ABSTRACT
This study addresses the challenge of quantifying the complex and non-linear correlations between the technical properties of ultra-high-Performance concrete (UHPC) and its mixture composition through conventional statistical methods. To tackle this, this study employs five advanced machine learning methods: Random Forest (RF), Long-Short-Term Memory, Convolutional Neural Network, K-Nearest Neighbor, and Adaboost. A dataset comprising 308 UHPC compressive strength tests with varying amounts of cement, silica fumes, superplasticizers, ground granulated blast furnace slag, fine aggregates, coarse aggregates, age, and temperature was used to train these models. The models’ performance was validated using statistical metrics, regression error characteristic curve, SHAP analysis, and uncertainty analysis. Among the models, the RF model achieved the highest prediction accuracy with an R2 of 0.9877 and an RMSE of 0.0323. This study also introduces an open-source graphical user interface based on the RF model, providing engineers with a practical tool for estimating the compressive strength of UHPC and facilitating mix proportioning decisions. This interface aims to bridge the gap between machine learning predictions and practical applications in UHPC mix design.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Supplementary material
Supplemental data for this article can be accessed online at https://doi.org/10.1080/24705314.2024.2385206.