Method of Ground Settlement Prediction in Urban Tunnel Construction Based on ARMA

Abstract

Deformation of surrounding rock during tunnel excavation could be regarded as a time sequence. In this paper, the auto regressive moving average (ARMA) process of tunnel deformation prediction during construction, which can, to some extent, modulate model parameters according to the input and output data, adjust them to its optimal values in some statistical conditions through the iterative algorithm. Firstly, the raw monitoring data is dealt with by difference operation (d=1) and the stationary error time series is obtained. Secondly, the randomicity and the stabilization of the error series have been analyzed to confirm the feasibility of ARMA model. Both the autocorrelation function (ACF) and partial autocorrelation function (PACF) of the error series are tested, and the order of model ARMA is obtained based on the AIC criterion. Finally, the method of least square is adopted for parameter estimation. As a test, this modeling is used to predict the surface settlement of a Shenzhen metro line 2 of China. The results of engineering case indicate that ARMA is reliable in deformation prediction. In addition, the new-information model idea has been put into the model to achieve the real-time forecast of settlement.