Evaluating unconfined compressive strength of cohesive soils stabilized with geopolymer: a computational intelligence approach

Soil stabilization using geopolymers is a new technique for improvement of weak cohesive soils. Evaluating behavior of improved soils requires an initial estimation of strength parameters. In this study, extensive experimental results on geopolymer-stabilized soil specimens were collected and analyzed. A model was then developed using group method of data handling (GMDH) and employing particle-swarm optimization algorithm to estimate the unconfined compressive strength (UCS) of stabilized cohesive soils using geopolymers. Type of additives and their compositions as well as soil characteristics were taken as the influential parameters on the UCS of soil specimens. Subsequently, sensitivity analysis was carried out to verify the performance of the proposed UCS model. Finally, the developed GMDH-based model was compared with artificial neural network model to predict unconfined compressive strength of stabilized soils. The results clearly illustrate the reasonable accuracy of the developed computational Intelligence-based model for estimating the unconfined compressive strength of geopolymer-stabilized cohesive soils.