Prediction of the load-shortening curve of CFST columns using ANN-based models

Abstract

Artificial neural network (ANN) as a machine learning (ML) technique has been successfully applied in engineering applications such as structural dynamics and structural design. It has also received considerable attention for the design of concrete-filled steel tube (CFST) columns. However, the application of ANN method to CFST columns is mainly restricted to the prediction of the ultimate strength. In this paper, a novel approach to predict and plot the complete axial load-shortening curve of concentrically loaded rectangular and circular CFST columns using ANN method is presented. To train the networks, a database including 392 test results of rectangular and circular CFST columns with their corresponding data points of the load-deflection curves is compiled. In addition, 1152 finite element models (FEMs) are generated and analysed using ABAQUS to expand the training data and address data gaps in the experimental database. The validity of the developed FEMs is verified by experimental data. Based on the trained ANN models, a MATLAB-based graphical user interface (GUI) is also developed to provide a convenient tool for users to predict and plot the axial load-shortening response of rectangular and circular CFST columns. Using the developed GUI, a parametric study is also conducted to verify the accuracy of the ANN models in predicting the behaviour of CFST columns fabricated with different geometric and material properties. The results show that the developed ANN models can accurately predict the load-deflection response of CFST columns.