Introduction

The accelerated development of textile industries is enhancing higher rate of water pollution in the environment (Nidheesh et al. 2013). Dye molecules consist of two key segments: chromophores, helps in producing color and auxochromes which are accountable for enhancing affinity toward the fiber and render the molecule soluble in water (Nidheesh et al. 2013). The most significant chromophores are azo (–N=N–), carbonyl (–C=O), methine (–CH=), nitro (–NO2) and quinoid groups. The most important auxochromes are amine (–NH3), carboxyl (–COOH), sulfonate (–SO3H) and hydroxyl (–OH) (Dos Santos et al. 2007). The auxochromes can be owned to the classes of reactive, acid, direct, basic, mordant, disperse, pigment, vat, ingrain, sulfur and solvent dye (Welham 2000).

Enormous quantities of wastewater with significant concentrations of chemical oxygen demand (COD), suspended solids, refractory organics and color are introduced to the environment by textile industries (Garca-Montano et al. 2006). Discharge of textile processed effluents has created unfavorable conditions in the environment (Emami et al. 2010) like reduction in solar energy penetration to the aquatic bodies, reduced reoxygenation capacity and being highly toxic toward aquatic flora and fauna (Clarke and Anliker 1980). Additionally, they are resilient to degradation due to their complex chemical structure and xenobiotic properties (Cristovao et al. 2011).

Azo dyes consist of –N=N– bond and are the most commonly used dyes (Meric et al. 2004). Degradation of azo dyes which contributes about 70% of all used dyes is crucial due to their high structural diversity (Nidheesh et al. 2013). Moreover, the reductive cleavage of azo bonds is responsible for the formation of metabolites of amines, which are classified as toxic and carcinogenic (Puvaneswari et al. 2006) and more noxious than intact dye molecules (Chung et al. 1978). Viewing the inimical effects on environment due to dyes, it becomes necessary to have an adequate treatment for textile wastewaters.

There are numerous physicochemical techniques used like adsorption by activated carbon (Nawaz and Ahsan 2014), photocatalysis and photocatalytic oxidation (Alinsafi et al. 2007), chemical coagulation (Yadav et al. 2013), ozonation (Sevimli and Sarikaya 2002) and chemical precipitation (Tunay et al. 1996) which have been reviewed for treatment of textile wastewater and their COD and color removal efficiencies are given in Table 1. However, physical and chemical treatment techniques are powerful in color removal but allow only a phase transfer of pollutant requiring additional treatment or disposal (Blanco et al. 2012). Moreover, most of the organic matter present in textile wastewater is non-biodegradable or toxic, and therefore, it becomes difficult for biological treatment to reach its required efficiency (Ugyur and Kök 1999).

Table 1 COD and color removal efficiencies of different physicochemical techniques
Full size table

In this frame, advanced oxidation processes (AOPs) are the ones involving the production of hydroxyl radical (·OH, E° = 2.8 V vs. NHE), efficient enough to destroy the recalcitrant compounds to CO2 and H2O (Blanco et al. 2012; Mansoorian et al. 2014). Among AOPs, Fenton oxidation has come out as a promising treatment technique for textile wastewaters (Nidheesh et al. 2013). Fenton reagent is a homogenous catalytic oxidation process that uses a mixture of ferrous ion and hydrogen peroxide in an acidic medium for generation of hydroxyl radicals (Bahmani et al. 2013). Chain reactions involved in Fenton oxidation process are shown below in Eqs. 1 and 2 (Bianco et al. 2011):

$${\text{Fe}}^{2 + } + {\text{H}}_{2} {\text{O}}_{2} \to {\text{Fe}}^{3 + } + {\text{HO}}^{ - } + {\text{HO}} \cdot \left( {{\text{chain}}\;{\text{initiation}}} \right)$$
(1)
$${\text{OH}} \cdot + {\text{Fe}}^{2 + } \to {\text{Fe}}^{3 + } + {\text{HO}}^{ - } \left( {{\text{chain}}\;{\text{termination}}} \right)$$
(2)

The advantages of Fenton reagent are high efficiency, easy applicability, low investment cost (Rodrigues et al. 2009), lack of residues, capacity to treat complex compounds (Bianco et al. 2011) and less environmental damages (Babuponnusami and Muthukumar 2014), whereas the limitations of this process are production of sludge containing high amounts of ferric ions as well as higher operational costs due to usage of chemical reagents (Meric et al. 2004; Sun et al. 2007).

