Biochemical Engineering Journal 27 (2005) 155–160

A new methodology for the optimal design of batch fermentation plants

Ricardo Simpson a,∗ , Carolina Astudillo b , Fernando Acevedo b
a Department of Chemical, Biotechnological and Environmental Processes, Universidad Técnica Federico Santa Marı́a, Casilla 110-V, Valparaı́so, Chile
b School of Biochemical Engineering, Pontificia Universidad Católica de Valparaı́so, Casilla 2950, Valparaı́so, Chile

Received 25 January 2005; received in revised form 13 June 2005; accepted 2 August 2005

Abstract

In the design of batch and fed-batch fermentation plants, the problem of determining the adequate combination of number and size of
the fermentors to be used, so to accomplish the desired production schedule must be faced. The problem has infinite solutions, as for any
fermentor size, a number of units of that size will do the work, but not all solutions are equal from an economical standpoint. The problem of
determining the optimum number of pieces of equipment and the optimum operation schedule has been addressed by several authors in the
process engineering field.
The optimal combination of size and number of fermentors will be the one that maximizes the net present value. The objective of this work
was to develop a methodology for the optimal design of batch fermentation plants based in the maximum net present value criterion and to
compare it with the more restricted criterion of minimum investment in production fermentors.
Results show that considering only the investment in production fermentors underestimates the optimum number of fermentation units,
probably because it only considers the investment in production fermentors and its assumptions of equal preparation and discharge times. On
the other hand, the method proposed in this work, although somewhat more complex, gives a more accurate result.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Batch fermentation plants; NPV; Production fermentor

1. Introduction for the production of fine biologicals. Continuous fermenta-

Industrial fermentation plants consist of three main sec-
tions: preparation, fermentation and product recovery. The
preparation section usually contains operations such us
medium preparation and sterilization and inoculum propaga-
tion. The fermentation section is the heart of the plant where
the transformation of raw material into products takes place.
The product recovery section encompasses the downstream
operations needed to obtain the product of interest with the
required purity [1].
Fermentations may be run in one of three modes of opera-
tion: batch, continuous or semi-continuous, also referred to as
fed-batch. Large-volume fermentations for the production of
commodities are usually run in the fed-batch or batch modes,
while batch operation is used for small-volume processes
∗ Corresponding author. Tel.: +56 32 654546; fax: +56 32 654478.
E-mail address: ricardo.simpson@usm.cl (R. Simpson).
tions are extensively used in research and development and in
effluent treatment processes, but are seldom used in industrial
fermentations, mainly because of the possibility of contami-
nation and mutation [2,3].
In the design of fed-batch and batch fermentation plants,
the problem of figuring out the adequate combination of num-
ber and size of the fermentors to be used, so to accomplish
the desired production schedule, must be faced. In princi-
ple, the problem has infinite solutions, as for any fermentor
size, a number of units of that size will do the work. Nev-
ertheless, not all solutions are equal from an economical
standpoint. The problem of determining the optimum number
of pieces of equipment and the optimum operation schedule
has been addressed by several authors in the process engi-
neering field [4–6]. Recently, Simpson el al. [7] presented
a complete study on the optimization of batch retort battery
design and operation in canneries. Nevertheless, an equiva-
lent approach has rarely been presented for the fermentation
industry.

1369-703X/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.bej.2005.08.003
156 R. Simpson et al. / Biochemical Engineering Journal 27 (2005) 155–160

