Reinforced concrete ultimate bond strength model using hybrid neural network-genetic algorithm

Abstract

The bond strength in reinforced concrete is defined as resistance to slipping of the reinforcing steel bars from the concrete. This slipping resistance is one of the most important features in the performance of the reinforced concrete structure, particularly to its failure mode and mechanisms. In this study, a hybrid model using Artificial Neural Network (ANN) and Genetic Algorithm (GA) has been developed to predict and optimize the ultimate bond strength (tu) between the reinforcing bar and the concrete based on numerous variables that influence this property. These variables include 28-day cube compressive strength f'c), concrete cover (c), the diameter of reinforcing bar (db), embedded length (Lm), rib height (hr), and rib spacing (sr). ANN was utilized into the prediction of bond property between the reinforcing bar and concrete based on the aforesaid input variables. The ultimate bond strength predicted by ANN model exhibited reasonably accurate and good agreement with the experimental values. On the other hand, GA was deployed in the search for the optimal combination of the input variables which resulted in high bond strength performance. Optimization results showed that smaller hr and sr developed high quality of the bond between the reinforcing steel bar and the concrete.