Académique Documents
Professionnel Documents
Culture Documents
MAY 2005
iii
ACKNOWLEDGEMENT
I would also like to record my thanks to Mohd. Sabri Sethpa and Nor Zalina
Othman of the Chemical Engineering Pilot Plant (CEPP), Universiti Teknologi Malaysia
for their assistance, support and sharing of their knowledge during this endeavor.
Thanks again to Dr. Firdausi Razali, Mohd. Sabri, Nor Zalina, my family and
friends for their significant contributions to my needs during the long hours.
v
ABSTRACT
The scale-up studies based on the constant oxygen transfer coefficient (kLa)
from 16 liter to 150 liter of aerated and agitated bioreactor were performed. The
studies included the investigation on the significance of hydrodynamic difference
between Rushton and marine impeller on the kLa at 16 liter scale. By employing
both static and dynamic gassing out techniques, the kLa values were calculated at
different sets of impeller speeds and air flow rates performed in various viscosities
and temperatures in the 16 liter and 150 liter BioengineeringTM stirred bioreactor.
Empirical correlation was employed to correlate and investigate the dependence of
kLa on specific power input and superficial air velocity. Our experimental results
discovered that the Rushton turbine was more effective in gas distribution and
provide a greater oxygen transfer rate than the marine impeller. In maintaining a
constant kLa upon scale-up from 16 to 150 liter, the specific power input and the
superficial air velocity cannot be maintained, adjustment has to be done. Specific
power input from 0.0001 to 4.2 kW/m3 and superficial air velocity within the range
of 9 x 10-4 to 7 x 10-3 m/s was tested to maintain a constant value of kLa upon scale-
up in distilled water and CMC solution model. The operating variables employed at
150 liter scale successfully gave a comparable kLa values as in 16 liter scale. Hence,
the calculated scaling-up factor for impeller speed and air flow rate were 0.28 and
3.1, respectively. In order to investigate the potential of employing scaling-up
protocol developed in this work, the kinetic profiles of E.coli batch fermentation at
16 and 150 liter were compared. By employing the scaling-up factors, the proposed
scale-up protocol managed to provide the similar trend of cell growth, glucose
consumption and oxygen uptake rate upon scale-up based on the constant kLa. It
may be concluded that the similar kLa for both scales was successfully achieved by
employing the proposed scale-up protocol.
vi
ABSTRAK
TABLE OF CONTENTS
Declaration ii
Dedication iii
Acknowledgement iv
Abstract v
Abstrak vi
Table of Contents vii
List of Tables xi
List of Figures xiv
List of Symbols xviii
Greek Letters xx
List of Appendices xxi
1 INTRODUCTION 1
2 LITERATURE REVIEW 7
3 METHODOLOGY 38
4.1 Introduction 61
4.2 Hydrodynamics Difference between Rushton and 62
Marine Impeller
4.2.1 Proportional Effect of Agitation and Aeration 62
Rates on KLa
x
6 REFERENCES 107
7 APPENDICES 112
xi
LIST OF TABLES
1.1 Values of parameter 'b' and 'c' from several works that 2
estimated from the empirical relationship proposed by
Cooper et al. (1944)
2.1 Different scale-up criteria and their consequences 32
3.1 Dimensions of 16 liter and 150 liter bioreactor 39
3.2 Operating conditions and techniques to determine the 41
oxygen transfer coefficient (kLa) reported in several works
3.3 Operating variables at 16 liter bioreactor 42
3.4 Operating variables at 150 liter bioreactor 49
3.5 Oswald-de Waele model at various CMC concentrations 54
3.6 Batch fermentation medium for production of E.coli 56
3.7 Operating conditions for E.coli fermentation at 150 liter 59
4.1 Increase of kLa values at higher operating temperature in 66
Rushton turbine and marine impeller at different impeller
speeds
4.2 Increase of kLa values at higher operating temperature in 66
Rushton turbine and marine impeller at different air flow
rates
4.3 Increase of kLa values at high broth viscosities in Rushton 68
turbine and marine impeller at different impeller speeds
4.4 Increase of kLa values at high broth viscosities in Rushton 68
turbine and marine impeller at different air flow rates
xii
4.18 The values of constant ‘b’ and ‘c’ upon scale-up from 16 95
liter to 150 liter at different operating temperature in air-
water system
4.19 The values of constant ‘b’ and ‘c’ upon scale-up from 16 95
liter to 150 liter at different liquid viscosities in air-viscous
system
4.20 The values of range of operating parameters varied upon 96
scale-up from 16 liter to 150 liter at different operating
temperature in air-water system
4.21 The values of range of operating parameters varied upon 96
scale-up from 16 liter to 150 liter at different liquid
viscosities in air-viscous system
4.22 Comparison of constant ‘b’ between E.coli culture broth 101
with air-water system in 16 liter and 150 liter
4.23 Comparison of experimental values of constant ‘b’ and ‘c’ 102
upon scale- up of E.coli fermentation from 16 liter to 150
liter
4.24 The values of range of operating parameters varied upon 103
scale-up from 16 liter to 150 liter in E.coli Fermentation
xiv
LIST OF FIGURES
3.6 Viscosity (kg/m.s) change with shear Rate (s-1) for CMC 54
solution
3.7 Deviation from Newtonian behaviour due to CMC 55
presence in the fluid at 30oC
3.8 Steps in E.coli fermentation at 150 liter scale 60
4.1 Dependence of kLa on impeller speed, N at different 63
temperature for Rushton turbine and marine impeller
4.2 Dependence of kLa on impeller speed, N at different 63
viscosities for Rushton turbine and marine impeller
4.3 Dependence of kLa on air flow rate, Q at different 65
temperature for Rushton turbine and marine impeller
4.4 Dependence of kLa on air flow rate, Q at different 65
viscosities for Rushton turbine and marine impeller
4.5 Dependence of kLa on volumetric power consumption, 70
Pg/VL at different temperature for Rushton turbine and
marine impeller
4.6 Dependence of kLa on volumetric power consumption, 70
Pg/VL at different viscosities for Rushton turbine and
marine impeller
4.7 Flow pattern produce by impellers. (a) axial-flow (b) 72
radial-flow
4.8 Dependence of kLa on volumetric superficial air velocity, 74
vg at different temperature for Rushton turbine and
marine impeller
4.9 Dependence of kLa on volumetric superficial air velocity, 74
vg at different viscosities for Rushton turbine and marine
impeller
4.10 Dependence of kLa on impeller speed in distilled water at 85
o o
different temperatures (a) T = 30 C (b) T = 40 C (c)T =
50oC
4.11 Dependence of kLa on impeller speed in CMC solution at 86
different concentrations (a) CMC 0.25%(w/v) (b) CMC
0.5%(w/v) (c) CMC 1% (w/v)
xvi
LIST OF SYMBOLS
GREEK LETTERS
LIST OF APPENDICES
INTRODUCTION
Cooper and his co-workers (1944) proposed that the kLa may be empirically
linked to the gassed power consumption per unit volume of broth (Pg/VL) and the
superficial air velocity (vg) as described by the following equation.
b
§ Pg ·
¸¸ v g
c
kLa a ' ¨¨ (1.1)
© VL ¹
2
In this equation, the values of the constants 'b' and 'c' may vary considerably,
depends on the bioreactor geometry and operating conditions. Data in Table 1.1
summarise the values of constant 'b' and 'c' from several works. Constant ‘b’
represents the level of dependence of kLa on the agitation, while, constant ‘c’
represents the level of dependence of kLa on the sparging rate applied to the system.
Table 1.1 Values of parameter 'b' and 'c' from several works that estimated from
the empirical relationship proposed by Cooper et al. (1944)
Author Constant Constant Type of Liquid Liquid
‘b’ ‘c’ impeller Model Volume
Cooper et al. 0.95 0.67 N/A Air-water 66 L
(1944) system
Shukla et al. 0.68 0.58 Disc turbine Air-water 5.125 L
(2001) and pitched system
blade
turbine
Shukla et 0.725 0.892 Disc turbine Yeast 5.125 L
al.((2001) and pitched fermented
blade broth
turbine
Badino Jr. et Flat-blade Aspergillus’s 10 L
al. (2001) 0.47 0.39 disc style fermented
turbine broth
Martinov & 0.84 0.4 Narcissus (2% w/v) 50 L
Vlaev (2002) blade CMC solution
Martinov & 0.82 0.4 Narcissus (0.5% w/v) 50 L
Vlaev (2002) blade Xanthan gum
solution
Arjunwadkar 0.68 0.4 Disc turbine (0.7% w/v) 5.125 L
et al. (1998) and pitched CMC solution
blade
turbine
speed (NDi), pump rate of impeller (Q), pump rate of impeller per unit volume (Q/V)
and Reynolds number.
1.2 Motivation
The oxygen transfer coefficient, kLa plays an important role towards carrying
out the design, scaling up and economic of the process. Efforts have been focused in
improving the design and scaling up studies to achieve adequate supply of oxygen at
higher scales (Martinov & Vlaev, 2001, Juarez & Orejas, 2001, Arjunwaadkar et al.,
1998). Their works employed the correlation proposed by Cooper et al. (1944) and
demonstrated the effects of agitation and aeration at different combination of
impellers in prediction of kLa values at the laboratory scales. The most commonly
methods in determining the kLa are the static and the dynamic gassing-out
techniques. As contrast to the static gassing-out technique, the live culture was used
in the dynamic gassing-out technique. Both of these techniques have been employed
by Martinov & Vlaev (2001), Juarez & Orejas (2001), Arjunwaadkar et al. (1998)
and Shukla et al. (2001).
Tragardh (1999) reported the influence of power input per unit mass on the
hydrodynamics of the bioreactor.
In order to achieve these objectives, the following scope of work shall be covered:
i) 16 liter and 150 liter at different viscosities namely 0.25, 0.5 and 1
%(w/v) of CMC solutions.
ii) 16 liter and 150 liter bioreactor at different temperatures namely 30o,
40o and 50 oC.
ii) 16 liter and 150 liter at different viscosities namely 0.25, 0.5 and 1
%(w/v) of CMC solutions.
iii) 16 liter and 150 liter bioreactor at different temperatures namely 30o,
40o and 50 oC.
6
LITERATURE REVIEW
In virtually all of these reactors, several phases are involved and substrates
and nutrients must be transferred from one phase to another. To be effective in
achieving the desired degree of conversion of reactants to products or supplying
sufficient nutrients for maintenance of cell viability, interphase heat and mass
8
transfer must occur to a sufficient extent. One of the key nutrients for all aerobic
cells is oxygen, which is sparingly soluble in water. Supply of oxygen from the gas
phase to the liquid phase is critical in most aerobic fermentations (Bailey and Ollis,
1986). These are to maximise the air sparging rate to place as much air (oxygen) into
contact with the liquid and also maximise the agitation of the liquid which will help
to break-up the air bubble to the smallest possible. Minimizing the bubble size,
maximise its surface area will minimise the thickness of the liquid film surrounding
the bubble and reduce the limiting factor of the oxygen transfer into liquid and
eventually to the organism (Michael, 1997) as illustrated in Figure 2.1. The transfer
of oxygen from gas bubble to cell is described in the following steps.
Figure 2.1 Steps for transfer of oxygen from gas bubble to cell (Doran, 1996)
The main components of the sensors are the oxygen permeable membrane,
the working electrode, the electrolyte solution and a possible reference electrode. A
voltage is applied between the gold (platinum or silver) cathode and the anode that
consists of either lead or silver (Ag/AgCl), and causes the oxygen to react
electrochemically. Increasing of the oxygen concentration will results a higher
electric current. The current in the sensor is measured and, after calibration,
converted into dissolved oxygen concentration. The reaction tends to produce
alkalinity in the medium together with a small amount of hydrogen peroxide. The
chloride ions are provided by the KCl electrolyte solution. For this reason and for
removing these alkaline hydroxide ions, the solution has to be replaced from time to
time (James, 1992). The other important parameter of the sensor is the response
time. It can be measured by making a step change in oxygen partial pressure in the
measurement medium and measuring the sensor response. The sensor can be
approximated as a first order system:
§ dc p ·
c cp W p ¨¨ ¸¸ (2.1)
© dt ¹
Where
c = oxygen concentration in the measurement sample
cp = oxygen concentration measured by the sensor
Wp = sensor time constant
The time constant, Wp is the time when the sensor response reaches 63.7% of
the ultimate response as shown in Figure 2.2. The significant of the sensor response
time has been reported by several authors namely Nielsen et al. (2003), Badino et al.
