Phenomena Affecting Multiphase (Dispersed) Reactor

Performance
Flow dynamics of the multi-phase dispersion
- Fluid holdups & holdup distribution
- Fluid and particle specific interfacial areas
- Bubble size & catalyst size distributions

Fluid macro-mixing
- PDF’s of RTDs for the various phases

Fluid micro-mixing Reactor

- Bubble coalescence & breakage Model or
- Catalyst particle agglomeration & attrition Scale-Up
Package
Heat transfer phenomena
- Liquid evaporation & condensation
- Fluid-to-wall, fluid-to-internal coils, etc.

Energy dissipation
- Power input from various sources
(e.g., stirrers, fluid-fluid interactions,…)
Chemical Processes: The Scale Issue
• Process scale-up is difficult mainly because the
flow patterns (hydrodynamics) and associated
transport effects are dependent on size and
capacity of process
• This is further complicated because we do not
have equally powerful flow imaging / diagnostic
Chemistry
experimental and computational tools at
different scales
Chemistry
+ Interface
Chemistry +
Transport
Interface
Transport +
Industrial
Scale Flow
Chemistry + Patterns
Interface
Transport +
Laboratory
Scale Flow
Patterns Chemistry +
Interface
Transport +
Pilot Plant
Scale Flow
Patterns
Modeling
Computational modelling is a method of representing a real world
process by equations (model)

Retrofits
B

Process A C R&D
Synthesis
Computational
Modelling

Process
Equipment E D Improvement
Design analysis
Coupling between Phases
• One-way coupling:
– Fluid phase influences particulate phase via aerodynamic drag and
turbulence transfer.
– No influence of particulate phase on the gas phase.
• Two-way coupling:
– Fluid phase influences particulate phase via aerodynamic drag and
turbulence transfer.
– Particulate phase reduces mean momentum and turbulent kinetic
energy in fluid phase.
• Four-way coupling:
– Includes all two-way coupling.
– Particle-particle collisions create particle pressure and viscous
stresses.
Types of Interactions in Gas Solid
Dispersed Flow
Turbulence
Gas Mean Gas Fluctuating
Motion Motion

Closure Flux of kinetic

Drag energy
Problem
Particle Particle
Mean Fluctuating
Motion Motion
Kinetic Theory of Granular
Flow (KTGF)
Multi Level Modeling Concept
(Ultimate Wish)
Experimental Continuum Models
Large Scale Simulations
Validation (Macroscopic)

Experimental Discrete particle model Particle-particle interaction

Validation (Mesoscopic) closure laws

Experimental Lattice Boltzmann Model Fluid-particle interaction

Validation (microscopic) closure laws
Models for Different Types
Multiphase Flow
Modeling Approach
• Empirical correlations

Increased complexity
• Lagrangian Track
- Track individual point particle
- Particles do not interact
• Algebraic slip model
– Dispersed phase in a continuous phase.
– Solve one momentum equation for the mixture.
• Two-fluids theory (multi-fluids)
– Eulerian models.
– Solve as many momentum equations as there are phases.
• Discrete element method
– Solve the trajectories of individual objects and their collisions, inside
a continuous phase.
• Fully resolved and coupled
Algebraic Slip Model
(ASM)
• Solves one set of momentum equations for the mass
averaged velocity and tracks volume fraction of each
fluid throughout domain.
• Assumes an empirically derived relation for the relative
velocity of the phases.
• For turbulent flows, single set of turbulence transport
equations solved.
• This approach works well for flow fields where both
phases generally flow in the same direction.
ASM Equations
• Solves one equation for continuity of the mixture:
  ( ui )
 0
t xi
• Solves for the transport of volume fraction of one phase:
2 um ,i 2
 0
t xi
• Solves one equation for the momentum of the mixture:

  P
( um , j )   mum ,i um , j  
t xi x j
 um ,i um , j  n

xi
 eff (
x j

xi
)  m g j  Fj 
xi
 k k k ,i k , j
 
k 1
u r
u r
ASM Equations
• Average density: m   1  1   2  2
 
 ρ 1 α 1 u1  ρ 2 α 2 u 2
• Mass weighted average velocity: um 
ρ1α1  ρ 2 α 2

• Velocity and density of each phase:  

u1 , u2, 1 ,  2

• Drift velocity: r  
u1 u1  um

• Effective viscosity:  eff   1 1   2 2

 Slip Velocity and Drag
• Uses an empirical correlation to calculate the slip velocity between phases.
  '
u urel a p

    u m
a ( g  (u m u m  ))
t
(  m   p )d p2
 'p 
• fdrag is the drag function. 18 f f drag

1  0.15 Re0.687 if Re 1000

f drag 
 0.0175 Re if Re 1000
Restrictions
• Applicable to low particle relaxation times

(  m   p )d p2
 'p f drag  < 0.001 - 0.01s.
18 f

• One continuous phase and one dispersed phase.

