Characterization of model uncertainty of adhesively bonded CFRP-to-steel joints

Abstract

This paper presents a critical characterization of the uncertainty of bond strength models of adhesively bonded carbon fiber-reinforced polymer (CFRP)-to-steel joints. A model factor ε was defined as the ratio of a tested ultimate load to the corresponding predicted value. A database consisted of 414 CFRP-to-steel single/double-lap shear tests was collected and two frequently used bond strength models were used in the evaluation. The statistics of ε indicated a high level of both the mean value and coefficient of variation (COV). The model factor was found to be highly dependent on input parameters rather than purely random. A systematic part (f) was consequently introduced by a multiple regression analysis to remove this kind of correlation, leaving a residual lognormally distributed random variable (ε*). After the critical characterization, performance of the model factor was significantly improved. Afterwards, a reliability analysis was performed to compare the model performance given the condition whether the model factor was properly considered or not. The merit of the critical calibration of the model factor is highlighted in the reliability analysis. The resistance factor for the two bond strength models was calibrated given a target reliability index βT of 3.5. The sensitivity analysis indicated that the adhesive tensile strength is the most influential parameter to the reliability index among all the input parameters.