Hostname: page-component-7b7567c8c-tnkh4 Total loading time: 0 Render date: 2026-11-16T05:46:44.420Z Has data issue: false hasContentIssue false

The dynamics of settling particles in vertical channel flows: gravity, lift and particle clusters

Published online by Cambridge University Press:  14 May 2021

Amir Esteghamatian
Affiliation:
Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA
Tamer A. Zaki*
Affiliation:
Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA
*
†Email address for correspondence: t.zaki@jhu.edu

Abstract

The dynamics of settling finite-size particles in vertical channel flows of Newtonian and viscoelastic carrier fluids is examined using particle resolved simulations. Comparison with neutrally buoyant particles in the same configuration highlights the effect of settling. The particle volume fraction is $5\,\%$, and a gravity field acts counter to the flow direction. Despite a modest density ratio ($\rho _r = 1.15$), qualitative changes arise due to the relative velocity between the particle and fluid phases. While dense particles are homogeneously distributed in the core of the channel, the mean concentration profile peaks at approximately two particle diameters from the wall due to a competition between shear- and rotation-induced lift forces. These forces act in the cross-stream directions, and are analysed by evaluating conditional averages along individual particle trajectories. The correlation between the angular and translational velocities of the particles highlights the significance of the Magnus lift force in both the spanwise and wall-normal directions. The collective behaviour of the particles is also intriguing. Using a Voronoï analysis, strong clustering is identified in dense particles near the wall, which is shown to alter their streamwise velocities. This clustering is attributed to the preferential transport of aggregated particles towards the wall. The practical implication of the non-uniformity of particle distribution is a significant increase in drag. When the carrier fluid is viscoelastic, the particle migration is enhanced which leads to larger stresses, thus negating the capacity of viscoelasticity to reduce turbulent drag.

Information

Type
JFM Papers
Copyright
© The Author(s), 2021. Published by Cambridge University Press
Figure 0

Figure 1.Schematic of the particle-laden simulations. No-slip boundary conditions $\boldsymbol {u}_f = 0$ are imposed at $y = \{0, 2\}$, and periodicity is enforced in the $x$ and $z$ directions.

Figure 1

Table 1.Physical and computational parameters of the simulations. Reynolds number ${\textit {Re}} \equiv hU_b/\nu$ and Weissenberg number $Wi \equiv \lambda U_b/h$ are based on the channel half-height $h$, the bulk velocity $U_b$, total kinematic viscosity $\nu$ and viscoelastic fluid relaxation time $\lambda$. The domain sizes are $L_{\{x,y,z\}}$ in $\{x,y,z\}$ directions, and the numbers of grid cells are $N_{\{x,y,z\}}$.

Figure 2

Figure 2.(a) Instantaneous snapshot of particle positions coloured by their velocity and contours of $u_f$ for case W0P5D. For clarity only particles located at $y \le 1$ are displayed. (b) A subset of the domain is magnified and reproduced with isosurfaces of $u_f=0.7$. (c) Contours of $u_f/\langle u_f \rangle$ in the $(x,y)$ plane in the vicinity of one particle.

Figure 3

Figure 3.Profiles of average (a,d) particle velocity relative to the fluid, (b,e) particle concentration and (c,f) fluid velocity. Suspensions: neutrally buoyant (circle – line, black), dense particles (filled circle – line, black); single-phase condition (—–, blue). The fluid in panels (a–c) is Newtonian and in panels (d–f) is viscoelastic. Vertical dashed lines in panels (b,c) and panels (e,f) mark the locations of the peak particles concentration.

Figure 4

Figure 4.Profiles of average spanwise (a,c) fluid vorticity and (b,d) particle angular velocity. Suspensions: neutrally buoyant (circle – line, black), dense particles (filled circle – line, black); single-phase condition (—–, blue). The fluid in the panels (a,b) is Newtonian and in panels (c,d) is viscoelastic. The marked area is magnified in the inset, and shaded area in the inset of panels (b,d) indicates the $y$ locations where $\langle \varOmega _{f,z}\rangle >0$.

Figure 5

Figure 5.Profiles of conditionally averaged (a,c) wall-normal particle flux and (b,d) particle concentration. Here $\varOmega _{p,z} > 0$ (—–, red); $\varOmega _{p,z} < 0$ (- - - - -, blue); unconditional statistics (– – –, black). The fluid in panels (a,b) is Newtonian and in panels (c,d) is viscoelastic.

