Lecture # 1: Introduction to Multiphase Flow

Overview
• Multiphase Flow Basics
– General Features and Challenges
– Characteristics and definitions
• Conservation Equations and Modeling Approaches
– Fully Resolved
– Eulerian-Lagrangian
– Eulerian-Eulerian
• Averaging & closure
– When to use what approach?
• Preferential concentration
• Examples
• Modified instability of a Shear Layer
• Sediment suspension in a turbulent channel flow
• Numerical simulation example: Mesh-free methods in multiphase flow
What is a multiphase flow?
• In the broadest sense, it is a flow in which two or more
phases of matter are dynamically interacting
– Distinguish multiphase and/or multicomponent
• Liquid/Gas or Gas/Liquid
• Gas/Solid
• Liquid/Liquid
– Technically, two immiscible liquids are “multi-fluid”, but are often referred to as
a “multiphase” flow due to their similarity in behavior

Single component Multi-component

Water Air
Single phase
Pure nitrogen H20+oil emulsions
Steam bubble in H20 Coal particles in air
Multi-phase
Ice slurry Sand particle in H20
Dispersed/Interfacial
• Flows are also generally categorized by distribution of the
components
– “separated” or “interfacial”
• both fluids are more or less contiguous
throughout the domain

– “dispersed”
• One of the fluids is dispersed as non-
contiguous isolated regions within the
other (continuous) phase.
• The former is the “dispersed” phase,
while the latter is the “carrier” phase.
• One can now describe/classify the
geometry of the dispersion:
• Size & geometry
• Volume fraction
Gas-Liquid Flow
Bubbly Pipe Flow – heat exchangers in power plants, A/C units
Gas-Liquid Flow (cont)
Aeration:
-produced by wave action
- used as reactor in chemical processing
- enhanced gas-liquid mass transfer
Gas-Liquid Flow (cont)
Ship wakes – detectability
Cavitation – noise, erosion of structures
Liquid-Gas Flow

Weather – cloud formation

Biomedical – inhalant drug delivery

Vukasinovic, Glezer, Smith (2000)

http://www.mywindpowersystem.com/2009/07/wind-power-when-nature-gets-angry-the-worst-wind-disasters-of-the-world/
Gas-Liquid Flow
Energy production – liquid fuel combustion
Biomedical – inhalant drug delivery

http://convergecfd.com/applications/engine/sparkignited/
Image courtesy A. Aliseda

Album of fluid motion, Van Dyke

Gas-Solid Flow
Environmental – avalanche, pyroclastic flow, ash plume, turbidity currents
Gas-Solid (dense)
Granular Flow – collision dominated dynamics; chemical processing

http://www.its.caltech.edu/~granflow/homepage.html

http://jfi.uchicago.edu/~jaeger/group/
Liquid-Liquid
Chemical production – mixing and reaction of immiscible liquids

http://www.physics.emory.edu/students/kdesmond/2DEmulsion.html
Solid-Liquid

Sediment Transport –
pollution, erosion of beaches,
drainage and flood control
Solid-Liquid

Settling/sedimentation,
turbidity currents

http://www.physics.utoronto.ca/~nonlin/turbidity/turbidity.html
Solid-Gas
Material processing – generation of particles & composite materials
Energy production – coal combustion
Solid-Gas
Aerosol formation – generation of particles & environmental safety
Classification by regime
• Features/challenges
– Dissimilar materials (density, viscosity, etc)
– Mobile and possibly stochastic interface boundary
– Typically turbulent conditions for bulk motion
• Coupling
• One-way coupling: Sufficiently dilute such that
fluid feels no effect from presence of particles.
Particles move in dynamic response to fluid
motion.

• Two-way coupling: Enough particles are present

such that momentum exchange between dispersed
and carrier phase interfaces alters dynamics of the
carrier phase.

• Four-way coupling: Flow is dense enough that

dispersed phase collisions are significant
momentum exchange mechanism
• Depends on particle size, relative velocity, volume
fraction
Viscous response time
• To first order, viscous drag is usually the dominant force on the
dispersed phase
r
dv p 0 for t < 0
 3Du  v p 
r r u  
mp u, vp
dt U for t  0
dv p 18 1 U
dt

 p D2
 p   U  v p 
U  v 
p
 

v p  U 1 expt  p  
t

– This defines the typical particle “viscous response time”
  p D2
p 
18
• Can be altered for finite Re drag effects, added mass, etc. as appropriate

• Stokes number:
 p
– ratio of particle response time to fluid time scale: St 
f
Modeling approach?

