Towards Reconstruction of Complex Flow Fields Using Unit Flows

Paul J Kristo,Mark L Kimber,Sharath S Girimaji

doi:10.3390/fluids6070255

Abstract

Many complex turbulent flows in nature and engineering can be qualitatively regarded as being constituted of multiple simpler unit flows. The objective of this work is to characterize the coherent structures in such complex flows as a combination of constituent unitary flow structures for the purpose of reduced-order representation. While turbulence is clearly a non-linear phenomenon, we aim to establish the degree to which the optimally weighted superposition of unitary flow structures can represent the complex flow structures. The rationale for investigating such superposition stems from the fact that the large-scale coherent structures are generated by underlying flow instabilities that may be reasonably described using linear analysis. Clearly, the degree of validity of superposition will depend on the flow under consideration. In this work, we take the first step toward establishing a procedure for investigating superposition. Experimental data of single and triple tandem jets in crossflow are used to demonstrate the procedure. A composite triple tandem jet flow field is generated from optimal superposition of single jet data and compared against ‘true’ triple jet data. Direct comparisons between the true and composite fields are made for spatial, temporal, and kinetic energy content. The large-scale features (obtained from proper orthogonal decomposition or POD) of true and composite tandem jet wakes exhibit nearly 70% agreement in terms of modal eigenvector correlation. Corresponding eigenvalues reveal that the kinetic energy of the flow is also emulated with only a slight overprediction. Temporal frequency features are also examined in an effort to completely characterize POD modes. The proposed method serves as a foundation for more rigorous and robust dimensional reduction in complex flows based on unit flow modes.

Highlights

Introduction and MotivationFluid mechanics research has traditionally employed canonical flows as a means of determining the dominant mechanisms of naturally occurring flow phenomena, relying exclusively on scientific discovery from first principles
Regarding more complex flows found in nature and engineering, high-fidelity parameterization of geometric, kinematic, and dynamic quantities of interest have proven costly for iterative optimization, for both experiments and simulations alike [1]
The objective of the present work is to examine the use of spatial superposition of linearly weighted unit flows as a means of reconstructing a complex flow by optimizing certain objective functions

Summary

Introduction

Fluid mechanics research has traditionally employed canonical flows as a means of determining the dominant mechanisms of naturally occurring flow phenomena, relying exclusively on scientific discovery from first principles. Regarding more complex flows found in nature and engineering, high-fidelity parameterization of geometric, kinematic, and dynamic quantities of interest have proven costly for iterative optimization, for both experiments and simulations alike [1]. There exists a significant area of possible contributions for data-driven methods that seeks not to replace, but to compliment traditional means of understanding, modeling, optimization, and control of complex fluid flows

Objectives

Results

Conclusion