Design of Alkali-Activated Slag-Fly Ash Concrete Mixtures Using Machine Learning

So far, alkali-activated concrete has primarily focused on the effect of source material properties and ratio of mixture proportions on the compressive strength development. A little research has focused on developing a standard mixture design procedure for alkali-activated concrete for a range of compressive strength grades. This study developed a standard mixture design procedure for alkali-activated slag-fly ash (low-calcium, Class F) blended concrete using two machine learning techniques: artificial neural networks (AMV) and multivariate adaptive regression spline (MARS). The algorithm for the predictive modelfbr concrete mixture design was developed using MAT AR programming environment by considering the five key input parameters: water/solid ratio, alkaline activator/binder ratio, Na-Silicate/NaOH ratio, fly ash/slag ratio, and NaOH molarity. The targeted compressive strengths ranging from 25 to 45 MPa (3.63 to 6.53 ksi) at 28 days were achieved with laboratory testing using the proposed machine learning mixture design procedure. Thus, this tool has the capability to provide a novel approach for the design of slag-fly ash blended alkali-activated concrete grades matching to the requirements of in-place, field constructions.