Figure 6

Figure 6.Sample particle trajectories projected onto the $xz$ plane for particles with a positive (—–, red) or negative (—–, blue) wall-normal angular velocity $\varOmega _{p,y}$; (a) W0P5 (b) W0P5D. Particles velocity relative to the fluid is in the negative $x$ direction and the positive $y$ axis points into the page. All particles are located at approximately the same $z$ location at the beginning of the observation time window, during which $\varOmega _{p,y}$ does not change sign.

Figure 7

Figure 7.Ensemble-averaged properties along the trajectories $\langle \cdot \rangle _{tr}$ for particles with a positive (—–, red) or negative (- - - - -, blue) wall-normal angular velocity $\varOmega _{p,y}$ at $t_0$, where $t_0$ denotes the time at which the condition is satisfied; (a) W0P5D (b) W15P5D. The shifted time is denoted $t'\equiv t-t_0$, and the particle properties from top to bottom are wall-normal angular velocity, spanwise translational velocity and spanwise hydrodynamic force exerted on the particle and normalized by the particle buoyancy force $F_b\equiv (\rho _p-\rho _f)V_p|\boldsymbol {g}|$. Particles are located within $0.4 \le y \le 1$ at $t'=0$.

Figure 8

Figure 8.Sample particle trajectories projected onto the $xy$ plane for particles with a positive (—–, red) or negative (—–, blue) spanwise angular velocity $\varOmega _{p,z}$ in W0P5D. Particles velocity relative to the fluid is in the negative $x$ direction and the positive $z$ axis points out of the page. All particles are located at approximately the same $y$ location at the beginning of the observation time window, during which $\varOmega _{p,z}$ does not change sign. The solid grey line marks the geometric limit for particles centre in $y$ direction.

Figure 9

Figure 9.Ensemble-averaged properties along the trajectories $\langle \cdot \rangle _{tr}$ for particles with a positive (—–, red) or negative (- - - - -, blue) spanwise angular velocity $\varOmega _{p,z}$ at $t_0$, where $t_0$ denotes the time at which the condition is satisfied; (a) W0P5D (b) W15P5D. The shifted time is denoted $t'\equiv t-t_0$, and the particle properties from top to bottom are spanwise angular velocity, wall-normal translational velocity and wall-normal hydrodynamic force exerted on the particle and normalized by the particle buoyancy force $F_b\equiv (\rho _p-\rho _f)V_p|\boldsymbol {g}|$. Particles are located within $y \le 0.4$ at $t'=0$.

Figure 10

Figure 10.Voronoï tessellation of (a) the $xz$ plane located at $y = 3d_p/4$ and (b) the volume $0.8 \le y \le 1.2$ corresponding to one snapshot of particle centres in W0P5D. In panel (b) only a subset of the domain is visualized for the sake of clarity.

Figure 11

Figure 11.The p.d.f. of normalized Voronoï cell areas for particles in the wall region, i.e. $d_p \le y \le 1.5d_p$ in W15P5D and $0.5d_p \le y \le d_p$ in all other cases; (a) W0P5, (b) W0P5D, (c) W15P5, (d) W15P5D. For each case, p.d.f. of the same quantity is also plotted for a random distribution of non-overlapping particles with a matching particle volume fraction (grey dashed lines (- - -)).

Figure 12

Figure 12.(a) Schematic of the rotation-induced lift forces acting on the aggregated and isolated particles within $d_p, and the preferential transport of aggregated particles towards the wall. (b) The p.d.f. of normalized Voronoï cell areas $\mathcal {A}/\langle \mathcal {A} \rangle$ for case W0P5D and particles located at $y \le d_p$. The data are conditioned based on ${\rm \Delta} t\equiv t-t_e$, where $t_e$ denotes the time at which a particle entered $y \le d_p$ and $t$ is the time of data collection. The red dashed line (- - -) is the unconditional p.d.f. and the grey dashed line (- - -) shows the p.d.f. for a random distribution of non-overlapping particles. Note that the average Voronoï cell area $\langle \mathcal {A}\rangle$ is not conditioned. The shaded area shows the range of $\mathcal {A}/\langle \mathcal {A} \rangle$ for which the particles are clustered.