• How to treat such a wide range of behavior?

– A single approach has not proved viable

• Fully Resolved : complete physics

Modeling Effort
Computational Effort
• Eulerian-Lagrangian : idealized isolated particle

• Eulerian-Eulerian : two co-existing fluids

Fully Resolved Approach
• Solve conservation laws in coupled domains
1. separate fluids
• Each contiguous domain uses appropriate transport coefficients
• Apply boundary jump conditions at interface
• Boundary is moving and may be deformable
2. single fluid with discontinuous properties
• Boundary becomes a source term
• Examples
– Stokes flow of single liquid drop
• Simple analytical solution

– Small numbers of bubbles/drop

• Quiescent or weakly turbulent flow G Tryggvason, S Thomas, J Lu, B Aboulhasanzadeh (2010)
Eulerian-Lagrangian
• Dispersed phase tracked via individual particles
– Averaging must be performed to give field properties
• (concentration, average and r.m.s. velocity, etc.)
• Carrier phase is represented as an Eulerian single fluid
– Two-way coupling must be implemented as distributed source term

Collins & Keswani (2004)

M. Garcia,
http://www.cerfacs.fr/cfd/FIGURES/IMAGES/vort_stokes2-
MG.jpg
 Particle Motion: tracer particle
• Equation of motion for spherical particle at small Rep:
r v dvr  v   r
dv p 1 
Du Du
 3Du  v p  m f  
r r
mp p
 m
 f  m 
1
p 
g

g

dt 2 Dt dt  Dt   p 
Inertia Viscous drag Added mass Pressure
gradient buoyancy
– Where
D3 r
 mp  p , the particle mass u  fluid velocity
r
6 v p  particle velocity
D3 g  fluid density
mf  f , fluid mass of same volume as particle
6
D  particle diameter
 p  particle material density
  fluid viscosity

– Possible alterations: 
 • Finite Rep drag corrections
• Influence of local velocity gradients (Faxen Corrections)
• Lift force (near solid boundary, finite Rep)
Two-Fluid Equations
• Apply averaging operator to mass and momentum equations
– Drew (1983), Simonin (1991)
• Phase indicator function

 

 x
,
kit



1
if
xiinside
phase
k

k

t

uI,j

x
k
0
0
if
xioutside
phase
k j

• Averaging operator

 
G
, volume
k
g
fraction
k

g gG kk
,ij kij ij ij 2
,
ij

– Assume no inter-phase mass flux, incompressible carrier phase


• Mass
 
 k k    k kU k, j  0 
aver
vis
str
in
te
pk
t x j k
,
ij

• Momentum I
k
,
iMean
interph
momtrae
(less
mpressu
ean
contr
on)
 U U k,i  P1  
 k k 
 t

k,i
 U k, j 

x j 
  k
x i
  k  k gi 
x j
 k  k,ij 
x j

 k k ui uj k  I k,i
Two-Fluid Equations (cont)
• Interphase momentum transport
– For large particle/fluid density ratios, quasi-steady viscous drag is by far
the dominant term
– For small density ratios, additional force terms can be relevant
• Added mass
• Pressure term
• Bassett history term
– For sediment, 2/1 ~ 2.5 > 1 (k =1 for fluid, k =2 for dispersed phase)
• Drag still first order effect, but other terms will likely also contribute

I
2
,i
I
1
,i
3
k 1
4
CD
d
v
rvr
,
i 
2

241 0.15Re 0.687  1vr d

ii p
p Re
v u i CD 
u
r
, 2
, 1
,
Re p 1
Closure requirements
• Closure
– Closure is needed for:
• Particle fluctuations
• Particle/fluid cross-correlations
• Fluid fluctuations

– Historically, the earliest models used a gradient transport model

• Shown to be inconsistent for many applications

– Alternative: Provide separate evolution equation for each set of terms

• Particle kinetic stress equation
• Particle/fluid covariance equation
• Fluid kinetic stress equation
– Also required for single-phase RANS (Reynolds-averaged Navier Stokes) models
• Also will require third-moment correlations models to complete the closure
Simpler Two-Fluid Models
• For St < 1, the particles tend to follow the fluid motion with
greater fidelity
– Asymptotic expansions on the equation of motion lead to a closed
expression for the particle field velocity, in terms of the local fluid velocity
and spatial derivatives (Ferry & Balachandran, 2001)

– Where p 3
St  w  pg 
f 2  1

– This is referred to as the “Eulerian Equilibrium” regime (Balachandar 2009).