(2001) and Martinov & Vlaev (2002).
11
1.0 1.0
cp cp
1
c c Area above the curve
0.36 is Wp
Wp time time
(a) (b)
Figure 2.2 (a) Sensor response time measurement (b) Integral method for
measuring the sensor time constant
Figure 2.3 Concentration gradient for gas-liquid oxygen transfer (Doran, 1996)
Oxygen is transferred from the gas phase into the liquid phase. The
concentration of oxygen in the bulk liquid phase is CAL and at the gas-liquid interface
in the liquid is CALi. The concentration of oxygen in the bulk gas phase is CAG and at
the gas-liquid interface in the gas is CAGi. The rate of mass transfer of the oxygen
through the gas boundary layer is:
Similarly, the rate of oxygen transfer through the liquid boundary is:
where
kG = gas phase oxygen transfer coefficient
kL = liquid phase oxygen transfer coefficient
Incorporating the equilibrium relationship into 2.2 and 2.3 we obtain the results:
§ 1 m ·
N A ¨¨ ¸¸ C AG mC AL (2.5)
k
© G a k a
L ¹
and
§ 1 1 · C AG
N A ¨¨ ¸¸ C AL (2.6)
© mk G a k L a ¹ m
1 1 m
(2.7)
KGa kG a kLa
Similarly, the overall liquid phase oxygen transfer coefficient, KL is defined by the
equation:
1 1 1
(2.8)
KLa mk G a k L a
§C ·
NA K L a¨ AG C AL ¸ (2.9)
© m ¹
C AG
C * AL (2.10)
m
Where
NA = rate of oxygen transfer per unit volume of fluid (g mol/m3s)
kL = liquid-phase oxygen transfer coefficient (m/s)
a = gas-liquid interfacial area per unit volume (m2/m3)
CAL = actual dissolved oxygen concentration in the broth (g mol/m3s)
C*AL = oxygen concentration in the equilibrium with the gas phase
(g mol/m3s), also known as the solubility
The difference (C*AL – CAL) between the maximum possible and actual
oxygen concentrations in the liquid represents the concentration difference driving
force for mass transfer.
15
The effects of the bubble size have been investigated by Al-Masry (1999).
The most important property of air bubbles in the bioreactor is their size. More
interfacial area (a) is provided if the gas is disintegrated into small bubbles and
provide a better gas dispersion in the bioreactor. As illustrated in Figure 2.4, the
flow patterns in agitated air sparged bioreactors is a function of both impeller speed
(N) and the air sparging rate (Q), both of which also affect the bubble size and the
number of bubbles. Therefore this, affects the interfacial area (a) available for mass
transfer. For a given volume of gas, more interfacial area (a) is provided if the gas is
dispersed into many small bubbles rather than a few large ones. Since the efficiency
of oxygen transport is approximately proportional to the ratio of the bubble surface
area to the bubble volume, the smaller size of the micro bubbles increased oxygen
transfer rate in the bioreactor. In addition, smaller bubbles have a longer dwell time
in liquid because of their slower bubble-rise velocities, allowing more time for the
oxygen to dissolve (Doran, 1996).
The total pressure and the oxygen partial pressure used during the aeration of
the broth affect the value of the solubility of the dissolved oxygen. For solutions, the
equilibrium relationship between these parameters follows Henry’s Law.
where
PAG = partial pressure of oxygen (kPa)
PT = total pressure of the system (kPa)
yAG = mole fraction of oxygen in the gas phase (dimensionless)
H = Henry’s Law constant (kPa.L/mg)
*
C AL = solubility of oxygen in the liquid (mg/L)
17
PA P
CA yA (2.14)
H H
Where
CA = concentration in the liquid phase (mg/L)
PA = gas phase partial pressure of oxygen (atm)
H = Henry’s coefficient (atm-liter/mg)
yA = mole fraction of oxygen in the gas phase
b
§P ·
a ' ¨¨ g ¸¸ v g
c
kLa (2.15)
© VL ¹
Where the values of the constants b and c may vary considerably, depends on the
system geometry, the range of variables covered and the experimental methodology
used.
19
Gas flow out
Oxygen out
LIQUID
PHASE
Oxygen transfer
= KLa (C* - C)
GAS
PHASE Oxygen absorbed by
organism
= x Q O2
Gas flow in
Oxygen in
Figure 2.5 Mass balance of oxygen transfer during aerobic fermentation (Charles
& Wilson, 1994)
Air Off
CL N2 On
Air On
N2 Off
In practical, experiments are carried out commencing from an oxygen free liquid and
the oxygen concentration will rise such that:
dC L
NA (2.16)
dt
Assuming that the liquid is well mixed and there is no oxygen uptake. Combining
both Equation 2.12 and 2.16 we obtain:
dC L
dt
kL a C * CL (2.17)
Basically, C* was assumed to be constant and integrating both side of Equation 2.17
with respect to time in order to calculate the kLa value:
CL t
1
³o C * C L dC L k L a ³ dt
0
(2.18)
C L
1 § C* CoL ·
kLa ln¨ ¸¸ (2.19)
t ¨© C * C L ¹
21
dC L
rO2 C x (2.20)
dt
Where
CL = actual dissolved oxygen concentration in the broth
(g mol/m3s)
rO2 = specific respiration rate (mmoles O2/g cells.h)
Cx = dry weight of cells per volume (g cell mass/m3)
After a time, the air supply is switched back on and the oxygen concentration will
rise again until it returns to the initial steady state value. The oxygen material
balance in an aerated batch bioreactor with growing organisms is given by:
22
dC L
dt
k L a C * C L rO2 C x (2.21)
Where
kL a = liquid-phase oxygen transfer coefficient (hr-1)
C* = oxygen concentration in the equilibrium with the gas phase
(g mol/m3s), also known as the solubility
ª§ 1 ·§ dC L ·º
CL C * «¨¨ ¸¸¨ rO2 C x ¸» (2.22)
¬© k L a ¹© dt ¹¼
The term, rO2Cx, can be obtained by measuring the slope of the CL v/s time
curve in Figure 2.7. Therefore, from Equation 2.22, the plot of CL versus (dCL/dt
+rO2Cx) will results in a straight line which has the slope of (-1/kLa) and the y-axis
intercept of C*. Similar dynamic gassing-out technique was performed by Badino et
al. (2001) and Garcia-Ochoa et al. (2000).
Air Off
CL
dC L
Air On
dt
NDi2 U L
N Re (2.23)
PL
The Reynolds number describes the flow only at periphery of the stirrer. To
distribute the turbulence homogenously within the entire reactor, an impeller of
appropriate shape and diameter must be used. The flow rate is crucial to the
distribution of turbulence. However, as turbulence may damage filamentous
organisms, there are frequently limitations on how fast a system can be stirred. The
24
NDi2 U L
N Re (2.24)
P app
Rushton et al. (1950) summarised that the power consumption for a given
system cannot be predicted theoretically, therefore empirical correlations have been
developed to predict the power required. The power consumption for stirring non-
aerated fluids depends upon fluid properties, UL and PL, the stirrer rotation rate, N
and the impeller diameter, Di. The latter is expected to vary with impeller Reynolds
number in a different manner for each flow regime: laminar, transition or turbulent.
Ungassed power consumption (Po) was obtained through a graph of power number
(Np) as a function of Reynolds number (NRe) for both Newtonian and non-Newtonian
fluid in a different type of flow regime.
§ Po ·
¨ ¸ Np (2.25)
¨ U N 3D ¸
5
© L i ¹
25
Where
Po = power in ungassed system (W)
Ni = impeller speed (rps)
UL = liquid density (kg/m3)
Di = impeller diameter (m)
The power number was correlated with Reynolds number for several types of stirrer
as shown in Figure 2.8. In the turbulent regime, the Reynolds number is large,
inertial forces dominate viscous forces and thus the dependence of the power number
on the Reynolds number will vanish. Therefore power input is independent of
Reynolds number.
Po v N i3 Di5 (2.26)
Where Np = constant
1
Po v N i2 Di3 or N p v (2.27)
Re
N 2 Di
N Fr (2.28)
g
Where
g = gravitational acceleration, 9.81 m s-2
26
Figure 2.8 Power number Vs Reynolds number for various impeller geometries
(Aiba et al., 1973)
The power required to agitate gassed liquid systems is less than for ungassed
liquids since the apparent density and viscosity of the liquid phase decrease upon
gassing. When gas is introduced into an agitated tank, the bulk liquid density and
viscosity in the tank decrease as a result of the formation of gas dispersion. This
alteration of the fluid properties is reflected in changes to the power required for
agitation. At low impeller speeds, the gas rises through the impeller region without
being effectively dispersed. This condition is known as flooding. At higher impeller
speeds beyond the point of flooding, there is a reduction in the power drawn by the
impeller due to the change in the overall liquid density. This reduction in power has
been correlated with the aeration number. The reduction in gassed system power,
Pg/Po is generally given as a function of the ratio of the superficial air velocity to the
impeller tip speed (Bailey and Ollis, 1986).
Pg
f N A (2.29)
Po
27
Q
Where aeration number, N A (2.30)
NDi3
4Q
and superficial air velocity, v g (2.31)
SDt2
0.45
§ P 2 ND 3 ·
Pg m¨¨ o 0.56 i ¸¸ (2.32)
© Q ¹
Where
m = constant (depends on impeller type and geometric form)
Q = volumetric gas flow rate (m3/s)
The degree of agitation has been demonstrated to have profound effect on the
oxygen transfer efficiency of an agitated bioreactor. Banks (1977) claimed that
agitation assisted oxygen transfer in many ways. Agitation increases the area
available for oxygen transfer by dispersing the air in the culture fluid in the form of
small bubbles. Agitation also delays the escape and prevents coalescence of air-
bubbles. Banks (1977) also reported that agitation decreases the thickness of liquid
film at gas-liquid interface by creating turbulence in the culture fluid. Aeration
supplies the necessary oxygen to the microorganism agitation maintains uniform
conditions within the bioreactor. Another reason for air sparging and mechanical
28
mixing in bioreactor is to remove carbon dioxide and other possible toxic gaseous
metabolic byproducts which are produced in the broth (Hensirisak, 1997).
Geankoplis (1993) has proved the importance of baffles in increasing the turbulence
within bioreactor and prevent entrainment of bubbles in the vortex that would
damage the cells.
Martinov & Vlaev (2002), Shukla et al. (2001) and Arjunwadkar et al. (1998)
demonstrated the influenced of aeration and agitation on oxygen transfer of
bioreactor at various scale. Hensirisak (1997) observed that direct sparging of gases
into a stirred bioreactor may be most critical than agitation to the health of cultured
cells. In most cases, the impeller must provide enough mixing to keep cells or gasses
homogenously while creating as little fluid force as possible. These effects need to
be minimized by reducing the impeller tip speed and power input per unit volume as
much as feasible. Stanbury and Whitaker (1984) reported the flooding phenomenon
occurred at high flow rates will results in extremely low oxygen transfer rates. Chia-
Hua Hsu (2003) addressed that, for mechanical agitated bioreactor, the air flow rate
employed rarely falls outside the range of 0.5 to 1.5 volume of air per volume of
medium per minute (vvm).
Bailey and Ollis (1986) highlight the influence of broth rheology on power
consumption, mixing, heat and mass transfer rates. Norwood (1960) claimed that
non-Newtonian viscosity can have very important practical effects on bulk flow and
on heat and mass transfer. For an example, in a stirred bioreactor, the shear rate is
highest near the impeller and decrease rather sharply with distance from it. The
simulated pseudoplastic broth employed in this work was to demonstrate the
deviation of CMC solution from the Newtonian fluid behaviour. Theoretically, the
deviation of the pseudoplastic fluid was shown in Figure 2.9. The behaviour of
pseudoplastic fluid has been successfully investigated by Garcia-Ochoa et al. (2000),
Martinov & Vlaev (2002), Shukla et al. (2001) and Arjunwadkar et al. (1998).
29
Newtonian Fluids
Deviation
Shear rate, V
Pseudoplastic Fluids
Shear stress, W
Their work described the flow behaviour and determined the power-law
quantities by employing the Oswald-de Waele model via Equation 2.33 (Garcia-
Ochoa, 2000).
