Predicting bond strength between EBROG‐FRP and concrete using ANN

The rehabilitation of deteriorated concrete structures using fiber‐reinforced polymer (FRP) sheets bonded to concrete through the externally bonded reinforcement on grooves (EBROG) method has become increasingly popular, with its effectiveness closely tied to its bond with concrete. A novel artificial neural network (ANN) method has been successfully produced to predict bond strength with an excellent accuracy rate. This model utilizes a dataset of approximately 445 data points and considers dominant parameters, such as compressive strength of concrete, stiffness of FRP sheet, bond length, width, and number of grooves, as well as groove dimensions. Using software, the ANN model was built, trained, and tested with seven input variables and one output. Descriptive statistical analysis, including Taylor charts, helped validate the model's performance and stipulate its superiority over existing literature models. The proposed ANN model reached a solution prior to 181 epochs while exhibiting excellent predictive performance with minimal root mean square error. The coefficient of correlation ( R ) for fitting and prediction datasets was estimated at 0.95 and 0.89, respectively. The trends of the predicted bond strength align well with those experimentally observed or provided by various literature models. Detailed statistical and sensitivity analyses were carried out to prove the practicality of this model in estimating bond strength in EBROG‐FRP joints.