Translator Disclaimer
Presentation + Paper
22 March 2021 Deep-learning assisted finite element model of a galloping piezoelectric energy harvester
Author Affiliations +
Abstract
One challenge in modelling a galloping piezoelectric energy harvester (GPEH) is the representation of the highly nonlinear aerodynamic force. The existing work in the literature employed various polynomial functions to fit the aerodynamic coefficient curve for simplicity, though their approximation capabilities are limited. In this paper, we propose to use the deep-learning technique to capture the aerodynamic force behaviour of a bluff body. Replacing the widely adopted third-order polynomial function by a welltrained artificial neural network (ANN) for aerodynamic force representation in modelling a GPEH, the feasibility of the proposed approach is preliminarily validated. To further improve the modelling accuracy, the electromechanical structure of the GPEH is then modelled using the finite element method. The trained ANN is integrated with the established finite element model to predict and update the aerodynamic fore applied on the bluff body in the real-time simulation. The aeroelastic motion and the electrical output of the galloping piezoelectric energy harvester are successfully predicted. Finally, based on a collection of experimental data, a welltrained artificial neural network (ANN) is proved to behave with a much better curve fitting performance than a third-order polynomial function. General procedures for using the deep learning technique to help model a general GPEH with complex geometric shapes are proposed.
Conference Presentation
© (2021) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Guobiao Hu, Junlei Wang, Chunbo Lan, Lihua Tang, and Junrui Liang "Deep-learning assisted finite element model of a galloping piezoelectric energy harvester", Proc. SPIE 11588, Active and Passive Smart Structures and Integrated Systems XV, 115880H (22 March 2021); https://doi.org/10.1117/12.2582939
PROCEEDINGS
9 PAGES + PRESENTATION

SHARE
Advertisement
Advertisement
Back to Top