Lecture 5:

Particulate Flows
15.0 Release

Advanced Multiphase Course

1 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Outline
• Applications

• Characteristic of particulate flows

• Modeling approaches
• Euler-Granular Model
• Dense Discrete Phase Model (DDPM)
• Discrete Element Method (DEM)
•
• Examples

2 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Applications from wide range of industries

• Oil and gas

• Energy
• Chemical processing
• Food processing
• Pharmaceutical
• Automotive
• Metallurgy and mining
• Agriculture
• Others

3 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Characteristic of particulate flows
Dilute vs. Dense flows and Fluid-Particle coupling
Inter-particle spacing
100 10 1

Dilute disperse Dense disperse

One-Way Coupling Two-Way Coupling Four-Way Coupling

10-8 10-6 10-4 10-1

Volume fraction 
4 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Characteristic of particulate flows
Particle-Particle interactions

Collision
dominated flow

Intermediate Flows

Dense Flows
Dilute Flows

Collision free Frictional

flow dominated flow

5 © 2013 ANSYS, Inc. May 2, 2014

Concentration
ANSYS Confidential
 Characteristic of particulate flows
Regime map for granular flows

Quasi-static regime Intermediate regime Inertial regime

Rapid flow
Slow flow Strain rate dependent
𝜎 ∝ 𝛾𝑛, 0 < 𝑛 < 2
Strain rate independent
𝜎 ∝ 𝛾2
𝜎≠𝑓 𝛾

Ref: V. Vidyapati and S. Subramaniam, Granular rheology and phase transition: DEM simulations and
order-parameter based constitutive model, Chem. Eng. Sci., 72 (2012) 20-34.

6 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Overview of modeling approaches
Model Numerical Approach Particle Fluid Particle Particle Particle Size Distribution (PSD)
Interaction Interaction

DPM Fluid – Eulerian Empirical models for Neglected Easy to include PSD because of
Particles – Lagrangian sub-grid particles Lagrangian description

DDPM - KTGF Fluid – Eulerian Empirical models for Approximate P-P Easy to include PSD because of
Particles – Lagrangian sub-grid particles interactions determined by Lagrangian description
granular models
DDPM - DEM Fluid – Eulerian Empirical models for Accurate determination of Easy to include PSD because of
Particles – Lagrangian sub-grid particles P-P interactions. Lagrangian description

Macroscopic Fluid – Eulerian Interactions are Accurate determination of Easy to include PSD; if particles
Particle Model Particles – Lagrangian determined as part of P-P interactions. become smaller than the mesh, uses
solution; particles an empiricial model
span many fluid cells
Euler - Granular Fluid – Eulerian Empirical models for P-P interactions modeled Different phases to account for a PSD;
Model Particles – Eulerian sub-grid particles by fluid properties, such as when size change operations happen
granular pressure, use population balance models
viscosity, drag, etc.

7 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Overview of modeling approaches
• Euler-Granular Model
• Treats continuous fluid (primary phase) as well as dispersed
solids (secondary phase) as interpenetrating continua
• Effects of Particle-Particle interactions are accounted based on
Kinetic Theory of Granular Flow (KTGF)

• Applicable from dilute to dense particulate flows. Particle size

distribution can also be accounted by assigning a separate
secondary phase for each particle diameter

• Compatible with species transport, homogeneous and

heterogeneous reactions Fluidized bed simulation: Contours of
volume fraction of particles

8 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Overview of modeling approaches
DDPM-DEM:
Particles colored by volume fraction

• Dense Discrete Phase Model (DDPM)

• Treats secondary phase solids as discrete particles dispersed in
continuous fluid
• Particle-Particle collisions are either modeled (KTGF based approach) or
explicitly resolved (DEM based approach)
• Applicable from dilute to dense particulate flows with wide particle size
distribution
• Compatible with species transport, homogeneous and heterogeneous
reactions

• Discrete Element Method (DEM)

DEM

• Soft-sphere contact model to explicitly resolve particle-particle collisions

• Efficiently handles dense and near packing limit particulate flows

9 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular Model

15.0 Release

10 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular model: Brief description
• Based on kinetic theory of granular flow
• Analogous to kinetic theory of gases

• Accounts for the effect of collisional particle-particle interactions

• Fluctuating velocity of particles
• Granular temperature as a measure of kinetic energy contained in the fluctuating velocity of
particles
• Stresses in secondary solids phase
• Constitutive models for solids stress : Solids pressure, solids shear and bulk viscosity
• Constitutive models depend on solids volume fraction, coefficient of normal restitution and
granular temperature.

