Flutter Control via Data-Driven Models

Active flutter suppression is a promising technology for advanced flight vehicles, but greatly relies on accurate and simple models of fluid-structure interaction. For example, it is a great challenge to establish such a model for the flutter controller of an aircraft subject to a transonic flow, which is nonlinear by nature. The lecture addresses how to construct a surrogate model from CFD data to predict the transonic flutter of an aircraft wing and to actively suppress it in a wide range of flow regime. The lecture presents how to use simple physical knowledge to improve both interpretability and generalizability of the surrogate model, which are two essential issues in data-driven modeling. The lecture also shows the design of a flutter controller via machine learning.