Self-adaptive Adjustment of Shield Tunneling Parameters and Land Subsidence Control Based on Regional Geological Information

Abstract

Objective At present, it has become a trend that the method of machine learning is used to predict land subsidence. However, most of the land subsidence prediction models established based on geological parameters and shield tunneling parameters are simple in terms of method, and can not be used for the actual control of land subsidence. Therefore, a method of self-adaptive adjustment of shield tunneling parameters and land subsidence control based on the regional geological information is proposed. Method Based on Nanjing Metro Line 9, the construction and implementation process of the above-metioned method is described into details. The control method adopts the fusion algorithm based on multiple machine leaning methods and contains two stages. In the first stage of model construction, Model A (land subsidence prediction model) is firstly established to train the overall samples of the construction database, and the optimal algorithm of Model A is obtained by fusion algorithm. Then, Model B (shield tunneling parameter optimization model) is established, and the land subsidence is taken as the control index to screen high quality construction data. The optimal algorithm of Model B is obtained by training and fusion algorithm. In the second stage i.e. construction stage, the relevant parameters of unexcavated sections or sections with large subsidence potential are input into Model B using the optimal algorithm, and the optimized shield tunneling parameters are output, thus realizing the timely adjustment of shield tunneling parameters and the control of land subsidence. The application effect of the above method is verified through simulation calculation based on Nanjing Metro Line 9. Result & Conclusion By comparing the simulated results and the measured data, the subsidence is reduced by the maximum of 21.51 mm after optimization, verifying the effectiveness of the method.

Keywords

• shield tunneling
• geological parameters
• machine learning
• land subsidence