• Solves set of conservation equations (continuity, momentum and energy) for primary and
secondary phases along with equation for granular temperature.

11 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Kinetic Theory of Granular Flow (KTGF)
• Particle-particle interactions are fundamentally non-conservative unlike molecular
interactions
• Dissipation of fluctuating energy due to inelastic deformation
• Dissipation also due to friction of particles with the fluid

Kinetic Transport Collisional Transport

• Two principal mechanisms of transport :

• Kinetic transport during free flight between collision
• Collisional transport during collisions
12 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Euler-Granular model: Averaged Equations
• Particle phase continuity equation
 
( s  s )    ( s  s u s )  m
 fs Mass transfer
t

• Particle phase momentum equation

    n   
( s  s u s )    ( s  s u s u s )   s p f     s   ( R fs  m
 fs u fs )  Fs
t s 1

Fluid pressure Solid stress tensor Phase interaction term

Solids stress tensor

  
S  1  u  (u )T  Strain rate
   P I  2  S   (  2  )  u I where,
2 s s 
s s s s s s 3 s s
P Solids Pressure
s
 , Solids bulk and shear viscosity
s s
13 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Granular Temperature (GT)
• Granular temperature (θ𝑠 ) is defined as average of the square of the
fluctuating particle velocities.
1
θ𝑠 = 𝐶 ∙𝐶 where 𝐶𝑠 = 𝑢𝑠 − 𝑉 𝑠 fluctuating velocity of the particle
3 𝑠 𝑠
𝑉 𝑠 = 𝑢𝑠 average particle velocity

• Granular temperature is obtained by solving its transport equation

• Algebraic form (convection and diffusion terms are neglected)
• Partial Differential Equation – full equation solved
Production
3    s s s  
     s sus s   τs : us
2 t  Exchange
    s s    s  lm  fs
Diffusion Dissipation
14 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Solids Pressure
• Pressure exerted on the containing wall due to the presence of particles
• Measure of the momentum transfer due to streaming motion of the particles and collision
• The granular flow becomes incompressible when the volume fraction of solids reaches its
maximum. The solid pressure de-couples from the volume fraction

• ANSYS FLUENT implementation for solids pressure:

Lun et al. Ps   s  s s  2  s s (1  ess ) s2 g os
Kinetic collisional contribution
𝑷𝒔
Syamlal et al. Ps  2  s s (1  ess ) s2 g os
𝝆𝒔 𝜽𝒔
• Incudes only collisional contribution
1
Ma and Ahmadi Ps   s  s s [(1  4 s g os )  (1  ess )(1  ess  2 fr )]
2
• Includes dependence on frictional viscosity
15 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Radial Distribution Function
• The radial distribution function g o ( s ) is a correction factor that modifies the
probability of collision close to packing.

• ANSYS FLUENT implementation

• Lun et al.
• Syamlal-O’Brian
• Ma and Ahmadi
• Arastoopour

16 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Granular Viscosity
• Shear viscosity due to kinetic motion and collisional interaction of particles
 s   s ,kin   s ,coll

• Collisional contribution 1
8   2
• Lun et al. for all models  s ,coll   s2  s d s g os  s 
5  
𝝁𝒔

• Kinetic contribution 𝒅𝒔 𝝆𝒔 𝜽𝒔
𝟏/𝟐

1
 d  (  ) 2  8 
• Syamlal et al. s,kin  s s s s
12(2   ) 1   (3  2) s gos 
 5 

1
5d s  s ( s ) 2
 8 
• Gidaspow et al.  s ,kin 
96g os 1  5 g os s 

• Bulk viscosity accounts for particle resistance

to expansion and compression 8  
1
2
s   s2  s d s g os  s 
• Lun et al. 3  
17 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Frictional Stress Modeling
• Frictional stresses are negligible below a certain solid volume fraction
(friction packing limit) but become important as the solids volume fraction
approaches the packing limit. The stresses are determined from soil mechanics
(Schaeffer, 1987)

• Packing limit is defined as the maximum achievable s,max  0.63

volume fraction of the granular phase
(0.63 for packed spheres)
• Particles do not collide much but rather rub against each other.
  
