You are on page 1of 8

Proceedings of the Twenty-second (2012) International Offshore and Polar Engineering Conference www.isope.

org
Rhodes, Greece, June 17–22, 2012
Copyright © 2012 by the International Society of Offshore and Polar Engineers (ISOPE)
ISBN 978-1-880653-94–4 (Set); ISSN 1098-6189 (Set)

Progress in the Application Of Lattice Boltzmann Method For Turbulent Flows

Gaurav Kumar and Sharath. S. Girimaji


Department of Aerospace Engineering, Texas A&M University
College Station, TX, USA

method (LBM). In this method, the nearly infinite degree of freedom


ABSTRACT
Boltzmann equation is approximated to a much smaller system with an
order of ten degrees of freedom. These restricted degrees of freedom
We present a review of applications of the Lattice Boltzmann Method
form a lattice in velocity space and hence the name lattice Boltzmann
in turbulent flows. Canonical flows of interest to off-shore (coastal)
method. The next major advancement of this method came with the
engineering are highlighted. Such flows include turbulence with
proof that the lattice Boltzmann equation is in fact equivalent to non-
thermal gradients, turbulence with frame rotation, turbulence subject to
traditional discretized Navier-Stokes equation. This discretization
periodic shear, two-fluid turbulent jets and flows past bluff bodies.
allows for many computational advantages not possible in conventional
LBM turbulence calculations at all levels-direct numerical simulations
discretized methods. Particularly, the advective operator is represented
(DNS), large eddy simulations (LES) and Reynolds averaged Navier-
nearly exactly in this approach. This represents a major advantage for
Stokes simulations (RANS)-are investigated. The theoretical challenges
LBM over conventional CFD for the simulation of turbulent flows,
and computational advantages of LBM turbulence calculations are
wherein much of the numerical difficulties arise due to the advective
detailed.
term. Further, the close connection with Navier-Stokes methods
enables LBM to be applied without any change to all flows including
KEY WORDS: Lattice Boltzmann Method, turbulence, frame those of fluids. Therefore, LBM has the potential for becoming a
rotation, periodic shear, turbulent jets, bluff-bodies, temperature frontline CFD tool for the ocean engineering community.
fluctuations, decaying turbulence, homogeneous shear, oscillating
shear, two-fluid models, binary mixtures. In this paper, we will focus on the various turbulence calculations
performed using LBM and other kinetic theory-based methods. The
INTRODUCTION procedure for filtering Boltzmann equation for filtered turbulence
description is provided in Girimaji (2007). The focus here will be
Kinetic Boltzmann equation describes the evolution of a single particle mostly on direct numerical simulations (DNS) and large eddy
velocity probability distribution function subject to collisions with simulations (LES) of various turbulent flows. This review of kinetic-
other particles in the flow. The kinetic Boltzmann equation provides theory based turbulence calculations is divided into three classes based
flow description at the mesoscale level, which is between microscale on flow complexity: (i) decaying isotropic turbulence (DIT) with
and macroscale in degree of detail. The mesoscale Boltzmann equation rotation and temperature fluctuation effects; (ii) homogeneous shear
is valid over the entire range of Knudsen numbers and a broader range turbulence with constant and periodic shear; and (iii) complex
of physics than the macroscale Navier-Stokes equations. For gases, this inhomogeneous flows.
equation enjoys a wide range of applicability-rarefied gas dynamics,
hypersonic non-equilibrium flow, astrophysical gas dynamics and flow A REVIEW OF LBM APPLICATIONS
in nano-scale devices. This equation is also valid in many instances
wherein constitutive relations are not available, as fluxes can be
Decaying isotropic turbulence
directly computed from the particle distribution function. Despite a
wider range of applicability than the Navier-Stokes equations,
Over the last decade, application of kinetic theory based methods in
Boltzmann equations, until recently, had not been used for
turbulence calculations has taken great strides. Kinetic theory based
computational fluid dynamics (CFD) computations due to the fact that
LBM have been applied to a wide range of fluid dynamic simulations.
appropriate numerical strategies were not yet developed. The last two
McNamara and Zanetti (1988) and Higuera and Jimenez (1989) were
decades have witnessed important progress in many areas that has led
amongst the first to point out the lattice Boltzmann equation (LBE) as a
to widespread use of kinetic based methods for a variety of CFD
possible efficient tool to simulate two and three-dimensional fluid
applications. The single important development is the lattice-based
flows. Benzi and Succi (J. Phys., 1990) investigate in detail, the ability
solution of the Boltzmann equation-the so-called lattice Boltzmann

