J. Gen. Appl. Microbiol.

, 41, 461-473 (1995)

PERFORMANCE OF A MULTISTAGE FERMENTOR

WITH CELL FILTERING AND RECYCLING FOR
CONTINUOUS ACETIC ACID PRODUCTION

AKIO NISHIWAKI* AND JOHN R. BOURNE'

Department of Materials Chemistry and Bioengineering, Oyama

National College of Technology, Oyama 323, Japan
' Technisch-Chemisches Laboratorium ETH , CH-8092 Zurich, Switzerland

(Received October 17, 1994; Accepted September 19, 1995)

The steady-state performance of a multistage fermentor with cell filtering

through a membrane and recycling of concentrated cells is studied nu-
merically for continuous production of acetic acid. The tanks-in-series
model is used to describe the flow behaviour of culture broth in the
multistage fermentor. Kinetic expressions and parameter values are taken
from the literature. The effects of the total number of stages, the dilution
rate, the bleed ratio, the recycle ratio and the feed concentrations as
operating conditions on fermentor characteristics such as the concentra-
tions of total and viable cells as well as substrate and product in each
stage, the cell viability, the acetic acid productivity and the substrate
conversion were examined under the conditions of equal tank volumes
and a recycle ratio equal to or greater than one. An increase in the
number of stages and decreases in the bleed and recycle ratios and in the
feed concentration of product enhance the productivity. The relations
between the maximum productivity and the corresponding optimum
dilution rate and between the corresponding outlet concentrations of the
four components and the optimum dilution rate are presented as figures.
Equations for the average cell viability and the critical dilution rate
causing cell washout are also obtained. Furthermore, the minimum feed
concentration of substrate for its complete consumption is mentioned.
Compared to a single tank fermentor, staging and cell recycle increase
substrate utilisation and acetic acid productivity. The computed perform-
ance of the fermentor configuration proposed in this study is sufficiently
promising for the continuous production of acetic acid that it should now
be checked experimentally.

* Address reprint requests to: Dr . Akio Nishiwaki, Department of Materials Chemistry and
Bioengineering, Oyama National College of Technology, 771 Nakakuki, Oyama 323, Japan.

461
462 NISHIWAKI and BOURNE VOL. 41

In recent years, various attempts to improve the productivity of acetic acid by

continuous fermentation have been made (3-6). One way of obtaining high
production rates is to maintain more viable microbial cells in a fermentor compared
to conventional batch or continuous production of acetic acid. The uses of
immobilized cells and cell recycle with membrane separation are favourable alter-
natives for this purpose. Although these two methods for high cell-density culture
are popular in researches aimed at efficient ethanol fermentation, however, the
latter method is not yet familiar for the case of acetic acid production. Most
recently, Park and Toda (4) investigated high cell-density fermentation of acetic
acid by using a membrane recycle bioreactor system and reported that a theoretical
maximum acetic acid productivity of 123.1 kg m-3 h- I at a dilution rate of 3.52 h '
without bleeding was obtained in their simulation. In such a reactor system, a
membrane filter module is attached to a single continuous stirred tank fermentor
with a semi-closed loop for cell recycling.
The use of a tanks-in-series or a multistage column fermentor instead of the
single tank in the above system may be expected to bring an overall increase in
acetic acid productivity. For ethanol fermentation, Ghose and Tyagi (1) showed
that productivity was increased by using two stirred tanks in series in place of a
single tank fermentor. Also, Lee et al. (2) demonstrated that ethanol productivity
could be further improved by using cell recycling together with one or two tanks.
However, the effect of product inhibition, especially on specific product formation
rate, in acetic acid fermentation is quite different from that in ethanol fermentation,
thus resulting in a remarkable difference between these two processes. Although
Park and Toda (4) also recommend to reuse the bleed which is rich in viable cells
from the first cell-recycle fermentor in the following fermentor for further acetic
acid production, it was not shown quantitatively how high productivity can be
attained by a multiple stage system. With recent notable developments in mem-
brane performance, multistage fermentor-separator operations will become more
important as a practical method to enhance acetic acid productivity. Therefore, an
analysis of the reactor performance of multistage systems is required for rational
design and operations.
The purpose of this work was to elucidate numerically the effects of various
operating parameters on the characteristics of tanks-in-series and multistage
column fermentors with cell separation and recycle for acetic acid production.

