Figure 3.
Removal of different dyes by EC (Ci = 100 mg/L, i = 54.61 A/m2, pHi = 5.5, CNaCl= 26 mM, DC mode).
It can be noticed that the behavior of the three dyes considered is not much different, either EC is used to remove an acid dye, such as Acid Blue 74, or a basic dye such as Basic Red1. Moreover, the aqueous solutions containing the most refractory dye considered in this study, Reactive Black 5 decolorized the fastest. Due to its higher persistence compared to the other studied dyes, Acid Blue 74 was chosen as the pollutant model in the present FFD study.
Figure 4.
Evolution of color removal efficiency during
EC/CAG coupling (experimental conditions are depicted in
Table 3; solid line—
DC, dashed line—
APC; light blue, blue and navy blue lines—low, center and high values of dye concentration).
Figure 5.
Evolution of unit energy demand (
UED) during
EC/CAG coupling (experimental conditions are depicted in
Table 3; solid line—
DC, dashed line—
APC; light blue, blue and navy blue lines—low, center and high values of dye concentration).
Figure 6.
Evolution of unit electrode material demand
(UEMD) during
EC/CAG coupling (experimental conditions are depicted in
Table 3; solid line—
DC, dashed line—
APC; light blue, blue and navy blue lines—low, center and high values of dye concentration).
3.1. Effects on the Color Removal Efficiency
Values of color removal efficiency obtained in the randomized 22 runs lie in the range of 11.58% and 100.0%. In order to identify the statistical significant factors and interactions, analysis of variance was employed. It is worth mentioning that main effects represent the difference of the averaged responses for the two levels (+1,−1) of a given factor [
34,
35]. A two-factor interaction effect can be determined as half the difference between the main effects of one factor at the two levels of the second one [
35].
Figure 7 depicts the normal plot of the standardized effects and the standardized Pareto chart,
i.e., main factors and interactions as a function of the standardized effects. On the Pareto chart (
Figure 7b), the standardized effect values of significant (α = 0.95) factors and interactions are higher than the critical value [
36]. Though, the Pareto chart allows one to compare the absolute values of the effects of each factor or interaction of the considered
FFD, normal plot of standardized effects is more accurate in determining, respectively, the significance and insignificance of each effect (
Figure 7a). On a normal probability plot of effects, the non-significant ones fall along a straight line,
i.e., normal distribution, and tend to be centered near zero. In contrast, the following factors caused significant deviations from the straight line [
34]. Current density (
A), contact time (
C), and initial concentration of dye (
F) have the most important effects on color removal efficiency. Other significant factors include pH (
B), concentration of electrolyte support (
E), GAC dose (
D), interaction between pH and GAC dose (
BD), and interaction between current density, GAC dose, and pH (
ABD).
Figure 7.
Normal plot of the (a) standardized effects; and (b) standardized Pareto chart for Y response.
Figure 7.
Normal plot of the (a) standardized effects; and (b) standardized Pareto chart for Y response.
Only a few authors approached wastewater treatment by conventional
EC operated in
APC mode [
15,
16,
17]. Since their results are rather contradictory, one of the goals of the present study was to establish the influence of current type on the performance of GAC-enhanced
EC technique. To this aim, the main effect plot and interactions plot shown in
Figure 8 allow one to analyze in depth the effects of factors considered in the
FFD.
Figure 8.
(a) Main effects; and (b) interactions plots pointing out the effects on color removal efficiency. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors, and the right ends depicts higher levels.
Figure 8.
(a) Main effects; and (b) interactions plots pointing out the effects on color removal efficiency. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors, and the right ends depicts higher levels.
Table 4 presents a statistical summary of the mathematical models suggested for the considered responses.
Table 4.
Statistical summary of Y, UED and UEMD models.
Table 4.
Statistical summary of Y, UED and UEMD models.