1104
of LBE to simulate two-dimensional turbulence. They compare the LES approaches with N-S. With an appropriate choice of the
scheme against pseudo-spectral simulation for the case of Smagorinsky constant Cs, even a 323 point coarse grid adequately
homogeneous forced two-dimensional turbulence. They also compare captures large scale motions and evolution of kinetic energy in LBM-
energy spectra and time evolution of energy and enstrophy to LES as shown in Figs. 3 and 4, respectively. The low wave-number
demonstrate an excellent match with pseudo-spectral method. scaling is shown for two different initial spectra with m = 2,4 in Fig. 5.
They establish LBE as a reliable tool for DNS and LES simulations of
Yu, Girimaji and Luo (2005) investigate DIT in inertial and rotating turbulence.
frames in order to assess the effectiveness of LBE as a computational
tool to perform DNS of turbulent flows. In the inertial frame they find
decay exponents, dissipation rate and low-wave number scaling of the
energy spectra agree well with established classical results. In Fig. 1
kinetic energy decays monotonically in absence of production, while
dissipation peaks first due to gradient-steepening and decays
subsequently.

Fig. 3: The instantaneous energy spectra at t' = 0.06079: LBE-LES


(323), NS-LES (323), and LBE-DNS (1283).

Fig. 1: Time evolution of the normalized kinetic energy k/k0 (solid


lines) and normalized dissipation rate ε/ε0 (dashed lines) for 643 (thin
lines) and 1283 (thick lines) domain.

The simulations begin with energy spectra given as (m=4):


⎪⎧0.038κ m e−0.14κ , κ ∈ [1,8]
2

Eˆ (κ , 0) = ⎨ (1)
⎪⎩0, κ ∉ [1,8]
The compensated spectra Eˆ (κ , t ′) / κ 4 for 1283 simulation is shown in Fig. 4: Evolution of the normalized kinetic energy: LBE-LES 323, NS-
Fig. 2. The spectra shows a dependence of low-wave number scaling LES 323, and LBE-DNS 1283.
( Eˆ (κ , t ′)ٛκ 4 ) on the initial spectrum (m = 4) which concurs with
previous results (Mansour and Wray, 1994; Huang and Leonard, 1994).

Fig. 5: Dependence of low-κ scaling on initial spectrum for 1283 grid.


The dash lines are initial spectra (t' = 0) and the solid lines are Eˆ (κ , t ′)
at t' = 0.022. (a) m = 2 and (b) m = 4 for the initial spectrum.
Fig. 2: The compensated energy spectrum for an 1283 simulation with
Djenidi (2006) performs DNS of grid-generated turbulence using LBM
urms = 0.023 and Reλ = 141 , at early times, t =0.0, 0.022, 0.044, in order to improve comparison between experimental and numerical
0.066 and 0.088. results on approximate isotropic turbulence. The grid for experiment
was made up of 4X4 floating flat square elements in an aligned
Yu, Girimaji and Luo (2005) further assess the applicability of LBE as arrangement. The numerical results from LBM-DNS matched closely
a computational tool for LES and DNS of turbulence. By comparing with experimental data of grid turbulence. They point out that extra
LBE-DNS and LBE-LES, they find LBM to adequately capture large caution should be exercised when one is trying to assess the existence
scale flow behavior. NS-LES and LBE-LES for DIT were compared of a power-law decay, both in numerics and experiments. They discuss
and LBE-DNS was found to preserve instantaneous flow fields the effect of mesh resolution and boundary condition on LBM
somewhat more accurately. They conclude that the Smagorinsky simulations. They show that by not accounting for large scale motion
constant Cs should be smaller than the typical value used in traditional (domain-size) or resolving the small-scales (mesh resolution)

1105
adequately, the ends of spectra are poorly resolved.

