Active learning-based research of foaming agent for EPB shield soil conditioning in gravel stratum

Injecting soil conditioner into the soil during EPB shield construction is crucial for soil enhancement. Foam agent consumption is common but developing their composition and qualities is time-consuming. Hence, there is an urgent need for a novel approach to material design. This study presents the development of an active learningbased model for foam agents to enhance soil conditioning in gravel strata during EPB shield operations. Foaming volume and half-lift time were forecasted using six machine learning algorithms. The Extra Tree Regression (ETR) model was the most effective for volume prediction, while CatBoost Integration (CB) model performed best for time prediction. The new foam has a half-life of 945 s and the Foam Expansion Ratio (FER) of 21, suitable for EPB shield soil conditioning. The results of slump value and shear strength experiments show a positive correlation between soil slump value, shear strength, and an increase in Foam Inject Ratio (FIR).