Artificial Neural Network (ANN) Based Prediction of Ultimate Axial Load Capacity of Concrete-Filled Steel Tube Columns (CFSTCs)

Concrete-filled steel tube columns (CFSTCs) are preferred due to enhanced ductility and energy absorption. The capability of an artificial neural network (ANN) based analytical model on estimating the ultimate load capacity of circular CFSTCs under axial loadings has been investigated in this study. To provide a better prediction in modeling, 150 comprehensive experimental data were obtained from circular CFSTC's geometrical and mechanical properties, such as height, diameter, thickness, the yield stress of steel, unconfined concrete strength, Young's modulus of steel and concrete, etc., were examined. The backpropagation-training practice available in ANN was used to update the weights of each layer based on the network output error. For feedforward-backpropagation, the Levenberg-Marquardt algorithm was employed. The effectiveness of the ANN model was developed using general-purpose software MATLAB (R) by training and predicting the ultimate load capacity of circular CFSTCs. Finally, about 75% of the data were used for ANN training, and the remaining 25% was used for testing the ANN model. The results show that the predicted values of ultimate load capacity using the ANN model agree well with that of the corresponding experimental observations, and the percentage difference is about +/- 10%. Additionally, a new engineering index, a20-index, was predicted to further verify the reliability of the model. The findings of this article are new and will significantly contribute to the existing technology of ANN-based modeling in composite construction.