Raw wastewater was obtained from the food processing of pickling, curing, and preserving fruits and vegetables in the north of Phayao province, Thailand. In general, raw wastewater was contaminated with acetic acid and eliminated by pouring into the sludge drying area with other waste and let it evaporate naturally, which can affect humans and the environment inevitably. The preliminary characteristics of raw wastewater are presented in [Table 1].

[Figure 1] shows the physical characteristics of the raw wastewater used. It can be seen that the color is yellow-brown and unpleasant smell, which wastewater comes from the process of pickling, acetic acid is added to the salt solution (NaCl) to make the pickling brine. Acetic acid acts as a preservative and diffuses into the vegetables, while salt solution keeps the vegetables from swelling [44]. However, to control the lab-scale experiment, this project only focused on the removal of acetic acid at a low pH via Fenton oxidation process.

Figure 1

Brine wastewater from food processing industry

The finding of optimum conditions for pre-treatment of acetic acid in terms of four independent variables were selected for the initial pH, H₂O₂ concentration, Fe²⁺ ions concentration and reaction time along with a fixed temperature of 25 ^oC, was further investigated using with the advanced Fenton oxidation process. In this step of the experimental design, experimental range and levels of center composite design were analyzed by response surface methodology (RSM) via the design of experiment (DOE) software tools called central composite design (CCD). The optimal response of the operation factors was analyzed to obtain the statistical techniques (empirical regression model) and graphical data (ANOVA, analysis of variance) for pre-treatment of acetic acid through RSM. Then, the second-degree polynomial was computed through a four-factor of CCD in order to estimate the advanced design for 30 experimental runs. The experimental runs (N) were accomplished by using 16 (2ⁿ) factorial points, 8 (2n) axial points at the center run and 6 (N₀) central points, which were attained according to eq. (12):

$N=2^n+2n+N_0=(2^4)+\left(2\times 4\right)+\left(6\right)=30,$ (12)

where N is the experimental runs, n is the numbers of operational factors of the initial pH, H₂O₂ concentration, Fe²⁺ ions concentration and reaction time involved, 2ⁿ is the number of factorial points, 2n is the number of axial points and N₀ is the centre points.

The eq. (13), a second-degree polynomial regression model was obtained from the experimental data of 30 runs that were analyzed through the CCD matrix:

$y={\beta }_0+{\displaystyle \sum }_{i=1}^k{\beta }_i{x}_i+{\displaystyle \sum }_{i=1}^k{\beta }_{ii}{x}_i^2+{\displaystyle \sum }_{i=1}^{}{\displaystyle \sum }_{i\ne j=1}^{}{\beta }_{ij}{x}_i{x}_{ij}+\epsilon ,$ (13)

where y is the predicted response, x_i and x_j are the real or code values of the variable parameter obtained from pre-treatment acetic acid runs, ε is the random error and β₀, β_i, β_ii and β_ij are the constant regression coefficient, linear regression coefficient, quadratic regression coefficient and interaction regression coefficient respectively.

For the pre-treatment of acetic acid study via Fenton oxidation process from food processing, CCD allowed four numeric factors were set to range in five levels: low level (-, minus factorial point), high (+, plus factorial point ), center point level (0), and two axial points (plus and minus alpha) corresponding to -1 and +1. In terms of plus and minus alpha are not shown here due to it is the same range and levels of factorial point. The overall range and levels of CCD are illustrated in [Table 2].

