Applying neural networks for estimating the compressive strength of confined circular concrete columns with FRP sheets

Most of existing reinforced concrete columns are in need of retrofitting and strengthening for various reasons, including errors during the construction phase, changing the type of applications in structures, changes in design codes, occurrence of strong beam-weak column mechanism and the damages due to natural disasters. One of the most common ways of strengthening the columns is the confinement of the reinforced concrete columns. So far, several experiments have been conducted on concrete columns confined with fibre-reinforced polymer sheets and the results show that the use of fibre-reinforced polymer sheets increases the compressive strength of the concrete columns effectively. Different models in order to determine the compressive strength of the fibre-reinforced-polymer-confined concrete columns are provided in the previous researches. In this study, a large set of experimental data regarding circular columns confined with different types of FRP has been collected. Two neural network prediction methods were also used for determining the compressive strength of the confined circular columns with FRP sheets. These methods are back propagation (BP) and radial basis functions (RBF). Finally, the defined neural networks were examined with the available estimation models based on four standard error testing criteria. Results show that the neural networks could estimate the compressive strength of the confined columns with FRP with more accuracy rather than the existing analytical models.