Research on the deformation prediction model of surrounding rock based on SSA-VMD-GRU
Surrounding rock deformation prediction can provide decision-makers with future deformation information, which enhances construction safety. Aiming at the problems of nonlinearity, high dynamics, and low prediction accuracy of surrounding rock deformation, this article proposes a time series predict...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
De Gruyter
2025-07-01
|
| Series: | Open Geosciences |
| Subjects: | |
| Online Access: | https://doi.org/10.1515/geo-2025-0842 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Surrounding rock deformation prediction can provide decision-makers with future deformation information, which enhances construction safety. Aiming at the problems of nonlinearity, high dynamics, and low prediction accuracy of surrounding rock deformation, this article proposes a time series prediction method based on the sparrow search algorithm (SSA)-variational mode decomposition (VMD)-gated recurrent unit (GRU) deep learning model. First, aiming at the problem that parameters in VMD are difficult to determine, a VMD evaluation standard in the field of surrounding rock deformation is proposed, and the SSA is used to find the optimal combination of decomposition parameters under this standard. Second, the VMD with optimal parameters is used to decompose the surrounding rock deformation series into the trend and random term displacement. Finally, a GRU neural network with memory and feedback capabilities is built to predict the displacement components separately, and superimposed reconstruction is performed to obtain the final predicted values. The proposed method is applied to the prediction of peripheral rock deformation in a tunnel and compared with the traditional model. The results show that the proposed SSA-VMD-GRU model can accurately predict rock deformation, and the prediction accuracy of each displacement component is high, and the error is small. The final predicted R
2 of the two groups of deformation sequences were 0.9265 and 0.9119, root mean square error were 0.1262 and 0.1243 mm, and mean absolute error were 0.1101 and 0.1062 mm, respectively. This research provides an efficient and reliable solution for the prediction of surrounding rock deformation. |
|---|---|
| ISSN: | 2391-5447 |