 2.63% to 14.54%, which was less than 15%. Under the planned path, the average de-icing rate of the cantilever assembly reached approximately 79.3%. The results indicate that the cylindrical PTFE rod impact method can effectively fracture and detach ice accretion from catenary cantilever assemblies, and that the proposed path planning method provides a reference for end-effector design and operation path generation in automated catenary cantilever de-icing equipment.

Abstract:To investigate tunnel face stability under complex working conditions, a three-dimensional active failure model incorporating soil non-homogeneity and seismic effects is constructed based on the linear Mohr-Coulomb criterion, utilizing spatial discretization and "point-generation" techniques. A non-homogeneity coefficient is introduced to characterize the linear distribution of cohesion with depth, and an analytical upper-bound solution for the limit support pressure is derived and optimized via MATLAB. Comparative validation demonstrates that the proposed method possesses good calculation accuracy. The results indicate that seismic effects significantly reduce tunnel face stability; the limit support pressure increases with the rise of the horizontal seismic acceleration coefficient and the vertical seismic proportionality coefficient, with the vertical seismic action exhibiting a non-linear amplification effect. Conversely, soil non-homogeneity contributes to enhancing self-stability capabilities; the limit support pressure shows a significant linear negative correlation with the non-homogeneity coefficient, suggesting that fully utilizing the strength increase characteristics of deep soil can optimize support design. However, strong seismic action weakens the safety benefits derived from soil non-homogeneity, necessitating increased seismic redundancy in engineering designs for high-intensity earthquake zones.

Abstract:Urban low-altitude logistics delivery faces constraints such as dense buildings and complex airspace. Existing studies insufficiently characterize the geographical environment and treat task allocation and trajectory planning separately. Based on complex urban environments, this paper proposes a bi-level planning model for multi-UAV collaborative delivery. First, a 3D grid environment incorporating static obstacles and dynamic risk fields is constructed. Second, a bi-level coupled optimization model is established. The upper-level model aims at total cost and time satisfaction, employing an improved GA-SA algorithm for task allocation; the lower-level model introduces grid flight risk and integrates a 3D A* algorithm for collision-avoidance trajectory planning. Experimental results show that the model can effectively generate 3D cooperative obstacle-avoidance trajectories in complex urban scenarios, balancing timeliness and flight safety.

Abstract:For low-carbon multimodal transport networks under uncertain transportation demand and transit time conditions, considering factors such as waiting costs, time windows, and service frequency, this study explores the multi-objective optimization problem of container multimodal transport networks. Robust optimization is employed to handle fluctuations in transportation demand, Monte Carlo simulation is used to represent uncertainties in transportation time, and mixed time window constraints are introduced to construct a stochastic robust optimization model targeting total cost, total carbon emissions, and total time. To improve algorithm convergence and maintain population diversity, an improved adaptive fast non-dominated sorting genetic algorithm II integrating multiple crossover and mutation strategies is designed. To avoid risks from parameter uncertainty affecting enterprise operations, multimodal transport operators identify the most preferred "satisfactory solutions" on the Pareto frontier according to decision-maker preferences, with decision plans favoring low-carbon transport. The improved NSGA-II algorithm effectively addresses multi-objective and uncertainty challenges in freight network optimization, providing decision-makers with a comprehensive perspective on cost, time, and environmental impact under varying robustness levels, demonstrating its potential and effectiveness in solving complex practical transportation problems. In light of their internal and

Abstract:To address fatigue crack propagation and life assessment of coupler knuckles under heavy-haul conditions, a crack growth life prediction method based on stress intensity factor range (ΔK) evolution is proposed. First, statistical analysis of in-service locomotive coupler data is performed to identify crack-prone regions and spatial distribution. A three-dimensional finite element model of the coupler knuckle is established to obtain stress field under loading conditions, and key analysis regions are determined with crack initiation locations. A three-dimensional surface crack model is introduced to evaluate stress intensity factor distribution along the crack front. Crack growth behavior is described using Paris law, and an incremental integration-based fatigue life prediction model is developed. Results show that cracks are mainly concentrated at the pin hole edge and traction lug transition region, consistent with high-stress regions from finite element analysis. The stress intensity factor range ΔK shows non-uniform distribution along the crack front, with peak concentration within about 1 mm beneath the surface. With increasing load eccentricity, the ΔK distribution is reconfigured, leading to higher crack growth rate and reduced fatigue life. The proposed method enables fatigue life prediction from crack initiation to failure. Compared with traditional nominal stress or empirical damage models, the proposed method describes three-dimensional non-uniform evolution of ΔK at the crack front and establishes relation between crack driving force and remaining life under load eccentricity conditions.