ABSTRACT
Reinforced concrete structures traditionally rely on Bernoulli-Euler beam theory, which assumes linear stress distribution. However, in D-regions areas near support or geometric discontinuities these assumptions break down, and more advanced methods are required, Strut and Tie method (STM) is one of the most popular. STM models rely heavily on mechanical judgment of the engineer, to determine optimal configurations, creating barriers to efficient design. This study examines machine learning algorithms to predict internal forces in deep beams with consider as D-regions. A hybrid ML/AI data bases were evaluated using standard formulation, and validated for three ML/AI approaches, linear regression, with and without polynomial expansion (2nd and 3rd order), and Artificial Neural Networks (ANN). The most advanced linear regression model and the ANN model achieved excellent accuracy, which is appropriate for structural design. The results demonstrate that machine learning can effectively automate STM analysis for deep beams, eliminating expert-dependency and enabling efficient structural design more types of D-regions.
