Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T13:32:47.397Z Has data issue: false hasContentIssue false

Wavelet-based adaptive large-eddy simulation of supersonic channel flow

Published online by Cambridge University Press:  25 August 2020

Giuliano De Stefano*
Affiliation:
Department of Engineering, University of Campania, Aversa, I-81031, Italy
Eric Brown-Dymkoski
Affiliation:
Space Exploration Technologies Corp. (SpaceX), Hawthorne, CA90250, USA
Oleg V. Vasilyev*
Affiliation:
Adaptive Wavelet Technologies, LLC, Superior, CO80027, USAKeldysh Institute of Applied Mathematics of Russian Academy of Sciences, Moscow, 125047, Russia

Abstract

The wavelet-based adaptive large-eddy simulation method is extended for computational modelling of compressible wall-bounded attached turbulent flows. The wavelet-threshold filtered compressible Navier–Stokes equations are derived. The unclosed terms in the governing equations are approximated by using eddy-viscosity and eddy-conductivity modelling procedures based on the anisotropic minimum-dissipation approach. The proposed filtering procedure is integrated with the adaptive anisotropic wavelet collocation method, which allows for the appropriate mesh stretching in the wall-normal direction. The performance of the method is assessed by conducting adaptive numerical simulations of fully developed supersonic flow in a plane channel with isothermal walls, which represents a well-established benchmark for wall-bounded turbulent compressible flows. The present results demonstrate both the feasibility and the effectiveness of the novel wavelet-based adaptive method in the high-speed compressible regime, showing good agreement with reference numerical solutions.

Information

Type
JFM Papers
Copyright
© The Author(s), 2020. Published by Cambridge University Press
Figure 0

Figure 1.Adaptive spatial grid and temperature contours, at a given time instant, for the simulation with ${Ma} = 1.5$ (case I), for four different equispaced cross-sections of the channel ($X,Y,Z$ represent the streamwise, wall-normal and spanwise directions).

Figure 1

Figure 2.Streamwise velocity contours, at a given time instant, for the simulation with ${Ma} = 1.5$ (case I), for the channel sections at $x_1/H=0$, $x_2/H=-0.96$ and $x_3/H=4{\rm \pi} /3$ ($X,Y, Z$ represent the streamwise, wall-normal and spanwise directions).

Figure 2

Figure 3.Adaptive spatial grid and temperature contours, at a given time instant, for the simulation with ${Ma} = 3$ (case II), for four different equispaced cross-sections of the channel ($X,Y,Z$ represent the streamwise, wall-normal and spanwise directions).

Figure 3

Table 1.Time-averaged grid compression and mesh spacings for different levels of resolution, for the simulation with ${Ma} = 1.5$ (case I).

Figure 4

Table 2.Time-averaged grid compression and mesh spacings for different levels of resolution, at a given time instant, for the simulation with ${Ma} = 3$ (case II).

Figure 5

Figure 4.Mean temperature profiles: present WA-LES compared to reference DNS (Coleman et al.1995), for the simulations with (a) ${Ma}=1.5$ (case I) and (b) ${Ma}=3$ (case II).

Figure 6

Figure 5.Mean density profiles: present WA-LES compared to reference DNS (Coleman etal.1995), for the simulations with (a) ${Ma}=1.5$ (case I) and (b) ${Ma}=3$ (case II).

Figure 7

Figure 6.Mean streamwise momentum profiles: present WA-LES compared to reference DNS (Coleman et al.1995), for the simulations with ${Ma}=1.5$ (case I) and ${Ma}=3$ (case II).

Figure 8

Table 3.Numerical resolution and mean flow results at ${Ma}=1.5$ and ${Re}=3000$ (case I): present WA-LES compared to reference DNS by Coleman et al. (1995) (1), DNS by Modesti & Pirozzoli (2016) (2), and LES by Brun et al. (2008) (3). For WA-LES, the resolution corresponding to an equivalent non-adaptive grid is shown.

Figure 9

Table 4.Numerical resolution and mean flow results at ${Ma}=3$ and ${Re}=4880$ (case II): present WA-LES compared to reference DNS by Coleman et al. (1995) (1), DNS by Modesti & Pirozzoli (2016) (2), and LES by Brun et al. (2008) (3). For WA-LES, the resolution corresponding to an equivalent non-adaptive grid is shown.

Figure 10

Figure 7.Turbulent shear stress (a) and streamwise normal stress (b) profiles: present WA-LES compared to reference DNS (Coleman et al.1995), for the simulations with ${Ma}=1.5$ (case I) and ${Ma}=3$ (case II).

Figure 11

Figure 8.Turbulent wall-normal (a) and spanwise (b) normal stress profiles: present WA-LES compared to reference DNS (Coleman et al.1995), for the simulations with ${Ma}=1.5$ (case I) and ${Ma}=3$ (case II).

Figure 12

Figure 9.Plane-averaged ratios of eddy viscosity to molecular viscosity and eddy conductivity to molecular conductivity, for the simulations with ${Ma}=1.5$ (case I) and ${Ma}=3$ (case II).

Figure 13

Figure 10.Mean-square temperature fluctuation profiles: present WA-LES compared to reference DNS (Coleman et al.1995), for the simulations with (a) ${Ma}=1.5$ (case I) and (b) ${Ma}=3$ (case II).

Figure 14

Figure 11.Time history of the grid compression for different thresholding levels.

Figure 15

Table 5.Time-averaged grid compression and mean flow results for different thresholding levels.

Figure 16

Figure 12.Mean temperature (a) and mean density (b) profiles for WA-LES with different thresholding levels, compared to reference DNS (Coleman et al.1995).

Figure 17

Figure 13.Turbulent shear stress (a) and streamwise normal stress (b) profiles for WA-LES with different thresholding levels, compared to reference DNS (Coleman et al.1995).

Figure 18

Figure 14.Turbulent wall-normal (a) and spanwise (b) normal stress profiles for WA-LES with different thresholding levels, compared to reference DNS (Coleman et al.1995).