关键词: 纤维增强复合材料梁, 分层, 损伤识别, 直观图像法, 人工神经网络

Abstract: Delaminations in fiber reinforced polymer (FRP) beams were detected and assessed through the changes in natural frequencies before and after the damage occurred in beams. To assess the three delamination parameters, namely, the interface, lengthwise location and size, two different inverse algorithms, i.e. graphical technique and artificial neural network (ANN), were developed to inversely predict the delamination from a series of known frequency shifts. To verify the prediction efficiency and accuracy of the inverse algorithms, both numerical and experimental validation were conducted and the results were compared. The numerical validation results show that both algorithms can predict the delamination parameters successfully, although the accuracy of the graphical technique is noticed to be higher than that of the ANN. Experimental validation using the measured frequencies in literature shows that the graphical technique can predict the delamination with satisfactory accuracy using the measured frequencies which is noise polluted, while ANN is very sensitive to the experimental errors and can hardly predict the delamination with experimental data. In conclusion, it is recommended to use the graphical technique rather than artificial neural network to assess the delamination in FRP beam through frequency shifts.

Key words: FRP beam, delamination, damage detection, graphical technique, artificial neural network