Precise design and characteristics prediction of Ultra-High Performance Concrete (UHPC) based on artificial intelligence techniques

Recently, with the rapid development of artificial intelligence (AI) techniques, there is a strong motivation to promote the intelligent development of the Ultra-High Performance Concrete (UHPC). To achieve this goal, this paper addresses an approach for precise design and characteristics prediction of UHPC by employing the Modified Andreasen and Andersen (MAA) model and Genetic Algorithm based Artificial Neural Network (GAANN) technique. Herein, 80 mixtures in total are conducted as a training dataset, and then a GA-ANN model is created for characteristics prediction of UHPC, which exhibits significant superiorities in fitting goodness and prediction accuracy compared to other classical prediction models. Furthermore, a prediction software based on GA-ANN technology with an efficient Graphical User Interface (GUI) is developed. Finally, a novel method for mix-design of UHPC by the use of MAA and GA-ANN models is proposed as follows: 1) conduct preliminary mixture design by MAA model; 2) further optimize the mixture by GA-ANN according to the property requirements. In general, a new UHPC with dense particle packing skeleton can be precisely designed by this method, which effectively verifies the feasibility of the application of AI technique in the UHPC field.