Micromechanics-integrated machine learning approaches to predict the mechanical behaviors of concrete containing crushed clay brick aggregates

Abstract

In this paper, we present micromechanics-integrated machine learning studies of concrete containing crushed clay brick as a coarse aggregate. Initially, the stress-strain responses of normal and crushed clay brick aggregates were quantitatively measured. Concrete specimens with various water/cement ratios and replacement rates of the aggregate were then fabricated and their stress-strain responses were evaluated. Concrete mixed with crushed clay brick aggregate showed lower compressive strength compared to concrete containing normal aggregate, and it was difficult to predict the material performance based on existing specifications and models. Based on results from an experimental assessment, micromechanics-integrated machine learning approaches that can effectively predict the properties of concrete mixed with crushed clay brick aggregates was developed. Several parametric studies and comparisons are carried out to verify the potential capacity of the proposed model framework.