Vous êtes sur la page 1sur 13

c=3

Regime analysis and scale-down: tools to


investigate the performance of bioreactors
A. P. J. SWEERE*, K. Ch. A. M. LUYBEN*
and N. W. F. KOSSENt
* Department of Biochemical Engineering, Delft
University of Technology, Julianalaan 67, 2628 B C Delft,
The Netherlands
t Gist-Brocades, P.O. Box 1, 2600 M A Delft, The
Netherlands

Summary. Scale-up and optimization of biotechnological processes on a large scale tend to be more methodically approached
than the application of rules of thumb, experience, and trial and
error. Methods frequently used in chemical engineering are
adopted in biochemical engineering and are employed with great
effect. A summary is given of methods and rules of thumb used in
scaling up chemical processes. A procedure to scale up and optimize bioreactors is presented. It is based on the so-called scaledown approach. Some elements of this procedure, viz. theoretical
regime analysis and small-scale investigations, are extensively
demonstrated by examples. It is shown that a regime analysis
based on characteristic times can give a quick estimation of the
performance of bioreactors. Small-scale experiments based on
the result of such analysis or on the results of a dimensional
analysis can give valuable information for scale-up or optimization fermentation processes.

Keywords: Regime analysis; rules of thumb; scale-up; bioreactor


design

Introduction
Many large-scale fermentation processes give a lower
yield than is expected from laboratory-scale experiments. This is caused by differences in reactor performance at various scales. Laboratory-scale fermenters
can be used with a high power input, resulting in a
rapid mixing of the fermentation broth and a high
mass-transfer rate. Only shear-sensitive systems, like
plant cells and systems with a high viscosity, like mycelium broths, will give mass, heat and momentum transfer
problems on a small scale. On a production-scale, the
power input is restricted for economical and mechanical
reasons, causing mixing and mass and heat transfer
problems. So, it can be concluded that scale-up of fermentation processes introduces these problems. In fact,
not enough is known about the hydrodynamics and the
interaction of the hydrodynamics and other mechanisms in production-scale bioreactors. There are two
ways to solve the problems arising during scaling up:
(a) By acquiring more knowledge about the hydrodynamics and the interaction with other mechanisms in
order to get a complete description of what is happening inside large-scale bioreactors;

386

Enzyme Microb. Technol., 1987, vol. 9, July

(b) By developing scale-up procedures which give an


adequate estimation of the performance of productionscale fermenters based on small-scale investigations.
Much research has been done with respect to the
hydrodynamics of bioreactors. 1 However, the route to
an adequate description of large-scale bioreactors is still
very long, especially as most research is done on a small
scale.
For years scale-up has been more or less an art in
which a lot of experience, rules of thumb, and trial and
error have been used to attain a proper result. At
present, scale-up tends to a more methodical procedure.
Oosterhuis 2 and Kossen and Oosterhuis 3 have reviewed
the different scale-up methods and discussed their
advantages and disadvantages. The methods include:
(1) fundamental method;
(2) semi-fundamental method;
(3) dimensional analysis/regime analysis ;3.9 11.13
(4) rules of thumb ;.-6
(5) scale-down approach/regime analysis; 2'7
(6) trial and error;
(7) multiplication of elements.
The literature gives many rules of thumb which can
be used to scale up processes. Table 1 gives some rules
of thumb from Jordan 4 for the scale-up of chemical
process equipment. From this table the conclusion can
be drawn that, while a constant power per volume ratio
can be used in almost every scale-up problem, an
important exception is mixing. Regarding Table 1 it
must be remarked that rules like ks = constant or
k~a = constant can not be used as such, but have to be
translated to rules based on operational parameters, for
example N and vs .
Einsele 5'6 has surveyed the scale-up rules generally
used in the European fermentation industries (Table 2).
He concluded that these rules are not suitable as the
sole scale-up procedure for microbial growth and production processes under large-scale conditions.
Often, none of the above-mentioned methods is, as
such, adequate to scale up bioreactors, but combinations of different methods can yield good results.
The semi-fundamental method in combination with
rules of thumb is the most widespread method.
However, with a few exceptions the flow models are
based on small-scale experiments. So, scale-up is based
on extrapolation, which makes it very risky.
0141-0229/87/070386-13 $03.00
1987 Butterworth Publishers

Regime analysis a n d s c a l e - d o w n : A. P. J. Sweere et al.


T a b l e 1.

Some rules of thumb to scale up chemical reactors 4

Process
1 Blending of liquids
l a Continuous, low viscosity

Scale-up rule

Remarks

Geometrical similarity

tr, depends on reactor scale;


tr, = constant requires too much energy on a large scale;
a change of T/tr, may cause a change of regime during scale-up

P/V = constant
/tr~ = constant or
l b Continuous, high viscosity

= constant if z > trn


Geometrical similarity
P/V = constant
t m = constant

l c Continuous, non-Newtonian liquid


Dispersion of non-mixable
liquids
3

Suspension of solids

Geometrical similarity

P/V = constant
"L"or 7:/trn= constant
Geometrical similarity

P/V = constant
4

Heat transfer of the liquid


phase

Overall heat transfer

Gas absorption in
mechanically stirred
reactors

7
8

Mass transfer to suspended


particles
Chemical reactors

Fluidized-bed reactors

Only reasonable mixing times can be achieved with helix-shaped


impellers
laminar flow: N x tm= constant (not checked on a large scale)
Only rapid mixing with relatively big impellers (high-energy input
or special shaped impellers, e.g. helix)

Geometrical similarity

P/V = constant or
h = constant or
vt~p = constant
Geometrical similarity
V t i p = constant
AT = constant
Geometrical similarity
k~ a = constant or
vt~o = constant or
a = constant or
P/V = constant or
v, = constant
k s = constant
See blending of liquids

Propeller is most effective; if the agitator has also another


function, e.g. gas dispersion, then another agitator can be chosen
If the resistance to heat transfer of the process is > 8 0 %
of the total resistance, relationship between h and Re number
must be known

U 1/L1/Z(L = characteristic length)


the ratio of volume to heat transfer area is important
k~ a = constant is the best rule; however k. a = f(P/Vvs)

k s is constant if P/V and dp are constant and geometrical


similarity
One has to take into account other characteristic times than r and
tin; a kinetic regime (small- and large-scale) has no scale-up
problem
Development at pilot-plant scale

T a b l e 2 Rules of thumb used as scale-up criteria in the fermentation industry 5 (Reproduced from Margariter, A. and Zajie, J. E.
Biotechnol. Bioeng. 1979, 20, 939-1001, by permission of Wiley
Interscience, New York ~))
Scale-up criteria
Constant
Constant
Constant
Constant

P/V
k~ a
vt~p
Po2

production scale
regime
analysis

% of industries

application

30
30
20
20
simu[afi0n

Dimensional analysis combined with regime analysis


and small-scale experiments is much used to solve scaleup problems, especially mass, heat and momentum
transfer problems.
Scale-down of rate-limiting mechanisms based on the
results of regime analysis is a powerful tool to solve
scale-up problems.
This review examines some elements of scale-down
research, particularly its application to scale-up of fermentation processes, and especially regime analysis
based on characteristic times and the scale-down of
rate-limiting mechanisms.

Scale-down procedure
Based on regime analysis
Pace 7 proposed a method to scale up bioreactors
using experiments on a laboratory-scale or pilot-plant
scale with the same behaviour as the full-scale plant.
Oosterhuis 2 combined this method with regime analysis
and applied the method to the optimization of the glu-

~--

optimization
m0del[ing

laboratory scale
Figure 1

Scale-down procedure

conic acid fermentation. Figure 1 shows the scale-down


procedure as proposed by Oosterhuis 2. In this procedure, four parts can be distinguished: (1) regime
analysis of the process at production scale; (2) simulation of the rate-limiting mechanisms at laboratory scale;
(3) optimization and modelling of the process at laboratory scale; (4) optimization of the process at production scale by translation of the optimized laboratory
conditions. These four parts can be elaborated as
follows:
(1) Regime analysis must give answers to several
questions. Which mechanisms are rate limiting, in other
words which is the ruling regime? Is the regime ruled by
one mechanism (pure regime) or more mechanisms
(mixed regime)? Will there be a change in regime?
The analysis has to allow for changes of scale,
changes in process parameters and the course of the
process. An important factor in the performance of the

Enzyme Microb.

