Predicting tunnel seepage field is a critical challenge in the construction of underground engineering projects. While traditional analytical solutions and numerical methods struggle with complex geometric boundaries, standard Physics-Informed Neural Networks (PINNs) encounter additional challenges in tunnel seepage problems, including training instability, boundary handling difficulties, and low sampling efficiency. This paper develops an enhanced PINN framework designed specifically for predicting tunnel seepage field by integrating Exponential Moving Average (EMA) weight stabilization, Residual Adaptive Refinement with Distribution (RAR-D) sampling, and Bayesian optimization for collaborative training. The framework introduces a trial function method based on an approximate distance function (ADF) to address the precise handling of circular tunnel boundaries. The results demonstrate that the enhanced PINN framework achieves an exceptional prediction accuracy with an overall average relative error of 5 × 10⁻⁵. More importantly, the method demonstrates excellent practical applicability in data-scarce scenarios, maintaining acceptable prediction accuracy even when monitoring points are severely limited. Bayesian optimization reveals the critical insight that loss weight configuration is more important than network architecture in physics-constrained problems. This study is a systematic application of PINNs to prediction of tunnel seepage field and holds significant value for tunnel construction monitoring and risk assessment.

Silt is often selected as the filling soil in Northwest and North China. “Pot cover effect” or rainfall infiltration are easily to cause localized wetting during the service period of airport. The response characteristics of pavement derived from aircraft loading under wetting condition are the basis to explore the evolution law of pavement damage. To investigate this response characteristics for different parameters of pavement under wetting condition, a self-developed model test system was developed, and the verified numerical model was then established. The interaction mechanisms of pavement slabs, and the influence of wetting on mechanical response in loading area of pavement under different pavement parameters were analyzed. The results showed that the increased vertical deformation depends on water-holding capacity of silt, and the vertical deformation under loading is little affected by pavement parameters. The peak shear stress and the increased peak shear stress derived from wetting at the bottom of pavement layers are largely influenced by elastic modulus of base course. For cement concrete pavement, the increase in load transfer capacity of joints cannot decrease the increase of peak stress, including horizontal stress and shear stress, derived from wetting. The increase in load transfer capacity of joints can decrease the stress concentration in surface layer, while the elastic modulus of surface layer has little influence. For bituminous pavement, optimal parameters of base course can be obtained to decrease the peak stress and the increase of peak stress derived from wetting in subgrade. The research can provide technical basis for the structure design, optimization, and disease treatment of pavement in airport engineering.

Determining the deformation trend of silt subsoil under long-term aircraft loading by conventional numerical methods based on finite elements is challenging and poses several limitations. In this study, a boundary surface model for remolded saturated silt considering the influence of the soil dry density was developed, and an explicit integral algorithm with error control was used to incorporate the model into a user-defined material subroutine that the finite element software (ABAQUS 6.14) could call. In this way, the consolidated undrained dynamic triaxial test of a soil unit was established for simulation and model validation, which corroborated that the model could describe the dynamic properties of the saturated silt. Then, a numerical model of the runway with layered compaction and different compaction degrees was also developed to numerically analyze the deformation of the subsoil under cyclic aircraft loading. The results showed that the subsoil deformation increased continuously with the increase of cycle number. However, the deformation rate decreased gradually, and the silt subsoil deformation remained stable after 50 loading cycles. After the same number of loading cycles, the cumulative plastic deformation of the subsoil model with the overall compaction degree of 94% was smaller than that of the model with layered compaction. It was also shown that different aircraft speeds have minimal effect on the cumulative plastic deformation of the subsoil. Nevertheless, the ultimate cumulative plastic deformation is larger, as the loading duration is longer at low aircraft speeds. It indicates that strictly controlling of the compaction degree within a certain range of load influence is imperative in practical engineering, as it reduces the associated costs.

Based on a large section drilling and blasting excavation project, the dynamic response characteristics of civil air defense tunnels are analyzed by combining field monitoring and numerical simulation. The dynamic response features include particle vibration velocity, main frequency, displacement, and stress, and the stability criterion of the tunnel is analyzed. A safety criterion model based on the ultimate tensile strength of materials is established. The results show that the frequency of the X , Y , and Z directions is mainly distributed in 90-140 Hz. The effective stress increases first and then decreases along the axis of the roadway. The stress near the explosion source is large and the relative reduction is also large. By fitting the relationship between blasting vibration velocity and maximum principal stress, the safe vibration velocity criterion based on tensile strength is obtained, and the safe threshold of vibration velocity is 19.62 cm/s. It can be assumed that blasting does not affect the structure.