To address the insufficient bearing capacity of existing slab culverts under increased fill heights in highway widening projects, a reinforcement method integrating an additional deck slab with a grouted steel pipe pile-supported composite foundation is proposed. A segmented elastic foundation beam model is developed to derive the analytical solution for slab deflection and calculate the load acting on the culvert roof. A three-dimensional finite element model is established in Abaqus, in which the soil is modeled using the Mohr–Coulomb elastoplastic constitutive model to investigate the effects of pile arrangement, pile diameter, pile length, pile spacing, and fill height on reinforcement performance. The results show that culvert displacement and stress decrease with reduced pile spacing and increased pile diameter and pile length, while the best reinforcement performance among the investigated cases is achieved when the piles extend 4 m below the culvert base slab Field tests from the Beijing–Hong Kong–Macao Expressway widening project demonstrate the applicability of the proposed reinforcement method and show that pressure grouting increases the coefficient of subgrade reaction by 7–10%. Based on theoretical and field investigations, recommended reinforcement parameters are proposed for a representative culvert. The proposed method provides an efficient solution for culvert reinforcement in highway widening projects.

The widespread application of the drill-and-blast method in tunnel construction generates instantaneous high-concentration dust, posing severe threats to workers’ health and safety. However, existing contact-based monitoring techniques, such as filter membrane weighing and light-scattering sensors, are hindered by operational complexity, maintenance challenges, data latency, and an inability to capture whole-field distribution. To address these limitations, this study conducted in situ experiments to construct a dust image dataset containing rich “real-world noise”. Analysis revealed significant variations in visibility, contrast, and light-scattering intensity across different concentration levels, establishing a physical basis for visual feature extraction. Consequently, a dust concentration prediction method based on transfer learning and CNN was proposed. Comparative training of six network models, including ResNet50, demonstrated that VGG16 achieved the best performance with an R² of 0.9684, verifying the method’s feasibility. Furthermore, hyperparameter optimization (learning rate: 0.001; batch size: 32; dataset split: 8:1:1) was shown to further enhance prediction accuracy.