Reynolds stress decay modeling informed by anisotropically forced homogeneous turbulence

TL;DR

This paper develops an improved Reynolds stress decay model using anisotropic forcing in homogeneous turbulence, enabling better model selection and tuning with comprehensive data and demonstrating superior accuracy over existing models.

Contribution

It introduces a novel framework for decay modeling in turbulence using anisotropic forcing, enhancing model selection and accuracy in Reynolds stress predictions.

Findings

The new model outperforms popular existing models in accuracy.

Anisotropic forcing allows exploration of previously unprobed turbulence states.

The framework facilitates robust model coefficient selection and tuning.

Abstract

Models for solving the Reynolds-averaged Navier-Stokes equations are popular tools for predicting complex turbulent flows due to their computational affordability and ability to provide or estimate quantities of engineering interest. However, results depend on a proper treatment of unclosed terms, which require progress in the development and assessment of model forms. In this study, we consider the Reynolds stress transport equations as a framework for second-moment turbulence closure modeling. We specifically focus on the terms responsible for decay of the Reynolds stresses, which can be isolated and evaluated separately from other terms in a canonical setup of homogeneous turbulence. We show that by using anisotropic forcing of the momentum equation, we can access states of turbulence traditionally not probed in a triply-periodic domain. The resulting data span a wide range of anisotropic turbulent behavior in a more comprehensive manner than extant literature. We then consider a variety of model forms for which these data allow us to perform a robust selection of model coefficients and select an optimal model that extends to cubic terms when expressed in terms of the principal coordinate Reynolds stresses. Performance of the selected decay model is then examined relative to the simulation data and popular models from the literature, demonstrating the superior accuracy of the developed model and, in turn, the efficacy of this framework for model selection and tuning.