Multiple response optimization analysis for pretreatments of Tequila’s stillages for VFAs and hydrogen production

Multiple response optimization analysis for pretreatments of Tequila’s
stillages for VFAs and hydrogen production
Froylań M. Espinoza-Escalante a
, Carlos Pelayo-Ortiz a
, Humberto Gutie´rrez-Pulido b
,
Vıćtor Gonza´lez-A´ lvarez a
, Vıćtor Alcaraz-Gonza´lez a
, Andre´ Bories c,*
a
University of Guadalajara, Department of Chemical Engineering, Guadalajara, Mexico
b
University of Guadalajara, Department of Mathematics, Guadalajara, Mexico
c
INRA-Unite´ Expe´rimentale de Pech-Rouge, 11430 Gruissan, France
Received 17 January 2007; received in revised form 27 September 2007; accepted 3 October 2007
Available online 11 December 2007
Abstract
The objective of this work was study the effect of three pretreatments (alkalinization, thermical treatment, and sonication) on Tequi-
la’s stillages hydrolysis process in acidogenesis stage, through the following response variables: soluble chemical oxygen demand (CODs),
total sugar and volatile fatty acids profile and the hydrogen production at the time. The stillages were subject to these pretreatments
(according to a 23
factorial design); afterward they were transferred to a batch reactor at 35 °C and inoculated with an anaerobic digestor
sludge. Multiple response optimization (MRO) analysis was done to find the global optimum for the response variables described above.
This optimum is able to maximize simultaneously all these variables. It was found adequate to be useful hydrolyzing the organic matter
present in Tequila’s stillages. Mathematical models were fitted to observe the estimated effects of pretreatments on each response vari-
able, then the MRO was applied.
Ó 2007 Elsevier Ltd. All rights reserved.
Keywords: Multiple response optimization; Tequila’s stillage; Hydrolysis; Volatile fatty acids; Hydrogen
1. Introduction
Wastewaters from food industries present generally high
organic load and therefore the anaerobic digestion is con-
sidered as the best available technology for their treatment
(Angenent et al., 2004). Alcohol producing industries
affront real problems cause of their high strength stillages.
Of special interest for this study are the Tequila’s stillages,
given the proximal organic load of 60 g/l as chemical oxy-
gen demand due to remaining organic suspended solids and
the dissolved organic compounds (Espinoza-Escalante
et al., 2006). In Mexico, the Bureau for the Regulation of
Tequila, reported a production of 253,000 m3
of Tequila
during 2006 (http://www.crt.og.mx), and it is known that
when producing Tequila, 10 l of stillages are produced
for each liter of this beverage, and then 2.5 million m3
of
Tequila’s stillages were generated for this year. Considering
the high organic load of the wastewater and the high vol-
ume generated per year, this becomes a real problem of
pollution for Mexico.
Given the complex composition of Tequila’s stillages,
the hydrolysis is the rate limiting step in the anaerobic
digestion process. Pretretaments are often required to pro-
mote solubilization of organic matter, such as: chemical
addition, thermal pretreatment, disintegration or mechani-
cal or ultrasonic oxidation, enzymatic or microbial pre-
treatment (Carballa et al., 2004; Benabdallah et al., 2007).
Thermal hydrolysis is referred as the process where
wastes are heated in a range varying from 130 to 180 °C
during 30 to 60 min at the corresponding vapor pressure,
however, according to several authors optimal temperature
is around 170–200 °C (Gosh, 1991; Tanaka et al., 1997).
0960-8524/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2007.10.008
*
Corresponding author. Tel.: +33 (0) 468494400; fax: +33 (0)
468494402.
E-mail address: bories@supagro.inra.fr (A. Bories).
Available online at www.sciencedirect.com
Bioresource Technology 99 (2008) 5822–5829

This process yields a partially solubilized waste and biolog-
ical cells are broken (Bougrier et al., 2004; Carballa et al.,
2004). Even when thermal hydrolysis is energetically
demanding, electrical needs are actually low. The energy
consumed during heating can be optimized in such way
that the energetic balance becomes positive compared to
the conventional treatment of wastes.
In chemical hydrolysis (alcalinization), the waste pH is
increased up to 12, and this process may be used to hydro-
lyze and decompose lipids, hydrocarbons and proteins into
smaller soluble substances such as aliphatic acids, polysac-
charides and aminoacids (Carballa et al., 2004; Chiu et al.,
1997; Luo et al., 2002).
In ultrasonic treatment, cavitational collapse produces
intense local heating and high pressure on liquid–gas inter-
face, turbulence and high shearing phenomena in the liquid
phase, but also formation of OH–
, HO2 and H+
(Bougrier
et al., 2004).
The physical process occurring during cavitation is
almost the same as boiling. The major difference between
these two processes is how is affected the phase change.
In boiling, the vapor pressure of the liquid is increased over
the local pressure to cause a change to gas phase, mean-
while cavitation is caused by a local pressure decrement
under the vapor pressure of the liquid.
