ABSTRACT
There are many prediction models for shield tunneling parameters, and there is a lack of comparative study on the applicability of various prediction models in the same stratum. This paper combines the three models of SVR, linear regression, and BP neural network in MATLAB software to train and learn the shield tunneling parameter data of the Shuai-Nei section of Hohhot Rail Transit Line 2 and predict the tunneling speed of the shield machine in this section. The results show that the SVR regression model has the lowest prediction accuracy and is unsuitable for predicting shield tunneling speed in this stratum. After the noise reduction of the input tunneling parameters, the BP neural network model and the linear regression model can better predict the tunneling speed of the shield machine in this stratum with the accuracy of the test set being 87%. Applying the BP neural network model and linear regression model to simulate the tunneling parameters of EPB shield in the water-rich round gravel stratum is good. Among them, the accuracy rate of the training set of the BP neural network regression model is 98%, which indicates that the nonlinear mapping ability and generalization application of the BP neural network are excellent.