• Frictional stress in Newtonian form:  f  Pf I   f (us  (us )T )
Frictional Frictional
pressure viscosity

• When solids volume fraction exceeds friction packing limit frictional stress is added to
solids stress
 s ,total   s ,ktgf   s , f
18 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Frictional Stress Modeling
• ANSYS Fluent implementation
• Frictional pressure • Frictional viscosity
Ps  Ps,kin  Pf s  s,kin  s,coll   f

• Johnson and Jackson (1987) • Coulomb law (Schaeffer model)

Pf sin 
 Fr
 s   s,min n f 
 s,max   s p
Pf
2 I2D
Fr  0.1 s , n  2 and p  5 • Johnson and Jackson (1987)
• Syamlal et al. (1993)  f  Pf sin 


Pf  A s  s,min n 

19 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Interphase forces: Drag force
• Formulation is based on forces on a single particle corrected for effects such as
concentration, clustering, particle shape and mass transfer effects.
• Use the one that correctly predicts the terminal velocity in dilute flow
• In bubbling beds ensure that the minimum fluidized velocity is correct
• It depends strongly on the particle diameter: correct diameter for non-spherical particles
and/or to include clustering effects

20 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Drag force for particulate flows
• Wen and Yu (1966)
• Extension of Schiller-Naumann drag law to high particle concentration

• Gidaspow (1992)
• Uses Wen and Yu for dilute concentrations. Ergun law for high concentration
• Recommended for dense fluidized beds

• Huilin-Gidaspow (2003)
• Similar to Gidaspow but with a better blending function when moving from the dense packing limit
to the dilute flow limit

• Syamlal and O’Brien (1989)

• Based on correlation for minimum fluidization velocity
• Parametric Syamlal-O’Brien model adjusts parameters based on the fluid flow properties and the
expected minimum fluidization velocity

• Gibilaro (1995) See appendix for further

• Useful for circulating fluidized beds details on drag force…
21 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Parametric Syamlal O’Brian Drag Model
• Syamlal O’Brian drag model is used in granular flows to compute the drag forces
between fluid-solid phases.
2
 
 α g α s ρ g vslip  Re  4.8 
 
vr 0.5 A0.06 Re  0.06 Re 2  0.12 Re 2 B  A  A2 
 CD  
3 s CD   0.63   '  
k gs  0.8 1.28 for  0.85
4  ds  vr3  Re  A g4.41, B
 g g
   vr  
2.65 for  0.85
 0.8 g g

• The drag law assumes values of 0.8 and 2.65 for calculation of factor B in equations
tend to under/over-predict the bed expansion
• In order to overcome this drawback, a simple yet effective has been developed to
adjust the above two coefficients according to fluid flow properties and the expected
minimum fluidization velocity

22 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Comparison of drag force models
𝜶𝒔 = 𝟎. 𝟏 𝜶𝒔 = 𝟎. 𝟐 𝜶𝒔 = 𝟎. 𝟒
𝜶𝒔 = 𝟎.2

𝜶𝒔 = 𝟎.3

Here F (dimensionless drag force) 𝐹=

𝐾𝑠𝑓
is calculated as 18 1 − 𝛼𝑠 2 𝛼𝑠 𝜇𝑓 𝑑2

23 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Interphase heat transfer

Qpq  hpq Tp  Tq 

Non-Granular Granular

6 q q p Nu p 6 f  f  s Nu s
h pq  hsf 
d p2 d s2

 
Nu s  7  10 f  5 2f 1  0.7 Re 0s .2 Pr1f / 3 
Nu p  2.0  0.6 Re
  1.2 Re
1/ 2 1/ 3
Pr
p q
 1.33  2.4 f 2
f
0.7
s Pr1f / 3

Ranz-Marshall Gunn
0.35   f  1, Res  105

24 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular Model Setup

15.0 Release

25 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Secondary phase settings

• Enable granular option

• Granular temperature model
• Diameter
• Granular viscosity
• Granular bulk viscosity
• Solid pressure with radial distribution
function
• Fictional viscosity, Frictional pressure, angle
of internal friction and friction packing limit

26 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Phase interaction models

Normal coefficient
of restitution

Interphase Homogeneous and

heat transfer heterogeneous reactions

Drag
Models

• Evaluate relative importance of non-drag forces for a given fluid-particle system

before enabling those models.