This offer the possibility of analyze the effect of pretreat-
ments in solubilizing the complex compounds (mainly poly-
meric sugars) present in Tequila’s stillages, and also to study
their aid in the acidogenic process known already as the first
stage of the anaerobic digestion process in which hydrogen
is produced. It has been shown by several studies that the
response surface methodology (RSM) is a suitable tool to
investigate the acidogenic process (Hwang et al., 2001; Hu
et al., 2006; Wang et al., 2005). However, as we have several
response variables, besides of use RSM it is necessary to use
MRO. Which gives the opportunity of optimize multiple
process variables at once and facilitates the process of tak-
ing decisions at same time that offers a general view of the
anaerobic digestion process of this novel wastewater.
The MRO methodology is oriented to the industrial
processes depending on multiple variables (X1,X2,. . .,Xp),
which have to been controlled in order to achieve a final
goal; this statistical tool leads to achieve a global optimum
for each the individual variables of interest. In this tech-
nique two or more variables are evaluated at once, these
are correlated to generate the global optimum that satisfies
as much as possible the individual optimum of each vari-
able considering acceptable all those values of the variable
inside of the ‘‘desired’’ limits.
The desirability function approach is one of the most
widely used methods for the optimization of multiple
response processes. It is based on the idea that the ‘‘quality’’
of a product or process that has multiple quality character-
istics, with one of them outside of some ‘‘desired’’ limits, is
completely unacceptable. The method finds operating con-
ditions x that provide the ‘‘most desirable’’ response values
(http://www.itl.nist.gov/div898/handbook/, 2006).
The objective of this work was determine the effect of
three pretreatment factors (thermical treatment, alkaliniza-
tion and sonication) on the Tequila’s stillages hydrolysis
process in the acidogenesis stage, through the following
response variables: soluble chemical oxygen demand
(CODs), total sugar and volatile fatty acids profile and
the hydrogen production at the time.
2. Methods
2.1. Tequila’s stillage
Stillages coming from a Tequila distillery of Jalisco,
Mexico, processing Tequila 100% Agave, were frozen at
À20 °C until their utilization. The Tequila’s stillage used
in this study has a total COD of 64,000 mg/l, a sugars con-
tent of 19 g/l and a VFAs concentration of 3 g/l.
2.2. Analytical methods
The pH was measured using a potentiometer (Orion
520A+). For VFAs analysis, samples were centrifuged to
10,000 rpm for 10 min, a 1 ml supernatant aliquot was
taken into a clean 2 ml vial then 1 ll of extern standard
(2-ethyl butyric acid) was added. Samples were thoroughly
mixed with vortex; 1 ll of sample was used for chromatog-
raphy. VFAs were detected and quantified with a Perkin–
Elmer Autosystem XL chromatographer using a capillary
column (WCOT Fused Silica 25 m · 0.32 mm ID Coating
CP-Wax 25CB DF 0.2) with nitrogen as carrier gas; the
temperature of the injector was 205 °C, flame ionization
detector (FID) temperature was 250 °C. The soluble
COD was measured according to the colorimetrical
method of Knetchel (1978), samples were centrifuged to
10,000 rpm for 10 min, a 20 ll aliquot of supernatant was
taken for CODs analysis. Total reducing sugars were mea-
sured with the Phenol–Sulphuric method (Dubois et al.,
1956) and total solids and volatile soluble solids were ana-
lyzed according to the APHA (1998). Gases were measured
using a Perkin–Elmer Autosystem XL with a thermal con-
ductivity detector (TCD) chromatographer using molesieve
13X column with nitrogen as carrier gas.
2.3. Fermentation tests
A 1 l glass vessel was used as model reactor with a work-
ing volume of 540 ml. A 90% vinasse:10% inoculum mix-
ture was added to the reactor and then purged with
gaseous nitrogen for 15 min and then sealed. A tedlar bag
was used for gas capture. Temperature was controlled with
a water bath to 35 °C. The pH was adjusted manually twice
a day, keeping it into a pH range varying from 6.5 to 7.5.
2.4. Inoculum
A digested sludge (14,158 mgVSS/l and 1.74 mgCOD/
mgVSS) from a mesophilic reactor from a brewing plant
was used as inoculum.
F.M. Espinoza-Escalante et al. / Bioresource Technology 99 (2008) 5822–5829 5823

2.5. Experimental design
Tequila’s stillages were pretreated according to 23
exper-
imental design (Neter et al., 1996). Based on previous stud-
ies, three independent variables (factors) were selected: X1
for Alcalinization, pH was adjusted from 3.5 up to 12
and kept for 24 h period, as reported previously by other
authors (Carballa et al., 2004); X2 for Thermal treatment,
stillages were heated up to 150 °C using a block heater
for 30 min, this considering a medium point to that
reported previously; X3 for Ultrasonic Sonication, stillages
were sonicated to 47 kHz for a 30 min period using an
Ultrasonic 104X (NDI) cleaner, since ultrasonic cleaner
can help to destroy cells as observed in previous experi-
ences (Espinoza-Escalante et al., 2006). All these levels
are considered as the high level of a multifactor design
(Table 1). The low levels were equal to current industrial
characteristics of the stillages: pH 7 (even when it is initially
3.5, an adjustment to pH 7 was considered as low level),
temperature of 70 °C and without sonication (see Table
1). The analytical methods for determine each of the inter-
est parameters were described above. Before seed the reac-
tors, stillage’s pH was adjusted up to 7.0 with NaOH. The
models were fitted to estimate the effect of pretreatments on
each response variable through the MRS. To fit the models
were used coded levels for the three factors (see Table 1).
