Civil Eng, Malayer University
Due to lack of theory of elasticity, estimation of ultimate torsional strength of reinforcement concrete beams is a difficult task. Therefore, the finite element methods could be applied for determination of strength of concrete beams. Furthermore, for complicated, highly nonlinear and ambiguous status, artificial neural networks are appropriate tools for prediction of behavior of such states. The main purpose of this paper is evaluation of ultimate torsional strength of rectangular concrete beams using two different approaches: nonlinear finite element modeling and artificial neural network prediction. This aim is achieved by creating a three-dimensional finite element models that employing the brittle failure criterion of concrete and an artificial neural network which training is carried out through experimental data. Numerical modeling is accomplished using the commercial software and its validation is exhibited by selected data from experimental tests. The test specimens were solid rectangular beams that were subjected to pure torsion. All of the experimental data for training of the network has been comprised. Data are divided into three categories: training, testing and validating. Three-layer perceptron network with a back propagation error algorithm is used for training. This study shows that, the results of numerical models are more accurate than neural network models to predict torsional strength of reinforced concrete beams.