Professional Documents
Culture Documents
Pfenosil
Biological Reaction Engineering
Dynamic Modelling Fundamentals
with Simulation Examples
Biological
Reaction Engineering
Dynamic Modelling Fundamentals
with Simulation Examples
WILEY-
VCH
WILEY-VCH GmbH & Co. KGaA
Dr. Irving J. Dunn This book was carefully produced. Nevertheless,
ETH Zurich authors and publisher do not warrant the informa-
Department of Chemical Engineering tion contained therein to be free of errors. Rea-
CH-8092 Zurich ders are advised to keep in mind that statements,
Switzerland data, illustrations, procedural details or other
items may inadvertently be inaccurate.
Professor Dr. Elmar Heinzle
University of Saarland
Department of Technical Biochemistry
P.O. Box 15 11 50 First Edition 1992
D-66041 Saarbrucken Second, Completely Revised Edition 2003
Germany
Dr.JiriE.Prenosil
ETH Zurich
Department of Chemical Engineering Bibliographic information published by Die Deut-
CH-8092 Zurich sche Bibliothek. Die Deutsche Bibliothek lists this
Switzerland publication in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in the
Internet at <http://dnb.ddb.de>.
ISBN 3-527-30759-1
Table of Contents
TABLE OF CONTENTS V
PREFACE XI
1 MODELLING PRINCIPLES 9
1.1 FUNDAMENTALS OF MODELLING 9
7.7.7 Use of Models for Understanding, Design and Optimization of Bioreactors 9
1.1.2 General Aspects of the Modelling Approach 10
1.1.3 General Modelling Procedure..... 72
1.1.4 Simulation Tools 75
7.7.5 Teaching Applications 75
1.2 DEVELOPMENT AND MEANING OF DYNAMC DIFFEREOTTAL BALANCES 16
1.3 FORMULATION OF BALANCE EQUATIONS ..21
7.5.7 Types of Mass Balance Equations 27
1.3.2 Balancing Procedure 23
1.3.2.1 Case A. Continuous Stirred Tank Bioreactor 24
1.3.2.2 CaseB. Tubular Reactor 24
1.3.2.3 Case C. River with Eddy Current 25
1.3.3 Total Mass Balances 33
1.3.4 Component Balances for Reacting Systems 34
1.3.4.1 Case A. Constant Volume Continuous Stirred Tank Reactor 35
1.3.4.2 Case B. Semi-continuous Reactor with Volume Change 37
1.3.4.3 Case C. Steady-State Oxygen Balancing in Fermentation 38
1.3.4.4 Case D. Inert Gas Balance to Calculate Flow Rates 39
7.5.5 Stoichiometry, Elemental Balancing and the Yield Coefficient Concept.. 40
1.3.5.1 Simple Stoichiometry 40
1.3.5.2 Elemental Balancing 42
L3.5.3 Mass Yield Coefficients 44
VI Table of Contents
11 INDEX 499
Preface
Our goal in this textbook is to teach, through modelling and simulation, the
quantitative description of bioreaction processes to scientists and engineers. In
working through the many simulation examples, you, the reader, will learn to
apply mass and energy balances to describe a variety of dynamic bioreactor
systems. For your efforts, you will be rewarded with a greater understanding of
biological rate processes. The many example applications will help you to gain
confidence in modelling, and you will find that the simulation language used,
Berkeley Madonna, is a powerful tool for developing your own simulation
models. Your new abilities will be valuable for designing experiments, for
extracting kinetic data from experiments, in designing and optimizing
biological reaction systems, and for developing bioreactor control strategies.
This book is based on part of our successful course, "Biological Reaction
Engineering", which has been held annually in the Swiss mountain resort of
Braunwald for the past twenty five years and which is now known, throughout
European biotechnology circles as the "Braunwald Course". More details can
be found at our website www.braunwaldcourse.ch. Modelling is often
unfamiliar to biologists and chemists, who nevertheless need modelling
techniques in their work. The general field of biochemical reaction engineering
is one that requires a very close interdisciplinary interaction between applied
microbiologists, biochemists, biochemical engineers, engineers and managers; a
large degree of collaboration and mutual understanding is therefore important.
Professional microbiologists and biochemists often lack the formal training
needed to analyze laboratory kinetic data in its most meaningful sense, and
they may sometimes experience difficulty in participating in engineering
design decisions and in communicating with engineers. These are just the very
types of activity required in the multi-disciplinary field of biotechnology.
Chemical engineering's greatest strength is its well-developed modelling
concepts, based on mass and energy balances, combined with rate processes.
Biochemical engineering is a discipline closely related to conventional
chemical engineering, in that it attempts to apply physical principles to the
solution of biological problems. This approach may be applied to the
measurement and interpretation of laboratory kinetic data or as well to the
design of large-scale fermentation, enzymatic or waste treatment processes. The
necessary interdisciplinary cooperation requires the biological scientists and
chemical engineers involved to have at least a partial understanding of each
other's field. The purpose of this book is to provide the mathematical tools
necessary for the quantitative analysis of biological kinetics and other
biological process phenomena. More generally, the mathematical modelling
XII Preface
The book is divided into two parts: a presentation of the background theory in
Part I and the computer simulation exercises in Part II. The function of the text
in Part I is to provide the basic theory required to fully understand and to make
full use of the computer examples and simulation exercises. Numerous case
studies provide illustration to the theory. Part II constitutes the main part of this
book, where the simulation examples provide an excellent instructional and
self-learning tool. Each of the more than fifty examples is self-contained,
including a model description, model equations, exercises, computer program
listing, nomenclature and references. The exercises range from simple
parameter-changing investigations to suggestions for writing a new program.
The combined book thus represents a synthesis of basic theory and computer-
based simulation examples.
Quite apart from the educational value of the text, the introduction and use
of the Berkeley Madonna software provides the reader with the considerable
practical advantage of a differential equation solution package. In the appendix
a screenshot guide is found concerning the use of the software.
Part I: "Principles of Bioreactor Modelling" covers the basic theory
necessary for understanding the computer simulation examples. This section
presents the basic concepts of mass balancing, and their combination with
kinetic relationships, to establish simple biological reactor models, carefully
presented in a way that should be understandable to biologists. In fact,
engineers may also find this rigorous presentation of balancing to be valuable.
XIV Preface
In order to achieve this aim, the main emphasis of the text is placed on an
understanding of the physical meaning and significance of each term in the
model equations. The aim in presenting the relevant theory is thus not to be
exhaustive, but simply to provide a basic introduction to the theory required for
a proper understanding of the modelling methodology.
Chapter 1 deals with the basic concepts of modelling, the basic principles,
development and significance of differential balances and the formulation of
mass and energy balance relationships. Emphasis is given to physical
understanding. The text is accompanied by example cases to illustrate the
application of the material.
Chapter 2 serves to introduce the varied operational characteristics of the
various types of bioreactors and their differing modes of operation, with the
aim of giving a qualitative insight into the quantitative behavior of the
computer simulation examples.
Chapter 3 provides an introduction to enzyme and microbial kinetics. A
particular feature of the kinetic treatment is the emphasis on the use of more
complex structured models. Such models require much more consideration to
be given to the biology of the system during the modelling procedure, but
despite their added complexity can nevertheless also be solved with relative
ease. They serve as a reminder that biological reactions are really infinitely
complex.
Chapter 4 is used to derive general mass balance equations, covering all types
of fermentation tank reactors. These generalized equations are then simplified
to show their application to the differing modes of stirred tank bioreactor
operation, discussed previously and which are illustrated by the simulation
examples.
Chapter 5 explains the basic theory of interfacial mass transfer as applied to
fermentation systems and shows how equations for rates of mass transfer can be
combined with mass balances, for both liquid and gas phases. A particular
extension of this approach is the combination of transfer rate and material
balance equations to models of increased geometrical complexity, as
represented by large-scale air-lift and multiple-impeller fermenters.
Chapter 6 treats the cases of external diffusion to a solid surface and internal
diffusion combined with biochemical reaction, with practical application to
immobilized biocatalyst and biofilm systems. Emphasized here is the
conceptual ease of handling a complex reaction in a solid biocatalyst matrix.
The resulting sets of tractable differential-difference equations are solved by
simulation techniques in several examples.
Chapter 7 describes the importance of control and summarizes control
strategies used for bioreaction processes. Here the fundamentals of feedback
control systems and their characteristic responses are discussed. This material
forms the basis for performing the many recommended control exercises in the
simulation examples. It also will allow the reader-simulator to develop his or
her own control models and simulation programs.
Preface XV
Part II, "Dynamic Bioprocess Simulation Examples and the Berkeley Madonna
simulation language" comprises Chapter 8, with the computer simulation
examples, and Chapter 9, which gives the instructions for using Madonna, Each
example in Chapter 8 includes a description of its physical system, the model
equations, that were developed in Part I, and a list of suggested exercises. The
programs are found on the CD-ROM. These example exercises can be carried
out in order to explore the model system in detail, and it is suggested that work
on the computer exercises be done in close reference to the model equations
and their physical meaning, as described in the text. The exercises, however, are
provided simply as an idea for what might be done and are by no means
mandatory or restrictive. Working through a particular example will often
suggest an interesting variation, such as a control loop, which can then be
programmed and inserted. The examples cover a wide range of application and
can easily be extended by reference to the literature. They are robust and are
well tested by a variety of undergraduate and graduate students and by also the
350 participants, or so, who have previously attended the Braunwald course. In
tackling the exercises, we hope you will soon come to share our conviction that,
besides being very useful, computer simulation is also fun to do.
For the second edition, the text was thoroughly revised and some of our
earlier, less relevant material was omitted. On the other hand, a number of new
examples resulting mainly from the authors' latest research and teaching work
were added. There was also an opportunity in this new edition to eliminate
most of the past errors and to avoid new ones as much as possible. Most
importantly, the examples have been rewritten in Berkeley Madonna, which all
of our reader-simulators will greatly appreciate.
Our book has a number of special characteristics. It will be obvious, in
reading it through, that we concentrate only on those topics of biological
reaction engineering that lend themselves to modelling and simulation and do
not attempt to cover the area completely. Our own research work is used to
illustrate theoretical points and from it many simulation examples are drawn. A
list of suggested books for supplementary reading is found at the end of
Chapter 6, together with the list of cited references. The diversity of the
simulation examples made it necessary to use separate nomenclature for each.
The symbols used in Chapters 1 - 6 are defined at the end of Part I. The
authors' four nationalities and three mother tongues, made it difficult to settle
on American or British spelling. Somehow we like "modelling" better than
"modeling".
We are confident that the book will be useful to all life scientists wishing to
obtain an understanding of biochemical engineering and also to those chemical
and biochemical engineers wanting to sharpen their modelling skills and
wishing to gain a better understanding of biochemical process phenomena. We
hope that teachers with an interest in modelling will find this to be a useful
textbook for undergraduate and graduate biochemical engineering and
biotechnological courses.
XVI Preface
Acknowledgements
Symbols Units
A Area m2
A Magnitude of controller input signal various
a Specific area m2/m3
a Constant in Logistic Equation 1/h
b Constant in Luedeking-Piret relation 1/h
b Constant in Logistic Equation m3/kg h
B Magnitude of controller output signal various
c Fraction carbon converted to biomass
C Concentration kg/m3, kmol/m3
CP Heat capacity kJ/kg K, kJ/mol K
CPR Carbon dioxide production rate mol/h
D Diffusivity m2/h
D Dilution rate 1/h
d Differential operator and diameter -, m
d Fraction carbon converted to product
DO Dissolved oxygen g/m3, 9 air sat.
E Enzyme concentration g/m3
E Ethanol kg/m3
ES Enzyme-substrate concentration g/m3
f Fraction carbon converted to CO2
f Frequency in the ultimate gain method 1/h
F Flow rate m3/h and m3/s
G Gas flow rate m3
G Intracellular storage product kg/m3
hi Partial molar enthalpy kJ/mol
H Henry's Law constant bar m3/kg
AH Enthalpy change kJ/mol or kJ/kg
I Inhibiting component concentration kg/m3
I Cell compartment masses kg/m3
j Mass flux kg/m2h, mol/m2h
K Mass transfer coefficient 1/h
K Constant in Cohen-Coon method various
KD Acid-base dissociation constant
k Constant various
Kca Gas-liquid mass transfer coefficient 1/h
kGa Gas film mass transfer coefficient 1/h
KI Inhibition constant kg/m3, kmol/m3
KLa Gas-liquid mass transfer coefficient 1/h
kLa Liquid film mass transfer coefficient 1/h
KM Michaelis-Menten constant kg/m3, kmol/m3
Kp Proportional controller gain constant various
KW Dissociation constant of water
KS Monod saturation coefficient kg/m3
L Length m
M Mass kg or mol
m Maintenance coefficient 1/h
N Mass flux kg/m2 h
N Molar flow rate mol/h
n Number of mols
n Reaction order _
OTR Oxygen transfer rate mol/h and kg/h
OUR Oxygen uptake rate mol/h and kg/h
P Pressure bar
P Product concentration kg/m3 and g/m3
P Output control signal various
Q Total transfer rate kg/h and mol/h
q Specific rate kg/kg biomass h
R Ideal gas constant bar m3/ K mol
R Recycle flow rate m3/h
R Residual active biomass kg/m3
r Reaction rate kg/m3h, kmol/m3h
fi Reaction rate of component i kg i/m3h
r
i/j Reaction rate of component i to j kg /m3h, kmol/m3h
RQ Respiration quotient mol CO2/mol 2
rx Growth rate kg biomass/m3 h
S Concentration of substrate kg/m3, kmol/m3
S Slope of process reaction curve various
s Stoichiometric coefficient
T Temperature CorK
T Enzyme activity kg/m3
T Time lag h, min. or s
L
t Time h, min and s
Tr Transfer rate mol/m3 h
U Heat transfer coefficient kJ/m2 C h
V Volume m3
V Flow velocity m/h
v
max Maximum reaction rate kmol/m3 h
w Wastage stream flow rate m3/h
Nomenclature
w Mass fraction
X Biomass concentration kg/m3
Y Yield coefficient kg/kg
Yi Yield of i from j kg i/kg j,mol i/mol j
y Mol fraction in gas
Z Length variable m
Greek
Indices
* Refers to equilibrium concentration
0 Refers to initial, inlet, external, and zero order
1 Refers to time ti, outlet, component 1, tank 1, and first order
2 Refers to tank 2, time t2 and component 2
1,2,..., n Refers to stream, volume elements and stages
A Refers to component A, anions and bulk
A- Refers to anions
a Refers to ambient
Ac Refers to acetoin and acetoin formation
aer Refers to aerobic
agit Refers to agitation
anaer Refers to anaerobic
app Refers to apparent
ATP/S,Ac Refers to ATP yield from reaction glucose --> Ac
ATP/S,CO2 Refers to ATP yield from glucose oxidation
ATP/NADH Refers to ATP produced from NADH
ATP/X Refers to consumption rate ATP > biomass
Nomenclature
then be used to further redefine or refine the model until good agreement is
obtained. Once the model is established it can then be used, with reasonable
confidence, to predict performance under differing process conditions, and it
can also be used for such purposes as process design, optimization and control.
An input of plant or experimental data is, of course, required in order to
establish or validate the model, but the quantity of experimental data required,
as compared to that of the empirical approach is considerably reduced. Apart
from this, the major advantage obtained, however, is the increased
understanding of the process that one obtains simply by carrying out the
modelling exercise.
These ideas are summarized below.
Empirical Approach: Measure productivity for all combinations of reactor
operating conditions, and make correlations.
- Advantage: Little thought is necessary
- Disadvantage: Many experiments are required.
uptake rate (qo2)> specific carbon dioxide production rate (qco2)> may also be
derived and used to provide a complete kinetic description of, say, a simple
batch fermentation.
For complex fermentations, involving product formation, the specific
product production rate (qp) is often correlated as a complex function of
fermentation conditions, e.g., stirrer speed, air flow rate, pH, dissolved oxygen
content and substrate concentration. In other cases, simple kinetic models can
also be used to describe the functional dependence of productivity on cell
density and cell growth rate.
A more detailed "structured kinetic model" may be required to give an
adequate description of the process, since cell composition may change in
response to changes in the local environment within the bioreactor. The greater
the complexity of the model, however, the greater is then the difficulty in
identifying the numerical values for the increased number of model parameters,
and one of the skills of modelling is to derive the simplest possible model that
is capable of a realistic representation of the process.
A basic use of a process model is thus to analyze experimental data and to
use this to characterize the process, by assigning numerical values to the
important process variables. The model can then also be solved with
appropriate numerical data values and the model predictions compared with
actual practical results. This procedure is known as simulation and may be
used to confirm that the model and the appropriate parameter values are
"correct". Simulations, however, can also be used in a predictive manner to test
probable behavior under varying conditions; this leads on to the use of models
for process optimization and their use in advanced control strategies.
The application of a combined modelling and simulation approach leads to
the following advantages:
effect and hence that these effects therefore can be neglected both from
the model and from the experimental program.
3. Models may be used predictively for design and control. Once the
model has been established, it should be capable of predicting
performance under differing sets of process conditions. Mathematical
models can also be used for the design of relatively sophisticated control
algorithms, and the model, itself, can often form an integral part of the
control algorithm. Both mathematical and knowledge based models can
be used in designing and optimizing new processes.
(i) The first stage involves the proper definition of the problem and hence the
goals and objectives of the study. These may include process analysis,
improvement, optimization, design and control, and it is important that the aims
of the modelling procedure are properly defined. All the relevant theory must
then be assessed in combination with any practical experience with the process,
1.1 Fundamentals of Modelling 13
Physical Model
New
LJt
Mathematical Model
Revise ideas
and equations
NO
experiments
Experimental Data
Comparison
Solution: C = f(t)
OK?
YES
and perhaps alternative physical models for the process need to be developed
and examined. At this stage, it is often helpful to start with the simplest possible
conception of the process and to introduce complexities as the development
proceeds, rather than trying to formulate the full model with all its complexities
at the beginning of the modelling procedure.
(iii) Having developed a model, the model equations must then be solved.
Mathematical models of biological systems are usually quite complex and
highly non-linear and are such that the mathematical complexity of the
14 1 Modelling Principles
(iv) The validity of the computer prediction must be checked and steps (i) to
(iii) will often need to be revised at frequent intervals during the modelling
procedure. The validity of the model depends on the correct choice of the
available theory (physical and mathematical model), the ability to identify the
model parameters correctly and the accuracy of the numerical solution method.
In many cases, owing to the complexity and very interactive nature of
biological processes, the system will not be fully understood, thus leaving large
areas of uncertainty in the model. Also the relevant theory may be very
difficult to apply. In such cases, it is then often very necessary to make rather
gross simplifying assumptions, which may subsequently be eliminated or
improved as a better understanding is subsequently obtained. Care and
judgement must also be used such that the model does not become over
complex and so that it is not defined in terms of too many immeasurable
parameters. Often a lack of agreement between the model and practice can be
caused by an incorrect choice of parameter values. This can even lead to quite
different trends being observed in the variation of particular parameters during
the simulation.
It should be noted, however, that often the results of a simulation model do
not have to give an exact fit to the experimental data, and often it is sufficient to
simply have a qualitative agreement. Thus a very useful qualitative
understanding of the process and its natural cause-and-effect relationships is
obtained.
1.1 Fundamentals of Modelling 15
Many different digital simulation software packages are available on the market
for PC and Mac application. Modern tools are numerically powerful, highly
interactive and allow sophisticated types of graphical and numerical output.
Most packages also allow optimisation and parameter estimation. BERKELEY
MADONNA is very user-friendly and very fast. We have chosen it for use in
this book, and details can be found in the Appendix. With it data fitting and
optimisation can be done very easily. MODELMAKER is also a more recent,
powerful and easy to use program, which also allows optimisation and
parameter estimation. ACSL-OPTIMIZE has quite a long history of
application in the control field, and also for chemical reaction engineering.
MATLAB-SIMULINK is a popular and powerful software for dynamic
simulation and includes many powerful algorithms for non-linear optimisation,
which can also be applied for parameter estimation.
As indicated in Section 1.1, many models for biological systems are expressed
in terms of sets of differential equations, which arise mainly as a result of the
predominantly time-dependent nature of the process phenomena concerned.
For many people and especially for many students in the life sciences, the
mention of differential equations can cause substantial difficulty. This section
is therefore intended, hopefully, to bring the question of differential equations
into perspective. The differential equations arise in the model formulation,
simply by having to express rates of change of material, due to flow effects or
chemical and biological reaction effects. The method for solution of the
differential equations will be handled automatically by the computer. It is
hoped that much of the difficulty can be overcome by considering the
following case. In this section a simple example, based on the filling of a tank
of water, is used to develop the derivation of a mass balance equation from the
basic physical model and thereby to give meaning to the terms in the equations.
Following the detailed derivation, a short-cut method based on rates is given to
derive the dynamic balance equations.
Consider a tank into which water is flowing at a constant rate F (m3/s), as
shown in Fig. 1.2. At any time t, the volume of water in the tank is V (m3) and
the density of water is p (kg/m3).
Figure 1.2. Tank of water being filled by stream with flow rate F.
During the time interval At (s), a mass of water p F At (kg) flows into the tank.
As long as no water leaves the tank, the mass of water in the tank will increase
by a quantity p F At, causing a corresponding increase in volume, AV.
Equating the accumulation of mass in the tank to the mass that entered the tank
during the time interval A t gives,
pAV = p F A t
Since p is constant,
-Fh
At -
1.2 Development and Meaning of Dynamic Differential Equations 17
Applying this to very small differential time intervals (At > dt) and replacing
the A signs by the differential operator "d", gives the following simple first
order differential equation, to describe the tank filling operation,
dV
dT = F
What do we know about the solution of this equation? That is, how does the
volume change with time or in model terms, how does the dependent variable,
V, change with respect to the independent variable, t? To answer this, we can
rearrange the equation and integrate it between appropriate limits to give,
t
o
or for constant F,
= F Jtf l l dt= F(ti-to)
o
Integration is equivalent to summing all the contributions, such that the total
change of volume is equal to the total volume of water added to the tank,
IV = IF At
For the case of constant F, it is clear that the analytical solution to the
differential equation is,
V = F t + constant
In this case, as shown in Fig. 1.3, the constant of integration is the initial volume
of water in the tank, VQ, at time t = 0.
dt
Vo
Figure 1.3. Volume change with time for constant flow rate.
18 1 Modelling Principles
Note that the slope in the variation of V with respect to t, dV/dt, is constant, and
that from the differential equation it can be seen that the slope is equal to F.
Suppose F is not constant but varies linearly with time.
F = Fo-kt
Integrating analytically,
kt"
V = FO t - + constant
The solution is,
v = FO t - kt'
+ V0
."t
Figure 1.4. Variation of F and V for the tank-filling problem.
Note that the dependent variable starts at the initial condition, (Vo), and that the
slope is always F. When F becomes zero, the slope of the curve relating V and t
also becomes zero. In other words, the volume in the tank remains constant
and does not change any further as long as the value of F remains zero.
Thus, the rate of accumulation of mass within the tank can be written directly as
dM/dt where the mass M is equal to p V. The rate of mass entering the tank is
given by p F, where both sides of the equation have units of kg/h.
=
PF
and
d(oV)
= pF
Thus this approach leads directly to a differential equation model, which is the
desired form for dynamic simulation. Note that both terms in the above
relationship are expressed in mass quantities per unit time or kg/h.
dV
dT = F
which is to be solved for the initial condition, that at time t = 0, V=Vo and for a
variation in flow rate, given by,
F = F0-kt
which is valid until F = 0.
These two equations, plus the initial condition, form the mathematical
representation or the mathematical model of the physical model, represented
by the tank filling with an entering flow of water. Thus this approach leads
directly to a differential equation model, which is the desired form for
simulation. This approach can be applied not only to the total mass but also to
the mass of any component.
We have seen an analytical solution to this model, but it is also interesting to
consider how a computer solution can be obtained by a numerical integration
of the model equations. This is important since analytical integration is seldom
possible in the case of real complex problems.
Computer Solution
dV
20 1 Modelling Principles
Figure 1.5. Graphical portrayal of numerical integration, showing slopes and approximated
values of V at each time interval.
1.3 Formulation of Balance Equations 21
Using such a numerical integration procedure, the computer can thus be used
to generate data concerning the time variations of both F and V. In practice,
more complex numerical procedures are employed in digital simulation
languages to give improved accuracy and speed of solution than illustrated by
the above simplified integration technique.
Steady-State Balances
Many bioreactor applications are, however, such that conditions are in fact
changing with respect to time. Under these circumstances, a steady-state mass
balance is inappropriate and must be replaced by a dynamic or unsteady-state
mass balance, which can be expressed as:
( Rate of accumulation of ^
mass in the system J
( Rate of ^
^mass flow inj
( Rate of ^
^mass flow out J
Here the rate of accumulation term represents the rate of change in the total
mass of the system, with respect to time, and at steady-state this is equal to zero.
Thus the steady-state mass balance represented earlier is seen to be a
simplification of the more general dynamic balance, involving the rate of
accumulation.
At steady-state:
22 1 Modelling Principles
( Rate of ^
= 0 = (Mass flow in) - (Mass flow out)
I accumulation of mass .
Component Balances
The previous discussion has been in terms of the total mass of the system, but
most fluid streams, encountered in practice, contain more than one chemical or
biological species. Provided no chemical change occurs, the generalized
dynamic equation for the conservation of mass can also be applied to each
component. Thus for any particular component:
Rate of
Mass flow of A Mass flow of >
accumulation of mass
of component
in the system
(the component _
into the system J (
the component out
of the system ,
Where chemical or biological reactions occur, this can be taken into account by
the addition of a further reaction rate term into the generalized component
balance. Thus in the case of material produced by the reaction:
Elemental Balances
The principle of the mass balance can also be extended to the atomic level and
applied to particular elements. Thus in the case of bioreactor operation, the
general mass balance equation can also be applied to the four main elements,
carbon, hydrogen, oxygen and nitrogen and also to other elements if relevant
to the particular problem. Thus for the case of carbon:
Note the elemental balances do not involve reaction terms since the elements do
not change by reaction.
The computer example PENFERM, is based on the use of elemental mass
balance equations for C, H, O and N which, when combined with other
empirical rate data, provide a working model for a penicillin production
process.
While the principle of the mass balance is very simple, its application can
often be quite difficult. It is important therefore to have a clear understanding
of both the nature of the system (physical model), which is to be modelled by
means of the mass balance equations, and also of the methodology of
modelling.
The methodology described below outlines six steps, I through VI, to establish
the model balances. The first task is to define the system by choosing the
balance or control region. This is done using the following procedure:
The balance region may be a reactor, a reactor region, a single phase within a
reactor, a single cell, or a region within a cell, but will always be based on a
region of assumed constant composition. Generally the modelling exercises will
involve some prior simplification. Often the system being modelled is usually
considered to be composed of either systems of tanks (stagewise or lumped
24 1 Modelling Principles
AO
Total mass = pV
Mass of A = C VA
Balance region
If the tank is well-mixed, the concentrations and density of the tank contents
are uniform throughout. This means that the outlet stream properties are
identical with the tank properties, in this case CA and p. The balance region can
therefore be taken around the whole tank.
The total mass in the system is given by the product of the volume of the
tank contents V (m3) multiplied by the density p (kg/m3), thus Vp (kg). The
mass of any component A in the tank is given as the product of V times the
concentration of A, CA (kg of A/m3 or kmol of A /m3), thus V CA (kg or kmol).
Balance region
'AO A1
In the case of tubular reactors, the concentrations of the products and reactants
will vary continuously along the length of the reactor, even when the reactor is
operating at steady-state. This variation can be regarded as being equivalent to
1.3 Formulation of Balance Equations 25
that of the time of passage of material as it flows along the reactor and is
equivalent to the time available for reaction to occur. Under steady-state
conditions the concentration at any position along the reactor will be constant
with respect to time, though not with position. This type of behavior can be
approximated by choosing the balance regions sufficiently small so that the
concentration of any component within a region can be assumed to be
approximately uniform. Thus in this case, many uniform property subsystems
(well-stirred tanks or increments of different volume but of uniform
concentration) comprise the total reactor volume.
For this example, the combined principles of both the stirred tank and
differential tubular modelling approaches need to be applied. As shown in Fig.
1.8 the main flow along the river is very analogous to that of a column or
tubular process, whereas the eddy region can be approximated by the behavior
of a well-mixed tank. The interaction between the main flow of the river and
the eddy, with flow into the eddy from the river and flow out from the eddy
back into the river's main flow, must be included in any realistic model.
The real-life and rather complex behavior of the eddying flow of the river,
might thus be represented, by a series of many well-mixed subsystems (or
tanks) representing the main flow of the river. This interacts at some particular
stage of the river with a single well-mixed tank, representing the turbulent eddy.
In modelling this system by means of mass balance equations, it would be
necessary to draw boundary regions around each of the individual subsystems
representing the main river flow, sections 1 to 8 in Fig. 1.9, and also around the
tank system representing the eddy. This would lead to a very minimum of nine
River
Eddy
>
* 1 ',
i
2
2 ',
i
3
3
1
i
4
4
1
i
, 5
5
1
i
, 66
1
i
, 7
7
1
i
, 8
8 > fc C
r
>k
Flow interaction
V
f
fitSliM^B
:i
iK&^'ffK<&y^
Figure 1.9. A multi-tank model for the complex river flow system.
Having defined the balance regions, the next task is to identify all the relevant
inputs and outputs to the system. These may be well-defined physical flow
rates (convective streams), diffusive fluxes, and also interphase transfer rates.
It is important to assume a direction of transfer and to specify this by means of
an arrow. This direction might reverse itself, but will be accomodated by a
reversal in sign.
Out by diffusion
Convective
flow out
Convective flow
in
In by diffusion
Figure 1.10. Balance region showing convective and diffusive flows in and out.
This is given by the derivative of the mass of the system, or the mass of some
component within the system, with respect to time. Hence:
dMi
(Rate of accumulation of mass of component i within the system) =
dMj _ d(CjV)
=
dt dt
_ = ni
c = Pi yiP
i T RT = "RT
where yi is the mol fraction of the component in the gas phase and p is the total
pressure.
The accumulation term for the gas phase can be written as,
/piV
dMj _ d(CjV) _ d(QV) _ . d _
dM _ d(p V)
dt = dt
with units
m3 s
Total mass flow rates are given by the product of volumetric flow multiplied by
density. Component mass flows are given by the product of volumetric flow
rates times concentration.
Mass \
(VolumeJ
kg m 3 kg
s - s m3
A stream leaving a well-mixed region, such as a well stirred tank, has the same
properties as the system volume as a whole, since for perfect mixing the
contents of the tank will have uniform properties, identical to the properties of
the fluid leaving at the outlet. Thus, the concentrations of component i both
within the tank and in the tank effluent are equal to Qi, as shown in Fig. 1.11.
1.3 Formulation of Balance Equations 29
lilf
C. Diffusion of Components
Ji = -i dZ
Figure 1.12. Diffusion flux j j driven by concentration gradient (Qo - Cji) / AZ through
surface area A.
In accordance with Pick's Law, diffusive flow always occurs in the direction of
decreasing concentration and at a rate proportional to the concentration
gradient. Under true conditions of molecular diffusion, the constant of
proportionality is equal to the molecular diffusivity for the system, Dj (m2/h).
For other cases, such as diffusion in porous matrices and turbulent diffusion, an
30 1 Modelling Principles
kg 2 _ m 2 kg 2 _ kg
sm 2 m
" s m4 m " T
D. Interphase Transport
Interphase mass transport also represents a possible flow into or out of the
system. In bioreactor modelling applications, this is most frequently
represented by the case of oxygen transfer from air to the liquid medium,
followed by oxygen taken up by the cells during respiration. In this case, the
transfer of oxygen occurs across the gas liquid interface, which exists between
the surface of the air bubbles and the surrounding liquid medium, as shown in
Fig. 1.13.