Laboratory-scale, the degradation of acetic acid, the Fenton oxidation process were carried out in 1.5 L batch reactor. 1 L of raw wastewater was adjusted to the desired level (pH 2.0 ± 0.1 – pH 5.0 ± 0.1) using 1.0 N H₂SO₄ and/or NaOH (both was purchased from Ajax Finechem) before adding the ratios of hydrogen peroxide oxide (H₂O₂, 0.01 – 0.05 mol/L, Sigma-Aldrich) and Ferrous (FeSO₄.7H₂O, 0.02 – 0.08 mol/L, analytical grade, Ajax Finechem) reagent with the heating of 40 °C temperature and then followed by changing the reaction time of 30 – 150 min. After that, aliquots were withdrawn with adding 10.0 mL NaSO₃ (sodium sulfite, 25% w/v, Sigma-Aldrich)) to stop the reaction immediately for each contact time and filtered through 0.45 μm syringe filters (Glass fiber filter, GA-55). COD (Chemical Oxygen Demand, mg/L, HANNA, HI 96727) and color (Pt-Co, HANNA, HI 96727) were measured as a surrogate parameter to obtain the optimization via Fenton oxidation process for sustainable water reuse. However, all batch experiments were run by following the CCD matrix (code and actual variables), as shown in [Table 6].

The computing of the empirical quadratic models was obtained by using 30 observed responses from the experimental runs that were formulated via the RSM. The percent of pre-treatment efficiency, the highest order polynomial of both empirical quadratic models for color and COD were remarked as suggested, and the model is not aliased by the software as shown in [Table 3].

The optimum condition of pre-treatment for acetic acid, the best percent of removal efficiency for both color and COD was correlated with four factors of H₂O₂ concentration (x₁), Fe²⁺ ions concentration (x₂), initial pH (x₃), and reaction time (x₄). The final empirical models of the suggested quadratic model in terms of the four actual variable factors are shown by eq. (14) and eq. (15). Both suggested quadratic models of color and COD, the negative signs in front of the terms show antagonistic effects, and the positive sign indicates synergistic effects [43].

$\text{Color},\%=48.87+626.81{\text{x}}_1-1838.58{\text{x}}_2+36{\text{x}}_3-0.15\text{x}4-1.04{\text{x}}_1{\text{x}}_2+28.06{\text{x}}_1{\text{x}}_3+0.54{\text{x}}_1{\text{x}}_4+10.18{\text{x}}_2{\text{x}}_3+0.52{\text{x}}_2{\text{x}}_4+0.03{\text{x}}_3{\text{x}}_4-0722.15{\text{x}}_1{}^2+17517.93{\text{x}}_2{}^2-5.45{\text{x}}_3{}^2+2.95\times {10}^{-4}{\text{x}}_4{}^2$ (14)

$\text{COD},\%=49.96+177.70{\text{x}}_1+1077.63{\text{x}}_2-0.75{\text{x}}_3+0.06{\text{x}}_4+42.71{\text{x}}_1{\text{x}}_2+38.73{\text{x}}_1{\text{x}}_3+2.06{\text{x}}_1{\text{x}}_4-85.01{\text{x}}_2{\text{x}}_3+0.09{\text{x}}_2{\text{x}}_4-0.02{\text{x}}_3{\text{x}}_4-8939.25{\text{x}}_1{}^2-7650.78{\text{x}}_2{}^2+1.48{\text{x}}_3{}^2+1.23\times {10}^{-4}{\text{x}}_4{}^2$ (15)

The ANOVA results analysis parameters for the quadratic regression models of Eqs. (14) and (15) are obtainable in [Table 4]. The t-test follows a Student’s t-distribution was used as a tool to verify the Fisher's F statistic (test statistic for F-test) value with a low p-value Prob>F less than 0.0500, indicating the model terms are significant for the regression models. It can be suggested that the regression models indicated the p-value Prob > F more than 0.0500, illustrating the model terms for the response surface quadratic model of eq. (14) and eq. (15) is insignificant. Therefore, the regression models needed to be reduced as shown in eq. (16) and eq. (17).

