Model-based scale up of solid-

state cultivation bioreactors:

A case study

David Mitchell
Department of Biochemistry and Molecular Biology,
Federal University of Paraná, Curitiba, Brazil

davidmitchell@ufpr.br
1
I will present a case study:
The use of modeling to guide
the scale up of a packed-bed bioreactor
...for the production of pectinases
...by solid-state cultivation
...in citrus waste biorefineries

Aims of the talk

To demonstrate that models can be important tools
for guiding the scale-up and operation of bioreactors
To illustrate what is involved in modeling bioreactors
• the type of thinking
• the type of experiments and data required

2
The context is that of a packed-bed bioreactor,
• modeling other bioreactors will follow a similar
process
• although particular phenomena will be different

You may not have a mathematical background

• focus on the phenomena, not the equations
• even if you won’t do any modeling yourself, it is still
important to understand the modeling process – so
that you can interact with the “modeler"

3
 Our work with pectinase production was motivated by
the idea of a citrus waste biorefinery

• Brazil is the largest producer of

orange juice in the world
• The Brazilian orange juice industry
produces over 1 million tonnes (dry
matter) of citrus pulp per year
• This waste is currently...
– used to produce pectin
(the market is too small)
– dried for use in animal feed
www.agrossim.com.br (drying is costly)
– disposed of in landfills
Pelleted (or worse...)
citrus
pulp

www.agry.com.br 4
 So, what to do with all this citrus pulp?

It has potential to be used as a feedstock for biorefineries...

5
 A possible scheme for a citrus waste biorefinery
Citrus pulp Separation
peels bagasse
vapor
Recovery Distillation
residue Fermentation
D-Limonene Acid seeds
hydrolysis

solid Anaerobic Product

Cellulose
residue digestion recovery
removal
supernatant
Biogas Products
Precipitation soluble
sugars
pectin
Pectinase
production pectinases Enzymatic
by solid-state hydrolysis
cultivation D-galacturonic acid
Mucic acid or
Recovery Conversion
L-galactonic acid
6
 A possible scheme for a citrus waste biorefinery
Citrus pulp Separation
peels bagasse
vapor
Recovery Distillation
residue Fermentation
D-Limonene Acid seeds
hydrolysis

solid Anaerobic Product

Cellulose
residue digestion recovery
removal Study of pectinase
supernatant production in a Products
pilot-
Biogas
soluble
Precipitation scale bioreactor for use
sugars
pectin in pectin hydrolysis
Pectinase
production pectinases Enzymatic
by solid-state hydrolysis
cultivation D-galacturonic acid
Mucic acid or
Recovery Conversion
L-galactonic acid
7
 Our pilot-scale solid-state cultivation bioreactor

Packed-bed bioreactor – with the possibility for intermittent

mixing events (through rotation around the central axis)
a baffle to
help during
agitation
air outlet
100 cm

50 cm
bed capacity
of 200 L

air inlet

60 cm
70 cm
8
Details of the aeration system

temperature and
RH of the outlet air

humidification
air from tower
blower temperatures in
various
positions within
the substrate
bed
 cool temperature and RH
water  of the inlet air
EITHER
pump 1 on, valve 1 open,
 warm
water  pump 2 off, valve 2 off
OR
pump 1 off, valve 1 off
pump 2 on, valve 2 on

9
 Filter Flow meter Blower

Water tanks
Humidification tower (one cool, one warm)
10
 Bioreactor
air outlet (to gas washer) thermocouple modules
air hose from
humidification
tower

lid
(open)

Motor
(to rotate the
bioreactor)

air hoses for the inlet and outlet air differential pressure sensor
(disconnected) with temperature (gives pressure drop across the bed)
and relative humidity sensors 11
Inside of the bioreactor 2 thermocouple sleeves
(4 thermocouples in each sleeve)

air outlet
(valve
closed)

central axis perforated plate, covered by a wire screen

12
So, what is the scenario?

We are interested in producing pectinases by SSC

(for later use in the hydrolysis of pectin to produce
D-galacturonic acid)

