Chemical Engineering Journal 167 (2011) 7783

Contents lists available at ScienceDirect

Chemical Engineering Journal

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

Modelling and optimization of coagulation of highly concentrated industrial grade leather dye by response surface methodology
M. Khayet a , A.Y. Zahrim b , N. Hilal b,
a b

Department of Applied Physics I, Faculty of Physics, University Complutense of Madrid, Spain The Centre for Water Advanced Technologies and Environmental Research (CWATER), College of Engineering, Swansea University, Swansea SA2 8PP, UK

a r t i c l e

i n f o

a b s t r a c t
High consumption of process water and water scarcity has motivated industry to reuse their wastewater. Membrane processes are vital to produce water for reuse from dyeing baths in the tanning industry. In this regard, synthetic dye was recognised as the major foulant. To minimise the membrane fouling, coagulation/occulation process is an important pre-treatment. Due to the complex nature of the process involving dyes-coagulant, the modelling is challenging. In this study, statistical experimental design and response surface methodology, RSM, have been applied to optimize removal of C.I. Acid Black 210 dye from highly concentrated solutions by means of a coagulation/oculation process. Aluminium sulphate was used as the coagulant. Central composite design (CCD) using as input variables the experimental temperature, the concentration of aluminium sulphate and the initial pH of the solution have been considered. Based on the design of experiment the quadratic response surface models have been developed to link the output response, which is the dye removal factor, with the input variables via mathematical relationships. The constructed response model has been tested using the analysis of variance (ANOVA). A Monte Carlo simulation method has been conducted to determine the optimum operating conditions. The obtained optimal point corresponds to a temperature of 40 C, a concentration of aluminium sulphate of 0.82 g/L and an initial pH value of 5.61. The maximal value of the dye removal obtained under optimal process conditions has been conrmed experimentally. 2010 Elsevier B.V. All rights reserved.

Article history: Received 15 October 2010 Received in revised form 30 November 2010 Accepted 30 November 2010 Keywords: Dye removal Coagulation/occulation Tanning industry Design of experiment Response surface methodology

1. Introduction The tanning industry employs huge amounts of water and this leads to the generation of enormous amounts of wastewater [1]. Due to human activities, water crisis has become a problem in many countries. This phenomenon has caused water reuse, especially for the water consuming industry such as tanning industry, to gain in importance. To cater for the needs of the tanning industry, water reuse via membrane technology has been proposed [2]. However, a major problem for membrane technology is fouling. High concentrations of dye are reported to cause severe fouling on the membrane [3] and caused failure at the end-pipe treatment of tanning wastewater [4]. To minimise the fouling, coagulation has been reported as an efcient and cost-effective pre-treatment [3].

Abbreviations: ANOVA, analysis of variances; CCD, central composite design; DF, degree of freedom; MLR, multi-linear regression method; MS, mean square; SS, sum of squares. Corresponding author. Tel.: +44 01792 606644; fax: +44 01792 295676. E-mail addresses: abuzahrim@yahoo.com (A.Y. Zahrim), n.hilal@swansea.ac.uk (N. Hilal). 1385-8947/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cej.2010.11.108

Leather making is comprised of several processes and the dyeing process is one of the important steps. At the present time, most tanneries use synthetic dyestuffs for their leather [2,5]. The history of the modern synthetic dyestuffs industry was begun with the development of mauveine by Perkin in 1856. During that time, the primary market was textiles, but the leather industry eventually took advantage half a century later, especially when it was realised that bright deep shades could be achieved with the new tannage with chromium (III) [5]. Nevertheless, the presence of residual dyes in surface water is aesthetically undesirable and causes problems for the aquatic biosphere due to the reduction of sunlight penetration and depletion of the dissolved oxygen. Some dyes are toxic and mutagenic and have the potential to release carcinogenic amines [1]. Furthermore, dyeing wastewater from the leather industry is high in organics and has high temperature. Moreover, dyeing wastewater contribute about 38% of the total volume of discharge wastewater [2]. Typical characteristics of dyeing wastewater from the leather industry are shown in Table 1 [2]. Moreover, industrial dyestuffs are typically not pure compounds and the purity differs from one batch to the other. Generally, dyes are marketed with diluent (a solid or liquid to dilute another) such as sodium chloride, sodium carbonate, dextrin, sulphite cellulose, naphthalene sulfonate, sodium sulphite, etc. [5].