[Table 5] demonstrates the analysis of variance table for response surface reduced quadratic model, the statistical significance for the regression coefficient of the model terms were confirmed by the results of the p-value Prob>F less than 0.0500. The results found that the model terms of color and COD were of high significance. It can be suggested that these model terms can be successfully applied to predict the optimum conditions for the response of the maximum percent removal efficiency via Fenton oxidation process. Hence, the reduced regression model of eq. (16) and eq. (17) was obtained as follows:

$\text{Color},\%=33.15+726.27{\text{x}}_1-1791.41{\text{x}}_2+38.72{\text{x}}_3+0.04{\text{x}}_4-8938.35{\text{x}}_1{}^2+17866.29{\text{x}}_2{}^2-5.31{\text{x}}_3{}^2$ (16)

$\text{COD},\%=49.50+179.83{\text{x}}_1+1086.88{\text{x}}_2-4.75{\text{x}}_3+0.06{\text{x}}_4+38.73{\text{x}}_1{\text{x}}_3+2.06{\text{x}}_1{\text{x}}_4-85.01{\text{x}}_2{\text{x}}_3-0.02{\text{x}}_3{\text{x}}_4-8939.25{\text{x}}_1{}^2-7650.78{\text{x}}_2{}^2+1.48{\text{x}}_3{}^2+1.23\times {10}^{-4}{\text{x}}_4{}^2$ (17)

As shown in [Table 6], both experimental and predicted percent removal of colure and COD on the Fenton oxidation process. Each of the experimental of percent removal was compared with the predicted value in terms of the errors. The comparison of percent errors with the mean square residual (MSR, [Table 5], 67.80 for color and 42.40 for COD) illustrated that for most of the percent errors of each runs not exceeded twofold as compared to the root mean square residual (RMSR). These results confirmed great appropriateness of the reduced quadratic model for pre-treatment of acetic acid in terms of color and COD as presented in eq. (16) and eq. (17).

[Figure 2] demonstrates the plots of actual values against predicted values of color and COD removal efficiency attained from the reduced quadratic model via the Fenton oxidation process. It can be signified that the responses from the reduced regression quadratic models for all values indicated well sufficiently complying with visually confirming the conclusion as shown in [Figure 2].

[Figure 3] represents the normal percent probability versus externally studentized plot of the residuals, and it confirmed that there did not show any apparent problems with the normal probability plot. A good relationship between input and output variables could be derived from the reduced quadratic model that was established by forming a normal probability plot of the residuals. The normality assumptions were assured as the residual plot approximated along a straight line [44] for both color and COD respectively, which indicates that the reduced quadratic model is accurate for decision.

Figure 2

Actual versus predicted values for color and COD removal efficiency obtained from the reduced quadratic model for food processing wastewater treatment

Figure 3

Normal percent probability of residuals versus externally studentized residuals plots for color and COD removal efficiency derived from the reduced quadratic model for food processing wastewater treatment

Adequate precision was computed the signal to noise ratio and found that a ratio of color (7.71) and COD (6.71) is greater than 4, indicating an appropriated signal to noise ratio [35]. Hence, this reduced quadratic model could be applied to investigate the design space and to obtain the optimal operating conditions via the Fenton oxidation process. Moreover, model F-value of 2.71 indicates the reduced quadratic model is significant with only a 3.47% chance that an F-value this large could occur due to noise as shown in [Table 5] for the color. While COD, the model F-value of 2.46 suggests, the reduced quadratic model is significant with 4.39% chance that an F-value this large could occur due to noise. These results confirmed that the reduced quadratic models had enough precision for the removal of acetic acid via the Fenton oxidation process using RSM. Also, p-value Probe>F of the models showed less than 0.0500 for both color (0.0347) and COD (0.0439), indicating the models present the statistical significance. As the results. It can be seen that the classified reduced quadratic models were applied to display the response surface in three-dimensional (3D) plots and to assess the optimal operating conditions for color and COD removal efficiency via the Fenton oxidation process.

[Figures 4] and [5] illustrate that the reduced quadratic model predicted color and COD efficiency through the Fenton oxidation process. A comparison of the effects of all variable factors at the optimal operating conditions in terms of percent range for color and COD removal efficiency via Fenton oxidation process was achieved by using the perturbation plots as shown in [Figure 6]. However, the steep curvature of Fe²⁺ (B) concentration dose demonstrated the color removal efficiency. On the other hand, pH (C) and reaction time (D) illustrated the COD removal efficiency was extremely affected by those variables, which are clearly shown in [Figures 4] and [5] with the results previously discussed.

