Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-16T07:59:48.772Z Has data issue: false hasContentIssue false

Turbulent flow of non-Newtonian fluid in rough channels

Published online by Cambridge University Press:  27 November 2024

C. Narayanan
Affiliation:
AFRY AMS, Herostrasse 12, 8048 Zürich, Switzerland
J.S. Singh
Affiliation:
AFRY AMS, Herostrasse 12, 8048 Zürich, Switzerland
S. Nauer
Affiliation:
AFRY AMS, Herostrasse 12, 8048 Zürich, Switzerland
R.J. Belt
Affiliation:
TotalEnergies, Pôle d'Etudes et de Recherche de Lacq (PERL), Pôle Economique 2 – B.P. 47, 64170 Lacq, France
T. Palermo
Affiliation:
TotalEnergies, Pôle d'Etudes et de Recherche de Lacq (PERL), Pôle Economique 2 – B.P. 47, 64170 Lacq, France
D. Lakehal*
Affiliation:
AFRY AMS, Herostrasse 12, 8048 Zürich, Switzerland
*
†Email address for correspondence: djamel.lakehal@afry.com

Abstract

Direct numerical simulations of the turbulence of a Herschel–Bulkley (HB) fluid in a rough channel are performed at a shear Reynolds number $Re_{\tau } \approx 300$ and a Bingham number ${Bn} \approx 0.9$. For the type of rough surface used in this study, the results indicate that Townsend's wall similarity hypothesis also holds for HB fluids. However, there are notable differences compared with the effect of roughness on Newtonian fluids. More specifically, the effect of roughness appears to be slightly stronger for HB fluids, in the sense that the bulk Reynolds number, based on the viscosity at the wall, is reduced further due to the increase in viscosity in the troughs of the roughness surface induced by the low shear. At the same time, for the simulated rough surface, the contribution of form drag to the total pressure drop is reduced from 1/4 to about 1/5 due to the persistence of viscous shear in the boundary layer, reducing its shielding effect. As for the friction factor, due to the nonlinearity of the HB constitutive relation, its use with the wall shear rate from the mean wall shear stress underpredicts the minimum viscosity at the wall by up to 18 %. This inevitably leads to uncertainties in the prediction of the friction factor. Finally, it is observed that the rough surface is unable to break the peculiar near-wall flow structure of HB fluids, which consists of long persistent low-speed streaks occupying the entire domain. This means that the small-scale energy is significantly reduced for HB fluids, even in rough channels, with the energy more concentrated in the lower wavenumber range, implying an increase in the slope of the power spectrum to $-7/2$ in the inertial range, as shown by Mitishita et al. (J. Non-Newtonian Fluid Mech., vol. 293, 2021, 104570).

Information

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

Table 1.Reynolds numbers achieved in the simulations.

Figure 1

Figure 1.Surface used for rough channel simulations.

Figure 2

Figure 2.Comparison of the roughness distribution density function to Busse et al. (2017).

Figure 3

Table 2.Mesh used for NS and NNS simulations.

Figure 4

Table 3.Mesh used for NR and NNR simulations.

Figure 5

Figure 3.Vertical slice of instantaneous velocity with detail of the wall neighbourhood (zoomed in). (a) NR, (b) NNR.

Figure 6

Figure 4.Profiles for the NNS compared with pipe flow results of Rudman & Blackburn (2006). (a) Mean velocity profiles, (b) streamwise velocity fluctuations, (c) wall-normal velocity fluctuations, (d) spanwise velocity fluctuations, (e) Reynolds stress profiles.

Figure 7

Figure 5.Comparison of mean streamwise velocity profiles. (a) Newtonian versus non-Newtonian, (b) NNS versus NNR.

Figure 8

Figure 6.Variation of viscosity for non-Newtonian flow. (a) Outer scaling. (b) Inner scaling.

Figure 9

Figure 7.Indicators for the analytical form of the mean velocity profile. (a) Diagnostic for log law $\gamma$. (b) Diagnostic for power law $\beta$.