• No interaction inside dispersed phase.

• Volume fraction of discrete phase should be less than 10%

• One velocity field can be used to describe both phases.

– No countercurrent flow.

– No sedimentation.
Bubble Column Example - ASM

Unstable flow in a 3-D bubble column with rectangular cross section.

 Eulerian-Eulerian (Two Fluid) Model
• Solves momentum equations for each phase and additional volume
fraction equations.
• Appropriate for modeling fluidized beds, risers, pneumatic lines,
hoppers, standpipes, and particle-laden flows in which phases mix
or separate.
• Discrete phase volume fractions from 0 to ~ 60%.
• Several choices for drag laws. Appropriate drag laws can be chosen
for different processes.
• Several kinetic-theory based formulas for the granular stress in the
viscous regime.
• Frictional viscosity based formulation for the plastic regime stresses.
• Added mass and lift force.
Granular Flow Regimes

Elastic Regime Plastic Regime Viscous

Regime.
Stagnant Slow flow Rapid flow
Stress is strain Strain rate Strain rate
dependent independent
dependent
Elasticity Soil mechanics
Kinetic theory
Kinetic Theory of Granular Flow

Kinetic
KineticTransport
Transport
Collisional
CollisionalTransport
Transport
Granular Multiphase Model:
Description
• Application of the kinetic theory of granular flow
Jenkins and Savage (1983), Lun et al. (1984), Ding and
Gidaspow (1990).
• Collisional particle interaction follows Chapman-Enskog
approach for dense gases (Chapman and Cowling, 1970).
– Velocity fluctuation of solids is much smaller than their mean
velocity.
– Dissipation of fluctuating energy due to inelastic deformation.

– Dissipation also due to friction of particles with the fluid.

Granular Multiphase Model: Description
• Particle velocity is decomposed into a mean local velocity and a
superimposed fluctuating random velocity

• Analogous to the thermodynamic temperature of the gas, the

‘granular temperature’ is associated with this random fluctuation of
fluctuating velocity of the solid particles.

• The source of the particle fluctuations come from collision with

neighboring particles
2
s  ks
3
ks=Kinetic Energy due to solids velocity fluctuation per unit mass
Gas Molecules and Particle Differences
• Solid particles are a few orders of magnitude larger.

• Velocity fluctuations of solids are much smaller than their

mean velocity.
• The kinetic part of solids fluctuation is anisotropic.

• Velocity fluctuations of solids dissipates into heat rather fast

as a result of inter particle collision.
• Granular temperature is a byproduct of flow.
 Conservation Equations: Two-fluid Model

Continuity:

 q  q   q  q vq 0
t

Phase denoted by: q, s Density: 

Volume fraction:  Velocity: v

Momentum:

 q  q vq   q  q vq vq   qp   q   q  q g  K sq vs  vq  F
t

 s  s vs   s  s vs vs    sp  ps   s   s  s g  K qs vq  vs  F
t
Solids pressure
Forces
ps  s  s  s  2  s 1  ess  s g 0  s
2
  
F Fext  Flift  Fvm
Interphase momentum exchange
Stress-strain tensor   f
coefficient K qs K sq  s s External body
  
  2  
 s  s  s vs  vs T   s  s   s  vs I
3 
s force Lift force

Virtual mass force
Stress-Strain Tensor for Continuous Phase
 T 2  
   (v  (v ) )      )  (.v ) I
3 
Dilatational Viscosity

 v x   2 
 xx    2        (.v)
 x   3 

 v y v x 
 xy     
 x y 
 Solids Stress-Strain Tensor
  
 s  s  s vs  vs T   s  s    2  
 s  vs I
3 


Solids shear viscosity

 s  col   kin  fr

Frictional Viscosity

Collisional shear viscosity Kinetic viscosity: Syamlal

1
4   2  d   s  2 
 col   s  s d s g 0 1  ess  s   kin  s s s 1  1  ess 3ess  1 s g 0 
5    63  ess   5 

Bulk viscosity
1
4   2
s   s  s d s g 0 1  ess  s 
3   
 Solids Pressure
ps  s  s  s  2  s 1  ess  s g 0  s
2

Lun et al

Granular Temperature
3   
        v  
2  t
s s s s s s s
 Radial distribution
  
 
 ps I   s : vs   k  s  s   s function
 1

1
Coefficient of
restitution
 s 3
g 0 1  


 
  s ,max  


Collisional
Generation of energy dissipation of energy
by solids stress tensor
Diffusion
coefficient:
Gidaspow,
Syamlal et al
Interphase Momentum Exchange Coefficient
 s s f
K qs K sq 
s