Figure 4

3D, color removal efficiency plot of Fe²⁺ vs. H₂O₂, pH vs. H₂O₂, Reaction time vs. H₂O₂, pH vs. Fe²⁺, Reaction time vs. Fe²⁺ and Reaction time vs. pH for food processing wastewater treatment

Figure 5

3D, COD removal efficiency plot of Fe²⁺ vs. H₂O₂, pH vs. H₂O₂, Reaction time vs. H₂O₂, pH vs. Fe²⁺, Reaction time vs. Fe²⁺ and Reaction time vs. pH for food processing wastewater treatment

Figure 6

Perturbation plot of color and COD removal efficiency via Fenton process at the optimal conditions: H₂O₂ = 0.03 mol/L (A), Fe²⁺ = 0.05 mol/L (B), pH = 3.5 (C) and reaction time 90 min (D) for food processing wastewater treatment

The optimization of the variable operating parameters for treated acetic acid during Fenton oxidation process was achieved with the target of desirability the highest removal efficiency for color and COD. Hence, various operational conditions were computed by using the design of the experiment software tool in covering the range of study. Obtaining the maximum percent removal efficiency was established following the desirable goals and the variable parameters of H₂O₂ concentration, Fe²⁺ concentration, initial pH, and reaction time as within the range. The desirability functions/ramps (d) were chosen to consider the various optimum conditions for the Fenton oxidation process, where the d value was set equal to 1.0. The numerical functional analysis and optimization examined that the maximum target percent removal 95.2 % for color could be obtained at H₂O₂ concentration of 0.004 mol/L, Fe²⁺ ions concentration of 0.02 mol/L, initial pH of 3.45, and reaction time of 149.08 min as shown in [Figure 7]. While, COD had a maximum target percent removal of around 84.7 % that could be obtained at H₂O₂ concentration of 0.014 mol/L, Fe²⁺ concentration of 0.051 mol/L, initial pH of 2.04, and reaction time of 144.58 min as shown in [Figure 8].

It can be concluded that Hydroxyl radicals (HO•) are an important role of reactive species, which might be the reactive species produced by the reaction of mixed Fe²⁺ or Fe³⁺ and H₂O₂ in water. Which it depends on the mentioned conditions as above from the viewpoint of the removal efficiency. Eq. (18) shows the mechanism of Fenton-like reaction (Fe²⁺:H₂O₂) at very slow of diffusion-controlled reaction rates (k = 0.001-0.01 M⁻¹ s⁻¹) [47] exists that a few short-chain carboxylic acids (-COOH) are the major by-products (originating from acetic acid) decomposition during the treatment process at a certain level of pH and reaction time:

${\text{CH}}_{\text{3}}\text{COOH}+{\text{H}}_{\text{2}}{\text{O}}_{\text{2}}\stackrel{{\text{Fe}}^{\text{2+}}}{\to }{\text{CO}}_2+{\text{H}}_{\text{2}}\text{O}$ (18)

The significance of this study remarked that the COD 84.7% and color 95.2% removal efficiency were almost higher than of these literatures [45], [46] and similar to this research [37] as compared to triplicated runs via the Fenton process at the laboratory level.

To confirm the agreement of the reduced quadratic model and experimental results achieved, the optimum condition agreement was examined with the experimental data and found the highest percent removal of color and COD obtained. These results were achieved close to those estimated by using the reduced quadratic models as shown in eq. (16) and eq. (17). It can be concluded that using RSM module in the design of expert was appropriate for optimizing the operational conditions for color and COD removal through Fenton oxidation process.

Figure 7

Desirability ramps of the Fenton oxidation process at the maximum of color removal efficiency for food processing wastewater treatment

Figure 8

Desirability ramps of the Fenton oxidation process at the maximum of COD removal efficiency for food processing wastewater treatment