We have a small pilot-scale packed bed bioreactor

We think that it might be interesting to scale up the

production of pectinases to even larger scales

We think that scale-up should be guided by models

13
Seven steps of modeling
1. Why model? Know what you want to use your model
for and what type of model will be appropriate
2. Draw/describe your system. Decide how you will
represent your system. Make appropriate assumptions
3. Write the equations
4. Determine the parameters of the equations
5. Solve the model
6. Calibrate/validate the model using experimental
results. Explore its sensitivity to the parameters and
operating variables
7. Use the model as a tool for the intended purpose,
always being ready to refine it further, if appropriate
14
Seven steps of modeling
1. Why model? Know what you want to use your model
for and what type of model will be appropriate
2. Draw/describe your system. Decide how you will
represent your system. Make appropriate assumptions
The modeling process is not necessarily
3. Write the equations linear
4. Determine the be
Steps may parameters of the equations
done simultaneously or in a
5. Solve the model different order
You may need to
6. Calibrate/validate thebacktrack fromexperimental
model using later steps
to rethink
results. Explorewhat you did you
its sensitivity earlier
to the steps
parameters and
operating variables
7. Use the model as a tool for the intended purpose,
always being ready to refine it further, if appropriate
15
Step 1: Why model? Know what you want to use your
model for and what type of model will be appropriate
We want a model to guide scale up of pectinase
production in a packed-bed bioreactor
We will focus on controlling the temperature in the bed
(minimizing deviations from the optimum temperature for growth)

We want to use the model to guide decisions

about design and operation
scale up
• what bed heights can be used?

• what air flow rates will be necessary?

(this will affect the sizing of the aeration system)

16
Step 1: Why model? Know what you want to use your
model for and what type of model will be appropriate
At this moment, we are NOT interested in describing
microscale phenomena in detail.
Describing microscale phenomena would
• be too complex
• require the determination of too many parameters,
...without improving the usefulness of the model as a tool for guiding
scale up

We are NOT interested in describing biomass production

(i.e. growth) and pectinase production accurately – our
specific interest is in temperature control
With these decisions, it is clear we are aiming for a highly
simplified model (it can obviously be improved!)
17
Why would it be too complex to try to describe microscale
phenomena?
Potentially, we could describe the growth rate as
depending on the local concentrations of nutrients and O2

In submerged liquid fermentation, this is simple to do

dX  max S
= X If the liquid is well
dt K S +S mixed, each
variable represents
dS 1 dX the value at all
=− points within the
dt YXS dt fermentation broth

18
What if we try to do the same in SSC?
• The biomass experiences local concentrations – within individual
particles
• The individual particle is NOT a well-mixed system

dX  max S
r
= r
X
dt r K S +S r
r

S DS   2 S  1 dX
=− 2 r  −
t r r r  dr  r YXS dt r

S and X are functions of the radial position “r” (O2 would be too!)
The model will already need to describe the fact that the biomass
concentration varies with height within the bed!
This complexity is usually avoided by using empirical equations
19
 Step 2: Draw/describe your system. Decide how you will
represent your system. Make appropriate assumptions

We were interested in modeling traditional packed beds

A large-scale packed bed will be several meters wide...

scale up

our pilot
Likely large-scale design
bioreactor
If a Zymotis type bioreactor will not be
used at large scale, then “horizontal”
heat transfer will be negligible

20
Assumption: air flow is Consequence: The model will not
uniform across the bed describe “horizontal” heat transfer
There is only one spatial
variable, the bed height, “z”

Assumption: in a
wide packed bed,
heat transfer
through the side
walls is negligible

21
Assumption: air flow is Consequence: The model will not
uniform across the bed describe “horizontal” heat transfer
There is only one spatial
variable, the bed height, “z”

Convection and evaporation to the flowing air

(which flows in the vertical “z direction”) will
make the major contribution to heat removal
Assumption: in a
There is no advantage in developing
wide packed bed, a model
to describe both verticalheat
andtransfer
horizontal heat
through the side
transfer
walls is negligible

22
Axial conduction will occur down the temperature gradient and
therefore will oppose convective heat removal

Assumption: Axial conduction

makes a negligible contribution
to heat removal

It should be
negligible if
reasonable air
flow rates are
used!

