Abstract: Slurry level control achieved by adjusting the parameters of the slurry circulation system is crucial for slurry balance shield tunneling. Owing to the longdistance pipeline transportation of slurry, the system exhibits considerable time delays, with the control characteristics changing depending on the geological conditions and pipeline length. Realizing accurate control is challenging with traditional manual control methods. To address this issue and ensure a stable excavation face pressure during shield tunneling, the authors propose the use of an intelligent prediction and control method for estimating the shield slurry chamber level based on a Transformer network. The proposed method leverages the Transformer network′s ability to model the dynamic characteristics of the slurry circulation system using an iterative multistep prediction strategy to forecast future slurry levels for a given sequence of control parameters. Additionally, to optimize control performance and meet system constraints, an adaptive gradient descent method is employed to solve the optimization problem and its constraints, thereby obtaining the optimal control parameters for the system. The proposed method is validated through simulation using a construction dataset of the deep drainage tunnel in the Suzhou river section. Experimental results indicate that the proposed control method substantially improves the control effectiveness of the slurry circulation system during slurry shield tunneling in silty clay layers. Compared to traditional control methods, this intelligent prediction and control method offers higher accuracy and stability, demonstrating its potential for practical applications in shield construction.

Key words:

slurry balance shield, slurry level control, iterative multistep prediction, Transformer network, intelligent control