Compressive strength modeling of blended concrete based on empirical and artificial neural network techniques

Abstract

ABSTRACT This article exhibits an experimental study carried out to investigate the combined effect of Silica Fume (SF) and Metakaolin (MK) on the fresh and hardened properties of concrete. The replacement levels of SF were adopted as 5%, 10%, 15%, 20% while that of MK were 5%, 10%, 15%, 20% and 25%. The results for cube testing revealed that the use of SF and MK produces considerably good strength concrete. Based on the experimental observations, an approach to predict the compressive strength using regression modeling was suggested. The result of which rendered a reasonable agreement with the available test data. Moreover, two Artificial Neural Network (ANN) models have also been proposed to predict the compressive strength of concrete using the data obtained from the experimental exercise. Randomized stratification method was used to divide the data samples into training and testing subsets in line with machine learning best practices. The results of the ANN models were found to be in good agreement with experimental values. When compared with the results of empirical models, the respective ANN models gave a better prediction for the compressive strength. This substantiates the reliability of ANN over the empirical models.