W kJ n (2.33)
Where
W = shear stress (N/m2)
J = shear rate (s-1)
k = consistency index (Pa.sn)
n = flow behaviour index (-)
Power law models are very useful from the engineering stand point,
especially when compared to the non linear and unquantified multi parametric
equations of state which has been developed for molecular considerations. However,
power law models fail to predict the Newtonian behaviour frequently observed at
very high and very low shear rates and the equations are not dimensionally sound.
Different composition of broth has different degrees of psedoplasticity and the value
of effective shear rate in the bioreactor was determined according to the equation
proposed by Metzner and Otto (1962):
J AN (2.34)
30
k >AN @
n 1
P app (2.35)
Where
A = Metzner and Otto constant (depend on impeller type). It was assume to
be 11.5, for Rushton turbine (Garcia-Ochoa et al., 2000) and 10, for
marine impeller (Nagata, 1975)
N = impeller speed (rpm)
It was found that the apparent viscosity (Papp) decreases with increasing of
shear rate for pseudoplastic fluids as reported by Bailey and Ollis (1986).
Bailey and Ollis (1986) discovered that optimal process conditions found at
laboratory scale may not be optimal in larger bioreactors. There is a serious scaling
problem. Hensirisak (1997) defined the scale-up as a procedure for the design and
construction of a large scale system on the basis of a result of experiments with small
scale equipment. Kossen and Oosterhuis (1985) divided the approach for scale-up
into four widely recognized steps namely fundamental methods, semifundamental
methods, dimensional analysis and rules of thumb.
Table 2.1 Different scale-up criteria and their consequences (Kossen and
Oosterhuis, 1985)
Scale-up Criterion Value at 10 m3 ( V = 10 L )
P P/V N ND Re N/D
Equal P/V 103 1 0.22 2.15 21.5 0.022
Equal N 105 102 1 10 102 0.1
Equal tip speed 102 0.1 0.1 1 10 10-2
Equal Re number 0.1 10-4 10-2 0.1 1 10-3
Equal shear to flow ratio 108 105 10 102 103 1
Scale-up on the basis of constant impeller tip speed would be obviously the
method of choice when an organism sensitive to a mechanical damage and shear rate
was the most important consideration. The work of Steel and Maxon (1962,1966)
has suggested that this method may be applicable to the viscous non-Newtonian
cultures of filamentous microorganisms, which are of such great industrial
importance. Scale-up on the basis of constant measured volumetric transfer
coefficient is time consuming in that it does not allow for prediction of results,
although it can be used in conjunction with other methods (Aiba et al., 1965) to
obviate this difficulty. The fact that none of the commonly employed physical or
chemical techniques for determination of kLa are suitable for use with pilot and
33
This criterion is important and most commonly used to evaluate the aeration
efficiency in aerobic fermentation where supplying sufficient oxygen is the intention
to satisfy the need of the microorganism as reported by Hubbard et al. (1994), Herbst
& Schumpe (1992) and Ju & Chase (1992). In this method, aeration efficiency is
measured on the small scale under conditions, which have been previously
established as optimal for product formation by using one of the recognized
techniques for determination of kLa. Employing the similar technique, conditions are
then found by experiment on the large scale, which will support the same aeration
efficiency. Shukla et al. (2001) successfully demonstrated the scale-up of bioreactor
on a basis of constant kLa in the biotransformation medium. For the scaling up of
aerobic fermentation, the effect of gas liquid mass transport is the most significant
factor (Hubbard et al., 1994). Therefore, scale-up in aerobic fermentation is often
performed on the basis of keeping the value of kLa constant.
34
Scale-up on the basis of constant power input per unit liquid volume is widely
practiced. This method is based on assumed proportionality between power
consumption per unit liquid volume and aeration efficiency. Such proportionality is
almost certainly of more limited application than is generally realized, even under of
fully turbulent flow. Unfortunately, it is not always possible to work under
conditions of turbulent flow. When highly viscous cultures (e.g. fungi and
streptomycetes) are employed, it becomes impossible to maintain turbulent flow even
at very high agitator shafts (Mohamad et al., 2001). Geankoplis (1993) stated that, if
the same power consumption is obtained on both scales of operation, and assuming
proportionality between power consumption per unit liquid volume and aeration
efficiency, then similar aeration efficiency should result on both scales.
Po N p U L N 3 Di5 (2.36)
0.45
§ P 2 ND 3 ·
Pg m¨¨ o 0.56 i ¸¸ (2.37)
© Q ¹
It can be shown that by combining both Equations 2.36 and 2.37, it will yield an
expression for both ungassed and gassed power consumption. The divergence
between values of the exponent on N and Di in the expressions is sufficiently small
to be ignored for the practical purpose.
0.45
§ N p2 U L2 N 7 Di13 ·
Pg m¨ ¸ (2.38)
¨ Q 0.56 ¸
© ¹
35
Pg
mN U 0.9
p
0.9
L N 3.15 Di5.85
(2.39)
Q 0.252
Pg1 Pg 2
(2.40)
V1 V2
The subscripts 1 and 2 refer to the small and large scales respectively. If
geometrically similar vessels are employed in scale-up, the value of the geometry
dependent constant will be the same for both scales, thus:
Pg1 Pg 2
2
(2.41)
SD H T 1
T1 SDT22 H T 2
4 4
Pg1 Pg 2
(2.42)
DT21 H T 1 DT22 H T 2
Substituting Pg1 and Pg2 for both scales into Equation 2.42.
mN U 0.9
p
0.9
L N 13.15 Di51.85 mN U 0.9
p
0.9
L N 23.15 Di52.85
(2.43)
Q10.252 DT21 H T 1 Q20.252 DT22 H T 2
The following equation is used for scaling-up on basis of constant power input per
unit liquid volume:
v g1 vg 2 (2.45)
4Q1 4Q2
(2.46)
SDT21 SDT22
Thus, with geometrically similar vessels, the required volumetric air flowrate is:
2
§D ·
Q2 Q1 ¨¨ T 2 ¸¸ (2.47)
© DT 1 ¹
37
The basis of this method is that the impeller tip speed is maintained at a
constant value during scale-up. Thus:
Where
V’i= impeller tip speed, (S N Di).
§D ·
N2 N 1 ¨¨ i1 ¸¸ (2.50)
© Di 2 ¹
METHODOLOGY
This procedure was performed in both 16 liter and 150 liter scale. During
start-up period, the bioreactor was allowed to run at impeller speed of 600 rpm and at
air flowrate of 9 l/min in the 16 liter scale. At 150 liter scale, the bioreactor was
39
warmed up at impeller speed of 150 rpm and the air flow rate was set at 30 l/min.
The bioreactor was warmed up for at least 30 minutes before performing any
experimental works. The bioreactor was operated when all the operating parameters
such as temperature, pH, dissolved oxygen concentration, impeller speed and air
flow rate are stable.
The geometry of the bioreactor and the design of the impeller used in the bioreactor
are illustrated in Figure 3.1 and Figure 3.2, respectively.
DT
Di
'i
HT
'C
DS
Di
Di
Figure 3.2 Type of agitator (a) Marine impeller (b) Rushton turbine
41
Table 3.2 Operating conditions and techniques to determine the oxygen transfer
coefficient (kLa) reported in several works
Impeller speed Air flow rate Technique Used
Cooper et al. N/A N/A Sulfite oxidation
(1944)
Shukla et al. 50 – 300 rpm 0.293 – 1.56 vvm Dynamic gassing
(2001) out
Badino Jr. et al. 300 – 700 rpm 0.2 – 1 vvm Modified dynamic
(2001)
Martinov & 0.1 < Pg/VL < 2 3.3 x 10-3 < vg < 6.6 Static gassing out
Vlaev (2002) kW m-3 x 10-3 ms-1
Arjunwadkar et 400 – 750 rpm 0.29 – 0.975 vvm Dynamic gassing
al. (1998) out
The operating variables for the experimental work in the 16 liter scale vessel
were given in Table 3.3. For each combination of impeller speeds and air flow rates,
the experiments were performed on distilled water at temperature of 30oC, 40oC and
at 50oC. The initial pH was set to operate at 7 + 0.03 for the entire experiment. The
liquid viscosity was increased by dissolving the Carboxy Methyl Cellulose to
obtained the concentrations of 0.25%(w/v), 0.5%(w/v) and 1%(w/v) at 30oC. The
experiment was repeated by using the marine impeller for comparison. The physical
properties of CMC solution and distilled water are given in Table A.2 and A.3 in
Appendix A 1, respectively.
42
The probe response time was determined to correct the transmission delay
and the lag in response of polarographic membrane oxygen probe used. This was
done on 10 liter cell-free distilled water in 16 liter bioreactor. The time constant
given in the specifications was less than 45 seconds for approximately 98% of
dissolved oxygen saturation concentration.
The impeller was set at 600 rpm and the air supply was set at 9 l/min of air.
The bioreactor was ensured to run at a steady temperature of 30oC. The change in
the dissolved oxygen concentration was allowed to reach 100% of saturation value.
Then, both of the agitation and aeration were switched off and the inlet valve of the
nitrogen tank was opened immediately until the oxygen probe indicated complete
removal of oxygen. Then, the inlet valve of nitrogen tank was closed and both
impeller and air supply was turned on to allow 100% oxygen saturation.
The rate of oxygen transfer from air bubbles to a liquid in a batch stirred
bioreactor was given by the following relationship.
dCL
k L a(C * CL ) (2.17)
dt
The oxygen transfer rate (OTR) was determined by implementing 'the static gassing
out' method as explained by Stanbury and Whitaker (1984). The change in dissolved
oxygen concentration (CL) in the liquid phase was detected by using a polarographic
oxygen probe. The nitrogen tank was connected into the feed port of the bioreactor
and was set to deliver an inlet pressure of 10 psig.
The experimental work was started after the dissolved oxygen concentration
reached 100% saturation value. The aeration of the bioreactor was interrupted by
switching off the air flow and the medium was purged with nitrogen until 0% oxygen
saturation was reached. As the dissolved oxygen concentration reaches below 10%
of saturation value, the air flow was restarted and the inlet valve of the nitrogen tank
was closed. Simultaneously, the chart is marked to show time zero. The oxygen
tension was allowed to increase and approaches 100% saturation value. When the
dissolved oxygen tension stabilizes at 100% saturation value, the above steps were
repeated for reproducibility checking. The delayed response time was offset by the
electrode response time.
In order to calculate the kLa, Equation 2.17 was firstly integrated with respect
to the time taken for the oxygen concentration to reach the saturation level from the
lowest point.
1 § C* C oL ·
t ln¨ ¸ (3.1)
k L a ¨© C * CL ¸¹
44
Then, the kLa value is determined by reciprocating the slope obtained from the semi
§ C* C oL ·
logarithmic plot of time (t) versus ¨¨ * ¸¸ . Note that the dissolved oxygen
© C CL ¹
electrode records oxygen concentration as percentage of maximum dissolved
oxygen. In this research, the dissolved oxygen saturation concentration in the liquid
or C* that calculated from the Henry's Law was quoted from the table that presented
in Perry and Green (1997). The oxygen saturation concentrations are presented in
Table A.4 in Appendix A1.
The scale-up protocol applied involved the application of rule of thumb, trial
and error, interpolation and extrapolation on the basis of keeping the value of kLa
constant as the scale increases. The protocol was summarised in Figure 3.3. The
scale-up performed by firstly, investigate the kLa values in 16 liter vessel. Upon
obtaining the kLa values in the 16 liter scale, the limitations for the operating
variables in the 150 liter bioreactor were computed. In order to design the
operational conditions at 150 liter scale, scale-up on the basis of constant power
consumption per unit liquid volume, Pg/VL, constant superficial velocity, vg and
constant impeller tip speed, SNDi was performed using the scale-up equations.
45
Investigation
of kLa at 16 liter scale
Interpolation and
extrapolation to achieve
the same value of kLa
No
Similar kLa value ?
Yes
Figure 3.3 Scale-up protocol based on constant oxygen transfer coefficient, kLa
46
2
§D ·
Q2 Q1 ¨¨ T 2 ¸¸ (3.3)
© DT 1 ¹
§D ·
N2 N 1 ¨¨ i1 ¸¸ (3.4)
© Di 2 ¹
By knowing the impeller speeds, N1 and air flow rates, Q1 at 16 liter scale,
Equation 3.2, 3.3 and 3.4 was used to determine the impeller speeds and air flow
rates at 150 liter scale. At different scale-up criterions, the allowable operating range
at 150 liter vessel was predicted. Incorporating Equation 3.3 and Equation 3.4 into
Equation 3.2 will yield the scale-up equations in predicting the impeller speeds and
the air flow rates at 150 liter scale. The equations are as follows:
(1) Constant power consumption per unit liquid volume, Pg/VL with constant
superficial velocity, vg.
