Assessment and prediction of cement paste flow behavior; Marsh-funnel flow time and mini-slump values

Abstract

In this study, the parameters affecting Marsh-funnel flow time and mini-slump of the paste mixtures were determined through experimental and modelling studies. Marsh-funnel flow times were modelled through artificial intelligence and regression methods. A novel model was used to train the coefficients of artificial neural networks (ANN) with the Teaching-Learning Based Artificial Bee Colony (TLABC) Algorithm. Accuracy of this method was investigated through ANN-Back Propagation, ANN-Teaching Learning Based Optimization Algorithm, ANN-Artificial Bee Colony, Multivariate Adaptive Regression Splines and Classical Regression Analysis methods. ANN-TLABC method showed the best results among the applied models. The admixture content, cement fineness, solid material content of admixture and C3A content of cement were found to be the most important parameters affecting the flowability of the paste. However, C2S, equivalent alkali, C4AF and C3S contents of the cement were observed to have no considerable effect on the Marsh-funnel flowability of paste.