Identification of FRP Debonding of an I-Girder Bridge Using Statistical Analysis of Impulse-Induced Vibration Responses

In aging civil structures, applying fiber-reinforced polymer (FRP) for retrofitting has been found to be effective. However, debonding of it from the structure may lead to catastrophic failure. Some identification methods for FRP debonding locations have been researched recently. In this paper, the method described uses a non-model-based, unsupervised, and output-only method using statistical analysis. Here, only FRP debonding or damage location was investigated; severity was not measured. 3D finite element modeling of an FRP-bonded I-girder bridge is performed in Abaqus. The damage situation has been simulated by the debonding of FRP. The excitation is an impulse load, and consecutive structural responses are collected at specified sensor-mounted points along the bridge span in flaw-induced damaged and undamaged conditions. A second-order statistical moment has been calculated for both sets of data (damaged and undamaged) in the frequency domain. The differences between the statistical moments of the responses of these damaged and undamaged conditions are calculated. The curvatures of these statistical moment differences (CSMD) are calculated and plotted against the sensor position. From the plotted curve, the FRP debonding location has been identified. Different responses’ (displacement, velocity, acceleration, and strain) performance to identify FRP debonding location has also been studied.