A dual-task parallel machine learning framework was developed by integrating a convolutional autoencoder （CAE） and a fully connected neural network （FCNN） via the gradient-coupled mechanism， enabling simultaneous data compression-reconstruction and structural damage identification. Under the condition where 40% of the sensor nodes are missing， the model successfully reconstructs the full sensor network with an R² of 0.916 and normalized root mean square error （NRMSE） of 0.028 8. Even under significant noise contamination with an SNR of 12 dB， the model maintains strong reconstruction performance， achieving a R² of 0.910 and NRMSE of 0.025 3. Forty-six structural damage scenarios were simulated using the scaled bridge model. The accuracy of spatial localization and quantification of the damage severity using the framework exceeds 99.3%. The proposed framework reduces the training time by 54.4% and iteration counts by 45.5% compared to conventional two-stage machine learning approaches， demonstrating the efficiency of gradient-coupled optimization.

Based on the surrounding rock arching and hinge-less arch structure theories, a theoretical formula for the minimum overburden thickness was derived. By substituting different mechanical parameters of multiple tunnels at home and abroad into this formula, minimum self-supporting arch formulas under different surrounding rock classes were obtained. Based on the actual engineering case of a dual-mode shield tunnel, a numerical model for the tunnel boring machine excavation mode was established to verify the theoretical formulas. Next, three surrounding rock classes, four soil layer thickness gradients, and twelve overburden thickness gradients were designed, resulting in 144 models formed by the combination of the three factors. Uniform tests were conducted, and the pressure arch heights under different surrounding rock classes were obtained. The results show that in the theoretical formulas, the tunnel radius has a linear positive correlation with the pressure arch height, while the tunnel depth has a linear positive correlation with the square of the pressure arch height. According to numerical simulation results, the pressure arch height increases with the increase of the overburden thickness and then tends toward a critical value of twice the tunnel diameter. Finally, the results of the numerical model are in good agreement with those calculated using the theoretical formulas, verifying the rationality of the established theoretical formulas.

The virtual preassembly of super-high steel bridge towers faces a challenge in the efficient and precise extraction of complex cross-sectional features. Factors such as fabrication errors， gravity-induced deformations， and temperature fluctuations can compromise the accuracy of contour extraction. To address these limitations， an improved Alpha-shape-based point cloud contour extraction method is proposed. The proposed approach uses a hierarchical strategy to process three-dimensional laser scanning point clouds. The processed data are then subjected to curvature-adaptive voxel filtering to reduce acquisition noise. In addition， an enhanced iterative closest point （ICP） variant with correspondence validation accurately aligns the discrete point cloud segments. The proposed curvature-responsive Alpha-shape framework enables multiscale contour delineation through topology-adaptive threshold modulation， which resolves boundary ambiguities in geometrically complex cross-sections. The method was experimentally validated using field-acquired measurement datasets from the Zhangjinggao Yangtze River Bridge tower segments， confirming its capability to reconstruct noncanonical cross-sectional geometries. Three contour extraction methods， including Poisson reconstruction， the conventional Alpha-shape algorithm， and random sample consensus with ICP （RANSAC-ICP）， were compared to evaluate the performance of the proposed Alpha-shape algorithm. The results demonstrate that the proposed method achieves superior contour extraction accuracy and data reduction efficiency， highlighting its effectiveness in contour extraction tasks.