that gives the highest net present value of the investment in

Nomenclature equipment for the complete plant as optimal.
The objective of this research is to develop a methodology
A proportionality constant, Eq. (11)
for the optimal design of batch fermentation plants based
a cost exponent, Eq. (6)
on the maximum net present value criterion and to compare
B boiler size
it with the criterion of minimum investment in production
b cost exponent, Eq. (27)
fermentors.
C product concentration
Cf fermentor cost
CP product cost per m3 of broth
2. Optimization methodology based on minimum
F hourly flow rate of broth to downstream sec-
investment in production fermentors
tion, Eq. (1)
F yearly flow rate of broth to downstream sec-
The following is the methodology developed by Borzani
tion, Eq. (33)
[8]. This approach requires that the following set of condi-
i interest rate
tions apply:
I total investment
Ia total investment in ancillary equipment (a) Continuous operation of the downstream section of the
Ib investment in the boiler plants is desirable, irrespectively of the batch operation of
Ic investment in pipe and fittings the fermentors. This can be achieved by having a battery
IP investment in propagators of Nf fermentors scheduled in such a way that there is
If total investment in fermentors, propagators always one of them discharging.
included (b) All production fermentors are of the same size.
j number of periods, Eq. (31) (c) In each fermentor, the discharge time (td ) equals the
K proportionality constant, Eq. (36) preparation time (tp ).
Kf proportionality constant, Eq. (6) (d) The cost of the fermentors can be represented by
M mass of product produced per year Cf = Kf Va [9], where Kf and a are constants in the con-
Nf number of production fermentors sidered volume range.
Nf * optimum number of production fermentors
NPV net present value, Eq. (14) The continuous flow of fermented broth to the downstream
n number of propagation steps, Eq. (22) section will be given by:
P sell price per m3 of broth M
ty hours of operation per year F= (1)
ty Cη
td fermentor discharge time
tf fermentation time The discharge time will be:
tp Batch preparation time VL
V total volume of each production fermentor td = (2)
F
V* optimal volume of each production fermentor
VL liquid volume of each production fermentor The operation cycle of one fermentor (tt ) equals:
Vp propagator volume
tt = tp + tf + td (3)
Greek symbols In order to comply with condition (a), the number of fer-
α proportionality constant, Eq. (16) mentors must be given by:
βA , total benefit
βj annual benefit tp + tf + td
Nf = (4)
η product recovery efficiency td
φ proportionality constant, Eq. (26)
Using Eq. (2) and applying condition (c) to Eq. (4):
ϕ proportionality constant, Eq. (25)
λ cost exponent, Eq. (26) Ftf
Nf = 2 + (5)
VL
The cost of one fermentor, as stated in condition (d) is
The optimal combination of size and number of fermen- given by:
tors can be considered as the one that requires the lowest
Cf = Kf V a (6)
investment in equipment. A restricted interpretation of this
condition would be to circumscribe the investment to that The objective function to be minimized is:
in production fermentors only, as proposed by Borzani [8].
A more general approach would be to consider the solution CT = Kf V a Nf (7)
 R. Simpson et al. / Biochemical Engineering Journal 27 (2005) 155–160 157

Neglecting the difference between V and VL and replacing Using Eqs. (6), (13) and (16):
the value of V in Eq. (5) in Eq. (7) renders:

a

a Ftf
Ftf Cf = Kf α (17)
CT = Kf Nf (8) Nf − 1 − A
Nf − 2
Propagation stage corresponds to the generation and prop-
The optimal number of fermentors is obtained by making agation of the inoculum required for each large-scale fermen-
the derivative of Eq. (8) equal to zero: tation and is considered to be 10% of the liquid volume of the
2 production fermentor. So n propagation steps of increasing
Nf ∗ = (9) volumes are required to obtain the inoculum for the fermen-
1−a
tor, and n will depend on the liquid volume and the number
Replacing Eq. (9) into Eq. (5) gives the optimal volume of production fermentors. So,
of each fermentor:
Vn−1 αVL,n−1
Ftf (VP )n = = (18)
V∗ = (1 − a) (10) 10 10
2a
So, the volume of each propagator will be:
In the case of industrial fermentors, the value of the expo-
nent a usually ranges from 0.35 to 0.74 [8–11]. (VP )i = 10(VP )i−1 (19)
The total cost of propagation will depend on the number
of propagation lines to be considered:
3. Optimization methodology based on maximum net
present value n

IP = NP Kf [(VP )i ]a (20)
We propose to generalize Borzani’s method by reliev- i=1
ing constraint (c) and expanding the optimization criterion From Eqs. (18)–(20), the following expression can be
from minimizing the investment in production fermentors to obtained:
maximizing the net present value of the investment (NPV)
a
Ftf 1 − (1/10a )n
considering the production and propagation fermentors, the IP = NP Kf α (21)
Nf − 1 − A 10a − 1
ancillary equipment and the boiler needed to generate steril-
ization steam. Downstream equipment will not be considered, As established before, the number of propagators depend
as it will remain the same for all the alternatives. on the liquid volume of the production fermentor:
Now we have that:
VL = V0 10n+1 (22)
tp = Atd (11)
So,
where A, a dimensionless proportionality constant, can be 
  