Response | Y, % | UED, kWh/kg | UEMD, kg/kg |
---|
Transformation | None | Log | Log |
Lack of Fit p value | 0.336 | 0.124 | 0.196 |
Model p value | <0.0001 | <0.0001 | <0.0001 |
Model F value | 101.71 | 3969.8 | 1628.9 |
Curvature p-value | <0.0001 | 0.002 | <0.0001 |
Significant model terms | – | – | – |
* | A, B, C, D, E, F, ABD | A, B, C, D, E, F, G, AC, AD, AE, AF, BD | A, B, C, D, E, F, AC, AD, AE, AF, BD |
** | BD | AB, ABD | ABD |
*** | – | – | G, AD |
R2 | 0.9854 | 0.9998 | 0.9996 |
Radj2 | 0.9693 | 0.9993 | 0.8342 |
Rpred2 | 0.835 | 0.9187 | 0.9982 |
PRESSa | 2586 | 0.9342 | 0.8183 |
Sb | 4.788 | 0.0199 | 0.0204 |
The mean value of color removal efficiency,
Y, obtained in corner points of
FFD (
Figure 8a) tends to decrease in the case of
APC use. In contrast, the mean value of color removal efficiency corresponding to the runs performed in the center point increases in the case of
APC use. The interactions plots (shown in
Figure 8b) and ANOVA test (
Table 4) shed some light on this issue. For instance, according to the ANOVA test, the interaction between GAC dose and pH parameter (
BD) is statistically significant for a 95.0% confidence level. GAC dose is more significant in the case of a high value of pH, which is due to the residual acidity of Pica L27 [
25]; taking into account that most textile effluents are of an alkaline character [
4,
5] and that effluents treated by
EC result in higher values of pH. The benefits of adding a certain amount of this kind of GAC consist in a decrease of pH value, especially at low values of current density, and, therefore, a faster removal of dyes.
According to
FFD generation structure, this interaction is confounded with
CF and
EG interactions [
29]. The interaction between the current type (
G) and the concentration of electrolyte support (
E) reveals that GAC-enhanced
EC operated in
DC mode provides higher values for
Y response at the low level of
E. This might be explained by the fact that in
DC mode the flow of the ions is not perturbed. In contrast, in
APC mode, the change of electrode polarity leads to a change in the direction of electrophoretic transport of charged particles. Statistically averaged, the current type factor seems to have no influence on the color removal efficiency (
Figure 7). However, from
Figure 4c and
Figure 8a it is clear that the use of
APC mode has positive effects on
Y response. In the center of the experimental region, the treated effluent has an adequate conductivity that favors the electrophoretic transport. The middle value of current density range also enables emphasizing the positive effect of
APC. For instance, at the high level of current density factor, the separation process becomes very fast, and the factor of current type becomes marginal.
Compared to the initial state (when all the terms are included), quality-of-fit indicators of the model obtained after removing the insignificant terms (
G,
AC,
AE,
AF and
AG), improved remarkably. As more terms are added to the model, the coefficient of multiple determination, R
2, increases, which is the reason why we obtained a lower value for this coefficient (0.9911
vs. 0.9854) after removing the insignificant terms [
36]. In contrast, when significant terms are added or insignificant ones are removed, this slightly improved (0.9627
vs. 0.9693) the adjusted determination coefficient, R
adj2, The final suggested model for
Y response also provides a lower value for the square root of the mean square error (5.727
vs. 4.788), a much lower value for PRESS (102087
vs. 2586), a good value for R
pred2 (0
vs. 0.835) and a high
p value of the lack of fit (0.336 > 0.05),
i.e., inadequacy is not significant. Therefore, the model suggested for color removal efficiency correctly describes the experimental data.
The final equation obtained for the decolorization of Acid Blue 74 aqueous solutions by GAC-enhanced
EC is described by Equation (5) given in
Table 5. Though the effects of
AB and
AD interactions are not statistically significant, the significant
ABD interaction implies that these terms should be kept in order to obtain a hierarchical model.
In order to emphasize the statistical validation of the model for color removal efficiency, experimental data were plotted against the predicted ones as presented in
Figure 9a.
The experimental values vary within the confidence interval of the values predicted by the suggested model, which supports the fact that the lack of fit is not significant. Therefore, it can be concluded that the model obtained to describe the response of color removal efficiency is adequate. However,
Table 5 pinpoints the existence of curvature in the polynomial model. In order to correctly describe the entire experimental region, a second order polynomial model obtained by response surface methodology would be recommended. Only the factors with the highest effects on the response should be considered. However, this task is beyond the goals of the present work.
Table 5.
Model equations (coded values)*.
Table 5.
Model equations (coded values)*.
Response | Expression | Equation Number |
---|
Y = | 64.71 + 19.96A − 6.55B + 16.31C + 4.45D + 6.01E − 10.55F + 0.04AB − 0.84AD + 3.64BD + 5.81ABD | (5) |
Log(UED) = | −1.0872 + 0.0575A + 0.0519B + 3.67 × 10−3C − 0.119D − 7.53 × 10−3E − 2.16 × 10−3F − 0.093G + 1.76 × 10−4AB + 3.46 × 10−5AC + 0.0277AD − 4.11 × 10−4AE − 3.47 × 10−5AF + 4.82 × 10−3AG − 0.06BD − 0.002ABD | (6) |
Log(UEMD) = | 0.597 + 0.341A + 0.058B + 0.0332C − 0.0524D − 0.03E − 0.251F − 0.009G − 0.0051AB + 0.05AC + 0.011AD − 0.021AE − 0.028AF + 0.041AG − 0.051BD − 0.018ABD | (7) |
Figure 9.
