Lecture Notes on Population Balance Modelling

Lecture 1: Introduction to Population Balance

Modelling

Dr. Ho Yong Kuen

School of Engineering (Chemical Engineering)
Monash University Malaysia
Why Population Balance?
Many quantities in real life are distributed,
e.g. a distribution of particle size

?
crushed
Number Density

Particles
Poly-disperse

?
hydrolyzed

Size, x
Starch
Describing these things as a lumped entity squelches
the distributed nature of the system!
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 2
What is Population Balance Modelling?

A mathematical tool to describe the evolution of population-related

quantities.
Basically anything that has a
Example Applications (Ramkrishna
population distribution!
and Singh 2014):
 Polymerization Some benefits:
 Obtain important properties
 Cell growth, division, differentiation and
of the population (number-
death
average particle size, weight-
 Milling average particle size)
 Crystallization  Study interaction within
populations at the molecular
 And many more….
level (how an enzyme cuts a
Ramkrishna, D., & Singh, M. R. (2014). Population Balance Modeling:
Current Status and Future Prospects. Annual Review of Chemical and
polymer)
Biomolecular Engineering, 5(1), 123-146.
 Probing of the change in
crystal morphology that is
challenging to trace
experimentally
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 3
Particle State Vector

 The particle state vector tells you the STATE of the particle!

 If we have a bunch of polymers in which the Degree of Polymerization (DP)

changes with time, then the DP of the polymer is the particle state.

 If we have a population of cells that do not divide until after a certain age, then
the cell age is the particle state.

 Choice of particle state is determined by variables needed to specify:

• The rate of change of those of direct interest to the application

• The birth and death processes.
 There can be more than one particle state, e.g. if we have a bunch of cylindrical
crystals in which the diameter and length vary throughout, the particle states
can be length and diameter.

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 4

Particle State Vector
 In general, the particle state vector accounts for both the internal and external
coordinates:

 Internal (size, age, DP etc.): x   x1 , x2 ,..., xd 

 External (position in physical space): r   r1 , r2 , r3 

 Therefore, the particle population may be regarded as being randomly

distributed in the particle state space, which includes physical space and the
space of internal coordinates.

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 5

The Continuous Phase Vector

Dispersed phase If the dispersed fraction is very

minimal, we can neglect its effect
Continuous phase on the continuous phase..

 The continuous phase may affect the behavior of the each particle and can be
collated into:
Y  r , t   Y1  r , t  , Y2  r , t  ,..., Yc  r , t  

 Governed by the laws of transport and interaction with particles.

 Continuous phase balance can be omitted if the interaction with particles does
not bring about any change in the continuous phase, e.g. dispersed phase is
non-soluble in the continuous phase. This shall be our focus here.

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 6

The Number Density Function
 In PBM, we talk about average number density function:

f1  x , r , t   E  n  x , r , t   Why do we consider the expected value?

 In the 1-D case and assuming no spatial variations, we have:

Number Density, f1(x, t) When (xi+1 – xi ) approximates zero,
we have a sufficiently smooth
number density function f1(x, t).

Imagine you can measure the average

number of particles at a particular x, and
do the same for all x’s!

Size, x
xi xi+1
 We will loosely refer f1(x, r, t) as the number density function.

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 7

The Number Density Function
 If we have the number density, we can calculate several important properties
of the population. Consider the 1-D case with no spatial variations:

 f1  0
 t   0 f1  x, t  dx (total number - 0th moment)

 f1 1
 t   0 xf1  x, t  dx (total mass - 1st moment)

 f1  2
 t   0 x 2 f1  x, t  dx (2nd moment)

 f1 n
 t   0 x n f1  x, t  dx (n-th moment)

