A deep learning-based method for predicting surface settlement induced by shield tunnel construction
The nonlinear feature relationships and bidirectional time-series information of data can not be obtained at the same time in the existing methods for predicting surface settlement triggered by shield tunnel construction.A deep learning-based method(CNN-BiLSTM-SA)for maximum surface settlement prediction was proposed by fusing convolutional neural network(CNN).Bidirectional long and short-term memory(BiLSTM)and self-attention(SA).In CNN-BiLSTM-SA,CNN was first used to analyse the nonlinear feature relationships among the network input data,and BiLSTM network was used to extract the bi-directional time series information of the input data.And then SA was introduced to assign corresponding weights to the features extracted by CNN to effectively capture the key information in the time series.Finally,the final surface settlement prediction results were output through the fully connected layer.The surface settlement dataset was constructed based on the Hunan Wanjiali Road power shield tunnel project,and the four models,ANN,RNN,LSTM and BiLSTM,were selected to carry out comparative analysis experiments.The results show that the four evaluation indexes of mean absolute error(MAE),root mean square error(RMSE),determination coefficient(R2),and mean absolute percentage error(MAPE)of CNN-BiLSTM-SA are optimal,indicating that the proposed model has better surface settlement.
万方数据
维普
