Compressive strength prediction for concrete materials using machine learning

Abstract

Concrete is one of the most widely used construction materials. Properties of concrete are determined by the mix proportion of component materials, such as water cement ratio. Previous studies suggest that machine learning is applicable to predict concrete properties based on concrete mix proportions. This paper proposed a model to build a prediction model of concrete compressive strength based on machine learning. And the model was validated with experimental results. The dataset was prepared based on papers published by the Japan Concrete Institute. Specimens for experimental validation were prepared with mix proportions included/excluded in the dataset. As a result, the build model could accurately derive compressive strength within the dataset. On the other hand, in case of outside the dataset, obtained results showed appropriately predicted compressive strength, while some of them included unneglectable error.