VL (Ftf /(Nf − 1 − A))
equal or higher than one. Taking Eq. (11) into consideration, n = log − 1 = log −1 (23)
V0 V0
Eq. (4) is modified to:
tf Using Eqs. (21) and (23):
Nf = 1 + +A (12)
a
td Ftf
IP = NP Kf α
From Eq. (2) and (12): Nf − 1 − A
Ftf 1 − (1/10a ) log[((Ftf /(Nf − 1 − A))/V0 )] − 1
VL = (13) ×
Nf − 1 − A 10a − 1
(24)
The NPV is given by:
m
 βj
NPV = −I + (14) The connection cost (pipes and fittings) can be expressed
j=1
(1 + i)j as a fraction of the fermentor cost:
Ic = ϕIf 0<ϕ<1 (25)
3.1. Total investment
The size of the boiler, B, is related to the size of each fer-
Total investment can be expressed as: mentor, as well as with the time schedule of the fermentations.
This means that there will be peaks of steam consumption.
I = If + Ic + Ib + Ia + Ip (15)
In the determination of that schedule, it is assumed that each
The cost of each fermentor can be estimated from Eq. (6) fermentor has a lag equal to the preparation time, tp , so B will
considering that V: depend mainly on the size of the fermentor, so:
V = αVL (16) B = f (V ) = φV λ (26)
158 R. Simpson et al. / Biochemical Engineering Journal 27 (2005) 155–160

 
Then the cost of the boiler can be expressed as: 1 − (1/10a )n
NPV = − Nf Kf (αVL )a + NP Kf (αVL )a
10a − 1
I b = KB B b 0<b<1 (27)
 b
× (1 + ϕ) − KB f (αVL ) − Ia + KFTy (P − CP )
Replacing B in Eq. (27) utilizing Eqs. (13) and (26):


b (38)
Ftf
Ib = KB f α (28)
Nf − 1 − A
Eq. (38) is the objective function to be maximized by mak-
Now the total investment I can be written: ing its derivative equal to zero. It can be seen that:


a
Ftf NPV = f (Nf ) (39)
I = Nf Kf α
Nf − 1 − A dNPV

a  =0 (40)
Ftf 1 − (1/10a )n dNf
+ NP Kf α (1 + ϕ)
Nf − 1 − A 10a − 1 Utilizing Eq. (14):


b
Ftf dNPV dI dβA
+ KB f α + Ia (29) =− + =0 (41)
Nf − 1 − A dNf dNf dNf
Eq. (41) implies that maximizing NPV is equivalent to
As can be seen in Eq. (29), the total investment is a function minimizing the total investment given by Eq. (42):
of the number of production fermentors and fermentation
a
time: αFtf
I = Kf (1 + ϕ)
Nf − 1 − A
I = f (Nf , tf ) (30)


NP 1
× Nf + a 1 − an
10 − 1 10
3.2. Benefits and NPV


λ b
Ftf
The total benefit may be calculated as: + KB φ α + Ia (42)
Nf − 1 − A
j
 βj
βA = (31) Eq. (42) contains a discrete variable (n), so the minimum
j=1
(1 + i)j value of I can be found graphically with A, the factor that
relates preparation and discharge times, as parameter.
Furthermore, the benefits over one period can be calcu-
lated from:
4. Example calculations
βj = F  (P − Cp ) (32)

The annual volumetric production, F , can be calculated 4.1. Design basis

from Eq. (32) considering ty hours of operation per year.
The production of lactase from Kluyveromyces frag-
F  = Fty (33) ilis (currently named Kluyveromyces marxianus) grown in
cheese whey will be considered using the data of Ordenes
Then, the annual profit is: and Ortega [12]. The plant capacity was 1.252 × 1011 IU
lactase/year and the overall efficiency of the downstream
βj = Fty (P − CP ) (34)
operations was 33%.
So, Lactase specific activity was 5000 IU/g cell dry weight
and maximum cell concentration was 25 g/l.
β1 = β2 = β3 = · · · = βm = Fty (P − CP ) (35) Enzyme produced in the fermentors:
1
βA = KFty (P − CP ) (36) 1.252 × 1011 (IU/year) = 3.794 × 1011 (IU/year)
33%
K can be calculated from Eq. (37): (43)
j
 1 (1 + i)j − 1
K= = (37) Fermented broth:
j=1
(1 + i)j i(1 + i)j
3.794 × 1011 (IU/year)· m3
= 3035 (44)
Replacing Eqs. (29) and (36) in Eq. (14): 5000(IU/g) × 25, 000(g/m3 ) year
 R. Simpson et al. / Biochemical Engineering Journal 27 (2005) 155–160 159

