Chengshi guidao jiaotong yanjiu (Aug 2024)
Land Subsidence Prediction Model of Rail Transit Based on High-frequency Combination Segment-Gene Expression Programming Algorithm
Abstract
Objective Land subsidence prediction and control is one of the most concerned issues in rail transit shield tunnel construction. In order to solve the complex and poor interpretable problem of the model expression in the existing land subsidence prediction and control, an interpretable model that is concise, clear，and capable of describing complex problems is needed. GEP (gene expression programming) algorithm provides this possibility, so it is necessary to study in depth the rail transit land subsidence prediction model based on HFS (high frequency segment)-GEP algorithm. Method Based on the shield tunnel project of a certain shield tunnel section in Hangzhou-Shaoxing Intercity Railway, parameters such as earth chamber pressure, cutterhead torque, cutterhead speed, advancing speed, total thrust, tunnel buried depth and shield tail grouting amount during shield construction are selected as key input construction parameters, with land subsidence as the output construction parameter. Through alternative formula set screening and HFS selection, a land subsidence prediction model of rail transit based on HFS-GEP algorithm is established. Key construction parameters of the 180th-210th section are optimized and adjusted with the model, and effect of the shield construction parameters change on the final land subsidence is analyzed. Result & Conclusion The land subsidence prediction model of rail transit based on HFS-GEP algorithm can reflect the explicit relationship between shield construction parameters and final land subsidence. Compared with the traditional GEP algorithm model, this model has higher accuracy, simpler structure and faster convergence. By optimizing and adjusting the key construction parameters of the shield, the final subsidence of the 180th-210th section can be controlled within 10 mm.
Keywords
