Laser structural training, artificial intelligence-based acoustic emission localization and structural/noise signal distinguishment in a thick FCEV fuel tank

Abstract

Acoustic emission (AE) is present when transient elastic waves from structures are generated by various causes, such as structural cracks, fiber breakage, debonding of fibers and matrix, temperature changes, and fatigue. In AE-based structural health monitoring, the simple event counting method is unable to determine where AE occurs, so it is possible to discard a structure even if it is safe or not discard it when it is not safe. Much research on AE localization has been conducted to solve these problems. However, most of the methods have limitations with respect to isotropic material or near field conditions and cannot be applied when there is a change in the boundary conditions of the structure or obstacles. Thus, to solve these problems, a Q-switched laser capable of generating elastic waves has been used to scan and train the structures. Although this method worked effectively on thin specimens, a more advanced method is required for thick and complex structures, such as a fuel tank of a fuel cell electric vehicle (FCEV). Therefore, we propose a novel method based on artificial intelligence (AI) that can be applied to a real FCEV fuel tank fabricated with a filament winding composite. More specifically, this technique modulates the difference in characteristics between AE and laser-induced elastic waves in the frequency domain with AI. AE is simulated by a pencil lead break of the Hsu-Nielsen source. Then, AE localization is performed through cross-correlation in the time-.frequency domains between a generated AE signal and modulated laser-induced signals obtained from AI. In addition, an experiment conducted to localize the AE that occurs at arbitrary points in real time confirms that AE localization can be performed within 2 s. Finally, an AI algorithm is proposed to distinguish between structural AE and unwanted noise to consider real-world applications and visualize the features of these two types of signals. (C) 2021 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.