Figure 2

Flowchart of the proposed ANN deformation prediction model. There are 5 steps in the procedure of the ANN model in prediction deformation. Step 1, perform wavelet decomposition, decompose the deformation time series into two components. Step 2, use Copula method to analyze the correlation of the two components with different variables, and determine and optimize the input variables of the ANN model. Step 3, input the obtained data of the input variables into the CNN and LSTM neural networks, and establish CNN and LSTM prediction models by training these neural networks. Step 4, use the CNN and LSTM to predict the two components for the future. Step 5, add the two components to obtain the final deformation prediction values.