Selective optimization in thermophilic acidogenesis of cheese-whey wastewater to acetic and butyric acids: partial acidification and methanation.

Abstract

For partial acidogenesis of cheese-whey wastewater, a set of experiments were carried out to produce short-chain volatile fatty acids (VFA) in laboratory-scale continuously stirred tank reactors (CSTR). The maximum rate of acetic and butyric acid production associated with simultaneous changes in hydraulic retention time (HRT), pH, and temperature was investigated, in which the degree of acidification of the whey to the short-chain VFAs was less than 20% of the influent chemical oxygen demand (COD) concentration. Response surface methodology was successfully applied to determine the optimum physiological conditions where the maximum rates of acetic and butyric acid production occurred. These were 0.40-day HRT, pH 6.0 at 54.1 degrees C and 0.22-day HRT, pH 6.5 at 51.9 degrees C, respectively. The optimum conditions for acetic acid production were selected for partial acidification of cheese-whey wastewater because of a higher rate in combined productions of acetic and butyric acids than that at optimum conditions for butyric acid production. A thermophilic two-phase process with the partial acidification followed by a methanation step was operated. Performance of the two-phase process was compared to the single-phase anaerobic system. The two-phase process clearly showed a better performance in management of cheese-whey wastewater over the single-phase system. Maximum rate of COD removal and the rate of methane production in the two-phase process were, respectively, 116% and 43% higher than those of the single-phase system.

Full text

Page 1

Water Research 37 (2003) 2467–2477
Selective optimization in thermophilic acidogenesis of
cheese-whey wastewater to acetic and butyric acids:
partial acidification and methanation
Keunyoung Yang, Youngseob Yu, Seokhwan Hwang*
School of Environmental Science and Engineering, Pohang University of Science and Technology, San 31, Hyoja-dong, Nam-gu, Pohang,
Kyungbuk 790-784, Republic of Korea
Received 10 June 2002; received in revised form 9 December 2002; accepted 29 December 2002
Abstract
For partial acidogenesis of cheese-whey wastewater, a set of experiments were carried out to produce short-chain
volatile fatty acids (VFA) in laboratory-scale continuously stirred tank reactors (CSTR). The maximum rate of acetic
and butyric acid production associated with simultaneous changes in hydraulic retention time (HRT), pH, and
temperature was investigated, in which the degree of acidification of the whey to the short-chain VFAs was less than
20% of the influent chemical oxygen demand (COD) concentration. Response surface methodology was successfully
applied to determine the optimum physiological conditions where the maximum rates of acetic and butyric acid
production occurred. These were 0.40-day HRT, pH 6.0 at 54.1?C and 0.22-day HRT, pH 6.5 at 51.9?C, respectively.
The optimum conditions for acetic acid production were selected for partial acidification of cheese-whey wastewater
because of a higher rate in combined productions of acetic and butyric acids than that at optimum conditions for
butyric acid production. A thermophilic two-phase process with the partial acidification followed by a methanation step
was operated. Performance of the two-phase process was compared to the single-phase anaerobic system. The two-
phase process clearly showed a better performance in management of cheese-whey wastewater over the single-phase
system. Maximum rate of COD removal and the rate of methane production in the two-phase process were,
respectively, 116% and 43% higher than those of the single-phase system.
r 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Two-phase anaerobic digestion; Cheese whey; Response surface method; Optimization; Partial acidification
1. Introduction
Thermophilic anaerobic digestion offers a potential
cost-effective solution for the treatment of high strength
organic waste, such as food processing wastewater,
because of a higher rate of methane formation compared
to the mesophilic anaerobic process. Thermophilic
anaerobiosis of wastes is a three-stage process in which
the complex organic components of the waste are
solubilized, broken down, and fermented into inter-
mediate products that are subsequently reduced to
methane and carbon dioxide. These processes involve
many different species of symbiotic microorganisms,
which are broadly divided into two groups: acidogenic
and methanogenic bacteria [1–3].
These two groups of microorganisms differ widely in
their physiology, biokinetics, and growth environment.
Several researchers have claimed that optimization of
each of these phases would enhance the overall rate of
waste stabilization if the biphasic ecosystem could be
maintained in separate digesters in series: one for acid
production and one for methane production [4–7].
*Corresponding author. Tel.: +82-0-542792282; fax: +82-0-
542798299.
E-mail address: shwang@postech.ac.kr (S. Hwang).
0043-1354/03/$-see front matter r 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0043-1354(03)00006-X

