Actual SWW samples were taken from selected provincially licensed meat processing plants directly from their source in Ontario, Canada [7]. A 30% w/w hydrogen peroxide solution was purchased from Sigma-Aldrich, whereas NaOH (99%) and H₂SO₄ (99%) were obtained from EMD Millipore for pH adjustment.

The main source of SWW is the feces, urine, blood, lint, fat, carcasses, and non-digested food in the intestines of the slaughtered animals, the production leftovers, and the cleaning of the facilities [8]. The SWW composition varies according to the industrial process and water demand. Nevertheless, they usually contain high levels of organics and nutrients, typically measured as biochemical oxygen demand (BOD), chemical oxygen demand (COD), total organic carbon (TOC), total suspended solids (TSS), total nitrogen (TN), and phosphorus (TP).

Slaughterhouse effluents are considered detrimental worldwide due to its complex composition of fats, proteins, and fibers, as well as the presence of organics, nutrients, pathogenic and non-pathogenic microorganisms, detergents and disinfectants used for cleaning activities, and pharmaceuticals for veterinary purposes [9]. Therefore, the treatment and disposal of wastewater from slaughterhouses and meat processing plants are an economic and public health necessity [10]. Table 1 attempts to summarize the typical characteristics of the slaughterhouse effluents in Ontario, Canada. The SWW features and common ranges are listed as BOD, COD, TOC, TSS, TN, and pH.

An anaerobic baffled reactor (ABR), followed by an aerobic activated sludge (AS) reactor, and a UV/H₂O₂ photoreactor, operated in continuous mode, were used in a combined system at the laboratory scale for SWW treatment. The schematic diagram of the experimental setup for the combined ABR-AS-UV/H₂O₂ processes is illustrated in Figure 2.

The 50 L combined ABR-AS-UV/H₂O₂system consisted of a 36-L ABR with five equal-volume chambers integrated with individual headspaces, biogas collection piping, and a 13-L aerobic AS reactor with a monitored air flow rate, and a 1-L photoreactor with recycle and uniform light distribution. A 45˚ slanted-edge baffle within each ABR chamber permits the down- and up-flow of the SWW, providing effective mixing and contact time between the SWW and the biomass. The AS air flow rate was set at 2 L/min to guarantee nitrifying bacteria growth and dissolved oxygen (DO) concentrations over 2.0 mg/L.

Anaerobic and aerobic sludge seeds were loaded into the anaerobic and aerobic bioreactors, respectively. The inoculum was acclimatized in two months by feeding the actual SWW continuously into the reactors at a constant flow rate (75 mL/min) while gradually increasing its concentration.

The stainless steel cylindrical photoreactor (Barrier SL-1S-Siemens Inc., Markham, ON) had an external diameter of 8 cm and a length of 34 cm with a 2.5 cm diameter UV-C lamp and output power of 6 W with 254 nm wavelength was inserted into the center of the photoreactor. A quartz sleeve covered the UV-C lamp to protect the lamp from fouling and maintain a uniform UV radiation emission.

TOC concentrations were analyzed for each sample using an automated TOC analyzer (Teledyne Tekmar Apollo 9000, Mason, OH). Temperature and pH were measured daily using a pH meter with atemperature probe (Thermo Scientific Orion 230A+, Ottawa, ON). The H₂O₂ residuals were measured with a UV-Visible Spectrophotometer (Ultrospec 1100 pro-Amersham Biosciences, Amersham, UK) at 454 nm using neocuproine and copper [11]. All experiments were repeated in triplicates, and the average values were reported. Furthermore, three replicates were made for each analytical measurement.

A four-factor along with five-level CCD in conjunction with RSM was used to maximize percent TOC removal and minimize percent H₂O₂ residuals in the effluent. The influent concentration of TOC (X₁), flow rate (X₂), H₂O₂ dosage (X₃), and pH (X₄) were used as independent factors in the DOE; whereas, the percent TOC removal (Y₁) and H₂O₂ residual (Y₂) were considered process responses. Thus, each factor was coded at five levels, from −2 to +2, as shown in Table 2. Previous studies [1] [7] [11] [12] were used to determine and select the critical ranges of the factors.

