See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/255567826

Modeling of Batch Alcohol Fermentation with Free and Immobilized Yeasts

Saccharomyces Cerevisiae 46 EVD

Article  in  Biotechnology & Biotechnological Equipment · October 2012

DOI: 10.5504/BBEQ.2012.0025

CITATIONS READS

19 794

6 authors, including:

Georgi Kostov Silviya B. Popova

University Of Food Technology - Plovdiv Bulgarian Academy of Sciences
95 PUBLICATIONS   173 CITATIONS    55 PUBLICATIONS   211 CITATIONS   

SEE PROFILE SEE PROFILE

Velizar Gochev Petia D. Koprinkova-Hristova

Plovdiv University "Paisii Hilendarski" Bulgarian Academy of Sciences
109 PUBLICATIONS   856 CITATIONS    109 PUBLICATIONS   536 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Innovative TechNologies for Utilization of the Wave and Wind Energy in the COastal Zone - INWECO View project

STUDY OF NANO-MODIFIED COATINGS PRODUCED BY MANUAL ARC OVERLAY WELDING View project

All content following this page was uploaded by Georgi Kostov on 05 June 2014.

The user has requested enhancement of the downloaded file.

 Article DOI: 10.5504/bbeq.2012.0025 BE
BIOTECHNOLOGICAL EQUIPMENT
Modeling of Batch Alcohol Fermentation with Free and
Immobilized Yeasts Saccharomyces cerevisiae 46 EVD
Georgi Kostov1, Silviya Popova2, Velizar Gochev3, Petia Koprinkova-Hristova2, Mihail Angelov4, Atanaska Georgieva5
1
University of Food Technologies, Department of Wine and Brewing Technology, Plovdiv, Bulgaria
2
Bulgarian Academy of Sciences, Institute of System Engineering and Robotics, Sofia, Bulgaria
3
Plovdiv University “Paisii Hiledarski”, Department of Biochemistry and Microbiology, Plovdiv, Bulgaria
4
University of Food Technologies, Department of Biotechnology, Plovdiv, Bulgaria
5
University “Paisii Hiledarski”, Department of mathematical analysis, Plovdiv, Bulgaria
Correspondence to: Velizar Gochev
E-mail: vgochev@uni-plovdiv.bg

ABSTRACT
Seven of the mathematical structures known by far were applied for modeling of batch alcoholic fermentation with free and
immobilized cells of Saccharomyces cerevisiae strain 46 EVD using experimental data. The obtained models were analyzed
and compared with respect to their quality and applicability for process simulation. Conclusions about the influence of cell
immobilization on the batch process intensification were drawn based on analysis of the obtained model parameters. It was found
that Monod-type model and logistic equation in combination with Ludeking-Piert model are the most appropriate structures for
description of both free- and immobilized-cell fermentations for the investigated yeast strain.

