Reliability analysis of strength models for short-concrete columns under concentric loading with FRP rebars through Artificial Neural Network

Abstract

Over the last decade, the utilization of fiber-reinforced polymers (FRP) has been increased due to their versatile properties in concrete columns as a replacement of steel bars and their contribution to the axial load-carrying capacity of short concrete columns (SCC). Different researchers proposed equations to understand the load-carrying capacity of FRP rebars in SCC at the ultimate limit state (ULS). However, the current design practices have their reservation on the use (or taking the contribution) of FRP bars as the main vertical reinforcement in SCC. The present study aims to provide reliability analysis of all well-known physical models (for predicting the effect of FRP in SCC under concentric loading at ULS) through Artificial Neural Network (ANN) models (which do not base on mechanics) and new proposed equation (having a constant parameter to incorporate the lateral confinement effect). For this purpose, a database of 108 samples of SCC with FRP bars under concentric loading only, with detailed information (i.e., cross-section Ag, length of column L, Elastic Modulus of FRP Ef, compressive strength of concrete fc (MPa), longitudinal reinforcement ratio ρl (%), transverse reinforcement ratio ρt (%), and the ultimate axial load Pexp (kN), is collected from previous studies. The predicted axial load values (Ppred) from the ANN model (R = 0.94 and RMSE = 0.32) and proposed equation (R = 0.94 and RMSE = 0.32) exhibited closer results to the experimental values (Pexp)as compared to counterpart physical models. Comparative studies of ratio Pexp/Ppred against the critical parameters exhibited better accuracy of the ANN model and proposed equation as compared to counterpart physical models.