Research On The Prediction Method For Ground Settlement Of Metro Station In PBA Method

China will increase about 3000 km mileage of metro transit operation in the next five years, which brings new opportunities to development of the metro. As the construction of metro always in the center of cities, most of metro projects have to use underground excavation method. In the process of metro construction, it is inevitably that metro station encountering the city municipal pipe network, underground building and structure, the road that the metro impact on. How to effectively control the ground settlement during the construction in PBA method is the hotspot and difficulty in our study. This paper discussing the problem in the context of Beijing metro construction in PBA method, mainly research as the following aspects:1. For the analysis of monitor data, ground settlement curves are drawn and the ground settlement data of monitoring points are organized.According to characteristics of the PBA method and the engineering, this paper analyzes the reasons of ground settlement and put forward the measures to reduce the ground settlement.2. Midas/GTS software is used to establish a three-dimensional finite element model that simulate the whole process of PBA method construction.In the paper, four typical stages are discussed to analysis each stage of ground settlement value and the proportion. The pilot tunnel stage is the main stage that caused the ground settlement.3. By numerical method for simulating of six different sequence project of pilot tunnel excavation,Analytic Hierarchy is used to establish hierarchy structure of the system and evaluation index matrix. Optimal excavation order of the pilot tunnel excavation sequence is lower-upper, edge-central.4. The station ground settlement value were fitted and predicted by the time series model. Analyzing the calculation process of ARMA and Artificial neural network models and combining the monitor data, this paper use two models to forecast ground settlement by Matlab mathematical software. Comparing the prediction precision of the two models, artificial neural network has better prediction accuracy and adaptability.