Computer-Aided Civil and Infrastructure Engineering最新文献

To fully leverage connected automated vehicle (CAV) technology for improving traffic flow at signalized intersections, this paper addresses the scalability limitations of traditional microscopic control methods. We propose a macroscopic connected automated flow control (CAFC) framework based on the cell transmission model (CTM), which formulates the vehicle sorting problem as a computationally efficient Mixed-Integer Quadratically Constrained Program (MIQCP). Numerical experiments, comparing our CAFC strategy against a traditional dedicated-lane benchmark, demonstrate a throughput improvement of approximately 63%. The framework also shows strong robustness in dynamic scenarios with mismatched traffic demand and signal timings, consistently outperforming a stronger, demand-responsive baseline. The results indicate that macroscopic flow control offers a scalable and highly effective alternative to microscopic methods for real-time traffic management in pure CAV environments.

Shield tunnels in water-rich strata are usually large in scale, long in length, and located in complex operating environments. The diagnostic results for tunnel structural uplift are easily affected by complex environmental factors. How to extract the key information related to the structural uplift of the tunnels from the massive monitoring data of distributed optical fiber sensors in this complex water-rich environment and accurately diagnose the structural uplift of the tunnels remains a difficult problem that urgently needs to be solved. This paper proposes a diagnostic approach for structural uplift in water-rich shield tunnels, which utilizes the spatiotemporal features of the densely distributed strain data to address this challenge. On the one hand, the spatial interdependence of the densely distributed strain data is analyzed. By combining k-means clustering with the artificial bee colony algorithm and referring to the distribution characteristics of the tunnel surrounding rock, a clustering algorithm for the dense strain measurements along the length direction of the tunnel is proposed. Then, a spatial interdependence model for the densely distributed measurements is established based on a one-dimensional convolutional neural network. On the other hand, the spatial interdependence features of the strain data are analyzed in the time domain. The factors influencing the spatial interdependence residuals of the strain data within each category are analyzed by using the principal component analysis algorithm and a diagnosis index for the structural uplift of the tunnel is constructed on the basis of the aforementioned residuals, thereby achieving a diagnosis of the structural uplift of the water-rich shield tunnel. Finally, the proposed method is validated using a synthetic numerical simulation and field monitoring data from an actual tunnel project.

Automated crack segmentation models are vital for infrastructure monitoring but fail when deployed in new domains. Overcoming this domain shift without costly re-annotation is vital. This paper presents a novel unsupervised domain adaptation framework that uniquely integrates Fourier-based style transfer with targeted morphological operators and a robust Uncertainty-guided self-training scheme. Specifically, its Fourier–Morphology blending aligns visual styles and crack geometries between domains through controllable image processing operations governed by two intuitive parameters. This is paired with an Uncertainty-guided dual-network training scheme that safely leverages unlabeled target data for robust self-training. Experiments on public and industrial data sets show state-of-the-art performance, improving the $��F�1�$ score by up to 18.5% over competitive baselines in challenging cross-domain scenarios.

Multi-Train Optimal Control (MTOC) addresses the cooperative control problem of multi-trains running on railway tracks through centralized or distributed controllers. However, two critical challenges emerge in solving MTOC problems: (1) the dynamic system dimensionality caused by time-varying train numbers during station arrivals and departures and (2) the strong inter-train command correlations in dense traffic scenarios. These complexities lead to computational challenges when scaling to extended railway networks with growing train populations, rendering conventional rule-based methods ineffective. To address these challenges, we propose Graph Attention Soft Actor-Critic (GASAC), a novel graph reinforcement learning algorithm integrating two core components: (1) A graph attention network (GAT) for efficient information aggregation from high-dimensional train observations, and (2) A Soft Actor-Critic (SAC) architecture serving as the centralized decision-maker. The GAT module performs dimensionality reduction through feature attention mechanisms, effectively supporting the SAC module in deriving optimal control policies. Comparative evaluations against multi-agent deep reinforcement learning baselines demonstrate that GASAC successfully synthesizes distributed train information to generate control commands, ensuring collision-free and on-time operations. Further sensitivity analysis shows the adaptability of the algorithm to different parameters.

Automatic condition assessment of road pavements is important for efficient pavement management. Most previous studies targeted highway, national, and state road routes and adopted their own imaging vehicles to evaluate pavement conditions. This study focuses on street view images for large-scale, fully automated, and comprehensive condition evaluation including local municipality roads. This study proposes a low-cost, efficient, and accurate method combining geographic information system (GIS), street view images, and state-of-the-art deep learning and clustering methods. The three contributions are: (1) Automatic high-quality data acquisition method is established. Geographic positions and road directions are estimated by image processing of GIS maps. Google Cloud Application Programming Interface is adopted. Street view images are screened to remove interior and building façade images applying a road segmentation U-Net trained with the CityScapes dataset. (2) Previous road damage dataset RDD2020 is augmented to adjust to street view images with shadows from adjacent objects, walls, pedestrian road tiles, and logos. You only look once version 8 (YOLOv8) is adopted to detect damages and classify the conditions at each location. (3) It was first revealed at a fine local scale in large administrative areas that the damages show spatial cluster patterns on GIS maps. Time histories are analyzed to depict deterioration process. To validate the method, about 144,000 images were collected in four wards in the Tokyo districts. It costs $0 to $100 and 5 to 10 h for one ward. A fine-tuned YOLOv8 model achieved about 95% classification accuracy. Damage maps varied while curves were similar, which are effective in practice, reflecting each municipality's pavement condition and meeting the inspection standards.

