A Hybrid Wavelet Transform and Bidirectional GRU Approach for TBM Advance Speed Prediction in Super-Large-Diameter Tunneling
Accurate prediction of tunnel boring machine (TBM) advance speed (AS) is important for construction decision-making, because this parameter governs both excavation safety and efficiency. This study examines a hybrid framework that combines wavelet transform (WT) with a bidirectional gated recurrent unit (Bi-GRU) network for AS prediction in super-large-diameter tunneling. WT is used to suppress high-frequency measurement noise in the raw monitoring series while keeping the trend components that carry the excavation-cycle information. The Bi-GRU network then extracts nonlinear relationships between the tunneling parameters and AS within a fixed historical input window. The framework was evaluated on field monitoring data from the Wuhu Yangtze River Tunnel. In this case, the WT-Bi-GRU model reached R2 = 0.82, RMSE = 1.07 mm/min and MAE = 0.77 mm/min and performed better than Bi-GRU, GRU, LSTM, WT-GRU and WT-LSTM models trained on the same data. Because the evaluation covers a single project, the results indicate a potential for supporting TBM operational assessment rather than a demonstrated improvement in tunneling safety; broader applicability still requires cross-project validation and prospective testing in operation.
Authors
- Junwei Jin (ORCID: https://orcid.org/0000-0002-0486-4477)
- Jicheng Shu
- Mingyu Li (ORCID: https://orcid.org/0000-0002-7241-2261)
- Jian Chen
- Qianqian Jin
- Juncai Wu
- Hengxin Zhen
Institutions
- Zhengzhou University (CN)
Publication Details
- Journal
- Applied Sciences
- Published
- 2026-11-12
- DOI
- https://doi.org/10.3390/app162211212
- Primary Topic
- Tunneling and Rock Mechanics
- Type
- article
- Field-Weighted Citation Impact
- 0.00