Identifiability of the Parameters Contained in a Damaging Cyclic Cohesive Zone Model for Bonded CFRP-Steel Interfaces

The use of externally bonded (EB) carbon fiber-reinforced polymers (CFRP) is an effective solution for strengthening metallic elements, restoring their load-bearing capacity and, in particular, extending the fatigue lifetime of existing elements. In these applications, failure usually occurs with debonding of the CFRP plate from the steel substrate, making the interfacial response crucial to the effectiveness of the bonded system. Cohesive zone models represent a valid numerical approach to describe the interface between two adherends, such as steel and composite. However, these models depend on many parameters, whose correct calibration is necessary to guarantee a reliable use. This paper investigates the parameter identifiability of a damaging cyclic cohesive zone model using an inverse analysis approach. Virtual data, generated via a finite element (FE) model of the experimental setup, are employed to rigorously assess the capacity of the inverse operator to identify a unique set of model parameters, based on given experimental information. Specifically, the FE model simulates the response of single-lap direct shear tests and the experimental data used as input to the inverse algorithm consists of axial strain distributions, measured on the CFRP external surface, and global slip measurements. The adopted stochastic procedure is based on the use of noise perturbation and metamodel reduction techniques, which reduce the computational cost that is particularly high in fatigue simulations. The results offer clear criteria for optimizing laboratory measurement setups by determining the minimum experimental data needed for the present identification purposes. While global slip measurements generally yielded a well-posed inverse problem, adding CFRP axial strain measurements proved crucial in a few cases to correctly identify the sought parameters.