抄録
In the analysis of unsteady flow field, low-dimensionalization methods have been often used. Those methods are based on a linearization method and have several issues in application to a flow field with a strong nonlinearity. In this paper, we focus on deep learning, which is a state-of-the-art nonlinear mapping method, and apply it to a low-dimensionalization method for unsteady flow field. In addition, we extend the dynamic network visualization method proposed by Elzen et al to a visualization of a time series data of unsteady flow field. Our proposed method is applied to the analysis of 3D unsteady flow around a sphere, and we shows the effectiveness in analyzing the unsteady flow.