Direct visual understanding of construction entities, such as tools and materials (T&M), underpin construction management and resource scheduling. Traditional supervised learning methods suffer from high annotation cost, severe computational demands, and limited datasets. In contrast, training-free approaches offer an effective alternative well-suited for construction scenarios constrained by data scarcity and limited resources. Besides, vision-language models (VLMs) can directly learn image semantics through natural language supervision and also demonstrate strong zero-shot detection capabilities without requiring retraining. Existing methods often exhibit limited image–text semantic alignment in construction scenarios, which restricts their effectiveness in construction tasks. Therefore, there is an urgent need for approaches that can enhance cross-modal understanding in such domain-specific contexts. To address this challenge, this paper proposes a training-free, knowledge-enhanced VLM to recognize T&M in construction tasks. The proposed approach leverages image matching and image–text knowledge alignment strategies, thereby utilizing the training-free nature of existing VLMs while benefiting from enhanced performance brought by knowledge integration. This method offers a novel solution for construction management and robotic collaboration tasks that are traditionally constrained by data and computational resource dependencies.

The topological networks with evacuation information of most macro evacuation models are created manually, which is a repetitive and time-consuming work. Meanwhile, micro-evacuation simulation software often pre-inputs path networks to reduce computational costs. To address these issues, this study developed a semantic segmentation model to recognize doors and rooms from building floor plans. Then, a series of morphological image processing techniques is proposed to further improve the accuracy of the results. Various types of functional nodes in topological network are extracted from the results, and improved A-star algorithm is adopted to find out the interconnectivity, that is, the lines among functional nodes. The topological network with evacuation information is generated. The proposed method achieves fast and automated generation of macro evacuation network models from building floor plans and finds possible congestion nodes in the building. Additionally, this study proposes a micro-level visualization method for abstract macro-simulation results. In the case study presented, the generated macro evacuation network model and the simplified micro-level simulation results transforming method both demonstrate excellent performance.

Deploying cables into the frame structure is an effective method to enhance its structural stiffness. The efficacy of cables is highly dependent on their placement, posing the core challenge of accurately identifying the optimal deployment positions from a vast array of feasible options. However, there exists a significant research gap in the field of structural optimization concerning cable arrangement. In current engineering practice, cable layout primarily relies on experience-based methods grounded in mechanical concepts (such as regions of large deformation), making it difficult to identify a globally optimal solution. To address this, an automatically accurate identification method is proposed to find the optimal deployment of high-performance cables within an exponentially large solution space. Leveraging the graph neural networks (GNNs) architecture, an intelligent generative cable optimal deployment (IGCOD) model is presented, which embeds a finite element physical model. This model utilizes the GNNs as a topology generation and discrimination engine, constructing an end-to-end closed-loop framework through the following steps: topology feature extraction, automated cable generation, and optimal scheme identification. By directly embedding the mechanical response of the physical model into the network prediction, a fully automated design is achieved without labeling the pre-training data. In various topological configurations of frame structures, the IGCOD model accurately identified the optimal cable placement within tens of thousands of feasible solutions, thereby maximizing structural stiffness performance. In the cases of irregular multi-story and high-rise frame structures, the maximum optimization effect of three pairs of cables increased by 40% and 21%, respectively, and the corresponding time cost is 717 and 6384 s. This research presents a systematic and transferable artificial intelligence (AI)-driven paradigm for the high-performance reinforcement of existing buildings, thereby reducing design costs and maximizing structural performance.

In structural defects inspection, the quantitative detection of slender cracks remains a significant challenge. Existing methods suffer from low segmentation accuracy for complex boundaries and high computational demands for high-resolution (HR) images, making them unsuitable for the current scenarios where unmanned devices are widely deployed. To address the above-mentioned limitations, a crack detection and quantification framework based on multi-scale convolution-enhanced Mamba (MCMamba) and an HR image calibration method is proposed. The MCMamba is designed based on the Mamba architecture and the calibration method using variable step-size moving least squares is proposed to fit the scale field of HR images, enabling precise crack segmentation and quantification. The MCMamba is trained on an established dataset, and the framework is further field-tested using a climbing robot and Unmanned Aerial Vehicle (UAV), achieving accuracy with less than 10% error for cracks thinner than 0.2 mm. This framework improves crack detection accuracy and demonstrates its advantages in quantifying slender cracks on large-scale bridges in engineering practice.

The cover image is based on the article Exploring the unjamming transition of meso-mechanical shear failure behavior in asphalt mixture by Geng Chen et al., https://doi.org/10.1111/mice.70089.