• Also, similar to “dusty gas” formulation
 by Marble (1970)
 
• For larger St, the dispersed phase velocity at a point can be
multivalued!
 • When is a given approach best?
– What approach is best, depends on:
• D/h: Particle size and fluid length scales (typically Kolmogorov)
 p/f: Particle response time and fluid time scales
• Total number of particles: Scale of system
 , F = p:Loading of the dispersed phase (volume or mass fraction)

p/f = 1000
p/f = 25
p/f = 2.5
p/f = 0
2
 p 2  1 1 d 
  
f 36  Re h 



Balachandar (2009)
 Preferential Concentration
• From early studies, it was observed that inertial particles can be
segregated in turbulent flows
– Heavy particles are ejected from regions of strong vorticity
– Light particles are attracted to vortex cores

• Small St approx. shows trend

– Taking divergence of velocity…

 
St = 1.33 St = 8.1
2
 u p  St1   Sij  
2



Wood, Hwang & Eaton (2005)

 PDF formulation
Fevier, Simonin & Squires (2005)
• Consequences of inertia
– Implies history of particle matters
– Particles can have non-unique velocity
• How can models account for this?
• Probability distribution function
– f(x,u,t) = phase space pdf

f
  x  u p f   u  a p f 
t p

dx p
up   x      x i  u p      u p,i
dt
 r
1    r  f
v v
du p 1 Du dv p  Du
3Du  v p  m f  
r r
ap    m f  m p  g
1 g 

dt m p 
 2 Dt dt  Dt   p  

 

– Instantaneous point quantities come as moments of 
 the pdf over velocity phase space up


 f x p ,u p ,t du p

n x,t  

uˆ p x,t    u f x
p p ,u p ,t du p


Examples
• Effect of particles on shear layer instability

• Particle-Fluid Coupling in sediment transport

• Case studies in interface tracking methods

Effect of particles on Kelvin-Helmholtz instability
• How does the presence of a dynamics dispersed phase influence
the instability growth of a mixing layer?
U
y


x

DU
Results
• Effect of particles
– At small St, particles follow flow
exactly, and there is no dynamic
response. Flow is simply a heavier
fluid.

– As St is increased, dynamic slip

becomes prevalent, and helps damp
the instability

– At large St, particles have no

response to perturbation and are
static

– Effect is stronger, for higher loadings,

but shear layer remains weakly
unstable

1/St
Mechanism
• Particles damp instability
Fluid streamfunction
– Particles act as a mechanism to
redistribute vorticity from the core
back to the braid, in opposition to
the K-H instability Dispersed concentration

 u p


• Limitations
– Results at large St do not capture
effects of multi-value velocity

Meiburg et al. (2000)

Sediment Transport in Channel Flow

g
2h
y

x
Concentration Mean Velocity

• Planar Horizontal Water Channel

– 4  36  488 cm, recirculating flow
– Pressure gradient measurements show fully-developed by x = 250 cm
– Particles introduce to settling chamber outlet across span
Experimental Conditions
• Both single-phase and two-phase experiments conducted
• Carrier Fluid Conditions
– Water, Q = 7.6 l/s
– Uc = 59 cm/s, u = 2.8 cm/s, Re = 570
– Flowrate kept the same for two-phase experiments
– Tracer particles: 10 m silver-coated, hollow glass spheres, SG = 1.4

• Dispersed Phase Conditions

– Glass beads: (specific gravity, SG = 2.5)
– Standard sieve size range: 180 < D < 212 m
– Settling velocity, vs = 2.2 to 2.6 cm/s
– Corrected Particle Response Time,  = 4.5 ms
– St+ = p/+ ~ 4 p

– Bulk Mass Loading: dM/dt = 4 gm/s, Mp/Mf ~ 5  10-4

– Bulk Volume Fraction,  = 2  10-4
Mean Concentration Profile
– Concentration follows a power law
• Equivalent to Rouse distribution for
infinite depth

a a
C
()
y hyy y 


o



o


C
(
y) yh
o 
yo y

v 2
.44
cm/
s
 
 
s
a 2.
07
u (
0.
40
)(
2
.95
cm/s
)