2
§D ·
Q2 Q1 ¨¨ T 2 ¸¸ (3.5)
© DT 1 ¹
47
§ 1 ·
N2
>N 1
3.15
Di1
5.85
Q2
0.252 2
DT 2 H 2 @¨ ¸
© 3.15 ¹
(3.6)
Q 1
0.252 2
DT 1 H 1 Di 2
5.85
(2) Constant power consumption per unit liquid volume, Pg/VL with constant
impeller tip speed, SNDi.
§D ·
N2 N 1 ¨¨ i1 ¸¸ (3.7)
© Di 2 ¹
§ 1 ·
Q2
>N 2
3.15
Di 2
5.85
Q1
0.252 2
DT 1 H 1 @ ¨ ¸
© 0.252 ¹
(3.8)
N 1
3.15 2
DT 2 H 2 Di1
5.85
The operating variables (N2 and Q2) achieved in solving Equation 3.5, 3.6, 3.7
and 3.8 was used a base line in ‘trial-and-error’ step upon scaling-up a bioreactor
from 16 liter to 150 liter on a basis of constant kLa. The ‘trial-and-error’ loop in
determination of the operating conditions at 150 liter scale was shown in Figure 3.4.
Figure 3.4 The ‘trial-and-error’ loop at 150 liter scale in the scale-up protocol
48
(1) Interpolation
Operating conditions (N or Q) kLa value attained
x1 y1
x y
x2 y2
ª§ y y1 · º
x «¨¨ ¸¸x 2 x1 » x1 (3.9)
¬© y 2 y1 ¹ ¼
(2) Extrapolation
kLa value attained Operating conditions (N or Q)
x1 y1
x y
§ y1 ·
y ¨¨ ¸¸ x (3.10)
© x2 ¹
The scaling-up factor upon scaling-up based on constant kLa from 16 liter to
150 liter bioreactor were calculated by using the equations below:
N2 Q2
R1 (3.11) R2 (3.12)
N1 Q1
where subscripts 1 and 2 for impeller speed, N and air flow rate, Q
refer to small (16 liter) and large (150 liter) scales respectively.
49
With the scale-up criteria selected (kLa), superficial air velocity, vg and the
specific power data, Pg/VL were correlated using the equation proposed by Cooper et
al. (1944) via Equation 3.5.
b
§P ·
a ' ¨¨ g ¸¸ v g
c
kLa (3.13)
© VL ¹
§ Po ·
¨ ¸ Np (3.14)
¨ U N 3D ¸
5
© L i ¹
The Reynolds number (NRe) was calculated from the following equation.
§ ND i 2 U L ·
N RE ¨ ¸ (3.15)
¨ PL ¸
© ¹
NDi2 U L
N Re (3.16)
P app
0.45
§ Po 2 NDi 3 ·
Pg m¨¨ 0.56
¸
¸ (3.17)
© Q ¹
Where m depends on the impeller geometry for which in this case, the value of m
is 0.832 for both disc turbine impeller and marine impeller (Badino Jr. et al., 2001).
As for the superficial air velocity (vg), it was calculated by using the equation below:
4Q
vg (3.18)
SDt2
Note that, these calculations were performed for both type of impellers at 16 liter
and set of impeller speeds and air flow rates attained at 150 liter from interpolation
and extrapolation upon scale-up on basis of constant kLa. In order to determine the
51
dependence of kLa on stirrer speed and aeration, the following correlations should be
firstly determined.
Further analysis was done by constructing a logarithmic plot of kLa values with
respect to impeller speed, Pg/VL and gas superficial velocity, vg to observe the effect
of agitation, aeration and power consumption in bioreactor on kLa.
k >AN @
n 1
P app (3.20)
Where A, for turbine stirrer type value was assumed to be 11.5 (Garcia-Ochoa et al.,
2000) and value for marine impeller was assumed to be 10 (Nagata, 1975). The
calculation for the apparent viscosity, Papp was performed at different stirrer speeds
as previously mentioned.
The trend in Figure 3.6 demonstrates that the decrease of shear rate (J) at
different spindle speeds proves that the CMC exhibits pseudoplastic characteristic.
This characteristic matches with the rheology of the filamentous cultures as cited in
Brooke & Halsall (2002). This was easily analyzed by plotting the shear rate (J)
against the shear stress (W) for different concentration of CMC concentrations as
illustrated in Figure 3.6. As shown in Figure 3.6, all the CMC solutions showed a
shear thinning and non-Newtonian pseudoplastic behaviour. As would be expected,
increases in the concentration of CMC lead to an increase in the apparent viscosity of
the solutions. The pseudoplastic nature of the CMC solutions is caused by
interactions between the CMC molecules. The analysis on the CMC solutions is
given in Appendix B.
54
0.07
0.03
0.02
0.01
0
0 0.2 0.4 0.6 0.8 1 1.2
shear rate, J
Figure 3.6 Viscosity (kg/m.s) change with shear rate (s-1) for CMC solution
The Power Law indices of the CMC solutions, shows that with more CMC
present, the power law index of the solutions decreases. So as the amount of CMC
was increased, the solution becomes more non-Newtonian. This is illustrated in
Figure 3.7.
55
1.2
0.8
Shear rate, J (s )
-1
0.25%(w/v) CMC
0.4
1%(w/v) CMC
0.2
Newtonian Fluids
0
0 0.01 0.02 0.03 0.04 0.05 0.06
2
Shear stress, W (N/m )
Figure 3.7 Deviation from Newtonian behaviour due to CMC presence in the
fluid at 30oC
3.7.1 Microorganism
The fermentation process was carried out by using the Recombinant E.coli
ARP012 without the protein expression. Stock cultures were maintained on LB agar
slants containing 10 g/L bactotryptone (Merck, Germany), 5 g/L sodium chloride
(Merck, Germany), 5 g/L yeast extract (Merck, Germany) and 15 g/L agar (Hamburg
Chemicals, Germany). The prepared slants were maintained at 4oC and subcultured
every three months.
56
Batch fermentations were carried out in the 16 liter bioreactor with total broth
volume of 11.5 liter. The medium was autoclaved in situ at 121oC for 20 minutes.
0.5 liter of glucose solution was sterilized separately and was pumped in after the
vessel cooled down. 1 liter of seed culture was transferred into 10 liter bioreactor.
6M NaOH, 2M H2SO4 and silicon-based antifoam was prepared and connected to the
bioreactor. Throughout the fermentation, the medium pH was maintained at 7.0 +
57
0.1. The temperature was controlled at 37 oC and the dissolved oxygen concentration
was maintained above 20% saturation value. Antifoam was added manually when
foaming occurred.
The following mass balance equation was used for the dissolved oxygen in
batch fermentation:
dC L
dt
k L a C * C L rO2 C x (3.21)
where the first term on the right hand side of Equation 3.21 is the oxygen transfer
rate (OTR) and the second term is the oxygen uptake rate of the culture (OUR). The
measurement of OUR and OTR was made using dynamic technique proposed by
Taguchi and Humphrey (1966) in two stages. In the first stage, the inlet of airflow
was shut down and a decrease of oxygen dissolved concentration due to cellular
respiration was observed, which was recorded by a polarographic oxygen probe. The
58
OUR was determined by change in the dissolved oxygen concentration after stopping
air flow. In this condition, Equation 3.21 can be simplified to:
dC L
rO2 C x (3.22)
dt
In the second stage, the inlet of airflow to the bioreactor was restarted at
predetermined values and will increase the dissolved oxygen concentration. Under
these conditions, Equation 3.21 can be rearranged to result in a linear relationship as:
ª§ 1 ·§ dC L ·º
CL C * «¨¨ ¸¸¨ rO2 C x ¸» (3.23)
¬© k L a ¹© dt ¹¼
From Equation 3.23, the plot of CL versus (dCL/dt +rO2Cx) will result in a
straight line which has the slope of (-1/kLa) and the y-axis intercept of C*. This
technique was employed at different combinations of impeller speed and airflow
rates as mentioned previously. The calculation for the static and the dynamic
gassing-out technique is shown in Appendix C1 and Appendix C2, respectively.
59
Constant kLa was selected as scale-up criteria in this work. Media and
inoculum protocol were as in 16 liter bioreactor. Steps for the E.coli fermentation at
150 liter scale were illustrated in Figure 3.8. The scale-up was done according to the
previous scale-up protocol employed using cell-free distilled water and CMC
solutions. The impeller speeds and air flow rates at 150 liter scale were determined
using the scale-up ratio attained. Upon achieving the scale-up ratio for impeller
speeds and air flow rates, the operating conditions for E.coli fermentation was
designed. The operating conditions for E.coli fermentation was shown in Table 3.7.
INOCULUM DEVELOPMENT
Seed Bioreactor
Stock Culture Shake Flask 16 liter
Production Bioreactor
150 liter
4.1 Introduction
The result shows that the oxygen transfer coefficient, kLa was a strong
function of the agitation in both Rushton turbine and marine impeller. This is
illustrated in Figure 4.1 and 4.2. Based on the trend achieved in Figure 4.1 and 4.2, it
was seen that agitation rate in the bioreactor has a proportional effect on kLa. This
shows that an increase on the impeller speed will increase the oxygen transfer rate in
the bioreactor. As illustrated in Figure 4.2, a very low kLa values (lower than 0.01)
was attained especially in the CMC solution at low agitation rate of 300 rpm. It
proved that the effect of agitation was hardly notice at agitation below than 300 rpm.
Under this condition, the agitation system was incapable of maintaining the turbulent
flow conditions and hence, unable to enhanced the oxygen transfer in the bioreactor.
Nevertheless, the oxygen transfer rate increase from 20% to 40% as the agitation rate
was increased. The results obtained successfully confirmed that the fact of increase
in impeller speed will increase the kLa in the bioreactor as cited by Martinov &
Vlaev (2002).
63
0.1
R2R = 0.97
kLa (s )
-1
R2M = 0.96
R2M = 0.93
Rushton 30°C
R2M = 0.91 Rushton 40°C
Rushton 50°C
Marine 30°C
Marine 40°C
Marine 50°C
0.001
100
200 400 600 800 1000
Impeller speed, N (rpm)
0.1 1
1% - Rushton
0.5% - Rushton
0.25% - Rushton
1% - Marine
0.5 - Marine
0.25% - Marine
0.1
kLa (s ) Rushton
kLa (s ) Marine
0.01
-1
-1
R 2R = 0.92 0.01
R 2R = 0.90
R 2R = 0.99
R2M = 0.93
R2M = 0.97
R 2M = 0.98
0.001 0.001
100 200 400 600 800 1000
Impeller speed, N (rpm)
The increase of the kLa due to the agitation and sparging was due to the
decrease of bubble size and subsequently led to the increase of the specific interfacial
area resulting in higher interfacial contact between the gaseous and liquid phase for
the oxygen transfer. Increase in residence time of the bubbles in the liquid due to
agitation also contributed to increase kLa. Agitation delays the escapes of air bubbles
from the liquid, breaks the bubbles and prevents it from coalescence. Most
importantly, higher agitation provides a better oxygen transfer because its decreases
the thickness of the gas liquid film at the gas-liquid interface by creating turbulence
in the culture. This is consistent with what was reported in the literature review
where agitation and aeration rates have a proportional effect on kLa (Arjunwadkar et
al., 1998, Shukla et al., 2001, and Martinov & Vlaev, 2002).
