Predicting axial load capacity of CFRP fire-damaged RC columns through DANN
Muhammad Noman, M. Yaqub, Muhammad Salman, Muhammad Faizan, Shakouri Mahmoudabadi, Afaq Ahmad
Abstract
Abstract This paper presents the results of a deep neural network (DNN) model predicting the axial load-carrying capacity of fire-damaged concrete square and rectangular reinforced concrete columns subsequently repaired using carbon fiber-reinforced polymer (CFRP) composites. An experimental database comprising 143 column specimen samples was collected from the literature. Thirteen parameters were noted, including Length of Columns (mm), Width of Columns (mm), Corner Radius (mm), Height of Column (mm), Initial Compressive Strength of Concrete (MPa), Initial Tensile Strength of Steel (MPa), Longitudinal Reinforcement Ratio (As/Ag, %), Temperature of Fire (°C), Fire Exposure Time (Minutes), Number of Layers of CFRP(Count), Thickness of CFRP (mm), and Tensile Elastic Modulus of CFRP (GPa). These parameters served as inputs for the model, with the final axial compressive strength of the CFRP-repaired column considered as the target/output. DNN models were calibrated for numerous numbers of neurons in the hidden layers to achieve optimization. The proposed DNN model demonstrated strong agreement with the experimental database, exhibiting an overall correlation factor (R) of 0.99946. The predicted outcomes of the proposed model closely align with the experimental findings, as evidenced by a thorough comparison with prior experimental investigations. Consequently, this model holds potential value for engineers seeking speedy assistance in repairing fire-damaged structures using CFRP composites.