Evaluation and prediction of bond strength of GFRP-bar reinforced concrete using artificial neural network optimized with genetic algorithm

Assessment of bond behavior of glass fiber-reinforced polymer (GFRP) bars to concrete plays an important role in design and implementation of the polymer-matrix composites (PMCs). This study develops an optimized modeling strategy that harnesses the strong nonlinear mapping ability of artificial neural network (ANN) with the global searching ability of genetic algorithm (GA) for bond strength prediction. The factors that affect the bond strength were identified from the test data of 157 beam-test specimens in the literature, in terms of bar conditions (bar diameter, surface, position and embedment length), concrete (thickness of concrete cover and concrete compressive strength), and confinement from transverse reinforcements. Comparison of the bond strengths predicted by the proposed optimized ANN-GA model with test results showed a higher accuracy with less scatter compared to the conventional ANN model.