The intrinsic interaction mechanism of flutter between the flow and structure of a rectangular plate remains a mystery from the viewpoint of unsteady flow. The present study provides a novel insight into this interaction mechanism based on an adequate understanding of the formation and evolution of the flapping leading-edge vortex (LEV). A series of wind tunnel tests was conducted to investigate the nonlinear flutter instability of an 8∶1 rectangular plate. The complete flow fields around the model throughout the flutter process were obtained by a particle image velocimetry (PIV) technique using two synchronous cameras with an interpolation and resampling method. To acquire the flow structures corresponding to the characteristic frequency of flutter, the spectral proper orthogonal decomposition (SPOD) method was extended to a noninertial frame to reconstruct the low-rank flow field during flutter and extract the characteristic flow pattern coupled with oscillations. It was found that when the 8∶1 rectangular plate undergoes flutter, the LEVs exhibit a periodic flapping phenomenon induced by the structure oscillations. A two-dimensional correlation analysis of the flapping LEVs was conducted for different inflow velocities. The results demonstrate that there is a substantial phase lead phenomenon in the LEV evolution downstream for a higher inflow velocity. This phenomenon may be related to a phase offset of aerodynamic forces, and finally, it gives rise to flutter.

Reducing carbon emissions is fundamental to achieving carbon neutrality. Existing studies have typically estimated emissions by predicting fossil fuel consumption across sectors under different socioeconomic scenarios; however, uncertainties in future development often lead to deviations from these assumptions. To address this limitation, this study proposes a data-driven approach for evaluating national carbon emissions using historical data. Countries with similar energy consumption patterns were selected as reference samples, and their emission pathways were analyzed to predict future emissions for countries that have not yet reached their peak. Key indicators, including peak levels, timing, plateau duration, and post-peak decline rates, were identified. The results indicate that the trends in unpeaked economies can be effectively assessed based on the emission patterns of countries with comparable energy structures. Applying this framework to China suggests a carbon peak between 2027 and 2030, in the range of 14.207 to 16.234 Gt, followed by a gradual decline from 2031 to 2036. Compared with the average results of the existing studies, the predicted minimum and maximum emissions show error margins of 10.1% and 1.41%, respectively. This study proposes a top-down methodology that provides a transparent, reproducible, and empirical framework for forecasting carbon emission pathways, thereby offering a scientific basis for assessing countries that have not yet reached their emissions peak.

Extreme traffic loads significantly challenge the safety and cost-effectiveness of highway bridges, especially under site-specific traffic conditions. Conventional assessments often rely on overly conservative load models, leading to excessive structural design. In this study, a framework for the prediction of maximum bending moments in simply supported bridges is developed by integrating weigh-in-motion (WIM) data, traffic microsimulation, and generalized extreme value (GEV) regression modeling to establish relationships between the GEV parameters (μ, σ, ξ) and traffic factors—heavy vehicle proportion, bridge span length, vehicle speed, headway, and traffic volume. Using one-year WIM data from 7.4 million vehicles, the developed models for μ and σ exhibit high predictive accuracy (R²>0.95) and are validated through leave-one-out cross-validation. The prediction of ξ is less accurate (R² ≈ 0.6), requiring further improvement. Applying these models to a 1 000-year return level yields a reliable, data-driven extrapolation, supporting optimized bridge design and safety assessment under varying traffic conditions.

To address the issue that static densest subgraph mining algorithms often exhibit low efficiency when handling large scale dynamic graphs, this paper proposes a heuristic approximation algorithm. The algorithm approximates the densest k-subgraphs of the entire graph through four steps: partitioning the large-scale dynamic graph, constructing a partial set of the densest k-subgraphs, heuristically merging the subgraph sets, and finally extracting the densest k-subgraphs. This approach significantly reduces the computational time for large-scale dynamic graphs while simultaneously improving the quality of the resulting subgraphs. This algorithm is applicable to various definitions of “density” and can accommodate diverse requirements on the number of edges. When integrated with existing static densest subgraph detection algorithms, it achieves scalability and computational efficiency. Theoretical analysis demonstrates that the optimal density of the densest k-subgraphs extracted by the proposed algorithm reaches 0.9. To evaluate the performance of the algorithm, experiments were conducted on four billion-scale datasets: Friendster, Orkut, YouTube, and DBLP. The results indicate that the proposed algorithm outperforms static methods in both runtime efficiency and subgraph quality on large-scale dynamic graphs.

