基于机器学习的EPB盾构土仓压力预测方法研究  

Prediction method of in-chamber earth pressure of EPB shield based on machine learning

作　　者：王伟 王兴 徐亮[1] 王美艳 WANG Wei;WANG Xing;XU Liang;WANG Meiyan(School of Civil Engineering,Changchun Institute of Technology,Changchun 130012,China;College of Construction Engineering,Jilin University,Changchun 130026,China)

机构地区：[1]长春工程学院土木工程学院,吉林长春130012 [2]吉林大学建设工程学院,吉林长春130026

出　　处：《人民长江》2023年第12期241-247,266,共8页Yangtze River

基　　金：吉林省科技发展计划项目(20220203058SF);中交路桥建设有限公司科技研发项目(ZJLJ-2018-44)。

摘　　要：为了避免因土压平衡盾构(EPB)土仓压力失衡造成的掌子面失稳、地层缺失和地表沉降等后果,将GA算法嵌入到PSO算法进行参数优化,结合灰色理论建立基于灰色最小二乘支持向量机的土仓压力预测模型,并以实际工程为案例进行了验证。研究结果表明:GA-PSO-GLSSVM土仓压力预测模型将总推力、刀盘扭矩、推进速度、螺旋机转速、螺旋机扭矩、注浆量6种掘进参数作为输入集,综合考虑了盾构掘进参数之间的相互影响,使预测模型更符合实际;该预测模型综合了GA算法的全局搜索能力、PSO算法的快速收敛能力和GM模型抗扰动能力,提高了复杂地层EPB土仓压力预测的精度;与其他预测模型预测结果的对比表明,GA-PSO-GLSSVM模型预测结果的均方根误差(RMSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)、可决系数(R^(2))明显优于其他模型,预测结果拟合优度和精确度更高,对砂卵石这类复杂地层中EPB盾构土仓压力的预测具有显著的适用性。研究结果可为砂卵石地层EPB盾构施工掘进参数控制提供参考。To avoid consequences caused by the unbalance of earth pressure of the in-chamber of earth pressure balance(EPB)shields,such as instability of tunnel face,the loss of stratum and surface settlement,the GA algorithm was embedded into the PSO algorithm for parameter optimization,and a prediction model of chamber earth pressure based on grey least squares support vector machine was established combined with grey theory,which was verified by an actual project.Research indicates that the GA-PSO-GLSSVM chamber earth pressure prediction model takes six driving parameters as input set,including total thrust,cutter torque,propulsion speed,screw speed,screw torque and grouting amount,and comprehensively considers the interaction between shield driving parameters,making the prediction model more realistic.The prediction model integrates the global search ability of the GA algorithm,the fast convergence of the PSO algorithm and the anti-disturbance ability of the GM model,so improves the prediction accuracy of EPB chamber earth pressure in complex strata.Compared with the prediction results of other prediction models,the root mean square error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE)and determination coefficient(R^(2))of the GA-PSO-GLSSVM model is significantly better.The prediction results have higher goodness of fit and accuracy,and have significant applicability to the prediction of EPB shield chamber earth pressure in sand and pebble complex strata.The research results can provide a reference for EPB shield tunneling parameter control in the sandy pebble stratum.

关 键 词：砂卵石地层 土压平衡盾构 土仓压力 GLSSVM PSO GA 

分 类 号：U455.43[建筑科学—桥梁与隧道工程]