Biotechnol. & Biotechnol. Eq. 2012, 26(3), 3021-3030
Keywords: bio-fuels, ethanol fermentation, yeasts,
mathematical modeling
Introduction
The environmental problems caused by the use of fossil
feedstock as an energy source and the rapid increase of the
oil-based fuels prices are the main reasons that have motivated
the production of bio-fuels (9). Bio-ethanol, being a clean, safe
and renewable resource, has been considered as a potential
alternative to the ever-decreasing fossil fuels. Ethanol
production has increased dramatically during the last years,
because it is considered as a renewable and environmentally
friendly alternative (1, 4, 18). However, the economic
feasibility of the ethanol industry is still questioned and much
effort should be put into improving the process, especially
regarding rapid fermentation and resistance to the main
inhibition factors.
To eliminate inhibition caused by high concentrations of
the substrate and product as well as to enhance ethanol yield,
cell immobilization approaches have been applied in ethanol
production (13, 18). The advantages of immobilized cells
over free-cell systems have been extensively reported (17,
19). Immobilized-cell fermentation has been shown to be
more effective than the free yeast process, mainly due to the
enhanced fermentation productivity, feasibility for continuous
processing, cell stability and lower costs of recovery and
recycling and downstream processing. However, immobilized
cells still have limited industrial application as there is a
strong demand for development of large-scale immobilization
procedures (11). The process of immobilization changes
not only the envionment, but also the physiological and
Biotechnol. & Biotechnol. Eq. 26/2012/3
morphological characteristics of cells, and the catalytic
activity of enzymes; therefore, the fermentation conditions
(specially the kinetics) of the free-yeast fermentation and of
the immobilized-cell process are different (14). The main
advantage of immobilized-cell systems is that the solid support
allows a combination of high biocatalyst concentration and a
high reactor load, which can lead to smaller reactor volumes as
compared to suspended-cell processes (16).
Simulation investigations are proven to be powerful
tools for evaluating the fermentation processes alternatives
that decrease spending of expenses on pilot experiments
for design, control synthesis and scaling-up (4, 18, 21). The
quality of the simulation itself depends on the quality of the
underlying mathematical model used for prediction of the
responses of a given system to changes in environmental and
operating conditions. Hence the mathematical models should
describe with sufficient accuracy the mechanisms of the
processes under consideration. For the purpose of bioprocesses
simulation, kinetic models based on mass balance of the main
compounds in the bioreactor – biomass, target product and
main substrates, are usually applied. Concerning ethanol
fermentation various model structures have been developed
and investigated (4, 12, 18). In the experiments of Birol et
al. (2) eleven modeling structures of ethanol fermentation
by the yeast strain Saccharomyces cerevisiae ATCC 9763
immobilized in Ca-alginate gel beds were considered. They
describe the main factors affecting the ethanol concentration
– substrate limitation, substrate inhibition, product inhibition
and cell death – but none of them can account simultaneously
for all of these factors. Prakash (20) applied a simple logistic
growth equation combined with Luedeking-Piret equation (15)
for modeling of ethanol fermentation by mutant Neurospora
3021
crassa. However, there is no universal model structure that
could perfectly suit ethanol fermentation by all possible kinds
of strains since each particular strain has its specifics that
require an individual approach to kinetics modeling.
The aim of our study was to carry out a comparative
analysis of seven of the mathematical structures known by
far for modeling of batch alcoholic fermentation with free
and immobilized cells of Saccharomyces cerevisiae strain
46 EVD using real experimental data. The obtained models
were analyzed and compared with respect to their quality
and applicability for process simulation. The two types of
processes (with free and immobilized cells) were compared
with respect to the main model parameters that determine
the main interactions in the culture – cell growth limitation
and inhibition and transformation of sugars to ethanol and
biomass. Conclusions were drawn about the influence of
cell immobilization on the batch process intensification. The
main purpose of this study was: first, to investigate how cell
immobilization influences the process productivity, and,
second, to choose the best model that will be further refined and
used for control synthesis of the process in order to increase its
productivity.
Materials and Methods
Yeast strain
The strain-producer used was dry yeast Saccharomyces
cerevisiae 46 EVD provided by the Martin Vialatte OEnologie
company, France. The yeasts were stored at 4-6 °C. Before
utilization they were re-hydrated in 4% sugar solution at 30 °C.
The amount of this solution was 1:10 with respect to the dry
biomass. The amount of inoculating material was 1% of the
bio-reactor working volume because the aim was to achieve
107 CFU∙cm-3. During the immobilization the amount of used
biomass was determined such that we achieved 107 CFU∙g-1
preparations.
Nutrition medium
The nutrition medium was composed based on the results of
our previous experiments (10). The aim was to enhance the
product (ethanol) formation, which could be done by promoting
the reaction of glucose transformation to ethanol rather than to
biomass. Since the main cell growth factors are P and N, the
nutrition media was chosen to have the lowest possible amount
of these two elements. Another reason to choose a medium
with low concentration of P was that higher P concentrations
can destroy the alginate pearls in the case of immobilized-cell
fermentation. The composition of the nutrition medium used
