In situ damage classification for composite laminates using Gaussian discriminant analysis

Abstract

An investigation was performed to develop a damage classification method to characterize sensor data from built-in piezoelectric actuators on laminated composites in terms of matrix micro-cracking and delamination. Traditional signal processing techniques (time-domain analysis and short-time Fourier transform) combined with Gaussian discriminant analysis were proposed to characterize damage in composite plates in this study. Composite coupons of different layup configurations with surfaced mounted arrays of piezoelectric actuators and sensors were subjected to cyclic loading to induce matrix micro-cracking and delamination. Sensor data as well as X-ray images were acquired through the coupon’s life. These experimental data were used to train a classifier as well as test the performance and robustness of the learned model. Classification performance was measured via 75–25 holdout and leave-one-sample-out cross-validation. Results show that the proposed method has a misclassification rate of 21%, with high precision and recall values, and it is sensitive to layup configuration.