Kerimo and Girimaji (2007) apply Boltzmann BGK based Gas Kinetic
Method (GKM), MRT-LBM and N-S equations for weakly
compressible turbulence. They evaluate the viability of extending the
application of GKM in simulating turbulent flows. GKM is compared
against LBM and N-S to evaluate the accuracy and robustness of
schemes. They perform three-dimensional DNS of DIT and compare
the evolution of kinetic energy and dissipation rate obtained using
GKM, N-S and LBM in Fig. 6. One-dimensional energy spectra at
eddy-turnover time t'=0.23 and t'=0.64 are compared in Figs. 7 and 8
respectively, to demonstrate the extent of scales captured by schemes.
A qualitative comparison of iso-surfaces of x-component of vorticity at Fig. 8: Normalized energy spectrum (Ê(κ)/k0) as a function of wave
t'=0.64 is shown in Fig. 9. An almost exact match in statistical number κ at t′ = 0.64 for LBM (solid line), GKM (square) and NS
quantities (k,ε), energy spectra and iso-surfaces of ωx strongly support (circle) and at t′ = 0 (dashed line).
the viability of LBM and GKM for turbulence calculations.

(a) (b)

Fig. 6: Evolution of normalized kinetic energy (k/k0) and the dissipation


rate ε/ε0 for GKM, N-S and LBM in eddy-turnover time t ′ = t (ε 0 k0 ) .

(c)
Fig. 9: Iso-surfaces of x-component of vorticity ωx at t' = 0.64 for: (a)
GKM (b) LBM and (c) N-S (Red: ωx =0.013; Green: ωx =-0.013).

where κ p = (κ max − κ min ) / 2 characterizes the energy containing wave


number. The effect of rotation depends on energy containing scales and
is enhanced by increasing rotation rate ω or decreasing Rossby number.
Prior studies by Clark and Zemarch (2002 and Yamazaki, Kaneda, and
Rubinstein (2002) have established slowdown of cascade and delayed
approach to equi-partition due to rotation. In Fig. 10, DNS with LBM
captures slowdown in decay of kinetic energy with decreasing Rossby
number in DIT simulations

Fig. 7: Normalized energy spectrum (Ê(κ)/k0) as a function of wave


number κ at t′ = 0.23 for LBM (solid line), GKM (square) and NS
(circle) and at t′ = 0 (dashed line).

LBM with frame rotation

Applicability of LBM in rotating reference frame is evaluated by Yu,


Girimaji and Luo (2005). Simulations with LBM show that the decay
rate of kinetic energy decreases with Rossby number as the energy
cascade is inhibited by rotation. The Rossby number is defined as:
κ p urms .
Fig. 10: Kinetic energy decay in 1283 LBE-DNS with different Rossby
Ro = (2)
ω (Ro) numbers.

1106
He and Doolen (1997) used an interpolation-based strategy to extend
the LBM for general curvilinear coordinate systems in-order to study
flow around a circular cylinder. Compared with previous lattice
Boltzmann simulations of the same problem, their new approach
greatly enhanced the computational efficiency of the scheme. Zhang et
al (2007) proposed an algorithm for rotationally invariant lattice
Boltzmann method (RILBM). Their approach overcomes discrete
artifacts presented in the standard lattice Bhatnagar, Gross, and Krook
(LBGK) model by introducing a generalized particle collision operator Fig. 11: Schematic of the interaction between kinetic and thermal
in arbitrarily rotated frames. They further demonstrate an exact modes.
recovery of the N-S equations through the Chapman-Enskog expansion.
The independence of numerical results relative to the lattice orientation Table 1: Simulation cases
is shown using numerical simulations: (i) decay of sinusoidal wave
involving two distinct orientations and (ii) propagation of sound wave. Cases Isothermal Case I Case II
Zhang et al (2010) proposed a generalized lattice Boltzmann based Domain size 1283 1283 1283
approach for sliding-mesh local reference frame. Their scheme exactly r
conserves the hydrodynamic fluxes across local reference frame v field κ ∈ [1 − 8] κ ∈ [1 − 8] κ ∈ [1 − 8]
interface. They demonstrate the accuracy and robustness of the scheme T field κ ∈ [1 − 4] κ ∈ [1 − 4] κ ∈ [5 − 8]
using benchmark problems: (i) DNS of two-dimensional rotating
Reλ , Pr 45, 0.6 45, 0.6 45, 0.6
cylinder, (ii) DNS of two-dimensional blade in a cross flow and (iii)
three-dimensional flow past D4119 propeller. The benchmark
validations establish the effectiveness and accuracy of the scheme. One
of the key features of their algorithm is that it can be parallelized easily
and achieve almost linear scalability across massive computing
clusters.