78

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783

Nomenclature b0 , bi , bii , bij regression coefcients b (u 1) vector of regression coefcients b0 intercept regression coefcient C aluminium sulphate concentration (g/L) C0 initial dye concentration (g/L) nal dye concentration (g/L) Cf C* optimal value of aluminium sulphate concentration (g/L) F ratio of variances N the number of experimental runs n number of factors (independent variables) pH* optimal pH value R2 coefcient of multiple determination Radj 2 adjusted statistic coefcient SSRegression sum of squares of the regression model SSResidual sum of squares of the residual T temperature ( C) T* optimal temperature value ( C) u the number of signicant regression coefcients in the response surface model X (N u) matrix of the independent variables levels x1 , x2 , x3 , xI the coded levels of the factors (independent or control variables) Y (N 1) vector of the experimental response Y response (experimental value) Greek letters statistical error in response surface model

Coagulation of dye-containing wastewater has been used for many years either as a main or pretreatment due to its low capital cost [1]. Coagulation shows higher efciency for azo dye [3] and this is an advantageous since about 70% of dyes currently in use are of azo type [1]. Nevertheless, Beltran-Heredia et al. [6] stated that the mathematical modelling of coagulation/occulation of dyes is very difcult due to: (1) The complex nature of the phenomenon, which implies various physicochemical interaction between coagulant-dye molecules. (2) The intrinsic composition of the dye molecules and the coagulant is not completely known. Coagulation in acidic conditions has been reported to have several advantages. Klimiuk et al. [7] reported that the ocs have a better structure and are more stable. Moreover, at an alkali initial solution, the dosage of alum to achieve the highest colour removal is relatively higher than at an acidic solution. Consequently, the
Table 1 Typical characteristics of dyeing wastewater from leather industry. Parameters pH Temperature, C Sedimentable materials, mg/l Total suspended solids, mg/l BOD5 , mg/l COD, mg/l Chromium (III), mg/l Chlorides, mg/l Oil and fats, mg/l Chlorinated solvents, mg/l Surfactants, mg/l Min 4 20 100 10,000 6000 15,000 0 5000 20,000 0 500 Max 10 60 500 20,000 15,000 75,000 3000 10,000 50,000 250 2000

production of sludge will also increase and the higher hydroxide content and higher TOC content of the sludges generally contribute to the poorer dewatering characteristics [8]. Dosage of metal coagulant, pH and temperature is an important factor for coagulation. Coagulant reactions and metal coagulant chemistry are strongly affected by temperature. For example, with decreasing water temperature, the minimum solubility of aluminium hydroxide species shift to higher pH, and the optimum operating pH value also shifts to higher pH [8]. However, there is lack of studies on the effect of high temperature (greater than 30 C) during coagulation of dye wastewater [3]. In this regard, several studies have been carried out with reactive dyes. Koprivanac et al. [9] investigated the removal of Reactive Blue 204 in the range between 20 and 80 C using ferric chloride. In this study, the authors concluded that the effective temperature for decolourization should be less than 30 C. However, Joo et al. [10] found that at 40 C alum/ferric salt plus an organic polymer can achieve 100% decolourization of several reactive dyes (Black 5, Blue 2, Red 2 and Yellow 2). Acid dyes are the most commonly used in the leather industry, particularly for chromed tanned leather. The properties can be summarized as follows [5]: (i) relatively small, typically hydrophilic molecules, (ii) used for penetrating dyeing, producing level shades, (iii) anionically charged, therefore high afnity for cationic leather, (iv) xed with acidication, due to the presence of sulfonate group, (v) react predominatly through electrostatic reaction between their sulfonate groups and the protonated amino group of lysine, (vi) secondary reaction is via hydrogen bonding through auxochrome groups, (vii) some dyes may react with the bound chrome; using it as a mordant, (viii) good fastness properties: less complex molecules offer fewer opportunities for structural changes by free radical mechanisms, and (ix) Wide range of colours, offering bright deep shades. It is to be noted that the major part of the reported studies on dye removal deal with the conventional method of experimentation, which involves changing one of the independent parameters, maintaining the others xed. This classical or conventional method of experimentation requires many experimental runs, which are time-consuming, ignores interaction effects between the operating parameters and leads to a low efciency in optimization. These limitations of the classical method can be avoided by applying the response surface methodology (RSM) that involves statistical design of experiments (DoE) in which all factors are varied together over a set of experimental runs. In fact, RSM is a collection of mathematical and statistical techniques useful for developing, improving and optimizing processes, and can be used to evaluate the relative signicance of several affecting factors even in the presence of complex interactions [11]. The statistical method of experimental design offers several advantages over the frequently used conventional experimental method being rapid and reliable, helps understanding the interaction effects between factors and reduces the total number of experiments tremendously resulting in saving time and costs of experimentation. RSM has been applied successfully in various scientic and technical elds such as applied chemistry and physics, biochemistry and biology, chemical engineering, environmental protection, membrane science and technology [1121]. Some recent publications have shown the effectiveness of RSM modelling for dye removal [2226]. The performance of a low cost coagulant, i.e. waterworks sludge (FCS: ferric chloride sludge) for the removal of acid red 119 (AR119) dye from aqueous solutions was carried out [22]. RSM was applied in this case to optimize three operating variables of coagulation/occulation process including initial pH, coagulant dosage and initial dye concentration. The optimum initial pH, ferric chloride sludge dosage and initial dye concentration were found to be 3.5, 236.68 mg dried FCS/L and 65.91 mg/L, respectively. Dye removal of 96.53% was obtained and found to be close to the predicted RSM results [22].