65
0.1
R2R = 0.97
R2R = 0.90
R2R = 0.84
kLa (s )
-1
0.01
R2M = 0.88
R2M = 0.84
Figure 4.3 Dependence of kLa on air flow rate, Q at different temperature for
Rushton turbine and marine impeller
0.1 0.1
R2R = 0.83
kLa (s ) Rushton
kLa (s ) Marine
R2R = 0.98
R2M = 0.90
R2R = 0.97
-1
0.01 0.01
-1
R2M = 0.87
R2M = 0.94
1% - Rushton
0.5% - Rushton
0.25 - Rushton
1% - Marine
0.5% - Marine
0.25% - Marine
0.001 0.001
1 10 100
Air Flow Rate, Q (l/min)
Figure 4.4 Dependence of kLa on air flow rate, Q at different viscosities for
Rushton turbine and marine impeller
66
Increase of temperature improve the oxygen transfer rate and reduce the
oxygen solubility as tabulated in Table A.4 in Appendix A1. Investigation on the net
effect was found to be crucial and was performed at different type of impellers
namely Rushton turbine and marine impeller by varying the temperature from 30oC
to 50oC. The increase of kLa in the bioreactor by operating at higher temperature in
Rushton turbine and marine impeller at different agitation rates are presented in
Table 4.1. A proportional increase of temperature and kLa was also achieved at
different air flow rates for both Ruhton and marine impellers as presented in Table
4.2.
It is demonstrated from the results tabulated in Table 4.1 and 4.2 that the kLa
value increases with increase of temperature. Oxygen solubility in water decreases
from 7.55 mg/L to 5.61 mg/L as the temperature is increased from 30oC to 50oC.
This is clearly observed in Table A.4 in Appendix A1. Although the oxygen
solubility significantly decreases at higher temperature, the net increase of the kLa
observed in this experiment was due to the improved oxygen transfer through a series
of transport resistances between the bubbles and the liquid. Our data were in
agreement with the report published by Nielsen et al. (2003) because of the increase
in the oxygen transfer coefficient, kLa.
In order to provide sufficient oxygen in the liquid phase, the system will
require greater amounts of oxygen to saturate. One way to accomplish this was to
increase the aeration rate in the bioreactor. However, by referring to Equation 2.22,
enhanced of oxygen transfer rate was due to an increase in dCL/dt, which result to an
increase in kLa or an increase in the driving force term, C* - C. The difference in
oxygen solubility in the liquid phase is the driving force that increases the oxygen
transfer rate in bioreactor. Though the pattern of increase in temperature with
respect to kLa were comparable for Rushton turbine and marine impeller, the effect
was insignificant compared to the effect of other operational parameters namely
agitation, liquid viscosity and the volumetric power consumption on kLa.
The gas-liquid interface is one of the barriers in the biocatalytic system and is
frequently the rate-determining step in gas-liquid transfer process. The strong
influence of broth viscosities on the oxygen transfer coefficient, kLa is shown in
Table 4.3 and 4.4. This result is also depicted in Figure 4.2 and 4.4. The results are
consistent with what being reported by Martinov & Valev (2002).
68
Table 4.3 Increase of kLa values at high broth viscosities in Rushton turbine and
marine impeller at different impeller speeds
Rushton Turbine Impeller Speed, N (rpm)
CMC solution ('P) 200 400 600 800 1000
kLa (s-1) at 0.25%w/v 0.0034 0.0173 0.0214 0.0257 0.0387
kLa (s-1) at 0.5%w/v 0.0027 0.0154 0.0203 0.0222 0.0305
kLa (s-1) at 1%w/v 0.0013 0.0063 0.012 0.019 0.0257
Table 4.4 Increase of kLa values at high broth viscosities in Rushton turbine and
marine impeller at different air flow rates
Rushton Turbine Air Flow Rate, Q (l/min)
CMC solution ('P) 3 6 9 12 15
kLa (s-1) at 0.25%w/v 0.0211 0.0213 0.0216 0.0283 0.0346
kLa (s-1) at 0.5%w/v 0.0145 0.0171 0.0207 0.0265 0.0321
kLa (s-1) at 1%w/v 0.0068 0.0102 0.0121 0.0163 0.0188
A sharp decrease of kLa up to 40% in both Rushton and marine impellers was
observed. The decrease in kLa was caused by an increase of broth apparent
viscosities from 0.25%(w/v) to 1%(w/v) of CMC solution. The result also shows
that under the same agitation and aeration rates, the kLa is highest in the air-water
system than that in the CMC solutions. In the Rushton turbine, maximum value of
kLa was 304.2 hr-1 at air-water system at 50oC and lowest kLa value attained was 4.68
hr-1 at 1%(w/v) CMC solutions at 30oC. As for the marine impeller, maximum value
of kLa was 117 hr-1 at air-water system at 50oC and lowest kLa value attained was
6.12 hr-1 at 1%(w/v) CMC solutions at 30oC. A viscous solution does not take up
oxygen as well as water. Identical results were obtained for both Rushton and
marine impeller.
69
Correlation proposed by Michel and Miller (1962) in Equation 3.9 was used
to compute the significance difference of volumetric power consumption per unit
volume by the Rushton and marine impellers on the kLa. The dependence of kLa on
volumetric power consumption for Rushton and marine impellers at different
temperature and viscosities is illustrated in Figure 4.5 and 4.6, respectively. It is
clear from the plots that the kLa depends on the power consumption, bearing a close
resemblance to the correlations found in the literature (Shukla et al., 2001, Martinov
& Vlaev, 2002, and Wernersson & Tragardh, 1999). The trend in Figure 4.5 and 4.6
shows that the kLa increase as the specific power input increases. It is evident that
the Rushton turbine provided a greater oxygen transfer rate compared to that of
marine impeller. In order to reach kLa ~ 26.28 hr-1 in air-water system at 30oC,
marine impeller requires 22.4 Wm-3 while Rushton turbine requires 12.7 Wm-3. On
the other hand, to reach kLa ~ 33.48 hr-1 in strong pseudoplastic fluid the values are
461 Wm-3 and 227 Wm-3 for the marine impeller and Rushton turbine, respectively.
Consequently, the same oxygen transfer rate was obtained by the Rushton turbine at
lower power consumption.
70
0.1 1
0.1
kL a (s ) Marine
kL a (s ) Rushton
R2R = 0.97
-1
R2M = 0.93
-1
R2R = 0.94
R2M = 0.91
0.01
Rushton 30°C
Rushton 40°C
Rushton 50°C
Marine 30°C
Marine 40°C
Marine 50°C
0.001 0.001
1 100 10000
3
P g/VL (W/m )
0.1 1
1% - Rushton
0.5% - Rushton
0.25% - Rushton R2R = 0.92
1% - Marine
0.5 - Marine R2R = 0.89
0.25% - Marine
R2R = 0.99
0.1
kLa (s ) Marine
kLa (s ) Rushton
-1
0.01 2
-1
R M = 0.93
2
RM = 0.97
R
2
M = 0.98 0.01
0.001 0.001
1 100 10000
3
Pg /VL (W/m )
However, the power required for the marine impeller to provide the same
agitation and aeration rates was much lower as compared to the Rushton turbine.
The measured volumetric power consumption for marine and Rushton impeller was
in the range of 0.002 kW/m3 to 0.5 kW/m3 and from 0.01 kW/m3 to 2 kW/m3
respectively. As illustrated in the power curve shown in Figure 2.8, power provided
by Rushton turbine is 5 times higher than the marine impeller one. Under the same
Reynolds number, Rushton turbine offers better local mixing and a greater kLa than
the marine impeller. This fact may look strange on the general conclusion that equal
power per unit volume and superficial gas velocity leads to the same kLa regardless
of the impeller type (Geankoplis, 1993).
Interestingly, power reduction (up to 5%) in the specific power input for both
Rushton and marine impeller was also observed. This power reduction was more
dominant especially on introduction of the gas in a viscous liquid, where there are
tendency for the gas to get accumulate behind the impeller blade and form a cavity
(Stanbury and Whitaker, 1984). The reduction in Pg/Po in the increase of liquid
viscosities however, does not significantly affect the oxygen transfer rate in the
bioreactor.
Agitation creates velocities in the fluid, which will result in a pumping flow
and a circulation flow in the tank. It is the convective flow that transports heat and
mass in the bioreactor over long distances, whereas the turbulent part of the flow
creates local mixing. Different flow pattern by Rushton and marine impellers as
illustrated in Figure 4.4 significantly affects the mixing capacity and kLa in the
bioreactor. Radial flow pattern by the Rushton turbine drives the liquid radially from
the impeller causes a compartmentalization problem when strong pseudoplastic
fluids are introduced into the liquid which results in a development of stagnant zones
away from the impellers. The axial-flow pattern created by the marine impeller
produced a higher turbulence compared to the Rushton turbine at the same agitation
rates. Under this condition, a better bulk mixing is in favours of marine impeller.
72
(a) (b)
Figure 4.7 Flow pattern produce by impellers. (a) axial-flow (b) radial-flow
The impeller speeds employed was to maintain the turbulence region in both
Newtonian and non-Newtonian. However, at low agitation rates in the viscous liquid
system, the two phases did not appear to be completely dispersed. Large gas bubbles
were observed in the CMC solutions at concentration of 0.5%(w/v) and 1%(w/v),
indicating that the gas and the liquid phases were not well dispersed. The Reynolds
number which indicates the turbulence in the bioreactor for Rushton turbine and
marine impeller are compared in Table 4.5. In the air-water system, the flow was
always in the developed turbulent regime and the Reynolds number exceeds 2 x 104.
Unlike in air-water system, the flow regime for mixing of the CMC solutions at
concentration of 1%(w/v) was transitional ( 460 < NRE < 2600) because of the high
apparent viscosities. Although a very low power number, Np (0.5 to 0.35) was
observed, the marine impeller manages to provide a high turbulence regime with a
Reynolds number as high as 2 x 105. This value is nearly a tenfold higher compared
to the value attained for the Rushton turbine.
73
Table 4.5 Turbulence parameter in the 16 liter bioreactor for Rushton turbine
and marine impeller at different impeller speeds and air flow rates
Air-water system Air-viscous system
Rushton Marine Rushton Marine
turbine impeller turbine impeller
Power number, Np 5 0.35 - 0.5 3.5 - 5 0.35 – 0.5
Reynolds number, 20156-146980 26443-191566 467-117487 616-127500
NRE
Impeller speed, N 200 – 1000 rpm 200 – 1000 rpm
Air flow rate, Q 3 – 15 l/min 3 – 15 l/min
A larger diameter of marine impeller would give better bulk mixing, however,
Rushton turbine are preferable for breaking up gas bubbles and promoting oxygen
transfer to the liquid. This will favour turbulent mixing over bulk mixing and
increased the kLa. Hence, a higher kLa values was attained in Rusthon turbine
compared to the marine impeller. This is illustrated in Figure 4.8 and 4.9. Although
liquid phase mixing is crucial, the effect of aeration rate on kLa values was also due
to the shearing and dispersing of gas bubbles, which leads to an increased of kLa and
contact times. The results attained consistent with the reports by other works
(Shukla et al., 2001, Badino Jr. et al., 2001, and Arjunwadkar et al., 1998).
74
0.1
R2R = 0.98
R2R = 0.90
R2R = 0.84
kL a (s )
-1
0.01
R2M = 0.88
R2M = 0.84
R2M = 0.96
Rushton 30°C
Rushton 40°C
Rushton 50°C
Marine 30°C
Marine 40°C
Marine 50°C
0.001
0.001 0.005
0.005 0.01
vg (m/s)
0.1 0.1
R2R = 0.83
kLa (s ) Rushton
R2M = 0.90
1% - Rushton
R2M = 0.87 0.5% - Rushton
R2M = 0.94 0.25 - Rushton
1% - Marine
0.5% - Marine
0.25% - Marine
0.001 0.001
0.001 0.005 0.01
vg (m/s)
The kLa values obtained were correlated with respect to the superficial air
velocity, vg and the specific power consumption, Pg/VL, as shown in Equation 3.5.
b
§ Pg ·
a ' ¨¨ ¸¸ v g
c
kLa (3.5)
V
© L¹
The correlation proposed by Cooper et al. (1944) was less complex because it only
considered the effect of volumetric power consumption, Pg/VL and superficial air
velocity, vg on kLa. The analysis in correlating the kLa values with the operating
variables are summarised in Appendix D1 and Appendix D2 for Rushton and marine
impellers, respectively. Results in Table 4.6 summarise the constants in the
empirical correlation obtained at different operating temperatures for Rushton and
marine impellers. The experimental values of constant ‘b’ and ‘c’ between Rushton
turbine and marine impeller in different liquid viscosities are compared and
presented in Table 4.7.
