A method using information theory to select and rank existing FRP/concrete bond strength models

•The presented method can be used to choose suitable bond strength models•A score based on information theory may be employed as model weight.•The effects of various factors on bond strength models are estimated.•Using the information theory, model suitability is measured. In order to achieve the desired resistance in reinforced concrete structures externally strengthened with fiber-reinforced polymer (FRP), an appropriate design for strengthening is necessary. The designer engineer must have a precise estimate of the bond strength between FRP and concrete before proposing a strengthening design. The bond strength is significantly lower than the FRP composite’s capacity due to the premature debonding of FRP from the concrete surface. Numerous studies have examined the bond strength of FRP/concrete and its influencing factors. As a result, a number of models have been proposed to evaluate the bond strength of FRP/concrete. The methodology, parameters, and data used in these models are distinct; consequently, the accuracy of evaluating these models differs. The selection of an accurate bond strength model presents a challenge for design engineers. Using information theory and the concept of relative entropy, this study compares the relative sufficiency of one model to another. For this comparison, 20 bond strength models and 410 experimental data are selected and compiled. None of these data is utilized in the development of the chosen models. According to the scores each model received from the information theory, the ranking results of the models have been established. The model with the highest score is ranked 1st and the model with the lowest score is ranked 20th. The superior model provides more information about the random parameter uncertainty. The results indicate that five of top seven models are provision models. The majority of the models that are among the 10 models with the highest score (rank 1 to 10) are based on fracture mechanics, whereas the 10 models with the lowest score (rank 11 to 20) are based on regression. Consideration of the effective bond length, along with the number and source of tests used to derive each model, has a significant impact on the ranking and accuracy of the models.