Integration of FEM-based permeation analysis and AI-based predictive models for improved chemical grout permeation assessment in heterogeneous soils

• FEM and AI models improve evaluation of chemical grout permeation in heterogeneous soils. • Low permeability zones significantly affect grout flow, requiring predictive modeling. • AI-based models accurately predict permeation risk, reducing computational requirements. This study proposes an integrated framework combining Finite Element Method (FEM)-based permeation analysis with Artificial Intelligence (AI)-based predictive models for assessing chemical grout permeation behavior in heterogeneous sandy soils containing low-permeability zones. FEM analysis was conducted to investigate grout flow patterns and permeation risks in such soils, revealing that proximity to low-permeability zones significantly influences flow velocity and grout distribution. Simplified regression equations were developed to efficiently predict permeation risks and reduce computational complexity. Additionally, AI models, including neural networks and gradient boosting decision trees, were trained on FEM-derived datasets to predict permeation velocities and ranges with high accuracy (R² = 0.849). Results showed that even with 5.5 % low-permeability content, average fill rates of 94.5 % (FEM) and 96 % (AI) were achieved, with worst-case scenarios dropping to approximately 81 % and 83 %, respectively. The findings highlight the effectiveness of integrating FEM, regression models, and AI-based modeling to optimize chemical grouting strategies and mitigate permeation risks due to soil heterogeneity. The proposed framework offers an effective approach to enhancing chemical grouting practices in heterogeneous ground conditions.