Technol.,

1 9 8 7 , v o l . 9, J u l y

387

Review
microorganisms is whether the existing regime will be
characterized by substrate or nutrient limitation, fluctuations in the environment of the cells, e.g. in the concentration of a component, in the pH, in the
temperature or shear rate, thus affecting the yield of
biomass or products.
From the knowledge originating from regime
analysis it can be concluded which mechanisms or features need further investigation on a small scale.
(2) The most important requirements for experiments
on a laboratory scale is that they have to be representative of the conditions applying on a production scale.
This determines the possibilities and the limits of smallscale experiments (see below).
(3) Optimization of the process on a laboratory scale
and modelling of the investigated features form the
third part of the scale-down procedure. In optimization
one has to keep in mind that the optimized situation
has to be translated to the production scale. Consequently, not all the results of optimization can be used.
With regard to the influence of fluctuations, the following remarks can be made. Fluctuations tend to
increase during scale-up (decreased mixing), resulting in
transient conditions for the microorganisms.
Much is known about modelling of balanced growth
and product formation. However, little is known about
growth and product formation under transient conditions. Barford 8 has reviewed some literature about
modelling transients and lag phases. A distinction must
be made between empirical and mechanistic models.
Most of the models described in the literature are
empirical, incapable of prediction and add little to the
fundamental understanding of microbial dynamics. Few
mechanistic models can be found in the literature. This
is due to the fact that the biochemical and physiological
information, which forms the basis of the mechanistic
models, is unsufficient. Therefore, experimental procedures are still very important.
(4) In the fourth step, the optimized laboratory conditions are translated to a production scale. Models
formed in the previous step can be used for this
purpose. Rules used to scale down the process can now
be used in scaling up the experimental conditions. The
success of the scale-up depends on whether one has succeeded in designing representative scale-down experiments.

Based on process analysis


In fact, the scale-down step, i.e. the step from regime
analysis to experimental simulation, is based on constant characteristic times. As such, it is only a variation
of the scale-down based on constant dimensionless
numbers. There are two main reasons for the use of
characteristic parameters instead of dimensionless
numbers.
First, for biotechnological processes much is
unknown about the behaviour of microbial cultures
under large-scale (transient) conditions. This makes it
impossible to express all mechanisms in dimensionless
numbers, while it is often possible to make an estimation of, for example, the relaxation time of mechanisms

(Figure 2).
Second, the reaction rate is often one of the ratelimiting mechanisms in fermentation processes. In
scaling down, however, the reaction is mostly conceived

388

Enzyme Microb. Technol., 1987, vol. 9, July

10-6 10'~,
I
I
Mass
action Caw
IL

10-2 100
I

10 2

10 t~ 106

characteristic

time

(s}

AItosteric controls
"~

Changes in enzymic
'=concentrations

m-RNA

Selection

control

Mixing probtems
(fed-)bafch
dynamics,CC-transients
,,m

Figure 2 Relaxationtimes of mechanismsinside microorganisms


and of their environment in a bioreactor

mechanistic I
analysis

dimensional

analysis

~I

regime

analysis

l yes
experimental
design

rules of
thumb
Figure

literature
data and
correlations

experience

3 Interactionof the methods used in process analysis

as an established fact. For instance, changes in pressure


or temperature in order to influence the reaction rate, as
used in chemical engineering, are hardly ever used in
biochemical engineering. This means that the advantage
of using dimensionless numbers, i.e. by changing the
rate of mechanisms without changing the ratio of their
rates, disappears if the reaction rate is present in the
dimensionless numbers.
In conclusion, if the behaviour of the microbial
culture is not the bottleneck of the process, an analysis
of engineering problems may be based on dimensionless
numbers. It must be remarked, however, that in engineering problems a complete description by means of
dimensionless numbers may be impossible.
This results in an extension of the scheme presented
in Figure 1. The first step in the scale-down procedure
not only includes regime analysis, but also dimensional
analysis, mechanistic analysis and the similarity principle. Therefore, it is better described as process
analysis. Figure 3 shows the interactions of the mechanisms used in the analysis of the process. Complementary knowledge may be supplied from rules of thumb,
literature data and correlations and experience.
Process analysis (Figure 3) can be seen as a systematic method, not restricted to regime analysis alone, to

Regime analysis and scale-down: A. P. J. Sweere et al.


analyse the performance of a large-scale process,
resulting in an experimental design for small-scale simulation of large-scale conditions.
Dimensional analysis is a widespread tool in chemical engineering and has been extensively treated in the
literature.a,9,10 12 The potentials and limitations of this
method have been discussed by Kossen and
Oosterhuis 2 and Zlokarnik. la
While dimensional analysis starts from a list of relevant parameters, mechanistic analysis is based on a list
of involved mechanisms. The rate of these mechanisms
has to be expressed in characteristic parameters, like
fluxes, pressures, heights or times. Further analysis may
be based either on the characteristic parameters or on
the ratio of these parameters, i.e. dimensionless
numbers.
To obtain the same behaviour of systems on different
scales the systems have to be similar. Four similarity
states are important in chemical engineering: geometrical, mechanical, thermal and chemical, each state necessitating the previous ones. However, in engineering
problems it is, on the whole, impossible to satisfy the
similarity demand, resulting in the fact that neither all
the characteristic parameters nor all the dimensionless
numbers can be kept constant during scale-up or scaledown. So it has to be decided which mechanisms are the
most important of the system under investigation: in
other words, what is the ruling regime?
Dimensional analysis always results in geometrical
similarity. In mechanistic analysis, geometrical similarity is not necessary, which offers a significant advantage. Regime analysis may however, reject the
geometrical similarity which resulted from dimensional
analysis.

Regime analysis based on characteristic


times
Introduction
The regime concept was introduced by Johnstone
and Thring 9 and is a necessary ingredient for the solution of engineering problems by means of dimensional
analysis based on the similarity principle.
In fact, the regime concept is very similar to analyses
based on the rate-controlling step used in chemical
engineering. Although not demonstrated for scale-up
purposes, the use of microbalances in combination with
the rate-controlling step is extensively treated in literature. 14 This approach is mainly used for mass transfer
with chemical reaction in gas-solid and liquid-solid
systems, liquid-liquid and gas-liquid systems, liquidliquid-solid and gas-liquid-solid systems, and systems
with solid-phase catalysis. In some cases it can also be
used as the method in scaling up (fundamental method).
As mentioned above, regime analysis must provide
information about the performance of the process. This
is based on a comparison of the rate of the mechanisms
which may play a role in the process, and the comparison can be made experimentally or theoretically, qualitatively or quantitatively.
Examples of experimental regime analyses have been
given by LevenspieP 4 and Kossen and Oosterhuis. a
Also, the method presented by Moes 15 to analyse the
liquid flow in (large-scale) fermenters can be conceived
as experimental regime analysis.

Experimental regime analysis cannot answer all the


questions which have to be solved, e.g. whether change
of regime will occur during scale-up. This question can
be solved by theoretical regime analysis.
Theoretical regime analysis can be subdivided into
regime analysis based on characteristic parameters (e.g.
characteristic times) and parameter sensitivity analysis.
Here, only regime analysis based on characteristic times
will be treated. It is based on a comparison of the rates
of different mechanisms, expressed in characteristic
parameters which can often be obtained accurately
enough by the use of rules of thumb. A comparison of
these characteristic parameters will yield the ratelimiting mechanisms, or regime of the process. In this
review, regime analysis based on characteristic times
will be demonstrated by means of some examples.