Table 4.6 Comparison of experimental values of constant ‘b’ and ‘c’ between
Rushton turbine and marine impeller in different operating temperatures
Liquid Temperature Constant ‘b’ Constant ‘c’
system (oC) Marine Rushton Marine Rushton
Impeller Turbine Impeller Turbine
30 0.356 0.420 0.773 0.406
Water-air 40 0.261 0.356 0.427 0.501
50 0.266 0.318 0.693 0.605
76
Table 4.7 Comparison of experimental values of constant ‘b’ and ‘c’ between
Rushton turbine and marine impeller in different liquid viscosities
Liquid Temperature Constant ‘b’ Constant ‘c’
system (oC) Marine Rushton Marine Rushton
Impeller Turbine Impeller Turbine
Water-air 30 0.356 0.420 0.773 0.406
0.25%(w/v) 30 0.439 0.450 1.084 0.278
CMC - air
0.5%(w/v) 30 0.381 0.431 0.555 0.485
CMC - air
1%(w/v) 30 0.368 0.563 0.375 0.626
CMC - air
The parameter estimates (constant ‘b’ and ‘c’) reflect the influence of
volumetric power consumption and superficial air velocity on kLa. Constant ‘b’ is
the slope of the graph at constant air flow rate. The magnitude of ‘b’ represents the
level of dependence of kLa on the agitation. Constant ‘c’ is the slope of the graph at
constant agitation speed. The magnitude of ‘c’ represents the level of dependence of
kLa on the sparging rate applied to the system. A higher value of constant ‘b’ and ‘c’
means that a stronger dependency of kLa on operating variables and a steeper slope in
the logarithmic plots. A strong dependence means that a small change in the
operating variables (power input or superficial velocity) will significantly affect the
kLa values. Observing the results shown in Table 4.6 and 4.7, Rushton turbine
showed a strong dependence of the kLa on the volumetric power consumption and
the marine impeller has a strong dependence of kLa on superficial air velocity. The
results also indicate that for Rushton turbine, a variation in temperatures and
viscosities will results in an insignificant decrease and small increase in constant ‘b’,
respectively. Also noted that the constant ‘c’ increased with the increased of
temperatures and viscosities. However, in the marine impeller the values of constant
‘c’ drops significantly with the increased of broth viscosity.
Reviewing the parameter estimates presented in Table 4.6 and Table 4.7, it is
observed that for air-water system, ranges of values for Rushton turbine are within
0.32 < b < 0.42 and 0.4 < c < 0.6. For air-viscous liquid system, the constants are
0.45 < b < 0.56 and 0.27 < c < 0.63. As for the marine impeller, the constants
attained are within 0.26 < b < 0.36 and 0.43 < c < 0.77 for air-water system. For air-
77
viscous liquid system, the constants are 0.36 < b < 0.44 and 0.37 < c < 1.08. These
values are tested for specific power consumption within range from 0.01 to 2 kW/m3
and 0.002 to 0.5 kW/m3 for Rushton and marine impellers respectively. The
operating variables were correlated at corresponding ranges of superficial air velocity
of 1.5 x 10-3 to 8 x 10-3 and at apparent viscosity of 0.81 to 35 cP for Rushton and
marine impellers. The predictions of the empirical equation agreed well with the
measured data within 15% and 30% of average and maximum standard error,
respectively.
The values attained are unique to the bioreactor used and fell within the
acceptable range if published data in Table 1.1 (section 1.1) are taken as a
comparison. Kawase and Moo-Young (1988) proposed that in promoting a good
oxygen transfer in the bioreactor, the value of constant ‘b’ and ‘c’ should not fall
beyond the range of 0.37 < b < 0.8 and 0.2 < c < 0.84, respectively. By employing
the empirical correlation proposed by Cooper et al. (1944) at different impeller
speeds and air flow rates, the values of constant ‘b’ and ‘c’ achieved are within the
range proposed by Kawase and Moo-Young (1988). This proved that the operating
variables selected are suitable and applicable in determining a wide range of kLa
values and promotes a good oxygen transfer rate in the bioreactor. The straight-line
trend (indicated by excellent regression coefficients) of the kLa with respect to the
operating variable values in the logarithmic plots presented in Figure 4.1, 4.2, 4.3,
4.4, 4.5, 4.6. 4.8 and 4.9 signifies that the experimental work matched with the
published results.
The performance of 16 liter and 150 liter bioreactor on the oxygen transfer
rate was compared at various temperatures and viscosities by employing the
proposed scale-up protocol as illustrated in Figure 3.3. Similar values of kLa attained
at 16 liter bioreactor are maintained upon scale-up to 150 liter by manipulation of
power input and aeration rates. The kLa at 150 liter was matched with was obtained
at 16 liter by adjusting the impeller rotation speed and air flow rates. The scale-up
was performed by considering the dynamic similarity in both scales (16 liter and 150
liter) and assumed equality in turbulence in hydrodynamic. The efficiency and
consequences in employing the scale-up protocol was evaluated in the following
section.
79
The scale-up equations was applied in order to determine the minimum and
the maximum value of the operating variables (impeller speeds and air flow rates)
upon scale-up on a basis of constant kLa from 16 liter to 150 liter bioreactor. The
results for the determination of air flow rates and impeller speeds at 150 liter scale on
the basis of constant volumetric power input with superficial velocity are tabulated in
Table 4.10 and Table 4.11, respectively. The results for the determination of
impeller speeds and air flow rates at 150 liter scale on the basis of constant
volumetric power input with impeller tip speed are presented in Table 4.12 and Table
4.13, respectively.
Table 4.10 Determination of air flow rates at 150 liter scale on the basis of
constant volumetric power input with superficial velocity
Q1 (l/min) Q1(m3/s) DT2 (m) DT1 (m) Q2 (m3/s) Q2 (l/min)
3 0.0005 0.41 0.2 0.00021 12.6075
6 0.0001 0.41 0.2 0.00021 25.215
9 0.00015 0.41 0.2 0.00021 37.8225
12 0.0002 0.41 0.2 0.00021 50.43
15 0.00025 0.41 0.2 0.00021 63.0375
The operating variables achieved at 150 liter scale were calculated based on
different types of scale-up criterion. It cannot be adopted directly in order to achieve
a similar kLa values at 150 liter bioreactor. However, the results attained are used as
a base line in determining the operating variables at 150 liter to achieve the similar
value of kLa as in the 16 liter scale. These operating variables were set as the upper
level and the lower level in the ‘trial-and-error’ step. It represents the inner-loop of
the proposed scale-up protocol. The base line and the constraint of the operating
variables at 150 liter bioreactor are shown in Table 4.14. In referring to the operating
variables achieved and the limitation in allowable operating range at 150 liter vessel,
it was found that the value of air flow rates are a bit out of range. Therefore, two sets
of operating variables were proposed to obtain a similar kLa values as in the 16 liter
bioreactor. The proposed operating variables at 150 liter bioreactor are presented in
Table 4.15.
81
Table 4.14 Base line in determining the operating variables at 150 liter scale
Operating Scale-up criteria Allowable
Variables Constant Pg/VL and Constant Pg/VL and Operating Range at
SND vg 150 liter
Impeller 65.2 – 326 rpm 70 – 350 rpm 50 – 600 rpm
Speed, N2
Air Flow Rate, 12.6 – 63 l/min 30.7 – 153 l/min 5 – 100 l/min
Q2
From the proposed agitation and aeration rates, the kLa attained at 150 liter
are closely matched with the kLa in the 16 liter scale. The objective of the ‘trial-and-
error’ step in the scale-up protocol was to achieve a comparable operating condition
in both scales and to determine the scaling-up factor upon scale-up from 16 liter to
150 liter bioreactor. By manipulation of the power input and the superficial air
velocity, a comparable kLa values was successfully achieved. The results of the
‘trial-and-error’ step in distilled water at 30oC are shown in Table 4.16. Similar
results are obtained for other operational parameters namely at different viscosities
and temperatures. The results are depicted in Table D.1, D.2, D.3, D.4 and D.5 in
Appendix D3.
82
It was known that the geometry of the bioreactor is the same in both scales.
However, the dimensions are difference as the scale increases. Different in mixing
and liquid rheology may also cause a difficulty in scaling-up of a bioreactor (Al-
Masry, 1999). Thus, the CMC solution and the air-water solution were used as a
model solution for non-Newtonian fluid and Newtonian fluid, respectively.
Parameters like pH, temperature and liquid viscosity are kept under tight control and
remain the same upon scaling-up on a basis of constant kLa. At different type of
liquid solution and operating variables, the scaling-up factor was determined. The
new operating variables at 150 liter scale and the scaling-up factor are shown in
Table 4.17.
Table 4.17 Operating variables at 150 liter scale on a basis of constant kLa
Operating Liquid System
Variables Air-water Air-Viscous
System System
Impeller Speed, N2 59 – 395 rpm 30 – 311 rpm
Air Flow Rate, Q2 9.3 – 54 l/min 7 – 50 l/min
Scale-up Factor, R
Scale-up Impeller 0.318 0.245
Speed, R1
Scale-up Air Flow Rate, 3.41 2.85
R2
that the greater the scale of operation, the harder to maintained the oxygen transfer
rate in the bioreactor.
Scale-up of the bioreactor from 16 liter to 150 liter scale must meet the
oxygen transfer requirements while maintaining a low variation in power input and a
controlled flow pattern. Technically, in order to obtain a similar value of kLa in both
scales, the differential in the constant ‘b’ and ‘c’ was kept as low as possible.
Therefore, to lower the hydrodynamics difference upon scale-up, the scaling-up
factor was computed. The scaling-up factor is the normalized value of operating
variables (impeller speeds and air flow rates) at larger scale to the smaller scale. The
scaling-up factor was calculated to observe the influence on the operating conditions
at 150 liter scale if the impeller speed and the air flow rate in the 16 liter was varied.
Upon achieving the impeller speeds and air flow rates at 150 liter scale, several
hydrodynamic parameters namely the impeller Reynolds number, the volumetric
power input and the superficial air velocity in the bioreactor were computed to define
the bioreactor operating conditions at larger scales. The significance and the
consequences of the hydrodynamic difference upon scale-up were evaluated based
on the dependence of the kLa on the operating variables. These will be further
discussed in the next section.
85
0.1 0.1
k La (s-1 )
kL a (s )
-1
0.0 1 0.01
2 2
R = 0.99 R = 0.97
R2 = 0.99 R2 = 0.94
(a) (b)
0.1
kL a (s )
-1
N (16 liter)
N (150 liter)
0.001
10 100 1000
Impeller speed, N (rpm)
(c)
0.1 0.1
N (16 liter) N (16 liter)
kL a (s ) N (150 liter) N (150 liter)
kL a (s -1)
-1
0.01 0.01
R2 = 0.94 R2 = 0.92
2 2
R = 0.98 R = 0.90
0.001 0.001
10 100 1000 10 100 1000
Impeller speed, N (rpm) Impeller speed, N (rpm)
(a) (b)
0.1
N (16 liter)
N (150 liter)
kL a (s )
-1
0.01
R2 = 0.98 R2 = 0.99
0.001
10 100 1000
Impeller speed, N (rpm)
(c)
0.1 0.1
2
R = 0.99
2
R = 0.99
kL a (s )
-1
2 2
kLa (s -1)
R = 0.94 R = 0.98
0.01
2
kL a (s )
R =1
-1
(c)
Figure 4.12 Dependence of kLa on volumetric power consumption in distilled
water at different temperatures (a) T = 30oC (b) T = 40oC (c)T = 50oC
The results in Figure 4.12 proved that upon scaling-up, a similar slope and the
plots are coincide with each other meaning that equal kLa dependency on power
input was successfully achieved. In comparison with the dependence of kLa on
volumetric power consumption in the Newtonian fluid, the non-Newtonian fluid
which is the air-viscous system caused a slight variation on the dependence of kLa on
the volumetric power consumption upon scale-up from 16 to 150 liter scale. From
observation in the bioreactor upon operation, it was seen that stagnant zones
developed was larger at 150 liter scale due to the compartmentalization. The
88
0.1 0.1
2
R = 0.95 2
R = 0.98
kL a (s )
R = 0.92
-1
kL a (s )
2
-1
0.01 R = 0.89
0.01
(a) (b)
0.1
2
R = 0.99
0.01
0.001
0.1 1 10 100 1000 10000
3
Pg /VL (W/m )
(c)
As illustrated in Figure 4.12 and 4.13, the specific power input required to
keep the kLa value constant at 150 liter was lower compared to the 16 liter scale
experiment. The volumetric power consumption computed at 150 liter scale is 53%
lower than the power attained at 16 liter bioreactor. This may be resulted from the
increase in the liquid volume and the gassing effect at larger scale significantly
changes the power consumption upon scale-up (Michel and Miller, 1962). This
agrees with the general opinion that not everything can be maintained constant upon
scale-up (Kossen and Oosterhuis, 1985).
