School of Process, Environmental

& Materials Engineering

Ali Hassanpour

Institute of Particle Science and Engineering, School of Process, Environmental and Materials Engineering, University of Leeds

Application of Discrete Element

Modelling for the Development of Particulate Processes:

linking materials properties to performance

1

Modelling Approach

(discrete/continuum)

Sensitivity

Analysis

Insufficient

materials/not

easily accessible

Parameters cannot

be measured

experimentally

Single Particle

Properties

Virtual

experiments, minimising trial and error

Bulk Behaviour

Characterisation of the bulk behaviour based on single particle properties

is of strategic importance in many processes involving particulate solids:

e.g. transportation, filling, mixing, compaction, milling and granulation.

Linking materials properties (single

particles) to performance (bulk

behaviour)

2

The term Discrete Element Method (DEM) is referred to a family of

numerical methods for computing the motion of a large number of

particles based on Newtonian laws of motion (Cundall and Strack,

1979).

MI

Fma

Solving Newtonian

equations of motion

particles

Solving Newtonian

equations of motion for a

large number of particles

Discrete Element Method (DEM)

3

MI

Fma

Fn

Fs

Normal stiffness

Normal force

Kn Fnd

Dashpot Ks

Ftd

Slider

Tangential force

mf

w Tangential stiffness

Cundall, P.A. and Strack, O.D.L.; Geotechnique (1979).

Modelling of Bulk Behaviour using

Distinct Element Method (DEM)

4

Effect of single particle properties and process variables on bulk

behaviour in particular processes needs to be understood......

sensitivity analysis

In a number of applications there is insufficient material for testing

or material is not easily accessible, e.g. pharmaceutical and

nuclear industries

Some parameters can not be measured or quantified in the -

experiments, e.g. internal particle flow and stresses

Scale-up: moving from lab scale to pilot plant and industrial scales

requires extensive trial and error….

Modelling is a mean to interpret experimental results

How DEM Modelling Can be Useful..

5

Both beads free-flowing

6

Analysis of Segregation of Mixtures: Sensitivity analysis

How to avoid segregation of light fine particles from dense course particles?

Dense Coarse Beads Cohesive

Surface Energy (G) = 0.5 J/m2

Understanding the cause of segregation during heap formation

7

Segregation during Heap formation • Interpretation of Experimental Results

• Sensitivity Analysis

Initial packing of sample: a randomly mixed system

Colours represent volume (related to size) of particles

5 mm

4 mm

3mm

2 mm

8

DEM Modelling of Segregation during Heap

Formation: effect of cohesion on

segregation

Colours represent average size of particles in each bin

5 mm

4 mm

3mm

2 mm

Fine particles are cohesive

All particles are

cohesive

All particles are free

flowing

9

DEM Modelling of Segregation during Heap Formation: effect of cohesion on segregation

10

Formation of Seeded Granules in

Cyclomix High Shear Mixer: Interpretation of Experimental Results

Seeded granules occur in Cyclomix high shear mixer under certain

operating conditions

11

DEM Simulation of Granulation in

Cyclomix

Seeded granules are quickly formed in the high shear region (middle part) and

break as soon as they approach the top part 12

DEM Simulation of Granulation in

Cyclomix: formation of seeded

granules

Modelling the agglomerate and compact breakage

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.02 0.04 0.06 0.08 0.1

Displacement (mm)

Pre

ssu

re (

kP

a)

Pre-consolidaition

pressure= 6 KPa

Load

Stainless

Steel

Stainless

Steel

Predicting Breakage of Granules,

Compacts

13

Particle motion analysis in a

paddle mixer

14

DEM of Mixing of Particulate Solids: Minimising Trial and Errors; Insufficient Available

Materials; Parameters Difficult to Measure

6l paddle mixer was seeded with a positron

charged tracer particle.

The mixer was run at various condition over

a period of time (usually 20-30 min).

Position of the tracer is continuously

recorded against time.

Bulk flow properties per trial is analysed

from the temporal velocity and occurrence

frequency of the tracer.

