A Data-Driven Model to Predict the Self-healing Performance of Ultra High-Performance Concrete

Ultra High-Performance Concrete (UHPC) demonstrates superior mechanical properties and durability compared to ordinary concrete. The self-healing ability of UHPC has also been widely observed in numerous studies. Consequently, the development of a predictive model capable of forecasting the evolution of UHPC's self-healing performance under diverse exposure conditions has become crucial in order to reliably and effectively exploit this potential in the framework of a durability-based design of UHPC structural applications. In this study, an extensive dataset was collected and established through experimental tests focusing on the crack-sealing performance of UHPC. These tests simulated the conditions in which pre-cracked UHPC specimens were subjected to sustained tensile stress, while simultaneously exposed to exposure environments, including fresh water, salt water and geothermal water. Based on this dataset, a mathematical regression model was constructed, incorporating significant factors such as crack width, exposure environment, and exposure time for UHPC. The results demonstrate the remarkable accuracy of the proposed mathematical model in predicting the self-healing ability of UHPC.