Figure 1.13. Transfer of oxygen across a gas-liquid interface of specific area "a" into a liquid
phase of volume V.
/ Rate of \
accumulation /Mass flow\ /Massflow\ / of interfacial
Rate \
of the of the of the mass transfer
mass of oxygen = oxygen - oxygen - from the gas
in the gas into the from the phase into
V phase system J Vgas phase ) ^ gas phase> V the liquid s
This form of transfer rate equation will be examined in much more detail in
Chapter 5. Suffice it to say here that the rate of transfer can be expressed in the
form shown below:
= KaACV
where, a is a specific area for mass transfer, A/V (m2/m3), A is the total
interfacial area for mass transfer (m2), V is the liquid phase volume (m3), AC is
the concentration driving force (kmol/m3 or kg/m3) and, K is the overall mass
transfer coefficient (1/s). Mass transfer rate expressions are usually expressed
in terms of kmol/s, and can be converted to mass flows (kg/s), if desired.
The units of the terms in the equation (with appropriate mass quantity units)
are:
kg 1 kg
mj
Production Rate
This term in the component balance equation allows for the production or
consumption of material bv
by reaction and is incorporated into the component
balance equation. Thus,
Chemical production rates are often expressed on a molar basis and, as in the
case of the interfacial mass transfer rate expressions, can be easily converted to
mass flow quantities (kg/s). The production rate can then be expressed as
32 1 Modelling Principles
kg __ kg
m3
s m3
Equivalent molar quantities may also be used. The quantity r^ is positive when
A is formed as product, and TA is negative when a reactant A is consumed.
The growth rate for cells can be expressed in the same manner, using the
symbol rx- Thus,
kg _ kg
mj
s m3
The consumption rate of substrate, r$, is often directly related to the cell growth
rate by means of a constant yield coefficient YX/S, which has the units of kg
biomass produced per kg substrate consumed. Thus,
kg ~ kg biomass kg substrate
s m3 m =
s m3 kg biomass
The system mass balance equations are often the most important elements of
any modelling exercise, but are themselves rarely sufficient to completely
formulate the model. Other relationships are therefore needed to supplement
the material balance relations, both to complete the model in terms of other
important aspects of behavior and to satisfy the mathematical rigor of the
modelling, such that the number of unknown variables must be equal to the
number of defining equations.
1.3 Formulation of Balance Equations 33
Examples of this type of relationships which are not based on balances, but
which nevertheless form an important part of any model are:
How these and other relationships are incorporated within the development of
particular modelling instances are shown later in the cases given throughout the
text and in the simulation examples.
In this section the application of the total mass balance principles will be
presented. Consider some arbitrary balance region, as shown in Fig. 1.14 by
the shaded area. Mass accumulates within the system at a rate dM/dt, owing to
the competing effects of a convective flow input (mass flow rate in) and an
output stream (mass flow rate out).
34 1 Modelling Principles
dM
= Mass rate in - Mass rate out
d(p V) system = F
3t oPo-FiPi
When densities are equal, as in the case of water flowing in and out of a tank,
dV
dT = F O-FI
The steady-state condition of constant volume in the tank (dV/dt = 0) occurs
when the volumetric flow in, FQ, is exactly balanced by the volumetric flow out,
FI. Total mass balances therefore are mostly important for those bioreactor
modelling situations in which volumes are subject to change.
Species i
Species i outflow
inflow
Thus for any species i, involved in the system, the component mass balance is
given by:
,3*6.
m~
_ m^ m3
ill 3 _
m-
F
FQ C AO C BO 1 C A 1 C B1
>f
I! i ll
ilili
;|i
Figure 1.16. Continuous stirred tank reactor with reaction A > 2B.
Here it is convenient to use molar masses, such that each term has the units of
kmol/h.
Under constant volume conditions:
d(VCA) = VdC A
d(VCB) = VdC B
and in addition FQ = FI. Thus the two model equations, then simplify to give:
dCAi F
= - CAI) +
V
and
dCBi =
F
~dT~ V ( c BO~C B i)
In these two balances there are four unknowns CAI, Q*i, rAl an(^ rB l 8
kinetics are assumed to be first order, as often found in biological systems at
low concentration. Then:
r
Al = -
According to the molar stoichiometry,
1.3 Formulation of Balance Equations 37
rfil = -2r M = + 2 k C A !
Together with the kinetic relations there are 4 equations and 4 unknowns, thus
satisfying the conditions necessary for the model solution. With the initial
conditions, CAI and CBI at time t = 0, specified, the solution to these two
simultaneous equations, combined with the two kinetic relations, will give the
resulting changes of concentrations CAI and CBI as functions of time. The
simulation example ENZCON, is similar to the situation of Case A.
The chemical reaction and reaction rate data are the same as in the preceding
example, but now the reactor has no effluent stream. The operation of the
reactor is therefore semi-continuous.
AO
*
2B
rB = - 2 rA = + 2 k CA
The component balances with no flow of material leaving the reactor are now:
= FC AO + r A V
d(V C )
atB
= IB v
38 1 Modelling Principles
The number of unknowns is now five and the number of equations is four, so
that an additional defining relationship is required for solution. Note that V
must remain within the differential, because the volume of the reactor contents
is now also a variable and must be determined by a total mass balance.
Assuming constant density p, this gives the defining equation as:
= FF
dt
With initial conditions for the initial molar quantities of A and B, (VGA,
and the initial volume of the contents, V, at time t = 0 specified, the resulting
system of equations can be solved to obtain the time varying quantities VCA(t),
VCs(t), V(t) and hence also concentrations CA and CB as functions of time.
Similar variable volume situations are found in examples FEDBAT, and
VARVOL.
Air
F
0'yO'T0'P0
Figure 1.18. Entering air and exit gas during the continuous aeration of a bioreactor.
Writing a balance around the combined gas and liquid phases in the reactor
gives,
At steady-state, the accumulation terms for both phases are zero and
For gaseous systems, the quantities are often expressed in terms of molar
quantities.
Often only the inlet air flow rate FQ and the mol fraction of 62 in the outlet
gas, yi9 are measured. It is often assumed that the total molar flow rate of gas is
constant. This is a valid assumption as long as the number of carbon dioxide
mols produced is nearly equal to the number of oxygen mols consumed or if
the amounts of oxygen consumed are very small, relative to the total flow of
gas.
Converting to molar quantities, using the Ideal Gas Law,
pV = nRT
or in flow terms:
pF = N R T
where N is the molar flow rate, R is the gas constant and F is the volumetric flow
rate. Thus, for the inlet gas flow:
P
where NO is molar flow rate of the oxygen entering. Note that the pressure, po,
and temperature, TO, are measured at the point of flow measurement.
Assuming NO = NI, then measurement of NO gives enough information to
calculate oxygen uptake rate, OUR, from the steady-state balance. Thus,
0 = yo NO - yi NI - ro2 VL
OUR = ro2 VL = yo NO - yi NI
If NO is not equal to NI, then this equation will give large errors in oxygen
uptake rate, and NI must be measured, or determined indirectly by an inert
balance. This is explained in the Sec. 1.3.4.4 below.
Differences in the inlet and outlet gas flow rates of a tank fermenter can be
calculated by measuring one gas flow rate and the mol fraction of an inert gas
in the gas stream. Since inert gases, such as nitrogen or argon, are not
consumed or produced within the system (rinert = 0), their mass rates must
40 1 Modelling Principles
therefore be equal at the inlet and outlet streams of the reactor, assuming
steady-state conditions apply. Then for nitrogen
NO YO inert = NI yi inert
Since the inlet mol fraction for nitrogen in air is known, the outlet mol fraction,
yi inert' must be measured. This is often done by difference, having measured
the mol fraction of oxygen and carbon dioxide concentration in the exit gas.
This relation indicates that 1 mol of pyruvic acid reacts with 1 mol of NADH to
produce 1 mol of lactic acid.
Another example of stoichiometry is that of the oxidative decarboxylation of
pyruvic acid to yield acetyl-CoA
C3H4O3 + CoA-SH + NAD+ -> CH3CO-S-CoA + CO2 + NADH + H+
Pyruvic Acid Acetyl-CoA
where 1 mol of glucose reacted, consumes 7 mol of water and produces 6 mol
of carbon dioxide. Here the molar quantities of NADPH and ATP produced
and consumed, respectively, are shown.
For many complex biological reactions, however, not all the elementary
reactions and their contributions to the overall observed reaction stoichiometry
are known (Roels, 1983; Bailey and Ollis, 1986; Moser, 1988).
Thus the case of a general fermentation is usually approximated by an
overall reaction equation, where
Substrate + Nitrogen source + O2 -> Product + CO2 + H2O
where c, d and f are the fractions of carbon converted to biomass, product and
CO2, respectively.
Elemental balances for C, H, O and N give
C 1 = c +d +f
H m+3 a = c p + dr + 2e
O l+2b =cn + ds + e + 2f
N a =cq+dt
In this general problem there are too many unknowns for the solution method
to be taken further, since the elemental balances provide only four equations
and hence can be solved for only four unknowns. Assuming that the elemental
formulae for substrate, biomass and product and hence 1, m, n, p, q, r, s and t are
defined, there still remain six unknown stoichiometric coefficients a, b, c, d, e
and f and only four elemental balance equations. Thus the elemental balances
need supplementation by other measurable quantities such as substrate, oxygen
and ammonia consumption rates (assuming controlled pH conditions), and
carbon dioxide or biomass production rates, such that the condition is satisfied
that the number of unknowns is equal to the number of defining equations. In
principle the problem then becomes solvable. In practice, there can be
considerable difficulties and inaccuracies involved, although the technique of
elemental balancing can still provide useful data. The application of so-called
macroscopic principles (Roels, 1980, 1982 and 1983; Heijnen and Roels, 1981)
introduces a more strict systematic system of analysis. This is depicted in Fig.
1.19.
1.3 Formulation of Balance Equations 43
<P2 Substrate
H
C32 b2 Oc2 Nd2
04 N Source
The system is represented here in terms of the various flow inputs, where f is the
corresponding flow vector
The steady state balance for the system is then represented by: <|) E = 0
where E is the elemental composition matrix
E = a 4 b4 c 4 d 4 O4
0 0 2 0 O5
1 0 2 0 O6
0 2 1 0 <D7
substrate (S) ys = 4 + m - 2 1
biomass (X) yx = 4 + p - 2 n - 3 q
product (P) yp = 4 + r - 2 s - 3 t
The reductances for NHs, f^O and CC>2 are of course zero.
Often the elemental composition of the substrate is not known and then the
reductance method may be supplemented by the following regularities, which
apply to a wide variety of organic molecules.
Yield coefficients are biological variables, which are used to relate the ratio
between various consumption and production rates of mass and energy. They
are typically assumed to be time-independent and are calculated on an overall
basis. This concept should not be confused with the overall yield of a reaction
or a process. The biomass yield coefficient on substrate (Yx/s) is defined as:
v
Y
x/S = rs
In batch systems, reaction rates are equal to accumulation rates, and therefore
/dX\
Y
IdTj dX
X/S = - TdST = - dS"
IdTj
X(t)-X(t=0)
S(t=0)-S(t)
rs
SQ-S!
where index 0 and 1 indicate feed and effluent values, respectively.
In the literature, yield coefficients for biomass with respect to nutrients are
most often used (e.g. Dekkers, 1983; Mou and Cooney, 1983; Roels, 1983;
Moser, 1988). In many cases this is very useful because the biomass
composition is quite uniform, and often product selectivity does not change
very much during an experiment involving exponential growth and associated
production. Some useful typical values are given in Table 1.1.
Table 1.1. Typical mass and energy yield values (Roels, 1983; Atkinson and
Mavituna, 1991).
Y
X/S,aer c-mol / c-mol 0.4-0.7
Yx/S,anaer c-mol / c-mol 0.1-0.2
Yx/02 (Glucose) c-mol / mol 1-2
YX/ATP c-mol / mol 0.35
Y
Q/O2 kJ / mol 380-490
YQ/C02 kJ / mol 460
YQ/x,aer (Glucose) kJ / c-mol 325-500
Yq/x,anaer kJ / c-mol 120-190
In many biological systems, processes with large ranges of time constants have
to be described. Usually it is important to start with a simplification of a system,
focusing on the most important time constant or rate. For example, if the
growth of an organism is to be modelled with a time constant of the order of
hours, it is very useful to ignore all aspects of biological evolution with time
constants of years. Also fast equilibrium reactions or conformational changes
of proteins having time constants below milliseconds should be ignored. Fast
reactions can, however, be very important when considering allosteric activation
or deactivation of proteins or simply pH-changes during biochemical reactions.
1.3 Formulation of Balance Equations 47
pH changes can have dramatic effects on the enzyme and microbial activity but
can also strongly influence absorption and desorption of carbon dioxide.
A typical equilibrium reactions is the dissociation of a receptor-protein ligand
complex, RL, into the free protein, L, and the receptor protein, P
CLP k_j
In most cases such relationships can be used to express the concentration of all
concentrations in explicit form using, e.g. a protein balance.
C Ptot r1 _i_ c*
~ ^P ~r ^LP
^ ^ KD
- ^ptot Cr + K
D
Acid Base' + H+
where CAcid is the concentration of the undissociated acid and CBase" is the
concentration of the corresponding base (salt).
An ion charge balance can be written
In the pH range of interest (usually around pH = 7) all strong acids and strong
bases are completely dissociated. Moderately strong acids and bases exist in
both the dissociated and non-dissociated forms,
In the usual pH range the sum of the cations are much larger than the H+
ions.
ICK+CH+
where ]CK+ is the total cation concentration.
Negative ions originate mainly from strong acids (e.g. Cl% SO42') but also
arise from weak acids (Ac", Pr, Bu~, HCO3'). The concentration of CO32' is
always much smaller than that of
The ion balance reduces to
KBj C C
V1 KAi
KW Btot,i+ K+ =
where KAI are the acid dissociation constants (e.g. KAC); KBI are the base
dissociation constants (e.g. KNHS); KW is the dissociation constant of water;
Cfitot,i are the total concentrations of base i; CAtot,i are the total concentrations
of acid i and EC An" is the sum of the anions.
The pH can be estimated from the above equation for any situation by
solving the resulting non-linear implicit algebraic equation, provided the total
concentrations of the weak acids, CAtot,i> weak bases, CBtot,i cations of strong
bases, CK+, and anions of strong acids, CAIT> are known.
1.3 Formulation of Balance Equations 49
The example ANAMEAS, Sec. 8.8.6 includes this ion balance for pH
calculation. This equation represents an algebraic loop in a dynamic simulation
which is solved by iteration at each time interval until 8 approaches zero. This is
accomplished with the root-finding feature of Berkeley Madonna.
If there is pH control, then strong base or acid would be usually added. The
addition of strong alkali for pH control would cause an increase in CK+
which in accordance with the above equation would result in a decrease of CH+.
An alternative approach, which avoids an algebraic loop, is to treat the
instantaneous equilibrium reactions as reactions with finite forward and
backward rates. These rates must be adjusted with their kinetic constants to
maintain the equilibrium for the particular system; that is, these rates must be
very fast compared with the other rates of the model. This approach replaces
the algebraic loop iteration with a stiff er and larger set of differential equations.
This could be an advantage in some cases.
Accumu-^ Rate of^ 'Rate of ^ 'Rate of" 'Rate of > 'Rate of >
lation energy energy energy energy energy
rate of in by out by out by generated added by
^Energy , ^flow , flow <transfer> ^by reactiony ^agitation ^
50 1 Modelling Principles
The above balance in word form is now applied to the measurable energy
quantities of the continuous reactor shown in Fig. 1.20.
AL H
agit
\
M)' PO P1
' '
U,A,T S
l!
Figure 1.20. A continuous tank fermenter showing only the energy-related variables.
An exact derivation of the energy balance was given by Aris (1989) as,
S
"dT = - h n ) ) + U A (Ta - TI) + rQ V + AHagit
agi
where ni is the number of moles of component i, cpi are the partial molar heat
capacities and hi are the partial molar enthalpies. In this equation the rate of
heat production, TQ, takes place at temperature TI. If the heat capacities, cpi, are
independent of temperature, the enthalpies at TI can be expressed in terms of
heat capacities as
hn = hio + Cpi (TI -TO)
and with
S S
2>i0 - 2X = vp
1=1 1=1
Thus with these simplifications,
The units of each term of the equation are energy per time (kJ/h or kcal/h).
1.3 Formulation of Balance Equations 51
Accumulation Term
Densities and heat capacities of liquids can be taken as essentially constant.
dT
VpcPdF
has units:
m3 (kg/m3) (J/kg K) K _ kJ
s ~ s
Thus the accumulation term has the units of energy/time (e.g. J/s)
Flow Terms
The flow term is F p CP (T0 - TI)
energy =
energy
~B55~ area time degree (area) (degree)
The sign of the temperature difference determines the direction of heat flow.
Here if T a > TI heat flows into the reactor.
energy _ energy
volume time ( volume ) - time
rq = rs YQ/S
rq = ro2YQ/Q2
In terms of a heat of reaction per mol of substrate and a substrate uptake rate,
rQ = AHr,s rs
Here rs is the substrate uptake rate and AHr?s is the heat of reaction for the
substrate, for example J/mol or kcal/kg. The rs AHr>s term therefore has
dimensions of (energy/time volume) and is equal to TQ.
For aerobic fermentation, the heats of reaction per unit volume of reactor are
usually directly related to the oxygen uptake rate, ro2-
Thus for a constant-volume batch reaction with no agitation heat effects, the
general energy balance is
where YQ/Q2 often has a value near 460 kJ/mol 2, as given in Table 1.1.
1.3 Formulation of Balance Equations 53
UA(Ti-Ta) = r 0 2YQ/o 2 V
Both physical and biological information are required in the design and
interpretation of biological reactor performance, as indicated in Fig. 2.1.
Physical factors that affect the general hydrodynamic environment of the
bioreactor include such parameters as liquid flow pattern and circulation time,
air distribution efficiency and gas holdup volume, oxygen mass transfer rates,
intensity of mixing and the effects of shear. These factors are affected by the
bioreactor geometry and that of the agitator (agitator speed, effect of baffles)
and by physical property effects, such as liquid viscosity and interfacial tension.
Both can have a large effect on gas bubble size and a corresponding effect on
both liquid and gas phase hydrodynamics. The biokinetic input involves such
factors as cell growth rate, cell productivity and substrate uptake rate. Often this
information may come from laboratory data, obtained under conditions which
are often far removed from those actually existing in the large scale bioreactor.
Although shown as separate inputs in Fig. 2.1, there are, in fact, considerable
interactions between the bioreactor hydrodynamic conditions and the cell
biokinetics, morphology and physiology, and one of the arts of modelling is to
make proper allowance for such effects. Thus in the large scale bioreactor,
some cells may suffer local starvation of essential nutrients owing to a
combination of long liquid circulation time and an inadequate rate of nutrient
supply, caused by inadequate mixing or inefficient mass transfer. Agitation and
shear effects can affect cell morphology and hence liquid viscosity, which will
also vary with cell density. This means that the processes of cell growth affect
the bioreactor hydrodynamics in a very complex and interactive manner.
Changes in the cell physiology, such that the cell processes are switched from
production of further biomass to that of a secondary metabolite or product, can
also be affected by selective limitation on the quantity and rate of supply of
some essential nutrient in the medium. This can in turn be influenced by the
bioreactor hydrodynamics and also by the mode of the operation of the
bioreactor.
The overall problem is therefore very complex, but as seen in Figure 2.1,
when all the information is combined successfully in a realistic and well
founded Bioreactor Model, the results obtained can be quite impressive and
may enable such factors as cell and product production rates, product
selectivities, optimum process control and process optimization to be
determined with some considerable degree of confidence.
Production rate
Selectivity
Control
The rates of cell growth and product formation are, in the main, dependent on
the concentration levels of nutrients and products within the bioreactor. The
concentration dependencies of the reaction or production rate are often quite
simple, but may also be very complex. The magnitude of the rates, however,
depend upon the level of concentrations, and it will be seen that concentration
levels within the bioreactor depend very much on its type and mode of
operation. Differing modes of operation for the bioreactor can therefore lead
to differing rates of cell growth, to differing rates of product formation and
hence to substantially differing productivities.
Generally, the various types of bioreactor can be classified as either stirred
tank or tubular and column devices and according to the mode of operation as
batch, semi-continuous or continuous operation.
2.2 Bioreactor Operation 57
Most industrial bioreactors are operated under batch conditions. In this, the
bioreactor is first charged with medium, inoculated with cells, and the cells are
allowed to grow for a sufficient time, such that the cells achieve the required
cell density or optimum product concentrations. The bioreactor contents are
discharged, and the bioreactor is prepared for a fresh charge of medium.
Operation is thus characterized by three periods of time: the filling period, the
cell growth and cell production period and the final emptying period as
depicted in Fig. 2.2. It is only during the reacting period, that the bioreactor is
productive. During the period of cell growth, strictly speaking, no additional
material is either added to or removed from the bioreactor, apart from minor
adjustments needed for control of pH or foam, small additions of essential
precursors, the removal of samples and, of course, a continuous supply of air
needed for aerobic fermentation. Concentrations of biomass, cell nutrients and
cell products thus change continuously with respect to time, as the various
constituents are either produced or consumed during the time course of the
fermentation, as seen in Fig. 2.3.
concentration
A biomass
ubstrate
product
time
During the reaction period, there are changes in substrate and product
concentration with time, and the other time periods are effectively lost as
regards production.
Since there is no flow in or out of the bioreactor, during normal operation,
the biomass and substrate balances both take the form,
For fed-batch operation, the cell balance follows the same form as for batch
operation, but since additional substrate feeding to the reactor now occurs, the
substrate balance takes the form:
Rate
( <* "| ( Substrate \ ( Substrate >|
accumulation = (f^d [n) _ consumption
V of substrate J \ rate )
One other balance equation, however, is also necessary, i.e. the total mass
balance,
Further extensions of fed batch operation are possible, such as the cyclic or
repeated fed batch, which involves changing volume with a filling and
emptying period. The changing reactor concentrations repeat themselves with
each cycle. This operation has similarities with continuous operation and
approaches most closely to continuous operation, when the amount withdrawn
is small and the cycle time is short. The simulation examples FEDBAT, Sec.
8.1.3 and in Sec. 8.3 (VARVOL, PENFERM, PENOXY, ETHFERM, REPFED)
allow detailed investigations of fed batch performance to be made on the
computer.
As shown later, these two differing forms of continuous reactor operation have
quite different operational characteristics. Both however are characterized by
the fact that after a short transient period, during which conditions within the
bioreactor change with time, the bioreactor will then achieve a steady state. This
means that operating conditions, both within the bioreactor and at the
bioreactor outlet, then remain constant, as shown in Fig. 2.6.
time
Tank Tube
So So
Cone. Cone.
distance distance
Figure 2.7. Profiles of substrate and product in steady state continuous tank and tubular
reactors.
lowest concentration at the reactor outlet. The product concentration rises from
inlet to outlet. These differences arise because in the tank reactor the entering
feed is continuously being mixed with the reactor bulk contents and therefore
being diluted by the tank contents. The feed to the tubular reactor, however, is
not subject to mixing and is transformed only by reaction, as material moves
down the reactor.
No real situation will exactly correspond to the above idealized cases of
perfect mixing or zero mixing (plug flow), although the actual behavior of
tanks and tubes tends in the limit towards the corresponding idealized model.
The characteristics of continuous operation are as follows:
The balance equations at steady state for a well-mixed tank reactor have the
form
0 = (Input) - (Output) + (Production)
since at steady-state the rate of accumulation and therefore the rate of change is
zero.
This equation predicts that the reaction rate causes a depletion of substrate
from the feed condition to the outlet, (the product will increase) and that the
rate of production can be obtained from this simple balance:
Cone.
distance
Table 2.2 lists the main operating parameters for the three differing modes of
bioreactor operation.
Stirring rate
2.2 Bioreactor Operation 65
As explained in Sec. 2.1, a realistic bioprocess model will usually require the
input of kinetic rate data. In the case of even simple chemical reactions, this
data has to be obtained by laboratory experiment. Since biochemical reactions
are controlled by enzymes, it is appropriate to start with a consideration of
simple enzyme kinetics (Sec. 3.1), In the case of modeling the behavior of
enzyme reactors, knowledge of the enzyme reaction kinetics is most important.
The sheer complexity of the biological reactions, occurring in a living cell,
seem to imply an almost impossible task in obtaining meaningful rate data for
biological modelling applications. Fortunately this is not the case and, as
shown in Section 3.2, a quite reasonable overall description of cell growth rate
data is possible, based on an overall empirical relationship, the Monod
Equation, which has been found to give a good fit to many general
observations of cell growth. This overall view, based on the net result of many
simultaneously occurring and highly interacting biochemical reactions, of
course represents an incredible oversimplification of the actual situation.
Fortunately it seems to work in many instances and can also be easily modified
to allow for the uptake of substrate by the cells and to include such additional
effects, as substrate limitation, multiple substrate limitation and product
inhibition. It is interesting, that the basic enzyme rate equation, or Michaelis-
Menten equation, based on the theory indicated in Sec. 3.1, is of the same basic
form as the empirically-based Monod equation for the growth of micro-
organisms.
When used in this manner, the cell kinetics are completely devoid of any
mechanistic interpretation and constitute what is known as an unstructured
kinetic model (Fig.3.1 A). In other cases, it may be necessary to look in some
detail at individual cell processes and reactions, in order to obtain a more
realistic description, thus leading on to the use of structured kinetic models
(Fig. 3.IB) as described in Sec. 3.3. In a most simple case, the cells are
composed of a catalytic part comprising proteins, RNA, DNA and other cellular
compounds and of a storage part, e.g. poly-hydroxy-alkanoic acids (PHAs) or
inclusion bodies of recombinant proteins. A most simple segregated model
considers different stages of cells and therefore a distribution of cell stages in a
culture without structuring the cell composition (Fig. 3.1C). In the most
realistic, but most complex situation the model is structured and segregated
(Fig. 3.ID). For the purpose of this book, the differences of these models can
be best described by their different balance regions.
Non-structured Structured
B
3CO
t
o
Figure 3.1. Types of kinetic models for cells. Balance regions: A Total cell biomass, B Cell
parts, C biomass parts, D Biomass and cell parts.
>
S 4- E < ES
ES - P + E
= - ki S E + k_i (ES)
dt
S = S0 (ES) = 0
The concentration changes for a batch reactor are shown qualitatively in Fig.
3.2. While the enzyme concentration is usually much lower than that of the
substrate, most of the enzyme is present during the reaction in the form of the
enzyme-substrate complex, ES.
Analytical solution is then possible by assuming a quasi-steady state for the
enzyme-substrate complex, ES,
d(ES)
-= 0
dt
This assumption is valid for E SQ.
Using the total enzyme mass balance,
EO = E + (ES)
the above equations can be solved for the unknown concentrations E and ES to
give,
and
ki S E k.i + k2
E = E = E
0 ~ k_i .+ k2 k_i + k2
70 3 Biological Kinetics
o
1
Q)
O
-^ time
Figure 3.2. Concentration changes of the reaction species for a simple enzymatic reaction
taking place in a batch reactor.
dS _ k2SEo
dt ~
where the parameters in terms of the mechanistic model are for the maximum
reaction rate (kmol/m3min):
v max = k 2 E 0
k-i + k2
KM - ki
The Michaelis-Menten equation exhibits three distinct regions for the reaction
rate. At very low and very high substrate concentrations the rs versus S curve is
essentially linear, as seen in Fig. 3.3.
3.1 Enzyme Kinetics 71
vm
rS
Michaehs-Menten region
vm/2""
0 KM 5 KM 10 KM 15 KM
Figure 3.3. Reaction rate versus substrate concentration for the Michaelis-Menten equation.
_ VmaxS
1 KM I
rs v
max S max
Typical values of the enzyme kinetic constants are given in Table 3.1. It is
interesting to note that the rate of formation of the enzyme-substrate complex
can be extremely fast, with the constant ki approaching 1 x 1010 L/mol s. This
is the maximum value for a rate constant of a reaction that is limited by
diffusion of a small substrate molecule in aqueous solution.
E+SES-P
c c non-competitive (i + I/KI) (KM + S)
El ESI
Usually, the substance I is the substrate or the product, and the reaction kinetics
are known as substrate or product inhibition, respectively.
Allosteric Kinetics
A simple model to describe allosteric inhibition is given, in which the enzyme
can bind to more than one substrate molecule. Thus:
nS + E
z
ESn ) nP + E
Vmax S n
s
" K M n + Sn
r =
K
S,H+CH+CH/ K I,H
76 3 Biological Kinetics
3.1.3 Deactivation
Considering that for a batch reactor, dE/dt = r^, the integrated form can be
written as
E = E 0 e- kdt
= MOJL kdt
KM+S
This equation suggests an exponential decrease of reaction rate regardless of
substrate concentration. The simulation example DEACTENZ, Sec. 8.4.12
illustrates this.
Engineering models for the kinetics of deactivation are given by Prenosil et
al. (1987).
3.1.4 Sterilization
Similar to enzyme deactivation, sterilization kinetics can be viewed as a process
of inactivation or the removal of viable organisms or cells from the system.
Inactivation can be achieved by using heat, radiation or chemicals. It is a
statistical process, with the rate of killing being usually proportional to the
number of the organisms at any time. Therefore it can be described again by
first-order kinetics:
rd = - k d X
where
x
_ ,,-kdt
where XQ is the initial live biomass concentration, X is the viable biomass after
the treatment time t, and kd is the specific deactivation constant (1/s).
The sterilization time will depend on the initial level of contamination. For
this purpose the D-value is defined as the treatment time required to reduce the
population by a factor of ten. This time is related to the rate constant by
2.3
Under ideal conditions for growth, when a batch fermentation is carried out, it
can be observed experimentally that the quantity of biomass, and therefore also
the concentration, increases exponentially with respect to time. This phenomena
can be explained by the fact that all cells have the same probability to multiply.
Thus the overall rate of biomass formation is proportional to the biomass itself
This leads to an autocatalytic reaction, which is described by a first order rate
expression as
where rx is the rate of cell growth (kg cell/m3 s), X is the cell concentration
(kg cell/m3) and k is a kinetic growth constant (1/s). For a batch system, this is
equivalent to,
where dX/dt is the rate of change of cell concentration with respect to time
(kg cell/m3 s). The analytical solution of this simple, first-order differential
equation is of the form
In X = k t + In X0
or,
= ekt
Plotting experimental growth data in the form of the natural logarithm of cell
concentration versus time will often yield a straight line over a large portion of
the curve, as shown below in Fig. 3.6.
InS
Limitation \Stationary
InX
X Death
Exponential
Lag
time
In the range from ti to t2 the logarithmic curve is linear, and this is the region
of exponential growth. Three other regions can be identified: between t = 0
and ti, there exists a period of cell adaptation or lag phase, and before t2 there
is a region where the growth is limited by the lack of a particular substance,
which is known as the limiting substrate.
The slope of the linear part of the curve between ti and t2 is the growth rate
per unit mass of cells or specific growth rate and is given the symbol \i.
1 dX
=
X" "dT = = specific growth rate = |i
In many processes cells begin to die (after ts), because of lack of nutrients,
toxic effects or cell aging. This process can typically be described by a first
order decay,
rd = - kd X
where rd is the death rate and k^ is the specific death rate, with the same
dimensions as the specific growth rate. This expression is identical with
sterilization kinetics, Sec. 3.1.4.
