Abstract
A slump test is a standard method to evaluate fluidity of fresh concrete as one of the quality control items in concrete production. With an increasing demand for improved productivity in the field of concrete, research has been conducted to establish a technology to predict the slump of concrete from video information of concrete mixing by using AI. However, no sufficient quantitative evaluation has been made on its prediction accuracy and variation. In this study, AI slump prediction was performed by using image data of concrete mixing inside a mixer for training, and the predicted values were analyzed to determine the accuracy rate and statistically evaluate the variation. The results showed that the AI-predicted slump values had a high accuracy rate within the tolerance specified by the JIS, with a smaller variation compared to that in slump test results. After verification at our factory, the AI slump prediction system has been completed as a product of joint development with Pacific System Co., Ltd. and launched under the name of PreSLump AIⓇ in April 2023.
