Lamb wave-based delamination identification of composite material during fatigue degradation using the mixture discriminant analysis

This paper studies the delamination identification of composite materials dur ing fatigue degradation. This is realized by active Lamb wave measurements, followed by signal feature extraction, such as the signal maximum amplitude and time of flight. It is found that the signal features for the cases with and without delamination are both scattered and the boundary of features between the two cases is strongly non-linear and complex; thus, the previous Gaussian discriminant analysis (GDA) is not proper for the considered problem. Therefore, this paper proposes to use the mixture discriminant analysis (MDA), which is a generalization of GDA by assuming that the features follow a mixture Gaussian distribution. The classification boundary of MDA is more flexible and thus better fits the guided wave signal features. The proposed delamination identification method is demonstrated via experimental data. It is shown that MDA outperforms the GDA and other classical machine learning methods.