Decaying turbulence with temperature fluctuations

In compressible turbulent flows, activation of thermal mode and


subsequent coupling with dilatational kinetic energy of the flow field
introduces new complicated physics not observed in incompressible
flows. Lee, Yu and Girimaji (2006) perform DNS of low Mach number Fig. 12: Temporal variation of accumulated kinetic energy budgets.
decaying isotropic turbulence, with a randomly distributed initial PW=pressure work, DD=dilatational dissipation, SD=solenoidal
temperature field, using hybrid thermal LBM (HTLBM) (Lallemand dissipation.
and Luo, 2003). The popularity of HTLBM is attributed to elimination
of unphysical coupling between energy and shear modes of LBM and
improvement in numerical stability. Lee et. al study the effect of
temperature fluctuations on turbulence. The interaction between kinetic
and thermal energy modes and the small-scale flow structures is
attributed to the key role played by pressure as depicted in schematic of
Fig. 11. Pressure can lead to two way exchange of energy between
kinetic and thermal modes unlike viscosity for which the exchange is
uni-directional. Three simulation cases were studied in which different
length-scales of the initial incompressible velocity and temperature
fields were used as given in table 1. All simulation cases were shown to
conserve energy and satisfy budget of the turbulent kinetic energy
equation, which in-turn establishes the fidelity of LBM simulations.
The incompressible/solenoidal component of the velocity field is found
to be unaffected by the temperature fluctuations while the Fig. 13: Alignments of vorticity vector with eigenvectors of strain-rate
compressible/dilatational component is affected through pressure work tensor (t'=0.4).
and dilatational dissipation as shown in Fig. 12. They analyze the
structure of velocity gradient tensor and find that the vorticity vector Magneto-Hydro-Dynamics flows
aligns with the intermediate eigenvector of the strain-rate tensor as in
isothermal flows. In Fig. 13, distribution of the angle ζ1, ζ2 and ζ3 In this section we discuss simulations of MHD fluid flows (plasma,
between the vorticity vector and eigenvalues of the strain-rate tensor (α liquid metals, electrolytes) using LBM. Deller (2002) modified the low
> β > γ) shows alignment for all three cases. But, the alignments among Mach number LBM to include the magnetic Lorentz force. The
vorticity vector and the gradients of pressure, density and temperature magnetic field (B) is represented by a separate vector-valued magnetic
are considerably altered by temperature fluctuations as depicted in Fig distribution function which obeys a vector Boltzmann--BGK equation
14. Their study shows LBM to be a viable tool for simulating and recovers the correct macroscopic equation for the magnetic field.
compressible turbulence with a thermal field. However, in the current The MHD-LBM scheme preserves a divergence free magnetic field
form the HTLBM is limited to low Mach number flows. (∇⋅B=0). Simulations by the Deller for Orszag-Tang vortex, Hartmann

1107
observed as seen from Figs. 17 and 18, respectively. Throughout the
decay process there is an exchange of energy between kinetic and
magnetic modes through Alfen waves, while the total energy decreases
monotonically. Lorentz forces play a key role in exchange of energy
between kinetic and magnetic modes. The exchange of energy is
oscillatory with intermittent large fluctuations. They find that magnetic
field effects dominate the evolution of enstrophy and when interaction
parameter N is high vortex compression is as likely as vortex stretching
(see Fig. 19).

