Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-16T06:12:26.122Z Has data issue: false hasContentIssue false

Near- and far-field structure of shallow mixing layers between parallel streams

Published online by Cambridge University Press:  07 October 2020

Zhengyang Cheng
Affiliation:
Department of Civil and Environmental Engineering, The University of Iowa, Iowa City, IA52242, USAHydrologic Research Center, San Diego, CA92127, USA
George Constantinescu*
Affiliation:
Department of Civil and Environmental Engineering, The University of Iowa, Iowa City, IA52242, USA
*
†Email address for correspondence: sconstan@engineering.uiowa.edu

Abstract

The dynamics of coherent structures forming in a turbulent shallow mixing layer (ML) between two parallel streams advancing in a constant-depth, open channel is investigated using three-dimensional, time-accurate simulations. The large channel length to flow depth ratio ($L_{x}/H = 400\text{--}800$) allows characterization of the spatial evolution of shallow MLs until the mean velocity difference between the two streams becomes less than 3% of the initial value at the end of the splitter plate. Away from the ML origin, the dynamics and coherence of the Kelvin–Helmholtz (KH) billows are affected by the destabilizing effect of the mean shear between the two streams and by the stabilizing effect of the bed friction. A linear decay of the entrainment coefficient α with the bed-friction factor, S, applies only over the region where merging of neighbouring KH billows is still observed (transition to quasi-equilibrium regime). At larger distances from the origin, where the billows are severely stretched in the streamwise direction before being destroyed, the rates of increase of the ML width, δ, and centreline shift, lcy, become very small and α is exponentially decaying with increasing S toward zero (quasi-equilibrium regime). During the initial stages of the quasi-equilibrium regime where the KH vortices are severely stretched, the ML assumes an undulatory shape in horizontal planes. New relationships are proposed to characterize the downstream variation of the non-dimensional ML width and centreline shift over the transition and quasi-equilibrium regimes. During the transition to equilibrium regime, the ML boundary on the fast-stream side remains close to vertical, while that on the slow-stream side becomes strongly tilted. The ML boundary on the slow-stream side becomes again close to vertical once the quasi-equilibrium regime is reached. During the transition to the equilibrium regime, the passage of the KH billows and the generation of streamwise cells of secondary flow generate regions of high instantaneous bed shear stress, such that the region where the erosive capacity of the flow peaks does not correspond to the fast stream. The paper also investigates the effects of flow shallowness and initial velocity ratio between the two streams on the turbulent kinetic energy inside the ML, the depth-averaged lateral momentum fluxes, the passage frequency and size of the KH billows and the wavelength and period of the undulatory motions of the ML during the early stages of the quasi-equilibrium regime.

Information

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

Figure 1.Sketch of the computational domain showing the presence of KH billows inside the shallow mixing layer forming between two parallel streams of unequal velocities.

Figure 1

Figure 2.Scalar concentration in the mean and instantaneous flow and non-dimensional turbulent kinetic energy, $k/U_0^2$, near the free surface. (a) H = D, VR = 2.2 case; (b) H = D/2, VR = 2.2 case; (c) H = D, VR = 3.0 case. The dotted line visualizes the ML centreline near the free surface. The non-dimensional concentration of the scalar introduced at the end of the splitter plate is C = 100 but because of strong dilution near the end of the splitter plate much lower maximum values of C are used in the legend to visualize the mixing layer and its coherent structures in the three cases.

Figure 2

Figure 3.ML structure near the free surface in the region where the ML assumes an undulatory shape. The ML is visualized using non-dimensional scalar concentration contours in an instantaneous flow field.

Figure 3

Figure 4.Non-dimensional scalar concentration in the mean flow in several cross-sections. (a) x/D = 164 and x/D = 373 for the H = D, VR = 2.2 case; (b) x/D = 30 and x/D = 164 for the H = D/2, VR = 2.2 case. The high-speed stream is situated on the left side of each panel. The non-dimensional concentration of the scalar introduced at the end of the splitter plate is C = 100.

Figure 4

Figure 5.Distribution of depth-averaged lateral momentum fluxes. (a) H = D, VR = 2.2 case; (b) H = D/2, VR = 2.2 case; (c) H = D, VR = 3.0 case. The frames show the lateral momentum flux by the mean flow, $\langle U\rangle \langle V\rangle /U_0^2$, the dispersive lateral momentum flux $[\langle UV\rangle - \langle U\rangle \langle V\rangle ]/U_0^2$ and the turbulent lateral momentum flux $\bigl\langle {\overline {u^{\prime}v^{\prime}} } \bigr\rangle /U_0^2$. The dotted line visualizes the ML centreline near the free surface.

Figure 5

Figure 6.Non-dimensional velocity difference between the two streams in a horizontal plane situated close to the free surface (z/H = 0.85). (a) ${(U_2^2 - U_1^2)^\ast }$ vs. x*; (b) $\mathrm{\Delta }{U^\ast }$ vs. ${x^\ast }$. The solid line shows the theoretical solution ${(U_2^2 - U_1^2)^\ast } = {\textrm{e}^{ - {x^\ast }}}$ and $\mathrm{\Delta }{U^\ast } = {\textrm{e}^{ - {x^\ast }}}$.

