Investigation of thermal conductivity of rubberized concrete as an energy-efficient building material and modeling by artificial intelligence

The main aim of this research is to investigate and mathematically express the relationship between the mixture proportions of rubberized concrete and its thermal conductivity performance. For that purpose, a dataset with a wide range of experimental variables was compiled from the studies available in the literature and one of the most important and widely used machine learning methods, called Artificial Neural Networks, was chosen to establish this mathematical expression strongly and consistently. Two important criteria were taken into consideration when compiling the dataset: firstly, the aggregate had to be of natural normal weight and secondly, the rubber aggregate had to be derived from waste tire and not treated. A reliable, functional, and robust empirical model to estimate the thermal conductivity coefficient of the rubberized concrete was generated in the scope of this study based on the input parameters like cement content (c), water-to-cement ratio ( w/c ), natural aggregate-to-cement ratio ( na/c ), rubber aggregate-to-cement ratio ( ra/c ), and rubber type ( rt ). The estimation capability of the model was validated using a dataset that the model never faced and was evaluated based on some statistical metrics like R 2 , MAPE , MSE , etc. The R 2 , MAPE , and MSE values of the trained model were about 0.984, 4.62%, and 0.002, respectively. Both validation and statistical evaluation results revealed that the model can accurately and reliably estimate the thermal conductivity coefficient of the rubberized concrete. Besides, the statistical metrics of the developed model were in the acceptable range for such models.