Locating and sizing of delamination in composite laminates using computational and experimental methods

This paper describes an application of Strip Element Method (SEM) and adaptive Multilayer Perceptron Networks (MLP) for inverse identification of interfacial delaminations in carbon/epoxy laminated composite beams. Displacement responses calculated using SEM for laminated beams containing predetermined delamination parameters (i.e. delamination location, depth and length) are used as training data for the MLP. Once the MLP is trained, the MLP networks are then employed for inverse determination of delamination using experimental displacement responses measured with a scanning laser vibrometer. The outputs of the initially trained MLP, which are the reconstructed delamination parameters, are then compared with the experiments data. The MLP will be re-trained and re-used for parameter reconstruction until satisfactory results are obtained. Examples show that the procedure performs well for the determination of a wide range of values for the location, depth and length of delamination. (C) 2001 Elsevier Science Ltd. All rights reserved.