Coherent structure identification in turbulent channel flow using Latent Dirichlet Allocation

TL;DR

This paper introduces a Latent Dirichlet Allocation-based clustering method to identify and analyze coherent structures in turbulent channel flow, revealing hierarchical structures and enabling data reconstruction and synthetic field generation.

Contribution

The paper presents a novel application of LDA for turbulence structure identification, providing a hierarchical description and capabilities for data compression and synthetic field generation.

Findings

LDA reveals hierarchical, wall-attached eddy structures consistent with Townsend's hypothesis.

The method achieves field reconstruction complexity comparable to POD.

Synthetic fields generated by LDA replicate original flow statistics.

Abstract

Identification of coherent structures is an essential step to describe and model turbulence generation mechanisms in wall-bounded flows. To this end we present a clustering method based on Latent Dirichlet Allocation (LDA), a generative probabilistic model for collections of discrete data. The method is applied for structure identification to the instantaneous Reynolds stress in turbulent channel flow at moderate Reynolds number R {\tau} = 590. LDA computes a robust flow description in terms of a hierarchy of vertically connected structure fragments, the characteristics of which scale with the wall distance, in agreement with the wall-attached eddy hypothesis of Townsend (1961). We show that the model can be used for field reconstruction with a complexity that can be compared with that of Proper Orthogonal Decomposition (POD). It can also be used to generate synthetic fields, the statistics of which mimic those of the original database. These findings highlight the potential of LDA for data analysis, compression and generation.