Confining Pressure Forecasting of Shield Tunnel Lining Based on GRU Model and RNN Model

Area of Science:

• Geotechnical Engineering
• Tunneling and Underground Construction
• Predictive Modeling in Civil Engineering

Background:

• Confining pressure significantly influences tunnel internal forces and construction safety.
• Variations in groundwater levels and applied loads dynamically alter confining pressure during construction.
• Accurate estimation of confining pressure is essential for ensuring tunnel structural integrity.

Purpose of the Study:

• To develop and compare predictive models for tunnel confining pressure.
• To evaluate the performance of Gated Recurrent Unit (GRU) and Recurrent Neural Network (RNN) models.
• To enhance the understanding of soil pressure dynamics for safer tunnel construction.

Main Methods:

• Acquisition of a high-frequency time-series dataset of tunnel confining pressure via field monitoring.
• Segmentation of monitoring data into training and testing sets for model development.
• Implementation and comparative analysis of GRU and RNN models for confining pressure prediction.

Main Results:

• The GRU model exhibited significantly faster training speeds and higher prediction accuracy compared to the RNN model.
• The RNN model demonstrated slower training performance and lower accuracy in confining pressure prediction.
• GRU model's effectiveness validates its suitability for dynamic soil pressure prediction in tunneling.

Conclusions:

• GRU models offer a more efficient and accurate approach for predicting tunnel confining pressure.
• The study underscores the importance of selecting appropriate predictive models that balance speed and accuracy.
• Findings contribute to developing advanced tools for improving tunnel construction safety and reliability.