27 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Recommended settings

Recommended settings for

Not relevant if solids
the first step before
volume fraction
trying other variations remains below 0.55

28 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Solver settings
Transient Steady state
• Default solver settings work well for most
of the problems.
• Operating density:
• If lighter phase density
• Constant
• Set operating density to lighter phase
density
• Compressible (ideal gas)
• Set operating density to zero
• For isothermal calculations, deselect
energy equation from solution controls >
equations
• Start with conservative solution control
settings
29 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Solution strategies
• Start solution calculation with
• Conservative solution control settings
• First order discretization schemes
• Gradually increase phase velocities over certain flow time calculations
in the beginning
• Once flow field is established, gradually increase under relaxation
factors and switch to higher order discretization schemes
• Consider using adaptive time step option to speed up the simulations

30 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Dense Discrete Phase Model (DDPM) and
Discrete Element Method (DEM)

15.0 Release

31 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Dense Discrete Phase model (DDPM)
• A general framework in which the continuous phase is solved
on an Eulerian grid and the particulate phase in a Lagrangian
frame.
• Extends application range of Discrete Phase Model (DPM)
from dilute to dense particulate flows.
• Accounts
• Effect of volume fraction of particle phase
• Particle-Particle interactions
• Fluid-Particle coupling (mass, momentum and
energy exchange)
• Wide particle size distribution Fluidized bed simulation using DDPM-DEM.
Particles colored by volume fraction of particles.
• Not limited to gas/particle flows and can be applied to
gas/liquid, liquid/gas/particle problem

32 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DPM solution procedure


Equation of motion d (m p u p )     
 Fdrag  F pressure  Fvirtual mass  F gravitation  Fother
for particles dt

Navier-Stokes equations
for continuous phase

33 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM solution procedure

Equation of motion
for particles

Averaging over discrete phase in a cell From

Particle-Particle
interactions

Mass flux
Volume fraction

Equations
for continuous phase

34 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM: Particle-Particle interactions
• Modeled / Approximate algorithms
• Based on Kinetic Theory of Granular Flow (KTGF)
• Interaction force model based on solids stress
tensor given by kinetic theory of granular flow.
• DDPM-KTGF approach
KTGF based collision
• Other models
• Sommerfeld, O’Rourke, Oesterle, etc.

• Explicitly resolved
• using Discrete Element Method (DEM)
• Soft-sphere contact model to explicitly resolve
particle-particle collisions.
• DDPM-DEM approach DEM based collision

35 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-KTGF approach
• Suitable for dilute to moderately dense particulate
flows
• Faster computations due to modeling of particle
interaction effects KTGF based collision
• Compatible with DPM’s char combustion and multiple
surface reaction framework
• Two ways to compute granular temperature for solids
stress calculation
• Algebraic formulation
• Based on particle statistics
• Predicts particle-particle collisions without
full DEM.

36 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-DEM approach
• Efficiently handles dense and near packing limit
particulate flows

• Higher accuracy at the cost of slower

computations
DEM based collision
• Compatible with DPM’s char combustion and
multiple surface reaction framework

• Pure DEM calculations also possible by

decoupling fluid-particle motion Pure DEM
simulation

37 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Discrete Element Method (DEM)
• DEM implementation based on work of
Cundall and Strack (1979)
Forces acting on a particle
• Soft Sphere Approach
• Contact forces computed from deformation
• Overlap of ideal spheres used as the measure 
u p,1
for deformation. 
u p ,2
• Newtons 2nd law integrated in time mp,1  
• Allows for N-body interaction Ffriction F m p,2
• Rigidity of materials determines time scale for
integration

d (mp,1up,1 )      
 Fdrag  Fpressure  Fvirtual mass  Fgravitation  F  Ffriction
dt