2.6. Computing of the multiple response optimization
Multifactor variance analysis (ANOVA), linear regres-
sion models and response surface plots for each variable were
computed using a statistical software program (Statgraphics
Centurion XV). The best model for each response variable
was obtained by the forward stepwise regression procedure
(Neter et al., 1996). Then the optimum pretreatment condi-
tion for each variable was obtained using that model.
After the analysis for each response variable, a MRO
analysis (Derringer and Suich, 1980) was done; an optimiza-
tion analysis was computed and plotted using Statgraphics
Centurion XV. For MRO analysis, a desirability function
was considered, it involves transformation of each estimated
response variable ðbY iÞ to a desirability value di, where
0 6 di 6 1. The value of di increases as the ‘‘desirability’’ of
the corresponding response increases. The individual desir-
abilities are then combined using the geometric mean,
D ¼ ðd1ðY 1Þ Â d2ðY 2Þ Â . . . Â dkðY kÞÞ1=k
ð1Þ
with k denoting the number of responses. This single value
of D gives the overall assessment of the desirability of the
combined response levels. Clearly the range of D will fall
in the interval [0,1] and D will increase as the balance of
the properties becomes more favorable. D also has the
property that if any di = 0 (that is, if one of the response
variables is unacceptable) then D = 0 (that is, the overall
product is unacceptable). It is for these reasons that the
geometric mean, rather than some other function of the
di’s such as the arithmetic mean, is used (Derringer and
Suich, 1980).
3. Results and discussion
It was analyzed the effect of pretreatments on the solu-
bilization of the raw material present in the Tequila’s stil-
lages (increment of the CODs, initially representing
60.5% of the total COD, 64 g/l), as well as total reducing
sugars (TRS) increment, (reducing sugars, expressed as glu-
cose, contribute to the COD by 1.066 g O2/g TRS, then
TRS contribute to 53% approximately of the CODs). Both
response variables help to evaluate the degree of hydroly-
sis, as the CODs leads to determine the hydrolysis/dissolu-
tion of the organic compounds (polysaccharides and other
organic polymers, yeast cells, etc.) present in the Tequila’s
stillages and the TRS show the hydrolysis of complex sug-
ars, which is related to the availability of the main substrate
for the acidogenic fermentation which produces the meth-
ane precursors. Finally, the effect of pretreatments was
analyzed on the VFAs behavior profile as consequence of
the hydrolysis of the raw material; each of three main acids
Table 1
Experimental design matrix and CODs increments, total reducing sugar (TRS) increments and CODs due to TRS and other compounds, increment in
VFAs, H2 production and H2 yield
Pretreatment
coded values
Pretreatment
real values
DCODs
(g/l)
DTRS
(g/l)
CODs (g/l) due to
reducing sugars
CODs (g/l) due to
other compounds
DTotal
VFAs
(g/l)
Hydrogen
production
(ml)
Total
consumed
sugar (g)*
H2 yield
(mol/mol
glucose)
X1 X2 X3 X1 X2 X3
À1 À1 À1 7 70 0 0 0.0 20.6 18.2 2.8 1172 8.0 1.07
À1 À1 +1 7 70 47 0 0.0 20.6 18.2 0.88 765 7.6 0.74
+1 À1 À1 12 70 0 7.9 0.0 20.6 26.1 6.01 1490 7.1 1.55
À1 +1 À1 7 150 0 29.7 1.6 22.3 36.2 6.18 1080 9.0 0.89
+1 À1 +1 12 70 47 15.9 5.6 26.5 28.2 3.28 65 10.5 0.05
À1 +1 +1 7 150 47 22.3 5.7 26.7 34.5 3.11 1670 11.5 1.07
+1 +1 À1 12 150 0 20.2 5.9 26.9 32.1 0.62 1250 12.3 0.75
+1 +1 +1 12 150 47 14.5 3.3 24.1 29.2 7.2 255 10.9 0.17
X1, alcalinization; X2, thermal pretreatment (°C); X3, cavitation (kHz); DCODs, increment in g/l of the CODs respect to the initial CODs of the control;
DTRS, increment in g/l of the TRS respect to the initial TRS of the control; DVFAs, increment rate in g/l of the VFAs respect to the initial VFAs of the
control.
