Feature Extraction from Turbulent Channel Flow Databases via Composite DMD Analysis

Abstract

In this contribution we consider the Dynamic Mode Decomposition (DMD) framework as a purely data-driven tool to investigate a Reτ & 950 turbulent channel database. Specifically, composite-based DMD analyses are conducted, with hybrid snapshots composed by skin friction and Reynolds stress. A small number of dynamic modes (less than 1% of the number of snapshots) is found to be able to recover accurately the DNS Reynolds stresses near the wall, with a weighted factor as an indicator for the modes selections. As a possibility of analysis large turbulent database, we conclude that composite DMD is an attractive, purely data-driven, feature extraction tool to study turbulent flows.