The exponential and limiting regions can be described by a single relation,
that sets JLI equal to a function of substrate concentration. It is observed
experimentally that |a is at a maximum when the particular limiting substrate
concentration S is large, and for low concentration ja is proportional to S. Over
the whole range from low to high S, |i is described by the following Monod
equation.
3.2 Simple Microbial Kinetics 79
Thus |i varies with S in the same fashion as does the enzymatic rate of
Michaelis-Menten kinetics.
Again, this is a two-parameter equation involving two constants, the
maximum specific growth rate |im and the saturation constant KS. It is best
considered to be an empirical relation, but since it has the same form as the
Michaelis-Menten enzyme kinetics equation, it is sometimes taken to be related
to a limiting enzymatic step. Although very simple, it often describes
experimental data for growth rates very well. The form of this relation is shown
in Fig. 3.7.
M
Monod Relation
Figure 3.7. Specific growth rate versus limiting substrate concentration according to the
Monod relation.
S - 0,
Urn
S = KS, Jl = ~
The first introductory simulations in Sec. 8.1 are based on Monod kinetics.
When two substrates can be limiting, it is often the case that a double Monod
type relationship can be used, as given in Sec. 3.2.4 and as shown by the
simulation examples NITRIF, Sec. 8.5.3, and BIOFILM, Sec. 8.7.1.
80 3 Biological Kinetics
1.0 -,
0.8 -
K
0.6 -
I
0.0
Figure 3.7. Substrate inhibition kinetics for various values of KS and Kj. The parameters used
are as follows: For all curves |im = 1.0 1/h. Curve A: KS = 1 and KI = 10, Curve B: KS = 0.1 and
KI = 10; Curve C: KS = 1 and KI = 20; Curve D: KS = 0.1 and KI = 20. The units of KS, KI and S
are g/m^.
3.2 Simple Microbial Kinetics 81
The substrate inhibition kinetic curve has the following characteristics, which
result from the kinetic equation:
1) When S = Ks
_
- - 2 + Ks/Kj
2) When S = KI
- 2 + Ks/Ki
Mm
M- =
2 (Ks/Ki)-5 + 1
When the formation of product inhibits the rate of cell growth, the basic Monod
equation can be modified, by the addition of a product inhibition term P/Kj.
Thus,
- KX + S
encompasses exponential growth and the levelling off to zero growth rate at
high X. For a batch fermentation the biomass balance is,
= aX-bX2
dX
When X is large,
0 = a X - b X2
and thus
X = a/b
In this way either substrate may be limiting under conditions when the other
substrate is in excess. Note that the multiplicative effect gives for S\ = K\ and
3.2 Simple Microbial Kinetics 83
Double-Monod kinetics
can also be written for two substrates as parallel reactions, according to
lSi k2S2 V 1
This form gives an additive, fractional contribution for each substrate. Thus for
Si = KI and 82 = K2, the result is |Li = |Lim/2. For the case Si = 0 and 82 large,
then ji = |iim k2/(ki+k2). Each substrate thus allows a different maximal growth
rate. If both Si and 82 are large then |ii = |iim. Note that the flexibility of this
kinetic form requires twice as many kinetic parameters as the simpler double
Monod kinetics. An example of this kinetics is the parallel use of alternative
substrates, such as various types of sugars.
Diauxic Monod Growth can be modelled for two substrates by the relation
K 2 + S2 + S / K !
in this way the consumption of substrate 82 will be inhibited until Si is
exhausted, for suitably low values of Kj. Diauxic growth can be observed in
many organisms. An example is E. Coli, where the uptake of lactose is
repressed in the presence of glucose. The simulation example SUBTILIS, Sec.
8.9.2 uses this kinetic form.
where rs is the rate of substrate uptake by the cells (kg substrate/m3 s). As
explained in Section 1.3.5, YX/S is the stoichiometric factor or yield coefficient
and relates mass cells/mass substrate.
84 3 Biological Kinetics
For the specific substrate uptake rate (kg S/kg biomass h),
rs
qs = x
For the specific oxygen uptake rate (kg C>2/kg biomass h),
r
O2
qo2 =
For the specific carbon dioxide uptake rate (kg CCVkg biomass h),
qco2 = "IT"
For the specific product production rate (kg P/kg biomass h),
qp =
Specific rate quantities may take simple or complicated forms, for example, for
the case of substrate:
rs = y^" - m X
then,
qs =
where |i is also a function of S.
3.2 Simple Microbial Kinetics 85
where rp is the rate of product formation (kg product/m3 s). YX/P is a yield
factor (kg cell produced/kg product produced), relating the growth and product
stoichiometry in the "growth associated" term of the Luedeking-Piret equation,
and b is a non-growth related term and is important for cultures which produce
product independent of growth. Often both coefficients of the above equation
(YX/P and b) are not constant but are functions of substrate or product
concentration.
When little is known about the detailed kinetics of product formation, a more
general expression rp = qp X is used, where qp will usually vary with culture
conditions and concentrations.
86 3 Biological Kinetics
Interaction-type Organisms
A B
Neutralism 0 0
Commensalism 0 +
Mutualism
Predator-prey
Predator-prey
Amensalism 0
Amensalism 0
Competition
Predator-Prey Kinetics
Organism A consumes substrate S, and organism B consumes organism A.
Commensalism
Organism A uses substrate 82 to produce product P; organism B uses substrate
Si to produce product 82, which benefits organism A since product 82 acts as
its substrate.
A
81
The following processes with the compound 82, shown in brackets, involve this
form of commensalism:
- nitrification (NC>2~ )
- anaerobic digestion (organic acids)
- methanogenation (H2, 62) as found in simulation example ANAMEAS,
Sec. 8.8.7.
This effect may be found in the removal of toxic wastes in mixed cultures with
multiple carbon sources. An example is found in ANAMEAS in which the
hydrogen substrate of the methanogens (A) inhibits the acetogenic organisms
(B).
Organism A utilizes the product from organism B, which also helps B because
PB inhibits its growth.
An example of kinetic modelling is presented for this case, in which the growth
of the two organisms, A and B, takes place in a batch reactor with initial
substrate concentrations SIQ and 820- The growth rate is expressed by Monod-
type kinetics and constant yield factors are used to express the substrate uptake
and product formation rates.
Substrate Si balance,
_
dt -
Substrate 82 balance.
_
dt -
BS2
Product PB balance,
dPB
IT A PB
Species A balance,
dXA
"3f" =
Species B balance,
3.2 Simple Microbial Kinetics 89
dXB =
~3T ^B XB
The kinetics are given by Monod-type relations, with a double form of Monod
equation employed for species A and a product inhibition term employed for
B,
Si PB
HA - MmA KI + Si
B = K2 + S2 + (PB/KI)
Other examples of interacting microorganism effects are given in the
simulation examples ACTNITR (neutralism), Sec. 8.4.3, COMPASM
(competitive assimilation and commensalism), Sec. 8.8.2, MIXPOP (predator-
prey population dynamics), Sec. 8.8.4 and TWOONE (competition between
organisms), Sec. 8.8.5.
(0
'55
CO
g0)
TJ
"5
o>
W
'(0
0)
0)
D)
O
I
O
Figure 3.9. Reaction scheme of anaerobic degradation. The symbols are: Poly - polymer
material (proteins, fats, hydrocarbons, etc.); XJJY - Biomass hydrolyzing Poly; Mono -
monomeric materials from hydrolysis of Poly; XAG ~acid generating biomass; HPr - propionic
acid; Pr" - propionate; Xpr - biomass growing on propionate; HBu - butyric acid; Bu" - butyrate;
XBU - biomass growing on butyrate; HAc - acetic acid; Ac" - acetate; XAC - biomass growing on
acetate; XH - biomass growing on hydrogen and carbon dioxide. Dashed arrows indicate gaseous
compounds transfer to the liquid phase. T - gaseous compounds transferred to liquid gas phase.
The respective reaction rates rj for the production of biomass Xi, for the
consumption of substrate Si and for the formation of product Pi in each step
are:
rxi = Hi Xi - kdi Xi
3.3 Structured Kinetic Models 91
Hi Xj
rpi =
where k^ is the specific death rate, including maintenance metabolism, and the
specific growth rates take the Monod form,
Mimax Si
W - KSi + Si
Mimax Si
These kinetics can then be combined with the mass balances as discussed in Ch.
4 for each component, Si and Pi, and for the biomass balances for each
organism type, Xi
Following this approach a model was established (Denac et al., 1988) and
combined with particular control algorithms to simulate a continuous anaerobic
digestor with feed rate control. This included a gas phase balance,
thermodynamic equilibrium constraints and acid-base equilibria using an ion
charge balance (Sec. 1.3.6.2). The simulations were used to adjust the control
parameters, which were employed on laboratory reactors (Heinzle et al., 1992).
The simulation example ANAEMEAS, Sec. 8.8.7, gives details concerning this
model.
Membrane transport.
- Accumulation of storage materials (Heinzle and Lafferty, 1980; Heinzle
et al., 1983). See example PHB, Sec. 8.2.4.
- Morphological changes, e.g., branching of filamentous organisms,
volume to surface ratio of yeast cells (Fig. 3.10), Furukawa et al., 1983.
2.0 -
L/d
1.5
S/V
1.0
Figure 3.10. Dissolved oxygen concentration, DO, influenced the shape of yeast in
continuous culture as given by the ratios of length/diameter (L/d) and surface/volume (S/V).
3.3 Structured Kinetic Models 93
Fig. 3.11 represents the process of cell growth and the synthesis of intracellular
product PHB. Residual biomass (R) is the difference between total cell dry
mass (X) and product PHB (P). Synthesis of PHB occurs with a single limiting
substrate NH4+ (S) and constant dissolved gas concentrations of H2, C>2 and
CO2 (So). Mass flows are indicated by solid lines, and regulatory mechanisms
are symbolized by dashed lines.
inhibiting
inhibiting
Figure 3.11. Schematic diagram of growth and synthesis of the intracellular product PHB (P)
with constant concentrations of dissolved gases H2, O2, and CO2 (SG) X is the total biomass
(X=R+P).
94 3 Biological Kinetics
It can be seen that the catalytically active biomass, R, is produced from both S
and SQ. During exponential growth the PHB content is constant, and thus the
rate of intracellular product formation is proportional to the rate of formation
of the residual biomass. On this basis, the basic mass balance equations for a
batch process can be formulated as shown in the simulation example PHB, Sec.
8.2.4. This model was used successfully in describing experimental batch
growth and the PHB product formation (Heinzle and Lafferty, 1980), as shown
in Fig. 3.12.
3 -
2 -
1 -
Figure 3.12. Comparison of simulation results from the structured PHB model with
experimental data (Heinzle and Lafferty, 1980).
15 10
10 0.5
X [g/L]
'1 DO [mg/L]
0.05
0.0
0.5
0
E [g/L] S [mg/L]
2
0
100
2.0
50 RQ
1.5
q X [mmol/h L]
1.0
0 10 20
t[h]
Figure 3.13. Oscillating profiles from a continuous culture of S. cerevisiae. (Heinzle et al.,
1983). Symbols used are X (total biomass), DO (dissolved oxygen), E (ethanol), S (glucose),
QCO2 and QO2 (specific gas reaction rates), and RQ (respiratory quotient).
(g/L)
r
10 15
Time (h) Time (h)
D=0.05 h'1
Figure 3.15. Simulation of the Baker's yeast model (simulation example YEASTOSC, Sec.
8.8.8, showing oscillations of all the components, Q.
3.3 Structured Kinetic Models 97
- 3
10
- 2
- 1
Figure 3.17. Growth and product formation of Bacillus subtilis at constant DO. Gl - glucose,
X - biomass, Ac - acetoin, Bu - butanediol.
The formulation starts with a steady state ATP balance, which assumes that all
energy-producing steps are balanced by those that consume energy. The form
of this balance is as fpllows:
i A ' I' U
NADH
Cells
6CO 2
NADH
ATP Acetoin ^ > Butanediol
NADH ^ NAD+ r
NAD"1"
ADP ^ ATP
Figure 3.18. Biochemical pathways for the production of acetoin and butanediol.
Using the steady-state approximation that the ATP level does not vary
significantly, allows setting the condition that dATP/dt = 0. The steady state
ATP balance is then solved for qATP/X, which is the rate of ATP available for
growth. The required yields can be calculated from the reactions given in Fig.
3.17. In these calculations it was assumed that at high oxygen concentration 1
mol NADH was converted to 3 mol ATP in the respiratory chain. At low
oxygen concentrations, the conversion equivalent was assumed to be a function
of oxygen as determined by parameter estimation, based on the experimental
data.
The glucose substrate balance can be written in terms of the rates at which
substrate was consumed for complete oxidation (qs/GO2) fr biomass
(qATP/X YX/ATP / YX/S) and for product formation (qs/Ac)'
dS ,. qATP/x YX/ATP x
X
3.3 Structured Kinetic Models 99
dAc
"dt" = ( qS/Ac ~ qAc/Bu + qBu/Ac )
and,
dBu
=
~dt~ ( ^Ac/flu ~ qBu/Ac ) X
In the above balances, all specific rate terms, q, are in the units
(mol/g biomass h). All concentrations (ATP, S, Ac, Bu) are in mol/L units
except X (g/L). All yield coefficients Y are in mol/mol units except when
involving biomass, e.g. YX/S is in units of g/mol.
Empirical Monod-type kinetic relationships, not given here, were established
to calculate the rate of glucose to acetoin, qs/Ac an(i the reversible acetoin to
butanediol rates, qAc/Bu and qBu/Ac as a function of reactant concentrations for
glucose, S, for acetoin, Ac, for butanediol, Bu, and for dissolved oxygen, DO.
Additional empirical kinetic terms were needed to fit the following
experimental observations:
The many kinetic parameters were determined partly by direct experiments and
partly by fitting the data using a parameter estimation computer program. The
influence of oxygen was determined using data from experiments at controlled
oxygen conditions and determining the best values of the oxygen sensitive rates
by parameter estimation procedures. A simple graphical procedure then
allowed determination of the appropriate constants.
100 3 Biological Kinetics
The quantities YATP/NADH and YX/ATP are linearly dependent on each other and
could therefore not be determined from experimental data. The maximum
value of YATP/NADH was arbitrarily fixed at the maximum theoretical value of
3.0, which has a direct influence on the estimation of YX/ATP (here 5.7 g/mol).
Good agreement of the batch curves with the model at constant DO was
achieved as shown in Fig. 3.18. From these results it is seen that the model
predicts the metabolite and biomass profiles. The model was quite versatile and
reasonably accurate, considering the large differences in biomass formation at
high DO (X = 3.4 g/L) and low DO (X = 2.5 g/L), as well as the variation of the
butanediol formation at high DO (Bu = 0.2 g/L) and low DO (Bu = 2 g/L),
X(g/L) GI(g/L)
t(h)
Figure 3.19. Comparison of simulation results with the Bacillus subtilis fermentation (Moes
et al., 1986). X=biomass, Ac=acetoin, Bu=butanediol, Gl=glucose.
Fermentation systems obey the same fundamental mass and energy balance
relationships as do chemical reaction systems, but special difficulties arise in
biological reactor modelling, owing to uncertainties in the kinetic rate
expression and the reaction stoichiometry. In what follows, material balance
equations are derived for the total mass, the mass of substrate and the cell mass
for the case of the stirred tank bioreactor system as shown in Fig. 4.1.
' X0! F0 1f F1
In this generalized case, feed enters the reactor at a volumetric flow rate FQ, with
cell concentration, XQ, and substrate concentration, SQ. The vessel contents,
which are well-mixed are defined by volume V, substrate concentration Si and
cell concentration X\. These concentrations are identical to those of the outlet
stream, which has a volumetric flow rate FI.
Organism balance:
d(V Xi)
J J t = FO XQ - FI X j + rx V
where the units are: V (m3), p (kg/m3), F (m3/s), S (kg/m3), X (kg/m3) with rs
and rx (kg/m3 s).
The rate expressions can be simply:
= KS + S!
-*x
fs =
Yxl
Jiiilii
Total balance:
Substrate balance:
-
dSi
V^r = r s V
Organism balance:
V
T = rxV
Suitable rate expressions for r$ and rx and the specification of the initial
conditions would complete the batch fermenter model, which describes the
exponential and limiting growth phases but not the lag phase. The simulation
example BATFERM, Sec. 8.1.1, demonstrates use of this model.
104 4 Bioreactor Modelling
A chemostat, as shown in Fig. 4.3, normally operates with sterile feed (Xo = 0)
at constant volume steady state conditions, meaning that dV/dt = 0,
d(VSi)/dt = 0, d(VXi)/dt = 0. In addition constant density conditions can be
taken to apply
UN
o = FQ-
The dynamic component balances are then,
Substrate balance
VdSi
= F(S 0 -Si)
Cell balance
VdXi
= F(X0-Xi)
0 = F (S0 - Si) + rs V
4.1 General Balances for Tank-Type Biological Reactors 105
Chemostats normally operate with sterile feed, XQ = 0, and hence for the cell
balance,
0 = - F X! + rx V
Here D is the dilution rate and is equal to 1/T, where i = V/F and is equal to the
tank mean residence time.
Thus the specific growth rate in a chemostat is controlled by the feed flowrate,
since [I is equal to D at steady state conditions.
Since |Li, the specific growth rate, is a function of the substrate concentration,
and |Li is also determined by dilution rate, the flow rate F then also determines
the outlet substrate concentration Si. The last equation is, of course, also
simply a statement that the quantity of cells produced is proportional to the
quantity of substrate consumed, as related by the yield factor YX/S-
The curves in Fig. 4.4 represent solutions to the steady-state chemostat
model as obtained from the simulation example CHEMOSTA, Sec. 8.4.1, with
KS = 1.0.
The variables Xi, Si, as well as the productivity DXi are plotted versus D.
Thus as flow rate is increased, D also increases and causes the steady state value
of Si to increase and the corresponding value of Xi to decrease. It is seen
when D nears |im, Xi becomes zero and Si rises to the inlet value SQ. This
corresponds to a complete removal of the cells by flow out of the tank; a
phenomenon known as "washout". Fig. 4.4 shows washout occurring for D well
below |jm, which is caused by the large value of KS. . When D is nearly zero
(low flow rates) then Si approaches zero, and Xi approaches YSo The
productivity of biomass, DXi, (kg X/m3 h) passes through a maximum rather
close to the washout region.
106 4 Bioreactor Modelling
X1,S1,DX1
10.0 T
5.0 -
Figure 4.4. Variation of the steady state variables in a chemostat with Monod kinetics as a
function of dilution rate.
As shown in Fig 4.5 the outlet is zero for a fed batch fermenter, and the inlet
flow, FO, may be variable. As a result the reactor volume will change with time.
4.1 General Balances for Tank-Type Biological Reactors 107
S.F,
o1 ' o
T7
F
dT = 0
d(VSi)
dt = F0 S0 + rs V
d(V XQ
dt = rxV
Expanding the differential terms, which are products of the two variables V
and Si and V and Xi, respectively, and substituting for dV/dt = FQ gives:
VdSi
= F 0 (S 0 -Si)
The above equations are identical to those for a constant volume chemostat
reactor. It can be shown by simulation that a quasi-steady state can be reached
where dXi/dt = 0 and fi = F/V (Dunn and Mor, 1975) as seen in the Fig. 4.7.
Since V increases, therefore n must decrease, and thus the reactor moves
through a series of changing steady states for which |a = D, during which Si
and p decrease, and Xi remains constant. This is analogous to a constant
volume reactor with slowly decreasing F. These phenomena are demonstrated
by the simulation examples FEDBAT, Sec. 8.1.3, and VARVOL, Sec. 8.3.1.
108 4 Bioreactor Modelling
A fed batch fermenter, in which the inlet feed rate is very low, will not exhibit
a large increase in volume and will not reach a quasi-steady state, unless X is
very high. Assuming V to be approximately constant, the general equations
can be integrated analytically for the case of |j = constant, giving an
exponential increase in X. The constant |u condition is maintained by constant
Si, which can be obtained via exponential feeding. Another phenomenon can
be proven from these equations for the case of constant feed rate and
essentially constant V; this is the linear growth situation, where X increases
linearly with time. As shown in Fig. 4.6, the slope of the curve is related to the
feed rate and the yield coefficient. If V changes as a consequence of dilute
feed, then the total quantity of biomass (VX) will increase linearly.
FS
0 Y X/S
Figure 4.6. Linear growth under conditions of feed limitation with constant volume.
Figure 4.7. Repeated fed batch operation in terms of dimensionless variables for substrate
inhibition kinetics. Two cycles of operation are shown. The dimensionless variables are
defined in the simulation example VARVOL, Sec. 8.3.1.
4.1 General Balances for Tank-Type Biological Reactors 109
The specific biomass production rate for a chemostat, DXi, (kg biomass/m3 h)
can be calculated by applying the above model equations. Thus,
D X i = DY x /s(So-Si) = DYx/s
The conditions for the maximum value of DXi as shown in Fig. 4.4, can be
obtained by setting
d(DXi)
=
dD
S^X^F+R
Cell recycle
The mass balances around the entire system are as follows: Biomass
accumulates both within the reactor of volume Vr and also within the separation
unit with volume Vs. Assuming that biomass leaves only in the wastage stream
and that growth occurs only in the reactor, the balance is then
dXi
=0-WXR
Thus, the wastage rate of biomass must equal its production rate, otherwise Xi
will change. Wastage rate is an important control parameter in wastewater
treatment, where the separator is usually a sedimentation tank.
For the substrate, which is consumed only in the reactor section,
dSi dSR
At steady state,
0 = F So FI Si W Sj + r$ Vr
Here it is seen that the uptake rate is equal to the difference between the inlet
and outlet mass flows. The efficiency of the continuous biomass separation
determines XI/XR, which controls the recycle ratio, R/(F+R).
Considering the fermentation tank only, the balances are as follows:
For the biomass,
At steady state,
0 =RXR-(F + R)X!+rxVr
Cell separation and recycle lead to high cell concentrations in the reactor,
which, when neglecting the contribution by growth, would be XR/(F+R). Since
the rates are proportional to Xi, an increase in reactor efficiency is obtained.
Assuming, rx = |n Xi gives,
MX,
where, XR > Xi , and D = F/V is the nominal dilution rate. This equation means
that the specific growth rate is decreased from the chemostat value, D. This is
due to the reduction in substrate concentration, Si, which is caused by the
higher biomass concentration, resulting from the cell recycle. Washout is
impossible due to the complete biomass retention, and for this reason flow rates
greater than in a chemostat are possible.
The substrate balance gives
JQ
Vr == F S0 + R Si - (F + R) Si + rs V
dt
which shows that at steady state
112 4 Bioreactor Modelling
This would also be the case without cell recycle, since the substrate is assumed
to pass unchanged in concentration through the separator.
The above equation can be written as
where the right hand side of this equation is the F/M (food/biomass) ratio. This
gives a theoretical basis to the F/M concept, which is well known for the control
of waste treatment plants. The simulation example ACTNITR, Sec. 8.4.3,
enables the main operating characteristics of cell recycle systems to be studied.
A related simulation example MEMINH, Sec. 8.9.1, considers the retention of
enzyme using a membrane, and SUBTILIS, Sec. 8.9.2, involves the retention of
biomass.
:
: :';: y~: : w> Vv
illi 81 S2
> . 1
111
>
11 S;^;:-);'::;HaKS illi
Figure 4.9. Tanks-in-series bioreactor.
where i\ = Vi/F and is the mean residence time of the liquid in tank 1.
The balances for tanks 2 and 3 have the same form except for the subscripts:
4.2 Modelling Tubular Plug Flow Bioreacrors 113
For known flow rate, F, and known tank volumes, there are six unknowns in
these three equations. Note that different tank sizes may be accounted for by
differing values for the tank residence times TI, 12 and 13.
If the kinetic terms r$ are only dependent on S, then the above equations can
be solved without any further balances. It is often the case that enzymatic rate
equations of the form given below can be used for each tank n = 1, 2 and 3:
This gives now six equations and six unknowns, and the problem is solvable by
simulation methods.
If the situation is more complex, such that r$ depends on other components,
for example P, in the case of product inhibition or biomass X, then additional
balance equations for these components must be included in the model. When
combined with equations for the complete kinetics description (rates as a
function of all the influential concentrations), the model can be solved to obtain
the dynamics of the system and also the final steady state values. It can be
shown that a three tanks-in-series reactor system will provide a good
approximation to the performance of a corresponding tubular reactor, except
for very high conversions.
Figure 4.10. Plug flow idealization of the tubular reactor with no axial mixing.
A reaction will cause a steady state axial concentration profile as shown in Fig.
4.11. Thus at steady state, the concentrations vary with distance in a manner
which is analogous to the time-varying concentrations that occur in a batch
reactor.
Concentration
This means that steady-state tubular reactor behavior can be modelled by direct
analogy to that of a simple batch reactor. Thus using the batch reactor
substrate balance (p = constant),
dS
dF = fS
where v = F/A and F is the volumetric flow rate through the tube with cross-
sectional area A. Thus substituting for dt,
dS_ rs_
=
dZ v
4.2 Modelling Tubular Plug Flow Bioreacrors 115
)))))>
Figure 4.12. Finite-differencing the tubular reactor.
Figure 4.13. Balancing the difference segment n for the tubular reactor.
dSn = A(SF)
+ rsn
dt AV
Setting AV AdZ and AS > dS gives the partial differential equation, which
describes changes in time and distance, as,
38
- =
1 d(SF)
-- x J_
at A az
When the volumetric flow, F, is constant,
as _F as + rs
at ~ ~ A az
At steady state,
and
V + fS
dZ
which is the steady state equation derived earlier in Section 4.2.1.
To model the dynamics by simulation methods, the partial differential
equation must be written in difference form as,
dt AV/F
The gas phase is dispersed as gas bubbles within the liquid phase. Mass transfer
occurs across the gas-liquid interface, out of the gas into the liquid, where the
reaction occurs. The typical example is aeration of the bioreactor broth and
the supply of oxygen to the cells as shown in Fig. 5.1.
Figure 5.1. Absorption of oxygen from an air bubble to the liquid medium.
In this case, a liquid phase is in contact with solid biocatalyst. Substrates A and
B diffuse from the liquid to the reaction sites on the surface of the solid, where
reaction occurs. The product C must similarly be transferred away from the
solid reaction surface, as shown in Fig. 5.3. Examples are found with
immobilized enzyme and cell systems. In Sec. 6.1 the modelling aspects of this
type of system are considered in detail.
5.2 Interface Gas-Liquid Mass Transfer 119
Concentration gradients are the driving forces for mass transfer. Actual
concentration gradients (Fig. 5.5) in the very near vicinity of the gas-liquid
120 5 Mass Transfer
interface, under mass transfer conditions, are very complex. They result from
an interaction between the mass transfer process and the local fluid
hydrodynamics, which change gradually from stagnant flow, close to the
interface, to perhaps fully-developed turbulence within each of the bulk phases.
According to the Two-Film Theory, the actual concentration profiles, as
represented in Fig. 5.5 can be approximated by linear gradients, as shown in
Fig. 5.6.
A thin film of fluid is assumed to exist at either side of the interface. Away
from these films, each fluid is assumed to be in fully developed turbulent flow.
There is therefore no resistance to mass transfer within the bulk phases, and the
concentrations, CG and CL, are uniform throughout each relevant phase. At the
phase interface itself, it is assumed there is no resistance to mass transfer, and
the interfacial concentrations, CGI and CLI, are therefore in local equilibrium
with each other. All the resistance to mass transfer must, therefore, occur within
the films. In each film, the flow of fluid is assumed to be stagnant, and mass
transfer is assumed to occur only by molecular diffusion and therefore to be
Interface
Gas
described by Pick's law, which says that the flux JA (mol/s m2) for the molecular
diffusion of some component A is given by,
dZ
5.2 Interface Gas-Liquid Mass Transfer 121
Interface
Gas Liquid
where D is the molecular diffusion coefficient (m2/s) and dC/dZ is the steady
state concentration gradient (mol/m3). Thus applying the same concept to mass
transfer across the two films,
JA = DG Z Z
G L
where DG and DL are the effective diffusivities of each film, and ZG and ZL are
the respective thicknesses of the two films.
The above equations can be expressed in terms of mass transfer coefficients
kc and kL (m2/s) for the gas and liquid films,
JA = k G (C G -C G i) = k L (C Li -C L )
Q = JAA = jA(aV)
where "A" is the total interfacial area available for mass transfer, and "a" is
defined as the specific area for mass transfer or interfacial area per unit liquid
volume (m2/m3). Thus for the total rate of mass transfer:
Q = k G A(C G -C G i ) = k L A(C L i -C L )
Since the mass transfer coefficient, k, and the specific interfacial area, a, depend
on the same hydrodynamic conditions and system physical properties, they are
frequently combined and referred to as a "ka value" or more properly a mass
transfer capacity coefficient.
In the above theory, the interfacial concentrations CGI and CLI cannot be
measured, and are therefore of relatively little use, even if the values of the film
coefficients are known. For this reason, by analogy to the film equations,
overall mass transfer rate equations are defined, based on overall coefficients of
mass transfer, KG and KL, and overall concentration driving force terms, where:
Here, CG* and CL* are the respective equilibrium concentrations, corresponding
to the bulk phase concentrations, CL and CG, respectively, as shown in Fig. 5.6.
Equilibrium relationships for gas-liquid systems, at low concentrations of
component A usually obey Henry's law, which is a linear relation between gas
partial pressure, PA, and equilibrium liquid phase concentration, CLA*:
PA=
where HA (bar m3/kg) is the Henry's law constant for component A in the
medium. Henry's law is generally accurate for gases with low solubility, such as
the solubility of oxygen in water or in fermentation media. Thus from this
relation, as shown in Fig. 5.7, the corresponding equilibrium concentrations can
be easily established.
CLC*
For gases of low solubility, e.g., oxygen and carbon dioxide in water, the
concentration gradient through the gas film is very small, as compared to that
within the liquid film, as illustrated in Fig. 5.6. This results from the relatively
5.3 General Oxygen Balances for Gas-Liquid Transfer 123
low resistance to mass transfer in the gas film, as compared to the much greater
resistance to mass transfer in the liquid film. The main resistance to mass
transfer is predominantly within the liquid film. This causes a large change in
concentration (Cy - CL), since the resistance is almost entirely on the liquid
side of the interface.
At the interface, the liquid concentration, Cy, is in equilibrium with that of
the gas, CGI, and since CGI is very close in magnitude to the bulk gas
concentration, CLI must then be very nearly in equilibrium with the bulk gas
phase concentration, CG- This is known as liquid film control and corresponds
to the situation where the overall resistance to mass transfer resides almost
entirely within the liquid phase. The overall mass transfer capacity coefficient
is KLa. Hence the overall mass transfer rate equation used for slightly soluble
gases in terms of the specific area is
Q = KLa (C L *-C L )V L
C G = HCL*,
Gas CQO GO
Figure 5.8. The balance regions for well-mixed gas and liquid phases in a continuous reactor.
For the gas phase the oxygen balance can be developed as follows:
dCGi
K L a(C L i*- C L i)V L
where VQ represents the volume of gas in the dispersed phase, or the gas
holdup.
For the liquid phase,
dCLi
- LiC L i + K L a(C L 1 *-C L i)V L -
Typical units are as follows: CG and CL (kg/m3); G and L (m3/s); K^a (1/s);
VG and VL (m3); qO2 (kg/kg s); X (kg/m3).
In the next sections, the general equations, given above, will be applied to
important special situations.