Figure 10

Figure 8.Instantaneous contours of normalized streamwise velocity and wall friction. Note that the same colour scheme is used for different quantities with vastly different ranges. (a) NNS, (b) NR, (c) NNR.

Figure 11

Figure 9.Time variation of normalized viscous drag and form drag for the NNR case.

Figure 12

Figure 10.Comparison of the form-drag coefficient versus the effective slope with data from Napoli et al. (2008).

Figure 13

Table 4.Comparison between DNS and Colebrook–White predictions using $Re_w$.

Figure 14

Table 5.Comparison between DNS and Colebrook–White predictions using $Re_{{gen,HB}}$.

Figure 15

Figure 11.Comparison of turbulent stresses between Newtonian and non-Newtonian fluid flows, for the smooth-wall case. (a) Streamwise velocity fluctuations. (b) Wall-normal velocity fluctuations. (c) Spanwise velocity fluctuations. (d) Reynolds stress.

Figure 16

Figure 12.Comparison of turbulent stresses between NNS and NNR cases. (a) Streamwise velocity fluctuations. (b) Vertical velocity fluctuations. (c) Spanwise velocity fluctuations. (d) Reynolds stress.

Figure 17

Figure 13.Mean shear stress balances for non-Newtonian fluid. (a) NNS, (b) NNR.

Figure 18

Figure 14.Comparison of turbulence anisotropy. (a) NS and NNS, (b) NNS and NNR.

Figure 19

Figure 15.Non-Newtonian (variable viscosity) terms in the TKE balance. (a) Smooth wall. (b) Rough wall.

Figure 20

Figure 16.Turbulent kinetic energy balance. (a) NS, (b) NNS, (c) NNS – $\chi _{vv}$ as production, (d) NNR.

Figure 21

Figure 17.Comparison of the pressure diffusion term between Newtonian and non-Newtonian flow.

Figure 22

Figure 18.Ratio of turbulence production to mean viscous dissipation. (a) All cases, (b) NNS case comparison to $P/\varepsilon _T$ and $(P+\chi _{vv})/(\varepsilon + \varepsilon _{vv})$.

Figure 23

Figure 19.Instantaneous axial flow velocity in the smooth channel at different heights for Newtonian and non-Newtonian fluids. (a) Newtonian $y^+ = 5$, (b) non-Newtonian $y^+ = 5$, (c) Newtonian $y^+ = 10$, (d) non-Newtonian $y^+ = 10$, (e) Newtonian $y^+ = 20$, ( f) non-Newtonian $y^+ = 20$, (g) non-Newtonian $y^+ = 30$, (h) non-Newtonian $y^+ = 30$, (i) non-Newtonian $y^+ = 50$, ( j) non-Newtonian $y^+ = 50$.

Figure 24

Figure 20.Instantaneous axial flow velocity in the rough channel at different heights for Newtonian and non-Newtonian fluids. (a) Newtonian $y^+ = 5$, (b) non-Newtonian $y^+ = 5$, (c) Newtonian $y^+ = 10$, (d) non-Newtonian $y^+ = 10$, (e) Newtonian $y^+ = 20$, ( f) non-Newtonian $y^+ = 20$, (g) Newtonian $y^+ = 30$, (h) non-Newtonian $y^+ = 30$, (i) Newtonian $y^+ = 50$, ( j) non-Newtonian $y^+ = 50$.

Figure 25

Figure 21.Contour of the Q-criterion with $\textrm {p.d.f.} = 0.005$, with $y+_{max} = 150$. (a) NS, (b) NNS, (c) NR, (d) NNR.

Figure 26

Figure 22.Power spectra in the streamwise $x$ direction. (a) Streamwise velocity, NS vs NNS. (b) Spanwise velocity, NS vs NNS. (c) Streamwise velocity, NNS vs NNR. (d) Spanwise velocity, NNS vs NNR.

Figure 27

Figure 23.Power spectra in spanwise $z$ direction. (a) Streamwise velocity, NS vs NNS. (b) Spanwise velocity, NS vs NNS. (c) Streamwise velocity, NNS vs NNR. (d) Spanwise velocity, NNS vs NNR.