Estimating Rice Husk Ash Concrete Compressive Strength Using Hybrid Machine Learning Methodology

Abstract

The earth's temperature is gradually increasing due to greenhouse gases (GHGs), which have caused the global warming phenomenon. Carbon dioxide (CO2) has accounted for a large majority of such greenhouse gases. Due to the high CO2 emissions caused by cement production, several alternative cementitious materials have been discovered, such as Rice Husk Ash (RHA). This study suggests a hybrid data-driven machine learning framework approach for predicting RHA concrete compressive strength. The framework utilizes Bayesian Regularization-based artificial neural network (ANN) and genetic expression programming (GEP) to develop a generalized prediction mathematical model that avoids overfitting. The model included six compressive strength prediction parameters: age, cement content, rice husk ash content, water, superplasticizer, and aggregate. Based on the literature's 192 compressive strength-tested specimens, the proposed framework has been trained. The proposed model resulted in average calculated-to-measured ratios of 1.00, a standard deviation of 0.15, and a COV of 15%. In addition, a parametric study has been conducted to evaluate the effect of each contributing parameter. The analysis shows that varying the RHA content from 0 to 200 kg/m3 will increase the compressive strength by up to 50%.