Fig. 14: Alignments among vorticity vector and gradient vectors of


thermodynamic variables (t'=0.4).

flow and the doubly periodic coalescence instability using MHD-LBM


show favorable agreement with spectral results. Dellar's magneto-
hydro-dynamic LBM (MHD-LBM) is used widely in its original or Fig. 15: Rectangular jet flow profiles for Re = 150 without magnetic
modified form (Deller, 2002; Breyiannis and Valougeorgis, 2004; field (N=0).
Premnath and Pattison, 2005; Pattison, Premnath, Morley andAbdou,
2008; Riley, Richard and Girimaji, 2008).

To account for the effect of the magnetic on velocity field two possible
LBM formulations are: (i) body-force formulation (BFF) where the
magnetic field effects manifest as an external acceleration and (ii)
extended equilibrium formulation (EFF) where the effect appears
through a modified equilibrium distribution function. Also, two models
are available for the velocity field: single-relaxation time (SRT) and Fig. 16: Rectangular jet flow profiles for Re = 150 with magnetic field:
multi-relaxation time (MRT) model. Riley, Richard and Girimaji β =3.35, N=3.0 and Rm = 5.
(2008) developed the implementation of MRT-EEF for MHD-LBM
and compared with SRT-BFF, SRT-EEF and MRT-BFF. They find
MRT-EEF to be the most robust and accurate amongst the MHD-LBM
computational schemes examined.

Riley, Girimaji, Richard and Lee (2009) perform simulations of MHD


plasma jets and investigate axis-switching and flow instabilities in
rectangular jets with/without an external magnetic field. In the absence
of an externally applied magnetic field, rectangular jets switch axes due
to differential spreading along the major and minor axes as shown in
Fig 15. An externally applied axial magnetic field decelerates the jet
(thereby enhancing spreading), inhibits instabilities and prevents axis
switching as shown in Fig 16. The external magnetic field is
characterized by an interaction parameter (N) and stability parameter β
which is ratio of hydrodynamic to magnetic pressure given as Fig. 17: Kinetic energy decay in eddy-turnover time (τ): Case I: B=0,
solid line; Case II: N=0.0, dash-dotted line; Case III: N=0.3, dotted line;
σ B2 H 2μ ρ u 2 2 R Case IV: N=0.05: dashed line.
N≡ ; β ≡ 02 = m (3)
ρu B N
where, σ is conductivity, H is jet height, ρ is density, u is characteristic
macroscopic speed, Rm is magnetic Reynolds number and μ0 is the
magnetic permeability of free space. The Lorentz forces arising due to
the magnetic field reverse the natural circulation of vortices. The
vorticity reversal makes the secondary flow in the transverse plane to
be inward along the minor axis and outward along the major axis. This
reversal in vorticity prevents axis-switching. The study also points
towards transfer of energy between kinetic and magnetic modes as
another key mechanism for the changes. Richard, Riley and Girimaji
(2011) apply MHD-LBM to perform DNS to study energy and
enstrophy in decaying MHD turbulence subject to initially uniform or
random magnetic fields. They also investigate kinetic-magnetic energy
exchange, velocity field anisotropy, action of Lorentz force, behavior
of enstrophy and helicity, and internal structure of the small scales. In Fig. 18: Magnetic energy decay in eddy-turnover time (τ): Case I: B=0,
both simulations, with initially uniform or random magnetic fields, a solid line; Case II: N=0.0, dash-dotted line; Case III: N=0.3, dotted line;
tendency of equi-partition between kinetic and magnetic energy is Case IV: N=0.05: dashed line.)

1108
grows and turbulence is sustained while for ω> ωcr kinetic energy
decays and turbulence dies. At very high frequency (ωcr>10),
periodically forced turbulence behaves similar to DIT as shown in Fig.
24.

