Simulating USBR4908 by ANN modeling to analyse the effect of mineral admixture with ordinary and pozzolanic cements on the sulfate resistance of concrete

One of the available tests that can be used to evaluate the sulfate resistance of concrete is a procedure for length change of hardened concrete exposed to alkali sulfates (USBR4908). However, there are deficiencies in this test method including a lengthy measuring period, insensitivity of the measurement tool to the progression of sulfate attack. Moreover it is difficult to obtain experimental expansion due to time and cost limitations. A reasonable prediction for the expansion in USBR4908 is basically required. This study focuses on the artificial neural network (ANN) as an alternative approach to evaluate the sulfate resistance of concrete. A total of 273 different data for three types of Portland cement combined with fly ash (FA) or silica fume (SF) concrete mixes, along with different w/c ratios of 0.35, 0.45 and 0.55 were collected from the experimental program. ANN models were developed. The data used in the ANN model consisted of five input parameters which include W/B ratio, cement content(CC), FA or SF content, tricalcium aluminate content (C3A), and exposure duration (D). Output parameter is determined as expansion (E). Back propagation (BP) algorithm was employed for the ANN training in which a Tansig function was used as the nonlinear transfer function. Through the comparison of the estimated results from the ANN models and experimental data, it was clear that the ANN models give high prediction accuracy. In addition, the research results demonstrate that using ANN models to predict the expansion in concrete cylinders is practical and beneficial.