Struvite (MAP, magnesium ammonium phosphate hexahydrate, MgNH4PO4.6H2O) precipitation-aided Fenton's OXidation (MAPFOX process) was explored in the treatment of high-strength real sheep slaughterhouse wastewater (RSSW) for the first time under a comprehensive soft-computing-based modeling study. The experimental results showed that under the highest-efficiency conditions (chemical combination of MgCl2.6H2O + NaH2PO4.2H2O, a molar ratio of Mg2+:NH4+-N:PO43–P = 1.2:1:1, a reaction pH of 9.0 ± 0.10, [NH4+-N]0 = 240 ± 20 mg/L, and a reaction time of 15 min), MAP precipitation could effectively remove more than 80 %, 60 %, 55 %, and 70 % of color, total chemical oxygen demand (TCOD), soluble COD (SCOD), and ammonium nitrogen (NH4+-N) from the raw RSSW. The results of the Fenton's oxidation used as the post-treatment unit of the proposed MAPFOX system indicated that the integrated advanced oxidation process (AOP) was able to reduce the residual pollutant levels in the MAP-pretreated RSSW to the relevant discharge standards. Under the subsequent condition of [Fe2+]0 = 90 mmol/L, [H2O2]0 = 180 mmol/L, reaction pH = 3.25, and total reaction time = 60 min, more than 97 % of color, TCOD, SCOD, and NH4+-N could be removed from the RSSW via the Fenton's oxidation after the MAP-based physicochemical treatment. According to SEM micrographs, surface morphology of dehydrated struvite exhibited irregular-shaped and overlapping sharp-edged particles of various sizes with an average size of about 50.9 μm. The Fourier Transform Infrared (FTIR) spectroscopy confirmed the active functional groups and type of bonds for the high-strength RSSW-oriented struvite (heated) within the spectral range of 4000–450 cm−1. Thermogravimetric Analysis (TGA), Derivative Thermogravimetry (DTG), Differential Thermal Analysis (DTA), and Differential Scanning Calorimetry (DSC) of the dehydrated struvite revealed that the weight loss occurred in three temperature zones, the maximum weight loss rate of 0.252 mg/min was recorded at around 224 °C and at time of 20.83 min, and the sample had strong endothermic and medium exothermic peaks at about 241 °C and 679 °C, respectively. The predictive successes of the implemented soft-computing approaches were benchmarked in terms of various statistical goodness-of-fit parameters. The performance assessment indices corroborated the superiority of the support vector machines-Pearson VII universal kernel function (SVM-PUKF)-based model (correlation coefficient (CC) = 0.9999–1.0000), mean absolute error (MAE) = 0.0222–0.0389 %, mean absolute percentage error (MAPE) = 0.0270–0.0506 %, root mean squared error (RMSE) = 0.0258–0.0415 %, coefficient of variation of RMSE (CV(RMSE) = 0.0003–0.0008, Nash–Sutcliffe efficiency (NSE) = 0.9998–1.000, Legates and McCabe's index (LMI) = 0.9894–0.9952) over other data-intelligent models in predicting the pollutant removal efficiencies. The computational results also indicated that the narrowest uncertainty bands (±1.96Se = 0.0537–0.1483 %) and the lowest amounts of expanded uncertainty (U95 = 3.1224–5.3124 %) values for all efficiency sets were achieved for the applied SVM-PUKF-based strategy. This study demonstrated the first-ever and successful application of the proposed MAPFOX process in treatability of the RSSW and capability of the implemented soft-computing strategy in modeling a highly nonlinear treatment system.
Read full abstract