Hostname: page-component-7b7567c8c-z4zkx Total loading time: 0 Render date: 2026-11-16T03:43:53.157Z Has data issue: false hasContentIssue false

Direct numerical simulations of Rayleigh–Bénard convection in water with non-Oberbeck–Boussinesq effects

Published online by Cambridge University Press:  28 October 2019

Andreas D. Demou
Affiliation:
UCY-CompSci, Department of Mechanical and Manufacturing Engineering, University of Cyprus, 1 Panepistimiou Avenue, 2109 Aglantzia, Nicosia, Cyprus
Dimokratis G. E. Grigoriadis*
Affiliation:
UCY-CompSci, Department of Mechanical and Manufacturing Engineering, University of Cyprus, 1 Panepistimiou Avenue, 2109 Aglantzia, Nicosia, Cyprus
*
†Email address for correspondence: grigoria@ucy.ac.cy

Abstract

Rayleigh–Bénard convection in water is studied by means of direct numerical simulations, taking into account the variation of properties. The simulations considered a three-dimensional (3-D) cavity with a square cross-section and its two-dimensional (2-D) equivalent, covering a Rayleigh number range of $10^{6}\leqslant Ra\leqslant 10^{9}$ and using temperature differences up to 60 K. The main objectives of this study are (i) to investigate and report differences obtained by 2-D and 3-D simulations and (ii) to provide a first appreciation of the non-Oberbeck–Boussinesq (NOB) effects on the near-wall time-averaged and root-mean-squared (r.m.s.) temperature fields. The Nusselt number and the thermal boundary layer thickness exhibit the most pronounced differences when calculated in two dimensions and three dimensions, even though the $Ra$ scaling exponents are similar. These differences are closely related to the modification of the large-scale circulation pattern and become less pronounced when the NOB values are normalised with the respective Oberbeck–Boussinesq (OB) values. It is also demonstrated that NOB effects modify the near-wall temperature statistics, promoting the breaking of the top–bottom symmetry which characterises the OB approximation. The most prominent NOB effect in the near-wall region is the modification of the maximum r.m.s. values of temperature, which are found to increase at the top and decrease at the bottom of the cavity.

Information

Type
JFM Papers
Copyright
© 2019 Cambridge University Press 
Figure 0

Table 1.Coefficients $a_{n}$ of the polynomials $(X-X_{0})/X_{0}=\sum _{n=1}^{3}a_{n}(T-T_{0})^{n}$ describing the temperature dependence of a property $X$ of water. The reference temperature is $T_{0}=313~\text{K}$ and the polynomials are accurate over the range $283 (Ahlers et al.2006).

Figure 1

Table 2.Physical and numerical parameters used for each simulation. Cases with $\unicode[STIX]{x0394}T\approx 0$ were computed using the OB approximation. $h$ denotes the local grid spacing and $h_{BL}$ the maximum grid spacing inside the thermal BLs. $h_{BL}^{th}$ denotes a theoretical estimate for the maximum grid spacing inside the thermal BLs as suggested by Shishkina et al. (2010). $\unicode[STIX]{x1D702}_{B}$ denotes the Batchelor length scale. $\unicode[STIX]{x0394}\unicode[STIX]{x1D70F}_{s}$ represents the time interval used for statistical sampling.

Figure 2

Figure 1.Cuboid cavity used for the 3-D simulations, with $L_{x}=L_{z}$.

Figure 3

Figure 2.Two-point correlations of the wall-normal velocity components $R_{uu}$ and $R_{ww}$ for case $A40$. The calculations were carried out at the locations of the maximum temperature r.m.s., in the vicinity of the top and bottom walls: dashed line, $(x,z)=(0.5,0.934)$; solid line, $(x,z)=(0.5,0.053)$.

Figure 4

Figure 3.Case D40-2D. (a) Temporal variation of the Nusselt number at the bottom wall and (b) ratios of temporally averaged Nusselt number, described by (2.13) and (2.14) (inset). For this case, the flow reaches a statistically steady state at $\unicode[STIX]{x1D70F}_{s}\approx 300$, and a sample period of $\unicode[STIX]{x0394}\unicode[STIX]{x1D70F}_{s}\approx 400$ is sufficient for statistical sampling.