38 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Contact forces
Particle 2:
• Implemented Force Laws Particle 1: mass m2
• Spring mass m1 position x2
• Spring-Dashpot position x1 velocity v2
• Friction velocity v1
r2
• Spring Force (Linear repulsive force)
r1
𝐹1 = 𝐾𝛿𝑒12 F2
Where 𝐊 𝐒𝐩𝐫𝐢𝐧𝐠 𝐜𝐨𝐧𝐬𝐭𝐚𝐧𝐭 𝐍 𝐦
𝛅<𝟎 𝐎𝐯𝐞𝐫𝐥𝐚𝐩 𝐦 F1
𝐞𝟏𝟐 𝐔𝐧𝐢𝐭 𝐯𝐞𝐜𝐭𝐨𝐫 𝐟𝐫𝐨𝐦 𝐩𝐚𝐫𝐭𝐢𝐜𝐥𝐞 𝟏 𝐭𝐨 𝟐

overlap*
𝐅𝟐 = −𝐅𝟏 * not to scale, greatly exaggerated

39 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Contact forces
• Dashpot (Linear dissipative force)
Where  is coefficient of restitution
Loss factor f loss   2  ln 2 
0  1
m1 m 2
Reduced mass m12  Perfectly elastic collision : η  1  γ  0
m1  m 2
m12
Collision time scale t coll  f loss Particle time step must be smaller than t coll
K
m ln 
Damping coefficient   2 12
t coll

• Combined Spring-Dashpot
𝐹1 = 𝐾𝛿 + 𝛾 𝑣12 ∙ 𝑒12 𝑒12 ,  0 since ln  0
𝐹2 = −𝐹1

40 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Contact forces
The friction coefficient is a function of the
• Coulomb Friction law: relative tangential velocity magnitude.
𝐹𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 = 𝜇𝐹𝑛𝑜𝑟𝑚𝑎𝑙
where 𝑭𝒏𝒐𝒓𝒎𝒂𝒍 is the force normal to the surface.
Direction of friction force opposite to relative tangential
motion.

• May or may not inhibit relative tangential

motion depending on
• size of relative tangential momentum
• size of other tangential forces
(e.g., tangential components from
gravity, drag, etc.)

41 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 What are Parcels in the Context of DEM?
• For real applications, cannot afford to track all particles individually
• Put several particles of same properties into one parcel
• Track this parcel by a representative particle

DEM can make use of the same concept, but:

• The mass used in collisions is that of the entire parcel.
• Radius of parcel is obtained from mass of a parcel and
density of parcel that is considered to be same as
particle density.
• Note: This way, close packing of parcels gives parcel 1 parcel 2
the
correct volume fraction for sphere packing.
42 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 DDPM and DEM Setup

15.0 Release

43 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-KTGF setup
Step 1: Enable DDPM Model
Step 2: Enable
Granular model in
secondary phase
panel to account KTGF
based interactions

Single discrete phase Recommended

even with wide PSD settings

Step 3: Create
Once DDPM model is enabled, make sure to specify
- Correct fluid-particle drag force
injection and
- Particle-Particle interaction option (KTGF or DEM) assign phase to it.
Either use KTGF or DEM but not both!

44 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-DEM setup
Step 1: Enable DDPM Model Step 3: Create injection and make
sure to assign phase and collision
Step 2: Enable DEM in DPM panel partner

Select a collision partner and define how particles of this

kind will collide: Among themselves and with others

Step 4: Specify DEM

collision properties

Particle interaction defined

45 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential between pairs of collision partners.
 Enabling drag and non-drag forces

Injection Panel
DPM Panel

Non-Drag
Forces
Drag
Force Models

• Evaluate relative importance of non-drag forces for a given fluid-particle

system before enabling those models.
46 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Particle time step
• DDPM-KTGF simulations
 Use “Track with Fluid Time Step” option for calculation efficiency

• DDPM-DEM simulations
 Specify particle time step smaller than fluid time step for
DEM based collision calculations.

1. Deselect DPM default

2. Choose desired time step.

47 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Particle Injection
• Volume Injection (Patching of particles as an initial condition)
• Irrespective of method used to patch particles, avoid excessive initial overlap of particles
specifically when using DDPM-DEM or DEM.

Use “file” injection option to injection particles

Use custom python script to in first time step
generate injection file

Specify inputs in panel

Panel appears after executing

python script in Workbench
48 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Particle Injection
• Continuous injection from boundary zones or user generated surfaces

Injection Panel
Suggestions for DEM based simulations
• Use Constant-mass or Constant-diameter of a parcel
option to ensure parcel size < smallest cell size in
mesh.
• To avoid excessive overlap of particles at injection
surface
• Consider staggering of particles using Fluent
TUI: define/models/dpm/options/particle-staggering

• Use UDF to randomize or rearrange particle

locations on surface .
49 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Particle Injection
With DEM, injection setup suggests a collision partner name

• The first suggested name

is based on the material.

