Nonlinear Reduced Order Modeling for Aeroelastic Simulation with Neural Networks

摘要

In the last few years reduced order modeling (ROM) of aerodynamics became more and more popular in the aeroelasticity community. Different reduced order modeling approaches are developed, for instance harmonic balance, center manifold, normal form and numerical continuation methods. An introduction to these methods can be found in Henshaw et al. [2]. Other methods create a linear state-space formulation via the eigenvalue realization algorithm (ERA) used for example by Lucia et al. [11]. Also proper orthogonal decomposition (POD) is a widely used reduction technique in aeroelasticity and flow analysis, for instance demonstrated by Willcox et al. [18]. Chen et al. [1] use a flow solver based nonlinear POD approach for aeroelastic investigations.