5824 F.M. Espinoza-Escalante et al. / Bioresource Technology 99 (2008) 5822–5829

were analyzed separately (acetic, propionic and butyric
acids) since their presence in the reactor is considered for
controlling the anaerobic digestion reactors and hydrogen
yield was also evaluated as considered one of the most
important energetic sources nowadays.
3.1. Pretreatments effect on CODs and total reducing sugars
In Table 1 are shown the results of the 23
experimental
design. The increments in soluble chemical oxygen demand
(DCODs) demonstrate that effectively the non-dissolved
compounds in the residual water were hydrolyzed and/or
dissolved in water. In order to evaluate the effect of pre-
treatments on CODs increment rate was fitted a multiple
regression model and applying the forward stepwise regres-
sion procedure the following equation was obtained:
CODs inc ¼ 13:81 þ 7:86 Ã X2 À 5:14 Ã X1 Ã X2
À 2:64 Ã X2 Ã X3 ð2Þ
The determination coefficient R2
was 0.968, the high value
of R2
implies a good model fitting for CODs increase. All
terms in Eq. (2) are statistically significant with a compared
versus a p-value of 0.05. The important effects that emerge
from this analysis are the main effect of temperature (X2)
and the X1X2 and X2X3 interactions. Then temperature
and its interaction with the other factors have an important
effect on increasing CODs values, then hydrolyzing the or-
ganic material.
A simple optimization analysis showed that the maxi-
mum increment in CODs is achieved with temperature
(X2) at the high level, and alkalinization (X1) and cavitation
(X3) both at the low level. And as all low levels in the fac-
tors of Table 1 are equal to current industrial characteris-
tics of the stillages, then only higher temperature is
required to achieve the maximum increment in CODs.
The experimental design matrix with the corresponding
TRS increments (TRS_Inc) in Table 1 were subject to
regression analysis, generating the following equation:
TRS inc ¼ 2:76 þ 0:94 Ã X1 þ 1:36 Ã X2 þ 0:89 Ã X3
À 1:54 Ã X1 Ã X2 Ã X3 ð3Þ
where the determination coefficient (R2
) was 0.922. This va-
lue of R2
indicates that the model had a high accuracy for
fit the experimental results. all the coefficients were statisti-
cally significant because of lower p-values than a with a of
0.05. Then the three factors have an influence on TRS
increments. An optimization analysis showed that maxi-
mum increments can be obtained when alkalinization (X1)
and temperature (X2) are applied without applying cavita-
tion. In this treatment a maximal increase rate of 5.71 g/l
was estimated compared to the 5.9 g/l observed, Table 1.
3.2. Pretreatments effect on acidogenesis
In the acidogenesis of carbohydrates, H2 is also
generated as an important product in addition to VFA
and alcohols (Wang et al., 2005). Some stoichometric equa-
tions has been defined to describe the metabolism of glu-
cose for H2 production (acidogenesis) where acetate and
butyrate are the final products depending on the environ-
mental conditions (Angenent et al., 2004), this is why,
TRS consumption is evaluated in this study. From results
of Table 1 it is observed that total consumed sugars (g)
can be described in two groups, the first one when no pre-
treatment or only one pretreatment is applied the total con-
sumed sugar are below 10 g (18.5 g/l), and the second
group is observed when two or more pretreatments are
applied with a consumption of about 11 g (21 g/l), this rein-
force the idea that sugars are hydrolyzed and therefore
more consumed.
The experimental design matrix and the corresponding
sugar consumption (g), Table 1, were subject to regression
analysis, the model was then depurated by the forward
stepwise regression procedure, the obtained equation is
shown below Eq. (4); the determination coefficient was
0.981, the p-values for each coefficient are below to a with
a of 0.05
TRS cons ¼ 9:61 þ 0:59 Ã X1 þ 1:31 Ã X2 þ 0:51 Ã X3
À 0:96 Ã X1 Ã X2 Ã X3 ð4Þ
Then in the best treatment (+1,+1,À1), the estimated max-
imal total consumed sugars was 11.96 g, considering the
working reactor volume in this study (540 ml), the total
consumed sugars was 22.8 g/l, which is in good agreement
with that reported by other authors. In their work Wang
et al. (2005), reported that a sucrose concentration of
24.2 g/l is advised for hydrogen production at mesophilic
temperature (33.5 °C). By their own, Cheong and Hansen
(2006), reported an optimal glucose concentration in the
reactor of 21.3 g/l for hydrogen production when working
in batch after optimization trials were carried out. Despite
these results, not all the process variables are considered in
these studies.
Respect to the volatile fatty acids production, Table 1 is
more difficult to observe a behavior pattern since in some
cases the methanization process has began earlier, as
shown in Figs. 1 and 2 (continuous line), and therefore at
the end of 96 h period time (time of the experiment) some
of the produced VFAs have been consomed (e.g., acetic
acid) whilst in other cases they remained or were increased
until the end of the experiment Fig. 2 (dotted line), Figs. 3
and 4. A complete profile of VFAs can be observed in
Fig. 5.
