Artificial intelligence assisted optimization of rammed aggregate pier supported raft foundation systems based on parametric three-dimensional finite element analysis

The increased usage of rammed aggregate piers has demanded optimizing the design parameters of this technique. In this study, numerous parametric three-dimensional finite element analyses were conducted to optimize the parameters of the rammed aggregate pier-supported raft foundation with novel methods. Three-dimensional finite element simulations were validated using a field loading test for a single rammed aggregate pier at Neola, Iowa (USA). After the validation, elasticity modules of the rammed aggregate piers, raft foundation thicknesses, and column spacings were varied. The raft foundation-rammed aggregate pier groups system in 81 different combinations was subjected to parametric three-dimensional finite element analyses to provide data for the optimization study. The effect of the changes in the elasticity modulus of the rammed aggregate piers, the raft foundation thickness, and column spacing on the settlement was investigated. Optimization analyses were performed using the artificial intelligence-supported novel Goal Attainment Method and the Response Surface Method. The novel coding developed in this study, which automatically selects the most reasonable prediction function with the support of artificial intelligence, created a prediction function with a high correlation coefficient used in the optimization analyses. Two different optimization methods specified in the optimization study performed as multi-objectives were compared, and the solution system with the highest desirability value was selected. Therefore, optimum design parameters were reached by using novel methods for general raft foundation-rammed aggregate pier group systems to guide researchers.

Automated summary

This study optimized the design parameters of rammed aggregate pier-supported raft foundations through parametric three-dimensional finite element analyses. By validating the simulations and varying parameters, the researchers obtained the optimum design parameters and provided guidance for future research.