Figure 5

Figure 4.Instantaneous temperature contours with velocity vectors at the $x$–$z$ plane for 2-D (a,c,e,g) and 3-D (b,d,f,h) simulations with $\unicode[STIX]{x0394}T=40~\text{K}$, at $Ra=10^{6}$ (a,b), $Ra=10^{7}$ (c,d), $Ra=10^{8}$ (e,f) and $Ra=10^{9}$ (g,h). These instantaneous fields are snapshots of the flow after statistical stationarity was reached. For the 3-D cases, the snapshots were taken at $y=0$. The range $[-0.3,0.3]$ is used in the colour bars instead of the full normalised temperature range $[-0.5,0.5]$ so that the structures of the flow are better illustrated. For clarity, the velocity vectors are only shown every second, sixth and twelfth grid node in each direction for (c,d), (e,f) and (g,h) respectively.

Figure 6

Figure 5.Time-averaged velocity vectors in the $y$–$z$ planes located at (a,b) $x=0.1$, (c,d) $x=0.5$ and (e,f) $x=0.9$, for the 3-D cases with $\unicode[STIX]{x0394}T=40~\text{K}$, at $Ra=10^{6}$ (a,c,e), $Ra=10^{9}$ (b,d,f). For clarity, the presented results were interpolated on a coarser grid of $N_{y}\times N_{z}=60\times 20$.

Figure 7

Figure 6.Time- and space-averaged velocity distributions for $\unicode[STIX]{x0394}T=40~\text{K}$ and (a) $Ra=10^{6}$, (b) $Ra=10^{7}$, (c) $Ra=10^{8}$ and (d) $Ra=10^{9}$. Solid lines: present 3-D results of $\langle v\rangle _{t,x,y}$ within one LSC structure. Since there are two counter-rotating LSC structures along the spanwise direction, one of these structures was mirrored across the $z$-axis and a mean distribution for both LSCs is presented. Dashed lines: 2-D results of $\langle u\rangle _{t,x,y}$ presented by Sugiyama et al. (2009) who followed a conditional time-averaging procedure to address the random reversals of the LSC structure in two dimensions.

Figure 8

Figure 7.(a) Reduced Nusselt number values, scaled with $Ra^{1/3}$ and (b) Nusselt number values normalised with the respective OB values, as a function of $Ra$ for $\unicode[STIX]{x0394}T=40~\text{K}$: open squares, present 2-D results; solid squares, present 3-D results; open circles, 2-D numerical results reported by Sugiyama et al. (2009); diamonds, numerical results in a cylindrical domain reported by Horn & Shishkina (2014).

Figure 9

Table 3.Time- and space-averaged temperature gradients along the top ($\unicode[STIX]{x2202}\langle \unicode[STIX]{x1D6E9}\rangle _{x,y}/\unicode[STIX]{x2202}z|_{z=1}$) and bottom ($(\unicode[STIX]{x2202}\langle \unicode[STIX]{x1D6E9}\rangle _{x,y}/\unicode[STIX]{x2202}z)|_{z=0}$) walls, for $\unicode[STIX]{x0394}T=40~\text{K}$.

Figure 10

Figure 8.(a) Nusselt number values and (b) Nusselt number values normalised with the respective OB values as a function of $\unicode[STIX]{x0394}T$ for $Ra=10^{9}$: open squares, present 2-D results; solid squares, present 3-D results; open circles, 2-D numerical results reported by Sugiyama et al. (2009) at $Ra=10^{8}$; diamonds, 3-D numerical results in a cylindrical domain reported by Horn & Shishkina (2014).

Figure 11

Figure 9.Variation of centre temperature $\unicode[STIX]{x1D6E9}_{c}$ with (a) $Ra$ for $\unicode[STIX]{x0394}T=40~\text{K}$ and (b) $\unicode[STIX]{x0394}T$ for $Ra=10^{9}$: open squares, present 2-D results; solid squares, present 3-D results; open circles, 2-D numerical results reported by Sugiyama et al. (2009); diamonds, 3-D numerical results in a cylindrical domain reported by Horn & Shishkina (2014); dashed line, extended BL theory for variable property flows, developed by Ahlers et al. (2006). In panel (b) the 2-D numerical results of Sugiyama et al. (2009) correspond to $Ra=10^{8}$.

Figure 12

Figure 10.Temperature drop $\unicode[STIX]{x1D6E5}$ across the thermal BLs at $Ra=10^{9}$: open upward pointing triangles, $\unicode[STIX]{x1D6E5}_{t}$ – two dimensions; open downward pointing triangles, $\unicode[STIX]{x1D6E5}_{b}$ – two dimensions; solid upward pointing triangles, $\unicode[STIX]{x1D6E5}_{t}$ – three dimensions; solid downward pointing triangles, $\unicode[STIX]{x1D6E5}_{b}$ – three dimensions.

