[3N4-GS-10] AI application: model

[3N4-GS-10-03] Construction of Reservoir Computing Reduced-order Model based on Time Series Velocity Field around Airfoil

〇Yuto Iwasaki¹, Kumi Nakai¹, Takayuki Nagata¹, Taku Nonomura¹, Keisuke Asai¹, Masanobu Inubushi² (1. Tohoku University, 2. Tokyo University of Science)

Keywords:Reservoir Computing, Reduced-order Model, Proper Orthogonal Decomposition

This study proposes a reservoir computing reduced-order model (RCROM) based on time series velocity field of post-stall flow around a NACA0015 airfoil. The low-dimensional description of the velocity field is obtained by applying a proper orthogonal decomposition (POD) to the time series velocity field. The time development of the low-dimensional velocity field, which is the amplitude of the POD bases, is learned by a reservoir computing approach using an echo state network. The estimation accuracy of RCROM is evaluated by one-step ahead prediction compared to a linear reduced-order model (LROM) which estimates the amplitude of the POD bases by a linear equation. The results show that the estimation accuracy of RCROM is higher than that of LROM by 12.2%.