To uncover the decision-making mechanisms and evolutionary dynamics of multiple stakeholders in highway noise pollution control, a three-party evolutionary game model involving the government, operators, and the public is constructed. The operation period is divided into different stages for differentiated analysis. A simulation analysis was performed on the Lituo sinking section of the Beijing-Hong Kong-Macao Highway to assess the impact of variations in critical elements on the system. The results indicate that the Lituo sinking section of the Beijing-Hong Kong-Macao Highway is currently in its early stage of development, with the corresponding strategies being active regulation, excessive emissions, and supervision. When the cost of the government’s active regulation decreases from 1×10⁵ to 5×10⁴ yuan, the system converges more rapidly toward the active regulation strategy. When the cost of the operator’s excessive emissions increases from 14.08×10⁶ to 20.00×10⁶ yuan, the system drives the operator toward the standardized emission strategy. In addition, when the cost of public supervision decreases from 15×10⁴ to 5×10⁴ yuan and the compensation paid by operators to the public increases from 1.288×10⁶ to 2.576×10⁶ yuan, the system converges more quickly toward the supervision strategy. The cost of the operator’s excessive emissions serves as the core decision variable for achieving the ideal equilibrium in the three-party game involving government active regulation, operator standardized emissions, and public supervision.

To promote the application of green recycled construction materials in civil engineering, this study presents a statistical damage constitutive model for polypropylene fiber recycled fine aggregate concrete (PRFAC), based on the strain equivalence principle and the assumption that microelement strength follows a Weibull statistical distribution. The proposed model incorporates the Drucker‐Prager failure criterion. By examining the influence of Weibull distribution parameters m and S₀ on the stress‐strain response, empirical relationships were established between the fine aggregate replacement ratio and the distribution parameters. This enabled the derivation of a theoretical stress‐strain curve accounting for variable recycled fine aggregate (RFA) replacement ratios. The experimental results show that the proposed model exhibits high agreement with measured data and effectively captures the increased brittleness of PRFAC with higher RFA replacement ratios. Moreover, increasing the replacement rate accelerates internal crack propagation, reduces deformability and toughness, and significantly hastens the accumulation of internal damage in PRFAC.

Deep coal mining rock support structures using rock bolts face complex geological conditions such as high ground temperatures and groundwater. Rock mass deformation and failure caused by bolt failure frequently occur, making it crucial to enhance the anchoring performance of rock bolts. First, the stress state of the anchor rod under axial loading across five stages of any anchored segment is analyzed. The shear stress patterns at the anchoring interface during different stages are elucidated. A refined mechanical model of the anchoring interface incorporating surface rib parameters is established. A failure criterion for the anchoring interface under the influence of ground temperature or groundwater is derived and validated. Second, the influence of anchor rib parameters on anchoring force is analyzed, and in-situ shear tests are conducted. Results indicate that increasing the rib angle and optimizing rib spacing can enhance anchoring force. To minimize the shear component of axial force at the anchor interface, the rib angle of the anchor bolt should not be less than 70°. When the anchor grout possesses high inherent strength, the spacing between ribs on the anchor bolt surface may be increased (to 24 mm or greater). Finally, methods for enhancing the anchoring performance of bolts in deep complex strata are proposed, providing technical references for the safe and efficient support of tunnel rock masses in similar geological conditions.

Spaceborne optomechanical systems face the dual challenges of extreme thermal disturbances and millikelvin-level temperature control precision during orbital operations, demanding robust control strategies. To address the performance limitations of conventional fixed-parameter active disturbance rejection control (ADRC) under complex operating conditions, this work proposes a Q-learning-enhanced adaptive ADRC framework. A thermal-transfer model incorporating multisource disturbances (solar radiation, structural conduction, and contact thermal resistance) is established, coupled with a reinforcement learning-driven parameter optimization mechanism. The ε-greedy policy dynamically adjusts observer bandwidth (ω_o ∈ [0.01, 0.2]) and controller bandwidth (ω_c ∈ [0.01, 0.1]) to enable real-time estimation and compensation of total disturbances. Simulation results demonstrate significant improvements over fixed-parameter ADRC and a self-tuning internal model control proportional-integral (SIMC-PI) controller: 31.3% and 15.4% reduction in settling time during setpoint responses, respectively; 21.8% lower integral absolute error (IAE) than the fixed-parameter ADRC during setpoint step responses; 12.7% and 52.5% enhancement in control precision over conventional fixed-parameter and SIMC-PI controllers, respectively, under ±10 K periodic and step thermal disturbances. Monte Carlo robustness tests reveal smaller fluctuation ranges of IAE, settling time, and overshoot under ±5% parameter perturbations. This methodology establishes a new paradigm for millikelvin-level thermal control in space optical payloads.