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783

79

Other low cost coagulant for treatment of dye wastewater was also studied in [23]. The investigation was focussed on the steel industrial wastewater (SIWW) FeCl3 rich as an original coagulant to remove the synthetic textile wastewater. RSM was used to study the effects of the coagulant dosage, initial pH of dye solution and dye concentration, and to optimize the process conditions for the decolourization and Chemical Oxygen Demand (COD) reduction of the disperse blue 79 solution. For obtaining the mutual interaction between the variables and optimizing these variables, a 23 full factorial central composite rotatable design of experiments was employed. The efciencies of decolourization and COD reduction for disperse blue 79 solution were accomplished at the optimum conditions 99% and 94%, respectively [23]. The decolourization of C.I. Acid Red 14 (AR14) azo dye by electrocoagulation (EC) process was studied in a batch reactor [24]. RSM was applied to evaluate the simple and combined effects of the three main independent parameters, current density, time of electrolysis and initial pH of the dye solution on the colour removal efciency and optimizing the operating conditions of the treatment process. A 23 full factorial central composite face centered (CCF) experimental design was employed. Analysis of variance (ANOVA) showed a high coefcient of determination value (R2 = 0.928) and a satisfactory prediction second-order regression model was derived. Maximum colour removal efciency was predicted and experimentally validated. Under optimal value of process parameters, high removal (> 91%) was obtained for Acid Red 14. The study clearly showed that RSM was a suitable methods to optimize the operating conditions and maximize the dye removal. Graphical response surface and contour plots were used to determine the optimum point [24]. The effects of pH and initial dye concentration on dye removal by coagulation/occulation process with natural coagulant, i.e. Moringa oleifera seed extract have been investigated [25]. The study was carried out by using RSM in an orthogonal and rotatable design of experiments. Three types of dyes were considered: anthraquinonic (Alizarin Violet 3R); indigoid (indigo Carmine) and azoic (Palatine Fast Black WAN). It was observed that the interaction of the two variables studied is higher in the case of azo dye, while it is negligible in the case of anthraquinonic dye. Indigoid dye presents an intermediate situation. The mineralization of C.I. Acid Orange 7 (AO7) azo dye by UV/H2 O2 advanced oxidation was studied in [26]. An experimental design based on RSM was applied to evaluate the simple and combined effects of parameters on mineralization efciency and to optimize the operating conditions of the treatment process. A 23 full factorial central composite face centered (CCF) experimental design was employed. ANOVA analysis showed a high coefcient of determination value and a satisfactory second-order regression model was derived. Graphical response surface and contour plots were used to localize the region of the optimum point. Maximal dye mineralization performance was predicted and experimentally validated. Under optimal value of process parameters, high mineralization of AO7 dye (87.07%) was achieved [26]. In the present study, aluminium sulphate has been used for coagulation/occulation to remove C.I. Acid Black 210 dye from highly concentrated solutions at different temperatures and initial pH values. The central composite experimental design (CCD) and RSM has been applied. The main objective is to maximize the dye removal applying the adequate optimum operating conditions. 2. Materials and methods 2.1. Materials Durapel Black NT which consists of C.I. Acid Black 210 dye (purity > 30%) was purchased from Town End (Leeds) plc, (United

Table 2 Actual and coded values of independent variables used for experimental design. Variable Symbol Real values of coded levels a T ( C) pH C (g/L) x1 x2 x3 20 5 0.8 1 21.8 5.09 0.86 0 30 5.5 1.15 +1 38.2 5.91 1.44 +a 40 6 1.50

a = 1.215 (star or axial point for orthogonal CCD in the case of 3 independent variables).