89
The slopes in the logarithmic plots in Figure 4.13 are similar in both scales
even though the plots are not coinciding with each other. The dependency of kLa on
the power input in non-Newtonian fluid was compared in order to observe the
differences upon employing the scaling-up factor in scaling-up on a basis of constant
kLa. As seen in Figure 4.12 and Figure 4.13, equal volumetric power consumption
may be achieve if higher power input was employed. Nevertheless, by doing so, it
will create a deviation in the kLa value upon scale-up from 16 liter to 150 liter.
Different impeller speed at higher scale will result in a different power input upon
scale-up. However, the dependence of kLa on the volumetric power consumption
was equivalent in both Newtonian and non-Newtonian fluids upon scale-up from 16
to 150 liter vessel.
Similarly as the impeller speed, the air flow rate was significant altered at
higher scale. This is illustrated in Figure 4.14. Figure 4.14 shows the dependence of
kLa on the air flow rate upon scale-up from 16 liter to 150 liter bioreactor based on
constant kLa in different liquid viscosities and operating temperatures. The air flow
rate was proportionally increased as the scale increases. Based on the logarithmic
plots in Figure 4.14, by implying the scaling-up factor for the air flow rate, similar
kLa values as in 16 liter scale was successfully achieved in the 150 liter bioreactor.
The increase of air flow rates necessitates in compensating with increase of
bioreactor volume. A greater volume at 150 liter and changes in surface to volume
ratio upon scale-up, a higher air flow rates was employed. In order to maintain the
kLa at 150 liter scale, it requires increasing the air flow rates up to three times higher.
The air flow rates were also increased at 150 liter scale to create similar turbulence
and provide adequate oxygen supply as in the 16 liter scale. This finding was
consistent with the results reported by Maranga et al. (2004).
90
0.1 0.1
16 liter 16 liter
150 liter 150 liter
kLa (s )
kLa (s )
-1
-1
R2 = 0.90 R2 = 0.95
2 2
R = 0.84 R = 0.99
0.01 0.01
1 10 100 1 10 100
Air flow rate,Q (l/min) Air flow rate,Q (l/min)
(a) (b)
0.1 0.1
16 liter 16 liter
150 liter 150 liter
kLa (s )
kLa (s )
-1
-1
R2 = 0.98 R2 = 0.98
R2 = 0.83 R2 = 0.99
0.01 0.01
1 10 100 1 10 100
Air flow rate,Q (l/min) Air flow rate,Q (l/min)
(c) (d)
0.1 0.1
16 liter 16 liter
150 liter 150 liter
kLa (s )
kLa (s )
-1
-1
0.01
R2 = 0.98 R2 = 0.98
R2 = 0.92 R2 = 0.99
0.01 0.001
1 10 100 1 10 100
Air flow rate,Q (l/min) Air flow rate,Q (l/min)
(e) (f)
Figure 4.14 Dependence of kLa on air flow rate in distilled water at different
temperatures (a) T = 30oC (b) T = 40oC (c)T = 50oC and CMC solution at different
concentrations (d) CMC 0.25%(w/v) (e) CMC 0.5%(w/v) (f) CMC 1% (w/v)
91
The air flow rates at 150 liter scale may be different from the air flow rates at
16 liter scale, however, the superficial air velocity are almost identical in both scales.
This is illustrated in Figure 4.15 and Figure 4.16 for Newtonian fluid and non-
Newtonian fluid, respectively.
0.1 0.1
2
R = 0.95
2
R = 0.99
k L a (s )
kL a (s )
-1
-1
R = 0.90
2
R = 0.84
(a) (b)
0.1
2
R = 0.98
2
R = 0.98
k L a (s -1 )
(c)
0.1 0.1
2
R = 0.99 2
R = 0.99
kL a (s )
-1
kL a (s )
-1
2 2
R = 0.83 R = 0.98
(a) (b)
0.1
16 liter 150 liter
kL a (s )
-1
2
R =1
2
R = 0.97
0.01
0.001 v g (m/s) 0.01
(c)
In referring to the logarithmic plots in Figure 4.15 and Figure 4.16, unlike the
dependence of kLa on the volumetric power consumption, the dependence of kLa on
the superficial air velocity are not coinciding with each other. It was found that the
superficial air velocity in the 16 liter vessel was higher compared to the 150 liter
ones. However, the slopes of the trend achieved are the same in both scales of
operation. In practicing the scale-up protocol, a maximum aeration of 1.5 vvm was
applied at 16 liter bioreactor and only 0.5 vvm of aeration was employed at 150 liter
scale. This may be due to the high residence time of bubbles, greater volume and
higher vessel in 150 liter scales (Maranga et al., 2004). Interestingly, by
manipulating the operating variables a similar turbulence was successfully achieved
in promoting a similar oxygen transfer rate in both scales. The turbulence in both
scales was compared at different impeller speeds as shown in Figure 4.17.
93
1000000 1000000
16 liter 16 liter
150 liter 150 liter
Reynolds number, N RE
Reynolds number, N RE
100000 100000
R2 = 1
R2 = 1
R2 = 1
R2 = 1
10000 10000
10 100 1000 10 100 1000
Impeller speed, N (rpm) Impeller speed, N (rpm)
(a) (b)
1000000 1000000
16 liter 16 liter
150 liter 150 liter
Reynolds number, N RE
Reynolds number, N RE
100000 100000
2
R =1
R2 = 0.99
R2 = 1
R2 = 1
10000 10000
10 100 1000 10 100 1000
Impeller speed, N (rpm) Impeller speed, N (rpm)
(c) (d)
100000 10000
16 liter 16 liter
150 liter 150 liter
Reynolds number, N RE
Reynolds number, N RE
10000 1000
R2 = 1 R2 = 1 R2 = 1
R2 = 1
1000 100
10 100 1000 10 100 1000
Impeller speed, N (rpm) Impeller speed, N (rpm)
(e) (f)
Figure 4.17 Dependence of kLa on Reynolds number in (a) water (T=30oC) (b)
water (T=40oC) (c) water (T=50oC) (d) CMC 0.25%(w/v) (e) CMC 0.5%(w/v) (f)
CMC 1% (w/v)
94
Based on the results presented in the previous logarithmic plots, equal liquid
motion was attained and the corresponding velocities are approximately the same in
both scales. Hence, equal mixing capacity was also achieved for both Newtonian
and non-Newtonian fluids upon scale-up from 16 liter to 150 liter bioreactor. It was
discovered that the operating temperature and the liquid viscosities are independent
of scale. Similar temperature was employed at 150 liter scale and it did not show a
significant effect on kLa upon scale-up (see Figure 4.12 and 4.15). A variation of kLa
dependence on the operating parameters in the non-Newtonian fluid showed that it is
difficult to maintain a similar hydrodynamics in the non-Newtonian fluid compared
to the Newtonian ones. The impeller speeds, air flow rates, volumetric power
consumption and the superficial air velocity was known to be the manipulate
variables and scale-dependent in employing the scale-up factor upon scale-up from
16 liter to 150 liter bioreactor on a basis of constant kLa. The following section will
be focussing on the significance of the constants in the empirical correlation upon
scaling-up the bioreactor.
Upon scaling-up from 16 liter to 150 liter bioreactor, a constant kLa was
successfully achieved in both scales. Similar empirical correlation was employed in
both scales and the parameter estimates (constant ‘b’ and ‘c’) was compared to
observe the significance of these constants upon scale-up based on constant kLa. The
operating variables namely the volumetric power consumption and the superficial air
velocity were correlated with the kLa values attained in the 150 liter scale. The
results are summarised in Appendix D4. It was previously illustrated in the
logarithmic plots that the trend of kLa dependency on the operating variables was
almost identical in both scales. However, to examine how close the slope of the
logarithmic plots was, the constant ‘b’ and ‘c’ in both scales were compared. Table
4.18 summarised the value of constant ‘b and ‘c’ in both scales at different operating
95
temperature in air-water system. Where else, Table 4.19 showed the value of
constant ‘b and ‘c’ in both scales at different liquid viscosities in air-viscous system.
Table 4.18 The values of constant ‘b’ and ‘c’ upon scale-up from 16 liter to 150
liter at different operating temperature in air-water system
Liquid system Temperature Constant ‘b’ Constant ‘c’
(oC) 16 liter 150 liter 16 liter 150 liter
30 0.4196 0.388 0.4063 0.5493
Water-air 40 0.3561 0.3541 0.5009 0.4491
50 0.3179 0.3207 0.6046 0.5257
Table 4.19 The values of constant ‘b’ and ‘c’ upon scale-up from 16 liter to 150
liter at different liquid viscosities in air-viscous system
Liquid system Temperature Constant ‘b’ Constant ‘c’
(oC) 16 liter 150 liter 16 liter 150 liter
0.25%(w/v) CMC 30 0.4485 0.4282 0.278 0.4458
- air
0.5%(w/v) CMC - 30 0.4305 0.4421 0.4849 0.4611
air
1%(w/v) CMC - 30 0.5626 0.4177 0.6264 0.5449
air
In order to compare the performance of 16 liter and 150 liter bioreactor, the
operating variable namely impeller speed and air flow rate was changed in order to
maintain the kLa value constant at higher scale. The range of operating parameters
varied upon scale-up from 16 liter to 150 liter vessel to obtain a similar kLa value are
presented in Table 4.20 and Table 4.21 in the Newtonian and non-Newtonian fluid,
respectively. The results attained are consistent with the literature reported by
Wernersson and Tragardh (1999).
Table 4.20 The values of range of operating parameters varied upon scale-up
from 16 liter to 150 liter at different operating temperature in air-water system
Scale of operation
Manipulated Variables 16 liter 150 liter
Range of constant ‘b’ 0.32 < b < 0.42 0.32 < b < 0.39
Range of constant ‘c’ 0.4 < c < 0.6 0.44 < c < 0.54
3
Range of Pg/VL 0.0013 < Pg/VL < 2 kW/m 0.001 < Pg/VL < 4.2
kW/m3
Range of vg 1.6 x 10-3 < vg < 8 x 10-3 1.7 x 10-3 < vg < 7 x 10-3
m/s m/s
Table 4.21 The values of range of operating parameters varied upon scale-up
from 16 liter to 150 liter at different liquid viscosities in air-viscous system
Scale of operation
Manipulated Variables 16 liter 150 liter
Range of constant ‘b’ 0.45 < b < 0.56 0.41 < b < 0.44
Range of constant ‘c’ 0.27 < c < 0.63 0.44 < c < 0.54
Range of Pg/VL 0.0013 < Pg/VL < 2 kW/m3 0.0001 < Pg/VL < 1.7
kW/m3
Range of vg 1.6 x 10-3 < vg < 8 x 10-3 9 x 10 < vg < 6 x 10-3 m/s
-4
m/s
It was discovered that the constant ‘b’ may influence the volumetric power
consumption and the constant ‘c’ may significantly effects the superficial air velocity.
By employing the same empirical correlation in both scales, both of these values
need to be matched upon scale-up. Hence, the impeller speeds and the air flow rates
are important manipulated variables in scaling-up on a basis of constant kLa. In
predicting the kLa values, an average deviation of 10% and maximum deviation of
20% standard error was expected. The experiment data fitted well with the empirical
correlation proposed by Cooper et al. (1944) with a high correlation coefficient, R2.
97
4.5 The Performance of E.coli Batch Fermentation at 16 and 150 Liter Scale
5 25
4.5
4 20
3.5
Dry Cell Weight (g/L
3 15 Substrate (g/L)
2.5
2 10
1.5
The specific oxygen uptake rate (qO2) of the E.coli strain is illustrated in
Figure 4.19. The maximum specific oxygen consumption rate was found to be
431.55 mg O2 per g cell.h and 630.81 mg O2 per g cell.h in 16 liter and 150 liter
bioreactor, respectively. It was observed that the consumption rate was higher at the
beginning of the exponential phase and maintained almost a constant rate as the
fermentation proceeds.