Fig. 19: PDF of the different components of the total time rate of
change of enstrophy at eddy-turnover time τ =0.25 (N=0.3).
Fig. 21: The results of b12 obtained from DNS using LBM and N-S
Homogeneous turbulence in uniform shear equations by Jacobitz, Sarkar and Van Atta (1997) with the initial
values of Reλ=44.72 and S*=2.0.
Yu and Girimaji (2005) study homogeneous turbulence subject to
uniform shear using LBM. They perform DNS to investigate the
dependence of the asymptotic state of turbulence on the initial
Reynolds number and strain rate. The validity and accuracy of LBM for
sheared turbulence is established by comparing evolution of normalized
Reynolds stresses and Reynolds stress anisotropy against classical
experimental data and N-S DNS results in Fig. 20 and 21, respectively.
In the low Reynolds number regime, the asymptotic state of turbulence
depends weakly on initial shear S* but strongly on initial Reynolds
number Reλ. Initial normalized shear S* has a relatively small effect on
the asymptotic values of S*, P/ε and b12 while the effect on normal
stress anisotropies are relatively larger. Fig. 22 shows evolution of
Reynolds stress anisotropy b12 to be weakly dependent on S* for a fixed
initial Reλ. The dependence on Reynolds number diminishes with
increasing Reynolds number. For fixed S*, evolution of b12 is strongly
dependent on initial Reλ as shown in Fig. 23. Fig. 22: Time evolution of b12 for different S* with initial Reλ=84. b12
(asymptotic) =−0.16.

Fig. 20: Evolution of normal Reynolds stress in homogeneous turbulent


shear flow. NS-DNS data are from Rogers (1986) with initial values of
Reλ≈35 and S*= 1.2. Experimental data are from Tavoularis and Corrsin Fig. 23: Time evolution of b12 for different initial Reλ with S*=6.9.
(1981) with Reλ= 284 at St = 8.6.
They show in Fig. 25 that net production of turbulence per cycle of
Homogeneous turbulence in oscillating shear oscillating shear decreases with increasing ω. This decrease in
production is attributed to an asymptotic decrease in phase difference
Yu and Girimaji (2006 perform DNS of homogeneous turbulence between applied strain and Reynolds stress from π in constant shear ( ω
subject to periodic shear using LBM. A periodic body force is applied = 0) to π/2 in very high frequency shear cases. Figs. 26 and 27 show
to produce the required shear which is given by: change in phase difference between anisotropy component b12 and
S = Smax sin(ωt ) (4) shear, for ω = 0.5 and ω = 1.0, respectively. Thus, when the production
falls below dissipation, turbulence decays. Evolution of kinetic energy
where ω is the forcing frequency and Smax is the magnitude of
maximum shear. They find turbulence statistics to be a strong function and dissipation for (ω/Smax = 0.5) is shown in detail in Figs. 28. The
of the forcing frequency with a bifurcation point in behavior of frequency of k is twice that of shear and evolution of k and ε are in
turbulence at a critical frequency ωcr/Smax≈0.5. For ω<ωcr kinetic energy phase. This observation has important modeling implications.

1109
Fig. 24: Evolution of k and ε in DIT and homogeneous shear Fig. 27: Evolution of b12 in ω/Smax = 1.0 case.
turbulence with ωcr =10.

Fig. 25: Evolution of k in various ω/Smax cases.


Fig. 28: Evolution of k and ε in the ω/Smax = 0.5 case.

Fig. 26: Evolution of b12 in ω/Smax = 0.5 case.

