Hostname: page-component-7b7567c8c-wxvwp Total loading time: 0 Render date: 2026-11-16T05:04:15.513Z Has data issue: false hasContentIssue false

Turbulent planar wakes of viscoelastic fluids analysed by direct numerical simulations

Published online by Cambridge University Press:  08 August 2022

Mateus C. Guimarães
Affiliation:
IDMEC/LAETA, Instituto Superior Técnico, Universidade de Lisboa, 1049-001 Lisboa, Portugal
Fernando T. Pinho
Affiliation:
CEFT, Faculdade de Engenharia, Universidade do Porto, 4200-465 Porto, PortugalALiCE, Faculdade de Engenharia, Universidade do Porto, 4200-465 Porto, Portugal
Carlos B. da Silva*
Affiliation:
IDMEC/LAETA, Instituto Superior Técnico, Universidade de Lisboa, 1049-001 Lisboa, Portugal
*
†Email address for correspondence: carlos.silva@tecnico.ulisboa.pt

Abstract

Direct numerical simulations employing the finitely extensible nonlinear elastic constitutive model closed with Peterlin's approximation (FENE-P) are used to investigate the far-field region of turbulent planar wakes of viscoelastic fluids and to develop the theory describing these flows. The theoretical results display excellent agreement with the simulations and provide new scaling laws for the evolution of the shear layer thickness $\delta (x) \sim x^{1/2}$, mean velocity deficit ${\rm \Delta} U(x) \sim x^{-1/2}$ and, for very high Deborah numbers, of the maximum polymer shear stresses $\sigma ^{[p]}_c(x) \sim x^{-2}$ and averaged polymer chain extension $\textrm {tr}(\bar {C}(x) - \boldsymbol{\mathsf{I}}) \sim x^{-2}$, where $x$ is the streamwise distance from the solid body generating the wake. The theory is able to show the existence of self-similarity for the profiles of mean velocity, mean polymer shear stress, averaged polymer chain extension and the conditions for similarity of the turbulent shear stress, and is very well supported by the numerical simulations. Similarly to the case of viscoelastic turbulent planar jets (Guimarães et al., J. Fluid Mech., vol. 899, 2020, p. A11), when the inlet Weissenberg and Deborah numbers are sufficiently large, turbulent viscoelastic wakes exhibit a considerable reduction of the spreading rate and of the normalised Reynolds stresses. However, for very large downstream locations the turbulent viscoelastic wake recovers the classical evolution laws observed for Newtonian turbulent planar wakes.

Information

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

Table 1.Physical and computational parameters of the DNS of viscoelastic turbulent planar wakes: global/inlet Weissenberg number ($Wi=\tau _p {\rm \Delta} U_0/d$); polymer relaxation time ($\tau _p$); Taylor-based Reynolds number at the far-field region ($Re_\lambda$); spreading rate constant $A_{\delta }$; centreline velocity deficit decay rate $A_{{\rm \Delta} U}$; polymer stresses decay constant $A_{\sigma _c}$; polymer extension decay constant $A_{c_{ii}}$; grid spacing normalised by the Kolmogorov microscale at the middle of the computational domain ($y/d=0$ and $x/d=L_x/2d=42$) (${\rm \Delta} x/\eta _{42d}$).

Figure 1

Figure 1.Contours of vorticity magnitude in the midplane ($z=0$) of the computational domain for turbulent planar wakes of viscoelastic fluids with different values of the Weissenberg number $Wi=1.1,2.1$ and $3.2$, together with the reference (Newtonian) case ($Wi=0$). Each simulation contains the entire domain (with $0 \le L_x/d \le 84$), however, different values of the vorticity magnitude threshold had to be used for each simulation in order to allow the visualisation of the flow in the entire computational domain (every subdomain using the same threshold is delimited by a vertical line). The figures do not show the total extent of the computational domain used in the lateral ($y$) dimension.

Figure 2

Figure 2.Streamwise evolution of the (a) squared shear layer thickness $\delta (x)^{2}$ normalised by the inlet momentum thickness $\theta$ and (b) centreline velocity defect ${\rm \Delta} U(x)$ normalised by the free stream velocity $U_\infty$, for the DNS of viscoelastic turbulent planar wakes with Weissenberg numbers equal to $Wi=1.1$, $2.1$, $3.2$ and $4.3$ ($Wi=0$ is the reference Newtonian simulation). The solid lines indicate straight line fits to the cases $Wi=0$ (Newtonian) and $Wi=4.3$.

Figure 3

Figure 3.Streamwise evolution of the (a) local Deborah number $De (x) = \tau _p {\rm \Delta} U/\delta$, (b) solvent dissipation reduction $SDR$ and (c) local Weissenberg number $Wi_{\eta } = \tau _p / \tau _{\eta }$ (computed at the centreline) for all the viscoelastic wake simulations carried in the present work.

Figure 4

Figure 4.Streamwise evolution of the r.m.s. velocity components $\sqrt {\overline {u'^{2}}}$, $\sqrt {\overline {v'^{2}}}$ and $\sqrt {\overline {w'^{2}}}$ normalised by the mean velocity deficit ${\rm \Delta} U(x)$ (at the centreline) for the viscoelastic turbulent wakes with different Weissenberg numbers. The reference Newtonian case ($Wi=0$) is also shown.

Figure 5

Figure 5.Transverse profiles of (a,b) mean (streamwise) velocity deficit ${\rm \Delta} \bar {u}$ and (c,d) streamwise Reynolds stresses $\sqrt {\overline {u'^{2}}}$ at (a,c) $x/\theta =200$ and (b,d) $x/\theta =315$ for the simulations with $Wi=0$ (Newtonian), $Wi=1.1,2.1,3.2$ and $4.3$, normalised with the velocity defect ${\rm \Delta} U(x)$ and the half-width $\delta (x)$.

