基于岭回归分析的TBM净掘进速率预测模型研究
时健1张仕林1,2范作松2孔德森1
TBM Net Advance Rate Prediction Model Based on Ridge Regression Analysis
SHI Jian1ZHANG Shilin1,2FAN Zuosong2KONG Desen1
-
作者信息:1.山东科技大学土木工程与建筑学院， 266590， 青岛
2.青岛地铁集团有限公司第三建设分公司， 266071， 青岛
-
Affiliation:1.College of Civil Engineering and Architecture, Shandong University of Science and Technology, 266590, Qingdao, China
2.The Third Branch of Qingdao Metro Group Co., Ltd., 266071, Qingdao, China
-
关键词:
-
Key words:
-
DOI:10.16037/j.1007-869x.20230594
-
中图分类号/CLCN:U455.44
-
栏目/Col:土建工程
摘要:
［目的］精确预测TBM净掘进速率对城市隧道施工方法选择、工程进度规划和建设成本控制有重要的参考价值。［方法］以青岛地铁1号线双护盾TBM施工为背景，对TBM净掘进速率预测模型的输入变量开展特征选取，研究TBM净掘进速率与各输入变量间的相关关系，并对输入变量进行共线性诊断。建立了基于岭回归分析的TBM净掘进速率预测模型（以下简称“岭回归预测模型”），并验证了该模型的预测效果。［结果及结论］TBM净掘进速率与岩石单轴抗压强度、完整性系数、刀盘推力和刀盘转速之间呈正相关，且相关程度较为显著；TBM净掘进速率预测模型的输入变量间存在一定程度的多重共线性，影响偏回归系数的取值，使得部分输入变量偏回归系数的检验结果失去统计学意义；岭回归预测模型预测精度稍差，但偏回归系数估值趋于合理，使得该模型稳定性更高；岭回归预测模型的绝对误差在5 mm/min以内，能够满足工程预测的需要。
Abstracts:
［Objective］ Accurate prediction of TBM (tunnel boring machine) net advance rate is of significant reference value for selecting urban tunneling methods, planning construction schedules, and controlling construction costs.［Method］ Taking the dual-shield TBM construction of Qingdao Metro Line 1 as the background, feature selection is conducted for the input variables of the TBM net advance rate prediction model. The correlations between the TBM net advance rate and various input variables are analyzed, and collinearity diagnostics are performed on the input variables. A TBM net advance rate prediction model based on ridge regression analysis (hereinafter referred to as the ′ridge regression prediction model′) is established, and its predictive performance is validated.［Result & Conclusion］ The TBM net advance rate shows a positive and relatively strong correlation with uniaxial compressive strength of the rock, rock integrity coefficient, cutterhead thrust, and cutterhead rotation speed. There exists a certain degree of multicollinearity among the input variables of the prediction model, which affects the estimation of partial regression coefficients, rendering some of them statistically insignificant. Although the prediction accuracy of the ridge regression prediction model is slightly lower, its estimated partial regression coefficients tend to be more reasonable, resulting in higher model stability. The absolute prediction error of the ridge regression prediction model is within 5 mm/min, which meets the requirements of engineering prediction.
过刊浏览
资料下载
友情链接