The process parameters are required to determine the performance of Fenton reaction (Zhu et al. 2011). Response surface methodology (RSM) is a useful statistical method widely used for experimental design and widely applied for optimization of wastewater treatment (Rosales et al. 2012). Rare studies for optimization of simulated textile wastewater with a mixture of 5 azo dyes with Fenton process have been performed. The most relevant part of RSM is the selection of applicable design of experiment that has higher impact on response surface and its accuracy on prediction (Sakkas et al. 2010). The application of Box–Behnken design (BBD) is a standard, stable and favored design used for examining the interaction between variables. BBD is usually very competent due to reduced number of experiments (Nair et al. 2014) and due to inconsideration of points at the vertices of a cube which could be considered as an advantage as experiments carried out at these points may turn out to be either expensive or unfeasible (Montogomery 2010). As BBD designs are economical, they are widely used for research purposes.

In this study, Box–Behnken design and response surface methodology were applied to design the experiments and assess the effects of variables like pH, iron and hydrogen peroxide concentration on application of Fenton oxidation treatment to simulated textile wastewater. Two regression models were developed with the experimental data, i.e., one for COD removal and one for color removal. The significance of each variable on the efficiency of Fenton process was studied, and optimal values were acquired and ratified. This study was carried out at Sardar Vallabhbhai National Institute of Technology, Surat, Gujarat, India during August 2014–April 2015.

Materials and methods

Simulated textile wastewater

In this study, simulated textile wastewater was used as the composition of real textile wastewater changes day to day owing to different operations taking place in textile industry. The simulated textile wastewater was prepared in agreement with the information presented in Table 2, and the characteristics obtained are shown in Table 3.

Table 2 Constituents of simulated textile wastewater.
Full size table
Table 3 Characteristics of simulated textile wastewater
Full size table

Samples of the auxiliary products and dyes as presented in Table 2 were taken from Isik and Sponza 2008. The mixture of azo dyes was prepared with concentration 50 mg/L each of Reactive Black 5, Direct Red 80, Direct Black 38, Direct Blue 71 and Direct Fast Yellow 5GL.

Chemicals

All chemicals used in the study were of highest commercially available grade. Analytical grade hydrogen peroxide (30% (w/v), Finar Limited, India) and FeSO4·7H2O (RFCL, India) were used to generate hydroxyl radical in aqueous solution. pH adjustment was done using reagent grade sulfuric acid and sodium hydroxide (RFCL, India) solutions. All the solutions were prepared in distilled water.

Fenton oxidation

Batch experiments of Fenton oxidation were carried out in glass beakers having 1 liter of operating capacity at room temperature (25–30 °C). pH of the sample was adjusted in the range of 2.5–3.5 using 0.5 M H2SO4. After adjusting pH, catalyst FeSO4·7H2O was added and reaction proceeded with further addition of H2O2 producing hydroxyl radicals. During the reaction, sample solution was constantly stirred using magnetic stirrer at 150 rpm for 60 min. After 60 min, residual hydrogen peroxide was eliminated by addition of excess sodium sulfite (Finar Limited, India) and by adjusting pH > 7 before analysis. Settling of samples was carried out in Imhoff cone for 2 h, and samples were analyzed thereafter.

Analytical procedures

The color of the samples was analyzed by measuring the absorbance of supernatant sample at wavelength corresponding to the maximum absorbance, i.e., 478 nm in UV–Vis Spectrophotometer (Hach DR 6000™). As the sample absorbance varied with pH, it was adjusted similar for all supernatant samples. Analytical methods carried out for remaining parameters followed standard methods. COD of the samples was measured using Open Reflux Method (Method 5220 B). pH of the sample was determined using pH-meter.