Figure 6

Figure 6.Budgets of turbulent kinetic energy (3.3) for (a) the Newtonian wake, (b) a viscoelastic wake at low Deborah number and (c) a viscoelastic wake at high Deborah number. In these figures all the terms in (3.3) have been normalised by ${\rm \Delta} U(x)^{3}/\delta (x)$.

Figure 7

Figure 7.(a) Streamwise evolution of the normalised centreline viscous dissipation rate of the solvent and (b) 2-D kinetic spectrum of turbulent kinetic energy $E(k_{2d})$ at a fixed station $x/\theta =200$ for different Weissenberg numbers.

Figure 8

Figure 8.Streamwise evolution of the (mean) diagonal terms of the conformation tensor (computed in the centreline) for the present DNS of viscoelastic turbulent wakes.

Figure 9

Figure 9.Transverse profiles of the terms from the mean streamwise momentum equation, with the thin shear layer approximation (4.1), for the reference Newtonian simulation at $x/\theta =200$ (a), and for the viscoelastic simulation with $Wi=3.2$ at $x/\theta =200$ (b) and $x/\theta =315$ (c). The terms are normalised by the velocity deficit ${\rm \Delta} U(x)$ and by the thickness of the shear layer $\delta (x)$, and the dotted lines represent the sum of all the terms shown in the figures excluding the term $\rho ^{-1}\partial \overline {\sigma _{xx}}^{[p]} /\partial x$, which is neglected in the thin shear layer approximation.

Figure 10

Figure 10.Transverse profiles of normalised mean velocity deficit at several stations $x/\theta$ for the Newtonian ($Wi=0$) and viscoelastic simulations with Weissenberg numbers $Wi=1.1$, $2.1$ and $4.3$.

Figure 11

Figure 11.Transverse profiles of normalised normal mean velocity $\bar {v}(x,y)$ at several stations $x/\theta$ for the Newtonian ($Wi=0$) and viscoelastic simulations with Weissenberg numbers $Wi=1.1$, $2.1$ and $4.3$.

Figure 12

Figure 12.Streamwise evolution of (a) the maximum value of the normalised mean polymer shear stress and (b) the maximum value of the normalised averaged extension of the polymeric chains, for the simulations with $Wi=3.2$ and $4.3$. Both evolutions are compared with the scaling relations expressed in (4.18) and (4.21). The profiles for $Wi=4.3$ have been shifted downwards by (a) 800 units and (b) 600 units for clarity. The results from the additional simulations of Appendix B are also shown, vertically shifted by 1200 units, accessing the influence of the discretisation scheme of $\boldsymbol {u} \boldsymbol {\cdot} \boldsymbol {\nabla } \boldsymbol{\mathsf{C}}$ on these quantities.

Figure 13

Figure 13.Transverse profiles of mean polymer shear stress $\overline {\sigma _{xy}}^{[p]}(x,y)$ for the simulation with $Wi=4.3$ normalised by $\rho U_\infty ^{2}$ (a) and by $\sigma _c^{[p]}$, the characteristic scale of the polymer shear stress as derived in expression (4.16) (b). The inset shows the influence of the discretisation scheme of $\boldsymbol {u} \boldsymbol {\cdot}\boldsymbol {\nabla } \boldsymbol{\mathsf{C}}$ for the additional simulations discussed in Appendix B.

Figure 14

Figure 14.Transverse profiles of mean extension of the polymer chains $\text {tr} (\bar {C} - \boldsymbol{\mathsf{I}})$ for the simulation with $Wi=4.3$ with no normalisation (a) and normalised using the characteristic scale derived in expression (4.19) (b). The inset shows the influence of the discretisation scheme of $\boldsymbol {u} \boldsymbol {\cdot}\boldsymbol {\nabla } \boldsymbol{\mathsf{C}}$ for the additional simulations discussed in Appendix B.

Figure 15

Figure 15.Transverse profiles of Reynolds shear stresses $\overline {u'v'}$ from the DNS with $Wi=2.1$, $3.2$ and $4.3$: (a) at the same station $x/\theta =315$ (and with different values of local $De(x)$) normalised according to the classical theory for turbulent (Newtonian) wakes; (b) with the same value of the local $De(x) = 2$ (so at different stations $x/\theta$) and normalised as in (4.23).

Figure 16

Figure 16.Streamwise evolution of several statistics from the three Newtonian turbulent wakes with different computational domain heights $L_y/d$, and different Reynolds numbers (see inset in panel (d) for details): (a) wake half-widths; (b) wake velocity deficits; (c) viscous dissipation rate; (d–f) normal Reynolds stresses.

Figure 17

Figure 17.Profiles of the Reynolds stress and mean velocity deficit at the far-field region for the three Newtonian planar wakes discussed in this Appendix, using different inlet conditions, normalised with the classical theory. The results are compared also with the numerical and experimental data available from the literature (described in the main text), and with the reference Newtonian DNS used in the present work.

Figure 18

Figure 18.Streamwise evolution of several turbulence statistics from three DNS of turbulent viscoelastic wakes (for Weissenberg number $Wi=4.3$) carried out with different numerical schemes for the advection term of the conformation tensor transport equation and different domain widths $L_z$. The results from one Newtonian turbulent wake are also shown for comparison: (a) wake half-width; (b) wake velocity deficit; (c–e) square root of the normal Reynolds stresses; (f–i) mean conformation tensor components (see inset in panel (g) for details).

Figure 19

Figure 19.Viscoelastic characteristic length $X_{elastic}$, marking the start of the return to a Newtonian spreading rate, i.e. the end of the flow region where a strong spreading rate reduction caused by viscoelasticity can be observed in turbulent viscoelastic wakes.