in all experiments was (g∙dm-3): glucose – 118.40; (NH4)2SO4
– 2; KH2PO4 – 2.72; MgSO4×7Н2О – 0.5; yeast extract – 1.0.
Before inoculation the medium was sterilized for 20 min at
121 °C in an autoclave.
Yeast cell immobilization
For yeast cell immobilization 2% solution of Na-alginate was
used. It was prepared by dissolving alginate in distilled water
3022
at constant stirring until a homogeneous solution was obtained.
After that the solution was sterilized for 20 min at 121 °C. For
jellification 2% solution of CaCl2 also sterilized for 20 min at
121 °C was applied.
For immobilization the equipment shown in Fig. 1 was
used. It consists of two sterile vessels each with a volume of
500 cm3. One of these vessels contains the suspension and
the other, the gelification solution. During the immobilization
process constant stirring rate of the cells–alginate suspension
was maintained in order to obtain homogenous solution. The
suspension was transported via peristaltic pump and through
the injector nozzle it was poured out drop by drop into the
gelification solution. This mixture was stirred at a low speed
using a magnetic stirrer. Thus pearls with a mean diameter of
2 mm were obtained.
Fig. 1. Installation for Ca-alginate pearls preparation with peristaltic pump.
The obtained pearls were left for 30 min in CaCl2 solution.
After that they were washed with distilled water and kept in
saline or sterile water in the refrigerator for later use. The
immobilization method used is described in greater detail by
Willaert (24).
Bioreactor and cultivation conditions
The laboratory bioreactor used is a glass cylinder with
geometrical volume of 2.0 dm3 and working volume of
1.7 dm3. It is equipped with a six-blade turbine stirrer and four
baffles. On the top of its head plate orifices for feeding in of
nutrition media and air, leading out of gases, inserting of heat-
exchangers and sensors for pH, temperature and dissolved
oxygen are mounted. The installation includes also measuring
devices and controllers for the main process variables (Fig. 2).
The culture media temperature is controlled via two channels:
cold water for cooling and heater for heating. The active acidity
(pH) is measured by combined glass-silver chloride electrode
(Ingold, Switzerland). The temperature and pH controllers
are integrated into the control device (Applikon, Holland)
equipped with precise controllers.
For maintaining constant pH of the culture media sterilized
reagent – 20% KOH solution – was supplied via peristaltic
pump. After reaching the desired pH value the pH controller
was switched on in order to maintain pH 4.5 ± 0.05 for both
types of processes – with free and immobilized cells. The
temperature was maintained at 28 °C ± 0.1.
Biotechnol. & Biotechnol. Eq. 26/2012/3
 The bioreactor was sterilized in “cold” conditions using
0.3% solution of neomycin for 24 hours. After that it was
washed out with sterile water. The suspension was fed in via
peristaltic pump used also for pH control. The immobilized
preparation was washed out with sterile physiological solution
and then it was fed into the bioreactor at sterile conditions.
After that the nutrition medium was inserted into the apparatus.
The amount of immobilized material was 10% of the nutrition
medium volume.
Fig. 2. Scheme of the laboratory bioreactor. 1: apparatus with geometrical
volume 2 dm3; 2: blades; 3: thermo-resistance Pt100; 4: heater; 5: cold water
heat exchanger; 6: turbine stirrer; 7: рН electrode; 8: output for the СО2; 9:
filter; 10: peristaltic pump for pH control; 11: pH control reagent – 20% КОН;
12: motor; 13: control connections; 14: control device “Applikon”; 15: lid.
Samples for analysis of the main process variables –
concentration of glucose, biomass and product – were taken
from the bioreactor every 2 hours.
Analytical procedures
Concentration of ethanol and glucose in the culture
medium
The accumulated ethanol and glucose concentrations were
determined by a density meter (Anton Paar DMA 4500,
Austria). The method includes density and refraction index
measurement for determination of glucose and ethanol
concentrations. The density is measured by the oscillating
U-tube sensor. The working ranges of Anton Paar DMA 4500
are: density – 0-3 g∙cm-3; temperature – 0-90 °C. Integrated
alcohol and sugars tables were used. This is a standard method
recommended by the European Brewery Convention (5).
Biomass concentration in the culture medium
The biomass concentration in the culture medium was
determined spectrophotometrically at 620 nm using Spekol
221 (Germany).
Biotechnol. & Biotechnol. Eq. 26/2012/3
Biomass concentration in the immobilized phase
To determine the biomass concentration in the immobilized phase
1.0 g of sample was added to 1.0% solution of sodium citrate;
for the control probe 1 g of pure alginate pearls were added to
1% sodium citrate; after full dissolving of the pearls the biomass
concentration was determined spectrophotometrically (26).
Results and Discussion
Mathematical models
Here we chose simplified mass-balance mathematical
models that reflect only the kinetic rates of the main process
reactions: biomass growth, ethanol production and substrate
consumption for biomass and product formation. In the
case of immobilized cell cultivation there is complementary
influence of the diffusion resistances on the process dynamics.
The cultivation can be described by adding to the model mass
balance equations accounting for diffusion and convective
mass exchange inside and at the surface of the jelly pearls
containing immobilized yeasts. The reason not to include such
complications in the model of immobilized-cell fermentation
was that our primary aim was to compare it with free-cell
fermentation and to reveal the differences. The same approach
was used in Birol et al. (2), where model structures applied
for immobilized-cell fermentation also did not account for
diffusion effects. Moreover, as it was shown by Viktor et al.
(22), in order to develop a more complicated model including