Characteristic times
In comparing rate processes, time is the most convenient characteristic parameter. For other processes,
stresses, heights (e.g. H E T P , the height equivalent to a
theoretical plate), or other parameters may give a better
insight into the problem. ~'4"16.1v
Characteristic time is a measure of the rate of a
mechanism and can be considered as the time needed
by that mechanism to smooth out a change to a certain
fraction. A low value of a characteristic time means a
fast mechanism; a high value means a slow mechanism.
In literature terms like time constant, process time
(constant) and relaxation time are also used. The term
time constant is commonly used for first-order processes only, and is equal to the time needed for a
mechanism to proceed to 63% conversion. The time
constant of a process can be composed from the time
constants of several mechanisms. To prevent confusion,
the term characteristic time is introduced to characterize the rate of mechanisms. Using the definition for the
characteristic time given above it is possible to characterize non-linear processes and processes of a higher
order with only one characteristic time. An example is
the characterization of liquid mixing in fermenters by
means of a mixing time. Characteristic times can be
determined experimentally or theoretically. Examples of
experimental determination of characteristic times are
liquid mixing and liquid circulation times in mechanically stirred fermenters. 2Ja
Theoretical values of characteristic times may be
determined in several ways (Table 3). The characteristic
time based on the ratio of a capacity and a flow seems a
very useful definition to make a rapid estimation of the
rate of various mechanisms. Parameters like dispersion
coefficients and mass transfer coefficients have to be

Table 3

Theoretical methods to determine characteristic times

Method

Example

1 Rules o f t h u m b and

t m = 4tc~ r
tc~r = V / ( 2 H r

literature correlations
2 Differential equations:
(a) Mass, heat and m o m e n t u m
balance
(b) A c c u m u l a t i o n by one
mechanism o n l y

E) d2C '
L 2 dx '2

N~2D 2)

v de'
L dx'

r
C

dC
d2C
dt - [~ dx 2

(Co.g/m - Coj )

3 Ratio o f capacity and f l o w


tmt J --

k, a(Co.g/m - Co.,)

Enzyme Microb. Technol., 1987, vol. 9, July

389

Review
estimated by correlations from the literature or from
small-scale experiments. If the parameters are not
known from the literature, the reaction kinetics also
have to be measured. The characteristic times from differential equations seem to be more accurate, but
require solution of the equations and have the drawback that they are based on an estimation of the parameters and a simplified model.
It is clear from the above that the characteristic times
of physical mechanisms in bioreactors can be estimated
fairly easily by correlations found in the literature.
However, the performance of a bioreactor may also be
ruled by physiological mechanisms. Characteristic times
of substrate consumption or product formation can be
calculated by means of the integrated MichaelisMenten kinetics for batch growth:
dCs
--

dt

#max CsCx
--

(1)

Y~x K~+Cs

or by the ratio of capacity and flow:


tsc -

Cs
rs

Y~,

- -

~maxCx

( K s + C,)

(2)

Michaelis-Menten kinetics only give a description of


balanced growth. If they are used to give an estimation
of the characteristic times under dynamic conditions
one has to be aware of the deviation Michaelis-Menten
kinetics may give, especially at low substrate concentrations. Under dynamic conditions, mechanisms inside
the cells (Figure 2) may change the performance of
microbial cultures, resulting in a behaviour distinct
from balanced growth. F r o m a comparison of characteristic times of mechanisms inside microorganisms and
of mechanisms in their environment it can be concluded
that there are mechanisms inside microorganisms with
characteristic times of the same order of magnitude as
those of changes in the environment to which they will
be exposed. So, the behaviour of the microorganisms
may change during the process, thus influencing growth
and product formation.
There are several examples of the use of characteristic times.

Gluconic acid production


Oosterhuis 2 used gluconic acid production by Gluconobacter oxydans as a model system to study the
influence of large-scale conditions upon a microbial
system. The study was aimed at the optimization of gluconic acid production using the scale-down method
(Figure 1). Gluconic acid is produced in an aerobic
batch process at a scale of ~ 2 5 m 3. The fermenter is
equipped with two Rushton-type impellers, a starshaped sparger just below the lowest impeller and four
baffles. Table 4 gives the relative dimensions of the
production-scale fermenter. To get an impression of the
rate-limiting steps, regime analysis was carried out.
Table 5 gives the characteristic times of the different
mechanisms which play a role in the process. The following conclusions can be drawn from the table. Substrate consumption and growth have no influence on
the performance of the process. The times for oxygen
consumption and oxygen transfer to the liquid phase
are of the same order of magnitude. Therefore, oxygen
limitations may occur. Also, the liquid circulation time

390

Enzyme Microb.

Technol.,

1 9 8 7 , v o l . 9, J u l y

T a b l e 4 Dimensions of production-scale fermenter for gluconic


acid production 2
Impeller/vessel diameter
Number of impellers
Impeller blade width/impeller diameter
Impeller speed
Baffle diameter/vessel diameter
Liquid height/vessel diameter
Gas flow (reactor volume time)
Liquid volume

0.32
2
0.2
1.3 or 2.6 litre s -1
0.09
up to 1.8
up to 0.5 vvm
up to 25 m 3

T a b l e 5 Characteristic times (s) of the mechanisms which are


important in gluconic acid production 2

Transport phenomena
Oxygen transfer
Circulation of the liquid
Gas residence
Transfer of oxygen from a
gas bubble
Heat transfer
Conversion
Oxygen consumption, zero order
first order
Substrate consumption
Growth
Heat production

5.5 (non-coal)-11.2 (coal)


12.3
20.6
290 (non-coal)-593 (coal)
330-650
16
0.7
5.5 x 104
1.2 x 104
350

is of the same order of magnitude, so the conclusion can


be drawn that oxygen gradients are likely to occur.
F r o m a comparison of the gas residence time and the
time for oxygen transfer from the gas phase it is clear
that no exhaustion of the gas phase will occur. Heat
transfer will balance heat production. F r o m a comparison with the liquid circulation time it can be concluded
that no temperature gradient will be present in the fermenter.
It has to be remembered that the correlations used
only give a rough estimation of the characteristic times,
so only the order of magnitude can be considered and
compared.
This regime analysis has formed the basis of smallscale investigations of the influence of fluctuating
oxygen concentrations on gluconic acid production and
in mass transfer studies.

Design of an installation for the


microbial desulphurization of coal
Huber 19 and Bos z used regime analysis based on
characteristic times to design an installation for the
microbial desulphurization of coal. For optimal design
of the reactor two major conditions must be met.
Biomass is the catalyst for the oxidation of pyrite to
sulphate, so that biomass limitation must be prevented.
The environment of the microorganisms must be
optimal for pyrite oxidation.
In relation to the first point, Huber 19 has stated that
a reactor configuration consisting of a mixed-flow
reactor followed by a plug-flow reactor would be most
adequate, because of the first-order pyrite oxidation
process. ~4 For this plug-flow reactor, a series of
Pachuka tanks was suggested. A plug-flow reactor as
such would result in washout of biomass. Therefore, an
intensively mixed fermenter has to be used in front of it
to generate an effective inoculum. Limitation of biomass
can then be avoided if the residence time in this wellstirred fermenter is greater than the characteristic time

Regime analysis and scale-down: A. P. J. Sweere et al.


Table 6 Characteristic times (s) relevant to the microbial desulphurization of coaP 9,20

Liquid mixing
Oxygen transfer
Oxygen consumption
Settling of the particles
Growth

60-118
80-641
3.4 x 105
2.5 x 104
8.6 x 104

Configuration
Capacity
Gas velocity
Diameter pachuka tank reactor
Height pachuka tank reactor
Pyrite content
Pyrite removal

106

--

.s
s

,.~

.~

/.
. ~ . .

__

s
p

"~

~.~. ""

s
s

T a b l e 7 Specification of the installation for the microbial desulphurization of coal ~9,2o

I
s

tp

.,,..~

Cascade of 10 pachuka
tank reactors
100 000 ton y 0.02-0.003 m s -1
10 m
20 m
0.5%
90%

102-

of biomass growth:
z > tx(= 1//~)

Optimal conditions for pyrite oxidation demand that


depletion of oxygen and settling of the coal particles has
to be prevented throughout the fermenter. To achieve
this, three conditions have to be met:
no overall oxygen depletion:
(4)

tmt,1 < toc

no gradients in oxygen concentration:


tm, 1 < trot,1

(5)

no sedimentation of the coal:


tm, I <

L-~

(3)

(6)

tsett

If the kinetics of the desulphurization reaction are


known the oxygen demand can be calculatedJ 9 Based
on the required capacity of the installation and the conditions in equations (3)-(6) the reactor can be designed.
Table 6 gives the values of the characteristic times.
From these values it can be concluded that mass transfer determines the performance of the reactor. The time
for oxygen consumption is much greater than the time
for oxygen transfer, so no oxygen limitation will occur.
Also, oxygen gradients and sedimentation of the coal
particles will not be a problem. The analysis has
resulted in the design of an installation for the desulphurization of coal with the specifications given in

Table 7.

