The data-driven research on the autogenous shrinkage of ultra-high performance concrete (UHPC) based on machine learning

This paper performs the data-driven research on the autogenous shrinkage of Ultra-High Performance Concrete (UHPC) based on multiple machine learning algorithms. The autogenous shrinkage prediction model of UHPC with high prediction accuracy (R2 = 0.89) and strong generalization capability is established based on Gradient Boosting (GB) algorithm. The Graphic User Interface (GUI) of UHPC autogenous shrinkage prediction is designed, which assists in scientific research and engineering applications. Cement, silica fume, water, superabsorbent polymer, and expansive agent contents that are closely related to hydration process have prominent impacts on autogenous shrinkage of UHPC, while the fly ash, steel fibers, and sand contents show the slight impacts. The additions of superabsorbent polymer, expansive agent, steel fibers, sand, and superplasticizer effectively reduce the autogenous shrinkage of UHPC. Lower water to cement ratio causes larger autogenous shrinkage of UHPC.