Radial Basis Function Network-Based Approach for Determining Interaction Behavior of Reinforced Concrete Rectangular Columns

Abstract

Determining the structural behavior using neural network (NN)-based approach is becoming common since it is found as a quicker and an easier method. The objective of this study was to investigate whether NN-based model can be used to determine the interaction diagram of confined and unconfined reinforced concrete (RC) columns. In the application of NN model, the radial basis function network (RBFN) was preferred after various tests. RBFN model was developed, trained and tested in MATLAB-based program. The training and validation data sets were obtained from a commercial software package, Xtract, in which cross-sectional capacity of RC sections can be calculated for both confined and unconfined case. The validity of the proposed RBFN model was verified by comparing the results obtained from RBFN models and those obtained from Xtract software package. It was demonstrated that RBFN-based model has high accuracy to determine interaction diagram of confined and unconfined RC columns with rectangular cross section. As a result of this study, it became clear that using RBFN model is more suitable for confined sections. The proposed model can determine interaction diagram for both confined and unconfined sections in the safety range required by the codes.