Page 2

2. Acidogenesis in an anaerobic process
Since the methanogenic reaction is believed to be the
rate-limiting step in the overall anaerobic digestion
process, proper control of the methanogenic phase has
been a key factor in the successful operation of most
anaerobic processes [4,8]. However, overall process
enhancement must be based on an understanding of
the optimum growth conditions and behavior of
acidogens in a two-phase process since they play the
primary role in producing major substrates, short-chain
organic acids, for methanogens. While organic acids are
important substrates in methanogenesis, high concen-
trations of these acids can cause anaerobic digestion
systems to fail [9].
It has been controversial whether complete or partial
acidification would improve treatment efficiency in an
anaerobic process, as well as the way in which partial
acidification should be achieved. Some reports show that
two-phase processes with complete acidification of
influent wastewater have advantages in organic loading
rate and gas production over single-phase processes [10–
12]. This type of two-phase system, however, may
involve a considerable increase in construction and
operational costs due to the need of an additional large
reactor for complete acidification of influent organics,
especially for lignocellulosic material that is in agricul-
tural wastewater. Despite the reported advantages of
acidification, complete acidification of organics in the
incoming wastewater could be detrimental for formation
of granular biomass in an anaerobic process [13]. On the
contrary, it was observed that adding a low level of
volatile fatty acids (VFA) (i.e., equivalent to as much as
25% of the influent chemical oxygen demand (COD)) to
an anaerobic digester treating synthetic wastewater
exerts a stimulating effect upon VFA degradation by
methanogens [14,15]. As the operating criteria and
design guidelines for pre-acidification reactors, it was
also recommended that the hydraulic retention time
(HRT) for acidogenic reactors be 6–24h, depending on
the characteristics of the wastewater, and the process
should be applied for pre-acidification of 20–50% of the
potential acidified matter instead of complete acidifica-
tion [16,17]. Thus, a partial-acidification strategy is
likely to enhance the overall efficiency of a two-phase
anaerobic process more than complete acidification
approach.
Therefore, it was hypothesized that the overall
performance of a thermophilic two-phase anaerobic
process would be enhanced if a fraction of the
biodegradable components in wastewater were mainly
converted to acetate and butyrate, which are known to
be the best precursors of methane formation [2,3,9]. This
partial-acidification process would require a smaller
digester volume than the complete acidification step in
conventional two-phase systems due to the shorter
HRT, thus lowering construction and operational costs
[14]. For this reason, the key factor was to find a
combination of operating parameters for a reactor that
maximized the production of acetic and butyric acids.
3. Empirical models for biological processes
There are many biological areas, like the acidogenic
microbial process, where basic knowledge of the
phenomenon is insufficient to build a mechanistic
model. In this case, empirical models and statistical
analysis play an extremely important role in elucidating
basic mechanisms in complex situations and thus
providing better process control.
Response surface methodology (RSM) is a collection
of mathematical and statistical techniques useful for
designing experiments, building models, evaluating
relative significance of several independent variables
(i.e., environmental factors), and determining optimum
conditions for desirable responses [18,19]. In most RSM
problems, the relationship between the response and the
independent variables is unknown. Thus, the first step in
RSM is to approximate the function ðfÞ using a low-
order polynomial in some region of the independent
variables. If the response is not linear, then a higher-
order polynomial must be used to approximate the
response. Once the higher-order polynomial is deter-
mined it can be used to locate the optimum, which is the
set of independent variables such that the partial
derivatives of the model response, Z; with respect to
the individual independent variables, xi; is equal to zero:
qZ
x1
x2
xk
¼qZ
¼ ? ¼qZ
¼ 0:
ð1Þ
This optimum response is called the stationary point
and it could represent a point of maximum response, a
point of minimum response, or a saddle point [18]. The
eventual objective of RSM is to determine the optimum
operating conditions for the system, or to determine a
region which satisfies the operating specifications. Al-
most all RSM problems utilize one or more of these
approximating polynomials [20,21].
4. Experimental purpose
This study was conducted to (1) develop continuous
response surfaces of thermophilic acetic and butyric acid
production with cheese-whey wastewater using mathe-
matical and statistical techniques, (2) find conditions to
maximize the production rates of the acids with respect
to the simultaneous effects of pH, temperature, and
HRT where a maximal 20% acidification of the waste-
water occurs, and (3) compare the process efficiency of
partial-acidification–methanation processwith the
K. Yang et al. / Water Research 37 (2003) 2467–2477
2468

