A hybrid neuro-swarm model for shear strength of steel fiber reinforced concrete deep beams

Abstract

Despite the wide application of steel fiber to augment the strength and durability of concrete, predicting the shear strength capacity of steel fiber-reinforced concrete in deep beams has been a challenge for civil engineering applications. This is due to the complex interactions of various parameters such as the beam dimensions, material strength, and reinforcements. To address this concern, a hybrid Neuro-Swarm model was developed using 116 experimental datasets. With the least root mean squared error (RMSE) of 10.02 and the highest correlation coefficient (R) of 0.997, the proposed model demonstrated remarkable performance indicators exhibiting superior prediction performance. The Analysis of Variance (ANOVA) test revealed no significant difference between predicted and experimental shear strength values. Furthermore, with relative importance values of 13.573% and 13.565%, the beam depth and yield strength of the reinforcing bars were found to be the most significant predictors influencing the shear strength. When compared to previously established shear strength prediction models in the literature, the proposed model emerged as the best and most robust. The results demonstrated that the new hybrid Neuro-Swarm model may be used to accurately forecast the shear strength of steel fiber reinforced concrete deep beams.