I
0.01

I
0.1

i
i
1
10
reacf0r volume (m 3)

i
1 00

Figure 4 Estimated values of the characteristic times in the


fluidized-bed reactor: 21 ta,=axial dispersion time; to,re = liquid
circulation time; tp = product formation time

Design of an IBE production process


Schoutens 21 has studied the design of a large-scale
process for the production of butanol from whey permeatc with immobilized Clostridium cells. For the
design of scale models of a gas-lift loop reactor (GLR)
and a fluidized-bed reactor (FBR) with liquid circulation, regime analysis of a large-scale reactor was carried
out. The characteristic times of liquid mixing, liquid circulation and product formation are given in Table 8.
From experiments in a continuous stirred-flow reactor
(CSTR) it was concluded that diffusion and reaction
inside the beads were of no importance to regime
analysis.
The criterion for designing a 10 litre FBR was a
comparable degree of mixing at a 50 m a and 10 litr
scale. Because production time is independent of scale,
this would result in the same concentration gradients.
For the 50 m 3 fermenter a height-to-diameter ratio
(H/D) of 3 was chosen. Figure 4 shows the characteristic
times as a function of reactor volume. To get the same
ratio of tciJtm, l an HID ratio of 25 has to be chosen for

Table 8 Specifications and characteristic times of a fluidized-bed reactor (FBR) with


recycle and t w o gas-lift loop reactors (GLR) for continuous IBE fermentation 21
Parameter

FBR

GLR1

GLR2

Volume (m 3)
Bead fraction (-)

50

65

0.010

0.45
0.3-0.5

0.35
0.2-0.4

0.35
0.2-0.4

5 x 10 -3

0.5-1.0

0.2-0.3

-(7.2-12.0) 103
600-1200

(1.0-3.0) x 10 -2
(9.0-18.0) x 103
20-40

(1.0-3.0) x 10 -2

(9.0-18.0) x 103
5-20

(6.~69.0) x 105

700-1000

60-500

Enzyme Microb.

Technol.,

Dilution rate (h -1)


superficial
Liquid velocity (m s - I )
superficial
Gas velocity (m s -1)
Production time (s)
Liquid circulation
time (s)
Axial dispersion
time (s)

1 9 8 7 , v o l . 9, J u l y

391

Review

the model system. The analysis resulted in the design of


an FBR with HID = 25, and H = 2 m. To simulate
large-scale gas production, a gas inlet was constructed.
F r o m the characteristic times of the 65 m 3 G L R
(Table 8) it can be concluded that a change of the
hydrodynamic regime is unlikely when scaling down to
laboratory scale. The one criterion that determines the
design of the model G L R is to avoid wall effects of the
bubbles and beads. Finally, a G L R with external loop
was chosen with V = 15 litre, H = 1 m and D r = D d =
0.08 m.

Bakers' yeast production


In designing or optimizing a fermentation process,
information can be gained from regime analysis if the
characteristic times are calculated as a function of
process parameters such as stirrer speed or gas flow
rate. These calculations can give information about
changing regimes during optimization or scale-up, or as
the process proceeds.
In a study of the influence of a continuously changing environment on growth and metabolite production
by microorganisms, fed-batch bakers' yeast production
has been chosen as a model system. 22
Bakers' yeast is sensitive to changes in glucose concentration (Crabtree effect) and changes in oxygen concentration (Pasteur effect). A high glucose concentration
or a low oxygen concentration will result in excretion of
ethanol by the yeast. Under anaerobic conditions, glycerol is produced and transient conditions may lead to
the production of glycerol, acetic acid, succinic acid,
pyruvic acid and other products. The production of
these metabolites, even if this is followed by their consumption in a later part of the process, 23 causes a
reduction of biomass yield. It is therefore important to
know whether limitations or gradients of oxygen, substrate or nutrients will occur.
Regime analysis is carried out for a bubble column
fermenter of 150 m 3 with an operational volume of
120 m 3. Air is supplied by means of perforated horizontal pipes near the bottom of the column.
Tables 9a and 9b give the characteristic times of the
mechanisms which may play a role in the performance
of the fed-batch fermentation. A distinction has been
made between mechanisms which have a clear effect on
the process and mechanisms which have a characteristic
time which is much smaller or much larger. The equations used to calculate the characteristic times are given
in Appendix A and the parameters used in these calculations are given in Table AI.
Figure 5 shows the characteristic times of liquid
mixing and circulation, mass transfer and oxygen consumption as a function of the superfical gas flow rate in
a bubble column. These graphs have been calculated for
the final situation in fed-batch production at a volume
of 120 m 3 and a height-to-diameter ratio of 4. Such a
plot allows easy comparison of the rates of the various
mechanisms. The time needed for oxygen transfer to the
liquid phase is much longer than that for oxygen consumption, so oxygen depletion will occur in the fermenter. If this is combined with a high mixing time for
the liquid phase and a zone with a high oxygen transfer
rate, for example near the sparger, then oxygen gradients are likely to occur.
The influence of the height-to-diameter ratio, H/D,

392

Enzyme Microb. T e c h n o l . , 1 9 8 7 , vol. 9, J u l y

Table 9 (a) Characteristic times of the relevant mechanisms in


bakers' yeast production in a bubble column fermenter
Mechanism
Mixing
Liquid

Definition

phase

Magnitude (s)

t m j = L 2 / D E,,

Gas phase
Liquid circulation
Gas flow
Oxygen transfer
Liquid phase
Gas phase
Substrate consumption
Oxygen consumption
Substrate addition

tm,g =

101 - 10 3

L 2/D E.o

1 O- 1 _ 103
101 - 102
1 - 102

to, r = 2L/vc~ r
tQ = (1 - c)L/v,
trn j = 1/k~ a
tmt.g = cm/(k~ a(1 - ~))
tsc = C,/r~
toc = Co,Jr o
t,a = V. Cd() v . C,o )

1 -- 1 0 2
103 - 106
101 - 102

1
101 - 102

(b) Characteristic times of yeast production which are not relevant to the final performance of the reactor

Mechanism

Definition

Magnitude

Fed-batch process

tp
t X = 1/ll
tht = V Pl ' c . / ( U A . A T )
thp = V ' P l ' C . / ( r h + P / V )

10 5
10 4
10 3
10 3

Diffusion

t o (see A p p e n d i x A )

Turbulent erosion

tte (see A p p e n d i x A )
t~s (see A p p e n d i x A )

10-2-1
1-10 2
10 -3

G r o w t h of biomass
Heat transfer
Heat production
Micro-mixing

Laminar stretching

103

102_ ~

u
E

liquid mixing

circulation

101- ~

"

~d

4-

~ . . . . . . ~ s s transfer

t..

100i

10 -1

oxygen consumption

50

100

1 SO

200

250

300

350

superficia[ gas vet0cify (mm/s)


Figure 5 Estimated values of the characteristic times of fedbatch bakers" yeast production as a function of the superficial gas
velocity in a 120 m3 bubble column reactor with a height-todiameter ratio of 4

on characteristic times is shown in Figure 6. If H/D is


< 0.3 or > 3 the reactor performance is governed by the
liquid mixing: a pure regime. If H/D is >0.3 and <3,
the reactor performance is governed by both liquid
mixing and mass transfer: a mixed regime. At HID = 1
a discontinuity occurs due to the fact that in the correlations used (see Appendix A) the smallest values of
height and diameter are used. This is based on the fact

Regime analysis and scale-down: A. P. J. Sweere et al.


10 ~

liquid mixin(.

