基于随机森林优化模型的土压平衡矩形顶管土仓压力预测  

Prediction of soil chamber pressure of rectangular pipe jacking with earth pressure balance based on random forest optimization model

作　　者：许有俊 单思远[1] 张博华 李选 XYU Youjun;SHAN Siyuan;ZHANG Bohua;LI Xuan(School of Civil Engineering,Inner Mongolia University of Science&Technology,Baotou 014010,China;Academician Workstation of Mine Safety and Underground Engineering,Inner Mongolia University of Science&Technology,Baotou 014010,China;Research Center for Urban Underground Engineering of Universities in Inner Mongolia Autonomous Region,Baotou 014010,China)

机构地区：[1]内蒙古科技大学土木工程学院,内蒙古包头014010 [2]内蒙古科技大学矿山安全与地下工程院士工作站,内蒙古包头014010 [3]内蒙古自治区高校城市地下工程技术研究中心,内蒙古包头014010

出　　处：《内蒙古科技大学学报》2024年第4期331-337,共7页Journal of Inner Mongolia University of Science and Technology

摘　　要：为解决土压平衡矩形顶管施工过程中土仓压力的调节完全依赖施工人员经验的问题,基于粒子群优化后的随机森林算法提出了土压平衡矩形顶管土仓压力的预测模型,通过施工参数间关联性的分析,确定对土仓压力影响较大的参数为螺机转速、刀盘转速、顶力和砾砂层厚度,将这些参数作为输入,土仓压力作为输出,建立土仓压力预测模型,并结合传统的支持向量机(SVM)预测模型进行对比分析。结果表明:PSO-RF模型相对于SVM模型预测的均方根误差减小了48.30%,且模型的平均误差更小,展现出了更强的学习预测能力,可以更好的应用于土压平衡矩形顶管的掘进参数预测中。To address the issue of soil chamber pressure adjustment during the construction of soil pressure balance rectangular pipe jacking,which is completely dependent on the experience of the construction personnel,a prediction model of soil chamber pressure for soil pressure balance rectangular pipe jacking is proposed based on the Random Forest Algorithm after Particle Swarm Optimization.A prediction model for the earth chamber pressure of earth pressure balance rectangular shield tunneling is proposed based on the Random Forest algorithm optimized by Particle Swarm Optimization(PSO).By analyzing the correlation between construction parameters that have greater influence on the soil chamber pressure are identified as the rotational speed of the screw machine,cutter rotational speed,the jacking force,and the thickness of gravel and sand layer.With these parameters as inputs and soil pressure as outputs,the soil pressure prediction model was established and compared with the traditional support vector machine(SVM)prediction model.The results show that the root-mean-square error of the PSO-RF model is reduced by 48.3%,compared with that of the SVM model,and the average error of the PSO-RF model is smaller,which shows stronger learning prediction ability and can be better applied to the prediction of digging parameters of the soil-pressure-balanced rectangular jacking pipe.

关 键 词：矩形顶管 随机森林 土仓压力 预测模型 

分 类 号：TU921[建筑科学—建筑设计及理论]