Dye decolorization and COD removal were calculated using expressions (3) and (4):

$$R_{1} \left( \% \right) = 100 \times \left( {A_{\text{i}} {-} \, A_{\text{t}} } \right)/A_{\text{i}}$$
(3)
$$R_{2} \left( \% \right) = 100 \times \left( {C_{\text{i}} {-}C_{\text{t}} } \right)/C_{\text{i}}$$
(4)

where R 1—decolorization efficiency (%), R 2—COD removal efficiency (%); A i and A t, initial and final absorbance at maximum wavelength; C i and C t, initial and final COD (mg/L) of simulated textile wastewater.

Experimental design, analysis and statistical validation

Response surface methodology (RSM) was used for the optimization of experiments. RSM is an extensively used statistical tool applied in engineering process (Xu et al. 2013). Whenever multiple variables and their interactions affect the responses, RSM can be applied to examine the relation between independent and dependent variables along with optimization of parameters (Baş and Boyaci 2007). Under RSM, for a selected response, it is possible to restrict the number of experiments for recognition of optimal conditions (Rosales et al. 2012). Optimization by RSM involves following steps: (1) selection of independent variables and their ranges, (2) experimental design selection, (3) estimation of mathematical coefficients using linear regression analysis technique and (4) verification of model adequacy and obtaining optimal conditions (Nair et al. 2014). Optimization of results is faster by using RSM rather than one factor at a time approach (Alim et al. 2008).

In this study, Box–Behnken design (BBD) which is widely used in RSM was employed for optimization of COD and color removal efficiencies in simulated textile wastewater. Here, a 23 factorial design was used to perceive the influence of 3 parameters including: H2O2 concentration (A), Fe2+ concentration (B) and pH (C). These factors were chosen as they might show some promising effect on COD and color removal efficiencies. A total of 17 experiments were conducted with 5 replicates at the central point, and the codification of values was done using Eq. 5:

$$x_{i} = z_{i} {-}z_{o} /\Delta z_{i}$$
(5)

where x i  = code level, z i  = uncoded value, z o corresponds to the uncoded value at the central point and Δz i  = step change value between low level (−1) and high level (+1) (Moghaddam et al. 2010). From preliminary experiments, the range of A, B and C was selected as shown in Table 4.

Table 4 Coded level and independent variables used in BBD
Full size table

A second-order regression model was employed for analysis and proves to be a good estimation of response surface (Zhu et al. 2011) and is expressed as shown in Eq. 6:

$$y = \beta_{0} + \sum\limits_{i = 1}^{k} {\beta_{i} X_{i} } + \sum\limits_{i = 1}^{k} {\beta_{ii} X_{i}^{2} } + \sum\limits_{i = 1}^{k = 1} {\sum\limits_{j = i + 1}^{k} {\beta_{ij} X_{i} X_{j} } }$$
(6)

where y = response; x i and x j  = input variables; β 0 = intercept constant; β i  = first-order regression coefficient; β ii  = second-order regression coefficient representing quadratic effect of factor i; and β ij  = coefficient of interaction between two factors i and j (Xu et al. 2013).

The analysis of variances (ANOVA) was carried out using Design Expert (version 9.0.0) to study the results and to determine the implication of fitted quadratic model. The fitted model was thereafter illuminated in the form of contour and surface plots to know the interaction between the variables and responses. The quality of model was checked using co-relation coefficient (R 2) (GilPavas et al. 2012). The model was further authenticated with respect to all the three variables within the design space.

Results and discussion

Preliminary experiments

In this work, before optimization preliminary experiments were carried out to know the range in which maximum color and COD removal took place. The range obtained for Fe2+ concentration was 450–750 mg/L, hydrogen peroxide range was 4400–8800 mg/L, pH was varied from 2.5 to 3.5 as maximum removal of COD, and color was obtained in these ranges. Further precision of doses was done with the help of RSM.

