Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-15T17:15:29.500Z Has data issue: false hasContentIssue false

Stable channel flow with spanwise heterogeneous surface temperature

Published online by Cambridge University Press:  06 January 2022

T. Bon*
Affiliation:
Department of Mechanical Engineering, KU Leuven, Celestijnenlaan 300, B3001 Leuven, Belgium
J. Meyers
Affiliation:
Department of Mechanical Engineering, KU Leuven, Celestijnenlaan 300, B3001 Leuven, Belgium
*
†Email address for correspondence: thijs.bon@kuleuven.be

Abstract

Recent studies have demonstrated that large secondary motions are excited by surface roughness with dominant spanwise length scales of the order of the flow's outer length scale. Inspired by this, we explore the effect of spanwise heterogeneous surface temperature in weakly to strongly stratified closed channel flow (at $Ri_\tau =120$, 960; $Re_\tau = 180$, 550) with direct numerical simulations. The configuration consists of equally sized strips of high and low temperature at the lower and upper boundaries, while an overall stable stratification is induced by imposing an average temperature difference between the top and bottom. We consider the influence of the width of the strips (${\rm \pi} /8 \leq \lambda /h \leq 4{\rm \pi} $), Reynolds number, stability and upper boundary condition on the mean flow structure, skin friction and heat transfer. Results indicate that secondary flows are excited, with alternating high- and low-momentum pathways and vortices, similar to the patterns induced by spanwise heterogeneous surface roughness. We find that the impact of the surface heterogeneity on the outer layer depends strongly on the spanwise heterogeneity length scale of the surface temperature. Comparison to stable channel flow with uniform temperature reveals that the heterogeneous surface temperature increases the global friction coefficient and reduces the global Nusselt number in most cases. However, for the high-Reynolds cases with $\lambda /h \geq {\rm \pi} /2$, we find a reduction of the friction coefficient. At stronger stability, the vertical extent of the vortices is reduced and the impact of the heterogeneous temperature on momentum and heat transfer is smaller.

Information

Type
JFM Papers
Copyright
© The Author(s), 2022. Published by Cambridge University Press
Figure 0

Table 1.Overview of simulations and some integral flow properties. Here $Re_b = u_b h/\nu$ is the bulk Reynolds number, while $Ri_b = \langle \Delta ^z\theta \rangle g h / (2\theta _0 u_b^2)$ is the bulk Richardson number. Further, $h/L_{MO}$ is the stability parameter (Nieuwstadt 2005), where $L_{MO} = u_\tau ^3/(\kappa ( g/\theta _0) \alpha (\partial \langle \bar {\theta }\rangle /\partial z)_w )$ is the Obukhov length. These bulk properties are reported for completeness and easy comparison to studies with a different set-up, all other quantities are described and discussed in the text.

Figure 1

Table 2.Domain size and resolution for the cases with $Re_\tau = 180$ and $Re_\tau =550$. Here $N_x$ and $N_y$ are the number of Fourier coefficients in the streamwise and spanwise directions, while $N_z$ is the number of grid points in the wall-normal direction. Grid resolutions are scaled with $\nu /u_\tau$ (e.g. $\Delta x^+ = \Delta x u_\tau /\nu$). The smaller and larger values of $\Delta z^+$ represent the vertical grid size at the walls and in the channel centre, respectively.

Figure 2

Figure 1.(a) General sketch of the (in-phase) configuration, with the high- and low-temperature patches indicated by the red and blue strips, respectively. The mean flow $u$ is in the $x$ direction. (b) Sketch of the surface temperature at the top (black line) and bottom (grey line) for the in-phase configuration. (c) Same as (b) but for the antiphase configuration. The subscripts $b$ and $t$ refer to bottom and top, $H$ and $L$ refer to the high and low value. Here $\Delta ^z\theta$ is the mean temperature difference between the bottom and top, while $\Delta ^y \theta$ is the difference between the high and low surface temperatures.

Figure 3

Figure 2.Contour plots of (a) non-dimensional streamwise- and time-averaged potential temperature$\langle \bar {\theta }\rangle _x /\Delta ^z\theta$ and (b) dispersive velocity deviation $\langle \bar {u}''\rangle _x$ normalized with the bulk velocity $u_b$, for case Het-Ri120-pi/2. The vectors in (b) represent the in-plane velocity vector ($\langle \bar {v}\rangle _x,\langle \bar {w}\rangle _x$). The thick lines on the lower and upper edge in (b) indicate the location of the high surface temperature patches. The spanwise variation of the friction velocity $u_\tau$ is plotted with a green line, where the dashed grey line indicates the mean value (1). Note that not the entire domain width is shown.

Figure 4

Figure 3.Friction coefficient $C_f$ (a) and Nusselt number Nu (b) vs surface temperature wavelength for all heterogeneous simulations, normalized by the corresponding homogeneous simulations. Errorbars are bootstrap 95 % confidence intervals as described in § 2.3.

Figure 5

Figure 4.Laminar, turbulent and dispersive contribution to the skin friction coefficient (a) and Nusselt number (b), calculated with (2.11) and (2.12), for the heterogeneous simulations with $Re_\tau =550$, $Ri_\tau =120$. The corresponding homogeneous simulation is plotted at $\lambda /h = 0$. The total values (red lines with 95 % bootstrap confidence intervals as errorbars) are also plotted for reference.

