Neuro-fuzzy inference system and white shark optimization of coagulation-flocculation of aquaculture wastewater treatment

BACKGROUND AND OBJECTIVES: Polymeric flocculants derived from the coagulation-flocculation procedure, such as chitosan, effectively remove algal-bacterial biomass from wastewater treatment, demonstrating efficiencies comparable to ferric salts at reduced dosages. For effective coagulation-flocculation in the treatment of aquaculture wastewater, it is crucial to determine accurate measurements for three primary input factors: potential of hydrogen, chitosan dose, and settling time. This study seeks to establish the most effective operating conditions for coagulation-flocculation by fine-tuning the three input parameters. METHODS: Artificial intelligence and recent optimzation algorithms are integrated to achieve the purpose of this study work. Initially, an adaptive neuro-fuzzy inference system model of the wastewater treatment process was built based on experimental data. The white shark optimizer is utilized to assess the most effective chitosan dosage, potential of hydrogen value, and settling time necessary for minimizing turbidity and salinity levels. Throughout the optimization procedure, the controlling parameters act as design variables, and the goal is to maximize the objective function, which is the combined total of turbidity and salinity removal. FINDINGS: The adaptive neuro-fuzzy inference system turbidity removal model has root mean square error values of 3.52e-05 and 1.51 for training and testing data, respectively. The R-squared values for testing and training are 0.76 and 1.0, respectively. The adaptive neuro-fuzzy inference system demonstrated a significant improvement over the analysis of variance, decreasing the root mean square error from 2.8 to 0.836, which represents a 70 percent decline. The estimated R-squared value rose by 11.76 percent, from 0.68 with analysis of variance to 0.76 with adaptive neuro-fuzzy inference system. In the context of salinity removal, the adaptive neuro-fuzzy inference system demonstrated root mean square error values of 7.13e-06 for the training phase and 2.045 for the testing phase. The R-squared values for training and testing are 1.0 and 0.82, respectively, and adaptive neuro-fuzzy inference system reduced the root mean square error from 6.8 with analysis of variance to 1.135 with adaptive neuro-fuzzy inference system achieving an 83.5 percent reduction. Through the analysis of variance, the R-squared value for prediction was 0.765, which improved by 7.2 percent to reach 0.82 when utilizing the adaptive neuro-fuzzy inference system. CONCLUSION: Adaptive neuro-fuzzy inference system succseded to present an accurate model of wastewater treatment process and white shark optimizer defined accutrelly the values of chitosan dose, pontential of hydrogen, and settling time to reduce turbidity and salinity. Thus, improving chitosan-based coagulation-flocculation presents substantial ecological benefits by lowering the use of toxic chemicals and decreasing sludge formation, ultimately contributing to reduced pollution.