The convective terms in the generalized liquid balance equation can usually be
neglected, owing to the low solubilities of oxygen in water (about 8 g/m3). This
gives the steady state liquid balance, dCL/dt=0, relation as:
Thus at steady-state, the oxygen transfer rate is effectively equal to the oxygen
uptake rate. Even during batch fermentations this is approximately true.
Substituting this relationship into the steady state gas balance gives,
=
K L a(CLi*-C L i)V L
5.3 General Oxygen Balances for Gas-Liquid Transfer 127
The classical dynamic KLa method assumes that K^a and CLI* are constant.
Under these conditions, the differential equation can be integrated analytically
to give the relationship:
r *(
CL* ^
CL - rCL\}
= K L at
Plotting the natural logarithmic concentration function on the left side of the
equation versus time, should, in principle, give a straight-line relationship, with
^a as the slope.
Usually deoxygenation is accomplished with nitrogen, so that initially the gas
phase consists of nitrogen, which is gradually displaced and mixed with air.
Under these conditions, CLI* is nt constant, and a gas balance must be
employed to calculate the variation in CGI versus t. Since the liquid phase
concentration, CLI, is measured by means of a membrane covered oxygen
electrode, the dynamics of measurement method usually cannot be neglected.
The dynamics of the measurement electrode can be described, approximately,
by a first-order lag equation,
dCE
time
Figure 5.9. Response of electrode for a step change in CE from zero to 100 % saturation
according to a first-order lag model.
Note that TE corresponds to the time for the electrode to reach 63 % of the final
response. The overall process dynamics involves thus the gas phase, the liquid
128 5 Mass Transfer
phase and the electrode response. The three responses might appear as shown
below:
time
Figure 5.10. Response of the gas, the liquid and the electrode measurement during a dynamic
KLa experiment.
The values of three individual time constants determine the process response.
These are TG = VG/G, (representing the dynamics of the gas phase), 1/KLa
(representing the dynamics of the liquid phase mass transfer process), and IE
(representing the measurement dynamics). This is illustrated in the simulation
example KLADYN, Sec. 8.5.5.
Low oxygen uptake rates, as exist in slow growing systems (plant and animal
cell cultures, aerobic sewage treatment processes, etc.), cannot easily be
measured by a gas balance method, since the measured difference between inlet
and outlet oxygen gas phase concentrations is usually very small. Due to the
low solubility of oxygen in the liquid media, quite small oxygen uptake rates
will cause measurably large changes in the dissolved oxygen concentration.
Thus it is possible to measure qo2 X either by taking a sample and placing it in
a small chamber or by turning off the reactor air supply, according to the
liquid balance equation
dCLi
5.3 General Oxygen Balances for Gas-Liquid Transfer 129
CL1
CLO
Thus the rate of oxygen supply via the liquid is equal to the rate of oxygen
uptake by the cells. This method provides a very sensitive way of measuring
low oxygen uptake rates (Keller et al., 1992, Tanaka et al., 1982). The case-
study H in Sec. 5.3.1.8 is an example of this use for an experimental reactor.
The simulation example FBR, Sec. 8.4.9, also demonstrates this method.
Knowing the flow rate L, the oxygen liquid concentrations CLO and CLI and the
outlet oxygen in the gas phase (to determine CLI*) permits the calculation of
5.3 General Oxygen Balances for Gas-Liquid Transfer 131
t
/"VX "N^
nM&^/m^M:WM.
-^^^ r-^^-~\_/~\^^x
Xiiiiiiiiiiji ^^^SRSSSS
^^^iO^wiBii
O
Wiiiiilmiiiiiii
Air
air
f Rate of \ ,^ f
accumulation of (Transfer rate of f N\ / Uptake rate of
=
V 02 in liquid ) l02 mto the hqmdj ^ by the cells
dCL
= K L a(CL*-C L )VL - k C L V L
A steady-state can be reached for which the mass transfer rate is equal to the
oxygen uptake rate by reaction:
Figure 5.14. Inlet and outlet oxygen mole fractions and total gas molar flow rates.
5.3 General Oxygen Balances for Gas-Liquid Transfer 133
From the ideal gas law as shown before, assuming a well-mixed gas phase in
steady state, N = (p / RT) F, where NO is the molar flow rate of air and F is the
air volumetric flow rate.
/ Rate of O2 \ / Rate of O2 \ / Transfer rate of \
0 =
V in by flow ) ~ Vout byflow) ~ \ C>2 to the liquid )
0 = yoN0-yiNi-KLa(CL*-CL)VL
where
Thus it is possible to distinguish between two different regimes for this system,
transfer control and reaction control:
1) Reaction rate control applies for low values of k/KLa, when re approaches
k CL*, and CL approaches CL*
2) Diffusion control applies for high values of k/KLa, when re approaches
KL& CL* and CL approaches 0.
3) If KLa = k, then rc = - (k/2) CL*, and CL approaches (1/2) CL*.
O2 -> NO3"
NH4+ + 2O 2 -
The reactor of volume, Vr, consisted of a conical sand bed column, which was
fluidized by the liquid recycle stream flowing up through the bed. The recycle
stream was oxygenated in a separate, baffled, tank contactor of volume VT, with
turbine impeller and air or oxygen sparging. The reactor and oxygenator were
thus separate parts of a recycle loop configuration. This could be operated
batchwise or with a continuous feed and effluent stream flow to and from the
system. When operating at high recycle rates, the whole system acted
effectively as one well-mixed tank system.
The reactor-oxygenator recycle loop can be analyzed as a total system or
broken down into its individual components as shown in Fig. 5.14. These
include liquid phase balance over the reactor and combined phase, liquid phase
and gas phase balances over the oxygenator and over the total system.
5.3 General Oxygen Balances for Gas-Liquid Transfer 135
Figure 5.15. Mass balancing regions for the fluidized bed reactor nitrification system.
The mass balances to be considered are those for oxygen and the nitrogen-
containing reactants and products. The oxygen balance taken over the total
system can be simplified by neglecting the accumulation terms and the liquid
flow terms, that will be small compared to the gas rates and the consumption by
reaction, owing to the relatively low solubility of oxygen in the liquid medium.
Thus the oxygen balance becomes,
0 =
VdCp
= F(C N i-C N 2 )
dt
and liquid phases as separate balance regions. The absorption tank can be
described by the oxygen balances for the liquid phase:
0 = F R (C L4 -C L3 ) + K L a(C L *-C L3 )V T
0 = KL
where ro2 is the oxygen uptake rate by the reaction. These equations, which
assume ideally mixed phases, are useful in designing the gas absorber
according to the required oxygen transfer coefficient.
Balancing the oxygen around the reactor gives
0 =
Since CL4 at the reactor outlet is usually very low, then,
FR CL3 = - ro2 Vr
which says that the oxygen uptake rate by reaction must be equal to the supply
rate from the oxygenation tank. This is the condition of reaction-rate
limitation by the oxygen transfer in the absorber.
From the stoichiometry, the relationships between the molar reaction rates
(rNH4> rO2 rH* r2,NO2 an^ fNO3) can be found. Thus, for example, the first
nitrification step gives
2
TNH4 = T r l , O 2 = ~ri,NO2
and the total rate for 02 is given by the sum of the rates for steps (1) and (2).
r
O2 = r l,O2 + r2,O2
d* = yH2p
where yo2 is the mole fraction of oxygen in the air and p is the total pressure at
some point in the tank.
The possibility that oxygen gas compositions, dissolved oxygen
concentrations, oxygen solubilities, gas holdup volumes, bubble sizes and other
transfer parameters can vary with depth in a tall fermenter introduces a much
greater degree of complexity to the problem of modelling the reactor. This
makes it difficult to obtain data on oxygen mass transfer coefficients.
Although it is impossible to give specific recommendations that apply to any
particular situation, a further discussion of possible models and their
underlying assumptions may help to define the problem. Incorporated into the
more complex models, discussed below, are such factors as gas and liquid phase
flow pattern, gas composition gradients and the effects of hydrostatic pressure.
Great caution and wisdom must be exercised to avoid creating a model that is
too complex to verify by experimentation. Experienced engineers will say
"Keep it simple!" and "Avoid too much model!".
All large scale reactors, whether multi-impeller tanks or column fermenters,
will display some axial dissolved oxygen concentration gradients. The most
general method for modelling is to represent the reactor using balances in a
series of sections or stages. Mass balances in multi-stage process are easy to
formulate, since both the liquid and gas phases may be assumed to be well-
mixed, for any given stage of the cascade.
Figure 5.16. A single gas-liquid stage with backmixing of the liquid phase.
The formulation of the mass balances for a single stage, as shown in Fig.
5.16, follows closely that described previously, except that now the reactor is
made up of many stages which are interconnected by the flows of gas and
liquid between stages and by diffusive mass transfer mechanisms.
5.4 Models for Oxygen Transfer in Large Scale Bioreactors 139
Gas
ill
GO L' LO
The oxygen balance equations for the gas and liquid phases of each stage are
as follows:
where,
P
r<~Ln * - 1
H rC r c * -
Gn or CLn
02
-77-
rl
and
rn = Qo2m |
L3 L2
G2
Gas
'LI
G1
FL+FB
Gas
F
I G- C GO \\>cu>
Inlet Gas Liquid Feed
To model this system, the liquid-phase impeller zones are assumed to be well-
mixed, and the plug-flow gas is described by a series of well-mixed phases,
together with an arithmetic-mean, concentration-driving-force approximation.
Here the flow rates and mass transfer coefficients are assumed constant.
Stage 1:
dCLi
VL -ar = FLCLO + FBCL2 - (FL + FB) CLI +
+ K L a(CLi*-C L i)VL + riV L
V dCGl
G
"""' xx-<
-CGI) -KLa(CLi -CLI)VL
f\ \ -rr- //~1 ^ /"I \A 7
where the plug flow nature of the gas is partially accounted for by
142 5 Mass Transfer
and
CLI
= - Qo2r
Stage 2:
dCL2
VL -gf = (FL+ FB) CLI + FBCL3 - (FL+ FB) CL2 - FB CL2
CL2
and n
r2 = -Qo2mK 0 + CL2
Stage 3:
dCL3
K L a(C L 3*-CL3)V L + r 3 V L
V
G - CG3) - KLa (CL3*- CL3) VL
where
CL3. .
CL3
and = -QO2n K0 + CL3
The above equations describe the dynamic oxygen concentrations in the multi-
impeller continuous bioreactor. Note that the liquid phase balances for the two
end stages 1 and 3 differ from that of the intermediate stage 2, owing to the
5.4 Models for Oxygen Transfer in Large Scale Bioreactors 143
The retention and immobilization of enzymes and cells usually requires the
presence of an additional solid carrier phase or flocculant cell mass. As
illustrated in Fig 6.1, in order to reach a reaction site, substrate S must first be
transported by convection from the bulk liquid to the exterior stagnant film
(point A). Then transport by diffusion must occur through the film (from A to
B) to the surface of the carrier (point B), where surface reaction can take place.
If further reaction sites are available within the carrier matrix, an additional
internal diffusion path (from B to C) is then also required. Similarly product P,
formed within the carrier matrix, must diffuse out of the matrix towards the
surface, and then away from the surface via the external mass transfer laminar
film to the bulk liquid.
Concentration
Diffusion film
i
Bulk Solid
liquid carrier
A B B
The stagnant film and the immobilization matrix constitute mass transfer
resistances which may slow the overall reaction rate, since reaction cannot
proceed at a rate greater than the rate at which substrate is supplied by the
mechanism of diffusion. The diffusional mass transfer process via the external
film is referred to as external mass transfer. Since the reaction site may often
be located within a gel, a porous solid, biofilm or biofloc, the transfer of
substrate or substrates from the exterior surface of the biocatalyst to reaction
sites, located within the internal structure of the carrier, is also usually
necessary. This process is therefore referred to as internal mass transfer or
intraparticle transfer. In what follows, external transport and internal transport
are considered separately, although, of course, the two effects can exert a
combined effect in reducing the effective reaction rate, compared to that which
would be obtained if there were no diffusional limitations.
Fig. 6.2 illustrates the substrate concentration profile, existing in the very near
region of an immobilized biocatalyst surface, supported on a non-porous
carrier. Also shown is the idealized concentration profile, as represented by the
film theory. As previously discussed in Sec. 5.2, the "film theory" assumes the
presence of a stagnant layer of liquid to exist at the solid-liquid interface. This
stagnant region is termed the diffusion film or Nernst-diffusion film and
constitutes the external resistance to mass transfer. It thus determines the rate of
supply of substrate to the surface, for subsequent reaction.
Substrate cone.
SA A
Figure 6.2. External diffusion model of substrate transport to a reactive enzyme immobilized
on a solid surface.
JS = ks L (SA-S B )
6.1 External Mass Transfer 147
where, js is the mass flux (mol/m2 s), ksL is the mass transfer coefficient (m/s)
and SA, SB are the substrate concentrations for the bulk and surface conditions
[mol/m3], respectively.
The steady-state balance can be written for the transport-reaction process,
SB = -SA
and hence
= ksSB=U^
l k S+ k SL
The intermediate situation is given by the full equation, for which the apparent
reaction rate is influenced by both the true kinetic rate constant ks and by the
diffusional mass transfer coefficient ksL-
For a zero-order reaction:
k S L(S A -S B ) = ks
where ks is now a zero-order kinetic rate constant. The concentration at the
reaction surface SB is thus,
SB = S A -ks/k S L
148 6 Diffusion and Biological Reaction in Immobilized Biocatalyst Systems
which indicates that the ratio of the magnitudes of the kinetic rate constant to
that of the mass transfer coefficient determines SB- If the reaction is zero-
order, the overall order of reaction rate is not influenced by diffusional
considerations, but the effective rate will still be reduced, owing to the lowered
concentration SB-
For Michaelis-Menten kinetics, which encompass the range between effective
zero and first order reaction kinetics, the relation between rate of supply and
the rate of reaction becomes,
kSL (SA - SB) = : > - = r app
After rearrangement, the resulting quadratic equation can be solved for SB, with
the solution indicating that, in general, the magnitudes of all the coefficients
can influence the overall reaction rate and also that the external transfer can
change the overall observed reaction kinetics. Thus they no longer follow the
Michaelis-Menten form with respect to bulk concentration, and the apparent
kinetics can differ substantially from the intrinsic true reaction kinetics. Under
these conditions, it is no longer correct to equate the Michaelis-Menten
constant, KM, to the substrate concentration at which the observed reaction rate
is equal to the half of the maximum observed rate. This can be most easily
seen from the above equation; when SB is low, the effective surface rate reduces
to the form, rapp = vm SB/KM- The overall reaction rate then becomes,
= (v m /K M )S A
app
(v m /K M k S L ) + l
showing that the apparent rate of reaction depends on the magnitude of the
mass transfer coefficient
It is only possible to measure true reaction kinetics, by operating
experiments in a truly kinetic regime, such that any influence of the external
diffusional mass transfer is negligible. This can be achieved by ensuring that
the ratio of vm/ksL is sufficiently low. Under these conditions,
dCA
jA = -D A
dZ
'AO
'AO
Diffusion
Diffusion
'AO
/Apparent rate\ =
/Rate of substrate^ /Net rate of reaction^
Vin bulk liquid) \entering matrix ) = v within matrix )
Vr a p p = (j|z=o)A = r a v g A L
where the units of each term are kg/s. Here ravg represents an average value in
the matrix, which will increase with higher internal substrate concentrations.
Regarding the influence of diffusion for a particular situation, it is possible
to arrive at some quantitative guidelines without considering any mathematical
details. Obviously the concentration profiles are caused by a competition
between reaction and diffusion. The ratio of the maximum intrinsic reaction
rate (not influenced by transfer) to maximum diffusion rate provides a useful
dimensionless parameter,
6.2 Internal Diffusion and Reaction within Biocatalyst 151
r
max A L r(C 0 ) A L maximum reaction rate
JA D (Co/L) A ~ maximum diffusion rate
For first order reaction, r = k CQ, this dimensionless group becomes k L2/D,
and for zero order reaction, r = k, it is k L2/D CQ.
Therefore for any kinetic form of equation, the distance coordinate or length
of diffusion path, L, plays an important role since the ratio of maximum
diffusion rate to maximum reaction rate varies according to L2. The higher the
value of this ratio, the greater in magnitude are the substrate gradients. With
this qualitative feeling for diffusion-reaction phenomena, more quantitative
aspects can be considered.
Liquid
jn-1 In I n+1
n-1 > ,
n n+1
A magnified view of element n is shown in Fig. 6.6, where the flux, jn, depends
on the local concentration gradient, and the reaction rate, rsn, depends on the
local concentration in element n.
Jn-1
AZ
Figure 6.6. A single element n of volume AV and thickness AZ, showing the diffusion fluxes.
A component mass balance is written for each segment and for each
component as
A AZ
dSn
~dT = Jn-1 A - jn A + rSn A AZ
(Sn-i ~ Sn)
s
and similarly for Jn gives,
n-l ~ Sn)
+rSnAAZ
6.2 Internal Diffusion and Reaction within Biocatalyst 153
Dividing by A AZ,
dSn (S n ,j -2S n + S n+ i)
s 5 + rsn
T
as a2s + rs
/Accumulation^ =
/External trans- A /Diffusion^ f Production \
V rate / V port rate in ) ~ \ rate out ) + Vrate by reaction/
Q = k S L (So-Si)A
where,
154 6 Diffusion and Biological Reaction in Immobilized Biocatalyst Systems
dS -
dZ I z=o =
The direction of the positive diffusion is into the biofilm, since dS/dZ is
negative and (SR - Si) /AZ is positive. Here SR corresponds to So in Fig. 6.4,
and Si is the concentration in the first element. The boundary condition is
S = SR at Z = 0.
F,SF F,SO
Figure 6.7. Coupling the biofilm model to the continuous tank model.
only between the values of zero and unity. Thus new dimensionless variables,
S=S/S 0 , Z = Z/L and dimensionless time, f = t / ( L I D ) , can be defined.
Substituting these new variables into the diffusion-reaction, partial differential
equation, for the case of a first order biochemical reaction, gives,
as
S 2 =
(L /D8)at
or
as _ _ _
at - az2 " DS s
Thus the solution depends only on the value of [lq L2/DsL which is a
dimensionless diffusion-reaction parameter. For zero-order reaction the
equation becomes,
as a2s2
at " az
where ko L2/Ds SQ is the governing parameter.
It is seen that the dimensionless parameters in the model have the same form
and significance as was derived from the qualitative reasoning presented earlier.
For heterogeneous reaction systems this dimensionless group is known as the
Damkohler Number, and its square root is called the Thiele Modulus.
In the above equations, all the terms, excepting that of the reaction term, have
dimensionless parameters of unity. On this basis, it can be said that if the
reaction parameter for a first order reaction, [ki L2/DsL has a value of 1.0 or
greater, then the reaction will have a large effect on the solution, that is, on the
concentration gradients. Similarly, ko L2/D$ SQ will govern a zero order
reaction. Such "order of magnitude analysis" is important for physical
understanding and also to obtain information from differential equations
without having to actually develop an analytical solution.
Dimensionless formulation of equations is also explained in the simulation
example VARVOL, Sec. 8.3.1, and KLADYN, Sec. 8.5.5.
Solutions of the above diffusion reaction model are available in the literature
for simple reaction orders (Satterfield and Sherwood, 1963). In Fig. 6.8 the
values of t| for zero, first and second reaction order have been plotted against
the Thiele Modulus ,<f>, where,
O = '\/L 2 kSo n - 1 /D
This figure shows that a zero order reaction is not influenced by concentration
gradients until the substrate falls to zero in the matrix, corresponding to O > \2
. The other reaction-types are influenced by low concentrations, as the curves
for T| indicate. It is seen, for example, that for a first order reaction a value of
O = l corresponds to t| = 0.8.
0.8 - A
A = zero-order
0.6 - B = first-order
C = second-order
0.4 -
0.2 -
For a biofilm or floe, whose oxygen uptake might be taken as a constant (zero
order), the corresponding group would be [L2 qo2 Xbi0fiim/D Co2L where qo2
Xbiofiim corresponds to the rate constant k and the oxygen concentration in the
outside liquid phase is Co2- Note that Xbiofiim is the biomass per unit of
biofilm volume and is not easy to measure. Substituting values obtained from
an aerobic biofilm nitrification experiment gives,
For this order-of-magnitude analysis, the value of 1.0 can be used to separate
the regions of reaction and diffusion dominance. Thus it is seen if L = 0.1 mm,
then the dimensionless group will have a value of 1.0, and it could therefore be
expected that a film or floe thickness greater than 0.1 mm would be oxygen
limited. From the exact solution, as seen in Fig. 6.8, gradients would appear for
a zero order reaction at a value of <&2 = 2.0, instead of 1.0. This example
shows how the Thiele Modulus can be used to make useful estimates for
diffusion reaction problems, providing rate and diffusion data are available.
The oxygen requirements for the first and second steps can be related to the
nitrogen content of NH4+ and NC>2~. These values are si = 3.5 mg 62 / mg
NH4+ - N and 82 = 1.1 mg 62 / mg NCV - N. The low yields and low growth
rates make it unnecessary to consider growth requirements and kinetics. In
previous work (Tanaka and Dunn, 1982) the intrinsic substrate uptake kinetics
for the two steps were shown to have a double Monod form for the first step,
rNH4 = v m i
. K NH4
and for the second step,
rN02 = vm2
where v m i and vm2 represent the maximum rates for a particular biomass
concentration and the chemical symbols represent concentrations.
Considering the diffusion phenomena in the biofilm to be represented by
one-dimensional diffusion with quasi-homogeneous reaction, differential
balance equations can be written for all reactants and products to describe the
concentration profile in the film. Proceeding as described in Sec. 6.2.1, a
component mass balance is written for segment n and for each component:
For NH4+,
3NH4+ a2NH4+
= DNH4 -wo - TNH4
For NO2",
aNQ2- = D
~3r NO2 ^2 + rNH4 - TNO2
6.2 Internal Diffusion and Reaction within Biocatalyst 159
For NO3%
3N03 32N03
For O2,
3202
" s l rNH4 ~
The stoichiometric oxygen requirements for the first and second reaction steps
are given by si and s2. The boundary conditions used represent the bulk liquid
phase or reactor concentrations and the zero gradient at the biofilm- solid
interface, as discussed earlier.
These equations can be written as differential-difference equations using the
finite-differencing technique (Sec. 6.2.1). Thus for each of N increments, four
component balances will be needed. Three simulation examples, BIOFILM,
ENZDYN, CELLDIFF in Sec. 8.7, demonstrate this approach.
This system was also analyzed in terms of dimensionless variables. A
comparison of the resulting dimensionless NH4+ and O2 balances reveals that,
when the second reaction is neglected, the equations are identical if
DNH4 = Do2 and if,
Q2R
where the subscript R refers to the concentrations in the bulk reactor liquid.
Under these conditions, to a good approximation, the penetration distances
of O2 and NH4+ would be the same. The ratio O2R/NH4+R, which can be varied
according to the reactor operating conditions, can thus be used as a criterion to
evaluate whether NtLj."1" or O2 might be penetration-limiting. The O2R/NH4+R
criterion indicates which component can be limiting, O2 if the ratio is less than
3.5 or NH4+ if the ratio is greater than 3.5. These conditions are not sufficient
for limitation, but indicate which component would be limiting. Simulation
results from a model that was developed using finite-differencing demonstrates
this phenomenon. The profiles in Fig. 6.9, are for the case O2R/NH4+R = 0.07.
160 6 Diffusion and Biological Reaction in Immobilized Biocatalyst Systems
100
H 20
80
60
NO
10
40
20
Figure 6.9. Steady state profiles for constant bulk concentrations showing incomplete
oxygen penetration.
Coupling the liquid and biofilm for a batch nitrification reactor as explained in
Fig. 6.7, gave the results in Fig. 6.10. Here the influence of oxygen limitation
caused the oxygen in the midpoint of the film to rise as the nitrogen substrates
were successively consumed.
100 80
12
80 64
10
60 48 8
40 32 6
4
20 16
2
0 0 0
120
t (min)
Figure 6.10. Simulated biofilm nitrification profiles in a batch reactor. The N-component
concentrations are in the bulk liquid. O2 is in the midpoint of the biofilm and indicates
limitation during the first 60 minutes.
7 Automatic Bioprocess Control
Fundamentals
Controller
mechanism
Desired
value
I Measured
I value
I
Thermo-
couple
Fermenter
Controller Load
mechanism
1 Com)jarator Controller Actuator Process
_ i' Controlled
Desired+c
value | *A/
1
1
J
$~ 1
K
1
1
-*U variable
Measuring
element
Measured k A A A A
A A A A A
value
^A^A^AAA^A
Figure 7.2. Block diagram of the feedback control system in Fig. 7.1.
7.2 Types of Controller Action 163
Similar temperature control systems are given for a simple water heater,
TEMPCONT, in Sec. 8.6.1 and for a batch fermenter, FERMTEMP, in Sec.
8.6.2
The most common and simplest type of control is an on-off or two position
action, sometimes called discontinuous control (Fig. 7.3). An example is a
contact thermometer, which closes or opens the heater circuit. The controller
changes the value of the controller output, or the manipulated variable, from
one extreme to the other, when the controlled variable moves above or below
the set point. This leads to oscillations that could become very fast, depending
on the speed of response of the process. The real on-off controller has
therefore a built in feature called a differential gap or a dead zone. It is a small
interval on either side of the set point, within which the controller does not
respond. When the controlled variable moves outside the dead zone, the
manipulated variable goes on or off. This is illustrated in Fig. 7.3. Such shifts
from the set point are known as offset. Such a controller is simple and
inexpensive, but the oscillatory nature of the action and the offset make it very
imperfect.
The usefulness of this type of control was demonstrated for a biological,
sequential batch process by Hediger and Prenosil, (1985). More sophisticated
function control modes consider the magnitude and time behavior of the
control error. Three principal functional modes of control generally employed
for process control are proportional (P), integral (I) and derivative (D) control.
164 7 Automatic Bioprocess Control Fundamentals
100
Manipulated
variable
0
Control Xl___4/_
variable .Set point
Differential gap
where Kp is the proportional gain, and P0 is the controller output for zero error.
An example of a level control is shown in Fig. 7.4. The action of this type of
controller is shown in Fig. 7.5 and 7.6.
F + AF
Setpoint
l
K
P = Kpe+ ?
TI
or for e = constant
dP
for t=Ti
dP
dF = K P
166 7 Automatic Bioprocess Control Fundamentals
where i\ is the integral time constant or reset time. This is the time required to
enable the controller to repeat the initial proportional output action (Fig. 7.6).
The integral part of the control mode eliminates the offset and it is especially
useful for correction of very small errors because the controller output P will
continue to change as long as an error persists. This can be understood by
considering a constant error, which would cause P to increase linearly (Fig. 7.6)
at a rate proportional to the error. This type of controller is found most often
and the simulation examples TEMPCONT, FERMTEMP, TURBCON and
CONTCON in Sec. 8.6 demonstrate the use of this control mode. The
examples also show the ease by which the programming of the PI controller
equations is made using the simulation language.
2Kp
Controlled
output PI action
Kp
P action
Error
signal
A controller with derivative function projects the error in the immediate future
and the controller output is proportional to the current rate of the error change.
The output signal varies only if the error is changing.
7.2 Types of Controller Action 167
d
P = P0 + Kp e + Kp TD gj-
P (alone)
The drawbacks of the derivative control mode standing alone are that a constant
error (e ^ 0) gives no response at all, since de/dt = 0, and therefore an
unnecessarily large response might occur as a result of small but fast error
changes.
KD r de
P = P0 + Kp e + ^ J dt+ Kp TD -gj-
TI
o
This combination gives the best control using conventional feedback
equipment. It retains the specific advantages of all three modes: proportional
correction (P), offset elimination (I) and stabilizing, quick-acting character,
which is especially suitable to overcome lag presence (D). The action of a PID-
controller as a response to a ramp function is shown in Fig, 7.7. The
performance of the different feedback control modes can also be seen in Fig.
7.8.
Controlled
variable
/ Uncontrolled
response
Figure 7.7. Response of controlled variable to a step change in error using different control
modes.
The selection of the best mode of control depends largely on the process
characteristics. Further information can be found in the recommended texts
listed in the reference section. Simulation methods are often used for testing
control methods.
7.3 Controller Tuning 169
The purpose of controller tuning is to choose the controller constants, such that
the desired performance is obtained. This usually means that the control
variables should be restored in an optimal way, following either a change in the
set point or as a result of an input disturbance to the system. Thus the
controller constants can be set by experimentation. A rational basis for such
experimental tuning is given in what follows. Other methods for tuning
combine process dynamic experimentation with theoretically-based control
methods; some of the standard methods are also described below. The
simulation example TEMPCONT, Sec. 8.6.1, provides exercises for controller
tuning using the methods explained below.
Controllers can be adjusted by changing the values of gain Kp, reset time i\ and
derivative time ID- By experimentation, either on the real system or by
simulation, the controller can be set by trial and error. Each time a disturbance
is made the response is noted. The following procedure might be used to test
the control with small set point or load changes:
1. Starting with a small value, Kp can be increased until the response is
unstable and oscillatory. This value is called the ultimate gain KPQ.
2. K p is then reduced by about 1/2.
3. Integral action is brought in with high i\ values; they are reduced by
factors of 2 until the response is oscillatory, and tj is set at 2 times this
value.
4. Include derivative action, increase ID until noise develops and set ID at
1/2 this value.
5. Increase Kp in small steps to achieve the best results.
important for this method are given by the normalized slope of the tangent
through the inflection point, S = slope/A and by its intersection with the time
axis (lag time TL), as determined graphically in Fig. 7.9. The actual tuning
relations which are based on empirical criteria for the "best" closed-loop
response are given in Table 7.1.
Manipulated
Variable
X
Slope
Time
Figure 7.9. Process reaction curve as a response to a step change in manipulated variable.
Cohen and Coon observed that the response of most uncontrolled (controller
disconnected) processes to a step change in the manipulated variable was a
sigmoidal shape curve. This can be modelled approximately by a first order
system with time lag TL, as given by the intersection of the tangent through the
inflection point with the time axis. The theoretical values of the controller
settings obtained by the analysis of this system are summarized in Table 7.1
The model parameters for a step change A to be used with Table 7.1 are
calculated as follows:
K = B/A T = B/S
where B is from Fig. 7.9 and S is the slope at the inflection point/A.
7.3 Controller Tuning 171
Controller Kp
Ziegler-Nichols
P 1/TLS
PI 0.9/TL S 3.33TL
PID 1.2/TLS 2TL TL/2
Co/ten-Coon
P T
TL \ 30 + 3TL/T
PI
4 TL \ 32 + 6TL/T
Ultimate Gain
P 0.5 Kpo
PI 0.45 Kp0 l/1.2fp0
PID 0.6 Kpo l/2fpo l/8fpO
In control situations with more then one measured variable but only one
manipulated variable, it is advantageous to use control loops for each measured
variable in a master - slave relationship. In this, the output of the primary
controller is usually used as a set point for the slave or secondary loop.
This may be relevant for some wastewater treatment plants where the high
concentration of some substrate may be toxic for the microorganisms. For
example, the simulation Example TURBCON, Sec. 8.6.3, could be easily
adapted to this situation if the substrate concentration were subject to significant
changes, as shown schematically in Fig. 7.10.
7.4 Advanced Control Strategies 173
m |
1
^ V^
lilllll
XI . .. N
! :
';S:'|||::||:|- illlll
/6oncentrationV/Concentriilonl
Controller )H f errand
x. ,s \ transmitter /
^
ill lit:
c Biomass
controller
J_
[ Turbidometer
prediction and the more complex "actual" process kinetics could then be taken
care of by a feedback control loop.
