Parametric study of retrofitted reinforced concrete columns with steel cages and predicting load distribution and compressive stress in columns using machine learning algorithms

Abstract

Recently, the use of reinforced concrete (RC) structures is becoming very common worldwide. Because of earthquakes or poor design, some of these structures need to be retrofitted. Among different methods of retrofitting a structure, we have utilized a steel cage to support a column under axial load. The numerical modeling of a retrofitted column with a steel cage is carried out by the finite-element method in ABAQUS, and the effectiveness of the number of strips, size of strips, size of angles, RC head, the strips’ thickness, and the steel cage’s mechanical properties are studied on 15 different case studies by the single factorial method. These parameters proved to be very effective on the load distribution of the column because by choosing the optimum case, lower amounts of force are born by the column. By increasing the number of strips, the steel cage would reach 52% of the total load. This value for the size of strips and angles’ size is 48 and 50%, respectively. However, the thickness of the strips does not have a significant effect on the load bearing of the column. In order to fully predict the load distribution of the retrofitted columns, the data of the present study are utilized to propose a predictive model for N c/P FEM and N c/P FEM using artificial neural networks. The model had an error of 1.56 (MAE), and the coefficient of determination was 0.97. This model proved to be so accurate that it could replace time-consuming numerical modeling and tedious experiments.