Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T01:19:50.645Z Has data issue: false hasContentIssue false

Revisiting the modelling framework for the unresolved scalar variance

Published online by Cambridge University Press:  25 March 2024

Z. Nikolaou*
Affiliation:
CORIA-CNRS, Normandie Université, INSA de Rouen Normandie, 685 Av. de l'Université, 76800 Saint-Étienne-du-Rouvray, France
P. Domingo
Affiliation:
CORIA-CNRS, Normandie Université, INSA de Rouen Normandie, 685 Av. de l'Université, 76800 Saint-Étienne-du-Rouvray, France
L. Vervisch
Affiliation:
CORIA-CNRS, Normandie Université, INSA de Rouen Normandie, 685 Av. de l'Université, 76800 Saint-Étienne-du-Rouvray, France
*
†Email address for correspondence: zacharias.nikolaou@insa-rouen.fr

Abstract

The unresolved scalar variance in large-eddy simulations of turbulent flows is a fundamental physical and modelling parameter. Despite its importance, relatively few algebraic models have been developed for this important variable with the most prominent models to date being the classic scale-similarity and gradient models. In this work a new generalized modelling framework based on reconstruction has been developed, which in contrast to classic modelling approaches allows the construction of base static variance models of arbitrary accuracy. It is demonstrated that higher-order reconstructions naturally lead to base static variance models of increased accuracy, and that the classic scale-similarity and gradient models are subsets of more general and higher-order models. The classic scale-similarity assumption for developing dynamic models is also revisited, and it is demonstrated that this can essentially be reinterpreted as a two-level reconstruction approach. Based on this result, a new general methodology is proposed that allows the construction of dynamic models for any given base static model, and a corresponding general reconstruction operator, algebraic or iterative. Consequently, improved static and dynamic models for the scalar variance are developed. The newly developed models are then thoroughly tested a priori using two high-fidelity direct numerical simulation databases corresponding to two substantially different flame and flow configurations, and are shown to outperform classic algebraic models for the variance.

Information

Type
JFM Papers
Copyright
© The Author(s), 2024. Published by Cambridge University Press
Figure 0

Table 1.Common explicit positive filters with their transfer functions and even moments $M_{2r}$.

Figure 1

Table 2.Operations involving the normal filter $\varDelta$ and the test filter $\hat {\varDelta }$.

Figure 2

Table 3.Optimised discrete explicit forward and inverse filter coefficients and properties for $\gamma =4$. Note that the filters are symmetric, hence, $g_{-l}=g_l$ and $\beta _{-l}=\beta _l$.

Figure 3

Table 4.Optimised discrete explicit forward and inverse filter coefficients and properties for $\gamma =8$. Note that the filters are symmetric, hence, $g_{-l}=g_l$ and $\beta _{-l}=\beta _l$.

Figure 4

Figure 1.Actual transfer function of a Gaussian filter $\hat {G}$, actual reconstructed transfer function $\hat {Q}^N$ obtained using $N=5$ van Cittert iterations, discrete transfer function of optimised filter $\hat {G}_d$, discrete reconstructed transfer function $\hat{Q}_d^N$ obtained using van Cittert iterations and the product between the discrete forward filter transfer function and the optimised inverse discrete filter transfer function $\hat {G}_d\hat{V}_d^N$: (a) $\gamma =4$ and (b) $\gamma =8$.

Figure 5

Table 5.Summary of the various models tested and the relevant equations used. A Gaussian filter is used for all models for which $a_2=\varDelta ^2 /24$. Note that the parameter $C^N_d$ is different for each model and calculated using the procedure outlined in § 5.2.

Figure 6

Figure 2.Progress-variable isosurface of the leading edge of the flame ($c=0.1$) for (a) the planar flame and (b) the V-flame. Dimensions are in mm.

Figure 7

Table 6.Important flame parameters for the planar (PB) and V (V97) flames.

Figure 8

Table 7.The DNS and LES meshes for $\varDelta /h =4$.

Figure 9

Figure 3.Conditionally averaged variance predictions of the static models against the reference variance at $\varDelta ^+=3$ for (a) the planar flame and (b) the V-flame. Note that these are normalised using the maximum of the reference conditional variance. Results are shown for (a) $\textrm {PB} - \varDelta ^+=3$, (b) $\textrm {V}97 - \varDelta ^+=3$.

Figure 10

Figure 4.Instantaneous variance predictions of the static models against the reference variance for case V97 and $\varDelta ^+=3$. The predictions are normalised using the maximum reference variance. Results are shown for (a) SM2, (b) GR, (c) SM4, (d) AD4, (e) DEIF.

Figure 11

Table 8.Mean-squared errors obtained using (6.1) for each of the static models.

Figure 12

Figure 5.Probability density functions of the different dynamic parameters: (a) $C^M_d$, (5.8); (b) $C^B_d$, (5.13); (c) $C^N_d$ for the SM2 model, (5.21); (d) $C^N_d$ for the SM4 model, (5.25); (e) $C^N_d$ for the AD4 model, (5.30) and (f) $C^N_d$ for the DEIF model, (5.30). The p.d.f.s are for case V97.

Figure 13

Figure 6.Conditionally averaged variance predictions of the dynamic models against the reference variance at $\varDelta ^+=3$ for (a) the planar flame and (b) the V-flame. Note that these are normalised using the maximum of the reference conditional variance. Results are shown for (a) $\textrm {PB} - \varDelta ^+=3$ and (b) $\textrm {V}97 - \varDelta ^+=3$.

Figure 14

Figure 7.Instantaneous variance predictions of the dynamic models against the reference variance for case V97 and $\varDelta ^+=3$. The predictions are normalised using the maximum reference variance. Results are shown for (a) DSM2-N, (b) DGR-B, (c) SM4, (d) AD4, (e) DEIF.

Figure 15

Table 9.Mean-squared errors obtained using (6.1) for each of the dynamic models.