Predicting flexural strength in fiber-reinforced UHPC via random forest

The incorporation of fibers into Ultra-High-Performance Concrete (UHPC) significantly enhances its flexural strength and ductility, making it a desirable material for high-performance structural applications. However, predicting the flexural behavior of fiber-reinforced UHPC remains challenging due to the complex interactions among constituent materials. This complexity increases further with the partial replacement of cement by supplementary cementitious materials (SCMs), which alter matrix reactivity and microstructure. To address these challenges, this study employs Random Forest (RF) regression to predict the ultimate flexural strength of UHPC, incorporating mixtures with diverse SCM combinations and up to two different fiber types. A dataset of 550 experimental mixtures, comprising 41 input variables, was used to train and validate the model. Results highlight the critical influence of fiber type and dosage, as well as matrix parameters such as cement and silica fume content, water-to-binder ratio, and aggregate size. Partial Dependence Plots (PDPs) were used to visualize these effects, offering interpretable insights into the flexural behavior of UHPC under bending loads.