Structural vibration-based classification and prediction of delamination in smart composite laminates using deep learning neural network
Asif Khan, Dae-Kwan Ko, Soo Chul Lim*, Heung Soo Kim* (* : corresponding author)
This paper proposes a Convolutional Neural Network (CNN) based approach for the classification and prediction of various types of in-plane and through-the-thickness delamination in smart composite laminates using low-frequency structural vibration outputs. An electromechanically coupled mathematical model is developed for the healthy and delaminated smart composite laminates, and their structural vibration responses are obtained in the time domain. Short Time Fourier Transform (STFT) is employed to transform the transient responses into two-dimensional spectral frame representation. A convolutional neural network is incorporated to distinguish between the damaged and undamaged states, as well as various types of damage of the laminated composites, by automatically extracting discriminative features from the vibration-based spectrograms. The CNN showed a classification accuracy of 90.1% on one healthy and 12 delaminated cases. The study of the confusion matrix of CNN provided further insights into the physics of the problem. The predictive performance of a pre-trained CNN classifier was also evaluated on unseen cases of delamination, and physically consistent results were obtained.