Equation (1) was used to predict the model responses as a quadratic model and estimate the parametrical

coefficients by correlating dependent and independent variables using the least-squares regression [11]:

where β_o, β_i, β_ii, and β_ij are the constant, linear, quadratic, and cross-factor interaction coefficients, respectively; X_i and X_j represent the independent variables; Y_i is the predicted response; and c and k are the residual term and the number of factors, respectively.

The Design-Expert 9.0.4.1 statistical software was employed for graphical and regression analysis to estimate the coefficients of the response functions. The significance of the independent variables, factor interactions, and model equations were examined by analysis of variance (ANOVA) at 95% confidence intervals (CI).

Three-dimensional (3D) surfaces and two-dimensional (2D) contour plots were obtained while keeping another factor constant in the quadratic models. Experiments were carried out to validate the statistical models for maximum percent TOC removal and minimum H₂O₂ residual.

Optimal operating conditions were estimated using the numerical optimization method built in the software. Lastly, an additional experimental run was carried out to validate the predicted optimal conditions for both response functions, the percent removal of TOC, and H₂O₂ residual.

The desirability multiple response method was used to combine the desirable ranges for each response to obtaining a simultaneous objective function that represents the geometric mean of all transformed responses as shown in Equation (2) [13]:

where D, d_i, and n are the desirability objective function, each response range, and the number of responses, respectively. If any of the analyzed responses is found to be outside of their desirability range, the overall desirability function becomes zero. Therefore, for a simultaneous optimization, each response is required to be assigned low and high values for optimization. In this case, the percent removal of TOC (d₁) is maximized while the H₂O₂ residual (d₂) is minimized.

Table 3 portrays the four-factor, five-level CCD with observed and predicted values for both percent TOC removal and H₂O₂ residual by the developed quadratic models related to the combined ABR-AS-UV/H₂O₂ system in a continuous photoreactor for SWW treatment.

RSM was employed for parameter estimation, indicating the relationship between the input factors and the responses, as shown in Equation (2). Thus, to predict the response functions for percent TOC removal and H₂O₂ residual, the second-order polynomial Equations (3) and (4) were developed, respectively:

Negative coefficients for the model components X₁, X₂, X₃, X₄, X₁X₂, X₁X₃, X₁X₄, X₃X₄, , , and in Y₁ and X₁, X₁X₂, X₁X₃, X₁X₄, X₂X₃, and in Y₂, indicate unfavorable effects on the percent TOC removal and the H₂O₂ residual, respectively. Whereas, positive coefficients for X₂X₃, X₂X₄, and in Y₁ and X₂, X₃, X₄, X₂X₄, X₃X₄, , , and in Y₂ indicate favorable effects on the percent TOC removal and the H₂O₂ residual, respectively. Since the coefficients with values close to zero represent lower relative intensity, do not intensely affect the TOC removal while X₁, X₁X₂, X₂X₃, and do not intensely affect H₂O₂ residual.

Although this evaluation provides a rapid analysis of the parametrical effect on the response variables, ANOVA with 95% CI was also applied to evaluate the statistical significance of the developed quadratic models for the percent TOC removal and the H₂O₂ residual. Thus, the statistical significance of each factor coefficient, as shown in Equations (3) and (4), was determined by the Fisher’s (F) exact test, comparing probability (p) values greater than F. Consequently, the model F-values of 287.69 and 116.90 for TOC removal and H₂O₂ residual, respectively, imply the models are significant.

Besides, small probability values (p < 0.05) indicate significant model terms, which confirm the accuracy of the developed models to predict the response functions. Conversely, p-values > 0.10 indicate the model terms are not significant, in this case, is not significant for both TOC removal and H₂O₂ residual. If the quadratic effect is not significant, then the optimal levels of the parameter are in the extremes of the experimental region [14].

The goodness of fit of the developed models was validated by the determination coefficient (R²) and the adjusted R² that ensures an adequate variation of the quadratic model to the experimental values. The values of R² and adjusted R² were found to be 0.9963 and 0.9928 for the percent TOC removal and 0.9909 and 0.9824 for the H₂O₂ residual, respectively, representing an adequate model’s significance.

