Detecting Coherent Structures in a Turbulent Wake by Using Delay Based Networks

Abstract

[Abstract] A delay based artificial neural network is used to analyze turbulent flow signals obtained by a hot wire anemometer along time in different points within the wake of a circular cylinder. Values of the Reynolds number range from 2000 to 8000, corresponding to late transitional or fully turbulent wake flows. The implemented trainable delay based artificial neural network is able to autonomously obtain the embedding dimension as well as the normalized embedding delay and permits performing short and long term predictions. The short term predictions are extremely accurate while the long term ones result in a sort of nonlinear filter able to extract the signal features concerned with the larger eddies and coherent structures present in the turbulent flow-field.