Prediction the Hardened Properties of Fiber-Reinforced Self-Compacting Concrete Using Artificial Neural Networks

Nowadays، intelligent methods inspired from nature are implemented to resolve complex problems، there are very popular too. The most common one is artificial neural network; they are capable to collect huge amount of complex information through experiments and tests. On the other hand، combination of fiber concrete with self-consolidating concrete produces a new product with high fluidity and good adhesion. Due to the presence of fibers، this type of concrete presents high-quality advantages such as high resistance to impact، high fatigue performance، low rate of erosion، enhances the tensile and flexural strengths and reduces the segregation. In this study by considering the components of the concrete mix design as networks input and two kinds of neural network modelling، a Radial Basis function and a recurrent neural network (NARX) for predicting the properties of hardened concrete، were used. To improve networks training، 40 concrete mix fibers reinforced self-consolidating concrete and three fiber types including steel fibers، glass and polypropylene، were prepared. Comparison of experimental results and network outputs indicate that both networks have the sufficient accuracy in estimating the hardened properties of self-consolidating concrete and recurrent neural network (NARX) error is less than Radial Basis Function Neural Network.