Page 3

single-phase anaerobic process. The production of
propionic acid, which is another common intermediate
in anaerobic digestion, was not considered because of its
toxicity to the system [3,9].
5. Materials and methods
5.1. Experimental materials and microbial culture
One batch (300L) of cheese-whey wastewater (Table
1) was obtained from a local cheddar cheese-processing
industry. The wastewater was homogeneously mixed
and divided into smaller portions and frozen at ?25?C
for further use. Because cheese-whey wastewater con-
tained most of the essential nutrients for microbial
growth [1], no additional nutrients were added. How-
ever, the content of ammonia–nitrogen and phosphorus
as orthophosphate were monitored in order to ensure
that these two critical nutrients were not limiting. The
raw wastewater was diluted with distilled water to give a
desired influent concentration (10,000mg COD/L) for
subsequent experiments.
Anaerobic seed sludge from a municipal wastewater
treatment plant in Pohang, Korea was cultivated in a
6.0L working volume lab-scale anaerobic continuously
stirred tank reactor (CSTR) to produce constant
inoculum for all experimental runs. The inoculum
system was operated at 10 days HRT at 55?C, a
thermophilic temperature range, and the pH was
maintained at 7.0 with 6.0N NaOH. The operating
conditions of the inoculum system remained constant
throughout the experiment. The effluent from this
system was used as seed culture for subsequent
experiments to minimize any confounding effect asso-
ciated with the use of inconsistent inoculum.
Two identical CSTRs with working volumes of
1L were used in the optimization of the partial-
acidification process. The absence or repression of
methanogenic activity was verified by the lack of CH4
production.
After the completion of the optimization study, a two-
phase anaerobic process was set up consisting of a
partial-acidification reactor with working volume of 1L
was directly connected to three identical CSTRs with
working volumes of 5L for methanation of the effluent.
Three separate peristaltic pumps, located between the
partial-acidification and methanation steps, were used to
achieve the desired HRTs for the methanation reactor.
Operational conditions for partial acidification were
based on the optimization study (see Section 6) while the
pH and temperature were maintained at 7.0 with 6.0N
NaOH and 55?C, respectively, for the methanation
reactors. Two identical single-phase anaerobic reactors
with working volumes of 5L were separately operated to
compare performance with that of a two-phase process.
All reactors were equipped with temperature and pH
controllers. The HRT was incrementally decreased (i.e.,
1.5 days) until washout of methanogenic activity was
observed, verified by the lack of CH4production. The
experiments were duplicated and randomized for statis-
tical purpose.
5.2. Analytical methods
A Hewlett-Packard gas chromatograph (Model 6890
plus) equipped with an Innowax capillary column and a
flame ionization detector was used to determine the
concentrations of short-chain organic acids. Helium was
the carrier gas at a flow rate of 2.5mL/min with a split
ratio of 10:1. The same gas chromatograph with an HP-
5 capillary column (film thickness, 0.25mm; length, 30m;
ID, 0.53mm; phase ratio, 3) and a thermal conductivity
detector was used to quantify methane in the biogas.
Helium was the carrier gas at a flow rate of 8mL/min
with a split ratio of 70:1.
The COD concentration was determined according to
the procedures in Standard Methods. The concentra-
tions of protein and total carbohydrate of the waste-
water were measured according to the Kjeldahl and the
phenol–sulfuric acid methods, respectively [22]. A
Shimadzu TOC analyzer (5000-A) was used to measure
the total organic carbon concentration. The concentra-
tions of cations and anions in the sample were measured
by using two identical Dionex ion chromatographs (DX-
120) with IonPac AS14 and IonPac CS12A columns,
respectively.
Table 1
Composition of cheese-whey wastewater
ParameterConcentration (mg/L)
Total chemical oxygen demand
Soluble chemical oxygen demand
Total solid
Volatile suspended solid
TKN
Ammonia nitrogen
Proteinb
Total organic carbon
Total carbohydrate
Acetate
Propionate
Butyrate
Phosphate
Potassium
Sodium
pH
71410 (10)a
56732 (194)
62440 (2927.4)
7560 (580)
1613 (110)
161 (1)
9063.8
35400 (10)
44373 (875.5)
836.3 (54.1)
74.2 (3.7)
52.4 (3.8)
1632.3 (22.6)
2248 (45.3)
455.5 (5.9)
5.92
aStandard deviation.
bProtein content=(TKN?ammonia nitrogen)6.38.
K. Yang et al. / Water Research 37 (2003) 2467–2477
2469

