Electrochemical oxidation of table olive processing wastewater over boron-doped diamond electrodes: Treatment optimization by factorial design

ARTICLE IN PRESS WAT E R R E S E A R C H

42 (2008) 1229 – 1237

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Electrochemical oxidation of table olive processing wastewater over boron-doped diamond electrodes: Treatment optimization by factorial design Anastasios Deligiorgis, Nikolaos P. Xekoukoulotakis, Evan Diamadopoulos, Dionissios Mantzavinos Department of Environmental Engineering, Technical University of Crete, GR-73100 Chania, Greece

art i cle info

ab st rac t

Article history:

The electrochemical treatment of an effluent from edible olive processing over boron-

Received 2 August 2007

doped diamond electrodes was investigated. The effect of operating conditions, such as

Received in revised form

initial organic loading (from 1340 to 5370 mg/L chemical oxygen demand (COD)), reaction

19 September 2007

time (from 30 to 120 min), current intensity (from 5 to 14 A), initial pH (from 3 to 7) and the

Accepted 20 September 2007

use of 500 mg/L H2O2 as an additional oxidant, on treatment efficiency was assessed

Available online 25 September 2007

implementing a factorial experimental design. Of the five parameters tested, the first three

Keywords: Boron-doped diamond Electrolysis Factorial design Olive processing Wastewater

had a considerable effect on COD and total phenols removal, while the other two were statistically insignificant. In most cases, high levels of phenols degradation and decolorization were achieved followed by moderate mineralization. The analysis was repeated at more intense conditions, i.e., initial COD up to 10,000 mg/L, reaction times up to 240 min and current up to 30 A; at this level, the effect of treatment time and applied current was far more important than the starting COD concentration. Treatment for 14 h at optimal conditions (30 A and an initial loading of about 10,000 mg/L) led to 73% COD removal with a zero-order kinetic constant of 8.5 mg/(L min) and an energy consumption efficiency of 16.3 g COD/(m3 A h). & 2007 Elsevier Ltd. All rights reserved.

1.

Introduction

Wastewater generated from olive processing comes from either olive milling for oil extraction or edible olives manufacturing. Both agro-industrial activities, which are of particular economic importance to Mediterranean countries like Spain, Italy and Greece, result in effluents that are characterized by high organic content, seasonal generation and the presence of classes of pollutants such as polyphenols that may exhibit antimicrobial, ecotoxic and phytotoxic properties (Kyriacou et al., 2005; Aggelis et al., 2001). The treatment of olive mill wastewaters (OMW) has received enormous attention over the past several years and various decontamination Corresponding author. Tel.: +30 28210 37797; fax: +30 28210 37852.

E-mail address: [email protected] (D. Mantzavinos). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.09.014

technologies based on biological (aerobic and anaerobic), advanced oxidation, chemical and separation processes have been proposed by several research groups as summarized in recent review publications (Paraskeva and Diamadopoulos, 2006; Mantzavinos and Kalogerakis, 2005). Conversely, relatively fewer studies have been dealing with wastewaters from table olive processing wastewater (TOPW). Table olive processing occurs through a series of steps, namely initial olive cleaning, debittering, washing, fermentation and packing; all these steps generate waste streams which, alongside the wash waters from tanks, machinery, etc., result in TOPW quantities of about 3.9–7.5 and 0.9–1.9 m3 per ton of green and black olives, respectively

