Quantile-Based Design and Optimization of Shallow Foundation on Cohesionless Soil Using Adaptive Kriging Surrogates

This paper utilized an advanced surrogate modeling technique known as adaptive kriging (AK) for the probabilistic design of foundations. The quantile-based approach accounted for the variability in soil and foundation dimensions with an adequate level of safety, and the AK algorithm replaced the theoretical or numerical models with surrogate models to improve the computational efficiency. The AK model had the advantage of considering the design dimensions and their variability in its optimization algorithm. For the design of shallow foundations, the variability in the design dimensions and its optimization was carried out by introducing the width-to-length ratio as the design parameter. This meant that a surrogate model for different foundation dimensions did not need to be developed separately. This paper showed that the computational time for the AK model was much less compared with the probabilistic analysis that used numerical models. A quantile-based design and optimization procedure that used the AK model was illustrated through a shallow foundation example and provided economical design parameters.