Position of particle

generation

Impellers

Experimental Measurement using

(Position Emission Particle Tracing (PEPT)

15

0

0.02

0.04

0.06

0.08

0.1

0.12

0.00 0.50 1.00 1.50 2.00 2.50

Fre

qu

en

cy

Velocity (m/s)

DEM modeliing

PEPT measurements

DEM: data on all particles

PEPT: the time averaged data for one tracer but over a long period of running

time (excess of 10 minutes)

Average normalised velocity from PEPT: 0.43

Average normalised velocity from DEM: 0.41

0.36 0.72 1.08 1.44

Normalised Velocity (-)

16

Quantitative Comparison of Powder

Flow between DEM and PEPT

5 l, 5 Hz Tip Speed 3.52 m/s

1 l, 7 Hz Tip Speed 3.17 m/s

Scale-up w2/w1=(d1/d2)n n = 1 Tip speed constant

n = 0.5 Froude number constant

m/s 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 50 l, 2 Hz

Tip Speed 4.13 m/s

DEM Modelling of Cyclomix High Shear

Mixer Granulators: Scale-up

The coefficient of variation of shear stress decreases as the impeller tip speed is increased.

DEM Modelling of Cyclomix High Shear

Mixer Granulators: Scale-up

0

0.2

0.4

0.6

0.8

1

1.2

50 5 1

Co

eff

icie

nt

of

Va

ria

tio

n o

f S

he

ar

Str

es

s (-

)

Granulator Size (l)

Tip Speed (n=1)

n=0.8

Froude (n=0.5)

Tip speed

increases

Tip speed

increases

Modelling of Particle Milling: • Minimising Trial and Errors

• Insufficient Available Materials

19

1- Modelling particle

breakage: particle made

of agglomerates, clusters

of smaller elements bond

Computationally very

expensive, difficult to

model full scale mill

2- Predicting the mill

performance: collisional

energy, stress magnitude

and distribution

DEMV Simulations

20

Modelling of Ball Milling

DEM simulation at 25 Hz of milling frequency in the single ball mill

n

j

jn vmE1

2

2

1 *

• Milling power (Pn ) is deduced from:

• Milling energy (En ) is deduced from the relative velocity (v ) and

reduced mass (m* ) of the two objects in contact by:

t

EP n

n

K p = 0.1218 P n H / K c2

R2 = 0.9826

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0 0.1 0.2 0.3 0.4 0.5 0.6

P n H / K c2 (m

2 s

-1)

Mil

lin

g R

ate

Co

nst

an

t, K

p (s

-1)

MCC-18 HzMCC-25 Hz

αLM-25 HzStarch-25 Hz

Sucrose-12 HzSucrose+Aerosil-25 Hz

21

Unification of Results

Material property group

Milling Power

Solid-fluid flow modelling

Fluidization Sedimentation/

re-suspension

DEM + Continuum Method (CFD); Full Fluid-Solid coupling

Solid/fluid interaction

22 22

Dispersion

Pneumatic conveying,

powder dispersion and

fluid diffusion

Solid-fluid Flow Modelling:

DEM-CFD coupling

23

Agglomerate dispersion

With an increase in bonding interface energy it becomes increasingly difficult

to disintegrate particle clusters.

0.2 J/m2

0.6 J/m2

1.0 J/m2

t = 1×10-4 s t = 2×10-4 s t = 3×10-4 s t = 4×10-4 s

Solid-fluid Flow Modelling:

agglomerate dispersion

The dispersion ratio (DR); i.e. ratio of the number of broken bonds to the initial number of bonds, (DR = 1 means all bonds are broken)

0.0

0.2

0.4

0.6

0.8

1.0

0 50 100 150

DR

(-)

Relative velocity (m/s)

0.1 J/m²

0.2 J/m²

0.3 J/m²

0.4 J/m²

0.5 J/m²

1.0 J/m²

0

t0

N

N-NDR

Solid-fluid Flow Modelling:

agglomerate dispersion

DR shown as a function of relative velocity between the fluid and particles.

26

DEM provides useful information in understanding particulate

processes and obtaining parameters difficult to measure by

experiment.

DEM analysis shows good capabilities of interpretation of

experimental data

Numerical modelling capabilities can enable virtual experiments

instead of extensive trial and errors:

Particulate Process Development

Process Optimisation

Process Scale-up

Challenges and Opportunities

Realistic and Complex Models

High Performance Computing (CPU&GPU)

Model Calibration

Concluding Remarks

2002 (5,000) 2005 (40,000) 2008 (500,000) 2012 (10 million)

Development of Modelling Capabilities (Desktop Workstation)

28

Prof. Mojtaba Ghadiri

Dr Colin Hare

Dr Graham Calvert

Dr Hongsing Tan

Dr Andrew Bayly

Dr Boonho Ng

Dr Chi Kwan

Dr Hossein Ahmadian

Mr Massih Pasha

Mr Umair Zafar

Mr Mohammad Afkhami And colleagues in Ghadiri research group

Acknowledgements

Thank you for your attention ! 29