ARTICLE IN PRESS 1230

WAT E R R E S E A R C H

42 (2008) 1229– 1237

(Kopsidas, 1992). The composition of TOPW is similar to that of OMW although the organic strength of the latter is far greater than that of the former. TOPW may have chemical oxygen demand (COD), biochemical oxygen demand) and suspended solids contents up to about 18, 6.6 and 0.4 g/L respectively (Kopsidas, 1992), while typical respective values for OMW are 170, 110 and 9 g/L (Paraskeva and Diamadopoulos, 2006). Unlike OMW, TOPW also contain high concentrations of sodium chloride and sodium hydroxide which are used for olive debittering and fermentation. The treatability of TOPW by biological means was investigated in several previous studies under aerobic (Kyriacou et al., 2005; Aggelis et al., 2001; Beltran-Heredia et al., 2000a, b; Brenes et al., 2000) and anaerobic (Aggelis et al., 2001) conditions. Depending on the operating conditions in question, aerobic treatment was capable of reducing the COD content by as much as about 70–90% and this was accompanied by low-tomoderate degradation of phenols and other aromatics. Conversely, anaerobic treatment led to partial (i.e. about 50%) COD removal accompanied by a 12% phenolic reduction. In an attempt to improve the biodegradability of TOPW, several studies have dealt with the use of advanced oxidation processes as a suitable pretreatment method to reduce the effluent’s COD and phenolic contents. Given the high alkalinity of table olive processing effluents, ozonation appears to be an attractive treatment option since ozone can react rapidly with polyphenols through a combination of direct attack and hydroxyl-radical-induced reactions. This has been demonstrated in several studies where ozone pretreatment was capable of removing most of the phenolic compounds present in TOPW and subsequently enhancing biodegradability (Benitez et al., 2002, 2001a; Beltran-Heredia et al., 2000c; Rivas et al., 2000; Beltran et al., 1999). Coupling ozonation with UV irradiation (with or without H2O2) has also been studied (Benitez et al., 2002, 2001a; Beltran et al., 1999) and found capable of improving the efficiency of single ozonation; this was attributed to increased occurrence of hydroxyl radical reactions. Besides ozonation, Fenton and photo-Fenton oxidation (Rivas et al., 2003; Benitez et al., 2001b) as well as wet air oxidation (Rivas et al., 2001) have also been employed to treat TOPW and, in all cases, preoxidized effluents were more readily biodegradable than the original effluent. Electrochemical technologies such as electrooxidation, electrocoagulation and electroflotation have been widely used in water and wastewater treatment and several applications have been recently reviewed by Chen (2004). Over the past two decades, electrooxidation has been widely investigated for wastewater treatment over electrodes made of several different materials such as graphite, Pt, TiO2, SnO2, IrO2, RuO2, PbO2 and several Ti-based alloys. In recent years, a new type of electrode material, namely boron-doped diamond (BDD) has received growing attention for pollutants oxidation since it exhibits significant chemical and electrochemical stability, good conductivity as well as it achieves increased rates of effluent mineralization with very high current efficiencies (Canizares et al., 2007; Chen et al., 2005). Although the electrochemical oxidation over BDD electrodes of several classes of organic pollutants has been studied extensively (Chen et al., 2005; Chen, 2004), process application for agro-industrial effluent treatment is limited; only recently has the electrooxidation of OMW over

BDD anodes been reported (Canizares et al., 2007, 2006). Nevertheless and to the best of our knowledge, the electrochemical treatment of TOPW over BDD electrodes or, indeed, any other anodic materials has not been investigated yet. The aim of this work was to study the electrochemical oxidation of TOPW over a BDD anode regarding the effect of various operating conditions such as current intensity, initial concentration, effluent pH, contact time and the addition of hydrogen peroxide on the conversion of COD, phenols, aromatics and color. A factorial design methodology was adopted to determine the statistical significance of each parameter as well as optimal treatment conditions.

2.

Materials and methods

2.1.

TOPW

The effluent used in this study was taken from a table olivemanufacturing plant located in the region of Chania, Western Crete, Greece. The process through which the effluent was generated comprises mixing 130–140 kg of black olives of the Kalamai variety, 10 kg of sodium chloride, 0.24 kg of calcium chloride and 1 kg of lactic acid in 90–100 kg of water. The original effluent’s major properties are as follows: COD ¼ 60,000 mg/L, total phenols (TP) ¼ 5200 mg/L, conductivity ¼ 111.5 mS/cm and pH ¼ 4.5. The effluent has a dark brown–black color and contains a substantial fraction of aromatic compounds. In all cases, the original effluent was diluted with water to achieve the desirable initial concentration and then fed to the electrolytic cell. For those samples where the starting concentration was less than about 5500 mg/L COD, the effluent’s pH was neutral.

2.2.