Current image inpainting models are primarily designed to achieve a large receptive field （RF） using refinement networks to incorporate different scales. However, these models fail to adapt the use of different RFs to the specific patterns of image damage, resulting in artifacts and semantic information confusion in repaired images. To address the problems of artifacts and semantic information confusion, inspired by different sensitivities of different RFs to inpainting the same image damaged patterns, this study proposes an image inpainting method based on multiple receptive fields （MRFs） and dynamic matching of damaged patterns. First, the parallel filter banks are used to extract the MRF feature groups. Second, the features are dynamically weighted and screened, guided by the mask image, to construct a relationship that adaptively matches the most relevant RF to each specific damaged pattern. A fast Fourier convolution based decoder is used to enhance the fusion of global contextual features during the reconstruction of high dimensional features into low dimensional images. Comparative experimental results show that the proposed method achieves better subjective and objective inpainting results on three public datasets: Paris StreetView, CelebA‐HQ, and Places2.

To enhance the heat-dissipation capacity of infra-red (IR) stealth structures in high-temperature environments, a selective heat emitter with multi-band thermal management is fabricated. This emitter comprises a high-temperature-resistant titanium dioxide (TiO₂)/hafnium dioxide (HfO₂)/Cr/Ge/Mo multi-film-layer structure. Additionally, the thickness of each layer is determined by the transfer-matrix algorithm. The emissivity of the structure across the IR band is simulated, and its electric field distributions are analyzed across different wavelengths. The stealth-and heat-dissipation bands of the structure are calculated to confirm its overall stealth and heat-dissipation capabilities. The results reveal that the average emissivities of the fabricated TiO₂/HfO₂/Cr/Ge/Mo multi-film-layer structure decrease to 0.21 and 0.27 within 3-5 and 8-14 µm atmospheric window bands, respectively, achieving the IR concealment effect. Conversely, the average emissivities of the structure increase to 0.56 and 0.80 within the 2.5-3 and 5-8 μm non-atmospheric window (NAW) bands, respectively. These high-emissivity bands enhance radiative heat dissipation to reduce heat accumulation and further weaken the detection and characterization of thermal signals. The simulated thermal images confirm the IR-stealth effect of the structure within a wide temperature range. Moreover, its efficient NAW heat-dissipation capability improves its operating life in high-temperature environments.

Information collaboration is crucial for optimizing resource allocation and improving diagnostic efficiency across hospital tiers through enhanced information technology capacity. To characterize the dynamic decision-making mechanism between general hospitals (GHs) and primary healthcare centers (PHCs), a two-player differential game model was constructed to analyze the relationship between optimal investment levels and corresponding payoffs and explore how GHs can incentivize collaboration by adjusting their investment intensity and sharing PHCs’ costs. The results indicate that information collaboration is a win-win strategy. Its dynamic equilibrium shows that GHs make intensive efforts in the early stage of digital construction. However, such investment decreases over time as patient information accessibility becomes limited. Under the collaboration mode, although GHs’ digital investment is lower than that in the independent operation, the total system payoff significantly increases. This improvement arises because PHCs, with their locational and informational advantages, undertake major digitalization tasks, allowing GHs to focus resources on disease treatment. The introduction of collaboration incentives strengthens this performance improvement.