Application of a novel pipe roofing method in metro station: Construction disturbance characteristics and prediction model

Abstract

Pipe roofing method, distinguished by its low disturbance, high adaptability, and excellent structural integrity, has emerged as a key solution for shallow-buried, large-span projects. Conventional pipe roofing methods face persistent challenges such as difficult inter-pipe soil removal and complex connection arrangements, which limit their efficiency and applicability. To overcome these limitations, this study proposes a novel pipe roof–concrete slab (PRCS) composite structure and demonstrates its application in the construction of a metro station. The characteristics of surface settlement, tunnel deformation, and surrounding building settlement throughout the construction process, particularly during critical stages, were systematically analysed. The results show that the demolition of the pilot tunnel wall and construction of the roof slab induced the most significant disturbance, resulting in a maximum cumulative surface settlement of 24.1 mm, a maximum building settlement of 7.08 mm, a tunnel crown settlement not exceeding 4 mm, and a clearance convergence below 5 mm. Besides, a surface settlement prediction model based on a VMD–KPCA–CNN–LSTM hybrid neural network was developed, which comprises five steps: data collection, anomaly detection and preprocessing, feature extraction, feature dimensionality reduction, and model prediction. The model achieved outstanding prediction accuracy (RMSE = 0.89 mm, MAE = 0.34 mm, MAPE = 11.21 %, R² = 0.98), significantly outperforming benchmark models such as LSTM, XGBoost, GRU, MLP, and BP. This research provides an innovative construction method and a high-precision predictive model for safety assessment in urban underground space development.