Electrochemical degradation experiments

Experiments were conducted in a DiaCells (type 100) singlecompartment electrolytic flow-cell manufactured by Adamant Technologies (Switzerland). Two circular electrodes of 0.1 m diameter made of BDD on silicon were used as the anode and cathode; each electrode area was 70 cm2 and the distance between them 0.01 m. In a typical run, the diluted effluent was batch loaded in a vessel and continuously recirculated in the cell through a peristaltic pump operating at a maximum flow rate of 0.02 m3/ min. In all cases, the working volume was 10 L. A spiral coil immersed in the liquid and connected to tap water supply was used to remove the heat liberated from the reaction. All experiments were conducted at ambient temperature; nonetheless, temperature was found to increase slowly with treatment time with the extent of temperature rise being dependent of the operating conditions employed. However, temperature never exceeded 34 1C at the end of each experiment which, in most cases, lasted for 120 min. For those experiments where the diluted effluent’s ambient pH (which was 7 at initial concentrations of less than about 5500 mg/L COD) was adjusted to acidic conditions, the appropriate amount of 98% w/w H2SO4 was added. In those cases where hydrogen peroxide was used as an extra oxidant, the appropriate amount of a 35% w/w solution was added to achieve a 500 mg/L H2O2 initial

ARTICLE IN PRESS WAT E R R E S E A R C H

cially available digestion solution containing potassium dichromate, sulfuric acid and mercuric sulfate (Hach Europe, Belgium) and the mixture was then incubated for 120 min at 150 1C in a COD reactor (Model 45600-Hach Company, USA).

concentration. Experiments were conducted at current intensity values between 5 and 30 A, initial TOPW concentrations between 1340 and 10,000 mg/L COD, reaction times up to 14 h and pH values between 3 and 7.

2.3.

Analytical methods

2.3.1.

Chemical oxygen demand (COD)

Table 3 – Percentage of COD, TP, aromatics and color removal at various conditions

COD was determined by the dichromate method. The appropriate amount of sample was introduced into commer-

Table 1 – Independent variables of the 25 experimental design Level of value  +

1231

42 (2008) 1229 – 1237

Initial COD (mg/L)

Current (A)

pH0

Reaction time (min)

Oxidant concentration (mg/L)

1340 5370

5 14

3 7

30 120

0 500

Order of experiments

COD

TP

Aromatics

Color

30 12 2 21 20 5 9 7

12.7 22 42.2 14.8 11.4 21.9 40.4 17.8

84.5 88.5 100 91 56.1 71.4 84.5 89.1

24.4 45.3 43.7 53.7 35.4 51.8 52.9 44.8

78.9 93.6 89.5 89.4 87.3 92.1 93.8 89.3

Order of experiments as in Table 2.

Table 2 – Design matrix of the 25 factorial experimental design and observed response (Y1: mass of COD removed per liter; Y2: mass of TP removed per liter) Order of running experiments

19 26 18 22 3 23 30 8 28 24 10 17 12 6 2 21 27 4 29 14 11 31 20 1 25 32 13 16 5 15 9 7

Level value of each variable in the experimental run

COD

Current

pH0

Time

Oxidant

 +  +  +  +  +  +  +  +  +  +  +  +  +  +  +  +

  + +   + +   + +   + +   + +   + +   + +   + +

    + + + +     + + + +     + + + +     + + + +

        + + + + + + + +         + + + + + + + +

                + + + + + + + + + + + + + + + +

Y1 (mg/L) of COD removed

Y2 (mg/L) of TP removed

75 234 176 596 64 85 160 192 437 469 623 1427 272 320 533 767 91 362 208 469 80 256 160 256 362 873 639 1257 304 533 565 1001

24 47 61 135 46 44 110 177 60 160 100 428 115 237 130 473 59 92 78 138 43 112 73 195 116 260 119 445 93 294 110 463

ARTICLE IN PRESS 1232

WAT E R R E S E A R C H

42 (2008) 1229– 1237

COD concentration was measured colorimetrically using a DR/2010 spectrophotometer (Hach Company, USA).

2.3.2.

Total phenols (TP)

The TP content was determined colorimetrically at 725 nm on a Shimadzu UV 1240 spectrophotometer using the Folin– Ciocalteau reagent according to the procedures described in detail elsewhere (Atanassova et al., 2005). Gallic acid monohydrate was used as standard to quantify the concentration of phenols in TOPW.

2.3.3.

UV–Vis absorbance

Sample absorbance was scanned in the 200–800 nm wavelength band on a Shimadzu UV 1240 spectrophotometer. Changes in sample absorbance at two specific wavelengths, i.e., 567 and 275 nm were monitored to assess the extent of decolorization and aromatics removal respectively that had occurred during treatment.

3.