PERFORMANCE EQUATIONS

The continuous steady-state operation of a multistage fermentor with cell

filtering and recycling for acetic acid production is considered. As shown in Fig. 1,
a fermentor in this system is composed of either N completely mixed tanks
connected in series with equal working volumes or a column partitioned into N
stages with equal spacings. Fresh medium supplied to the first stage (or tank) is
considered to be sterile. In the fermentor, cell growth and death, substrate
1995 Multistage Fermentor for Acetic Acid 463

(a) (b)

Fig. 1 Scheme of a multistage fermentor with cell filtering and recycling.

(a) N tanks-in-series fermentor. (b) Multistage column fermentor.

consumption and product formation proceed. The culture stream from the Nth
stage is introduced to a filtering module to separate cells in the culture through a
membrane and the concentrated cells are then recycled to the first stage. The
recycle flow rate is denoted by rF. To maintain a stable operation, part of the
culture broth is withdrawn from the Nth stage. The bleed flow rate is denoted by
BE Without such bleeding, a steady-state in total cells cannot be achieved because
of an unlimited accumulation of non-viable cells.
In deriving the mass balance equations for viable and total cells, substrate and
product, the following assumptions were made:
(1) Kinetic expressions for acetic acid fermentation and their parameter
values are available in the literature (4) and are stated below.
(2) For the column-type fermentor, the culture broth in each stage is
perfectly mixed and backmixing between adjacent stages is negligible. Hence, the
flow behaviour of the culture in the whole column as well as that in multiple tanks
is expressed by the tanks-in-series model.
(3) Membrane separation is perfect and thus the filtrate stream contains no
cells.
(4) The liquid residence time in the recycle loop is negligible.
(5) No resistance to mass transfer exists and the rate of fermentation is
determined by its kinetics.
According to Park and Toda (4), the specific growth rate can be expressed by
the following equation with product inhibition and without the substrate saturation
constant in the Monod expression:

[ -,im {1- (P~Pm)n} (1)

where ,um= O.26h ', Pm= 63.5 kg m 3 and n = 3.61.
464 NISHIWAKI
and BOURNE VOL.
41

Also, the specific rates of cell death, acetic acid production and ethanol
consumption are represented by Park and Toda (4) as follows, respectively:

y=klexp(k2D) (2)
where k1-0.059h -1 and k2-0.325h.
v=a+b,u-cp2 (3)
where a =1.92 h-1, b = 386.8 and c =1,347 h.

qs=v/Yps (4)
where Yp~s =1.18 kg kg 1, implying that this yield coefficient was constant. Park
and Toda (4) used Sf=47.4 kg m-3 and Pf= l0 kg m-3 as feed concentrations in
their experiments and determined the above kinetic parameter values.
Based on the above assumptions and kinetics, the steady-state mass balances
for total cells, viable cells, substrate and product in the first and the ith stages can
be described as follows:

i1, r$XrN- (l +r)X~1+ ND

D 1X1=0 (5)

r$XN-(l+r)X1+ ND
1 (u1-y)X1=0 (6)

S 1
f+rSN- (1 +r)S1- ND gSlX1 0 (7)

Pf+rPN-(l+r)P1+ ND
1 L1X1=0 (8 )

i2,3, ...,N, (l+r)Xr1-1-(l+r)Xt~+ ND

1 ~c,Xi=O (9 )

l+r X,_1- l+r X~+ l ,ul-r X,=O 10

(1+r)S1-1-(l+r)Si- ND
l gs,X1=O (11)

