Aiming at some stress-induced failure phenomena in surrounding rock that occur during the construction of super-long and deep-buried tunnels, a method for estimating the in situ stress in the tunnels based on multivariate information integration is proposed, which uses a small amount of in situ stress measurement, stereographic projection technology, and a numerical simulation method. Firstly, by conducting a macroscopic analysis of the regional geological structure, topography, and pre-excavated small tunnels (such as exploration of adits and pilot tunnels), the strength of the tectonic stress field and the orientation of the principal stresses in the tunnel sections are preliminarily determined. Secondly, the reliability of the in situ stress measurement data were analyzed using full-space stereographic projection and the plane stress projection method. Then, some representative measurement points that reflected the distribution characteristics of in situ stress in the project area, on the whole, were determined. Thirdly, the finite difference (FDM) and multiple regression analysis (MRA) methods were used to inverse the in situ stress field in the project area. The proposed method was applied to a super-long and deep-buried tunnel project in Qinling, and the in situ stress distribution characteristics of the tunnel sections at different mileages were obtained. The results show that both the calculated principal stress values and the azimuth angle of the maximum horizontal principal stress are in good agreement with the measured ones, indicating that the method used in this study is reasonable. Finally, the typical surrounding rock failure phenomena encountered during the excavation of the project were investigated, and targeted treatment measures were proposed. The research results can provide references for support design and disaster management of surrounding rock in deep-buried long tunnels.

In response to the problem of significant post-construction settlement that may occur in a motor racing circuit (MRC), two representative composite foundation testing areas, PHC pile (pre-tensioned spun high-strength concrete pile) and CFG pile (cement fly ash gravel pile), were selected for field tests to obtain the deformation law of pile–soil. Then, finite element numerical simulation was used to carry out back analysis on the geological mechanical parameters of the testing areas. The results showed that the error of soil settlement between the piles in the PHC pile and CFG pile testing areas were 8.2% and 9.6%, respectively, with good inversion precision. The obtained geological mechanical parameters can be used to predict the settlement of the rest of the MRC. On this basis, a finite element numerical model was constructed to analyze the bearing and deformation characteristics of the foundation of the MRC under five types of working conditions that may cause significant post-construction settlement. It showed that the settlement of the embankment was large in the middle and small on both sides after the consolidation of the embankment. The maximum settlement was about 27.0 mm, and the maximum longitudinal uneven settlement ratio of the embankment was 1.3/4000. The axial force of piles in the PHC pile and CFG pile composite foundations increased first and then decreased with depth. The maximum bending moment was located at the foot of slopes or at the boundary of strata, which was relatively small in the middle of the embankment. The deformation of the embankment and the bearing capacity of the piles could meet engineering requirements. This study has certain guiding significance for the design and construction of similar pile-net composite foundations.

Current methods for predicting deep excavation deformation suffer from insufficient accuracy and limited generalization capability. Moreover, the applicability of these methods to different types of monitoring data also requires in-depth analysis. To address this, a machine learning-based prediction model, i.e., the VMD-GWO-CNN model, integrating Variational Mode Decomposition (VMD), the Grey Wolf Optimizer (GWO), and the Convolutional Neural Network (CNN), is proposed to predict various types of monitoring data. The GWO algorithm optimizes both the key parameters of VMD and the hyperparameters of the CNN. The optimized CNN model predicts each subsequence decomposed by VMD, and the final prediction is obtained by superimposing these results. Furthermore, the prediction performance of the proposed model is evaluated against the LSTM, CNN, and GWO-CNN models using four metrics (RMSE, MAE, MAPE, R²). The results indicate that all four algorithms possess effective predictive capability for the monitoring data, in which the VMD-GWO-CNN model demonstrates the best performance across all metrics. Specifically, its RMSE for surface settlement prediction is reduced by 59.2%, 34.1%, and 33.0% compared to the LSTM, CNN, and GWO-CNN models, respectively. Moreover, the VMD-GWO-CNN model exhibits strong predictive performance for deformation in slope engineering and subgrade engineering, demonstrating its good applicability across different geotechnical engineering. The findings provide a scientific basis for safe excavation construction and contribute to efficient and rapid execution of foundation pit projects.