Abstract Accurate and rapid prediction of deep excavation deformation is crucial for construction safety and environmental protection. Traditional finite element analysis is time‐consuming, while single‐objective optimization (SOO) tends to cause parameter compensation effects. This paper proposes a deep learning and multi‐objective optimization (MOO) approach for excavation deformation prediction. A triple‐attention convolutional network (TACN) is constructed to capture the complex interactions among soil parameters, deformation locations, and excavation stages. Integrating the proposed TACN, a TACN‐MOO optimization framework is established to perform rapid parameter identification and deformation prediction by simultaneously considering wall deflection and ground settlement. Validation through a Shanghai excavation project shows: (1) TACN effectively captures nonlinear soil‐deformation relationships with higher accuracy than convolutional neural network models; (2) the MOO framework effectively mitigates parameter compensation effects while reducing computation time from 8+ h to 1–2 min; (3) engineering applications demonstrate that the method achieves high accuracy in wall deflection prediction and good agreement in settlement estimation with excellent transferability. This research provides an efficient and reliable technical framework for intelligent prediction and dynamic control of deep excavation deformation.

Abstract Accurate and rapid assessment of pavement surface condition is essential for maintaining transportation safety and minimizing vehicle wear. Manual pavement inspections are subjective and time‐consuming, and machine learning methods typically require large labeled datasets. This study introduces an innovative zero‐shot learning method that leverages large language models’ (LLMs) image analysis and natural‐language understanding capabilities for accurate road condition assessment. Prompts were designed in alignment with the pavement surface condition index criteria to generate multiple evaluation models, which were then compared against official scores to identify an optimized configuration. Tests conducted using Google Street View imagery indicate that the optimized LLM‐based model achieves a mean absolute error of 1.07 on a 0–10 scale, outperforming expert evaluations. The proposed approach enables rapid, accurate, and consistent assessments without the need for labeled data, demonstrating the transformative role of LLMs in automating infrastructure monitoring and emphasizing the importance of structured prompt engineering for reliable performance.

Abstract Accurate rutting prediction is crucial for traffic safety and road maintenance, enabling timely interventions and cost‐effective strategies. Such prediction remains challenging, especially with limited data across road segments. As traditional methods struggle in the case of data scarcity and complexity, in this study, a Double‐T model is developed by merging TimeGAN and TimesNet. TimeGAN is used to augment the dataset from 2925 to 7578 records, while TimesNet is applied to capture multi‐scale periodic features. The model has achieved determination coefficients ( R 2 ) of 0.939, 0.935, and 0.893 for the training, validation, and testing subsets of the dataset. Increasing the prediction intervals has led to a decline in the model performance. Comparative experiments demonstrate the superior performance of the Double‐T model over conventional regression models, pinpointing that key factors influencing rutting include pavement age, traffic load, pavement thickness and temperature.

The accuracy of road crack detection is susceptible to factors such as illumination variations and tiny crack sizes. Additionally, the loss function in road crack detection is prone to fluctuations during the training process, leading to instability. To address these issues, this paper proposes a multimodal crack detection network to improve detection robustness and accuracy. First, a feature extraction module, named channel–spatial Kolmogorov–Arnold network (CSKAN), is designed specifically to handle challenges posed by illumination variations, enhancing the stability of feature extraction. (KAN is a mathematical framework that decomposes complex multivariate functions into learnable univariate components, enabling adaptive nonlinear modeling.) Second, an efficient multi-scale attention mechanism (Laplacian-adaptive weighted attention) based on the Laplace operator is introduced to address detection difficulties caused by tiny crack sizes and to improve the network's ability to capture small-scale crack features. Finally, a decoupling method for physical information neural networks (DPINNs) is proposed for the first time, ensuring better smoothness in the loss function during the training process. Experimental results demonstrate that our proposed detection framework outperforms seven state-of-the-art detection methods.

Abstract Most post‐disaster damage classifiers perform best when destructive forces leave clear spectral or structural signatures. However, these signatures are often subtle or absent after inundation, where damage may be nonstructural and difficult to detect. Consequently, existing models perform poorly at identifying flood‐related building damage. The model presented in this study, Flood‐DamageSense, addresses this gap as the first deep learning framework purpose‐built for building‐level flood‐damage assessment. The architecture fuses pre‐ and post‐event synthetic aperture radar/interferometric synthetic aperture radar (SAR/InSAR) scenes with very high‐resolution optical basemaps and an inherent flood‐risk layer that encodes long‐term exposure probabilities, guiding the network toward plausibly affected structures even when compositional change is minimal. A multimodal Mamba backbone with a semi‐Siamese encoder and task‐specific decoders jointly predicts (1) graded building‐damage states, (2) floodwater extent, and (3) building footprints. Training and evaluation on Hurricane Harvey (2017) imagery from Harris County, Texas—supported by insurance‐derived property‐damage extents—show a mean F1 improvement of up to 19 percentage points over state‐of‐the‐art baselines, with the largest gains in the frequently misclassified “minor” and “moderate” damage categories. Ablation studies identify the inherent‐risk feature as the single most significant contributor to this performance boost. An end‐to‐end post‐processing pipeline converts pixel‐level outputs to actionable, building‐scale damage maps within minutes of image acquisition. By combining risk‐aware modeling with SAR's all‐weather capability, Flood‐DamageSense delivers faster, finer‐grained, and more reliable flood‐damage intelligence to support post‐disaster decision‐making and resource allocation.