富水圆砾地层土压平衡盾构掘进速度预测模型对比研究

Comparative Study on Prediction Models of Excavation Speed of Soil Pressure Balance Shield in Water-rich Gravel Stratum

目前,有多个盾构掘进参数预测模型,但同一地层各预测模型适用性对比尚未见报道。依托内蒙古呼和浩特市地铁2号线帅家营站—内大南校区站区间盾构隧道工程,基于该工程掘进参数,采用SVR回归模型、线性回归模型、BP神经网络回归模型进行了训练和学习。结果表明:3种预测模型中,SVR回归模型预测的盾构掘进速度与真实值相差较大;通过对输入的掘进参数进行降噪处理后,BP神经网络回归模型和线性回归模型均能较好地预测出该地层盾构机的掘进速度,2种模型的测试集预测准确率皆为87%,而BP神经网络回归模型的训练集预测准确率高达98%,说明BP神经网络回归模型在经过学习和训练后有较强的预测盾构机掘进速度的能力。

There are many kinds of prediction models for shield tunneling parameters. However, the applicability comparison of each prediction model at the same stratum has not been reported. Based on the interval shield tunnel project between Shuaijiaying Station and Inner Mongolia University South Campus Station of Metro Line 2 in Hohhot,Inner Mongolia and the tunneling parameters, SVR regression model, linear regression model and BP neural network regression model were used to training and learning. The results show that the shield tunneling speed predicted by SVR regression model is quite different from the actual one among the three prediction models. After the input tunneling parameters were denoised, both the BP neural network regression model and the linear regression model can better predict the tunneling speed of the shield machine in this stratum whose prediction accuracy of the validation set are 87%. The prediction accuracy of the BP neural network regression model is as high as 98% which indicates the BP neural network regression model can better predict the tunneling speed of shield machines after being learned and trained.