Nomenclature: a, parameter in Eq. (3) (h-'); b, parameter in Eq. (3); B, bleed ratio; c, pa-
rameter in Eq. (3) (h); D, dilution rate (h-'); F, volumetric flow rate of medium feed (m3h-'); k1,
parameter in Eq. (2) (h-'); k2, parameter in Eq. (2) (h); n, parameter in Eq. (1); N, total number
of stages or tanks; P, product (acetic acid) concentration (kg m-3); Pr, productivity of acetic acid
(kg m-3h');~qs, specific consumption rate of ethanol (kg kg-' h-' ); r, recycle ratio; S, substrate
(ethanol) concentration (kg m-3); v, cell viability; X, viable cell concentration (kg m-3); X~, total cell
concentration (kg m-3); Yp~s,yield of acetic acid based on ethanol consumed (kg kg-'); Z, dimen-
sionless distance. [Greek letters] j3, factor by which cells in the recycle stream are concentrated by
membrane filter; r, specific death rate (h-'); i, specific growth rate (h-'); v, specific production rate
of acetic acid (kg kg-' h-'). [Subscripts] av, average; c, critical; f, feed; i, ith stage or tank; m,
maximum; min, minimum; opt, optimum.
1995 Multistage Fermentor for Acetic Acid 465

(l+r)P,_1-(1+r)P,+ ND
1 v~X;=O (12 )

In these balance equations, the dilution rate D refers to the volumetric feed
rate F divided by the total working volume of either all tanks in series or the whole
column. Also, Lt, v; and qs~in the ith stage are denoted in Eqs. (1), (3) and (4) with
P =P~, respectively.
From the mass balance for either viable or total cells around the filter module,
the following relation is obtained:

This expression shows that the cell separator factor $ is fixed by selecting both
the bleed ratio B and the recycle ratio r as operating conditions and increases with
the decreases of B and r.
The concentrations of total and viable cells, substrate and product in each
stage for a given set of N, D, B, r, Sf and Pf are obtained by solving numerically Eqs.
(5)-(12) combined with Eqs. (1)-(4) and (13). As a numerical technique, the
Newton-Raphson iterative method is applicable. The ith-stage substrate concentra-
tion S~is also calculated from the corresponding ith-stage product concentration P,
by means of the constant mass yield of product from substrate Yp~s:
Y P; -Pf
5= S (14)
f-S;
which is derived from Eqs. (4), (7), (8), (11) and (12).
When N ==1,the following equations (4) are obtained by setting N= 1 in Eqs.
(5)-(8) and using Eqs. (1) and (13) :
=BD+r (15)
P=Pm {1-(IJ/,1m)} l~n (16)
X =D(P-Pf)/v (17)
X~_ (BD + y)X/BD (18)
S=Sf-q5X/D (19)
The concentrations of the four components for N=1 are calculated from Eqs.
(16)-(19) in turn after obtaining the values of y and , t from Eqs. (2) and (15),
respectively. These concentrations are independent of the recycle ratio r.
When r - DOas a special case of the multistage system, the resultant equations
derived from Eqs. (5)-(12) by setting r = DOand using Eq. (13) are identical to
Eqs. (15) and (17)-(19).
The acetic acid productivity Pr is defined as follows:
Pr-D(PN-Pf) (20)
466 NIsHIwMU and BOURNE VOL. 41

This definition means that acetic acid in the bleed stream is recoverable as
product.
Operating parameters are the total number of stages N, the bleed ratio B, the
recycle ratio r and the feed concentrations Sf and Pf. The following fermentor
analysis is carried out under the conditions of equal tank volumes and r >_ 1. Also,
the ranges 25-47.4 and 0-15 kg m-3 are used for Sf and Pf, respectively.

RESULTS AND DISCUSSION

Concentrations in stages
Figure 2 shows an example of the calculated concentrations of total cells,
viable cells, substrate and product, Xt;, X,, Si and F, respectively, in each stage, with
the total number of stages N as a parameter. The abscissa Z refers to the
dimensionless distance from the top of a column fermentor as shown in Fig, lb.
For a tanks-in-series fermentor, the ith tank is equivalent to the ith stage of the
column. Xr~ and X~ in each stage are larger than those for N=1 and increase with
an increase of N. Also, their stepwise concentration profiles for each N along the
whole fermentor are almost flat. On the other hand, the corresponding axial
profiles of S; and P~ are greater for a larger N. As can be seen from the values of

Fig. 2. Concentrations of total and viable cells, substrate and product in each
stage with total number of stages as a parameter.
D=1.2 h-`, B=0.1, r=2, Sf=47.4kgm_3 and Pf=0. Broken lines indicate N=1.
1995 Multistage Fermentor for Acetic Acid 467

