Conventional and Ensemble Machine Learning Models to Predict the Compressive Strength of Fly Ash Based Geopolymer Concrete

Abstract

Cement is the primary component of concrete, a material that is widely used in the construction world. Due to the release of numerous gases, cement production and its use harm the environment. To mitigate this negative impact, the researchers have developed a solution named “Geopolymer Concrete (GPC)”. Among the various mechanical properties of concrete, the Compressive Strength (CS) of concrete is regarded as one of the crucial criteria in many design codes and standards and its dependency is highly non-linear on its ingredients including supplementary cementitious materials (SCMs). SCMs are pozzolanic materials, and their use in concrete results in the enhancement of its properties. Among the various SCMs, Fly Ash (FlA) is one of the most effective and sustainable material to improve concrete's CS. So, to predict the influence of the materials used along with FlA on the CS of GPC, either it requires the extensive experiments, requiring money and time or use soft computing techniques. With the help of various soft computing techniques, a more reliable model for the prediction of CS was developed, under which the assets can be saved. Also, with the help of these models for CS, we can receive recommendations for how to schedule the construction process and remove the formwork. In this study, the two Conventional Machine Learning (CML) models as Linear Regression, Artificial Neural Network and AdaBoost as an Ensemble Machine Learning (EML) model were developed to predict the CS of FlA based GPC. On comparison, between the developed models, AdaBoost model was found to be the most effective for an accurate CS prediction with Correlation coefficient (R2) as 0.944, Root Mean Squared Error (RMSE) as 2.506 and Mean Absolute Error (MAE) as 1.259. Linear Regression model was deemed to be inferior with R2 as 0.701, RMSE as 5.805 and MAE as 4.502.