Experimental design by BBD under RSM

According to 23 BBD factorial design experiments were designed and carried out. Among 17 experiments, five experiments were repetition of central point (Run 2, 11, 14, 15 and 16). These runs enclosed the variables which were at central region of the cube. Proximity of the results for these runs shows the symbol of precision maintained during the experimentation work (Azami et al. 2012). The results obtained from experimental design are shown in Table 5.

Table 5 Box–Behnken design matrix along with experimental and predicted values
Full size table

Regression model and statistical testing

For determination of optimal values for decolorization and COD removal within the experimental runs, a second-order polynomial fitted for the results of color and COD removal. The polynomials obtained are shown in Eqs. 7 and 8:

$$R_{1} = 82.22 - 3.81{\text{A}} + 2.46{\text{B}} + 2.07{\text{C}}{-}24.64{\text{A}}^{2} - 8.79{\text{B}}^{2} - 10.4 5 {\text{C}}^{2}$$
(7)
$$R_{2} = 88.26 - 6.80{\text{A}} + 2.76{\text{B}} - 3.40{\text{AB}} - 29.54{\text{A}}^{2} - 6.37{\text{B}}^{2} - 7.12{\text{C}}^{2}$$
(8)

where R 1 and R 2 accounts for COD and color removal, respectively. Tables 6 and 7 shows the ANOVA report generated for this model. In Eqs. 7 and 8, positive effect of a factor means that the response is improved when the factor level increases and a negative effect of the factor means that the response is not improved when the factor level increases (Saldańa-Robles et al. 2014). Values of probability < 0.05 indicate that model terms are significant and values > 0.1000 indicate that model terms are not significant (Rosales et al. 2012). In the case of COD removal, it can be seen that A, B, C, A2, B2 and C2 were significant terms, and rest of the terms were not significant and therefore, not included in the equation for COD removal while for color removal, it was concluded that A, B, AB, A2, B2 and C2 proved to be significant. Thus, statistical analysis showed that not all variables set in the grounding of model within the tested limitations and had noteworthy effect.

Table 6 ANOVA for response surface second-order model for COD removal
Full size table
Table 7 ANOVA for response surface second-order model for color removal
Full size table

The impact of each factor and interactions between each other is checked by Fisher test. The better the magnitude of F value and likewise the lesser the ‘p > F’, the more considerable are equivalent model and individual coefficients (Montogomery 2010). From ANOVA, it was observed that the corresponding F values for COD and color removal efficiencies were 80.39 and 211.39 with a very low probability value (p value < 0.0001) which indicated there was only 0.01% chance that such model could occur due to noise. The F values for COD and color removal indicate that both models were extremely significant and adequate as they were much larger than the critical F value of 2.42 (Rosales et al. 2012).

The overall performance of a model is generally explained by coefficient of determination (R 2) (Nair et al. 2014). For a model to be highly competent, the value of R 2 should be nearest to 1.0. The R 2 evaluates the correlation between experimental data and predicted responses (Fathinia et al. 2010). R 2 values for COD and color removal were obtained as 0.9904 and 0.9963 which indicated that the predicted values matched the experimental ones quite realistically. The models of COD and color removal indicated that 99.04 and 99.63% of the total variability could be explained by these models leaving 0.96 and 0.37% of variability owing to random error. Figure 1 suggests that the predicted values conformed well to the experimental values of COD and color removal of simulated textile wastewater. The model prediction competency should not be examined by R 2 value alone (Sarabia and Ortiz 2009) as R 2 value increases with the increase in number of terms in model, irrespective of its statistical significance. R 2 should be related to adjusted R 2 which reflects the number of factors in experiment (Montogomery 2010). The predicted value of R 2 and the adjusted value of R 2 were in reasonable agreement for both COD and color removal because the difference between both the values was < 0.2. The value of predicted R 2 for COD and color removal was 0.8949 and 0.9744, while the adjusted R 2 values were 0.9781 and 0.9916. As the difference between predicted and adjusted R 2 was less for both models, adequacy of model seemed to be eminently convincing with high values of adjusted R 2 confirming model of strong implication.