S, and P,, the substrate consumption and product formation in the first stage for the
multistage fermentor are reduced compared to N=1. Thus, because product
inhibition in the first stage is reduced, the cell growth proceeds further and viable
cells increase with increasing N compared to N=1. Such increases lead to substrate
consumption and acetic acid production to a high level in the following stages. As
a result, the outlet concentrations SN and PN decrease and increase with increasing
N, respectively.
Examining the concentrations of the four components in each stage for B less
than that in Fig. 2, it is found that the corresponding X, for each N in this case
increases slightly, while X<, becomes considerably larger because of further accumu-
lation of non-viable cells. However, the increase of X; in this case brings higher and
lower concentrations of the corresponding FN and SN, respectively.
Figure 3 shows an example of the calculated effect of the dilution rate D on the
final-stage concentrations of total cells, viable cells, substrate and product, XtN, XN,
SN and PN, respectively, with the total number of stages N as a parameter. For each
N, as D becomes large, both XtN and XN increase, approaching their respective
maximum concentrations, and then decrease promptly. At a certain critical
dilution rate D, the washout of cells occurs because of insufficient residence time
for cell growth. The relation between such a D~ and other parameters will be stated

Fig. 3. Effect of dilution rate on final-stage concentrations of total and viable

cells, substrate and product with total number of stages as a parameter.
B =0.05, r=2, Sf=47.4 kg m_3 and Pf=O. Broken lines indicate N=1N
468 NIsHIWAKI and BOURNE VOL. 41

later. At any constant D less than D~, except D very close to D~, both X1N and X~
are higher when more stages are used. PN decreases with increases in D and alsc
becomes higher for larger N values, as in the above case of cell concentrations. The
effects of D and N on SN are the inverse of those on PN.

Cell viability
The viability of cells is the fraction of viable cell concentration relative to total
cell concentration. The cell viability v; in each stage refers to X1/Xr~and the average
cell viability vavin the whole fermentor to the arithmetic mean of v,. Figure 4 shows
a calculation example of the viabilities v~and vas. For each N, the stepwise profile
of v; along the entire fermentor tends to a convex one with increasing N and thus
vN becomes less than the corresponding vas. However, differences between these
two values for each N and also between vavvalues for different N stages are withir.
10o, although vavincreases slightly with the increase of N. Examining the effects of
the dilution rate D, the bleed ratio B and the recycle ratio r on vas, it is found that
the value of vavincreases with increasing D and B and is almost independent of r.
The equation for giving vas, which is defined here as a constant v~ (-X/X1)
can be obtained by setting X, =Xrwav in Eqs. (5), (6), (9) and (10) and combining
them to eliminate Lt and then Xr~by substitution. Using Eq. (13), moreover, thf
resultant equation for vavis represented as follows:
11N
ND ( l+r) 1- 1- l+B r
vav- B IN (21)
ND(1+r) 1- 1- l+
r +y
When N-1 or r= oo, Eq. (21) is reduced to:

V= BD+
BD
y (22)

Fig. 4. Cell viability in each stage and average cell viability in the whole
fermentor with total number of stages as a parameter.
D = l.2 h-', B =0.1, r=1, Sf=47.4 kg m-3 and Pf=0. Solid and dotted lines
indicate viability in each stage and average viability, respectively. The broken line
indicates N=1.
 1995 Multistage Fermentor for Acetic Acid 469

This equation is identical to that obtained from Eq. (18) for N=1. The values
of vav calculated from Eq. (21) agree well with the arithmetic means of v; in Fig. 4
and also under other conditions.

Critical dilution rate

When the value of D becomes larger than a certain critical dilution rate D~, as
mentioned above in Fig. 3, cell washout occurs. From calculation results for
various conditions, it can be seen that the effects of N and r on D~ are not significant,
while the effect of B on D~ is great.
The equation for D~ can be derived by setting P; =Pf fora; in Eqs. (6) and (10)
and combining them to delete X~ by substitution. Furthermore, using Eq. (13), the
resultant expression for D. is written as:

D~= 'tT ,~N (23)

N(1+r) 1- 1- lB+
r
where u refers to Eq. (1) with P =Pf and y to Eq. (2) with D =Dr. Thus, since the
right-hand side of Eq. (23) also contains D~, the value of D~ is determined by a
trial-and-error method. When N =1 or r = c, Eq. (23) results in Eq. (15) with D
= D~. The values of D~ calculated from Eq. (23) agree exactly with those in Fig. 3.
It is found that Eq. (23) is also applicable for other conditions.