10 3 10 2.
I

10 2-

liquid mixing

101
mass t r a n s f e r

liquid circulation

QJ
E
.u

"-~ 10 0
L-

t_

oxygen consumption

substrafe consumption
subsfrafe addition

~.J
to

L-

qO
0

101

height diameter ratio


Figure 6 Estimated values of the characteristic times of fedbatch bakers' yeast production as a function of the heightto-diameter ratio of a bubble column reactor (V= 120 m3; ~g=
1 vvm)

10

time (h}

--~m.~

that macro-mixing of the liquid phase in the reactor is


determined by the smallest dimension of the column.
The characteristic times for mass transfer and oxygen
consumption are independent of the height of the
column. From the ratio of these times the conclusion
can be drawn that oxygen limitation will always occur.
A similar analysis can be made for substrate concentration in the fermenter (Figure 7). This figure shows
some characteristic times as a function of batch time. In
these calculations the gas flow rate was 5 cm s-1 and
the volume increased from 80 to 120 m 3 in 10 h. Liquid
circulation time and liquid mixing time increase, due to
the volume increase. The times for substrate addition
and substrate consumption decrease, due to an increase
in the biomass concentration. From the ratio of liquid
mixing time and the liquid circulation time to substrate
addition time and substrate consumption time it can be
concluded that substrate gradients will occur and that
the intensity of these substrate gradients increases
during the fermentation. A characteristic of a fed-batch
process is the limitation of the added component. From
the ratio of substrate consumption and addition time it
can be concluded that substrate limitation will probably
occur.
In conclusion it can be said that the following
aspects need further investigation: liquid mixing; mass
transfer; and the influence of periodic changes on
microorganisms.
Much research has been done with respect to liquid
mixing and mass transfer in bubble column reactors.
This has resulted in models describing oxygen transfer
and liquid mixing in laboratory-scale fermenters. More
research is needed on the performance of large-scale fermenters, e.g. on liquid mixing patterns and concentra-

12

Figure 7 Estimated values of the characteristic times of fedbatch bakers" yeast production as a function of batch time in a
bubble column reactor (V = 80-120 m3; at V = 120 m 3, H/D = 4,
v, = 0.05 m s- 1 )

tion profiles. However, large-scale experiments are


needed for some of these investigations. The third
aspect which needs further investigation is well suited to
laboratory experimentation, provided that the experimental set-up creates the same conditions as those met
on a production scale.

Conclusions and discussion


From the examples presented above, it can be concluded that regime analysis is a suitable tool in the optimization, design and scale-up of bioreactors. The
analysis can also play an important role in the experimental set-up. Regime analysis can be based on various
characteristic quantities, but in the analysis of bioreactors the characteristic time seems to be an optimal
choice. This is due to the fact that the mechanisms compared are very diverse and because the theoretically
determined characteristic times can be easily compared
with experimentally determined times.
A quantitative regime analysis is only possible if the
relationships between the operational parameters and
the dependent parameters are available to calculate the
characteristic times of the mechanisms involved. Otherwise, knowledge about the system has to be completed
by experimental research, often on a small scale, or
necessitates working with qualitative analysis. Knowledge of the reaction kinetics is especially important and
has to be measured, in general.
Optimization provides the advantages of an existing
process, which opens the possibility of large-scale
experiments. However, production-scale experiments
are very time-consuming and expensive and do not
always lead to the desired results. This may be caused

Enzyme M i c r o b . T e c h n o l . , 1 9 8 7 , vol. 9, J u l y

393

Review
by the fact that the methods of measurement designed
or tested on a small scale are not suitable on a large
scale and have to be adjusted. Consequently, theoretical
analysis of the process is also very important.
Despite the fact that regime analysis seems very
promising and the fact that the results seem very realistic, the method has to be proven further by experimental verification at a production scale. Oosterhuis z
predicted oxygen limitation and oxygen gradients from
the results of regime analysis of gluconic acid production. The oxygen gradients and local oxygen depletion were confirmed by measuring dissolved oxygen
profiles during fermentation in a 25 m 3 bioreactor. Prediction of circulation time was checked by means of
flow-follower (radio-pill) experiments.
The results of regime analyses used to design installations for the desulphurization of coal and to design a
reactor for the IBE fermentation can only be verified
after an installation of the predicted scale has been
built.
In the case of the bakers' yeast production, experimental verification of the conclusions on a large scale
from regime analysis and of the calculated characteristic
times has not yet been possible.

Experimental simulation of large-scale conditions


Introduction

Process analysis leads to design of small-scale experiments for the simulation of large-scale conditions. This
may result in scale models of the large-scale fermenter
or segment models. It is clear that scale models are
aimed at an overall study of the reactor performance
whereas segment models are aimed at a more thorough
investigation of parts of the process. Examples of both
kinds of models are treated below.
Hydrodynamics and cell physiology are both characterized by a lack of sufficient knowledge, leading to
problems in scaling up bioreactors.
If it is assumed that the biomass cannot be changed
or replaced, then most problems are connected with the
hydrodynamics of the system and the interaction
between the hydrodynamics and the physiology of the
biomass. To obtain relevant information for the solution of these problems the different subjects and their
interaction need further investigation.
Scale-down based on dimensionless numbers

As mentioned above, dimensional analysis is often


used to solve scale-up problems. If all dimensionless
numbers can be kept constant, small-scale experiments
are easy to design and translation of their results is
uncomplicated. However, this is an ideal situation.
Usually, not all dimensionless numbers can be kept
constant unless unusual experimental set-ups are used
or the number of dimensionless numbers is reduced by
application of regime analysis. Let us look at some
examples, though not all are based on bioreactor
design.
In the scale-down of gas-liquid flow through pipes,
six dimensionless numbers (Re, Fr, We, Eu, p', r/') have
to be kept constant, and the systems have to be geometrically similar. 24 The dimensionless numbers can be
derived from dimensional analysis or from the dimensionless Navier-Stokes equation, the equation of con-

394

Enzyme Microb. Technol., 1987, vol. 9, July

tinuity and the state equation. If the gravitational


acceleration is constant, a scale factor of only three can
be achieved. However, if a rotating experimental set-up
is used, a scale factor of 10 can be reached. 24
A nice example of the scale-down procedure and the
power of small-scale experiments is the investigation of
Bolle et al. 25 into the performance of a large-scale
(1340m 3) upflow anaerobic sludge blanket (UASB)
reactor.
The content of a UASB reactor can be mixed by the
gas produced, provided that the influent is well distributed over the sludge bed. Results obtained from experiments with a 800 m 3 UASB reactor have shown that an
influent distribution system with one inlet point per
10 m 2, placed near the bottom of the reactor, in combination with a constant injection velocity is insufficient. This set-up resulted in large dead-spaces and large
short-circuiting fows. To study these phenomena, a
scale model was designed and built. The model design
was based on a dimensional analysis resulting in 14
dimensionless numbers. As expected, not all numbers
can be kept constant during scale-down. Since most
problems with full-scale reactors originate from problems with the distribution of the influent in the sludgebed, most attention was paid to the behaviour of the
liquid flow in the sludge bed. In particular, the modified
Froude number had to be kept constant during scaledown.
Based on the assumption Amode~/A
. . . . t o r = 1/400, this
analysis resulted in the design of a scale model with the
dimensions: H = 0.26 m; L = 0.8 m; W = 0.6 m; dv =
0.1 mm. Besides this scale model, a model segment was
used to study the sludge bed around an influent inlet
point. Small-scale experiments were carried out with a
continuous influent flow, on/off switching of the influent
flow and minimum/maximum switching of the influent
flow.
Based on the results of these experiments and on a
substrate consumption time of 4.5 min on a full scale, a
scheme was proposed of 2 min maximum flow and
2 min minimum flow. Stable short-circuiting flows were
avoided using these switching times. These results were
verified by means of large-scale experiments from which
it was concluded that min/max switching of the inlet
flow gave negligibly small short-circuiting flows and no
dead spaces. In contrast, on/off switching resulted in
less efficient use of the sludge bed. Continuous influent
supply led to overloading and washout of the sludge
particles.
Scale-down based on characteristic times

Regime analysis based on characteristic times gives


information about the rate-limiting mechanisms of the
process and the presence of nutrient limitations and
concentration profiles. From this, it can be concluded
what complementary knowledge about the system and
what experiments are needed to optimize the process.
From regime analysis of gluconic acid production
(see above) Oosterhuis concluded 2 that the cells were
exposed to a continually changing oxygen concentration. Scale-down was based on constant characteristic
times for liquid circulation and oxygen consumption.
Two 'segment' models were used for experimental simulation. One model consisted of a batch fermenter with a
periodically changing gas inlet concentration. The inlet

Regime analysis and scale-down: A. P. J. Sweere et al.