these effects we need first a model of free-cell fermentation
that can be upgraded further with mass exchange dependences.
That is why in the preliminary investigation presented here we
chose to make all these simplifications and to use the models
described below by systems of Equations 1 and 2.
For the modeling we applied two different model structures.
The first one as in Birol et al. (2) is as follows:
(1)
It is based on the assumption that the substrate denoted by
S is consumed with a rate proportional to the rates of increase
of the cells (X) and product (P) concentrations. The constants
Yx/s and Yp/s are yield coefficients [g/g] that reflect the ratio of
biomass and product obtained from the substrate consumed for
their synthesis; μ and q denote the specific biomass growth and
ethanol production rates respectively. Since in the literature
there is a large variety of mathematical dependences for the
two main process kinetic rates μ and q, the choice of proper
dependences is usually done by trial and error. In the present
investigation we tried six (given in Table 1 below) of the
eleven dependences investigated before (2). They include
substrate limitation (numbers 1 and 2 from the table), substrate
inhibition (number 3) and product inhibition (numbers 4, 5 and
6) effects.
3023
 The second model structure from Luedeking and Piret (15)
is based on the logistic equation for biomass growth rate and
Ludeking-Piret equation that implies product and substrate
dynamics linear dependence on the biomass concentration and
biomass growth rate as follows:
(2)
Here μm, β, qp0, K, δ and γ are model parameters.
In both cases the parametric identification of the models
was carried out in MATLAB environment with Optimization
toolbox functions for model parameters identification using
the least squares error minimization method and the explicit
Runge-Kutta formula of 4-5 order for solving the differential
Equation Systems 1 and with direct analytical integration of
the equation in System 2 using the methodics presented by
Wang et al. (23).
The influence of cell immobilization is described by the
efficiency coefficients ημ (related to the maximum specific
growth rate) and ηq (related to the maximum specific ethanol
production rate) defined as follows:
(3)
The choice of the two structures of mathematical models
1 and 2 and the equations in Table 1 was made on occount
of the following considerations. The system differential
Equations 1 is a classic description of various biotechnological
processes. The essential terms for the final selection of model
structures are the models for the specific growth rate μ and
the specific product accumulation rate q. Birol et al. (2) tested
11 mathematical models obtained by different authors and
applied for different biotechnological processes. The careful
analysis of that work and our experimental results restricted
our choice to 6 of these 11 mathematical models. For some of
the discarded models unusual and unexplained constant values,
even from a biological point of view, were obtained, while in
other models it was impossible to identify the parameters. The
mathematical models chosen as a result of the pre-selection
are listed in detail in Table 1. Some features of the selected
mathematical relationships will be further commented below.
The System Equations 2 has a very significant advantage
– it could be integrated analytically and its constants could
be determined directly from the experimental data. Another
important advantage of this system of equations is the biological
meaning of its constants. Through them one can determine
what part of the product accumulation is associated with the
cell growth; whether there occurred significant physiological
changes in the culture – defined by the constant β; how much
of the substrate was assimilated for biomass growth and how
3024
much of the substrate was transformed into a metabolic product
– defined by the constants δ and γ.
The work of Mi-Young et al. (17) quoted modification of
the first equation in System 2. The modification takes into
account the effect of lag-phase and the product inhibition of
cell growth (the equation is applied for obtaining lactic acid
by lactobacilli). As it will be seen below, when using ethanol
tolerant yeasts, the product inhibition has no significant impact
and therefore the classical System 2 can be used. The usage of
the proposed modification is necessary because the Ludeking-
Piret model did not describe well the asymptotic increase of
lactic acid and the substrate transformation into the metabolic
product (17). The investigated process of ethanol fermentation
had a clear end and full transformation of the substrate (Fig.
3) (10). Since the equation in the work of Mi-Young et al. (17)
gives interesting guidelines, it will be applied in future studies
of the alcoholic fermentation by yeasts.
Experimental data
The experimental data from the batch cultivations carried out
using free and immobilized yeasts are shown in Fig. 3. They
are arithmetic mean of three independent experiments. The
immobilized-cell fermentation resulted in faster achieving of
maximum ethanol concentration after 24 hours cultivation. In
the case of free cells the same ethanol concentration was reached
after 30 hours of cultivation. In the case of immobilized-cell
fermentation the ethanol yield was 83% of the theoretically
possible outcome, while in the case of free-cell fermentation
it was about 88% of the theoretical outcome. This could be
explained by the influence of internal and external diffusion
resistances in the immobilization pearls.
Fig. 3. Dynamics of alcohol fermentation with free and immobilized cells of
S. cerevisiae 46 EDV.
From the experimental data it becomes clear that the
biomass concentration in the case of immobilized-cell
fermentation is 3 times lower than the one in the case of free-
cell fermentation. This could be explained by the decision to
start each fermentation with 107 CFU∙cm-3 active free cells. As
for inoculums preparation we used dry yeasts biomass with
Biotechnol. & Biotechnol. Eq. 26/2012/3