Figure 13

Figure 13.Joint probability density functions (j.p.d.f.) of normalized Voronoï areas $\log _{10}(\mathcal {A}/\langle \mathcal {A}\rangle )$ and mean particle relative velocity $(u_p -\langle u_f\rangle ) / \langle u_p \rangle$ for particles at $y \le 0.2$; (a) W0P5, (b) W0P5D, (c) W15P5, (d) W15P5D.

Figure 14

Figure 14.Particle conditioned mean flow field normalized by the unconditional fluid velocity $\langle u_f \rangle _{pc}/\langle u_f \rangle$ and plotted in the particle coordinates ($\tilde {\boldsymbol {x}} = \boldsymbol {x} - \boldsymbol {P}_{ref}$); (a,c) W0P5D, (b,d) W15P5D. In panels (c,d) particles are further conditioned to have $\varOmega _{p,y}>0$. The conditional averages are performed for particles located at $y \le 0.4$.

Figure 15

Figure 15.Particle-pair distribution function $q(r,\psi ,\theta )$ defined in (3.1), averaged in the range $-{\rm \pi} /8 \le \psi - {\rm \pi}/2 \le {\rm \pi}/8$. Reference particle is located at $y \le 0.4$.

Figure 16

Figure 16.(a,c) Particle-pair distribution function $q({\rm \Delta} x,{\rm \Delta} z)$ defined in (3.5), and (d,e) two-point correlation of $u'_p$ defined in (3.6); (a,b) W0P5D, (c,d) W15P5D. Dashed lines mark equidistant locations from the reference particle which is located at $y \le 0.4$.

Figure 17

Figure 17.(a) Contribution of different stress components to the wall drag, integrated between $0\le y \le 1$. Wall-normal profiles of the stresses defined in (3.7) for cases (b) W0P5D and (c) W15P5D: viscous stress $\tau _\mu$ (—–, black); turbulent Reynolds stress $\tau _{Re}$ ($\cdot \cdot \cdot \cdot \cdot$, black); polymer stress $\tau _\beta$(–$\cdot$–$\cdot$, black); particle stress $\tau _\phi$ ($\square$); gravity stress $\tau _g$ ($\blacksquare$); right-hand side of (3.7) (—–, blue); case designations are listed in table 1.

Figure 18

Figure 18.Mean profiles of (a,d) turbulent kinetic energy of the unconditional velocity field. Suspensions: neutrally buoyant (circle–line, black), dense particles (filled circle–line, black); single-phase condition (—–, blue). (b,e) Mean squared streamwise velocity fluctuations and its constituents as defined in (3.8), $\langle u'u' \rangle$ (—–, black), $\mathcal {C}_p$ ($\cdot \cdot \cdot \cdot \cdot$, red), $\mathcal {C}_f$ (–$\cdot$–$\cdot$, blue), $\mathcal {C}_i$(- - -, green). (c,f) Fluid fluctuations (—–, black) further conditioned to near-particle (filled square–solid line, black) and far-field regions (square–solid line, black). The fluid in panels (a–c) is Newtonian and in panels (d–f) is viscoelastic.

Figure 19

Figure 19.(a) Isosurfaces of streamwise fluid velocity fluctuations $u'_f$ and particle positions located at $y=2d_p$. (b) Two-point correlation of $u'_f$ defined in (3.9) as a function of (top) streamwise ${\rm \Delta} x$ and (bottom) spanwise ${\rm \Delta} z$ spacing for cases W0P5D (—–, black) and W15P5D(– – –, black). Here $R_{u'_fu'_f}$ is evaluated at $y=2d_p$. Case designations are listed in table 1.

Figure 20

Figure 20.(a–d) The p.d.f.s of normalized Voronoï cell volumes $\mathcal {V}/\langle \mathcal {V}\rangle$ for particles in the bulk of the flow $0.8 \le y \le 1.2$; (a) W0P5, (b) W0P5D, (c) W15P5, (d) W15P5D. For each case, p.d.f. of the same quantity is also plotted for a random distribution of non-overlapping particles with a matching particle volume fraction profile (grey dashed lines (- - -)). (e–h) Particle-pair distribution function $q(r,\psi ,\theta )$ defined in (3.1), averaged in the range $-{\rm \pi} /8 \le \psi$–${\rm \pi} /2 \le {\rm \pi}/8$. Reference particle is located in $0.8 \le y \le 1.2$.