• Important: Material and

collision partner are
independent!
• E.g., changing the material has
no influence on the choice of
the collision partner.

• Choose “none” to make

particles ignore DEM collisions

50 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Wall boundary conditions
With DEM enabled, boundary conditions suggest a name for the
collision partner, if the DPM boundary condition type is “reflect”.

E.g., for walls, it will suggest a name

using the wall material name Only apply to
none-DEM particles.
E.g., massless particles.

51 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM simulation best practices
• Maintain parcel size sufficiently smaller than cell size in
mesh
• Convergence issues will arise if fluid volume fraction
• becomes zero due to Volume fraction Volume fraction
standard average node based average
• Parcel size bigger than cell size
• By default, Fluent attributes all mass of the parcel to the CFD
mesh cell where the center of the parcel is.
• Suggestion: Enable “Node Based Averaging” in DPM panel -
Numerics tab.

• Too soft parcels (low K in force law) squeezing in a cell

• Many may squeeze together into a CFD cell and leave no room
for the fluid.
• Suggestion: Make force laws rigid enough to avoid this.

52 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM simulation best practices
• Maintain parcel size sufficiently smaller than cell size in
mesh
• Additional suggestions
• Geometry and mesh
• Simplify as much as possible:
• Remove small geometrical features
• Close any gaps that are irrelevant
• Mesh size needs to be small enough to capture
remaining
• features and gaps

• Also consider smaller parcel size

• Typically you want the largest possible parcel size
• CFD mesh must be larger than largest parcel
• This can be a difficult balance!

53 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM simulation recommended settings
Transient:
• PC-SIMPLE
Steady state: Swap URF’s for U and P
• Multiphase coupled solver URF_U = 0.3
• PC-SIMPLE URF_P = 0.7
Swap URF’s for U and P • Node based or LSQ gradients
URF_U = 0.3 • Node based averaging.
URF_P = 0.7 • Second order interpolation or QUICK
• Node based or LSQ gradients DDPM-KTGF:
• Node based averaging. • Algebraic granular temperature or particle
statistics based.
• One particle time step per fluid time step for
efficiency.

DDPM-DEM:
• Multiple particle time steps per fluid time step
for collision calculations.
54 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 DEM based simulation best practices
• Estimation of particle time step
m12
• Collision time scale is the time period measured from t coll  f loss
initial contact to complete detachment. K

• Select particle time step from this range t p  t coll 5 or t coll 5 0

• while considering factors like
• Number of parcels to track
• Total flow time calculations to perform
• Availability of computing resources

• Some suggestions to increase t coll and t p

• Consider bigger parcel
m12  and K  t coll  hence tp 
• Consider softer parcels (low K in force law)
• Without any better idea, start experimenting with K = 100 ... 1000 N/m

55 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Applications

15.0 Release

56 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular: Circulating Fluidized Bed
• A fully 3D circulating fluidized bed is modeled.
• 74,000 cell hybrid mesh
• Gas/Solids dilute flow (average solids volume fraction around 7%)

Outlet

Riser
Cyclone
Separator
Downcomer
Inlet

57 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular: Circulating fluidized bed
• Solid volume fraction contours

58 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular: Stirred tank with solid
suspension
• Turbulence enhancement example water and glass beads in a stirred closed vessel

Local Volume Fraction

Normalized bed height

of Solids

Normalized Solids Concentration

59 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Euler-Granular: Stirred tank with solid
suspension
• Example: Unbaffled vessel equipped with four A310 impellers

Vertical tank dimension

The local volume fraction
of solids

60 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular: Bubbling fluidized bed

Porosity (2.5 cm from center) Porosity (10 cm from center)

Y= 10 cm

Porosity (17 cm from center) Solids Y Velocity (at 10 cm)

2 cm, 10 cm, 17 cm

61 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-DEM: Segregation
• Bed dimensions: 150 X 15 X 700 mm
• 50-50 mixture of 1.5 mm and 2.5 mm particles
• Particle density: 2523 kg/m3
• Initial bed height: 7.5 cm
• Minimum fluidization velocities
•1.5 mm particles – 0.8 m/s
•2.5 mm particles – 1.25 m/s
• Segregation studied with a superficial velocity of 1.1 m/s

62 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-DEM: Filtration modeling

Inlet Outlet

Filter: Allows particles below a threshold to pass through,

0.002m in this case. Filter represented by a internal boundary
condition.