Productivity
Figure 5 shows the acetic acid productivity Pr as a function of the dilution rate
D with the total number of stages N and the feed concentration of product Pf as
parameters. As D increases, Pr for each N and Pf increases and has its maximum
at a certain L). Then Pr decreases steeply. As shown by the dotted lines, the

Fig. 5. Effect of dilution rate on acetic acid productivity with total number of
stages and feed concentration of product as parameters.
B=0.05, r=2, Sf=47.4kgm-3 and Pf=10 (A) and 0 (B) kgm-3. Broken lines
indicate N=1. Dotted lines refer to the relation between maximum productivity and
optimum dilution rate.
470 NISHIWAKI and BOURNE VOL. 41

attainable maximum productivity Prm becomes higher with increasing N and with
decreasing Pf. Also, the optimum dilution rate Doptgiving Prm is almost independ-
ent of N and is larger for a smaller Pf. Hence, multistage fermentors are superior
to single-stage fermentors.
Such values of Prm and Doptwere determined for various sets of N, B, r, Sf and
Pf as operating conditions. Figure 6 presents the relation between Prm and Doptwith
N, B and r as parameters in the case of Sf= 47.4 kg m- 3 and Pf= 0. For N = 3 and
5, only the lines of r =1 are shown in the figure. Prm for multiple stages is higher
at any B and r compared to that for N=1 and increases with decreasing both B and
r. Although Prm becomes high with increasing N, the increase of Prm per stage for
multiple stages is largest when N = 2. Both Prm and D0 at r = for each N are
identical to those for N=1 at the same B. The effects of N and r on Dopt are
comparatively small or negligible.
Figure 7 shows the corresponding final-stage concentrations of product and
substrate, PN and SN, respectively, i.e., equal to the concentrations in the filtrate and
bleed streams, at the same Dopt as those in Fig. 6 for Prm. PN at a constant B
increases with increasing N and with decreasing r. Also, except for the range of B
larger than about 0.2, PN at constant N and r becomes slightly higher for a smaller
B. PN in most cases is over 40 kg m-3, which is the minimum acetic acid concentra-
tion used for vinegar (4). The relation between SN and Doptseems nearly symmetric
with respect to a horizontal line of about 27 kg m-3 against the case of PN.
Similarly, Fig. 8 shows the corresponding final-stage concentrations of total cells
and viable cells, XtNand XN, respectively, at the same Doptas those in Fig. 6 for Prm.
Ac R h,.rr mc.ccm~ll Y .. ra~cac firct ararlrlallu and than rc r rlly rlrla tc crrnwlnn

0opt (h-

Fig. 6. Relation between maximum productivity and optimum dilution rate with
total number of stages, bleed ratio and recycle ratio as parameters.
Sf 47.4kgm_3 and P 0. The broken line refers to N=1.
1995 Multistage Fermentor for Acetic Acid 471

Fig. 7. Corresponding final-stage concentrations of product and substrate at the

same optimum dilution rates as those in Fig. 6.

Fig. 8. Corresponding final-stage concentrations of total and viable cells at the

same optimum dilution rates as those in Fig. 6.

accumulation of non-viable cells. AtN also increases with increasing IV and with
decreasing r. The effects of N, B and r on XN are similar with those on Prm in Fig.
6 and the relation between XN and D0 for constant N and r values is almost linear.
In the case that the feed concentration of product Pf is nonzero, both Prm and
D0Pt are found to decrease with increasing Pf. Also, the influences of B and r on
them for a constant Pf are similar to those in Fig. 6. On the other hand, there is no
effect of the feed concentration of substrate Sf on Prm and D0 because Eqs. (5), (6),
(9) and (l0) are independent of Sf and S1.
 472 NISHIWAKI and BOURNE VOL. 41