23
With these assumptions, the mass and energy balances
will need to describe the following phenomena

A representative section of the bed

convective flow of
water vapor and
thermal energy

no heat
solids-air growth and removal
water transfer metabolic heat
production
solids-air
heat transfer

24
Key variables and parameters will be

solids moisture content

air humidity (dry basis)
kg-vapor kg-dry-air-1 kg-H2O kg-dry-solids-1
g solids-air s
water transfer parameters
coefficient (k) related to the rate
of metabolic heat
solids-air production
Tg heat transfer Ts
gas coefficient (h) solids
temperature temperature

FG = mass flow rate of the air

kg-dry-air m2-bed s-1
25
 Step 3: Write the equations
Step 4: Determine the parameters of the equations
The decisions in steps 1 & 2
lead to what bioreactor model? Operating variables
related to the inlet air
Growth submodel Mass flow rate (Fg)
• empirical kinetic equation Temperature (Tgin)
• growth depends on Ts and s Humidity (gin)

values of Rates of production of

Ts and s • metabolic heat
• metabolic water

Mass and energy balance submodel: key variables as f (t, z)

Ts solids temperature How these are affected by
Tg gas temperature • convective flow of heat and water vapor
s solids moisture content • solids-gas heat and water transfer
g gas humidity • microbial production of heat and water
26
Step 3: Write the equations
Step 4: Determine the parameters of the equations

This will be covered as

Steps 3A and 4A: The submodel describing growth

kinetics – equations and parameters

Steps 3B and 4B: The submodel describing heat and

mass transfer in the bed – equations and parameters

27
 Steps 3A and 4A: The submodel describing growth
kinetics – equations and parameters
Temperature control will be the key issue at large scale
The main aim of our growth kinetics submodel is to
describe realistic heat production rates
In aerobic processes, metabolic heat production is
typically directly related to oxygen uptake

We worked with Aspergillus niger,

but similar work could be done with
other organisms

https://commons.wikimedia.org/wiki/File:Aspergillus_niger_meaox.png 28
 We did oxygen uptake studies in Raimbault columns
𝑑𝑄
= 𝑌𝑄𝑂2 𝑂𝑈𝑅
blower 𝑑𝑡
*YQO2 = 520 J mmol-O2-1

100

Heat production rate (J kg-1 s-1)

O2 uptake rate (mmol kg-1 s-1)
analyzer 80
0.15

60
0.10
40

0.05
20

0.00 0
Waterbath 0 5 10 15 20 25
Time (h)

*Bailey & Ollis, Biochemical Engineering Fundamentals,

2nd ed., McGraw-Hill, New York, 1986. 29
We needed to find an empirical kinetic equation that would give a
bell-shaped profile for the heat production rate

The logistic equation is a good approximation

𝑑𝑋 𝑋 X is the biomass content of the solids
= max 𝑋 1 − (kg-dry-biomass kg-dry-solids-1)
𝑑𝑡 𝑋max

the growth rate accelerates when integrated, it gives this

and decelerates shape for the growth profile

𝑑𝑋 X
𝑑𝑡

time time
30
 The rate of metabolic heat production was assumed to be directly
proportional to the growth rate (i.e. no maintenance metabolism)

𝑑𝑄 𝑑𝑋 𝑋
100 = 𝑌𝑄𝑋 = 𝑌𝑄𝑋 max 𝑋 1 −
Heat production rate (J kg-1 s-1)

𝑑𝑡 𝑑𝑡 𝑋max
80

This fit was obtained with

60
Parameter from the literature*
40
YQX = 1.54107 J kg-dry-biomassa-1
20
Fitted parameters
0 Xo = 0.001 kg-dry-biomass kg-dry-solids-1
0 5 10 15 20 25
Xmax = 0.25 kg-dry-biomass kg-dry-solids-1
Time (h)
max = 0.42 h-1

*Pajan et al. (1997) https://doi.org/10.1007/978-94-017-1711-3_19

31
What are the effects of key process variables on the
maximum specific growth rate constant, max ?
• the temperature
• the water activity (related to the water content)

This was done by using empirical equations for fractional

specific growth rates based on temperature and water activity

max = max(opt)  fT  fW

32
 The effect of temperature was described by an empirical equation adapted
from the literature*

1.0

0.8

fT 0.6 −70225
8.40651011 . exp
𝑅 𝑇𝑠 + 281
𝑓T =
−283356
0.4 1 + 1.31047 . exp
𝑅 𝑇𝑠 + 281

0.2

0.0
20 30 40 50
Temperature (°C)

*Saucedo-Castañeda et al. (1990) https://doi.org/10.1002/bit.260350808 33

The effect of water activity was described by an empirical equation fitted
to literature data*
1.0

0.8 3
618.9218𝑎ws
−1863.527𝑎 2
0.6 𝑓W = 𝑒𝑥𝑝 ws
+1865.097𝑎ws
fW −620.6684
0.4

0.2 We need an
isotherm to describe
the water activity of
0.0 the solids as a
0.85 0.90 0.95 1.0
function of their
Water activity moisture content

*Glenn & Rogers (1988) Australian Journal of Biotechnology 2, 50–57.

