Research On Deformation Prediction Of The Three-lanes Tunnel Based On Neural Network In Soft Area

In the process of tunnel construction with NATM (New Austrian Tunneling Method) , measuring the deformation of surrounding rock of the tunnel is the key content which can estimate whether the surrounding rocks lose stabilization or support methods are reasonable. In many former engineering projects, the accidents were mainly caused by impertinent support order and support parameters due to an incomplete mastery of the deformation of surrounding rock of the tunnel. Therefore how to predict the value of the deformation of the surrounding rock, which can be useful in making construction decisions, is an important issue in tunnel construction with NATM. This paper takes the building tunnels in HangQian Expressway in Zhejiang Province as the research background. Firstly, based on the special characters of these tunnels (soft rock surrounding & three lanes) and the construction method, the problems happening frequently in this project and the methods to solve these problems were discussed. On the basis of the monitoring of these tunnels, the prediction method of the deformation of soft rock tunnel was researched. A new BP model considering the actual factors which can do effects on the deformation of the tunnel is proposed. Then, based on BP model, the software which can be used to forecast the displacement of tunnel according to some data of monitoring was coded by VB. The application results of this software in the tunnel monitoring indicate that it can benefit the safe construction of these tunnels and the stabilization of tunnel. The research work of this paper could be summarized as follows:(1)Based on the construction of the tunnels in HangQian Expressway, this paper firstly analyzes the geologic and engineering characters of soft rock, and then comprehensively analyzes the major influence factors of the stabilization of soft rock tunnel.(2)According to characters of soft rock tunnel, the construction measures and mechanisms of excavating and supporting are deeply discussed. Because of the heavy section of three lanes tunnel and poor geological conditions, problems like deformation, cracking and water-leakage often occur in supporting of soft rock tunnel. Therefore, it should close inverted arch and reinforce the support as soon as possible to assure safety of construction of tunnel. Seep, crushed zone, limestone cave, slope slide and landslide frequently occurring in the excavation of tunnel were discussed, and the means to solve them were enumerated. Several kinds of methods to enhance the support of tunnel could provide experiences to some other projects.(3)According to the theory of neural network, the monitoring data was considered in analyzing the factors which have influence on the deformation of the tunnel, then based on BP neural network model improvement method, adds the momentum item to the model to enhance speed of convergence of the model, and has established the corresponding neural network forecast program. Abundant monitoring data had been taken as the neural network study sample, and then the forecast deformation was contrast to the practical data which measured in the next step. The number of the study samples and the actual situation which have influence on the BP neural network were also discussed.(4) The model for prediction is applied to the tunnels of HangQian Expressway, And the deformation of rock surrounding was predicted based on the 3DTOS software, the predicting result provides some valuable suggestion for construction, it is proved that this model have some significance in tunnel engineering.