Short-term tunnel-settlement prediction based on Bayesian wavelet: a probability analysis method

Abstract

As urbanization accelerates, the metro has become an important means of transportation. Considering the safety problems caused by metro construction, ground settlement needs to be monitored and predicted regularly, especially when a new metro line crosses an existing one. In this paper, we propose a settlement-probability prediction model with a Bayesian emulator (BE) based on the Gaussian prior (GP), that is, a GPBE. In addition, considering the distortion characteristics of monitoring data, the data is denoised using wavelet decomposition (WD), so the final prediction model is WD-GPBE. In particular, the effects of different prediction ratios and moving windows on prediction performance are explored, and the optimal number of moving windows is determined. In addition, the predicted value for GPBE based on the original data is compared with the predicted value for WD-GPBE based on the denoised data. One year of settlement-monitoring data collected by a structural health monitoring (SHM) system installed on the Nanjing Metro is used to demonstrate the effectiveness of WD-GPBE and GPBE for predicting settlement.