Statistical validation of the models for (a) Y; (b) UED; (c) UEMD responses.
Figure 9.
Statistical validation of the models for (a) Y; (b) UED; (c) UEMD responses.
3.2. Effects on the UED Response
Unit Energy Demand (
UED) response is defined by Equation (2). Values obtained for
UED response range between 0.03 and 30.04 kWh/kg, and the ratio of maximum to minimum is greater than 10 (898.9), which indicates that a transformation is required [
34]. According to the Box-Cox plot for power transformation, Minitab software indicated that the logarithm of response values is the proper transformation.
Figure 10 presents the normal plot and Pareto chart. It can be noted in the normal plot of
UED response (
Figure 10a) that no straight line crosses the points representing the effects of the coefficients in the model. The Pareto plot in
Figure 10b pinpoints that all the main and interaction effects of 2
7-3 FFD have significant effects on
UED response.
Figure 11 presents the main effects and interaction plots for
UED response.
The current density and contact time factors present the most important effects on the UED response. The concentration of background electrolyte factor has a greater effect on UED response compared to the initial concentration of dye factor. The current type effect is also significant.
Figure 10.
Normal plot of the (a) standardized effects; and (b) standardized Pareto chart for UED response.
Figure 10.
Normal plot of the (a) standardized effects; and (b) standardized Pareto chart for UED response.
Figure 11.
(a) Main effects; and (b) interaction plots for the effects on UED response. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors, and the right ends depict the higher levels.
Figure 11.
(a) Main effects; and (b) interaction plots for the effects on UED response. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors, and the right ends depict the higher levels.
Statistical data support the experimental observations. According to these results, the use of
APC in
EC leads to lower energy consumption. As shown in
Figure 11a, the mean of
UED experimental data for
DC mode is 5.12 kWh/kg. This mean value decreases 10% to 4.65 kWh/kg in the case of
EC operated in
APC mode. Therefore, we claim that the use of
APC in
EC leads to the diminution of
UED compared to
DC mode. In terms of averaged effect, the current type has no significant effect on the response of color removal efficiency. Hence, the advantage of using
APC in
EC technology consists of a significant reduction in electrical energy consumption.
The addition of a certain amount of GAC leads to a faster removal of dye. The energy consumption increases proportionally with the GAC dose. This might be due to the adherence between GAC and metal-electrode surface that results in an increase of the total active electrode surface. Even with this slight increase in the energy consumption, the beneficial effect of a faster separation of the pollutant is dominant. Consequently,
UED,
i.e., energy consumption related to the separated amount of dye, decreases with GAC dose. When adding 0.1 g/L of GAC, an experimental mean value of 5.48 kWh/kg was obtained for
UED response (
Figure 10a), while adding 0.5 g/L GAC results in a reduction of 20.8% (4.34 kWh/kg).
Likewise, an acid initial pH favors the removal of dye and leads to a decrease in
UED response. The mean of the
UED response values, corresponding to an initial pH of 9, is 5.88 kWh/kg, while in the case of initial pH of 3, the mean is 4.05 kWh/kg. This can be correlated also with the effect of the addition of GAC such as Pica L27 that has an acid character [
25]. In case of conventional EC, the magnification in hydroxyl ions generated from water reduction at the cathode, leads to the pH solution increase. When GAC is added into the system, there is a competition between the release of protons due to the acidic functional surface groups of L27 and the generation of hydroxyl ions at the cathode. This is due to water electrolysis and the significantly slower pH increase in the solution.
The adequacy of the model for
UED is relatively good as emphasized by the statistical summary shown in
Table 4 and the experimental
versus predicted values plot shown in
Figure 9b. Equation (6) of the suggested model for
UED response is given in
Table 5.
3.3. Effects on the UEMD Response
UEMD represents the consumption of electrode material in relation to the mass unity of removed dye as defined by Equation (3). Experimental values obtained for this response range between 0.042 and 47.43 kg/kg (ratio of maximum to minimum is 113.7) requiring also the logarithmic transformation.
The normal plot of standardized effects (
Figure 12a) and standardized Pareto chart (
Figure 12b) for
UEMD response outline that the current type factor and
AB and
AD interactions represent the only terms of the model that are not significant. However, taking into account that the superior terms
ABD and
AG are significant, insignificant terms were kept in the model to preserve the hierarchic characteristic.
Figure 12.
Normal plot of (a) the standardized effects; and (b) standardized Pareto chart for UEMD.
