【摘要】 为实现对地铁车致地面振动的高效预测，建立车辆-轨道-隧道-土体三维耦合动力学模型，通过模型计算搭建了涵盖典型车速、轴重、轨道类型、隧道埋深、土体剪切波速和距线路中心距离取值的数据库，分别采用K近邻(KNearest Neighbors,KNN)、随机森林(Random Forest,RF)、极限梯度提升树(Extreme Gradient Boosting,XGBoost)以及轻量梯度提升机(Light Gradient Boosting Machine,LightGBM)建立地面振动预测模型，对比分析4种模型的预测效果，验证LightGBM模型的泛化能力。结果表明：现场实测与数值模型计算所得地面最大Z振级分别为68.7、70.3 dB，相对误差仅为2.3%，验证数值模型的可靠性；4种模型中，LightGBM模型预测效果最佳，验证集标准差为0.034，决定系数为0.91;LightGBM模型泛化能力良好，地面振动实测数据与模型预测结果之间吻合度高，最大Z振级相对误差分别为3.7%、4.3%。此预测方法具有较高的可靠性，可用于实际地铁车致地面振动预测。更多还原

【Abstract】 To achieve efficient prediction of ground vibration induced by metro trains operation, a three-dimensional dynamic model of the vehicle-track-tunnel-soil coupled system was established, and a database covering typical values of train speed, axle weight, track type, tunnel depth, soil shear wave speed and distance from the line center was constructed through the model calculations. The ground vibration prediction models were established using K-Nearest Neighbors(KNN),Random Forest(RF), Extreme Gradient Boosting(XGBoost) and Light Gradient Boosting Machine(LightGBM)respectively. The predictive performance of the four models was compared. Finally the generalization capability of the LightGBM model was validated. The results show that: the maximum Z-levels of ground vibration obtained from the field measurement and numerical model calculation are 68.7 dB and 70.3 dB, respectively, with a relative error of only 2.3 %,confirming the reliability of the numerical model. Among the four models, the LightGBM model has the best prediction effect, and the standard deviation of the validation set was 0.034, and the coefficient of determination was 0.91; The LightGBM model exhibited strong generalization capability, with higher consistency between field-measured ground vibration data and model predictions, and the relative errors are 3.7 % and 4.3 % for the maximum Z-levels. This prediction method is highly reliable and can be applied to predict ground vibration induced by actual metro trains.更多还原