Abstract
AbstractInterface crack propagation of FRP (fiber-reinforced polymer) strengthened reinforced concrete (RC) flexural member is often initiated from the toes of the intermediate cracks and propagates towards the supports. This type of FRP delamination is commonly termed intermediate crack (IC) debonding and is common for flexural members with high shear span-to-depth ratios. If the ultimate FRP strain at IC debonding failure is known, the moment capacity of the member can be obtained through a simple section analysis. This research deals with the prediction of ultimate FRP strain at IC debonding, using neural networks and regression models. Basic information on neural networks and the types of neural networks most suitable for the analysis of experimental results are given. A set of experimental data for FRP-strengthened RC beams and one-way slabs, covering a large range of parameters, for the training and testing of neural networks is used. The available test results were not only compared with current code provisions but with equations proposed by other researchers as well. The prediction models based on neural network are presented. A new design equation is also suggested.