Figure 13

Figure 11.(a) Variation of thermal BL thickness $\unicode[STIX]{x1D706}^{\unicode[STIX]{x1D703}}$ with $Ra$ for $\unicode[STIX]{x0394}T=40~\text{K}$: open upward pointing triangles, $\unicode[STIX]{x1D706}_{t}^{\unicode[STIX]{x1D703}}$ – two dimensions; open downward pointing triangles, $\unicode[STIX]{x1D706}_{b}^{\unicode[STIX]{x1D703}}$ – two dimensions; solid upward pointing triangles, $\unicode[STIX]{x1D706}_{t}^{\unicode[STIX]{x1D703}}$ – three dimensions; solid downward pointing triangles, $\unicode[STIX]{x1D706}_{b}^{\unicode[STIX]{x1D703}}$ – three dimensions. (b) Variation of $\unicode[STIX]{x1D706}_{2D}^{\unicode[STIX]{x1D703}}/\unicode[STIX]{x1D706}_{3D}^{\unicode[STIX]{x1D703}}$ with $Ra$ for $\unicode[STIX]{x0394}T=40~\text{K}$: downward pointing triangles, bottom BL; upward pointing triangles, top BL.

Figure 14

Table 4.Scaling of thermal BL thickness $\unicode[STIX]{x1D706}^{\unicode[STIX]{x1D703}}$ at the top and the bottom of the cavity for both the 2-D and the 3-D cases, at $\unicode[STIX]{x0394}T=40~\text{K}$. The power laws were calculated using the least-squares method.

Figure 15

Figure 12.Variation of (a) thermal BL thickness $\unicode[STIX]{x1D706}^{\unicode[STIX]{x1D703}}$ and (b) $\unicode[STIX]{x1D706}_{2D}^{\unicode[STIX]{x1D703}}/\unicode[STIX]{x1D706}_{3D}^{\unicode[STIX]{x1D703}}$ with $\unicode[STIX]{x0394}T$ for $Ra=10^{9}$. The notation is the same as in figure 11.

Figure 16

Figure 13.Variation of $\unicode[STIX]{x1D706}_{NOB}^{\unicode[STIX]{x1D703}}/\unicode[STIX]{x1D706}_{OB}^{\unicode[STIX]{x1D703}}$ as a function of (a) $Ra$ for $\unicode[STIX]{x0394}T=40~\text{K}$ and (b) $\unicode[STIX]{x0394}T$ for $Ra=10^{9}$: open upward pointing triangles, top BL – two dimensions; open downward pointing triangles, bottom BL – two dimensions; solid upward pointing triangles, top BL – three dimensions; solid downward pointing triangles, bottom BL – three dimensions; diamonds, 3-D numerical results in a cylindrical domain at $Ra=10^{9}$ reported by Horn & Shishkina (2014).

Figure 17

Figure 14.Effect of $Ra$ on the vertical distribution of the plane-averaged mean temperature $\langle \unicode[STIX]{x1D6E9}\rangle _{t,x,y}$ (a,b) and r.m.s. values of temperature $\langle \unicode[STIX]{x1D6E9}_{rms}\rangle _{x,y}$ (c,d) in the vicinity of the bottom (a,c) and the top (b,d) walls: solid line, $Ra=10^{6}$; dashed line, $Ra=10^{7}$; dashed-dotted line, $Ra=10^{8}$; dashed-dotted-dotted line, $Ra=10^{9}$. In all cases $\unicode[STIX]{x0394}T=40~\text{K}$.

Figure 18

Figure 15.Effect of $\unicode[STIX]{x0394}T$ on the vertical distribution of the plane-averaged mean temperature $\langle \unicode[STIX]{x1D6E9}\rangle _{t,x,y}$ (a,b) and r.m.s. values of temperature $\langle \unicode[STIX]{x1D6E9}_{rms}\rangle _{x,y}$ (c,d) in the vicinity of the bottom (a,c) and the top (b,d) walls: solid line, OB; dashed line, $\unicode[STIX]{x0394}T=20~\text{K}$; dashed-dotted line, $\unicode[STIX]{x0394}T=40~\text{K}$; dashed-dotted-dotted line, $\unicode[STIX]{x0394}T=60~\text{K}$. In all cases$Ra=10^{9}$.

Figure 19

Table 5.Ratio $\unicode[STIX]{x1D706}^{rms}/\unicode[STIX]{x1D706}^{\unicode[STIX]{x1D6E9}}$ at $Ra=10^{9}$, for different values of $\unicode[STIX]{x0394}T$.