Routes to stratified turbulence revealed by unsupervised classification of experimental data

TL;DR

This paper introduces a data-driven approach combining dimensionality reduction and unsupervised clustering to identify and classify diverse turbulent states in stratified shear flows, revealing multiple transition pathways.

Contribution

It presents a novel method for analyzing turbulence using experimental shadowgraph data, uncovering multiple energetic and intermittent turbulent states at high Reynolds numbers.

Findings

Identification of multiple energetic turbulence types

Discovery of intermittent turbulence cycling between states

Revelation of distinct transition pathways

Abstract

Modeling fluid turbulence using a 'skeleton' of coherent structures has traditionally progressed by focusing on a few canonical experiments, such as pipe flow and Taylor-Couette flow. We here consider an alternative canonical experiment, the stratified inclined duct, a sustained shear flow whose density stratification allows for the exploration of a wealth of new coherent and intermittent states at significantly higher Reynolds numbers than in unstratified flows. We automatically identify the underlying turbulent skeleton of this experiment with a novel data-driven method combining dimensionality reduction and unsupervised clustering of shadowgraph visualizations. We demonstrate the existence of multiple types of energetic turbulence across parameter space, as well as intermittent turbulence that cycles between these types, revealing distinct transition pathways. Our method and results pave the way for new reduced-order models of multi-physics turbulence.