Table 10 Comparison of a 50 m 3 fluidized-bed reactor (FBR)
and a 65 m 3 gas-lift loop reactor (GLR) 21

Table 11 Characteristic times of fluctuations to which microorganisms are exposed on a large-scale and small-scale

Property

FB R

G LR

Investigation

Time (s)

Remarks (ref)

Reactor productivity
Biocatalyst attrition rate
Reactor construction

+/-

Domestic sewage treatment

105

Construction bottom plate

+
-

+/+

Ease of operation

Repeated fed-batch
Mixing problems
Dropwise feeding in
continuous culture
Pulse feeding
Two-fermenter system
Stimulus response analysis
Step signal

104-105
10-102
1-102

Day and night rhythms


in feed supply and
temperature (48)
(49)
(2, 5, 41, 42, 46, 47)
(38, 39)

Foaming

Gas release
Combination with simultaneous
recovery processes

+
+

+, Favourable;

-,

unfavourable or attention needed

gas flow changed from air (20 s) to nitrogen gas (60180 s). Not only did a reduction in productivity occur
due to the anaerobic periods but negative effect on the
potential capacity to produce gluconic acid was also
observed.
The other model was a configuration with two fermenters and an exchange flow between them. One fermenter was sparged with air and one with nitrogen gas.
While circulating through the aerobic and anaerobic
fermenter, no reduction of the potential capacity of the
cells to produce gluconic acid was observed. The
reduction of overall productivity could be related to the
fraction of the volume which was anaerobic.
Regime analysis of the IBE process 21 resulted in the
design of two scale-models: a 10 litre fluidized-bed
reactor (FBR) with recycle and a 15 litre gas-lift loop
reactor (GLR). The design of these models was based on
constant characteristic times of production, and a constant ratio of the times of liquid circulation and axial
dispersion of the liquid (see section above).
Hydrodynamic studies resulted in liquid-mixing
models consisting of a 10 tanks in series model (FBR)
and a 100 tanks in series model (GLR). Combination of
the FBR mixing model, a plug-flow model with recycle
and a CSTR model with a kinetic model, shown that
this overall model only had low sensitivity for the
hydrodynamic parameters. However, experimental comparison of the FBR and G L R showed a better performance of the FBR, leading to 22% higher
production rate per unit reactor volume and a slightly
higher outlet product concentration. These differences
were caused by the higher solids hold-up in the FBR
and the plug-flow with recycle character of the liquid
mixing. For these and other reasons (Table 10) the FBR
was preferred for large-scale application in the IBE
process with immobilized Clostridia.

Interaction between cell physiology and


liquid hydrodynamics
Much has been published about the hydrodynamics
of mechanically stirred 26'27 and gas stirred fermenters. 28-33 Little, however, has been published about
the performance of large-scale bioreactors. Therefore,
complementary research has to be carried out on a
large scale unless the conditions can be scaled down to
the laboratory scale. However, the hydrodynamics are
difficult for adequate scale-down. It is possible that
interesting scale-down experiments have been reported
in the literature, but because they cannot be found
under this subject heading they are difficult to find.
It is clear that it will be no problem in the study of
cell physiology, especially growth and metabolite pro-

10
10-102
102-10 s
10z-105

(43)
(2)
(50-52)

(39, 53-55)

duction, at a small scale. However, it is commonly


found that the environment to which the cells are
exposed on a large scale is not similar to that on a small
scale, the more so because periodic fluctuations in the
environment may occur on a large scale.
The conclusion can be drawn that the behaviour of
cells studied on a laboratory scale may be quite different from that on a large scale.
This usually results in a reduction of the yield of
biomass and the required metabolite. It is also possible,
though, that a periodically changing process parameter
results in increased formation of certain metabolites. 34
In these cases, fluctuations have to be optimized for
optimal product yield. The characteristic times of these
fluctuations are usually of another order of magnitude
than those of liquid mixing. So it is not expected that
imperfect mixing will give a significant increase in
product yield.
Much research has been done concerning the
response of microorganisms to fluctuations in their
environment. Reviews have been published by Barford, s
Harrison and Topiwala, 3s Cooney et al. 36 and
Pickett. 37 In investigating the influence of a changing
environment on microorganisms, a distinction has to be
made between a single change (for example impulse or
step signal, batch) and continuous changes. The latter
can be divided into periodical (e.g. sinusoidal, block
signal, repeated fed-batch) and not-periodical (e.g. noise,
fed-batch). Table 11 gives a survey of changes and their
characteristic times to which microorganisms are
exposed in order to study their metabolism. Just a few
studies are concerned with the influence of imperfect
mixing on microbial metabolism. With respect to
mixing problems, transients of a (fed-)batch and a continuous culture are not very relevant, unless there is an
interaction between these transients and the transients
due to imperfect mixing. Therefore, only periodic transients will be treated.
Small-scale fermenters used for physiological investigations are commonly well mixed, so that fluctuations
to simulate mixing effect have to be created artificially.
Brooks and Meers 3s'39 have shown the influence of a
dropwise feeding of a continuous culture: a drop frequency of 2.4 min-1 can reduce the biomass yield by
10%. More features described for microorganisms measured in physiological studies can probably be
explained by imperfect mixing and its effect on the
metabolism of the microorganisms. 4
Different experimental set-ups can be used to simulate the effect of imperfect mixing on a microbial culture

(Table 12).

Enzyme Microb.

Technol.,

1 9 8 7 , v o l . 9, J u l y

395

Review
Table 12 Experimental set-ups which can be used to simulate
the behaviour of large-scale bioreactors
Type

Ref.

Well-mixed fermenter
Coupling of well-mixed fermenters
Plug-flow- reactor with recirculation
Coupling of plug-flow and well-mixed
fermenters
Bubble column reactors
Loop reactors

2, 41, 43-45
2
48, 49

A well-mixed fermenter can be used, but to simulate


bad mixing it has to be used with increased viscosity or
a continually changing parameter.
Vardar and Lilly41 used a cycling top-pressure to
simulate cycling dissolved oxygen tension (DOT) and
hydrostatic pressure gradients in large-scale vessels used
for penicillin production. The top pressure was changed
sinusoidally between 1 and 2 bar with a period of
2 min. This resulted in a cycling D O T between 23%
and 37% air saturation. It was found that the fluctuating D O T led the organisms to a metabolism which corresponded to a lower D O T than the mean value
imposed. Fluctuating conditions did not result in an
irreversible change in metabolic activity.
Sokolov 42 studied the influence of changes in pH
and dissolved oxygen tension on the specific growth
rate of Candida utilis, Saccharomyces cerevisiae and
Candida scottii. During a phase of exponential growth,
cultures of these organisms were exposed to fluctuations
with periods of 2-60 min. No negative influences, but
rather accelerated growth, were observed.
Periodic fluctuations of glucose concentrations in a
continuous culture of bakers' yeast, Saccharomyces cerevisiae, caused a reduction in biomass yield.43 Also, hysteresis was observed in the biomass concentration,
going from low to high dilution rates and back. Hysteresis was also observed by Furukawa, 44 who investigated the influence of oxygen concentration on yeast
metabolism in a continuous culture without any
imposed fluctuations. So, it might be concluded that
continuous culture rather than imposed fluctuations are
responsible for the hysteresis.
By coupling well-mixed fermenters, the residence
time distribution (RTD) or circulation time distribution
(CTD), which will occur in large-scale fermenters, can
be simulated. In each fermenter different conditions can
be established, corresponding to the conditions in different zones of large-scale fermenters. 2
It is clear that in plug-flow reactors gradients will
occur. But, as mentioned above, pure plug-flow will
cause washout of the cells, so some degree of backmixing or recirculation is needed.
Another possibility is the combination of a well
mixed fermenter and a plug-flow reactor with or
without recirculation. The so-called maximummixedness fermenters, consisting of a zone with intense
mixing and a zone with little mixing, can be regarded as
a combination of a well-mixed and a plug-flow reactor.
Tsai 45 has concluded from theoretical considerations
that in such systems the substrate utilization will be
optimal. However, this is based on the Monod equation
to describe microbial growth, and does not take into
account the influence of the periodically changing
environment in this system on the microorganisms.
Katinger 46 has worked with a maximum-mixedness