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 8

The Population Balance Equation
(1-D Case and No Spatial Variation)
Flux In Flux Out
x  x, t  x  x, t 
f1  x, t  f1  x, t 
t t
h  x, t  dx
b
xa
a
x b

x
a b
 b x  x, t  x  x, t  b

 f1 x , t  dx  f1 x , t   f1  x, t    h  x, t  dx
t a t xa
t x b a
ACCUMULATION
FLUX ENTERING x  a FLUX LEAVING x  b NET BIRTH RATE

f1  x, t    x  x, t   b
f1  x, t   dx   h  x, t  dx
b b
Leibniz Rule a t
dx   
a

x  t  a

f1  x, t    x  x, t   b
f1  x, t   dx   h  x, t  dx  0
b b
a t
dx  
a

x  t  a

f1  x, t    x  x, t  
f1  x, t    h  x, t  dx  0
b
 a t
 
x  t 
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 9
The Population Balance Equation
(1-D Case and No Spatial Variation)
f1  x, t    x  x, t  
f1  x, t    h  x, t  dx  0
b
a t
 
x  t 
 From the above then we can write:

f1  x, t    x  x, t  
  f  x , t    h  x, t 
t x  t
1


This is the PBE for 1-D with no spatial variation!

 If we could model the following phenomenological components accurately:

x  x, t  Growth rate of
h  x, t  Breakage/Aggregation
t particle of size x

Then we could predict the evolution of the number

density with time!

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 10

The Population Balance Equation
(General Case)
 Drawing from the 1-D example, the general population balance equation can
be written as:

  x  x , r , t    r  x , r , t  
f1  x , r , t    x  f1  x , r , t     r  f1  x , r , t    h  x , r , t 
t  t   t 

 Like any PDEs, the equation has to be solved with appropriate initial condition
and boundary conditions. We will talk about these when we encounter specific
examples.
  x  x , r , t  
 Well-mixed batch reactor: f1  x , r , t    x  f1  x , r , t    h  x , r , t 
t  t 

 Zero growth and well-mixed batch reactor: f1  x , r , t   h  x , r , t 
t

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 11

PBE Involving Pure Breakage

 Breakage is an important phenomenon occurring in many processes, e.g.

milling, depolymerization, etc.

 Let us use a 1-D example with no spatial variation and zero growth rate for
simplicity:

f1  x, t   h  x, t 
t

We are concerned with this NET BIRTH

RATE, which may be expressed as the
difference between a source and a sink
term.
h  x, t   h   x, t   h   x , t 
SOURCE SINK

Our objective is to give the general expression of the source

and sink terms.
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 12
PBE Involving Pure Breakage


f1  x, t   h   x, t   h   x, t 
t
SOURCE SINK

 SINK term: Breakage of particles of size x into smaller pieces

h   x, t   b  x, t  f1  x, t  The breakage frequency has

BREAKAGE
FREQUENCY
dimensions of reciprocal time

 SOURCE term: Sum of breakage of all particles of size > x into x


h 
 x, t   x v  x ', t  b  x ', t  P  x | x ', t  f1  x ', t  dx '
AVERAGE BREAKAGE PDF FOR BREAKUP
NUMBER FREQUENCY OF PARTICLE OF
OF PARTICLES STATE x' TO x
FORMED

Binary scission, v = 2 Decides the rate Decides the type of

of scission scission
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 13
PBE Involving Pure Breakage

 In many cases, only the number density changes with time, so:
 
f1  x, t    v  x ' b  x ' P  x | x ' f1  x ', t  dx '  b  x  f1  x , t 
t x
AVERAGE BREAKAGE PDF FOR BREAKUP BREAKAGE
NUMBER FREQUENCY OF PARTICLE OF FREQUENCY
OF PARTICLES STATE x' TO x
FORMED

 Additional constraints:
x'
P  x | x ' dx  1 Normalization condition: Area under the curve for a
0 probability density function should be 1

P  x | x '  0, x  x ' Conservation of mass

Conservation of mass: Equality holds if there is no

x '  v  x '  xP  x | x '  dx
x'

0 loss of mass during breakage

P  x | x '  P  x ' x | x '  Symmetry condition

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 14

Conservation of Mass
 Can be checked using the first moment:

Scission

First Moment  3  2  2  3  12 Conserved!