Complex inhomogeneous flows Fig. 29: Half-width contours at Re=150 at different locations
downstream. The jet orifice is in dashed line.
LBM has been applied extensively for solving complex flows like: jets,
flow over bluff bodies, cavity flows. Higuera and Succi (1989) were the Luo and Girimaji (2003) developed a two-fluid lattice Boltzmann
first to demonstrate the validity of LBE as a numerical scheme for fluid model for binary mixtures. In the proposed model, viscosity and mass
flows past bluff bodies at moderate Reynolds numbers (Re <100) for diffusion coefficients can be varied independently through mutual and
two-dimensional flow past a cylinder. Wu et. al (2006) simulate flow self-collision relaxation time-scales. The LBM model can be used for
past a square cylinder in a channel using the multi-relaxation-time miscible and immiscible fluids. Yu and Girimaji (2005) study
(MRT) model in the parallel lattice Boltzmann BGK method (LBGK). instability in rectangular jets and axis-switching using LBM. The
Chen, Kandasamy, Orszag, Shock, Succi and Yakhot (2003) suggest aspect ratio (AR) of the rectangular jet studied is 1.5. They perform
that complex fluid physics can be modeled using an extended kinetic simulations of jet flow for Reynolds numbers 10, 100, 150 and 200. At
equation more efficiently than N-S. They explain LBM for modeling low Reynolds number (Re = 10, 100), the jet flow is laminar and stable.
fluid turbulence and demonstrate its computational effectiveness. The For Re=150, flow is still laminar and stable but axis-switching occurs
premise of their approach is that, it is simpler to model complex downstream of the flow as shown in Fig. 29. Axis-switching is closely
physics at the Boltzmann (or kinetic) level first, and then use coarse related to the dynamics of corner vortices. Fig. 30 shows half-width
graining procedures. They study NACA 4412 and flow past a real car contours at higher Reynolds number Re=200, where jet becomes
geometry and show an excellent match with experimental data.