Results and discussion

There are two ways one can investigate the effect of a large number of variables. The most commonly used method involves the variation of one variable while keeping the other variables constant, until all variables have been studied. This methodology has two disadvantages: First, a large number of experiments are required, and second it is likely that the combined effect of two or more variables may not be identified. In this work, a statistical approach was chosen based on a factorial experimental design that would allow us to infer about the effect of the variables with a relatively small number of experiments. The independent variables of the experimental design are presented in Table 1. Each one of the five variables received two values, a high value (indicated by the+sign) and a low value (indicated by the–sign). Regarding the initial pH which took values of 7 (natural pH of the diluted effluent) and 3 (after adding H2SO4), it should be noticed that the solution was not buffered to the aforementioned values. However, pH was monitored constantly throughout the reaction (with a Toledo 225 pH meter) showing that only marginal changes (of 0.2 pH units) had occurred between the initial and final (i.e. after 120 min) solutions. The experimental design followed in this work was a full 25 experimental set, which required 32 experiments. The order each experiment was performed was selected randomly. This order of experimental runs, along with values of each independent variable for each run, is presented in Table 2. Table 2 also shows the obtained results in terms of mg/L of COD oxidized (depended variable or response Y1) and mg/L of TP oxidized (depended variable or response Y2). On the assumption that TP are represented by gallic acid (monohydrate), the stoichiometry of its reaction to carbon dioxide and water dictates that 100 mg of gallic acid would require 102 mg oxygen for the complete oxidation; therefore, the last column of Table 2 practically corresponds to the concentration of COD oxidized due to the phenolic fraction of the effluent. During the early stages of the reaction (i.e. at low current intensities and/or short treatment times), TP appear to degrade readily and a significant part of the recorded

overall COD reduction is due to the removal of TP; for instance, this is more pronounced for the runs 1, 3, 6, 8, 15, 27 and 30 where COD reduction due to TP accounts for about 65–90% of the total COD decrease. As the electrochemical reactions proceed, other compounds that were originally present in TOPW or formed as secondary reaction byproducts also undergo oxidation and this contributes to the increased COD reduction. Table 3 shows the extent of aromatics degradation and effluent decolorization for representative experiments of Table 2 alongside the respective extent of TP degradation and COD removal. As clearly seen, TP are easily electrochemically oxidized, while other aromatic compounds present in the effluent are quite resistant to oxidation. Noticeably, increased TP degradation is accompanied by a consistently high effluent decolorization; this is

Table 4 – Estimated effects of the 25 factorial design for the oxidation of COD and TP Effect

Value of effect Oxidation of COD

Oxidation of TP

Average effect

432.69717.46

Main effects COD Current pH Reaction time Oxidant

271.75734.91 263.25734.91 171.88734.91 432.37734.91 61.62734.91

147.69710.80 89.56710.80 24.56710.80 135.56710.80 21.44710.80

Two-factor interactions COD  current COD  pH COD  time COD  oxidant Current  pH Current  time Current  oxidant pH  time pH  oxidant Time  oxidant

90.87734.91 112.75734.91 92.25734.91 53.00734.91 48.25734.91 142.00734.91 51.50734.91 52.12734.91 33.62734.91 24.12734.91

61.44710.80 11.69710.80 91.94710.80 15.81710.80 3.81710.80 27.06710.80 20.56710.80 3.81710.80 15.06710.80 3.19710.80

Three-factor interactions COD  current  pH COD  current  time COD  current  oxidant COD  pH  time COD  pH  oxidant COD  time  oxidant Current  pH  time Current  pH  oxidant Current  time  oxidant pH  time  oxidant

50.37734.91 68.13734.91 62.87734.91 14.50734.91 22.25734.91 31.50734.91 2.25734.91 38.75734.91 6.25734.91 8.37734.91

0.44710.80 36.44710.80 9.69710.80 3.44710.80 11.06710.80 0.56710.80 11.19710.80 1.94710.80 2.56710.80 8.31710.80

Four-factor interactions COD  current  pH  time COD  current  pH  oxidant COD  current  time  oxidant COD  pH  time  oxidant Current  pH  time  oxidant

