Machine Learning Analysis of Effect of Aggregate Properties on Drying Shrinkage in Concrete

Abstract

Drying shrinkage of concrete is affected by mix proportions such as unit water content and water-cement ratio. In addition to these, many publications have reported in recent years that aggregate itself also shrinks and that this affects the shrinkage of concrete. The effect of aggregate is qualitatively included in formulas proposed by academic societies such as the Japan Society of Civil Engineers and the Architectural Institute of Japan, however it is still not sufficiently clarified due to its complexity. This study aims to analyze and clarify the effect of aggregate properties on drying shrinkage in concrete by machine learning, which can identify dominant regimes directly from data without explicit programming, and clarify the effects of aggregate properties on concrete shrinkage. The results of the analysis show that the average specific surface area of aggregate has a significant effect and affects shrinkage of concrete in relation to mix proportions.