Transport and Communications Science Journal, Vol 71, Issue 2 (02/2020), 154-166 154 Transport and Communications Science Journal ARTIFICIAL NEURAL NETWORK BASED MODELING OF THE AXIAL CAPACITY OF RECTANGULAR CONCRETE FILLED STEEL TUBES Hai-Bang Ly*, Thuy-Anh Nguyen University of Transport Technology, 54 Trieu Khuc, Thanh Xuan, Hanoi, Vietnam ARTICLE INFO TYPE: Research Article Received: 13/02/2020 Revised: 28/02/2020 Accepted: 28/02/2020 Published online: 29/02/2020 https://doi.org/10.25073/tcsj.71.2.10 * Corresponding author Email: banglh@utt.edu.vn; Tel: 0966661185 Abstract. Concrete filled steel tubes (CFST) with many advantages compared with conventional structural members made by steel or reinforced concrete has been widely applied in the field of civil engineering. The axial capacity of CFST, which is considered as the most important mechanical property, depends on many constituting factors such as the mechanical properties of materials or the cross-section of the tubes. In this study, the Artificial Neural Network (ANN) with the Levenberg-Marquardt algorithm is used to predict the axial capacity of CFST with rectangular cross-section. A number of 99 samples, collected from published international studies, is divided into two parts: the training part (69 samples), used to construct the BPNN black-box, and 30 samples in the testing part to evaluate the performance of the BPNN model. The main input parameters used in this study are the height of steel tube, width of steel tube, the thickness of steel tube, length of the column, the yield stress of steel and the compressive strength of concrete. The results show that the ANN algorithm is a good predictor for the problem with low error and high correlation (R=0.99) between the experimental and predicted results derived from the ANN algorithm. Keywords: concrete filled steel tube, artificial neural network, axial compression, axial capacity. © 2020 University of Transport and Communications