APPLICATION OF A NEURAL NETWORK IN CALCULATIONS OF PUNCHING SHEAR CAPACITY THROUGH FLAT SLABS OF REINFORCED CONCRETE SLABS WITHOUT TRANSVERSE REINFORCEMENT

Abstract

The punching shear capacity of reinforced concrete floor slabs is one of the most difficult kinds of resistance of reinforced concrete structures because it has the brittle form of destruction with the instantaneous separation of the slab from the column. The accuracy of punching shear capacity estimation has the key importance for the design of reinforced concrete structures. In spite of the fact that experimental and theoretical researches of shear resistance under punching have been carried out for more than a hundred years, a unified and reliable calculation model hasn't been worked out yet. This can be explained by the complexity of the stress-strain state which occurs under load at the junction of the floor slab and the column. Semi-empirical and fully empirical computational models are most commonly used to estimate punching shear capacity because of their simplicity and competitive prediction accuracy. This paper investigates the modeling of punching shear resistance using a fully coupled neural network, considered as an analytical alternative to existing computational models. Using a database compiled from numerous experimental works, the validity of existing calculation models introduced in some existing regulatory documents and punching shear capacity predictions obtained with a neural network has been evaluated. The punching shear capacity values obtained by using a neural network were more accurate than those obtained by using the calculation models discussed in this paper.