Fig. 1
figure 1

Predicted versus actual plot for a COD removal, b Color removal

Full size image

Another statistical tool observed for model adequacy is adequate precision. It scrutinizes the range of the predicted values at the design points to the average prediction error, in other words a signal-to-noise ratio (Mason et al. 2003). The values of adequate precision are in range with a precision of 22.242 for COD removal and 36.883 for color removal. The obtained ratio indicates an adequate signal as values > 4 are desirable. Thus, this model can be used to navigate the design space.

The coefficient of variation is defined as the ratio of standard deviation to the mean value of observed response. This factor demonstrates the repeatability and reproducibility of the models generated (Rosales et al. 2012). Low values of C.V, i.e., 3.77 and 2.33 for COD and color removal indicated high accuracy and dependability of experiments as the values are <10%.

The sufficiency of the models was also evaluated by the residuals (difference between the observed and the predicted responses). Residuals act as components of variation unidentified by the fitted model, and so, they should have a normal distribution (Khataee et al. 2012). Normal probability plots are formed in order to check normality presumption (Nair et al. 2014). If the plots are normally distributed, the residual plots will maintain linearity with some moderate irregularity of points showing appropriate fitting of model (Carlson and Carlson 2005; Anderson and Whitcomb 2007). The points should distribute uniformly along 45° line maintaining linearity, whereas the points above and under the line indicate areas of over or under prediction (Anderson and Whitcomb 2007). Thus, a review was made for COD and color removal of simulated textile wastewater using normalized plot of residuals as seen in Fig. 2 which showed satisfying normality assumption as the points for both COD and color maintained linearity with the values.

Fig. 2
figure 2

Normal probability plots of studentized residual of the linear model for a COD removal, b Color removal

Full size image

Chemical oxidation by Fenton

Response surface and contour plots for simulated textile wastewater treatment

The response surface (3D) and contour plots (2D) of the model-predicted responses obtained from Design Expert software (version 9.0.0) supported in determining the interactive relationships between the process variables and treatment outputs.

Effect of Fe2+ concentrations on COD and color removal efficiency

Figure 3 exhibits the 3D and 2D plots for the effect of iron concentration on COD removal. As it can be seen from response surface and contour plots, the peak values for both the plots were attained at the values in middle ranges of iron. The values of COD removal increased from 50.5 to 71% for iron doses of 450–600 mg/L and then decreased from 67.8 to 46.4% for iron doses of 650–750 mg/L at hydrogen peroxide dose of 4400 mg/L. Removal of COD can be attributed to the fact that at higher concentration of iron, decomposition of hydrogen peroxide is accelerated leading to the formation of per-hydroxyl radicals (HO2·) (Mansoorian et al. 2014), while at lower iron concentration, the reaction with hydrogen peroxide leads to the formation of hydroxyl radicals which are more reactive than per-hydroxyl radicals. It was reported that methylene red removal was increased from 45 to 75% in the presence of iron ions (Zhou et al. 2007). Wang et al. 2010 described that iron ions reduced COD value by increasing the concentration of iron by 0.33 mM where COD removal reached to 43 from 19%.

Fig. 3
figure 3

Response surface and contour plots for the effect of iron and hydrogen peroxide concentration on COD removal

Full size image

Color removal efficiencies were higher in the range of 500–600 mg/L with values of 71.7 and 80.1% and then decreased to 48% at 750 mg/L at hydrogen peroxide dose of 4400 mg/L as seen in Fig. 5. As the concentration of iron increases, self-scavenging effect takes place at a faster rate than the formation of hydroxyl radical leading to reduction in COD and decrease in degradation rate of pollutants as seen in Eq. 9 (Arslan-Alaton et al. 2009).

$${\text{Fe}}^{2 + } + {\text{HO}} \cdot \to {\text{Fe}}^{ + 3} + {\text{HO}}^{ - }$$
(9)

