Application of artificial neural networks to predict the bond strength of FRP-to-concrete joints

A Back-Propagation Neural Network (BPNN) model for predicting the bond strength of FRP-to-concrete joints is proposed. Published single-lap shear test specimens were used to predict the bond strength of externally bonded FRP systems adhered to concrete prisms. A database of one hundred and fifty experimental data points from several sources was used for training and testing the BPNN. The data used in the BPNN are arranged in a format of six input parameters including: width of concrete prism; concrete cylinder compressive strength; FRP thickness; bond length; bond width (i.e. FRP width); and FRP modulus of elasticity. The one corresponding output parameter is the bond strength. A parametric study was carried out using BPNN to study the influence of each parameter on the bond strength and to compare results with common existing analytical models. The results of this study indicate that the BPNN provides an efficient alternative method for predicting the bond strength of FRP-to-concrete joints when compared to experimental results and those from existing analytical models.

► A BPNN model for predicting the bond strength of FRP-to-concrete joints is proposed.
► The results of the BPNN compared with experimental results and those from five existing analytical models.
► BPNN is a convenient tool for developing a parametric study.
► The application of BPNN techniques provide a reliable and simple model for predicting the bond strength.