Prediction of Shear Behavior of Fiber Reinforced Concrete Beams Using Artificial Neural Networks

Abstract

Due to the shortage of clear equations in the building codes that explain shear strength for fiber reinforced concrete (FRC) beams; there is a need to develop a numerical approach that can be used to predict shear behavior in FRC. The main objective of this research is to develop an artificial Neural Network (ANN) that can predict shear strength and simplify its use through developing a Graphic User Interface (GUI). Moreover, shear behavior in fiber reinforced concrete beams (FRCBs) is quantified by compressive strength of concrete, longitudinal steel, size effect, fiber's type, content and aspect ratio. The research methodology is based on collecting experimental results of technical investigations carried out to predict shear behavior in FRCBs. ANN aims at reducing the amount of computing time required in the numerous iterations involving structural analysis and experimental work. For this, two back-propagation neural networks have been experimented by MATLAB program; their types have been fitting (1st network) and pattern recognition (2nd network) which are used to classify failure of FRC beams into 6 categories. Through simulation study, the optimum architectures for the individual ANNs have been determined. The training algorithms used feed forward back propagation. The ANNs model has been assessed in comparison with exact values and deduces a good correlation with it. Finally, a software program is developed as an evaluation system for predicting resistance of FRC beams to shear forces, and to expect the failure pattern in order to avoid its occurrence.