Page 4

5.3. Optimization protocol
In this research we examined the production rates of
acetic and butyric acids associated with simultaneous
changes in HRT, pH, and temperature in the partial-
acidification phase, where the degree of acidification of
the wastewater to the short-chain VFAs (namely acetic,
propionic, butyric, valeric, and carproic acids) was less
than 20% of influent COD concentration. The initial
starting points of the independent variables were
selected as close as possible to the optimum values
based on literaturevalues
[3,20,23,24]. An iterative procedure was used to locate
optimum conditions for production of acetic and butyric
acids. In this experiment, ‘‘optimum conditions’’ meant
the operating conditions for maximizing the rate of
corresponding acid production within the investigated
space of the independent variables. A first-order
polynomial was tried to approximate the first region of
the independent variables. If there was a curvature in the
system, then a polynomial of higher degree, such as
quadratic or partial cubic model, was used. The method
of least squares was used to estimate the parameters in
the approximating polynomials.
The reason for using production rate instead of acid
concentration was to prevent converging the operational
conditions to the endogenous decay phase. If the
product concentration was considered a major depen-
dent variable it could result in the system approaching
the endogenous decay phase because the concentration
generally increases as the HRT increases. Microbial
activity is lowest in the endogenous decay phase [25];
therefore, it is necessary not to operate the system deep
in the endogenous decay region.
of similarconditions
6. Results and discussions
6.1. Optimizing partial acidogenesis
Nineteen trials were run to locate optimum conditions
for the production of acetic and butyric acids. Influent
substrate concentration was maintained at 10,000mg
COD/L for all trials and steady-state was assumed after
five turnovers.
The first region of exploration for the first-order
model was decided as, [0.2, 0.6] days HRT, [5.5, 6.5] pH,
and [50, 60]?C, because acidogens were thought to favor
slightly acidic conditions and showed mmof 9.9d?1on
lactose and the system performance was unstable near a
washout point of 0.1 days [20,26]. The experimental
conditions and data are shown in Table 2. The
orthogonal design, a 23factorial augmented by five
center point runs (from trials 1 to 9 in Table 2), was used
to collect data. The operating condition at the center
point was 0.4-day HRT, 6.0 pH, and 55?C. Repeat
observations at the center were used to estimate the
experimental error.
Thirteen trials were run first. A first-order model, Zi¼
b0þ b1x1þ b2x2þ b3x3; was tried to fit this data by
least squares. The following models were obtained on
each acid:
Zacetate¼ 1954:8 ? 50:4x1? 59:1x2? 17:1x3;
ð2Þ
Zbutyrate¼ 763:7 ? 401:8x1þ 10:5x2? 4:3x3;
ð3Þ
where Ziis the rate of production of acid i (mg i/L/d,
where i=acetic and butyric acids in order), bj the
coefficient values of jth term (mg i /L/d/xj; where
b0=constant; j=HRT (days), pH, temperature (?C) in
Table 2
Experimental conditions and results of the central composite design
Trial HRT (day)pHTemperature (?C) Acetate (mg/L/d)Butyrate (mg/L/d)Conversion ratiob(%)
1
2
3
4
5
6
7
8
0.6
0.6
0.2
0.2
0.6
0.2
0.2
0.6
0.4
6.5
5.5
6.5
5.5
5.5
5.5
6.5
6.5
6.0
50
60
60
60
50
50
50
60
55
544.3
162.6
287.7
513.4
558.8
577.2
355.8
387.7
982.2 (95.7)
130.1
532.4
342.2
177.7
154.3
362.8
686.9
109.8
616.6 (33.5)
17.6
13.2
5.6
5.7
18.5
7.7
8.1
15.0
17.1 (1.9)
9a
10
11
12
13
14
15
0.7
0.4
0.4
0.1
0.4
0.4
6.0
6.0
6.7
6.0
6.0
5.3
55
48
55
55
62
55
246.1
514.8
845.4
489.4
283.4
450.1
203.0
285.0
446.9
412.9
311.6
406.2
16.0
8.6
21.1
5.2
11.1
10.8
aThe experiment was replicated 5 times and the response represented average values (standard deviation).
bCalculations were based on COD equivalents of the short-chain organic acids produced.
K. Yang et al. / Water Research 37 (2003) 2467–2477
2470