In order to evaluate the effect of pretreatments on
VFAs (acetic, propionic and butyric acids) increment rates,
a multiple regression model was fitted to each variable;
then forward stepwise regression procedures were per-
formed. The three equations representing each acid
increment rate are shown below. All coefficients in these
equations are statistically significant with a below
a = 0.05:

Acetic acid ¼ 1:59 þ 0:52 Ã X1 À 0:36 Ã X2
þ 0:40 Ã X1 Ã X3 þ 0:29 Ã X2 Ã X3 ð5Þ
Propionic acid ¼ 1:41 þ 0:48 Ã X1 À 0:45 Ã X2
À 0:26 Ã X3 þ 0:37 Ã X2 Ã X3 ð6Þ
Butyric acid ¼ 2:20 þ 0:12 Ã X3 À 0:28 Ã X1 Ã X2
þ 0:24 Ã X1 Ã X3 þ 0:14 Ã X2 Ã X3 ð7Þ
According to these equations, alkalinization (X1) has a po-
sitive effect on acetic and propionic acid and no effect on
butyric acid production; temperature (X2) has a negative
effect on acetic and propionic acids accumulation; cavita-
tion (X3) has a negative effect on propionic acid increment.
Combined pretreatments have positive effect on VFAs
increments, but alkalinization–temperature (X1 * X2) has
a negative effect on butyric acid accumulation. This could
mean in general that combined pretreatments have positive
effect in hydrolysis thus in VFAs production.
Maximal VFAs increment rates were calculated around
250–300%. According to the results of Kim et al. (2004),
the use of thermal pretreatment combined with enzymatic
pretreatment, a VFAs production increase of 340–380%
are observed when fermenting food wastes. Differences
between the results of our study and those of Kim et al.
(2004) could be due to the fact that these lasts calculated
total VFAs increase and applied other pretreatments than
the used in this study. However, Kim et al. (2004) reported
a maximal VFAs accumulation at 5 days (120 h), similar
results were observed in our study.
3.3. Pretreatments effect on hydrogen production
An important objective of this study was to determine
the pretreatments which improve hydrogen production.
The experimental results of the H2 production (ml) are
Fig. 1. Volatile fatty acids profile (acetic, propionic and butyric) through
the time in different pretreatments: continuous line (X1 = À1, X2 = +1,
X3 = À1) and dotted lines (X1 = À1, X2 = +1, X3 = +1).
the time in different pretreatments: continuous line (X1 = +1, X2 = +1,
X3 = À1) and dotted lines (X1 = +1, X2 = +1, X3 = +1).
the time in different pretreatments: continuous line (X1 = À1, X2 = À1,
X3 = À1) and dotted lines (X1 = À1, X2 = À1, X3 = +1).
the time in different pretreatments: continuous line (X1 = +1, X2 = À1,
X3 = À1) and dotted lines (X1 = +1, X2 = À1, X3 = +1).

shown in Table 1. Applying multiple regression analysis,
the final model (Eq. (8)) was obtained
H2ðmlÞ ¼ 968:38 À 203:38 Ã X1 À 279:62 Ã X3
À 325:38X1 Ã X3 ð8Þ
The corresponding determination coefficient was 0.792
with all terms being statistically significant because of
lower p-values than a with a of 0.05. According to the sta-
tistical analysis, the model shows that applying alkaliniza-
tion at high level (X1 = +1) and cavitation at low level
cavitation (X3 = À1) is possible to get a higher H2 produc-
tion, with a slight difference to that treatment were alkalin-
ization is set to À1 and cavitation to +1, then opposed
levels of alkalinization and cavitation are advised in order
to get high hydrogen productions. This agrees with the fact
that acetic acid accumulation, the main substrate for H2
production (Lin and Chang, 1999; Lay, 2000), is mainly af-
fected by alkalinization and cavitation, Eq. (5).
Among the newer works related to hydrogen produc-
tion, Lee et al. (2006), developed a carrier-induced granular
sludge bed (CIGSB) bioreactor able to significantly
enhance H2-producing efficiency by gaining a high biomass
concentration of up to 26 g VSS/l even when it was oper-
ated at a low HRT of 0.5 h, achieving hydrogen yields
varying from 2 to 4 mol H2/mol sucrose. Mu et al. (2006)
in their experiments achieved maximal hydrogen yields in
the range of 1–2 mol H2/mol glucose. Fang et al. (2006)
reported a hydrogen production rate of 346 ml H2/g of car-
bohydrate, using rice as model carbohydrate. Zhang et al.
(2006), reported hydrogen yields of 0.9 mol H2/mol of glu-
cose in a trickle bed reactor; working in CSTR’s Iyer et al.
