A new and robust hybrid artificial bee colony algorithm – ANN model for FRP-concrete bond strength evaluation

Abstract

The present study describes the bond strength estimation between the FRP strip and the substrate concrete using machine learning approaches. To provide a precise model, a database including 656 single and double-lap direct shear test results was compiled from the literature. Artificial neural networks (ANN) and the combination of ANN with the artificial bee colony optimization algorithm (ABC-ANN) were implemented for the model development. The concrete compressive strength, the elasticity modulus of FRP strip, FRP thickness, the width of the FRP strip, and the width of concrete block were considered as input parameters of the models. The results of the ANN and the hybrid ABC-ANN were compared with those of existing models and international procedures. It was indicated that the accuracy of the proposed models outperforms the existing models. The correlation coefficients of the ABC-ANN and ANN approaches are 0.97 and 0.93, respectively. The robustness of the models was investigated using the distribution of absolute error values. It was demonstrated that the proposed ABC-ANN model is robust and resulted in the least error values, distributed in small ranges (less than 10%), compared to ANN and other methods. Using the hybrid ABC-ANN, a straightforward formulation for bond strength evaluation was proposed.