34
Summarizing our kinetic model

effects of temperature and the

water activity of the solids are
combined empirically

Logistic 𝑑𝑋 𝑋
= max 𝑋 1 −
equation 𝑑𝑡 𝑋max

the rate of metabolic heat production is

directly proportional to the growth rate

𝑑𝑄 𝑑𝑋 𝑋
= 𝑌𝑄𝑋 = 𝑌𝑄𝑋 max 𝑋 1 −
𝑑𝑡 𝑑𝑡 𝑋max

(The aim was to describe, empirically, the heat production rate

inferred from O2 uptake rates measured in Raimbault columns)
35
Steps 3B and 4B: The submodel describing heat and
mass transfer in the bed – equations and parameters
Four “balance equations” are necessary
• a mass (water) balance on the solids phase
• a mass (water) balance on the gas phase
• an energy balance on the solids phase
• an energy balance on the gas phase

These equations contain two key parameters

• k, the air-solids mass (water) transfer coefficient
• h, the air-solids heat transfer coefficient

We estimated h and k experimentally, in cooling, heating

and drying experiments done in the absence of growth

36
Cooling experiment

50
cool, saturated air
Solids temperature (°C)

z = 18 cm
40

30

z = 5 cm
20
0 4 8 12 16 20
Time (min) 37
Heating experiment

warm, saturated air 44

z = 5 cm
Solids temperature (°C)

40

36

32 z = 18 cm
28

24

20
0 4 8 12 16 20
Time (min) 38
Drying experiment

36
z = 5 cm

Solids temperature (°C)

32

28

24

20
z = 18 cm
16
hot, dry air 0 200 400 600 800 1000
Time (min)

the experiment takes much longer!

39
 36
z = 5 cm

Solids temperature (°C)

32

28

24

20
z = 18 cm
16
hot, dry air 0 200 400 600 800 1000
Time (min)

the bed initially cools to the

wet bulb temperature of the air
40
 36
z = 5 cm

Solids temperature (°C)

32

28

24

20
z = 18 cm
16
hot, dry air 0 200 400 600 800 1000
Time (min)

a drying front moves through

the bed and the bed heats
41
 the dried solids heat to the
temperature of the hot dry air

36
z = 5 cm

Solids temperature (°C)

32

28

24

20
z = 18 cm
16
hot, dry air 0 200 400 600 800 1000
Time (min)

42
 50 44
z = 5 cm
Solids temperature (°C)

Solids temperature (°C)

40
z = 18 cm
40 36

32 z = 18 cm
30 28

24

20
z = 5 cm 20
0 4 8 12 16 20 0 4 8 12 16 20
Time (min) Time (min)

36
z = 5 cm

Solids temperature (°C)

We estimated the heat and 32

mass transfer parameters 28

(h and k) by fitting a model
24
of heat and mass transfer
(in the absence of growth) 20
z = 18 cm
to the data 16
0 200 400 600 800 1000
Time (min) 43
 A model of heat and mass transfer in the bed
(no growth)

The mass (water) balance on the solids phase

moisture content bulk density
(kg-dry-solids kg-dry-solids-1) (kg-dry-solids m-3)

rate of change of
the amount of 𝑑s 𝑆 driving force for
water held in the = −𝑘(s − ∗s ) evaporation
solids (kg m-3 s-1)
𝑑𝑡

air-solids mass (water)

transfer coefficient

44
How to interpret the driving force for evaporation?

𝑑s 𝑆
= −𝑘(s − ∗s )
𝑑𝑡

Moisture content of Moisture content that

the solids the solids would need to
have to be in equilibrium
The model keeps
with the gas phase
track of this
In order to know this, we
need the sorption
isotherm of the solids...
45
 With the isotherm of the solids, and knowing g from the
mass balance on the gas phase, we can calculate ∗S
𝑑s 𝑆
= −𝑘(s − ∗s )
𝑑𝑡

curve described by
0.6
Isotherm for the Oswin equation
0.488
wheat bran ∗s 0.4 s = 0.095
∗
𝑎wg
from the 1 − 𝑎wg
literature# 0.2