Kingdom) and used without further purication. A dye mass of 20 g in powder form was dissolved in Milli-Q Plus, 18.2 M cm (Millipore) water to make 5 l solution at a concentration of 4 g/L. The initial absorbance at 460 nm for this solution is about 0.7 (100 dilution). The laboratory reagent aluminium sulphate hexadecahydrate (Al2 (SO4 )3 16H2 O) (molecular mass of 630.39 g/mol, purity >96%) was purchased from Fisher Scientic UK Ltd. (United Kingdom). Aluminium sulphate solutions were prepared fresh everyday by dissolving appropriate amounts of powder aluminium sulphate in Milli-Q Plus (Millipore) water. 2.2. Jar tests An appropriate volume of dye solution was transferred into the round jar and pH was adjusted accordingly. The pH adjustment was made under vigorous stirring with a magnetic stir bar using solutions of 2 M HCl and 2 M NaOH. The initial temperature was increased by heating the solution using magnetic hotplate (Fisher Scientic UK Ltd., United Kingdom). Subsequently, aluminium sulphate was added to the dye solution in the jar, making the total volume of 500 ml. In every experiment, aluminium sulphate solution was less than 5% v/v in aqueous solution. This was done to prevent unnecessary dilution effects. A standard jar-test apparatus (Bibby-Stuart Flocculator SW6) equipped with stainless steel paddles and stirrer was used for the coagulation/occulation tests. The aqueous solution was then rapidly mixed at a paddle speed of 250 rpm for 3 min followed by slow mixing for 20 min at 30 rpm. After allowing settling to occur (120 min), about 25 ml of the liquid was withdrawn using a pipette from a height of about 3 cm below the liquid surface in each jar. This height comprises the rst 40% of the total height. Experiments were all triplicate to test their reproducibility. 2.3. Experimental design The statistical design of experiments (DoE) is a structured and systematized method of experimentation in which all factors are varied simultaneously over a set of experimental runs in order to determine the relationship between factors affecting the output response of the process. As stated earlier, the statistical experimental design employed has been carried out considering three factors (controllable variables), namely, the solution temperature (T), the concentration of the used aluminium sulphate (C) and the initial pH value (pH). Table 2 shows the controllable variables (factors) and their levels in coded and actual values. The output response is the dye removal (%) determined by means of Eq. (1). In order to describe the factors effects on the responses an orthogonal central composite design (CCD) with star points was employed with 3 factors and 5 levels. The central composite experimental design (CCD) consists of 16 experiments with 8 orthogonal design points (factorial points), 6 star points to form a central composite design with = 1.215 and 2 center points for replication. The experimental design matrix is summarized in Table 3.

80

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783 Table 4 Analysis of variance (ANOVA) for the developed RSM model. Source Model Residual Total
a b c

Table 3 CCD experimental design (DoE) for dye removal by coagulation/occulation. Run number (N) and typea 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
a b

T ( C) x1 b

pH x2 b +1 +1 1 1 +1 +1 1 1 0 0 + 0 0 0 0

C (g/L) x3 b +1 +1 +1 +1 1 1 1 1 0 0 0 0 + 0 0

Dye removal (%)

DFa 9 6 15

SSb 2868.817 174.101 3042.918

MSc 318.757 20.0168

F-value 10.985

F-tab 3.8

R2

Radj 2

0.943 0.857

O1 O2 O3 O4 O5 O6 O7 O8 S1 S2 S3 S4 S5 S6 C1 C2

+1 1 +1 1 +1 1 +1 1 + 0 0 0 0 0 0

71.96 88.06 77.71 97.23 96.32 94.85 95.59 92.83 90.38 93.03 59.05 63.72 81.11 87.49 61.24 60.36

DF degree of freedom. SS sum of squares. MS mean square.

order to minimise the sum of squares of the residuals. The least square estimations of the regression coefcients were calculated by the following matrix equation [11,12,1921,2739].
T ) = (X X b 1 T Y X

(3)

O = orthogonal design points, C = center points, S = star or axial points. 1 = low value, 0 = center value, +1 = high value, +/ = star point value.

2.4. Residual concentration of dye analysis Residual concentration of dye (without ltering or centrifuging) was analysed with a UV/vis-spectrophotometer, (UVmini-1240, Shimadzu) by measuring the absorbance at the max (460 nm) and nal pH of the solution. The absorbance was measured using water Milli-Q as background and the concentration of dye was computed from calibration curves preliminary determined at different pH values. If the reading of absorbance was greater than 1.0, then necessary dilution was made (normally 100 times of dilution). After every experiment, precision cell (10.00 mm, quartz SUPRASIL (Hellma GmbH & Co, Germany)) was cleaned by soaking it with methanol HPLC Grade (Fischer Scientic UK Ltd, United Kingdom) overnight. Glassware was cleaned by rinsing with 0.1 M NaOH. The values of the initial and nal concentrations of the dye measured as outlined above were used to calculate the removal percentage of the dye (i.e. dye removal factor) using Eq. (1). dye removal (%) = C0 Cf 100 C0 (1)

is a (u 1) vector of regression coefcients, X is a (N u) where b is a (N 1) vector of matrix of the independent variables levels, Y the experimental runs according to the arrangements points into the experimental design, N is the number of experimental runs and u is the number of regression coefcients in the response model. In our case (n = 3 input variables) and the empirical model in terms of coded factors can be written as: = b0 + b1 x1 + b2 x2 + b3 x3 + b11 x2 + b22 x2 + b33 x2 + b12 x1 x2 Y 1 2 3 +b13 x1 x3 + b23 x2 x3 (4)