10.38734.91 54.13734.91 17.62734.91 11.00734.91 24.25734.91

5.06710.80 0.94710.80 4.69710.80 5.19710.80 4.94710.80

31.63734.91

1.94710.80

Five-factor interactions COD  current  pH  time  oxidant

157.4175.40

ARTICLE IN PRESS

expected since the dark color of effluents like OMW and TOPW is associated, to a great degree, with the presence of polyphenolic compounds (Kotta et al., 2007). Estimation of the average effect, the main effects (i.e. the effect of each individual variable on the response) and the two and higher order interactions was made by means of the statistical package Minitab 14. The results are presented in Table 4. To assess the significance of the effects, an estimate of the standard error is required. An estimate of the standard error is usually made by performing repeat runs. Alternatively, three and higher order interactions can be used, since these interactions may be considered negligible and may measure differences arising from experimental error (Box et al., 1978). The variance of each effect would then be Variance_of_effects ¼

1233

42 (2008) 1229 – 1237

WAT E R R E S E A R C H

P ðthree_and_higher_order_effectÞ2 . Number_of_three_and_higher_order_effects

(1) The standard error is then the square root of the variance (half this amount for the average). If an effect is about or below the standard error, it may be considered insignificant (or in other terms, not different from zero). The contribution of a variable, however, whose effect appears different from zero, is not necessarily very large. One way to identify the most important effects is to construct the normal probability plot (Box et al., 1978; Daniel, 1976). All effects that are small can be explained as white noise, following a normal distribution with a mean of zero. In the normal probability plot, these effects will appear on a straight line. Any effects with a significant contribution will lie away from the normal probability line. The normal probability plot for the oxidation of COD appears in Fig. 1. There are basically only three effects which lie away from the straight line: in order of significance, the reaction time, the initial COD concentration and the

current. These effects are the most important factors affecting the oxidation of COD. The presence of oxidant and the initial solution pH, along with all interactions, are not significant and may be explained as random noise. All three significant effects are positive indicating that an increase in their level brings about an increase in the amount of COD oxidized. It was then decided to proceed with a new factorial design to further investigate the effect of these three variables. The three independent variables and their respective levels are presented in Table 5. Each one of the three variables received two values, higher than the original ones. The experimental design followed in this case was a full 23 experimental set, which required 8 experiments. The order each experiment was performed was selected randomly. This order of experimental runs, along with values of each independent variable for each run, is presented in Table 6. Table 6 also shows the obtained results in terms of mg/L of COD oxidized (depended variable or response Y1) and mg/L of TP oxidized (depended variable or response Y2). Estimation of the average effect, the main effects and the two and higher order interactions was made by means of

Table 5 – Independent variables of the 23 factorial experimental design Level of value

Initial COD (mg/L)

Current (A)

Reaction time (min)

5000 10,000 7500

15 30 22.5

60 240 150

 + 0

99 D

95

A B

90

Percent

80 70 60 50 40 30 Factor A B C D E

20 10 5

Name COD I(A) pH Time Oxidant

C

1 -200

-100

0

100

200

300

400

Effect Fig. 1 – Normal probability plot of the effects for the removal of COD.

500

ARTICLE IN PRESS 1234

WAT E R R E S E A R C H

42 (2008) 1229– 1237

Table 6 – Design matrix of the 23 experimental design and observed response (Y1: mass of COD removed per liter; Y2: mass of TP removed per liter) Order of running experiments

3 5 4 1 6 8 2 7

Level value of each variable in the experimental run

Y2 (mg/L) of TP removed

COD

Current

Time

 +  +  +  +

  + +   + +

    + + + +

341 639 575 1022 1236 1406 2237 2428

498 180 447 813 601 784 504 1096

0 0 0 0

0 0 0 0

0 0 0 0

1236 1022 1449 1108

655 843 652 696

Table 7 – Estimated effects of the 23 factorial design for the oxidation of COD and TP

sponse was constructed as follows: Y 1 ¼ 1235:5 þ

Effect

Y1 (mg/L) of COD removed

Value of effect Oxidation of COD

Oxidation of TP

Average effect

1235.5792.65

615745.05

Main effects COD Current Reaction time

276.97185.3 660.47185.3 1182.37185.3

205.9790.10 199.2790.10 261.7790.10

Two-factor interactions COD  current COD  time Current  time

42.67185.3 95.97185.3 351.57185.3

273.3790.10 181.8790.10 91.8790.10

Three-factor interactions COD  current  time

32.07185.3

69.0790.10

the statistical package Minitab 14. Estimates of the effects of the three variables and their interactions appear in Table 7. An estimate of the standard error was obtained by performing repeat runs at the center point of the factorial design, i.e., the variables receive mean values between their high and low levels. The standard error (standard deviation) is also shown in Table 7 (half this amount for the average effect). Once again, the reaction time is the most significant variable, since its effect has the highest value and it is twice as high as the second effect corresponding to the current. The effect of the initial COD concentration is much less significant as it is only slightly higher than the standard error. Among the interactions, only the (current)  (reaction time) interaction is significant. Based on the variables and interaction which are statistically significant, a model describing the experimental re-