• Based on mixing length theory, but still

gives good agreement
Mean Velocity

– Particles alter mean fluid profile

• Skin friction increased by 7%; qualitatively similar to effect of fixed roughness
– Particles lag fluid over most of flow
• Observed in gas/solid flow (much large Stokes number… likely not same reasons)
• Particles on average reside in slower moving fluid regions?
– Reported by Kaftori et al, 1995 for p/f = 1.05 (current is heavier ~ 2.5)
– Organization of particles to low speed side of structures – a la Wang & Maxey (1993)?
– Particles begin to lead fluid near inner region – transport lag across strong gradient
Particle Slip Velocity, uf up p uf up p

uf up vf vp vf vp

p
p
p

uf up
p

vs

– Streamwise direction
• Particle-conditioned slip (+) is generally small in outer flow
• Mean slip (•) and particle conditioned slip are similar in near wall region
– Wall-normal direction
• Mean slip (•) is negligible
• Particle-conditioned slip (+) approximately 40% of steady-state settling velocity (2.4 cm/s)
Particle Conditioned Fluid Velocity

uf vf
p,up

uf
p,down
vf

uf vf
p,up
p,down

– Average fluid motion at particle locations:

• Upward moving particles are in fluid regions moving slower than mean fluid
• Downward moving particles are in fluid regions which on average are the same as the fluid
• Indicates preferential structure interaction of particle suspension
Suspension Quadrant Analysis

– Conditionally sampled fluid velocity fluctuations

• Upward moving particles primarily in quadrant II
• Downward moving particles are almost equally split in quadrant III and IV

– Persistent behavior
• Similar quadrant behavior in far outer region
• Distribution tends towards axisymmetric case in outer region
Expected Structure: Hairpin “packets”

• Visualization of PIV data in single-phase boundary layer

– Adrian, Meinhart & Tomkins (2000), JFM
– Use “swirl strength” to find head of hairpin structures
• Eigenvalues of 2-D deformation rate tensor, swirl strength is indicated by magnitude of complex
component

Hairpin
packets

Swirl
strength
contours

– Spacing ~ 200 wall units

– Packet growth angle can increase or decrease, +10° on average
– Packets were observed in 80% of images (Re = 7705)
Event structures: Quadrant II hairpin

Swirl
Strength
Q2 & Q4
contribution
s

• Similar structures found

– Appropriate spacing
– Not as frequent
• Re effects? (Re = 1183)
• Smaller field of view?
• Evidence suggests packets contribute to particle suspension
Particle Kinetic Stress
• Turbulence budget for particle stresses
• (Wang, Squires, Simonin,1998)
0
  
  U 2,m  u2,i u2, j  P2,ij  D2,ij   d2,ij   2,ij
p

t x m  2

U 2, j U 2,i
P2,ij   u2,i u2,m  u2, j u2,m
– Production by mean shear x m x m


– Transport by fluctuations
D2,ij  
1
 2 x m

 2 u2,i u2, j u2,m 2


1 3 Cd
 d2,ij   v u u
– Momentum coupling to fluid  2 2 d r 2,i 2, j 2
– (destruction) 
1 3 Cd
 2,ij
p
 v u u  u u 
– Momentum couplingto fluid  2 4 d r 1,i 2, j 1, j 2,i 2
– (production)
Particle Kinetic Stress Budget
• Streamwise Particle/Fluid Coupling: d2,11, p2,11
– Compare results to Wang, Squires, & Simonin (1998)
• Gas/solid flow (2/1=2118), Re = 180, No gravity, St+~700
• Computations, all 4 terms are computed; Experiments, all but D 2,ijcomputed

Experiment, solid/liquid 0.2 Wang, et al, solid/gas

0.1

-0.1

-0.2

– Interphase terms are qualitatively similar

Similar general shapes, d11 > 
p
11
– Quantitative difference
• Magnitudes different: d11 /  11~1.3 vs 3, overall magnitudes are 10 to 20 times greater
p

– Interphase terms are expected to increase with decreased St+

• Dominant interphase transfer () greatly diminishes importance of mean shear (P)
• Turbulent transport (D) has opposite sign because of small shear production (P)
Summary
• We have looked at examples of multi phase flow in real
life situations
• We have also experienced the analysis of problems in
multi phase flow
• It should be emphasized that there are many
approaches in solving problems in multi phase flow
because of the nature of the conditions
• The next lecture will look at specific multi phase flow
phenomenon