Data-driven ultimate condition prediction of UHPC columns subjected lateral confinement using Artificial Neural Network (ANN) approach

Despite the superior mechanical properties of Ultra-High Performance Concrete (UHPC) in comparison to conventional concrete, UHPC is still not widely used in the construction industry due to the lack of design standards. This research mainly focuses on using advanced computational tools for the prediction of ultimate stress and its corresponding strain for confined UHPC columns, to facilitate the process of analysis and design, and to support further development of standards. Characteristic parameters of confined UHPC column samples were extracted from available literature as the input, and the corresponding ultimate stress and strain as the output. Artificial Neural Network (ANN) technique was adopted to process the input parameters in order to train and predict the output. In the regression analysis, the trained networks exhibited the coefficients of determination (R2) of 0.995 and 0.964 for the prediction of ultimate stress and corresponding strain respectively. The weights and biases within the best performance ANN from the k-fold cross validation can be extracted to perform mathematical solutions for the prediction in design.