Figure 6

Figure 7.Comparison of non-dimensional mean streamwise velocity profiles, (u − U1)/(U2 − U1), predicted by DES (solid lines, 0.05 m < x < 20 m, z/H = 0.85) with the experimental data of Uijttewaal & Booij (2000) (solid circles, x < 11 m, z/H = 0.85). The velocity profiles are plotted versus the non-dimensional lateral distance, (y − lcy)/δ. The dashed red line represents the error-function curve corresponding to the self-similar solution for a free ML.

Figure 7

Figure 8.Shift of the ML centreline near the free surface (z/H = 0.85). (a) lcy/H vs x/D, (b) ${l^{\prime}_{cy}} = {l_{cy}}/D(\Delta U(0)/\bar{U}(0))$ vs x*; (c) $l_{cy}^\ast = {l_{cy}}/D(\Delta U(x)/\bar{U}(x))$ vs x* in log-log scale. Also shown in (a) are the experimental results of Uijttewaal & Booij (2000).

Figure 8

Figure 9.Mixing layer width near the free surface (z/H = 0.85). (a) δ/H vs. x/H; (b) ${\delta ^\ast } = {c_f}\delta /H(\Delta U(x)/\bar{U}(x))$ vs. x* in linear scale; (c) ${\delta ^\ast }$ vs. x* in log–log scale. Also shown in (a) are the experimental results of Uijttewaal & Booij (2000). The two lines in (c) show best fit, power-law curves for x* < 1.5 and x* > 2. The solid lines in (a) show ML width predictions using the equation proposed by Van Prooijen & Uijttewaal (2002), $\delta (x/H)/H = \alpha (\Delta U(0)/\bar{U}(0))(1/{c_f})(1 - {\textrm{e}^{ - {x^\ast }}}) + \delta (0)/H$. The horizontal dotted lines in (a) show the equilibrium values (x, x*→∞) predicted by the same equation.

Figure 9

Figure 10.Entrainment coefficient, α, as a function of x* (a) and the bed-friction number S (b,c) near the free surface (z/H = 0.85) for the H = D/2, VR = 2.2 case. Also shown in (d) are the estimations of α vs. S based on the experimental results of Uijttewaal & Booij (2000) for x < 11 m and based on simulation results for the H = D and $H = D/2$ cases. The dashed lines in (b) show best fit curves, α ~ e−γS, with γ = 14.7 for S < 0.15 and γ = 3.3 for S > 0.15. These curves are visualized as straight lines in log–linear scale in (c). The solid straight lines in (b,d) correspond to the initial quasi-linear decay of α with S.

Figure 10

Figure 11.Power spectra of the spanwise velocity fluctuations at points situated close to the centreline of the ML and the free surface (z/H = 0.85). (a) H = D, VR = 2.2 case; (b) H = D/2, VR = 2.2 case; (c) H = D, VR = 3.0 case.

Figure 11

Figure 12.Autocorrelation function of the spanwise velocity fluctuations, Rvv. (a) x/D = 36 and x/D = 164; (b) x/D = 299. The autocorrelation function is evaluated close to the ML centreline and the free surface (z/H = 0.85). Also plotted in (a) is the autocorrelation function for the base case from the experiment of Uijttewaal & Booij (2000). The vertical lines denote the location of the first maximum of Rvv in the H = D, VR = 2.2 case (solid lines) and H = D/2, VR = 2.2 case (dashed lines).

Figure 12

Figure 13.Distribution of the normalized bed shear stress. (a) $\bar{\tau }/{\tau _0}$, mean flow H = D, VR = 2.2 case; (b) root mean square of the bed shear stress fluctuations, ${\tau ^{rms}}/{\tau _0}$, H = D, VR = 2.2 case; (c) $\tau /{\tau _0}$, instantaneous flow, H = D, VR = 2.2 case; (d) $\tau /{\tau _0}$, instantaneous flow, H = D/2, VR = 2.2 case; (e) $\tau /{\tau _0}$, instantaneous flow, H = D, VR = 3.0 case. Also shown in (f) is the instantaneous streamwise velocity distribution, u/U0, and 2-D streamline patterns in the x/D = 122 cross-section for the H = D, VR = 3.0 case. The dotted line visualizes the ML centreline near the free surface. The black and blue arrows point toward patches of large bed shear stress induced by the KH billows and by streamwise cells of secondary flow, respectively. The x/D = 122 cross-section shown in (f) cuts through the elongated streak of high $\tau /{\tau _0}$ situated between x/D = −1 and x/D = 0.3 in (e). The streamwise cell of secondary flow pushes high velocity fluid toward the bed, which induces the formation of a region of high $\tau /{\tau _0}$.