Evaluating punching shear strength of slabs without shear reinforcement using artificial neural networks

Punching shear failure of concrete slabs poses a significant risk in many concrete structures. This mode of failure can be brittle and catastrophic. The ability to accurately estimate the punching shear capacity of slab column connections in existing structures is essential, especially in evaluating the suitability to new loads added to a building. Punching shear has been studied, both experimentally and analytically. However, due to the number of parameters involved and the complexities in modeling, current approaches used to estimate the punching shear capacity of reinforced concrete (RC) slabs include mechanical models and design code equations. Mechanical models are complex, while design code equations are empirical. This study investigates the ability of artificial neural networks (ANN) to predict the punching shear strength of concrete slabs. The parameters considered to be the most significant in punching shear resistance of RC slabs were: concrete strength, slab depth, shear span to depth ratio, column size to slab effective depth ratio and flexure reinforcement ratio. Using a large and homogenous database from existing experimental data reported in the literature, the ANN model is able to predict the punching shear capacity of slabs more accurately than were the code design equations.