Mathematical dependences for the specific kinetic rates
№ Model name
1 Monod
2 Tiessier
3 Andrews and Noack
4 Hinshelwood
5 Aiba
6 Ghose and Tyagi
109 CFU∙cm-3, this means that free-cell fermentation actually
started with 3 g∙dm-3 initial concentration of viable cells. Since
we aimed to compare two fermentations, we had to ensure
the same initial cell concentration in the case of immobilized
cells. However, inoculation of 107 CFU∙cm-3 active cells after
immobilization led to about 1.0 g∙dm-3 biomass concentration
in the bioreactor, i.e. 3 times lower than that in the case of free
cells. In fact this is one of the advantages of immobilized-cell
fermentations – they are faster in the presence of lower cell
concentrations. Moreover, the immobilized preparations could
be used several times depending on their operation stability,
which varies from 3 to 6 months. This can also save time for
initial inoculum preparation before the fermentation start.
Another difference visible from the experimental data plots
is that the substrate and ethanol dynamics were also different.
In the case of immobilized-cell fermentation the substrate
was almost exhausted by the 12th hour but ethanol formation
continued till the 24th hour. This could be explained by the
diffusion process into and out of the alginate pearls which is
driven by the differences in the concentrations of substances
inside and outside these pearls: the substrate (glucose) enters
the pearls reaching the yeast cells that produce ethanol; in
turn ethanol must go out from the pearls. Since the ethanol
production rate of immobilized yeasts is 4-5 times higher,
the glucose concentration gradient on the pearls surface stays
constant. However, by the end of the fermentation the ethanol
concentration becomes higher which increases its gradient on
Biotechnol. & Biotechnol. Eq. 26/2012/3
Table 1
μ q
the pearls surface causing ethanol diffusion from their inside
towards the outside even after glucose exhaustion. When the
ethanol concentration gradient becomes smaller it prevents
its diffusion into the culture medium leaving some amount
inside the pearls and stopping its increase on the outside. This
is a sign that the fermentation should be stopped in order to
prevent undesired yeast lysis or undesirable change in their
metabolism that could lead to ethanol consumption.
Another explanation given by Kourkoutas et al. (11) is
that there are possible changes in the physiological state and
activity of the yeasts resulting from the immobilization which
could lead to an increase in their viability and activity. These
changes depend to a very great extent on the immobilization
method. However, the published data about these changes
are controversial. For example, Jamai et al. (7) reported
insignificant physiology change of S. cerevisiae; Buzas et al.
(3) and Galazzo and Bailey (6) observed that the fermentation
activity of immobilized cells in alginates was independent of
the pH of the culture medium; Galazzo and Bailey (6) reported
reduction of the intracellular pH of the immobilized yeast
cells. The reduced intracellular pH value of the immobilized
cells could lead to increased enzyme activity and productivity.
It was also attributed to increased permeability of cell
membranes, which leads to increased glycolytic activity
and glucose uptake rate. These observations can explain the
resulting profile of substrate consumption of the process with
the immobilized cells shown in Fig. 3.
3025
 Table 2
Identified model parameters and fitting errors