Moreover, the adequate precision for the percent TOC removal and H₂O₂ residual models were found to be 77.49 and 51.54, respectively (Table 4). Since both values were greater than 4.00, the model can be used to navigate the CCD design space [15]. The lack of fit was calculated to assess how well the model fits the data. The lack of fit p-values of the percent TOC removal and the H₂O₂ residual were found to be 0.6059 and 0.1145, respectively. A not significant lack of fit (p > 0.10) indicates that the model fits the data well.

On the other hand, the assumption of the constant variance was verified by plotting the internally studentized residual versus predicted values (Figure 3(a) and Figure 3(b)). The studentized residuals were found dividing the residuals by their standard deviations showing a randomly scattered pattern within the outlier detection limits −3 and +3. Therefore, model predictions, described in Equations (3) and (4), for both the percent TOC removal and the H₂O₂ residual, respectively, are satisfactory.

Moreover, the normal probability plot of residuals, shown in Figure 4(a) and Figure 4(b) for the TOC removal

and the H₂O₂ residual, respectively, showed a straight line pattern followed by the points on the plot, not an S-shaped curve. Consequently, a transformation of the response is not required because of the normal distribution of the residuals [11].

The correlation between the observed and predicted values is shown in Figure 5(a) and Figure 5(b) for the TOC removal and the H₂O₂ residual, respectively. As a result, minor discrepancies are represented by a straight line trend, which indicates a good agreement between observed and predicted values. Hence, the quadratic model predictions for both percent TOC removal and H₂O₂ residual responses are satisfactory.

The significance of each model factor was also evaluated using the F-exact test and p-values for each factor including linear, quadratic, and cross-factor interaction. All four independent variables including influent TOC (X₁), flow rate (X₂), H₂O₂ dosage (X₃), and pH (X₄) have a significant effect on both responses since their p- values are lower than 0.05. Besides, the cross-factor interactions of all model parameters, including the influent TOC concentration and flow rate (X₁X₂), influent TOC concentration and H₂O₂ dosage (X₁X₃), influent TOC concentration and pH (X₁X₄), flow rate and H₂O₂ dosage (X₂X₃), flow rate and pH (X₂X₄), and H₂O₂ dosage and pH (X₃X₄) showed a significant effect on both TOC removal and H₂O₂ residual. The cross-factor interaction effects with the highest significance as per their p-values < 0.0001 are illustrated in Figure 6.

The RSM was used to calculate the optimum conditions for the four independent variables to get maximum percent TOC removal and minimum H₂O₂ residual. Equations (3) and (4) were defined as objective functions for percent TOC removal and H₂O₂ residual, respectively, and the independent factors in their range were used as model constraints. Thus, the following optimum conditions to reach a maximum TOC removal of 98.9% and minimum H₂O₂ residual of 1.4% were found: influent TOC of 50 mg/L, flow rate of 15 mL/min, H₂O₂ dosage of 344 mg/L, and pH of 7.2. The obtained optimal operating conditions were used in an additional run to validate the predicted values. Obtaining a TOC removal of 97.8% and H₂O₂ residual of 1.3% were obtained experimentally, confirming the reliability of the model since the values are within the 95% CI.

^aDegrees of freedom (df). ^bFisher’s (F) exact test value. ^cA probability value (p) < 0.05 is considered to be significant, a p-value > 0.10 is considered not significant. ^dTotal sum of squares corrected for the mean.

^*Corresponding author.

The interaction effects of the influent TOC concentration, flow rate, H₂O₂ dosage, and pH had a significant effect on both TOC removal and H₂O₂ residual. Optimum conditions were found for each variable to achieve maximum TOC removal with minimum H₂O₂ residual. The developed mathematical models provided a comprehensive exploration of the cross-factor interactive effects of the independent variables on the responses. The proposed models explaining the treatment of SWW by the continuous ABR-AS-UV/H₂O₂ system were found suitable for future studies on reactor design, modeling, and scale-up.