276:9 660:4 1182:3 351:5 X1 þ X2 þ X3 þ X2 X3 , 2 2 2 2

(2)

where Y1 is the mass of COD oxidized (mg/L), Xi are the transformed forms of the independent variables according to Xi ¼

Zi  ðZhigh þ Zlow =2Þ ðZhigh  Zlow =2Þ

and Zi are the original (untransformed) values of the variables. The coefficients that appear in Eq. (2) are half the calculated effects, since a change of X ¼ 1 to 1 is a change of two units along the X-axis. The model predicts a linear dependency of the mass of COD oxidized on the operating variables. The values of residuals (i.e. observed minus predicted values of Y1) can then be plotted in a normal probability plot (Fig. 2). All the points from this residual plot lie close to the straight line confirming the conjecture that effects other than those considered in the model may be readily explained by random noise. Further treatment optimization in terms of the initial COD concentration and the current is not possible for the following reasons:

 The current used is close to the maximum permissible



current recommended by the electrode manufacturer (35 A). The high-level value of the initial COD concentration is in the same order of magnitude as the undiluted wastewater COD, while a further increase in concentration would result in unacceptably high temperatures in the reactor.

In order to further investigate the effect of reaction time on the amount of COD oxidized, a sample with an initial COD concentration of 9350 mg/L was treated at 30 A for an extended period of time (14 h). The results are presented in

ARTICLE IN PRESS WAT E R R E S E A R C H

1235

42 (2008) 1229 – 1237

99

95 90

Percent

80 70 60 50 40 30 20

Mean 0 StDev 255,7 N 8 AD 0,518 P-Value 0,128

10 5

1 -1000

-500

0

500

1000

Residuals, COD model Fig. 2 – Normal probability plot of the residuals; COD removal model at optimal treatment conditions. 10000

COD, mg/L

8000 6000 4000 2000 0 0

2

4

6

8

10

12

14

Time, hours

Fig. 3 – COD concentration as a function of time under optimal treatment conditions (current: 30 A; initial COD concentration: 9350 mg/L).

Fig. 3. A linear reduction in the COD was observed throughout the treatment period in accordance to the trend predicted by the linear model, despite the fact that the linear model considered an upper level of treatment time of 4 h. After 14 h of treatment the COD reduction reached 73%. In addition, the linear reduction in COD indicates zero-order kinetics. Based on the experimental results of Fig. 3, the estimated zero-order kinetic constant was 8.5 mg/(L min) with the correlation coefficient (r2) of the linear fitting being 99%. In terms of energy consumption, the removal efficiency of the process was 16.3 g COD/(m3 A h) under optimal conditions. The specific energy consumption, defined as the amount of energy consumed per unit mass of COD removed, was 174 kWh/kg COD; therefore, cost of electricity was about 17.4 h/kg COD (based on 0.1 h/kWh). Rivas et al. (2001) who studied the wet air oxidation of TOPW at 180 1C, 5 MPa total pressure and about 12,500 mg/L

initial COD reported a 30% COD reduction after 6 h of treatment; this increased to 60% when copper at a concentration of 420 mg/L was used as a homogeneous catalyst. In other studies, Beltran-Heredia et al. (2000a, b) reported that ozonation of TOPW at about 6000 mg/L starting COD for 3 h resulted in 24–55% COD reduction; performance was mainly dependent of solution pH and increased levels of decontamination could be achieved at alkaline rather than neutral or acidic conditions. Eight hours of combined UV irradiation and ozonation (Benitez et al., 2002) at alkaline conditions led to 88% COD removal for a TOPW of 4400 mg/L initial COD, while nearly complete mineralization was achieved when the initial concentration decreased to 3050 mg/L. Nonetheless, treatment of diluted TOPW (2300 mg/L COD) for 6 h by combined UV irradiation and the Fenton process led to 28–67% COD removal depending on the concentration of hydrogen peroxide and iron ions employed (Benitez et al., 2001b). Regarding the removal of TP, the normal probability plot of the effects (Fig. 4) showed that the most important factors affecting their removal by electrolytic oxidation were the same as in the COD. In addition, some of two-factor interactions, such as the (COD)  (reaction time) and the (COD)  (current), were important, indicating that quadratic terms of the variables may be needed in the relevant model, in order to adequately represent the removal of TP at interpolated levels of the variables. Under optimal (for COD removal) conditions, as depicted in Table 7, all three variables and the two-factor interactions (COD)  (reaction time) and the (COD)  (current) have values higher than the estimated standard error. The levels of the variables which yielded the maximum TP oxidation were the same as in the removal of COD, i.e., all variables were at their highest level (initial COD ¼ 10,000 mg/L; current ¼ 30 A; reaction time ¼ 4 h). The mathematical model that describes the oxidation of TP as a

