Mix design development for geopolymer treated expansive subgrades using artificial neural network

Abstract

Expansive soils have been inevitable during road constructions due to its widespread availability. Conventional stabilizers; cement and lime have not been quite reliable in terms of economic, environmental, and durability perspectives. Thus, geopolymer has come into limelight as a sustainable alternative. However, the complexity of geopolymer-soil behaviour makes it challenging to predict their strength development patterns and it requires a large number of trial and error tests in laboratory scale due to the soil specific nature of such mixes. To date, no research has been attempted to develop design charts for geopolymer-expansive soil stabilization field applications. This study proposes design charts using Artificial neural networks (ANN) to predict the 28-day unconfined compressive strength (UCS) of Class F fly ash (FA) based geopolymer stabilized expansive soil (PI = 32% – 46%) to support field implementation of this novel geopolymer based soil stabilization. The model was developed based on a robust collection of literature data; FA/Soil, Activator/FA, Na2SiO3/NaOH, alkaline molarity, and their respective 28-day UCS values. Three contour plots for mix design: FA/soil vs. NaOH molarity, activator/FA vs. Na2SiO3/NaOH, and Na2SiO3/NaOH vs. NaOH molarity were proposed based on the model predictions. The proposed mix design charts were validated through the in-house tests which showed reasonable match in comparison with the predicted 28-day UCS using the contour plots. The proposed mix design charts can be employed for in-situ geopolymer-expansive subgrade stabilization to achieve strengths between 0.6 and 2.1 MPa.