Estimation of sulfate expansion level of PC mortar using statistical and neural approaches

Abstract Prediction of sulfate resistance is a keynote issue for the structural evaluation of cementitious systems. Concrete structures may experience a range of sulfate attack depending on several factors. Among these, C3A content, C3S/C2S ratio of the cement, type of mineral admixture and its inclusion level, as well as type of sulfate and its concentration may be listed. In this investigation, an experimental study was undertaken to characterize sulfate expansion of PC mortar with related parameters, and comprehensive numerical analyses were conducted for the estimation of the sulfate expansion levels. In the experimental study, twenty seven mortar mixtures were prepared and tested in accordance with ASTM C1012 testing procedure. Forty-five experimental expansion values obtained from sulfate expansion tests were statistically analyzed in detail and seventeen linear and nonlinear regression models were established for the characterization of target mapping. Apart, neural network (NN) methodology was also employed for the identification of considered nonlinear relationship. Results of this study revealed that NN model exhibited better performance over regression models to predict the sulfate expansion of various cements containing natural pozzolan and fly ash.