Figure 12.
Normal plot of (a) the standardized effects; and (b) standardized Pareto chart for UEMD.
The Pareto chart also emphasizes the strong influence of current density and time factors as well as that of initial dye concentration on the
UEMD response. This is due to the fact that
UEMD response is directly dependent on current density and inversely related to electrolysis time and initial concentration of dye. The effects of pH and GAC dose as well as
AG interaction are significant. This means that electrode material consumption might be diminished if GAC-enhanced
EC is operated at a certain current density. In order to elucidate the interaction effects, the main effects and interactions plots (
Figure 13) obtained for
UEMD response were analyzed.
Figure 13.
(a) Main effects; and (b) interaction plots for the effects on UEMD response. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors and the right ends depict the higher levels.
Figure 13.
(a) Main effects; and (b) interaction plots for the effects on UEMD response. Solid lines represent low black levels (−1 and DC) of the factors; dashed green lines represent high levels (1 and APC); dashed red lines represent the center levels (0); left ends of the lines in each plot means low level of underlying factors and the right ends depict the higher levels.
As shown in the interaction plot (
Figure 13b), while operating
EC at the low level of current density, the mean of
UEMD values obtained in
DC mode is 1.348 kg/kg (
Figure 13b), while for APC mode the
UEMD mean is 1.07 kg/kg. This represents a reduction in the specific consumption of the electrode material of approximately 20%. In contrast,
EC operated at the high level of current density leads to an increase in
UEMD mean of 16.4% in case of
APC mode compared to
DC mode. Therefore, the beneficial effect of using
APC mode decreases to about 4% less material consumed on average. This makes the entire main effect of current type (G) to be statistically significant only at a confidence level of 90%.
The interaction of
AG is confounded with
BF and
CD interactions [
29].
Figure 13b shows also that a higher initial concentration of dye requires less electrode material dissolved at a low pH value. A higher GAC dose leads to a diminution of
UEMD at short durations. This is due to the important increase in electrode material consumed at longer durations in relation to the effect of GAC dose.
Table 4 and
Figure 9c support that the model suggested for
UEMD response (
Table 5) in Equation (7) is adequate.
3.4. Multi-Objective Optimization
Although the suggested models by 27-3 FFD are not very reliable to interpolate precisely the entire experimental region, one can use them to estimate the local optimum, which might serve further as a possible center point of the experimental region in a Response Surface Methodology design.
The goals of the optimization of GAC-enhanced
EC system consist in maximizing the response of color removal efficiency and minimizing
UED and
UEMD responses. To solve this kind of multi-objective optimization problem, Derringer and Suich [
37] suggested the desirability function (Equation 8) that is one of the most appropriate methods. The overall desirability function,
D, is the geometric mean of the individual desirability functions [
38]:
with
di denoting the individual desirability function for each response, and
k the number of responses,
i.e.,
k = 3.
The optimization of the overall desirability function,
D, implies the maximization of the response of color removal efficiency and the minimization of
UED and
UEMD responses. This function is subject to the constraint. In this regard, all the factors take values in the limit of the experimental region explored. Bezerra
et al. [
38] described in detail the methodology of desirability function.
Table 6 presents the goals, criteria, optimal values of responses, and values obtained for global and individual desirability functions. Experimental tests were performed to verify the predicted values of responses.
Table 6.
Optimization criteria and obtained results.
Table 6.
Optimization criteria and obtained results.
Goals | Criteria | Desirability | Results |
---|
Predicted | Experimental |
---|
– | A,B,C,D,E,F,G∈ Ω | Da = 0.215 | – | – |
Max(Y) | dY = 0.432 | 94.32 | 92.24 (s.d.b 2.5) |
Min(UED) | dlog(UED) = 0.346 | 0.144 | 0.178 (s.d.b 0.01) |
Min(UEMD) | dlog(UEMD) = 0.242 | 4.209 | 4.683 (s.d.b 0.11) |
The algorithm of multi-objective optimization Minitab software allowed us to obtain the predicted optimal values. Three confirmation runs were carried out in order to check experimentally the optimal point. Optimal values of responses correspond to a current density of 2.73 A/m2, pH value of 3, GAC dose of 0.5 g/L, salt concentration of 50 mM, dye initial concentration of 50 mg/L, duration of 180 min and APC mode. The optimal predicted values are in good agreement with the experimental ones.
Based on these results, a central composite design can be developed in order to optimize the GAC-enhanced EC system. In our future work, we will consider APC mode of EC operation as an established improvement of this system. Only the continuous flow feature will allows us to estimate the reduction in the consumptions and costs of energy and electrode material in correlation with those of GAC material added.