where x1 , x2 and x3 are the coded values of the independent variables. The obtained regression equation in terms of the coded variables is: = 65.799 3.159x1 4.781x3 + 16.548x2 4.009x2 + 11.528x2 Y 1 2 3 4.982x1 x3 (5)

where C0 is the initial dye concentration and Cf is the nal dye concentration. Measurement of solution pH was done using a Jenaway 3540 pH meter. Baseline atness and wavelength accuracy for the UVspectrophotometer and pH calibration were carried out daily. The obtained dye removal factors (output responses) for the experimental design matrix are presented in Table 3. 3. Results and discussion Based on the CCD experimental design results shown in Table 3, the RSM has been applied to develop the polynomial regression equations and nd out the relation between the output response, dye removal, and the input factors. The response surface models of second-order were developed as follows [11,14,1821,2739]:
n n n 2 bii xi + i =1 i<j

It must be mentioned that is the star point in the experimental design that gives the limits of the valid region, [xj ; = {xj | xj + }; j = 1, 2, 3], of experimentation. The regression coefcients were tested for signicance using Students t-test [13,16]. Therefore, in Eq. (4) only the signicant coefcients were maintained. The regression coefcient b2 and the interaction terms b12 and b23 are found to be negligible. The nal empirical model obtained in terms of actual parameters is determined and written in general form as follows: = 296.008 12.642T + 262.338pH 269.136C + 0.244T 2 Y 23.849pH2 + 137.075C 2 2.087TC 20 C T 40 C; (6)

= b0 + Y
i=1

bi xi +

bij xi xj +

(2)

is the predicted response, xi refers to the coded variables, where Y b0 , bi , bii , bij are the regression coefcients and is the statistical error. The regression coefcients of the response model were computed by means of Multiple Linear Regression (MLR) method in

where: 5 pH 6 and 0.8 g/L C 1.5 g/L. The response surface models were validated statistically for adequacy by analysis of variance (ANOVA). The results are presented in Table 4. The mathematical relationships used for computation of statistical estimators (i.e., F-value, R2 , Radj 2 ) can be found elsewhere, for example in Refs [13,16]. The statistical signicance of the second-order regression model was determined by F-value, which is a measurement of variance of data about the mean, based on the ratio of mean square of group variance due to error. If the model gives a good prediction of the experimental data then the calculated F-value should be greater than the tabulated F-value, 3.8 [11,12]. In this case, according to analysis of variance the calculated Fvalue is found to be greater than the tabulated one (F-tab = 3.8). This means that the obtained response model is validated from a statistical standpoint and is a good predictor of the experimental data. Moreover, the R2 -value is 0.943, which is desirable. This implies that more than 94.3% of the data deviation can be explained by the developed empirical models. Furthermore, the predicted R2 values

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783

81

100 90 80 70 Dye Removal (%) 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 N 9 10 11 12 13 14 15 Experiment RS-Model

Fig. 1. Comparison of experimental and predicted dye removal values by response surface model (RS-Model).

are in agreement with the adjusted statistics Radj 2 . This means that signicant terms have been included in the empirical model. The response values determined by means of the empirical models were compared to the experimental data designed in Table 3. The results are shown in Fig. 1. As can be seen the response model shows good ts to the experimental data. Therefore, the model can be considered adequate for the predictions and optimization. The graphical representations of the response surfaces were plotted based on this model, i.e. Eq. (6). Some relevant response surface plots and the corresponding contour plots are reported in Figs. 24. The plots shown in Figs. 2 and 4 indicate the inuence of the initial pH on the dye removal. Similar gures can be plotted for other aluminium sulphate concentrations and temperatures. The results given by the response model indicate that the increase of the initial pH leads to an increase of the dye removal up to a maximum. For higher pH values the trends are declined suggesting the existence of an optimal dye removal factor. It is to be noted that the effect of the temperature upon the output response is the greatest one. From Figs. 2 and 3, it can be observed gradual reductions of the dye removal factor to minimum values with the increase of the feed temperature from 20 C to about 30 C and then an increase with further temperature enhancements. Similarly, from Figs. 3 and 4, minimum dye removal factors are obtained for different pH values and temperatures and higher dye removal factors are observed at low aluminium sulphate concentration (i.e. 0.8 g/L). Zahrim et al. [3] have previously shown that aluminium based coagulant is superior to metal based coagulant during coagulation of anionic dyes (e.g. acid and reactive dyes) in acidic condition. Previous work on coagulation of several reactive dyes have shown that at the same dosage, aluminium based coagulant works better at acidic pH while ferric, magnesium and calcium based coagulant works better at basic solution [40]. Akbari et al. [41] stated that acid dyes might produce wastewater with acidic condition, therefore it is expected that the acid dye efuent, in the paper, is in acidic condition. Thus we can conclude that aluminium based coagulant is the best coagulant for leather dyeing bath. In a later study, Gaydardzhiev et al. [42] found that aluminium sulphate is better than ferric chloride or aluminium chloride for the removal of C.I. Acid Blue 113 dye due to the corresponding less consumed dosage. In the present study, high dosage of aluminium sulphate is expected due to very high concentration of initial dye, i.e. 4 g/l. It has also been shown in previous work [43] that greater amount of aluminium sulphate

Fig. 2. Response surface plot (a) and contour plot (b) of predicted dye removal factor as function of the operating temperature and initial pH at C = 1.15 g/L.