396

E n z y m e M i c r o b . T e c h n o l . , 1 9 8 7 , v o l . 9, J u l y

fermenter (a tubular closed-loop fermenter) to simulate


the mixing phenomena of a large-scale recycle fermenter
used for single cell protein production from n-paraffin.
The recycle time in the small-scale system was in the
order of 1 min up to several minutes. Candida tropicalis
utilizing pure n-alkanes was used as a model system.
With a more intensified transient limitation, Katinger
observed an increase of the yield of biomass on substrate Y~x, a reduced oxygen demand Yox, and a declining respiration coefficient RQ. More efficient energy
generation or utilization of substrate energy must have
been responsible for the increased Y~xand decreased Yox.
In her study of the influence of liquid-phase bulk
mixing and mass transfer on penicillin production,
Tuffnel147 used a reactor configuration consisting of a
well-mixed reactor and bubble column fermenter. Thus,
in simulation of a large-scale fermenter two zones can
be distinguished: a well-mixed zone in which agitation
is very good, and a bubble zone in which agitation is
very poor. This system can also be regarded as a
maximum-mixedness fermenter, although the plug-flow
reactor will be characterized by a considerable amount
of back-mixing.
Loop reactors and bubble column reactors can be
conceived as a combination of well-mixed and/or plugflow reactors. A disadvantage of these systems is the
more complicated hydrodynamics.

Conclusions and discussion


It is important to base small-scale experiments on largescale data, or estimations about the conditions at a production scale.
Many experimental set-ups can be used to simulate
large-scale fermentation processes on a small scale. The
choice of set-up depends on: the system which has to be
simulated; the equipment present and financial
resources; and the expertise with specific fermenters
with regard to both experimental work and knowledge
(including modelling) of the hydrodynamics of the
system.
Mixing effects can increase as well as decrease the
yield of biomass and the required metabolite. This will
depend on: the character of the fluctuations (parameter,
amplitude, frequency); the process (batch, fed-batch,
continuous); and the organism. The expression of fluctuations depends on the reduction of product yield or
production rate, and formation of by-products, or even
loss of viability.
Little attention has been paid to the formation of byproducts, although this can give considerable information on the metabolism of the organisms studied under
transient conditions. A better understanding of growth
and (by-)product formation is needed to develop adequate models to describe the behaviour of the microorganisms under large-scale conditions.
The advantage of the scale-down procedure is that, if
it works well, pilot-plant investigation can be circumvented. A distinction must be made between optimizing
an existing process and designing and scaling up a new
process. In the second case, in particular, the pilot plant
has some important tasks apart from process development, namely the training of operators and other personnel who have to deal with the new process;
production of a small amount of the product for market

Regime analysis and scale-down: A. P. J. Sweere et al.

research; safety studies; study of accumulation and pollution characteristics; study of resources and constructing materials.
There are, also, however, limits to small-scale investigations which make it necessary to use a scale larger
than laboratory scale. The scale may be limited by
sample volume, installation of probes, viscosity of the
liquid and inhomogeneities.
It can be concluded that, although pilot-plant investigation is very time-consuming and costly it offers
some possibilities not offered by laboratory-scale
studies. The final decision as to whether a pilot plant
will be built or not has to be an economic one.
Nomenclature
A
a

C
Cp

D
D
DE

dp
Eu
e

Fr
9
H
Hr
h
kIa
K
ks
L
Ls
ls
M
m

N
P
P
R
Re
Fh
F

T
t
ti

U
V
/)b
/)s
/)tip

We

ro,
.Y~x
x

area (m 2)
specific area based on liquid volume (m- 1)
concentration (kg m-3)
heat capacity (J kg- 1 K - 1)
diameter (m)
diffusion coefficient (m 2 s 1)
effective dispersion coefficient (m 2 s- 1)
particle diameter (m)
Euler number (--)
energy input per unit mass (J kg- 1)
Froude number (-)
gravitational acceleration (m s-2)
height (m)
height of impeller blade (m)
heat transfer coefficient (W m 2 K - 1)
volumetric mass transfer
coefficient based on liquid volume (s- 1)
Michaelis-Menten constant (kg m-3)
mass transfer coefficient for mass
transfer to the solid phase (m s- 1)
length (m)
scale of the reactor (m)
scale of a volume element caused
by turbulent erosion and laminar
stretching of the volume element (m)
molecular weight (kg mol- 1)
gas-liquid distribution coefficient (kg kg- 1)
stirrer speed (s- 1)
power input (W)
pressure (N m 5 2)
gas law constant (J mol- 1 K - 1)
Reynolds number (-)
heat production rate (J m -3 s-1)
reaction rate (kg m - 3 s- 1)
temperature (K)
time (s)
characteristic time (s)
overall heat transfer coefficient (W m - 2 K-1)
volume (m 3)
bubble rise velocity (m s- 1)
superficial gas velocity (m s- 1)
impeller tip speed (m s- 1)
Weber number (-)
yield of biomass on oxygen (kg dry wt kg- 1)
yield of biomass on substrate (kg dry wt kg- 1)
length (m)

Greek symbols
q
dynamic viscosity (Pa s)
#max

(maximum) specific growth rate (s h)

p
z
qbg
q~v
to

kinematic viscosity ( m 2 s - 1)
density (kg m-3)
mean residence time (s)
gas flow rate (m 3 s-1)
feed flow rate (m 3 s- 1)
heat production per kg oxygen
(J kg 1)

consumed

Subscripts
cir
D
d
g
hp
ht
1
ls
m
mt
o
oc
p
r
s
sa
sc
sett
sO
te
x

liquid circulation
diffusion
downcomer section of loop reactor
gas phase
heat production
heat transfer
liquid phase
laminar stretching
mixing
mass transfer
oxygen
oxygen consumption
process
riser section of loop reactor
substrate
substrate addition
substrate consumption
settling of particles
feed substrate
turbulent erosion
biomass

Superscripts
'

dimensionless parameter

References
1 BHRA Proc. Int. Conf. Bioreactor Fluid Dynamics Cambridge,
15-17 April 1986
2 Oosterhuis, N. M. G., PhD Thesis, Delft University of Technology, 1984
3 Kossen, N. W. F. and Oosterhuis, N. M. G. in Biotechnology
(Rehm, H. J. and Reed, G., eds) VCH Verlagsgesellschaft, Weinheim, 1985, vol. 2, 572-605
4 Jordan, D. G. Chemical Process Development (Schoen, H. M. and
McKetta, J. J., eds) Interscience Library of Chemical Engineering
and Processing, Wiley, Interscience, New York, 1968, vol. 6, parts
1 and 2
5 Einsele, A., paper presented at 5th Int. Ferment. Syrup. Berlin, 28
June-3 July 1976
6 Einsele, A. Process Biochem., 1978, 13(7), 13-14
7 Pace, G. W. Adv. School Microbiol. Biotechnol. Queen Elizabeth
College London, 14-20 September 1980
8 Barford, J. P., Pamment, N. B. and Hall, R. J. in Microbial Population Dynamics (Bazin, M. J., ed.) CRC Press, Boca Raton, 1982,
pp. 55-89
9 Johnstone, R. E. and Thring, M. W. Pilot Plant, Models and
Scale-up Methods in Chemical Engineering McGraw-Hill, New
York, 1957
l0 Pawlowski, J. Die Aehnlichkeitstheorie in der physikalischtechnischen Forschung: Grundlagen und Anwendung SpringerVerlag, Berlin, 1971
l l Zlokarnik, M. Aehnlichkeitstheorie in der Verfahrenstechnik
Bayer, 1974
12 Becker, H. A. Dimensionless Parameter: Theory and Methodology
Applied Science Publishers, London, 1976
13 Zlokarnik, M. Ger. Chem. Eng. 1984, 7, 150-159
14 Levenspiel, O. Chemical Reaction Engineering 2nd edn, John
Wiley & Sons, New York, 1972
15 Moes, J., Griot, M., Keller, J., Heinzle, E., Dunn, I. J. and
Bourne, J. R. Biotechnol. Bioeng. 1985, 27, 482~489
16 Coulson, J. M. and Richardson, J. F. Chemical Engineering: Unit
Operations 2nd edn, Pergamon Press, Oxford, 1968, vol. 2