(2004), reported H2 yields of 1.6 mol H2/mol glucose;
working in batch, Wang et al. (2005) found an optimal
hydrogen yield of 3.728 mol H2/mol sucrose. Finally, in
our study it has been demonstrated that Tequila’s stillages
have a great potential to be used as substrate for hydrogen
production, since a production rate of 1.55 mol H2/mol
glucose can be achieved when the optimal pretreatment is
applied, result that is in good agreement with the results
reported by other authors as mentioned above.
Given that Tequila’s stillages have a great potential for
hydrogen production and considering the economical anal-
ysis presented by Van Ginkel et al. (2005) and the annual
generation of this wastewater (50 million liters;
www.crt.org.mx), the hydrogen produced could represent
an annual income of $67,000 USD. Moreover, the purify-
ing process must be considered.
3.4. Multiple response analysis for acidogenesis
The MRO was applied in order to determine the global
optimum treatment, and then to achieve the objective of
the study: minimize propionic acid increment, usually
desired in anaerobic digestion processes as reported by
authors like Wang et al. (2006), and maximize all the
remaining variables, considering as the most important
variables hydrogen production. With these restrictions
was transformed each estimated response variable Eq. (2)
to Eq. (8) to a desirability value di where 0 6 di 6 1. And
then according to Eq. (1) was obtained the desirability
function.
According to the optimization analysis, it was observed
that the higher desirability value, 0.724, is achieved with
the following combination of pretreatments in codified lev-
els: X1 = À1, X2 = +1 and X3 = +1, this is, Tequila’s stil-
lages must be subject to alkalinization up to pH 7,
thermical treatment (Temperature) of 150 °C/30 min and
cavitation treatment at 47 kHz/30 min; time periods can
be addressed since only the value of the parameter per se
is taking in count for the analysis. According to predictions
with Eq. (8) (hydrogen production) for the desirability
function, it was observed that from the eight possible treat-
ments (only vertices), in six cases desirability value equals
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
HAc
HPr
HBu
TotalAGV's
-1-1-1 -1-1+1 +1-1-1 -1+1-1 +1-1+1 -1+1+1 +1+1-1 +1+1+1
Pretreatment
Production(g/L)
Produced/Consumed
Fig. 5. Volatile fatty acids profile after the fermentation process. Note: negative columns show a consumption.

to zero (e.g., the low level of the three factors X1 = À1,
X2 = À1, X3 = À1), but a high values of desirability can
be observed when temperature (X2) and cavitation (X3)
are applied when alkalinization is set at low level. Fig. 6
shows the contour plot where each curve connecting points
of temperature (X2) and Cavitation (X3) where the response
has a same particular value. It can be seen that X2 has a
high impact on desirability increasing desirability as the
level of X2 increases and slight increase on desirability is
observed as X3 level does.
In previous studies oriented to the optimization of the
anaerobic digestion process, and mainly to hydrogen pro-
duction optimization, it has been proposed a two-step
anaerobic digestion concept coming from the fact that this
biological process involves generally two stages: the acido-
genesis and the phase of methane production (Alexiou and
Panter, 2004). The two stages design takes advantage of the
phase’s separation phenomenon occurring naturally at dif-
ferent kinetic values, and then the use of separated units for
hydrolysis and methanogenesis would imply cost reduc-
tions and efficiency improvements in hydrogen and meth-
ane production as well as in the stability of the anaerobic
digestion plant (Alexiou and Panter, 2004).
In addition to the above mentioned advantages of the
two-stage process, operational differences can be attached.
In classical anaerobic digestion process (mono-stage),
organic load must be controlled in order to avoid the VFAs
accumulation in the reactor, preventing a fermentation
accident in the system, as other organic acids different from
acetic acid are slower degraded to methane than acetic acid
is. In the counterpart, in the two-step anaerobic digestion,
in the first stage (acetogenesis) VFAs accumulation in
important quantities is permitted and desired (Hu et al.,
2006; Hwang et al., 2001; Mu et al., 2006; Wang et al.,
2005), so the organic load to treat can be higher, this profile
corresponds to the fermentation accident mentioned above,
preparing the methanogenic archeas to a mixture of acids
(propionic and butyric) and a noteworthy development of
the obligated hydrogen producing acetogens can be
observed, these will degrade the VFAs in syntrophic asso-
ciation with methanogenic archeas in second stage (Bories,
1980).
The acetogenesis stage however, is not as robust as it
could be thought, even when it is feasible to treat high
organic loads and VFAs accumulation is permitted, and
the hydrolysis of the raw material is a primary concern in
this stage. This is why some considerations must be done
in order to: (i) achieve the maximum hydrolysis of the poly-
meric compounds for the maximum acidification (main
indicator of the acetogenesis) in the reactor, and (ii) find
the best VFAs profile for the desired objective.
