Clayshock mechanism and application to shield tunneling through existing tunnels: Settlement prediction using artificial intelligence

In this study, we investigated the clayshock mechanism during the construction of a new shield tunnel that passes through an existing shield tunnel. The experimental findings demonstrated a significant reduction of approximately 12% in the friction between the shield machine and clay and silt layers. Once the clayshock concentration surpasses 400 kg/m³, it may successfully impede the movement of the slurry, completely fill the space between the formation and the shield, and minimize surface settling. Furthermore, the Plackett–Burman test was used to assess the sensitivity of the primary shield characteristics affecting settlement. The findings indicate that settlement is significantly influenced by the soil pressure, grouting quantity, and clayshock efficiency. A full-factor analysis revealed an intricate link between the shield properties and settlement. If the volume of grouting is less than 6.5 m³, it is very difficult to regulate the settling value within -1.0 mm. We utilized the following neural network models: gate recurrent unit networks, long short-term memory networks, and the Elman recurrent neural network. The training input data consists of geometric parameters, geological parameters, and shield parameters, while the prediction output data serves as the settlement value. Researchers can produce accurate and reliable predictive outcomes by improving the model training process.