Compressive strength prediction models for cementitious composites with fly ash using machine learning techniques

Abstract

In this study, it was proposed a novel prediction model to predict compressive strength of mortar samples having different properties. For this purpose, 8 different fly ashes were used in mortar mixture as a replacement of cement by weight. Mortars including different ashes were prepared with addition of 10%, 20%, 30% and 40% fly ash. Compressive strength of the produced mortar samples were evaluated at 1, 3, 7, 28, 90 and 365 days. Totally 196 test samples were produced and mechanically tested. The relation between compressive strength values (dependent value) and SiO2 + Al2O3 + Fe2O3 content, age, and fly ash replacement ratios (independent values) were predicted by machine learning techniques such as Artificial Neural Networks (ANN) and Adaptive-Network Based Fuzzy Inference Systems (ANFIS). The findings were compared with traditional statistical method Multi-Linear Regression (MLR) to prove proposed models. According to test results it has an incentive effect for future studies to know that GA based Anfis model produce better results to estimate compressive strength using chemical composition of fly as in terms of SiO2 + Al2O3 + Fe2O3, fly ashsubstation ratio in the mortar and age of the sample.