ARTICLE IN PRESS 1236

WAT E R R E S E A R C H

42 (2008) 1229– 1237

99 A

95

D AD B

90

Percent

80 E

70 60 50 40 30

BD C

ABD

AB

Factor Name A COD B I(A) C pH D Time E Oxidant

20 10 5 BE

1 0

40

80

120

160

Effect Fig. 4 – Normal probability plot of the effects for the removal of total phenols (TP).

99

95 90 80

Percent

70 60 50 40 30 20 Mean 0,5 StDev 61,47 N 8 AD 0,387 P-Value 0,296

10 5

1 -200

-100

0 Residuals, TP model

100

200

Fig. 5 – Normal probability plot of the residuals; TP removal model at optimal treatment conditions.

function of the reduced variables is Y 2 ¼ 615 þ

considered in the model may be readily explained by random noise.

205:9 199:2 261:7 273:3 181:8 X1 þ X2 þ X3 þ X1 X2 þ X1 X3 , 2 2 2 2 2

(3) where Y2 is the mass of TP oxidized (mg/L). Adequacy of the model was also checked by means of constructing the normal plot of the residuals (Fig. 5). Once again, all points from this residual plot lie close to the straight line confirming the conjecture that affects other than those

4.

Conclusions

The conclusions drawn from this study can be summarized as follows: (1) Electrochemical oxidation over BDD electrodes is a relative innovative process in industrial wastewater treatment. In

ARTICLE IN PRESS WAT E R R E S E A R C H

42 (2008) 1229 – 1237

this view, a strong effluent from edible olive processing was chosen to be treated electrochemically with emphasis given on the effect of various operating conditions, such as initial COD loading, contact time, starting effluent pH, applied current and the use of H2O2 as an extra oxidant on treatment efficiency. In general, the process was capable of achieving satisfactory levels of phenols degradation and decolorization followed by relatively low mineralization at short treatment times. (2) To evaluate the importance of the various parameters on treatment efficiency, a factorial design approach was followed. Of the five parameters tested, contact time, applied current and the initial effluent COD affected COD and phenols removal at a statistically significant level. At high initial COD loadings (i.e. over 5000 mg/L), its effect on performance became much less important implying a zero-order reaction rate with respect to COD concentration. (3) To assess the potential of this method to treat heavily polluted wastes, the effluent at an initial concentration of about 10,000 mg/L COD was subject to electrochemical treatment at 30 A for 14 h; the process resulted in 73% COD removal at 16.3 g COD/(m3 A h) energy efficiency.

R E F E R E N C E S

Aggelis, G.G., Gavala, H.N., Lyberatos, G., 2001. Combined and separate aerobic and anaerobic biotreatment of green olive debittering wastewater. J. Agric. Eng. Res. 80 (3), 283–292. Atanassova, D., Kefalas, P., Psillakis, E., 2005. Measuring the antioxidant activity of olive oil mill wastewater using chemiluminescence. Environ. Int. 31 (2), 275–280. Beltran, F.J., Garcia-Araya, J.F., Frades, J., Alvarez, P., Gimeno, O., 1999. Effects of single and combined ozonation with hydrogen peroxide or UV radiation on the chemical degradation and biodegradability of debittering table olive industrial wastewaters. Water Res. 33 (3), 723–732. Beltran-Heredia, J., Torregrosa, J., Dominguez, J.R., Garcia, J., 2000a. Treatment of black olive wastewaters by ozonation and aerobic biological degradation. Water Res. 34 (14), 3515–3522. Beltran-Heredia, J., Torregrosa, J., Dominguez, J.R., Garcia, J., 2000b. Ozonation of black-table-olive industrial wastewaters: effect of an aerobic biological pretreatment. J. Chem. Technol. Biotechnol. 75 (7), 561–568. Beltran-Heredia, J., Torregrosa, J., Dominguez, J.R., Garcia, J., 2000c. Aerobic biological treatment of black table olive washing wastewaters: effect of an ozonation stage. Process Biochem. 35 (10), 1183–1190. Benitez, F.J., Acero, J.L., Gonzalez, T., Garcia, J., 2001a. Ozonation and biodegradation processes in batch reactors treating black

