Predicting the Punching Shear Capacity of RC Slab-Column Connections with FRP Bars Using Machine Learning Based Algorithms

Abstract

In this study, two novel predictive models are proposed for predicting the punching shear capacity of reinforced concrete (RC) slab-column connections with fiber reinforced polymers (FRP) as longitudinal bars. These models were developed using Gene Expression Programming (GEP) and Multi Expression Programming (MEP) algorithms. The models were derived using a dataset of 141 experimental specimens gathered via a literature review. It includes input parameters such as the concrete compressive strength used in the slab, effective slab thickness, punching perimeter, percentage of FRP bars, and the modulus of elasticity of the FRP. In addition, the punching shear capacity was also predicted using 20 existing models, including those from design codes and widely cited literature. The predicted results obtained were compared with the experimental results. The models were then statistically analyzed according to this comparison. The coefficient of determination, coefficient of variation, mean absolute percentage error and root mean square error statistical results of the GEP model are 0.953, 0.183, 14.881 and 67.396 in the training phase, 0.957, 0.177, 14.394 and 64.011 in test phase, respectively. The results indicate that the proposed GEP model outperforms the other models in terms of prediction accuracy and reliability. Following the GEP model, the MEP model achieved the second-highest prediction accuracy models evaluated. The predictions obtained from the proposed GEP and MEP models were found to be statistically similar. Finally, sensitivity and parametric analyses were conducted to evaluate the influence of each input parameter on the predicted punching shear capacity.