Figure 6

Figure 5.(a,b) Horizontally averaged streamwise velocity (a) and potential temperature (b) for the simulations with $Re_\tau = 550$ and $Ri_\tau = 120$. The insets are a zoom of the near-wall region, with the dashed line in (a) representing $u^+ = z^+$ and the markers in (b) representing the location of the first three grid points. The horizontal axis in the inset in (a) is scaled in wall units ($z^+ = u_\tau z /\nu$). (c,d) Horizontally averaged profiles of turbulent (-, full lines), dispersive (- -, dashed lines) and viscous (-.-, dash-dotted lines) momentum (a) and heat flux (b).

Figure 7

Figure 6.Contour plots of (a) streamwise- and time-averaged non-dimensional potential temperature $\langle \bar {\theta } \rangle _x/\Delta ^z\theta$ and (b) dispersive velocity deviation $\bar {u}''$ normalized with the bulk velocity $u_b$, for case Het-Ri960-pi/2. The vectors in (b) represent the in-plane velocity vector ($\langle \bar {v} \rangle _x, \langle \bar {w} \rangle _x$). The thick lines on the lower edge in (b) indicate the location of the high surface temperature patches. The spanwise variation of the friction velocity $u_\tau$ is plotted with a green line, where the dashed grey line indicates the mean value (1). Note that only the lower channel half and not the entire domain width is shown.

Figure 8

Figure 7.Time evolution of $C_f$ (a) and Nu (b) over the averaging period between $tu_\tau /h = 60$ and $tu_\tau /h = 90$.

Figure 9

Figure 8.Time-height plots of square instantaneous vertical velocity profiles over the averaging period between $t^* = 60$ and $t^* = 90$. (a) The homogeneous case with $Ri_\tau = 960$ and $Re_\tau = 180$, (b) the heterogeneous case with $Ri_\tau = 960$ and $Re_\tau = 180$ and $\lambda /h = {\rm \pi} /2$.

Figure 10

Figure 9.Instantaneous contours at $tu_\tau /h = 90$ of the vertical velocity deviation $w''$ at a horizontal $x$–$y$ plane located at $z^+ = 15$ ($z/h = 0.083$). Plots (a) and (c) are heterogeneous cases with $\lambda /h = {\rm \pi} /2$ and $Ri_\tau = 960$ and 120, respectively, where the thick lines on the left and right edges indicate the location of the high surface temperature patches. Plots (b) and (d) are homogeneous cases with $Ri_\tau = 960$ and 120, respectively.

Figure 11

Figure 10.Contour plots of (a) streamwise- and time-averaged non-dimensional potential temperature $\langle \bar {\theta } \rangle _x/\Delta ^z\theta$ and (b) dispersive velocity deviation $\langle \bar {u}''\rangle _x$ normalized with the bulk velocity $u_b$, for case Het-Ri120-pi/2-A. The vectors in (b) represent the in-plane velocity vector ($\langle \bar {v} \rangle _x,\langle \bar {w} \rangle _x$). The thick lines on the lower and upper edge in (b) indicate the location of the high surface temperature patches. The spanwise variation of the friction velocity $u_\tau$ is plotted with a green line, where the dashed grey line indicates the mean value (1). Note that not the entire domain width is shown.

Figure 12

Figure 11.Contour plots of streamwise- and time-averaged dispersive velocity deviation $\langle \bar {u}''\rangle _x$ normalized with the bulk velocity $u_b$, for cases with surface temperature in-phase (Het-Ri120-4pi, a) and in antiphase (Het-Ri120-4pi-A, b). The vectors represent the in-plane velocity vector ($\langle \bar {v}\rangle _x,\langle \bar {w}\rangle _x$). The thick lines on the lower and upper edge indicate the location of the high surface temperature patches. The spanwise variation of the friction velocity $u_\tau$ is plotted with a green line, where the dashed grey line indicates the mean value (1).

Figure 13

Figure 12.Laminar, turbulent and dispersive contribution to the skin friction coefficient (a) and Nusselt number (b), calculated with (2.11) and (2.12), for the heterogeneous simulations with $Re_\tau =180$, $Ri_\tau =120$. The full and dashed lines indicate the respective in-phase and antiphase configuration. The corresponding homogeneous simulation is plotted at $\lambda /h = 0$. The total values (red lines with 95 % bootstrap confidence intervals as errorbars) are also plotted for reference.

Figure 14

Figure 13.Wall-normal profiles (a) mean velocity, (b) mean temperature, (c) vertical momentum flux and (d) vertical heat flux, for averaged in time and over the horizontal direction. Results from homogeneous stable channel simulations with SP-wind (full coloured lines) are compared with results from García-Villalba & del Álamo (2011) (open symbols) for $Re_\tau = 180$ and $Ri_\tau = 0$, 120, 480. Normalization is in wall variables $u_\tau$ and $\theta _\tau$.

Figure 15

Figure 14.Same as figure 14, but for simulations with $Re_\tau =550$. Temperature is normalized by the vertical temperature difference $\Delta \theta$ instead of the friction temperature $\theta _\tau$.