Quantitative Analysis of Seasonal Uncertainty of Metro Tunnel’s Long-Term Longitudinal Settlement via Hierarchy Bayesian Network

Abstract

The control of long-term longitudinal settlement of metro tunnels has become a critical problem in the daily maintenance process. However, describing settlement behaviors accurately is a tough task because amount of uncertainty lies in the input parameters, models and even the monitoring data. The Hierarchy Bayesian Network is a dependable method to analyze the variability and randomness of model. In this paper, we utilized Hierarchy Bayesian Network to deal with the model bias caused by seasonal changes and Marko chain Monte Carlo simulation to conduct the numerical integration and parameters’ estimation. The data from Shanghai Metro Line 1 was employed to calibrate models. The results show that the maximum settlement difference of tunnel sections between the first and the second half year reached nearly 10 mm, due to seasonal variance. The settlement in different regions follows the same distribution with different mean values, which reflect the inter-region variability. The distribution parameters in every six months deviate from the benchmark with similar absolute values, reflecting the seasonal uncertainty. The model bias gradually reduces over time, especially after year 2011.