This control system can automatically modify its behavior according to the
changes in the system dynamics and disturbances. Especially systems with
nonlinear and unsteady characteristics call for use of this control strategy.
There are a number of actual adaptive control systems. Programmed or
scheduled adaptive control uses an auxiliary measured variable to identify
different process phases for which the control parameters can be either
programmed or scheduled. The "best" values of these parameters for each
process state must be known a priori. Sometimes adaptive controllers are used
to optimize two or more process outputs, by measuring these and fitting the
data with empirical functions, as employed on anaerobic treatment process, by
Ryhiner, et al. (1992).
Controlled
variable
Table 7.2. Examples of methods and strategies for the control of bioprocesses
(Heinzle and Saner, 1991).
Process Method Controlled Manipulated
and strategy variable(s) variable(s)
For constant value or set-point control usually constant control parameters are
used. Because of non-linearities or varying process dynamics (e.g. exponential
178 7 Automatic Bioprocess Control Fundamentals
Selection of a control strategy and its parameters (e.g. for a PID controller)
may be difficult, since the process and controller dynamics are often not well
understood. If possible, it is useful to use dynamic models to select a control
strategy, and to use it for testing and tuning. An example with anaerobic
digestion is given by Heinzle et al. (1992). In Fig. 7.12 are shown the results of
a simulation and a corresponding experiment for the control of the propionic
acid concentration by manipulation of the feed flow rate.
7.5 Concepts for Bioprocess Control 179
1
I
0 1 2 3 4
6(
Propionic acid [mg/l]
Feed flow [ml/min]
5(
4(
o
CD
O
"E
3( -
V2
i
I
20-
1(
2 3
Time [h]
Figure 7.12. Control of anaerobic digestion of whey wastewater. Simulation (A) and
experiment (B) of control after step change (Heinzle et al. (1992). Here the controlled variable
was the propionic acid concentration, and the manipulated variable was the feed flow rate.
References
Blanch, H.W. and Dunn, I.J. (1973) Modelling and Simulation in Biochemical
Engineering. Adv. Biochem. Eng, 3, 127-165.
Denac, M., Miguel, A., Dunn, I.J. (1988) Modeling Dynamic Experiments on
the Anaerobic Degradation of Molasses Wastewater. Biotechnol. Bioeng. 31, 1-
10.
Dunn, I.J. and Mor, J.R. (1975) Variable Volume Continuous Cultivation.
Biotechnol. Bioeng. 17, 1805-1822.
Furukawa, K., Heinzle, E., and Dunn, IJ. (1983) Influence of Oxygen on the
Growth of Saccharomyces cerevisiae in Continuous Culture. Biotechnol.
Bioeng. 25, 2293-2317.
Heinzle, E., Furukawa, K., Dunn, I.J., and Bourne, J.R. (1983) Experimental
Methods for On-line Mass Spectrometry in Fermentation Technology.
Bio/Technology 1, 14-16.
Heinzle, E., Dunn, IJ. and Ryhiner, G. (1992) Modelling and Control for
Anaerobic Wastewater Treatment, Adv. Biochem. Eng., Springer Verlag.
Ingham, J., and Dunn, I.J. (1991) "Bioreactor Off-Gas Analysis", in "Bioreactors
in Biotechnology", Ed. A. Scragg, Ellis-Horwood, Chichester, 195-220.
Keller, R. and Dunn, I.J. (1978) Fed Batch Microbial Culture: Models, Errors,
and Applications. J. Appl. Chem. Biotechnol. 28, 508-514.
Keller, J., Dunn, I. J. and Heinzle, E. (1991). A Fluidized Bed Reactor for
Animal Cell Culture, in preparation, Biotechnol. Bioeng.
Luyben, W.L. (1973) Process Modeling, Simulation, and Control for Chemical
Engineers. McGraw-Hill.
Meister, D., Post, T., Dunn, I.J. and Bourne, J.R. (1979) Design and
Characterization of a Multistage, Mechanically Stirred Column Absorber.
Chem. Eng. Sci., 34, 1376.
Moes, J., Griot, M., Keller, J., Heinzle, E., Dunn, I.J., and Bourne, J.R. (1985) A
Microbial Culture with Oxygen-sensitive Product Distribution as a Tool for
Characterizing Bioreactor Oxygen Transport. Biotechnol. Bioeng. 27, 482-
489.
Moes, J., Griot, M., Heinzle, E., Dunn, I.J., and Bourne, J.R. (1986) A Microbial
Culture as an Oxygen Sensor for Reactor Mixing Effects. Ann. N. Y. Acad. Sci.
469, 482-489.
Mona, R., Dunn, I.J. and Bourne, J.R. (1979) Activated Sludge Process
Dynamics with Continuous TOC and Oxygen Uptake Measurements.
Biotechnol. Bioeng. 21, 1561-1577.
Mou, D.G. and Cooney, C.L. (1983) Growth Monitoring and Control through
Computer-aided On-line Mass Balancing in a Fed-batch Penicillin
Fermentation. Biotechnol. Bioeng. 25, 225-255.
References 183
Prenosil, J., Dunn, I.J., and Heinzle, E. (1987) Biocatalyst Reaction Engineering,
in: Biotechnology Vol.7a, (Ed.: HJ.Rehm and R.Reed) VCH, Weinheim, 489-
545.
Ruchti, G., Dunn, I.J., Bourne, J.R. and v. Stockar, U. (1985) Practical
Guidelines for Determining Oxygen Transfer Coefficients with the Sulfite
Oxidation Method. Chem. Eng. J. 30, 29-38.
Ryhiner, G., Dunn, IJ, Heinzle, E, Rohani S., (1992) Adaptive On-line Optimal
Control of Bioreactors: Application to Anaerobic Degradation, J. BiotechnoL.
22, 89-106.
Saner, U., Bonvin, D., Heinzle, E. (1990) Application of Factor Analysis for
Elaboration of Stoichiometry and its On-line Application in Complex Medium
Fermentation of B. subtilis, in: Dechema Biotechnology Conferences Vol.3,
(Ed.: Behrens, D. and Driesel, A.J.), 775-778.
Satterfield, C.N. and Sherwood, T.K. (1963) The Role of Diffusion in Catalysis.
Addison-Wesley, N.Y.
Shioya, S., Dang, N.D.P. and Dunn, IJ. (1978). Bubble Column Fermenter
Modeling: A Comparison for Pressure Effects. Chem. Eng. Sci., 33, 1025 -
1030.
Tanaka, H., Uzman, S., Dunn, IJ. (1981). Kinetics of Nitrification Using a
Fluidized Sand Bed Bioreactor with Attached Growth. BiotechnoL Bioeng., 23,
1683 - 1702.
Ziegler, H., Meister, D., Dunn, IJ., Blanch, H.W., Russell, T.W.F. (1977). The
Tubular Loop Fermenter: Oxygen Transfer, Growth Kinetics and Design.
BiotechnoL Bioeng. 19, 507.
184 References
Biochemical Engineering
Aiba, S., Humphrey, A.E. and Millis, N.F. (1973) Biochemical Engineering.
Academic Press, N.Y.
Pirt, S. John (1975) Microbe and Cell Cultivation. Blackwell Scientific Publ.,
Oxford.
Wang, D.I.C., Cooney, Ch.L., Demain, A.L., Dunnill, P., Humphrey, A.E., Lilly,
M.D. (1979) Fermentation and Enzyme Technology. Wiley, New York.
Asenjo, J.A., Merchuk, J.C. (1995) Bioreactor system design; Marcel Dekker,
N.Y.
Hannon, B., Ruth, B. (1997) Modeling Dynamic Biological Systems; Springer-
Verlag, New York.
186 References
Sinclair, C.G., Kristiansen, B., Bu'Lock, J.D. (1987) Fermentation Kinetics and
Modelling. Open University Press, Milton Keynes.
van't Riet, K., Tramper, J. (1991) Basic Bioreactor Design. M. Dekker, New
York.
Webb, C., Black, G.M., and Atkinson, B. (1986) Process Engineering Aspects of
Immobilized Cell Systems.Pergamon Press Ltd., Oxford.
Metabolic Engineering
Satterfield, C.N. and Sherwood, T.K. (1963) The Role of Diffusion in Catalysis,
Addison-Wesley, New York.
Fish, N. M., Fox, R.L, and Thornhill, N.F. (1989) Computer applications in
fermentation technology: Modelling and control of biotechnological processes.
Elsevier, London.
System
The system is represented in Fig. 1, and the important variables are biological
dry mass or cell concentration, X, substrate concentration, S, and product
concentration, P. The reactor volume V is well-mixed, and growth is assumed
to follow kinetics described by the Monod equation, based on one limiting
substrate. Substrate consumption is related to cell growth by a constant yield
factor YX/S- Product formation is the result of both growth and non-growth
associated rates of production, where either term may be set to zero as required.
The lag and decline phases of cell growth are not included in the model.
Model
Mass Balances:
For substrate
V
dS = r
dF s
or
dS
dF = rS
For product
dP
V-3T =
or
dP
dF = r?
Kinetics:
rx = f i X
with the Monod relation, constant yield relation, and product formation
kinetics:
tx/s
rP = (ki + k2 |^) X
x
o Biomass X
^
Balance
r
T*
4_ x Growth
Rate
M
Monod
Kinetics
So Substrate 1 A s
^
Balance f'x
r
4_ s Substrate
Rate
PO Product p
Balance
r -^M
A
4
p Product
Rate ,-_
It is seen in that all the variables required for the solution of any one equation
block are obtained as the products of other blocks. The information flow
diagram thus emphasizes the complex inter-relationship involved in even this
very simple problem. Solution begins with the initial conditions XQ, SQ and PQ
at time t=0. The specific growth rate |i is calculated, enabling rs, rx and rp to
be calculated, and hence the initial gradients dX/dt, dS/dt and dP/dt. At this time
the integration routine takes over to estimate revised values of X, S and P over
the first integration step length. The procedure is repeated for succeeding step
lengths until the entire X, S and P concentration time profiles have been
calculated up to the required final time.
Program
{BATPERM}
{Batch growth with product formation}
{Constants}
UM=0 . 3
KS = 0 . 1 ;kg/m3
Kl=0.03 ;kgP/kgX h
K2=0.08 ;kgP/kgX h
Y=0. 8 ;kg X/kg S
196 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
{Initial Conditions}
INIT X=XO
INIT S=SO
INIT P=PO
(Mass Balances}
X'= RX ;BIOMASS BALANCE
S'= RS ;SUBSTRATE BALANCE
P' = RP ;PRODUCT BALANCE
{Kinetics}
RX = U*X ; BIOMASS RATE EQUATION, kg/m3 h
U = UM*S/(KS + S) ;MONOD EQUATION, 1/h
RS = -RX/Y ; SUBSTRATE RATE EQUATION, kg/m3 h
RP= (K1 + K2*U) *X /PRODUCT RATE EQUATION, g/m3 h
Limit S>=0.0
Nomenclature
Symbols
Indices
Exercises
2. Vary the product kinetics constants (Kj and K2>, and observe the
effects. Observe the P versus time curve when S reaches zero.
Results
The plots of X, S and P versus T in Fig. 3 show that when substrate is depleted,
the growth stops, and the product continues to increase, but only linearly. The
results of Fig. 4 are obtained by varying the product formation rate constants,
ki in three runs using a slider, which is defined in the Parameter Menu.
Figure 3. Plots of X, S and P versus time during batch growth and production.
"l"""'"*"'.^_ / 9
-8
** /
\ /* 7
-S:2
P:2
X / /' .5 cn
-S:3 -4
P:3 \ /s' \,"~'" 3
-'7^'^ 2
.**&'*' \ 1
-rp^ fm
vtf*EV \ .. 0
0 5 1 0 15 20 25 3()
TIME
Figure 4. Plots of P and S versus time created by varying the product formation rate constant
8.1 Introductory Examples 199
System
D,S F S,X
Model
For cells
dX
.= - D X + rx
200 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
For substrate
dS
3f = D (SF - S) + rs
For product
dP
df = -DP + rp
where D is the dilution rate and Sp the concentration of the limiting substrate in
the feed.
Program
{CHEMO}
(Chemostat startup and steady state. Startup as
batch reactor until time=tstart}
{Constants}
UM=0.3 ; 1/h
KS = 0.1 ; kg/m3
Kl = 0.03 ; kgP/kgX h
K2=0.08 ; kgP/kgX
Y = 0.8 ; kg X/kg S
X0 = 0.01 ; Initial biomass inoculum, kg/m3
S0=10 ; Initial substrate cone., kg/m3
P0=0 ; Initial product conc.,kg/m3
SF = 10 ; Feed cone. ,kg/m3
Dl = 0.25 ; Dilution rate, 1/h
t start = 5 ; Start time for the feed
(Initial Conditions}
Init X=XO
Init S=SO
Init P=PO
{Mass Balances}
X'=-D*X+RX ; BIOMASS BALANCE EQUATION
S =D* (SF-S) +RS ; SUBSTRATE BALANCE EQUATION
P'=-D*P+RP ; PRODUCT BALANCE EQUATION
8.1 Introductory Examples 201
{Kinetics}
RX = U*X ; BIOMASS RATE EQUATION, kg/m3 h
U = U M * S / ( K S + S) ; MONOD EQUATION, 1/h
RS=-RX/Y ; SUBSTRATE RATE EQUATION, kg/m3 h
RP= (K1 + K2*U) *X ;PRODUCT RATE EQUATION, kg/m3 h
Nomenclature
Symbols
Indices
F Refers to feed
MONOD Refers to Monod kinetics
P Refers to product
S Refers to substrate
X Refers to biomass
202 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Exercises
Results
The graphical output in Fig. 2 shows three startups of the fermenter under
initially batch growth conditions, using three values for D l . The break in the
concentration-time dependency as feeding starts is quite apparent, and the new
transient then continues up to the eventual steady state chemostat operating
condition or washout in the case of one run. For the results of Fig. 3 the
program was changed by adding the line PROD = X*D, and the final, steady
state value of production rate was plotted versus Dl for twenty runs, using the
Parameter Plot feature of Madonna.
8.1 Introductory Examples 203
Figure 2. Startups of the chemostat after initial batch growth for 3 values of Dl.
Figure 3. Productivity in a chemostat. Steady states are shown for 20 runs using the Parameter
Plot.
204 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
In this case the model equations allow for the continuous feeding of sterile
substrate, the absence of outflow from the fermenter and the increase in volume
(accumulation of total mass) in the fermenter, schematically as shown in Fig. 1.
Simulation of fed batch fermenters can be used to demonstrate the important
characteristics of quasi-steady state, linear growth, and use of alternative feed
strategies.
F,SF
V
X
s
p
Model
Total balance
dV
F
dT =
For cells
For substrate
8.1 Introductory Examples 205
For product
where F is the volumetric feed rate, Sp is the feed concentration and V is the
volume of the fermenter contents at time t. Thus the mass quantities, VX, VS,
and VP are calculated and are divided by the volume at each time interval to
obtain the concentration terms required for the kinetic relationships. The
kinetics are taken to be the same as in BATFERM.
Program
The "IF" statement in the program causes the continuous feed to start when time
reaches tfeed, at which point batch operation stops and the fedbatch starts.
(FEDBAT)
{Fermentation with batch start up}
{ Constants}
UM=0.3 ; 1/h
KS = 0 . 1 ; kg/m3
Kl = 0.03 ; kgP/kgX h
K2 = 0.08 ; kgP/kgX
Y = 0.8 ; kg X/kg S
X0 = 0 .01 ; Initial biomass inoculum, kg/m3
S0 = 10 ; Initial substrate cone., kg/m3
P0 = 0 ; Initial product conc.,kg/m3
SF = 10 ; Feed conc.,kg/m3
Pl-1. 5 ; Feed flow rate, m3/h
tfeed=22.5 ; Start time for the feed
{Initial Conditions}
init V=l
init VX=V*XO
init VS=V*SO
init VP=V*PO
206 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
d/dt(VP)=RP*V {kg/h}
{Calculation of concentrations}
X=VX/V
S=VS/V
P=VP/V
{Kinetics}
RX=U*X
U=UM*S/(KS+S)
RS=-RX/Y
RP=(K1+K2*U)*X
Nomenclature
Symbols
Indices
F Refers to feed
P Refers to product
S Refers to substrate
X Refers to biomass
Exercises
Results
growth" situation in which the growth rate is limited by the feeding rate. In
Fig. 3 the values of X, S, and P are plotted versus T for a switch from batch
(F = 0) to fed batch (F = 5) at time T = 20 h. The product production rate
depends linearly on biomass concentration, and thus even when ja becomes very
low, P will continue to increase linearly in mg/m3 amounts.
TIME= 34.13 X = 12.34
10-.- .^
10 20 30 40 50 60 70 80 90 100
0.35-
0.3-
-~, I
3
0.25.
Q 0.2-
V
CO
0.15-
0.1- -J.
T
0.05-
0-
27 28 29 30 31 32 33 34 35 36 37
TIME
System
The intermediate enzyme-substrate complex is the basis for the simplest form
of enzymatic catalysis (Fig. 1):
E +S ^ ES *- E +P
k2
Model
dt
Using the steady state approximation for the change of active complex,
dt
the Michaelis-Menten equation is obtained.
_dS _
K
~ dt ~ M +S
210 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Program
Nomenclature
Symbols
Indices
Exercises
8.2 Batch Reactors 211
Results
Figs. 2 and 3 give the results of the full model and the Michaelis-Menten
simplification, respectively
Run 1:119 steps in 0.0167 seconds
0.007- \ / 0.7
v
:. f ^..]
0.003- \ 0.3
1
0.002 0.2
*i* fc
\. i
0.001 - i *v. " %^ 0.1
""""%-, ^*"'..._
.0
10 20 30 40 50 70 80 90 100
TIME
\
\
10 20 30 40 50 60 70 80 90 100
TIME
System
r
s = K^TS-
The inverse rate is plotted versus the inverse concentration (Fig. 1).
Comparison of this plot with the concentration-time plot together with the Km
value, demonstrates the importance of data in the Km region and the difficulty
of obtaining this in a batch reactor. It is useful to make specially-scaled graphs
in the KM region.
8.2 Batch Reactors 213
Model
Program
To make the Lineweaver-Burk plot, the inverse values of S and rs are calculated
in the program on the CD-ROM.
Nomenclature
Symbols
Indices
0 Refers to feed
m Refers to maximum
S Refers to substrate
Exercises
Results
The results are shown in Fig. 2 (rates and concentrations versus time) for a
range of Michaelis-Menten constants KM and in Fig. 3 the corresponding
Lineweaver-Burk plots.
8.2 Batch Reactors 215
0.45
.0.4
0.35
.0.3
.0.25
0.2
-0.15
0.1
0.05
140 160
Figure 2. Rate and concentration plots for KM = 0.2, 0.5, 1.0 and 2.0 (bottom to top curves).
Figure 3. Lineweaver-Burk plots for KM = 0.2, 0.5, 1.0 and 2.0 (bottom to top curves).
System
Some enzyme catalyzed reactions are very complex. For this reason their
rigorous modelling leads to complex kinetic equations with a large number of
constants. Such models are unwieldy and are usually not suitable for practical
216 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
La + E ^ ^ LaE ^- Ga + GI + E
Ga E + La ^ ** E + Tr
GaE + H2O ^ E + Ga
Figure 1. Complex and simplified models for the enzymatic hydrolysis of lactose, where the
symbols are La for lactose, Ga for galactose, Gl for glucose, Tr for trisaccharide and E for
enzyme.
K
!
1L.a
a to Ga T. V3II
f^i
K
1
La -i- Ga
fcaCl ^ Tr
i
K
2
The simulation of this model is easy, and the constants can be adjusted to
achieve good agreement with experimental data.
Model
dLa
-gj- - K! La - KI La Ga + K2 Tr
dGa
= Kj La - KI La Ga + K2 Tr
dTr
= KI La Ga - K2 Tr
Program
It was found that K2 must be two orders of magnitude greater than KI in order
to bring the simulation into agreement with the experimental data. The
program is on the CD-ROM.
Nomenclature
Symbols
Indices
Exercises
Results
The outputs in Figs. 3 and 4 show the influence of KI, KI and Lao on the sugar
concentration profiles.
90.
80.
70
60.
. 50.
40
30
20
10
0
20 100 180 200
TIME
80
( '
Reference
System
Heinzle and Lafferty (1980) have presented a structured model to describe the
batch culture of Alcaligenes eutrophus under chemolithoautotrophic growth
conditions, as discussed in Case C, Sec. 3.3.1. Growth and storage of PHB are
described as functions of limiting substrate S (NH4+), residual biomass R and
product P (PHB) concentrations.
220 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Model
In the model seen in Fig. 1 the whole cell dry mass (X) consists of two main
parts, namely PHB (P) and residual biomass (R), where R is calculated as the
difference between the total cell dry weight and the concentration of PHB (R =
X - P). R can be considered as the catalytically active biomass, including
proteins and nucleic acids. With constant concentrations of the dissolved gases,
two distinct phases can be recognized: growth and storage. During the growth
phase there is sufficient NH4+ to permit protein synthesis. When the limiting
substrate NH4+ (S) is exhausted, the protein synthesis ceases, and the
production rate of PHB is increased. During the storage phase only PHB is
produced. The limiting substrate NH4+ (S) is essential to produce R and limits
its synthesis at low concentrations.
For the batch process,
dR
= r
dF R = MR
where TR is the rate of synthesis of R and (j is the specific rate of synthesis of R,
where
S (S/Ks,2)n
+ S) + ^m,2 ! + (S/KS,2)n
where n is the empirical Hill coefficient (see Sec. 3.1.2), having a value of 4 in
this example.
This is based on the postulate that there are two different mechanisms for the
assimilation of NH4+ in procaryotes. This formulation is not a mechanistic one,
8.2 Batch Reactors 221
dP
df = rp = r P j + rP,2
where r P j = YP/R rR
The non-growth associated term of the synthesis of P(rp,2) is assumed to be a
function of the limiting substrate S, of the residual biomass R and of the
product P. When the PHB content in the cells is high, the rate of synthesis of P
is decreased, which can be formally described as an inhibition.
Program
Nomenclature
Symbols
Indices
1 Refers to reaction 1
2 Refers to reaction 2
m Refers to maximum
Exercises
8.2 Batch Reactors 223
Results
3.5- 14
/ ^"~'"~*" "
3- f -*' 12
y *
of - *"'"
'"" J
I -'
'*"
1.5- -6
y " f'
1-
/ / " '. .'
4
0.5-
0-
^^
*"^ V
_j__ *%
-2
_n
0 5 10 15 20 25 30 35 40
TIME
Figure 2* Profiles of residual biomass concentration R, substrate S and product P in the batch
fermentation.
Run 4: 41 6 steps in 0.01 67 seconds
35- ... 5
'"'V^
^.^.- 4.5
30-
'v -4
25- \ / .._.. 3:3(2,3) 3.5
\ / ~- P:3(2.3)
3
20- 1 / P:4 (5)
a \ -2.5 (/)
-*. \ /
15-
-2
"""-, \ f .*""r-'" -1.5
10-
5
"\ "" \"' /^i''^' -1
-0.5
0. ^^";:^ -n
0 5 10 15 20 25 30 35 40
TIME
References
System
Model
Volume,
dv
dT = FO
Substrate,
224 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
Model
Volume,
dv
dT = FO
Substrate,
- = F0S0
Biomass,
d(VX)
dt = rx
The kinetics are
rx = MX
|LimS
** - (Ks + S)
and
rx
rs = -Y
The dilution rate is defined as
In order to simplify the equations and to present the results more generally, the
model is written in dimensionless form. Defining the dimensionless variables:
v V
=
X
X< =
s s
=
F =
- _ JL
Mm
F
tf-
t' = t
226 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
d(V S) = V dS + S dV
and
d(V X) = V dX + X dV
Volume
dV'
Biomass
dX'
dt'
Substrate
dS'
dt = (l-S)D-jiX
The Monod equation is:
KS +S
Quasi- steady
Programs
Nomenclature
Symbols
Indices
Dimensionless Variables
Exercises
8.3 Fed Batch Reactors 229
Results
During the quasi-steady state, \l becomes equal to D, and this requires that S
must decrease steadily in order to maintain the quasi-steady state as the volume
increases (Fig. 3). Increasing flow rates from 0.01 to 1.0 causes a delay in the
onset of linear growth and causes the final biomass levels to be higher (Fig. 4).
Run 1:105 steps in 0 seconds
4.5
Figure 3. Fed batch concentration and growth rate profiles, showing quasi-steady state.
2.5 C/>
10
230 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Figure 4. Influence of flow rate on growth. Flow rate increase from 0.01 to 1.0.
References
Dunn, I.J., and Mor, J.R. (1975) Variable Volume Continuous Cultivation.
Biotechnol. Bioeng. 17, 1805.
Keller, R., and Dunn, I.J. (1978) Computer Simulation of the Biomass
Production Rate of Cyclic Fed Batch Continuous Culture. J. AppL Chem.
Biotechnol. 28, 784.
System
Oxygen
Precursor
Phenylacetic acid
Sulfuric acid
Ammonia
Model
a) Elemental Balancing
Knowing the composition of all chemical substances and the biomass mycelium
(Table 1) allows the following steady state balances of the elements in terms of
mol/h:
For carbon
6 RI + R6 + 16 R3 + 8 RH + 16 RH + R2 = 0
For oxygen
For nitrogen
0.16R 2 + 2 R 3 + 2R 4 + R8 = 0
For sulfur
0.00 46 R2 + R3 + R4 + R9 = 0
8.3 Fed Batch Reactors 233
For hydrogen
For phosphorus
0.0054 R2 + RIO = 0
- 3 1 9 R i o - 6 9 R n -68Ri 2 + rH = 0
- R i o = 0.0054 R2
To complete the model, equations for glucose uptake rate (-Ri), biomass
formation rate (R2), rate of penicillin formation (Rs), precursor consumption
rate (-Rn), and rate of penicillin hydrolysis (R4) must be known. Note that the
reaction rates are defined with respect to total broth weight, since the process is
the fed-batch type and broth weight is variable with respect to time.
-QlCiM2
where Y2 is the maximum growth yield and m is the maintenance rate factor
(mol glucose/mol mycelial biomass h).
Some sugar is used in the formation of the product. Hence,
- Rl = Yj R2 + m M2 + YJ (R3 + R*)
R2 = - Y 2 R i - Y 2 m M 2 - yf (Rs + 4)
- R l l = R 3 + R4
where - RH is the precursor consumption rate.
8.3 Fed Batch Reactors 235
R 3 = Q3 M2 - R4
R 4 = K 3 M3
c) Balance Equations
180 F
F F =
l = 500 2.78
where F = mol glucose /h.
Feed rate of NH3 stream (kg/h)
F8 = R 8 25Q = T4JT
dG F R9 Rg
+ +
"dT = Tn 235" TTTT
Component Balances
Expressed in mol/h the dynamic balances are,
236 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Glucose
~3T = R l + F
Biomass
dM2
~ar = R2
Penicillin
dM3 .
Penicilloic acid
dM4
M2
Tr
M3
M4
c4 =
where the masses MI, M2, M3 and M4 are in mol units.
d) Metabolism Relations
R2
f 2 = 0.16 R|
N2 fraction in penicillin
f3 = 1-F 2
f5 = 1 - F4
Program
The Madonna program covers a fermentation time of 200 h starting from the
initial conditions of 5500 mol glucose, 4000 mol biomass, 0 mol penicillin and
0.001 mol penicilloic acid in an initial broth weight of IxlO 5 kg. The program
is on the CD-ROM.
Nomenclature
Symbols
Indices
0 initial
1 glucose
2 biomass
3 penicillin
4 penicilloic acid
5 oxygen
6 carbon dioxide
8 ammonia
9 sulfuric acid
10 phosphoric acid
11 phenylacetic acid
12 water
Exercises
8.3 Fed Batch Reactors 239
Results
The results of Fig. 2 show the substrate MI to pass through a maximum, while
the penicillin M2 develops linearly, for this constant feeding situation.
Increasing the feeding linearly with time (F = 500 + 5* time) gave the results in
Fig. 3, where it is seen that maintenance accounts for about 70 % of glucose
consumption at the end of the fermentation.
Run 1:215 steps in 0 seconds
Figure 2. Penicillin fed batch fermentation with total masses of glucose (M]) and biomass
(M2).
240 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
0.8-
0.7-
0.6-
8
r-
- 0.4-
0.3-
0.2-
0.1-
0-
20 60 80 100 120
TIME
Reference
Heijnen, J., Roels, J. A., and Stouthamer, A.H. (1979). Application of Balancing
Methods in Modeling the Penicillin Fermentation. Biotechnol. and Bioeng., 21,
2175-2201.
System
Yeast exhibits diauxic behavior with respect to the glucose and ethanol in the
medium as alternative substrates. In addition, the glucose effect, when glucose
levels are high, will cause fermentation, instead of respirative oxidation, to take
place, such that the biomass yields are much reduced (Fig. 1). In this example
the constant a designates the fraction of respiring biomass and (1 - a) the
fraction of biomass that ferments. The rates of the process are controlled by
three enzymes.
8.3 Fed Batch Reactors 24 1
^^ C02 + X
Glucose
^^^*- Ethanol + X
Model
R, =
Glu+K sl
Fermentation to ethanol,
R2 = K2 (1 - a) X
Glu + KS2
R4 = - -rXEo
K S4 +Glu 3
R5 = K 5 X E i
The mass balances for the biomass, substrates and enzymes are those for a fed
batch with variable volume.
242 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
dt
The component balances are written by separating the accumulation term,
noting that
dt V
^o = _ E0Q
dt V
f -*-.-*
Program
Note that the program on the CD-ROM is formulated in terms of C-mol for the
biomass. This is defined as the formula weight written in terms of one C atom,
thus for yeast CHL667Oo.5No.i67-
8.3 Fed Batch Reactors 243
Nomenclature
Symbols
Indices
0 Refers to feed
1 Refers to reaction 1
2 Refers to reaction 2
3 Refers to reaction 3
4 Refers to reaction 4
5 Refers to reaction 5
Exercises
244 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
Seen in Fig. 3 are the simulation results giving the concentrations (glucose,
ethanol and biomass) during the fed batch process. In Fig. 4 the maximum in
ethanol concentration as a function of feedrate is given from a Parameter Plot.
Run 1: 605 steps in 0.0167 seconds
30
25
60
Reference
System
Model
The equations are the same as given in the example FEDBAT (Section 8.1.3),
where the balances for substrate and biomass are written in terms of masses,
instead of concentrations. The only difference is that an outlet stream is
considered here to empty the fermenter at the end of the production period.
Program
Nomenclature
Symbols
Indices
S Refers to substrate
X Refers to biomass
0 (zero) Refers to initial and inlet values
initial Refers to initial values
in Refers to inlet
out Refers to exit
Exercises
248 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
50- A pr^Ti
| "vsi|
/- 70
, go
/ I
40-
/ I /"( / .50
30-
/ I ' J -40
f I / \
20- / I / i /-V
/ I/ I / ^\
10-
\l XV ,* \ -10
0- --*'"-' L -Q
0 5 10 15 20 25 30 35 40 45 50
TIME
Figure 2. Masses of substrate and biomass during filling and emptying cycles.
8.3 Fed Batch Reactors 249
Figure 3. Concentrations of product, substrate and biomass during filling and emptying
cycles. The volume is also shown.
References
Keller, R., Dunn, I.J. (1978) Computer Simulation of the Biomass Production
Rate of Cyclic Fed Batch Continuous Culture J. AppL Chem. Biotechnol. 28,
784.
