Review of prediction and control for surface settlement caused by shield tunneling based on machine learning

Based on the research on the prediction and control of surface settlement caused by shield tunneling using machine learning methods, the development of choosing input parameters, prediction objectives, prediction algorithms and settlement control methods was discussed．Then, several key problems that need to be solved were summarized, and the prospect of machine learning on the prediction of surface settlement was explored． Previous studies indicate that geometric parameters, geological parameters and operational parameters are the most commonly used as the input parameter, in which geometric parameters mainly choose the tunnel depth, and geological parameters are considered based on the combination of their physical and spatial features．Correlation analysis is generally required before the selection of operational parameters, and the number of input parameters should to be determined carefully． The algorithm for predicting settlement which can consider the time series should be selected, combining with the optimization algorithm to optimize the model performance．The study on settlement control is at the preliminary stage, and further studies are suggested on the recommendation and adjustment methods of tunneling parameters. © 2022 Huazhong University of Science and Technology. All rights reserved.