Damage detection in smart composite structures using low frequency structural vibration

Abstract

Output-only based damage assessment of delaminated smart composite structures is increasingly appealing due to its easy availability in real engineering applications. In this work, structural vibration responses of the pristine and delaminated composite structures are processed via Fast Fourier Transform (FFT) and Convolutional Neural Network (CNN) for the classification of healthy and various damaged cases. The dynamic model for the healthy and delaminated smart composite laminates is developed by incorporating of improved layerwise theory, higher-order electric potential field, and finite element method. Structural vibration responses are obtained through a surface bonded piezoelectric sensor by solving the electromechanically coupled dynamic model in the time domain. FFT is used to construct vibration-based images from the transient responses of the sensor and CCN is used to classify those images into healthy and damaged classes. The confusion matrix of CNN showed physically consistent results and an overall classification accuracy of 90% was obtained. The pre-trained CNN was also tested to predict labels for new cases of delaminations in the smart composite laminates. The essence of the proposed method is that it requires only low-frequency structural vibration responses for the detection and localization of delamination in smart composite laminates.