System
Slow-growing animal and plant cell cultures require certain growth factors and
hormones which begin to limit growth after a period of time. To avoid this,
part of the entire culture is replaced with fresh medium. A single cycle of
repeated replacement culture is shown in Fig. 1. In this procedure part of the
medium volume (with cells) is removed after a certain replacement time and
replaced with fresh medium. Each cycle operates as a constant volume batch in
250 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Replacement
Final Conditions VX VS
Initial Conditions
Model
The equations are those of batch culture, where for convenience the total
masses are used.
dVS = r
"dT sV
dvx
* VR
f=
VX = (1 - ) VXF
8.3 Fed Batch Reactors 251
VS= ( l - f ) V S F + V R S 0
Program
The program as shown on the CD-ROM makes use of the PULSE function to
vary the biomass and substrate concentrations corresponding to the
replacement of a fraction F of the culture medium. The time for each batch is
the value of INTERVAL.
Nomenclature
Symbols
Indices
Exercises
Results
Fig. 2 shows how the biomass increases, until after six cycles the time profiles
become almost identical.
TIME= 19.29 X = 1.26
10 20 30 40 50 60 70 90 100
TIME
Figure 2. Oscillations of biomass and substrate concentrations with replacement cycles for
Interval 10 and F=0.8
8.3 Fed Batch Reactors 253
Model
Total mass
dt
Biomass:
d(MassX)
= Vr-X
dt
Substrate:
d(MassS)
= FSf+Vrs
dt
Product:
254 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
d(MassP) _.
= V fp
p
dt
Dissolved oxygen, neglecting the content of the inlet stream is calculated from
d(MassO)
= K L a*(O sat -0) + Vr 0
dt
The influence of biomass concentration on the oxygen transfer is
approximated here by
KX+X
The substrate uptake kinetics includes that amount used for growth, for product
and for maintenance
Product production involves two terms whose constants are turned on and off
according to the value of |ii, as seen in the program.
=-TT
Y
xo
8.3 Fed Batch Reactors 255
Program
Nomenclature
Symbols
Exercises
256 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
II!
References
Results
0.007
100
! I 0.006
80 i I 0.005
- 60
: I 0.004 O
Li I -0.003
40 I
-0.002
20
-0.001
0
0 20 40 60 80 100 120 140 160 180 200
TIME
Figure 3. Influence of initial KLa value from 100 to 160 h"^ on the S and O profiles.
System
-0.007
100 ^
1I 1 -0.006
80 I
I 1
-0.005
\ . !i i1
. ll -0.004 O
v i
- 60
'-, 1
40 \\1 1 |\ i
I i '\ *
-0.003
-0.002
20
\ \! \l -0.001
!
\\ \
-0
20 40 60 80 100 120 140 160 180 200
TIME
Figure 3. Influence of initial KLa value from 100 to 160 h"^ on the S and O profiles.
System
D,S
Model
The dynamic balance equations may be modified to apply only to the steady
state by setting the time derivatives equal to zero. The corresponding equations
are then:
For biomass,
0 = - D X + rx
For substrate,
0 = D (S0 - S) + rs
Growth kinetics,
rx = ^ X
\i = D
where S is determined by the kinetics
\i = f(S)
The Monod relation results in,
S =
The substrate balance gives,
X = Y(S 0 -S)
8.4 Continuous Reactors 259
The above equations represent the steady state model for a chemostat with
Monod kinetics. Using them it is possible to calculate the values of S and X,
which result from a particular value of D, and to investigate the influence of the
kinetic parameters.
Program
In Madonna programs, time can be used as a variable which will increase from
the starting time. Here it is renamed D. Thus equations will be solved for
increasing values of the dilution rate. Fortunately X and S can be explicitly
solved for in this problem. If not, the ROOT FINDER facility of Madonna can
be used. The program is found on the CD-ROM.
Nomenclature
Exercises
260 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
The steady state curves of X, S, and XD versus D are given Fig. 2. The results
in Fig. 3 were obtained by varying K$ in each run. An interesting effect can be
observed on the position of the washout point.
Run 1:113 steps in 0 seconds
10 - -4
9-
3.5
8-
*S**~\ !I -3
7-
m ^r Y.-I
\1 i
6-
f*r
i ~s;i
mm .
\! -2.5
v- 5
" S" 1 2 S
4-
1.5
3-
1
2-
// /{
1-
0.5
0- 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
System
If substrate concentrations are low, the term S2/Kj is lower in magnitude than
KS and S, and the inhibition function reduces to the Monod equation. In batch
cultures the term S2/Kj may be significant during the early stages of growth,
even for higher values of K[. The inhibition function passes through a
maximum at Smax = (Kg Ki)-5. A continuous inhibition culture will often lead
262 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
to two possible steady states, as defined by the steady state condition JLI = D and
as in shown Fig. 1.
D=
One of these steady states (A) can be shown to be stable and the other (B) to be
unstable. Thus, only state A and the washout state (S = SQ) are possible.
Model
+- F,S,X
VdX
jj- = ^ i V X - F X
or,
VdS ^VX
= F (S 0 -S) -~^f
or,
dS MX
dT = D (S0 - S) -
Program
When the system equations are solved dynamically, one of two distinct steady
state solutions are obtained, the stable condition A and the washout condition.
The initial substrate and organism concentrations in the reactor will determine
the result. This is best represented as a phase-plane plot X versus S. All results
indicate washout of the culture when the initial cell concentration is too low;
higher initial substrate concentrations increases the likelihood of washout.
Nomenclature
Symbols
Indices
0 Refers to inlet
I Refers to initial value
m Refers to maximum
Exercises
8.4 Continuous Reactors 265
Results
10 15 20 25 30 35 40
TIME
4.5 -
3.5
e/> 2.5
1.5
0.5
0
0.5 2.5
Figure 4. Phase-plane plot of X versus with varying ST from 0 to 5 kg/m3 using Batch Runs
with overlay.
266 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
4 - ^^^>
3:2 (0.5)
*" t**
1 3:3 (0.5857)
3.5- ^ *t %% _ - - 3:4(0.6714)
\ * * C^1 v 3:5(0.7571)
3 "| "% % V 3:6(0.8429)
\ ^ x 3:7(0.9286)
3:8(1.014)
</) 2 . 5 - i ! l 3:9(1.1)
f/ // // /;^',--^,
2
1.5-
1 / / J
. *x
j* S ^ "^ -s^^-
K
f f / t / ,- rf'^T "^"^>*i'^' ** -^
0.5- f
:' / / * / Jf*'^ ^^^*^^$^S? **
!
0 * \ f 1 C & ( ^^*^^ **
0.5
Figure 5. Phase plane plot of influence of the initial biomass Xi from 0.5 to 1.1 for = 0.0.
Steady states upper left and lower right.
Reference
Edwards, V.H, Ko, R.C. and Balogh, S.A. (1972). Dynamics and Control of
Continuous Microbial Propagators Subject to Substrate Inhibition Biotechnol.
and Bioeng. 14, 939-974.
8.4 Continuous Reactors 267
System
O,F O 2, F4
Reieto*
2, F3
Model
The dynamic balance equations can be written for all components around the
reactor and around the settler. The settler is simplified as a well-mixed system
with the effluent streams reflecting the cell separation.
Organic substrate balance for the reactor:
= F 0 So + F 2 S 2 -
268 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
R2Vt
= FQ AQ + F2 A2 - FI A
pj7 = p2 O2 FI QI + RI Vj_
l
di = F 2 N 2 - F i N i + R2Vi
V 2 dS 2
j^ = F i S i - F3S2 - F4S2
V 2 dA 2
= F A - F4A2
3t I I -
Balance for heterotrophic organisms in the settler:
V 2 d0 2
dt = Fl l - F3 02
V2 dN2
34 = F i N i - F 3 N 2
Recycle flowrate:
F2 = F 0 R
where R is the recycle factor.
Monod-type equations are used for the growth rates of the two organisms.
R =
l
R
l^2max
2 = ^Ni =
Program
Nomenclature
Symbols
Indices
Exercises
8.4 Continuous Reactors 271
Results
The results in Fig. 2 demonstrate the influence of flow rate on the effluent
organics 82- The ammonia in the effluent A2 is seen, in Fig. 3, to respond
similarly to FQ, but for a very high value of FQ = 1000 m3/h the nitrification
ceases, and A2 becomes the same as the inlet value AQ. This corresponds to
washout of the nitrifiers, which would be seen by plotting NI versus time.
0.8 -I ,/"
M if
If J
II /
It
0.3- rr
J 82:1(20)
02. II 82:2(180)
" 82:3(340)
0-1 JM | 82:4(500)
" .
6 2 4 6 8 10 12 14 16 18 20
TIME
0.09-
0.08-
0.04-
0.03-
0.02J
0 2 4 6 8 10 12 14 16 18 20
TIME
Figure 3. Ammonia in the effluent (A2) at various flow rates F0 (5 to lOOOm^/h, bottom to
top).
272 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
Enzyme
concentration
Enzyme distribution
Model
Substrate balance,
dS 1
=
dZ ~v
Kinetics,
The linear flow velocity is increased by the presence of the solid enzyme carrier
particles according to
8.4 Continuous Reactors 273
F
V7
L = ~
Ae
vm = KE
E = E0 + mZ
Program
Nomenclature
Symbols
Indices
0 Refers to inlet
S Refers to substrate
Exercises
Results
Flow rate is the primary operating variable, along with enzyme loading and
inlet concentration. In Fig. 2 the influence of F is seen in the steady-state, axial,
substrate profile.
8.4 Continuous Reactors 275
10 12 14 16 18 20
System
Model
= D
1
(SlFeed - Si) -
/O O \ I ^__i^^_ I i i "V
D (o2pee(j o2) lv YQO J ]LlA
where D = F/V.
The yield coefficients are assumed to vary with the carbon-nitrogen ratio in the
reactor.
Si
RATIO = ^
The yield coefficients are varied according to RATIO using the following logic:
The boundaries of the three growth regimes in Fig. 1 are defined by the
quantities BI and B2.
8.4 Continuous Reactors 277
XSi
0.8
r
XS1
B2
S2
Figure 1. Limitation regions for carbon and nitrogen showing influence on yield.
The yield coefficients for biomass on nitrogen and carbon take maximum or
minimum values when only one substrate is limiting and vary linearly with
opposing tendencies in the double-limitation region.
Program
Note that the programing of this example is rather more complicated than usual
owing to the need to allow for the logical conditions of carbon limitation,
nitrogen limitation or both substrates together causing limitation. A partial
listing is seen below and the full program is on the CD-ROM.
Nomenclature
Symbols
Indices
Exercises
8.4 Continuous Reactors 279
Results
The startup of a continuous culture is shown in Fig. 2. Note that the nitrogen
level 82 in the reactor drops to a low level after 15 h and causes a change in the
yield coefficients. The influence of dilution rate on the system was investigated
by varying D from 0 to 1.5 as shown in Fig. 3.
X 1.5
Reference
System
The process involves the removal of dichloromethane (DCM) from a gas stream
and the subsequent degradation by microbial action. The reactor consists of
biofilm sand bed column with circulation to an aeration tank, into which the
substrate and oxygen enters in the gas phase, or the substrate can be fed in a
liquid stream, as shown in Fig. 1. The column is approximated by a series of six
stirred tanks. The reaction is treated with homogeneous, double saturation
kinetics with dichloromethane (DCM) inhibition. The oxidation of one mole of
DCM produces 2 moles of HC1, making a hydrogen ion balance for pH im-
portant. The yield with respect to oxygen is 4.3 mg DCM/mg 62. In practice,
care must be taken to prevent stripping of DCM to the air stream.
8.4 Continuous Reactors 281
C
SR6>
SRin C jn , pH jn
Model
The model does not include a gas balance on the aeration tank, since it is
assumed that the gas phase dynamics are comparatively fast and hence an
equilibrium with the inlet concentration of oxygen and DCM may be assumed.
The biomass is assumed to grow slowly, and growth rates are therefore also not
modelled. The model for pH changes does not include buffering effects.
For the inlet section 1 at the bottom of the column the balances are as follows:
O2 balance,
dCQ1 ^
dt
282 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
DCM balance,
C
dC
Srl _ Srin~ C Srl
dt t
+
H ion balance,
dCHi i -CHI 2r S i
84900
Here T is the residence time of the liquid in one section of the column. The
constant 84,900 converts grams to moles and includes the stoichiometry.
V
maxCSrl -01
KI )
K L a 0 2(Co2eq-Coin)
dC R
-~ = (CSr6 - CSrin ) DCM (Cs2eq - Tr- (CSFO ~ CSrin
at V
Program
The program constants describe DCM entering the reactor in the gas stream.
The DCM concentration in the liquid feed is set to zero. The program is on the
CD-ROM.
8.4 Continuous Reactors 283
Nomenclature
Exercises
11
284 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
The concentrations in the stream leaving the top of the column (CSr6) during
startup of the fluidized bed are shown in Fig. 2 for four values of F (0.5 to 10)
The change of the pH for one flow rate (F = 0.5) is shown in Fig. 3.
8.4 Continuous Reactors 285
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
TIME
Figure 2. Fluidized bed startup for four values of F (0.5 to 10, bottom to top).
2.5
CSR6:1
: 2 ... PH6:1
1.5
0.5
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
TIME
Figure 3. Change of carbon substrate and pH in the top section 6 during startup.
Reference
System
Two chemostats are arranged in series (Fig. 1) with the intention that the first
operates at a relatively high rate of cell growth, while the second operates at low
growth rate, but high cell density, for secondary metabolite production.
Additional substrate may be fed to the second stage.
, 810
Hi
X1.S! US
Model
The balance equations are written for each component in each reactor.
= F[S O -S!] -
8.4 Continuous Reactors 287
Stage 2 with additional substrate feed and an input of cells and substrate from
Stage 1,
V2 - [F + Fi]X2
KS + S2
First stage,
Prodi =
V,
Both stages,
Prod2 =
Program
Nomenclature
Symbols
Indices
Exercises
Results
The results in Fig. 2 give biomass concentrations and productivities for both
tanks during a startup with a constant feed stream to the first tank (F = 0.5). In
Fig. 3 the influence on X2 of feed to the second tank Fl (0 to 1.0) with
constant F is shown.
8.4 Continuous Reactors 289
35 40
Figure 2. Biomass (Xj X2) and productivities for both tanks (F = 0.5).
4.5
3.5
32.5
2
1.5
0.5
0
10 15 20 25 30 35 40
TIME
Figure 3. Influence on X2 of feed to the second tank (Ft = 0 to 1.0, curves right to left).
290 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
Model
When product concentrations are low, the equation reduces to the Monod
equation.
The product kinetics are according to Luedeking and Piret, with dependence
on both growing and non-growing biomass,
an - a"Q
~ 1i-r+rPn
_ _rxn
Mass balances:
Stage 1,
j- = F[So-Si] +r S iVi
dX2
V 2 -gjT- = F Xj - [F + F!]X2 + rX2V2
dS2
V2 -gj- = F [Si - S2] + FI [Sio - S2] + rS2V2
dP2
- =FPl- [F + Fi]P2 + rp2V2
Prodi =
Both stages,
Program
Nomenclature
Symbols
Indices
n Refers to tank n
0 Refers to tank 1 inlet
1 Refers to tank 1 and inlet of tank 2
8.4 Continuous Reactors 293
Exercises
Results
The startup and approach to steady state for the two stages is shown in Fig. 2.
The influence of the inhibition can be tested by varying KI from 0.1 to 10.0, as
shown in Fig. 3. The higher the KI the lower is the degree of inhibition and the
greater is the product concentration P2-
294 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
3.5 ,
3
2.5
,
1.5
0.5
0
10 15
TIME
Figure 2. Startup and approach to steady state for the two stages.
1.3
1.2.
1.1
I
1
0.9.
0.8
0.7 J
10 12 14 16 18 20 22 24 26
TIME
Reference
System
? So , Fn
Fluidized
Bed
F,S
Model
The model balance equations are developed by considering the individual tank
stages and the absorber separately. The gas phase in the absorber is assumed to
be air.
Substrate balances:
For the absorption tank
dS FR
dF =
For each stage n
dSn FR
-3T = -^(Sn-!-Sn)- rsn
Oxygen balances:
For the absorption tank
Program
Nomenclature
Symbols
Indices
0 Refers to feed
l,2,3,n Refer to the stage numbers
m Refers to maximum
O Refers to oxygen
S Refers to substrate
T Refers to aeration tank
X Refers to biomass
Exercises
298 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
Note from the results below that the steady state for oxygen is reached rather
quickly, compared to that of substrate.
Figure 2. Oxygen concentrations in fluidized bed reactor. Top of column is the lower curve.
8.4 Continuous Reactors 299
tf
System
NO2~ + O2 - NO3~
The overall reaction is thus
Both steps are influenced by dissolved oxygen and the corresponding nitrogen
substrate concentration. Owing to the relatively slow growth rates of nitrifiers,
treatment processes benefit greatly from biomass retention.
300 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Model
03.
Fluidized
bed
In the absorber, oxygen is transferred from the air to the liquid phase. The
nitrogen compounds are referred to as Si, 82, and 83, respectively. Dissolved
8.4 Continuous Reactors 301
Similarly for the absorption tank, the balance for the nitrogen-containing
components include the input and output of the additional feed and effluent
streams, giving
The oxygen balance in the absorption tank must account for mass transfer from
the air, but neglects the low rates of oxygen supply and removal of the feed and
effluent streams. This gives
For the first and second biological nitrification rate steps, the reaction kinetics
for any stage n were found to be described by
v
r = ml Sin Qn
K + S K + O
V
r2n = m2 S2n n
K S K
2+ 2n O2+n
The oxygen uptake rate is related to the above reaction rates by means of the
constant yield coefficients, YI and 2, according to
ron = - H n Y i -r 2 n Y 2
The reaction stoichiometry provides the yield coefficient for the first step
302 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Program
Nomenclature
Symbols
Indices
Exercises
304 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
10 15 20 25 30 35 40 45 50
270
260
250
1^240.
5
,230-
c
<
220
M I
210
200 0.5
190
180
15 20 25 30 35 40
KLA
Figure 3. Parametric run of continuous operation showing oxygen and ammonia in the effluent
versus
8.4 Continuous Reactors 305
System
EO.FE I S 1f P 1 F1
Model
The mass balance equations are formulated by noting the two separate feed
streams and the fact that the enzyme does not react but is conserved.
Total flow:
FS + FE =
Mass balances:
dSi
= FsSo-FiS1+rsV
306 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
r = -FiP1+rpV
r
S - - v mK M + S + (P/Ki)
vm = EI K2
rP = - 2 r s
Program
Nomenclature
Symbols
Indices
Exercises
Results
1.6
--- P1:2(0.2)
P1:3(0.3)
0.8
s
.0.6 / / ,-'"
0.4 tS
0.2
0 10 20 30 40 50 60 70 80
TIME
Figure 2. Performance for three values of FE.
3.5
3 x--'
2.5
.2
<*/r
jft
I . P1:2(1.5)
.. P1:3(2)
0 10 20 30 40 50 60 70 80
TIME
System
Biocatalysts usually deactivate during their use, and this has to be considered in
the bioreactor design. One of the methods to keep productivity fluctuations
low, and hence to efficiently utilize the biocatalyst, is to use a series of reactors
with biocatalyst batches having different times-on-stream in each reactor. In
8.4 Continuous Reactors 309
this example a series of three stirred tanks of a constant equal volume with
biocatalyst deactivating by first order reaction kinetics is investigated (Fig. 1).
After a time period ILAG? ^e biocatalyst from the tank with longest time-on-
stream (first tank in the cascade) is discarded and replaced by a fresh batch.
The streams are switched over so that tank 1 becomes tank 3, the last reactor in
the series. Other tanks are switched over correspondingly. This is equivalent to
replacing the used enzyme with fresh enzyme in tank 3 and moving the used
enzyme upstream from tank 3 to tank 2 to tank 1, which is easier to simulate.
(3-galactosidase was taken as an example of the biocatalyst. This obeys
Michaelis-Menten kinetics with competitive product inhibition, and the kinetic
constants were determined with considerable accuracy. The same constants are
used also in this substrate inhibition model.
F,S 0
Model
Using the stoichiometry, S > P, the mass balances for the ith tank (i = 1, 2, 3)
with the volume V can be written
Substrate
Product
= F(P M -Pi)
dt
Enzyme (active)
310 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
=r Ei v
where rate of substrate consumption is given by product inhibition
(competitive)
s
r i
Si = ' v max b i -7-Z~^\
K
V inh
R P i = -R Si
rEi = - kD EI
This equation can be applied by changing the initial conditions for each tank
when the enzyme is moved from tank to tank. Thus the final value in tank n
becomes the initial condition in tank i-1. The initial conditions can also be
calculated by analytical integration of the enzyme deactivation equation at
times corresponding to the respective ages of the biocatalysts in the respective
reactors (multiples of TLAG)- Fresh enzyme with the activity EQ is in the third
tank. The other tanks start with the following enzyme activities:
EI = E0 e C- (3 - i) ko TLAG]
Program
In the program on the CD-ROM note that the cost calculation at the end of the
program is included only as a comment but could be incorporated into the
program with the corresponding values for the constants.
8.4 Continuous Reactors 311
Nomenclature
Symbols
Indices
Exercises
312 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
0.45-
\
s *** -3500
0.4- \ j**
% r* -3000
0.35- *. i**
"% ^
a ' 03 x
x "'' -2500 3
I
-2000 Q_
.0.25-
s . E2:1
"* 0.2- -. -' _- Totalproduct:1 -1500 p
*"">cr H
^r "'"'"-'"*. |
0.15-
-1000
0.1- _/ j+ " . -i.. .500
0.05- i...
0- ^ -" -0
0 100 200 300 400 500 600 700 800 900 1000
TIME
Figure 2. Exponential biocatalyst deactivation and total product during one run.
120- 100
100-
80
8
l
C/l
-
a-
^ 60
cn
40
40-
20
20
0
0 100 200 300 400 500 600 700 800 900 1000
TIME
References
Prenosil, I.E., Peter, J., Bourne, J.R. (1980). Hydrolytische Spaltung des
Milchzuckers der Molke durch immobilisierte Enzyme im Festbett-Reaktor.
Verfahrenstechnik 14, 392.
Prenosil, J.E. (1981). Optimaler Betrieb fur einen Festbett- und einen Fliessbett-
Reaktor mit desaktivierendem Katalysator. Chimia 35, 226 .
System
Sfeed,
> 82, F0
Model
The details of the structured model will not be repeated here (See PHB). The
biomass consists of a synthesis part R and the intracellular product P. The
biomass growth rate of R is proportional to the specific growth rate, which is
given by a two-part expression
S (S/Ks,2)n
(KS,i + S) -*
The model requires component balances for P, R and S for both tanks, as seen
in the program. The relative reactor volumes are determined by the parameter
Vrat. The volumetric productivities are calculated to compare the results.
Program
Nomenclature
Symbols
Indices
Exercises
8.4 Continuous Reactors 317
Results
90 100
Figure 2. A run showing the dynamic approach to steady state for X, S, P in both tanks.
0.2
Figure 3. Here with FO set at the optimum value of 1.24, the influence of VRAT is investigated
giving a value for the maximum in PROD corresponding to the OPTIMIZE results. VRAT is seen
not to be very important. Thus equal-sized tanks are adequate.
318 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
The influence of gassing rate and stirrer speed on an enzymatic, aerated reactor,
as shown in Fig. 1, is to be investigated. The outlet gas is assumed to be
essentially air, which eliminates the need for a gas balance for the well-mixed
gas phase.
gas
Hi
Ill + 02
ii
air
Model
_S CL
dS
= r
dF S
dCL *
= K L a(C L * -CL) - r s Y o /s
dP
=
dF ~ r s Y P/s
KLa varies with stirring speed (N) and aeration rate (G) according to:
KLa = kN 3 G- 5
Program
Nomenclature
Symbols
Indices
0 Refers to feed
o Refers to oxygen
p Refers to product
s Refers to substrate
Exercises
Results
The results in Fig. 2 show the influence of stirrer speed N on the dissolved
oxygen level. Variations from 30,000 to 5,000 1/h were made with the Batch
Run facility. Runs to obtain the results in Fig. 3 were made by varying the gas
flow rate G from 25 to 5 m3/h (curves top to bottom).
8.5 Oxygen Uptake Systems 321
7s"
8
7 -
_..-....._..-..._ --'/r
f
*
.*** *
6 -
5
i
CL1 (3e+4) *
CL:2 (2.1676+4) 1
3 -- CL:3 (1.3336+4) /
-^_CL:4 (5000) 4
2 -
{ -"-''
0
C 1 2 3 4 5 6 7 8 9 1 0
TIME
<j 6.5
1 2 3 4 5 6 7
TIME
System
Cell growth is limited by the oxygen mass transfer rate, and hence by the
dissolved oxygen concentration. It is also inhibited by an inhibitory substrate S.
Liquid phase balances for X, S and 62 in the liquid phase are therefore used,
together with a gas phase oxygen balance to determine the rate of O2 supply.
322 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
To avoid washout of cells, it is important that the reactor should never enter the
range of inhibitory behavior. Schematic representation of a continuous aerated
fermenter is given in Fig. 1.
liquid F,S 2 , X
air, G, CG1
Model
For biomass,
dX
rxVL
For substrate,
dS2
For oxygen,
dCL2
L2
KS+S2+(S22/KI)K0+CL2
rx = |a X
8.5 Oxygen Uptake Systems 323
rs =
~
The oxygen equilibrium relates the concentration in the gas phase to the liquid
phase saturation concentration,
CL2* = MC G2
The gas holdup fraction is,
VG = eV L
Proportional control of the feed rate, based on exit substrate concentration, can
be added with,
F = Fo + KpE
with E = S2set ~ $2- Here the sign must be adjusted depending on the substrate
region above or below the maximum kinetic rate.
Program
Nomenclature
Symbols
Indices
0 Refers to feed
1 Refers to inlet
2 Refers to outlet
G Refers to gas
I Refers to inhibitor
L Refers to liquid
m Refers to maximum
O Refers to oxygen
P Refers to product
S Refers to substrate
X Refers to biomass
Refers to equilibrium
Exercises
8.5 Oxygen Uptake Systems 325
326 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
^ 0.09
7. . L
a'4-
|N ^^
% *%- N**
82:4(10)
-CL2:4(10)
-0.05
0.04
3- -0.03
1. 0
i) 1 2 3 4 5 6 7 8 9 K)
TIME
DJ
2 3 4 5
TIME
Figure 3. Influence of the control on the reactor. The setpoint 82 is 5.0 kg/m^.
8.5 Oxygen Uptake Systems 327
System
The stoichiometry is for the first step YI = 3.5 g O2/ (g NPfy-N) and for the
second step Y2 = 1.1 g O2/(g NO2-N).
Model
Neglecting the details of the biofilm diffusion, the apparent kinetics of this
biofilm process can be approximately described with homogeneous kinetics
that follow a double Monod limitation:
Si CL
*
I + S| KOI +C
S2 CL
= vm2
K 2 +S 2
ForNH 4 + (Si):
328 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
~dT = - f i
For NO2~ (S2):
dS2
dt = ri - r2
For NO3- (S3):
dS3
~3T = r2
For oxygen (CL):
dCL
= - Y i r i - Y 2 r 2 + K L a(C L *-C L )
Program
The program is found on the CD-ROM.
Nomenclature
Symbols
Indices
0 Refers to feed
1,2,3 Refer to reaction steps
O Refers to oxygen
s Refers to substrate
8.5 Oxygen Uptake Systems 329
Exercises
330 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
70
...X'-''
CO 60
$ 50
5) 40
30
20
10
0
1.5
TIME
Figure 2. NH4+, NO2~ and NC>3" and dissolved oxygen in a batch nitrification with KLa = 40 h"1/
2.5
Figure 3. NH 4 + and dissolved oxygen in batch nitrification using three values of KLa from 20
8.5 Oxygen Uptake Systems 331
System
Model
Specific OUR,
qo2m CL
102 - KQ + CL
Oxygen balance,
dCL *
T = K L a(C L *-C L ) -OUR
332 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
C (mg/L)
6
0 20 40 60 80 time(s)
Figure 1. Typical response of the batch oxygen uptake and reaeration experiment.
Measurement dynamics for the liquid film may be important with a viscous
culture,
dCp CL - Cp
TF
and for the electrode lag,
dt =
Program
Experimental data, in the file OXDYNDATA, and the program are found on
the CD-ROM.
Nomenclature
Symbols
Indices
E Refers to electrode
F Refers to liquid film
L Refers to liquid
m Refers to maximum
02,0 Refer to oxygen
S Refers to saturation
* Refers to saturation
Exercises
334 g Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
The results shown in Fig. 2 demonstrate the effect of changing K^a. Runs
varying the electrode time constant TE gave the results of Fig. 3.
70 80 90 100
10 20 30 40 50
TIME
System
VQ.CQ
Hii
V
G> CGO
As shown below, the influence of three quite distinct dynamic processes play a
role in the overall measured oxygen concentration response curve. These are
the processes of the dilution of nitrogen gas with air, the gas-liquid transfer and
the electrode response characteristic, respectively. Whether all of these processes
need to be taken into account when calculating K^a can be determined by
examining the mathematical model and carrying out simulations.
Measurement
CF
Model
The model relationships include the mass balance equations for the gas and
liquid phases and equations representing the measurement dynamics.
Oxygen Balances
The oxygen balance for the well-mixed flowing gas phase is described by
dCG_ _
VG = G (CGO - CG) - KLa (CL* - CL) VL
dt
8.5 Oxygen Uptake Systems 337
= K L a(C L *-C L )V L
dt
The equilibrium oxygen concentration CL* is given by the combination of
Henry's law and the Ideal Gas Law equation where
RT
r
CL * = rCG
and CL* is the oxygen concentration in equilibrium with the gas concentration,
CG- The above equations can be solved in this form as in simulation example
KLAFIT. It is also useful to solve the equations in dimensionless form.
dC F _ C L -C F
dt TF
and
dCg Cp ~Cg
dt " TE
Tp and TG are the time constants for the film and electrode lags, respectively. In
non- viscous water phases Tp can be expected to be very small, and the first lag
equation can, in fact, be ignored.
CG
C = CL =
CGO C GO (RT/H) TG
the component balance equations then become
338 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
u
dt' V ' VG H
and
t1 = 0 ; C' L =C' G =0
C
= L-CF
dt
and
dC C
E = F-CB
dt' TE/^G
where Cp is the dimensionless diffusion film concentration.
C GO (RT/H)
CE
C GO (RT/H)
where oc is the area above the CE versus t response curve, as shown in Fig. 3.
8.5 Oxygen Uptake Systems 339
1.0
CE'
Time (s)
Figure 3. Determination of the area a above the CE' versus time response curve.
Program
The program KLADYN can be used to investigate the influence of the various
experimental parameters on the method, and is formulated in dimensionless
form. The same model, but with dimensions, is used in program KLAFIT and
is particularly useful for determining K^a in fitting experimental data of CE
versus time. A set of experimental data in the text file KLADATA can be used
to experiment with the data fitting features of Madonna. All are on the CD-
ROM.
The program ELECTFIT is used to determine the electrode time constant in
the first-order lag model. The experiment involves bringing CE to zero by first
purging oxygen from the water with nitrogen and then subjecting the electrode
to a step change by plunging it into fully aerated water. The value of the
electrode time constant, TE can be obtained by fitting the model to the set of
experimental data in the file, ELECTDATA. The value found in this
experiment can then be used as a constant in KLAFIT.
