关键词: 盾构隧道, 施工参数, 地表沉降, 推进速度, 随机森林, 多目标粒子群优化

Abstract: In urban building areas, accurate prediction and control of surface settlement caused by earth pressure balance (EPB) shield tunneling is essential to construction safety and environmental impacts. Thus, the authors propose a surface settlement prediction method integrating random forest (RF), genetic algorithms (GA), and the multi-objective particle swarm optimization algorithm based on crowding distance (MOPSO-CD) to optimize critical construction parameters. First, data collected from EPB shield tunneling projects are preprocessed and analyzed to establish a comprehensive database. A RF regression model is then trained on this database, with GA used to optimize hyperparameters, resulting in an intelligent surface settlement prediction model. Feature importance analysis identifies construction parameters with significant influence on settlement prediction, reducing the parameter set to eight key factors for constructing fitness functions. Constraints for each parameter are determined, and the MOPSO-CD algorithm is applied to achieve dual optimization objectives: minimizing surface settlement and maximizing driving speed. The proposed settlement prediction model achieves an R² value of up to 0.937 and a root mean square error of 11.7 mm, enabling real-time, accurate surface settlement predictions during shield tunneling. The multi-objective optimization model provides optimal parameter ranges, reducing average surface settlement to －4.28 mm, only 3.5% of the original value, while maintaining an average driving speed at 58% of the initial level.

Key words: shield tunnel, construction parameters, surface settlement, driving speed, random forest, multi-objective particle swarm optimization