Fig. 3. Response surface plot (a) and contour plot (b) of predicted dye removal factor as function of the operating temperature and concentration of aluminium sulphate at a pH value of 5.5.

82

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783 Table 5 Optimal point in terms of the actual operating variables and the output response, dye removal (%). T* ( C) pH* C* (g/L) Dye removal (%) Predicted 40
*

Experimental 97.78 0.63

5.61

0.802

115.71

Optimal predicted value by Monte Carlo simulation.

to obtain the optimum point and not the optimum region of experimentation that can only guarantee a dye removal factor of 100%. The conrmation run was carried out in order to check experimentally the optimal point. Table 5 shows the obtained experimental dye removal factor. The experimental value of the output response determined considering the obtained optimal conditions represent the best (maximal) values throughout all the conducted experimental tests inside the region of experimentation (Table 3). In other words, the obtained combination corresponding to the optimum point is the best one compared to the data presented in Table 3. 4. Conclusions Coagulation/occulation using aluminium sulphate is a suitable process for the removal of highly concentrated leather dye especially when the operating parameters are optimized. In fact, failure in optimizing the process parameters may lead to inefciencies of the coagulation process. The design of experiment (DoE) and response surface methodology (RSM) proved to be an effective method for optimization of coagulation/occulation process with a view of maximization of the C.I. Acid Black 210 dye removal from highly concentrated solutions. The developed response model has been tested using the analysis of variance (ANOVA). The temperature exhibits the strongest effects and high interaction effects exist between the temperature and the concentration of aluminium sulphate. On the contrary, the interaction effects between the pH and the temperature and between the pH and the aluminium sulphate are negligible. The optimal operational conditions are as follows: a feed temperature of 40 C, an initial pH value of 5.61 and a concentration of aluminium sulphate of 0.82 g/L. By applying these parameter values, maximal dye removal has been predicted and conrmed experimentally. References
[1] Y. Anjaneyulu, N.S. Chary, D.S.S. Raj, Decolourization of industrial efuents available methods and emerging technologies a review, Rev. Environ. Sci. Biotechnol. 4 (4) (2005) 245273. [2] A. Cassano, R. Molinari, M. Romano, E. Drioli, Treatment of aqueous efuents of the leather industry by membrane processes a review, J. Membr. Sci. 181 (1) (2001) 111126. [3] A.Y. Zahrim, C. Tizaoui, N. Hilal, Coagulation with polymers for nanoltration pre-treatment of highly concentrated dyes: a review, Desalination 266 (13) (2010) 116. [4] A.R. Rakmi, M.N. NorZaini, Y. Abu-Zahrim, H.A. Zalina, Chemibiological treatment of difcult wastewaters: case study on tanning wastewater, in: Z. Ujang, M. Henze (Eds.), Environmental Technology: Advancement in Water and wastewater Applications in the Tropics, IWA Publishing, London, 2004, pp. 315322. [5] A.D. Covington, Tanning Chemistry: The Science of Leather, Royal Society of Chemistry, Cambridge, 2009. [6] J. Beltran-Heredia, J. Sanchez-Martin, A. Delgado-Regalado, Removal of carmine indigo dye with Moringa oleifera seed extract, Ind. Eng. Chem. Res. 48 (14) (2009) 65126520. [7] E. Klimiuk, A. Filipkowska, A. Korzeniowska, Effects of pH and coagulant dosage on effectiveness of coagulation of reactive dyes from model wastewater by polyaluminium chloride (PAC), Polish J. Environ. Studies 8 (2) (1999) 7379. [8] J. Bratby, Coagulation, Flocculation in Water and Wastewater Treatment, 2nd ed., IWA Publishing, London, 2006. [9] N. Koprivanac, G. Bosanac, Z. Grabaric, S. Papic, Treatment of wastewaters from dye industry, Environ. Technol. 14 (4) (1993) 385390.