63 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DDPM-DEM/KTGF: NETL Fluidization challenge
problem
A blind challenge problem on modeling
a bubbling fluidized bed of FCC particles
with a Particle Size Distribution
Submitted results using the Euler-
Granular & DDPM + KTGF model
We present results here with the DDPM
+ DEM model with the R14
enhancements.
Able to simulate 16s of flow time in a
day on 12 processors with a complete
PSD
12% fines 3% fines
64 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 DEM: Flow in a rotating drum

65 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Flow through a hopper
Flow rate of particles in a hopper
independent of material head
• Depends on appropriate friction models for
particle-particle and particle-wall interaction
• Rate comparable to experimental data

66 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Appendix

15.0 Release

67 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• Drag models are based forces in single particle motion, corrected for effects

such as:

• Nearness of other particles in a cloud

• Particle shape

68 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• Schiller and Naumann (1935) is the oldest model drag model. The gas–solid moment exchange
coefficient Kgs is defined as following:

Where

• The Schiller–Naumann drag model is the default model in FLUENT and has been established by
investigations of the flow field around a single sphere

69 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• The Syamlal and O'Brien(1989) drag model is appropriate when the solid shear stresses are
defined according to Syamlal et al. (1993). The drag function f and gas solid exchange coefficients
are defined as:

Where

• This model is based on measurements of terminal velocities of particles in fluidized or settling

beds, with correlations that are a function of the volume fraction and relative Reynolds number

70 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• The Wen and Yu (1966) drag model is recommended for diluted systems:

(a) experiment of Kuipers , (b) simulation using the interphase

drag coefficient of Wen and Yu, and (c) simulation using the
interphase drag coefficient of Syamlal et al.

and

• The Wen and Yu drag model yields better agreement with experimental data of Kuipers for both
bubble shape and size . The drag model of Syamlal under predicts the bubble size and produces
bubble that is more circular in shape than in observed in experiments
71 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Gas–solid exchange coefficient
• The Gidaspow (1994) drag model is a combination of the Wen and Yu (1966) model and the Ergun
(1952) equation. For very low particle concentrations (αg >0.8), kgs is defined as following:

and

72 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• For high particle concentrations αg ≤0.8, the momentum exchange coefficient is calculated by the
Ergun (1952) equation, which is based on experimental data for the pressure drop over fixed,
dense beds of mono disperse particles:

and

73 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• The Huilin-Gidaspow model is also a combination of the Wen and Yu model and the Ergun
equation). However, the two correlations, Wen and Yu expression with Ergun equation, are
‘stitched’ together using a generalised function

Where

74 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Gas–solid exchange coefficient
• Gibilaro et al. (1985) modified the Ergun equation and proposed an alternative pressure drop
equation on the basis of theoretical considerations. The equation is a combination of two
equations, one for the laminar regime and the other for fully turbulent flow

• The Gibilaro drag model gives an accurate prediction of bed expansion characteristics for both the
laminar and turbulent regimes
• Particulate bed expansion (Geldart group A) is more adequately described by Gibilaro drag model

75 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Inter-phase (Solid-Solid) drag

• Solid-solid momentum interaction

-Drag function derived from kinetic theory (Syamlal et al, 1993)

 2
3(1  elm )(  Clm ) l  l m  m ( d l  d m ) 2 g olm
2 8  
K lm  | ul  u m |
2 (  l d   m d )
l
3 3
m

76 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Euler-Granular: Packed Bed simulation

• Packed bed of solids:

Step 1: Phases > packed-solids > Edit > Packed bed
Enable Granular and then Packed Bed
option Inlet
Step 2: In cell zone conditions set packed-
solids velocity components to zero
Step 2
• Advantages: Step 1
 This is the best PHYSICAL representation of
packed beds (better than porous media).
 Access to homogeneous and
heterogeneous reactions framework.

77 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM Limitations
DEM does not work with periodic boundaries.
DEM does not work well with sliding boundaries.
DEM particles do not rotate.
DEM particles do not transfer heat during contact.
• Heat exchange works with surrounding fluid as in DPM.