Page 5

order), and xjthe corresponding variable term (j=HRT,
pH, and temperature in order).
The regression coefficient, the lack of fit of the first-
order model, and p-values of parameter estimations were
used to validate the models. The p-values of the lack of
fit for acetic and butyric acids were, respectively, 0.0048
and 0.0003, which were significant at a 1% a level. The
regression coefficients for both acids were not significant
at a 5% a level. Therefore, it was concluded that the
first-order model was not an adequate approximation
for both organic acids. The curvature in the response of
the acids might indicate that the conditions were near
the optimum.
At this point, additional trials were conducted to
locate the optimum conditions for both acids more
precisely. A second-order model could not be fit using
the data from trials 1 to 9 in Table 2 due to the lack of
axial data points. An additional six trials (from 10 to 15
in Table 2) were augmented with previous trials to make
a central composite design. Fig. 1 shows the design
boundary for linear and higher-order models.
In order to find an optimum condition for acetic acid
production, various models from linear to quadratic
were tested with these augmented trials. Eqs. (4)–(6)
represent first, interaction, and second-order polynomial
models, respectively.
Za1¼ 1268:9 ? 178:2x1þ 53:2x2? 16:9x3;
ð4Þ
Za2¼5965:8 ? 2216:9x1? 922:6x2? 66:5x3
þ 822:1x1x2? 52:6x1x3þ 11:8x2x3;
ð5Þ
Za3¼ ? 38846:3 þ 2275:2x1þ 4317:7x2
þ 972x3þ 822:1x1x2? 52:6x1x3
þ 11:8x2x3? 5615:1x2
1? 436:7x2
2? 9:4x2
3;
ð6Þ
where Zaiis the rate of acetic acid production with the
model i (i=first, interaction, and quadratic models in
order)
p-Values of regression, lack of fit, and corresponding
coefficients were tested. Lack of fit was not significant,
but regression was significant at a 1% a level only for the
quadratic model which indicated this model fits the
response surface. Therefore, the quadratic model was
selected to describe the response surface of acetic acid
within this region.
Two- and three-dimensional response surfaces of the
quadratic model for acetic acid production with
corresponding estimated optimums are shown in Figs. 2
and 3. The response surface of acetic acid showed a clear
peak, which indicated the optimum condition was well
inside the design boundary. The conditions that max-
imized the production rate of acetic acid, defined as the
optimum conditions earlier, were calculated by setting
the partial derivatives of the function to zero with
respect to the corresponding variables. The optimum
conditions for acetic acid production were 0.40 days
HRT, pH 6.0, and 54.1?C.
The convex hulls in Figs. 3 and 5 are the minimum N-
dimensional volumes, which contain all the trials for
which data have been collected. Contour plots are two-
dimensional slices through these multidimensional de-
sign spaces [27]. The limits of the convex hull within the
contour plots are shown as straight lines forming a
polygon on the plots. The response surface of acetic acid
with respect to HRT and pH showed a rounded ridge
running diagonally on the plot from lower right to the
upper left (Figs. 2 and 3). From statistical inspection of
the coefficients of the models (Eq. (6)), it was shown that
the interaction terms in acetic acid production were not
significant at the 5% level, while other terms were
significant at a 1% confidence level. This meant that the
two variables (HRT and pH) were slightly interdepen-
dent, but the effect on the response surface was not
significant.
The same method was applied to describe the
response surface of the rate of butyric acid production
using the first, interaction, and second-order models
with the augmented trials. The p-values of regression,
lack of fit, and corresponding coefficients were also
tested. Lack of fit was not significant and regression was
significant at a 0.1% a level only for the quadratic
model; therefore, the quadratic model was selected to
describe the response surface of the rate of butyric acid
production which was
Zbutyrate¼ ? 34652:3 þ 2784:7x1þ 5187:4x2
þ 699x3? 1169:3x1x2þ 111x1x3
? 27:9x2x3? 2818:7x2
1? 264x2
2? 5:3x2
3:
ð7Þ
Figs. 4 and 5 show two- and three-dimensional
response surfaces of the quadratic model for butyric
45
50
55
60
65
0.0
0.2
0.4
HRT (d)
0.6
0.8
5.5
6.0
pH
6.5
7.0
Temperature (°C)
2nd order
1st order
Center point
Fig. 1. Experimental design boundary of orthogonal design for
linear and quadratic models.
K. Yang et al. / Water Research 37 (2003) 2467–2477
2471