Applications and Good Practice (Prof. Andrea Ianiro)

Описание: This lecture was given by Prof. Andrea Ianiro, Universidad Carlos III de Madrid, Spain, von Karman Institute for Fluid Dynamics, Belgium in the framework of the von Karman Lecture Series on Machine Learning for Fluid Mechanics organized by the von Karman Institute and the Université libre de Bruxelles in February 2020.

This lecture develops a theoretical background for the definition of good practice (number of samples and sampling time versus the effect of noise) for data-driven modal analysis and shows example applications to experimental and computational datasets. Snapshot POD is employed as a benchmark but the same concepts can be applied also to the other data-driven modal analysis tools, presented in previous lectures.
Beyond the estimation of the effect of noise on the estimated modes, this lecture also presents several application examples showing how hidden information can be extracted after a well-converged POD. In particular, it is shown how temporal modes of non-timeresolved data can provide detailed phase information and how extended POD modes can provide a linear stochastic estimation of correlated events. This is, in particular, found
to be useful for applications which involve flow sensing and for the study of convection problems.