Search academic texts

Information × Registration Number 2125U004018, Article popup.category Стаття Title Machine learning for enhanced durability: A robust BR-ANN model for predicting concrete shrinkage (AI translated) popup.author Quoc P.D.Long T.N.Anh P.T.Thuc P.X.Ha N.T.Quoc P.D.Long T.N.Anh P.T.Thuc P.X.Ha N.T. popup.publication 01-01-2025 popup.source_user Сумський державний університет popup.source https://essuir.sumdu.edu.ua/handle/123456789/100705 popup.publisher Sumy State University Description Concrete shrinkage poses a significant threat to the durability of concrete structures, particularly in hot and humid climates where conventional models often fail. This study focuses on traditional concrete (the most prevalent construction material in regions like Vietnam) to develop a robust, data-driven framework for predicting 365-day shrinkage. A Bayesian regularization artificial neural network (BR-ANN) was trained using a high-quality, regionspecific dataset of 120 experiments. The model utilized 11 input parameters encompassing concrete composition, mixture proportions, mechanical properties, and specimen geometry. Its performance was rigorously validated, demonstrating exceptional predictive accuracy with a coefficient of determination of 0.92 and a root mean square error of 52.4 με microstrain. To further affirm its robustness and generalizability despite the limited dataset size, a 10-fold cross-validation was performed, yielding stable and consistent results. The BR-ANN model significantly outperforms established international codes (CEB-FIP, Eurocode 2, and ACI 209R-92), which proved unreliable for local conditions. To bridge the gap between advanced modeling and practical application, a simplified mathematical formula and a user-friendly graphical user interface (GUI) were derived. Overall, the research not only provides a reliable predictive tool but also highlights the significant potential of localized, data-driven approaches for solving regionspecific engineering challenges, contributing to the design of more durable infrastructure. popup.nrat_date 2026-04-05 Close