0.2 0.4 0.6 0.8 1.0

The mass balance on the calculate awg

gas phase calculates g

#Casciatori et al. (2015) https://doi.org/10.1016/j.indcrop.2014.11.034 46

 The mass (water) balance on the gas phase
humidity mass of gas per m3 of bed
(kg vapor kg-dry-air-1) (kg-dry-air m-3-bed)
evaporation
rate of change of
the amount of 𝑑g 𝐺 𝑑g
= −𝐹G + 𝑘(s − s∗ )
water held in the 𝑑𝑡 𝑑𝑧
air (kg m-3 s-1)

mass flow rate of air

(kg-dry-air m-2 s-1) axial gradient of water vapor
in the gas phase (related to
the amount of vapor in the
gas entering and leaving
each position in the bed)

47
 convective
flow of vapor in
the air stream evaporation
rate of change of
the amount of 𝑑g 𝐺 𝑑g
= −𝐹G + 𝑘(s − s∗ )
water held in the 𝑑𝑡 𝑑𝑧
air (kg m-3 s-1)

48
The energy balance on the solids phase

𝑑𝑇s
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘(s − ∗s )
𝑑𝑡

SCPsTs is the amount of thermal energy held in the

dry solids within the solids phase (J m-3-bed)
SCPw sTs is the amount of thermal energy held in the
liquid water held within the solids phase (J m-3-bed)

Remember the equation H = mCPT?

On the left-hand side, we are calculating H = mCP(T-Tref)
where the reference temperature is 0 °C
49
 𝑑𝑇s
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘(s − ∗s )
𝑑𝑡

h(Tg-Ts) describes the transfer of sensible heat from the gas

phase to the solids phase
h is the solids-air heat transfer coefficient (J s-1 m-3-bed °C-1)
Its value depends on the total surface area
of contact between the gas phase and the
solids phase (m2 of contact area per m3 of bed)
But it is impossible to know the surface
area of contact
in this sense, it is like kLa in gas-liquid O2
transfer – which we treat as a single parameter
The driving force is the temperature difference between the gas
and the solids
50
 𝑑𝑇s
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘(s − ∗s )
𝑑𝑡

Hvapk(s - s* ) is the removal of energy in the form of the latent

heat of vaporization as water evaporates from the solids phase
to the gas phase
k is the solids-air mass (water) transfer
coefficient (“kg-water s-1 m-3-bed” evaporated
per “kg-water/kg-dry-solids” of driving force)
As with h, its value depends on the total
surface area of contact between the gas
phase and the solids phase (m2 of contact
area per m3 of bed).
Again, it is impossible to know the surface area of contact.
We have already talked about the driving force
51
The energy balance on the gas phase

𝑑Tg 𝑑Tg
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ Tg − Ts
𝑑𝑡 𝑑𝑧

GCPgTg is the solids-air heat transfer phase (J m-3-bed)

SCPv gTg is the amount of thermal energy held in the

water vapor held within the gas phase (J m-3-bed)

52
The energy balance on the gas phase

𝑑𝑇g 𝑑𝑇g
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ 𝑇g − 𝑇s
𝑑𝑡 𝑑𝑧
h(Tg-Ts) describes the transfer of sensible heat from the gas
phase to the solids phase (we have already seen it in the solids
balance)
𝑑𝑇g
𝐹G 𝐶Pg + 𝐶Pv g describes the convective flow of thermal
𝑑𝑧
energy with the gas phase (related to the temperature of the gas
entering and leaving each position in the bed)

The evaporation term does not appear in the energy balance

because evaporation removes thermal energy from the solids
phase, but the water vapor that enters the gas phase does not
increase the temperature of the gas phase
53
 rate of change of the convective flow of
amount of thermal energy thermal energy in solids-air heat
held in the air (J m-3 s-1) the air stream transfer
𝑑𝑇g 𝑑𝑇g
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ 𝑇g − 𝑇s
𝑑𝑡 𝑑𝑧

