From: A hybrid RNN-GPOD surrogate model for real-time settlement predictions in mechanised tunnelling
Offline stage | |
1. Generate a numerical model representing the tunnelling process within a selected section of the project from time step 1 to n | |
2. Define investigated input parameters and the corresponding range of values | |
3. Run numerical simulations with different input parameters | |
4. Store the numerical results of displacements of the complete surface area of the analysis section | |
5. Provide data of selected monitoring points for the training of a RNN | |
6. Provide data of the complete displacements field for training of the POD-RBF | |
Online stage | |
1. Input: an arbitrary set of input parameters | |
2. Approximate the complete displacement field from time step 1 to n (by POD-RBF) | |
3. Predict the displacements of selected monitoring points for the next time step n+1 (by RNN) | |
4. Predict the displacements field for the next time step n+1 (by GPOD) | |
5. Update the complete displacements field from time step 1 to n+1 | |
6. Repeat steps 3–5 |