Improving Shear Strength Prediction in Steel Fiber Reinforced Concrete Beams: Stacked Ensemble Machine Learning Modeling and Practical Applications

Existing machine learning (ML) models often encounter challenges in accurately predicting the shear strength of steel fiber reinforced concrete (SFRC) beams, mainly due to a lack of generalization. This study introduces an advanced stacked ensemble ML architecture to overcome this limitation by utilizing a comprehensive data set of 394 experimental observations and a 20-feature matrix. The model exhibits exceptional performance with a mean absolute error of 0.391 and a correlation coefficient (R2) of 93.7%, and surpasses traditional ML algorithms. Furthermore, a sensitivity analysis of the developed model yields that shear strength is highly responsive to the shear span-to-effective depth ratio, with an increase from 1 to 4 resulting in a significant reduction (about 50%) in strength. Increasing the percentage of longitudinal steel from 1 to 2% leads to a 14.6% gain, whereas doubling its yield strength has a more modest 3.7% effect. Increasing the compressive strength of concrete from 25 to 50 MPa, notably increases the shear strength by 19.6%. Fiber length, diameter, and aspect ratio exhibit varying impacts, with shear strength most influenced by the fiber volume fraction, which leads to a peak enhancement of 30.7% at 2% fibrous volume; however, the tensile strength of fibers minimally affects the shear strength. Additionally, this research presents a simplified empirical model to predict the shear strength of SFRC beams based on the key determinants. This model employs the iterative Gauss–Newton algorithm, demonstrates reasonable predictive capability, and boasts an R2 of 83.3% and mean prediction-tested strengths of around 1.039. The practical implications of these findings are substantial for the construction industry as they enable a more accurate and reliable design of SFRC beams, optimize material usage, and potentially reduce construction costs as well as enhance structural safety.