Evaluation of Compressive Strength of High-Volume Fly Ash Concretes by using Multivariate Adaptive Regression Splines

Abstract

Fly ash, an industrial by-product, a supplementary cementitious material used in civil engineering construction. Addition of fly ash reduces production of carbon dioxide and disposal problem. This paper presents an advanced statistical model, Multivariate adaptive regression splines (MARS) to predict the compressive strength of fly ash concretes. MARS is formulated based on nonlinear and nonparametric regression approach. Dependent variables have been identified from the extensive experimental investigations on fly ash concrete. According to the output and input parameters adopted, MARS establishes appropriate relations by using the concept of divide and conquers strategy. Using the compressive strength data on various concrete mixes, MARS models were developed. About  of the experimental data has been used for development of MARS models and remaining data for verification and validation of the developed models. In this models, all associated and interrelated variables and their limits have been taken into account. Three MARS models have been developed to evaluate the compressive strength of fly ash concrete at 28, 56 and 90 days. It is observed from the validation studies that the predicted compressive strength concrete for various periods exhibits good agreement with the experimental results.

Keywords:Fly ash; Concrete; Compressive strength; Multivariate adaptive regression splines.