Abstract: Cutterhead torque (T) and cutterhead thrust (F) are key parameters to ensure the normal tunneling of shield machine,and their accurate prediction can effectively guide equipment operation.The research datasets are collected from earth pressure balance (EPB) shield machine on line 19 of the Chengdu Metro.By analyzing the characteristics of EPB data,we develop a standard data preprocessing algorithm that includes data segmentation,outlier processing,data filtering and data compilation.Based on Butterworth filter,we establish the prediction model of EPB tunneling parameters by gated recurrent unit (GRU) algorithm,and then comprehensively assess the prediction effect of the model by RMSE and MAE.Results manifest that the proposed model can achieve good prediction for T and F under different geological conditions,and the prediction accuracy of the GRU model in fusion with Butterworth filter is better than that of the unfiltered model.In sandstone formation,the prediction error of the model for T is the smallest,and the RMSE and MAE are 4.91 and 3.86,respectively.The prediction of tunneling parameters based on GRU algorithm can significantly improve the judgment level of shield tunneling state,which is conducive to the optimization and adjustment of construction parameters.

Key words: tunneling parameters prediction, data preprocessing, Butterworth filter, GRU algorithm, earth pressure balance shield machine