Prediction of axial bearing capacity of confined concrete columns by using artificial neural networks

The axial bearing capacity of concrete columns confined by stirrups is a significant mechanical property in design and evaluation of concrete structures. However, due to the complicate mechanics between concrete, longitudinal reinforcement and stirrup, it is relatively difficult to accurately estimate the axial bearing capacity of confined concrete columns through traditional analysis methods. A new and robust approach for predicting axial bearing capacity of confined concrete columns using artificial neural networks (ANN) was explored in this paper with a reliable test database containing 208 specimens conducted in this paper and published literatures. The test database had a large application range, covering unconfined concrete compressive strength with 27-115.9 MPa and stirrup yield strength with 288-1000 MPa. Seven key parameters were considered as input variables, including size of core concrete, unconfined concrete compressive strength, stirrup volumetric ratio, stirrup yield strength, longitudinal reinforcement ratio, yield strength of longitudinal reinforcement, and confinement type. Moreover, to optimize ANN model configuration, the importance of different input parameters on axial bearing capacity of confined concrete columns was investigated. Furthermore, ANN calculating formulas about axial bearing capacity of concrete columns confined by stirrups were established based on the proposed ANN model. Finally, the ANN model and the proposed ANN calculating formulas were evaluated and verified compared with experimental results and four existing theoretical models, indicating that the proposed calculating formulas have enough accuracy, reliability and application in predicting of the axial bearing capacity of confined concrete columns.