3 OF DP2 2 OF DP3


f 1
 t   0 xf  x, t  dx Continuous form
Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 15
Conservation of Mass
 Can be checked using the first moment using the PBE:
    
xf1  x, t  dx   x  v  x ' b  x '  P  x | x '  f1  x ', t  dx ' dx   xb  x  f1  x, t  dx
t 0 0 x 0

f1   t 
1
  
  x  v  x ' b  x ' P  x | x '  f1  x ', t  dx ' dx   xb  x  f1  x, t  dx
t 0 x 0

 Let us then study the first integral on the RHS on the (x, x’) plane:
x’ = x x’ = x

x’ x’

x x

 x  v  x ' b  x ' P  x | x ' f  x ', t  dx ' dx
  x'

 x  v  x ' b  x ' P  x | x ' f  x ', t  dx ' dx

0 x 1 0 0 1

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 16

Conservation of Mass
f1   t 
1
 
  x  v  x ' b  x ' P  x | x '  f1  x ', t  dx ' dx   xb  x  f1  x, t  dx
x'

t 0 0 0

 
  dx ' b  x ' f1  x ', t  v  x '   xP  x | x '  dx   xb  x  f1  x, t  dx
x'

0 0 0



 x'
 
dx ' b  x ' f1  x ', t  v  x '   xP  x | x '  dx   xb  x  f1  x, t  dx
0  0  0
 
  dx ' b  x ' f1  x ', t  x '   xb  x  f1  x, t  dx
0 0

0

First moment is always constant!

 One critical aspect of solving PBE is that the first moment must be conserved.
Various researchers are developing numerical solvers which could conserve the
zeroth and the first moments.

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 17

Breakage PBE: Random Scission

 
f1  x, t    v  x ' b  x '  P  x | x '  f1  x ', t  dx '  b  x  f1  x, t 
t x

Polymer chain with 6

monomer and 5 bonds

Equal
probability

 Let x and x’ be the number of mers in the polymer:

1
P  x | x ' 
1
P  x | x '  OR for particle breakage
x ' 1 x'

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 18

Breakage PBE: Random Scission

 
f1  x, t    v  x ' b  x '  P  x | x '  f1  x ', t  dx '  b  x  f1  x, t 
t x

1 1
Stoichiometric kernel: P  x | x '  or Question: How many
x ' 1 x' initial conditions and
Binary scission: v  x  2 boundary conditions do
you need to solve this?
Rate kernel: b  x   k p (constant rate kernel)

 The resulting PBE for random scission will be:

  2
f1  x, t   k p  f  x ', t  dx '  k p f1  x, t  Polymer scission
t x x ' 1 1

  2
f1  x, t   k p  f  x ', t  dx '  k p f1  x, t  Particle breakage
t x x' 1

These are the equation to solve for random scission (uniform breakage)

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 19

Breakage PBE: Chain-End Scission

Polymer chain with 6

monomer and 5 bonds

Let x and y be the number of mers in the

polymer:

P  x | x '    x   x ' 1

 Called the Dirac delta kernel

 Exists only at x = x’ – 1
…

 Very challenging kernel to solve

A separate treatment will be needed to

solve this problem!

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 20

Common Methods to Solve PBE

 Analytical solutions  Numerical solutions

 Only for simple cases
 Discretization of the PBE
 Method of moments
• Finite difference, finite volume, finite
 Solves only for the moments element techniques
and not the complete
distribution • Sectional methods (Fixed Pivot
technique, Moving Pivot technique etc.)
  
x f1  x, t  dx
t 0 • Solves for the complete distribution
 
  x   v  x ' b  x ' P  x | x '  f1  x ', t  dx ' dx • Our group will prioritize on this
0 x
 particular class of solution methods!
  x b  x  f1  x, t  dx
0

Lecture Notes on PBM - Dr. Ho Yong Kuen 2016 21