基于BO-XGBoost的盾构掘进地表变形预测

SURFACE DEFORMATION PREDICTION OF SHIELD TUNNELING BASED ON BAYESIAN OPTIMIZATION XGBOOST

在城市地铁建设中,土压平衡盾构掘进引起的地表变形大小受到众多因素的影响,因此往往难以准确预测。针对这一问题,提出了贝叶斯优化(BO)的极端梯度提升(XGBoost)预测模型。模型输入考虑了地质参数、施工参数与隧道几何参数,并通过特征选择分析了各特征的重要性,同时降低了输入参数的维度。采用贝叶斯优化算法确定了最优超参数,并将预测结果与手动调参的XGBoost、随机森林(RF)、支持向量机(SVM)模型作对比。结果表明集成算法在预测中比传统SVM表现更好,贝叶斯优化算法提升了XGBoost模型的预测性能;隧道埋深与直径对XGBoost预测模型影响最大,隧道上方淤泥质土层、黏土层、粉砂土层、风化岩层对模型的影响依次降低,刀盘转速和平均土仓压力的相对重要度也较高。

In urban subway construction,the surface deformation caused by EPB shield tunneling was affected by many factors,so it was always difficult to predict accurately.To solve this problem,a Xtreme Gradient Boosting(XGBoost)prediction model based on Bayesian optimization was proposed.Geological parameters,construction parameters,and tunnel geometric parameters were considered in the model input.The importance of each feature was analyzed through feature selection,and the dimension of input parameters was reduced.The optimal hyperparameters were determined by Bayesian optimization algorithm,and the prediction results were compared with the manually adjusted XGBoost,Random Forest(RF),and Support Vector Machines(SVM)models.The results show that the integrated algorithm performs better than the traditional SVM in prediction.Bayesian Optimization algorithm improves the prediction performance of XGBoost model.The buried depth and diameter of the tunnel have the greatest impact on the XGBoost prediction model,the influence of muddy soil layer,clay layer,silt soil layer,and weathered rock layer above the tunnel on the model decreases in turn;the relative importance of cutter head speed and average chamber pressure is also high.