Model parameters Efficiency coefficients Model error

μmax Ksx qpmax Ksp Yx/s Yp/s Kpx Kpp ημ ηq
h-1 g∙dm-3 g∙(g∙h)-1 g∙dm-3 - - g∙dm-3 g∙dm-3 - -
Monod
Free cells
0.613 700 0.432 0.413 0.029 200 - - 0.548
0.219 5.25
Immobilized cells
0.134 55.4 2.27 0.009 0.006 34.13 - - 0.458
Tiessier
Free cells
0.469 561.69 0.422 0.001 0.021 10 - - 0.574
0.936 6.82
Immobilized cells
0.439 349.93 2.88 3.07 0.010 0.829 - - 0.993
Andrews and Noack
Free cells
0.026 700 3.15 100 0.022 150 3.73 0.138 0.631
3.33 1.6
Immobilized cells
0.088 438.03 5.04 0.004 0.006 6.85 0.97 0.45 0.419
Hinshelwood
Free cells
0.254 203.46 0.452 2.154 0.022 200 0.019 0 0.715
0.469 4.95
Immobilized cells
0.119 42.67 2.237 0.0185 0.006 200 0 0 0.436
Aiba
Free cells
0.18 18.31 0.4 2.206 0.0314 2.795 0.115 0 1.12
0.844 6.36
Immobilized cells
0.152 27.61 2.544 0.006 0.007 2.11 0.04 0 0.695
Ghose and Tyagi
μmax Pxmax qpmax Ppmax Yx/s Yp/s - -
h-1
g∙dm -3
g∙(g∙h) -1
g∙dm -3
- - - -
Free cells
3.131 1.97
0.122 18.63 0.653 144.7 0.1 0.324 - - 14.8
Immobilized cells
0.382 21.98 1.28 156.47 0.1 0.356 - - 7.38
Logistic equation and Ludeking-Piret model
μmax β qpmax K δ γ

h-1
m ∙(kg∙h)
3 -1
g∙(g∙h) -1
g∙(g∙h)-1 - -
Free cells
0.982 5.31
0.385 0.066 0.386 0 0.178 35.26 0.651
Immobilized cells
0.378 0.32 2.052 0 6.365 64.4 0.512