78 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 DEM: Cartesian Collision Mesh
Problem: For N parcels, direct force evaluation requires inspecting O(N2) pairs. Too costly.
Approach: Bin parcels into a suitable cartesian mesh and inspect pair collisions for neighbor
mesh cells only.
Edge length is multiple of biggest parcel diameter.

Example: Force evaluation

for only with , as
others are a priori known
to be out of reach.

79 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Using DEM: Miscellaneous
Automatically adapts the mesh width of the cartesian
collision mesh to the largest parcel diameter
… times the edge scale factor

Alternatively, one may specify a fixed collision mesh width.

Limit the maximum particle velocity to a physically plausible

range.
Attention: Requires implicit tracking scheme (cf. Numerics
Tab)

80 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Flow of solids in a riser

16 m tall and ID of 0.35 m

FCC particles are of 75 microns diameter
At “steady state” about 1.5 million parcels are handled
Approximately 1s of flow time simulated in a day in a
single processor
Each parcel represents 20000 particles

81 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Flow of solids in a riser

82 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Proppant Transport

Volume fraction of proppant

Velocity of proppant showing motion only at the top

83 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Structure of the bed for various gas velocities

Particles colored by VOF of solids Pressure at the inlet

84 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 The fluidization curve

85 © 2013 ANSYS, Inc. May 2, 2014

0.26 < Theoretical Umf < 0.37
ANSYS Confidential
 Validation for FCC fluidization

86 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Cloud Height Prediction for Solid Suspension in a
Stirred Tank
• Material Properties
– Liquid Density: 1000 kg/m3
– Liquid Viscosity: 0.001 Pa-s
– Solid Density: 2630 kg/m3
– Particle Diameter: 180 micron

• Operating Conditions
– Solid Concentration: 10% wt and 15% wt
– Agitation Rate: 150 RPM to 450 RPM in the
steps of 50 RPM

Experimental Details from BHR Group, UK

87 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Turbulent Dispersion Force (TDF)
The interphase exchange term is given by

    
K fs (v f  vs )  K fs (U f  U s )  K fs dr
Instantaneous Slip Average Slip
• where the drift velocity vdr results from turbulent fluctuations in the volume fraction (based
on gradient eddy diffusion hypothesis by Simonin)

  Df Ds 
 dr  

 f   s 

  fs f  fs s 
• Where
• Df = Ds = Dt,fs : Turbulent diffusivities
•  fs : Dispersion Prandtl number = 0.75

88 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Comparison with Experimental Data

Single Impeller 400 RPM 10% w/w solid loading

The double-headed arrows marked “EXP” represents the location of experimental observation with
arrow thickness representing the observed 8% spread in the experimental data.
89 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Comparison with Experimental Data

Dual Impeller 150 RPM 15% w/w solid loading

The double-headed arrows marked “EXP” represents the location of experimental observation with
arrow thickness representing the observed 8% spread in the experimental data.
90 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Complete Data Set
Plot of Normalized Cloud Height against Normalized Impeller Speed

10% (w/w) loading 15% (w/w) loading

N : Agitation rate in rpm

Njs : Experimentally observed just suspension speed in rpm
Hc : Measured cloud height in CFD in meter
H : Reactor height in meter
91 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential
 Dense Cyclone: Multiphase Pressure Drop
• Case settings:
• Block structured hex mesh: 72K cells
0.2m
• Gas velocity: 20 m/s
• Particle velocity: 3 m/s

Pressure drop (Pa)

• Multiphase Model: DDPM
• DEM Settings:
– Spring constant: 100 N/m
– Coefficient of restitution: 0.3
– Friction coefficient: 0.3
• Drag: Gidaspow
Solid loading (kg per kg of air)
• Turbulence: Dispersed RSM

• Pressure drop matches reasonably.

Geometry and Mesh
• Mass weighted average of (static
pressure + phase-1 dynamic pressure)

92 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential

 Particle Tracks Colored by Axial Velocity

(a) (b) (c) (d) (e)

Solid Loading (kg per kg of air) (a) 0.5 (b) 1.0 (c) 1.5 (d) 2.0 (e) 2.5

93 © 2013 ANSYS, Inc. May 2, 2014 ANSYS Confidential