Machine Learning of Concrete Temperature Development for Quality Control of Field Curing

Abstract

Understanding the relationship between concrete temperature development and field curing time helps to control material quality, improve construction efficiency, and enhance research on concrete design. However, it is difficult to precisely predict temperature trends when placing concrete because there are many influencing factors and uncontrollable ambient variables in the curing process. To forecast the short-term temperature trends reliably and automatically, this research proposes a temperature measurement and quality prediction (TMQP) system to proactively evaluate the development trajectory of concrete quality and the temperature changes at the center and surface of the cross section of concrete structural members. The TMQP system includes radio-frequency identification (RFID) temperature sensors for recording the temperature data and Big Data analytics (BDA) combined with the machine-learning method of classification and regression tree (CART) for measuring and predicting of temperature development. The results indicate that the system has over 98% reliability on the correlation coefficients between the predicted temperatures and actual temperatures based on 240 h of continuous experiments and 190 h of documented data. This entire research design is applicable to various concrete construction projects and sheds light on how BDA and machine learning can help construction engineers and managers to control concrete curing and take preventive measures to avoid concrete surface cracks.