Nomenclature
Symbols
Indices
E Refers to electrode
F Refers to film
G Refers to gas phase
L Refers to liquid
Prime denotes dimensionless variables
Exercises
8.5 Oxygen Uptake Systems 341
Results
TIME
Figure 5. A fit of experimental data (open circles) as dimensionless CE versus time (s) to
determine KLA using KLAFIT, which gives 0.136 1/s.
References
Dang, N.D.P., Karrer, D.A. and Dunn, IJ. (1977).Oxygen Transfer Coefficients
by Dynamic Model Moment Analysis, Biotechnol. Bioeng. 19, 853.
Ruchti, G. Dunn, IJ. and Bourne J.R. (1981). Comparison of Dynamic Oxygen
Electrode Methods for the Measurement of KLa, Biotechnol. Bioeng., 13, 277.
System
L, S
r
G, C1T6
Tank6 Transfer
wifsp#K
iS:Tl;niiiS?'
Gas
TriT6 Illllpll
G.Q1T5
'"';:"H? H'MK'SK
*
Transfer
Tanks illlii
Gas
llllll
I
Tank4
Transfer
tlwilffi
Gas I l
lll
lll
I
fankS
Transfer
illlll
i
Gas Illlllll
I
Tank 2
Transfer
^
: : ' : . ; : - : ^ v \ r : : ^ -^ ;.;ji: ' ;' ';':'
^^iiiriS
Gas i;|tii^|;i;||
I
Tank1
Transfer
^^iinS^p:
I
^
Gas TriT1 OliiulllI
L, S1T1
Such columns can be run with a liquid phase flow (bio-trickling filter) or only
with moist packing (biofilter). The work of Deshusses et al. (1995) investigated
the removal of two ketones, methyl isobutyl ketone (MIBK) and methyl ethyl
ketone (MEK), in such a biofilter. The kinetics of this multi-substrate system is
especially interesting since both substances exhibit mutual inhibitory effects on
their rates of degradation.
344 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Model
The model requires stagewise mass balances in the gas and liquid phases for
both components. Transfer takes place from the gas to liquid phases with
reaction in the liquid phase. The symbols used for the concentrations of
substrates 1 and 2 in the n th tank are for the liquid phase Sixn and S2Tn and
for the gas phase CiTn and C2Tn- The reaction rates are RiTn and R2Tn and the
transfer rates are designated TriTn and Tr2Tn-
G, C1Tn, C2Tn L, S1Tn+1, S2Tn+1
m
nth
Figure 2. Single n stage for the biofiltration countercurrent column.
Referring to above figure, the mass balances for a single tank can be written as:
Gas phase
^OL~(G(C2Tn.1-C2Tn)-Tr2Tn
Liquid phase
1Q 1
n =
d ^~ (L(S2Tn+l - S2Tn ) + Tr2Tn -r2Tn VS )
Vs = (1 - EG - e L ) ^E.
For the transfer terms
Vc
Tr
2Tn = K L a ( S 2EQn - S2Tn)~
For the reaction rate terms the following equations are used to describe the
mutual inhibition. Note that oxygen is assumed to be in excess.
Program
Nomenclature
Symbols
Indices
Exercises
Reference
Results
0.0025-
^ 0.002- 0.8 ^
. 0.0015- 0.6
&-
0.001 - ,0.4 W
H
'0.2 (0
0 5e+5 1e+6 1.5e+6 2e+6 2.5e+6 3e+6 3.5e+6 4e+6 4.5e+6 5e+6
TIME
5e-5 1e-4 1.5e-4 2e-4 2.5e-4 3e-4 3.5e-4 4e-4 4.5e-4 5e-4
System
Microtiter plate
Filter
Light
Model
dCT
dt
where CL is the dissolved oxygen concentration, CL* is the saturation value and
OUR is the oxygen uptake rate in mM/min.
The experiment starts with high values of dissolved oxygen concentration,
CL After addition of dithionite OUR increases as calculated by
OUR = ko CL CD
dC D ^ 2 f
dt "" 3
In order to account for some time delay of the sensor a first order equation was
used
dC E ^C L -C E
dt TE
The time constant TE was estimated to be about 1 s.
Program
Two separate programs are given on the CDROM: TITERDYN for the chemical
oxidation with re-aeration and TITERBIO for the biological oxidation and re-
aeration dynamics during a cultivation in a microplate reader. For the program
TITERDYN there is experimental data on the file TITERDYNDATA available
to allow fitting the value of KLa. In TITERBIO KLa during measurement is a
fraction of KLa during shaking and is determined by the parameter kmax. KLa
during measurement is defined as,
KLameasure=KLashaking*(kmax-l)/kmax.
In the original model settings, kmax has a value of 2. The larger kmax, the
larger the error of KLa or OUR estimation.
Nomenclature
Symbols
CD Dithionite concentration mM
CL Oxygen concentration mM
ko Rate constant for dithionite reaction 1/min mM
^a Transfer coefficient 1/s
KQ Saturation constant for oxygen mM
OUR Oxygen uptake rate mM/s
Q Specific oxygen uptake rate mM/ s
TE Time constant for measurement s
Indices
E Refers to electrode
D Refers to dithionite
L Refers to liquid
S and * Refer to saturation
352 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Exercises
8.5 Oxygen Uptake Systems 353
Results
0 100 200 300 400 500 600 700 800 900 1000
TIME
it 'VtV'tj " *
Reference
John, G.T., Klimant, I., Wittmann, C., Heinzle, E. (2003). Integrated Optical
Sensing of Dissolved Oxygen in Microtiter Plates - A Novel Tool for Microbial
Cultivation, Biotechnol. Bioeng., 81, 829-836.
354 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
T 0 ,F
IT*
F,T R
ip
Model
where Q is the delayed heat input from the heater represented by a first order
lag
dt TQ
Program
Nomenclature
Symbols
Indices
C Refers to controller
R Refers to reactor
sens Refers to sensor
set Refers to setpoint
0 Refers to inlet or initial
Exercises
8.6 Controlled Reactors 357
Results
6000
3000
1000
70 80 90 100
1e+4
8000
100.
6000
80 4000
2000
60
0
-2000
-4000
20 40 80 100 120 140 160 180 200
TIME
Figure 3. Response to a step change in the inlet temperature TO at 120 h. The controller
constant Kp was set higher than in the run of Fig. 2.
System
Water
Model
dS -H
dt = Y
^ = Ks+S
where,
TQ = ^XY Q S /Y
For the well-mixed cooling coil, the energy balance equation is:
F = F0 + K P 8 +
Program
As seen on the CD-ROM and below, the control equations can be written in
terms of the error and its integral.
Nomenclature
Symbols
Indices
Exercises
362 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
Figure 2. Cooling flow starts when TR > Tset (25 C); after batch growth finishes at time=4.6 h
the reactor cools. Here Kp=0.6 and TI = 0.6.
8.6 Controlled Reactors 363
Figure 3. With Parameter Plot, the integral of the error and minimum water temperature versus
Kp for a fixed value of Tj=9.
System
F,S0
Feed pump
X,S
Turbidometer
Model
dS M_X
dt" = F(So-S) - -y~
dX FX
dT = -IT
Considering product production with Luedeking-Piret kinetics:
dP FP
dT = -"V" +
8 = (X-X S et)
Program
Nomenclature
Symbols
A Growth-associated constant
B Nongrowth-associated constant 1/h
F Flow rate m3/h
Fo Normal feed flow rate m3/h
Kp Proportional controller constant m6/h kg
KS Saturation constant kg/m3
P Product concentration kg/m3
S Substrate concentration kg/m3
V Reactor volume m3
X Biomass concentration kg/m3
Y Yield coefficient kg/kg
e Error kg/m3
Specific growth rate 1/h
Integral control time constant
Indices
m Refers to maximum
P Refers to proportional control
S and set Refers to setpoint
0 Refers to inlet stream
366 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Exercises
Results
,3.5 c/>
1
5
TIME
45
3.
4
2.5-
-3.5
2" _X:1
-3 C^
X" F:
1.5. " '
-2.5
1 ,m -2
n e
\
VJ.O
r i T... -1.5
System
F,S,
Feed
pump X,S
' Substrate
measurement
Controller
Model
The biomass and substrate mass balances are the same as in the previous
TURBCON model.
Kinetics:
Biomass balance,
V = VX-FX
dt
or,
f
where D is the dilution rate (= F/V). Thus steady-state behaviour, where dX/dt
= 0, is represented by the conditions that |u = D.
Substrate mass balance,
dt
or,
f "><*>--f
8.6 Controlled Reactors 369
where Y is the yield factor for biomass from substrate. Also from this equation
at steady state, since (j = D and dS/dt = 0, the steady-state cell concentration is
given by
X = Y(S 0 -S)
A continuous inhibition culture will often lead to two possible steady states, as
defined by the steady-state condition (a = D, as shown in Fig. 2.
Control equations:
=
Sset- S
Kp r
F = F0 + KP e + I edt
>
Program
When the system equations are solved dynamically, one of two distinct steady-
state solutions is obtained, i.e., the reactor passes through an initial transient but
then ends up under steady-state conditions either at the stable operating
condition, or at the washout condition, for which X=0. The initial
concentrations for the reactor will influence the final steady state obtained. A PI
controller has been added to the program, and it can be used to control a
substrate setpoint below Smax. The controller can be turned on setting by
Kp>0. The control constants Kp, and the time delay tp can be adjusted by the
use of sliders to obtain the best results. Appropriate values of control constants
might be found in the range 0.1 to 10 for Kp and 0.1 to 10 for TJ. Note that
the control does not pass Smax even though the setpoint may be above Smax.
Another feature of the controller is a time delay function to remove chatter.
The program comments on the CD-ROM should be consulted for full details.
Nomenclature
Symbols
Indices
0 Refers to inlet
I Refers to initial value
m, max Refers to maximum
Exercises
8.7 Diffusion Systems 371
Results
0.25
0.2
1.5
0.15
TIME
References
System
Results
0.25
0.2
1.5
0.15
TIME
References
System
the matrix is divided into segments, and the diffusion flux, j, from segment to
segment, is expressed in terms of the concentration difference driving force.
Liquid
j n-1 jn
n-1 ^_ n ^ n+1
^
Model
dSn
' F = J n - l A - j n A +r S n A A Z
Using Pick's law,
S n -l ~ Sn
n-1 =
AZ
and dividing by A AZ,
8.7 Diffusion Systems 373
Thus N dynamic equations are obtained for each component at each position,
one for each element. The boundary conditions are for the above case dS/dZ =
0 at Z = L and S = So at Z = 0. The equations for the first and last elements
must be written accordingly.
The kinetics used here consider carbon-substrate inhibition and oxygen
limitation. Thus,
S O
At steady state, the overall reaction rate or consumption rate of substrate can be
calculated from the gradient at the outer surface. To find the resulting change
of bulk concentration, the liquid phase can be coupled with suitable mass
balances. For a well-mixed, continuous-flow, liquid the resulting balance
equation would be
dS0 F SQ-SI
So) - a DS ^Z
dO0
= K L a(Os-0 Q ) - a Do AZ
Program
As seen below, the program on the CD-ROM uses the array-vector form which
permits plotting the values at time=Stoptime versus the distance index. Also the
number of finite-difference elements N can be varied.
Nomenclature
Symbols
Indices
Exercises
8.7 Diffusion Systems 375
376 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
Figure 2. Oxygen and substrate time profiles for a step change in KLA.
1.4-
0 1 2 3 4 5 6 7 8 10
Figure 3. Oxygen and substrate distance profiles at the end of the run in Fig. 2.
8.7 Diffusion Systems 377
Figure 4. Dynamic response of oxygen and substrate mid-points caused by a step change in
KLA (as Fig. 2) followed at 3 h by a step reduction in Sfeed.
System
Biocatalyst Matrix
Substrate diffusion
diffusion
X = L -* X = 0 ^ X=L
Model
dX
A quasi-homogeneous form for the reaction term is assumed.
The boundary conditions are given by:
At X = L:
S = S0 , P = P0
At X = 0:
dX dX
8.7 Diffusion Systems 379
dS dP
=
SdX -PdX
which on integration gives
DS
P = (So-S)
S< P
=^ ' ' =^ and X' =
gives
d^' L2R'
dX'2 DSS0 ~
where,
kES'
R' =
(K M /S 0 )(1 +(S 0 F/K I ))
and,
P = (1 - S')
X' = 0 dS'/dX' = 0
DS Sp (dSVdX')x=l
=
^ L2R0
380 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
kES0
~K M (i +P O /K I ) + SO
Program
The dimensionless model equations are used in the program on the CD-ROM.
Since only two boundary conditions are known, i.e., S at X = L and dSVdX' at
X' = 0, the problem is of a split-boundary type and therefore requires a trial
and error method of solution. Since the gradients are symmetrical, as shown in
Fig. 1, only one-half of the slab must be considered. Thus starting at the mid-
point of the slab at X1 = 0, where dSVdX' = 0, an initial value for S1 is assumed
(SGUESS). After integrating twice, the computed value of S is compared with
the known value of SQ at X' = 1. A revised guess for S' at Xf = 1 is then made.
This is repeated until convergence is achieved.
Nomenclature
Symbols
Indices
I Refers to inhibition
M Refers to Michaelis-Menten
8.7 Diffusion Systems 381
P Refers to product
S Refers to substrate
!
Refers to dimensionless variables
0 Refers to bulk concentration
GUESS Refers to assumed value
Exercises
382 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
1.86
1.84
Q. 1.82
1.8
1.78
1.76
1.74
System
Model
With complex kinetics a steady state split boundary problem of the type of
Example ENZSPLIT may not converge satisfactorily, and the problem may be
reformulated in the more natural dynamical form. Expressed in dynamic
terms, the model relations become,
3S
dt =
ap a2p +R
dX~dX
384 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Outside Center
S2
S3
S4
Using finite differencing techniques (refer to Sec. 6.2.1), these relations may be
expressed in semi-dimensionless form for any given element n by
f 1
'P
dP'n r
n+l OP
^r n _i_ P'
"* r n+ A R'n
ar AX' 2 ' + S
I
where
D S S 0 1-S
L 2 R n AX'
and
S'n = Sn/SI; Fn = Pn/Si andAX' = AX/L
(a) the effectiveness factor based on the ratio of actual rate to maximum rate
(here for eight segments).
R R
Ri + R29 + Ra3 + RA4 + RS5 + R*
-J 6 + 77 + a8.
R
o
(b) an estimate of the slope of the substrate concentration at the solid surface
o I c)
_ D OQ 1~ O1
^"L^RO AX'
Where the rate at the bulk conditions is
kES 0
+ P 0 /K I ) + S0
Program
Nomenclature
Symbols
AX Increment of length m
r|2 Effectiveness factor based on rates -
386 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Indices
refers to segment n
Exercises
References
Results
0.9
* 0.8
B?' 0.7
0.6
81:1
,.. 82:1
.. 83:1
tf"
tfO.3
. 84:1
- 85:1
_ _S6:1
, 87:1
_S8:1
W0.2
</)
0.1
0
30 50 80
TIME
P5:1
-- P6:1
-ST-J7U.
_ -P8:1
, P3:1
-P4:1
10 20 30 50 60 70 80 90 100
TIME
System
Product N
I/I
Substrate AV
Rp
Rp
Diffusion and reaction takes place within a spherical bead of volume = 4/37cRp3
and area =47iRp2. It is of interest to find the penetration distance of oxygen for
given specific activities and bead diameters. As shown, the system is modelled
by dividing the bead into shell-like segments of equal thickness. The problem
is equivalent to dividing a rectangular solid into segments, except that here the
volumes and areas are a function of the radial position. Thus each shell has a
volume of 4/3 n (rn3 - rn_i3). The outside area of the nth shell segment is 4n rn2
and its inside area is 4n r n _i 2 .
8.7 Diffusion Systems 389
in
Figure 2. The diffusion fluxes entering and leaving the spherical shell with outside radius rn
and inside radius r n _j.
Model
Ar
S
n~ S n-l
jn-l =
Ar
Substitution gives
dSn 3D
dt "'
The balance for the central increment 1 (solid sphere not a shell) is
4 3 dS = J.. ,, 22 4
4 3
= -
-
dS 3D
l _ S /q o N , p
I f - ^ S 2- S l) +R sl
r
S
RSn = - (
where X is the biomass concentration (cell number/m3) in the bead, OUR is the
specific oxygen uptake rate (mol/cell s) and Sn is the oxygen concentration
(mol/m3) in shell n.
Program
{Shells 2 to Array-1}
a/at (S [2. . (Array-1) ] )=3*D*( ((r[i]**2)*(S[i-l]-
/(deltar*( (r[i]**3)-(r[i+l]**3))
Nomenclature
Symbols
Indices
1 Refers to segment 1
2 Refers to segment 2
n Refers to segment n
P Refers to particle
S Refers to substrate
Exercises
392 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
0 10 20 30 40 50 60 70 80 90 100
Figure 4. Doubling the bead radius causes oxygen deficiency inside the bead (lower curve) as
these radial profiles show.
Reference
Both steps are influenced by dissolved oxygen and the corresponding substrate
concentration and are catalyzed by two different organism species. Since their
growth rates are very low, nitrification as a wastewater treatment process benefits
greatly from biomass retention.
In this example, a biofilm column reactor for nitrification is modelled as three
tanks-in-series with a recycle loop (Fig. 1). Oxygen is supplied only in an
external contactor and circulates to the reaction column in dissolved form.
This is similar to the example NITBED. However, in this case the reaction takes
place within an immobilized biofilm, similar to the single tank example
BIOFILM.
394 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
O3, Si3
Fluidized
bed
SiA
Model
To tank n+1
t
S2ntO]
S3ntO]
On[0]
Because of the complexity involving four components in two phases and four
regions care must be taken with the nomenclature. The nitrogen compounds are
referred to as Si, 82, and 83, respectively. Dissolved oxygen is referred to as O.
Referring to the above figure, a single tank n is shown with the four
components. [0] refers to the liquid phase in contact with the solid. Transfer to
the biofilm is by diffusion to the first section, denoted [1].
Further diffusion brings substrate to all the biofilm sections, as shown above, for
a single substrate in section i. The reactions occur in these sections.
In the absorber, oxygen is transferred from the air to the liquid phase.
Additional subscripts, as seen in Fig. 1, identify the feed (F), recycle (R) and the
flows to and from the tanks 1, 2 and 3, each with volume V, and the absorption
tank with volume VA-
The fluidised bed reactor is modelled by considering the component balances
for the three nitrogen components (i) and also for dissolved oxygen. For each
stage n, the liquid phase component balance equations have the form
dSin[0] =
dt
396 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
-J0[0]
For the absorption tank, the balance for the nitrogen containing components
include the input and output of the additional feed and effluent streams, giving
ds
iA _(S/QiF -SQiA )\
-
The oxygen balance in the absorption tank must account for mass transfer from
the air, but neglects the low rates of oxygen supply and removal by the
convective streams. This gives
For the first and second biological nitrification rate steps, the reaction kinetics
for any stage n are given by
v
= ml Slni ni
Kl+ S lni K O i + O ni
V
rs2n = m2 S2ni
K
2+ S 2ni K
02+ni
The oxygen uptake rate is related to the above reaction rates by means of the
constant yield coefficients, YI and Y2, according to
i -r S 2niY2
The reaction stoichiometry provides the yield coefficient for the first step
Nomenclature
Symbols
Indices
Exercises
Program
Results
'4.5
*C
4 3
A
I 60 ' .3.5 g
j,g 5 0 - 13
a?
3
40- 2.5
.2 S
1.5
520.
1
55 io- 5
o-
50 60 70 80 90 100
TIME
Figure 4. Time profiles of the nitrogen component bulk concentrations in the first tank and
the oxygen bulk concentrations in the three tanks.
4.5
4
3.5
0.5
0
0 10 20 30 40 50 60 70 80 90 100
TIME
Figure 5. Time profiles of the oxygen concentrations within the 10 segments of biofilm in
the first tank.
400 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
XA XB
- The overall individual growth rate of each species at any time is the sum of
the rate of growth on glucose plus the rate on citrate.
- The specific growth rate on each substrate depends on the concentration
level of some key enzyme responsible for the rate-controlling step E.
- The key enzyme for the preferred substrate is assumed to be constitutive.
- The production of the key enzyme for the secondary substrate is subject to
induction and repression by the preferred substrate.
- An inhibitor I is produced from the growth of K. oxytoca on glucose and
inhibits the growth of P. aeruginosa on citrate. The inhibitor is thus a
growth-associated product.
Model
The growth rates, jny, for each organism are the sums of the growth rates on
glucose and citrate. The subscripts i and j have the following meaning: i refers
to the organisms (K. oxytoca - A and P. aeruginosa = B) and j refers to the
substrate (glucose = Y and citrate = Z). The levels of the key enzymes are
denoted by E.
dXA =
(MAY + MAZ) XA
dXB
I
The specific growth rate equations for the two organisms on each substrate are
given by:
A*maxAYSYEAY
K
SAY + S Y
MmaxAZSZEAZ
K
SAZ +S Z
A*maxBZSZEBZ I K
I
M-BZ - K--
S K"FT
SBZ+ Z V I +
dS a
Y = !
Y Y Y
SAY I ) SBY
dSz
XB XA
dt - - YSBZ " YSAZ
Inhibitor (I) production is growth associated to organism A, and its decay is
proportional to the cell concentration. The balance for the inhibitor is
dl
gj- = oc|u AY X A -pX A
The balances for the key enzymes, which control the growth on secondary
substrates are given by:
Sz KRAZ
-
T-T - --
" *^--rt\iu
kpAZf\z^
EAZ
and
dE
BY SY KRBY
- " --Y KbY
OUR =
OAY OAZ
OBY OBZ
CER =
Y Y
CAY CAZ
CBY CBZ
8.8 Multi-Organism Systems 403
OUR
DOT = 100 1-
K L aC 0
FA=-
X A + XB
FB = 1 - F A
Program
Nomenclature
Symbols
Indices
A Refers to K. oxytoca
B Refers to P. aeruginosa
C Refers to carbon dioxide
I Refers to inhibitor
M Refers to maximum
O Refers to oxygen
P Refers to dilution due to cell division
R Refers to repression
S Refers to substrate
Y Refers to glucose
Z Refers to citrate
Exercises
8.8 Multi-Organism Systems 405
Results
The graphical results in Fig. 2 show the dynamic changes in biomass fractions
FA and FB for two values of a: 0.007 kg/kg and 0.0007 kg/kg .
0.9.
0.8
0.7
0.6
e
s-0.45
0.3
0.2
0.1
0
4 5
TIME
Figure 2. Dynamic changes in biomass fractions FA and FB for a = 0.007 and 0.0007.
Reference
This example was developed from the original paper by J. Lang, ETH-Zurich.
406 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
The interactions between two microbial species (Ma and Mb) in a mixed
continuous culture are considered (Miura et al., 1980). The population
dynamics of the two microbes, is described by competitive assimilation of
substrate Si and commensalism, with the participation of growth factor Ga that
is excreted by microbe Ma and required by microbe Mb for growth. Mb also
consumes a second substrate 82 from the medium. These interactions are
represented in the Fig. 1.
,G
Model
For the chemostat shown above the unsteady-state material balances are as
follows:
Dilution rate:
8.8 Multi-Organism Systems 407
Organism Ma:
dXa
j- = ftia-D)Xa
Organism Mb:
dXb
-gj- = (jib - D) Xb
ar = - "a
dS2 M-b Xb + D (S2
-ar = - ~YT" "S2)
The yields for organism Mb on the two substrates are assumed here, for
simplicity, to have the same values, Yb.
dGaa "P 11 Y
Jib Xb T-X f*
The mass balance for the growth factor, Ga, is formulated by assuming a
formation rate, Pa |ia Xa, and consumption rate, (|Lib Xb)/Ybg. Here Xa and Xb
are the concentrations of microorganisms Ma and Mb, respectively. The
constants Pa and Ybg are the biological yield constants for the formation and
consumption of Ga, respectively. The specific growth rates of microbes Ma and
M b are expressed by:
Organism Ma:
Si
KSa+Si
Organism Mb:
Si Ga
82
K Sb 2 + S2 Kg + Ga
408 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
where K$a and K$bi are the saturation constants of Ma and Mb for substrate Si,
Ksb2 is the saturation constant of Mb for substrate 82, and Kg is the saturation
constant of Mb for growth factor Ga. Setting \imb2 = 0 corresponds to the
consumption of only one substrate Si.
A rigorous stability analysis of the system has been carried out by Miura et.
al. (1980). This involves linearizing the mass balances by Taylor's method in
the vicinity of the steady state solution and determining the characteristic
eigenvalues of the resultant matrix. The following relationship for co-existence
of the two microbes can be derived for the case of a single substrate.
Also, a critical dilution rate, where the maximum dilution rates of the two
organisms cross-over can be written:
.
Cnt _
"
Sa Sbl
Four particular cases depending on the values of the maximum specific growth
rate and saturation constants of both microbes can be simulated for the single
substrate case (|imb2 = 0).
1- M-ma > l^mbli Ksa > K$bi: Coexistence below a certain value of D
2. |ima > Mmbi; Ksa < KSbi: No coexistence range
3. |ima < |imbi; Ksa > Ksbi: Coexistence with wider range of stable focus
K$a > Coexistence at higher D and wider range of
Program
Nomenclature
Symbols
Indices
0 Refers to feed
1 Refers to substrate 1
2 Refers to substrate 2
a Refers to organism a
b Refers to organism b
bl Refers to organism b growing on substrate 1
b2 Refers to organism b growing on substrate 2
g Refers to growth factor
m Refers to maximum
Exercises
410 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
A simulation is given in Fig. 2 for the parameters as given in the program with
feed flow rate F = 0.24 m3/h, and the feed concentrations SIQ = 500 mg/L and
820 = 1500 mg/L. The oscillating concentrations are given for Xa and S\
versus time. It is seen that this solution is stable and homes into a steady state,
corresponding to case 1.
250
200
150 CD
100
300
Figure 2. Competition and commensalism of two organisms (F = 0.24 m3/h, S10 = 500 mg/L
and S2o = 1500 mg/L), showing the biomass concentrations.
8.8 Multi-Organism Systems 411
Reference
System
S.XLX2
Model
dx2
2 - D X2
ds
g^ = D (SQ - s) - (Lli (s) Xi - |I2 (s) X2
"SIT
Z2)
8.8 Multi-Organism Systems 413
Thus:
dz
= P(S-ZI)
where p = ai/a2, and cxi and a2 are the adaptability factors or inverse time
constants. The effect of (3 is to delay the substrate for growth in the wild and
engineered organisms according to their first order time constants. For p > 1 the
wild type is delayed with a shorter time constant. At high values of oci and OC2,
the model describes an undelayed, instantaneous Monod growth model. It is
assumed, that (li^ > H2m .
The probability factor p represents the probability (or fraction) of
conversion to the wild strain during growth of the engineered strain. Thus the
growth rate of the engineered strain is multiplied by [1 - p].
Program
Nomenclature
Symbols
Indices
Exercises
Results
The output in Fig. 2 gives the substrate oscillations created by the square wave
feed concentrations, showing the engineered organism X2 being washed out. A
similar situation for sine wave feeding is given in Fig. 3.
416 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Figure 3. Sine wave feeding. Similar to Fig. 2 but with longer period and a higher b value.
References
Aiba, S., Imanaka, T. (1981) in Annals of the New York Acad. of Sciences, 369,
1-15.
8.8 Multi-Organism Systems 417
System
Model
The model involves the chemostat balances for each species with the
corresponding kinetics. The variables are given in Fig. 2.
418 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
S0 D Sl,X-|,X2
^*
1
II
Substrate balance,
dS
= D(So-Si) -
Species 1 (prey) balance,
^2X2
dX2
- DX 2
~"v
The kinetics are given by Monod relations,
Program
Nomenclature
Symbols
Indices
Exercises
420 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
Stable steady states for the system are shown in Fig. 3 for |LLmi = 0.5 and
|Lim2 = 0.11. Oscillations in the biomass populations are achieved by setting the
specific growth rates nearly equal (\im\ = 0.5 and |Lim2 = 0.49) as shown in
Fig. 4 and also by the phase plane plot of Fig. 5.
3.5
2.5
1.5
0.5
50 100 150 200 250 300 350 400 450 500
TIME
450
System
Consider organism A and organism B with their respective specific growth rates,
(HA and ILLB, which both grow independently on substrate S. Assume:
S/(KSA + S)
S/(KSB + S)
Depending on the values of |LIM and K$, these two functions may occur in two
different forms, as shown in Fig. 1.
B M
inter
inter
Figure 1. Comparison of growth rate curves for the competitive chemostat growth.
It is clear that the curves B and A will cross each other if (IMB < MMA and KSB
< KSA- In Fig. 1, the situation on the left indicates that B will grow fastest at
any value of S. For this case, in chemostat cultures with dilution rate DI, after
an initial start up period, a substrate concentration S i will be reached at which
(LIB = DI and for which |LIA < DI. Organism A will then be washed out, and
only organism B will remain in the reactor.
The situation in the right of Fig. 1 shows (IB crossing (IA- The point of
intersection can be found easily by simple algebra where:
8.8 Multi-Organism Systems 423
Wnter =
=
Sinter
For this case a chemostat can theoretically operate stably at D = Jiinter such
that both A and B will coexist in the reactor. This however is an unrealistic
metastable condition, and with D < Hunter* A will wash out. With D > Hunter A
will grow faster, causing B to be washed out.
Model
The equations for the operation of chemostat with this competitive situation are,
d*A ,
dXB
jp = 0 - D XB + JIB XB
dT = D(S0 - S) -
In addition, the Monod relations, |IA = f(S) and |LLB = f(S), are required.
Solution of these equations will simulate the approach to steady state of A and
B competing for a single substrate.
Program
Nomenclature
Symbols
Indices
A Refers to organism A
B Refers to organism B
M Refers to maximum
0 Refers to inlet stream
inter Refers to the intersection of the ju versus S curves
Exercises
8.8 Multi-Organism Systems 425
Results
4.5
3.5
3
!
2.5 '
1.5
0.5
9 10
System
Wastewater with toxic chemicals is often treated directly at the source with
specialized microbial cultures in small-scale biofilm reactors. A model may
help to understand, optimize, and control such reactors. In a paper by Soda et
al. a simple biofilm model was developed to simulate the competition between
two microorganisms for a common inhibitory substrate. In the model the
following assumptions were made: (i) the biofilm has a uniform thickness and
is composed of 5 segments, (ii) each microorganism A and B utilizes a
common substrate, and the growth rates are expressed by Haldane kinetics with
a spatial limitation term but is otherwise independent of the other
microorganism and (iii) the diffusion of the substrate, movement of the
microorganisms, and continuous loss of the biomass by shearing are expressed
by Pick's law-type equations.
8.8 Multi-Organism Systems 425
Results
4.5
3.5
3
!
2.5 '
1.5
0.5
9 10
System
Wastewater with toxic chemicals is often treated directly at the source with
specialized microbial cultures in small-scale biofilm reactors. A model may
help to understand, optimize, and control such reactors. In a paper by Soda et
al. a simple biofilm model was developed to simulate the competition between
two microorganisms for a common inhibitory substrate. In the model the
following assumptions were made: (i) the biofilm has a uniform thickness and
is composed of 5 segments, (ii) each microorganism A and B utilizes a
common substrate, and the growth rates are expressed by Haldane kinetics with
a spatial limitation term but is otherwise independent of the other
microorganism and (iii) the diffusion of the substrate, movement of the
microorganisms, and continuous loss of the biomass by shearing are expressed
by Pick's law-type equations.
Sf S0
Wall
Figure 1. Schematic of the continuous reactor with biofilm, showing the descretization into
five layers.