Effect of hydrogen peroxide concentration on COD and color removal efficiency

As seen in Fig. 4, COD removal efficiencies increased up to 62% at hydrogen peroxide dose of 4400–6600 mg/L and then decreased to 56.9% at dose of 8800 mg/L while color removal efficiencies were higher up to with value of 71% at dose of 7700 mg/L and decreased to 68% at dose of 8800 mg/L with constant iron dose value of 450 mg/L. Degradation rate of organic compounds increases with increased dose of hydrogen peroxide but up to a certain value until a critical H2O2 concentration is reached. However, when a concentration higher than critical concentration is used, the degradation was condensed due to scavenging effect as seen in Eqs. 10, 11 and 12 (Hsueh et al. 2005, Lodha and Chaudhari 2007). From Eq. 10 it can be conferred that due to the production of peroxyl radial as well as the scavenging effect of hydrogen peroxide, the removal efficiency of COD and color decreases. The unused quantity of H2O2 also leads to higher values of COD and might be the possible reason for more COD values and lesser removal of COD at higher concentration of H2O2.

Fig. 4
figure 4

Response surface and contour plots for the effect of iron and hydrogen peroxide concentration on color removal

Full size image
$${\text{HO}} \cdot + {\text{H}}_{2} {\text{O}}_{2} \to {\text{HO}}_{2} \cdot + {\text{H}}_{ 2} {\text{O}}$$
(10)
$${\text{HO}}_{2} \cdot + {\text{ HO}} \cdot \to {\text{O}}_{2} + {\text{H}}_{ 2} {\text{O}}$$
(11)
$${\text{HO}} \cdot + {\text{ HO}} \cdot \to {\text{H}}_{ 2} {\text{O}}_{ 2}$$
(12)

Also, at higher dosages of H2O2, the deactivation of Fe2+ takes place which allows the formation of brownish yellow precipitates and thereby increasing the color of the wastewater after the critical concentration. Additionally, it was found that the ratio for influent COD and hydrogen peroxide concentration varied in the range of 0.44–0.89, where it was concluded that maximum removal for COD and color was obtained at a ratio of 0.65–0.7 and then it decreased as the concentration of hydrogen peroxide was not enough for breaking up of recalcitrant dye molecules.

The ratio of hydrogen peroxide concentration to iron dose was seen in the range of 9.8–11.8 where it was found that as the ratio increased, the removal was seen in an ascending manner and then it decreased after a ratio of 11 which concluded that as the ratio increased more peroxyl radical was produced leading to degradation of organic molecules but after the critical concentration of hydrogen peroxide was reached, the removal was lowered.

According to Mansoorian et al. (2014), the COD removal efficiency increased at H2O2 concentration of 2–5 mM, and then the efficiency decreased after the concentration was more than 5 mM. Also, Zhang et al. (2005) observed that efficiency of hydrogen peroxide for removal of organic materials in the leachate decreased with the increase in Fenton reagent dose.

Effect of pH on COD and color removal efficiency

As seen in Figs. 5 and 6, maximum COD and color removal were obtained in the range of 3–3.2. Efficiency for COD removal increased from 47.5 to 61% at pH from 2.5 to 3 and then decreased to 56% at pH 3.5 at iron dose of 450 mg/L. Color removal efficiency increased from 58 to 68% at pH of 2.5–3 and then decreased to 61% at pH of 3.5. The values obtained are sustained by previous studies as Lucas and Peres (2006) observed an optimum pH of 3 for the decolorization of dyes. Meric et al. (2004) depicted more than 99% color removal in the pH range of 3–3.5. Previous studies show that more hydroxyl radicals are generated in acidic range and this might be the possible reason of decolorization and COD removal. Under highly alkaline conditions, the precipitation of Fe(OH)3 occurs; therefore, the concentration of Fe+3 ions decreases in wastewater. Also, in alkaline conditions hydrogen peroxide is less stable, and therefore, lesser is the production of hydroxyl radicals which leads to the decrement in efficiency of process for color and COD removal (Rodrigues et al. 2009). At pH below 3, hydrogen peroxide will also take a proton to form oxonium ion (H3O2)+ which produces more stable hydrogen peroxide and lower its reactivity with ferrous ions (Kavitha and Palanivelu 2005; Kwon et al. 1999). Under such conditions when pH is < 3, reduction in the formation of hydroxyl radicals is due to decrease in soluble amount of Fe+3 that is in equilibrium with other iron species (Fe (OH)2) and Fe (OH)+2 (Lucas and Peres 2006).