54
We have a model of heat and water transfer in the bed
Mass (water) balance on the solids phase
𝑑s 𝑆 We can obtain
= −𝑘(s − s∗ ) these parameters
𝑑𝑡 (and isotherms)
from the literature
Mass (water) balance on the gas phase
𝑑g 𝐺 𝑑g
= −𝐹G + 𝑘(s − s∗ )
𝑑𝑡 𝑑𝑧
Energy balance on the solids phase
𝑑𝑇s
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘(s − s∗ )
𝑑𝑡
Energy balance on the gas phase
𝑑Tg 𝑑Tg
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ Tg − Ts
𝑑𝑡 𝑑𝑧
55
We have a model of heat and water transfer in the bed
Mass (water) balance on the solids phase
𝑑s 𝑆 We need to obtain
= −𝑘(s − s∗ ) estimates of
𝑑𝑡 h and k from the
cooling, heating and
Mass (water) balance on the gas phase
drying experiments
𝑑g 𝐺 𝑑g
= −𝐹G + 𝑘(s − s∗ )
𝑑𝑡 𝑑𝑧
Energy balance on the solids phase
𝑑𝑇s
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘(s − s∗ )
𝑑𝑡
Energy balance on the gas phase
𝑑Tg 𝑑Tg
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ Tg − Ts
𝑑𝑡 𝑑𝑧
56
 50 44
Solids temperature (°C)
z = 5 cm

Solids temperature (°C)

40

40 36

32
z = 18 cm
30 28
z = 18 cm
24

20
z = 5 cm 20
0 4 8 12 16 20 0 4 8 12 16 20
Time (min) Time (min)

36
Fits obtained with z = 5 cm

Solids temperature (°C)

32
𝐹G
ℎ = 16000
0.095 28

J s-1 m-3-bed °C-1 24

𝐹G s 20
𝑘 = 0.06 z = 18 cm
0.095 0.095 16

kg-water s-1 m-3-bed (kg-water/kg-dry-solids)-1 0 200 400 600 800 1000

Time (min) 57
In order for the model to fit well, it was necessary to express k as a
function of the moisture content of the solids (s)

k independent of s k depends on s

𝐹G 𝐹G s
𝑘 = 0.06 𝑘 = 0.06
0.095 0.095 0.095
36 z = 5 cm 36
z = 5 cm
Solids temperature (°C)

Solids temperature (°C)

32 32

28 28

24 24

20 20

z = 18 cm z = 18 cm
16 16
0 200 400 600 800 1000 0 200 400 600 800 1000
Time (min) Time (min)
poor fit
58
In other words, it appears that the drying of the solids follows a
“falling rate” drying expression
Falling rate drying occurs when
water diffusion in the particle k depends on s
limits the evaporation rate
𝐹G s
𝑘 = 0.06
0.095 0.095
36
z = 5 cm

Solids temperature (°C)

32

28
As the particle dries, the water needs
to diffuse over ever greater distances 24

within the particle, slowing the drying 20

z = 18 cm
16
0 200 400 600 800 1000
Time (min)

59
Step 5: Solve the model
We now have
• a growth kinetic model
• a model of heat and mass transfer in the bed

Combining the models, we have, for growth

Growth kinetics Logistic equation
(Calibrated with data from
𝑑𝑋 𝑋 O2 uptake experiments)
= max 𝑋 1 −
𝑑𝑡 𝑋max
YSX represents the grams of dry matter
Consumption of dry solids lost per gram of dry biomass produced
𝑑𝑆 𝑌SX 𝑑𝑋 It is negative: The overall mass of dry
= 𝑆 matter (biomass and residual substrate)
𝑑𝑡 1 − 𝑋𝑌SX 𝑑𝑡
decreases during the process!
60
...and for the heat and mass balances in the bed
Mass (water) balance on the solids phase

𝑑S 𝑆 𝑑𝑋𝑆 new term:

= −𝑘 S − S + 𝑌WX
∗ water production
𝑑𝑡 𝑑𝑡 due to growth
Mass (water) balance on the gas phase

𝑑g 𝐺 𝑑g
= −𝐹G + 𝑘(s − s∗ ) new term:
𝑑𝑡 𝑑𝑧 metabolic heat
Energy balance on the solids phase production
𝑑𝑇s 𝑑𝑋𝑆
𝑆 𝐶Ps + 𝐶Pw s = ℎ 𝑇g − 𝑇s − H𝑣𝑎𝑝 𝑘 s − s + 𝑌QX
∗
𝑑𝑡 𝑑𝑡

Energy balance on the gas phase

𝑑Tg 𝑑Tg
𝐺 𝐶Pg + 𝐶Pv g = −𝐹G 𝐶Pg + 𝐶Pv g − ℎ Tg − Ts
𝑑𝑡 𝑑𝑧
61
We have partial differential equations

The spatial coordinate is “discretized”

This generates a model with 6 ordinary

differential equations for each point in space

The “discretized” differential equations are solved using

numerical methods

There can be issues with the “numerical stability” of these

methods!
62
Solving the model, you can generate for each variable
X Dry biomass
Time course profiles
S Total dry solids
at particular positions in
s Moisture content of the solids the bioreactor
g Humidity of the gas phase OR