700
16 liter
600 150 liter
500
q O2 (mg O2 / g cell. h)
400
300
200
100
0
0 2 4 6 8 10
time (h)
Figure 4.19 Specific oxygen uptake rate of recombinant E.coli in 16 and 150 liter
99
180
160
Oxygen Transfer Rate, k La (h )
-1
140
120
100
80
60
40 16 liter
150 liter
20
0
0 2 4 6 8 10
time (h)
Figure 4.20 Oxygen transfer rate of recombinant E.coli in 16 and 150 liter
The kinetic profiles may not be coinciding with each other; however, F-test
for the equality of kinetic profiles for E.coli fermentation in 16 liter and 150 liter
scale was performed to demonstrate the similarity of these profiles. The results of
the F-test are presented in Appendix F. Similar kinetic profiles of cell growth,
glucose consumption, specific oxygen uptake rate and oxygen transfer rate achieved
in both scales demonstrated that the scale-up protocol was successfully employed in
the E.coli fermentation at 16 and 150 liter scale.
100
0.1 0.1
R2 = 0.99
R2 = 1
k L a (s -1)
kLa (s )
-1
R2 = 0.91 R2 = 0.92
(a) (b)
Figure 4.21 Dependence of kLa on (a) volumetric power consumption and (b)
superficial air velocity for recombinant E.coli fermentation
0.1 0.1
2
R = 0.99
R2 = 0.99
R2 = 1
k L a (s )
kL a (s )
-1
-1
(a) (b)
Figure 4.22 Comparison of dependence of kLa on volumetric power consumption
between recombinant E.coli fermentation and air-water system in (a) 16 liter (b) 150
liter
It is clear from the logarithmic plots in Figure 4.22; the trend of the kLa
dependency was comparable in the air-water system and the real culture. The slopes
achieved were also are not very different in both system and the values are presented
in Table 4.22.
Table 4.22 Comparison of constant ‘b’ between E.coli culture broth with air-
water system in 16 liter and 150 liter
Liquid system Temperature Constant ‘b’
(oC) 16 liter 150 liter
E.coli 37 0.5469 0.3016
Water-air 40 0.3561 0.3541
0.1 0.1
2
R2 = 0.92 R =1 R2 = 0.91 R2 = 0.95
k L a (s )
kL a (s )
-1
-1
16 liter-E.coli
150 liter-E.coli
16 liter-Water
150 liter-Water
0.01 0.01
0.001 0.01 0.1 0.001 vg (m/s) 0.01
v g (m/s)
(a) (b)
Figure 4.23 Comparison of dependence of kLa on superficial air velocity between
recombinant E.coli fermentation and air-water system in (a) 16 liter (b) 150 liter
Table 4.23 Comparison of experimental values of constant ‘b’ and ‘c’ upon scale-
up of E.coli fermentation from 16 liter to 150 liter
Liquid system Temperature Constant ‘b’ Constant ‘c’
(oC) 16 liter 150 liter 16 liter 150 liter
E.coli 37 0.5469 0.3016 0.8638 0.9359
Water-air 40 0.3561 0.3541 0.5009 0.4491
In predicting the kLa values, the empirical equations in E.coli system at both
scales are in good agreement with the experimental data with a 3% maximum
experimental error. The range of operating parameters varied upon scale-up of E.coli
fermentation from 16 liter to 150 liter vessel to obtain a similar kLa value is
presented in Table 4.24.
Table 4.24 The values of range of operating parameters varied upon scale-up
from 16 liter to 150 liter in E.coli Fermentation
Scale of operation
Manipulated Variables 16 liter 150 liter
Constant ‘b’ value 0.5469 0.3016
Constant ‘c’ value 0.8638 0.9359
Range of Pg/VL 0.27 < Pg/VL < 0.7 kW/m3 0.27 < Pg/VL < 1.6 kW/m3
Range of vg 5.3 x 10-3 < vg < 7.4 x 10-3 3 x 10-3 < vg < 4.6 x 10-3
m/s
Reynolds number, NRE 6.4 x 104 < NRE < 8.7 x 1.6 x 105 < NRE < 2.9 x
104 105
CHAPTER 5
5.1 Conclusions
Marine impeller provided a better bulk mixing, however gave low mixing
power. On the other hand, Rushton turbine provided a better mixing power but
resulted in compartmentalization problem. The significance of hydrodynamic
difference between Rushton turbine and marine impeller on the oxygen transfer rate
at 16 liter bioreactor was successfully confirmed. From the experimental data, it was
evident that the kLa increased with the volumetric gassed power input and superficial
air velocity. The results showed that the liquid temperature and viscosity
significantly affected the oxygen transfer rate in the bioreactor. The Rushton turbine
was more effective in gas distribution and gave greater oxygen transfer than that of
the marine impeller. However, in viscous environment, marine impeller provided
better mixing. By using the correlation introduced by Cooper et al. (1944) (i.e. kLa =
a’ (Pg/VL)b (vg)c), the agitation speed and airflow rate was empirically associated
with the oxygen transfer coefficient. The magnitude of ‘b’ and ‘c’ in this equation
represented the degree of dependence of kLa on the agitation and sparging rate
applied to the system respectively. The constant ‘a’ was dependent on the broth
conditions, impeller types and the geometry of the bioreactor. These parameter
values were estimated through experimental work. Hence, the values obtained were
unique. Our results indicated that the estimates agreed well with the published data
by Arjunwadkar et al. (1998), Shukla et al. (2001) and Martinov and Vlaev (2003)
(refer Table 1.1).
105
1. The changes of kLa due to agitation by Rushton turbine impeller was more
pronounced in comparison to the marine one. On the other hand, the effect of
sparging rate on kLa was more dominant in the marine system in comparison
to the turbine. Hence, the significance of hydrodynamic difference between
Rushton turbine and marine impeller on the oxygen transfer rate at 16 liter
bioreactor was successfully confirmed.
3. The manipulation of specific power input and superficial air velocity may be
a useful approach in maintaining a constant kLa value upon scale-up from 16
to 150 liter bioreactor.
106
3. The validity of the scale-up protocol proposed in this work may be further
tested in scales higher than 150 liter bioreactor.
4. The scale-up protocol proposed in this work was based on the rules of
thumb technique. Other scale-up approaches such as fundamentals
method, semi-fundamentals method, dimensional analysis and time-
regime analysis may be investigated in scaling-up stirred aerated
bioreactor on the basis of constant kLa.
107
REFERENCES
Aiba, S., Humphrey, A. E., and Mills, N. F. (1973). Biochemical Engineering. 2nd
ed. Tokyo: University Tokyo Press.
Aiba, S., Suzuki, K., and Kitai, S. (1965). Progress Report No. 27 of Biochemical
Engineering Laboratory. Japan: University of Tokyo.
Arjunwadkar, S. J., Sarvanan, K., Kulkarni, P. R., and Pandit, A. B. (1998). Gas-
Liquid Mass Transfer in Dual Impeller Bioreactor. Biochem. Eng. 1: 99-
106.
Badino Jr, A. C., Facciotti, M. C. R., and Schmidell, W. (2001). Volumetric oxygen
Transfer Coefficients (kLa) in Batch Cultivations involving Non-Newtonian
Broths. Biochem. Eng. 8: 111-119.
Bailey, J., and Ollis,O. (1986). Biochemical Engineering Fundamental. 2nd ed.
New York: Mc Graw-Hill.
Garcia-Ochoa, F., Gomez, E., and Santos, V. E. (2000). Oxygen Transfer and
Uptake Rates during Xanthan Gum Production. Enzy. Microbial. Technol.
27: 680-690.
Geankoplis, C. J. (1993). Transport Processes and Unit Operations. 3rd ed. New
York: Prentice Hall.
Kawase, Y., and Moo-Young, M. (1988). Chem. Eng. Res. Des. 66: 284.
Maranga, L., Cunha, A., Clemente, J., Cruz, P., and Carrondo, M. J. T. (2004).
Scale-up of Virus-like Particles Production: Effects of Sparging, Agitation
and Bioeractor Scale on Cell Growth, Infection Kinetics and Productivity.
Biotechnol. 107: 55-64.
Midler, M., and Finn, R. K. (1966). Model System for Evaluating Shear in the
Design of Stirred Fermenters. Biotechnol. Bioeng. 8: 71.
Mohamad, R., Ariff, A., Mohd. Yusoff, H., and Kamal, M. N. (2001). Fermenter
Design and Scale-up. University Putra Malaysia: Fermentation Unit
Technology. Bioscience Institute.
110
Perry, R. H., and Green, D. W. (1997). Perry’s Chemical Engineers Handbook. 7th
ed. New York: Mc Graw-Hill.
Pirt, S. J. (1975). Principles of Microbe and Cell Cultivation. Hulsted Press, New
York: John Wiley & Sons.
Shukla, V. B., Parasu, U., Kulkarni, P. R., and Pandit, A. B. (2001). Scale-up
Biotransformation Process in Stirred Tank Reactor using Dual Impeller
Bioreactor. Biochem. Eng. 8: 19-29.
Vantriet, K., and Tramper, J. (1991). Basic Biorector Design. New York: Marcel
Dekker.
Wang, D. I. C., Cooney, C. L., Demain, A. L., Dunnill, P., Humphrey and Lilly, M.
D. (1979). Fermentation and Enzyme Technology. New York: John Wiley
& Sons.
Wong, I., Garcia, M. A., Rodriguez, I., Ramos, L. B., and Olivera, V. (2003).
Fermentation Scale-up for Production of Antigen K88 Expressed in
Escherichia coli. Proc. Biochem. 38: 1295-1299.
Yagi, H., and Yoshida, F. (1975). Ind. Eng. Chem. Proc. Des. Dev. 14: 488.
Appendix A1
% x V Mx (1)
x
100
Table A.1 shows the calculations and measurements for the preparation of 0.25%(w/v),
0.5%(w/v) and 1%(w/v) of CMC solutions. The same calculation was repeated in preparing
the CMC solutions at respective concentrations for 10 liter and 100 liter scale.
M b c M b
U CMC (2)
VCMC
Table A.3 shows the physical properties of distilled water at various temperatures which has
been taken from “Transport Processes and Unit Operations 3rd Ed.”.
Table A.4: Oxygen Solubility in Air Saturated Pure Water in mg O2/L at an Overall
Pressure of a Water-Vapor Saturated Atmosphere of 760 mm Hg (Perry’s
Chemical Engineering Handbook 7th Edition)
Appendix A2
The following derivation was based on the concentric cylinder viscometer as shown
in Figure 3.1. The equations were derived to determine the shear rate, J value which
based on the concentric cylinder viscometer. For non-Newtonian viscous liquids, the
shear stress was taken to be function of the shear rate as was defined as:
W K.J (3)
For non-Newtonian viscous liquids, the apparent viscosity was also taken to be
function of shear rate and was defined as:
K f J (4)
In steady state, the torque measure on the inner cylinder (using the conservation of
momentum) was given by:
where W was the fluid shear stress at radius r. The shear rate was given by:
So, rearranging equation 5 for W, and substituting this together with equation 5 for J
into equation 3, we get:
This gives:
Z > @
*1 / 2.S .L.K * 1 / 2* 1 / r12 1 / r22 (9)
K >
*1 / 4.S .Z .L * 1 / r12 1 / r @
2
2
(10)
From equations 6 and 7 we can get an expression for the shear rate.
J >
2.Z / r 2 * r12 .r22 / r12 r22 @ (12)
142
Appendix D3
Table D.5 Results of the ‘trial-and-error’ step in 1%(w/v) CMC solution at 30oC
Appendix F
F-Test for Equality of Kinetic Profiles at 16 and 150 liter E.coli Fermentation
The objective of the F-test is to test if the standard deviations of two populations are
equal. The hypothesis that the two equal standard deviations is rejected if:
Type of kinetic profile : Specific Oxygen Uptake Rate, qOUR (mg O2/ g cell.h)
Scale 16 liter 150 liter
Mean 151.682 186.109
Variance 13159.985 30480.755
Observations 10 10
Degree of freedom 9 9
Significant level 0.05 0.05
F-test value 0.432
P (F<= f) one-tailed test 0.113
F critical one-tailed test 0.3146
160
Type of kinetic profile : Oxygen Uptake Rate, OUR (mg O2/ L.h)
Scale 16 liter 150 liter
Mean 374.004 360.18
Variance 19259.898 20233.303
Observations 10 10
Degree of freedom 9 9
Significant level 0.05 0.05
F-test value 0.952
P (F<= f) one-tailed test 0.471
F critical one-tailed test 0.3146