Machine Learning prediction of ultimate strain of CFRP/GFRP-RC column with lap spliced rebars subjected to cyclic loads

Abstract

Fiber Reinforced Polymers (FRPs) are widely being used to retrofit steel and concrete structures due to their high resistance to corrosion and high mechanical qualities. To extend the application of FRPs in the construction industry, there is a need to provide a powerful model to predict the load-carrying capacity of FRP concrete elements such as beams and columns. Herein, different techniques were applied to predict the ultimate strain of FRP rectangular concrete columns subjected to cyclic loads using machine-learning models. A comprehensive database of 318 specimens available in the literature was collected. Six Artificial Intelligence models including five machine learning models named as K-Nearest Neighbors (KNN), and Decision Tree (DT), CatBoost (CB), AdaBoost (AB), Random Forest (RF) and one deep learning model named Artificial Neural Network (ANN) were considered. The result showed that DT, and RF models are able to predict the ultimate strain of the column with high accuracy of 96.4% and 96.5%, respectively.