Substrate conversion
Using Eq. (14) with i =N and Eq. (20), overall substrate conversion at the
fermentor outlet is written as:
Sf-SN _ PN-Pf _ Pr 24
Sf YpsSf
I YpsSfD
I ( )
This equation means that the effects of N, B and r on the substrate conversion
for a constant Sf are the same as those either on PN at a constant Pf or on Pr at a
constant D. For example, as can be seen from Fig. 7 for SN and PN at D0 giving
Prm, the substrate conversion becomes high with increasing N and with decreasing
B and r. Also, from Fig. 5, by comparing Pr values at a constant D, the substrate
conversion is found to increase with decreasing Pf. In this figure, each left end of
the solid and broken lines corresponds to complete substrate consumption (See Fig.
3 for P -O).
Examining the effect of Sf on the substrate conversion, it is found that the
conversion increases with decreasing Sf and as a result, there exists a minimum feed
concentration of substrate Sfmin for complete substrate consumption in the final
stage. Hence, Sf for a feasible operation must be greater than Sfmin. Determining
such values of Sfminfor various given sets of N, D, B, r and Pf by changing Sf, these
values agree well with those from the following equation:

S _ Pr
fmin- Y
p5D (25)
I which is obtained by setting SN = 0 in Eq. (24). For the same conditions, except Sf,
as those in Fig. 6 for Prm and Dopt, Sf minincreases with increasing N and with
decreasing B and r. Also, examining the effect of Pf on Sfmin, it is found that Sfmin
increases with decreasing Pf.

CONCLUSIONS

Applying the tanks-in-series model and the kinetic expressions and parameter
values from the literature (4) to a multistage fermentor with cell separation and
recycle, its steady-state performance was investigated numerically for continuous
acetic acid production. The effects of operating parameters on various fermentor
characteristics were examined and the following results were obtained.
The viable cell concentration in each stage for multistage fermentors is higher
compared to a single fermentor and increases with increasing the total number of
stages and with decreasing the bleed ratio. This raises the overall substrate
consumption and acetic acid production. The difference between the cell viability
in each stage and the average cell viability in the whole fermentor is not significant
and the average viability is given by Eq. (21). The critical dilution rate for the
washout of cells is expressed by Eq. (23). As presented in Figs. 5 and 6, the
maximum acetic acid productivity increases with increasing the total number of
1995 Multistage Fermentor for Acetic Acid 473

stages and with decreasing the bleed ratio, the recycle ratio and the feed concentra-
tion of product. Also, the optimum dilution rate giving the maximum productivity
increases with decreasing the bleed ratio and the feed concentration of product.
Although the maximum productivity can be enhanced by reducing the bleed ratio,
the total cell concentration then becomes very high, as shown in Fig. 8. The
substrate conversion increases with decreasing the feed concentration of substrate.
There exists a minimum feed concentration of substrate for its complete consump-
tion and Eq. (25) holds.
The computed performance of the fermentor configuration proposed in Fig. 1
is based on the kinetic model of Park and Toda (4) and is sufficiently promising for
the continuous production of acetic acid that it should now be checked experimen-
tally.

REFERENCES

1) Ghose, T. K. and Tyagi, R. D., Rapid ethanol fermentation of cellulose hydrolysate. II. Product
and substrate inhibition and optimization of fermentor design. Biotechnol. Bioeng., 21, 1401-1420
(1979).
2) Lee, J. M., Pollard, J. F., and Coulman, G. A., Ethanol fermentation with cell recycling:
Computer simulation. Biotechnol. Bioeng., 25, 497-511 (1983).
3) Mori, A., Production of vinegar by immobilized cells. Process Biochem., 20, 67-74 (1985).
4) Park, Y. S. and Toda, K., Simulation study on bleed effect in cell-recycle culture of Acetobacter
aceti. J. Gen. App!. Microbiol., 36, 221-233 (1990).
5) Sun, Y. and Furusaki, S., Continuous production of acetic acid using immobilized Acetobacter aceti
in a three-phase fluidized bed bioreactor. J. Ferment. Bioeng., 69, 102-110 (1990).
6) Von Eysmondt, J., Vasic-Racki, D., and Wandrey, C., Acetic acid production by Acetogenium
kivui in continuous culture-Kinetic studies and computer simulations. App!. Microbiol. Biotech-
no!., 34, 344-349 (1990).