Mass concrete is a type of concrete used for structures with large dimensions that require precautionary measures to be taken in order to control the heat. The heat generated in the core of such a structure, due to hydration of cementitious materials, is dissipated at a slow rate leading to the formation of high temperature differential. This increases the risk of temperature-induced stresses and cracking that is dependent on many factors such as the materials and proportion of concrete mix, environmental and construction conditions, etc. In order to prevent cracking, a thermal control plan is essential before the placement of concrete that in turn requires prior knowledge of the temperature development in the mass concrete member. In this context, this study presents the analysis of a case in which the construction of a mass concrete bridge foundation was investigated and a computer program, ConcreteWorks (CW), was used to predict its overall thermal performance. Predictions of temperature development profile, temperature differential, maturity, and compressive strength were made using CW and were also validated with the measured data. The results suggest CW to be a useful tool for developing a thermal control plan and for the prevention of thermal cracking.

From the perspective of the rate of heat dissipation in a mass concrete element, thermal conductivity of concrete is an important parameter. Keeping other parameters the same, a concrete mix of high thermal conductivity can reduce the risk of temperature-induced cracking by increasing the rate of dissipation of heat. Therefore, in this study, the effects of various concrete materials, such as supplementary cementitious materials (SCMs), normal-weight, lightweight, and recycled aggregates, and steel and polypropylene (PP) fibers, on the thermal conductivity of concrete were experimentally determined. SCMs, lightweight, and recycled aggregates reduced conductivity of concrete while steel fiber was observed to improve it. In addition to the experimental measurements, a prediction model for thermal conductivity was also developed. For this purpose, a database was developed from published articles and various machine learning (ML) algorithms were evaluated for their prediction accuracies. Performance metrics indicated an artificial neural network to be the best ML algorithm for the developed dataset and a 14-6-1 model architecture was finally adopted. The robustness of this model was also evaluated on an unseen/independent dataset that furnished satisfactory results with good performance (R2 ~ 0.80).