1110
unstable and a rough axis-switching is observed. turbulence at low Reynolds numbers,” Physics of Fluids, Vol 6, pp
3765-3775.
Kerimo, J, and Girimaji, SS (2007). “Boltzmann-BGK approach to
simulating weakly compressible 3D turbulence: comparison between
lattice Boltzmann and gas kinetic methods,” Journal of Turbulence,
Vol 8, No 46, pp 1–16.
Lallemand, P, and Luo, L-S (2003). “Theory of the lattice Boltzmann
method: Acoustic and thermal properties in two and three dimensions,”
Physical Review E, Vol 68, No 3, pp 036706.
Lee, K, Yu, D, and Girimaji, SS (2006). “Lattice Boltzmann DNS of
decaying compressible isotropic turbulence with temperature
Fluctuations,” International Journal of Computational Fluid
Dynamics, Vol 20, No 6, pp 401–413.
Luo, L-S, and Girimaji, SS (2003). “Theory of the lattice Boltzmann
method: two-fluid model for binary mixtures,” Physical Review E, Vol
67, pp 036302.
Mansour, N, and Wray, A (1994). “Decay of isotropic turbulence at low
Reynolds number,” Physics of Fluids, Vol 8, pp 808-814.
Fig. 30: Half-width contours at Re=200 at different locations McNamara, GR, and Zanetti, G (1988). “Use of the Boltzmann equation
downstream. The jet orifice is in dashed line. to simulate lattice-gas automata,” Phys. Rev. Lett., Vol 61, pp 2332–
2335.
Near-field LES for low aspect ratio rectangular jets using LBM is Pattison, M, Premnath, K, Morley, N, and Abdou, M (2008). “Progress in
performed by Yu and Girimaji (2005). LES technique combines lattice Boltzmann methods for magnetohydrodynamic flows relevant
D3Q19 multiple relaxation time LBE with Smagorinsky model for the to fusion applications,” Fusion Engineering and Design, Vol 83, No 4,
sub-grid stress. They demonstrate that, for LES of turbulent flows, pp 557–572.
MRT-LBE model is more suitable than the widely used single- Peng, Y, Liao, W, Luo, L-S, and Wang, L-P (2010). “Comparison of the
relaxation-time LBE model. They compare the near-field behavior lattice Boltzmann and pseudo-spectral methods for decaying
simulated using MRT-LBM against experimental data. They find the turbulence: Low-order statistics,” Computers & Fluids, Vol 39, No 4,
statistical behavior of turbulent jets to be more sensitive to inflow pp 568-59.
velocity than transverse boundary conditions. Premnath, K, and Pattison, M (2005). “Computation of MHD flows
using the lattice Boltzmann method,” MetaHeuristics LLC.
CONCLUSIONS Richard, J, Riley, B, and Girimaji, SS (2011). “Magnetohydrodynamic
turbulence decay under the influence of uniform or random magnetic
In the last two decades, development of appropriate numerical fields,” Journal of Fluids Engineering, Vol 133, pp 081205:1-9.
strategies for LBM has brought about applications for a wide range of Riley, B, Girimaji, SS, Richard, J, and Lee, K (2009). “Magnetic field
flows. The ease of flux calculation, computational advantages in effects on axis-switching and instabilities in rectangular plasma jets,”
handling of advection term and a richer physics content compared to Flow Turbulence Combust, Vol 82, pp 375-390.
Navier-Stokes, puts LBM in the frontline of CFD tool. For complex Riley, B, Richard, J, and Girimaji, SS (2008). “Assessment of two
flows like ocean and off-shore engineering, LBM is yet to be fully magnetohydrodynamic lattice Boltzmann models in rectangular jets
exploited. Canonical flows of interest to coastal engineering have been and turbulence,” International Journal of Modern Physics C, Vol 19,
studied extensively using LBM and some of the relevant works have No 8, pp 1211-1222.
been reviewed in the paper. A great deal of success in variety of Yamazaki, Y, Kaneda, Y, and Rubinstein, R (2002). “Dynamics of
simulations for such canonical flows has set-up an excellent framework inviscid truncated model of rotating turbulence,” Journal of the
for future applications of LBM in coastal engineering applications. Physical Society of Japan, Vol 71, pp 81-92.
Yu, D, and Girimaji, SS (2005). “DNS of homogenous shear turbulence
REFERENCES revisited with the lattice Boltzmann method,” Journal of Turbulence,
Vol 6, No 6, pp 1-17.
Chen, H, Satheesh, K, Orszag, S, Shock, R, Succi, S, and Yakhot, V Yu, D, and Girimaji, SS (2006), “Direct numerical simulations of
(2003). “Extended Boltzmann Kinetic Equation for Turbulent Flows,” homogeneous turbulence subject to periodic shear,” Journal of Fluid
Science, Vol 301, No 5633, pp 633-636. Mechanics, Vol 566, pp 117-151.
Clark, T, and Zemarch, C (1998). “Symmetries and the approach to Yu, H, and Girimaji, SS (2005). “Near-field turbulent simulations of
statistical equilibrium in isotropic turbulence,” Physics of Fluids, Vol rectangular jets using lattice Boltzmann method,” Physics of Fluids,
10, pp 2846–2858. Vol 17, pp 125106.
Breyiannis, G, and Valougeorgis, D (2004). “Lattice kinetic simulations Yu, H, Girimaji, SS and Luo, L-S (2005). “DNS and LES of decaying
in three-dimensional magneto-hydro-dynamics,” Phys. Rev. E, Vol 69, isotropic turbulence with and without frame rotation using lattice
pp 065702. Boltzmann method,” Journal of Computational Physics, Vol 209, pp
Deller, P (2002). “Lattice kinetic schemes for magnetohydrodynamics,” 599-616.
Journal of Computational Physics, Vol 179, pp 95-126. Yu, H, Girimaji, SS, and Luo, L-S (2005). “Lattice Boltzmann
Guo, Z, Liu, H, Luo, L.-S, and Xu, K (2008). “A comparative study of simulations of decaying homogeneous isotropic turbulence,” Physical
the lbe and gks methods for 2d near incompressible laminar flows,” Review E, Vol 71, pp 016708.
Journal of Computational Physics, Vol 227, No 10, pp 4955–4976. Yu, H, and Girimaji, SS (2006). “Lattice Boltzmann equation simulation
Higuera, FJ, and Jimenez, J (1989). “Boltzmann approach to lattice gas of rectangular jet (AR=1.5) instability and axis-switching,” Physica A,
simulations,” Europhysics letters, Vol 9, No 7, pp 663–668. Vol 362, pp 151-157.
Huang, M-J, and Leonard, A (1994). “Power-law decay of homogeneous

1111

You might also like