The effect of simple operational parameters on acido-
genesis, is referred in literature (e.g., Decloux and Bories,
2002; Swamy et al., 1998), but a global optimization is
needed in order to observe the effect of a group of factors
over a group of variables at once. The MRO, however, is
able to respond to the increasing demands of the scientific
and technological questions oriented to process optimiza-
tion, given that this powerful statistical tool can identify
the best operational parameters, for the process under
study, in a well defined range of desired values for many
variables at once, as shown above, the MRO was an useful
tool to determine which combination of pretreatments
accomplished, as much as possible, all the desired objec-
tives of this study at once including the imposed restric-
tions, having an overall accomplishment of 72.4% when
considering the global optimum (X1 = À1, X2 = +1 and
X3 = +1). Optimal values for each variable under study
are shown in Table 2.
4. Conclusions
Tequila’s stillages represent an important pollution
problem in Mexico given the increasing demand of Tequila
in the international market, with the consequent generation
of stillages in the order of 10 liters per liter of produced
Tequila. According to the results presented in this work,
it was found that Tequila’s stillages can become a second-
ary income to the Mexican industries given their potential
for hydrogen production.
In our study, it was observed that temperature and cav-
itation pretreatments highly improve the digestion poten-
tial by hydrolyzing or dissolving the complex compounds
of the Tequila’s stillages and therefore the consumption
rate of reducing sugars and the acidogenesis process. The
aim of the statistical tools, in this case MRO, was useful
in helping to observe these results globally at same time
that facilitates the decision taking, as well it offers the
Fig. 6. Contour plot for desirability given the possible combinations of
two pretreatments: temperature and cavitation, when pH (alkalinization)
level is fixed to À1.0.
Table 2
Optimum values for the multiple response optimization
Response Optimum value
CODs increment (g/l) 21.51
Total sugar consumption (g/l) 21.87
Acetic acid increment rate 0.61
Propionic acid increment rate 0.60
Butyric acid increment rate 2.50
Hydrogen accumulated production (ml) 1217

opportunity to get models which correlate the available
information in order to make improving changes in the
process for the future or well changes oriented to the
desired objectives.
Acknowledgements
The stillages supply by Tequila HerraduraÒ
is really
appreciated. We acknowledge the technical work of Mr.
Jose´ Navarro Corona whose invaluable work helped to de-
velop this study. The gratitude of one of the authors is ex-
tended to CONACyT for the Ph.D. scholarship 184794.
References
Alexiou, I.E., Panter, K., 2004. A review of two phase applications to
define best practice for the treatment of various waste streams. In:
Anaerobic Digestion 10th World Congress, Sept 2004. Montreal,
Quebec, Canada.
Angenent, L.T., Karim, K., Al-Dahhan, M.H., Wrenn, B.A., Dominguez-
Espinoza, R., 2004. Production of bioenergy and biochemicals from
industrial and agricultural wastewater.
APHA, 1998. Standard Methods for the Examination of Water and
Wastewater, 20th ed. American Public Health Association, Washing-
ton DC.
Benabdallah, T.E.H., Dosta, J., Ma´rquez-Serrano, R., Mata-A´ lvarez, J.,
2007. Effect of ultrasound pretreatment in mesophilic and thermophilic
anaerobic digestion with emphasis on naphthalene and pyrene
removal. Water Res. 41, 87–94.
Bories, A., 1980. Fermentation me´thanique avec se´paration des phases
acidoge`ne et me´thanogene, appliqueé au traitement des effluents a` forte
charge polluante (distilleries). Ann. Technol. Agr. 29 (3), 509–528.
Bougrier, C., Carre`re, H., Battimelli, A., Delgene`s, J.P., 2004. Effect of
various pre-treatment on waste activated sludge in order to improve
matter solubilisation and anaerobic digestion. In: Anaerobic Digestion
10th World Congress, Sept 2004. Montreal, Quebec, Canada.
Carballa, M., Omil, F., Lema, J.M., 2004. Improvement of anaerobic
digestion operation and digested sludge characteristics using a chem-
ical and a thermal pretreatment. In: Anaerobic Digestion 10th World
Congress, Sept 2004. Montreal, Quebec, Canada.
Cheong, D.Y., Hansen, C.L., 2006. Acidogenesis characteristics of
natural, mixed anaerobes converting carbohydrate-rich synthetic
wastewater to hydrogen. Process Biochem. 41 (8), 1736–1745.
Chiu, Y.C., Chang, C.N., Lin, J.G., Huang, S.J., 1997. Alkaline and
ultrasonic pretreatment of sludge before anaerobic digestion. Water
Sci. Technol. 36 (11), 155–162.
Decloux, M., Bories, A., 2002. Sillage treatment in the French alcohol
fermentation industry. Int. Sugar. J. 104 (1247), 509–517.
Derringer, G., Suich, R., 1980. Simultaneous optimization of several
response variables. J. Qual. Technol. 12 (4), 214–219.
Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956.
Colorimetric method for determination of sugars and related sub-
stances. Anal. Chem. 28 (3), 350–356.
Espinoza-Escalante, F.M., Pelayo-Ortı´z, C., Bories, A., Ruı´z-Pe´rez, M.,
Navarro-Corona, J., Gonza´lez-A´ lvarez, V., A´ lcaraz-Gonza´lez, V.
