Bayesian networks modelling for predicting compressive strength and slump flow of self-compacting concrete

In this research paper, Bayesian networks were used to predict slump flow and compressive strength of self-compacting concrete (SCC) based on dune sand, which are essential values for evaluating this concrete’s rheological and mechanical properties. Slump flow and compressive strength values were predicted using a machine learning method based on a database extracted from previous literature (113 samples). These values were compared with the results of experimental work to determine the accuracy of the forecasting process. Based on the practical work, it is also possible to know the effect of replacing crushed sand with dune sand in the self-compacting concrete mixtures on its properties. The results showed that there is a notable convergence between the predictive and experimental values of the studied properties, as the percentage of the integrated absolute error is low and does not exceed 2.46% for slump flow and 1.49% for compressive resistance, demonstrating the effectiveness of the prediction approach employed in this study. It was also concluded that low to medium percentages (up to 50%) of dune sand have a positive impact on the rheological properties, improving filling and passing abilities, as well as segregation resistance. Moreover, it was observed that these rates of dune sand do not harm the mechanical properties of this concrete. These findings encourage the dual use of dune sand alongside crushed sand in the production of self-compacting concrete.