Multi-scale turbulence modeling and maximum information principle. Part 4

TL;DR

This paper develops a maximum information principle-based turbulence model that addresses negative energy spectrum issues and examines the constraints of conventional closure schemes in isotropic turbulence.

Contribution

It introduces a novel fourth-order turbulence model formulated as a second-order cone programming problem, incorporating constraints to improve closure schemes.

Findings

The model can satisfy physical constraints like non-negativity of energy spectrum.

It addresses the negative energy spectrum problem in classical turbulence models.

The approach provides a new framework for turbulence closure using optimization techniques.

Abstract

We explore incompressible homogeneous isotropic turbulence within the (fourth-order model) formulation of optimal control and optimization, in contrast to the classical works of Proudman and Reid (1954) and Tatsumi (1957), with the intention to fix specially their defect of negative energy spectrum values being developed and to examine generally the conventional closure schemes. The isotropic forms for the general and spatially degenerated fourth order correlations of fluctuating velocity are obtained and the corresponding primary dynamical equations are derived. The degenerated fourth order correlation contains four scalar functions $D_i$, $i=1,2,3,4$, whose determination is the focus of closure. We discuss the constraints of equality for these functions as required by the self-consistency of the definition of the degenerated. Furthermore, we develop the constraints of inequality for the scalar functions based on the application of the Cauchy-Schwarz inequality, the non-negativity of the variance of products, and the non-negativity of the turbulent energy spectrum. We intend to indicate the difficulty for a conventional scheme to satisfy all these constraints. As an alternative, we employ the turbulent energy per unit volume as the objective function to be maximized, under the constraints and the dynamical equations, with the four scalar functions as the control variables, which is a second-order cone programming problem. We then treat the asymptotic state solutions at large time and focus especially on the sub-model where the third order correlation is taken as the control variable, considering the computing resources available.