Title: A Population Balance Model for Agglomeration
1A Population Balance Model for Agglomeration
- M. Goodson, M. Kraft, S. Forrest, J. Bridgwater
2Powder Granulation
- Dry calcium carbonate powder particles
- Add aqueous polymer solution as binder
- Irregular aggregates formed
- Coagulation (sticking together)
- no well-established rate law
- Compaction (porosity reduction)
- first order porosity reduction
3Experimental Results Size Distributions
Size distributions are very similar after the
same number of blade revolutions
4Results Qualitative Observations
5Existing Population Balance Model 1
- Characterise particle with three independent
properties - solid volume, s,
- liquid volume, l, and
- air volume, a.
- Particle can then be characterised according to
whether it is - big (sla gt 500 ?m) or small
- wet ( ) or dry
- Conservation laws for two coagulating particles
are simple -
6Existing Population Balance Model 2
- Propose a property-independent collision rate
with a property-dependent coagulation
probability. - Observations lead to a table of coagulation
probabilities - PhD
Thesis PAL Wauters, TU Delft
7Existing Population Balance Model 3
- Compaction reduction of particle porosity, ?
- Treated by using empirically derived rate law
- In terms of the parameters of interest, porosity
can be written as - So the rate of compaction can be expressed as
-
- Which can lead to unphysical results
8A New Population Balance Model
- How do we characterise an aggregate?
- solid volume, s
- liquid volume, l
- pore volume, p
- total volume, v
- porosity, ?
- surface area, a
- How many of these are independent?
- How many of these are conserved?
9Two types of Coagulation
(1)
(2)
- Interactions can preserve surface area (1) or
pore volume (2). - Real interactions may fall between these two
extremes. - Need some way of predicting pore volume if area
is known (and vice versa)
10Characterising an Aggregate
Define theoretical radius that includes a total
volume equal to s p. Use fractal
dimensions Assume Relate pore volume to
surface area Now solid volume (and liquid
volume) conserved on aggregation so if surface
area is conserved, pore volume can be found (and
vice versa)
11New Coagulation Probabilities
Include a consideration of whether particle is
either soft ( ) or hard
12Experimental Results Size Distributions
Size distributions are very similar after the
same number of blade revolutions
13Proposed Coagulation Rate
- Set collision rate and compaction rate
proportional to blade rotation speed - Expect similar size distributions after same
number of rotations at different speeds - Make critical porosity (above which a particle
is considered to be soft) a function of blade
rotation speed - Expect to see differences in other particle
properties (e.g. porosity) at different rotation
speeds
?crit
?
14Multi-dimensional Population Balances
- pbe often size-dependent only
- can be readily solved using standard
deterministic techniques - This model is based on three independent
particle properties - Solving the population balance equation by
standard numerical methods gets prohibitively
computationally expensive
15Stochastic Simulation
- Convert rate equations into probabilities that a
certain event will happen within a given waiting
time - This is a function of the whole population of
particles at any given time - Update particle population according to either
coagulation or compaction jump - Stochastic treatment of coagulation can be
inefficient due to the need to consider n(n-1)/2
possible coagulating pairs. - Increment timestep for every collision rather
than every coagulation - Use table of probabilities to determine which
collisions result in coagulations - Complexity of problem significantly reduced
16Stochastic Particle System
17Simulation Results 1
After 120 revolutions, very similar size
distributions are observed at different critical
porosities (to model different blade rotation
speeds)
18Simulation Results 2
but differences can be observed in other
properties, such as the average particle porosity.
19Conclusions
- Some problems cannot be solved with a
one-dimensional population balance - Standard numerical methods become prohibitively
computationally expensive when applied to
multi-dimensional population balances - Stochastic methods can be extended from one
dimension to many without a significant loss of
efficiency - Stochastic simulation has been shown to
qualitatively predict the behaviour of
granulating powders - Refinements to the model will then lead to
quantitative prediction