Fig. 5
figure 5

Response surface and contour plots for the effect of pH on COD removal

Full size image
Fig. 6
figure 6

Response surface and contour plots for the effect of pH on color removal

Full size image

Perturbation plots

Perturbation plots shows the effect when all factors at optimal experimental conditions in the design space are compared at center point. When the factor curvature is sharper, this factor is more important for the response (Saldańa-Robles et al. 2014). Figure 7 shows the perturbation plots for COD and color removal. Here, in the plots, A represents iron dose, B represents hydrogen peroxide dose, and C represents pH. It can be seen from the graph that factor A is having the steepest slope as compared to factor B and C having relatively flat slopes for both COD and color removal. A steeper slope is showing that the response is more sensitive to that factor as compared to the other two factors taken for consideration. The plots generated are nothing but “one factor at a time” experimentation where only one factor is varied keeping others constant.

Fig. 7
figure 7

Perturbation plots for COD and color removal

Full size image

Optimization of Fenton treated simulated textile wastewater for COD and color removal

In case of assorted responses, RSM appoints set of specific working conditions that functions in maximizing all responses or at least keeps them in the desired ranges (Myers and Montgomery 2002). In this study, the chosen objectives in terms of color and COD removal efficiencies were represented as “maximize” to achieve maximum treatment while the process variables Fe2+ and H2O2 concentrations were selected to be “minimize” considering operating costs while pH variable was chosen with a condition of “maximize” with reference to less usage of chemicals. Subsequently, optimum working conditions and respective percent removal efficiencies were retrieved. The optimum values obtained with the application of these confined conditions were: Fe2+ dose—550 mg/L, H2O2 dose—5538 mg/L, pH—3.3 with COD and color removal of 73.86 and 81.35%.

An additional study was conducted applying the optimal conditions to certify the agreement of results achieved from models. It was found that the values obtained were in close agreement with the model values, i.e., COD removal −74.2% and color removal −80.4% with standard deviation of % ±1.23% and ±2.48 at an interval of ±5%. It can be known that COD and color removal response surface plots showed clear peaks which proved that the optimal conditions of Fe2+ dose, H2O2 dose and pH were within the design boundary. High R 2 values of the models indicated that RSM approach was suitable and precise for predicting Fenton parameters which was conferred well with the experimental results.

Conclusion

In this work, simulated textile wastewater was treated by Fenton method and optimization of operational parameters was studied. Box–Behnken design and response surface methodology were exercised to assess the independent and interactive effects of the parameters involved in study. The quadratic models generated showed immense significance with high values of R 2 showing that the predicted values matched the experimental ones quite fairly. Dosage of iron was the most significant factor as obtained from perturbation plots for both COD and color removal. Also, to check the normality assumption, plot of residuals were found in a straight manner showing satisfactory results for both models. Response surface and contour plots showed the effects of operational parameters in a logical manner. Based on the model prediction, the optimum conditions were obtained as Fe2+ dose—550 mg/L, H2O2 dose—5538 mg/L, pH—3.3 with COD and color removal efficiencies of 73.86 and 81.35%. Therefore, it was found relevant to use Fenton treatment for simulated textile wastewater. Viewing the results, Fenton treatment will prove as a reasonable method as a pretreatment to textile wastewater where direct biological oxidation is not possible due to the presence of recalcitrant organic molecules of dyes.

Lastly, a considerable attention can be given upon coupling of Fenton process with an optimized dosage with biological aerobic/anaerobic treatment which can lead to environmental friendly and feasible solution to such type of textile wastewaters.