Predicting the load-carrying capacity of GFRP-reinforced concrete columns using ANN and evolutionary strategy

Abstract Recently, glass fibre reinforced polymer (GFRP) rebars have been widely used by different researchers due to the high strength and deformability of polymer rebars. However, current concrete designing codes do not include design provisions for GFRP‐reinforced concrete columns, and neglect the contribution of GFRP bars when used as compression reinforcements. Therefore, this study intends to develop a new simple closed-form model to predict the load-bearing capacity of GFRP-reinforced concrete columns under an axial load considering the influence of longitudinal GFRP rebars and confinement conditions. This study was performed in two steps. In the first step, three artificial neural network (ANN) algorithms were used to predict the axial performance of GFRP-reinforced columns: radial basis function (RBF), support vector regression (SVR) and multilayer perceptron (MLP). Then, a sensitivity analysis was performed and the impact priority of each of the input parameters was examined. In the second step, a comparison between the obtained results and available models was performed, and a new highly accurate formula was proposed using the Covariance Matrix Adaptation- Evolution Strategy (CMA-ES) to predict the nominal compressive strength of GFRP-reinforced concrete columns. The results showed that the ultimate strength and strain of longitudinal and transverse rebars and their spacing were important parameters affecting the axial load-bearing capacity of GFRP-reinforced concrete columns and the previous models did not consider the influence of these parameters. Therefore, the presented ANN models in this study with high accuracy and low error could be used as beneficial tools to predict the axial performance of GFRP‐reinforced concrete columns. In addition, among the presented ANN methods, RBF showed the highest accuracy and efficiency. Also, the developed closed-form model with a high agreement with experimental results could be used as an efficient model to determine the nominal strength of GFRP-reinforced concrete columns.