Ts Temperature of the solids phase Spatial profiles

at particular times
Tg Temperature of the gas phase

63
Step 6: Final calibration / validation of the model
So, we now have a model, but does it describe the
experimental data?
We compared the model predictions with data from a
pilot-scale fermentation
34

Solids temperature (°C)

33
18 cm
32

31
5 cm
30

29
0 4 8 12 16 20
Time (h)
64
 50
Solids temperature (°C)

40

30

20
We did a “sensitivity analysis”
50
Solids temperature (°C)

How sensitive are the model

40 predictions about the cooling
30
experiment to the values of h,
k and Fg?
20
50
Solids temperature (°C)

40

30

20
0 4 8 12 16 20
Time (min) 65
Solids temperature (°C)
50 0,5h h 1,5h Predicted cooling profiles for the
18 cm bed height
40

50% changes in the solids-air heat

30
transfer coefficient have little effect
20
50 0,5k k 1,5k
Solids temperature (°C)

40 50% changes in the solids-air mass

(water) transfer coefficient have little effect
30

20
50 0,5Fg Fg 1,5Fg
Solids temperature (°C)

40
50% changes in the mass flow rate
of air have a large effect
30

20
0 4 8 12 16 20
Time (min) 66
Solids temperature (°C)
50 0,5h h 1,5h

40

30
Maybe the solids-air heat and mass
20
transfer coefficients are already very
50 0,5k k 1,5k high, such that the solids and air are
Solids temperature (°C)

in moisture and thermal equilibrium?

40

30

20
50 0,5Fg Fg 1,5Fg
Solids temperature (°C)

40

30

20
0 4 8 12 16 20
Time (min) 67
 We tried a DIFFERENT mathematical model, one that assumes
that, at each bed height...
• ...the air heats up to the temperature of the solids at that bed
height, and...
• ...water evaporates from the solids to saturate the air

50
Solids temperature (°C)

40
But we could not fit the temperature
profile at 18 cm in the cooling
experiment, even if we changed the
30 value of Fg (the mass flow rate of air)

20
0 4 8 12 16 20
Time (min)

68
 We tried a DIFFERENT mathematical model, one that assumes
that, at each bed height...
• ...the air heats up to the temperature of the solids at that bed
height, and...
• ...water evaporates from the solids to saturate the air

50
Solids temperature (°C)

40 Although changing h But andwe by 50%

k could not fithas
the temperature
little
profile at 18 cm in the cooling
effect on the model predictions, you need to
experiment, even if we changed the
30
use the model that wevaluedeveloped,
of Fg (the you
masscannot
flow rate of air)
simplify it with the assumptions above
20
0 4 8 12 16 20
Time (min)

69
Step 7: Use the model for the intended purpose

We don’t model just for modeling’s sake!

We want a model to guide scale up of pectinase production

How high can the bed be?

scale up

Our pilot
What mass flow
bioreactor
rate of air to use?

70
 The effect of simply
increasing the bed height

1m
40 cm

~10 cm s-1 ~10 cm s-1

saturated at 30 °C saturated at 30 °C
40 40
38 38
36 36
Temperature (°C)

Temperature (°C)
34 temperature 34
32 32
30
profiles plotted
30
28 every 10 cm 28
26 26
24 24
22 22
20 20
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time (h) 71
Time (h)
71
 Increasing the air velocity in
proportion to the bed height

1m
40 cm

~10 cm s-1 ~25 cm s-1

saturated at 30 °C saturated at 30 °C
40 40
38 38
36 36
Temperature (°C)

Temperature (°C)
34 temperature 34
32 32
30 profiles plotted 30
28 every 10 cm 28
26 26
24 24
22 22
20 20
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time (h) Time (h) 72
72
 Not increasing the air velocity
in proportion to the bed height,
but cooling the inlet air
1m
40 cm

~10 cm s-1 ~10 cm s-1

saturated at 30 °C saturated at varying
temperatures
40 40
38 38
36 36
Temperature (°C)

Temperature (°C)
34 temperature 34
32 32
30 profiles plotted 30
28 every 10 cm 28
26 26
24 24
22 22
20 20
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time (h) Time (h)
73
 Both increasing the air velocity
in proportion to the bed height,
and cooling the inlet air
1m
40 cm

~10 cm s-1 ~25 cm s-1

saturated at 30 °C saturated at varying
temperatures
40 40
38 38
36 36
Temperature (°C)

Temperature (°C)
34 temperature 34
32 32
30 profiles plotted 30
28 every 10 cm 28
26 26
24 24
22 22
20 20
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time (h) Time (h)
74
 So, what does this modeling work tell us about how to
scale up packed-bed bioreactors?