3026 Biotechnol. & Biotechnol. Eq. 26/2012/3

Quality of the mathematical models
The obtained parameters for all the investigated models
are given in Table 2. Fig. 4, Fig. 5 and Fig. 6 represent the
corresponding experimental data fitting by all the models.
Here we shall discuss the models quality and meaning of the
obtained parameter values.
b b
Fig. 4. Comparison of mathematical models for biomass concentration in the
cases of free cells (a) and immobilized cells (b).
Fig. 4 shows the model simulation results in comparison
with experimental data for the biomass concentration
modeling equations. As seen, all the identified models describe
with relatively good accuracy the batch ethanol fermentation
process in both cases (with and without cell immobilization). In
order to choose the best model it is important to consider how
well it describes the transition from exponential to stationary
phase of the process. Looking at Fig. 4a we can conclude that
in the case of free-cell fermentation the models of Aiba and
Ghose and Tyagi (numbers 5 and 6 in Table 1) do not describe
this transition well in comparison to the rest of the models. The
Ghose and Tyagi model supposes biomass growth inhibition
by the product starting from the initial product accumulation;
and the maximum concentration when the inhibition is
complete is defined by the parameter Pxmax = 18.63 g∙dm-3.
Biotechnol. & Biotechnol. Eq. 26/2012/3
It is obvious that this model does not suit the S. cerevisiae 46
EVD dynamics because the product inhibition effect cannot
be accounted for properly above this ethanol concentration.
Aiba’s model predicts higher biomass concentrations than the
experimental one but keeps the biomass trend characteristic
similar to the experimental data.
Fig. 5. Comparison of mathematical models for ethanol concentration in the
cases of free cells (a) and immobilized cells (b).
Most of the biomass models in the case of free-cell
fermentation have considerably high values of the parameter
KSX – in the range of 203.46 ÷ 700 g∙dm-3. These high values
could be explained by the fact that for alcohol yeasts it is
typical to produce relatively low biomass concentrations
– only about 15% of the substrate is consumed for biomass
growth (8, 26). This results in lower specific biomass growth
rates due to substrate limiting effect on biomass. Thus the
product accumulation into the culture medium is dominating.
Aiba’s model is an exception, having Ksx = 18.31 g∙dm-3,
which means lack of significant substrate limitation effect on
biomass growth, and respectively explains the higher predicted
yeast concentrations.
A similar analysis can be done for the models of the
second case of immobilized-cell fermentation (Fig. 4b).
3027
Since immobilized cells grow inside the jelly pearls, they are
restricted by diffusion resistances and have a considerably
lower specific growth rate. The efficiency coefficient (ημ)
varies in the range of 0.319 ÷ 4.57. The decreasing of the
specific growth rate (Aiba’s model is an exception) is due to
the internal diffusion resistances rather than due to the external
ones because the process is carried out with intensive stirring
and hence the main resistance is due to the substrate diffusion
through the alginate pores. In that case the models also have
considerably lower values of Ksx in comparison with those for
the free-cell fermentation models. Their values are in the range
of 27.61 ÷ 438.03 g∙dm-3. This could be explained with higher
concentration of viable yeasts cells in the bioreactor working
volume.
a obtained values for the parameter Ksp are considerably smaller
b from the ethanol influence.
Fig. 6. Comparison of mathematical models for glucose concentration in the
cases of free cells(a) and immobilized cells (b).
The ethanol accumulation and substrate consumption were
described with considerably high accuracy by all the identified
models, as shown in Fig. 5 and Fig. 6. In the case of free-cell
fermentation the only exception is the model of Ghose and Tyagi
(Fig. 5а). Its predictions are far from the experimental data
during the exponential cell growth as well as in the stationary
3028
phase. The best model for both cases is that of Tiessier. All
other models predict ethanol concentration increase after 24 h,
while according to the experimental data it should be stationary
and about 90% of the theoretical outcome. Nevertheless, we
can conclude that all the models except Ghose and Tyagi’s one
follow the free-cell fermentation trend well. The deviation of
this model from the experimental data is due to its structure,
which supposes product accumulation inhibition by itself
which was not the case with our experimental data. Hence,
Ghose and Tyagi’s model is not applicable to our modeling
purposes.
In the case of immobilized-cell cultivation the best
model is Monod type. The Tiessier model underestimates
the experimental data after 14 h. The logistic equation and
Ludeking-Piret model fits the experimental data with the
highest precision. As a whole, all the models are acceptable
excluding the Ghose and Tyagi model. This can be explained
by the structure of this model, which includes biomass growth
inhibition by the product. However, in our case the yeasts
grow at lower ethanol concentrations than the ones supposed
by the value of the parameter Ppmax = 156.47 g∙dm-3 maximum
inhibiting concentration.
The ethanol accumulation in the case of free-cell
fermentation has maximum production rates (qpmax) in the range
of 0.386 ÷ 3.15 g∙(g∙h)-1, while in the case of immobilized-cell
fermentation it is in the range of 1.28 ÷ 5.04 g∙(g∙h)-1, which
is 1.6 to 6.82 times higher. This fact proves the advantage of
the immobilized-cell cultivation. These product accumulation
rates are close to the maximum possible rates since the
than the substrate concenation (S) and hence S/(Ksp + S) ≈ 1.
The only exception is the model of Andrews and Noack with
Ksp = 100 g∙dm-3, which is approximately two times lower than
the maximum product accumulation rate.
The structures of Andrews and Noack, Hinshelwood and
Aiba models include inhibition of cells growth and product
accumulation from the current ethanol concentration via the
parameters Kpx and Kpp. The values of these two parameters
(see Table 2) are very low and in some cases even zero. This is
because the obtained ethanol concentration is much lower than
the concentration that causes complete cell growth inhibition
for our experimental investigations. From the Ghose and
Tyagi model it can be concluded that the inhibiting ethanol
concentrations are in the range of 144.7 ÷ 156.7 g∙dm-3. In the
case of immobilized-cell cultivation the values of these two
parameters are lower because the jelly pearls prevent the cells
In the case of logistic equation model the value of
the efficiency coefficient (ημ) is close to 1. This could be
explained by the fact that the fermentation is carried out in a
bioreactor with intense stirring that leads to minimal external
diffusion resistances. Since the alginate beds have relatively
big pores (9, 16), the diffusion through them is easy, i.e. the
immobilized cells grow in conditions similar to those of free
fermentation. This can also explain the similar values of the
Biotechnol. & Biotechnol. Eq. 26/2012/3
maximum growth rate parameter μmax in both cases. Another
interesting fact is that the coefficient of internal population
concurrence β is higher in the case of immobilized cells,
which can be explained by the restricted space in the alginate
pearls in comparison to the case with free-cell fermentation.
Hence, model 2 has a more appropriate structure for the
process because it accurately reflects the physiology changes
in immobilized cells due to entrapment in the carrier.
All models describe with high precision the sugar
consumption dynamics (Fig. 6). The obtained values of the
parameters Yx/s (see Table 2) are in the normal range for
alcohol yeasts. According to Yarovenko (25) about 10% to
