Novel informational bat-ANN model for predicting punching shear of RC flat slabs without shear reinforcement

While design codes provide guidelines to prevent brittle punching shear failures in flat reinforced concrete (RC) slabs, they are associated with high inaccuracy. This study scrutinizes existing design provisions, highlighting its features and limitations. Sensitivity analysis is then used to identify the influential mechanical and geometric parameters. Subsequently, an artificial neural network coupled with a metaheuristic Bat algorithm (Bat-ANN) is used to develop a hybrid model for estimating punching shear strength. Several statistical metrics revealed that the Bat-ANN model achieved superior predictive accuracy. The novel hybrid model was deployed to assess the influence of key parameters affecting punching shear strength, including the slab effective depth, concrete strength, reinforcement ratio, reinforcement yield strength, and width of the square loaded area. The analysis identified the importance of the flexural reinforcement, which is not typically considered in estimating punching shear strength. Subsequently, using the supervised machine learning method through the EUREQA software, a new regression expression was proposed to estimate the punching shear resistance of flat slabs. This hybrid computational intelligence model could be integrated in future automated design platforms of RC structures.