Estimating the punching shear capacities of concrete slabs reinforced by steel and FRP rebars with ANN-Based GUI toolbox

Abstract

In this paper, the punching shear capacity of reinforced concrete (RC) slabs reinforced by steel and fibre-reinforced polymer (FRP) rebars jointed to circular, square, and rectangular columns was estimated by various artificial neural networks (ANNs). A large experimental database, including 164 tests, was compiled to achieve this goal. The influential input parameters contained the cross-section area of the column, the perimeter of the critical section in the RC slab, the effective depth of the RC slab, the modulus of elasticity, the reinforcement ratio of steel and FRP rebars, and the compressive strength of concrete. The results showed that considering 8 neurons in the single hidden layer, named ANN-6-8-1, and respectively 15 and 5 neurons in the first and second hidden layers in multi-layer models, named ANN-6-15-5-1, were the optimized configurations. The results showed the ANN-6-15-5-1 model with an R value of 0.9925, and a MAPE error value of 7.48% is more accurate. Among the existing models, the Ospina et al. and Metwally models, respectively, with R values of 0.9473 and 0.9386 and MAPE values of 18.77% and 15.40 %, were the best ones. Eventually, a graphical user interface (GUI) toolbox is provided to enable the user to calculate the punching shear capacity of RC slabs in practice.