Table 1
Cost calculation dataa
Parameter Nomenclature Value
Initial cell concentration X0 2.5 g/l
Final cell concentration X 25 g/l
Specific growth rate µm 0.37 h−1
Fermentation time tF 6.22 h
Broth flow rate F 0.479 m3 /h
Inoculum volume of the first V0 0.01 m3
propagator
Proportionality constant of Kf US$ 12,025
fermentor cost equation
Exponent of fermentor cost a 0.5
equation
Proportionality constant of KB US$ 42.4
boiler cost equation Fig. 1. Total investment as a function of the number of fermentors and the
Exponent of boiler cost b 0.86 relation between preparation and discharge times using the criterion of max-
equation imum NPV.
Proportionality constant of φ 231.43
boiler size equation Table 2
Proportionality exponent of λ 0.697 Summary of the results of the optimization using criteria I and II
boiler size equation Nf * V* (l) Total volume (l)
Ratio of total fermentor to α 1.2
Criterion I 4 1490 5960
liquid volume
Proportionality constant of ϕ 0.12 Criterion II
fittings cost equation A = 1.0 5 993 4965
Number of propagation lines Np 1 A = 1.5 6 851 5106
a Based on data of references [12–14]. A = 2.0 9 497 4473

It will be considered that the plant operates 24 h/day
from Mondays through Fridays, which is equivalent to 264
days/year, so the flow to the downstream section will be
0.479 m3 /h.
Table 1 summarizes the values of the different variables
and parameters employed for the cost calculations.
4.2. Criterion I—minimum investment in production
fermentors
Using the data of Table 1 and Eqs. (9) and (10), the opti-
mum number of production fermentors turn out to be 4 units
of 14,90 l each.
5. Conclusions
Our results show that Borzani’s method, although being
very simple, underestimates the optimum number of fermen-
tation units and underestimates the total required volume,
probably because it does not considers the investment in prop-
agators and ancillary equipment and its assumption of equal
preparation and discharge times.
On the other hand, the proposed method, which includes
the total cost in propagation and production units and admits
different preparation and discharge times, is somewhat more
complex but gives a more accurate result. Lastly, it is inter-
esting to note that the assumption of equal preparation and
discharge times give the least number of production units and
total fermentation volume.

4.3. Criterion II—maximum NPV (minimum total References

investment)
Using Eq. (42) with the data of Table 1 for values of A
(Eq. (11)) between 1 and 2, Fig. 1 is obtained.
It can be seen that the optimum number of fermentors that
yields the minimum investment is affected by the relation
between preparation and discharge times (A), ranging from
5 to 9 units of equal volumes between 993 l (A = 1) and 496 l
(A = 2). Values of A less than 1 were not considered because
they represent unrealistic situations.
Table 2 summarizes the results obtained by the application
of each criterion.
[1] H.B. Reisman, Economic Analysis of Fermentation Processes, CRC
Press, Boca Raton, FL, 1988, Chapter 1.
[2] A.L. Demain, Biotechnol. Adv. 18 (2000) 499.
[3] K. Toda, J. Gen. Appl. Microbiol. 49 (2003) 219.
[4] D.W. Rippin, Comput. Chem. Eng. 17 (1993) 1.
[5] C. Kondili, R. Pantelides, H. Sargent, Comput. Chem. Eng. 17 (1993)
211.
[6] G.V. Reklaitis, Overview of scheduling and planning of batch process
operations, in: G.V. Reklaitis, A.K. Sunol, D.W. Rippin, O. Hortacsu
(Eds.), Batch Processing System Engineering, Springer, Berlin, 1996,
pp. 660–705.
[7] R. Simpson, S. Almonacid, A.A. Teixeira, J. Food Proc. Eng. 25
(2003) 515.
160 R. Simpson et al. / Biochemical Engineering Journal 27 (2005) 155–160
[8] W. Borzani, Fermentação descontinúa, in: W. Borzani, U. Lima,
E. Aquarone (Eds.), Engenharia Bioquı́mica, Blücher Editores, São
Paulo, 1975, pp. 105–111.
[9] K.M. Guthrie, Chem. Eng. 76 (3) (1969) 114.
[10] J.O. Kalk, A.F. Langlykke, Cost estimation for biotechnological
projects, in: A.L. Demain, N.A. Solomon (Eds.), Manual of Industrial
Microbiology and Biotechnology, American Society for Microbiol-
ogy, Washington, DC, 1986, pp. 363–385.
[11] H.W. Blanch, D.S. Clark, Biochemical Engineering, Marcel Dekker
Inc., New York, 1996, Chapter 8.
[12] C. Ordenes, M. Ortega, Thesis, School of Biochemical Engineering,
Catholic University of Valparaiso, 1995, 430 pp.
[13] P. Tello, G. Vergara, Thesis, School of Biochemical Engineering,
Catholic University of Valparaiso, 2001, 367 pp.
[14] R. Simpson, Optimising batch processing, in: P. Richardson (Ed.),
Improving Thermal Processing, Woodhead Publishing Ltd., Cam-
bridge, 2004, Chapter 4.