Group talk about the SINDy paper in PNAS

On Apr. 19, 2017, my group (modeling and optimization branch in Weiwei's Group, led by me) started the second talk. The topic of this talk is on discussing a paper from Dr. Steven L. Brunton and his colleagues from University of Washington. In this paper, the authors developed a general framework for discovering any nonlinear system's governing equations. This system identification method is called SINDy (Sparse Identification of Nonlinear Dynamics). I gave a talk anout this paper.

Firstly, the paper's content was reviewed. Secondly, some limitations disscussed by the authors were introduced and the succeeding works aiming to overcome these limitations were also noted. Finally, I gave some comments about our future work combining with the SINDy algorithm. It seems that there are also lots of work we can do to introduce and improve SINDy to aerodynamic system identification problems.

Fig. 1 PPT cover.

Figs. 2-3 Use SINDy to identify optimal basis functions of our multi-kernel neural networks.

Figs. 4 Use improved SINDy (adding exogenous inputs and varying parameters) to construct parameter-varying ARX models for aeroelastic problems across multiple Mach numbers.

Figs. 5 Use improved SINDy (adding exogenous inputs and other variations) to identify Hammerstein-type systems.