Hybrid FEM–machine learning framework for back-analysis of spatially varying soil parameters in super-large caisson foundation

Key Points

• Effective integration of finite element method and machine learning enhances modeling accuracy for soil parameters.
• A hybrid framework significantly improves back-analysis results with predictive models, achieving high precision.
• Utilization of advanced computational techniques in a real-world foundation context aids in soil behavior understanding.
• Highlights the potential for improved designs in large-scale infrastructure projects, suggesting broader applications.