Research on Deformation Prediction of Underground Structure Deep Foundation Pit Based on CNN-LSTM Hybrid Model

DOI

10.2991/978-94-6463-726-7_27How to use a DOI?

Keywords

foundation pit engineering of underground structure; pearson correlation analysis; multi-source heterogeneous data; spatio-temporal correlation effect; CNN-LSTM hybrid neural network model

Abstract

Reliable prediction of deformation during deep foundation pit excavation is an important guarantee for construction safety. In order to improve the calculation accuracy and stability of deformation prediction in foundation pit engineering, a CNN-LSTM combined neural network model based on multi-source heterogeneous monitoring data is proposed. Firstly, Pearson correlation analysis is carried out for multi-source isomorphic real-time monitoring data and multi-source heterogeneous monitoring data, and the prediction results of single-feature and multi-feature based on LSTM network model are compared. There are strong spatiotemporal characteristics among multi-source heterogeneous monitoring data, and the prediction accuracy of the multi-feature LSTM network model based on the multi-source heterogeneous data is higher. The important correlations with the top horizontal deformation are as follows: deep horizontal deformation > top vertical deformation > surface settlement > groundwater level > anchor cable tension. Then, the CNN-LSTM hybrid model was constructed that CNN network and LSTM network is used to extract the spatial features and temporal features of monitoring data. The results show that the prediction accuracy of MAE, MAPE, RMSE and R2 of CNN-LSTM hybrid model considering temporal and spatial correlation are 0.1159~0.1587, 0.0375~0.0463, 0.1256~0.1636 and 0.8543~0.9269, respectively. Compared with the single LSTM model, prediction accuracy of CNN-LSTM hybrid model is higher, which only considers the time correlation. All in all, the deformation characteristics of foundation pit engineering can be better predicted by the CNN-LSTM hybrid model.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Xinna Liu
AU  - Jie Dong
AU  - Peng Wang
AU  - Wei Zhang
AU  - Penghao Jiang
PY  - 2025
DA  - 2025/06/13
TI  - Research on Deformation Prediction of Underground Structure Deep Foundation Pit Based on CNN-LSTM Hybrid Model
BT  - Proceedings of the 2024 6th International Conference on Hydraulic, Civil and Construction Engineering (HCCE 2024)
PB  - Atlantis Press
SP  - 266
EP  - 278
SN  - 2589-4943
UR  - https://doi.org/10.2991/978-94-6463-726-7_27
DO  - 10.2991/978-94-6463-726-7_27
ID  - Liu2025
ER  -