Model
Fig. 1 illustrates an idealized flat biofilm with a uniform thickness Lf (m). The
biofilm is divided into N segments for simulation purposes and each has a
thickness of AZ = Lf/N (m). Wastewater containing the substrate is fed to the
reactor at a constant feed rate and a concentration Sf (mg/L). The bulk liquid in
the reactor is mixed throughout the tank and the substrate diffuses into the
biofilm. The substrate is transported from the bulk liquid having a
concentration S[0] (mg/L) to the surface of the biofilm having a concentration
S[l] (mg/L). A diffusion layer of a thickness Lj (m) is used to represent the
external mass transport resistance.
Using the same approach as in the example BIOFILM, the mass balances in
the bulk liquid for the substrate and microorganisms A and B with a continuous
flow are simply described as following:
8.8 Multi-Organism Systems 427
, XA[0]-XA[1] , }
- -DXA|0] - aDXA (mALO] - bA JXA[0]
/\/1
Reactions within the biofilm are described by diffusion reaction equations. The
mass balances of the surface segment are described as following:
= s[0]-sm
S S
dt LjAZ AZ2
Component mass balances are written for each segment (i = 2, .., N-l), where:
428 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
The mass balances of the boundary segment on the support wall are described
by the following equations:
dS[N] = p S[N-1]-S[N]
S
dt AZ2
.. m_
K IA
[i]=
K IB
where KI? Ks, and (im are inhibition constant (mg/L), half saturation constant
(mg/L), and maximum specific growth rate (day" ). Xm (mg/L) is the maximum
capacity of total biomass of microorganisms A and B in a segment.
The formulation of the spatial limitation term used here is the most simple
one possible with non-restricted growth at zero biomass concentration and zero
growth at maximal biomass concentration Xm.
Applying the above model it was found (Soda et al., 1999) that the
qualitative behavior of the biofilm reactor is characterized by 5 regions,
depending on the operating conditions, the substrate concentration in the feed
and the dilution rate. In region I, both microorganisms are washed out of the
biofilm reactor. In region II, microorganism B is washed out, and in region III,
microorganism A is washed out of the biofilm. In region IV, both
microorganisms coexist with one another. In region V, both microorganisms
coexist with a sustained oscillatory behavior. Convergence to regions I-V
depends on the initial conditions. In regions II-V, washout of either or both
microorganisms is also observed when the initial conditions are too far away.
Nomenclature
Symbols
Indices
A refers to microorganism A
B refers to microorganism B
Numbers in brackets
Exercises
8.8 Multi-Organism Systems 431
References
Program
Shown below is a portion of the program. The full program is on the CD-ROM.
Results
-0.03
\,
2 40-
-0.025
m /
X
0.02
\ XAmid:1
(A
. . XBmid:1
I Smid:1 -0.015
x 20 i
\ -0.01
g.
-0.005
---. '--.
10 20 30 40 50 60 70 80 90 100
TIME
0.045
0.04
0.035
0.03
0.025
(A
0.02
0.015
0.01
0.005
10 20 30 40 50 60 70 80 90 100
System
The model was developed to aid the design of the measurement system and in
the interpretation of the data.
Model
M-imax Si
Mi = KSi + Si
M^maxi Si
The individual equations for each substrate Si are given in the program,
Thermodynamic equilibrium constraints on the Step 3 reactions (Table 1)
are also included in the model.
8.8 Multi-Organism Systems 435
Reaction Equilibrium
In the acetogenic step (Step 3 reaction in Table 1), acetic acid, hydrogen and
carbon dioxide are produced from propionic and butyric acid. The
thermodynamic equilibria for these reactions are incorporated by estimating
the chemical equilibrium limits for butyric acid:
From this the equilibrium constant is KBU = 2.02 x 10~3 (mol4 nr12) given by
KBU
"
For propionic acid similarly,
The factor 4/3 is necessary because concentrations here are given in C-mol.
An empirical approach was chosen to slow the reactions down on
approaching the equilibrium, and they were not allowed to proceed to the right
side when the equilibrium condition was reached. Using the actual
concentrations, the parameters KBU* and Kpro* were estimated.
* C A c 2 C H2 2 CH +
KBU
~~ CBu
The ratio of these values to the true equilibrium constant, K*BU/KBU, and
K*pro/Kpro will be greater than unity if the equilibrium has not yet been
reached. Using these ratios with the empirical S-shaped curve of Fig. 1, the
factor FEQ was determined and was used to modify the growth rates. This
somewhat arbitrary function starts from FEQ = 0 at K*/K < 1 and rises to FEQ
= 1 at K*/K > 2. The factor FEQ causes the reaction to the right to stop when
436 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
a 0.6-
uj
u_
0.4-
0.2.
0.
-0.2-
K*/K
Figure 1. Equilibrium factors (FEQ) to slow the growth rates near equilibrium.
The kinetics of biomass growth butyric acid, and propionic acid were modified
by these empirical equilibrium factors, FEQ, according to
i =FEQi
I + CH+
Thus CH+ is varied iteratively until 8 becomes essentially zero. This numerical
solution is not always trivial using conventional methods for non-linear
8.8 Multi-Organism Systems 437
Program
(PH>
GUESS CHPLUS = le-4
ROOTS CHPLUS =
KW/CHPLUS+KdBu/(KdBu+CHPLUS)*BU/4+KdPr/(KdPr+CHPLUS)
*Pr/3 +KdAc/(KdAc+CHPLUS)*Ac/2 +KdC/(KdC+CHPLUS)*Cg
+KdBuf/(KdBuf+CHPLUS)*BUFFER-lonen-CHPLUS
Nomenclature
Symbols
C Concentration C-mol/m3
Cons Consumption rate C-mol /m3h
F Stoichiometric coefficients (-)
FEQ Equilibrium factor (-)
Ftitr Titration flow rate m3/h
438 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Indices
Results
The first of the three graphs in Fig. 2 shows dynamic profiles of substrate whey
(Mo), CH4 (CH), dissolved CO2 (CO) and dissolved hydrogen (Hy). The whey
is almost instantaneously consumed. Hy reaches a maximum very soon and is
then quickly reduced to almost zero. CH4 is produced with varying rates. CC>2
reaches a maximum, which is partly caused by pH changes and by
consumption by hydrogen-consuming organisms. The peaks in the CC>2 curve
originate from numerical inaccuracies in the stiff system. In the second graph,
Fig. 3, the total concentration of volatile acids acetate (Ac), propionate (Pr) and
butyrate (Bu) are given. The thermodynamic inhibition of acetogenesis is
clearly seen in the early phase of the experiment. Ac reaches a maximum much
later than Pr and Bu, since it is produced from these two acids. In the third
graph, Fig. 4, the pH versus time profile is given, exhibiting an early decrease,
followed by almost constant pH during the rest of the simulation.
8.8 Multi-Organism Systems 439
0.05 -0.25
0.04 -0.2
0.02 -0.1
0.01 0.05
Figure 2. Dynamic profiles of substrate whey (Mo), CH4 (CH), dissolved CO2 (CO) and
dissolved hydrogen (Hy). Zoomed to show the early period.
Run 1:4389 steps in 1.43 seconds
0.3
0.25
r 0.2
0.15 -
-0.05
0.35
Figure 3. Total concentration of volatile acids acetate (Ac), propionate (Pr, lower curve) and
butyrate (Bu). The whey (Mo) peak is hardly visible at T = 0.
440 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
6.16 0.18
6.14 0.16
6.12 .0.14
6.1 .0.12
: 6.08 0.1
6.06 0.08
6.04 0.06
..'
6.02 0.04
6 0.02
5.98 0
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Exercise
References
Heinzle, E., Dunn, I.J. and Ryhiner, G. (1993) "Modelling and Control for
Anaerobic Wastewater Treatment." Adv. Biochem. Eng. 48, 79-114.
System
Model
The balance equations for continuous culture with dilution rate D are as
follows:
dR
dE
= - D E + [ rGE(R,G,E) + rSE(S) - rEX(E) ] R
dG
= - D G + [ rSG(S,E) - rGE(R,G,E) ] R
dT
= - D T + [ rTi(R,G,E) - rT2(R,G,E) ] R
The species in parenthesis indicate the dependencies of the rates. The kinetic
expressions used in the balance equations are as follows:
r
GEm E
/KGX\n
l /
S
Ks + S
ME)
TEX = -
1 + ( KG/G + KET/E )n
rT2 =
Nomenclature
Symbols
Indices
E Refers to ethanol
G Refers to storage material
m Refers to maximum
S Refers to glucose
T Refers to enzyme
X Refers to biomass
Exercise
444 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
The influence of dilution rate is given below in plots from simulations: Fig. 2
with D = 0.05 and Fig. 3 with D = 0.1. In Fig. 4 the phase plane from the run
of Fig. 2 is shown.
0.4
30-
0.35
25-
0.3
20- 0.25
(
0.2
15-
-0.15
10
0.1
5-
0-
M..JL.JL...
20 40 60
0.05
-0
100 120 140 160 180 200
TIME
1.8
-1.6
1.4
1.2
-1 </>
0.8
.0.6
.0.4
0.2
0.08-
0.07-
0.06-
0.05-
i
0.04-
0.03-
0.02-
0.01 -
0-
12 16 18 20 22 24 26 28 30
X
Figure 4. Phase plane giving S versus X from the run of Fig. 3. Zoomed in for detail.
Reference
Heinzle, E., Dunn, I.J., Furukawa, K. and Tanner, R.D. (1983). Modelling of
sustained oscillations observed in continuous culture of Saccharomyces
cerevisiae. in Modelling and Control of Biotechnical Processes (ed. A.Halme),
Pergamon Press, London, p.57.
System
Rb+E2F Rb-P+E2F
Model
= V2 -Vl
dt ~
dK
dKr
dt
8.8 Multi-Organism Systems 447
dRE
dt
^=v 6 > r -v 6 , f
dk
drp
-= Va3 VA
4
di
k + k p + k p j =1
r + rp + r^ = 1
448 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
=1
i + A,kPI=l
re
0I= JPJL
kpi
Nomenclature
c Cyclin E concentration
e E2F concentration
i Concentration of cyclin E-c
k ckd2 concentration
kP Phosphorylated cyclin E-cdk2 complex
concentration
kP,I Concentration of phosphorylated cyclin
E-cdk2 complex bound to inhibitor
r Concentration of hypophosphorylated form
of pRb
rE Concentration of hypophosphorylated form
of pRb that binds to E2F
rP Concentration of hyperphosphorylated form
of pRb
g Ratio of total concentrations of cdk2 and cyclin E
s Ratio of total concentrations of pRb and E2F
1 Ratio of total concentrations of cdk2 and inhibitor
t Dimensionless time
8.8 Multi-Organism Systems 449
Exercises
Program
Results
Figure 2. Profiles of concentrations k, rp and c versus time as obtained from the rate constants
in the program.
450 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
0.32 0.33 0.34^ 0.35 0.36 0.37 0.38 0.39 0.4 0.41 0.42
Reference
System
Consider a reactor whose outlet stream passes through a membrane that retains
only the biomass as seen in Fig. 1.
The growth is assumed to follow substrate inhibition kinetics with constant
yields. The oxygen transfer rate influences the growth at high cell density
according to a Monod function for oxygen.
F,S 0
i Gas
illiiil
Air
Model
Biomass balance:
dX
dT = rx
Substrate balance:
dS F
-r s
dCL
-gj- =K L a(CLS-C L )-ro
Kinetics:
S __ CL
rx
~ ^m x
r
s = rx YS/X + MS X
r = r
o x YQ/X + MO X
MS _ YS/X
=
M0
Program
Nomenclature
Symbols
Indices
0 Refers to feed
1 Refers to reaction 1
m Refers to maximum
O Refers to oxygen
S Refers to substrate
X Refers to biomass
Exercises
454 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
20
\\.
0 2 4 6 10 12 14 16 18 20
TIME
Figure 3. Profiles of dissolved oxygen influenced by KLB (0.5 to 5, curves bottom to top).
System
The metabolic pathways for the production of acetoin and butanediol are well
known, as shown in Fig. 1. A kinetic model for a Bacillus subtilis strain has
been established from continuous culture experiments using an approach
involving overall stoichiometric relationships and energetic considerations. The
influence on the culture of product removal by pervaporation was investigated
by simulation methods.
Complex Compounds
Lactate Butanediol
Model
Growth
The sugar (S) and dissolved oxygen effects were described in terms of a double
Monod function and the product inhibition by a simple inhibition kinetic term.
Diauxic effects were observed, but the diauxic components were unknown, and
for simplification only one diauxic switchover was assumed. The preferred
8.9 Membrane and Cell Retention Reactors 457
where,
and
C2
*pl
Recycle Membrane
Feed
Pervaporation Module
Condenser
Bioreactor
Permeate
Fpe
ACpe , Bu
Pump
1 1
and
Bu_cc Bu
The inhibition constants, kinh,Ac and kinh,BU> an<l the exponents, and
were estimated from shake flask experiments.
The complete expression for specific growth rate was thus,
DO i i
1+ [_^_]Ac 1+ | Bu_iaBu
458 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
The products were assumed to be formed when the respiration was not
sufficient to cover the energy requirement for growth. While the distribution of
the two products was assumed to be dependent on oxygen, through a rapid
equilibrium, according to an S-shaped empirical function described by
Ac fAcBu,max (DO)2
Bu -
To avoid an algebraic loop, the kinetics for the rate of butanediol production
was assumed to be dependent on the deviation from equilibrium,
The constant kAcBu was set high enough to ensure equilibrium conditions.
The specific reaction rate q^3 for product formation was obtained as,
Reaction Rates
The specific reaction rates for each component were obtained from the specific
reaction rates q^ and the stoichiometric coefficients, Vy (component j in
reaction i) as follows,
i
The volumetric rates rj [mol L"1 Ir1] for components X, S, Ac, Bu, 62 and CC>2
were related to the specific rates as,
rj = qj X
V 2 , X MG X
where V2,x is the stoichiometric coefficient, and MGx is the C-mol mass of
biomass.
The rate for the respiration reaction, q^i, was found to be influenced by the
dissolved oxygen, and was described as follows:
qxi =
The yield coefficients for the complex components GI and C2 , were assumed
to be 1 g of each component for 1 g biomass, and the molecular weights of the
components were assumed to be the same as for the biomass. The initial
amounts of these components in the molasses medium were adjusted in the
simulation. The corresponding rates were proportioned according to the
growth rates as
Pervaporation Model
_ FO _ Fpu _
D = y- , Dpu = -^r > Dpe =
where D, Dpu, Dpe were the flows for the feed, purge, and permeate, respectively
intr 1 .
460 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Total mass
D = Dpu + Dpe
Biomass
dx
"3T = rx - X Dpu
Sugar
dS
df = S0 D - S Dpu - rs
Acetoin
dAc
~dT = rAc ~ Ac D
pu ~ Acpe Dpe
Butanediol
dBu
~dt~ = rBu
~ Bu D U
P " BuPe DPe
Complex components
dCi
~dT = Ci>0 D - Ci Dpu - rCi
where Q = 1 and 2.
Oxygen transfer from the gas phase and the oxygen uptake rate determine the
DO in percent saturation as,
dDO 100
-HT~
m
= KLa (100 - DO) ^r OUR
2.34xlO"4
Here the equation is in terms of percent saturation using the DO saturation
value at 30 C of 2.34 x 10''4 mol/ L.
Nomenclature
Symbols
Greek symbols
Indices
Ac Acetoin
Bu Butanediol
Q Component i (unknown components 1 or 2)
CO2 Carbon dioxide
i Refers to reactions, i (1 to 4)
j Refers to components j (Ac, Bu, C>2, CC>2, X and S)
inh Refers to inhibition
02 Oxygen uptake
pe Permeate
pu Purge
P Product
Ri Chemical reaction i (1 to 4)
S Sugar
X Biomass
0 Feed or initial concentration
Ai Refers to reactions 1 to 4
Reference
Dettwiler, B. Dunn I. J., Heinzle E., and Prenosil J. E. "A Simulation Model
for the Continuous Production of Acetoin and Butanediol Using B. subtilis with
Integrated Product Separation by Pervaporation" Biotechnol Bioeng. 41, 791
(1993).
Program
Results
The results of Fig. 3 show the influence of the FM/P ratio, which corresponds
to the membrane area per unit reactor volume, on the biomass concentrations.
8.9 Membrane and Cell Retention Reactors 463
Here the enrichment factor was kept constant (pAc = 2.0), corresponding to the
values found for one of the membranes. In a second set of runs the enrichment
factor was varied for constant FM/P ratio = 1.0.
100
60
50
6 8 10 12 14 16 18 20
0.15
..p.-'*"" j- *~ "*"
0.05
Figure 4. Influence of the enrichment factor on the acetoin and butanediol in the permeate.
(D=l, FMP=1 BETAAC=1, 3, 5).
464 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
System
(1-B2XB1F1+I?)
X 2 ,S 2 ,P 2
Model
The model assumes completely-mixed stages and complete cell separation. The
operating parameters are feed flow rates or dilution rates and the bleed stream
fractions. The bleed stream from the first fermenter is led to the second
fermenter. The model equations are developed neglecting the volume of the
lines and separators. Note that the retentate streams are returned to their
respective reactors and can be considered as part of the well-mixed system.
ii ii S KLP
1 1 . >
2+V2X2
Clt UC (JC (JC
Substrate Conversion
From stage 1
From stage 2
(B 1+ f)S 2
Xco =1 --
82
B^+fSfz
Total glucose conversion
Program
In the simulation product-free feed streams are assumed with flow rates between
10-30 kg m 3 . The program is on the CD-ROM.
Nomenclature
Exercises
8.9 Membrane and Cell Retention Reactors 469
References
Results
^Xt1:1
...Xt2:1 25
20-
... S2:1
20
P2:1
rsj 15-
X
gioH
10 CO
v
1
10 15 20 30 35 40
TIME
Figure 2. Results showing the steady state for biomass substrate and product in both stages.
470 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
140 i 0.9
| 120-I
0.8
100
B
\ -0.7 rf
X 80
-
0.6 tfT
60 X
1
-0.5
flf 40
A.. 0.4
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
B1
Figure 3. Results showing that the productivity changes slightly with bleed ratio but that the
biomass concentration and the substrate conversion depend highly on bleed ratio.
System
This tubular reactor- radial diffusion model assumes a series of nine well-mixed
tanks to describe a single hollow fiber module. Flow of lactose substrate passes
axially through the inner region of the fiber lumen. By diffusion the substrate
is transported radially outward from the lumen through the membrane and into
the cylindrical porous support surrounding the membrane. The reaction takes
place in this support region, where the immobilized enzyme is located. The
products of hydrolysis are glucose and galactose, which diffuse back toward the
liquid phase in the lumen. The parameter N can be used to adjust the number
of shells required. The module is assumed to consist of a large number of
identical fibers*
8.9 Membrane and Cell Retention Reactors 471
Model
FRfiber LAfiber
LAtank
FRfiber
Figure 1. Hollow-fiber module showing only a single fiber as modelled by nine stages.
472 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Fig. 3 shows a cross-section of the hollow fiber membrane showing the inner
hollow fiber region (white) and the outer porous support (shaded). The finite-
difference shell of volume V2 (white) is shown with diffusion fluxes of lactate
JLAI entering and leaving JLAI- It is important to account for the radial
variation of volumes and diffusional areas. Note that the segments are
numbered from outside to inside, 1 to N.
For each tank the component balances account for the accumulation, the flow
in and out and the diffusion in or out from segment N of the porous
membrane. For lactose in tank 1
8.9 Membrane and Cell Retention Reactors 473
LA FR
lumenl _ fiber */ T A T A x , JlAltN]* Aj[N]
H
ai
- v
v
v^^tank ' L/Mumenl ) + 77
v
lumen lumen
For the external recycle tank, the flow leaving all of the fibers enters it and also
the feed stream enters it. The total flow rate leaving must equal the sum of these
rates. Thus for lactate
Taking the component balances for each segment in the enzyme zone account
for accumulation, diffusion in and reaction.
Thus for lactose in the enzyme regions of the first tank
where Aj is the area available to diffusion and TLAI!^] is the reaction rate for
lactose in the ith enzyme segment of the first tank. Note that the above
equations do not include the balance for segment 1. The wall condition
requires that this balance contains only the rate of diffusion from segment 1 to
segment 2, as seen in the program.
*!__ LAt[i]
Here the units are mole lactose per cm3 of porous support volume per second.
Program
Repeated here for the first tank section are the lactose balances, fluxes and rates
as given in the program on the CD-ROM.
d / d t ( L A t a n k ) = ( Number * F R f i b e r / V t a n k ) * (LAlumen9-
L A t a n k ) + ( F f e e d / V t a n k ) * (LAf e e d - L A t a n k )
Dynamic lactose balance for the circulation tank
The geometry of the fiber is programmed such that the lumen radius and the
total fiber radius are given. The number of porous segments can be varied.
Nomenclature
Additional symbols for the geometrical factors are defined in the program on
the CD-ROM
Symbols
?
LA, GA, GL
Kinetic constant for vmax
Length of fiber,
Lactose, galactose and glucose
mol/mgE h cm3
cm
mol/cm3
Number Number of fibers
V Volumes cm
Maximum rate mol/cm3 h
Indices
Exercises
476 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Results
L
9.393e-5
Figure 5. Radial concentration gradients for the lactose and glucose. The left axis corresponds
to the outside of the fiber.
8.9 Membrane and Cell Retention Reactors 477
System
OXYGEN
j (MEASUREMENT
CHAMBER
SAMPLE
PORTM .
THERMOSTAT
VESSEL
j3, FR
Spent medium
Figure 2. Schematic of the model structure, where Ci refers to any component concentration.
For the tanks-in-series description of the column, 3 (or more) tanks in series are
used. The mass balance for one component in tank 2 is then
Here V is the volume of one tank and r is the reaction rate of the component.
The circulation flowrate is FR. This balance equation form would apply to all
components, but not for the biomass since it is immobilized.
The kinetic model assumes the following:
From the data in the dissertation of Keller (1) can be calculated the yield of
lactate with respect to glucose, giving Yiacc = 2.0 mmol lactate/mmole glucose
or 1.1 g sodium lactate/g glucose. This can be used to calculate the production
rate of lactate. Also calculated from the dissertation is YgiutG=:36mg
glutamine/g glucose.
Thus the glucose uptake rate becomes
"OX K lac
w
sox K sglut 4-C lac
8.9 Membrane and Cell Retention Reactors 479
Note that the last term in the equation models an inhibition by lactate.
The growth rates are based on the specific substrate uptake rates. Other specific
rates are related by yield coefficients and biomass concentration.
For example for growth rate,
rx = qG X Yxg
Program
Nomenclature
Exercises
8.9 Membrane and Cell Retention Reactors 481
References
Keller, J. and Dunn, I. J. "A Fluidized Bed Reactor for the Cultivation of
Animal Cells", In: Advances in Bioprocess Engineering, Eds. E. Galindo and O.
T. Ramirez, Kluver, pp 115-122 (1994).
Results
10'
9' 6 O
O
8
? 7
5
O
4 iliii X[3]:1
3
20 40 80 100 140 160
TIME
Figure 3. Batch run showing the difference in dissolved oxygen between the tank sections.
482 8 Simulation Examples of Biological Reaction Processes Using Berkeley Madonna
Figure 4. In this run with F=0.05 L/h first oxygen limitation develops and later glucose
limitation. The immobilized biomass in the three sections increases at different rates due to the
glucose gradient in the column.
Appendix: Using the Berkeley
Madonna Language
Computer Requirements
Two Berkeley Madonna versions are supplied with this book on a CD, one for
PC with Windows and one for the Power Macintosh. More information with
downloads can be found on the following website:
http://www.berkeleymadonna.com
Installation from CD
The files are compressed on the CD in the same form as they are available on
Internet. Information on registering Madonna is contained in the files.
Registration is optional since all the examples in the book can be run with the
unregistered version. Registration makes available a detailed manual and is
necessary for anyone who wants to develop his or her own programs.
Running Programs
Decide which plot might be interesting, based on the discussion in the text.
Adjust the graph by setting the legend with the legend button. Perhaps put
one of the variables on the right side of the graph with Graph/Choose
Variables.
Possibly select the range of the axes with Graph/Axis Settings. Choose
colors or line types with the buttons.
As seen at the end of the Screenshot Guide, Parameter Plot runs are very
useful to display the steady-state values as a function of the values of one
parameter. For this, one needs to be sure that the Stoptime is sufficient to reach
steady-state for all the runs.
When running a program with arrays, as found in the finite-differenced
examples, the X axis can be set with [i] and the Y axes with the variables of
interest. The resulting graph is a plot of the variable values at the Stoptime in
all of the array sections. For equal-sized segments, this is the equivalent of a
plot of the variables versus distance. If the steady-state has been reached then
9.1 A Short Guide to Berkeley MADONNA 485
the graph gives the steady-state profile with distance. More on running
programs is found in Sec. 2 of the Appendix.
Berkeley Madonna, like all programming languages, has certain functions and
characteristics that are worth noting and that do not appear elsewhere in this
book.
Editing text
The very convenient built-in editor is usually satisfactory. Also the program
can be written with a word processor and saved as a text-only file with the suffix
".mmd". Madonna can then open it.
Is a bracket missing?
Madonna tests for bracket pairs, and a missing bracket will be indicated.
Optimisation of a variable.
There is optimization available under Parameters/Optimize, but if it is
something simple with one or two parameters, then sliders can also be
effectively used. If the value of a maximum is sought as a function of a single
parameter value, then the Parameter Plot for maximum value can be used.
DTMAX higher than DT, but sometimes the resulting curves are not smooth if
DTMAX is too high. In most cases, good results are obtained with AUTO and
DT set to about 1/1000 of the smallest time constant.
If no success is found with AUTO, then try STIFF and adjust by the same
procedure. Oscillations can sometimes be seen by zooming in on a graph; often
these are a sign of integration problems. Sometimes some variables look OK
but others oscillate, so look at all of them if problems arise. Unfortunately
there is not a perfect recipe, but fortunately Madonna is very fast so the trial-
and-error method usually works out.
(CHEMQSTATST
(FILE/CHEMCr1)
{Constants}
UM=OJKS=OJK1
8F*10 D1=0.25
Stoptime=80
!nrfX=1
lnt P=0
{Mass Balances}
; 81OMASS BALANCE EQUATION
; SUBSTRATE Q^ANGE EQUATION
; PRODUCT BALANCE EQUATION
9.2 Screenshot Guide to Berkeley MADONNA 489
Figure 1. The example CHEMO has been opened and the Menu (From left: File, Edit, Flowchart
active only for flowchart programs, Model, Compute, Graph, Parameters, Window and Help) and
Graph Buttons (From left: Run, Lock, Overlay, Table, Legend, Parameters, Colors, Dashed Lines,
Data Points, Grid, Value Output and Zoom).
UP AND QPEFIAT1GW
3T6RTTIME
iTOPTIMI
Kl=Q.03K2=O.OBY=fl.B DT 0,02
BF=10 D1=OJS MTOUT 0
LJM 0.3
KS 0.1
{Conditional equation for D} Kl 0,03
D=iftime=tstirttnenDl o.oa
0.8
10
01 0,34
toil 8=10 5
initP=G HITX 1
NITS 10
{Mass Balances} N1TP 0
; EQ
; BALANCE
Figure 3. The Model/Equations was chosen. Seen here is also the Parameter Window.
Figure 4. If a new graph is chosen under Graph/New Window then the data must be selected
under Graph/Choose Variables.
9.2 Screenshot Guide to Berkeley MADONNA 491
4J-,
4- -9,5
3,5 9
3 8.5
2.5 -8
2- -7.5
1.5 7
f- -8.5
0,5-
0 5.5
20 40 80 SO 108 120 140 ISO 130 200
TIME
Figure 5. A graph window for variables on the left and right-side Y axis with Legend Button
selected.
7.
6-
-7
5- 6
X 4- 5 01
-4
3-
-3
4-
-2
1- 1
0
20 40 80 100 120 140 180 !80 200
Figure 6. An Overlay Graph for three values of Dl as selected in the Parameter Window.
492 9 Appendix: Using the Berkeley MADONNA Language
Figure 8. A graph of two slider runs, showing the Parameters Menu pulled down.
9.2 Screenshot Guide to Berkeley MADONNA 493
Figure 10. A Parametric Plot was chosen for 40 runs changing values of Dl to give the
final, steady-state values. The Data Button was pressed to give the points for each run.
494 9 Appendix: Using the Berkeley MADONNA Language
Figure 11. The Optimize Window, with the value of Dl being selected to minimize the
expression -D*X. The value found was 0.27, corresponding to the Parameter Plot results.
Figure 12. Two Parameter Plots overlaid showing the effect of reducing Y from 0.8 to 0.6.
9.2 Screenshot Guide to Berkeley MADONNA 495
Figure 13. A program written in an array form allows plotting all the values versus time by
choosing the variable vector, here S[ ] versus TIME for the program CELLDIFF.
Figure 14. From the same program as Fig. 13, radial profiles of three runs are plotted in an
overlay plot. The [i] values can be selected in the Choose Variables. Here the parameter
Radius has been changed to demonstrate the large influence of diffusion length.
496 9 Appendix: Using the Berkeley MADONNA Language
Figure 15. Here the program file KLAFTT is run and fitted to data in the text-file KLADATA.
The data consists of two columns: time and CE at equal intervals as seen by the open circles on
the plot. Note that the fit variable is CE and the parameter varied to minimize the difference in
least squares is KLA.
10 Alphabetical List of Examples
Electrode measurement dynamics, 127, 129 First order lag equation, 127, 337
Electrode membrane, 337 First order lag model, 127
Electrode response characteristic, 336 First-order, 132
Electrode time constant, 127, 129 First-order time lag, 354
Elemental balances, 23 Flocculant cell mass, 145
Energy balances, 49 Flotation, 110
Entrance, 113 Flow interaction, 140
Enzymatic, 112 Flow velocity, 114, 148
Enzyme, 112, 118 Fluid, 120
Enzyme loading, 148, 149 Fluid elements, 113
Enzyme reactor, 115 Fluidized bed, 133, 299, 149
Enzyme-substrate complex, 69 Flux, 120
Equations, 113 Food/biomass ratio, 112
Equilibrium, 10, 122 Fractional conversion, 126
Equilibrium oxygen concentration, 337 Fractional response, 127
Equilibrium relationships, 46 Free rise velocity, 137
Equilibrium value, 132 Functional modes of control, 163
Errors, 161 Gas, 117
Exit, 113 Gas absorber, 136
Experimental reactor, 130 Gas Absorption, 117
Exponential, 108 Gas and liquid films, 121
Exponential and limiting growth phases, 103 Gas balance, 125
External film, 145 Gas balance method, 128, 126
External mass transfer, 145 Gas bubbles, 117
External transport rate, 153 Gas concentrations, 125
Extraction, 118 Gas flow rates, 125
Fed Batch, 58, 64 Gas holdup, 55, 124
Feed Forward Control, 173 Gas inlet, 137
Feedback, 161 Gas phase, 117
Fermentation, 101 Gas-liquid, 117, 120
Fermentation media, 137 Gas-liquid systems, 122
Pick's Law, 29, 120 Gas-liquid transfer, 336
Film coefficients, 122 Gel, 145
Final, 109 Growth, 110
Final control element, 162 Growth rate, 32, 103
Finite difference, 30 Heat of agitation, 52
Finite difference Model, 151 Heat of fermentation, 49
Finite differencing, 388 Heat Transfer, 51
Finite differencing technique, 159 Henry coefficient, 340
Finite-differencing, 115, 153 Henry's law, 33, 122, 337
First order, 65 Henry's law constant, 122
Index 503