Abstract
Ultra-high-performance concrete (UHPC) is a new design material with superior physical and mechanical properties. It has gained more and more attention in many areas, especially civil engineering. The compressive strength (CS) of UHPC design is one of them. The mechanical properties of concrete can be evaluated by Machine learning (ML), a kind of artificial intelligence (AI). However, determining the appropriate parameters of the artificial neural network is a confusing and complicated task. Many metaheuristic algorithms have been used to optimize these parameters, and some of them have been selected to optimize the CS prediction of UHPC in this research. They are as follows: Arithmetic Optimization Algorithm, African Vulture Optimization Algorithm, Generalized Normal Distribution Optimization, Stochastic Paint Optimization. The best prediction accuracy of CS is obtained by the Stochastic Paint Optimization-ANN model, achieving the R2 0.8655 on the testing dataset. It is a superior method that enhances CS's ability to predict UHPC.