The best strategy to control bed temperatures is to increase the

air flow rate in proportion to bed height
Of
You want to increase the course,
bed height n-fold? you can
also
1m increase
the bed
40 cm width!

You found an air flow rate

(kg m-2 s-1) that gives good
Then increase the air flow
temperature control at small scale?
rate (kg m-2 s-1) n-fold!
75
So, what does this modeling work tell us about how to
scale up packed-bed bioreactors?

It might also be a good idea to cool the inlet air below the
optimum temperature for growth, as necessary, during the
cultivation (but this requires more sophisticated equipment)

measure the
temperature at the
top of the bed

1m
Texcess = Ttop – Toptimum

saturated at varying
temperatures
Tin = Topt – Texcess
76
 The model also explains why many authors report high
bed temperatures even at small scale

Many authors use air flow rates that are simply too low!

40 cm Superficial volumetric air flow rate (m3 s-1)

velocity (m s-1)
=
cross sectional area of bed (m2)

~10 cm s-1 saturated at 30 °C ~2 cm s-1 saturated at 30 °C

44 44

42 42
Temperature (°C)

Temperature (°C)

40 40

38 38

36 36

34 34

32 32

30 30
0 4 8 12 16 20 24 28 0 4 8 12 16 20 24 28 32 36 40 44 48
Time (h) Time (h)
77
We have seen that models are useful tools
A model can be used explore bioreactor performance
This is cheaper than doing experiments
The modeling work does not remove the need to do experiments, but
• it can be useful to explore what appear to be “good ideas”... to
see if they really do have potential to improve performance
• it can be used to identify the most promising cultivation strategies

... but is our model good enough?

Here are some thoughts to consider...

78
How to model the effect of temperature on growth?
In the traditional “isothermal” method, different cultures are incubated
at different temperatures and an empirical equation is fitted to the data

20 25 30 35 40 45
Biomass

time
max

empirical equation
max = f(T) 20 25 30 35 40 45
Temperature
79
...but, as we have seen, due to difficulties in heat removal,
the temperature of the bed reaches values above the
optimum temperature for growth

Temperature optimum temperature

Time

80
 ...so, does the traditional “isothermal” method make sense?

The dashed horizontal lines represent

the temperature profiles experienced
Temperature

by cultures incubated according to the

traditional “isothermal” method.
Although the temperature is the same,
the temperature history is different:
optimum temperature
• initially, the organism has been
Time recently exposed to near-optimal T
• later, the organism has been recently
exposed to high T
max

The underlying specific growth rate

constant would be expected to be
different in the two situations...
...but according to the curve obtained
Temperature by the “traditional isothermal method”,
for the same T, the value of max will be
the same! 81
What is the effect of growth on the isotherm of the solids?
The isotherm of the solids change significantly as some of the
substrate is converted into biomass
• this affects the water content that will give optimal growth
• the equation for the equilibrium moisture content s∗ changes
(s∗ is used to calculate the driving force for evaporation)

4 fungal
Water content (dry basis)

biomass
3
fermented
soybeans
2

1 unfermented
soybeans
0
0.80 0.84 0.88 0.92 0.96 1.0
water activity 82
What are the effects of growth on the solids-air heat and mass
transfer coefficients?

hyphal
growth

solids-air heat and or

mass transfer biofilm
growth
• It is common to determine transfer coefficients for “unfermented”
solids
• Since properties of the solids-air interface change, the heat and
mass transfer coefficients (h and k) can change – but this has
received no attention
• We need to characterize these effects! But how?
83
The take home message from the case study
I have presented a simple case study about one type of
bioreactor
I hope that you are convinced that models can be useful
tools
Similar thought processes and investigations will be
involved in modeling other types of bioreactors
Even if you are not a modeler, you should not try to scale
up your process just by guessing. Do modeling – by
cooperating with someone who has the modeling skills
If you are not a modeler – I hope that the case study will
help you to understand what is involved... so that you can
talk fruitfully with the modeler!
84
Thank you for your
attention!
(Obrigado pela sua
atenção!)

85