1237

table olives washing wastewaters. Ind. Eng. Chem. Res. 40 (14), 3144–3151. Benitez, F.J., Acero, J.L., Gonzalez, T., Garcia, J., 2001b. Organic matter removal from wastewaters of the black olive industry by chemical and biological procedures. Process Biochem. 37 (3), 257–265. Benitez, F.J., Acero, J.L., Gonzalez, T., Garcia, J., 2002. The use of ozone, ozone plus UV radiation, and aerobic microorganisms in the purification of some agro-industrial wastewaters. J. Environ. Eng. Sci. Health—Part A: Toxic/Hazardous Substances & Environ. Eng. 37 (7), 1307–1325. Box, G.E.P., Hunter, W.G., Hunter, J.S., 1978. Statistics for Experimenters. Wiley, New York. Brenes, M., Garcia, P., Romero, C., Garrido, A., 2000. Treatment of green table olive waste waters by an activated-sludge process. J. Chem. Technol. Biotechnol. 75 (6), 459–463. Canizares, P., Martinez, L., Paz, R., Saez, C., Lobato, J., Rodrigo, M.A., 2006. Treatment of Fenton-refractory olive oil mill wastes by electrochemical oxidation with boron-doped diamond anodes. J. Chem. Technol. Biotechnol. 81 (8), 1331–1337. Canizares, P., Lobato, J., Paz, R., Rodrigo, M.A., Saez, C., 2007. Advanced oxidation processes for the treatment of olive-oil mills wastewater. Chemosphere 67 (4), 832–838. Chen, G., 2004. Electrochemical technologies in wastewater treatment. Separ. Purif. Technol. 38 (1), 11–41. Chen, X., Gao, F., Chen, G., 2005. Comparison of Ti/BDD and Ti/SnO2–Sb2O5 electrodes for pollutant oxidation. J. Appl. Electrochem. 35 (2), 185–191. Daniel, C., 1976. Applications of Statistics to Industrial Experiments. Wiley, New York. Kopsidas, G.C., 1992. Wastewater from the preparation of table olives. Water Res. 26 (5), 629–631. Kotta, E., Kalogerakis, N., Mantzavinos, D., 2007. The effect of solids on the electrochemical treatment of olive mill effluents. J. Chem. Technol. Biotechnol. 82 (5), 504–511. Kyriacou, A., Lasaridi, K.E., Kotsou, M., Balis, C., Pilidis, G., 2005. Combined bioremediation and advanced oxidation of green table olive processing wastewater. Process Biochem. 40 (3–4), 1401–1408. Mantzavinos, D., Kalogerakis, N., 2005. Treatment of olive mill effluents: Part I. Organic matter degradation by chemical and biological processes—an overview. Environ. Int. 31 (2), 289–295. Paraskeva, P., Diamadopoulos, E., 2006. Technologies for olive mill wastewater (OMW) treatment: a review. J. Chem. Technol. Biotechnol. 81 (9), 1475–1485. Rivas, F.J., Beltran, F.J., Gimeno, O., 2000. Joint treatment of wastewater from table olive processing and urban wastewater. Integrated ozonation–aerobic oxidation. Chem. Eng. Technol. 23 (2), 177–181. Rivas, F.J., Beltran, F.J., Gimeno, O., Alvarez, P., 2001. Chemicalbiological treatment of table olive manufacturing wastewater. J. Environ. Eng.—ASCE 127 (7), 611–619. Rivas, F.J., Beltran, F.J., Gimeno, O., Alvarez, P., 2003. Optimisation of Fenton’s reagent usage as a pre-treatment for fermentation brines. J. Hazard. Mater. 96 (2–3), 277–290.