For concrete beams strengthened with bonded FRP plates, interfacial debonding of the plate is the major failure mechanism. The present investigation consists of two independent parts: the study of effect of flexural cracking on interfacial debonding and the development of an empirical design guideline for interfacial debonding....[ Read more ]

For concrete beams strengthened with bonded FRP plates, interfacial debonding of the plate is the major failure mechanism. The present investigation consists of two independent parts: the study of effect of flexural cracking on interfacial debonding and the development of an empirical design guideline for interfacial debonding.

Flexural cracking often occurs along the concrete beam, causing significant changes in the stress distribution at the FRP/concrete interface, particularly near the plate end region. In this thesis, experimental and analytical results are presented to illustrate the change of interfacial stress distributions in two different cases: (i) the case with flexural cracks approaching the plate end at failure load and (ii) the case with flexural crack right at the plate end at failure load. In the first case, the shear lag analysis for interfacial shear stress is performed with consideration of effective flexural stiffness reduction for the concrete beam. The significance of concrete section cracking on the plate end stress for various combinations of design parameters is studied and discussed. In the second case, based on observations from the experimental and analytical results, the average shear stress over a given distance from the plate end is found to be a potential parameter governing delamination failure. An approximate method to calculate such an average stress is proposed.

Comprehensive review of the literature shows that no existing analytical models can accurately predict delamination and debonding failures under general conditions. To develop an empirical design approach, the neural network is adopted. An experimental database is compiled to provide training and validating data sets for the neural network study. Finally, design charts with correction factors are developed based on the trained network and parametric studies are performed to examine the validity of the trends.