Thema der Dissertation:
Data-driven modeling of intermittent turbulence in the stably stratified atmospheric boundary layer Thema der Disputation:
A method for circulation-based topological decomposition of fluid-dynamical data
Data-driven modeling of intermittent turbulence in the stably stratified atmospheric boundary layer Thema der Disputation:
A method for circulation-based topological decomposition of fluid-dynamical data
Abstract: The dynamic motion of fluid flows often exhibits characteristic flow patterns which are commonly referred to as vortices. Over the years, researchers have proposed numerous definitions for a vortex in three dimensions although none reached a consensus. In the field of flow visualization, there is a wide range of methods to study and classify flow structures. In this talk, I will present a methodology that addresses flow analysis by formulating and solving an inverse problem. The idea is to classify the two-dimensional vortex vector field in terms of closed and smooth circulation paths. We will specifically look for a classification function that maximizes the total cluster-averaged circulation over all subsets. The data is constrained to a discrete 2D plane containing a snapshot of a 3D velocity vector field. I will show how to construct an inverse problem by applying Stokes' theorem in combination with 2D variational regularization. The solution to the inverse problem provides a threshold-free segmentation of vector-valued structures that can be interpreted as vortex structures through the definition of the Stokes theorem. The method is able to identify pairwise related structures whose relationships are given by their direction of rotation. Adding the circulation values of pairwise opposite structures gives a zero. This result stimulates considerations about the topological properties of the structures found. The performance of the method is demonstrated on numerous flows with steadily increasing complexity: time-independent double vortex, Bickley jet, and a velocity vector field of a direct numerical simulation of a turbulent channel flow.
Zeit & Ort
16.06.2022 | 14:00
Seminarraum 126
(Fachbereich Mathematik und Informatik, Arnimallee 6, 14195 Berlin)