Machine learning models for predicting the compressive strengths of ordinary Portland cement concrete and alkali-activated materials

Alkali-activated materials (AAMs) are a type of environmentally friendly concrete and blast furnace slag (BFS) and fly ash (FA) instead of cement are used for powder. BFS and FA are industrial byproducts, and their composition ratios vary depending on where they were created. There are two challenges to the further use of AAMs around the world. First, the compressive strength of AAMs depends on the composition ratios of the powder. Second, there are many factors that affect the compressive strength of AAMs, but the magnitude of the effect of each factor has not been understood. The purpose of this study is to develop a machine learning model considering composition ratios for predicting the compressive strength and to identify the key factors influencing it. In this study, four machine learning models are proposed to predict the compressive strengths of ordinary Portland cement concrete (OPC) and AAMs. Data set of OPC is used to demonstrate the effectiveness of using machine learning to predict the compressive strength of concrete. The models for OPC and AAMs were created using 202 and 287 test results, respectively. The performance of the models was evaluated with hold-out and k-fold cross-validation. This study revealed the following. The effect of the composition ratio of FA on the compressive strength of AAMs was greater than that of BFS. The prediction accuracy for AAMs was greatly improved by dividing AAMs into BFS-based AAMs and FA-based AAMs.