Fig. 4. Response surface plot (a) and contour plot (b) of predicted dye removal factor as function of the operating aluminium sulphate concentration and initial pH at T = 30 C.

was required to achieve a maximum removal in basic condition. Finally, the aim of this work is to reuse this treated water in the leather dyeing bath. In addition to this, the process water for the dyeing stage should be free from Fe [44]. Therefore, the present work is designed to avoid the application of Fe metal coagulant. It is interesting to mention that in all studied ranges of operating parameters, strong interaction effects exist between the temperature and the concentration of aluminium sulphate, while the interaction effects between the pH and the temperature from one side and the pH and the aluminium sulphate from the other side are negligible. This nding shows that at the equilibrium, the aggregation of dye and the solubility of aluminium suphate might be similar. Therefore the effects of pH interactions towards the two parameters temperature and concentration were negligible. In a previous study [43], at the same concentration range of aluminium sulphate, the obtained dye removal shows also similar values for pH values in the range 56. The main objective of this study is to determine the optimum operating conditions in order to maximize the dye removal factor. This has been carried out by means of Monte Carlo simulation (MCS) method. The MCS is a stochastic optimization technique that generates the random values of the input variables and correspondingly generates a response inside the valid region (i.e. experimental region) that comprises the values of the objective function according to the random values of variables [17]. Table 5 presents the coordinates of the obtained optimal point in terms of the actual operating variables as well as the predicted value of the dye removal factor. As can be observed the optimum predicted dye removal factor is found to be higher than 100%, which is not logical from the experimental point of view. This is because in the optimization step, a restriction condition for the response, 100%, was not considered. In any way, the restriction was not taken into consideration in order

M. Khayet et al. / Chemical Engineering Journal 167 (2011) 7783 [10] D.J. Joo, W.S. Shin, J.H. Choi, S.J. Choi, M.C. Kim, M.H. Han, T.W. Ha, Y.H. Kim, Decolorization of reactive dyes using inorganic coagulants and synthetic polymer, Dyes Pigm. 73 (1) (2007) 5964. [11] D.C. Montgomery, Design and Analysis of Experiments, 5th ed., John Wiley & Sons, New York, 2001. [12] R.H. Myers, D.C. Montgomery, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, 2nd ed., John Wiley & Sons, New York, 2002. [13] S. Akhnazarova, V. Kafarov, Experiment Optimization in Chemistry and Chemical Engineering, Mir Publishers, Moscow, 1982. [14] M. Khayet, C. Cojocaru, C. Garcia-Payo, Application of response surface methodology and experimental design in direct contact membrane distillation, Ind. Eng. Chem. Res. 46 (17) (2007) 56735685. [15] M. Khayet, C. Cojocaru, M.C. Garcia-Payo, Experimental design and optimization of asymmetric at-sheet membranes prepared for direct contact membrane distillation, J. Membr. Sci. 351 (12) (2010) 234245. [16] M. Tomescu, A. Isaicmaniu, S. Cretu, G. Rachiteanu, Application of statisticalmathematical methods to the operation of chemical processes, Rev. Chim. 35 (11) (1984) 10121017. [17] M. Redhe, M. Giger, L. Nilsson, An investigation of structural optimization in crashworthiness design using a stochastic approach a comparison of stochastic optimization and the response surface methodology, Struct. Multidisciplinary Opt. 27 (6) (2004) 446459. [18] C. Cojocaru, G. Zakrzewska-Trznadel, Response surface modeling and optimization of copper removal from aqua solutions using polymer assisted ultraltration, J. Membr. Sci. 298 (12) (2007) 5670. [19] K. Ravikumar, K. Pakshirajan, T. Swaminathan, K. Balu, Optimization of batch process parameters using response surface methodology for dye removal by a novel adsorbent, Chem. Eng. J. 105 (3) (2005) 131138. [20] O. Lacin, B. Bayrak, O. Korkut, E. Sayan, Modeling of adsorption and ultrasonic desorption of cadmium(II) and zinc(II) on local bentonite, J. Colloid Interface Sci. 292 (2) (2005) 330335. [21] D.R. Hamsaveni, S.G. Prapulla, S. Divakar, Response surface methodological approach for the synthesis of isobutyl isobutyrate, Process Biochem. 36 (11) (2001) 11031109. [22] S.S. Moghaddam, M.R.A. Moghaddam, M. Arami, Coagulation/occulation process for dye removal using sludge from water treatment plant: optimization through response surface methodology, J. Hazard. Mater. 175 (13) (2010) 651657. [23] A. Anouzla, Y. Abrouki, S. Souabi, M. Sa, H. Rhbal, Colour and COD removal of disperse dye solution by a novel coagulant: application of statistical design for the optimization and regression analysis, J. Hazard. Mater. 166 (23) (2009) 13021306. [24] A. Aleboyeh, N. Daneshvar, M.B. Kasiri, Optimization of C.I. Acid Red 14 azo dye removal by electrocoagulation batch process with response surface methodology, Chem. Eng. Process. 47 (5) (2008) 827832. [25] J. Beltran-Heredia, J. Sanchez-Martin, A. Delgado-Regalado, Removal of dyes by Moringa oleifera seed extract. Study through response surface methodology, J. Chem. Technol. Biotechnol. 84 (11) (2009) 16531659. [26] M.E. Olya, M.B. Kasiri, H. Aleboyeh, A. Aleboyeh, Application of response surface methodology to optimise CI Acid Orange 7 azo dye mineralization by UV/H2 O2 process, J. Adv. Oxidation Technol. 11 (3) (2008) 561567.

