Smart geotechnical design of green foundations using recycled materials and Nano-silica: Machine learning techniques and numerical validation

At a crossroads, the global construction industry is also one of resource depletion and growing waste production. In this research, we present a full-scale “Smart Geotechnical Design” framework for green foundations that employs Recycled Concrete Aggregate (RCA) and Crumb Rubber (CR) as sustainable alternatives to natural aggregates. While these materials present great environmental benefits, they also have complex mechanical behavior and high variability, which traditional empirical design methods have trouble with. To remediate structural issues and microstructural porosity in these recycled materials, we added nano-silica as a nanomechanical additive at precise doses (1% – 3%). We addressed this by developing an advanced predictive model that used Machine Learning (ML) algorithms, Random Forest (RF), Artificial Neural Networks (ANN) and Extreme Gradient Boosting (XGBoost). In the laboratory we collected a large set of 450 data points from triaxial and plate load tests. The XGBoost model outperformed all others to report an R² of 0.96 and RMSE of 0.042. To validate the ML results, we performed a nonlinear Finite Element Analysis in ANSYS Workbench. We modeled the soil-RCA-CR with the Drucker-Prager constitutive criterion, which we calibrated via experimental data. In our analysis, which included FEA, we found very good agreement with what we saw in the lab and also what the models predicted, which in turn proves the smart design framework’s value. Furthermore, we used SHapley Additive exPlanations (SHAP) and Sobol Global Sensitivity Analysis which identified replacement ratio and compaction energy as the most influential parameters. In this research, we present a Life Cycle Assessment (LCA) and a reliability-based design that we developed using Monte Carlo Simulations. We are filling in the gap between what the circular economy has to offer and reliable geotechnical engineering. We have put forth a data based tool which is to improve sustainable foundation systems at the rate of reducing carbon footprints by 35% at the same time we maintain structural integrity.