15% of the substrate is transformed into biomass and the rest
is used for product formation. The parameters Yp/s physically
describe the ratio between the produced ethanol and the sugars
consumed for this purpose. However, in the culture medium
there are free enzyme systems that additionally contribute to
the product formation. Hence, the values of these parameters
reflect not only the ethanol production by yeasts but also the
additional ethanol formation by the free enzymes. In the case
of immobilized-cell fermentation it is logical to have higher
values of YP/S due to the already mentioned fact that the culture
medium pH does not influence the fermentation activity of the
immobilized cells, which leads to higher glycolytic activity
of the yeasts. However, for most of the models we obtained
lower values for this parameter. A possible explanation could
be found in the diffusion resistances, which from some point
onward prevent ethanol from leaving the carrier. From the
parameters of the Ludeking-Piret model we can conclude that
ethanol produced by the biomass is accumulated during the
stationary phase (К = 0) while sugars are consumed during
both process phases.
From the obtained results it is difficult to choose a single
best fitting model. Good approximation potential was shown
by the models of Monod, Tiessier, Hinshelwood and logistic
equation in combination with Ludeking-Piret model. The four
mathematical relationships are characterized by simplicity
and good approximating potential. The aim of this work was
to make an initial selection of mathematical relationships,
which begin the description of both the batch fermentation
process with immobilized cells in the presence of diffusion
resistances and continuous fermentation with immobilized
cells. In terms of proper synthesis models of such systems
should be characterized by simplicity, as the four models are.
These relationships described the ethanol fermentation process
very well and gave a clear idea of the process parameters’
influence on the kinetic characteristics. The models of
Andrews and Noack, Aiba and Ghose and Tyagi could be
useful when a significant product and/or substrate inhibition
affect the process. In practice the Hinshelwood model could
be simplified to the Monod equation because of weak product
inhibition. These observations and results were in accordance
with the observation of Birol et al. (2) that the Monod and
Hinshelwood models were suitable to describe the process of
alcoholic fermentation.
Biotechnol. & Biotechnol. Eq. 26/2012/3
Thus, of the 6 proposed structural models for the System
of Equations 1, the choice fell on the three most simple
dependencies. For further research we selected the Monod-type
structure model (Monod and Hinshelwood) and the equation
of the logistic curve model with Ludeking-Piret model. These
models have the best approximating ability to describe the
alcoholic fermentation with Saccharomyces cerevisiae 46
EDV. These models were characterized by accuracy and clarity
of the obtained parameters and allow to generate new systems
of differential equations or intelligent methods to describe the
continuous fermentation process more easily. We have made a
preliminary analysis of one intelligent model using the Monod
model and neural networks, which gives a good starting point
for control of the continuous fermentation process. Another
group of models using the other two mathematical models is
under development. The results of these studies will be the
subject of future publications.
Conclusions
In the present investigation we considered modeling of
batch fermentations with free and immobilized yeasts
Saccharomyces cerevisiae 46 EDV in Ca-alginate. Seven
of the most popular batch cultivation model structures were
investigated. From the obtained results it is difficult to choose
a single best fitting model. Good approximation potential was
shown by the models of Monod, Tiessier, Hinshelwood and
logistic equation in combination with Ludeking-Piret model.
For further research we selected the Monod-type structure
model (Monod and Hinshelwood) and the equation of the
logistic curve model with Ludeking-Piret model.
Based on the physical meaning of the obtained model
parameters we compared the free- and immobilized-cell
fermentations. The results definitely proved the advantage of
cell immobilization leading to higher production rates.
Our conclusion is that the Monod and Ludeking-Piert model
in combination with logistic equation are the candidates for
the best model, which we shall improve further for modeling
of continuous alcohol fermentation with immobilized cells
in column bioreactor and for generation of control strategies
aimed at process intensification.
Acknowledgements
This work was supported by the National Science Fund of
Bulgaria, Project No. DТК 02/27 “Increasing the efficiency
of bio-fuel purpose ethanol production” and by the bilateral
agreement project “Advanced intelligent control in chemical
and bio-chemical industries” between ICSR–BAS and
Petroleum-Gas University, Ploiesti, Romania.
REFERENCES
1. Bai F., Anderson W.A. Moo-Young M. (2008) Biotech.
Adv., 26, 89-105.
3029
2. Birol G., Doruker P., Kardar B., Onsan Z.I., Ülgen K.
(1998) Process Biochem., 33, 763-771.
3. Buzas Z., Dallmann K., Szajani B. (1989) Biotech.
Bioeng., 34, 882-884.
4. Cardona C.A., Sanchez O.J. (2007) Bioresour. Techn.,
98, 2415-2457.
5. European Brewery Convention (2005) Analytica EBC,
Fachverlag Hans Carl, Nurnberg.
6. Galazzo J., Bailey J. (1990) Biotech. Bioeng., 36, 417-
426.
7. Jamai L., Sendide K., Ettayebi K., Errachidi F.,
Hamdouni-Alami O., Tahri-Jouti M.A., McDermott T.,
Ettayebi M. (2001) FEMS Microbiol. Lett., 204, 375-379.
8. Johnson E.A., Echavarri-Erasun C. (2011) In: The
Yeasts, 5th Ed. (J.W. Fell, C.P. Kurtzman, T. Boekhout,
Eds.), Elsevier Academic Press, 315-422.
9. Kosaric N., Vardar-Sukan F. (2001) In: The Biotechnology
of Ethanol (M. Roehr, Ed.), WILEY-VCH Verlag GmbH,
Weinheim, pp.87-90.
10. Kostov G. (2007) Investigation of Fermentation Systems
for Ethanol Production, PhD thesis, University of Food
Technologies, Plovdiv, Bulgaria, p.180. (In Bulgarian)
11. Kourkoutas Y., Bekatorou A., Banat I.M., Marchant
R., Koutinas A.A. (2004) Food Microb., 21, 377-397.
12. Lin Y., Tanaka S. (2006) Appl. Microb. Biotech., 69, 627-
642.
13. Liu C.Z., Wang F., Ou-Yang F. (2009) Biores. Techn.,
100, 878-882.
14. Liu R., Shen F. (2008) Biores. Techn., 99, 847-854.
15. Luedeking R., Piret E.L. (2000) Bioeng., 67, 636-644.
3030
View publication stats
16. Martins Dos Santos V.A.P., Tramper J., Wijffels
R. (2005) In: Fundamentals of Cell Immobilization
Biotechnology (V. Nedovic, R. Willaert, Eds.), Kluwert
Academic Publishers, 493-529.
17. Mi-Young H., Si-Wouk K., Yong-Woon L., Myong-Jun
K., Seong-Jun K. (2003) J. Biosci. Bioeng., 96, 134-140.
18. Mussatto S.I., Dragone G., Guimaraes P.M.R., Silva
J.P.A., Carneiro L.M., Roberto I.C., Vicente A.,
Domingues L., Teixeira J.A. (2010) Biotech. Adv., 28,
817-830.
19. Plessas S., Bekatorou A., Koutinas A.A., Soupioni M.,
Banat I.M., Marchant R. (2007) Biores. Techn., 98, 860-
865.
20. Prakash M.S. (2010) Int. J. Chem. Eng. Res., 2(1), 13-21.
21. Sablani S., Datta M.A.K., Rahman S., Mujumdar
A.S. (2006) Handbook of Food and Bioprocess Modeling
Techniques, Taylor and Francis Group, LLC, p. 610.
22. Viktor A.P., Dos Santos M., Tramper J., Wijffels
R. (2004) In: Fundamentals of Cell Immobilization
Biotechnology (V. Nedovic, R. Willaert, Eds.), Springer,
493-526.
23. Wang X., Gai Z., Yu B., Feng J., Xu C., Yuan Y., Lin, Z.,
Xu P. (2007) Appl. Environ. Microbiol., 73, 6421-6428.
24. Willaert R.G. (2001) Immobilized Cell, Springer, Berlin,
p. 120.
25. Yarovenko V. (2002) Technology of Ethanol, Kolos-Press,
Moskow, p. 602. (In Russian)
26. Zhou Y., Martins E., Groboillot A., Champagne C.P.,
Neufeld R. (1998) J. Appl. Microbiol., 84, 342-348.
Biotechnol. & Biotechnol. Eq. 26/2012/3