83

[27] A. Palamakula, M.T.H. Nutan, M.A. Khan, Response surface methodology for optimization and characterization of limonene-based coenzyme Q10 selfnanoemulsied capsule dosage form, AAPS PharmSciTech 5 (4) (2004) e66. [28] H.B. Zhao, R.G. Tonkyn, S.E. Barlow, C.H.F. Peden, B.E. Koel, Fractional factorial study of HCN removal over a 0.5% Pt/Al2 O3 catalyst: effects of temperature, gas ow rate, and reactant partial pressure, Ind. Eng. Chem. Res. 45 (3) (2006) 934939. [29] R. Lemoine, A. Behkish, B.I. Morsi, Hydrodynamic and mass-transfer characteristics in organic liquid mixtures in a large-scale bubble column reactor for the toluene oxidation process, Ind. Eng. Chem. Res. 43 (19) (2004) 61956212. [30] M. Ahmadi, F. Vahabzadeh, B. Bonakdarpour, E. Mofarrah, M. Mehranian, Application of the central composite design and response surface methodology to the advanced treatment of olive oil processing wastewater using Fentons peroxidation, J. Hazard. Mater. 123 (13) (2005) 187195. [31] J.H. Ramirez, C.A. Costa, L.M. Madeira, Experimental design to optimize the degradation of the synthetic dye Orange II using Fentons reagent, Catal. Today 107108 (2005) 6876. [32] P. Rana-Madaria, M. Nagarajan, C. Rajagopal, B.S. Garg, Removal of chromium from aqueous solutions by treatment with carbon aerogel electrodes using response surface methodology, Ind. Eng. Chem. Res. 44 (17) (2005) 65496559. [33] X.K. Gao, T.S. Low, Z.J. Liu, S.X. Chen, Robust design for torque optimization using response surface methodology, IEEE Trans. Magn. 38 (2) (2002) 11411144. [34] I. Xiarchos, A. Jaworska, G. Zakrzewska-Trznadel, Response surface methodology for the modelling of copper removal from aqueous solutions using micellar-enhanced ultraltration, J. Membr. Sci. 321 (2) (2008) 222231. [35] A. Santafe-Moros, J.M. Gozalvez-Zafrilla, J. Lora-Garcia, J.C. Garcia-Diaz, Mixture design applied to describe the inuence of ionic composition on the removal of nitrate ions using nanoltration, Desalination 185 (13) (2005) 289296. [36] A.F. Ismail, P.Y. Lai, Development of defect-free asymmetric polysulfone membranes for gas separation using response surface methodology, Sep. Purif. Technol. 40 (2) (2004) 191207. [37] M. Khayet, C. Cojocaru, G. Zakrzewska-Trznadel, Response surface modelling and optimization in pervaporation, J. Membr. Sci. 321 (2) (2008) 272283. [38] C. Cojocaru, M. Khayet, G. Zakrzewska-Trznadel, A. Jaworska, Modeling and multi-response optimization of pervaporation of organic aqueous solutions using desirability function approach, J. Hazard. Mater. 167 (13) (2009) 5263. [39] M. Khayet, M.N. Abu Seman, N. Hilal, Response surface modeling and optimization of composite nanoltration modied membranes, J. Membr. Sci. 349 (12) (2010) 113122. [40] S. Petrov, P.A. Stoichev, Reagent ultraltration purication of water contaminated with reactive dyes, Filtration Separation 39 (8) (2002) 3538. [41] A. Akbari, J.C. Remigy, P. Aptel, Treatment of textile dye efuent using a polyamide-based nanoltration membrane, Chem. Eng. Process. 41 (7) (2002) 601609. [42] S. Gaydardzhiev, J. Karthikeyan, P. Ay, Colour removal from model solutions by coagulation surface charge and oc characterisation aspects, Environ. Technol. 27 (2) (2006) 193199. [43] A.Y. Zahrim, N. Hilal, C. Tizaoui, Evaluation of several commercial synthetic polymers as occulant aids for removal of highly concentrated C.I. Acid Black 210 dye, J. Hazard. Mater. 182 (13) (2010) 624630. [44] C.H. Spiers, Chemical engineering in the leather industry, Chem. Eng. Res. Des. 14 (1936) 129142.