Particulate relaxation time

sds2
s 
18 q C D Re s  q
f 
24v 2 r , s

Reynolds number
Drag function:  
 q d s v s  vq
Syamlal-O’Brien Re s 
 
2 q
 
4.8
C D  0.63  
 Re s 
 vr , s  Diameter
 
Terminal velocity
correlation for Viscosity
It is this drag function which solid phase
makes different models for
exchange coefficient
Kinetic Theory of Granular Flow
(KTGF)
• Granular Temperature Granular temperature is a flow dependent quantity
as against thermodynamic temperature
2
s  ks
3
ks=Kinetic Energy due to solids velocity fluctuation per unit mass

• Transport of solids fluctuating kinetic energy

1

2 4

3
1:Generation of energy by solid stress tensor
2:Diffusion of energy
3:Collisional dissipation of energy
4:Interphase energy exchange 27
(Chapman and Cowling,, The Mathmatical Theory of Non-
Models for Different Types
Multiphase Flow
Eulerian-Lagrange
Approach
• Fluid phase treated as a “Continuum”

• Dispersed phase is tracked in a Lagrangian way

• Newton’s equation of motion is solved for dispersed

phase
• Particle-particle collisions are included and modeled
through spring dash-pot model.

29
Discrete Element Modeling is an outgrowth of molecular
dynamics simulations used in computational statistical
physics.

However, the discrete element method was independently

developed by P. Cundall in the 1970’s.

30
Basic Idea of DEM Modeling

Approximate the collisional interactions between particles using

idealized force models that dissipate energy.

Integrate system equations of motion  Determine individual

particle positions and velocities.

Compute relevant transport quantities, bulk properties and

analyze evolving microstructure.

31
Basic Flow Chart
Initial positions, orientations and velocities

Update particle link-list(find new or broken contacts)

“F = ma”
“T = Ia”
Calculate the force and torque on each particle
LOOP

Integrate the equation of motion to calculate the new

positions, velocities and orientation

Accumulate statistics to calculate transport properties

Time increment: t = t + dt

32
 In contrast to the energy conservation of molecular systems-energy
dissipation is a critical characteristic of granular systems, and
consequently, it is necessary to employ realistic approximations to
model energy loss in colliding particles.

For this purpose, there are essentially two basic approaches.

‘Hard’ Particle Models

By considering particles to be infinitely stiff, “hard particle” models assume
instantaneous, binary collisions governed by a collision operator, which is a
function of particle properties (i.e., friction, normal and tangential restitution
coefficients) and the pre- and post collisional velocities and spins. Such an
approach is appropriate in collision-dominated systems, where continuous
and/or multiple contacts are not characteristic.

‘Soft’ Particle Models

The interaction is a function of an allowed overlap between colliding
particles that is intended to model the plastic deformation* at the
contact.
33
Elementary Hard Particle Model
• When particles collide, some of the kinetic energy is lost –
that is, a portion of the initial kinetic energy goes into
deforming the objects.
• Thus, a ball that is dropped and hits the ground will not
rebound to the same height from which is was released.
• As a first approach, this energy loss is modeled through a
“coefficient of restitution”, denoted by e.
e=1  No energy loss (perfectly elastic)
e=0  Complete energy loss (plastic)
0<e<1  A portion of the incident energy is loss

34
The coefficient of restitution is related to the velocities of
the two particles by the relation

VB V A VB
v B  v A VA
A B A B
e
vA  vB
Before After collision
collision

Consider a particle A of mass mA that has a velocity vA ,and

collides with a stationary particle B of mass m B.
What is the kinetic energy lost ΔKE during the impact?
It can be shown that …
m AmB
KE  v 2A (1  e 2 )
2(m A  mB )
35
 DEM (Soft Sphere
Gas Phase
Approach)

V=Volume of fluid cell

Solid Phase
Contact Drag
Force Force

Van-der
Waal’s Force

36
 Modeling of Contact between Two
Particles
Soft sphere approach (Spring dashpot model) (Popularly known as
‘Distinct element method’ DEM)

Dashpot
spring

slider

F Force (N)
k Spring constant [N/m]
Fn ,ab  kn n nab  n vn ,ab
ξ Displacement [m]
Ft ,ab  ktt  t vt ,ab Ft ,ab  Fn ,ab
η Damping coefficient [Ns/m]
Ft ,ab   Fn ,ab tab v Velocity [m/s]
μ Coefficient of friction [-]
Ref:- Cundall and Strack, Géotechnique, 37
Typical DEM Simulation for
Fluidized Bed

(a) (b) (c) 38

39
40
41
42
43
44
45
46
47
48
49
Analogy to Kinetic Theory of Gases

Velocity distribution Pair distribution

function function

Collisions are brief

and momentarily.
No interstitial fluid
effect.

Free streaming Collision

51
52