Damage localization for composite structure using guided wave signals with Gramian angular field image coding and convolutional neural networks

Abstract

Accurate detection of damage in composite structures is of great significance to ensure safe service and avoid catastrophic accidents. In this paper, a novel damage diagnosis method, integrating guided wave (GW)-based structural health monitoring (SHM) with Gramian angular field (GAF) image coding and convolutional neural networks (CNNs), is proposed to improve the accuracy of damage localization in composite structures. Firstly, an improved piecewise aggregate approximation (PAA) algorithm is proposed to achieve guided wave data compression and obtain a series of damage indexes (DIs) characterizing the damage for effectively improving the localization accuracy. Then, the multi-path fused DI sequences are converted into two-dimensional (2-D) images using GAF to take full advantage of the significant benefits of CNN in machine vision. Secondly, a CNN model is constructed to learn high-level feature representations and conduct damage location regression. Finally, the proposed method is tested and assessed on measured GW signal datasets with the addition of zero-mean Gaussian noise. The experimental results show that the mean relative error of the proposed method for damage localization on the invisible datasets is 3.59%. The proposed method demonstrates better generalization and localization performance than other selected state-of-the-art damage localization techniques.