2006. Efecto del pretratamiento de vinazas tequileras en el proceso
de digestioń anaerobia. In: Memorias del XXVII Encuentro Nacional
de la AMIDIQ, May 2006, Ixtapa, Mexico.
Fang, H.H.P., Li, C., Zhang, T., 2006. Acidophilic biohydrogen produc-
tion from rice slurry. Int. J. Hydrogen Energ. 31 (6), 683–692.
Gosh, S., 1991. Pilot-scale demonstration of two-phase anaerobic diges-
tion of activated sludge. Water Sci. Technol. 23, 1179–1188.
Hu, Z.H., Yu, H.Q., Zheng, J.C., 2006. Application of response surface
methodology fro optimization of acidogenesis of cattail by rumen
cultures. Bioresource Technol. 97, 2103–2109.
Hwang, S., Lee, Y., Yang, K., 2001. Maximization of Acetic Acid
Production in Partial Acidogenesis of Swine Wastewater. John Wiley
and Sons, Inc., pp. 521–529.
Iyer, P., Bruns, M.A., Zhang, H., Van Ginkel, S., Logan, B.E., 2004. H2-
producing bacterial communities from a heat-treated soil inoculum.
Appl. Microbiol. Biotechnol. 66, 166–173.
Kim, H.J., Choi, Y.G., Kim, D.Y., Chung, T.H., 2004. Effect of
pretreatment on acid fermentation of organic solid waste. In: Anaer-
obic Digestion 10th World Congress, Sept 2004. Montreal, Quebec,
Canada.
Knetchel, 1978. A more economical method for the determination of
chemical oxygen demand. Water Poll. Control 116, 25–27.
Lay, J.J., 2000. Modeling and optimization of anaerobic digested sludge
converting starch to hydrogen. Biotechnol. Bioeng. 68, 269–278.
Lee, K.S., Lo, Y.C., Lin, P.J., Chang, J.S., 2006. Improving biohydrogen
production in a carrier-induced granular sludge bed by altering
physical configuration and agitation pattern of the bioreactor. Int. J.
Hydrogen Energ. 31, 1641–1657.
Lin, C.Y., Chang, R.C., 1999. Hydrogen production during the anaerobic
acidogenic conversion of glucose. J. Chem. Technol. Biotechnol. 74
(6), 498–500.
Luo, H.W., Zhang, H., Suzuki, T., Hattori, S., Kamagata, Y., 2002.
Differential expression of methanogenesis genes of Methanothermob-
acter thermoautotrophicus (Formerly Methanobacterium thermoauto-
trophicum) in pure culture and in cocultures with fatty acid – oxidizing
syntrophs. Appl. Env. Microbiol. 68 (3), 1173–1179.
Mu, Y., Zheng, X.J., Yu, H.Q., Zhu, R.F., 2006. Biological hydrogen
production by anaerobic sludge at various temperatures. Int. J.
Hydrogen Energ. 31 (6), 780–785.
Neter, J., Kutner, M.H., Nachtsheim, C.J., Wasserman, W., 1996. Applied
Linear Statistical Models. McGraw-Hill, New York.
Swamy, C.N., Jain, N., Kumar, P.y., Shrivastava, A.K., 1998. Effect of
culture parameters on degradation of distillery wastewater by activated
sludge microorganisms. Adv. Food Sci. 20 (1/2), 59–63.
Tanaka, S., Kobayashi, T., Kamiyama, K., Signey-Bildan, M.L., 1997.
Effects of thermochemical pretreatment on the anaerobic digestion of
waste activated sludge. Water Sci. Technol. 35 (8), 209–215.
Van Ginkel, S.W., Oh, S.E., Logan, B.E., 2005. Biohydrogen gas
production from food processing and domestic wastewaters. Int. J.
Hydrogen Energ. 30, 1535–1542.
Wang, G., Mu, Y., Yu, H.Q., 2005. Response surface analysis to evaluate
the influence of pH, temperature and substrate concentration on the
acidogenesis of sucrose–rich wastewater. Biochem. Eng. J. 23, 175–
184.
Wang, L., Zhou, Q., Li, F.T., 2006. Avoiding propionic acid accumulation
in the anaerobic process for biohydrogen production. Biomass
Bioenerg. 30, 177–182.
Zhang, H., Bruns, M.A., Logan, B.E., 2006. Biological hydrogen
production by Clostridium acetobutylicum in an insaturated flow
reactor. Water Res. 40 (2006), 728–734.

Multiple response optimization analysis for pretreatments of Tequila’s stillages for VFAs and hydrogen production

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Multiple response optimization analysis for pretreatments of Tequila’s stillages for VFAs and hydrogen production

Similar to Multiple response optimization analysis for pretreatments of Tequila’s stillages for VFAs and hydrogen production (20)

Recently uploaded

Recently uploaded (20)

Multiple response optimization analysis for pretreatments of Tequila’s stillages for VFAs and hydrogen production