Enzyme Microb. Technol., 1987, vol. 9, July

397

Review
17 Bemer, G. G. PhD Thesis, Delft University of Technology, 1979
18 Einsele, A. Chem. Rundschau 1976, 29 (25), 53
19 Huber, T. F., Kossen, N. W. F., Bos, P. and Kuenen, J. G. Proc.
Symp. Biotechnnol. Res. in the Netherlands Delft, 22 November
1983
20 Bos, P., Huber, T. F., Kos, C. H., Ras, C. and Kuenen, J. G.
paper presented at Int. Syrup. Biohydrometallurgy Vancouver,
August 1985
21 Schoutens, G. H. PhD Thesis, Delft University of Technology,
1986
22 Sweere, A. P. J., Mesters, J. R., Kossen, N. W. F. and Luyben, K.
Ch. A. M. Proc. Int. Conf. Bioreactor Fluid Dynamics BHRA,
Cambridge, 15-17 April 1986, pp. 217-230
23 Geurts, Th. G. E., de Kok, H. E. and Roels, J. A. Biotechnnol.
Bioeng. 1980, 22, 2031-2043
24 Chester, A. K. Int. J. Multiphase Flow 1977, 3, 235-241
25 Bolle, W. L., Van Breugel, J., Van Eybergen, G. Ch., Kossen, N.
W. F. and Zoetemeyer, R. J. submitted for publication
26 Midoux, N. and Charpentier, J.-C. Int. Chem. Eng. 1984, 24,
249 287
27 Midoux, N. and Charpentier, J.-C. Int. Chem. Eng. 1984, 24,
452481
28 Heynen, J. J. and van 't Riet, K. Chem. Eng. J. 1984, 28, B21-B42
29 Joshi, J. B. and Shah, Y. T. Chem. Eng. Commun. 1981, 11,
165-199
30 Shah, Y. T., Kelkar, B. G., Godbole, S. P. and Deckwer, W. D.
AIChE J. 1982, 28, 353-379
31 Schuegerl, K., Luecke, J. and Oels, U. Adv. Biochem. Eng. 1977, 7,
1 84
32 Schuegerl, K., Luecke, J., Lehnmann, J. and Wagner, Adv.
Biochem. Eng. 1978, 8, 63-132
33 Schuegerl, K. Adv. Biochem. Eng. 1982, 22, 93-224
34 Pickett, A. M., Topiwala, H. H. and Bazin, M. J. Process
Biochem. 1979, 14(ll), l0
35 Harrison, D. E. F. and Topiwala, H. H. Adv. Biochem. Eng. 1974,
3, 167 219
36 Cooney, C. L, Koplov, H. M. and Haeggstroem, M. in Continuous Cultures of Cells (Calcott, P. H., ed.) CRC Press, Boca
Raton, 1981, vol. l, pp. 143-168
37 Pickett, A. M. in Microbial Population Dynamics (Bazin, M. J.,
ed.) CRC Press, Boca Raton, 1982, pp. 91-124
38 Brooks, J. D. and Meers, J. L. J. Gen. Microbiol. 1973, 77, 513
39 Meets, J. L. and Brooks, J. D. J. Gen. Microbiol. 1973, 75, VIII
40 Van Dijken, J. P., and Scheffers, W. A. FEMS Microbiol. Rev.
1986, 32(3/4), 199-224
41 Vardar, F. and Lilly, M. D. Eur. J. Appl. Microbiol. Biotechnol.
1982, 14, 203 211
42 Sokolov, D. P., Lirova, S. A. and Sokolova, E. A. Microbiologiya
1983, 52, 909-916
43 Heinzle, E., Moes, J. and Dunn. I. J. Biotechnol. Lett. 1985, 7,
235-240
44 Furukawa, K., Heinzle, E. and Dunn, I. J. Biotechnol. Bioeng.
1983, 25, 2293-2317
45 Tsai, B. I., Erickson, L. F. and Fan L. T. Biotechnol. Bioeng. 1969,
II, 181-205
46 Katinger, H. W. D. Eur. J. Appl. Microbiol. 1976, 3, 103-114
47 Tuffnell, J. M., Deford Atmettla, D. and Street, G. paper presentTable A1

49
50
51
52
53
54
55
56

ed at ACHEMA '85 Frankfurt am Main, 9-15 June 1985


Borzani, W., Gregori, E. and Vairo, M. L. R. Biotechnol. Bioeng.
1976, 18, 623-632
Weigand, W. A. Biotechnol. Bioeng. 1981, 23, 249-266
Soni, V. K. and Ghose, T. K. J. Ferment. Technol. 1974, 52, 551
Young, T. B., Bruley, H. R. and Bungay, H. R. Biotechnol. Bioeng.
1970, 12, 747
Gilley, J. W. and Bungay, H. R. Biotechnol. Bioeng. 1968, 10, 99
Lievense, J. C. PhD Thesis, Purdue University, 1984
Bellgardt, K.-H. PhD Thesis, University of Hannover, 1984
Katterer, L., Allemann, H., Kaeppeli, O. and Fiechter, A. Biotechnol. Bioeng. 1986, 28, 146-150
Villermaux, J. and David, R. Chem. Eng. Commun. 1983, 12,
105-122

Appendix A: Parameters and equations used


in regime analysis of the bakers'
yeast production
T h e p a r a m e t e r s o f t h e r e g i m e a n a l y s i s a r e g i v e n in
T a b l e A1. T h e f o l l o w i n g e q u a t i o n s h a v e b e e n u s e d :
m i x i n g a n d m a s s t r a n s f e r : 2a
D E j = 0.36(gD4vs) 1/3
DE,g = 78(0 s D) 3/2
vci r = 0.9(oDvs) 1/3
1 -- e = vd(vb + /3cir)
k I a = 0.32v '7
Energy input:
P = pg (og R T / ( V M ) I n ( p l / p 2 )

rh = (/)#max Cx/Yox
Kinetics:
r, = ,/~max/YsxC s Cx/(k, + Cs)

ro = ~am,x~YoxCo Cx/(ko + Co)


M i c r o - m i x i n g : t h e e q u a t i o n s f o r s t i r r e d t a n k r e a c t o r s 56
h a s b e e n u s e d as a n e s t i m a t i o n o f t h e m a c r o - m i x i n g i n a
bubble column reactor:
tls >~ (2v/e) 1/2

tte =

3.85ls vs/12D - 2/3 e - 1/4

tD =

3/5(12/6D)

12/D = 24(L2/e) + 6(v/e)ln(v/D)

Values of parameters used in regime analysis of the bakers' yeast p r o d u c t i o n

cp=
4 . 2 k J kg -1 K -1
Cs=
0 . 1 5 0 kg m 3
Cso = 2 5 0 . 0 kg m 3
C X= 4 5 . 0 kg m -3
C.(t=0)=
10.0kgm -3
g =
9.8 m s -2
ko =
5 . 1 0 - 5 kg m -3
ks=
0 . 5 kg m -3

398

48

Enzyme

Microb.

Technol.,

m=
3 0 . 0 k g k g -1
R =
8 . 3 1 4 J mo1-1 K -1
U =
1.3 k W m -2 K -1
V = 120.0 m 3
V(t=0) = 80.0m 3
vb=
0 . 2 5 m s -1
Yox =
1.0 kg kg -1

1987,

vol.

9, July

Y,, =
0 . 5 3 kg kg -1
Pm=x =
0.4 h -1
pg =
1.2 kg m -3
Pl = 1 0 0 0 . 0 kg m -3
~v =
4 . 0 m 3 h -1
to =
14.2 x 103 kJ (kg 0 2 ) -1
D =
0 . 6 7 3 x 10 -~ m 2 s -1

Vous aimerez peut-être aussi